A UNITED STATES DEPARTMENT OF COMMERCE PUBLICATION ^<°'c°* *f4t,s Of * U.S. DEPARTMENT OF COMMERCE National Oceanic and Atmospheric Administration COLLECTED REPRINTS-1972 Volume I ATLANTIC OCEANOGRAPHIC AND METEOROLOGICAL LABORATORIES r«p #**&**. r*»ENT Of C U.S. DEPARTMENT OF COMMERCE Frederick B. Dent, Secretary NATIONAL OCEANIC AND ATMOSPHERIC ADMINISTRATION Robert M. White, Administrator ENVIRONMENTAL RESEARCH LABORATORIES Wilmot N. Hess, Director Collected Reprints-1972 Volume I ATLANTIC OCEANOGRAPHIC AND METEOROLOGICAL LABORATORIES ISSUED JULY 1973 a 8 Q a a Atlantic Oceanographic and Meteorological Laboratories Miami, Florida 33149 For sale by the Superintendent of Documents, U. S. Government Printing Office, Washington, D. C. 20402 FOREWORD The end product of research is new knowledge. To be useful to any- one other than the investigator himself, however, this knowledge must somehow be made widely available. This routinely is done through the publication of research results in scientific journals. The published papers of the researchers at NOAA"s Atlantic Oceanographic and Meteoro- logical Laboratories appeared in so many different journals that many people who could benefit from the knowledge were in fact unaware it ex- isted. For this reason we began in 1966 to bring these papers together as a set of collected reprints. The response from the scientific com- munity has justified continuing this publication, and this volume, the seventh in the series, contains the published research results of NOAA's Atlantic Oceanographic and Meteorological Laboratories for the year 1972 V Harris B. Stewart, Jr. Director, Atlantic Oceanographjfx and Meteorological Laboratory m Digitized by the Internet Archive in 2012 with funding from LYRASIS Members and Sloan Foundation http://archive.org/details/collectedreprint1972atla CONTENTS VOLUME I Page Foreword i i i Table of Contents v General 1 . Carrodus , R. L . History of Maps: Muse News III, No. 12, 479~512 1 2. Cox, Doak C, and Harris B. Stewart, Jr. Technical Evaluation of the Seismic Sea Wave Warning System: The Great Alaska Earthquake of \Sdh\ Oceanography and Coastal Engineering ISBN 0-309-1 605~3 . National Academy of Sciences, 229-2^5. 7 3. Henderson, J. Welles, and Harris B. Stewart, Jr. A Recently Discovered CHALLENGER Sketchbook: Proceedings of the Royal Society of Edinburgh, Section B, 72, No. 20, 223-229. 2k k. Stewart, Harris B., Jr. Cicar's NOAA-CARIB, a New Experiment in International Scientific Cooperation: Environmental Data Service, February, 10-12. 36 5. Stewart, Harris B., Jr. Eyewitness Pictures of the CHALLENGER Expedition Discovered: Environmental Data Service, October, 3-7. 39 Physical Oceanography 6. Apel , J. R. Editor Sea Surface Topography from Space: Volume I - NOAA TR ERL 228-A0ML 7, Volume II - NOAA TR ERL 228-A0ML 7~2 (Summary) hk 7 . Chew, Frank, and George A. Berberian Neighbor Diffusivity as Related to Lateral Shear in the Florida Current: Deep-Sea Research 19, ^93~506 50 8. Chew Frank, William S. Richardson, and George A. Berberian A Comparison of Direct and Electric-Current Measurements in the Florida Current: Journal of Marine Research 29, No. 3, 339-3^6. 6k 9. Hansen, Donald V., and Maurice Rattray, Jr. Estuarine Circulation Induced by Diffusion: Journal of Marine Research 30, No. 3, 281-29^. 72 10. Maul, George A., and Donald V. Hansen An Observation of the Gulf Stream Surface Front Structure by Ship, Aircraft, and Satellite: Remote Sensing of Environment 2, No. 2, 109-116. 86 11. Maul, George A., and Miriam Sidran Comment on "Estimation of Sea Surface Temperature from Space" by D. Anding and R. Kauth: Remote Sensing of Environment 2, I65-I69. 9^ 12. Molinari, Robert L., and Donald V. Hansen Formulation of Drifting Limited Capability Buoy Placement and Retrieval Concepts: NOAA Tech Memo ERL A0ML-18. (Originally published in National Data Buoy Center's Statement of work OH^EC.) 99 13. Zetler, Bernard D. Comments on Paper by A. A. Nowroozi , "Long-Term Measure- ments of Pelagic Tidal Height Off the Coast of Northern California": Journal of Geophysical Research 77, No. 2k, 4590. 129 ]k. Zetler, Bernard D., and Robert A. Cummings Tidal Observations Near an Amphidrome: Geophysical Surveys 1, 85-98. 130 Meteorology 15. Anthes, R. A, Non-Developing Experiments with a Three-Level Axisymmetric Hurricane Model : NOAA Tech Memo ERL NHRL-97. \kk 16. Black, Peter G Some Observations from Hurricane Reconnaissance Aircraft of Sea-Surface Cooling Produced by Hurricane Ginger (1971): Mariners Weather Log 16, No. 5, 288-293. 164 VI 17. Black, Peter G., H. V. Senn, and C. L. Courtright Airborne Radar Observations of Eye Configuration Changes, Bright Band Distribution, and Precipitation Tilt During the 1969 Multiple Seeding Experiments in Hurricane Debbie: Monthly Weather Review 100, No. 3, 208-217. 170 18. Carlson, T. N., and J. M. Prospero The Large-Scale Movement of Saharan Air Outbreaks Over the Northern Equatorial Atlantic: Journal of Applied Meteorology 11, No. 2, 283-297. 180 19. Hawkins, H. F., Jr. Development of a Seven-Level, Balanced, Diagnostic Model and Its Application to Three Disparate Tropical Distur- bances: NOAA Tech Memo ERL NHRL-98 (Summary Only) 195 20. Hawkins, H. F., Jr., K. R. Kurfis, B. M. Lewis, and H. G. Ostlund Successful Test of an Airborne Gas Chromatograph : Journal of Applied Meteorology 11, No. 1, 221-226. 197 21. Lamb, D., and W. D. Scott Linear Growth Rates of Ice Crystals Grown from the Vapor Phase: Journal of Crystal Growth 12, 21-31. 203 22. Lubin, S. J., and B. M. Lewis Effects of Weather on Airborne Omega: Navigation Journal of the Institute of Navigation 19, No. 2, 1 75~1 80 . 2 1 4 23. Moss, M. S., and S. L. Rosenthal On the Role of the Organizational Periods in the NHRL Circularly Symmetric Hurricane Model: NOAA Tech Memo ERL NHRL-99. 220 2*t. Scott, Wil 1 iam D. An Assessment of the Present Instrumentation for the Measurement of Cloud Elements and our Needs: Second Symposium on Meteorological Observations and Instruments, March 27"30, 1972, San Diego, California, 205-216. 2kk 25. Scott, Wil 1 iams D. Details for Constructing a Miniature Solid State Electro- meter Probe: The Review of Scientific Instruments 43, No. 1, 152-153. 256 VI 1 26. Scott, Williams D., and Z. Levin Open Channels in Sea Ice (Leads) as Ion Sources: Science 177, klS-hld. 258 27. Smith, Clark L. On the Intensification of Hurricane Cel ia (1970): NOAA Tech Memo ERL NHRL-100. 259 28. Staff Project STORMFURY 1971 Annual Report. 296 Vol ume I I Marine Geology and Geophysics 29. Cronan, D. S., T. H. Van Andel , G. R. Heath, M. G. Dinkelman, R. H. Bennett, and D. Bukry Iron-Rich Basal Sediments from the Eastern Equatorial Pacific: Leg 16, Deep Sea Drilling Project: Science 175, 61-63. ^77 30. Dietz, Robert S. Book Review of: The Face of the Deep, by B . C. Heezen and C. D. Hoi lister: EOS 53, No. 2, 200-201. kQO 31 . Dietz, Robert S. Geosynclines, Mountains and Continent-building: Scientific American 226, No. 3, 30-38. 481 32. Dietz, Robert S. Plate Tectonics, Sea-Floor Spreading and Continental Drift: Journal College Science Teaching 2, No. 1, 16-19- ^90 33- Dietz, Robert S. Shatter Cones (Shock Fractures) in Astroblemes: 24th IGC Section 15, 112-118. k$k 34. Dietz, Robert S. Sudbury Astrobleme, Splash Emplaced Sub-Layer and Possible Cosmogenic Ores: New Developments in Sudbury Geology, J. Guy-Bray, Editor. Geological Association of Canada, Special Paper 10, 29"/*0. 501 vi 1 1 35. Dietz, Robert S. Walter H. Bucher Medal, Fourth Presentation to Robert S. Dietz, Citation by J. T. Wilson and REPLY by Dietz: EOS 52, No. 7, 5^0-5^1 . 513 36. Dietz, Robert S., and D. C. Krause Book Review of: The Earth's Tectonosphere . Its Past Development and Present Behavior, by J . H. Tatsch: EOS 53, No. 11 , 940-942. 515 37. Dietz, Robert S., and J. F. McHone, Jr. Laguna Guatavita: Not Meteoritic, Probable Salt Collapse Crater: Meteoritics 7, No. 3, 303-307. 51 6 38. Dietz, Robert S., and W. P. Sprol 1 Overlaps and Underlaps in the North America to Africa Continental Drift Fit: ICSU/SCOR Working Party 31 Symposium, Cambridge 1970: > The Geology of the East Atlantic Continental Margin, edited by F. M. Delany, 1970. Institute of Geological Sciences Report No. 70/13, 143-151. 521 39- Dietz, Robert S., J. C. Holden, and W. P. Sproll Antarctica and Continental Drift: Proceedings SCAR/IUGS Symposium on Antarctic Geology and Solid Earth Geophysics, Oslo, Norway 1970, 837-842. 529 40. Duane, David B., Michael E. Field, Edward P. Meisburger, Donald J. P. Swift, and S. Jeffress Williams Linear Shoals on the Atlantic Inner Continental Shelf, Florida to Long Island: Shelf Sediment Transport, edited by Swift, Duane, and Pi 1 key © Dowden , Hutchinson and Ross, Inc., Stroudsburg, Pa., 447-498. 544 41. Freeland, G. L., and Robert S. Dietz Plate Tectonics in the Caribbean: A Reply (Mattson's Critique): Nature 235, 156-157. 596 42. Grim, P. J., G. H. Keller, and R. J. Barday- Computer Produced Profiles of Microtopography as a Supplement to Contour Maps: International Hydrographic Review XL ix , No. 1, 81-94. 598 43. Holden, J. C, and R. S. Dietz Galapagos Gore, NazCoPac Triple Junction and Carnegie/Cocos Ridges: Nature 235, 266-269. 612 IX 44. Kel ler, George H . Engineering Properties of Greenland and Norwegian Basin Sediments: Proceedings First International Conference on Port and Ocean Engineering Under Arctic Conditions II, 1285-1311. 616 45. Keller, George H., Douglas N. Lambert, Richard H. Bennett, and James B. Rucker Mass Physical Properties of Tobago Trough Sediments: VI Conferencias Geologica del Caribe, Margarita, Venezuela, Memorias - 405-408. 643 46. Lambert, D. N., and R. H. Bennett Tables for Determining Porosity of Deep-Sea Sediments from Water Content and Average Grain Density Measurements: NOAA Tech Memo ERL AOML-17. (Summary Only) 647 47. Long, L. T., S. R. Bridges, and L. M. Dorman Simple Bouguer Gravity Map of Georgia: Geological Survey of Georgia, State of Georgia Department of Natural Resources. 652 48. McDonald, V. J., R. E. Olson, A. F. Richards, and G. H. Keller Measurement and Control System for Deep Sea Vane-Shear Device: Ocean 72, IEEE International Conference on Engineering in the Ocean Environment, 474-479. 654 49. Peter, George Geology and Geophysics of the Venezuelan Continental Margin Between Blanquilla and Orchil la Islands: NOAA TR ERL 226- AOML 6. 660 50. Prospero, J. M., and T. N Carlson Vertical and Areal Distribution of Saharan Dust Over the Western Equatorial North Atlantic Ocean: Journal of Geophysical Research 77, No. 27, 5255~5265. 747 51 . Rona, P. A. Exploration Methods for the Continental Shelf: Geology, Geophysics, Geochemistry: NOAA TR ERL 238-AOML 8. 758 52. Scott, R. B., P. A. Rona, L. W. Butler, A. J. Nalwalk, and M. R. Scott Manganese Crusts of the Atlantis Fracture Zone: Nature Physical Sciences 239, No. 92, 77"79. 807 53. Shideler, G. L., and Donald J. P. Swift Seismic Reconnaissance of Post-Miocene Deposits, Middle Atlantic Continental Shelf-Cape Henry, Virginia to Cape Hatteras, North Carolina: Marine Geology 12, 165-185 8U 5k. Shideler, G. L., Donald J. P. Swift, G. H. Johnson, and B. W. Holl iday Late Quarternary Stratigraphy of the Inner Virginia Shelf: A Proposed Standard Section: Geological Society of America Bulletin 83, 1787-1804. 835 55. Swift, Donald J. P. Implications of Sediment Dispersal from Bottom Current Measurements; Some Specific Problems in Understanding Bottom Sediment Distribution and Dispersal on the Continental Shelf-a Discussion of Two Papers: Shelf Sediment Transport, edited by Swift, Duane, and Pilkey © Dowden , Hutchinson and Ross, Inc., Stroudsburg, Pa., 363-371. 853 56. Swift, Donald J. P., and W. R. Boehmer Brown and Gray Sands on the Virginia Shelf: Color as a Function of Grain Size: Bulletin of Geological Society of America 83, No. k, 877-883. 862 57. Swift, Donald J. P., John C. Ludwick, and W. Richard Boehmer Shelf Sediment Transport: A Probability Model: Shelf Sediment Transport, edited by Swift, Duane and Pilkey© Dowden, Hutchinson and Ross, Inc., Stroudsburg, Pa., 195-223. 869 58. Swift, Donald J. P., J. R. Schubel , and R. W. Sheldon Size Analysis of Fine-Grained Suspended Sediments: A Review: Journal of Sedimentary Petrology 42, No. 1, 122-13**. 897 59. Van Andel , T. H., G. R. Heath, Richard H. Bennett, S. Charleston, D. W. Cronan, Kelvin S. Roldolfo, R. S. Yeats, D. Bukry, M. Dinkelman, and A. Kaneps Deep Sea Drilling Project - Leg 16: Geotimes, 12-14. 910 xi Sea-Air Interaction 60. Hanson, Ki rby J . Remote Sensing of the Ocean: Remote Sensing of the Troposphere, edited by V. E. Derr, U. S. Dept. of Commerce and the Electrical Engineering Dept., Univ. of Colo., Boulder, Colo. 913 61. Hanson, Ki rby J., and Monte F. Poindexter Attenuation of Broad-Band Solar Irradiance in the Ocean: Conference on Atmospheric Radiation, August 7-9, 1972, Fort Collins, Colo. (AMS , Boston, Mass.). 969 62. Hanson, Ki rby J., and Monte F. Poindexter The Solar Irradiance Environment of Florida Coastal Water During Flare: NOAA Tech Memo ERL A0ML-16. 973 63. Ostapoff, Feodor Observation of Laminae in the Thermocline of the Tropical Atlantic: Presented at the ICES Symposium on "Physical Variability in the North Atlantic" held in Dublin, Ireland, September 1969- Paper 8. 997 XI 1 Reprinted from Muse News ill, No. 12, 479-512 NORTH itsr The oldest known map, now on exhibition in the Semitic Museum of Harvard University, was discovered in excavating the ruined city of Ga Sur, about 200 miles north of Babylon* THE HISTORYOFMAPS By Robert L. Carrodus The oldest known map, made about 2500 B.C., is a small clay tablet show- ing a mountain-lined Babylonian val- ley. The Egyptians made maps as early as 1300 B.C., but the first accurate maps were drawn by the Greeks. In the 500s B.C., the Greeks became the first people to recognize that the world was round, rather than flat. The Romans used maps to set up taxation systems and to plan military moves. In the A.D. 300s, they mapped a network of roads from southern England to the Ganges Valley in India. The most fam- ous ancient maps, depicting all the known world, were made about A.D. 150 by Claudius Ptolemy, an Egyptian scolar. The next great map makers were the Moslems. A Moslem must face the holy city of Mecca, Arabia, while pray- ing, hence he must know the direction of that city. The Moslems were also interested in trade on land and sea, thus needing information about dis- tant lands in order to plan their jour- neys. In the Middle Ages, the people of Europe had only the poorest of maps, some of which they made by reading the Bible, which states that Jerusalem is surrounded by cities; therefore they put Jerusalem in the center of their maps. The Bible also speaks of the "circuit" of the earth, which caused some map makers to draw up oval- shaped maps. Other maps were made square because the Bible also speaks of the four corners of the earth. As people began to go on long pilgrim- ages, maps were gradually improved. During the Crusades, men began to use the compass in sailing. Compass direc- tions also helped to make the early maps more accurate. World maps were begun in the 1400s. Prince Henry the Navigator, the ruler of Portugal, set up a schoo'forsea captains and gathered together the best maps that could be found. Soon European sailors were discovering dis- tant lands around the world. Christo- pher Columbus studied the best maps of his time before he set out on his great voyage and Ferdinand Magelian started to sail around the world thirty years later. We are still making discoveries and putting new information on maps today. During World War II, it was found that there were almost no accurate maps available for many of the islands of the Pacific. Cartography, or the making of maps and charts, is a pre- cise and exact science which is steadily developing and expanding. There are many different kinds of maps. Some maps may show where the highest mountain ranges are. Others may show the location of cities, roads and railways, and the great currents of the oceans, plus maps used to illustrate various types of science, such as meteorology, satellites, geography, geology and oceanography. Every community uses maps to set- tle problems. Businessmen need maps OUR A UTHOR is a scientific meteoro- logical illustrator at NOAA 'S National Hurricane Research Laboratory, Coral Cables. A native of Ohio, he spent ten years as an Air Force weather observer and forecaster prior to his graduation from University of Miami in 1961, where, he is an adjunct professor. He is a professional member of American Meteorological Society. This is his first appearance in Muse News. 479 to show them a good place to sell their products and how to ship their goods. Aviation maps are necessary for safe flying. Maps are necessary for workers in special fields and for all people who follow world events. Maps and globes make large things appear small. Studying a map or globe is somewhat like looking at an object through the wrong end of a telescope. Areas of the world which are much too large to be seen at one time, even from the air, can be studied this way. The advancements in photography, aviation and the printing processes, together with the development of the electronic computer and photogram- metric methods, are extensively used in the preparation of nearly all pub- lished large-scale topographic maps. Today, although costs are consider- able, high-speed, multicolor processes are capable of handling any kind of cartographic problems. Modern map construction with its scribing, photo lettering and infrared photography by satellite, is as different from the map making of fifty years ago as is the automobile from the horse and buggy. Maps have increasingly come to be based on accurate measuring, counting and computing, rather than mainly on speculation and guesswork. Modern cartography has stressed the scientific and practical functions of maps at the expense, if not the total neglect, of their aesthetic functions. The development of scientific data has enormously en- larged the variety and quantity of mappable facts and consequently of maps and their uses. So many and so diverse are these uses today that maps have become indispensable. How to Use a Map In order to make intelligent use of a map you must know map language, made up of symbols of various kinds, each symbol standing for a condition or a feature of the landscape. In the process of map making the number of symbols used has become very great. Learning to translate them is some- what like learning to translate a for- eign language. But it is impossible to put a vast amount of information into a single map. The symbols are "signs that save the mind a great many words". There are two kinds of symbols used on maps. One kind is used to show man-made features of the land- scape, also called cultural features. The other stands for natural features. A dot may be used to represent a city, or the size of a circle may vary with the size of the city; capital cities being shown by a star inside the circle or a line drawn under the city's name. Special lines are used to show roads, railroads, ship lines and air routes. Canals, tunnels, bridges, dams and other works of man have standard symbols. Boundaries and borders are shown by dots and lines. A good map always has a pictorial legend which is the key to the various symbols used. Natural features may include the drainage and relief features of a region. Fairly large-scale maps may show fea- tures such as streams, water holes, springs, rapids and falls. Where color is used, water bodies are usually shown in blue. Lines which mark off areas of equal elevation or depression are usual- ly brown. There is an international agreement on the standard use of color in physical maps. Green may be used for land less than 1,000 feet above sea level, yellow for land elevation be- tween 1,000 and 2,000 feet, tan be- tween 2,000 and 5,000 feet, orange for land between 5,000 and 10,000 feet and brown means an elevation more than 10,000 feet above sea level. Glaciers, marshes, sand dunes and other special natural features are shown only on large-scale maps. The most useful single map for in- formation about the character of the landscape is a good political-physical map. But there are any number of other maps, that are used for special purposes. A map must be drawn to scale in order to be accurate. The scale on a map shows how much any real-earth distance has been reduced in order to be shown on the map. The map scale must be shown so that the distance 480 and the areas on the map can be trans- lated into real distances and areas on the earth's surface, or the map loses much of its usefulness. Distances on the maps are often shown by a line scale at the bottom of the map. Scale is shown in three dif- ferent ways, but it is not necessary that all three ways be shown on every map. Representative Fraction Method, such as 1:10,000 or 1/10,000. This means that one unit of measurement (1 inch) on the map represents 10,000 of the same units on the earth's sur- face. This means that if the distance shown on the map were multiplied 10,000 times, it would be its true size. Every earth distance on this scale of map is thus shrunk to 1/10,000 of its real size. Inches to the Mile. A common way of expressing scale is in the inch-to- the-mile system. This appears at the bottom of the map for example, as 1 inch = 8 miles or 1/8 inch = 1 mile. This system makes it easier to figure distances. Graphic Representation. On many maps, scale is represented geographic- ally by means of a straight line drawn in the margin of the map. They repre- sent a certain number of miles on the surface of the earth. In order to find locations on maps, a network of accurately placed lines is necessary so that a person reading a map can describe the location of a cer- tain place. This network is called the geographic grid. It is formed by regu- larly spaced, true east-and-west and true north-and-south lines. The lines running east and west are called para- llels of latitude. The lines running north and south are called meridians of longitude. There are various forms of maps, such as the globe. The globe presents the most nearly perfect picture of the earth. The oldest known globe in ex- istence was made by Martin ^^-haim in Nurnberg, Germany in 1492. A globe of the earth is called a W.rrestial globe. The universe may also be shown by a celestial globe, which shows the posi- tion of the earth and the heavenly bodies. Only a globe can show all parts of the earth true to scale. Directions and distances may be given accurately and ocean routes can be easily measured. The globe is a preferred map for many purposes. The best way to get an un- derstanding of basic world geography is to study the globe. Any map drawn on a flat surface using the globe's network of parallels and meridians is called a projection. It is not possible to draw a flat map of the earth without some kind of squeez- ing, stretching or tearing. It is like trying to flatten out half a rubber ball. It will wrinkle or crack. If only a very small area is shown on a map, the error is not important. But when large areas of the surface of the earth are put onto flat maps the areas nearer the poles will be greatly pulled out of shape, or distorted. Flat maps are al- ways somewhat distorted. It is impor- tant to realize that there is no one per- fect map projection. The following descriptions indicate nine kinds of map projections. Although there are other special projections, the nine indicated are most widely used. 1. Mercator projection. This type of map was invented by Gerhard Kremer, a geographer who used the Latinized name of Mercator. A group of Dutch merchants asked Mercator for an improved map to be used by their navigators. Mercator finished his map in 1569. The meridians are straight lines, equal distance and true on the equator. The parallels are straight lines at right angles to the mer- idians and are progressively farther apart away from the equator. Uses for the Mercator projection are for navi- gators' plotting charts; they can be used for any purpose if the area is at or near the equator. 2. Conic projections.. A good pro- jection for showing s single continent. A conic map has its greatest accuracy along the 35° latitude line. The meri- dians are straight lines, truly spaced on each parallel. The parallels, equally (Continued on page 511) 481 MAPS (Continued from page 481) spaced on the meridians, are con- centric circles or arcs or concentric circles. This projection is used for Atlas maps and for maps of mid-lati- tude areas. 3. Lambert's Conformal projec- tion.. This type of map, invented by a German mathematician named Johann Lambert, is often used for military purposes. It was used to map the Western Front during World War I and is the present basis of the United States sectional aeronautical charts. It is also used in many of the scientific fields, such as National Weather Service base maps, mainly because the error is very small. For instance, if we use standard parallels of 29° and 45° the error is almost evenly divided (1% in the central U.S. and 1.2% in Florida). The meridians are converging straight lines and are spaced truly on two parallels, which are known as the standard parallels. All the parallels are sections of concentric circles (circles placed one within another and having a common center). The parallels are spaced so that small areas are true to shape on the globe. The amount of error is very small. 4. Polyconic projection. This map, worked out by an American, Dr. Ferdinand Hassler, has been widely used for topographical maps of the United States Geological Survey. It is a very important map projection because of its wide usefulness for maps of small areas. The meridians are curved lines (except the center meridian), spaced truly on each parallel. The parallels are non-concentric circles or arcs of non- concentric circles, spaced along the standard meridian. The spacing of the parallels increases with distance away from the standard meridians. 5. Gnomonic projection. In this type of conic projection map, the grid is set up by projecting the surface of the globe upon a plane from the center of the globe. The map is always limited to less than a hemisphere. Its distortions of shape, area and scale are very great. The parallels are concentric circles placed so that the distance from the pole increases. At the equator the meridians are vertical straight lines. The distance between meridians in- creases as the distance from the central meridian increases. It is used by navi- gators, because the shortest route be- tween two points is along the arc of a great circle; it is also used for polar flights. 6. Sinusoidal projection. This map has straight horizontal parallels spaced equally at their true distance. It has a straight, vertical central meridian and curved meridians drawn through true divisions of each parallel. Squeezing causes a large distortion in high lati- tudes, but it is a good projection in low latitude. Africa, Australia and South America are often drawn on this projection, in use since about 1650. 7. Jiomolographic projection. The maps are drawn with horizontal straight parallels. They gradually be- come closer together as the distance from the equator increases. The meridians are equally spaced circles somewhat flattened. Distortion of shape is great at edges. For this reason the projection has not been very popu- lar. It was designed by Karl Mollweide of Germany in 1805. 8. Interrupted Homolosine projec- tion. This map is widely used. It was worked out by Dr. J. P. Goode of the University of Chicago in 1925. The land-form shapes are very good and it is especially good for showing distribu- tion of land use. The equator is di- vided true to scale. Parallels are straight lines. Each continent has its own central meridian from which other meridians are laid off to the east and west. The shapes are quite good. The projection is especially good for show- ing distributions. The fact that the oceans are cut or interrupted is a dis- advantage for certain distributions. 9. Polar Equidistant projection. No study of projections is complete with- 511 out calling attention to the polar pro- jection, often called "Map for the Air Age". It may be centered on the North Pole or the South Pole. The polar pro- jection map was used as far back as the 1500s. The meridians are radiating straight lines. The parallels are con- centric circles equally spaced at the true distances. The scale is correct only along meridians and there is a great exaggeration of scale which in- creases at the edges. South America, Australia and New Zealand are hor- ribly distorted. The projection is es- pecially valuable as an "air map" for plotting routes and showing map dis- tances. In an effort to name various other projections, they are listed as I, II, III, IV. |_ Meridians and parallels straight and at right angles. 1. Rectangular 2. Stereographic Cylindrical 3. Orthographic Cylindrical II. Parallels straight lines; meridians Maps (Continued from Page 512) The oldest known map, now on exhibition in the Semitic Museum of Harvard University, was discovered in excavating the ruined city of Ga Sur, about 200 miles north of Babylon, where the excavators found a baked clay tablet showing a river valley, perhaps the Euphrates, with moun- tains indicated the fish-scale fashion on each side. curved 1. Parabolic equal area 2. Goode's homolosine pro- jection 3. The Eckert projection III. Meridians straight lines; parallels concentric circles 1. Alber's conic equal area 2. Polar orthographic 3. Polar equidistant IV. Curved meridians straight lines; parallels concentric circles 1. Meridional stereographic 2. Azimuthal equidistant meridional globular 3. Bonne's projection In conclusion: a map projection is a systematic drawing of lines repre- senting the meridians and parallels of the earth sphere or any portion of it on a plane surface. The map is really the end result. It is a plane surface representation of a portion or all of the spherical surface of the earth. (Continued on Page 518) North, east and west are indicated with inscribed circles, indicating that maps were aligned with the cardinal directions then as now. Though broken, the tablet, small enough to be held in the hollow of one's hand, is remarkably fresh looking, and the clearness of the minute cuneiform characters would hardly suggest its venerable age of about 4500 years. Our cover is a freehand drawing by the author from a published picture of the oldest map. •••*••••••• 512 2 Technical Evaluation of the Seismic Sea Wave Warning System DOAK C. COX UNIVERSITY OF HAWAII HARRIS B. STEWART, JR. ENVIRONMENTAL SCIENCE SERVICES ADMINISTRATION* Reprinted from The Great Alaska Earthquake of 1964: Oceanography and Coastal Engineering ISBN 0-309-1605-3 NATIONAL ACADEMY OF SCIENCES Washington, DC 1972 OBJECTIVES OF THE SSWWS ABSTRACT: The objective of the Seismic Sea Wave Warning System is taken to be to minimize the hazards of tsunamis, especially hazards to human life and health, through issuance of timely warnings for dis- semination by participating agencies. A tsunami warning was issued by the system on the occasion of the 1964 Alaska earthquake; neverthe- less, loss of life suggests room for improvement. Means of increasing the speed of seismic analyses and their data, the speed of wave detection, and the information from hydrodynamic analyses used in the system are discussed. Many of these have been adopted since the Alaska earthquake. In the reduction of hazard on a particular coast, not only the timeliness and accuracy of the warning issued on a single instance are important, but also the record of suc- cesses and failures of previous warnings on that coast. A policy of re- gional evaluation, adopted in 1966, will permit an increased ratio of successes to failures on each coast. A special regional warning system has been established for Alaska since the earthquake. Even with the special system, there will be dif- ficulties connected with the high frequency of earthquakes, low fre- quency of significant tsunamis, and short travel times to vulnerable coasts from areas of potential tsunami generation. *Now, in part, the National Ocean Survey, NOAA. The performance of the Seismic Sea Wave Warning System (sswws) in relation to the threat of tsunamis accompanying the great 1964 Alaska earthquake has been described in detail by Spaeth and Berkman (1967, and this volume). Any evalua- tion of this performance must depend on the objectives as- sumed for the system. In terms of the simple objective of issuing warnings of tsunamis, which has sometimes been con- sidered to be the objective of the sswws (for discussion of the various objectives stated for the sswws, see Cox, 1968), the system of course functioned perfectly. This is, however, neither the final nor the most important objective. A requirement of the sswws that distinguishes it from other tsunami warning systems is that the generation of a tsunami be confirmed, if possible, by marigraphic evidence before a warning is issued. Because of unavoidable imperfec- tions in the array of marigraphic stations participating in the system and the communications system linking them with the Honolulu Observatory, which serves as the control center of the sswws, this requirement introduces the significant probability that some small tsunamis will occur without warning. The sswws was, however, rightfully considered as constituting a considerable improvement over an earlier sys- tem that had used only seismographic information (Finch, 1924; Cox, 1964). That system, which had previously been discontinued, could have been more certain of issuing warn- ings of every tsunami by a suitable choice of threshold seis- mic magnitude, but there would have been less certainty of tsunami generation for every warning issued. The greater im- portance attached to issuing warnings more likely to be fol- lowed by tsunamis than to issuing warnings of all possible tsunamis is an indication of a humanitarian intent involving human response. As stated by Comdr. (later Capt.) Elliott B. Roberts of the U.S. Coast and Geodetic Survey, under whose leadership the sswws was initially planned, the operation of the system was intended to be effective ". . . so that the people of seacoast towns such as Hilo could be warned and escape death" (Roberts, 1950). 229 230 TSUNAMIS The U.S. Coast and Geodetic Survey has never actually been responsible for the dissemination of warnings directly to the public. It has relied on state and local Civil Defense and police organizations for dissemination of warnings in the United States and its territories, and on foreign agencies for reevaluation and dissemination in foreign countries. In view of the essential cooperation of these other agencies, the real objective of the sswws apparently is to minimize the haz- ards of tsunamis-especially hazards to human life and health- through the issuance of timely warnings for dissemination by participating agencies. In relation to this goal, the performance of the sswws on the occasion of the great earthquake of March 1964 should be evaluated not only in itself but also in relation to the capabilities of the agencies to which the warning informa- tion was issued and the understanding of the public whose appropriate response was intended to be induced by the warnings. The capabilities and performance of the agencies responsi- ble for the dissemination of the warnings and the understand- ing and response of the public to the warnings are summarized by Weller ( 1972, this volume) and are treated in detail else- where (Norton and Haas, 1970; Yutzy, 1964; Anderson, 1970). Hence, these topics will not be discussed at length here, except for the preconditioning of the public by the performance of the sswws prior to 1964; their essential importance must, however, be kept in mind. MARCH 1964 OPERATIONS For convenience in the evaluation of the operation of the sswws at the time of the March 1964 earthquake and tsu- nami, an abbreviated chronology, compiled from the detailed log of the Honolulu Observatory (HO) (Spaeth and Berkman, 1967, and this volume) and from tide-gage and other reports, is given below: Time (GMT) March 28, 1964 Time (GMT) March 28, 1964 Event Event 03:36 03:44 04:10± 04:05 04:13-04:18 04:19-04:59 Occurrence of earthquake [March 27, 17:36 Alaska Standard Time (AST) j . First waves of local tsu- namis arrived within minutes at Valdez, Whittier, and Seward. Seismograph alarm triggered at HO. Tsunami warning, based on reports of Coast Guard at Cape Chiniak, informally spread in town of Kodiak (March 27, 18:10 AST). HO informally notified Coast and Geodetic Survey and Civil Defense (CD) personnel in Hawaii. HO requested readings from cooperative seismograph stations. Seismic readings, except from Alaska, received by HO. 04:20-04:35 04:32 04:49 04:36-04:54 04:52 05:01 05:02 05:30 05:39-05:50 05:50+ 05:52-05:55 05:55 06:11 06:30 06:37-06:45 07:15 07:39 08:10 08:22 08:26 08:42 09:00 09:15 09:23 09:42 09:45 11:00 Rise of first tsunami wave at Womens Bay, Kodiak (March 27, 18:20-18:35 AST). Tide observer sent message to HO at 04:35. HO received report on quake felt and damage to tide gage on Kodiak. Seismic readings sent by HO to Japan Meteorological Agency HO received Federal Aviation Agency reports of Alaskan communications failures. Epicenter and magnitude determined by HO. HO received report of tsunami headed for Kodiak. HO issued advisory bulletin giving estimated time of arrival (ETA) at Honolulu if tsunami generated. HO issued second advisory bulletin giving ETA's at other places, including Alaska. Marigraphic readings requested from Alaska stations. Maximum inundation by second wave in town of Kodiak (March 27, 19:50+ AST). Marigraphic readings requested from British Colum- bia and California. HO received 04:35 report on tsunami from Kodiak tide observer. Information passed to CD at 05:58. HO received beginning of second message on tsunami from Kodiak tide observer. HO received complete second message on tsunami from Kodiak tide observer. (Message apparently originated not long after 05:07.) HO issued warning, repeating original ETA's. HO provided CD with ETA's for other Hawaiian ports besides Honolulu. First rise began at Crescent City, California [March 27, 23:39 Pacific Standard Time (PST)] . HO received report from Kodiak that by 07:15 wave action had decreased and was only slight. HO received report from Coast Guard on tsunami heights at various points between Alaska and northern California. HO received report from Torino, Canada, on first crest at 07: 10 (March 27, 23:10 PST). HO received Sitka report of waves to 07:45. First rise began at Hilo, Hawaii [March 27, 23:00 Hawaiian Standard Time (HST[ . CD passed first Hawaiian wave reports to HO. HO passed various wave reports to Japan Meteorologi- cal Agency. Highest wave (second) hit Hilo (March 27, 23:42 HST). Highest wave (fourth) hit Crescent City, California (March 28, 01:45 PST). HO issued all clear for Hawaii, and advised all-clear status elsewhere 2 hours after local ETA's in absence of local contraindications. WARNING The first and most important comment to be made on the sswws operation on the occasion of the March 1964 earth- TECHNICAL EVALUATION OF THE SEISMIC SEA WAVE WARNING SYSTEM 231 quake is that it issued a warning which reached many coasts on the borders and on islands of the Pacific Ocean, resulting unquestionably in a reduction of casualties at Crescent City, California, and perhaps at other places in Washington, Oregon, California, and Hawaii. Clearly, then, the Warning System fulfilled its mission in part, even though the loss of life that did occur suggests that the risk had not been reduced to a minimum and that further improvement should be sought. One limitation of the effectiveness of the warning lay in its timing. As shown by Table 1, the warning was issued by the sswws at 06:37 GMT about 3 hours after the occurrence of the earthquake, almost the same length of time after the attack of the first of the local, slump-generated tsunamis at some Alaska towns such as Valdez, Seward, and Whittier, and 2Va hours after the arrival of the front of the major tsu- nami at Kodiak and probably about V* hour after the arrival of the highest crest there. The warning of the sswws was issued % hour before the arrival of the tsunami late at night on the coast of Washington, and over 1 hour before its arrival at the coasts of Oregon and northern California. The relay of the warning information to the public consumed most of the lead time on these coasts, however, so that warnings were publicly disseminated at about the time of arrival of the tsunami. The highest crests, however, came after midnight \Vi to 2 hours after the warning had been given to the public. The tsunami front arrived in Hawaii 2 hours after the warning had been issued by the sswws and 1 Vi hours after the public dissemination of the warning. The timeliness of warnings has always been a problem for the sswws, particularly in the case of tsunamis from the Aleutian Islands, which may arrive in the Hawaiian Islands in not much over 4 hours. Because the Hawaii State Civil Defense Division believes that it needs 1 hour, and preferably should have 2 hours, to make certain that a warning is dis- seminated as well as possible, the sswws has in recent years agreed to issue any warning for Hawaii not later than 70 minutes before the earliest estimated Hawaiian time of ar- rival, even if tsunami generation cannot be confirmed by then. Ordinarily, however, confirmation can be obtained in less than 3 hours, In the case of the March 1964 tsunami, although the 3-hr interval from the time of the earthquake until issuance of the warning by HO did not exceed the elapsed intervals in some previous Aleutian tsunami warnings, it deserves comment. An interval of this length between the earthquake and the issuance of the tsunami warning becomes important when one realizes that ( 1 ) local tsunamis washed on shore at populated places almost immediately after the earthquake, (2) the major tsunami was observed at Cape Chiniak, Kodiak, only about half an hour after the earthquake, with a local warning being given in the town of Kodiak not long after, and (3) a report of the waves at Kodiak was received by HO less than 2lk hours after the earthquake. The slowness with which reports of the early waves in Alaska reached the Honolulu Observa- tory was, of course, due to the failure of most of the con- ventional lines of communication in Alaska. Undoubtedly the Observatory did not issue a warning at the time of its receipt of the first report of a tsunami ap- proaching Kodiak, because this report had come through an unusual channel. The Observatory had had considerable ex- perience with reports from agencies and through channels other than those officially connected with the sswws; many of these had proved to be highly exaggerated or even totally false. It is not clear, however, why a warning was not issued 2lA hours after the earthquake, following the receipt (at 05:55 GMT) by the Observatory of the first official message from TABLE 1 Times (GMT) of Issuance and Dissemination of Alaska Earthquake Tsunami Warnings, March 28, 1964 Time of Issuance of Warning by HO Earliest Arrival of Time of Receipt of Warning by Time of Local Warning Tsunami Highest Waves State State CD Dissemination Time Place Time Place Alaska - - 03:40" Valdez, Whittier, Seward - - - 04:10ft 04:20 Womens Bay, Kodiak 05:50 - Washington 07:13 07:18+ 07:18 Neah Bay - - Oregon 07:00 07:00+ 07:56c Astoria - - California 07:13 07:25-07:50 07:39 Crescent City 09:00-09:50 Crescent City Hawaii 06:43 07:00 08:33 Nawiliwili 09:28 Hilo 06:37 "Local slump-generated tsunamis, during or immediately after the main shock. "Warning originated locally, not from HO. ( The Coast Guard reported arrivals between 07:00 and 08:00. 232 TSUNAMIS the Kodiak tide observer that a seismic sea wave was expe- rienced at 04:35 GMT. There are several possible explana- tions, which may well have reinforced each other. The Observatory personnel had computed and broadcast an esti- mated arrival time of 05:30 GMT at Kodiak for the tsunami. They were probably skeptical of a report that a significant water-level change could have begun almost an hour earlier. The message did not give a measure of the change in water level, as is prescribed in tsunami reporting, but instead re- ported a height 10 to 12 ft above mean sea level. Actually, a water level that high above preearthquake mean sea level would not have been beyond the range of the normal tide, because of the subsidence of 5 ft at Kodiak, but the Observatory personnel could not have known of the sub- sidence; if the Kodiak tide range had been known, the ex- cursion reported should have been recognized as excessive. The picture was complicated by the fact, known to the Ob- servatory, that the Kodiak tide gage had been damaged and that any readings reported must have been based on visual observations. In hindsight, then, the Kodiak report received at 05:55 GMT should have served as the basis for issuance of a warning under the officially accepted criterion that a tsu- nami, of however small magnitude, had been observed. The beginning of the second message from the Kodiak tide observer, originating a little after 05:07, was received by HO at 06: 1 1 , but the transmission was cut off before any information additional to that of the first message had been received. The full message was not received until 06:30, and the warning was issued only 7 minutes later. The message delays point up the inadequacy of the communication equip- ment to handle the incoming and outgoing traffic during the peak period of an actual emergency. It appears probable that earlier issuance of the warning would not actually have made any sigificant difference in the number of casualties from the tsunami. The warning from Honolulu could not have reached any of the Alaska com- munities in time to effect shoreline evacuation, and in Oregon and California the disastrous high waves came some time after the warning had been disseminated. Special comment is due on the warning of Alaska com- munities. The tsunami warning services serving Japan and the Soviet Union are designed to issue warnings, on the basis of seismographic information alone, in time for them to be of value in the case of tsunamis generated locally offshore. The sswws, with its requirement of the marigraphic confirma- tion necessary to reduce false alarms, could not be expected to operate with sufficient speed to provide for local warn- ings. The manual of the system (Spaeth, 1962) specifically cautions against reliance on sswws warnings in the case of locally generated tsunamis on foreign coastlines; the cau- tion would have been just as pertinent to domestic coast- lines. It is somewhat questionable whether a local warning system is practicable on a coastline such as that of Alaska, with its sparse population, attenuated communications, and low frequency of tsunamis, although the U.S. Coast and Geo- detic Survey has now established a regional warning system in Alaska to do what it can. In any case, the earthquakes themselves provide natural warnings, and there will always be a place for the kind of in- telligent local initiative that must certainly be credited with a significant reduction in casualties from the 1964 tsunami at the town of Kodiak. ADVISORY BULLETINS While considering the timing of bulletins issued by HO, we should note (1) that the Civil Defense personnel in Hawaii were informally notified by HO of the occurrence of the earthquake less than half an hour after the earthquake and only 21 minutes after the seismograph alarm was triggered at the Observatory, and (2) that an official advisory bulletin was issued less than Vh. hours after the earthquake and \lA hours before the tsunami warning was issued. With the ad- vance notice that a warning might be issued, the Hawaii State CD and its various cooperating county agencies could be fully mobilized and capable of coordinating the siren warnings, radio broadcasts, and police action all along the Hawaiian shores. Warning-dissemination agencies in areas other than Hawaii probably did not begin to mobilize until after the issuance of the first official advisory bulletin by the Observatory about \xh hours after the earthquake and \xh hours before tsunami warning. It seems probable that the advisory bulletin was de- layed pending determinations by the Observatory of an epi- central location and a magnitude for the earthquake, proof that the quake was of such size and location as to be of con- cern. These determinations were made W* hours after the earthquake, and the bulletin was issued 10 minutes later. SEISMOGRAPHIC AND MARIGRAPHIC ANALYSIS The length of time required for determining epicentral location and magnitude deserves some comment. It should be recog- nized, of course, that no analysis could begin until after the seismograph alarm was triggered in the Observatory, which was 8 minutes after the earthquake began in the Prince William Sound area of Alaska. Still, over an hour was re- quired for the analysis. Cause of some of the delay probably lies in part in the sophistication of the seismographic equip- ment used. There was a visual recorder in the Observatory, but only the onset of the primary (P) phase of the earthquake could be distinugished on that. Other, more sensitive instru- ments recorded photographically, and the records had to be removed and developed before they could be used. To some extent, this sophistication was a needless handicap. For the major earthquakes of principal concern to the sswws, high magnification of the seismic wave does nothing but confuse the recording, but simple, low-magnification, visual-recording 10 TECHNICAL EVALUATION OF THE SEISMIC SEA WAVE WARNING SYSTEM 233 seismographs, available decades before the modern precise equipment was developed, are no longer used. Another source of slowness in epicentral determination lies in the method of /'-phase-arrival analysis ordinarily used. This method has advantages in precision and in lack of am- biguity, but it requires reports of arrival times of the /'-phase of an earthquake from several stations. Reports from sta- tions around the Pacific began to arrive at HO within 45 minutes after the beginning of the earthquake. However, an epicentral distance, at least, probably could have been com- puted from the difference in arrival times of successive phases of an earthquake on the records of the Honolulu Ob- servatory alone, if the clutter of the traces on the records had been reduced by low magnification. Local estimates of the direction of the epicenter and of the earthquake mag- nitude could also have been made after the photographic records were developed. A potentially serious result of slowness of epicentral de- termination is a delay in the request for marigraphic data. A request to a tide station has to indicate an approximate eta at the station. Travel times to every participating tide station have been worked out in advance, but it is necessary to know earthquake time and epicentral location. It should be noted that several tide stations in the system, including three in the Aleutian Islands, are equipped with automatic tsunami detectors and alarms, so that queries from HO are not necessary in the case of large waves. However, waves that are damaging elsewhere are commonly too small at these stations to trigger the detector alarms. COMMUNICATIONS Of course, the major contributor to delays in all the operations of the warning system was the partial failure of the most important communications links. Those who would com- pare the transmission times of March 1964 sswws messages with the times for normal transocean or transcontinental telephone communication should recognize, first, that most of the stations reporting to HO, especially the tide stations, are in isolated localities far from normal communication lines. Indeed, a system of communications, very ingenious and intricate but inevitably requiring multiple relaying, has had to be worked out. At the time of a major catastrophe, even though alternate routes were prearranged, considerable delays were to be expected; indeed, considering the general collapse of communications in Alaska immediately after the 1964 earthquake, it is a wonder that any of the initial mes- sages were successfully transmitted. ESTIMATED TIMES OF ARRIVAL TABLE 2 Comparison of Estimated and Actual Times (GMT) of Arrival of the 1 964 Tsunami Estimated Actual Kodiak Unalaska Adak Attu Sitka Tofino Neah Bay Crescent City Midway Honolulu 05:30 06:30 07:00 07:45 05:30 07:30 07:30 08:00 08:45 09:00 04:20 06:06 07:00 07:27 05:06 07:00 07:18 07:39 08:27 08:53 uation of Kodiak's early reports has already been mentioned. Other estimated arrival times were also in error, although not so seriously (Table 2). The discrepancies were due to the assumption of a point source for the tsunami. The epicenter determined by HO at 61°N, 147'/4°W was near northern Prince William Sound, about 100 mi from the coast. The tsunami travel times were computed from a point on the coast near Seward. However, the uplifted portion of the Continental Shelf that served as the tsunami-generating area extended some 600 km to the south, 1 00 km to the east, and 500 km to the west of the assumed source, giving the wave front in all directions a con- siderable lead. WAVE-HEIGHT INFORMATION The possible importance of the nonstandard height parameter used in the first official report of the tsunami from Kodiak has already been mentioned. It is worth noting that other wave reports received by HO, including some transmitted by HO to other warning agencies, used nonstandard parameters, in general not identified. Initial rise, double amplitude or range, runup height on land, and height above mean sea level were all used in reports to HO, and these four at least were passed on to other agencies. It is unlikely that the confusion of wave-height param- eters, rarely involving errors exceeding a factor of two, made any significant difference in the effectiveness of the warning in March 1964, with the exception already noted. However, such confusion has caused problems in earlier warnings, and it must certainly be eliminated before even generalized quan- titative tsunami forecasting is attempted. The recently re- vised wave-reporting manual for tide observers, which gives explicit instructions on scaling and reporting wave heights, should reduce this confusion. MANPOWER AND EQUIPMENT LIMITATIONS It is of interest to look into some other deficiencies of the warning information. The error in estimate of arrival of the Without a doubt, the few possibly serious deficiencies in the tsunami at Kodiak and its possible importance in HO's eval- operation of the sswws and some deficiencies that might 11 234 TSUNAMIS have proved serious under other circumstances are due, in very large measure, to limitations in manpower and equip- ment. The system was originally established by the U.S. Coast and Geodetic Survey through improvisation without specific authorization and without special funding. The Honolulu Ob- servatory has assumed its tsunami-warning duties in addition to its pre- 1948 duties with no significant increase in man- power. A crew ranging from two to four is responsible during a tsunami alert for developing and reading seismograph rec- ords, analyzing seismic data, sending data requests and re- ceiving data from distant stations, transmitting advisory and warning bulletins, and consulting with civil and military warning authorities, as well as makingjudgments of an ob- viously critical nature. Improvement in the recording, com- munication, and analytical equipment available has been slow because of financial limitations. Urxk-r tire circum- stances, the degree of success of the sswws must be judged remarkable. PAST PERFORMANCE OF THE sswws Undoubtedly, the effectiveness of the warning issued for the 1964 Alaska tsunami was influenced by the experience that the agencies and the public had had with previous warnings and the degree of confidence felt by both the officials and the public. According to Yutzy (1964), the relative ineffective- ness of the warning in Crescent City was due largely to a low level of confidence resulting from previous experience with false or seemingly false alarms; a qualitatively similar limita- tion in effectiveness could probably have been demonstrated at most other places for which warnings were issued. The performance of the sswws from its initiation in August 1948 through August 1967 has been discussed at length in a report by Cox (1968). In the 1 5 Vi years from September 1 , 1948, through February 1964, there occurred in and oa the borders of the Pacific 224 earthquakes of suf- ficient magnitude to have been accompanied by tsunamis (Table 3). At least 45 of these earthquakes were, in fact, ac- companied by tsunamis, including 16 large enough (generally, near their origins) to have been capable of doing damage (that is, with maximum runup heights of 2 m or more). In the same period, 14 warnings were issued by the sswws. The correlation between warnings and significant tsunamis was far from perfect, however, as will be shown. Table 4 summarizes data for all Pacific tsunamis that had maximum runups greater than 2 m and all sswws warnings for the period September 1948 through March 1964. This table shows that no tsunami causing significant damage on United States coasts had occurred without warnings and, in- deed, that warnings had been issued for the only three tsu- namis that were readily observable on any United States coast after crossing the ocean. These three were that of 1952 from Kamchatka, 1957 from the Aleutians, and 1960 from Chile. That casualties had been reduced by the oper- ation of the sswws prior to the Alaska earthquake seems beyond question. However, warnings had also been issued on the occasion in 1958 when the only significant wave motion occurred within Lituya Bay in southeastern Alaska, and on 1 0 other occasions when no significant wave activity was observed on any United States coastline, including two when the waves did not run up on any Pacific coastline as high as 2 m, another occasion when the generation of a tsunami is ques- TABLE 3 Summary of SSWWS Action in Relation to Pacific Tsunamis and Large Pacific Earthquakes, from August 1948 through February 1964a Tsunamis or Earthquakes SSWWS Action6 None Watch Warning Total Moderate to large tsunamis (magnitude Small tsunamis (magnitude < 1) Total of certain tsunamis Questionable tsunamis Earthquakes without tsunamis Total 5 31 5e 164 200 2 2_ 4 0 6 10 9c 10 1 3^ 14 16 29 45 6 173 224 "Tsunami magnitude as defined by lida and others ( 1967) = log, Hmax, where Hmax is the maximum runup in meters at a coastline near the origin. The "warning" column includes events for which warnings were issued by the Honolulu Observatory; the "watch" column, those for which ad- visory bulletins, but not warnings, were issued by the Honolulu Observatory; the "none" colum.i, those for which no public releases were made by the Honolulu Observatory, although there may have been much diagnostic activity. 'An enormous tsunami within Lituya Bay, Alaska, resulting (in July 19S8) from a quake-triggered landslide, is included as a large tsunami, although the tsunami tectonically generated on the Alaska Continental Shelf on the same occasion was very small. S3 small tsunamis generated by volcanic explosions in the Bonin Islands in 1952-1953 are omitted. e Questionable tsunamis occurring without earthquakes at Midway Island in October 1957 and in Dixon Entrance, British Columbia, in March 1963 are omitted. =S 2. 1 i ° 5 u. a: SO -a 3 o 3 00 3 < E o "B E 3 a: * IS 3 •» u ae -o o •* -J < =s £ 0- —. ^5 S. * 2= 1 5 5<2 s o 3 !t: < y s^ d. ■= a il 5 u-2 — ro ~ w, '" o Hi S < U s^ ™ a> o E ■= O a. LU uu on -S i4 sags S -a £ E x o < < VO — — *i ' _ h»0 O r*i ^ *-* a * 5 I — . rv» — . < u O f> 00 00 D- 0> 5 < 2 S J3 3 < ^ 2 E ■a e =3 3 » .e M ^ o 5 J| 5 I s s 235 13 =S 2. % £ <5& a o: "5 o£ =*;*:* £ I £ u: ^ S X 3 = OD S 0. U — — rt « < < a x Z E u 0 i£, •a » « c«J >. 3^ T, . d 3 E = 6 S ea > = C ■ i 2 .8 * 5 v o § 3 - "° c <= 5S.233 6 S c ■ 2 £■- 6 ■ S '" 2 n « lis ^ H* "w ■o : 5 8 »3 £ V . ~ .* 0. r a a 00 „ o x 2 a c E- E-O u I c : a «_ i E 0 a c S 2 "Si" c >. <" S i E = » > c 3 « 08 (^ UJ ■" '" § 2 - 2 a» c ■ i 2 c 8 2 « E£ I 6 2 c- = g< E ■ « k "S T C E H e < H 236 14 TECHNICAL EVALUATION OF THE SEISMIC SEA WAVE WARNING SYSTEM 237 tionable, and three more occasions when there was no tsu- nami. In contrast, warnings were not issued on the occasions of seven tsunamis that had runups equal to or exceeding 2 m somewhere in the Pacific, including one generated immed- iately off the coast of Puna, Hawaii, which had a very local runup of 2Yi m. All the false warnings (those of May 25, I960; November 13, 1%0; and December 20, 1962), as well as the warning on November I 2, 1958, on the occasion of a questionable tsu- nami, were the result of faulty marigraphic reporting. The water-level changes on which the warnings were based were probably not greater than the normal background oscillations at the stations in question on all these occasions except that of May 25, 1960, when the abnormal oscillations probably represented the tail of the major tsunami that had been gen- erated 3 days before. The importance of the problem of faulty marigraphic reporting was noted in a Natural Disaster Warning Survey in 1965 (Keutschenreuter, 1965), which as- cribed it in part to inadequate training of frequently rotated personnel at the tide stations. The actual false warnings are, however, not the only ones that cause a loss of public confidence. So far as the populace of any particular coast is concerned, a warning seems false if it is not followed by tsunami effects at least readily visible on that coast. Because every tsunami varies greatly in its effects from one coastal region to another (and even in short dis- tances on the same coast), it will never be possible to elimi- nate completely these seemingly false warnings. Of the 16 tsunamis that occurred from September 1, 1948, through February 1964 and that had runups anywhere in the Pacific as much as 2 m, only five had runups as great as 1 m outside the immediate region of their generation. Table 5 includes two tsunamis that have occurred since: in March 1964 and in February 1965. Table 5 indicates the distribution of the significant effects (runups S* 1 m) of these seven in 13 coastal regions of the Pacific. It will be seen that, of the seven, only two had sig- nificant effects in as many as half the regions listed (not counting respective regions of generation). These were the Kamchatka tsunami of November 1952 and the Chile tsunami of May 22, 1960. Alaska and Kamchatka were, as far as is known, not significantly affected by any tsunami not gener- ated off their own coastlines. In contrast, Peru and Chile were each significantly affected by three distant tsunamis, the Washington-Oregon-California coast by four, and Japan and Hawaii each by five of the seven tsunamis of interregional significance. With a policy of issuing a warning uniformly to all coasts of the Pacific whenever a tsunami had been generated, re- gardless of its size in the area of generation or the location of the generating area, the Seismic Sea Wave Warning System must inevitably have continued to issue a large number of seemingly false warnings to most coasts of the Pacific. In Crescent City, Yutzy (1964) found that false alarms had contributed greatly to the lack of response to the warning of the March 1964 Alaska tsunami. Yet Crescent City had been damaged by a runup of 3.7 m from the May 1960 Chile tsunami. How much poorer a response might have been ex- pected in Alaska, which had never had any significant effects from a distant tsunami! IMPROVEMENTS Since the Alaska earthquake, a considerable number of im- provements have been made in the sswws, some resulting from deficiencies disclosed by the performance of the system at the time of the earthquake and some made possible by TABLE 5 Distribution of Known Significant Runups of Tsunamis outside Their Areas of Origin, 1948-1965" 1952 1957 1958 1960 1963 1964 1965 Area November 4 March 9 November 6 May 22 October 13 March 28 February 4 Japan X X X X X _ _ Kuril Islands X - O X 0 - - Kamchatka o - - - - _ - Aleutian Islands X o - X - _ O Alaska - — - - - O - British Columbia - - - - - X - Washington-Oregon-California X X - X - X - Mexico - - - X - X - Ecuador - - - X - - - Peru X - - X - X - Chile X X - 0 - X - Hawaii X X - X - X X Samoa X - - X - - - "Only tsunamis that had runups of 1 m or more outside their regions of origin are listed. O, the region of origin of each tsunami; X, other regions in which the tsunami had a runup of I m or more. 15 238 TSUNAMIS special appropriations for reconstruction after the earthquake. The most important change, as far as Alaska is concerned, is the creation of the Alaska Regional Tsunami Warning Sys- tem, supplementing the original Seismic Sea Wave Warning System and primarily intended to cope with the problem of issuing warnings of tsunamis generated locally in Alaska. The operating center of the new system is at Palmer, Alaska, where a tripartite seismograph array has been constructed. Additional arrays have been constructed at other Alaska seismograph stations, a two-part array at Sitka, and a small tripartite array on Adak Island. All the participating seismo- graph and marigraph stations in Alaska and the Aleutian Islands (except Attu) are now required to report information useful in tsunami warning to the Palmer Observatory as well as to the Honolulu Observatory. This service has been further improved by telemetering seismic and marigraphic data from several Alaska stations to Palmer; this permits rapid determi- nation of earthquake epicenters and early detection of tsu- namis generated in Alaska. Another array, a quadripartite one, has been constructed on the island of Oahu, Hawaii. The records of the array, tele- metered to the Honolulu Observatory, increase the capability of HO for rapid determination of earthquake epicentral location. In the network of seismographic and marigraphic stations reporting to the sswws, 3 new seismographs and 10 new marigraph stations have been added. However, the Canton marigraph station has been withdrawn, owing to the closing of communications facilities on Canton Island. Two addi- tional teletype machines have been installed at the Honolulu Observatory. This added capability in communications should aid in the more rapid receipt and transmission of tsunami information and warning bulletins. The major change, as far as U.S. Pacific coasts other than Alaska's are concerned, is the institution of a new policy of selective cancellation of warnings in which warnings may be kept in force for selected coastal regions and canceled early in other regions where significant tsunami effects are not ex- pected. The idea that regional selectivity in warning was not only needed but was already probably feasible, at least for Hawaii, emerged during a tsunami symposium sponsored by the Panel on Oceanography of the Committee on the Alaska Earthquake and held in Menlo Park in May 1965. The time seemed propitious for a change in policy. The U.S. Coast and Geodetic Survey, with its competence in managing the sswws, and the U.S. Weather Bureau, with its regional ap- proach to forecasting and warning, had just been merged to form the new Environmental Science Services Administration (ESSA). The Governor of the State of Hawaii recommended in June 1965, and more explicitly in September of that year, the creation of a capability for regional evaluation of tsunami risk for Hawaii. The matter was thoroughly discussed by Hawaiian tsunami scientists with ESSA officials through con- ferences and correspondence. In April 1966, the U.S. Coast and Geodetic Survey and Hawaii Civil Defense Division an- nounced a new policy whereby regional evaluation would be involved in the issuance and particularly in the cancellation of tsunami warnings in Hawaii. In May 1966, the U.S. Coast and Geodetic Survey had arranged to supply the necessary competence for regional evaluation from its newly established tsunami research group at the University of Hawaii. In June 1966, the following new specification was issued in the Com- munication Plan for the Warning System (Spaeth, 1962) (Change 18 to Communication Plan for sswws): "A watch will be canceled when HO determines that a wave has not been generated. A warning will be canceled if it is issued on the basis of erroneous data or if HO determines from subse- quent information that only an insignificant wave has been generated. In addition, a warning maybe canceledon a selec- tive basis when a significant wave which has been generated clearly poses no threat to one or more of the areas HO warns, either because of intervening continents or islands which screen them or because the orientation of the generating area causes the tsunami to be directed away from the areas warned." This specification still does not explicitly provide for regional selectivity in the original issuance of warnings; but if a warning can be canceled in some areas immediately after issuance, on the basis that the tsunami poses no threat in those areas, then presumably nonissuance for those areas is implied. The adoption of the principle of regional selectivity in issuance and cancellation of warnings is not of great conse- quence on coastlines served by foreign national tsunami warning systems, because those systems are capable of evalu- ating and free to evaluate for those coasts the warnings of remotely generated tsunamis which the sswws supplies. More important for foreign areas are the improvements in international coordination for which firm plans were first made by a working group established by the Intergovernmen- tal Oceanographic Commission meeting in Honolulu in April 1965 (Stewart, 1965). Most significantly, the group recom- mended the establishment in Honolulu of a permanent Inter- national Tsunami Information Center to augment the services of the Honolulu Observatory and the appointment of an In- ternational Coordinating Group to effect technical liaison, to ensure the due exchange of information, and to coordinate efforts with concerned international bodies. Although ham- pered by a lack of funds, the U.S. Coast and Geodetic Survey began in 1967 the collection and collation of records to initi- ate the International Tsunami Information Center, and a first meeting of the International Coordinating Group was held in the spring of 1968. Perhaps indicative of its undertaking of increased interna- tional responsibilities, the name of the Seismic Sea Wave Warning System was changed in 1967 to one that was more in accord with international technical usage, the Tsunami 16 TECHNICAL EVALUATION OF THE SEISMIC SEA WAVE WARNING SYSTEM 239 Warning System (Change 24 to Communications Plan for sswws). SPECIAL PROBLEMS WITH TSUNAMI WARNING IN ALASKA Some problems of tsunami warning in Alaska deserve special attention, particularly in consideration of the establishment of the Alaska Regional Tsunami Warning System with head- quarters at Palmer. This system was developed explicitly to handle regional analysis for Alaska in connection with warn- ings of tsunamis of local origin. The most serious problems are: (1) how to provide, in time to be of use, warnings of significantly greater reliability than natural warnings afforded by the earthquakes; and (2) how to maintain public confi- dence and public understanding in spite of the low-recurrence frequency of actual hazard. A thorough analysis of these problems would require a much better record of Alaska tsu- namis and their effects than is available, but the historical data at hand are adequate to indicate their severity. The history of tsunamis in Alaska has been examined in detail by Cox and Pararas-Carayannis(1969), and the record of tsunamis generated in Alaska is summarized in Table 6. As noted by these authors, the record was unquestionably quite incomplete prior to the 20th century because of the paucity of civilized settlements. In fact, it approaches adequacy only for the period after 1946 when several new tide gages were established as components in the Seismic Sea Wave Warning System of the U.S. Coast and Geodetic Survey. Tsunamis locally generated by landslides, submarine slumps, and ice- falls in the isolated straits, sounds, and fiords of Alaska, and even major tsunamis tectonically generated off the Aleutian Islands might easily have escaped the historic record. How- ever, there is little reason to believe that tsunamis other than those historically recorded had any significant effects at the few old and continuously occupied settlements. These old settlements include Old Harbor, established in 1784, Kodiak in 1792, Sitka in 1 799, Wrangell in 1834, Juneau in 1880, Cordova about 1890, Valdez in 1898, and Seward in 1902. On 12 (or 13) of the 20 to 23 occasions of Alaska and Aleutian tsunami generation listed in Table 6, the tsunamis were generated off the coast in connection with earthquakes. Seven of these tsunamis were generated off the Aleutian Islands, and the other five off the Alaska mainland. Only half of these offshore tsunamis are known to have had significant runup heights anywhere (equal to or exceeding 1 m), four originating off the Aleutian Islands (1878, 1946, 1957, and 1965) and two off the mainland ( 1 788 and 1964). On the occasions of two of the earthquakes associated with offshore tsunamis (1958 and 1964) and on the occasions of four or perhaps six other earthquakes, tsunamis were gen- erated in straits, sounds, or fiords. On the six occasions when tsunamis were certainly generated locally, they were reported to have had significant runup heights (> 1 m) to distances of a few tens of miles at the most from the points of generation. Some of the tsunamis were tectonically generated (the Yakutat tsunami of 1899, the tectonovolcanic Augustine tsunami of 1883, and the possible tectonovolcanic tsunami of 1901). Others were generated by landslides or submarine slumps (several additional tsunamis in 1899, the Lituya tsu- nami of 1958 that occurred at the same time as a small tec- tonic offshore tsunami, and several tsunamis occurring at the same time as the major offshore tsunami in 1964). On four occasions, significant tsunamis (runup height > 1 m) were generated in fiords by icefalls or by landslides without earthquakes. On none of these occasions was the tsunami of any consequence outside of the fiord in which it was generated. Table 7 contains the record of tsunamis generated else- where for which measurements or observations have been re- ported in Alaska. The records of the 1868 Chilean tsunami were taken from a tide gage established at Kodiak soon after the United States assumed sovereignty over Alaska. Unfor- tunately the records appear to be lost, and the amplitude is not known. All the other records resulted from the modern program of tide gaging. Of the 10 or 1 1 tsunamis recorded in Alaska since the beginning of 1944, only two had amplitudes exceeding 1 m. These, from Kamchatka in 1952 and Chile in 1960, had recorded amplitudes between 1 and 2 m both at Massacre Bay, Attu, and at Sweeper Cove, Adak, but no amplitudes or runups of as much as 1 m were recorded else- where. There is no record of damage from a distant tsunami in either Alaska or the Aleutians. Since the beginning of 1946, 17 or 18 tsunamis, of which six have been of practical consequence, have been observed in Alaska and the Aleutians. Of these six, two were of Aleu- tian origin, two of Alaskan origin, and two of distant origin. This frequency of six significant tsunamis in 22 years is es- sentially the same as that in Hawaii (7 in the same period). However, the combined Aleutian -Alaskan coastline is much too long to treat as a single region, as far as public warning is concerned. A resident of Sitka would surely consider false a warning of a tsunami whose runup exceeded 1 m on no shore closer than Unalaska. In Table 8, the distribution of significant tsunamis-those with runup heights of 1 m or more— is shown in relation to four Alaska-Aleutian coastal segments, each about 1,000 km long. It will be seen that only four earthquakes were accom- panied by tsunamis that had significant effects in more than one of these segments, and none were accompanied by tsu- namis with such effects in more than two segments. The 1 788 tsunami had significant effects on coasts (totaling no more than a few hundred kilometers in length) that happened to span the arbitrary boundary between the eastern Aleutian and western Alaska segments. The 1957 tsunami, originating 17 D. 3 B 3 3 < ii g o 2 5? J3 -o „ LU S X UJ 3t r ili — o - 1j o3 E c "Z -s C 3 3 iC ^ 1 c 3 03 ■a c M aj 03 Wj 3 o3 cr — O ^ -1 > o 2 — T3 c __- E C c c c C 03 3 03 >. T3 C ~ D. x 03 XI a> 03 '-> O -J x a* a. NO ON w x: 00 00 r- c/l C OS 1-2 §-.s o O a -a o a. nO Uj -J < E A S a: o J2 E w < 3 oa c a. 3 >. _2 c -o c c 03 -o c rrl CO 03 3 03 i CJ 3 >. 03 E 03 pa .id "3 03 3 c c 03 ca ■a i 2 03 c C/0 a. M J* o3 3 CD a. o3 X sa 03 03 _2 e3 o3 3 ■a a. 03 3 T3 X 0 Oij CO •5 ■a .* 3 > 3 03 ^ c 3 c M c 0 o 1> 0- c o 03 o 03 03 o :_> < u u. > -J _1 3 a. > U > -J ^ > > z o o. ^ s u x; 1 8 a CO c o c >• CJ O x> ■o i- c x> 3 O « 03 T£ Xt 0> Q- 2 0> .* £ >> 03 •S =5 >- S 2 « 3 « 03 i?5 03 c ° « t; ^ .= 3 o3 o3 E m 03 _03 < C 3 « C J< z B> 3 5 3 3 O U. 8 S 2 -o O ^ u DO a. 3 uj >< a >> s "3 > « « « S «> t: x ^^ o ro 1 — 1 ^O 0> O o s. H. ^ m n « E S I < O n in m " "> ■* oo f^ On ON r- r^ 00 ON ON oo 00 oo OO 00 x> I ? 3 I 2 " H i X) o. x> s u 3 to 5> o ^ Q 4 u. (/■> u. S -^ "rt OO - V) o O O O »-1 * ■o c T)i{ 00 -o c S &« .a 2f-o 5- S™«sqO^-5o«j«3.s Jivi^O^^^viUU^^w oa oa O O I a 3 Q (NVOO rg _, _ (NINMN- 3 K "O Ml So 3 Sa £ X) OS fc > o Q. i 3 fc D. CO J= o Z < < a 00 so Os C7\ r~ m t ■* ■* «-> os os Os OS Os S3 »> w fl a 3 is £ U ij | « s V 00 — E 4= J J a ■2 E s 2 c -o S E £ E si o.2f Si a 3 a « Q 5 241 19 242 TSUNAMIS TABLE 7 Distant Tsunamis Reported Observed in Alaska, 1868-1966 Date 1868 (1883 1944 1952 1952 1953 1956 1958 71958 1959 1960 1963 1966 August 1 3 August 27) December 7 March 4 November 4 November 25 March 30 November 6 November 1 2 May 4 May 22 October 1 3 October 17 Area of Origin Maximum Runup (m)a Place Chile (Indonesia, Krakatoa volcano eruption) Japan Japan Kamchatka Japan Kamchatka Kuril Islands Kuril Islands Kamchatka Chile Kuril Islands Peru 14 to 21 7.5 3.8 4 to 18. 3 7 3 to 4 7 7 20 to 25 4.4 1.1 Arica Kumanoura Kiritappu Paramushir Island Choshi Shikotan Isla Mocha Urup Callao Maximum Runup and all Runups > 1 m in Alaska (m) Runup Small Small 0.3 0.2 1.5 1.1 0.2 0.3 0.2 0.2 0.2 1.5 1.8 0.4 0.1 Place Kodiak Kodiak6 Massacre Bay, Massacre Bay, Massacre Bay, Sweeper Cove Massacre Bay, Massacre Bay, Massacre Bay, Massacre Bay, Massacre Bay, Sweeper Cove Massacre Bay, Massacre Bay, Massacre Bay, Attu Attu Attu Adak Attu Attu Attu Attu Attu Adak Attu Attu Attu Runup is measured above contemporaneous sea level. Small waves observed at Kodiak were generated by atmospheric-pressure disturbances and did not represent the tsunami that caused damage in the Sunda Strait. in the Andreanof Islands, was reported to have had a 1 2-m runup at Scotch Cap, Unalaska, although this seems question- able. There were several separate tsunamis generated in inlets as accompaniments of the 1899 Yakutat earthquake. Only the major tsunami accompanying the 1964 earthquake had significant effects certainly extending across major parts of of two coastal segments-all the way from the Trinity Islands in the south central Alaska segment to the southern bound- ary of Alaska (and far south of that). As with the local tsunamis accompanying the 1964 earth- quake, which were disastrous at Valdez, Whittier, Chenega, and Seward, the effects of tsunamis accompanying the 1899 Yakutat Bay earthquake were practically restricted to the in- dividual bays, Yakutat Bay, Lituya Bay, and Valdez Arm. The tsunami accompanying the 1958 Lituya Bay earthquake, three other tsunamis generated by landslides in Lituya Bay, and two tsunamis generated by icefalls in Yakutat Bay were restricted to the bays in which they originated. What may be deduced concerning recurrence intervals for significant tsunamis on the four coastal segments and at settlements with long histories may be summarized as follows. (It will be recognized that the probable error of most of the estimates is very great.) Western Aleutian Islands: Two tsunamis generated locally offshore and two distant tsunamis of significance have oc- curred in the last 20± years, that is, the average recurrence interval seems to be about 5 years. No damage has been re- corded. Eastern Aleutian Islands: One or two significant and dam- aging tsunamis locally generated offshore but no significant distant tsunamis have occurred in the last 20± years, that is, the recurrence interval seems to be on the order of 10 to 20 years. (This is probably more reliable than the 45- to 50-year recurrence interval indicated by the record since 1788.) At Unalaska, no damage or casualties have occurred in 192 years. South central Alaska: In the last 20± years, only one significant and damaging tsunami of local offshore origin has occurred, that of 1964. On the same occasion, several separate tsunamis generated in inlets had significant effects confined to those inlets. Other tsunamis generated in in- lets occurred 69 and 84 years ago. A distant tsunami 100 years ago had observable effects but, so far as is known, did no damage. The only other«significant and damaging tsunami of local offshore origin occurred 179 years ago. The recur- rence interval of damaging offshore and distant tsunamis ap- pears to be at least 20 years and perhaps as much as 90 years. At Kodiak, only one tsunami causing damage and casualties has occurred in 184 years. At Valdez, two damaging or po- tentially damaging tsunamis have been locally generated in less than 70 years. One damaging tsunami has occurred in the 78-year record at Cordova. Southeastern Alaska: Only one offshore or distant tsunami has occurred in the last 20± years or in the historical record. However, one damaging tsunami was generated in Lituya Bay in the period of 20± years and at least four damaging or po- tentially damaging tsunamis have occurred in the last 1 15 years, suggesting a recurrence frequency of less than 30 years. Local tsunamis have similarly affected Yakutat Bay, with an average recurrence interval of less than 45 years. Wrangell has 20 TECHNICAL EVALUATION OK THE SEISMIC SEA WAVE WARNING SYSTEM 243 TABLE 8 Distribution of Tsunami Runup Heights of 1 Meter or More in Alaska and Aleutian Islands, 1788-1965 Western Eastern Southcentral Southeastern Aleutians Aleutians Alaska Alaska (Near Islands to (Fox Islands to (Alaska Peninsula to (Cape St. Elias Date Andreanof Islands) Shumagin Islands) Prince William Sound) to Ketchikan) 1788 July 27 - 0 () - 1845 - - - 1 1853- 1854 - - - I 1868 August 1 3 - - D - 1874 7 - - - 1 1878 August 29 - O - - 1883 October 6 - - 1 - 1899 September 10 - - 1 I 1905 July 4 - - - 1 1936 October 27 - - - 1 1946 April 1 - o - - 1952 November 4 D - - - 1957 March 9 O 0^ - - 1958 July 10 - - - 1 1960 May 22 D - - - 1964 March 28 - - 01 0 1965 February 4 O - — - I, tsunami or tsunamis generated in inlet, fiord, or strait in Alaska or the Aleutians; O, tsunamis generated offshore in Alaska or the Aleutians; D, distant tsunami. not experienced tsunami damage or casualties in 133 years. At Sitka a dock collapsed in the 1964 tsunami, but otherwise there have been no tsunami damage or casualties in 163 years. Very long recurrence intervals for the tsunamis generated by vertical crustal displacements associated with the largest earthquakes are suggested by the record, in the stratigraphy of the Copper River Delta and the terrace history of Middle- ton Island, of but three major sudden uplifts during the last 1 ,700 years (Reimnitz, this volume). Considering the difficulties in maintaining public under- standing in Hawaii, where there are considerably higher re- currence frequencies of significant tsunamis, the region to be served by the recently established Alaska Regional Tsunami Warning System should expect even greater difficulties be- cause of the lower recurrence frequencies in Alaska, as in- dicated above. Further, to be successful, the Tsunami Warning System will have to provide warnings more reliable than those pro- vided naturally by the public's direct observation of the earthquakes themselves. Although the probability of tsunami generation by crustal displacement may in some cases be better indicated by the Warning System's epicentral location and magnitude determination than by the simple observation of the earthquake in coastal areas, it must be recognized that, at least for a major earthquake like that of 1964, the epicen- tral location may be quite misleading as to the probability of tsunami generation. Further, the epicentral location would give no indication of the extent of the area in which tsunamis might be locally generated by submarine slumps. In particular, tsunamis generated by slumps might very well cause serious inundation within minutes after the earthquake (as was the case in 1964), far too soon for a warning to be formulated and disseminated by the Warning System. In addition, as demonstrated in 1964, there is a high probability that damage caused by a large earthquake will seriously interfere with the communications required by the system for the receipt of in- formation and for the dissemination of warnings. The observation of earthquakes is, of course, of no value in warning of tsunamis that occur independently of earth- quakes for example, the tsunamis locally generated by land- slides, as in Lituya Bay, or by icefalls, as in Yakutat Bay. However, the warning system will be of no value in these cases either, because the system is normally triggered by earthquake recording. It would seem that the practical feasibility of operation of a regional tsunami warning system for Alaska should be closely re-examined in the light of the low-recurrence fre- quencies of significant tsunamis in any of the Alaskan or Aleutian coastal segments and the difficulties of producing warnings more reliable than those provided by the simple observation of earthquakes. CONCLUSIONS In the light of the objective of issuing timely warnings of im- pending tsunamis, the operation of the Seismic Sea Wave Warning System in relation to the tsunami associated with 21 244 TSUNAMIS the Alaska earthquake was an unqualified success. However, some qualifications should be made in the light of the perhaps more appropriate objective of minimizing the hazards of tsu- namis, especially hazards to human life and health, through the issuance of warnings for dissemination by cooperating agencies. As far as the operation of the system at the time of the Alaska earthquake is concerned, the principal deficiency was in the time of the issuance of the warning and earlier bulletins. Considering the system's requirement of mari- graphic confirmation before issuance of a warning, no warn- ings could have been issued in time to be of value in the macroseismic area of the earthquake. Although warnings for the coasts of Washington, Oregon, and California were some- what delayed, they reached residents on those coasts an hour and a half or more before arrival of the highest crests of the major tsunami. The major deficiency in the effectiveness of the system arose not through its operations at the time of the Alaska earthquake, but from confusion and lack of confidence (stemming from earlier operations) on the part of those re- ceiving the warnings. To speed the seismological analysis with which the warn- ing process usually begins, maximum use should be made of seismographic information available at the Honolulu Ob- servatory. The quadripartite seismographic installation, es- tablished on Oahu by ESSA since the earthquake, should be of considerable assistance, especially if and when arrange- ments can be made to analyze its output by computer. How- ever, the utility of simple equipment, especially visibly recording seismographs of low magnification, for speedy determination of the most essential data, should not be overlooked. In connection with the Alaska tsunami warning, there was some confusion in marigraphic data; in the history of the Warning System, faulty marigraphic data have been a major contributor to false alarms. Since 1964, the U.S. Coast and Geodetic Survey has taken a number of steps to increase the speed of transmission, reliability, and significance of mari- graphic information by installation of additional sensors on exposed coastlines; arrangement for remote recording from some sensors to communication centers; arrangement for long-distance telemetering in the Alaska Regional System; and general improvement and simplification of communica- tions. Further improvement in each of these areas should be promoted. On-call telemetering direct to the Honolulu Ob- servatory should be arranged wherever feasible. Parameters and units used in transmitting marigraphic information should be standardized or explicitly indicated. The vulner- ability of the communications system to earthquake damage, clearly demonstrated during the Alaska earthquake, must be taken into account in estimating the reliance to be placed on marigraphic data from the macroseismic zone of an earth- quake. The development of midocean tsunami gages, which has been under way for several years, should be pushed to com- pletion as rapidly as possible; installation of such gages in midocean makes it possible for them to record tsunamis un- distorted by coastal effects. To avoid errors in estimated times of arrival of a major tsunami generated on the Continental Shelf, such as occurred in the case of the Alaska tsunami, propagation should be as- sumed from the edge of a hypothetical generating area whose size depends on the earthquake magnitude and whose location is suggested by the local geologic structure considered in re- lation to the epicenter of the earthquake. The policy whereby regional evaluation of the risk from a tsunami is involved in the issuance of tsunami warnings in Hawaii will assist greatly in the elimination of the seemingly false alarms that have contributed so heavily to lack of public confidence in the warning system. Presumably the Palmer Observatory will serve the same function with respect to regional analysis for Alaska that the Honolulu Observatory does for Hawaii. Provision must still be made, however, for the analyses of regional risk that should be made before warn- ings are issued in Washington, Oregon, or California. In addition, it must be recognized that the existence of a certain probability of tsunami arrival in a region is not in it- self a sufficient basis for the issuance of a warning. The hazard from the tsunami must at least be sufficient basis for the is- suance of a warning. The hazard from the tsunami must at least be sufficient to balance the unquestionable hazards that stem from the warning operations, including the future jeop- ardy that may occur through the destruction of confidence if the warning later appears to have been unjustified. Logi- cally the whole Tsunami Warning System, as it is now known, should be subjected to an economic justification. The bene- fits of the system are clearly the values associated with the reduction of casualties together with the reduction of damage to movable property. The costs include those of the equip- ment and operation of the official parts of the system with the seismic and marigraphic detection and analysis, com- munications, and decision-making component, as well as the public dissemination and evacuation-policing component. They also include the cost to the public of the disruption of normal activities during warning periods, whether or not the damaging waves arrive. This includes the costs of accidents attributable to the disruption. The issuance of a warning should involve a similar analysis or at least estimation of (1) the values associated with the probable reduction in casualties to be achieved on the im- mediate occasion, and (2) the costs both of the potential warning and of the future jeopardy that such a warning might induce. In Alaska, the low frequency of tsunami recurrence makes the successful operation of a formal warning system appear very difficult, both from the standpoint of economics and 22 TECHNICAL EVALUATION Oh THE SEISMIC SEA WAVE WARNING SYSTEM 245 from that of maintaining public confidence. In order to be judged successful, the formal system would have to provide warnings more reliable than those naturally provided by the direct observation of the earthquakes by the public. The vulnerability of communications systems in the macroseismic area is a contributory problem. It is important to guard against the rejection of authoriza- tion or support for those components of the Tsunami Warn- ing System that are effective and economically feasible, while at the same time withholding authorization for other com- ponents that are not effective. REFERENCES Anderson, William A., 1970. Tsunami warning in Crescent City, Cali- fornia, and Hilo, Hawaii in The Great Alaska Earthquake of 1964: Human Ecology. NAS Pub. 1607. Washington: National Academy of Sciences, p. 116-124. Cox, Doak C, 1964. Tsunami forecasting. Report HIG-43. Hono- lulu: University of Hawaii, Institute of Geophysics. 15 p. Cox, Doak C, 1968. Performance of the Seismic Sea Wave Warning System, 1948-1967. Report HIG-68-2. Honolulu: University of Hawaii, Institute of Geophysics, March 25. 79 p. Cox, Doak C, and George Pararas-Carayannis, 1969. Catalog of tsu- namis in Alaska. World Data Center A: Tsunami. Washington: Environmental Science Services Administration, U.S. Coast and Geodetic Survey, May. 39 p. Finch, R. H., 1924. On the prediction of tidal waves. Monthly Weather Review (V.S. Weather Bureau), v. 52, p. 147-148. lida, Kumizi, Doak C. Cox, and George Pararas-Carayannis, 1967. Preliminary catalog of tsunamis occurring in the Pacific Ocean. Data Report No. 5, HIG-67-10. Honolulu: University of Hawaii, Institute of Geophysics, August. 270 p. Keutschenreuter, P. H. (Chairman), 1965. A proposed nation-wide disaster warning system. Report of Natural Disaster Warning Survey Group. Washington: Environmental Science Services Administration, October. 1 1 3 p. Norton, Frank R. B., and J. Eugene Haas, 1970. The human response in selected communities in The Great Alaska Earthquake of 1964: Human Ecology. NAS Pub. 1607. Washington: National Academy of Sciences, p. 248-399. Reimnitz, Erk, 1972. Effects in the Copper River Delta in The Great Alaska Earthquake of 1964: Oceanography and Coastal Engineer- ing. NAS Pub. 1605. Washington: National Academy of Sciences. Roberts, Elliott B., 1950. A seismic sea wave warning system for the Pacific. Journal of the Coast and Geodetic Survey, v. 3, p. 74-79. Spaeth, Mark G., 1962. Communication plan for Seismic Sea Wave Warning System. Journal of the Coast and Geodetic Survey, 3 (November), 1-85. Spaeth, Mark G., and Saul C. Berkman, 1967. The tsunami of March 28, 1964, as recorded at tide stations. Environmental Science Services Administration Technical Report C&GS 33. Washington: Govern- ment Printing Office. 86 p. Also in The Great Alaska Earthquake of 1964: Oceanography and Coastal Engineering. NAS Pub. 1605. Washington: National Academy of Sciences, 1972. Stewart, Harris B., Jr. (Chairman), 1965. Report of the working group meeting on the International Aspects of the Tsunami Warn- ing System in the Pacific. Honolulu: Intergovernmental Ocean- ographic Commission, April. 17 p. Weller, Jack M., 1972. Human response to tsunami warnings in The Great Alaska Earthquake of 1964: Oceanography and Coastal Engineering. NAS Pub. 1605. Washington: National Academy of Sciences. Yutzy, Daniel, 1964. Aesop 1964: Contingencies affecting the issuing of public disaster warnings at Crescent City, California. Disaster Research Center Research Note 4. Columbus: The Ohio State University, Disaster Research Center. 8 p. 23 3 Reprinted from Proceedings of the Royal Society of Edinburgh, Section B, 72, No. 20, 223-229. 20. — A Recently Discovered Challenger Sketchbook. By J. Welles Henderson, Philadelphia Maritime Museum, Pennsylvania, USA and Harris B. Stewart, Jr., NOAA Atlantic Oceanographic and Meteorological Laboratories, Miami, Florida, USA. (With 5 plates) Synopsis In 1968 the senior author discovered in a Boston, Massachusetts, bookstore a sketchbook of some 33 watercolours of the Challenger Expedition painted by Benjamin Shephard, a cooper assigned to the Challenger on November 15, 1872, at the age of 31. The paintings were all accomplished between the ship's departure from Gibraltar on January 26, 1873, and the completion of her February 1874 work in the Antarctic before departing for Melbourne, Australia. Shephard's paintings are elliptical in shape and measure 9-75 by 5-5 inches, each within a painted 'frame' resembling a leather belt with a 'buckle' at the lower left. The paintings themselves are skilfully done and each shows the Challenger at a different location during the voyage. Since all but one of the paintings were dated by the artist, it has been possible to go back to the official narrative of the expedition and to the published logs kept by the Assistant Engineer, W. J. J. Spry, and by Sub-Lieutenant Lord George G. Campbell to determine just what was happening aboard at the time each painting was made. Excerpts from these logs provide the narrative accompany- ing the sketches which are to be published in book form to coincide with the Challenger Expedition Centenary celebrations. Black and white photographs of the sketchbook's cover and of four of the 33 sketches are included as part of this paper (Plates 1-5). The discovery of the Challenger sketchbook in a bookshop in Boston, Massa- chusetts, by J. Welles Henderson in 1968 adds a delightful artistic postscript to the volumes already written about what is still considered the greatest of all oceanographic expeditions. Drawn between January 1873 and February 1874, these 33 pen-and-ink drawings with added watercolour cover only one-third of the total expedition. The artist, Benjamin Shephard, stayed aboard for the remaining two years of the expedi- tion, and it is hoped that he was able to continue his watercolours and that additional sketchbooks are lying unnoticed waiting to be found and appreciated for what they are. Like the sailors on most oceanographic expeditions, those aboard the H.M.S. Challenger, although intrigued by the work of the scientists, were more interested in the ports which punctuated the long periods of observations at sea. Thus Shephard, with few exceptions, concentrated on painting not the scientific work at sea but rather the Challenger at her various ports of call. Starting with one showing the ship's departure from Gibraltar on January 26, 1873, Shephard's watercolours cover St Thomas, Bermuda, Halifax, St Michael's, St Vincent, St Paul's Rocks, Fernando Noronha, Tristan da Cunha, Capetown, Prince Edward Island, Crozet Island, Kerguelen Island and McDonald Island. In addition, there is a delightful and fanciful dredging scene, but this, plus one of a storm and six reflecting his fascination with icebergs, are the only ones of his 33 not related to specific landfalls. In reading through two of the logs maintained during the expedition to try to obtain some background for each of the paintings, it was soon realised that the journals 24 224 J. Welles Henderson and Harris B. Stewart, Jr. kept by men aboard the H.M.S. Challenger during the trip provided much more of the true flavour of the expedition than could ever be recaptured almost one hundred years later. They were even better than the official narrative of the expedition (Thomson and Murray 1885). With this as the rationale, this paper shamelessly borrows from the extremely informative and well- written journals of the Assistant Engineer, W. J. J. Spry (Spry 1877), and of Sub-Lieutenant Lord George G. Campbell (Campbell 1877). Any student of oceanography or of the history of science in general would do well to read these two accounts of a great expedition. They supplement each other nicely, one going into great detail of incidents on the expedition which the other brushes over only casually or omits altogether. The descriptions of Spry and Lord Campbell provide a delightful contemporary descriptive background to most of the water- colours, for as members of the officer-crew complement of the Challenger they, like the artist, were more interested in the ports than in the scientific work aboard. In all fairness, however, it must be added that in both journals the authors exhibit a com- mendable familiarity with and respect for the scientific activities aboard. Shephard's renditions of the ship itself, are extremely accurate with the single exception that — as a true man of sail — he steadfastly refuses to show in the majority of his paintings that the Challenger had a stack. Although she was a spardecked corvette of 2306 tons displacement, she did have a 1234-horsepower engine that was used from time to time during the expedition. Generally, when on-station, she would drop her canvas and stay head-on into the seas, maintaining station with power. Shephard, however, refused to make this concession to progress, and few of his paintings even suggest that this magnificent sailing-ship was also capable of using coal as a means of propulsion. Other than in the paintings themselves, Shephard had a delightful disregard for accuracy. The English vessel towed by the Challenger into St Thomas on March 23, 1873, for example, was named the Varuna. Shephard, however, not only called her the 'Baruner' in the title of one painting, but he missed the date by four days and even spelled the name of the island incorrectly (St Thomass). To his casual approach to the accuracy of dates and spelling must also be added a rare dedication to the doctrine of inconsistency. Three of his sketches depict the ship's activities at Kerguelen Island, yet in no two of them does he spell the island's name the same way. It is 'Kerguelen' in sketch No. 24, 'Kerguelan' in No. 25, and 'Kergulen' in sketch No. 26 (PI. 4). These, however, are the comments of an incurable nit-picker. Shephard's water- colours are magnificently done and extremely well preserved. They provide to oceanographers a new feel for and insight into the Challenger Expedition. This expedition, some one hundred years later, is still regarded as the great marine science expedition of all time. Through Shephard's paintings, today's oceanographers will realise that the Challenger Expedition was not too different from those of today. We feel closer to Sir Charles Wyville Thomson, Moseley, Murray and the others of the scientific party and to Captain George S. Nares, his officers and crew because of Shephard's translation of his love for the sea, a ship, and an expedition into a series of watercolours that have been lost and now are found. For some unknown reason, the voluminous official documentation on the Challenger Expedition nowhere lists the crew members by name and rate. Scientists and ship's officers are well known and referred to regularly but the members of the crew appear A Recently Discovered Challenger Sketchbook 225 consigned to anonymity. For this reason, it was extremely difficult to discover if one, B. Shephard (PL 1) was even aboard the Challenger. No pertinent records could be found in the United States. Through the great assistance of Dr J. D. H. Wiseman of the British Museum, the first glimmer of hope was seen when he discovered in the Murray Library — the collection of John Murray's papers relating to the expedition — a pencilled note to one 'Shepherd', spelled with an 'e', in connection with some lithographic plates. Shortly thereafter, Commander D. P. D. Scott, M.B.E. (R.N.), of the British Hydrographic Department, who had also been working on the problem, wrote to report the complete success of his research with the files of Naval Service Certificates in the Public Record Office in London. Commander Scott has graciously had copies made of some seven sheets of official records all relating to the service of Benjamin Shephard (also spelled with an 'e' in one place in these records) who did indeed serve on the Challenger. Through these records and Commander Scott's additional research, the following information has been compiled about the artist. Benjamin Shephard was born September 18, 1841, at Brixton in Surrey. He entered service on May 9, 1862, at the age of 20, and when he signed up for a ten-year period of service the following October, the records show that he stood 5 feet 7 inches, was of fair complexion, with brown hair and blue eyes. He entered as a cooper and two years later was promoted to coopers' crew. Evidently he found work not particularly to his liking, as he was disrated and promoted several times during his 25-year career. His final rate, however, was still cooper. On October 28, 1867, he deserted but returned the following October 3, to sign up again for ten years. His first assignment was a 15-week gaol sentence for desertion, but from then on, for the rest of his career, he served on many ships as cooper or coopers' crew. Shephard was assigned to H.M.S. Challenger on November 15, 1872, at the age of 31, as a cooper. Again he was promoted to coopers' crew on December 1, 1874, but was demoted back to cooper again on June 1, 1875, while the Challenger was in Japan. The expedition returned to England the last week of May 1876, and Shephard was transferred to H.M.S. Pembroke on June 12, 1876. In succession he served on H.M.S. Pembroke (June to August 1876), H.M.S. Penelope (August 1876 to March 1879), H.M.S. Pembroke again (March to October 1879), H.M.S. Cornus (October 1879 to March 1884), back to H.M.S. Pembroke (March 1884 to January 1885), H.M.S. Tyne (January 1885 to June 1887) and finally H.M.S. Nelson. He transferred from H.M.S. Tyne to H.M.S. Nelson on June 20, 1887, at Albany, W. Australia, where he died of phthisis (pulmonary tuberculosis) three days later at the age of 45. Although not numbered by the artist, the sketches are in a bound sketchbook and in dated chronological order, so the present authors have assigned sequential numbers to each for ease in discussion. Sketch No. 6 (PL 2) shows the Challenger 'leaving Halifax, Nova Scotia, May 19th, 1873, and H.M.S. Royal Alfred cheering us out'. After making an oceanographic section across the Gulf Stream between Bermuda and New York (the Challenger changed course for Halifax some 100 miles off Long Island), she sounded and dredged her way up the coast, crossing 'curious veins of warm and cold water' (Campbell 1877, 22), to arrive at Halifax, Nova Scotia, in the afternoon of May 8, 1873. Evidently the crew was glad to be ashore where the countryside was more like PROC R.S.E. (B) Vol. 72. 1972. 15 26 226 /. Welles Henderson and Harris B. Stewart, Jr. their homeland and where the inhabitants greeted them most warmly. Spry described Halifax and Dartmouth in considerable detail and with great affection. Even as today's oceanographic ships are often opened for inspection at foreign ports and local scientists invited aboard, so too was the Challenger at Halifax. Spry concludes his description of the ship's stay at Halifax thus (Spry 1877, 74-75): 'During our stay, as we lay alongside the Naval Yard every facility was afforded our Halifax friends to visit the ship. Many availed themselves of the opportunity, and evinced the greatest desire to see and examine the many submarine wonders that had up to this date been collected. 'The members of the Halifax Institute of Natural Science mustered in strong numbers, and appeared to take a special interest in the work already accomplished. 'The blind crustacean zoophytes, the varieties of rare and new forms of corals and sponges, were well scanned ; while for the geologists, amongst other things attracting their attention, was a large boulder, which had been brought up in the dredge some 300 miles south of the coast. This was carefully examined, and eventually recognized as a piece of Shelburne granite, which perhaps was carried off to sea in long past ages, on an iceberg detached from the coast glacier of Nova Scotia, and deposited where we had found it, to be again recovered after such a lapse of time, and to help the solution of the glacial theory, according to which, at one time, ice held Nova Scotia in as close an embrace as it does Iceland and Greenland at the present. 'The weather had not been of the best; cold winds, with occasional snow and rain, greeted us during the time at our disposal here; yet we would fain have made a longer stay amongst such kind friends, of whom it is a pleasure to speak. There was a goodness and cordiality with their hospitality and warmheartedness that can never be forgotten by those who knew them. 'On the 19th of May, we steamed out of the harbour, and before nightfall the coast was out of sight.' There follow the fine sketches of the Challenger in a gale, at St Michael's, two at St Vincent, three at St Paul's Rocks, and one at Fernando Noronha. Sketch No. 15 (PI. 3) shows the Challenger off the island of Tristan da Cunha in the South Atlantic. Twenty days after leaving Bahia, Brazil, the Challenger arrived at Tristan da Cunha on October 15, even though Shephard's title says October 14. They were well along on their Transatlantic profile, and it had been a good run down from Bahia. This painting is the only one showing dolphins playing ahead of the ship, and they may reflect the general feeling of pleasurable comfort now that the ship was in cooler latitudes. A section was now commenced across the Atlantic to the Cape of Good Hope. 'When clear of the land [Brazil, Ed.], sail was made, and with a pleasant breeze, we raced on into cooler and healthier latitudes. It had been intended to sight and make a short stay off the island of Trinidad, a rocky and barren spot, surrounded with a dangerous shore of almost unapproachable, sharp, rugged rock, over which generally a rough and turbulent surf breaks, affording security to innumerable sea-birds, for whose refuge it seems expressly formed. 'Owing, however, to unfavourable winds and other causes, we were unable to get nearer than 300 miles; so our course was altered for Tristan d'Acunha. During our passage the usual programme of sounding and trawling was carried out when opportunities offered. The ocean seems teeming with animated organisms. The A Recently Discovered Challenger Sketchbook 227 drift nets, which are always trailing behind us, get filled in a short time with immense numbers of little lively creatures, pretty-looking red and blue cockles, sea-nettle, and various other inhabitants of the deep, many of the most minute size and delicate form and tint. 'In the work-room is disclosed, by aid of the microscope, to the observer, an entirely new world in the economy of nature as displayed in animal life from the surface of the sea. 'On the 6th October, in lat. 30° south, we picked up the commencement of the "westerlies", and by their influence we made short work of the 900 miles still separating us from the islands. On the morning of the 15th, land was in sight, a little speck, at first rising up dark and rugged out of the sea, growing larger and larger as we neared, terminating at length in a huge conical peak some 8000 feet in height, covered with snow' (Spry 1877, 92-94). After some 20 days at sea, the magnificent snow-capped peak of Tristan da Cunha must have been a welcome sight, and Shephard has done it justice in this fine painting of the island. Shephard continued to paint, and his sketches record her at Inaccessible Island and at Nightingale Island (both near Tristan da Cunha), at Simons Bay and Table Bay at the Cape of Good Hope, and then on south towards the Antarctic with paint- ings at Marion Island, the Crozet Islands in the south-western Indian Ocean at about 45°S, and three at Kerguelen Island at 50°S. On Shephard's three fine paintings of the Challenger at Kerguelen, each title has a different spelling of the island's name. One of these (PI. 4), however, has the distinction of having also the wrong date. The Challenger left Christmas Harbour on January 31, and it was not until the following day, February 1, that she reached the southernmost tip of Kerguelen Island, a point which the Expedition named Cape Challenger. These minor inaccuracies of date and spelling are, however, magnificiently unim- portant, for the painting itself is a beauty. Shephard shows the Challenger in rough seas and strong winds moving along under only her fore and main topsails, the foresail and mainsail being furled, and the others not yet in evidence. As suggested by one painting of the ship tied up to Saint Paul's Rocks (No. 12), Shephard and the official expedition artist, J. J. Wild, undoubtedly worked together from time to time or both worked from the same photograph. Wild's engraving of the Challenger off Cape Challenger appears in the Thomson and Murray official narrative of the expedition (1885, 1, 343). In comparing the two, one is immediately struck by the fact that the peaks of Cape Challenger are exactly the same in the two — every crag and pinnacle is in exactly the same position in both. However, in the Wild version, the Challenger is on the opposite tack, and the ship is headed towards the right. From the detailed track chart of the expedition during this particular section (Thomson and Murray 1885, sheet 21, facing page 337), it would appear that it is Shephard in this case who is right, for the direction in which she is sailing is indeed the course on which she passed Cape Challenger on that stormy afternoon in 1874. Campbell tells of the passage this way (Campbell 1877, 105): 'Feb. 1.— Steaming and sailing along the south shore to "fix" the southern point of the island, which having done (calling it Cape Challenger), we made sail and 28 228 /. Welles Henderson and Harris B. Stewart, Jn. shaped our course for Heard Island. We had a fine view of the southern coast and mountain ranges in the evening, the highest mountain of which was calculated by our surveyors to be 6,180 feet high, and called by us Mount Ross, after Captain Sir John Ross.' This naming of newly discovered or newly described geographic features for famous people or for one's friends was as common on the Challenger Expedition as it is today on oceanographic expeditions or in Antarctica. For today's oceanographers, however, about the only previously undiscovered features remaining are seamounts, and these are generally given the name of the ship that first found them. However, the far-ranging oceanographic ships of today such as the Vema, Atlantis, Melville, Discoverer, Researcher, Meteor, Discovery and the many Soviet oceanographic ships discover so many seamounts that personal names are often used, although the Board on Geographic Names prefers that such names be limited to those already dead. The Challenger, however, had no such restrictions, and on Kerguelen they named the Wyville Thompson Range after the Chief Scientist of the expedition, Mount Campbell after Sub-Lieutenant Lord George G. Campbell, Aldrich Channel after Lieutenant Pelham Aldrich, the surveyor and magnetics man aboard, and Mount Tizard after Commander Tizard who was in charge of the navigation and hydrography. The name of the ship itself was reserved for the southernmost cape on Kerguelen, and it is still known as Cape Challenger. After one painting of the ship at MacDonald or Heard Island, Shephard encountered his first iceberg, or 'ice burgh' as he spelled it, on February 11, 1874, at 60°30'S. For the next three weeks they were almost constantly with floating ice, and Shephard was evidently fascinated with it, for he did some six sketches showing the Challenger among the icebergs. The final sketch in the book (No. 33) is shown as PI. 5, and Shephard has titled it 'H.M.S. Challenger Dredging Inside Antarctic Curcle, Feby. 1874'. It is impossible to determine just which dredging operation Shephard is depicting here. Actually, the Challenger did no dredging during her half-day excursion below the Antarctic Circle (February 16, 1874). Spry gave an account of the ship's short penetration below the Antarctic Circle (Spry 1877, 136-137): 'Feb. 16th — The weather was remarkably fine, such as is but seldom experienced in these high latitudes — bright sun and blue sky, with but little wind; so had recourse to steam, passing some magnificent ice bergs, extending in all directions and in every conceivable shape and form; for the most part having flat tops covered with snow, glistening in the sun, with smooth, inaccessible sides, beautifully tinted with every shade of blue and green. It was about 1.30 p.m. when we crossed the barrier of the Antarctic Circle (latitude 66°30' south), in longitude 78° east, situated about 1400 miles from the South Pole. The sight was indeed a grand one as we threaded our way through the pack ice and up through avenues of vast bergs, over a course never before taken by explorers ; all this left an impression of those icy desolate regions that can never be forgotten. It seems most difficult to attempt a description, for all I could say would convey but little of the reality to the imagination of one who has not been similarly situated. Proceeding on to latitude 66°40' south, the course was altered, and the horizon scanned in all directions for land; the weather was unusually clear, so that we should certainly have seen it had any existed within 29 Henderson and Stewari Plate 1 U £ rfji i Henderson ani> Stewart Plate 2 2 -3 > 0 I "I s G/j i c 31 Henderson and Stewart Plate 3 a "O u c o c o S. 03 O J5 O 3 u •o H te o I a ± <— o so c 1 V) 32 Henderson and Stewart Plate 4 c- u I •J C C BO a c "3 ji jjG so Si Below: the NOAA ship DISCOVERER. training centers in Colombia, Mexico, Argentina, and Venezuela. These centers train marine scien- tists, engineers, and technicians from many Latin American countries, with particular emphasis on marine sciences instrumentation. The work of these centers will assist in the intercallbration of marine sciences instrumentation and the standardization of data formats and reporting; it should improve the capabilities of many Latin American countries par- ticipating in the CICAR program. In addition to the AID effort, the University of Rhode Island has provided some excess gear to the Univer- sity of the West Indies in Trinidad, and the NOAA National Ocean Survey has loaned a shallow-water echo sounder to the Mexican Research Vessel , URIBE. These developments have helped, but the need for education and training at sea still remain unsatisfied. It was primarily to meet this particular need that the National Oceanic and Atmospheric Administration ( NOAA) conceived the idea of NOA A-C ARIB. Although the CARIB part of the expedition title can be considered as being derived either from the name ot tne sea or from the ancestral inhabitants of the area, it also stands for Clear Area Research and Instruction Break- through. Actually, the concept does represent a real breakthrough in the running of international coopera- tive expeditions. eontimti 37 Briefly, the rationale is this: To work with the CICAR National Coordinators in several CICAR coun- tries to develop joint research projects that would reflect the marine research interests of each of the countries concerned. Arrangements would be made to place six or eight scientists from CICAR country "X" aboard a United States oceanographic research ship to work with their United States counterparts in carrying out marine research projects of interest to country "X" as well as the United States. At the end of the project (about 1 week), the ship would put into a major port of country "X" and make a 1-day training and demonstration cruise, on which she would carry as many students, faculty members, administrators, government officials, local press representatives, etc. , as she could accomodate for 1 day. On this trip, scientists from country "X" who had been aboard for the previous week could act as instructors for the 1-day cruise participants. Not only would they speak their language, but they would also by then know the ship's capabilities, be more familiar with local training needs, and could tailor the teaching accordingly. If properly planned, such an operation could contribute valuable research infor- mation. With this concept as the basic framework for NOAA- CARIB, shiptime aboard the NOA A ship DISCOVERER was allotted to the program by NOAA's Atlantic ocea- nographic and Meteorological Laboratories in Miami, Fla. The scheduled time includes the 72 days between October 5, when the ship will sail from Miami, and December 15, 1972, when she will return. The pre- sent itinerary will take the DISCOVERER from Miami to Veracruz, Mexico; Kingston, Jamaica; San Juan, Puerto Rico; Port of Spain; Trinidad; Cumana-, Vene- zuela; Cartagena; Colombia; andbackto Miami. The plan is that some 7 to 10 scientists from each country will join the ship at the last port prior to the stop in their own country , and will carry out their own projects aboard enroute to their own port for the one-day trip. Several NOAA projects will also be carried out simul- taneously. Planning is already well underway for NOAA- CARIB, and the contacts in each of the six countries tobe visitedhave been established. If marine scien- tists in any of these countries have projects that they would I ike to have considered, they should submit their plans to their local NOAA-CARIB representative. (Names and addresses are listed following this article). The October-phase of the operation will coincide with CICAR Survey Month-Ill, a third quasi-synoptic survey, and, wherever possible, the standard CICAR sections that have been established throughout the region will be occupied for physical and biological observations. NOAA-CARIB thus has two main goals: to provide opportunities for education and training at sea for CICAR country nationals and to contribute to the overall CICAR scientific program. A secondary ob- jective is to provide, through scientist-to-scien- tist contacts, the mechanisms for continuing cooper- ative oceanographic activities in the Caribbean long after the formal CICAR program has terminated. As the details of the various projects take shape, it is NOAA's intention to report them in this publi- cation, so that all involved in Caribbean marine re- search will be kept informed. Senior United States oceanographers who wish to participate in a teaching capacity aboard the DECOVERER during one or more legs of NOAA-CARIB should have extens ive experience at sea and preferably, but not necessarily, a working knowledge of Spanish. Limited posi- tions are still available for marine fisheries specialists, marine biologists, physical ocean- ographers, marine geologists and geophysi- cists, and marine meteorologists who can pay their own way to and from the ship. Since marine chemistry is not taught almost univer- sities in the CICAR area, it is not planned now to have any specialists in marine chemistry aboard. Those who are interested and can qualify should write directly to U. S. National Coordinator for CICAR, NOAA Atlantic Ocean- ographic and Meteorological Laboratories, 901 South Miami Avenue. Miami, Fla. 33130. In Mexico, those who wish to propose scien- tific Drojects to be carried out aboard the DISCOVERER between Miami and Veracruz in early October of 1972 should submit their plans to Dr. Augustin Ayala Castanares, Instituto de Biologia, U. N. A. M, , Apartado Postal 70-233. Mexico 20 DF, Mexico. In Jamaica, proposed projects should be submitted to Dr. Edward Robinson, University of the West Indies, Department of Geology and Geography, Mona, Kingston, 7, Jamaica. In Puerto Rico, proposals should be submitted to Dr. Rolf Juhl, Department of Agriculture, Santurce, Puerto Rico 00936. In Trinidad and Tobago, proposals should be submitted to Dr. J. Kenny, Department of Bio- logical Sciences, University of West Indies, Trinidad and Tobago. In Venezuela, Dr. Luis E. Herrara is the NOAA-CARIB coordinator, and proposals should be addressed to Dr . H errara at Instituto Oceanografico, Cumana, Venezuela. In Colombia, proposed projects aboard the DBCOVERER for late November and early December should be submitted to Capt. Juan Pablo Rai ran Hernandez, ComisionColombiana de Oceanografia, Bogota, Colombia, Apartado Aereo 28466. NOAA-CARIB will make a good wind-up cruise for the Cooperative Investigation of the Adjacent Regions and pave the way for future cooperation in marine science in the Carib- bean. 12 38 Reprinted from Environmental Data Service, October, 3-7 5 /-. m *ml msm fit * — — •»• /lL)F(L)Q>Q/lL) . (1) For AL <§ L0, Stommel (1949) shows that F{L0+\AL)~{ALYI2At; where L0 is the initial separation, AL is the change in separation in the time interval A?, and the bars denote averaging over a group of particle pairs with separation interval (L0 + ^AL). Available data generally indicate the power law, F(L) = mU '3 , (2) where m ranges in order of magnitude from 0-001 to 0-1 cm2/3/sec. Besides sampling variability, the coefficient m is believed to be affected by the rate of energy dissipation and the size of the dispersing parcels. We require AL <^ (L0/10), thus (L0 + iAL)< (1-05)L0. Hence, because of the unknown magnitude of error implicit in the assump- tion of stationarity and homogeneity, we take as of sufficient accuracy, F(L0 + iAL) ~ F(L0) ~ (AL)*/2Ar . (3) By letting the east and north components of L0 and (AL) be (x0, y0) and (Ax, Ay), respectively, we can define two other diffusivities : F(x0) ~ (Axf/lAt F(y0) ~ (Ayf/2At . 3. DROGUES, DATA AND ANALYSIS The construction of the drogues is traditional in having two principal components : the surface buoy and the 8-5 m parachute drogue. They were linked together by BT wire weighted at the deep end with a 25 kg weight. The surface buoy was constructed from a 1 5 cm thick styrofoam disk sandwiched between two sheets of plywood reinforced with wood stud. The overall dimensions were 1-3 m in diameter, nearly 1/3 m thick, and carried a 3 m bamboo pole above the water. In addition to passive radar reflectors, lights, and painted identification marks, each buoy was equipped with a small radar transponder adapted from meteorological balloon equipment. The range of the transponder was unexpectedly limited to two ffliles. A feature in the use of drogues or similar floating objects in diffusion study is that such objects respond only to the horizontal components of any three-dimensional field 51 Neighbor diffusivity as related to lateral shear in the Florida Current 495 of motion. When, e.g. changing density stratification reduces vertical turbulence in favor of horizontal turbulence, they respond to the greater horizontal activity, not to the reduced vertical turbulence. The increase in horizontal mixing due to vertical shear as discussed by Bowden (1965) applies only to the use of dye substances or the like that are susceptible to vertical mixing. Another feature is that in the drift of a drogue formation, if one drogue oscillates vertically with respect to its companions, the effect of the oscillation will simulate the action of the twisting vortex tube term in the vorticity equation and will lead to spurious data. Some possible causes of oscillation are the circulation in Langmuir cells, orbital motions in wind and internal waves, and action of the surface buoy in heavy seas in conjunction with vertical shear of horizontal velocity. For a given length of wire, the exact depth of the drogue can be a complicated function of the drag forces on the different components of the drogue assembly. To check on the total effect, the depths of one drogue in each of lines 2 and 3 were monitored by a Benthos* time-depth recorder attached to the deep end of the weighted cable. On both occasions, the recorded depth was within 1 m of the intended 50 m depth. During the monitoring, the winds were northerly at 5-8 m/sec. It is likely that in situations of heavier seas accom- panying stronger northerlies, depth changed will be significant. The weather pattern during the entire experiment was dominated by a succession of cold front passages, with veering winds that were generally 5-12 m/sec, reaching 15 m/sec on occasions. However, the very limited range of the transponder, and the interference of sea clutters on the radar scope from heavy seas, confined observation of drogue positions during intervals of strong northerlies, mostly to tracking individual drogues separately for a good period at a time. This circumstance automatically deterred collection of doubtful data on the simultaneous positions of drogue pairs. From February 17 to March 11, 1969, we tracked 11 lines of three drogues (Table 1) each in the Straits of Florida off Miami (Figs. 1 and 2). A common feature in the drogue formation is the strong tendency for downstream elongation. Otherwise, Table 1 . Starting time, date, and duration of each line. Beginning Duration Line Date Time (1969) (GMT) (hours) It Feb. 15 1943 20-7 2 17 1921 220 3 25 1513 22-7 4 27 0130 240 5 28 1412 24-5 6 Mar. 2 0315 22-8 7 3 1923 17-6 8 5 0404 21-5 9 6 1440 22-8 :? 8 0208 24-4 9 1734 290 ♦Benthos, Inc., North Falmouth, Mass. 02556, U.S.A. tin this line, each drogue was tracked in turn over a period so that these data are excluded from the computation for neighbor diffusivity. The line is included only to show its mean path. 52 496 Frank Chew and George A. Berberian West Palm Beach 4 //^Bimini \ . \ . \ i \ . I* i i i Fig. 1. The mean path of each line of drogues. Numbering as in Table 1, indicates tracking sequence. Dotted portion of line means no observation. All lines were started near the southern boundary of the figure. the speeds of the observed currents varied from line to line, along each line, and between drogues in each line, giving rise to accelerations, lateral shears, and other kinematic deformations. Speeds in excess of 200 cm/sec were found in the last portions of Lines 5 and 1 1 . The mean speed for each line may be roughly estimated from the durations (Table 1) and the total distances (Fig. 1). Negative lateral shear in excess of the magnitude of the local Coriolis parameter was observed in line 8. Positive and negative lateral shear was observed in turn in line 4, whose mean path was well within the region where the mean shear is anticyclonic. Many mechanisms, including that of Ekman layer instability, may be involved in the evolution of these extremely interesting kinematic features. Bathythermograph traces obtained while tracking drogues show a cross-stream pattern that varied from strong temperature decreases with depth below a thin mixed layer of about 20 m in depth along the length of line 5 to a weaker temperature gradient topped by a thicker mixed layer that ranged from 80 m to 95 m along the length of line 4. On the basis of vertical temperature pattern and the positions of each line, we think all of line 5, and perhaps also parts of lines 3 and 6 were in the region where the mean lateral shear is positive (the cyclonic flank). The magnitude of vertical shear in the cyclonic flank is of the order of 10~2/sec, Duing and Johnson (1972). A two meter difference in depths between a drogue pair 53 00(5 ««cri 2245 "*C_\ M45 1015 0815 / A 5V V 0900' 0700 ' 0500' 0300' 0100' 2315' 2115' 1915' ,S7 V" Fig. 2a. Examples of drogue formations as a function of time for lines 4, 5, 6, 8, and 10 (left to right). Hours (GMT) of observation are shown to the left of each cluster. The compass direction and accompanying distance scale to the right apply only the separation and relative orientation between neighbors at a given time. (One drogue was lost in line 5, second from left). 0300 «C^^ 0200. V 00J0 V- T V "3o<7 ,mv /«jo^w 1> V IB 30 • , •"V '"V ~> V 1630 • « ""V "V "V Fig. 2b. Examples of drogue formations for lines 2, 3, 7, 9 and 1 1 . Notations as in Fig. 2a; for line 1 , see footnote, Table 1 . 54 498 Frank Chew and George A. Berberian persisting for a full day will lead to a difference in downstream displacement of nearly two kilometers. But this is unlikely to have occurred in the experiment, for any depth difference is likely to be random. Moreover, for most lines, the lateral shear evolved at a rate much in excess of that implied by a two meter depth difference. In the computer analysis, the separation between drogues for each line was first classified into 21 groups with center values, L0, and fixed limits equal to +0-1 L0 Table 2, column 1. Next, the observations in each group were screened for time difference, keeping only those whose differences were 15 minutes apart. Table 2. Neighbor diffusivities, (cm2/ sec) x 101 Lo, xo or yo F(Lo) N* F(xo) N* F(yt>) N* (km) 0-90 219 24 1-31 86 1-58 31 115 2-77 64 1-89 108 3 02 127 1-40 2-26 109 2-14 60 2-51 94 1-70 2-21 164 2-78 156 3-36 71 210 3-66 203 4-93 136 4-30 117 2-60 5-79 151 5-90 56 7-83 98 3-20 13-47 66 1002 59 9-27 33 3-90 8-79 50 — 2111 27 4-70 19-40 14 21-55 20 5-70 29-25 26 31-46 26 7 00 21-30 47 17-13 37 8-50 28-61 25 22-54 33 10-40 28-56 34 53-88 30 12-70 67-50 15 49-67 13 15-50 48-88 13 — — *The number of values in the entries to its immediate left. This important step is intended also to screen out doubtful data collected during heavy seas. Examination of the data log showed that while there were observations of individual drogue positions in heavy seas, successive simultaneous location of a drogue pair was never attained in the situation, so that the step is coincidentally an effective filter of doubtful data. A considerable number of observations are discarded in this step because of the many interruptions in the continuity of the observation. The same procedure is followed for the evaluation ofF(L0), F(x0), and F(y0). The time difference used in these calculations is 15 minutes, corresponding to our observational interval; and an estimate of diffusivities is computed only when the number of values is, arbitrarily, 13 or more. 4. EXPERIMENTAL RESULTS The first result is the estimates of F(L0) for all lines. They are for L0 of 900 m or more as listed in column 2, Table 2, and plotted in Fig. 3. The small departure from the 4/3 slope line may not be significant owing to the assumption of stationarity and homogeniety. Compared to some other results (Fig. 4) our estimates agree best with 55 Neighbor diffusivity as related to lateral shear in the Florida Current 499 ^ 6 (0 o iL 5 o O o 1 r T i/t1 o I9A 1 Xi iLirii if [ 95% CONFIDENCE LEVEL LOG>oLo 6 (CM) Fig. 3. The plot of neighbor diffusivities against initial neighbor separation, La, both in log scale. Vertical bars give intervals for the 95% confidence levels on the basis of Chi-square distributions. The slanting line has a 4/3 slope. 1 ' 1 ' 4// >> „/ Xr S6 — v/+& O d 5 - / /\r u_ J f • ♦ 0 . \+ 34 _ / • r ML MITCHELL 3 a DENNER O NEMCHENKO • 1 OZMIDOV 1 L0610L0(cm) Fig. 4. A comparison of neighbor diffusivities estimates. The Nemchenko and Ozmidov data as well as the different 4/3 slope lines are all adapted from Denner, Green and Snyder (1968). The different slope lines correspond to different values of m as defined in (2). Our Mt. Mitchell estimates are seen closest to those of Denner et al. 56 500 Frank Chew and George A. Berberian the summer observation of Denner (1968), who used the same size parachute drogues in the California Current, but lowered them to a depth of only 10 m. The second result is the estimates of F(x0) and F(y0) for all lines (Table 2, Columns 4 and 6, Fig. 5). Comparing all three diffusivities at corresponding scales, we see ?5- o CD O I 1 / & / — A » / * / A i c a* © A o / > NORTH COMPONENT o 0 1 I EAST COMPONENT 1 L0G|0(XQOR YQ) (cm) Fig. 5. Plot of neighbor diffusivities for the xo and yo component of Lo, against xo and yo, all in logarithmic scales. Triangles are for the yo or north-south component, circles for the xo or east- west component. (Table 2) that they all have the same trend. So that, again, these values fit the same 4/3 slope line fairly well. Because the average direction of this segment of the current does not depart much from the north, we shall, hereafter, take the east-pointing x axis as the cross-stream or lateral direction and the north-pointing y axis as the down- stream or longitudinal direction. The error involved is assumed small. In this con- vention, F(x0) and F(y0) correspond respectively to the lateral and longitudinal component of the neighbor diffusivity. An important feature of the estimates in Table 2, columns 4 and 6, is the difference in the range of scales for which there were sufficient data to estimate the two diffusivities. Where estimates for F(y0) extend to 13 km, estimates for F(x0) end at the 3-km scale. The initial drogue formations were at the vertices of triangles that were roughly equilateral. If the forces that disperse them laterally were as strong as those dispersing them downstream, we would expect horizontal isotropy, as our observations would yield enough data for the determination of both the F(x0) and F(y0) to nearly the same scales. But this possibility is at variance with the feature that emerged. The feature is one of horizontal anisotropy, characterised in most cases by smaller scale drogue formations during the early portions of the lines with rapid development toward downstream elongation in the later portion. The time interval from the start, when this transition occurs varies from line to line. For a rough average measure of this time interval, we consider the magnitude of the observed range of F(x0) relative to that of the observed range of F(L0), taking the former to represent 57 Neighbor diffusivity as related to lateral shear in the Florida Current 501 the smaller-scale formations. The range of F(x0) is 3-2 — 0-9 =2-3 km, and the range of F(L0) is 15-5 — 0-9 = 14-6 km. Assuming that the ratio of the former to the latter is proportional to the average time interval, T0, after start when horizontal asymmetry became evident, we have TQ = (2-3/14-6)22-9 = 3-6 hr, (5) where 22-9 hours, the average of Table 1, column 4, is the average duration of all lines. 5. A DIFFUSIVITY MODEL In a study conducted near the speed axis of the Gulf Stream off Cape Lookout, Chew and Berberian (1971) documented a situation where horizontal divergence was the major mechanism in neighbor dispersion. In general, vorticity, stretching and shearing deformations may also be important. Here we consider, in a crude fashion, a diffusivity model where the lateral velocity shear is the only contributor. The model is obviously a drastic simplification of the observed situation. Consider two drogues, 1 and 2, riding in a steady current with linear velocity shear. Drogue 1 has (x,y) velocity components («i,vi). Relative to drogue 1, drogue 2 has initial x,y coordinates A, B, (Fig. 6), and velocity components (t/2, V2) given by: 1'2 Ml, vi + 0>v/lx)A (6) Y n -X Fig. 6. A schematic diagram of position coordinates of Drogue No. 2 relative to Drogue No. 1 at two instants of time. At time zero, Drogue 1 is at position labelled 1, while Drogue 2 is at position 2. At time t, relative to Drogue 1, Drogue 2 is at position labelled 2'. The (x, y) axes are oriented positive eastward and northward, respectively. At time t, relative to drogue 1, the position of drogue 2 is at 2', with relative co- ordinates (A, B + b). The neighbor separation L at t is: where L2=A2+(B + bf, b = (v2 — vi)t — Q>v/lx)At, (7) (8) 58 502 Frank Chew and George A. Berberian which may be positive or negative ; for the present we take b b, so that L02 = 2A2 = 2B2. Solving for the root-mean-square value of the lateral shear, we get : [(Sv/dx)2] * = [(S/AtL02) (AL)2/2At] * . (14) For the right member, we make use of (2) with m = 0-003, as shown in Fig. 4. For the left member we consider, within each scale, an arithmetical mean shear, denoted by double overbar, such that when Reynolds rule is used, we have : (7>v/lx) = (Iv/lx) + (dv/dx)' and (15) (bv/ix)2 = [(Dv/ix)]2 + [(Zv/Dx)']2 . With the further assumption that the two terms on the right hand side are of equal magnitude, then, on taking L0 = 3 km, and recalling t = 900 sec, we have : {[(c)v/J)x)]2}» ~ (5-2 x 10-3)/2L1/3 = 3-9 x 10-5/sec. (16) 59 Neighbor diffusivity as related to lateral shear in the Florida Current 503 For this region of the current, Schmitz and Niiler (1969) report a magnitude of 2 to 3 x 10-5/sec for the mean lateral shear; but this is defined differently from (15). Nevertheless, (16) seems plausible. Since lateral shear will introduce horizontal asymmetry, this aspect may be used as another test for the diffusivity model. Suppose a drogue pair was set adrift initially with an east-west spacing A, and a north-south spacing of zero. When subsequently the drogue pair is oriented SW-NE or SE-NW, we have either A — b or A = —b. If we repeat this procedure for a good number of drogue pairs, then, on the average, the elapsed time T for A = b or A = —b is: T = {(iv/^x)2}-i = (1/7-72 x 10-5) sa 3-6 hr, (17) which is the reciprocal of twice the magnitude given in (16). The agreement with (5) is reassuring. A third test arises from the absence in our model of a mechanism for lateral dispersion. If the absence is a reflection of the condition in the current, we would expect the magnitude of the observed maximum lateral scale for all drogue lines to be of the same order as the initial separations. The plot of Fig. 2 and the content of Table 2, all suggest that this is largely the case. On the other hand, lateral shear of the down- stream velocity component is the only operating physical mechanism in the model; if this is a reflection of the condition in the current, we would expect horizontal anisotropy in the longitudinal direction. This is indeed the case. Finally, depending on the scale of averaging, what is laminar shear flow on one scale is turbulence on another (larger) scale. The theoretical basis for the 4/3 law of neighbor diffusivity is found on the 5/3 law of the inertial subrange part of the turbulence spectrum. On this basis, we can relate the intensity of the energy of lateral shear, as a turbulent mechanism to the magnitude of neighbor separation. For our model a relationship is obtained by returning to (14) and introduce (3) and (2), the latter with L = L0. Then solving for the r.m.s. value of the lateral shear, we see that its effectiveness as a turbulent mechanism in relative dispersion decreases as separation grows,* the decrease being proportional to L~2la. 6. DISCUSSION The root mean square particle separation is a function of the square root of time in Fickian diffusion, a function of some power of time in the neighbor diffusivity model, and a linear function of time in our linear lateral shear model. In addition there are other models; however, when the time interval involved is short, as in the present case, there is no way to distinguish one from the other. This may be a reason for the apparent success of neighbor diffusivity over the range of scales reported by Ichiye and Olson (1960). The diffusivity model reveals a possible pitfall in the interpretation of neighbor diffusivity. Consider, for example, the plan of deploying 25-30 drogues simultaneously across the width of the Florida Current off Miami, and photographing their successive positions. Among the many kinds of information the experiment may yield is the lateral component of the neighbor diffusivity at scales up to the width of the stream. *This was pointed out to us by one of the referees of the paper. 60 504 Frank Chew and George A. Berberian Let us restrict our attention to a width of the current where our model may apply. For an east-west line of drogues across this portion of the current, we have initially B = 0 and A = L, so that equation (7), on differentiating, becomes (dL/dr) = (b/L) (db/dt). Again for L m L0, the resulting diffusivity is : (ALf/2At = {(Ar)3(i)(J)v/5x)4} (A*/LQ*) oc A*IW = A2 . (18) That is, the diffusivity is proportional to the square of the east-west separation, A. At subsequent intervals, where (10) holds, we have, similarly, from (13): (AL)2/2A; oc (AB/L0f = (A~* + B'2)~l . (19) Hence, as long as our model remains valid, we find that the larger the initial cross- stream separation, the larger the resulting diffusivity. But these hypothetical findings will be artifacts of the experiment, not a measure of the activity of lateral dispersion. The reason is simply that our model, as stated in (6), contains no mechanism for lateral dispersion. Not having a lateral dispersion mechanism is a defect of the model, if such a mechanism is in the current. But finding the diffusivity to increase with the initial lateral separation does not necessarily mean the presence of lateral dispersion activity ! Thus diffusivity estimates must be interpreted with care. If the plan of cross-stream drogue deployment were undertaken, we would need to record the evolution of the neighbor separations among all pairs. For a given drogue cluster, no significant lateral dispersion is indicated when the scales of lateral separation that evolved in time remain substantially the same as the corresponding initial scales. Although both have the same dimensions, the concepts of neighbor diffusivity and kinematic eddy viscosity are different. The possible ambiguity in estimates of the former illustrates this difference, as well as underlines the condition of overlapping scales as a pre-requisite for magnitude transformation from one to the other. We denote by K% and Ky the east and north components of the kinematic eddy viscosity, respectively. Similarly, we denote by ax and ay the standard deviation of the east and north component of the separations of all drogues from their respective ' center of mass ' at a given time. The center of mass of each drogue formation at any given time is defined as the average (x,y) coordinates of all the drogues constituting each forma- tion. This is the definition used by Angell, Allen and Jessup (1971) in a study of relative diffusion estimates from Tetroon flights in the atmosphere. For long diffusion time, the Fickian diffusion model gives, *x2 = 2tKx <7y2=2tKy . (ZV> Now when the scales of the Fickian diffusion model and the neighbor diffusivity model overlap, in particular when x0 = ax and y0 = ay, Richardson (1926) suggests, F(x0) * 3 Kx F(y0) c* 3 Kv (21) Let Fyx and Fxy be the tangential stresses exerted upon a fluid element by the fluid in contact with it, where the first subscript indicates the direction of the normal to the 61 Neighbor diffusivity as related to lateral shear in the Florida Current 505 plane of action and the second the direction of action. It is fundamental that in the limit of a vanishingly small element, we have Fxy = Fyx . (22) Thus, approximating each stress in (22) by the product of kinematic eddy viscosity and velocity shear, as in Newton's formula for stress, we may postulate pKx(Zv/Zx) ~ pKy(m/7>y) , (23) where (bv/~bx) is the eastward shear of the mean northward velocity component, and (bu/iy) is the northward shear of the mean eastward component. To match the generally long averaging time for the mean shears, the diffusion time corresponding to Kx and Ky must be similarly long. In our case, this means the largest values of F(x0) and F(y0). Thus from Table 2, and using (21), we find K% < Ky. In turn, from (23), we find (bv/'dx) > (M/iy), which accords with observation. It is generally assumed that Kx ~ Ky, and (iv/ix) ~ (iu/iy) in the California Current, and moreover, that the magnitude of (iu/'by) is of the same order in the California and Florida currents. If the estimates by Denner, Green and Snyder (1968) satisfy (21), then Fig. 4 shows that thfe magnitude of Ky is also of the same order in both currents. Thus we infer that the magnitudes of the tangential stresses are of comparable order in the open regions of both the Florida and California currents. From Table 2, the largest F(x0) is 1 x 105 cm2/sec; corresponding to this, we have Kx = 3-3 x 104 cm2/sec. Our estimate of the stress in the open region of the current, thus comes only to, pKx(iv/ix) = (3-3 x 104)(3 x 10~5) = 1 dyne/cm2. (24) This is consistent with an earlier estimate by Chew and Berberian (1970), but is much smaller that the stress that can be inferred from the data given by Schmitz and Niiler (1969). But, as was pointed out earlier, the difference may be compatible. First of all, our estimate represents a Lagrangian measure, while the larger estimates are Eulerian statistics. Secondly, the larger estimates were based on instruments that respond only to water motion in the top 1-2 m layer of the sea surface, while our estimate is based on the motion of circular parcels 8-5 m in diameter at a depth of 50 m. Diff- erences in the scales of motion must be expected in these circumstances; for example, the present Lagrangian statistics describe only the interior of the current, while the Eulerian statistics most probably include perturbations of the stream as a whole. 7. CONCLUSION Our estimates of the neighbor diffusivity show two features: a consistency with the 4/3 power law, and a marked horizontal asymmetry. The diffusivity model in terms of lateral shear appears capable of explaining both features. The diffusivity estimates are comparable to those for the California Current, consequently we infer that the greater advective processes found in the Florida Current do not lead to greater lateral tan- gential stresses. REFERENCES Angell J. K., P. W. Allen and E. A. Jessup (1971) Mesoscale relative diffusion estimates from tetroon flights. J. appl. Met., 10, 43-46. 62 506 Frank Chew and George A. Berberian Bowden K. F. (1965) Horizontal mixing in the sea due to a shearing current. J. Fluid Mech., 21, 2, 83-95. Chew F. and G. A. Berberian (1970) Some measurements of current by shallow drogues in the Florida Current. Limnol. Oceanogr., 15, 88-99. Chew F. and G. A. Berberian (1971) A determination of horizontal divergence in the Gulf Stream off Cape Lookout. J. phys. Oceanogr., 1, 39-44. Denner W. W., T. Green and W. H. Snyder (1968) Large scale oceanic drogue diffusion. Nature, Loud., 219, July 27, 361-362. Duing W. and D. Johnson (1972) High resolution current profiling in the Straits of Florida. Deep-Sea Res., 19, 259-274. Ichiye T. and F. W. C. Olson (1960) Uber die neighbour diffusivity in ozean. Dt. hydrogr. Z., 13, 13-23. Okubo A. (1970) Horizontal dispersion of floating particles in the vicinity of velocity singularities such as convergences. Deep-Sea Res., 17, 445^154. Okubo A. (1971) Oceanic diffusion diagrams. Deep-Sea Res., 18, 789-802. Richardson L. F. (1926) Atmospheric diffusion shown on a distance-neighbor graph. Proc. R. Soc., (A) 110, 709-727. Schmitz W. J. and P. P. Niiler (1969) A note on the kinetic energy exchange between fluctuations and mean flow in the surface layer of the Florida Current. Tellus, 6, 814-819. Stommel H. (1949) Horizontal diffusion due to oceanic turbulence. J. mar. Res., 8, 199-225, Stommel H. (1958) The Gulf Stream. Univ. Calif. Press, 202 pp. 63 8 Reprinted from Journal of Marine Research 29, No. 3, 339-3^6 A Comparison of Direct and Electric-current Measurements in the Florida Current Frank Chew2 William S. Richardson3 George A. Berberian2 ABSTRACT Simultaneous measurements of surface velocity obtained in 1967 by means of geomagnetic electrokinetograph and free-instrument techniques aboard two ships in the Florida Current off Ft. Pierce, Florida have been compared. The results show a varying degree of agreement in magnitude across the stream and good agreement in direction. A surprising result is the high seabed conductance. Introduction. The many interfering influences that may affect the measure- ment of ocean currents by means of the geomagnetic electrokinetograph (GEK) have been discussed by Longuet-Higgins et al. (1954). Previous attempts to assess the magnitude of the errors have been based mainly on a comparison of data from ship drifts and GEK readings. These comparisons are usually not satisfactory because the drift cannot be determined with the required accuracy; or, where accuracy at the end of a transect is available, only an average value is obtained (cf. Hela and Wagner 1954). This is a report on a two-ship study of the Florida Current off Ft. Pierce, Florida, the express purpose of which has been to collect data for a point-by- point cross-stream comparison of direct and electrical surface-current measure- ments. The ships were the Gulf Stream (Nova University) and the U.S.C. & G.S. Ship Peirce. Measurements. Two transects of the Florida Current (Fig. 1) were made: the first between 1300 hr and 1940 hr on June 14, the second between 1950 hr on June 14 and 0320 hr on June 15, 1967.4 The winds were variable and 1. Accepted for publication and submitted to press 15 June 1971. 2. Atlantic Oceanographic and Meteorological Laboratories, National Oceanic and Atmospheric Administration, Miami, Florida 33030. 3. Physical Oceanographic Laboratory, Nova University, Dania, Florida 33004. 4. Times are Eastern Standard Time (Greenwich Mean Time plus 5 hours). 339 64 340 journal of Marine Research [29,3 Figure I. Location of the section and stations. generally less than 10 knots. The synchronized but separate measurements of the current were made at the same cross-stream positions; however, to avoid possible intereference with the reception of HiFix navigation signals on board the Gulf Stream, the northern limits of the stations occupied by the Peirce were 500 m upstream of the southern limits of the stations occupied by the Gulf Stream. The sailing plan of the Peirce for GEK measurements was similar to one that von Arx (1950) called A 2, with the difference that, at course changes, the ship always passed to the starboard to minimize the effect of the cross- stream velocity gradient on the downstream component of the measurement. The base course was directed east-west, with each leg on the fix course varying from 2 to 3 km, depending on the rapidity with which the GEK signal ap- 65 197 'J Chew, Richardson, and Berberian: Current Measurements 34* .2 S S E .ti c Q -o l^ o 5 £ 0 . to Si IN. £°- m— 'eoeMOioOTfOCT>eomcomcNr^cococM EX, E a 2 u 00 © — « r~ — in^ininiOin^H^NW^MOi'^,Lntooino N.C f-s inocooi-co-NNffimoiajNOO I 0 'tmi'TfmininninM'tni'^n * — co co co co co co co co co co o co co co co o — ffi o co o « Tf~ooma)r^omr^r-~a3CT)COco VQ^ * u a "o E 2 0000000 — — —, — — —, — 000000 66 342 Journal of Marine Research [29>3 proached a consistent steady trace on the recorder. Thus, a complete GEK measurement of four legs took about 15 minutes. The free-drop instrument technique employed on board the Gulf Stream has been described by Richard- son and Schmitz (1965). Generally, about 5 minutes were necessary for a determination of the surface velocity; for the current averaged over the whole water column, 5 to 25 minutes were required, depending on the water depth. Table I summarizes the results of the experiment and Fig. 2 shows the cross-stream profiles of the northward components of the quantities measured. For simplicity, only the results of the west-to-east transect are shown. There are two interesting features in the profiles of the direct measurements. First, in Fig. 2 there is the negative portion of the {vs~v) curve at the eastern stations where the northward component of the average current of the water column (v) exceeded the directly measured northward component of the sur- face current (v$). This is a somewhat atypical situation; v appears to be about like the average values given by Richardson et al. (1969), but vs is significantly lower than their average in this region. Second, Table I shows the profile of vs changing from one that has a relatively flat speed axis and a low speed on the far anticyclonic flank on the first transect to one of peaked axis and faster eastern flank on the return transect. Also, there are two features in the profiles for the surface current as measured by the GEK (vg): (i) the sign and magni- tude of vG are always positive and larger than (vs — v); (ii) the profile change between the two GEK transects is qualitatively in the same sense as the change in the dropsonde profiles. The general agreement in the directions of the meas- urements of the surface stream (last column of Table I) and the similarity in the relative change in the two sets of profiles between transects assures us of the consistency of the measurements as well as the reality of the changes in the current. The k-correction Factor. For the purpose of reducing the GEK data, von Arx (1950) has suggested the use of a k factor, which is defined as the ratio of the actual speed of the water to the speed indicated by the GEK. For the present data, k is listed in column 7 of Table I. At Sts. 103, 104, 113, and 105 (on the return transect), the current recorded by the GEK exceeded the current that was directly measured. However, because the direct-measurement and GEK stations were somewhat separated, the small differences can be reason- ably ascribed to spatial changes. Hence, for these four stations we have taken k as unity. The values of the k factor range from values near one at both sides of the stream to a high of 2.8 near the deepest part of the channel. Where k is the largest, the directional discrepancy is also the largest. The pattern of cross- stream variation in k (Fig. 3) is similar for both transects, but there is a numer- ical difference that may be significant where knowledge of the cross-stream gradient of velocity is desired. 67 1971] Chew, Richardson, and Berberian: Current Measurements 343 west station number east 103 104 105 106 107 108 109 )]Q ]]] ])2 113 -40 -1 1 1 r 10 20 30 40 50 60 70 80 Cross-stream Distance (km.) Figure 2. The cross-stream profiles of the northern (downstream) component of the directly measured surface velocity, -vg, the surface velocity from the GEK, vq, and the difference between i>£ and v, the averaged velocity from surface to bottom (first transect only). Seabed Conductivity. Our transects of the current were close to where the channel begins to broaden out downstream (Fig. i). If the structure of the Florida Current is not significantly affected by this broadening, then completion 68 344 Journal of Marine Research station number west 103 104 105 106 107 108 109 110 09,3 111 east 112 113 3.0. L. 0 S2-0J 1.0 JU o g west-to-east transect Q g east-to-west transect © 6- 0- .-tD-^Q 10 cross-stream distance (km) Figure 3. The cross-stream profile of the k factor for both transects. of the circuit for the flow of an induced electric current is controlled principally by two factors. When the seabed conductivity is negligibly small, completion of the circuit is wholly within the water and is dependent on only the mean speed of the stream. In this circumstance, the northward component of the GEK signal, vq, is given by (vs—v). However, it is clear from Fig. 2 that vg differs considerably from (vs— v) in magnitude, and, for the eastern stations, in sign also. We conclude, therefore, that the seabed conductivity is not at all negligible in this region. In terms of the equivalent electric-circuit analogy considered by Longuet- Higgins et al., significant conduction through the seabed reduces the electric effect of v. The reduction may be expressed in terms of m, a function of the internal seawater resistance (rS/h) relative to the external seabed resistance (r'SjH), as follows: m = {i+{rSlh)l{r'SIH)Yn (1) here r and r' are, respectively, the electric resistivity of seawater and of seabed, h and H are, respectively, the stream depth and the depth into the seabed to which the electric current spreads, and S is the stream width. The last column in Table I tabulates the values of m computed from m = (vs - vq)Iv, (2) with the requirement that o8IH)-i.S, (3) or, that the internal resistance through the sea water and the external resistance through the seabed are of comparable importance. For the Florida Current off Ft. Pierce, the ratio of the stream width, S, to stream depth, h, is about 200. If, following Longuet-Higgins et al., we take S = H, then, on the average, the ratio (rjr') of the electrical resistivity of the sea water to that of the seabed is about 7 times larger than the ratio that is thought to be applicable to the English Channel. The result given in (3) is supported by voltage measurements obtained by means of stationary electrodes across the Florida Current from Palm Beach, Florida, to the Little Bahamas Bank (Thomas Sanford, Woods Hole Oceano- graphic Institution, personal communication). On the other hand, the voltage recordings obtained from stationary electrodes placed on the sea bottom at 3.0 and 8.5 nautical miles east of Fowey Rocks off Miami, Florida, indicate other- wise. The voltage recorded continuously from 5 May to 17 November 1969 required, on the average, only a 5% correction for seabed conduction to bring the voltage-indicated transport of the section to the mean transport values given by Schmitz and Richardson (1966) (Harry DeFarrari, University of Miami, personal communication). The small correction required for data off Miami supports the premise of Chew's (1967) estimate of the cross-stream variation in the k factor. Off Miami, the ratio of stream width to stream depth is about 130 — not much different from the corresponding ratio off Ft. Pierce. The large change in seabed conductance over the 200-km distance from Miami to Ft. Pierce is surprising. Conclusion. The correction required for the GEK measurement of the sur- face velocity of the Florida Current off Ft. Pierce varies from o in the shallower waters to a factor of 2.8 in the deeper portion of the channel. Moreover, the k factor is not locally constant, but varies significantly over a period of a few hours. The situation is further complicated by the surprisingly high seabed electrical conductivity, so that even the trend of correction cannot be extended to other segments of the Current. Acknowledgment. We are grateful to Ensign J. W. Walsh, who assisted materially with the operation of GEK, and to the officers and crews of the Peirce and Gulf Stream for their fine efforts in our behalf. Financial sup- port to Nova University was provided by the Office of Naval Research. 70 346 "Journal of Marine Research [29, 3 REFERENCES Chew, Frank 1967. On the cross-stream variation of the k-factor from geomagnetic electrokinetograph data from the Florida Current off Miami. Limnol. Oceanogr., 12: 73-78. Hela, Ilmo, and L. P. Wagner 1954. Note on tidal fluctuations in the Florida Current, Sect. B: 14-24. Techn. Rep. Mar. Lab., Univ. of Miami, 54-7; 71 pp. Longuet-Higgins, M. S., I. Stern, and Henry Stommel 1954. The electrical field induced by ocean currents and waves, with applications to the method of towed electrodes. Pap. phys. Oceanogr., 13: 1-37. Richardson, W. S., and W. J. Schmitz, Jr. 1965. A technique for the direct measurement of transport with application to the Straits of Florida. J. mar. Res., 23: 172-185. Richardson, W. S., W. J. Schmitz, Jr., and P. P. Niiler 1969. The velocity structure of the Florida Current from the Straits of Florida to Cape Fear. Deep-sea Res., Suppl., 16: 225-231. Schmitz, Jr., W. J., and W. S. Richardson 1966. A preliminary report on Operation Strait Jacket. Techn. Rep. Mar. Lab., Univ. of Miami, 66-1; 222 pp. Von Arx, W. S. 1950. An electromagnetic method for measuring the velocities of ocean currents from a ship underway. Pap. phys. Oceanogr., 11: 1-62. Printed in Denmark for the Sears Foundation for Marine Research, Yale University, New Haven, Connecticut, U.S.A. Bianco Lunos Bogtrykkeri A/S, Copenhagen, Denmark 71 9 Reprinted from Journal of Marine Research 30, No. 3, 281-29** Estuarine Circulation Induced by Diffusion Donald V. Hansen Atlantic Oceanographic and Meteorological Laboratories, NOAA Miami, Florida 33130 Maurice Rattray, Jr. Department of Oceanography University of Washington Seattle, Washington 98105 ABSTRACT Similarity solutions of the equations for estuarine circulation and salt balance are presented for a circulation generated by diffusive modification of stratification maintained at the en- trance to an inlet by external dynamics rather than by fresh water discharged directly into the inlet. The explicit x dependence may be factored from the governing equations if the inlet geometry and longitudinal variation in turbulent exchange coefficients are expressible as power or exponential functions of distance along the inlet, and the ordinary differential equations so obtained are solved approximately by perturbation in a function of the Rayleigh number. Salinity in the model decreases inward along the bottom but increases inward along the surface of the inlet, and the flow is three-layered — inward at the top and bottom, out- ward at the mid-depth — as have been shown and inferred for Baltimore Harbor. Inclusion of bottom friction increases the salinity gradient along the bottom and decreases it along the surface, giving the salinity distribution some resemblance to that expected in conventional estuaries. The induced circulation in all cases is a strong function of total depth and a weaker function of length and turbulent exchange coefficients for salt and momentum. The model can serve as an aid to interpretation of observations in this type of estuary, and it provides a simple means of evaluating the circulation and mixing rates from salinity measurements to whatever precision the eddy viscosity can be estimated, or current measurement can be obtained to scale the flow. Introduction. Estuarine circulation is induced primarily by the density dif- ference between freshwater and seawater. River water typically enters near the head of an estuary, resulting in a two-layer circulation and stratification main- tained by a dynamic balance of advective and diffusive processes within the estuary. However, in inlets with negligible river discharge there is another type of circulation induced by mixing of an externally maintained density strat- I. Contribution No. 654. from the Department of Oceanography, University of Washington. Accepted for publication and submitted to press 14 June 1972. 28l 72 282 "Journal of Marine Research [30,3 2 4 6 8 10 12 14 16 DISTANCE -KILOMETERS FROM HEAD OF HARBOR ification. This mechanism was studied experimentally by Hachey (1934) in connection with studies of the Bay of Fundy, and it appears to be of primary importance in flushing Baltimore Harbor. Stratification is maintained in Chesa- peake Bay by dynamic processes associated with freshwater discharged into the system primarily by the Susquehanna River, but the small quantity of fresh- water discharged directly into the embayment that forms Baltimore Harbor is insufficient to maintain locally the stratification against vertical diffusion. The resulting distri- bution of salinity (Fig. 1) has in the deeper water the positive seaward gradient that is usually found in estuaries, but a nega- tive seaward gradient near the surface and the circulation in- ferred by Carpenter (1960) and by Pritchard and Carpenter (i960) from the movement of acid wastes discharged within the Harbor consists of inward flow near the surface and bottom and of outward flow at mid-depths. This circulation appears to be too weak for reliable measurement; nonetheless it was identified by Carpenter (i960) as the primary reason why the mean exchange rate in the Harbor exceeds by a factor of five the estimates by tidal exchange theory and why it operates more steadily than is expected from transient wind effects. Although this phenomenon doubtless occurs in many other embayments and conceivably in stratified lakes, there is to our knowledge no analytical basis for its quantitative appraisal. This paper presents some results of work toward mathematically modeling it. Figure Typical longitudinal section of the salinity distribution in Baltimore Harbor. Chesapeake Bay, at the mouth of the Harbor, is at the right end of the figure. The arrows show the net flow pattern (from Cameron and Prit- chard 1963). Notations .*•, z Av S,d g Q k L,ByD Rectangular space coordinates with origin in the free sur- face, positive seaward and downward. Dimensional and dimensionless stream function. Vertical turbulent viscosity. Horizontal and vertical turbulent eddy diffusivity. Salinity. Gravitational acceleration. Density. Q-'dg/dS. Length, width, and depth of inlet. Horizontal and vertical velocity components. 73 1972J Hansen and Rattray: Estuarine Circulation 283 e Subscript used to denote values taken at entrance to the inlet. T{, Bottom stress. c Drag coefficient. U0 Amplitude of tidal current. £, rj Horizontal and vertical dimensionless space coordinates. Fi Similarity scale function, /' = o or 1. a, /?, y, A, yiy j<2, x3 Exponents for scale function. K L2 P = — — — Relative isotropy number for turbulent mixing. KheDe l> ■dveKhe 45 UqD Inlet Rayleigh number. Dimensionless drag coefficuent. Formulation. As in an earlier paper on conventional estuarine circulation (Hansen and Rattray 1965, hereafter referred to as HR), the most essential variations are considered to occur in the vertical and longitudinal directions. We therefore limit consideration here to embayments that are sufficiently elongated or laterally homogeneous so that cross-channel mean values are useful representations of all properties. Pritchard's (1956) determination that nonlinear accelerations are dynamically unimportant in conventional estuarine circulation in coastal plain estuaries has provided the basis for the dynamical development given in HR. In view of the weak circulation anticipated and the importance of turbulent mixing, it is assumed that these factors are also unimportant here. Therefore, the governing equations are the same as those given in HR, but new solutions have resulted from the imposition of different boundary conditions. The equations derived in HR for the flow and salinity fields are: Jv + gkSx = o, (0 y>xSz-y>zSx = (BKh,Sx)x + (BKvSz)z, where the transport stream function %p is defined by Bw = ipx and Bu = — y>z; the subscripts denote partial differentiation with respect to the independent variables # and z. For boundary conditions we require conservation of water and salt at the boundaries, z = o, D (x), and at the head of the inlet. In the absence of river 74 284 yournal of Marine Research [30, 3 flow, the net transport through any section across the inlet must be zero. The circulation is driven by density stratification imposed at the entrance to the inlet; however, this stratification is not expected to be entirely independent of processes that occur within the inlet. To specify a clear mathematical problem, we make the plausible assumption that the salinities of the source waters — those found at the surface and bottom in the entrance to the inlet — Soe and Sbe, are maintained by external processes. Wind stress on the free surface can be expected to modify the weak circulation anticipated, but it is neglected here as being irrelevant to the mechanism of interest or, because of its intermittency, as being implicitly represented as a contributor to turbulent exchanges. These boundary conditions are expressed by: W{xi°) = V(X->D) = Wzz{x->°) = Sz(x,o) = S^x^D) = 0 1 \ (2a) Representation of bottom stress by a square-law of the form Xb= -c Ub\ Ub\ , where Ub is the current speed just above the bottom, has proved useful for estimating bottom stress in tidal currents. This equation was used by Abbott (i960) to determine the direction of the residual stress for a flow with an harmonic component of amplitude U0. It is easily shown from his result that the relationship between residual stress, f6, and the net current near the bot- tom, «j, is rb = --cU0Ub[i + o(UbIU0)2]. 71 Hence, for tidal current speeds that are large relative to the net circulation, the linearized bottom-stress condition, ^-i);^>-°> (2b) is applied at the bottom. In the absence of a dominant tidal current, flow will be weak in general, and (2b) is readily transformed to an exact representation of a linear stress- velocity relationship by substituting \U c = c\ where c' is a linear drag coefficient. The Similarity Functions. For simplicity we assume that the breadth of the inlet and the exchange coefficients vary in the x- direction only. We then make use of the function introduced in HR Ft{a;l;) = (l-i)e*-S + iF; 75 1972] Hansen and Rattray: Estuarine Circulation 285 here ; may be assigned values of zero or unity to portray particular kinds of inward behavior as appropriate in seeking similarity solutions of the form !P(*,«) - [BeKveLIDe]Ft(y; £)*(,), \ S(xtz) = S-S[i +6(o)-Ft(Xi W(*l)l> j where S = i/2(Sbe + Soe), S = il2(Sbe-Soe), I = */£, 17 = [z/DJ^Ca; f). The origin of coordinates can be chosen to make Fi = 1 at the entrance to the inlet; De and L are scale depth and length for the embayment. The rep- resentation of the embayment can be explicitly closed at the head only by the choice i = 1, in which case L is simply its physical length. However, it is at least mathematically possible to retain the option i = 0, which might apply near the entrance to an embayment of dynamically infinite physical length, in which case L will be a dynamic length scale analogous to that used in HR. Longitudinal variations in dimensions and mixing parameters are presumed to be expressible by D(x) = DeFi(- a; £), B(x) = BeFi{p; f), Jv(x) = JveFi(xl; |), Kv(x) = KveFi(x2; |), Kh{x) = KheFi(>{3; f). (4) These expressions allow an interesting variety of smooth variations of inlet geometry and exchange coefficients, but for separation of the explicit x depend- ence, the exponents must satisfy the conditions (P~y) = (^4« + ^-^) = - (1 + * + xJ - (f + «- X3). (5) The exponent X is to be determined by the distribution of salinity and flow within the inlet. We may logically assign inlet geometry (a, /?, i) and one of x2, x3; y, A, and the remaining mixing parameter variations are then specified by the physics of the problem and by the similarity conditions. We would of course desire the possibility of a priori specification of all configuration and mixing parameters, but problems of such generality are not accessible to anal- ysis by the method of similarity solutions. The philosophy adopted herein is to specify independently as many external parameters as is possible consistent with allowing the density distribution to be determined by advective and dif- fusive processes through an Eigenvalue condition, analogously to a conven- tional heat conduction problem. The remaining parameters must satisfy eqs. (5). Substitution of equations (3), (4) and (5) into (1) and (2) yields the set of ordinary differential equations and boundary conditions, 286 Journal of Marine Research [3°>3 P0"" + Ra(Xd + » = 0"(i)+/i0>(i) = o'(") = *(0 = °> 1 f8) Ra = gkS D^jJveKhe is a Rayleigh number similar to that defined in HR> /* = KveU\KheD^ is a relative isotropy number for turbulent mixing, equal to the ratio of horizontal to vertical mixing time scales, and /j, = (^c/n) (U0DJAV) is a dimensionless drag coefficient. The x dependence is factorable from eq. (2b) to obtain its counterpart in (8) only in the special cases of free-slip (/x = o) or no-slip (^u-1 = o) bottom-friction conditions or if U0Dcj Av = constant. Hence, the latter condition is presumed to be satisfied, except possibly for the special but interesting cases /u = o and yu_I = o, for which it is unnecessary. For convenience, from this point we restrict consideration to inlets of con- stant depth (a = o). We observe that{ if / = o, (6) and (7) have the solution A = o for arbitrary 0, indicating that, if the time scale for horizontal mixing is short relative to that for vertical mixing, an arbitrary external stratification is diffused into the inlet without generation of horizontal gradients or circulation. For vanishingly small Ray a result of vigorous mixing or shoal depth, solu- tions of (6) and (7) that satisfy the boundary conditions (8) are 0 = o and 0 = -cos p7irj, if A (A + y) = (/>tt/)% p = 1, 2, The positive roots of (8) define the longitudinal attenuation of a salinity field maintained by a balance of horizontal and vertical diffusion only. In the limit of vanishing /, we again obtain X = 0, so the external stratification again penetrates the inlet essentially undiminished. As / becomes large, so does X, indicating rapid attenuation of the external field inward from the entrance to the inlet. We have been unable to obtain an exact solution to the complete equations; rather, we have developed approximate solutions through perturbation of the diffusive model, using a series expansion of 0, d, and A in a suitable para- 0(rj) = 0o(r}) + E0i(ri) + e*02(r)) + , Hi) = 0o(v) + eBi(rj) + #8z(ri) + ...., X = X0 + eXI + e2X2 + Use of Ra as the perturbation parameter revealed a difficulty that was described also in HR; that is, the procedure leads quickly to useful forms for 0 and 6 but not for A in that terms to the first order in Ra portray the essential features of the salinity and stream-function profiles, and even some aspects of their interaction; but the approximation to A diverges at values of Ra well below that expected in nature. Higher-order approximations, as well as being 77 1972] Hansen and Rattray: Estuarine Circulation 287 not only computationally tedious, introduce polynomial and trignometric ex- pressions of a higher degree that produce features of smaller scale in the vertical than are significant in the context of using exchange coefficients explicitly in- dependent of the depth and stability. In the present case, however, no simple integral method has been found to extend the low-order approximation to large Ra without incurring excessive violation of the boundary conditions on 0. Instead, we apply a technique suggested by Bellman (1955) to improve the convergence properties of the approximation. For the perturbation parameter we use e = Ra/(F + Ra), where F is a disposable parameter, which we con- veniently define by F = I27t4(i +.33^/(1 +.28^). Then the perturbation equations to be solved are: 0O"" = O /-■ ®i"" = &i-j""-(FlP) 2A»0 j-n-it /20o" + Ao(Ao + y)0o = o, Pdj" + h(X0+y)6j = 2{/2(y0„0',-»-^'»0;-n) )-n M (9b) (10a) (10b) -(X0+y)X„dj-n-^n 2 (I2®'m+*m)0j-m-n}-> m = o subject to the boundary conditions 0}(o) = 0/'(o) = 0/(0) = 6/(1) - 0,(1) = */'(!) +/**/(!) - O, j - 0, 1,2, ; 0o(l)-0o(o) - 2,0,(1) = 0;(O), j = 1,2, ; These provide conditions for a determination of A at every level. Solutions of the zero-order set are described above, and 0O = o has already been used in (10b). Because the nonlinearity of the equations precludes superposition of solutions, expressly through the incompatibility of the similarity conditions with a multiplicity of A, we work with only the lowest stratification mode. While this simple solution cannot match an arbitrary external field, it does contain the basic stratification essential for the circulation mechanism of in- terest. Solutions of (9) and (10) to order e are: 0(rj) = e0i(rj) = XRa .-= / 1 2 (COS TIT] + 2 f\ - l)-27I2(2/j3- 47r4(3+/z)/2V v 1 1 J 3*?2 + r?) + 4y" (cos 7177 - 773+377- 1) -7r2/u(773-2 77I+ 77)} (») 78 288 ^Journal of Marine Research i Q{n) = Bo(r}) + eOi(rj) = - cos nr) + f {G far) + o + 1 1 %\G (i, r)-G{o,r)] cos jtrjj • (y7t& (r) sin jtr + + Xo0' (r) cos TrrWr, i?a (i +0.28,u))-1 where A0 is the positive root of + 0-33/")] ' A0(A0 + y) = 7i2/2, and Gfar) is the modified Green's function for (10), i.e., COS 717] [30,3 (12) (i3) (H) Gr„ = (r - i) sin Tir ' 71 cosjtm sin 71V7I + 77 > cos nr . 2 7t2 71 Eqs. (i i) and (12) were evaluated numerically on an IBM 7094 computer at the University of Washington. Discussion. To demonstrate the nature of the flow, it is necessary to specify some parameters. The embayment is approximated by a rectangular box model having a constant depth and width, open-ended at the entrance (x = L) and closed at the bayhead (x = o). Of the many possible postulates regarding tur- bulent mixing, we assume simply that the fundamental circulation-generating mechanism, vertical mixing of salt, is characterized by a Kv that is constant throughout the inlet. These conditions, which require a = /3 = x% = o, i = 1, yield y = 1, x3 = 2, and Xi = A — 2. Determination of A from equations (13) and (14), though now possible in principle, is of doubtful value due to the lack Table I. Effects of bottom friction* WPDrpntuJXRaLBKn) T) at y) = maximum 10* PDy>minaRaLBKve) t] at rp = minimum T) at y> = 0 X ._ [S(o,ri)-S]lS 0 5.36 0.24 -5.36 0.76 0.50 0.935 0. 1 6.42 0.26 -3.94 0.77 0.55 0.950 -0.108 10 8.81 0.29 -1.24 0.83 0.70 0.984 -0.349 9.94 0.31 -0.26 0.87 0.80 1. -0.465 * All other parameters as given for example in text. 79 1972] Hansen and Rattray: Estuarine Circulation 289 of good a priori estimates of the turbulent Austausch. We elect rather to apply the solution in a semi-inverse way, using directly measured quantities as fully as possible. The dimensionless flow is given by DPxp LBKveXRa XRa 0. Without bottom friction, the flow consists of a pair of half cells (Fig. 2a), symmetric about the mid-depth. With increasing bottom friction, the upper cell grows at the expense of the lower (Table I), and the cir- culation, while it ultimately resembles that found in conven- tional estuaries (Fig. 2 b), is of opposite sense (cf. HR), having predominantly inward flow in the upper half of the inlet and outward flow in the lower half. The maximum horizontal ve- locity generated by this circu- lation is Wl - B-*y>e(L,0) 0'(o) 4-XgkSD^ I O3 Ave L ' (15) For application to Baltimore Harbor, we use the observed values D = 1 3 m, L = 20 km, and S = 5%o (cf. Fig. 1) in (15) to obtain Avt Figure 2. Circulation stream function. Isopleths are iot PDxpjXRaLBKye for (A) zero-bottom stress (jj. = o) and (B) zero-tangential flow JX* Cu-^o). 20 cm3 sec-2. A ioo-hour cur- rent station occupied by the Coast and Geodetic Survey2 approximately 13 km from the head of the Harbor, near the south side of the main channel, in- dicated a net current of 3 cm/sec into the Harbor at 1.5 m (rj = O.i) and 2.5 cm/sec out of the Harbor at 4 m. However, nearly all of the observations were near the threshold of the current meters and must be accepted with caution. 2. Data on file with the Office of Marine Surveys and Maps, National Ocean Survey, NOAA, Rock- ville, Maryland 20852. 80 290 'Journal of Marine Research [3°,3 An advective flushing time (perhaps better termed the average time of re- plenishment), equal to the time required for the volume of the inlet to flow into and out of the entrance, can be defined for the model as ra = 5ZD/(Vmax-vmln) = D*lKve(0^x-0mln)~ioiL*Jvel(giSDn). (16) The shape of the Harbor is considerably more complicated, primarily by ex- tensive shoal areas on both sides of the main channel, especially near the entrance, but consideration can be limited to circulation in the channel. Car- penter (i960) has reported an exponential flushing rate of 10% per day for the Harbor; this corresponds to 99% renewal in 20 days compared with 100 days computed by Garland (1952), who used Ketchum's (1951) modified tidal- prism method. Setting Ave\X = 20/«max = 5 cm2/sec in (16), we find Ta = 20 days, which suggests that the model, simple as it is, may be useful for the derivation from a few relatively easy observations of advective exchange rates associated with this class of circulation. Inward attenuation of the externally imposed stratification is seen in Fig. 1 to be approximately linear, i.e., to imply a value of A near unity. Using A = 1, we obtain Jve = 5 cm2 /sec; and using the dependence of A on / and Ra shown in Fig. 6, we find Kvt*> 0.5 cm2/sec over a wide range of / and Ra on A = 1 . For the parameter range of interest for Baltimore Har- bor, X and all other features of the solutions are little sensitive to variations of K^e because the salt balance is maintained pri- marily by advective processes; only an ambiguous determina- tion of / and Ra is obtained by this inverse method. If we identify the dispersive process described by K^ with the mod- ified tidal prism process as de- fined by Ketchum (1951), then the five-fold increase in longi- tudinal transports, A0/A, due to the steady flow gives Ra rs 5000 from (13). The associated val- ues of Kh and / are then 5 x 1 o5 cm*/sec and 1.6, but the other Figure 3. Salinity distribution, (S — S)JS , for (A) zero- bottom stress (yU = o), and (B) zero-tangential flow (//_1 = o). 81 1972] Hansen and Rattray: Estuarine Circulation 291 -0.5 results are not dependent upon these determinations. We observe that a value of 2 for A would, in addition to the obvious effect on estimates of Avt, yield Xi = o from the similarity condition. This implies that Av <* xVy which appeals to intuition. Salinity distribu- tions that indicate a value for — 77"I / XRO A as large as 2 can perhaps be found among the several sets of Baltimore Harbor observa- tions, which show considerable variability within the consis- tent general pattern exempli- fied by Fig. 1, but this possi- bility has not been pursued. The exemplary value A = 1 requires Xi = — 1, which may be a mathematical artifact im- posed by the similarity condi- tions. Its physical meaning is that the flow must be atten- uated by increasing viscosity near the head of the inlet be- cause the pressure gradient does not vanish there. The salinity distribution can now be conveniently shown (Fig. 3) as (S-S)IS = ^d(r})-[i + d(o)]. Figure 4. Vertical variation in dimensionless longitudinal component of pressure gradient for zero-bottom stress. With no bottom friction (Fig. 3a), the salinity distribu- tion is antisymmetric about the mid-depth, increasing and decreasing inward in the upper and lower halves to become vertically homogeneous with a salinity equal to S at the head of the Harbor — an idealization of the distri- bution described by Carpenter (i960). With increasing bottom friction (Fig. 3b; Table I), an inward and downward flow of low-salinity water causes a pro- gressive reduction in the sectional mean salinity with distance into the inlet as well as a consequent decrease in the salinity gradient at the surface and an increase near the bottom, giving the salinity distribution some resemblence to that found in conventional estuaries. The restrictions of similarity force com- plete vertical mixing at the inner boundary except for the singular case, A = o; hence this feature will persist for all values of Ra and / except 1 = 0; but as A is reduced by large Ra or small /, the nearly vertically homogeneous water is limited to a region near the inner end. 82 292 yournal of Marine Research [30,3 -II [COS 7^+0] Each of the circulation half-cells of the model has zero net transport of water. The advective salt transport, however, resembles the diffusive salt transport in that it is inward in the lower portion and outward in the upper. The overall salt balance is maintained by vertical diffusion of salt from the lower to the upper half- cell. Figs. 1 and 3 and Table I suggest that the salinity distri- bution in Baltimore Harbor reflects some influence of bot- tom friction inasmuch as the vertical mean salinity decreases inward. A similar circulation and effect on the salinity dis- tribution might be expected to result from wind blowing into the Harbor, such as has been observed in East Sound (Ratt- ray 1967), but Carpenter (i960) has observed a qualita- tively similar distribution fol- lowing a sustained period of west wind (out of the Harbor). The vertical variation in the longitudinal pressure gradient associated with the circulation (Fig. 4) is different from that observed in association with the wind-driven three-layer flow in East Sound in that it is zero at two levels. The upper level of zero longitudinal pressure gradient occurs at about one- fifth of the total depth down from the upper surface, and a second such level occurs an equal distance above the bottom; the precise level of each depends only slightly upon the bottom-friction parameter. The perturbation of the basic cosinusoidal salinity profile depends upon the several parameters that enter the problem, but the perturbing influences of vertical and horizontal advection can be identified separately as the effects of terms associated with 0 and 0' respectively in (12) (Fig. 5). The effect of the convergent vertical flow is to sharpen the halocline, but the effect of horizontal advection of mixed water from up-inlet is nearly the opposite. The balance between the two effects is determined by the relative magnitudes of A and y, which is given by Figure 5. Advective perturbation of salinity profile for various bottom-friction conditions. Solid line is horizontal advective perturbation H- Xi dashed line is vertical advective perturbation -=-y. [(i+[a»//y]»)"»-i] 2 + Ra (1 +0.28/4) 6tt4(i +0.33//) (17) 83 1972] Hansen and Rattray: Estuarine Circulation 293 Figure 6. Dependence of X\y on (1 +.28/^)(i +-3Jfl)~1 Ra and I/y. Dashed line denotes exemplary locus of values obtained by varying AT^e only. The dependence of X on other parameters is shown in Fig. 6 for ranges of interest. When I/y is small or when Ra is large, as in a deep inlet, the vertical advection dominates, sharpening the halocline against relatively weak vertical mixing; when I/y is large and Ra is not, the horizontal ad- vection of mixed water pre- dominates, reinforcing the al- ready considerable effect of vertical diffusion in reducing stratification. With increasing bottom fric- tion, both perturbations take on positive mean values, which, because the salinity is made in- variable at the stratification source, effect a salinity reduc- tion almost everywhere inside the inlet because of the pre- ponderant inward and down- ward flow of low-salinity wa- ter. The magnitude of all of these effects is proportional to the strength of the dimensionless circulation, which varies as XRaI~* = XgkSD$/L2KvJv, a strong function of depth that is, however, partially reduced by the dependence of X on / and Ra when these parameters are large. Advection reduces the hori- zontal salinity gradient near the entrance to the inlet, as expressed by (13) and (14). Although bottom friction has a considerable influence on the circulation and salinity distribution, its influence on the horizontal salinity gradients is relatively minor; the factor containing [x in (13) and (17) varies by only 15% over the entire range, o < ju <«>• In dimensional terms, the circulation is proportional to XgkSBD*/ Ldvey which is also a strong function of depth. The inverse dependence of the strength of the circulation on Ave is weakly offset by the dependence of X on Ra as shown in Fig. 6. Increased vertical mixing of salt strengthens the circulation more or less uniformly, and horizontal mixing weakens it, primarily at low Ray only by their effect on X through I and Ra as expressed by (17), and Fig. 6. Too few observations of this type of circulation are available at present to encourage an attempt to relate the strength of the circulation to the bulk para- meters upon which it ultimately depends, as has been done for conventional estuarine circulation (Hansen and Rattray 1966). The model should be of value in interpreting future observations of such circulations, and, from obser- 84 294 "Journal of Marine Research [30, 3 vation of the salinity distribution and a single well-placed and reliable current measurement, the model should provide a simple means of evaluating the strength of the circulation and exchange rates in inlets having simple shapes. Acknowledgments. We wish to express our appreciation for the assistance given by Leonard Pietrafesa for checking the equations, Paul Lu and Mrs. M. Rona for computer programming, and by C. B. Taylor, who furnished current-measurement data. This research was supported in part by the National Science Foundation under grants CP-3549 and GS-5107. REFERENCES Abbott, M. R. i960. Salinity effects in estuaries. J. mar. Res., 18(2): 101-111. Bellman, Richard 1955. On perturbation methods involving expansions in terms of a parameter. Quart, appl. Math., 13(2): 195-200. Cameron, W. M., and D. W. Pritchard 1963. Estuaries. In The Sea, Vol. 2, pp. 306-324. M. N. Hill, Editor. Interscience, New York, N. Y. 554 pp. Carpenter, J. E. i960. The Chesapeake Bay Institute study of the Baltimore Harbor. Proc. 33rd Annu. Conf. Md.-Del. Water and Sewage Ass.; 62-78. Garland, C. F. 1952. A study of water quality in Baltimore Harbor. Publ. Chesapeake biol. Lab., 96; 132 pp. Hachey, H. B. 1934. Movements resulting from mixing of stratified water. J. biol. Bd. Canad., j(2): 133-143- Hansen, D. V., and Maurice Rattray, Jr. 1965. Gravitational circulation in straits and estuaries. J. mar. Res., 23(2): 104-122. 1966. New dimensions in estuary classification. Limnol. Oceanogr., 11(1): 319-326. Ketchum, B. H. 1951. The exchanges of fresh and salt waters in tidal estuaries. J. mar. Res., 10(1): 18-38. Pritchard, D. W. 1956. The dynamic structure of a coastal plain estuary. J. mar. Res., 15(1): 33-42. Pritchard, D. W., and J. H. Carpenter i960. Measurements of turbulent diffusion in estuarine and inshore waters. Bull, intern. Ass. Sci. Hydrol., 20; 37-50. Rattray Jr., Maurice 1967. Some aspects of the dynamics of circulation in fjords, In Estuaries, pp. 52-62. G. H. Lauff, Editor. Amer. Ass. Advance. Sci., Washington, D. C. 757 pp. 85 10 REMOTE SENSING OF ENVIRONMENT 2, 109-1 16 (1972) An Observation of the Gulf Stream Surface Front Structure by Ship, Aircraft, and Satellite GEORGE A. MAUL and DONALD V. HANSEN National Oceanic and Atmospheric Administration, Atlantic Oceanographic & Meteorological Laboratories, Physical Oceanography Laboratory, Miami, Florida 109 Observations of surface temperature and salinity along the western edge of the Gulf Stream were made from a ship while concurrent temperature observations were obtained by instrumented aircraft at six altitudes. The major feature along a five-kilometer line normal to the Stream's edge is a temperature gradient of about 0.75°C/kilometer within which are embedded two abrupt temperature increases of about 1.5°C. Temperature variations were compensated by salinity variations, yielding nearly constant density through the frontal zone; a sharp lateral current shear was associated with the thermohaline mixing region between the steps. The attenuation of surface temperature measured by the airborne radiometer was compared with a theoretical model. The analysis supports the view that a two-part correction technique is required: one part for bulk-skin temperature differences, and another for atmospheric attenuation of sea surface emission due to the mass and temperature of interfering gases. Electronic noise and scan geometry of the Nimbus II radiometer experiment prohibits any discussion of detailed oceanographic variability as seen from space, however, the gross features of the front are reproduced to a fair degree of accuracy. Introduction An experiment designed to compare the detec- tion of some thermal properties of the ocean observed simultaneously by ship, aircraft, and satellite was performed between 1600 and 1800 GMT on 12 October, 1966, 150 kilometers north- east of Cape Hatteras. The Environmental Science Services Administration's Coast and Geodetic Survey Ship Explorer, as part of a long-term Gulf Stream project (Hansen, 1970), obtained ground truth in the form of sea surface temperature (Ts), sea surface salinity, and bathy- thermograph (BT) sections across the surface outcrop of the Gulf Stream front. The Naval Oceanographic Office's Super Constellation NC- 121K and the National Aeronautics and Space Administration's (NASA) Convair 240-A flew in tandem and collected Airborne Radiation Thermometer (ART) data, Infrared (IR) Scanner data, vertical incidence aerial color photographs, and meteorological data, at six altitudes. The High Resolution Infrared Radiometer (HRIR) experiment was operating aboard the Nimbus II satellite which was in transit over the area during these observations. The analysis, however, uses the nighttime HRIR data, collected 12 hours earlier, because the Nimbus radiometer is influ- enced by reflected solar radiation. Fig. 1 . The location of the experiment is superimposed on computer generated contours of equivalent blackbody temperature from Nimbus II data. The contour interval is 1° Celsius. The ground truth is the hatched area near 36°10'N, 73°45'W. The dotted line through the ground truth area is the BT transect shown in Fig. 2b. The curving dashed line is the center scan of the Nimbus II HRIR scan spot data of Fig. 5. The location of the survey area is shown in Fig. 1. Computer generated contours in this figure are from Nimbus II HRIR data for orbit number 1995. The HRIR data were filtered Copyright © 1972 by American Elsevier Publishing Company, Inc. 86 110 73*48' 73*45' ^ \ 0 ^ *•*•*' 0 ' J .?<: . ** ^2?8 GEORGE A. MAUL AND DONALD V. HANSEN 73*45' 73*48' 73*45' - , TEMPERATURE CO SALINITY (%„) "t Fig. 2a. Sea surface temperature (left), in intervals of 0.2° Celsius; salinity (center), in intervals of 0.2 parts per thousand; and density (right), in intervals of 0.05ct, units [a, = (density - 1) x 103] for the area of the overflights. The dots are the sampling points. The sampling commenced in the lower right hand corner at 1635 GMT with the Explorer steaming NNW and was completed at 1 800 GMT at the lower left. The contours were positioned by a linear interpolation between sampling points. Jlk Fig. 2b. Temperature cross-section from the BT transect located in Fig. 1 . The time and location of each lowering in the section are given on the abscissa; the geographical position of the termini and one point in Fig. 2a are noted. All soundings were obtained with the same BT and are uncorrected for the difference between the bucket values and the instrumental surface values (the BT averaged 0.9°C higher). This section was obtained approximately 4 hours earlier than the data in Fig. 2a. The Gulf Stream flow is into the plane of the page. (McMillin, 1969) to reduce electronic noise and then spatially averaged so that approximately 10 scan spot readings constitute each element of the contour matrix. Data Collection and Instrumentation Prior to the rendezvous with the aircraft at local noon, the ship determined the position and orientation of the front using mechanical BT soundings. During the overflights, surface water samples for temperature and salinity determina- tion were obtained at 0.8-kilometer intervals. A linear interpolation analysis of the horizontal distribution of temperature, salinity, and density is presented in Fig. 2a. The BT section through the area of Fig. 2a is given in Fig. 2b. Surface meteorological observations were made at five-minute intervals during the over- flights. The wind velocity was 5.6 m/sec from 350° and steady; barometric pressure was constant at 1014.5 mb; dry bulb temperature averaged 18.8°C; relative humidity averaged 44%; visi- bility was more than 35 kilometers and the sky was cloudless. (Cloud-free conditions were reported within the area of Fig. 1 by all ship and shore stations on the 1 200 GMT synoptic weather map.) Radiosonde data were obtained from the Cape Hatteras 1000 and 2000 GMT releases and from the ship at 2200 GMT from a position 50 kilometers downstream of the experiment area. A special radiosonde release at 1600 GMT was made for the experiment but the humidity sensor failed. The vertical profiles from the radiosondes and the AN/AMQ- 17 Aerograph (Beckner, 1968) thermal data obtained by the Navy aircraft were nearly identical above 990 mb, indicating that the atmospheric temperature field was laterally uniform within 1 or 2°C for several hundred kilometers. The aircraft began their pattern of overflights 87 STRUCTURE OF THE GULF STREAM SURFACE FRONT 111 at 5200-meters altitude, flying quasi-normal to the surface front. At 4100 meters the flightline was parallel to the front. At 3000 meters, 2300 meters, 1 500 meters, and 600 meters, the aircraft flew both parallel and normal to the front. A final line at 150 meters was normal to the front and extended for 30 kilometers to the northwest. The aircraft passed directly over the ship on all over- flights made normal to the front. The ART was a Barnes Model 14-320 which is responsive in the 7.3-1 3.5-micron range and has a field of view of less than 3°. The IR scanner was a Singer Reconofax IV, responsive in the 8—14- micron range; it has an automatic gain control which enhances thermal contrasts but does not allow a grey scale calibration. Laboratory calibrations indicate that the ART has a noise equivalent temperature difference (NEJT) of ±0.2°C (Peloquin et al., 1964). A temperature controlled water bath and a mercury thermometer are used for calibration; the water surface is vigorously agitated by a submerged pump to insure that a bulk temperature reading is obtained. The ART lens temperature is con- tinuously monitored during the flight to correct the readings for the effects of the operating environment on the instrument. Nimbus II HRIR views the earth in the 3.5-4.1 micron window from approximately 1200- kilometers altitude in a sun-synchronous polar orbit. The i° field of view of the HRIR optics resolves a 10-kilometer-diameter ground spot at the subsatellite point; the spacecraft is gravity- gradient stabilized to ±1° (NASA, 1966). McMillin reports that the NEJT of the HRIR is 2°-3cC. Data Presentation The temperature, salinity, and density field observed by the ship during the overflights is shown in Fig. 2a. Two steps of high temperature gradient, separated by an intermediate zone of low gradient, characterizes the surface thermal conditions; the salinity distribution is similar to the temperature distribution. The temperature- salinity (T-S) relationship results in a remarkably uniform density field across the front. An abrupt increase in sea state occurred south of the dashed line in Fig. 2a. Such changes in sea state (called Siomes) are often associated with the edge of ocean currents (Uda, 1938). The region of locally disturbed water in these data was observed on the ship's 3-cm navigation radar as a straight narrow band. The increase in wave height was about 20 cm ; the horizontal transition occurred in less than 50 meters. Fig. 2b is the vertical temperature profile through Fig. 2a. The thermal field is characterized by an inversion within the thermocline which is frequently observed in this region. The isotherms in the upper 50 meters are compacted and posi- tioned in agreement with the data of Fig. 2a and the folding of these isotherms is such as to suggest an overrunning wind-drift current. ART TEMPERATURE CO (uncorrected) Fig. 3. T heuncorrected 3000 meter ART trace is super- imposed on the IR scanner image taken at the same altitude. The scales along the top and right side gives distance in kilometers; non-linearity of the scales is discussed in the text. There was a sharp increase in sea state south of the southernmost (19-20°C) thermal step. The ship was very close to 36°12'N, 73°46.6'W at the time of this image pass. The area of this figure corresponds to the 2-min square centered at 36°12'N, 73°46'W. The infrared line scan image of Fig. 3, flown at 3000 meters, is representative of the series. Super- imposed is the temperature trace from the ART, also from 3000 meters, which applies to the centerline of the scan image. The image is a negative print, that is, cold water appears lighter in tone. The form of the thermal field is in good qualitative agreement with the linear interpreta- tion in Fig. 2a. In comparing the two figures, one 88 112 GEORGE A. MAUL AND DONALD V. HANSEN must recognize that Fig. 3 covers a small region in the upper right of Fig. 2a. The transverse coordinate in IR scan images is linear in arc length along the cylindrical film platten. A point at an angle 8 off the nadir corresponds to a horizontal displacement Htand, where H is the height of the instrument.1 The longitudinal scale is the ratio of film speed to flight speed and is usually adjusted to be the same as the transverse scale for some 6. Fig. 4. ART traces normal to the direction of the Gulf Stream surface front. The altitude in meters at which the data were obtained is labeled at both ends of each trace. The data are instrument readouts, in degrees Celsius, uncorrected for lens temperature. Four ART traces, normal to the stream and adjusted to be the same scale, are presented in Fig. 4. The most obvious feature is the decrease of recorded temperature with altitude. There is also a more subtle but measurable attenuation of the temperature difference across the front. The total temperature difference across the front registered at 3000- meters altitude is about 1°C less than that registered at 1 50 meters. Recorded gradients in each step may be limited by the response time of the ART which is 4.2°C/km at normal flight speeds (Beckner, 1968). The lower triplet in Fig. 5 is three consecutive scans of the unfiltered, digitized, Nimbus II HRIR data through the study area. Scans are 1 It is implicit that the transverse scale also varies laterally across the image. If scale (5) is defined as 81/8L, where / (=R0, where R is the radius of the printer barrel) is the arc distance from the centerline of the image and L is the corresponding distance on the ground plane, then dl IdL s=WyrR/Hsec2*- which differs by 1/sec 6 from that erroneously given by Taylor and Stingelin (1969). Fig. 5. In the lower half of this figure are three scans of the Nimbus II HRIR across the region of interest for data orbit 1995, 12 October, 1966. The latitude, longitude, and nadir angle for the relevant portion of the middle scan are noted below. A smoothed composite scan, which is displaced vertically for clarity, is the lower profile in the upper triplet; the uppermost and middle profiles are filtered versions of the data by the authors noted below profile. parallel to each other, separated by about 8 kilometers, and the nadir angle of the data varies from —39° on the extreme left of the figure to —22° on the right. Azimuth of the scans is 278- 098°, the center one of which is the dashed line passing through the area of ground truth outlined in Fig. 1. The lower profile of the upper triplet is a smoothed scan formed by averaging the scan spots with the same nadir angle from each of the three scans, and further smoothing by a three- point running mean. The essential features are a rapid increase in radiation temperature as the scan spot sweeps from the land to the sea and a second rapid increase when sweeping onto the Gulf Stream. Filters to minimize noise of sundry origins designed by Vukovich (1970) and McMillin (1969) were applied to the scan that passed through the ground truth area (solid line in the lower triplet) and are the uppermost and middle profiles of the upper triplet, respectively. 89 STRUCTURE OF THE GULF STREAM SURFACE FRONT 113 Discussion Sea Surface Conditions Except for the lowered gradient as discussed for the ART data earlier, the data of Fig. 3 are in agreement with the ground truth of Fig. 2a. The northern thermal step appears sharper on the scanner image which is surprising since the change in sea state is associated with the southern step (there was no visual evidence of the northern thermal step). The horizontal scale of the inter- mediate thermal step in the ART data is essentially the same in both the IR scanner image and the ground truth. This intermediate zone is approxi- mately one kilometer wide and it retains its width and orientation for at least 17 kilometers downstream. The zone is parallel to the direction of the 15°C-isotherm at 200 meters in agreement with a statistical correlation of the surface and subsurface fronts by Hansen and Maul (1970). The mixture of surface water types in the intermediate zone is approximately three parts Gulf Stream to one part slope water as defined by the grouping of points on a T-S diagram. Another such intermediate step, with a tempera- ture reflecting a similar 3:1 mixture, has been reported (U.S. Naval Oceanographic Office, 1970). An estimate of the velocity shear across the frontal region was obtained from set and drift calculations based on Loran A navigation. Fixes were taken at 2-kilometer intervals. Calculated drifts were 50 cm/sec in the 21°C water north of the stream and 180 cm/sec in the 26°C water to the south; the set was 60° on both sides of the front which is in agreement with the orientation of the thermal field. The shear maximum occurs in proximity to the strong thermal gradients. The data do not permit further refinement of details of the surface velocity field, but the general situation is in agreement with observations on other occasions of a shear discontinuity in the ship's wake when crossing the edge of a Siome. IR scanner imagery shows the northern edge of the intermediate zone to be characterized by a saw-toothed pattern of cool water extending into the warmer water. The aspect ratio of the intru- sions is about 2:1, with lengths approximately 100 meters, and widths 50 meters. The streaks are oriented in the direction of the wind and demon- strated no deflection which tends to support the analysis of Krauss (1965) where an uncoupled solution was found to the Navier-Stokes equa- tions leading to a band structure similar to that observed in these data. IR Atmospheric Attenuation To investigate the attenuation of the ART data, the radiative transfer equation for a non- scattering atmosphere and a non-reflecting ocean (both valid assumptions in the 8-14 /u region of the spectrum) was solved using the trans- missivity model of Davis and Viezee (1964). The integrated form of Schwartzchild's equation transformed to optical wave number space is adopted from Craig (1965): /(«o) = J vLv(us)Tv(u0)dv + //*.i,Wd^d„d„ "1 «o where the optical depth (w) is positive downward from the height of the radiometer (w0) to the sea surface (us), I is the intensity of radiation received, Lv is the intensity of emission of a black body at a wave number v and is given by Planck's Law. t„(w0) is the total transmissivity of the atmosphere at u0 for a given wave number band (25 cm-1 in this model), and v is the response function of the ART filter. The first term on the right-hand side is the attenuation of the sea surface emission Lv(us) by the transmissivity of the intervening atmosphere t^Uq), intergated over the spectral response v of the radiometer, and the second term is the contribution to the received signal by the re-emission of the atmos- phere L„(u) from each level u along the vertical path of transfer, again integrated over the appropriate spectral interval; v is zero outside the interval vl to v2 and is some number less than or equal to unity within that interval. To convert from units of radiance to equivalent black body temperature (TBB), "2 J vLv(TBB)dv is solved. A slightly non-linear relation exists between / and TBB over the temperature range of interest, namely 260-3 10°K. 90 114 GEORGE A. MAUL AND DONALD V. HANSEN 14 8 18.8 22 8 28 8 TEMPERATURE (*C) 14 8 18 8 22 8 TEMPERATURE (°C) Fig. 6. Comparison of the observed (open circles) and theoretically predicted (solid lines) ART signal with altitude assuming the radiometer reading at 990 mb was the radiant input to the air column. The ART readings are corrected for lens temperature. The humidity data of the radiosonde released by the ship 50 km downstream and 20 km on the warm side of the front are used for (a). Cape Hatteras humidity data is used for (b); the atmospheric thermal field was essentially identical in both (a) and (b). Initially, in order to avoid questions concerning the emissivity of the sea surface, the equations were solved using the average ART reading at 150 meters (990 mb) as the input intensity [Lv(us)]. Results of the computations using the temperature and humidity profile from the 2200- GMT ship radiosonde are given in Fig. 6a for the warm (25.7°C) and cold (19.7°C) sides of the front ; C02 was assumed to be constant at 0.03 1 % by volume. The influence of the 7° air-sea temperature difference on the warm side is reflected in the larger slope of the theoretically predicted solid line. Comparison of the predicted and observed values gives differences of several degrees at 520 mb where the optical depth was 1.6 gm cm-2. Further, the observed values are higher than theoretical which is contrary to the anticipated results considering the assumptions in the model. After several possible explanations for the lack of agreement between the observed radiation temperatures and those calculated using data from the radiosonde released on the warm side of the Gulf Stream front, the radiosonde data for Cape Hatteras was obtained. As mentioned earlier, the temperature data of all atmospheric profiles was essentially identical. The radiative transfer equation was integrated using the mean Cape Hatteras sounding for the day. The results of those calculations are given in Fig. 6b which is in substantial agreement with the observations. It is more important, however, to note how sensitive the theoretical curve is to changing the total precipitable water vapor from 1 .6 to 1 .0 gm cm-2 over the air column. In particular, if 990 mb were sea level and the aircraft flew at 300 meters above sea level (i.e., 955 mb) an uncertainty of 0.4°C would result from lack of humidity measurements alone. Clearly, atmospheric water vapor must be considered in correction techniques if they are to give useful precision in other than average atmospheric conditions. An estimate of the radiometric temperature of the sea surface which would be observed by the ART at sea level was made. The radiative transfer equation was used to extrapolate the ART reading observed 1 50 meters above the Gulf Stream; atmospheric data collected by the ship during the overflights was employed. A tempera- ture difference (Ts minus ART) of 0.6°C is calculated. A similar comparison could not be made on the cold side due to lack of ship data. Depression of the skin temperature of a water body relative to its bulk temperature has been discussed by Ewing and McAlister (1960) and an estimating equation based on dimensional con- siderations was given by Saunders (1967). A value of 0.8°C is calculated for the bulk tempera- ture-skin temperature difference {AT) using Saunder's equation, and the proportionality constant (7) that Saunders suggested for the western North Atlantic. Hasse (1963) prepared 91 STRUCTURE OF THE GULF STREAM SURFACE FRONT 115 a figure from several sources that indicates AT '« 0.7°C for the air-sea temperature difference in this experiment. In comparing these results it should be noted that heat flux measurements using a two-channel radiometer (Foster et al., 1971 ) suggest that A T is dependent on the spectral band of the ART. The decrease of temperature difference across the front with increasing altitude is predicted by the theoretical model and is in agreement with Saunders (1970). This decrease arises in part because the fraction of the outgoing effective radiance due to atmospheric re-emission in- creases as the instrument views more and more atmosphere with increasing altitude. Primarily however, the decrease is caused by the air-sea temperature differences across the front and must be anticipated as a general effect which varies with optical depth and spectral band. Satellite Data The satellite data presented in Figs. 1 and 5 are corrected for instrument variations on an orbit by orbit basis. For any one orbit, a correction is computed from the inflight comparisons of the HRIR output with a reference black body and is applied to that data block. Corrections for clear sky attenuation of the earth's infrared emission for model atmospheres or nadir angle are not made [Barksdale (NASA), personal communica- tion]. Mean values of sea surface temperature on each side of the Gulf Stream front were calculated from the HRIR data. Ten unfiltered scan spot values with the same nadir angle from each of the three scans were used to arrive at the mean value. Warnecke et al. (1971) report that such areal averaging minimizes the effects of the NEJT on the data. A temperature difference of 7°C is observed by the HRIR across the front. This value is 1°C greater than the 150-meter altitude aircraft data, and is in contradistinction with the ART's recorded decrease with altitude. A higher temperature difference in the satellite data reflects that more of the cold water to the west of the Gulf Stream had to be included in order to obtain a statistically significant step function approximation although it does qual- itatively agree with Kunde's (1965) theoreti- cally predicted increase in Ts minus HRIR due to higher nadir angles to the west (limb darken- ing). 10 The average temperatures for the Gulf Stream and for the slope water were used to com- pare the HRIR and the ART. Both instruments recorded 24°C as the temperature in the Gulf Stream. The HRIR value for the slope water was 17°C; the ART gave 18°C. Kunde's curve for the ARDC standard atmosphere (which contains 1.16-gm cm"2 water vapor) was used to estimate the correction to the HRIR. For a nadir angle of 30° the corrected HRIR value for the Gulf Stream is 26°, which is essentially in agreement with the skin temperature deduced from the ART. This agreement is in part fortuitous however because the 200-Hz noise reported by McMillin increases an averaged reading. Hand contoured maps of the unfiltered and filtered Nimbus II HRIR scan spot listing for the experiment reflect in two dimensions the vari- ability described in Fig. 5. (See, for example, McMillin, 1969.) Typical patchiness in the un- filtered, and to a lesser degree in filtered, data is essentially lost in the averaging process used to produce Fig. 1 . In this figure a base scale was used which provides approximately 10 hltered scan spots averaged into each grid point; there are 20 grid points per square degree of latitude and longitude. The orientation of the thermal step across the Gulf Stream is reflected in the map, but due to spatial averaging the temperature gradient is reduced by a factor of four relative to the ship data in both the map and the scan spot data. The geographic location of the center of maximum temperature gradient from the data of Fig. 5 coincides with the map of Fig. 1 and is essentially identical with the position of the front determined by the ship. Summary and Concluding Remarks 1 . Dense sampling of temperature and salinity across the Gulf Stream front revealed two regions of high temperature and salinity gradient, separated by a low-gradient zone one kilometer wide. The density of the surface water was remark- ably uniform, but had a slight maximum near the edge of a Siome. These surface features are parallel to the deep structure of the Gulf Stream front and they coincide with the region of surface velocity shear maximum. 2. IR scanner imagery and ART data taken simultaneously by aircraft add detail to and are in quantitative agreement with the ship data. 116 GEORGE A. MAUL AND DONALD V. HANSEN Both the ART signal and the gradient are at- tenuated with altitude; the essential features of the attenuation are given by a theoretical model which demonstrated the sensitivity of the Barnes Model 14-320 radiometer to minor perturbations in the integrated mixing ratio below the aircraft. Downward extrapolation of the ART data supports the calculated depression of the skin- bulk sea surface temperature, and suggests the need for a two-part correction technique. 3. Under the cloud-free conditions of this day, the spacecraft data are in good agreement with the skin temperature when corrected for theor- etical attenuation by the atmosphere and agree with the geographical location of the Gulf Stream front. Step averaged temperature differ- ences across the Gulf Stream are in accord with the ship data but the gross gradient is reduced by a factor of four due to the relatively poor resolving power of the HRIR sampling pattern. Future experiments of this type should be done with infrared filters on aircraft and ships that are identical in response and range to those on orbiting vehicles in order to compare the data. This was not possible with these data because of the different windows used in the instruments. The 10.5-12. 5-micron filter used on Nimbus IV and 1TOS appears to be a favorable region of the spectrum for future ground truth measurements so that the data obtained from space can be evaluated. Acknowledgements The authors are indebted to the crews of the ship and aircraft involved in obtaining these data and especially to R. M. Nelson who initiated the project and coordinated the collections. The advice and assistance of W. L. McLeish, D. J. Pashinski, P. A. Bush, and H. V. Donn are gratefully acknowledged. This research was in part supported by the National Aeronautics and Space Administration's Earth Resources Survey Program through the National Environmental Satellite Service's Environmental Science Group. References C. F. Beckner, Jr., Comparisons of Remote Airborne Oceanographic Sensors, U.S. Naval Oceanographic Office, TR-204( 1968). R. A. Craig, The Upper Atmosphere: Meteorology and Physics, Academic Press, New York (1965), pp. 281-286. P. A. Davis and W. Viezee,/. Geophys. Res. 69, 3785 (1 964). G. Ewing and E. D. McAlister, Science 131, 1374 (1960). T. D. Foster, E. D. McAlister, and S. Rearwin, Trans. A.G.U. 52, 255(1971). D. V. Hansen, Deep-Sea Research 17, 495 (1970). D. V. Hansen and G. A. Maul, Remote Sens. Environ. 1, 161 (1970). L. Hasse, Tellus XV, 363 (1963). W. Krauss, Deutsch. Hydr. Zeit 18, 5 (1965). V. G. Kunde, Theoretical Relationship between Equivalent Blackbody Temperatures and Surface Temperatures Measured by the Nimbus High Resolution Infrared Radiometer, NASA SP-89 (1965), p. 23. L. M. McMillin, A Procedure to Eliminate Periodic Noise Found in Nimbus II High Resolution Infrared Radio- metric Measurements, Technical Report No. 9, NASA Contract NAS 5-10343 (1969). NASA, Nimbus II User's Guide, Goddard Spacecraft Center (1966). R. A. Peloquin, J. C. Wilkerson, and G. L. Hanssen, Bureau of Sport Fisheries and Wild Life Circular No. 202(1964). P. M. Saunders,/. Atmos. Sci. 2A, pp. 269-273 (1967). P. M. Saunders,/. Geophys. Res. 75, 7596 (1970). J. I. Taylor and R. W. Stingelin, /. Hydraulics Div., A.S.C.E.95, 175(1969). M. Uda, Geophysical Magazine VII, 307 (1938). U.S. Naval Oceanographic Office, The Gulf Stream, Monthly Summary 5 (1970). F. M. Vukovich, Physical Oceanography Feasibility Study Utilizing Satellite Data: Part II, Final Report to National Environmental Satellite Center, ESSA, Research Triangle Institute (1970). G. Warnecke, L. J. Allison, L. M. McMillin, and K. H. Szekielda, J. Phys. Oceanography, I, 45 (1971). 93 11 REMOTE SENSING OF ENVIRONMENT 2, 165-169 (1972) Comment on "Estimation of Sea Surface Temperature from Space' by D. Anding and R. Kauth GEORGE A. MAUL Physical Oceanography Laboratory, National Oceanic & Atmospheric Administration, Atlantic Oceanographic & Meteorological Laboratories, Miami, Florida 33130 MIRIAM SIDRAN* N.S.F. Science Faculty Fellow, National Oceanic & Atmospheric Administration, Southeast Fisheries Center, Miami, Florida 33149 165 In the course of investigating the remote sensing of sea surface temperature with the NOAA 1 scanning radiometer, the calculations of Anding and Kauth (1970) were checked using the atmospheric transmissivity model of Davis and Viezee (1964). We found that according to this model, the 11.0 /zm/9.1 ^m band pair proposed by Anding and Kauth will not provide the information needed to compensate for water vapor in a cloud-free atmosphere, because the transmissivities in these bands due to water vapor are almost identical. The model developed by Davis and Viezee uses analytic expressions for calculating the infrared transmissivity through water vapor and carbon dioxide. Coefficients for these expressions are tabulated in intervals of 25 cm-1 over the range 25-2150 cm-1; the environmental variables are temperature, pressure, and amount of absorbing gas. The mean transmissivities are computed over finite spectral intervals, and this makes it possible to solve the radiative transfer equation in the form given by Craig (1965) for an absorbing, emitting, but nonscattering atmosphere. The validity of the model for the lower atmosphere has been established by Saunders (1970) and others. Our calculations were carried out on the National Oceanic and Atmospheric Administra- tion's computer facility at Suitland, Maryland. Input variables for the program were sea surface temperature, and atmospheric data in radiosonde format, i.e., temperature and relative humidity as a function of pressure. Amounts of absorbing gases along the path were calculated from the radiosonde data; the atmosphere-centimeters of * On leave from New York Institute of Technology. carbon dioxide were calculated assuming that C02 is uniformly mixed 0.031 % by volume. The emissivity of the sea surface was assumed to be unity. A vehicle altitude of 1200 km was chosen as being typical of a sun-synchronous orbit. The curvature of the earth was taken into account. (At this altitude and at a 60° nadir angle, the path length is 20% longer than if the earth were assumed planar.) Transmissivities were calculated at one millibar intervals, from the surface up to one millibar (~48 km). . Anding and Kauth had plotted the calculated spectral radiance at 1 1 .0 ^.m versus that at 9. 1 fj.m (their Fig. 6) and had obtained a unique straight line for each surface temperature, thus correcting for water vapor effects. They found an rms error of 0.1 5°C in estimating sea surface temperature from this graph. As a first order check on these results, we did not attempt to use exactly the same band widths or central wave lengths. Instead, we plotted in Fig. 1 the spectral radiance at 912.5 cm-1 (10.96 yu,m) versus that at 1087.5 cm-1 (9.19 fxm). Our calculations covered a wide range of sea surface temperatures (Ts) and geographic- ally meaningful atmospheres (U.S. Standard Atmosphere Supplements, 1966), which are listed in Table I. In contrast to the results of Anding and Kauth all points in Fig. 1 fall on the same straight line. An explanation of our results is found by examin- ing the analytic expression used by Davis and Viezee in this region of the infrared spectrum, rv = exp [-(kv WP)a*], where rv is the transmissivity for the 25-cm_1 wave-number interval centered at wave number v, Copyright © 1972 by American Elsevier Publishing Company, Inc. 94 166 GEORGE A. MAUL AND MIRIAM SIDRAN I 550 •104 Radiance (X=9.l9^m)^Wcm'2sr'l/Am~l i E o ID CO Radiance (v =108 7.5cm ~l)/iWcm~2sr~lcm Fig. 1. Radiance at 912.5 cm"1 (10.96 /xm) versus radiance at 1087.5 cm-1 (9.19 iim). Even degree values of Ts are dots; odd degrees are crosses. For each value of Ts the upwelling radiance was computed through three model atmospheres and for each 10° of nadir angle from 0° to 60°. The spectral radiances from different values of Ts overlay one another which does not allow a unique solution for atmospheric correction. W is the amount of precipitable water vapor in cm, P is the ratio of in situ pressure to standard pressure, and k„ and a„ are constants. The values of kv and a„ for the wave number interval 900-925 cm-1 are 0.095 and 0.885 respectively; for the interval 1075-1 100 cm"1 they are 0.091 and 0.880 respectively. The near equality of these constants for each of these spectral bands means that the transmissivities due to water vapor are approxim- ately equal. To proceed with the explanation of our results, consider an atmosphere containing water vapor that absorbs but does not emit. Then the equation of radiative transfer in wavelength space reduces to N,d\ = T,Lx(Ts)d\, where Nx is the radiance of wavelength A, ta is the transmissivity of the entire atmosphere, and Lx{Ts) is the blackbody radiance at the tempera- ture Ts. The radiance (wavelength units) L9l)im is nearly equal to L110fim for realistic ocean surface temperatures. Since t9 Afim x Tn,0/im then N9.lltm&NllmQltm for all Ts. That is, a decrease in t due only to H20, would decrease the radiance in both channels by about the same amount. Thus a plot of N9lfim versus Nu.o /^m would yield one straight line (of slope about unity) for all Ts, and for all nadir angles and amounts of precipitable water. Emission by the atmosphere would change the slope some- what, as in Fig. 1. Davis and Viezee's transmissivity model was 95 COMMENT ON ANDING AND KAUTH 167 Table I Tabulation of the sea surface temperatures (7",) and the model U.S. Standard Atmospheres used in the calculations. For each atmosphere/7"s pair are listed : The amount of precipitable centimeters of water vapor, the atmosphere-centimeters of carbon dioxide, the temperature departure AT(AT=TS-TC where Tc is the calculated equivalent blackbody temperature at the top of the air column) at nadir viewing for the 900-925 cm-1 spectral interval, and the surface air temperature (7",,). T "S Atmos phere H?0 CO 2 AT Ta (°C) (pr.cm. ) (at . -cm . ) (°C) CO 2 8° 15°N Annual 4.0.5 248.5 6 .2 26.5 30°N January 2 .08 250.2 6.5 14.0 30°N July 4. 36 248.8 6.8 28.0 2 3° 30°iJ January 2.08 250.2 5.0 14 .0 30°N July 4. 36 248. 8 4.2 28.0 45°N July 2 .96 248.6 4.6 21.0 18° 30°N January 2 .08 250.2 3.5 14 .0 4S°N January 0.80 249.3 3.6 - 1.0 45°N July 2.96 248.6 2.6 21.0 13° 45°N January 0.80 249 . 3 3.0 - 1.0 45°N July 2.96 248.6 0.5 21.0 6 0°N July 2.07 247.5 2.9 14 .0 8° 45° N January 0.80 249.3 2 . 3 - 1.0 45°N July 2.96 248.6 -1.6 21.0 60°N July 2 .07 247.5 1. 3 14 .0 3° 45°N January 0.80 249. 3 1.6 - 1.0 60°N January 0. 36 248.2 1.5 -16 .0 60°N July 2.08 247.5 -0.3 14 .0 -2° 45°N January 0.80 249.3 0.9 - 1.0 6 0°.N January 0.36 248.2 1.2 -16.0 60°N July 2.08 247.5 -1.9 14.0 used to find line separation for different Ts. We retained the 912.5-cm-1 band as the basis of comparison, and plotted in Fig. 2 the radiance in this band versus that in the 1 162.5-cm-1 (8.60-/^m) band which is closer to the center of the water vapor absorption line. (For the 1 150-1 175-cm_1 interval, A „ = 0. 1 1 5 and a, = 0.795.) This band pair was the best compromise between maximum separation of T5 values and minimum scatter, according to our model. Our results are not as encouraging as those of Anding and Kauth; using their Eq. (4), the rms error in the estimated 28c surface temperature is 0.6°C. This is because each T± line is really a family of closely spaced lines each corresponding to a different tempera- ture-humidity profile. At first, we tried to explain the discrepancy between our results and those of Anding and Kauth as due to the inclusion of ozone in their model. However, Kauth (1972, personal com- munication) was kind enough to show that this was not the case. We now believe that the dis- crepancy is due to the use of different trans- missivity equations and that therefore both sets of results are model-dependent. The expression used by Anding et al.(\91\) for water vapor transmissivity (t) is t(AX) = exp[~(W ■ K[AX))l/2] where K(AX) is the spectral absorption coefficient and W is the equivalent optical depth due to water vapor, given by W i M dr where P is the atmospheric pressure (mm Hg), T is the atmospheric temperature (°K), r is the range (cm), p is the atmospheric density (g cm-3), M is the mixing ratio (g H20/g air), P0 is the 96 168 GEORGE A. MAUL AND MIRIAM SIDRAN Radiance (X = 8.60/zm)/i.Wcm'2sr'l/i.m'1 Radiance (v = \ l62.5crrrl)yu.Wcm"2sr"lcm Fig. 2. Radiance at 912.5 cm-1 (10.96 /im) versus radiance at 1162.5 cm-1 (8.60 ^m). Even degree values of Ts are dots; odd degrees are crosses. For each value of T„ the upwelling radiance was computed through three model atmospheres and for each 10c of nadir angle from 0° to 60°. The straight line through each Ts family is the least-squares best fit. The lines were not constrained to pass through the point of no atmospheric effect although McMillin (1972, personal communication) suggests they should. standard pressure and T0 is the standard tem- perature. From their table of spectral absorption coefficients we estimate that K(AX) = 2.72 x 10"3 for the 900-925 cirr1 interval, and K(A\) = 2.35 x 10~2 for the 1075-1 100 cm"1 interval. The large difference in K(A\) is the reason for the success of Anding and Kauth's model. It causes the t's to be unequal, and leads to the separation of lines in their Fig. 6. We checked the transmissivities reported by several other investigators, and computed their ratios. From Taylor and Yates (1957, Fig. 8) we scaled a transmission ratio tu m/t9 2 = 1.15 for a horizontal path containing 1.37-cm precipitable water in an atmosphere very near NTP. For these same conditions, Anding and Kauth's equation yielded t1096/t919 = 1.13,andDavisandViezee's expression gave t912 .5/t1087 5 = 0.996. Kond- rat'ev et ah (1966) used an expression r = exp(— KW) where we estimate Kgi2.s = 0.107 and ^087.5 = 0.090; again for W=l.37 cm, T9i2.s/Tio87.s = 0.977. Finally we scaled a ratio tii.o/t9.2 = 0.975 from Burch (1970, Fig. 3-1); this ratio is not for the same atmospheric con- 97 COMMENT ON ANDING AND KAUTH 169 ditions, which we kept constant for the four previous comparisons. Thus, each model yielded a different ratio of r for the two channels. From this we conclude that the degree of separation of lines of equal Ts depend on the transmissivity model used. In conclusion, we feel that the family of solu- tions suggested by Anding and Kauth is a signifi- cant step in the remote sensing of the sea surface temperature. However, our calculations suggest that the separation of lines of different Ts depends on the transmissivity model employed. The Sur- face Composition Mapping Radiometer (origin- ally designed for land use) to be flown on Nimbus E later this year, and the Sea Surface Temperature Imaging Radiometer proposed for the Earth Observatory Satellite both use the ll.0-ju.rn/ 9.1-/zm band pair. If the two channels have the limitations which our results would indicate, these instruments will not be useful for sea surface temperature measurements. In the meantime, we invite other investigators with different transmissivity models to check our results, and communicate their ideas on this subject. References Anding, D. and R. Kauth (1970), Estimation of sea surface temperature from space. Remote Sensing 1, 217-220. Anding, D., R. Kauth, and R. Turner (1971), Atmospheric effects on infrared multispectral sensing of sea surface temperature from space, NASA CR-1858, Washington, D.C. Burch, D. E. (1970), Semi-Annual Technical Report, investigation of the absorption of infrared radiation by atmospheric gases, Philco-Ford Corp., Aeronutronic Division, Newport Beach, Calif., p. 5-2. Craig, R. A. (1965), The Upper Atmosphere, Meteorology and Physics, Academic Press, New York. Davis, P. A. and W. Viezee ( 1 964), A model for computing infrared transmission through atmospheric water vapor and carbon dioxide, /. Geophys. Res. 69, 3785-3794. Kondrat'ev, K. Ya., Kh. Yu. Niilisk, and R. Yu. Noorma (1966), The spectral distribution of the infrared radia- tion in the free atmosphere, Izv. Atmos. & Ocean. Phys. 2, 121-136. Saunders, P. M. (1970), Corrections for airborne radiation thermometry, J. Geophys. Res. 75, 2796-7601. Taylor, J. H. and H. W. Yates (1957), Atmospheric transmission in the infrared, /. Opt. Soc. Am. 47, ■ 223-226. U.S. Standard Atmosphere Supplements (1966). Super- intendent of Documents, U.S. Government Printing Office, Washington, D.C. 12 U.S. DEPARTMENT OF COMMERCE National Oceanic and Atmospheric Administration Environmental Research Laboratories NOAA Technical Memorandum ERL AOML-18 FORMULATION OF DRIFTING LIMITED CAPABILITY BUOY PLACEMENT AND RETRIEVAL CONCEPTS Robert L. Molinari Donald V. Hansen Atlantic Oceanographic and Meteorological Laboratories /*% J°**i Miami, Florida * April 1973 % 99 TABLE OF CONTENTS Page ABSTRACT 1 1. INTRODUCTION 1 2. SUMMARY OF PREDICTION MODELS 2 2.1 Coast Guard Search and Rescue Plan CG 308 Procedure 2 2.2 SAR Center, Seventh Coast Guard District Procedure 6 2.3 FNWC Search and Rescue Computer Program 7 2.4 Coast Guard SARP Computer Program 8 2.5 Balloon Location 9 3. RECOMMENDATIONS FOR FUTURE DIRECTION 10 3.1 Introduction 10 3.2 Current Determinations 13 3.2.1 Pre-Launch Stage 13 3.2.2 Launch Stage 21 3.2.3 Drift Stage 23 3.2.4 Failure Stage 23 3.3 Recommendations for the DLCB Test Program 26 4. ACKNOWLEDGMENTS 2 8 5. REFERENCES 29 100 FORMULATION OF DRIFTING LIMITED CAPABILITY BUOY PLACEMENT AND RETRIEVAL CONCEPTS Robert L. Molinari Donald V. Hansen This report is submitted in response to the National Data Buoy Center's requirement for a "Formulation of Drifting Limited Capability- Buoy Placement and Retrieval Concepts". The search and rescue methodologies of the agencies engaged in finding lost objects at sea are sum- marized, and a method for predicting the drift of a buoy subject to surface current stresses is suggested. Finally, a test program for the Drifting Limited Capability Buoy is offered. 1 . INTRODUCTION The "Formulation of Drifting Limited Capability Buoy (DLCB) Placement and Retrieval Concepts" as delineated in National Data Buoy Center (NDBC) Statement of Work 0134EC, consists of two tasks. The first task requires a "summary of buoy location prediction models" and the second "recommenda- tions for future NDBC direction". The requirements have short term aspects related to the field testing of the buoys during the R and D phase of the program and long term opera- tional aspects. Primary emphasis in this report is given to drift prediction and buoy retrieval concepts as placement concepts are determined by the experiment being conducted, and will require detailed individual analysis. However, s suggestions concerning deployment during initial DLCB tests will be presented. The procedures of the United States Coast Guard and the United States Navy's Fleet Numerical Weather Central (FNWC) at Monterey, California, are summarized, as both groups are actively engaged in finding lost objects at sea. The Coast Guard Search and Rescue (SAR) plan, given in Coast Guard Manual CG-308, is operationally oriented and strives for an some 101 engineering handbook solution to the problem. Each step of the SAR procedure is presented with accompanying charts and figures to facilitate the computations. The FNWC has developed a computer program in support of the Coast Guard SAR operations. The program uses sea surface current and wind forecasts generated at the FNWC to determine the probable drift of a SAR object. The Coast Guard is presently instituting an operational SAR computer program which is more sophisticated than the FNWC model. The problems encountered in balloon location are similar to those found in buoy location. The work of Quinlan and Hoxit (1968) is summarized as an example of the meteorolo- gist's approach to balloon location. Other reports were reviewed which discussed short-term tracking of balloons and rockets (Young 1962, Rachelett and Armendary 1967, for example) but their procedure is too specialized for applica- tion to the buoy problem. Recommendations for future NDBC action are given in a following section. A procedure for predicting drift and re- trieving non-functioning buoys, as well as a test program for the DLCB are offered. The specialized nature of the DLCB's and the methods employed by various SAR agencies are considered in these recommendations. 2 . SUMMARY OF PREDICTION MDDELS 2.1 Coast Guard Search and Rescue Plan CG 308 Procedure Chapter 6 of CG-30S pertains to the "Determination of Search Areas". Relevant sections of this chapter discuss the means of estimating the probable position of a SAR object, and determining the search area. The position of the lost craft is based on a probable drift from the initial location of the SAR incident. The object's drift is computed as a function of three variables, the average sea current, the local wind current, and the wind's effect on the object (leeway). All variables are tabulated on maps, charts, or figures. For instance, the average sea current is obtained from one of three sources, 102 which in order of preference are U.S. Naval Oceanographic Office Atlases, Oceanographic Office Atlas of Surface Cur- rents, and Pilot Charts. Section I of the Oceanographic Atlas of the North Atlan- tic (Publ. No. 700) is an example of the first publication, and contains tides and currents of the region. Two surface flow charts are presented, one combining data taken in the summer months of July, August, and September and the other for the winter months of January, February, and March. Cur- rent directions are given by arrows. The color of the arrow indicates the variability of the flow, with only the pre- vailing currents to be used in the computations. Isotachs are plotted on overlays to the direction charts, permitting determination of the velocity vector at any point. The Atlas of Surface Currents, such as Publ. No. 576, are being replaced by Publ. No. 700-type publications, and are used only in areas where these publications do not exist. Monthly current representations are given for the region of the particular atlas. Pilot charts, which also give monthly currents, are mentioned as a last resort and only in regions where no other information is available. Technical reports of NOAA, the Naval Oceanographic Office, and interested re- search groups are recommended as a source of supplementary data. The wind induced current is obtained from figure 1, which is taken from James (1966). James considered the effect of wave transport as well as pure wind drift to arrive at the functional dependence of the current on the wind speed, the fetch over which the wind acts, and the duration of the wind. Although James considers the deviation of the drift from the wind to be 20° to the right in the Northern Hemisphere, the Coast Guard tabulates the drift as a function of latitude. Both wind speed and the "sail" characteristics of the lost object are used to determine leeway. The larger the sail presented to the wind, the greater the drift caused by the wind. Figure 2 is given to aid in leeway determinations, the curves representing the leeway of a liferaft with and without drogue. Small objects other than liferafts are as- sumed to have leeway curves falling between the two pre- sented. Although most boats will tend to drift off the downwind line, the direction of the leeway vector is consid- ered downwind. The three terms are combined vectorially to give the drift, and thus probable position, of the SAR object. Com- putations are continued to arrive at an estimate of the error 103 (WN) HD13d ^ 0r> O m § O u> o the FNWC so shown. LO o used h t are , — , K if) 0) c» T e Oi o o >, £^ rt3 ts o 5 11 o .J £§ (T> +^ ca o th and w ion to I CVJ 3 3 o ferafts oontrih ^ o O S) <3 0) o 3 <3i ca O o Figure 2. Le for percent in the computations. This quantity is a function of the error in the last known position of the object, the navi- gational errors of the search vessel and the drift error. This last error is taken as the total drift divided by- eight, that is a 12$ error is automatically assumed to exist in the drift computations. An important part of Coast Guard SAR strategy is a measurement of the drift at the last known position (datum point) of the object. The effectiveness of this strategy is expected to be a strong function of the promptness of search vehicles on the scene. Ideally, a datum marker buoy, an air- droppable floating beacon which transmits on UHF frequencies, is deployed at the datum point. Listed as less desirable methods of determining the datum drift are dye markers and/or drifting ships. 2.2 SAR Center, Seventh Coast Guard District Procedure In discussions at the SAR Center of the Seventh Coast Guard District, modifications to the CG-30# procedures were outlined. These modifications are attempts to cope with the diverse current regimes found in the large area of responsi- bility of this District. The particular SAR procedure applied is dependent on the current regime present and the data avail- able. The majority of SAR incidents in this district occur off the Florida East Coast in the area of the Florida Current. The SAR Center, in conjunction with oceanographers of the University of Miami, has developed a climatological chart of the Florida Current. The chart encompasses the entire Flori- da Strait. Geostrophic and direct current measurements taken by the University are averaged to give contours of average surface velocity. When the last position of the lost object is known, it's drift speed and direction are determined from the nearest isotach. If an exact position cannot be ascertained the ob- ject is put in "maximum peril", for which the core velocity of the current is used in the drift computation to delineate the search area. In regions other than the Florida Strait, monthly charts such as Publ. No. 576 and Pilot Charts are used, rather than Publ. No. 700. A comparison of the Pilot Chart and Publ. No. 576 velocities is made; and if differences exist, they are subjectively averaged to obtain the current velocity. The leeway component is tabulated as a function of the 6 106 lost object. A percentage of the wind speed, dependent on the sail area, is used as the velocity of the leeway. The deviation from downwind direction is also given in a table. For instance, the maximum deviation of a sailboat has been empirically determined as 60°, either to the right or to the left of the wind direction. The search area must be enlarged accordingly to compensate for this unknown drift direction component. This Center is compiling a store of observational cur- rent measurements as datum marker buoys are deployed during most search operations. The use of these buoys follows the procedures set down in CG-30#. One difficulty encountered in applying this technique is obtaining accurate positions when the plane cannot sight land. The Center personnel do not use the Monterey Search and Rescue program operationally. They indicate that the area prescribed by this program is often prohibitively large to launch a successful search because of the poor spatial res- olution afforded by this model relative to the SAR problems most frequently encountered at the Center. Occasionally they do use the program to verify their calculations. 2.3 FNWC Search and Rescue Computer Program Hubert, Hinman, and Mendenhall (1970) describe the com- puter program developed at the FNWC to predict the position of an object lost at sea. The program uses in-house generat- ed forecasts of surface current and wind to compute the prob- able drift of a craft. SAR missions are initiated upon re- quest from a suer who must provide the details of the SAR incident. Larson and Laevastu (1971) describe the procedure for forecasting surface currents from local temperature structure and wind stress. The density field is approximated by the average temperature of the upper 600 feet of the water column in the calculations to determine "the permanent flow compo- nent". The average temperature is a weighted mean of the sea-surface temperature and 600-foot temperatures which are both computed twice daily from ship reports. Horizontal tem- perature gradients are then substituted into the following simplified version of the thermal wind equation to arrive at the velocity, u = (-gz/fT) (BT/ay) v = ( gz/fT) (8T/dx) 107 where, x,y) = (east, north) directions u,v) = (east, north) components of velocity g = acceleration of gravity z = 600 feet T = average temperature f = Coriolis parameter. Witting' s formula is used to determine the wind-induced current. Empirically derived in the early 1900's from lightship observations (Defant, 196l), the induced flow is proportional to the wind speed. The deviation of this cur- rent, D, from the wind direction, W, is given by D = 40.0 - S.OW^ W< 25m/sec D = 0 W > 25m/sec where the deviation is to the right in the northern hemi- sphere. The FNWC approximations to the liferaft leeway curves determined by the Coast Guard are given on figure 2. Options are available for calculating the leeway of craft other than liferafts. A linear percentage of the speed, depending on the sail area of the object, is applied as the leeway com- ponent. Figure 2 also contains some percentage drift lines used in the FNWC program. Drift components are computed on a grid with 200-mile space increments. To obtain the drift at non-grid points, a non-linear interpolation is used. Drift forecasts are made every 12 hours at present, but it is planned to reduce the time interval to 6, then to 3 hours. 2.4 Coast Guard SARP Computer Program The Operations Analysis Branch of the Coast Guard's At- lantic Area Command has developed a search and rescue compu- ter program called SARP (Operations Analysis Branch, personal communication 1972). This program was initiated in response to the need for an automated SAR program, to the evolution of SAR methodology, and to the problems encountered during the applications of the FNWC model. SARP has been tested at Third Coast Guard District, New York, and due to its success will become operational in the near future. The present model adhers strictly to the SAR methodology of CG-308. Future models will include the newer techniques. SARP has a surface current file based on the climatolog- ical data of Publ. No. 700. Only currents with a steadiness 8 of greater than 55$ are on file. Supplementary information, such as the Florida Strait surface current chart constructed for the Miami SAR Center also will be input on the data file. An updated version of the SARP current file is being developed in conjunction with the Naval Oceanographic Office. Historical data are being recompiled to arrive at new monthly climatological charts. The computations also will result in a determination of the error bounds of the current means. The present method of applying a constant 12$ drift error is probably optimistic in regions of poorly known or highly variable currents. The option to override the Publ. No. 700 current file is available in the SARP program. If the user has knowledge of the current field in the region of the SAR incident, these data can be input to the SARP program. The local wind drift is determined from the chart of James (fig. 1). The program models the effect of wind shifts by considering the degradation of the previous wind drift as well as the onset of the new current. The technique described by James (1966) is used in these computations. The leeway component is computed by the CG-308 method. The user option to specify the percentage of wind to apply to the leeway speed also exists. The 200-nautical mile spacing between grid points of the Monterey program is too large for an operational SAR tool. The Coast Guard program, however, will have a space increment of approximately 30 nautical miles, with a 6-mile space in regions of numerous data such as the Florida Strait. The time increment of SARP is one hour. The Coast Guard, through a contract to a consulting firm, is developing another SAR computer program to be called CASP. This model will be more statistical in nature producing prob- ability maps of the search area. Environmental data, search plans and other possible factors will be assigned uncertainty values to arrive at these maps. The program will be tested operationally at the Third Coast Guard District, and if suc- cessful will become available to all the Coast Guard SAR Centers within two years. 2.5 Balloon Location Quinlan and Hoxit (1968) describe a technique developed to determine the trajectories of high altitude balloons. A "climatology of balloon positions" for packages launched 9 from Chico, California, to float for up to twenty hours at 90,000 feet was needed to plan recovery operations. The climatologies were computed for seven 2-week periods to ar- rive at the optimum launch time. The first step was to compute trajectories from winds aloft data stored at the National Weather Records Center Asheville, North Carolina. Displacement vectors at 5, 10, 15, and 20 hours after launch were computed from three consecutive twelve-hourly observations. Between 30 and 3& randomly se- lected sets of data were chosen for each two-week period. Target ellipses (climatologies) were then constructed to outline the probable area of impact. The two-dimensional displacement vectors were assumed to have a bivariate normal distribution with a probability density which falls off ex- ponentially in all directions from a mean value. When the standard deviations of the east and north components are equal, the distribution is circular, otherwise it is elliptical. The computational procedure involves determining the major and minor axes of the ellipses as well as the angle of orientation. The .90 and .99 probability ellipses are then drawn on a base map for use by the retrieval personnel. Ex- amples of these ellipses for a typical two-week period are given on figure 3. The optimum launch period has the small- est target area. To facilitate the retrieval operation, the probability that the balloon package will be within a circle of given radius centered at the mean position of the probability el- lipse can also be calculated. This computation results in a probability curve such as shown in figure 4. Although con- ceptually appropriate also for the drifting buoys, this technique is presently inapplicable because insufficient data on ocean current condition exist at present and, in any case, is likely to involve an impractically large search area. 3 . RECaiMENDATIONS FOR FUTURE DIRECTION 3.1 Introduction In contrast to the meteorologist's one-parameter problem, the SAR techniques of the Coast Guard and FNWC emphasize the importance of considering two variables, current and leeway, 10 110 uj +^ <3 to O to o u, K o r- 1 « CO Ci ca •> CO <^> (-J -P Si CO +1 ?H •»4 •P . o 0) to 0) 1*1 11 111 § 1 ■ > i \ \ o l-l o \ \ \ \ \ Oh o - \ \ \ \ 1 t 1 00 • :\ : \ : \ : \ f- \ • \ • \ : \ : \ w : \ '. \ \ \ • \ ; \ • o • (0 H 1 : \ : \ : \ : \ \ i June 1 ) July 1 \ \ \ © • •i oj % 1 1 1 , J, i i \ \ \ \ • © \ \ \ \ \ • © 00 © © o o o o o o o © * p> CM iH (ico sniavu ^> E « r« +i +i a tJ Ci) S^ ^ -^ K t) Q) ro <> • )n to •t^ 'Xi t) O •^> w Ss 0) s a. V ^ ^ -p 0) v 0) 3 3 <3i O C •^ S Qi rQ •\ 0 S •3 o -^ o r'^ -P ^ •^ +^> <-^> CO •^> •^> rCl tj « A^ o » s^ •^ C^4J O <1> -P ^ o •ri (5^ 12 112 in the drift computations. Intuition dictates that an object will drift at the velocity of the current in which it is floating, but it is not as clear as to how a buoy will sail in the wind. Thus, an example of the magnitude of the sail component under typical oceanic conditions is presented. Consider a 1 m/sec current flowing to the north and a 10 m/sec wind acting in the same direction. If the duration of this wind is two hours, an insignificant wind induced cur- rent will result. Assuming the DLCB approximates a liferaft with drogue, the 10 m/sec wind will produce a leeway of |- knot. This is a 25$ and 10 mile per day increase in the re- sultant drift. If this component is not considered in the drift compu- tations, the probability of failing to locate the drifter increases. If included, the additional mileage would shift the center of the search area. Furthermore, the search area would have to be enlarged to compensate for the possibility of non-downwind drift if the sail characteristics of the object are not known. The sailing characteristics of the DLCB are a function of the shape of the drifter, its height above and below the water, whether it is drogued or not, and other variables. These factors will have to be determined during the DLCB Test Program, and thus the leeway component of the drift can- not be realistically considered in this report. Therefore, the following sections include recommendations for determin- ing only the current components of drift and for conducting a test program for the DLCB. Specific examples, taken from studies conducted by AOML personnel, are used to argue for various procedures. 3.2 Current Determinations The knowledge of precisely where and when the DLCB failed and what the previous drift was greatly facilitate the task of locating the buoy. This section considers the DLCB drift as a function of only the current and offers a drift prediction plan consisting of pre-launch, launch, drift, and failure procedures. Some of these procedures are more easily applied during the test program than for operations, but may be operationally applicable in the context of large-scale geophysical experiments from which extensive supporting observations are available. 3.2.1 Pre-launch Stage Knowledge of the initial DLCB deployment area should be 13 113 used to predict the probable drift of the DLCB, and to devel- op the recovery plan for a non-functioning buoy. It is rec- ommended that all possible sources of information be consid- ered to satisfy the prediction and recovery requirements. While describing these data sources, the limitations and ad- vantages of various procedures given in Chapter II also are discussed. The ideal data source for the prediction and recovery operations is a research project taking observations in con- junction with the DLCB experiment. For instance, if a DLCB is to be launched during a MODE or GATE type experiment, all the information necessary to predict the drift of the buoy is being collected. However, if such an experiment does not exist, it is recommended that the NDBC ascertain if any groups are actively engaged in other descriptive studies of the area. Some examples of such data sources follow. Dr. W. D. Nowlin, (Personal communication) of Texas A&M University, in conjunction with the NDBC and the Environmen- tal Data Service (EDS), is attempting to develop a seven-day analysis program for the Gulf of Mexico. Temperature data collected from the Buoy Center's EEP buoys and from ships operating in the region would be forwarded to Dr. Nowlin from the EDS and NDBC. Computer programs have been developed which will produce contours of the depth of selected iso- therms. Since the temperature field closely parallels the density field in the Gulf, the flow pattern is approximated, at least in the region of the major currents. Continuing projects of the U.S. Naval Oceanographic Of- fice and the Coast Guard produce maps of surface temperature which, although not as valuable as the previous type of study, could be used to predict the surface flow. The Gulf Stream series of the Oceanographic Office presents monthly charts of the Gulf Stream Axis and mean surface temperature in the northwestern Atlantic Ocean. Also included are de- scriptions of any anomalous circulation features which were recently investigated. The Coast Guard disseminates the results of their monthly aircraft infrared thermometer mea- surements of the waters off the eastern seaboard. By means of such observations, the changing position of major currents can be detected and a semi-quantitative measure of current speeds can be obtained from figure 5. This figure was com- piled by James (1966) from data relating sea surface temper- ature gradients to current speed. Less desirable data must be obtained if these types of projects do not exist. The Coast Guard manual recommends the use of Atlases such as Publ. Nos. 576 and 700 to determine 14 114 SARGASSO GULF STREAM SUMMER WINTER LABRADOR SUMMER WINTER 0 0 *-l I I ' I ' I I I I ' I I ' I I ' I I I I I ' I i 0 5 10 15 20 25 TEMPERATURE GRADIENT (°F/I5NM) Figure 5. Estimation of geostrophic current from surface temperature gradients as given by James (1966) . 15 115 the permanent current component. However, this type of data compilation filters out all temporal variations with time scales less than the averaging period. Thus, the atlas pic- ture is seldom duplicated by an instantaneous snapshot of the circulation field. As an example of the type of differences that can occur, two AOML experiments are considered. The first of these examples is taken from a region of intense boundary currents. Approximate tracklines of buoys drogued at 40 m to minimize leeway are shown as figures 6 and 7. The average currents given on figure & are taken from Publ. No. 700 and Pilot Charts. The drift, particularly the speed components, predicted by these charts is quite differ- ent from the drift experienced by the buoys. A search based on these current representations would not have a high prob- ability of finding the drifter. The second example is taken from a drift buoy experiment still in progress as this report is being written. Five drifting buoys using the EOLE satellite positioning system were deployed in the southern Sargasso Sea, the same mid- ocean area wherein the MODE-I experiment is to be conducted in 1973. Figure ^ shows the regional current field as given by the Pilot Chart for October in comparison to that experi- enced by these drifting buoys which also have parachute drogues attached to minimize leeway. Again it is clear that the mean currents shown by such charts are not a good index to individual drift experiences. Therefore, it is recommended that historic oceanographic data from the drift area be acquired to determine the repre- sentativeness of the atlas current fields and to obtain snap- shots of various flow patterns. The National Oceanographic Data Center (NODC) and research groups working in the area are possible data sources. Upon request, the NODC provides a listing of cruises made in a particular area. Most research groups compile reports giving the cruise tracks and types of data collected. Dynamic height computations relative to a deep dynamic surface are the preferred type of data. Both the speed and direction of the current can be determined from contoured dynamic height fields. However, if this type of data is not plentiful, the use of a characteristic indicator to define the current regime should be considered. For instance, Leipper (1970) uses the depth of the 22°C isotherm to illustrate the circulation pat- tern in the Gulf of Mexico, while Hansen (1970) uses the 15°C isotherm at 200 m to depict the Gulf Stream axis. In areas other than the North Atlantic, it may also be possible to calibrate the temperature data to obtain current speeds as well as direction. 16 116 (a 0) Co l"H OS Sa I ■+i Q +i I CD I to •«4 64 17 117 t ! a i .!*& * : £•• ? ? !>£ 6^ e : j 1. ^ Si : 1 ■ 1 t • 1 hi 7 : i -f 1 ■1 i- 1 .» I 1 : -t -• i 1 . •-. 1 1 « «r*. 7^h. 7 •] " "^ . 1 7\ " <: I <\ :\ -. • g 3 1 a- 1 ; | \ \ \, -''j tx" "t 1 * - » « 1 i IX Oi t-H CO 3 s +i H s a •*» CO . 4i -^ is &£ * CD ca +i .§1 H ?> O Cu H *$ ix O CD «(-, 3 3 18 118 CD •t3 a a, a. w E +^> O £ ,s^ to to O §^ H o a, r-~s CO e to to to ts c § <3 to -. -Q -n •^ +i S^ O « C to +i to to ^ to O a a to o rSJ to +i '^ <(-, Q) to to to C Cu to to Ss Sh i « o to K to to • c> +i c^ S to rV 3 O £ 19 119 GO E +4 OS 0) CO •5 ^J o ^^ Q) CO 03 -P a^ » o (X, «K O y CO o A.' 3 1-3 03 a •^ 03 E +» • » 0) *<-» 20 1*4 120 Since little is known about long-term cycles in the ocean, it is advised that all pertinent data be considered, not just those taken during the month of the launch. The importance of this procedure will be illustrated later. If little or no data exist in the drift region, it will be necessary to use atlases to predict the probable DLCB drift. Supplementary information can be obtained from the FNWC temperature forecasts. As mentioned, the 200-nautical mile grid spacing of the FNWC program can define only large scale phenomena such as the Gulf Stream of Kuroshio. How- ever, the program does use the latest ship injection and bathythermograph reports to arrive at these features. A more accurate updated representation of the current field could result from a comparison of the two circulation pat- terns. Over a period of years, records of drift buoy move- ments can be expected to improve upon the present state of knowledge of the current systems. 3.2.2 Launch Stage The late August 1971, AOML drifter studies in the Gulf of Mexico included an XBT survey just prior to the launch of the buoys. An early October 1970 cruise and Publ. No. 700 depict a well defined Loop Current, extending deep into the Gulf of Mexico (figs. 8 and 10). Based on these data, it was decided to plant the drogues at the northern boundary of the Loop, at approximately 26040' N and 87°30,W. However, midway through the survey it was obvious that the historical and snapshot picture did not exist at that time, but rather that an eddy had detached from the flow and the axis of the Yucatan Current remained at 24°N (fig. 7). This circulation pattern had been observed before (Nowlin, 1972) but at earlier times of the year. A search plan based on either the previous or atlas data would have had little success in recovering a non-functioning buoy. Therefore, it is recommended that a pre-launch survey be conducted to determine the flow pattern in the drift area. It is recommended that buoys be launched in pairs during the test program. This procedure will be of scientific as well as practical value. Practically, two buoys insure the representativeness of the measurements. For instance, the rapid separation of two drogued buoys probably indicates a parachute failure. Drogue studies (Molinari and Starr, 1972) suggest that in many current regimes the rate of separation of neighboring buoys is not great. In case of buoy failure, the trackline of the other drifter would be a valuable tool 21 121 OCTOBER 1970 XBT Temperature (°C) STD Temperature CO Cayo Jutias Lighthouse Figure 10. Approximate drogue tracks, given by the triangle apexes and the temperature field at 200 m for the AOML October 1970 Gulf of Mexico drifter experiment. 22 122 in the recovery operation. The rate of separation of prop- erly functioning buoys will produce data on diffusive processes in the ocean. 3.2.3 Drift Stage An attempt to correlate the DLCB drift with the availa- ble data should be made. If the correlation is good, the prediction and recovery operations are simplified. If the the correlation is not good, an attempt to explain the dis- crepancy should be made in order to predict future drift. The drift data should be input both the the FNWC and to the Coast Guard SAR computer programs. If either or both of these programs are successful in predicting the future drift of the DLCB, the models can be used in case of a buoy trans- mission failure. 3.2.4 Failure Stage It is recommended that arrangements be made with the responsible Coast Guard District to consider the transmission failure of a DLCB as a SAR incident. In particular, the immediate deployment of a Coast Guard aircraft would increase the probability of locating the drifter. If a plane is dis- patched within a day or two, under most oceanic conditions, a simple extrapolation of the buoy path would suffice to locate the DLCB. However, if more than three days are required to initi- ate a search, and a plane cannot be used, other methods to drift determination must be considered. The longer time period would require a study of all the variables considered by the FNWC and Coast Guard. The procedures applied by these agencies during a SAR incident, as well as possible modifica- tions, are considered next. As described in the pre-launch procedures, the ideal data source for surface current determination is a simultaneously conducted experiment. The resulting data would eliminate any guesswork in determining both the wind-induced and permanent current components. As long as the buoys remained in the ex- periment area, which can be ascertained by current velocity computations, the time required to launch a recovery mission is not critical. If the buoys leave the study region, or if no experiment exists, the data compiled during the pre-launch, launch, and drift state must be reviewed to predict the probable trackline. 23 123 If correlation is good between the observed drift and the compiled data, an extrapolation based on the compilation would suffice to locate the DLCB. If the correlation is bad, factors that can cause differences must be considered. For instance, changes in the surface current forcing functions would produce different flow patterns. Large-scale wind shifts can alter the circulation. The exact form of the alteration cannot be predicted with the present knowledge of ocean dynamics. In such cases predictions of the drift would be based on a subjective choice of a theoretical or observational study. Both the Coast Guard and FNWC deduce the wind-induced current from empirically derived relationships. The Coast Guard method considers wind duration, fetch, and wind veloc- ity, while the FNWC program considers only the wind velocity for the prior 36-hour period. Figure 11 from James (I966) shows the results of some wind drift current formulas in- dicating the range of results possible. Again, the choice of a relationship is dependent on the biases of the user. If a drogue is not used, and the sail characteristics of the buoy have not been defined, the leeway effect also has to be determined subjectively. The Coast Guard solution to the problem appears satisfactory. Many of the choices just mentioned would not have to be made if two buoys were deployed together. If the rate of separation of the buoys were determined during the drift stage, and if no dramatic changes occurred in the drift of the functioning buoy, an extrapolation assuming a constant separation rate could determine the DLCB position. If a significant change occurred in the functioning buoy's drift, the rate of separation probably also changed (Molinari and Starr, 1972). It most probably is not feasible to use the buoys in pairs operationally. An operational counterpart of the pair- ing procedure is to infer movement of nonresponding buoys from that of adjacent buoys in an array. Successful use of this idea, however, requires much greater knowledge of the horizontal coherence of ocean currents than is presently a- available outside major currents such as the Gulf Stream. The present EOLE buoy experiment will provide initial data of this sort, and investigators associated with MODE-I are de- veloping new techniques for use of Lagrangian correlation data, but further drift experiments must be conducted in a variety of ocean environments to provide data needed for im- plementation of such techniques. 24 124 10 15 20 25 30 WIND SPEED (KNOTS) 35 Figure !!• Results of several wind drift studies as presented by James (1966) . 25 125 3.3 Recaimendations for the DLCB Test Proqram The five DLCB scenarios describing the possible uses of the drifter implicitly specify different modes of buoy drift. For instance, if the buoy is used in an experiment such as GATE or MODE, the participating oceanographers would require the buoy to be directly coupled to the ocean current, rather than have it actively respond to wind changes. However, in the hurricane experiment where the buoy is to be used as a floating weather station, the constraint of having the DLCB remain in the same parcel of water is not as severe. In either case, the response characteristics of the DLCB to wind, current, and wave action are required to evaluate the data received from the drifter. Computer simulation and tank studies can not duplicate the numerous combinations of wind, wave, and current situa- tions that are possible in the ocean. Therefore, it is recommended that the determination of the DLCB's drifting and sailing characteristics be the first priority of the test program. Furthermore, an efficient mechanism to couple the buoy and current will be required, and various drogue config- urations should be tested to evaluate their effect on the drift. An oceanographic research vessel, or at least a ship capable of taking over-the-side measurements, is necessary to conduct a comprehensive DLCB test. The vessel should have a complete suite of meteorological instruments; in particular, equipment to obtain accurate wind speed and direction is essential. An independent method of obtaining current veloc- ity is necessary to verify the drift obtained from the buoys. The continuous current profiler, developed by Duing of the University of Miami (Duing and Johnson, 1972), can fulfill this requirement. Accurate navigational control is an obvious necessity for the test vessel. Ship positions are needed to verify the DLCB's navigational system and to obtain the ship's drift during the current meter observations. This last measurement is necessary to eliminate the effect of ship motion from the current meter results. One method of obtaining the floating and sailing charac- teristics of the DLCB's is to launch two closely spaced buoys, one drogued and the other not. Since the depth of the drogue will be determined by the requirements of the particular ex- periment being conducted, various drogue depths should be used during the test. Simultaneous direct current and wind measurements should be taken regularly along the buoy's tracklines. 26 126 The rate of separation of the drogued and non-drogued buoys is a function of the buoy's response to the wind. The greater the sail area, the more rapidly will the DLCB's separate in a wind. The deviation from downwind drift, if any, should be determined at this time. The response of the drogued buoy to the current can be ascertained by a direct current measurement at the nominal depth of the drogue. Duing (personal communication) has used the continuous current profiler while in the drift mode, and he reports that the current measurements are representa- tive of the flow if the ship's drift is accurately known. The differences between the drogue and current meter veloc- ities are an indication of how the DLCB's react to a current. The DLCB's response curves to wind and current are probably non-linear necessitating a series of tests in dif- ferent oceanic and atmospheric situations. The optimum logistical area would be one where various wind and current regimes exist in a small region. The drifter studies of Molinari and Starr (1972) suggest that the western Caribbean Sea and Gulf of Mexico is such a region. Figure 7 summarizes the preliminary results of a July- August 1971 drifter study in the Caribbean Sea. The current speeds vary from 1.5 knots in the basin to over 3 knots at the Yucatan Strait, and the current directions range from due west to northeast. The current also accelerates dramat- ically near Cozumel Island. Testing the drifters in this area would subject them to both constant and accelerating currents, without extensive deploying redeploying of the DLCB's. For instance, the drogued buoy could be left in the water while only the non- drogued buoy was moved. This region has the logistical ad- vantage of currents flowing towards the staging area in Mississippi. The Gulf of Mexico also contains many dissimilar current regions. These range from the intense but usually slowly varying Loop Current in the Eastern Gulf to the weaker cur- rents in the Western Gulf. If the test is conducted in the winter, the Gulf has the advantage of having a more variable wind pattern than the Caribbean Sea. The occurrences of "northers" and other weather systems cause frequent shifts in the wind pattern over the Gulf of Mexico. This wind vari- ability will permit a more comprehensive study of the sail characteristics of the DLCB's. The drift prediction procedures outlined previously 27 127 should be tested during this time to verify their usefulness. A pre-launch survey is essential during the test to assure the most efficient deployment of the DLCB's in the short time available. The validity of the Coast Guard and FNWC drift predictions should be ascertained during the test. Finally, because the importance of ancillary data on the currents in the vicinity of the buoy test has appeared throughout our investigation of the problem, and present knowledge and predictability of instantaneous currents in most parts of the ocean are still very weak, it is advisable to associate the buoy tests with other ocean observation projects to the greatest extent possible. This will maximize the amount of current data available to those who may have to make estimates of buoy movement. 4. ACKNCWLE3X^1ENTS The authors would like to express their appreciation to Lt. Cdr. A. Shirvinski (USCG), Director, Operations Analysis Branch, Atlantic Area Coast Guard, and Lt. P. Hill (USCG), Seventh Coast Guard District, for their time. The Coast Guard's SARP computer program and the Seventh District search and rescue procedures were described to the authors by these officers. This work was funded through the National Data Buoy Center's Statement of Work 013 4EC. 5. REFERENCES Atlas of Pilot Charts, Central American Waters and South Atlantic Ocean, United States Naval Oceanographic Office (1955), HO Publ . Mo. 576, Supt. of Documents, Washington, D. C. 20402. Defant, A. (1961), Physical Oceanography, Vol. I, Pergamon Press, 729 pp. Duing, W. and D. Johnson (1972), High resolution current profilinq in the Straits of Florida, Deep-Sea Res. 19: 259-274. Hansen, D. V. (1970), Correlation of movements in the Western North Atlantic, Deep-Sea Ree. 17: 495-511. Hubert, W. E. K. G. Hinman, and B. R. Minderhall (1970) , The NFWC search and rescue planning program (NSAR) , Tech . Note 60 : 217 pp . , Fleet Numerical Weather Central, Monterey, California. James, R. W. (1966), Ocean thermal structure forecastinq, SP-105: 217 pp., U. S. Naval Oceanographic Office, Washington, D. C. Larson, S. and F. laevastu (1971) , Numerical analysis of ocean surface currents, Tech. Note No. 71-1: 21 pp.. Fleet Numerical Weather Central, Monterey, California. Leipper, D. F. (1970) , A sequence of current patterns in the Gulf of Mexico, J. Geophys. Res. 75: 637-657. Molinari, R. L. and R. B. Starr (1972) , Lagrangian current and hydro- graphic measurements in the Western Caribbean Sea, abstract in: Trane. Am. Geophys. Union 53: 392. National Search and Rescue Plan, United States Coast Guard, CH-308, Supt. of Documents, Washington, D. C. 20402. Oceanographic Atlas of the North Atlantic Ocean, United States Naval Oceanographic Office (1965), Section I, Tides and Currents, Publ. No. 700, Supt. of Documents, Washington, D. C. 20402. Ouinlan, F. T. and L. R. Hoxit (1968) , A technique for determination of a climatology of high altitude balloon trajectories, Tech. Memo EDSTM 6: 18 pp., Environmental Data Service, Silver Sprinq, Maryland. Rachelett, J. and M. Armendariz (1967) , Surface wind sampling for un- guided rocket impact prediction, J. Appl. Meteorol. 6: 516-518. Young, M. J. (1962) , Comparison of methods for determininq probable impact areas in planning short range instrumented balloon flights, J. Appl. Meteovol. 1: 531-536. 128 VOL. 77, NO. 24 JOURNAL OF GEOPHYSICAL RESEARCH AUGUST 20, 1972 13 Letters Comments on Paper by A. A. Nowroozi, 'Long-Term Measurements of Pelagic Tidal Height off the Coast of Northern California' Bernard D. Zetler Physical Oceanography Laboratory NO A A, Atlantic Oceanographic and Meteorological Laboratories Miami, Florida 33130 Nowroozi [1972] says, 'A comparison of my estimation of tidal constants, which are free from local coastal anomalies, with the predicted tidal constants given by the available charts [various published cotidal charts] will be deci- sive on the validity of these charts for the OBS 3 vicinity.' In that case it is important that the values accepted for the principal tidal con- stituents for OBS 3 be as reliable as possible. The published results are not reassuring in this regard. Although Nowroozi points out that the 'epochs obtained from the second year are anomalous for Nt and 01,' he does not hesitate to include this second year in either his mean of four annual values or his 4-year analysis. The annual Nt and Oi epochs vary by as much as 42° and 26°, respectively. In contrast to these values the analyses of fifteen 1-year tidal records at San Francisco between 1863 and 1935 [U.S. Coast and Geodetic Survey, 1942] show com- parable spreads of 5° and 3° for N2 and 0%. The San Francisco records are indeed not 'free from local anomalies.' They extend over a much longer period, and there are many more samples ; yet the total spreads are smaller by an order of magnitude. Nowroozi does not mention that the f\ amplitude for the second year is almost 50% larger than the /\ mean amplitude for the other three years, but surely this difference is another reason for rejecting the second year of data from the compilation. Unless the second year is rejected, I do not agree that 'For major con- stituents the yearly analysis of data gives stable results.' Furthermore, it would seem to be im- portant to identify the mechanism responsible for the anomalous results. He places considerable emphasis on the fact that 'The agreement is good between the average yearly tidal constants and the constants ob- tained from the four years of tidal heights.' This agreement is merely a check on his arithmetic and says nothing for the reliability of the data. I am also troubled by his statement that 'The tidal constants obtained from classical tidal analysis and those obtained from cross-correla- tion of the observed tides with the noise-free synchronous equilibrium tides agree very well.' Figure 5 shows a spread of about 25° in the M2 epoch for the 4 years, whereas his Table 2 values vary within 4°. In Figure 6 the major contribution to a large spread in the Oi epoch is due to the fourth year, whereas it is due to the second year in Table 2. The coherency for the fourth year in Figure 5 is noteworthy, being quite poor near M2 and unusually good near S2. I do not agree that the two forms of analysis agree very well, and, since the classical analysis is much more consistent, I do not think any- thing is achieved by introducing the cross- spectral procedure or the results. Finally, since the long-period constituents (Figure 4) are essentially buried in the con- tinuum, I suggest that the harmonic constants for these constituents be called unreliable rather than 'preliminary and subject to change.' References Nowroozi, A. A., Long-term measurements of pelagic tidal height off the coast of northern California, /. Geophys. Res., 77, 434-443, 1972. VS. Coast and Geodetic Survey, Tidal harmonic constants, Pacific and Indian oceans, Publ. TH- 2, Rockville, Md., 1942. Copyright <§) 1972 by the American Geophysical Union. (Received February 17, 1972.) 4590 129 14 Reprinted from Geophysical Surveys 1, 85-98. TIDAL OBSERVATIONS NEAR AN AMPHIDROME BERNARD D. ZETLER Physical Oceanography Laboratory, NOAA, Miami, Fla., 33130, U.S.A. and ROBERT A. CUMMINGS National Ocean Survey, NOAA, Rockville, Md., 20852, U.S.A. Abstract. The signal to noise ratio in tidal data in the diurnal and semidiurnal frequency bands is ordinarily so large that the noise contribution to the tidal harmonic constants is unimportant. How- ever, as the observational locations are selected progressively closer to an amphidrome (point of no tide), the signal to noise ratio decreases, making the tidal harmonic constants less dependable. Standard deviations in amplitude of Ma and S2 obtained from 1 2 29-day analyses of a year of tide data obtained at a standard tide station, estimated to be 280 and 550 km away from the amphidromes for these constituents in the eastern Caribbean, are roughly one-third of the mean amplitudes for these constituents; the standard deviations in epoch are 38° and 30° respectively. Therefore, a program to locate an amphidrome precisely is self-defeating and the location can only be approximated by a grid of tide observations spanning the geographic position and/or by longer series of observations, using higher resolution to increase the signal to noise ratio. Amplitudes of 0.64 cm and 1.24 cm were calculated for IVhand S2from a one-month series of pelagic observations obtained very close to an inferred position of the M2 amphidrome in the northeast Caribbean Sea. Abbreviations C&GS Coast and Geodetic Survey; CICAR Cooperative Investigation of the Caribbean and Adjacent Regions; I APSO International Association for the Physical Sciences of the Ocean ; ICOT Institute of Coastal Oceanography and Tides ; IDOE International Decade of Ocean Exploration; NOAA National Oceanic and Atmospheric Administration ; NOS National Ocean Survey; NSF National Science Foundation; SCOR Scientific Committee of Oceanic Research ; UNESCO United Nations Educational, Scientific, and Cultural Organization. 1. Introduction The state of the art in tide analysis and prediction has been reasonably good for almost a century, primarily because the important tidal frequencies are accurately determined from astronomical data and tidal theory and because the signal to noise ratio at these frequencies is ordinarily quite high. Because it is to be anticipated that the signal to noise ratio will be significantly lower in the vicinity of an amphidrome (point of no tide), this study was designed to investigate whether conventional ana- lysis procedures are adequate for an international program with a high priority on locating amphidromes precisely. Copyright © 1972 by D. Reidel Publishing Company, Dordrecht-Holland 130 86 BERNARD D. ZETLER AND ROBERT A. CUMMINGS The major portion of tidal variations in sea level can be attributed to just a few tidal constituents, each representing a periodic change or variation in the relative positions of the Earth, Moon and Sun. M2 and S2 are tidal constituent names assigned to periods associated with the transits (both upper and lower) of the Moon and Sun respectively across any particular meridian. These periods, 12.42 hr and 12 hr res- pectively, are ordinarily the most important of the semidaily variations. Kx and Ol5 with periods of 23.93 and 25.82 hr respectively, dominate the diurnal constituents. Although some tide predictions involve more than 60 constituents, the four listed above ordinarily con'ribute more than 90% of the predictable tidal energy. Other components referred to in the text are S,, Pj, N2, K2,R2, T2, and v2. Because land masses and irregular bottom topography in the oceans prevent the tides from circling the Earth as progressive waves as the Earth rotates on its axis, the tides observed at particular locations are a composite of progressive waves and standing waves related to basin dimensions. Oceanographers delineate the geographic distribution of amplitudes and epochs of individual tidal constituents by cotidal and co-range charts. Cotidal charts have lines passing through positions at which time of high water is thought to be simul- taneous; co-range charts have lines passing through positions of equal tidal range. Frequently, cotidal and co-range data are combined on a single chart. One of the most interesting features on these charts is the location of amphidromes, also called nodal points, at which theoretically there is no tide for the particular constituent. Characteristically, the cotidal lines radiate progressively around an amphidrome through all hours of the tidal cycle and the co-range lines encircle the point as con- tours, increasing in amplitude with distance from the amphidrome. Figures 19 and 20 in Munk et al. (1970) show comparative cotidal and co-range charts off California for M2 and Kj that have been prepared by various researchers such as Bogdanov (1962) and Dietrich (1963). Due to technological improvements in man's capability to measure tides on the sea floor at great depths, considerable interest has developed in expanding the effort in this direction. Until now, the global tidal charts were based on coastal measure- ments and a relatively few island tide gauges. The IAPSO/SCOR/UNESCO Working Group No. 27 on Deep-Sea Tides was organized in 1965 to encourage and co-ordinate an international program; this Working Group is concerned with analytical proce- dures as well as observational data and instrumentation. Cartwright et al. (1969) stated that since pelagic (open ocean) tide observations require considerable effort, logistics, and expense, "it is most important that the analysis of each set of good data be as comprehensive and meaningful as possible". The global tide program has gathered considerable support because a successful solution to the problem will also be significant to other geophysical problems. Munk and Zetler (1967) indicated that improved global tidal information could lead to increased understanding of tidal energy dissipation in the solid earth as well as in the oceans and to improved comprehension of variability in geomagnetism, seismology, etc. 131 TIDAL OBSERVATIONS NEAR AN AMPHIDROME 87 2. Tidal Analysis Procedures After Zetler and Lennon (1967) objectively tested several tidal analytical processes, the U.S. Coast and Geodetic Survey (now National Ocean Survey) accepted the Harris et al. (1963) least square procedure for the routine analysis of series of one year in length, replacing procedures described by Schureman (1941). One advantage of the change, in addition to the small improvement in accuracy, is that the procedure does not require any assumption of relative amplitudes and phase relationships within each species (number of cycles per day) whereas the traditional C&GS procedure made use of theoretical relationships. The following discussion in Schureman, p. 51, explains why certain lengths of series are considered optimum for tidal analysis; for example, with about a month of data, 29 days are analyzed: In selecting the length of series of observations for the purpose of the analysis, consideration has been given to the fact that the procedure is most effective in separating two constituents from each other when the length of series is an exact multiple of the synodic period of these constituents. By synodic period is meant the interval between two consecutive conjunctions of like phases. Thus, if the speeds of the two constituents in degrees per solar hour are represented by a and b, the synodic period will equal 360°/ \a-b\ hr. If there were only two constituents in the tide the best length of series could be easily fixed, but in the actual tide there are many constituents and the length of series most effective in one case may not be best adapted to another case. It is therefore necessary to adopt a length that is a compro- mise of the synodic periods involved, consideration being given to the relative importance of the differ- ent constituents. Fortunately, the exact length of series is not of essential importance and for convenience all series may be taken to include an integral number of days. Theoretically, different lengths of series should be used in seeking different constituents, but practically it is more convenient to use the same length for all constituents, an exception being made in the case of a very short series. The longer the series of observations the less important is its exact length. Also the greater the number of synodic periods of any two constituents the more nearly complete will be their separation from each other. Constituents like S2 and K2 which have nearly equal speeds and a synodic period of about 6 months will require a series of not less than 6 months for a satisfactory separation. On the other hand, two constituents differing greatly in speed such as a diurnal and a semidiurnal constituent may have a synodic period that will not greatly exceed a day, and a moderately short series of observations will include a relatively large number of synodic periods. For this reason, when selecting the length of series no special con- sideration need be given to the effect of a diurnal and a semidiurnal constituent upon each other. In a C&GS test of analysis methods for 29-day series, it was not clear that there was an improvement in accuracy achieved by changing from the traditional C&GS pro- cedure to a least squares method. The present NOS procedure uses computers to avoid the stencil summing method described by Schureman but otherwise the approach is basically unchanged although the analysis is now done by a computer. Accordingly, the semidaily M2, S2 and N2 are obtained by a modified Fourier calculation and then corrected for sideband effect in the traditional elimination procedures. The effect of other semidaily constituents is calculated through theoretical relationships for amplitude and by assuming a linear trend for phase. The same relationships are later used in inferring harmonic constants for other semidiurnal constituents from the corrected M2,N2 and S2 values. A similar approach is used for diurnal consti- tuents, using Fourier calculations for Kt and Oj. 132 88 BERNARD D. ZETLER AND ROBERT A. CUMMINGS A basic premise in the above calculations is that two frequencies can not be satis- factorily resolved unless the length of series is at least as long as the synodic period of the two frequencies. Munk and Hasslemann (1964) show in their paper on super- resolution of tides that, in principle, two frequencies can be resolved in a much shorter period if the noise level is sufficiently low. We have not been able to achieve super-resolution with real data but have not determined whether the difficulty is due to the noise level or to insufficient accurary (least count) in our analysis procedure. The Doodson (1928) analysis of 29-day series also assumes that nearby frequencies can not be resolved. It uses an assumption that the amplitude ratios and phase re- lationships are comparable at a nearby station for which these values have been well established from a much longer series. The Munk-Cartwright (1966) tidal analysis procedure assumes that the response of the ocean to the tide-producing forces is smooth over a narrow frequency band. Cartwright et al. (1969) used this method for analyzing a 37-day pelagic series but, in a procedure similar to the Doodson method, used as reference a longer series at a nearby coastal station. Inasmuch as there is a shift of 180° in phase in a very short distance across an amphidrome, it seems that this indicates that there is a sharp discontinuity in the ocean's response at this frequency. However, Munk (personal communication) pointed out that if the real and imaginary (in and out of phase) responses are plotted separately, each is quite smooth across the amphidrome, with both going through zero at the particular point. Hence, an amphidrome is not a special case for which the Munk-Cartwright assumption of smoothness of response across a narrow band is violated. Zetler et al. (1965) demonstrated that a least squares analysis procedure has an additional advantage over the traditional Fourier-type analysis in that equal spacing of observational data is not required and breaks in data series need not be interpolated. 3. Caribbean Tides CICAR (Cooperative Investigation of the Caribbean and Adjacent Regions) is an international effort to use the resources of interested nations in a combined oceano- graphic study of the Caribbean region. At the CICAR planning meeting in Mexico City, February 1970, Zetler, as CICAR subject leader for tides, included the following in the tide program: For the semidiurnal tides, there are also two amphidromic points in the region, one in the Caribbean south of Puerto Rico and one in the Gulf of Mexico between the Missisippi River delta and the northeast corner of the Yucatan Peninsula. In the Caribbean node there is an additional interesting aspect, since the S2 node appears to be somewhat south of the M2 node. The latter statement was based on the M2 and S2 tidal amplitudes measured in the Caribbean off South America and on the various islands bordering the eastern Carib- bean. Figure 1, a cotidal chart of the area by Michaelov et al. (1969), interprets the data similarly, the S2 amphidrome in the Eastern Caribbean being located significantly south of the M2 amphidrome. The CICAR tides program has some potential for 133 TIDAL OBSERVATIONS NEAR AN AMPHIDROME (C) 89 (d) Fig. 1. Michaelov ef al. (1969), cotidal (solid lines) and co-range (broken lines) charts for the Caribbean Sea and the Gulf of Mexico: (a) M2, (b) S2, (c) Ki and (d) Oi. The cotidal epoch is the angular retardation of the maximum of a constituent behind the maximum of the corresponding theoretical constituent at the Greenwich meridian. serving as a prototype for the global tide program inasmuch as the above amphidro- mes are reasonably well located from nearby data whereas the locations of mid-ocean amphidromes are more uncertain. 4. Tidal Observations at Magueyes Island The National Ocean Survey has analyzed a year of hourly heights at Magueyes Island (17°58.3'N, 67°02.8'W), just south of Puerto Rico. This location is roughly 280 km northwest of the estimated position of an M2 amphidrome and 550 km north northwest of an S2 amphidrome. The NOS analysis of the year's data produced M2 and S2 amplitudes of 0.70 and 0.76 cm, respectively. The very small amplitudes support the hypothesis of a semidaily amphidrome in the eastern Caribbean and, 134 90 BERNARD D. ZETLER AND ROBERT A.CUMMINGS since M2 is usually 2 to 4 times larger than S2, the values support the suggestion that the M2 amphidrome is considerably closer than the S2 to Magueyes Island. The data for the first 29 days in each month were analyzed separately by the NOS method to determine the variability in the results. Even though Magueyes is some distance from the M2 amphidrome, the large variability in both amplitude and phase indicates that a single 29-day series obtained near an amphidrome is not very useful in locating the amphidrome. Table I lists the results for the five principal constituents for each month as well as values from the 365-day analysis. The local epoch, k, is the angular retardation of the maximum of a constituent of the observed tide behind the maximum of the corresponding theoretical constituent at the longitude of the observations. The relative values of the semidaily constituents vary greatly from month to month in both amplitude and phase, the amplitude standard deviation a being roughly one third of M2 and S2 and one half of N2. The M2 and S2 phases vary during the year by more than 100°. N2 phases of 346° in January and 157° in December indicate that it is pointless to try to use the results to locate the N2 amphidrome. To insure that the above variations were not due to inadequacies in NOS analysis procedures, Francisco Grivel (Universidad Nacional Autonoma, Mexico) was asked to make similar analyses by the Doodson method as part of Mexico's contribution TABLE I Tidal constants - NOS analysis - Magueyes Island, 1969. H is in cm, k is in degrees. M2 s2 N2 Ki Oi H K H K H K H K H K 365 days 0.70 278.4 0.76 213.9 0.40 277.9 7.92 166.3 5.36 160.2 29 days Jan. 1.31 293.8 1.28 232.3 0.67 346.0 7.10 168.7 5.46 163.5 Feb. 0.76 315.2 1.01 217.6 0.82 311.8 6.95 166.5 5.67 155.5 Mar. 1.07 283.2 0.55 236.4 0.52 338.2 7.89 158.3 5.88 . 154.5 Apr. 1.10 335.9 1.31 211.3 1.19 238.1 8.56 157.2 5.55 156.3 May 1.10 239.7 0.67 195.3 0.85 218.2 8.63 163.9 5.46 157.1 June 0.49 231.7 1.07 277.4 0.46 369.1 7.68 164.3 5.18 162.9 July 1.13 312.1 0.40 158.3 0.46 211.9 8.47 162.6 5.46 162.4 Aug. 0.58 330.1 0.79 206.0 0.49 324.4 8.84 165.3 5.55 163.3 Sept. 1.16 286.0 0.88 189.6 0.27 284.4 9.11 173.3 5.46 161.8 Oct. 0.73 290.8 0.98 214.9 1.40 290.8 7.71 176.0 5.49 155.9 Nov. 1.55 236.5 0.61 225.8 0.61 215.9 7.35 170.6 4.66 160.6 Dec. 1.34 236.8 0.37 180.8 0.49 157.0 7.35 169.0 5.12 158.4 Mean of monthly values 1.03 282.6 0.83 212.1 0.69 275.5 7.97 166.3 5.41 159.4 15000 time steps) with the symmetric analog model under adiabatic and frictionless conditions. For a complete description of the model, see Anthes et al . (1971b). These experiments demonstrate numerical stability in the time and space integration scheme and in the vertical staggering of the variables. They also show the model is quite stable using the leapfrog (LF) time integration scheme that does not contain the damping of the high (time) frequency waves inherent in the Matsuno (1966) scheme. A more or less incidental result is the documentation of an instability in a t i me- smooth i ng procedure that has been used previously to avoid growth of the "computational mode" associated with the LF scheme. 2.1 Matsuno and Leapfrog Time Integration Experiment 1 is integrated for 17,280 time steps (192 hours) with the Matsuno ( 1 966) time integration scheme used in the normal experiments with the symmetric and asymmetric models. The initial conditions con- sist of a vortex in gradient balance with no radial or vertical motion. Without friction or heating, the radial and vertical velocity fields should remain zero for all time; however, truncation and roundoff errors produce radial velocities after one time step and generate internal and external gravity waves. The subsequent growth (or non-growth) of the radial velocities is a measure of the stability of the model. Figure 1 shows the time variation of the maximum boundary layer (level 3?) inflow (v ) in experiment 1 and demonstrates that the model is r 147 ADIABATIC. FRICTIONLESS EXPERIMENTS Figure I. Adiabatic and friotionless integrations with Matsuno and leapfrog time integration schemes . very stable for 17,000 time steps. The maximum radial velocities reach -8 x 10~10 m sec-1 after 800 time steps and remain bounded by -10 x 10~10 m sec 1 thereafter. Qualitatively there appears to be very little damp- ing associated with the Matsuno time integration. The oscillations re- flect the stable internal and external gravity waves that are present in the model . While experiment 1 has demonstrated the numerical stability of the model when integrated with the Matsuno scheme, it is interesting to test the stability of the model with a time integration scheme that contains no time damping. Figure 1 also shows the time variation of maximum boundary layer in- flow for experiment 2, which is identical to experiment 1, except that the LF time integration scheme is used instead of the Matsuno scheme. Experiment 2, therefore, contains no damping of the high (time) frequency waves. All experiments with the LF scheme begin with a forward time step, a common starting procedure in experiments utilizing this scheme (Gates, 1959)- Figure 1 shows that the model is also quite stable utilizing the LF scheme, even with no smoothing in time. The amplitude of the error is 3 148 quite similar to that associated with the Matsuno scheme. There is, how- ever, more high-frequency noise toward the end of Experiment 2. Long time integrations with the LF method without any time smoothing may develop fictitious solutions associated with the "computational mode" (Fischer, 1965; Kurihara, 1965; Lilly, 1965). Since experiment 2 contains no time smoothing, the computational mode is quite noticeable toward the end of the integration. Figure 2 shows the variation of radial velocity +10 LEAPFROG r k*3£ Vr-,i AT 15 KM, +5 i o UJ s I o rA NO SMOOTHING i» i \ i\i « in — i — i — r LEAPFROG (I = SMOOTHING INTERVAL) 1*5000 J 1 I 1 L 2000 4000 6000 6000 10,000 12P00 14,000 N TIME STEP Figure 3. Maximum boundary radial velocities in leapfrog experiments showing instability caused by forward time step following each time smoothing procedure . 6 151 unstable (Gates, 1959), it is surprising that such an infrequent applica- tion generates such a growth of error. 2.2 Experiments with Random Perturbations as Initial Conditions The initially small perturbation (lO"10m sec"1) associated with the initial conditions of the balanced vortex requires a rather long time to become a significant part of the mean flow under the instability cited above. For further study of the instability, initial conditions consist of random velocity components (amplitude 0.1 m sec"1) and random tempera- ture fluctuations (amplitude 0.1°C) superimposed on -a stagnant base state. Figure 4 shows the variation of maximum inflow for smoothing inter- vals of ^5, 100, 200, and ^00 time steps and confirms that the instabi- lity is related to the frequency of smoothing. Figure h also shows the result of one experiment in which the smoothing operator (1) was applied LEAPFROG FORWARD TIME STEP AFTER EACH SMOOTHING (I -SMOOTHING INTERVAL) L> 0 UJ 5 -5 -SO * -15 * -20 z 5 -25 I« 100 (NO FORWARD TIME STEP AFTER SMOOTHING 1-45-^1 -i 1 i I i I i _J ■■■■■■:■■'■■ _i_ -I i 1_ _l i I I i l_ 1000 2000 3000 4000 5000 6000 7000 8000 1 9000 10,000 11000 12.000 13,000 14.000 15,000 96 HOURS NUMBER OF TIME STEP (At =40 SEC) Figure 4. Time variation of maximum boundary layer inflow in experiments utilizing leapfrog time integration and initialized with random perturbations . Experiments which utilize the forward time step after each smoothing pro- cedure become unstable . Resuming the calculation (after smoothing ) with the leapfrog scheme is quite stable. 152 every 100 time steps, but the integration was resumed using the normal LF integration scheme, i.e., an+1 = an"1 + 2At §£ (a?) (3) The "perfect" stability of this experiment shows that the cause of the previous instabilities was the repeated use of the forward time step following each smooth i ng-appl i cat i on . Because the instability associated with the forward time step is fasrly weak, one might expect horizontal eddy viscosity to control the growth of the error. Figure 5 shows a comparison of two experiments that contain the normal horizontal eddy viscosity coefficients for heat and momentum (see Anthes et al . , 1971b). The friction terms are computed using forward time steps to avoid instability in these terms. Both ex- periments are smoothed according to (l) every 100 time steps. The experiment that resumes with the LF step (3) shows a steady damping of the initial perturbations. The experiment that utilizes the forward time step remains stable; however, the solutions diverge considerably, and the perturbations seem to have reached a quasi-steady state after 8000 steps. LEAPFROG HORIZONTAL DIFFUSION ONLY •iS FORWARD STEP AFTER SMOOTHING 2 O lOOO 2000 3000 4000 5000 6000 7000 8000 9000 10.000 11P00 12,000 11000 W0OO 15000 N TIME STEP Figure 5. Maximum boundary layer inflow with leapfrog scheme and horizontal diffusion of heat and momentum with and without the forward time step procedure. 8 153 The loss of energy associated with the eddy viscosity must be compensated after this time by the energy source associated with the unstable forward time step. Summarizing the adiabatic experiments, the model exhibits very good stability for long time integrations with either the Matsuno or LF time integration schemes. The computational mode is quite noticeable in the LF experiments after 1 7000 steps, if the solutions are not smoothed in time. Time smoothing at periodic intervals reduces the computational mode but care must be taken in restarting the integration after each smoothing procedure. A forward-time integration following each smoothing is unstable. Resuming the calculation with the LF scheme proves highly stable . 3. DIABATIC EXPERIMENTS: INHIBITING EFFECTS ON HURRICANE DEVELOPMENT Linear analysis (Rosenthal and Koss , 1968; Yamasaki , 1968a) and primitive equation models (Yamasaki, 1968b) have shown that the static stability is an important parameter in determining the growth of initially small disturbances into the hurricane scale vortex. Rosenthal (1970) in- vestigated the effect of the initial relative humidity on hurricane development and found more rapid growth for moister environments. This section summarizes the stabilizing effects of a) sea-surface temperature, b) static stability, c) initial relative humidity, and d) form of initial disturbance on hurricane development as forecast by the symmetric analog. The experiments demonstrate that the model is quite sensitive to these parameters over a relatively small range of synoptic conditions. 9 154 Table 1 lists the temperatures interpolated from the mean hurricane season soundings for the West Indies and Gulf of Mexico areas (Jordan, 1958; Hebert and Jordan, 1959) at the three levels in the model. Also listed are the initial temperatures and relative humidities for the typical (control) model experiment and the relative humidities for the mean Gulf of Mexico sounding. The initial temperatures in the model are very close to the mean Gulf of Mexico hurricane season sounding, while the relative humidities are somewhat higher than those of the mean sounding. The sea temperature for the control experiment is 302°K (28.8°C). . The initial conditions for all experiments discussed in this sec- tion are the strong vortex type, with an initial maximum wind speed of 18 m sec-1 . Table 1. Mean Hurricane Season Soundings 0 P(a) (July-Oct) Gulf Gulf Model Model (mb) W. 1 . T* of Mex i co Tt of Mexi co RH (Control exp.) T RH .22 223.3 223.7 224.0 224.5 .90 .67 680.0 280.4 280.8 • 55 280.4 .90 • 94 954.1 296.4 297.2 .81 297.3 .90 *From Jordan (1958) tFrom Hebert and Jordan (1959) 10 Table 2 summarizes the important characteristics of each experiment Experiment D1 is the control. Experiments D2 and D3 investigate the effect of reduced initial relative humidities. The sea-surface tempera- ture is reduced in experiments D^+ and D5 , and the effects of an increase in static stability are shown by experiments D6 and D7 . Experiments D8-D10 consider different initial conditions and are discussed in the next section. Figure 6 shows the time variation of minimum pressure for experi- ments D1-D7- Since the purpose of these experiments is the isolation of Table 2. Diabatic Experiments Experiment # T sea n i t i a 1 RH (%) ATk=H ATk=3i D1 (control) 302 D2 302 D3 302 Dk 301 D5 300 D6 302 D7 302 D8 302 D9 303 D10 303 90 80 70 90 90 90 90 90 99 99 0 0 0 0 0 +5.0 +2.5 0 -1 .0 -1 .0 0 0 0 0 0 0 0 0 Random perturbation v1 = 0.1 m/sec +1.0 Random perturbation v1 = 1.0 m/sec +1.0 Random perturbation v1 = 2.5 m/sec 11 156 stabilizing effects on the development of tropical cyclones, the struc- tures of each experiment are not presented. The control experiment (D1) shows a rapid development, reaching hurricane intensity (represented by approximately 990 mb) at about 3*t hours. Decreasing the initial relative humidities to 80 percent and 70 percent delays the attainment of hurricane intensity by 26 and 50 hours, respectively. The increased time required for development reflects the reduced water vapor supply available for convection and also the fact that more of the available water vapor is utilized in enriching the large-scale humidity rather than in heating the environment. EFFECT OF Tsea, RELATIVE HUMIDITY, AND STATIC STABILITY ON HURRICANE DEVELOPMENT CD 1015 1010 1005- 1000 Q- 995 Z 2 990 - 985- 980- - —I 1 — 1 1 I 1 1 1 1 ^^ \ \" \ \ - \ \ \ \ \ V \ 1 V-5 Tse*=301 - 1 CONTROL >"*"*'\ Tsea =302 \ RH=90% \ \ \ \=*Z5 \ \ 1 \ \ \ - \ \ \ \ \ \ \ \ \ > 1 1 1 1 \ 1 ' RH=70% - •.RH = 80% 9 0 10 20 30 40 50 60 70 80 90 100 TIME (HOURS) Figure 6. Stabilizing effects of sea temperature (T )3 initial relative humidity (RE) 3 and static stabili%ya (represented by increases in the initial upper level temperature) 3 on the symmetric analog hurricane model. 12 157 The sensitivity of the symmetric analog to variations in sea-tempera- ture over a 2° range during the mature stage has been discussed elsewhere (Anthes, 1972). Figure 6 shows that the development of the model storm is also greatly dependent on a high sea temperature. Decreasing the temperature 1°C is sufficient to prevent hurricane development, and a 2° decrease prevents any intensification whatsoever. Experiments D6 and D7 investigate increases in static stability, represented by increases of 5-0 and 2.5°C, respectively, at the upper level. Sheets (1969) found the standard deviation of temperature at this level to be about 2.4°C under hurricane conditions; thus, the increase of 2.5° appears within the range of synoptic variation. Figure 6 shows that an upper level increase of 2.5°C causes a much slower development, and hurricane intensity has not been reached after 96 hours. An increase of 5°C inhibits intensification altogether. Experiments D1-D7 demonstrate the sensitivity of model storm devel- opment to a range of parameters that lie within synoptic variation. Rapid hurricane development occurs only when the combination of sea tem- perature, relative humidity, and static stability equal or exceed the mean hurricane season values. The requirement for simultaneous achieve- ment of these conditions helps explain the relatively infrequent devel- opment of hurricanes. k. HURRICANE DEVELOPMENT FROM RANDOM PERTURBATIONS ON A STAGNANT BASE STATE Although linear analyses of tropical storm development have assumed a stagnant base state, initial conditions in hurricane models usually 13 158 take the form of a balanced cyclonic vortex of varying size and intensity (Ooyama, 1969; Yamasaki , 1968b; Rosenthal, 1970; Anthes et al . , 1970a; 1970b). In these experiments, considerable kinetic energy is present initially. Although CISK instability theory (Charney and Eliassen, 196*0 indicates that cyclone scale disturbances should develop from any per- turbation, it is not completely clear to what extent the introduction of a large-scale, cyclonic vortex with considerable kinetic energy biases the ultimate storm development. This section presents results starting from random temperature and velocity perturbations superimposed on a stagnant base state. If a CISK type of instability is present in the model, a tropical cyclone scale disturbance should eventually develop. Experiment D8 (see Table 2) consists of the same parameters as the control experiment, except that the model is initialized by introducing random tangential and radial velocity perturbations (amplitude 0.1 m sec-1) on a stagnant base state. Experiment D8 showed very little development out to 120 hours. After 120 hours a large but weak circula- tion had developed with upward motion over nearly the entire domain--a maximum tangential velocity of 0.2 m sec"1 occuring at 390 km (near the edge of the domain) and a maximum diabatic heating rate of 0.1°C/day in the vicinity of the tangential wind maximum. The solution appeared to be in nearly steady-state at this time; thus the initial kinetic energy of the environment appears to play an important role in the time required for development of the hurricane. Experiment D9 investigates development from random perturbations under slightly more favorable conditions than those in experiment D8. 14 159 The importance of the initial kinetic energy in development time is probably through the sea-air transfer of latent and sensible heat to the environment, since the rate of transfer is approximately proportional to the kinetic energy. To test this hypothesis, experiment D9 was repeated with initial random velocity perturbations of 2.5 m sec"1 instead of 1 m sec-1. The total initial kinetic energy, therefore, is roughly in- creased by a factor of five. Figure 7 shows that rapid intensification occurs about 6 days earlier in the experiment with the higher kinetic energy, but that the ultimate intensities of the two experiments are qui te s imi 1 ar . 5. SUMMARY Several experiments with the symmetric analog hurricane model have been made to investigate some numerical aspects of the time-differencing scheme and the vertical staggering of variables. The model is quite stable, under adiabatic and frictionless conditions for at least 17,000 time steps using either the Matsuno or LF time integrationi scheme. Smoothing in time in the LF integrations suppresses the "computational mode." Resuming the integration after each smoothing with a forward time step becomes unstable without friction. With friction, the calcula- tions are stable, but the forward time step adds energy to the system. Experiments with heating and friction show that the model is sensi- tive to small variations in sea-surface temperature, initial relative humidity, and static stability. Widely varying initial conditions are also tested. The hurricane scale vortex develops from random velocity perturbations superimposed on a stagnant base state. The time of 15 160 The initial static stability is reduced by increasing the low-level tem- perature and decreasing the upper level temperature both by 1°C. The initial relative humidity is increased to 99%, and the amplitude of the initial velocity perturbations is increased to 1 m sec" . Finally, the constant part of the lateral eddy viscosity coefficient for heat and momentum (see Anthes et al . , 1971b) is set at zero. Figure 7 shows the time variation of maximum wind speed and central pressure for experiment D9- After a long period (9 days) of very gradual development, sudden intensification occurs, and an intense hurricane is generated. Because of the extremely favorable environmental conditions, the storm is considerably more intense than the control. Similarity in structures, however, indicates that the hurricane scale disturbance is indeed the preferred scale for development. RANDOM PERTURBATIONS 920^ 1 I ■ ■ I i ■!■ i_J_ _i I i 100S 90 X 80 ^ 60 50 m rn 40° 30 £ 20 ft o 10 A 50 100 150 200 250 TIME (HOURS) 300 350 Figure 7. Time variation of central "pressure and maximum wind speed for experiments initialized by random temp- erature and velocity perturbations on a stagnant base state . 16 161 development, but not the structure or intensity of the mature state, is closely dependent on the amount of initial kinetic energy. 6. ACKNOWLEDGMENTS Dr. Stanley L. Rosenthal and Mr. Walter J. Koss made helpful sug- gestions concerning this research. 7. REFERENCES Anthes, R. A. (1972), The development of asymmetries in a three-di mens iona numerical model of the tropical cyclone, submitted for publication in the Monthly Weather Review. Anthes, R. A., S. L. Rosenthal, and J. W. Trout (1971a), Preliminary results from an asymmetric model of the tropical cyclone, to be published in the Monthly Weather Review, 99- Anthes, R. A., J. W. Trout, and S. L. Rosenthal (1971b), Comparisons of tropical cyclone simulations with and without the assumption of circular symmetry, to be published in the Monthly Weather Review, 99- Charney, J. G. and A. Eliassen (196*0, On the growth of the hurricane depression, Journal of the Atmospheric Sciences, Vol. 21, No. 1, 68-74. Fischer, G. (1965), A survey of finite difference approximations to the primitive equations, Monthly Weather Review, Vol. 93, No. 1, 1-10. Gates, W. L. (1959), On the truncation error, stability, and convergence of difference solutions of the barotropic vorticity equation, Journal of Meteorology, 16, No. 5, 556-568. Hebert, P. J. and C. L. Jordan (1959), Mean soundings for the Gulf of Mexico area, National Hurricane Research Project Report No. 30, Washington, D. C. , April, 10 pp. Holloway, J. L. and S. Manabe (1971), Simulation of climate by a global general circulation model, Monthly Weather Review, Vol. 99, No. 5, 335-370. Jordan, C. L. (1958), Mean soundings for the West Indies area, Journal of Meteorology, Vol. 15, No. 1, 91-97- 17 162 REFERENCES (continued) Kurihara, Y. (1965), On the use of implicit and iterative methods for the time integration of the wave equation, Monthly Weather Review, Vol. 93, No. 1 , 33-^6. Lilly, D. K. (1965), On the computational stability of numerical solu- tions of time -de pen dent non-linear geophysical fluid dynamic problems, Monthly Weather Review, Vol. 93, No. 1, 11-26. Matsuno, T. ( 1 966) , Numerical integrations of the primitive equations by a simulated backward difference method, Journal of the Meteorological Society of Japan, Ser. 2, Vol. 44, No. 1, 76-8*1. Ooyama, K. (1969), Numerical simulation of the life cycle of tropical cyclones, Journal of the Atmospheric Sciences, Vol. 26, No. 1, 3~40. Rosenthal, S. L. (1970), A circularly symmetric primitive equation model of tropical cyclone development containing an explicit water vapor cycle, Monthly Weather Review, Vol. 98, No. 9, 643-663. Rosenthal, S. L. and W. J. Koss (1968) , Linear analysis of a tropical cyclone model with increased vertical resolution, Monthly Weather Review, Vol. 96, No. 12, 858-866. Sheets, R. C. (1969), Some mean hurricane soundings, Journal of Applied Meteorology, Vol. 8, No. 1, 134-146. Yamasaki, M. (1968a), Numerical simulation of tropical cyclone develop- ment with the use of the primitive equations, Journal of the Meteorological Society of Japan, Vol. 46, No. 3, 178-201. Yamasaki, M. (1968b), A tropical cyclone model with parameterized verti- cal partition of released latent heat, Journal of the Meteorological Society of Japan, Vol. 46, No. 3, 202-214. 18 163 16 Reprinted from Mariners Weather Log 16, No. 5, 288-293 SOME OBSERVATIONS FROM HURRICANE RECONNAISSANCE AIRCRAFT OF SEA-SURFACE COOLING PRODUCED BY HURRICANE GINGER (1971) Peter G. Black Environmental Research Laboratories, NOAA Coral Gables, Fla. During the monitoring of the hurricane modification experiment on hurricane Ginger from September 26 to 29, 1971, an interesting set of oceanographic data was collected by low-flving aircraft of the U.S. Navv Weather Reconnaissance Squadron Four for Project STORMFl'RY. The data consisted of airborne infra- red thermometer measurements of sea-surface tem- perature (SST) along the various flight tracks and airborne expendable bathythermograph (AXBT) mea- surements of water temperature vs. depth along the flight tracks on September 27 and September 28. The intensity' of hurricane Ginger oscillated slightly during this period as can be seen from minimum pressure and maximum wind values shown in figure 12. Central pressures varied from 985 mb on Sep- tember 26 to 970 mb on September 27 and back up to 980 mb on September 28 and to 985 mb on September 29. Maximum winds ranged from 65 kt on September 26 to 70 kt on September 27 and 28, and to 80 kt on September 29. Satellite pictures of Ginger for September 26-28 are shown in figure 13. These cloud integrations of SE&T. 26 HURRICANE "GINGER" SEPTEMBER 25-29. 1971 Figure 12. — Profiles of maximum wind (dashed line through open circles) and minimum pressure (solid line through solid circles). Triangles are analyzed maximum wind values from various aircraft at the times of the sea-surface temperature analyses. Figure 13. — Cloud integrations for the times and dates shown using ATS-3 photographs taken approximately every 25 min. The cross hairs indicate the storm center with tick marks every 5° of latitude. 164 HURRICANE SEA SURFACE SEPTEMBER HURRICANE "GINGER" SEA SURFACE TEMPERATURE SEPTEMBER 29, 1971 25 26 «" 26 / Figure 14. —Sea-surface temperature (SST) analyses for September 21 (top), before hurricane Ginger's traverse of the area, and for September 29 (bot- tom), after hurricane Ginger's traverse of the area. The warm temperature tongues along the U.S. east coast and in the northern part of the area are por- tions of the Gulf Stream. Ginger's track is indicated by the dashed lines with the date /time indicated at the 0000 GMT positions (open circles). The solid circles indicate the 1200 GMT positions. ATS-3 photographs constructed by T. Fujita (Uni- versity of Chicago) are for the periods specified. The integrations are made by making multiple ex- posures on the same sheet of photographic paper of successive individual pictures and are used to bring out regions of persistent cloud cover. These pictures show a large eye (nearly 120 mi across) on all 3 days. The cloud structure appears to be better organized on September 27 and more ragged and poorly organized on September 26 and September 28, the 2 seeding days. The first step in the analysis was to try to determine if hurricane Ginger caused any large-scale changes in the sea-surface temperature structure over that portion of the Atlantic which it traversed. Therefore, before looking at the detailed aircraft measurements, some of the personnel at the lab decided to look at the large- scale sea-surface temperature structure over the portion of the North Atlantic from 25° N. to 40° N. and from 50° W. to 80° W. as indicated by available ship reports. Most of the area proved to be a region of relatively dense reports. This enabled daily com- posite analyses of SST reports to be made from Sep- tember 21 to September 30, inclusive. Figure 14 shows the analysis for September 21, before the storm entered the area, and for September 29, as the storm was leaving the area. Superimposed upon these an- alyses is the track of the storm. One can see that a portion of the area was traversed twice, once when the storm was moving northeast and once when it was moving west after it had made a loop. The only region of cold water at the initial time period is where the track crosses itself, suggesting that some storm - related cooling could have occurred from the first passage of the storm through this area. The September 29 analysis shows a drastically changed sea-surface temperature field. Cooling has occurred over about three-quarters of the track. The amount of cooling is variable from nearly 4° C to only 1° C. No cooling was evident along the first quarter of the track (0000 GMT, September 22 to 1200 GMT, September 23). Having identified the regions where cooling of the sea surface occurred in the wake of the storm, the aircraft measurements were examined in detail. The first problem was to calibrate the infrared sea-surface temperature measurements so they would be com- patible with sea temperatures measured by ships. A rough calibration for the infrared SST measurements was arrived at by determining an average temper- ature difference for the whole flight between AXBT surface observations and the infrared SST observa- tions at the AXBT observation time. On the days when AXBT observations were not made, average differ- ences were computed between nearby ship observations of SST and the infrared observations of SST. For the 4 days analyzed, the average correction to the in- frared SST's was about -2° C. The analyses for the above data plus available ship reports for September 26-29 are shown in figures 15 through 18. Figure 15 shows a pool of cold water cen- tered on the track near the 1200 storm position for September 24, the region where the storm slowed and upwelled cold water was brought to the surface. Only slight cooling is indicated near the storm position on September 26. The data on September 26 were mostly peripheral, and, hence, the dashed isotherms were used to indicate sparse data near the center on this day. The warm-water region (>27°C) to the north of the storm, which persists during the following 3 days (figs. 16-18), is probably an area of warm-water convergence and downwelling caused by warmer sur- face water being transported away from the storm region. Figure 16, the sea-surface temperature analysis for September 27, shows the persistence of the cold pool near the September 24 position. It also shows a new region of cooler water surrounding the storm center on this day. Of particular interest is the presence of a warm region within the eye of the storm. We speculated that existence of an exceptionally large eye (nearly 125 mi in diameter) has allowed time for solar radiation to warm the region within the eye since a given point on the ocean within the eye would be nearly cloud free for almost 24 hr. There are undoubtedly other possible explanations for this ob- servation, but more research is needed before they can be clearly formulated. The observation, however, is well documented, as will be shown later, and is 289 165 HURRICANE GINGER SEA SURFACE TEMPERATURE SEPT 26,1971 1100Z Figure 15. — Sea-surface temperature analysis for the area near hurricane Ginger on Sept. 26, 1971. The hurricane track is superimposed and every 12-hr position is shown by an open circle. The open box in- dicates the storm position at the time of the airborne SST measurements. Figure 16.— Same as figure 15 except for Sept. 27, 1971. The squares indicate the position of the AXBT data. The straight lines, lettered at each end, indicate the various cross section orientations. 290 166 HURRICANE GINGER SEA SURFACE TEMPERATURE SEPT. 28,1971 2000Z 70" _L ^L. 27 Figure 17. — Same as figure 16 except for Sept. 28, 1971. HURRICANE GINGER SEA SURFACE TEMPERATURE SEPT. 29,1971 2000Z Figure 18. —Same as figure 15 except for Sept. 29, 1971. 291 167 persistent in each of the analyses shown in figures 15 to 18, most prominently in figure 16 and 17. The warm tongue to the north and northwest of the storm, mentioned earlier, is again present. Figure 17 shows the SST analysis for September 28. This analysis shows that the cold pool mentioned in the previous two figures has drifted westward slightly and become even cooler, probably because it has continued to be under the influence of Ginger's large circulation and further cooling by evaporation has resulted. The cooling observed on September 27 around the center of Ginger does not appear to have persisted. This could possibly be due to the fact that sparse data in that area on September 28 did not allow it to be de- tected, or it could be due to some kind of gross measurement error on September 27. The AXBT data, to be discussed shortly, suggest that the former is the case. Some cooling (~ 2.5° C) is evident beneath the max- imum wind region on September 28, as was the case on September 27. The warm eye region is also evi- dent again. A somewhat more well-defined warm region surrounds the periphery of the cooler waters to the west, northwest, and north of the storm center. The analysis for September 29 (fig. 18), in a region displaced to the northwest of the previous analysis areas, shows a region cooled by about 2° C along the storm track in the wake of the storm. Again a region of warmer SST values is evident within the eye of the storm. Note also the presence of the Gulf Stream just ahead of the storm at this time, as well as a cold eddy between the storm center and the Gulf Stream. The effect of the storm on the ocean SST structure is markedly less on September 29 than on September 28 or September 27. The acquisition of AXBT data within hurricane Ginger is perhaps the most unique aspect of this study. To the author's knowledge, no such data have ever been gathered before in real time. These data have allowed a much higher degree of confidence to be put in the SST analyses just discussed than would other- wise have been justified Two days of AXBT data, comprising about 40 observations on September 27 and 28, have been analyzed. Figure 16 shows the locations of four cross sections that were constructed through and ahead of the storm on September 27. Sections A -A' and D-D' are shown in figures 19 and 20. Section D-D' is considered as representative of the undisturbed state since it is nearly 200 mi ahead of the storm center in a region of approximately 30-kt winds. This section shows an undisturbed mixed layer depth of about 100 ft. Section A -A' shows the changes in the subsurface structure brought about by the storm on September 27, when it had become almost stationary. Note that within 50 mi of the storm center (the approximate radius of maximum winds) the thermocline1 has bulged upward by about 15 ft, indicating upwelling in this region. Upwelling is also indicated by the upward bulging and vertical displacement of subthermocline isotherms. However, the cooler subsurface water has not reached the surface in this region and, in- 1 The thermocline is a vertical temperature gradient in some layer of a body of water, such as the ocean, which is appreciably greater than the gradients above and below it; also a layer in which such a gradient occurs. stead, is overlain by relatively warm water. Only in the region 50 to 70 mi from the center, especially in the right (northeast) quadrant (the region of highest winds), does cooler subsurface water reach the sur- face, with a gradual warming radially outward from this region. However, in this region the mixed layer depth has become much deeper than the undisturbed state, and the thermocline has become diffuse and spread out. The mixed layer depth has increased by about 100 ft, and the thermocline gradient has de- creased from 4° C/20 ft to 1 " C/20 ft. This structure suggests that intense mixing is probably the primary mechanism responsible for the sea-surface cooling detected by the infrared SST measurements. Perhaps the upwelling beneath the center made the mixing process more efficient as the region of maximum winds moved over the upwelling region previously beneath the eye. It is suggested that the processes at work in this case operate as follows. The unusually large eye in hurricane Ginger results in a larger radius of curvature for the wind. Hence, a longer fetch of the wind can be expected than in a "normal" hurricane with the same wind speed, resulting in greater wave action and a higher signifi- cant wave height. It has been shown that strong sur- face wave action leads to an instability in the ocean friction layer which is primarily responsible for the generation of long roll vortices in the ocean. These vortices have been referred to as Langmuir circula- tions. The vertical scale is limited first by wind speed and then by mixed layer depth. As instability increases because of increasing wave action via in- creasing wind speed, the vertical scale of the Lang- muir circulation works to increase in size and in so doing erodes and deepens the thermocline. Figure 17 shows the locations of three cross sections constructed through hurricane Ginger on September 28. Section F-F' is shown in figure 21. This section shows that on September 28 very little upwelling was detected. The thermocline beneath the storm center is at about the same depth as before the storm traversed the area. The mixing process ap- pears to be less efficient on this day also, with only about a 50-ft decrease in the thermocline occurring beneath the coldest surface waters at the radius of maximum winds (50 mi). Note that, as on September 27, warmer waters (perhaps due to downwelling) are present at the storm's periphery (the outer radius of 50-kt winds). Of special interest is the extremely deep thermo- cline located on the right side of figure 21 (140 mi to the east-northeast of the storm center). Here the thermocline is nearly 150 ft deeper than the undis- turbed cross section (fig. 20), but with a fairly tight gradient of about 2° C/20 ft. This is part of the region associated with the intense cooling of the sea surface which occurred on September 24. This is the only part of this particular region for which AXBT data were obtained. In view of these data, one might infer that intense mixing at the radius of maximum winds (for a large storm such as Ginger) will erode the elevated thermocline produced by the upwelling and leave, instead, a trough in the thermocline along the storm track. In other words, the intense mixing in a storm as big as Ginger, may wipe out the after- effects of upwelling, and leave the aftereffects of mixing (i.e., a deep, cold mixed layer). The explanation for the variable amount of cooling in the wake of a relatively steady state storm such as 292 168 Figure 21. — Same as figure F-F'. 20 except for section RADIAL DISTANCE Figure 19. — Temperature-depth cross section beneath hurricane Ginger along section A-A'. The thick vertical line indicates the center position of the hurricane and the thin vertical lines indicate the position of the AXBT's projected onto the section. RADIAL DISTANCE Figure 20. — Same as figure 19 except for section D-D', the undisturbed cross section. The thick vertical line in this case is the future hurricane track position. Ginger is still under investigation. However, recent studies suggest that the most critical parameter gov- erning cooling of the sea surface by a hurricane is the ratio of its forward speed to the internal wave speed at the thermoeline. The smaller the ratio the more intense the cooling. The hurricane Ginger results seem to bear this out. The strongest cooling was observed on September 24-25 and September 27, the two occasions when the storm became nearly station- ary. The size and intensity of the storm, as well as the subsurface ocean structure, are other parameters which would act to modulate the degree of cooling. Also now being studied is the possibility that the observed amount ot sea-surface cooling would act to decrease the hurricane intensity by reducing the energy transfer between the ocean and the air. The- oretical hurricane model studies predict that this should happen, but observational proof of this has been difficult to obtain. A third area now being studied is the possibility that sea-surface cooling in the wake of hurricanes may cause changes in fish habits. These changes may be a direct result of the temperature change or, more likely, a result of changes in the food supply of fish. This would probably have the effect of bringing fish which normally inhabit deeper regions of the ocean closer to the surface in the wake of a hurricane. It has been shown on at least one occasion that a hurri- cane caused increased amounts of phytoplankton in its wake near the surface. In conclusion, it is hoped that aircraft observations of the type described in this article can continue to augment surface ship observations, and that future studies of this type will be possible. It should be emphasized that surface ship reports are extremely important to meteorologists within 500 to 600 mi of hurricanes, especially observations of sea surface temperature. It is hoped that continued care in ob- serving and reporting these observations will be maintained. BIIUMI mi iimnil ■■■■■ inimimm mininiimiimiiimiiinif¥¥¥inniiiimmy 9 WE OF NOAA ARE MAKING USE OF THIS SMALL AMOUNT OF SPACE TO EXTEND OURS THANKS TO ALL THE SHIPS' OFFICERS WHO ROUTINELY TAKE SHIPBOARDS WEATHER OBSERVATIONS. TO US, THESE EXCELLENT OBSERVATIONS ARE PRICE-3 LESS. WE CERTAINLY DO APPRECIATE RECEIVING THEM ON A REGULAR BASIS.: 293 169 17 Reprinted from Monthly Weather Review 100, No. 3, 208-217. Airborne Radar Observations of Eye Configuration Changes, Bright Band Distribution, and Precipitation Tilt During the 1969 Multiple Seeding Experiments in Hurricane Debbie1 PETER G. BLACK — National Hurricane Research Laboratory, Environmental Research Laboratories, NOAA, Miami, Fla. HARRY V. SENN and CHARLES L. COURTRIGHT— Radar Meteorology Laboratory, Rosenstiel School of Marine and Atmospheric Sciences, University of Miami, Coral Gables, Fla. ABSTRACT — Project Stormfury radar precipitation data gathered before, during, and after the multiple seedings of the eyewall region of hurricane Debbie on Aug. 18 and 20, 1969, are used to study changes in the eye configuration, the characteristics of the radar bright band, and the precipitation tilt. Increases in the echo-free area within the eye followed each of the five seedings on the 18th, but followed only one seeding on the 20th. Changes in major axis orientation followed only one seeding on the 18th, but followed each seeding on the 20th. Similar studies conducted recently on unmodified storms suggest that such changes do not occur naturally. However, the studies do not exclude this possibility. Changes in the radius of maximum winds follow closely the changes in eyewall radius. It is suggested that the different results on the 2 days might be attributable to seeding beyond the radius of maximum winds on the 18th and inside the outer radius of maximum winds on the 20th. The bright band is found in all quadrants of the storm within 100 n.mi. of the eye, sloping slightly upward near the eyewall. The inferred shears are directed outward and slightly down band with height in both layers studied. The hurricane Debbie bright band and precipitation tilt data compared favorably with those gathered in Betsy of 1965 and Beulah and Heidi of 1967. 1. INTRODUCTION On Aug. 18 and 20, 1969, the first multiple seeding of a hurricane was carried out in hurricane Debbie. Freez- ing nuclei (silver iodide) were injected into the hurri- cane along a 15- to 25-n.mi. swath beginning at the inner edge of the northeast sector of the eyewall and extending outward at an altitude of 35,000 ft (—25° C). Seeding was repeated five times at intervals of 2 hr. The Stormfury experiment was designed to test the hypothesis that seeding the clouds radially outward from the inner edge of the eyewall (near the radius of max- imum winds) would cause a reduction and outward dis- placement of the maximum winds by releasing large amounts of latent heat, causing a pressure fall beneath the seeded region, and lessening the pressure gradient. An earlier and similar hypothesis was outlined by Simp- son et al. (1963) but was subsequently modified by Gentry (1969). Examination of recent numerical modification experi- ments by Rosenthal (1971) has led to a different hy- pothesis. The new hypothesis has resulted from the exchange of ideas by several people and is still being checked and improved. As summarized by Gentry (1971), it suggests that seeding the hurricane beyond the radius 1 Contribution No. 1402 from the University of Miami, Rosenstiel School of Marine and Atmospheric Sciences 208 / Vol. 100, No. 3 / Monthly Weather Review of maximum winds will cause the cumulus towers at radii from about 5 to 30 n.mi. from the inner edge of the eyewall to grow, thus building a new eyewall at a larger radius. The Stormfury Operations Plan (U.S. Navy Fleet Weather Facility 1969) has described in detail the flight patterns that were carried out by the NOAA-Research Flight Facility (RFF), Navy, and Air # Force aircraft involved in the monitoring of the storm before, during, and after the seeding operation. Gentry (1970) and Hawkins (1971) have described the results from the wind, temperature, etc., data collected by some of these aricraft. Their results were generally in the sense predicted by the new hypothesis. Another approach to the problem of proving or dis- proving this hypothesis is taken in this paper. Sheets (1970) has examined flight data collected in hurricanes by RFF aircraft druing the past 10 yr. He has concluded that, for mature hurricanes, the radius of maximum winds was most likely to be located about 3 n.mi. radially out- ward from the inner edge of the eyewall. In addition, Rosenthal (1971) has suggested that, in a steady-state mature storm, the position of the eyewall should be well correlated with the radius of maximum winds. There- fore, the radar data collected during the experiment were examined in an effort to detect changes in the eye- wall region that could be related to the seeding. Un- fortunately, because of a lack of high quality data, de- 170 finitive results were not entirely possible. Nevertheless, two studies were decided upon. One was to document the time variation of the eye configuration during the seed- ing operations on August 18 and 20 (data were avail- able for only 1 hr before until 1 hr after the seeding period on each day). The other was to document the bright band height and thickness as well as the precipitation tilt on August 20. The objective of the eye size study was to look for evidence to suggest an outward displacement and, hence, weakening of the maximum wind region, as predicted by the Stormfury hypothesis. Such an outward displace- ment was reported by Simpson and Malkus (1964), who stated that at some time after the single seeding attempt on hurricane Beulah on Aug. 24, 1963, the eye increased in radius from about 10 n.mi. to about 20 n.mi. However, at that time, they were not certain whether or not the fluctuations were natural or were changes caused by seeding. Although occasional observations of bright bands in small segments of hurricanes date back to 1945 (Atlas et al. 1963), Senn (19666) showed that this phenomenon was observed in almost all regions of hurricane Betsy of 1965. Unfortunately, it is not generally possible to make widespread observations of the bright band in most hurricanes because of the path of the storm, the lack of high quality, narrow beamwidth range-height indicator (RHI) radars, and/or the lack of qualified observers to man the airborne RHI radars in non-Stormfury recon- naissance missions. The present data on Debbie, there- fore, constitute another attempt to study such conditions simultaneously throughout a hurricane. For the same reasons, precipitation tilt data from RHI radars throughout hurricanes are rare. In a previous paper, Senn (19666) showed that tilt observations were often dismissed as erroneous or due to equipment problems when, in fact, the}- were probably real. A careful check of the equipment and analysis of the data from hurricane Betsy confirmed the validity of the observations of appreciable tilt at most levels and azimuths in that storm. The same equipment and basic methods were used to obtain the Debbie data during the Project Stormfury experiments. 2. RADAR DATA INTERPRETATION PROBLEMS The Operations Plan (U.S. Navy Fleet Weather Facility 1969) called for over 20 plan-position indicator (PPI) radarscopes to be photographed once every 10 s, and the four prime RHI radarscopes to be photographed once every 3-5 s. The characteristics of the radars used in this study are given in table 1. Less than 20 percent of the anticipated PPI data and less than 10 percent of the anticipated RHI data were considered usable for this research due mainly to below-normal performance of the radars, lack of antenna stabilization, range attenuation effects in the shorter wavelength radars, and poor azimuth documentation. These effects made it difficult to establish continuity on the various eyewall features, a process that 456-793 O - 72 - 5 Table 1. — Characteristics of radars used for eye configuration study APS-20- APS-16* APS-64f WP-101J Wavelength 10 cm 3.2 cm 3.2 cm 5.5 cm Peak power 2 Mw 450 Kw 40 Kw 75 Kw Beamwidth 1.5°/5.5° 3.1°/1.0° 1.5°/csc2 7.5°/7.5° (horizontal/vertical) Pulse length 2 ^s 1.8 ms 2.5 us 2.1 Ms Pulse repetition 270 s"1 450 s"1 200 s"1 400 s"1 frequency •WC-121N, Weather Reconnaissance Squadron Four, U.S. Navy fWB^J7, 53d Weather Reconnaissance Squadron, U.S. Air Force t DC-6. Research Flight Facility, NOAA was made more difficult by the rapidly changing nature of the eyewall itself. A comparison of the eyewall region in figure IB with the eyewall region in the radar composite constructed by Fujita and Black (1970) using the 3-cm APS-64 radar will demonstrate some of the problems encountered due to range attenuation when using short wavelength radars. In an attempt to minimize these problems, two methods of measuring the echo-free eye area were used. One method consisted of planimetering the echo-free eye region. The other method consisted of measuring the major and minor axes of the eye and computing its area from the ellipse formula. The two measurements were made independently by different people. There were periods when the echo- free eye was especially difficult to define and at these times there were disagreements between the two methods. In these data-poor regions, it is thought that measuring major and minor axes provided a more reliable measure of the eye size than did the planimetered area. A considerable number of calibration problems were also encountered. By analyzing ground targets after take- off and before landing, researchers found that many of the radars have range mark errors up to 10 percent and north-line errors of up to 25°. This complicated the eye size analysis. However, during most of the period of study, there were simultaneous data available from two radars and, for short periods of time, from up to four radars. After the calibrations were made, the data taken from different radars at the same time produced consistent results. 3. QUALITATIVE RADAR FEATURES OF HURRICANE DEBBIE Figure 1 shows a representative PPI display of hur- ricane Debbie on each of the two operational days. Since there was no single high-quality photograph of the storm, figure 1 is a composite constructed from photographs taken over a period of several minutes centered on the indicated time. Only the 10-cm APS-20 radar photo- graphs were used. Figure 1 illustrates the basic difference in the storm on the 2 days. On the 18th, only a single wall cloud was evident, except for occasional periods when a second, concentric wall cloud was attempting to March 1972 / Black, Senn, and Courtright / 209 171 Figure 1. — Typical APS- 20, 10-cm radar composites of hurricane Debbie on (A) Aug. 18 and (B) Aug. 20, 1969. The dotted areas indicate regions of faint echoes. S 8 1 S S 2 ? 8 ? ; 1250 1245 IJOO 1315 1350 15*5 1 B'Z 1415 1450 1445 1500 ?g ?s ?s ?'a si? 4' g i'k ?s i's ?s j's' 115 1550 1543 1500 101B 1550 1545 1700 1715 1750 1745 15O0 Figure 2. — Maximum height (103 ft) of radar echoes in each quad- rant of the hurricane Debbie eyewall on (A) August 18 and (B) August 20, obtained from the 3-cm APS-45 radar. The vertical dotted lines indicate the seeding times and are numbered relative to the first seeding. All times are qmt. form. However, on the 20th, concentric wall clouds were evident for the entire day. On the 18th, the storm was more asymmetric than on the 20th with numerous precip- itation echoes in a large band extending to the east and southeast of the center. The most intense precipitation echoes were located in the north and northeast quadrant while there was none outside the eyewall to the south- west and south. The southeast quadrant of the eyewall was open most of the day. On the 20th, Debbie appeared to have a more uniform distribution of echoes around the storm. Figure IB does not show all the echoes to the north of the storm that were observed later in the day, for reasons stated in the pre- vious section. Nevertheless, a fair idea of the echo dis- tribution can be shown. The outer eyewall was closed during the entire day and was generally more solid than the inner eye, which was often broken in appearance especially in the south quadrant. Again, the most intense precipitation was located in the northeast quadrant. The concentric wall cloud structure, or "double eye" as it is sometimes called, is apparently not an uncommon feature of hurricanes or typhoons. Fortner (1958) first described this feature in typhoon Sarah of 1956, and later Jordan arid Schatzle (1961) reported a double eye in Hurricane Donna of 1960. Many other cases are described in the Annual Typhoon Reports prepared by the Joint Typhoon Warning Center on Guam. However, the double eye is usually associated with more intense storms than Debbie. Figure 2 shows the distribution of the maximum echo heights in each quadrant of the eyewall on both days as estimated from the available APS-45, RHI data. One can see that on the 18th the maximum echo heights were lowest in the southwest quadrant (averaging 25,000-30,000 ft) and highest (averaging 45,000 ft) in the northeast quadrant. A similar picture is evident from the data on the 20th with the lowest maximum echo heights (averag- ing 30,000 ft) being located in the southwest quadrant 210 / Vol. 100, No. 3 / Monthly Weather Review Figure 3. — Eye configuration changes in hurricane Debbie from 1300 to 2300 gmt on August 18. Vertical lines indicate the seeding times. early in the day, shifting to the southeast later in the day. The highest echoes (averaging 45,000 ft) were again lo- cated in the northeast quadrant. Echo tops in hurricane Donna, although at greater ranges, were also higher to the right of the storm center (Senn and Hiser 1962). It is interesting to note that, despite the different radar configuration of Debbie on the 2 days, the maximum wind at 12,000 ft on both days before seeding was nearly the same, approximately 100 kt (Gentry 1970). 4. EYE STRUCTURE CHANGES ON AUG. 18, 1969 Figure 3 shows the changes in eye radius, echo-free area, major axis orientation, and eccentricity that occurred just prior to, during, and immediately following the multiple seedings, which are shown by vertical lines at 172 seeding times. The planimetered area is indicated by the dot-dashed line and the area computed from the measured radii is given by the solid line. Eye size changes on the 18th were much more pro- nounced than on the 20th. Unfortunately, radar data quality was poorer on the 18th. The data from 1500 gmt to 2000 gmt are considered most reliable because they consist of an average of measurements from two and sometimes three radars. Careful study of figure 3 leads to the following observa- tions. Approximately 1 hr and 15 min after the first seeding, the eye area began to increase rapidly. During a period of 30 min, the eye increased in area by 50 percent until it disappeared, and a new eye formed. Again, approxi- mately 1 hr after the second seeding, a rapid increase in the area occurred. This time, the area nearly tripled in a period of 15 min. The increase does not appear in the planimetered area due to a difference in interpretation of the echoes which made up the eyewall while it was in the process of re-forming. Just prior to the third seeding, a double eye structure clearly appeared, The larger eye continued to increase in area following the third seeding until it disappeared approximately 1 hr after the seeding. Meanwhile, the smaller eye slowly increased in area. Shortly after the fourth seeding the eye area again in- creased suddenly and disappeared. A smaller eye quickly re-formed and slowly began to increase in area until about 1 hr and 15 min after the fourth seeding when another sudden increase in area occurred. In that case, the area doubled in size in about 10 min. A larger increase may have occurred, but lack of data prevented further measure- ments until shortly after the fifth seeding. At that time, the data showed that a smaller eye had formed. No further data were available after 2300 gmt. The manner in which the sequence described above occurred is difficult to ascertain exactly because of poor definition in the radar photographs of the wall clouds. However, a schematic illustration of the process as it occurred following the third seeding is shown in figure 4. Each of the six views represents a composite of several radar film frames centered on the indicated time. Ap- parently the echoes in the northern and southern segments of the eyewall began moving outward, as shown in the sequence of sketches from 1730 to 1753 gmt. At about 1753, a protrusion in the western segment of the eyewall began moving inward rather quickly. As this movement continued and the echoes spiraled inward, a new and smaller wall cloud was formed as can be seen in the sequence of sketches from 1753 to 1820 gmt. Meanwhile, the northern segment of the original eyewall appeared to break up into smaller echoes as it moved outward and lost its identity while the southern segment moved out- ward and merged with a feeder band. This process is believed to be representative of what happened following the other seedings. However, this cannot be proven conclusively due to the poor quality of the data. The eye retained its elliptical shape during the day with an average eccentricity of about 0.7. The northeast to southwest orientation of the major axis remained nearly 1730 GMT 1745 GMT 1807 GMT 1753 GMT 1820 GMT Figure 4. — Process of eyewall expansion as it occurred after the third seeding of hurricane Debbie on August 18. North is toward the top of each panel. constant for most of the day with the exception of a sudden change between 1800 and 1900 gmt, the time during which the largest expansion of the eyewall occurred. 5. EYE STRUCTURE CHANGES ON AUG. 20, 1969 The changes in eye radius, echo-free area, major axis orientation, and eccentricity on Aug. 20, 1969, shown in figure 5, are more subtle than on the 18th. As was noted earlier, two concentric wall clouds existed on this day as opposed to a single wall cloud on the 18th. The larger eye had a mean radius of 22 n. mi. while the smaller, inner eye had a mean radius of 12 n. mi. The radar data quality was much better on the 20th than on the 18th, although far from optimum. As on the 18th, data exist only from 1 hr before the first seeding until 1 hr after the last seeding. From the second through the fourth seeding, two, three, and sometimes four radars were used to obtain measurements. All agreed rather well; this heightened confidence in the measurements. The first four seedings were conducted on the inner eyewall, but the last seeding was conducted on the outer eyewall. The areas and radii of the large and small eyes showed only minor changes after each seeding. The fol- lowing features are noteworthy. The area of the large eye showed a general trend to decrease in size during the day. The area of the small eye remained nearly constant until 1900 gmt when it began a slight increase, resulting in a much reduced separation between the two wall clouds by the end of the seeding operation. The eccentricity of the two eyes did not change markedly during the day, the outer eye having a mean value of about 0.4 and the inner eye, being more elliptical, having a mean value of about 0.6. The behavior of the major axis orientation was also different on the 20th than on the 18th. On the 20th, the axis did not remain fixed as it did on the 18th, but rotated. March 1972 / Black, Senn, and Courtright / 211 173 Table 2. — Comparison of changes in radar eye radius with changes in the radius of maximum wind on Aug. 18, 1969 Figure 5. — Eye configuration changes in hurricane Debbie from 1100 to 2100 omt on August 20. Vertical lines indicate the seeding times. + 0 MAJOR AXIS + 80 Tn +40 Figure 6. — Schematic illustration of the rotational acceleration and deceleration of the eyewall major axis following each seeding of hurricane Debbie on August 20. The interesting feature about the major axis orientation is that as it rotated it went through a definite cycle that had a period of about 2 hr. As can be seen from figure 5, four of these cycles were observed, one following each seeding. The orientation of the major axis was northwest to southeast at each seeding time, the seedings taking place in the direction of the minor axis. Figure 6 illustrates schematically the rotation cycle of the eyewall. Beginning shortly after each seeding, the rotation rate of the major axis decelerated so that during the first 40 min after seeding it had rotated through only 60°, a rate which continued for the next 40 min. Then rapid acceleration took place so that within the next 40 min the major axis rotated through nearly 240° before decelerating again. Thus, it appears that one effect of seeding was to slow the rotation rate of the major axis of 212 / Vol. 100, No. 3 / Monthly Weather Review Time (OMT) Wind profile Major ails Wind mai. Radar eye orientation orientation radius radius (n.ml.) (n.ml.) 1300-1600 040°-220° 050°- 230° 22 23 (along major axis) 1600-1700 120°-300° 050°-230° 20 17 (along minor axis) 1800-1830 040°-220° 120°-310° 25 11,21 (along minor axis) 1830- 1900 040°-220° 040°-220° 30 14, 30 (along major axis) 1900-2000 040°-220° 070°-250° 20 15 Table 3. — Comparison of changes in radar eye radius with changes in the radius of maximum wind on Aug. 20, 1969 Time (OMT) Wind profile orientation Major axis orientation Wind max. radius (n.ml.) Radar eye radius (n.ml.) 1130-1230 020°- 200° 070°-250° 11,24 12,24 (in between axes) 1450-1500 120°-300° 050°-230° 11,22 11,22 (along minor axis) 1830-1930 020°-200° 030°-210° 14,21 14,21 (along major axis) the elliptical eyewall. Although the rotation of the major axis of an elliptical eye was thought to be indicative of changes in the storm track in hurricane Donna (Sadowski 1961), Senn (1966a) found such rotation occurring in other storms without track changes. 6. COMPARISON OF EYEWALL RADIUS CHANGES WITH CHANGES IN THE MAXIMUM WIND RADIUS OBSERVED IN HURRICANE DEBBIE The process of eye expansion and re-formation is sup- ported to some extent by the behavior of the winds at 12,000 ft on the two seeding days that were monitored by the RFF, DC-6 aircraft and reported by Gentry (1970) and Hawkins (1971). Tables 2 and 3 summarize the comparisons between the orientation of the aircraft passes, the orientation of the major axis of the eye, the radius of maximum wind, and the radar eye radius on the 18th and 20th, respectively. Tables 2 and 3 show a high degree of correlation between the fluctuations in the radius of maximum winds and the eyewall radius along the aircraft track, certainly within the limits of aircraft navigation errors. In general, the eyewall radius is equal to or slightly less than the radius of maximum winds, indicating that the centers of the eyewall and maximum wind region are nearly coincident. The data in table 2 indicate that as the eyewall expands and contracts on the 18th, so does the radius of maximum winds. The data in table 3 show that both the larger 174 eyewall radius and the larger maximum wind radius are decreasing with time on the 20th. As noted earlier, the data in Figure 5 show the inner and outer eyewalls coming closer together with time. Not shown in table 3 is that at a later time, about 5 hr after the seedings had ended, the inner wind maximum had been eroded away and only a single maximum existed at a radius of about 20 n.mi. Therefore, the fluctuations in the radius of maximum wind on both days tend to support the fluctuations in the radar eyewall radius. It is interesting also to compare the changes in eyewall radius with recent numerical modification experiments carried out by Rosenthal (1971). In his experiments, he used a circularly symmetric model (Rosenthal 1970) with continuous enhanced heating (to simulate the effect of seeding) beyond the radius of maximum winds for a period of 10 hr. Comparisons with intermittent seeding in a real asymmetric storm are, therefore, difficult. Nevertheless, it is significant that the maximum wind region in the modi- fied model storm began shifting outward shortly after the enhanced heating had commenced. By 2 hr after the enhanced heating had ceased, the maximum wind region had shifted from a 10- to a 20-n.mi. radius. It would appear, therefore, that the timing of the outward shifting of the radar eyewall, beginning 80 min after each seeding, is not an unreasonable response time for the storm to react to the seeding. The fact that the eyewall expansion did not continue is significant. Perhaps in future seeding experiments the seedings should be repeated more frequently and at successively larger radii as the eyewall expands. In this way, possibly the eyewall and the maximum wind region could be main- tained at a larger than natural radius and thus prolong the reduction in wind speed. 7. EYE SIZE CHANGES IN UNSEEDED STORMS The question may be asked whether or not eye size changes described previously would have occurred if the storm had not been seeded. Hoecker and Brier (1970) have conducted a study of the eye size changes in Hurricanes Carla of 1961, Betsy of 1965, and Beulah of 1967, covering a continuous time period of about 24 hr for each storm. Airborne radar was used only for the Carla study, while ground based radar was used for all three storms. The data sample for Carla was the longest (40 hr). For this storm, the eye decreased in size fiom a 30 n.mi. diameter to 23 n.mi. diameter during the first 24 hr and remainod relatively constant thereafter. Superimposed upon this trend were shorter period fluctuations of the order of ±4 n.mi. in 4 hr. The Betsy and Beulah eye sizes behaved somewhat similarly. It should be mentioned that during the period of study, both Carla and Beulah had a double eye structure, while Betsy had a single eye. In the above data sample, there was no evidence of cyclic, short period (on the order of 2 hr) changes in the eye size or shape, or even any sudden individual changes of the magnitude and time scale observed in Debbie. However, sudden significant changes in eyewall structure probably occur at some time or other in the life history of Figure 7. — Typical APS-45 radar photograph showing the verti- cal structure of radar echoes distorted 10:1 in the vertical. Verti- cal lines are range marks at 20-n.mi. intervals. The horizontal line is the 20,000-ft height line. The sea return can be seen as the fuzzy horizontal echo at the bottom of the photograph. The bright band is the thin band echo just below the height line. The azimuth of the photograph is indicated by the pointer beside the data card, which in this case indicates 360°. most hurricanes. Such changes usually accompany rapid deepening or filling of the storm such as during landfall or while passing over warmer or colder than normal sea surface temperatures. Recently, Fujita (1971) documented the eye diameter changes in hurricanes Camille of 1969 and Celia 1970 for a 7%-hr period before and just after landfall. His data showed some evidence of a cyclic change in the major and minor axis diameters beginning about 2 hr before landfall. The eye diameters were nearly constant prior to this time. Both the period and amplitude of the cycle were about one half that found in Debbie. Therefore, the data studied by Hoecker and Brier as well as several unpublished accounts of hurricane radar observations suggest that cyclic eye size changes of the type observed in Debbie may be somewhat unique. However, further study of unseeded storms is necessary to be more certain of this. 8. DISTRIBUTION OF THE BRIGHT BAND IN HURRICANE DEBBIE One of the interesting features of the APS-45 RHI photographs taken in hurricane Debbie was the frequent occurrence of the bright band. This is a region of enhanced radar reflectivity at levels just below the 0°C isotherm caused by the melting of frozen precipitation. The signifi- cance of this feature is that it can only exist in regions of weak vertical motion (less than 1 m/s). Hence, its existence in a hurricane can be used to infer regions of weak vertical motion. A typical example of the bright band is seen in figure 7. The bright band is visible at times out to a range of 30 n.mi. from the radar. March 1972 / Black, Senn, and Courtright / 213 175 Figure 8. — Average bright-band heights/thicknesses (103 ft) in hurricane Debbie from APS-45 at 1,000 ft on August 20. The arrow in the upper left in this and subsequent figures is the storm direction of motion. The range marks are at 50-n.mi. intervals. Figure 9. — Average bright-band heights/thicknesses (103 ft) in hurricane Debbie from APS-45 at 10,000 ft on August 20. Figures 8 and 9 show the distribution of the bright band and its mean height and thickness as measured by APS-45 radars on two different aircraft. Lack of good stabilization of the radar antenna presented problems in analyzing the data. Although care was taken to select only those photographs where the sea return was level (indicating a properly oriented antenna platform), there was some scatter in the data. On the early low-level flight (1,000 ft) before seeding, the RHI radar was able to detect a well-defined bright band in most quadrants of the storm (fig. 8) except for the northwest quadrant beyond about 30 n.mi. from the storm center. Considering the number of RHI photos taken in Debbie and the concentration of the observer on specific areas of the storm for extended periods, we cannot accurately compare the areas where the bright band phenomenon is observed with those lacking this feature. However, probably 25 percent of the RHI photos indicated that a bright band was in evidence. Contrasting these values with the data obtained from the higher level flight (10,000 ft) during and after the seedings (fig. 9) in the area south of the storm center, we find that the bright bands observed from the high-level flight are higher by 3,000-4,000 ft than those from the low-level flight for unknown reasons. A small amount of over- lapping data from the two flights in the southwest quadrant about 30 n.mi. from the center at nearly the same times indicate that the height differences may be due to radar calibration. However, bright band heights and thicknesses are consistent for each radar. Due to 214 / Vol. 100, No. 3 / Monthly Weather Review radar problems on both flights, it was impossible to calibrate for height in an absolute sense; and neither operator succeeded in confirming an RHI picture of one of the Stormfury aircraft at a known altitude. In general, the bright band appears to be 1,000-2,000 ft higher near the outer periphery of the eyewall than at the 100-n.mi. radius. The thickness appears to be about 2,000 ft and uniform throughout the storm. The average thickness of the bright band is probably very close to the true value because it was often near the threshold of de- tection. It would not have suffered from even the normal beam width stretching of less than 1,000 ft at the short observation ranges (10-30 n.mi.). The 0°C isotherm was found by aircraft observations and dropsondes in various parts of Debbie to be near 15,000 ft at ranges of about 100 n.mi. to the west, south, and southeast of the eye and was estimated to be near 17,000 ft near the eyewall. One would expect the "melting level" represented by the bright band to be topped near those levels and to extend below them. Observations from the earlier aircraft indicated the top of the bright band to be near 14,000 ft at 100 n.mi. and about 16,000 ft near the eyewall. The later aircraft data indicated a bright band top height of 19,000 ft near the eyewall, which appears to be too high. Comparison of the Debbie data with those obtained in Betsy of 1965 (Senn 19666) and Beulah of 1967 shows relatively good agreement. The Beulah data were obtained at greater distances from the storm center and indicated that the average bright band height was 14,000 ft. In Betsy, the bright band also tended to be higher in and near the eyewall than at greater distances, and the phe- 176 nomenon was found in all azimuths and most ranges from the storm center. The Debbie data is also in relatively good agreement with the RDR-1 RHI data of Hawkins and Rubsam (1968o, 19686) in hurricane Hilda of 1964. The bright band was found in nearly all quadrants of the storm on three successive days. The top of the bright band was esti- mated to be at about 16,000 ft between radii of 50-70 n.mi. with some suggestion that the height increased near the eyewall. Hawkins noted also that convective clouds were present throughout the bright band region. Such widespread existence of the bright band in the hurricane, which is convectively driven, raises a rather basic question. How can the bright band, which is formed by frozen precipitation falling in less than a 1 m/s updraft and melting beneath the 0°C isotherm, exist in the pres- ence of numerous convective echoes? Atlas et al. (1963) have stated that the bright band is "characteristic of stratiform precipitation . . . and is never found in regions of strong convection." However, perhaps the data are not as contradictory as they may seem. The bright band observations in Debbie were frequently between the rainbands in regions of strati- form precipitation. It has been known for some time (Malkus et al. 1961) that convective bands are sometimes embedded in regions of stratiform precipitation in the hurricane. However, the bright band was also observed within convective towers in many cases. This is not too surprising, as the bright band has been observed within thunderstorms that have passed their mature stage (Battan 1959). In a hurricane, therefore, the convective towers are undoubtedly in different stages of maturity and there should be no reason why the bright band should not exist in a convective echo that is decaying, while nearby the echoes are intense and growing. In many cases, one "bubble," or part of a convective system, is growing while another part is decaying as is clearly shown by Byers and Braham (1949). The data from both Betsy of 1965 and Debbie of 1969 show such convective echoes existing in the company of large areas of stable precipi- tation and bright band echo formations. 9. PRECIPITATION TILT IN HURRICANE DEBBIE Another interesting measurement that can be made with the narrow beamwidth APS-45 RHI radar is the vertical tilt of the precipitation echoes. If one makes certain assumptions regarding drop sizes and fall velocities, the RHI data can be used to indicate the sense and a crude estimation of the magnitude of the wind shear in the precipitation layer observed. The echoes photographed are undoubtedly in various stages of development and hence the degree of echo tilt for a given wind shear will probably be variable, giving rise to a spectrum of inferred shears. However, the variability would be mainly in the magnitude of the shear because all the echoes should tilt to a greater or lesser degree in the direction of the shear. Even assuming a given echo is tilted significantly, only the component of tilt along the radar beam will be detected. In fact, perhaps 75 percent of the usable RHI photographs were taken looking toward or away from the eye so that mainly the radial component of tilt was measured. There- fore, due to radar orientation as well as to meteorological reasons, only about 50 percent of the total number of echoes studied had a significant tilt (greater than 0.5 n.mi. in 15,000 ft). It was possible to make measurements with a 0.5-n.mi. horizontal resolution and a 1,000-ft vertical resolution. Echo tilts were measured for three height intervals: surface to 16,000 ft, 16,000-30,000 ft, and above 30,000 ft. The largest data sample was in the lowest layer, where 115 tilts were measured. In the middle layer, 76 tilts were measured. Less than a dozen tilting echoes were measured in the upper layer and are not presented here. Figures 10 and 11 show the observed vector component of the precipitation tilt along the radar beam for the low and middle levels, respectively. The average magnitude of the tilts in both layers was nearly the same, being slightly more than 1 n.mi. hori- zontally in 3 n.mi. vertically. This corresponds to a tilt angle of 20°. The maximum tilt measured corresponded to an angle of 45°. The direction of the tilts in the low level was generally radially outward in the front and right quadrants and generally radially inward in the left and rear quadrants. The tilt in the middle level was generally outward and slightly upband (cyclonic tan- gential component) in the right quadrant and outward and downband (anticyclonic tangential component) in the left, front, and rear quadrants. The magnitude of the representative wind shear vectors in a hurricane can be roughly estimated from the precipi- tation tilt vectors using a procedure established by Senn (19666). He used reflectivities obtained in hurricane Betsy from a well-calibrated radar to infer subtropical raindrop concentrations from the data of Mueller and Jones (1960). Then using the terminal fall velocity data of Gunn and Kinzer (1949) which assume the drops to be falling through stagnant air, the wind shears were computed from the APS-45 RHI tilt data. For an echo tilt of 6,000 ft in a 16,000-ft layer, the wind shear is of the order of 6 kt. Using the precipitation tilt-wind shear relationship de- vised by Senn (19666), wind shears computed from air- craft data, where available, were converted to tilts. They are plotted in figures 10 and 11 as dotted arrows. In only a few instances were aircraft at 1,000 ft coincident in space and time with aircraft at 12,000 ft and likewise, only in three cases were aircraft at 31,000 ft coincident in space and time with aircraft at 12,000 ft. Nevertheless, wind shears were computed for these few cases, but they are not considered any more reliable than the precipitation tilt-inferred shears because the low-level Navy aircraft, the middle-level RFF aircraft, and the high-level Air Force aircraft all use different navigation and wind com- puting schemes, none of which is calibrated with respect to either of the others. However, fair agreement between the tilts inferred from the aircraft winds and the precipitation tilts in the lower level can be seen in figure 10. The wind shears computed from aircraft data ranged from 5 to 20 kt in the directions March 1972 / Black, Senn, and Courtright / 215 177 Figure 10. — Precipitation-tilt vector composite for the layer 0-16,000 ft in hurricane Debbie on August 20. The scale in the lower left gives the length of the echo tilt vectors in thousands of feet. The clotted arrows are tilts inferred from wind shears com- puted from aircraft wind measurements at 1,000 and 12,000 ft. Figure 11. — Precipitation-tilt vector composite for the layer 16,000-30,000 ft in hurricane Debbie on August 20. The dotted arrows are tilts inferred from wind shears computed from air- craft wind measurements at 12,000 and 31,000 ft. shown which was approximately the range in wind shear magnitude inferred from the tilt vectors in figure 10. A lesser degree of agreement is seen in the higher layer in figure 11 probably because the actual wind shears are larger in magnitude. The aircraft-derived wind shears ranged from 20 to 35 kt while the precipitation-inferred shears ranged from 5 to 20 kt. It appears that the tilt vectors in figures 10 and 11 average about half those found by Senn (19666) in hur- ricane Betsy of 1965 for an equivalent layer. The difference could be real, for Betsy was increasing from tropical storm to hurricane intensity while the observations were being made; whereas Debbie was already a respectable hurricane that appeared to decrease in intensity while under obser- vation. However, the authors must consider the radar problems and the distinct possibility that less echo tilt was noted in Debbie solely because the radar performance was below normal in comparison with the Betsy observa- tions. Echo tilts computed by the authors from Heidi of 1967 were equal in magnitude to the Betsy tilts or about twice as great as the Debbie tilts. 10. CONCLUSIONS Airborne radar photographs of hurricane Debbie were used to determine eye radius and echo-free area, major axis orientation, eye eccentricity, bright-band height and thickness, and precipitation tilt. The available PPI data on both days were used to make the eye size measure- ments at 5-min intervals beginning 1 hr before the first 216 / Vol. 100, No. 3 / Monthly Weather Review seeding and ending 1 hr after the last seeding. Results for August 18 th show sudden increases in echo-free are a at seed- ing time plus 1 hr and 15 min. Area increases ranged from 50 percent to threefold. Results for the 20th were quite different. A double eye structure was present on this day as opposed to a single eye on the 18th. The echo-free area within the smaller eye remained nearly constant throughout the day and the larger eye slowly decreased in area. The only evidence of seeding effects on the 20th as noted from the radar photographs was observed in the rotation rate of the major axis of the elliptical eye. A slowing of the rotation rate was observed within 10 min of each seeding followed 1 K hr later by a rapid increase in the rotation rate that continued until the next seeding time. The period of this cycle (the time required for one revolution of the major axis) was about 2 hr, the same period as the seedings. The changes in the radar eyewall radius were followed quite closely in time by changes in the radius of maximum winds. The hypothesis suggests that the large fluctuations observed on the 18th were due to seeding beyond the radius of maximum winds while the general lack, of such changes on the 20th might be attributed to seeding at smaller radii on that day. It is suggested, therefore, that, to obtain a more lasting modification of a storm, seeding be carried out more frequently, perhaps at 1-hr intervals and at successively larger radii as the eyewall expands. In the case of a storm with concentric wall clouds, seeding should probably be carried out beyond the outer wind maximum. 178 On the one day with usable RHI data, the echo tilts indicated that the shear is most often found in the lower levels of the storm. Most echoes leaned slightly back- ward, up spiral bands and radially outwaid as they in- creased in height. The average tilt was about 1 n.mi. horizontally in 3 n.mi. vertically, corresponding to a wind shear of 6 kt per 15,000 ft. The bright band was found in most quadrants and at most ranges from the storm center. It was slightly higher near the eyewall than at the 100-n.mi. radius. ACKNOWLEDGMENTS The authors gratefully thank the commander and personnel of U.S. Navy Weather Reconnaissance Squadron Four and the radar data advisors, without whose help and cooperation most of these data would not have been obtained. The assistance of the dedicated people of the U.S. Air Force and NOAA-Research Flight Facility was also necessary in gathering some of the radar and other data used. Finally, thanks are due R. C. Gentry, Director of the National Hurricane Research Labora- tory and Project Stormfury for his support and encouragement of this work. REFERENCES Atlas, David, Hardy, Kenneth R., Wexler, Raymond, and Boucher- Roland J., "On the Origin of Hurricane Spiral Bands," Proceedings of the Third Technical Conference on Hurricanes and Tropical Meteorology, Mexico City, Mexico, June 6-12, 1963, Geofisica Inter- national, Vol. 3, No. 3/4, Mexico City, July/Dec. 1963, pp. 123- 132. Battan, Louis J., Radar Meteorology, University of Chicago Press, Chicago. 111., 1959, 161 pp. (see p. 96). Byers, Horace R., and Braham, Roscoe R., Jr., The Thunderstorm, U.S. Department of Commerce, Washington, D.C., June 1949, 287 pp. Fortner, Limon E., Jr., "Typhoon Sarah 1956," Bulletin of the American Meteorological Society, Vol. 39, No. 12, Dec. 1958, pp. 633-639. Fujita, Tetsuya T., The University of Chicago, 111., July 1971 (per- sonal communication). Fujita, Tetsuya T., and Black, Petei G., "In- and Outflow Field of Hurricane Debbie as Revealed by Echo and Cloud Velocities From Airborne Radar and ATS-III Pictures," Proceedings of the Fourteenth Radar Meteorology Conference, Tucson, Arizona, November 17-20, 1970, American Meteorological Society, Boston, Mass.,. 1970, pp. 353-358. Gentry, R. Cecil, "Project STORMFURY," Bulletin of the Ameri- can Meteorological Society, Vol. 50, No. 6, June 1969, pp. 404-409. Gentry, R. Cecil, "Hurricane Debbie Modification Experiments, August 1969," Science Vol. 168, No. 3930, Apr. 24, 1970, pp. 473- 475. Gentry, R. Cecil, and Hawkins, Harry F., "A Hypothesis for the Modification of Hurricanes," Project STORMFURY Annual Re- port, 1970, Appendix B, Miami, Fla., May 1971, pp. B-l— B-15. Gunn, Ross, and Kinzer, Gilbert D., "The Terminal Velocity of Fall for Water Droplets in Stagnant Air," Journal of Meteorology, Vol. 6, No. 4, Aug. 1949, pp. 243-248. Hawkins, Harry F., "Comparison of Results of the Hurricane Debbie (1969) Modification Experiments With Those From Rosenthal's Numerical Model Simulation Experiments," Monthly Weather Review, Vol. 99, No. 5, May 1971, pp. 427-434. Hawkins, Harry F., and Rubsam, Daryl T., "Hurricane Hilda, 1964: I. Genesis, as Revealed by Satellite Photographs, Conven- tional and Aircraft Data," Monthly Weather Review, Vol. 96, No. 7, July 1968a, pp. 428-452. Hawkins, Harry F.. and Rubsam, Daryl T., "Hurricane Hilda, 1964 : II. Structure and Budgets of the Hurricane on Octobar 1, 1964." Monthly Weather Review, Vol. 96, No. 9, Sept. 19686, pp. 617-636. Hoecker, Walter, and Brier, Glenn, "Measurement of Hurricane Eye Diameter by Land-Based and Airborne Radar," Air Re- sources Laboratory, NOAA, Washington, D.C., 1970 (personal communication) . Jordan, Charles L., and Schatzle, Frank J., "Weather Note — The 'Double Eye' of Hurricane Donna," Monthly Weather Review, Vol. 89, No. 9, Sept. 1961, pp. 354-356. Malkus, Joanne S., Ronne, Claude, and Chaffee, Margaret, "Cloud Patterns in Hurricane Daisy, 1958," Tellus, Vol. 13, No. 1, Feb. 1961, pp. 8-30. Mueller. E. A., and Jones, D. M. A., "Drop-Size Distributions in Florida," Proceedings of the Eighth Weather Radar Conference, San Francisco, California, April 11-14, 1960, American Meteorological Society, Boston, Mass., 1960, pp. 299-305. Rosenthal, Stanley L., "A Circularly Symmetric Primitive Equation Model of Tropical Cyclone Development Containing an Explicit Water Vapor Cycle," Monthly Weather Review, Vol. 98, No. 9, Sept. 1970, pp. 643-663. Rosenthal, Stanley L., "A Circularly Symmetric, Primitive Equa- tion Model of Tropical Cyclones and Its Response to Artificial Enhancement of the Convective Heating Function," Monthly Weather Review, Vol. 99. No. 5. May 1971, pp. 414-426. Sadowski, Alexander, "Radar Analysis of Hurricane Donna's Recurvature," Proceedings of the Second Technical Conference on Hurricanes, Miami Beach, Florida, June 27-30, 1961, National Hurricane Research Laboiatory, U.S. Weather Bureau, Miami, Fla., Mar. 1962, pp. 63-69. Senn, Harry V., "Radar Hurricane Precipitation Patterns as Track Indicators," Proceedings of the Twelfth Conference on Radar Meteorology, Norman, Oklahoma, October 17-20, 1966, American Meteorological Society, Boston, Mass., 1966a, pp. 436-440. Senn, Harry V., "Precipitation Shear and Bright Band Observations in Hurricane Betsy 1965," Proceedings of the Twelfth Conference on Radar Meteorology, Norman, Oklahoma, October 17-20, 1966, Amer- ican Meteorological Society, Boston, Mass., 1966b, pp. 447-453. Senn, Harry V., and Hiser, Homer W., "Effectiveness of Various Radars in Tracking Hurricanes," Proceedings of the Second Technical Conference on Hurricanes, Miami Beach. Florida, June 27-30, 1961, National Hurricane Research Laboratory, U.S. Weather Bureau, Miami, Fla., Mar. 1962, pp. 101-114. Sheets, Robert C, National Hurricane Research Laboratory, NOAA, Coral Gables, Fla., Sept. 1970 (personal communication) . Simpson, Robert H., Ahrens, Merle R., and Dacker, Richard D., "A Cloud Seeding Experiment in Huiricane Esthei , 1961," National Hurricane Research Project Report No. 60, U .S. Weather Bureau, Washington, D.C., Apr. 1963, 30 pp. Simpson, Robert H., and Malkus, Joanna S., Hurricane Modifica- tion: Progress and Prospects, 1964, U.S. Weather Bureau, Wash- ington, D.C., Aug. 1964, 54 pp. United States Navy Fleet Weather Facility, STORMFURY Opera- tions Plan No. 1-69, Naval Air Station, Jacksonville, Fla., June 1969, 100 pp. [Received March 29, 1971; revised August 26, 1971] March 1972 / Black, Senn, and Courtright / 217 18 Reprinted from Journal of Applied Meteorology, Vol. 11, Xo. 2, March, 1972, pp. 283-297 American Meteorological Society Printed in U. S. A. The Large-Scale Movement of Saharan Air Outbreaks over the Northern Equatorial Atlantic1 Toby N. Carlson National Hurricane Research Laboratory, NOAA, Miami, Via. 33124 and Joseph M. Prospero University of Miami, Rosensliel School of Marine and Atmospheric Science, Miami, Fla. 3314V (Manuscript received 20 July 1971) ABSTRACT The intense and prolonged heating of air passing over the Sahara during the summer and early fall months forms a deep mixed layer which extends up to 15-20,000 ft during July, the warmest month. The dust-laden heated air emerges from West Africa as a series of large-scale anticyclonic eddies which move westward over the tropical Atlantic above the trade-wind moist layer, principally in the layer between 5000 and 15,000 ft (600-800 mb). Measurements made during BOMEX show that this Saharan air is characterized by high values of potential temperature, dust and radon-222 which confirm a desert origin. As the parcels of air within the layer proceed across the Atlantic the continuous fallout of particulate matter and the mixing at the base of the layer cause dust to be transferred to the lower levels where its concentration may become sufficiently great to produce dense haze at the surface over wide areas over the Atlantic and Caribbean in the latitude belt 10-25X. Nevertheless, measurements indicate that the dust concentration and associated haziness are greater at 10,000 ft than at the surface. The presence of Saharan air over the Caribbean can be recognized on conventional meteorological sound- ings as a virtually isentropic layer within which the potential temperature is about 40C; the mixing ratio within this layer generally remains fairly constant with height with typical mean values of 2-4 gm kg-1 The upper surface of the Saharan air layer, clearly visible from above as a sharply defined haze top, coincides with an inversion, above which the mixing ratio decreases rapidly with height. The lower portion of the isentropic Saharan air layer may be as much as 5-6C warmer than the normal tropical atmosphere; con- sequently, there is a strong suppressive inversion above the moist trade-wind layer. There is also a sharp horizontal temperature gradient between the Saharan dust plume and the normal tropical air mass; aircraft penetrations into Saharan air at heights of 700-800 mb show that the discontinuity between Saharan and non-Saharan air is front-like in character inasmuch as the potential temperature and mixing ratio may change by several degrees Celsius and several grams per kilogram, respectively, over a distance of just a few kilometers. Because of the steep (adiabatic) lapse rate in the Saharan air, the positive temperature anomaly diminishes rapidly with height and tends to vanish near 650 mb, above which the dusty air may be slightly cooler than the normal tropical environment. Associated with this large-scale temperature con- trast is a wind maximum of up to 40-50 kt in the Saharan air layer, usually between 600 and 700 mb. The westward speed of the Saharan air mass is usually about 15 kt, requiring about 5-6 days to cross the Atlantic. The leading edge of the Saharan air is often found immediately to the rear (east) of a large-ampli- tude African disturbance which also migrates from Africa to the Caribbean during the summer months at about the same forward speed. Normally the strongest winds and highest dust concentrations in the Saharan air are found in the southeasterly winds behind the disturbance and are therefore associated with the so- called "surges in the trades" which are often observed in the tropical Atlantic. Saharan air pulses tend to leave the continent of Africa with a potential temperature of 43-44C, about 3-4C higher than that found in the Saharan layer over the Caribbean. This apparent cooling is due to net radiation losses within the Saharan air which amount to about 0.7C per day. At the same time the Saharan air sinks by 50-100 mb (a mean descending motion of 1-2 mm sec-1) between Africa and the Caribbean. A model is proposed which depicts the movement of Saharan air from Africa to the Caribbean and its interaction with African disturbances. Although it is not known what effect, if any, the dust plume has upon the growth or suppression of disturbances, it is clear that the warmth of the Saharan air has a strong suppressive influence on cumulus convection and that, as a result, the advancing dust pulse is often marked by rapid clearing behind the disturbances. 1. Introduction the Caribbean, specifically to the region of Barbados, Recent work by Prospero and Carlson (1970) and West Indies (13N, 59W). Aerosol studies carried out at Prospero el al. (1970) has documented the westward .Contribution No. 1455 from the University of Miami movement of masses of dusty air from North Africa to Rosenstiel School of Marine and Atmospheric Science. 180 284 JOURNAL OF APPLIED METEOROLOGY VOLUML 11 -26 E oU ±1 JJ14. T? 1 \ TTT JUNE +IWf hi 1969 JULY 1970 Fig. 1. Atmospheric dust concentration, Barbados, West Indies, May-July 1969 and July 1970. The dust concentration in surface air was measured using the mesh technique described by Prospero (1968). The aerosol collectors used in this study have a low collection efficiency for particles below several microns radius; calculations based on independent measurements at Barbados of aerosol size distributions in the range of 0.3-5 n radius (Blifford, 1970) indicate that the actual dust load is about three to six times greater than the values presented here. Sampling was carried out continuously (i.e., 24 hr a day) during the periods shown. The individual data points for the 1969 period represent the average dust load (fig m~3) for a 24-hr period as measured from 1800 (all times local standard, LST) the preceeding day; for example, the data point for 20 May 1969 indicates that the average dust load for the period 1800 LST 19 May to 1800 LST 20 May was 19.4 fig per cubic meter of air. During 1970, the sampling time on some days was divided into two periods: from 1800 to 0800 the following day and from 0800 to 1800; for example, the data bar plotted for 11 Julv indicates that the average dust load from 1800 LST 10 July to 0800 LST 11 July was 16.2 Mg m"3 while the dust load from 0800 to" 1800 LST 1 1 July was 77.7 fig m~3. Open triangles indicate that no sample was taken. Barbados between 1965 and 1969 show that the amount of airborne dust reaching the island is highly variable from day to day, especially during the summer months when the atmospheric dust concentration is greatest (see Fig. 1). The average daily dust load during this four-year period was ~2 ^g m~3, although amounts in excess of 20 /*g m~3 are not uncommon (Prospero, 1968; Prospero el al., 1970). In contrast, during the winter the dust content of the atmosphere at Barbados is smaller by one or two orders of magnitude; the dust is also grey or black as compared to the light red-brown color of the summer dust. The onset of the maximum dust trans- port period is often rather abrupt (Prospero, 1968) and corresponds to the inception of airflow from the western Sahara. From August until October or November the average dustiness at Barbados is fairly high but ex- ceptionally dusty periods become less frequent and less intense, and there are extended periods when the air is essentially dust free. The probable origin of the red- brown dust transported during summer and early fall lppears to be the area encompassing southern Algeria, Mali, Mauritania, Spanish Sahara and Senegal (Prospero el al., 1970). From December through May, the source region of the black dust is primarily the semi- arid grasslands south of the Sahara (Delany el al., 1967). Although we have been able to relate several periods of unusually' high dust concentrations at Barbados to specific outbreaks of dust-storm activity over the Sahara (Prospero el al., 1970), conventional surface observations of dust-storm activity or haze over West Africa and the eastern Atlantic are not good predictors for the eventual arrival of African air (and dust) at Barbados. Instead, we have found that the presence of African air over the equatorial Atlantic is related to large-scale processes occurring over Africa and is readily detected on meteorological soundings as a layer of relatively warm air between 600 and 800 mb. The air within this layer has an almost constant potential temperature and slowly changing mixing ratio with height. Evidence which we gathered during BOMEX and on later field expeditions further shows that the variation in the dust content of the air at Barbados is related to the passage of tropical disturbances (easterly waves) and that the trajectories of the dust}' air parcels between Africa and the Caribbean are influenced by these travelling systems, called "African" disturbances by Carlson (1969a, b) because of their sub-Saharan origin. It is our purpose in this paper to describe the African dust layer and its movement across the Atlantic into the Caribbean. To accomplish this, we will utilize con- ventional meteorological observations combined with a number of specialized measurements made by us during BOMEX in 1969 and on the oceanographic vessel, the U.S.C.G. Discoverer, during the summer of 1970 as well as on the island of Barbados and at Miami. 2. Background The appearance over the Atlantic of African air parcels having uniquely identifiable characteristics can be more readily understood by considering the early 181 March 1972 CARLSON AND J. M. PROSPERO 285 history of the airstream over Africa. As air proceeds across the arid portions of the continent, it is intensely heated at the earth's surface because of the strong solar radiation, a disproportionately large fraction of which is redistributed to the lower atmosphere in the form of sensible (rather than latent) heat. This heating produces vigorous dry convection throughout a layer of air in convective contact with the ground. Because of the vastness of the desert terrain and the intensity of the heating, a column of air entering the Sahara remains in convective contact with the ground for a period of at least a few days and acquires a uniform potential temperature in the mixing layer, typically about 45C in July; this layer can extend to 500 mb in the warmest season (Carlson and Ludlam, 1965). As the layer of dry convection deepens, the rate of radiative heat loss gradually increases, eventually approaching the rate at which energy is gained by heating; consequently, the rate of warming becomes progressively smaller. The soundings in Fig. 2 show that in travelling from Tobruk, which lies on the coast upwind from the desert, to Tamanrasset in the interior, the air acquires a consider- able amount of sensible heat but little moisture. At Tamanrasset the potential temperature is almost con- stant at 317K (44C) up to 450 mb where it is capped by a stable layer. Similarly, the mixing ratio changes slowlv up to 550 mb, above which the air is saturated. The Saharan air is therefore especially warm and dry in the lower part. However, the air is cooler than the surrounding tropical air mass near the top where the rising thermals, because of vigorous mixing, overshoot the equilibrium level and incorporate air from above into the mixing layer. This results in the further deepening of the mixing layer and the formation of an inversion at its top. The relative humidity in the layer of dry mixing increases very rapidly with height and may achieve saturation near the top, as in the case for Tamanrasset (Fig. 2) ; however, saturation usually does not occur unless the layer is lifted subsequent to mixing. (Taman- rasset lies on a high mountain ridge and the air arriving there is subject to some lifting.) 3. Structure of the dust layer as revealed by air- borne measurements Radon-222 measurements made from aboard aircraft during BOMEX have been very useful in the study of Saharan air transport. Randon-222 is a radioactive (half-life 3.82 days) inert gas which is produced by the uranium-238 decay series. [In our work, we determined the air concentration of radon by filtering out and im- mediately measuring the activity of the short-lived daughter products of radon, principally Pb-214 and Bi-214. The procedures are discussed in Prospero and Carlson (1970).]] As is the case with sensible heat and water vapor, radon is introduced into the atmosphere at the earth's surface (Servant, 1966); since radon has about the same vertical eddy diffusivity as that of heat Fig. 2. Soundings for Tamanrasset (southern Algeria) at 1200 GMT 9 July 1959 and Tobruk, Lybia [(July mean), after Carlson (1965)]. Henceforth, the solid lines refer to the temperature and the dashed lines to the dewpoint (mixing ratio). Horizontal scale lines labelled in degrees Kelvin are the dry adiabats (isentropic surfaces). or water vapor (Pearson and Jones, 1966), the gas should achieve a similar vertical distribution throughout the mixing layer. However, the emanation rate of radon from the ocean is about 100 times less than that over land (Servant); consequently, oceanic air parcels which have not been in recent convective contact with a continental land surface (i.e., within several radon-222 half-lives, or roughly ten or more days) will have a relatively low radon concentration. The general validity of this reasoning is supported by radon-222 measure- ments made during a trans-Atlantic flight from Bar- bados to Africa and return (Prospero and Carlson, 1970); the radon concentration within areas of dense dust-produced haze was markedly higher than in clear areas or above the upper inversion. a. Case 1: 14 July 1969 The mesoscale structure of the African dust layer over the western tropical Atlantic can be illustrated using data gathered by us during BOMEX. On 14 July 1969, a weak African disturbance arrived over the BOMEX area and produced a dramatic increase in cloudiness, a common occurrence when easterly waves arrive at the eastern side of the semi-permanent upper cold trough in this area (Frank, 1969). On the 13th, con- siderable cloudiness formed just east of the wave axis (as defined by the wind shift line on streamline charts), which in Fig. 3 is shown near 56W. A number of research aircraft operated by the Research Flight Facility (RFF) of the National Oceanic and Atmospheric Administra- tion (NOAA) investigated the disturbance. The main cloud area (labelled "dense cloud, showers" in Fig. 3) consisted of altocumulus and altostratus having a base slightly above the upper flight level (10,000 ft); within this cloud area were embedded rising cumulus towers whose tops were less than 25,000 ft (as viewed on the range-height radar of the DC-6), except for one cell 182 286 JOURNAL OF APPLIED METEOROLOGY Volume 11 CLOUD and JULY 14 Fig. 3. Weather and haze distribution for approximately 1600 GMT 14 July 1969. The triangular DC-6 aircraft track is divided into half-hour segments by the rilled circles, the first and last half- hour intervals beginning at 1400 and 2330 GMT, respectively. The letters H and h refer to visual haze observations made by aircraft at 10,000 and 1300 ft, respectively. The letter N signifies that no haze was reported by aircraft. The 10,000 ft streamlines (solid barbed lines) and isotachs (thin dashed lines labelled in knots for speeds of 30 and 40 kt) are shown in the figure. Major cloud areas are enclosed by the various stippling and hatching and are labelled accordingly. The two cross-hatched areas located inside the main cloud band near the disturbance axis contain the majority of radar echoes observed by the DC-6 aircraft. Repre- sentative echo tops are indicated by the labels 25K (25,000 ft) and35K (35,000 ft). (near the flight track at 1530 GMT near 54W) which reached 35,000 ft. To the east of this cloud mass was a thin broken layer of altocumulus topped by some cirrus and, beyond that a virtually clear zone. After 1730 GMT the aircraft encountered a shallow layer of strato- cumulus along the eastern edge of the track from which it emerged near 1930. Radon concentrations (measured aboard the DC-6 RFF aircraft, 40C) are listed on Fig. 4 along the outside of the track in units of picocuries per standard cubic meter of air (pCi m~3; a picocurie is 10~12 curies or 2.2 disintegrations per minute). The color density of the filters used for the radon measurements provides a rough indication of the atmo- spheric dust concentration. We established a logarithmic color scale on the basis of the variations of color density observed between successive filters having a three-fold difference in sampling duration. (Series of samples in which sampling times alternated between 5 and 15 min were frequently taken whenever the aircraft maintained constant altitude for extended periods.) The color sort- ing of the approximately 1400 filters obtained during 300 hr of in-flight sampling during BOMEX yielded a graded color scale of 10; this scale was placed on a semi- quantitative basis by correlating the color density of a similar series of filter samples taken on Barbados with the dust loads measured there using the previously described mesh technique (Prospero, 1968). The scale of 10 represents a dust concentration variation which ranges from less than 1 fig m~3 to approximately 50-100 Hg m~3. The dust loading estimates derived from the filter color index for the flight of 14 July are shown along the inside of the track. Ahead of the wave axis, the radon concentrations were about 3—1 pCi m-3; these values are typical oceanic background levels and are consistent with the inter- pretation that the air parcel had not recently been in convective contact with a continental land surface. In addition, the filter color index was low. This situ- ation persisted until the wave axis had been passed and the plane began to level off near 10,000 ft after an ascent from 1300 ft. At that point the radon began to increase rapidly toward the east, reaching 41 pCi m~3 in the cloud-free zone near 49W. The filter color index increased in the same manner, showing the dustiest air to be slightly east of the highest radon concentra- tions and near the position of a warm temperature ridge. The onset of these higher values of radon and dust load (i.e., filter index) was accompanied by a marked increase in temperature in which the potential temperature rose by 3.5C over the course of ~100 mi, as indicated in Fig. 4. In this particular example, the light showers intermittently falling from the middle cloud deck (the stippled area of Fig. 3) may have led to an exaggeration of this temperature contrast by producing evaporative cooling of the air beneath the cloud layer and to anom- alous (wet bulb) cooling of the vortex temperature sensor; nonetheless, the large-scale temperature gradient is quite real. This warming is evident upon inspection of Fig. 5 in comparing the temperatures at points B or C which lie within the hazy air with those at point A or on the Barbados sounding which represent observations made outside the hazy regime. Fig. 3 shows that haze was prevalent east of the wave axis and north of the zone of cloudiness located between i i i i RADON -222 DISTRIBUTION « JOT.T 14 .1969 s,^ \" ( 5-.2ei6K — 4K1 I 2lVl UOK — 6K) 6/,V % 21\34 \ 6> ALTITUDE (FT) — — ■ 1 3 K TRANSITION • IOK _1_ 45* Fig. 4. Radon-222 concentrations measured along the DC-6 aircraft track shown in Fig. 3. The figures to the right (outside) of the track are the radon concentrations in picocuries per cubic meter and the numbers on the left (inside) of the track are the filter color index values which are roughly equivalent to the dust load expressed in micrograms per cubic meter of air. The duration of individual measurements ranged from 5 to 15 min (15 to 50 mi). Changes in aircraft altitude from the 1300- or 10,000-ft levels are indicated by the dotted lines and labelled accordingly. 183 March 1972 T. N. CARLSON AND J. M . PROSPERO 287 5 and ION which we can identify with the Intertropical Convergence Zone (ITC). However, no haze was ob- served by aircraft west of the disturbance. Haze was also visible at 1300 ft east of the disturbance (Fig. 4), but was considerably thinner than that reported at 10,000 ft. The layered character of the air in this region is revealed most strikingly by the extremely rapid decrease in the values of radon concentration and filter color density during the rapid descent of the aircraft from 10,000 to 1300 ft near 20N, 49W. The rapid increase in potential temperature (and decrease in mixing ratio) that occurred near 55W at about 1500 GMT and the subsequent gradual increase to a value of 40.5 C at the temperature ridge (see the respective segments AB and BC in Fig. 5) coincided with in- creases in radon concentration and filter index. An isentropic layer is visible on the sounding in Fig. 5 from C to D, representing the lower portion of the Saharan mixing layer penetrated by the aircraft as it descended from C to E and thence into the stratocumulus layer. The low radon concentrations and low dust loads (i.e., low filter color index) measured during the latter part of the flight clearly indicate that the air was of non- Saharan origin. The full depth of the Saharan air layer is shown on the sounding made at the Discoverer on the following day under hazy and suppressed conditions (Fig. 6). Saharan air, evidenced by the steep lapse rate of temperature and increasing relative humidity with height, is recognizable between 820 and 580 mb. Although the lapse rate is somewhat more stable than the dry adiabatic in the Saharan layer, the sounding is similar to the temperature profile obtained at Sal, Cape Verde Islands, shortly after the same wave disturbance had passed that island Fig. 5. Temperature and dewpoint soundings on 14 July 1969, as made by aircraft (heavy solid and dashed lines) and by the Barbados radiosonde (thin solid and dashed lines). The segments ABC (temperature) and A'B'C (dewpoint or mixing ratio) were recorded when the plane flew horizontally between 1500 and 1900 GMT. The most rapid warming was between A and B (see text) during the first half hour. The segments CDE and C'D'E' were recorded during descent at approximately 1930-2000 GMT. Saharan air is indicated between 720 and 790 mb on the aircraft sounding. The Barbados sounding was made west of the wave axis, while the aircraft record was obtained east of the wave axis. Fig. 6. Temperature and dewpoint sounding from the Discoverer and the temperature sounding for Sal, Cape Verde Islands. The vertical wind profile for the Discoverer (beaufort scale) is shown at the right. Saharan air is located on both soundings approxi- mately between 820 and 600 mb. There is a wind maximum of 46 kt near 670 mb at the Discoverer. some three days prior to entering the BOMEX area. In contrast, the Barbados sounding made ahead of the wave axis at 1200 GMT 14 July (Fig. 5) fails to exhibit any deep isentropic layer and may be devoid of Saharan air except for a possible thin, cool segment located between 700 and 760 mb. Moreover, the relative dryness at Barbados precludes the explanation that Saharan air had passed through the disturbance; if this had occurred, the air parcel would have had a much higher mixing ratio because of the effects of cumulus convection and rainfall. At Barbados, the dust concentration at the surface increased by a factor of 3 prom 3.5 to 10.5 pg m-3 (see Fig. 1)] following the passage of the wave disturbance. However, the increase was not as dramatic as that ob- served on the flight at 700 mb where, on the basis of the filter index, the dust load increased by a factor of about 27. Exceptionally strong winds having a jet maximum of ~45 kt near 650 mb were observed within the Saharan air layer; these winds are evident in the Discover sounding shown in Fig. 6. High winds were confined to the east of the disturbance where speeds in excess of 40 kt were reported by Navy aircraft at the 10,000-ft level (Fig. 3). Following the passage of the disturbance, the wind speeds at upper levels over the Antilles in- creased markedly, and the vertical wind profiles there came to resemble the one shown in Fig. 6. It seems quite likely, therefore, that the marked temperature contrast between the warm Saharan air and the rela- tively cooler tropical surroundings is somehow respon- sible for the strong winds inasmuch as the geostrophic thermal wind equation would imply strong vertical wind shear in the vicinity of the large-scale temperature contrast. Because of the extraordinary sharpness of the discontinuity, however, a particular point located within the zone of "sudden warming" shown in Fig. 4 may 184 288 JOURNAL OF APPLIED METEOROLOGY VoLUMli 11 -30"C -20 Lie. 7. Aircraft temperature and dewpoint soundings made by the 39C DC-6 (thin lines) on continuous ascent east of Barbados, W.I., and by the 40C aircraft (heavy lines) ascending to point A.V, flying horizontally through the dust front to BB', ascending behind the dust front to CC, and later descending to DD' and EE'. Saharan air is recognizable between 670 and 820 mb on the 40C sounding. not be experiencing even approximate geostrophic equilibrium. It is important to note here that the anomalous warmth of the Saharan layer is greatest at the base of the isentropic segment and may vanish above about 650 mb, where the layer becomes a little cooler than normal. As will be shown in subsequent examples, the wind speed also tends to diminish very rapidly with height above the 600-700 mb wind maximum. b. Case 2: 18 July 1969 In another episode during BOM EX, an aircraft flew into a large mass of hazy air to the rear of an African disturbance on 18 July 1969. The pair of soundings in Fig. 7 and the profiles of temperature, mixing ratio, wind velocity, radon concentration, and filter color index in Fig. 8 illustrate the exceedingly sharp definition of the interface between the Saharan air stream and its surroundings. In order to emphasize the steepness of this transition, we will hereafter refer to the leading edge of the Saharan dust plume as the "dust front" and the small-scale temperature-dewpoint contrast between air masses as the "Saharan air front"; the use of the term "front" is not meant to suggest a mid-latitude type of feature, however. The RFF DC-6 aircraft, 39C, ascended to 520 mb in the vicinity of 14N, 58W and proceeded eastward at this level. Throughout the ascent and thereafter, the radon concentrations remained relatively low (10-15 pCi m-3) and the filters were essentially clean. The temperature sounding produced during the ascent phase of this flight, shown in Fig. 7, is very close to that of the mean tropical sounding for July, although the humidity at middle levels was somewhat higher than the average indicating that the air may have interacted recently with cumulus convection. A cloud mass associ- ated with a travelling disburbance of African origin was located near the Lesser Antilles but the weather over Barbados and further east was undisturbed except for numerous trade wind cumuli and an occasional towering cumulus. Another DC-6 aircraft, the RFF 40C, flying below the 39C aircraft; proceeded eastward at 720 mb and en- countered the dust front near 56W (Fig. 8) ; within a few minutes (over a distance of ~20 mi) the potential temperature rose by 4C from 36 to 40C and the mixing ratio decreased by over 7 gm kg-1 to 3.6 gm kg-1, values which were maintained for at least several hun- dred miles inside the dusty air. In crossing the dust front, the radon concentrations rose by a factor of 3 to 30 pCi m~3 and the filter color index increased full scale to 81 (Fig. 8), a value attained on only a few occasions during BOMEX. During this time, the cloud cover became suppressed and the haze grew quite dense. In Fig. 7 the temperature and dewpoint contrast ex- perienced by the aircraft 40C in crossing the Saharan air front is indicated on the sounding by the segments AB and A'B', respectively. (The arrows indicate the chrono- logical sequence of events from the aircraft's initial ascent to 720 mb and later to its maximum altitude at 670 mb.) Aircraft 39C, flying at 520 mb, visually fixed the location of the Saharan air front at 56W but reported that the haze top was well below the flight level. Accordingly, there was no indication of Saharan air either in the radon concentrations or in the filter color index on this flight. Aircraft 40C, however, encountered a portion of the isentropic (and constant mixing ratio) layer between 720 and 670 mb [^represented by the segments BC (temperature) and B'C (mixing ratio) in Fig. 7J. Later, the aircraft descended to 820 mb near 51W and in so doing was able to measure the lower portion of the Saharan mixing layer given by the seg- ments BD and B'D' in Fig. 7. The aircraft continued to descend to 985 mb, obtaining the temperature and dewpoint profiles DE and D'E', respectively. The sound- ing from aircraft 40C shows the presence of a deep layer 53'W 5Z*W St"W LONGITUDE Fig. 8. Temperature, mixing ratio, wind velocity and aircraft altitude measured bv the 40C DC-6 aircraft on an eastward traverse, from 1100-1400 GMT 18 July 1969, through the Saharan air front at approximately 14N. The radon concentration (in picocuries per cubic meter) and the filter color index are listed at the top for the appropriate measurement intervals. 185 March 1072 T. N. CARLSON AND J. M. PROSPERO 289 of Saharan air having an almost, constant potential temperature (~40C) and mixing ratio (~3.6 gm kg-1). respectively, as shown by the segments CD and C'D' in Fig. 7. Except for a temporary reduction in filter color index and radon concentrations near 55W, the Saharan-air indicators remained uniformly high until about 53W, after which the filter color index de- creased steadily to a value of 5 at the easternmost end of the flight track (Fig. 8). After 51W the plane followed a straight line track southwestward to a point near 9N, 54W. Radon concentrations decreased abruptly near 850 mb during the descent but continued to be relatively high (20-24 pCi m-3) until the aircraft reached ION. The filter color index increased gradually from 5 to about 27 as the aircraft proceeded southwest- ward; at ION the dense haze vanished and the radon concentrations and filter color index decreased to low values comparable to those recorded at the beginning of the flight, indicating that the aircraft had emerged from the southern edge of the Saharan dust plume. It should be noted also that no haze had been reported in the ITC south of about ION on 14 July (see Fig. 3). (However, on 23 July, a case not discussed in this paper, Saharan air was observed well within the region of ITC cloudiness between 5 and ION.) Wind velocities also changed abruptly at the dust front (see Fig. 8), increasing in speed to about 40 kt and backing slightly to southeasterly within the plume. The winds reached a maximum near the dust front and thereafter gradually decreased further toward the east. In contrast, during the 14 July case the highest speeds occurred a few hundred miles behind the Saharan air front. 4. Large scale movement of the Saharan dust plume a. Over the eastern Atlantic During the morning of 30 July 1970, while cruising near the Cape Verde Islands, the Discoverer suddenly -30'C -20 Fig. 9. Soundings (1130 and 2300 GMT, 30 July 1970) made from the Discoverer near the Cape Verde Islands. The wind profile for the earlier sounding is listed at the right showing the presence of a wind speed maximum between 600 and 700 mb. A shallow laver of stratocumulus is indicated below 900 mb. Fig. 10. Sounding and wind profile for Dakar, Senegal (1200 GMT 5 July 1970), showing a layer of Saharan air between 540 and 820 mb with an almost uniform lapse rate of temperature and mixing ratio. Note in the wind profile the counterflow of maritime air along the coast below the Saharan air layer. encountered dense haze after a period of a day or so with little haziness. No Saharan air had been visible on the Discoverer sounding made 12 hr earlier, but the deep isentropic layer evidently appeared sometime prior to the 1130 GMT sounding where it is clearly recognizable between 540 and 780 mb (Fig. 9). During the following 12 hr the isentropic layer deepened by about 50 mb, the base remaining just above a very thin layer of stratocumulus. A wind speed of 40 kt is also visible between 600 and 700 mb. The temperature and dewpoint stratification visible in the soundings in Fig. 9 is commonly observed off the coast of Africa at this time of year. The lower (cooler and moister) portion of the sounding is attributable to the northerly maritime flow along the coast. These winds have a persistent landward component on the western coast of North Africa. Consequently, the Saharan air is undercut by the cool maritime current well inland from the coast as is evident from the Dakar sounding in Fig. 10. The arrival of this dusty air aloft is reflected in the deck-level measurements of radon concentration, atmo- spheric dust loading, and turbidity, all of which show an increase from the 29th to the 31st; except for radon, uniformly high values were maintained until the ship entered port three days later on 3 August (Fig. 11). The turbidity B, measured with a Volz photometer at 0.5 n wavelength (Flowers et al., 1969; Volz, 1970), rose to 0.28 on 31 July; this value is comparable to that measured in urban regions under conditions of dense haze usually attributable to human activity (Flowers et al, 1969; Volz, 1969). The onset of haziness at the Discoverer occurred one or two days after a large-amplitude African disturbance passed the vessel; the passage of the disturbance seemed to be associated with a narrow arc-like band of strato- cumulus which on satellite photographs appeared to be radiating from the African coast and which was first recognized shortly after the disturbance (shown at 186 290 JOURNAL OF APPLIED METEOROLOGY Volume 11 Olt ■ — to OIB 'O I B t I 9BR I MttT 1 DUSTY I HKT I HAZY ii UlLiii JULY tt I JULY 90 I JULY SI I AIM I I MM t MN,tOW ITM, tSm ITN, 23W MM, M* l*H. WW Fig. 1 1 . Dust and African air-indicator measurements from the Discoverer, 29 July-2 August 1970, showing dust load 0*g m-*), B (turbidity) and Rn r_radon-222 concentration (pCi m-*)]. The position of the data bars indicates the time of measurement; the width of the dust load data bar shows sample duration. Radon-222 samples are 30 min long. The vessel entered port on 3 August and measurements were terminated about 40W in Fig. 12) crossed the African coastline2 late on the 26th. The leading edge of this arc, delineated by the serrated border in Fig. 12, appears to form a discontinuity between the large-sized stratocumulus cells to the west and the smooth, small-celled strato- cumulus and clearing weather to the east.3 The winds at the ship gave no evidence of the passage of a second, but weak, low-latitude wave which passed south of the ship on the 30th. Instead, the winds strengthened at middle levels and backed gradually to the northeast by 2 August, immediately prior to the passage of the next major disturbance (shown at 12W in Fig. 12). As this next major wave approached the Discoverer, the haze and the associated indicators decreased and, with the exception of radon, ultimately attained low values again on 4 August. In contrast, the radon con- centrations increased markedly with the passage of the cloud arc which signaled the arrival of Saharan air aloft on the 30th; the radon then decreased abruptly about the time the skies cleared on 31 July and fluctu- ated thereafter. We have frequently observed these somewhat erratic fluctuations in radon concentration in surface and low- level air and interpret them as being indicative of the extent of mixing which has taken place between surface air and the air within the Saharan layer above. Nor- mally, the highly suppressed conditions beneath the Saharan air impede the exchange of air between these layers; however, in areas of convective cloudiness this exchange is facilitated and could lead to the observed abrupt and erratic fluctuations. On occasion, evidence of the effects of convection is seen in soundings from coastal stations of West Africa. For example the Dakar sounding data plotted at 15N, 17 W in Fig. 12 reveal the presence of a Saharan layer, but one which is relatively cool and moist indicating that cumulus convection may 1 For a description of how the disturbances are tracked, see Carlson (1969a). * When Saharan air overrides the maritime cloud cover, the latter becomes suppressed and, in some cases, vanishes entirely. Fresh pulses of Saharan air, therefore, may be accompanied by a sufficiently marked discontinuity in the cloud pattern to become visible on satellite photographs. It is this discontinuity which we have identified as the leading edge of the Saharan dust plume and which is delineated by the serrated border in Figs. 12-14. have partially modified the Saharan air at some earlier time. In general, deck-level radon measurements made off the coast of Africa during the past several years show concentrations that are comparable to the values found for surface air at Barbados and considerably less than those measured aloft within the Saharan air layer over the Caribbean. The same conclusions hold true for our dust measurements. These observations suggest that the Saharan layer is the principal source for radon and for aerosol particles found in the lower layers and that the dust concentration in the lower layers is determined principally by the settling rate of the particles and the degree of mixing between the low level air and air in the Saharan layer. In a sense, the Saharan air layer can be thought of as functioning as a somewhat leaky "duct" transporting dust to the western Atlantic. In another episode, intense haze was reported over much of the Caribbean and western equatorial Atlantic during the period 9-14 July 1970; the appearance of haze correlated with the arrival of a dust pulse which could be tracked back to the coast of Africa and there jm jo* to* 10* ■k-BbtV - F ^r 4^P «•» b **•/ -*nB |Ml/ • f/xJKl B ; JEftfe*<^&a&$9H ■ '.♦ ¥&m U; •*"•*?£ '^■w*- ^T^^»■ilfl5'ffV--*^'••-'--'* to* *■ AStfl fck ' 1 "£\ N.SF ml—** fcw .'$@^&?li to* - •f «** to* Fig. 12. ATS-III satellite photograph of the eastern equatorial Atlantic at 1600 GMT 30 July 1970. Superimposed on the photo are sounding data and surface streamlines for 1200 GMT. Each radiosonde station symbol contains a decimal figure which is the 700-mb potential temperature in degrees Celsius but with the tens digit deleted (i.e., 39.0»»9.0 and 43.2»3.2). The numbers at the top and bottom and the decimal number at the left refer, respec- tively, to the heights of the top and base of the Saharan air in centibars and the average mixing ratio in grams per kilogram. These three values are omitted when Saharan air cannot be found at the station. The wind barbs show direction and speed in the Saharan air at the level of maximum wind speed (or between 850 and 500 mb in the absence of Saharan air). Haze, dust and cloud observations are also included where reported. The hori- zontal visibility in miles (the number prefixed by the letter V) is also listed at surface locations reporting a visibility less than 10 mi. The axes of African wave disturbances are indicated by the heavy dash-dotted lines. The serrated border signifies the leading edge of the Saharan air. The data plotted at 17N, 24W refer to the 1200 GMT Discoverer sounding of Fig. 9. 187 March 1972 T. N. CARLSON AND J M P R O S P E R O 2<)\ Fig. 13. ATS-III satellite photographs of the equatorial At- lantic for 3 (a), 5 (b) and 6 (c) July 1970. The same display code is used as that in Fig. 12. In 13b the serrated edge marks the leading edge of the cloud discontinuity and the dotted line indicates the edge of the clear area, while 13c shows the larger view of the advancing cloud discontinuity. MOVEMENT OF WAVE AXIS. DUST FBUI.M1M CEITE1 .ABO AXIS OF MID-IEVEL JET JWVt tgl -13th . 1STO 5« tM 111 win cimiM ®>40i* 9**<>' fl'40.0* 9*41.0 D uvc »xi» J QUIT PULSE ««i« or ■ID-LCVCL JIT Fig. 14. Continuity chart showing the progress of the axis of an African wave disturbance and the leading edge of the dust plume over the period 1-13 July 1970. The positions of the easterly wind maximum in the Saharan air and the centers of maximum po- tential temperature are shown for the period 10-14 July. The location of the leading edge of the dust plume during the period 8-13 July was determined on the basis of where Saharan air could be found on the temperature soundings. associated with a front-like band of stratocumulus similar to the one shown in Fig. 12. Figs. 13 a-c show that this feature moved away from the African coast, following closely behind a large African disturbance which had crossed the coastline on 2 July. The leading edge of this discontinuity first became distinctly visible along the coast of Africa on 3 July, although there was some evidence that it may have emerged late on 2 July. As in the 30 July example, the highest potential temperatures found in the Saharan isentropic layer on 4 July over the African coast were about 43-44C and the top of the layer was at a height of about 540 mb. By 5-f3 July, however, the potential temperature of the Saharan air leaving the continent had increased to 46-47C and the top had risen slightly to a height of 500-520 mb (Figs. 13 b,c). Soundings along the African coast (such as Dakar or Sal) continued to show Saharan air aloft during this period without any definable inter- ruption in the airstream. Since there was no evidence for the passage of a Saharan air front along the African coast, the leading edge of the Saharan air may have beccme organized at some distance to the west.4 In this case, the earliest direct observation of the Saharan air front was made at 49W on 7 July by an Air Force reconnaissance aircraft flying across the band of strato- cumulus cloud which had continued to move westward behind the disturbance. Flying at a height of 660 mb along the track ABCD shown in the continuity chart (Fig. 14), the aircraft reported heavy cumulus activity at point A (near the axis of the disturbance) which cleared rapidly at B and was replaced between B and C by small cumulus and some upper layer cloud. At C, clear air turbulance was experienced followed by the 4 The inability to discern the passage of the Saharan air pulse at Sal may be due to the infrequency with which the sounding from Sal is received. 188 292 JOURNAL OK APPLIED METEOROLOGY Vnl.l'MK 11 093 t/> 23 .3 CD « 11 C 10 a: s 0 I s'hWE* 1 '"^VCl^UOY. SOW C.RRUS SHOWIM or Ul CC COMBINED AVE 1 2> 25 15 > o in : II ll II 1 1 1 Mil Q a i I 1 If 1 111 III! 1 1 , II H (/> 7/10 7/H 7/12 7/13 33N, 79W 33N.69W 3IN.MW SIN, G2W 1 ' ,iu SON, 39 W zbn.sow oas — too Q < n O -1 » CO K 3 n OVERCAST RAIN I MAZY IVtRYMAZY [EXTREMELY | CLOUOY J OVERCAST I CLOUOY CLOUOY UJ a z h- (/> UJ 5 O Q --JMmBammmmi*J§Sfflm. i ^^ wMmmm 09 3 CD Z li II ll 1 i < CD CC < 19 X 1 JlH II II ll 1 1 111 1 1 II III J T/e 7/9 7/10 7/11 7/12 7/13 7/14 7/15 OSS OVERCAST OVERCAST CLOUOY \* WEEKEND 4*MAZE8T MIN CIRRUS < Q cr O r _| so K^ I I 1 r I u. 3 04 1 ll > (06') I < 2 CD « H III 16 c 10 l|l .. In .„ Ill II! r/9 ' 1 7/9 1 1 1 | 1 7/10 7 DATE '"( | 1 1 1 | T 1 7/tZ V 1970) 13 m* r ■, 7/18 Fig. 15. Dust and African air-indicator measurements: ship Discoverer, Barbados, W. I., and Miami, Fla., 8-15 July 1970 (for coordinates of ship, see figure). Units are the same as in Fig. 11. The dust load data for Discoverer on the 11th and 12th are a composited average. The turbidity data for Miami on the 14th are questionable because of the possible effect of thin cirrus which may have been present. appearance of a haze layer and a 2C rise in temperature between C and D ; the haze top was reported to be at 640 nib at D. movement of the Saharan air over the Caribbean from 10-13 July. In these Caribbean analyses, the phase speed of the easterly wave, 16 kt westward, is subtracted from the wind vectors in order to depict the air flow relative to the moving dust pulse and wave axis. In contrast to the examples shown for the eastern Atlantic- analyses, where the data were not abundant, the western Atlantic analyses enable us to define the exact boundary of the dusty air mass and to chart the trajectory of the air travelling in the core of high winds within the Saharan air layer. Figs. 17 a-d clearly show the follow- ing features: 1) the leading edge of the Saharan air, 2) an area of relatively warm and deep Saharan air located some distance to the east of the leading edge of the pulse, and 3) a band of strong winds located between the warm center and the forward boundary of the pulse. (Note that the correction of the winds for relative movement reduces the speed of the absolute easterly- wind maxima and increases the strength of the rela- tive westerlies.) The wind barbs suggest an anticyclonic rotation about a center a little to the east of the warm dome.5 Surface observations of haze accompanied by low horizontal visibility (^10 mi) are infrequent in this region; however, during this 4-day period, one or two such reports were received daily and each was located near the center of maximum potential tempera- ture (noted in the filled circles in Figs. 17 a-d). The Discoverer briefly encountered the northern fringe of the dust plume on 12-13 July at about 31N, 63-64W, as indicated by the presence of Saharan air aloft on the soundings. Although the turbidity did in- crease moderately, haze was not reported at the ship. Also, the surface air dust load remained low and the radon concentration failed to rise above the oceanic "background" level (Fig. 15). We can only conclude b. In the Caribbean The above-mentioned disturbance passed through the Barbados area late on 8 July and was accompanied by a period of showery weather conditions which ended on 9 July. (The transit time from Africa to Barbados of about 6 days is typical.) The surface air dust loading at that island rose abruptly from virtually zero on the 7th to 5-9 Aig n-r3 on the 9th and 10th (Fig. 15) and 15-19 jug m~3 on the 11th and 12th with a brief period of exceptionally dusty conditions on the 11th when the dust load rose to 78 ng m~3 (see Figs. 11 and 15). Turbidity values also rose accordingly, approaching 0.3 on the 11th. Radon concentrations, however, actually decreased during this period, as would be anticipated by the advent of stable conditions at the base of a deep layer of Saharan air. The presence of a very deep and warm layer of Saharan air over Barbados on the 10th and 11th is evident from the sounding shown in Fig. 16. A characteristic wind maximum is evident near 700 mb on the sounding for that period (Fig. 16). The analyses in Figs. 17 a-d show the structure and Fig. 16. Soundings made at Barbados, 10-11 July 1970. The vertical wind profile for an intermediate time is shown at the right. 6 The anticyclonic rotation of the dust plume advancing west- ward behind a travelling disturbance is clearly depicted in a film loop made from a sequence of ATS-III satellite pictures for 1 July 1969. This incident is discussed in greater detail in a sub- sequent section of this paper. Makui 1°72 T. X. CARLSON AND J. M. PKOSPKUO _>»)3 f K _\ \ \ *. v. « IL^: \JUir it. t*ro J^ t~!J Pig. 17. ATS-III satellite photographs of the Caribbean, 10-13 July 1970 (a-d, respectively), upon which are superimposed the 1200 GMT sounding data and isotherms of the 700-mb potential temperatures (solid contours). The coding of the 700-mb potential temperature, mean mixing ratio in the Saharan air, andjheight of the base and top of the Saharan air is identical to that in Pig. 12. The winds, however, are taken relative to the moving system (16 kt westward) and are plotted only where there was a definable wind maximum in the 850-500 mb layer; the How, therefore, refers to the level of the middle-level (Saharan air) jet. Square station symbols signify that 12-hr off-time data have been plotted 3.2° of longitude east (or west) of the station to account for the dis- placement of the system. Trapezoidal station symbols indicate aircraft observations. The position of the dust front (the serrated border) separates Saharan from non-Saharan air (see l'ig. 14). The easterly wave axis is also indicated by the dash-dotted line. The small filled circles signify ship or land observations where haze was reported in the presence of horizontal visibility less than 10 mi. that there was no significant mixing between the air within the Saharan layer and the low-level maritime air which had not recently passed over a land mass. In contrast, at Miami, the ground-level atmospheric dust loading and turbidity rose dramatically (to 78 /ig m-3 and over 0.3, respectively) and the radon con- centration tripled following the arrival of the dust plume and the onset of heavy haze on the 13th. On ihe afternoon of the 13th, the RFF aircraft 39C, flying on a cloud seeding mission for the Experimental Meteorology Laboratory, NOAA, encountered suppressed conditions over the Florida Everglades. Observers reported an unusually dense red-brown haze which had a distinct top near 580 mb. The sounding produced by the air- craft's ascent to the 480 nib is shown in Fig. 18; ihe lop of ihe Saharan air layer is dearly discernible at 580 mb, in agreement with the reported height of ihe Pic IS. Sounding made by IX '-6 aircraft ascending jusl west of Miami and a portion of the dewpoinl sounding from the Miami radiosonde, 13 July 1970. 190 294 JOURNAL OF APPLIED METEOROLOGY Volume 11 Fig. 19. Movement of wave axis and dust pulse, 26 June-4 July 1969. Positions of the leading edge of the dust pulse were deter- mined from ATS-III satellite photographs; the placement of the wave axis was determined from the position of the apex of the "inverted V" on the photograph (see text). Extensive cloud cover west of the Antilles made it difficult to follow the progress of the dust plume after 3 Jul)'. haze top. An isentropic layer (0~ 40C) is visible between 580 and 760 mb, below which the clouds are obviously being suppressed by the warm base of the layer. It should be pointed out, in the various aircraft soundings we examined (e.g., Figs. 5, 7 and 18), that the mixing ratio was found to be almost unchanging with height throughout the Saharan isentropic layer; this was not always the case with the conventional sound- ings (Figs. 6, 9, 16 and 18, for example). This dis- crepancy is probably attributable to the radiosonde hygristor which, because of its poor response time (Morressey and Brousaides, 1970), is incapable of recording the sharp transition between the moist mixed layer and the dry conditions above the base of the Saharan air. In addition, the radiosonde hygristor may seriously underestimate the true moisture content of the layer once it has come into equilibrium with the new dry environment above the inversion (Morressey and Brousaides, 1970). c. Further evidence for the large-scale transport of African dust Generally, parcels of dust-laden air over the ocean are visible in satellite photographs only during the first few days of the transit across the Atlantic. By the time the large-scale pulse of dusty air has entered the western Atlantic, estimates of position and areal ex- tent based on photographs are somewhat ambiguous and must be verified by soundings showing the presence of Saharan air aloft ; diagrams such as Fig. 14 showing the progress of dust fronts are composited in this manner. An exception is the dramatic series of ATS-III satellite photos which show the progress of a dust pulse from the time it leaves the coast of Africa on 28 June 1969 until it reaches the Leeward Islands on 4 July.6 The positions shown in Fig. 19 delineate the forward edge of the areas of greyish "smudge" which we find to be 6 See Footnote 5. characteristic in photos of regions where there is a dense dust haze. Although only the dust front is indicated in the figure, the photos clearly show that the entire equatorial Atlantic from 10-25N is covered with dust for several days during early July. Unfortunately, the disturbance entered the BOMEX area during a rest interval between the third and fourth periods; conse- quently, there are no airborne dust measurements available. On 3 July the dust load at Barbados, situ- ated at the southernmost edge of the plume, increased sharply (Fig. 1) and remained relatively high for several days (except for the 5th when rainfall temporarily reduced the dust concentration). The position of the forward edge of the plume became diffuse on the satellite photographs after 4 July because of increased cloud cover west of 65W. However, high concentrations of dust were measured at Puerto Rico on 5 and 6 July (Volz, 1970) following the passage of this disturbance. Perhaps the strongest evidence supporting our belief in a common African origin for the airborne dust col- lected during episodes such as these is that the mineral- ogy of the dust collected in hazy areas is uniform re- gardless of the location of the sampling site. In the event of July 1970 (discussed above) and during much of the summer, the x-ray diffraction spectra of the aerosol samples collected at Miami (26N, 80W), Barbados (13N, 59W) and aboard the Discoverer are identical. In contrast the aerosol samples collected in Miami during the days prior to the arrival of an African air parcel consist almost entirely of calcite, as might be expected since there are large areas of exposed coral in this region and the soils are highly calcareous (Prospero and Carlson, 1971). The similarity in the composition of dust collected at Barbados and that collected aboard vessels off the coast of Africa during the summer months is a monotonously persistent phe- nomenon which we have noted from the inception of our shipboard aerosol sampling program during the summer of 1967 until the present; thus, there can be little question about the origin of this material. 5. A model of the African dust plume A schematic flow pattern showing the movement of the African dust plume during the summer months is presented in Fig. 20. The model is a conceptual com- posite based on our general knowledge of the African circulation patterns and on a number of detailed studies of individual dust episodes such as the ones described in this paper. The following features are incorporated in this model. 1) The low-latitude troughs identified as the axes of African wave disturbances which are often visible in the cloud pattern as an "inverted V" shape. 2) The dust front situated behind the trough, in the relatively cloud-free area. 3) The large-scale anticyclonic eddies behind the dust front at 600-700 mb. 191 Makch 1972 T. N. CARLSON AND J. M. PROSPERO 295 4) The low-level coastal How of dust-free air from northern latitudes. (The portion of this flow having an overland trajectory may pick up small but significant quantities of dust.) 5) Dust-free low-level air from the southern latitudes sweeping inland and also into the trough. 6) Upper level winds from the Sahara overriding the low-level winds along the coast. The figure is not intended to convey the suggestion that African disturbances actually generate the dust pulses by bringing about an increased lifting of dust. However, it is highly probable that the alternate weakening and strengthening of the pressure gradient over Africa, caused by the cyclic passage of the dis- turbances, contributes to the cyclic variation in low- level wind speeds which, in turn, may produce a cyclic fluctuation in dustiness. Indeed, we have accumulated some statistical evidence which shows a weak correlation between the frequency of haze or dust observations over West Africa and the passage of a travelling disturbance. However, the weakness of this correlation may be due to the fact that in many instances, such as the ones described in this paper (see Fig. 13, for example), the surface observations fail to show an)' systematic variations in the day-to-day occurrence of dust, haze or visibility along the African coast. It is also possible that the rainfall associated with the disturbance contributes to the fluctuating character of the dust loads by washing a large fraction of the dust from the air entering the disturbance, the Saharan air itself being modified by the cumulus mixing in these showers. In Fig. 17 a-d, for example, some to the tra- jectories of dusty air are shown to be passing close to the dense cloud cover to the east of the disturbance axis. Furthermore, although the Saharan air may serve as a strong suppressive influence on the convection, the intrusion of the airstream into an existing area of active cumulonimbus may result in a penetration of the Saharan inversion by the rising towers. We believe that the pulsating nature of the Saharan airstream is associated with the periodic interruption of this flow due to the passage of travelling disturbances from Africa. This situation is depicted in Fig. 20 by the streamlines which show a split in the flow of heated Saharan air which is continuously leaving the continent and being exported to middle latitudes over Europe (Carlson and Ludlam, 1968) and toward the Caribbean. With the passage of a disturbance over the African coast, an amount of moist, dust-free oceanic air is advected northward behind the wave axis; consequently, the stream of Saharan air, which would otherwise flow westward in the tropics, is diverted to join the north- ward-moving airstream along the African coast.7 Con- versely, the air immediately ahead of the wave axis in Fig. 20 would be drawn from the northern Atlantic 7 The presence of Saharan air containing high dust concentra- tions has been observed over Northern Europe (Stevenson, 1969;. TRAJECTORIES OF DUST-LADEN AIR NEAR 700BB TRAJECTORIES OF DUST-FREE AIR NEAR 700MB — TRAJECTORIES OF DUST-LADEN AIR NEAR SBOM6 TRAJECTORIES OF DUST- FREE AIR NEAR (BOMB. LOW- LEVEL AIR FLOW ALONO AFRICAN COAST f»«LSiiSS DISTURBED WEATHER Fig. 20. Schematic model of air motions accompanying the movement of African disturbances and the associated dust pulses from Africa. (either directly or on a relatively short overland route) rather than from the interior of the Sahara and, there- fore, would be relatively low in dust and not as warm as the primary stream of Saharan air.8 Consequently, there is a strong temperature contrast established in the 600- 800 mb layer between the wave axis and the Saharan air on either side of the disturbance. In accordance with this temperature gradient the thermal wind relationship for geostrophic flow implies the existence of a warm anticyclonic cell located between the disturbances at middle levels and a jet-like character of the wind in the region of strongest temperature contrast. Indeed, on an even larger scale, the clockwise turning of the Saharan air toward the north along the entire west coast of Africa is associated with the warm Saharan anti- cyclone and a semi-permanent upper trough which lies along the coast of the Euro-African land mass in summer and which extends to quite low latitudes. Following the passage of the disturbance off the African Coast and the return of easterly flow, the westward advance of Saharan air is once again resumed at low latitudes. 6. Radiational cooling and vertical motions in the dust layer Examination of the Caribbean analyses in Figs. 17 a-d shows that the maximum potential temperatures in the Saharan air progressively decrease from 10 to 13 July at a rate of almost 1C day-1; this decrease is equivalent to a daily cooling rate of about 0.7C as measured in units of conventional, not potential, temperature. A net potential temperature change of 8 African disturbances are known to be most intense in the vicinity of the African coast. The dust pulse, however, would remain despite later weakening of the disturbance. 192 296 JOURNAL OF APPI. I E I) METEOROLOGY Volume 1 1 Fig. 21. Soundings for Pt. Etienne (Pt. Nouadhibou), Mauri- tania, 2 July 1970, and San Juan, Puerto Rico, 11 July 1970. The vertical wind profile for San Juan is shown at the right. 3C is commonly observed when African soundings are compared with those made in the western Atlantic, implying a mean cooling rate of 0.5-0. 7 C day-1 (see, for example, Figs. 6 or 21). In most of the events we have studied, Saharan air leaves Africa in July with a mean potential temperature of 43^44C, a mixing ratio averaging between 2 and 4 gm kg-1, and a top at 540-560 mb. The air parcel arrives over the eastern Caribbean with a similar moisture content but with a potential temperature of about 40C (as in the two BOMEX cases described) and a top near 600 mb. The implied sinking motion of the Saharan air top is therefore about 50 mb in 4-5 days for a mean rate of 1 mm sec-1. In the previously described case during July 1970, when an unusually warm and deep outbreak of Saharan air occurred, the warm center (0~43C) was situated near the Antilles on the 10th (see Fig. 17). Assuming that this air had departed Africa on 5 or 6 July with a potential temperature of 46-47C, its arrival over the Antilles on the 10th with a potential temperature of 43C would be compatible with a mean cooling rate of 0.7C day-1 provided that the trajectory speed9 was about 30 kt. The implied sinking motion in this case was about 80-100 mb. Cox (1969), London (1952), and others10 have found that the normal infrared cooling of the middle tropical troposphere is about 2C day-1 under average weather conditions and perhaps a little less (~1.6C) under clear skies. London's (1952) results indicate that the net solar absorption at middle levels in the tropics is about 0.65-0.8C day-1 under widely varying cloud conditions. Therefore, the net cooling (net longwave radiation loss minus shortwave absorption) is likely to be 0.9-1. 2C day-1, a little larger than the 0.7C estimate which we 9 It is important to note that the trajectory speed of the Saharan air parcel can be substantially greater than the forward motion of the leading edge of the dust pulse as suggested in Fig. 20. 10 "Summary of findings of the Workshop on BOMEX Radia- tion and Particulate Investigations, 21-23 October 1970," Memo to participants. obtain by simply observing the temperature change in the African dust plume during its oceanic transit. Anthes (1970), using Sasamori's (1968) longwave (CO2 and water vapor) radiation model, has calculated in- frared cooling rates of about 2C day-1 for the mean sounding in the tropics and somewhat lower values for drier conditions. Using Anthes' program, we have made longwave radiation calculations based on several sound- ings such as the one shown in Fig. 21; our results show that for mixing ratios of 2-3 gm kg-1 in the isentropic layer, the net cooling was about 2.5C day-1 at the base of the Saharan layer and about 2C day-1 at the upper inversion, but was only about 1.6C day-1 averaged over the isentropic layer itself. Since the moisture distri- bution is so critical in these calculations it remains to be seen whether the peculiar temperature and humidity dis- tribution in the Saharan air is more important than the actual aerosol content of the atmosphere in accounting for discrepancies between observation and measurement which may have been noted in regions of haze.11 Recent observations made during BOMEX show, however, that the albedo, longwave emission and shortwave absorption are all enhanced by the presence of the Saharan haze layer.12 The top of the Saharan air arriving in the area south and east of the principal potential temperature maxima shown in Figs. 17 a-d appears to have descended to about 600 mb. Along the northwestern fringes of the Saharan plume, however, the Saharan air top appears to be somewhat higher, possibly because the descending motion in the Saharan plume is being reversed in the northward-moving part of the flow pattern, particu- larly near the middle latitude frontal zone where the isentropic surfaces would slope upward toward the north. (No vertical motions should be inferred from the displacements of the base of the Sarahan layer which would rise gradually between Africa and the Caribbean in response to a deepening of the trade wind moist layer produced by small-scale convection over the ocean.) 7. Concluding remarks In this paper we have attempted to show that pulses of warm, dust-laden air from the Sahara are commonly present over the northern equatorial Atlantic Ocean during the summer and early fall. These pulses appear to be generated near the African coast by an interrup- tion in the westward transport of dusty Saharan air caused by the passage of travelling disturbances from equatorial Africa; after the disturbances have passed, the westward flow of dusty air from the interior of the Sahara is resumed. A significant feature of the dust pulse is its anticyclonic rotation and strong winds (often with a distinctive jet maximum near 650 mb) which necessarily exist in the presence of the strong horizontal 11 See Footnote 10. 12 See Footnote 10. 193 March 1972 T . N . CARLSON AND J M PKOSFERO 297 temperature contrast between the Saharan pulse and surrounding air masses. Because the Saharan air departs Africa only in a layer between approximately 850-550 mb, these pulses are shallow middle-level phenomena which are not easily recognizable on surface or high- level wind charts. Nevertheless, in view of the large- scale nature of the Saharan air anticyclone, the peculiar- ities of wind, temperature and atmospheric stability may well have important implications for the meteor- ology of the tropical Atlantic. The dust concentration in the Saharan layer is high, being comparable to that in polluted areas over the United States (Blifford, 1970). Consequently, as far as aerosols are concerned, it is probably erroneous to think of this region in terms of a typical maritime atmosphere. We cannot state with any certainty what effect the presence of the dust has upon the meteorology of this region, although we have set forth an estimate of the atmospheric cooling in the Saharan air which may de- pend, in part, upon the optical and radiative properties of the dust. It also seems that in addition to modifying the radiation balance of the tropics the dust might play some secondary role in water vapor and ice nucleating processes. Acknowledgments. We thank the Research Flight Facility (NOAA), particularly the crews of aircrafts 39C and 40C, for their generous cooperation. We are especially grateful to Ann R. Prospero who participated in many of the flights and gathered much of the data presented here. We are indebted to Judge G. L. Taylor, Barbados, for the continued use of his land and resi- dence for our operations and to R. Clarke, D. Moore and C. Sealy for their assistance in the Barbados studies. We are also indebted to personnel aboard the Discoverer for their diligence in making various measurements cited in the text. We would also like to thank Dr. Harry Hawkins of NHRL for his encouragement and helpful criticisms. Mr. R. L. Carrodus drafted the figures. A portion of this research was supported by the Office of Naval Research under Contract No. 4008(02) and the National Science Foundation under Grant GA-25916. REFERENCES Anthes, R. A., 1970: A diagnostic model of the tropical cyclone in isentropic coordinates. ESSA Tech. Memo. ERLTM- NHRL-89, NHRL, Miami, Fla., 147 pp. Blifford, I. H., Jr., 1970: Tropospheric aerosols. J. Ceopliy. Res., 75, 3099-3103. Carlson, T. N., 1969a: Synoptic histories of African disturbances that developed into Atlantic hurricanes. Man. Wea. Rev., 97, 256-276. , 1969b: Some remarks on African disturbances and their progress over the tropical Atlantic. Man. Wea. Rev., 97, 716-726. , and F. H. Ludlam, 1965: Research on characteristics and effects of severe storms. Annual Summary Rept. No. 1, Grant AF-EOAR-64-60, Imperial College, London. , and , 1968: Conditions for the formation of severe local storms. Tellus, 20, 203-226. Cox, S. K., 1969: Observational evidence of anomalous infrared cooling in a clear tropical atmosphere. /. Almos. Set., 26, 1347-1349. Delany, A. C, A. C. Delany, D. W. Parkin, J. J. Griffin, E. D. Goldberg and B. E. F. Reimann, 1967 : Airborne dust collected at Barbados. Geochim. Cosmocliim. Acta, 31, 885-909. Flowers, E. C, R. A. McCormick and K. R. Kurfis, 1969: Atmo- spheric turbidity over the United States, 1961-1966. J . Appl. Meteor., 8, 955-962. Frank, N. L., 1969: The "Inverted V" cloud pattern — an easterly wave? Man. Wea. Rev., 97, 130-140. Junge, C, 1963: Air Chemistry and Radioactivity. New York, Academic Press, 382 pp. London, J., 1952: The distribution of radiation temperature change in the Northern Hemisphere during March. J. Meteor., 9, 145-151. Morressey, J. F., and F. J. Brousaides, 1970: Temperature-in- duced errors in the ML-476 humidity data. /. Appl. Meteor., 9, 805-808. Pearson, J. E., and G. E. Jones, 1966: Soil concentrations of emanating radium-226 and the emanation of radon-222 from soils and plants. Tellus, 18. 655-622. Prospero, J. M., 1968: Atmospheric dust studies on Barbados. Bull. Amer. Meteor. Soc, 49, 645-652. , and T. N. Carlson, 1970: Radon-222 in the north Atlantic trade winds: Its relationship to dust transport from Africa. Science, 167, 974-977. , and , 1971 : Mineralogy of aerosols collected at Miami, Florida: Evidence for the frequent presence of African dust (Abstract) Trans. Amer. Geophys. Union, 52, 370. , E. Bonatti, C. Schubert and T. N. Carlson, 1970: Dust in the Caribbean atmosphere traced to an African dust storm. Earth Planetary Sci. Letters, 9, 287-293. Sasamori, T., 1968: The radiative cooling calculation for appli- cation to general circulation experiments. /. Appl. Meteor., 1, n\-n9. Servant, J., 1966: Temporal and spatial variations of the concen- tration of the short-lived decay products of radon in the lower atmosphere. Tellus, 18, 663-670. Stevenson, C. E., 1969: The dust fall and severe storm of 1 July 1968. Weather, 24, 126-132. Volz, F. E., 1969: Some results of turbidity networks. Tellus, 21, 625-630. , 1970: Spectral skylight and solar radiance measurements in the Caribbean : Maritime aerosols and Sahara dust. /. Almos. Sci., 27, 1041-1047. 194 19 U.S. DEPARTMENT OF COMMERCE National Oceanic and Atmospheric Administration Environmental Research Laboratories NOAA Technical Memorandum ERL NHRL-98 DEVELOPMENT OF A SEVEN-LEVEL, BALANCED, DIAGNOSTIC MODEL AND ITS APPLICATION TO THREE DISPARATE TROPICAL DISTURBANCES Harry F. Hawkins, Jr, National Hurricane Research Laboratory Coral Gables, Florida January 1972 195 ABSTRACT Large scale vertical motions are computed using a 7-level diagnostic model in which the s treamf unct i on and contours are related through the balance equation. The model with grid-point spacing of about 205 km is designed primarily for synoptic-scale studies over the subtropical ocean and island areas where moisture is relatively abundant. Consequently, an idealized moisture supply, driven by surface frictional convergence, releases latent heat which is partitioned in the vertical in a modified Kuo fashion. This model has been applied to a late season, strong, cold low; an easterly wave deepening into a closed system, and, to a weaken- i ng eas te r ly 1 ow . If one considers the si mp lest form of the model, i.e., without surface frictional effects and wit ho ut latent heating, introduction of frictional effects increases central updraft velocities up to 50% in strong circulations. Further, the introduction of the release of latent heat increases the central updrafts by an additional factor of 2 to 5. With both of these effects included, grid scale vertical motions of the strong- -4 est circulations considered reached values of 10 cb/sec (about one cm/sec) and a tenth of this value in the weakest system. The vertical motions and fields of divergence associated with them relate to the observed weather in satisfactory fashion although they correlate very poorly with the kinematic divergences. Test runs were made in which the weightings of stream function (derived from the wind field analysis) and analyzed heights were varied. Diagnostic results were relatively stable and suggested that wind analyses alone are adequate for most purposes. The effects of latent heating are greatest at higher levels (250 and 400 mbs) with maximum heating equivalent to k or 5 C/12 hrs noted. Vertical motion accompanying the heat release is just about enough to counteract the heating through "adiabatic" expansion. The net heating or cooling is a very small residual. Vorticity budgets revealed that in the weaker disturbances where divergence is small almost all of the vorticity change can be attributed to advection. In stronger systems with stronger vertical velocities the divergence term may equal or even dominate the advective term at those levels where divergence is most marked. The vertical advection and twisting terms seldom become very large,but on occasion when the advective and divergence terms are weak they may contribute significantly to the vorticity change at that particular level. Correlations show that the calculated vorticity tendency relates to the observed tendency best in the more quiescent situations. In the strongly disturbed state we find evidence that up through the 550 mb level the cumulus scale activity seems to be effecting vorticity changes which are proportional to the vertical gradient of vorticity between 1000 mb and the level under consideration. The inference to be drawn is that in intensely active, convective situations the vertical transport of vorticity by synoptic scale vertical velocity is too small and that the transport must occur at something more nearly approaching cumulus scale velocities. 196 Reprinted from Journal of Applied Meteorology, Vol. 11, No. 1, February, 1972, pp. American Meteorological Society Printed in U. S. A. 221-226 20 Successful Test of an Airborne Gas Chromatograph1 Harry F. Hawkins National Hurricane Research Laboratory, NOAA, Coral Gables, Fla. Karl R. Kurfis Division of Meteorology, EPA, Raleigh, N. C. Billy M. Lewis National Hurricane Research Laboratory, NOAA, Coral Gables, Fla. AND H. GOTE OSTLUND Rosensliel School of Marine and Atmospheric Science, University of Miami, Miami, Fla. (Manuscript received 23 July 1971) ABSTRACT A portable gas chromatograph designed for real-time studies in diffusion has been successfully tested in an airborne application. A 10-mi low-level plume of colorless, odorless non-toxic sulfur hexafluoride (SF«) was laid down over northwest Andros Island, Commonwealth of Bahama Islands, by the Research Flight Facility's (NOAA) C-130. An accompanying DC-6 carrying the analyzer made 81 in-flight samples over a period of less than 3 hr. The chromatograph has a sensitivity limit of 0.03 parts per billion and completes the analysis for SF» less than 2 min after the sample is injected into the column. Of the more than 80 in- flight chromatograms, 1 1 indicated positive identification of the inert tracer gas. Confirmation of the on-site analyses was provided by bottle samples taken in flight and analyzed later in the laboratory. 1. Operation and description of the portable gas chromatograph Several tracer materials have been used for meteoro- logical studies in air mass transfer and circulation Saltzman el al., 1966). Most of these have been success- ful to some extent; however they were, with some degree of accuracy, limited to rather short-range experiments. Sampling and analyzing the materials was difficult and expensive, and certain of the materials were known to be toxic. Halogenated compounds such as freons and sulfur hexafluoride (SFe) have been success- fully used as tracers (Saltzman et al.; Niemeyer and McCormick, 1968). The analytical process by which these gaseous tracers are separated from air and analyzed for quantity is by gas chromatography. The gas selected was SF6 since it is chemically inert, non- toxic (Kirk-Othmer, 1963), very sensitive to the gas chromatograph process used, and is not highly subject to fallout, washout, impaction or other removal processes from the atmosphere (Saltzman el al.). Com- mercially, SF6 is used in some high-voltage switching equipment and transformers. Background levels, with the exception of areas near where the gas is manu- 1 Contribution No. 1441 from the Rosenstiel School of Marine and Atmospheric Science, University of Miami. factured or used in electrical equipment, is suspected to be approximately 1X10~15 (demons el al., 1968). The detection mode within the chromatograph is electron capture. The detector is an ionization type which uses a 300-mCi radioactive tritium source emitting low-energy beta particles. The detector is balanced electrically, and the introduction of electron capture materials disturbs the balance. The result of this disturbance appears as a chromatographic peak. The peak for SF6 is separated from other materials by use of a | inch (outside diameter) column, 42 inches long, packed with Alcoa F-l Alumina, 100-120 mesh. The chromatograph used for this work was designed and built by Analytical Instruments Limited, Fowlmere, England. The primary purpose of the instrument is for leak detection from closed vessels. When a gas sensitive to electron capture is introduced into a sealed container under slight pressure, the seams and weak points can be checked by using the chromatograph as a continuous "sniffing" analyzer. The instrument, operated in this mode, is not as sensitive, but quite acceptable for leak detection. A second mode of operation introduces a measured sample into the system and provides a much more sensitive chromatogram. A 0.5-cm3 loop attached to the sample valve enables the measured sample to enter the 222 JOURNAL OF APPLIED METEOROLOGY Volume 11 E PUMP • VENT t I DETECTOR m ♦ ♦ El ) ALUMINA COLUMN LU CONTINUOUS AIR SAMPLER SAMPLE VALVE LiJ 1 — «- -O CARRIER GAS GJ2 3 RECORDER Fig. 1. Schematic diagram of the gas chromatograph. Ambient air enters the system via air-scoop at [1] and unless sample valve [_3~] is depressed, is pulled through sample loop [4] by the pump [2~] and returned to the outside airstream. When the sample valve is depressed, the helium stream is diverted so that it drives the air in the sample loop [4] through a 42-inch column [5] of Alcoa F-l Alumina, (100-120 mesh). The alumina separates the "air" molecules from those of SFe so that as these arrive separately at the detector [6] where they all are exposed to a radioactive tritium source, the electrical balance (previously established under regular helium flow) is disturbed and the disturbance is amplified [7] and meas- ured [8, 9]. column detector array upon depressing a valve. Helium carrier gas pushes the sample through the column and detector (see schematic diagram, Fig. 1). The carrier gas flow was maintained at 35 ml min_I, the controlled flow for optimum instrument sensitivity, to provide optimum operating conditions for the system. Helium was selected as the carrier because it contributed to good separation between the air peak (described below) and the SF6 peak. The chromatograph amplifier and detector operate from four mercury cells establishing ±15 V, and the pump for pulling or pushing the sample is powered by a 6 V nickel cadmium rechargeable cell. The only outside power source required was for the recorder. In general, preparing the equipment for use aboard the aircraft took very little time. Since the unit is portable, small, and has a self-contained power supply, very few special requirements were necessary. The chromatograph and recorder were shock-mounted to minimize vibration. Helium carrier gas flow was maintained for about 12 hr prior to flight time to allow the unit to stabilize and assure a smooth performance during the flight. In the aircraft a rather large plastic tube (1| inches in diameter) extended from near the front of the fuselage to the center, the tube being tapped for intake and exhaust of sample. A slight positive pressure was noted through this tube while in flight, and a restrictor clamp was attached in front of the instrument intake and a teflon pump at the exhaust end of the instrument. This method allowed a controlled flow of sample which was regulated to 10 cm3 min-1. The sample loop was flushed with at least 20 cm3 of air prior to sampling. Sampling aboard the aircraft was accomplished in two ways. Analysis of 81 chromato- grams was made for SF6 content in flight. Real-time data with chromatograms in flight enabled the meteor- ologist and navigator to optimize use of the aircraft as a sampling platform. Grab samples were taken in small metal cylinders (~1 liter capacity) every one-half minute. The cylinders had been evacuated so that their vacuum enabled sampling when opening the cylinder valve. Care was exercised prior to the experiment to keep the cylinders isolated from any areas where SF« was stored. v> 2 - *y SENSITIVITY SETTING ■ 10X — SF, (0.5 PPB) AIR PEAK 0 - A"-S*MPLE INJECTION J- _l_ _l_ -1_ _l_ _1_ _l_ _l_ -J_ 0 10 20 30 40 50 60 70 80 90 100 Fig. 2a. Laboratory chromatogram of reference sample of 0.5 ppb SF, at 10X sensitivity setting. Note stability of base line. 0- I 1 1 1 1 1 1 SENSITIVITY SETTING «2X SF, (0.013 PPB) "? •-SAMPLE INJECTION 10 20 30 40 50 60 70 -L. 80 90 100 Fig. 2b. Laboratory chromatogram of grab sample bottle no. 341 at 2X sensitivity showing 0.013 ppb SF». 198 February 1972 HAWKINS, KURFIS, LEWIS AND OSTLUND 223 In the chromatograms (Figs. 2a, 2b, 2c) the first peak to appear is oxygen or air. The air peak in this type chromatogram is referred to as the composite air peak. In addition to oxygen, certain fixed gases present in the atmosphere may be contained under this peak. This peak usually appears within one-half minute from injection of a sample to the column. If the sample contains SF6 in an amount greater than the indicated instrument baseline sensitivity, the peak will appear about 1 min, 16 sec after sample injection. The height of this peak determines the quantity of SF6 in the sample. Fig. 2a is a reference standard gas mixture in dry air. This chromatogram is one of several reference chromatograms used to calibrate the system. The refer- ence mixtures are repeatable from day to day to within 2%. The concentrations in most of the reference mix- tures are such that they may be analyzed at different instrument sensitivity settings and thus provide a ratio for each sensitivity setting. These crosschecks indicate a direct comparison between the numbered settings on the chromatograph and the recorder out- puts (chromatograms). The chromatograph has at- tenuation settings of 50X, 20X, 10X, 5X, 2X and IX, the last being the most sensitive. A peak height obtained at a setting of 10 X would be 5 times larger at a setting of 2X. The six attenuation settings conven- iently allow for a reasonably wide range of sample concentrations. Instrument noise and poor stability prevented the use of sensitivity 1 X . The reference mixtures are used to establish a cali- bration curve. The final calibration curve (Fig. 3) is established by a smooth line between points of the standard reference gas mixtures. The calibration curve approximates a straight line for quantities to 2.5 parts per billion (ppb). Beyond 2.5 ppb the curve decreases rapidly and the system accuracy suffers considerably. Where very large amounts of SF6 are probable, a dilu- tion process may be used to bring the data within acceptable calibration limits (Saltzman et al., 1966). Fig. 2b is a chromatogram from the bottled air samples taken aboard the aircraft. These chromatograms were analyzed several weeks after the experiment. Samples OT 2 o- SENSITIVITY SETTING = 5X SF, (0.2 PPB) — SAMPLE INJECTION 10 20 30 40 50 70 80 90 100 Fig. 2c. An in-flight chromatogram at SX sensitivity of known reference sample. Note increased "grass" on base line which reduces airborne detection capabilities. The grass appears to be mainly due to mechanical vibration. 1800 1600- 1400- £ 1200- °- 1000 - I 800 t 600- 400 200 1.0 15 20 CONCENTRATION (PPB) Fig. 3. Final calibration :urve for the portable chromatograph established by using referenced samples. The curves used for this experiment were expanded versions of areas B and C. contained in the metal cylinders do not show any noticeable loss for moderate storage time before being analyzed. These samples were at pressures > 1 atm because the containers were filled from the cabin pres- surization line. They were taken from a different inlet than that used for the real-time data. This overpressure pushed the sample through the chromatograph for the laboratory analysis. The bottled samples that contained SF6 content above the instrument baseline sensitivity were analyzed three or four times to check peak height and assure that the data were run at the most sensitive setting. Reference mixtures were intermittently ana- lyzed to check for any changes in the calibration curve. Fig. 2c is a chromatogram taken aboard the aircraft. Vibration and perhaps some electrical noise aboard the aircraft contributed to less sensitive chromatograms in flight than in the laboratory by a factor close to 10. In the aircraft, we were able to determine SF6 content to 0.03 ppb, whereas determinations of 0.004 ppb could be made in the laboratory. Quantities of SF6 less than these stated limits were considered a trace since exact amounts were difficult to determine because of instru- ment noise. All of these chromatograms are from a recorder operating at 10 mV full scale and a chart speed of 1 inch min-1 with a rapid response. The calculation of the peak height was made from the center of the base line of the peak and the center of the peak. The units in this case are chart divisions at the ordinate and parts per billion at the abscissa. All SF6 data are given in parts per billion. 199 224 JOURNAL OF APPLIED METEOROLOGY Volume 11 2. Airborne test of the portable chromatograph A n airborne test of this portable analyzer was made on 30 October 1970 by the National Hurricane Re- search Laboratory utilizing resources of the Research Flight Facility. The instrument was couched in an im- provised shock mounting so located in the DC-6 that the operator could not readily determine what part of the flight track he was on. Meteorological conditions were quite stable with a weak anticyclonic south- easterly flow (variable at times) in the lower 10,000 ft. Cumulus activity over the open ocean was altogether missing, but small and swelling cumulus were located about j mi inland from the beach on Andros Island (Commonwealth of Bahama Islands) over which we proposed to conduct the experiment. At 1620 GMT the RFF C-130 laid an SF6 plume 10 mi in length along the beach beginning at a readily identifiable estuary (Wide Opening). Four tanks of SF6 were used emitting a total of 145±5 lb of sulfur hexafluoride at an altitude of 1500 ft. We planned to allow some time for diffusion and then to fly the DC-6 with the portable analyzer aboard through the dispersed plume, testing for SF6 on a real-time basis and at the same time taking grab samples in bottles to be analyzed later in the laboratory. Real-time readings could be taken about once every 2 min and the grab samples were taken once every 30 sec (in the early testing) as the plane travelled about 3-3.5 mi min-1. In order to provide some estimate of where the plume was located, we have made a rough calculation of its mean transport by the wind (at 150°, 6 kt) and its diffusion. At the suggestion of Dr. L. Machta of the Air Resources Laboratory, we have assumed the standard deviation a of a point source is given by ')] +- /4wtPSFt \1 ■AKJLnl {KhK.)*) — 1. where / is the time in seconds. The denominators are taken as the squares of the major (horizontal) and minor (vertical) axes of the ellipse formed in the x, z plane by the diffusing gas. The diffusion envelopes in Figs. 4H> are for concentrations of 0.004 ppb, the small- est measurable amount in the laboratory analysis. The first search pattern was flown at an aircraft Fig. 4. Western portion of Andros Island showing the original plume laid along the beach and the calculated 0.004 ppb concen- tration 40 min later. The aircraft was at altitudes above the calculated maximum diffusion height but measured SFj in cumuli nevertheless. altitude of 3800 ft and was centered in time ~ 40 min after the plume was laid down. Fig. 4 shows that the real-time analyzer reported a positive measurement of 0.36 ppb at 1716, in just about the middle of the calcu- lated plume. A trace was recorded on the next sample at 1718. Both of these samples were taken in small cumu- lus cells. The first reading was the highest concentra- tion recorded all day (as was the bottle sample taken at the same time). In addition, bottle samples showed their minimum measurable amount (0.004 ppb) in two measurements around 1705 which were presumably below the real-time detection capabilities. Another positive bottle reading about 1653 was apparently spurious. It may be noted that our initial flight altitude (for the track shown in Fig. 4) was above the maximum height calculated for the z axis at / = 40 min. This apparent discrepancy can be readily explained as due, first, to the lack of consideration of the original "instan- taneous" diffusion of the gas in the wake of the C-130 and, second, to the vertical mass transport of the SF6 upward in the small cumuli in which the positive measurements were obtained. The next run was made at 4900 ft (dashed track, Fig. 5). Once again measurable amounts were found (around 1737) in a small swelling cumulus within the calculated plume. The portable analyzer suggested 200 February 1972 HAWKINS, KURFIS, LEWIS AND OSTLUND 225 0.06 ppb while the bottle sample showed 0.024 ppb. A barely measurable bottle sample at 1746 showed the minimum detectable amount outside of the calculated plume in a cumulus cloud. These measurements were above the maximum calculated diffusion height for this time (3900 ft), so that the coincidence of measured amounts with the occurrence of cumulus penetrations is significant. Next a short run at a higher altitude (6000 ft) was carried out as indicated in Fig. 5. Positive measurements were recorded in selected swelling cumulus with fairly good agreement between portable and grab samples. (One must take note that the limit of sensitivity on the portable airborne analyzer was about 0.03 ppb.) The final track (Fig. 6) was flown at 2000 ft and was begun more than 2 hr after the plume was laid down. At 1834 a weak trace was barely perceptible on the airborne chromatograph and was judged to be accept- able but (when the position is considered) was in all probability spurious. Both the portable and grab samples showed measurable amounts at about 1841 in the calculated plume. No other traces were discovered except a questionable trace barely suggested by the airborne chromatograph at 1921 — a spurious deter- mination. Why no positive readings were found be- tween 1855 and 1900 when the plume was crossed is not immediately obvious. The winds were light and variable, enough so that the Doppler winds were not dependable and was occasionally on "memory." The Fig. 5. Subsequent flight altitudes and tracks relative to the 0.004 ppb plume calculated for 90 min after dispersal. If one allows for differences in frequency of observations and sensitivities the readings are in general agreement. Fig. 6. Final aircraft track and calculated plume (140 min after dispersal). The only measurable amounts of SFe were again in the plume and were in fair agreement with each other. mean displacement vector was calculated by consider- ing both the Doppler and omega navigation system winds. Also, significant amounts of the SF6 were evidently pumped aloft by the cumulus activity, possi- bly at the temporary expense of the lower level concentrations. It will be noted that the portable chromatograms con- sistently indicated greater quantities of SF6 than the bottle samples. A number of factors may account for this difference: 1) The calibration "in-flight" may have changed from the pre-flight calibration run made on the ground; because of time limitations no calibration run was made in-flight after the measurements. 2) Differences in samples must be expected because of differences in sampling times and techniques. 3) The Tygon tubing used in gathering the bottle samples is known to have a tendency to retain, at least temporarily, minute amounts of SF6 and may have weakened the bottle samples by this retention. It could possibly account for some of the weak samples which were seemingly "delayed" but we cannot, at present, make any definitive statement. 3. Conclusions The portable airborne chromatograph is a viable instrument that has many potential meteorological applications. For certain meteorological experiments of limited time and space scales (e.g., studies in cumulus dynamics) the current portable analyzer could provide very useful indications. Our own sphere of interest, that of hurricane dynamics, demands analyses to 201 226 JOURNAL OF APPLIED METEOROLOGY Volume 11 greater sensitivity (1 part in 10") ; such sensitivities can at this time only be achieved in permanent labora- tory installations where conditions can be more rigidly controlled and enrichment techniques can be used, if necessary. REFERENCES demons, Clarence A., Arthur I. Coleman and Bernard E. Saltz- man, 1968: Environ. Sci. Tech., 2, 551-556. Heffter, J. L., 1965: The variation of horizontal diffusion param- eters with time for travel periods of one hour or longer. J. Appl. Meteor.^ 4, 153-156. Kirk-Othmer, 1963: Encyclopedia oj Chemical Technology, Vol. 9, 2nd revised ed.( 664-670. Niemeyer, Lawrence E., and Robert A. McCormick, 1968: Some results of multiple tracer diffusion experiments at Cincinnati. Air Pollution Control Assoc. J., 18, 403-405. Saltzman, Bernard E., and Clarence A. demons, 1966: Andy. Chem., 38, 753-758. , Arthur I. Coleman and Clarence A. demons, 1966: Andy. Chem., 38, 800-801. 202 21 Reprinted from: Journal of Crystal Growth 12 (1972) 21-31 © North Holland Publishing Co. Printed in the Netherlands LINEAR GROWTH RATES OF ICE CRYSTALS GROWN FROM THE VAPOR PHASE D. LAMB* Atmospheric Sciences Department, University of Washington, Seattle, Washington, U.S.A. and W. D. SCOTT National Hurricane Research Laboratory, NOAA, Miami, Florida, U.S.A. Received 2 April 1972; revised manuscript received 26 September 1971 Measurements of the linear growth rates of individual basal and prism faces of ice grown on a substrate were made as functions of temperature, excess vapor pressure, and partial pressure of air. The crystals grown in an environment of pure water vapor appear relatively featureless and flat, but still exhibit well-defined specular faces, indicating a domination of surface kinetic effects. Addition of air to the system decreases the growth in proportion to the diffusivity of water vapor in air, so it appears that the effects of the environment are separable from the surface kinetic processes. Steps were observed on both the basal and prism faces and are an important feature of the growth mechanism. Trains of macroscopic steps were observed being "shunned" in an area packed by etch pits and initially inhibited faces were seen to assume the same linear growth velocity as overtaking adjoining faces. The tem- perature dependences of the linear growth rates of basal and prism faces are remarkable with local maxima and minima which are consistent with the observed alternation with temperature of the primary habit. The temperature trends are not readily explained, but they are correlated with previously measured values of the velocity and interaction distance of steps on the basal face. This correlation and the other observations of this study are consistent with a growth mechanism based on spiral steps arising from emergent screw dislocations. 1. Introduction The growth of ice from the vapor phase has been studied by numerous workers, particularly with a view toward explaining atmospheric phenomena. The laboratory investigations of Nakaya1) showed that the many habits of ice crystals grown in air are systemati- cally dependent on the temperature and supersatura- tion. The habits of ice crystals grown naturally in the atmosphere have been similarly classified2) and were found to correspond well with those of artificially grown crystals. A similar temperature dependence to the habits was found by Shaw and Mason3), growing single crystals of ice on a metal substrate; but super- saturation seemed to control only the overall rate of growth, not the relative rates of growth upon the prism and basal faces. Kobayashi4) confirmed this result and pointed out that the habits of ice are separ- able into two classes. The basic or primary habit of a * Present affiliation: Institut fur Meteorologie und Geophysik, Universitat Frankfurt/Main, Germany. crystal is defined by the shape (i.e., whether plate-like or column-like) of the "underlying" regular hexagonal prism; it correlates strongly with temperature. The secondary habits are features of growth which appear as modifications of the primary habits; they correlate with the supersaturation or rate of growth. It is now generally accepted that the primary habits of ice alter- nate (plates-columns-plates-columns) as the tempera- ture is lowered from the melting point, with the tran- sitions occurring at roughly -4, -9, and -22 °C. To investigate the mechanism of growth and habit formation quantitatively Hallett5) and Kobayashi6) used the optical interference technique of Bryant et al.7) to measure the velocity v at which individual steps were observed to propagate across the basal face of ice grown epitaxially on covellite. Both investigators found the velocity of steps, normalized to a height of 250 A, to be a remarkable function of temperature. Each found a trend which contained a local maximum and a local minimum, but the temperatures at which the respective maxima occurred differed by about 4 CC (see fig. 1). 21 203 22 D.LAMB AND W.D.SCOTT V(KOBAYASHI) -20 TEMPERATURE CO Fig. 1. Measured trends to the mean migration distance as and the velocity v of a 250 A step as a function of temperature. Since both workers were purportedly investigating the same parameter, this difference is probably not real8). In addition, Hallet found that the step velocity was inversely proportional to the step height and observed qualitatively that two approaching steps could interact with each other while still some distance apart, leading to a mutual reduction in the step velocities. From these results he attributed the existence of such an inter- action to the limitations imposed on the step velocity by the diffusion of adsorbed molecules across the sur- face of the ice. Mason et al.9) identified this critical interaction distance with twice the mean migration distance xs of the adsorbed molecules and measured it as a function of temperature. The variation of xs with temperature was found to be similar in form to the changes in v (see fig. 1). There is, as yet, no satisfactory explanation for the trends with temperature of either xs or v although several attempts have been made9-11). These trends are most likely linked with the temperature variation of the crystal habits of ice. Unfortunately, similar measurements on the prism face of ice have not yet been possible, so any discussion of habit formation based on these measurements is necessarily speculative. With this in mind, a series of experiments was per- formed in an attempt to discriminate between the basal and prism faces and reveal the physical basis for the growth of ice from the vapor phase. Specifically, the objectives of the work were : ( 1 ) to separate the processes taking place in the environment from those acting on the growing surface itself, (2) to quantify the rates of growth on the individual prism and basal faces, and (3) to determine the growth mechanism, if possible. To accomplish these objectives, the linear growth rates of individual faces were measured in a vacuum chamber under precisely controlled conditions of temperature and excess vapor pressure, in environments of pure water vapor and known amounts of air. The results are reported here. 2. Description of apparatus The single crystals of ice used in the present study were nucleated and grown on a stage beneath an optical microscope from a vapor source derived from poly- crystalline ice surrounding the growth stage. A cross sectional representation of the vacuum growth chamber is shown in fig. 2. The chamber was machined as a unit of stainless steel and immersed in a circulating bath (designated the "secondary" bath) of methanol supplied from a larger, "primary" bath whose temperature was CONNECTION PLATE MICROSCOPE GLASS COVER \a .GROWTH STAGE THERMOELECTRIC ELEMENT Fig. 2. Schematic cross section of the axially-symmetric growth chamber. 204 LINEAR GROWTH RATES OF ICE CRYSTALS GROWN FROM THE VAPOR PHASE 23 governed by a mercury contact thermostat. The top connection plate, containing a glass plate as well as electrical and pressure connectors, was fitted to the growth chamber by means of a Viton-A O-ring. Sandwiched between two copper plates beneath the growth surface was a thermoelectric element which provided the cooling of the growth substrate relative to the "moat". The bottom plate dissipated the heat from the thermoelectric element, whereas the top cylin- der minimized temperature gradients across the sub- strate. Current was supplied to the thermoelectric ele- ments by means of two electrical feed-throughs soldered to the lower copper plate. One other, dual feed-through (not shown) passed the wires of a thermocouple into the inner compartment. One junction for this thermocouple was imbedded in the top copper cylinder to measure the "absolute" temperature of the stage relative to a reference junction situated in a Dewar flask of ice water. A second thermocouple (38-gage, copper-constantan) was passed through the top connection plate, the sens- ing junction (not shown) of which was soldered to the center of the growth substrate. Control of the supersaturation over the growth substrate was effected by controlling the substrate temperature and the pressure of water vapor in the chamber. Although controlled independently, each parameter was differentially related to a common reference chamber, yielding in effect a single "double- differential" control system. The reference chamber was entirely submersed in the primary bath and was partially filled with polycrystalline ice to provide a refer- ence vapor pressure equal to the equilibrium vapor pres- sure of ice at the bath temperature, the reference for the temperature controller. The sensor for the differential temperature control system was the 38-gage thermo- couple soldered to the substrate ; the sensor for the pres- sure controller was a capacitance manometer (MKS In- struments, Inc.). After suitable amplification, the voltage from each sensor was fed into separate "proportional- plus-reset" control circuits built from solid state opera- tional amplifiers. The output of the temperature con- troller governed the amount of current flowing through the thermoelectric element to maintain the temperature of the substrate within ± 0.002 °C of the temperature of the reference chamber. Similarly, the current through a stainless steel heater tape (not shown) im- bedded in the "moat" ice was controlled to maintain the pressure of the pure vapor inside the reference chamber. The magnitude of the excess pressure at any temperature was specified by an electromechanical function generator calibrated against the factory-cali- brated capacitance manometer. The net control over the excess vapor pressure in a pure vapor system was limited by the pressure controller and maintainable to within ± 1.0 urn Hg of the desired value. System gases were evacuated by means of a dual- chamber mechanical vacuum pump, and a liquid nitrogen trap isolated the growth chamber from the pump. But, in a few experiments air was readmitted into the growth chamber through a drying tube con- taining CaS04, in which case the system pressure was measured with a conventional precision Bourdon-type pressure gage. In all cases, water vapor was supplied to the system by redistilling previously distilled water under vacuum from a glass flask blown as a unit into the system. 3. Procedures Prior to taking measurements of ice crystal growth rates, the growth chamber was cleaned repeatedly with water, acetone, and carbon tetrachloride. The entire system was then evacuated by pumping at least over- night, while a 100-watt incandescent lamp heated the growth chamber to approximately 100 °C. Then, water was distilled into the growth and reference chambers, the temperatures of which were maintained just above 0 °C. When enough water had been collected in each chamber, the temperature of the primary bath was lowered to about — 10 °C to freeze the water. Through- out the distillation, freezing, and crystal growth operations, the two chambers were exposed to the vacuum manifold through valves which were just "cracked" open to prevent the build-up of any out- gassed contaminants. The ice in the chambers was allowed to settle into the coldest places for at least another day before any growth measurements were attempted. After suitable adjustments to the differential tem- perature and pressure controllers had been made, ice crystals were nucleated on the growth stage with ap- parently random orientations. Just after nucleation, while still small, the ice crystals were usually polyhedral in form, exhibiting well-defined edges and faces with the geometry of a truncated hexagonal prism. Fig. 3 205 24 D.LAMB AND W.D.SCOTT YOUNG STAGE H h — 10, INTERMEDIATE STAGE MATURE STAGE Fig. 3. Three stages in the evolution of an ice crystal grown in an environment of pure water vapor. illustrates the typical evolution of an ice crystal as it matured in an environment of pure water vapor. It was during the "young" stage that positive identification of the crystal faces could be made. The orientation of a given face with respect to the substrate was determined by measuring the angle of inclination of a narrow light beam when it reflected specularly into the microscope. Under a specified set of environmental conditions the rate at which a given face advanced normal to itself was determined trigonometrically from the angle of inclination of the face and the rate of propagation of the lower edge of the crystal across the field of view. The lateral rate of propagation of the crystal edge was measured by noting its position relative to fiducial lines in a vernier eyepiece at specified intervals of time (usual- ly 30 sec). The resulting time-series of position points was then differentiated numerically using a criterion of least squares over 5 consecutive points, so obtaining a "running" derivative. When growth measurements were made as a function of the excess vapor pressure or as a function of the partial pressure of air in the chamber, the environmen- tal conditions were changed incrementally and the growth rate measured under steady state conditions. Measurements taken as a function of temperature, on the other hand, were made as the temperatures of the primary and secondary baths were changing linearly with time at about ± 2 °C hr_1. The controlled drift of the temperature was accomplished simply by having a timing motor rotate the set screw of the mercury thermostat which governed the temperature of the primary bath. 4. Quantitative results The linear growth rates were measured for three distinct sets of conditions, namely: (1) as a function of temperature in pure water vapor at constant excess pressure, (2) as a function of excess vapor pressure in pure vapor at constant temperature, and (3) as a func- tion of the partial pressure of air at constant temperature and ambient supersaturation. The measured tempera- ture dependence of the linear growth rates of the basal 1.0 1 - BASAL 1 /- _ FACE . i _ i J: - i i 1 - x • x . •' ** "H '. I 1 1 E 4 05 -20 -15 -10 TEMPERATURE -5 (°C) (a) — PRISM 1 I FACE • — - '• 7 — • •/ .. •* / '■■l — •/.• ': -V '.. \ I P :\ P r \ "T _ y. . \ J i " 10 E 4. — 05 o LE O < UJ -20 -15 -10 TEMPERATURE -5 C) (b) Fig. 4. The linear rate of ice as a function of temperature at a constant excess vapor pressure of 10 p.m Hg. (a) Basal face, (b) prism fase. 206 LINEAR GROWTH RATES OF ICE CRYSTALS GROWN FROM THE VAPOR PHASE 25 _ 08 (_> E -3-0.6 04 02 - 1 1 1 1 Face Prism „ X Temp -50°C y' y% SP1=10Fm-Hg s / Std Dev 0020Mm/sec / / y ,y / y y / y • / y X s / y / y / / yf s / y / y / — y y - y y y y' y y y / y . ^y .(L^l 1 1 1 1 08 E - 0.6 h 04 - 0.2 - 5 10 15 20 EXCESS PRESSURE (Mm-Hg) (a) 25 1 1 / /l // // . / / //' // / / / / / / / - / / / / /• Face Prism Temp -9.2°C SPi= 7Mm-Hg " / ./ Std Dev 0.035 /im/sec // -LS l 1 I 1.0 1 1 1 / ' - o <*> 0 8 Face Prism Temp.- -12 0°C. C2=2 3xl0~3Mm/se - Std Dev 0036/im c/^m /sec -Hg2 J / E / £ 06 < cc / - X S 04 o rr o - rr 2 02 z _j n " ./ 1 I - 5 10 15 20 EXCESS PRESSURE (^m-Hg) (b) 16 - £ 12 < rr x 5 0.8 o rr o 04 Face Basal Temp -14.3' rC2= 2 5^i.0'3fim/%ec/tirr\-Hqi <-Std Dev 0.068 ^m/sec 5 10 15 20 EXCESS PRESSURE (^m-Hg) (c) 25 5 10 15 20 EXCESS PRESSURE (^m-Hg) (d) 25 Face Prism Temp -16 8°C Std Dev 0 083^m/sec _L 5 10 15 20 25 (e) EXCESS PRESSURE (^m-Hg) Fig. 5. The dependence of linear growth rate on excess vapor pressure, (a) Fitted with the theory of Burton et al.12); (b) fitted with the theory of Burton et al.12); (c) showing a quadratic trend; (d) showing a quadratic trend (circled points); (e) fitted with a straight line. 26 D. LAMB AND W. D. SCOTT faces of the ice crystals is shown in fig. 4a*. Throughout the temperature range the excess vapor pressure was held constant at a value of (10 ± I) urn Hg. The analo- gous results for the prism faces are shown in fig. 4b. The solid line on each graph indicates a plausible average trend to the data. Note that the curves for growth on both the basal and prism faces are similar in overall shape, but that the maxima are shifted by about 7 °C. Most of the scatter observed in these data arose from actual variations in the rate of growth of a given face under steady conditions. Occasionally these variations in the growth rate were found to be as much as 50%. Fig. 5 shows how the growth of individual crystal faces was found to depend on excess vapor pressure at various constant temperatures. Figs. 5a and 5b suggest non-linear trends at low excess pressures and have been fitted with the spiral growth theory of Burton et al.12) (solid curves). The dashed lines represent the linear asymptotes to the solid curves. Figs. 5c and 5d show non-linear trends at all excess pressures in the range of the measurements and have been fitted with a simple * Where more than one growth rate was available at the same temperature, the arithmetic mean has been used. 10 01 001 0.1 Fig. 6. rate. 1 10 100 PARTIAL PRESSURE OF AIR (mm Hj) The effect of environmental air on the linear growth parabola. Only the circled points of fig. 5d were con- sidered in this case since the crystal face changed char- acter abruptly during the measurements and subse- quently grew at rates which were consistently higher and more scattered (solid points). The data of fig. 5e ex- hibited too much scatter over the entire range to be fitted with anything except a straight line. Some measurements of growth on arbitrary low- index faces in an environment of air are shown in fig. 6. .--- -<£ * ?iw*S* yrfB&Lji 4 TgKrHi^ ' 0 Fig. 7. The 3-dimensionaI character of an ice crystal grown in air. 208 LINEAR GROWTH RATES OF ICE CRYSTALS GROWN FROM THE VAPOR PHASE 27 The points represent data at constant temperatures and 5. Observed environmental interactions ambient supersaturations; different symbols indicate separate experimental runs. The solid lines represent In addition to the quantitative results on the linear reciprocal pressure relationships. growth rates just presented, a number of qualitative ob- Fig. 8. Examples of ice crystals grown in an environment of pure water vapor. In both figures: (I) ice crystal, (2) specular face, (3) substrate. 209 28 D. LAMB AND W. D. SCOTT servations on the forms and behavior of the ice crystals were made during the course of the investigation. The general appearance of the ice crystals grown in an en- vironment of pure water vapor, for instance, was found to differ considerably from that of crystals grown in the presence of air. Fig. 7 shows an ice crystal grown in air with the light being reflected specularly off the basal face and into the camera. The crystal edges and corners are seen to be sharp and to exhibit hexagonal geometry. A mature ice crystal grown in pure vapor, on the other hand, lacks many of these surface features and tends to grow two-dimensionally, as illustrated in fig. 3 (mature stage) and in fig. 8. (The low-index faces, being molecularly flat and oriented at an angle to the substrate, appear as bright patches in fig. 8 through specular reflection.) This tendency toward rounding of the extended features may be attributed to the uniform vapor supply above every point on the crystal surface and the warming which results from the release of latent heat. But one should note that the appearance of specular faces on the crystal near the substrate indicates that surface kinetics are limiting the rate of growth there. This is understandable since at the crystal- substrate boundary, where the growth measurements of figs. 4 and 5 were made, the effects of latent heat release should become insignificant. The influence of air on the growth is shown in fig. 6. The growth rates strongly tend to follow a reciprocal pressure relationship over nearly two orders of magni- tude, but deviate from it at high and low pressures. At low pressures, where the air makes up only a small fraction of the total vapor phase, the growth rate is very likely controlled by the rates of surface kinetic processes. As the partial pressure of the air is increased above a value comparable to that of the water vapor, diffusive resistance increases in the vapor phase and diminishes the net flux of water molecules to the growing ice surface. In the range of pressures in which molecular diffusion is important, the molecular fluxes are proportional to the binary diffusion coefficient which, in turn, varies inversely with pressure. As the pressure is increased still further, convection currents presumably develop within the chamber and partially offset the normal decreases of vapor fluxes due to pure molecular diffusion, leading to a more gradual fall-off of the growth rate with pressure. The above observations suggest that the effects of RAPIDLY MOVING FACE AFTER UNION Fig. 9. Two examples of inhibited growth, (a) Two separate basal faces; (b) two segments of a single crystal face, before and after union. the environment are separable from the surface kinet- ics. This means that the ideal growth function G = G(bp, T) may be considered to be the same in air and in pure vapor, if bp is taken as the local excess pressure immediately above the growing face and T is the temperature. 6. The importance of steps With crystals just recently nucleated on the growth stage, it was occasionally foundthat one or more low- index faces could be supersaturated by as much as 10 um Hg of excess vapor pressure with negligible growth occurring. Fig. 9a schematically illustrates one instance in which the basal faces of two neighboring and similarly oriented crystals each advanced at greatly differing rates. The large face remained virtually stationary, while the other advanced at several tenths of a micron per second past the large face. Furthermore, the inhibited face retained its characteristic hexagonal symmetry for as long as it remained inhibited. In another instance (fig. 9b), one segment of a crystal was apparently stationary as another segment of the same crystal approached the inhibited one from behind. When the moving segment caught up to the inhibited one, the two formed a visually smooth junction, 210 LINEAR GROWTH RATES OF ICE CRYSTALS GROWN FROM THE VAPOR PHASE 29 creating a single, continuous specular face. Very soon after this union occurred, the previously inhibited face advanced at the same rate as the other and remain- ed joined to it. The velocity of the initially moving face was apparently unaffected by the merger. Since the previously inhibited segment was already in a super- saturated environment, it must have lacked a mechanism for growth. Presumably the moving segment, upon merging with the other, supplied a train of steps. With both segments joined and provided with steps from the same source, it would be more than mere coincidence that the two subsequently advanced at identical rates. Such observations imply that the origins of steps on low-index faces are an important aspect of any growth mechanism. Certainly among the more commonly observed features of the surface structure of low-index faces of ice were laterally propagating steps themselves. The probability of observing steps on prism faces was about the same as on basal faces. Furthermore, each face exhibited both "slow" and "fast" steps. Although the actual speed of propagation of all steps varied qualita- tively in proportion to the excess pressure, "slow" st;ps almost invariably advanced in the direction of the local surface temperature gradient with decelerating speeds, whereas the "fast" steps were observed to emanate with relatively higher and more uniform speeds from generally well-defined points of disturb- ance at points of contact with another crystal or with something on the substrate. The behavior of the "slow" steps appeared to be a direct consequence of the self- heating of the growing crystals, since as the steps pro- pagated away from the substrate, they encounter areas which become progressively warmer. This warming should diminish the local excess pressure and so cause the steps to slow down and converge forming fewer, thicker steps that would be visible under a magnifica- tion of 100 x . Minute etch pits, not more than a few microns in size, also appeared commonly as dark spots on other- wise smooth white faces growing at very low excess pressures. Often the etch pits were found gathered together in small isolated groups, within which the pits appeared to be distributed randomly. The pits would frequently appear first near the cold, lower edge of a face and gradually move upward as the crystal face advanced. GROUP OF ETCH PITS MEAN POSITION OF WAVE FRONT TRAIL OF ETCH PITS MOTION OF FACE LIMIT OF SPECULAR FACE Fig. 10. Interactions between visible steps and thermal etch pits. (a) Macroscopic steps being "shunned" by a group of etch pits; (b) parallel trails of etch pits formed in the wake of an isolated macroscopic step. Occasionally, etch pits and steps were seen together and their interactions could be studied. One instance of interaction occurred on a specular face containing etch pits in an isolated group in the upper region of the face with "slow" steps propagating toward them from below (see fig. 10a). As the steps approached within about 10 urn of the pits, they appeared to be repelled by the pits, forming a wave front which vibrated rapidly about a mean position shown approximately by the dashed line in fig. 10a. On the same face, but after several seconds had elapsed since the passage of any visible steps, a dark "slow" step appeared and moved up the visually smooth face, leaving in its wake trails of etch pits all parallel to one another and normal to the step which appeared to create them (see fig. 10b). 7. Discussion of the mechanism for growth The appearance of etch pits on low-index faces has of- ten been considered indicative of the presence of disloca- tions which emerge at the crystal surface, since lattice molecules in the immediate vicinity of the emergent sites tend to be relatively highly stressed and so evap- orate preferentially13). If so, then the presence of a 211 30 D. LAMB AND W. D. SCOTT screw component in one or more of the dislocations, emerging perhaps on the face depicted in fig. 10a, would create steps which would propagate outward and interact with "slow" steps approaching from below. Meeting head-on, the two sets of steps would mutually "annihilate" each other at positions which depend on the local fluctuations in the velocities and frequencies of generation of the steps. These observations point toward the importance of steps and their sources in the mechanism by which vapor molecules become incorporated into the lattice of ice on low-index faces. In principle, measurements ot linear growth rates as a function of excess vapor pres- sure are valuable for providing insight into the nature of the growth mechanism but the large scatter in the data makes the present results inconclusive. Never- theless, the data shown in fig. 5 show that the growth rates of both faces were often found to vary quadratic- ally over some range of excess pressure. This overall observation is at least consistent with a mechanism based on the rotation of spiral steps, as postulated by Frank14). The distinction between purely quadratic trends (figs. 5c and 5d) and quadratic-linear trends (figs. 5a and 5b) is very likely a result of an insufficient data range. It is apparent that a parabolic trend cannot be continued indefinitely toward high excess pressures or the fraction of the total number of molecules impinging on the crystal surface which actually in- corporate into the lattice would eventually exceed unity. The high variability in the growth rate stiown in fig. 5e may be due to spurious sources of steps, perhaps introduced where the crystal face intersects the substrate, or to significant changes in the defect structure of the growing surface with time. Upon comparing the temperature dependence of the linear growth rate G of the basal face presented in this study (fig. 4a) with that of the velocity v and interaction distance 2xs of steps measured by others (fig. 1), one notes a general similarity in the overall form of the curves. The analogous experimental curve (fig. 4b) for the prism face shows a similar trend which is shifted about 7 °C toward colder temperature. The observed growth on the two crystalline faces (figs. 4a and 4b) does, indeed, predict the observed primary habits of ice. But, as mentioned, there are no satisfactory ex- planations for the trends with temperature. Never- theless, a comparison of these results through the qualitative proportionality that exists between the three trends in xs, v, and G provides insight into the likely mecanism of the crystal growth itself. At the first level of comparison, the proportionality between .v5 and v strongly suggests that the velocity at which a step can propagate is limited by the distance over which the step can gather adsorbed molecules by a transport mechan- ism based on surface diffusion. This idea was presented sometime ago by Hallett5) who found the step velocity to be inversely proportional to step height. Similarly, at the second level of comparison, the linear rate G at which a face advances as a whole seems to be limited by the velocity of the steps which are the elementary units of growth. For steps of a given average height, how- ever, it can be shown that the linear growth rate should be dependent only on the frequency at which the steps pass a given point, not on their velocity. Thus, there must exist an intimate link between the velocity and the generation rate of steps. One mechanism of growth which does provide a simple relationship between velocity and generation frequency is the spiral step mechanism already invoked as a possible explanation for other observations in this work. Hence, it seems reasonable that the basic growth mechanism of ice on the basal and the prism planes involves the impingement and adsorption of vapor molecules onto a surface containing emergent screw dislocations and spiral steps; the random, but limited, migration of the adsorbed molecules across the surface; and their final incorporation into the ice lattice at the steps. Although such a spiral step has been used to explain the growth of numerous substances, it has not been considered seriously with regard to ice growing from the vapor phase. One must naturally expect some deviation from the traditional theory12), since, as already pointed out, the temperature trends themselves are not explicable in terms of it. But, the theory ^s generally presented does not allow for variations with temperature in the structure of the growing surfaces. The magnitude of the step catchment distance xs, for instance, is certainly expected to be sensitive to the energetic structure of the surface and is also a most important parameter in the growth mechanism. A more complete theory should bring our concepts of the phys- ical processes at work on the growing surface into better quantitative agreement with the experimental data. 212 LINEAR GROWTH RATES OF ICE CRYSTALS GROWN FROM THE VAPOR PHASE 31 8. Conclusions (1) To a first approximation, the effects of the en- vironment on the growth of an ice crystal can be sep- arated from the surface kinetics. This means that there appears to be no effect of sorbed air on the growth properties themselves. (2) The linear growth rates of the basal and prism faces were found to vary with temperature in a peculiar manner, each having a local maximum and a local minimum. The temperatures at which the extrema occur on the prism face are shifted relative to those of the basal face by about 7 °C. The trend in the linear growth rates of the basal face follow the trends of previous measurements of the velocity and interaction distance of steps on the basal face. (3) Steps are an important and probably essential feature of ice crystal growth. It appears likely that the steps originate at the emergence of screw dislocations on the surface. The basic mechanism of spiral step growth may apply to the growth of the basal and prism faces, but the traditional treatment would have to be expanded to account for possible changes in the structure of the low-index surfaces with temperature. Acknowledgements The experimental work was performed at the At- mospheric Sciences Department of the University of Washington, Seattle, USA, under grants GA-11250 and GA- 17381 from the Atmospheric Sciences Section of the National Science Foundation. One author (D. L.) would like to acknowledge additional support from an NDEA Title IV Fellowship. The paper was written while both authors were associated with the National Hurricane Research Laboratory, NOAA, Miami, USA. This is Contribution No. 248 from the Atmospheric Sciences Department^ University of Washington. References 1) N. Nakaya, Snow Crystals: Natural and Artificial (Harvard Univ. Press, Cambridge, 1954). 2) C. Magono and C. W. Lee, J. Fac. Sci., Hokkaido Univ. 7 (1966) 2. 3) D. Shaw and B. J. Mac»-0w— < Jgjgf, • >g THINDAO- HAWAII (t-C) - I * • r '- / b v* ; - *s. i • -ir ; •■>-■ ■» !*• *V >fe k ■4- " pari 1 «!■ H~flel4 Afttewwi HypoboBc v gggg *Ko««t lor T*U*M&-ra«STMfttP4 V . . .v ». i. ■ t- * H ■*-•■• *% . * , * ! E- fit 14 Anttnwo HypoboKc Difftr«K«-Ome«« phote* for TKINDAD- HAWAII fc-C) 3^ J iflinii^nsrL; ■ L I' Kill ' i jLi^iiiaiuiui-LUi •Dhmii .*...*..*«*, S,JmX *JHJu ii*-» .^^ *,.JMIkjL.a.-*. . 1l^A#^ a.ajLBBBl._ZJBl^I.' '/ -~— A B C D Figurt 3. Rustrak Recordings , Aircraft of on altitude where air temperature it 00* C in thunderstorm. Hurricane Fern was penetrated by a DC-6 while it was in the western part of the Gulf of Mexico on the night of September 7, 1971. The aircraft with the H-field antenna was utilized for the storm reconnaissance. This storm had winds up to 70 knots, it was turbu- lent, there was rain but not of a particularly heavy nature, and the lightning was very in- tense in all quadrants. The aircraft was at 3000 feet during the entire mission and no precipita- tion static was recorded by the Omega Naviga- tion System's Rustrak recorder. Thunderstorm Experiment The above observations suggest that a con- trolled experiment should be designed to use one DC-6 aircraft with one Omega receiver attached to an E-field antenna and another attached to an H-field antenna. Each receiver would have its own Rustrack recorder and both recording on 10.2 Khertz. A flight plan was designed to make penetrations into a thunder- storm at altitudes where the temperatures were well above, near, and below freezing respec- tively. Several penetrations would be made at each level in order to determine if precipitation static would repeat itself in intensity. On the afternoon of October 5, 1971, a suitable thunderstorm (Fig. 1) was located in a tropical air mass over land near Naples, Florida. The flight level where the temperature was 8°C was selected for the first penetration. The Rus- trak tapes for the first and second penetrations are reproduced in Fig. 2. 217 Vol. 1 9, No. 2 Lubin & Lewis: Effects of Weather on Omega 179 TUm&Afi - HAWAII tfc-£l The Rustrak recording Trinidad-Hawaii (B-C) at the top of the figure which is the H-field antenna printed poorly. Note that no precipi- tation static took place at the beginning of the traces for stations Trinidad-Forestport, N. Y., (B-D) on the H-field antennas and on the two traces for the E-field antennas. At point "A" in Fig. 2, precipitation static begins to appear in both traces for the E-field antenna, that is, for stations Trinidad-Hawaii (B-C) and Trini- dad-Forestport, N. Y. (B-D). At point "B" the aircraft executed a 180 degree turn under the thunderstorm anvil cloud and repenetrated the cloud. It is clear that the width of the trace is considerably broader for the E-field traces before and after penetration as well as those shown for the above H-field antenna. After finishing the lower level investigation, the aircraft climbed to an altitude of near freezing temperatures. Fig. 3 shows precipita- tion static which occurred on penetrations be- tween points "A" and "B", as well as points "C" and "D" for the E-field traces. Again it is noted that the H-field traces continue to mark the lanes. The aircraft was then taken to an altitude where the temperature was — 7°C. At this alti- tude the cloud was not as dense as at lower alti- tudes. However, the aircraft did make a round- trip penetration in the cloud, each pass lasting for at least two minutes. The effect of precipita- tion static on the E-field antenna is shown in Fig. 4. The lane count is completely destroyed between points "A" and "B" as well as points 218 180 Navigation Summer 1 972 "C" and "D" as it was at the lower levels. The H-field antenna showed no effects of precipita- tion static at this level. Conclusion It is concluded from the above described flights into weather at mid-latitudes and trop- ical systems that Omega Navigation Systems equipped with E-field antennas will yield poor navigation results where severe weather is encountered. Further, if the aircraft is operating at altitudes where the temperature is below 0°C temperature, it is almost a certainty that an Omega System will lose its reception of the signals for extended periods in clouds. The experiment described clearly indicates that precipitation static can be controlled to a large degree by use of an H-field antenna with an automatic crossed-loop switching device con- nected to the aircraft compass system. However, skin mapping of the aircraft prior to installing the H-field antenna system greatly improves the chances of the system operating successfully. Each aircraft, even though the same model, must be skin mapped individually. The installation of null dischargers did not prevent the loss of Omega signals using an E-field antenna under extreme weather conditions. The lack of the complete loss of Omega signal as shown in Fig. 2 may have been due to the presence of the Granger Associates dischargers. References 1. Amason, M. P. and J. T. Kung, Lightning Cur- rent Transfer Characteristics of the P-Static Dis- charger Installations, Douglas Paper 5789, Douglas Aircraft Company, page 12, December 1970. 2. Eisenberg, R. L. and M. F. Williams, The Flight Performance of the NRL Mark HI Omega Air- craft N avigation Set, NRL Report 7004, Systems Integration and Instrumentation Branch Com- munications Sciences Division, Naval Research Laboratory, Washington, D. C, page 76, Octo- ber 1969. 3. Granger, John V. N., Precipitation Static Pro- tection for Jet Aircraft, Granger Associates, Palo Alto California, page 25, 1968. 4. Nanevicz, J. E. and R. L. Tanner, Some Techni- ques for the Elimination of Corona Discharger Noise in Aircraft Antennas, Proceedings of the IEE, pages 53-64, January, 1964. 5. Tanner, R. L. and J. E. Nanevicz, An Analysis of Corona-Generated Interference in Aircraft, Proceedings of the IEEE, pages 44-52, January, 1964. Notice The National Hurricane Research Laboratory, the Research Flight Facility, NOAA, and the Federal Aviation Agency do not approve, recom- mend, or endorse any proprietary product or proprietary material mentioned in this publica- tion. No reference should be made to either or- ganization or to this publication in any adver- tising or sales promotion which would indicate or imply that the National Hurricane Research Laboratory, the Research Flight Facility or the Federal Aviation Agency approves, recommends or endorses any proprietary product or material mentioned herein, or which has as its purpose an intent to cause directly or indirectly the advertised product to be used or purchased be- cause of this publication. 219 23 U.S. DEPARTMENT OF COMMERCE National Oceanic and Atmospheric Administration Environmental Research Laboratories NOAA Technical Memorandum ERL NHRL-99 ON THE ROLE OF THE ORGANIZATIONAL PERIOD IN THE NHRL CIRCULARLY SYMMETRIC HURRICANE MODEL Michael S. Moss Stanley L. Rosenthal National Hurricane Research Laboratory /*\/°*% Coral Gables, Florida f 0§TT " March 1972 % Pfj A 220 TABLE OF CONTENTS Page ABSTRACT 1 1. INTRODUCTION 1 2. EXPERIMENTS ALTERING TANGENTIAL WIND, RADIAL WIND, AND POTENTIAL TEMPERATURE 7 3. EXPERIMENTS ALTERING SPECIFIC HUMIDITY 12 k. SUMMARY AND CONCLUSIONS 20 5. ACKNOWLEDGMENTS 21 6. REFERENCES 22 221 ON THE ROLE OF THE ORGANIZATIONAL PERIOD IN THE NHRL CIRCULARLY SYMMETRIC HURRICANE MODEL Michael S. Moss and Stanley L. Rosenthal The NHRL seven-level, circularly symmetric hurricane model is used in an attempt to ascertain the processes by which a weak tropical cyclone undergoes rapid development into a hurricane. Experimental results reveal that rapid development is primarily a manifestation of the formation of a deep moist layer. 1. INTRODUCTION Numerous investigations (e.g., Riehl, 195*0 have revealed that an incipient, relatively weak tropical cyclone must undergo a period of organization before rapid development into a hurricane. Due to sparse and uneven distributions of data over the tropical oceans, it has been difficult to determine the vortex structure changes that occur during this time and that render a vortex unstable. By controlled experimenta- tion with numerical models, we may perhaps obtain a better understanding of this problem. It should be stressed that these models are highly idealized theoretical tools, and only very general qualitative compari- sons can be made with real hurricanes. Therefore, the model results presented in this paper are tentative and represent some sort of "average" or "typical" hurricane. The model used is the National Hurricane Research Laboratory (NHRL), seven-level, circularly symmetric hurricane model (Rosenthal, 1 970) - The model base state temperatures, potential temperatures, and relative humidities are nearly equal to those of the mean hurricane season (Hebert and Jordan, 1959) and are listed in table 1. 222 Table 1. Base state values of height, pressure, temperature, potential temperature, and relative humidity . Level Height (m) P (mb) r (°K) e (°K) Relati ve Humi di ty (*) 1 0 1015.0 301.3 300 90 2 1054 900.4 294.1 303 90 3 3187 699.4 282.6 313 54 4 5898 499.2 266.5 325 44 5 9697 299.2 240.8 340 30 6 12423 199.5 218.9 347 30 7 16621 101.1 203.1 391 30 The initial temperature field is specified by fi,j =fi +T* jcos (f) rj + if sin (f-) z. , (1) where i and j are the height and radial indices, respectively, Tj is the base state temperature, T* = 0.16°K, r = 440 km (the radial limit of the computational domain), and z is the height of level seven. The initial pressure at level seven is taken to be the standard, base state value (table 1), and the hydrostatic equation is integrated downward to obtain the remainder of the initial pressure field. The gradient wind equation is then solved to obtain the initial tangential wind while the radial and vertical motions are initially zero. The initial specific humidity is horizontally uniform and equal to that of the base state. 223 The equations of motion and the thermodynamic equation are, respect i vely , 3M .. _.. 9M _ 9M _ , 1 9 .- 3Mv at u IF WI7 fru+ = 3l (pKz ^ p + ^11_ jr3i_ (Vj| (2) r 9r i 9r VT 9u _ 9u 9u M ir H > fl 34 , 1 3 /—„ 9us IF" ""F^ 97 +7(f + 7-)-e# + -a7(PKz 37) r2 9r V 9r vH' and 9t 9r 9z ' r 9r 9r -p 9z *• 9z ) Other pertinent equations include the hydrostatic equation, ££ - £ (5) 9z e ' a form of the continuity equation given by 9pw 1 9 /- N (6) 9z r 9r K wi th ♦ - S, (^-)R/CP > (7) K Po and 6•• zi and parcel ascent from the level z. is conditionally unstable, convection will originate from that layer. The distinguishing aspect of this model is the parametric represen- tation of the diabatic heating (Q) produced by organized systems of cumulonimbi. Thhs parameterization is achieved by a convective adjust- ment that may be generalized as follows. Suppose that existing cumulonimbi transport upward and condense a certain amount of water vapor in some period of time. Part of this condensate will be reevapo- rated and enrich the macroscale humidity as clouds dissipate. The remaining condensate falls as rain, and thus provides latent heat to the macroscale system. The reader is referred to Rosenthal (1970) for a detailed explanation of the convective adjustment. 5 226 The model storm's evolution is shown in figure 1, on which the maximum surface wind speed is plotted against time. During the "organizational period" (from 0 to 168 hours), a wind minimum occurs at 120 hours. From 120 to 192 hours, a gradual increase in intensity is observed and rapid development commences at 192 hours. By restoring, at selected times, a dependent variable to its initial value we may isolate the relative importance of its variations during the period involved. Therefore, experiments were performed in which at hour 120 or at hour 192 of the model storm the tangential wind, radial wind, potential temperature, and specific humidity were each replaced by their respective initial values. Table 2 lists the various experiments. ~ 50 1 1 1 1 1 1 1 1 m - ■ O 1 *-~— ^^ Ul • • CO J v. 1 2 1 CONTROL / O 40 — UJ / Ul / 0. / CO / o / 5 30 / - * / UJ J o J < 1 u. „^ j C 20 I - z> J CO 2 z> 1 * 10 I - X ^** < v ^__-_^*^^"^ 2 ^r — ^* I 1 1 ... 1 1 1 1 1 48 96 144 192 240 TIME (HOURS) 288 336 384 Figure 1. Maximum surface wind as a function of time for the control experiment. 6 227 G2 no change G3 Do. Gk Do. G^tA Do. Table 2. Experiments (G-series) in which changes in the data field are introduced at 120 hours of the control experiment. Exp.* Tangential Radial Potential Specific Wind Wind Temperature Humidity Gl restored to no change no change no change initial val ues restored to Do. Do. initial val ues no change restored to Do. initial values Do. no change restored to initial values Do. Do. restored to initial values above boundary layer (>105*t m) ; no change in boundary 1 ayer "The H-series experiments are the same, except that the changes are intro- duced at 192 hours of the control experiment. 2. EXPERIMENTS ALTERING TANGENTIAL WIND, RADIAL WIND, AND POTENTIAL TEMPERATURE Experiments Gl and HI (fig. 2) reveal that replacing the tangential winds with initial values alters the rate of development but not the ultimate intensity of the model storm. Since the initial tangential winds are greater than those at 120 hours, frictional convergence of water vapor is increased (Rosenthal, 1971; Rosenthal and Moss, 1971) in experiment Gl , and development is more rapid than in the control. For experiment HI, the delayed development is explicable by the reverse of the arguments cited above. 7 228 These results are consistent with previous calculations. An experi- ment in which the initial winds were twice those of the control (Rosenthal, 1970 revealed that the vortex development was characterized by a much shorter organizational period but with only a small deviation in peak intensity from the control. Anthes et al . (1971) cite similar results obtained from other models. 50 1 1 — 1 1 I i r 1 ^■^ / / O « / * UJ i / • CO ' / • z « Gl AND HI ' / ° / • - ^^ / ° / ° o / ° UJ / • UJ ,' / Q. i f CO 30 ' / * ' / • — O « / o z ' / ° * ' / • ' / • UJ ' / • U 20 - 1 1 if © - < / / / ° U. and experiment HI circles, 8 229 The absence of transverse circulations at the start of experiments G2 and H2 does not significantly disrupt the storm's development (fig. 3) The generation of frictional inflow occurs very rapidly and is primarily determined by the strength of the tangential wind (Rosenthal and Moss, 1970- The mechanism by which this occurs may be demonstrated as fol lows . 144 192 240 TIME (HOURS) 288 336 Figure 3. Maximum surface wind as a function of time. Control data are solid lines 3 experiment G2 dashed lines, and experiment H2 circles. 9 230 In the absence of radial motion, (2) and (3) at the bottom surface become1 3M 1 3 ,- 3Mv KH 3 t 3 9 /M \l (11) and 3M . 1 3 ,- 3Mv H 3 jj ,M_\ l iH-il (f + M- ) -eii . (12) 3t r r2 3r Surface drag in (11) quickly produces significant negative angular momentum tendencies, and the rotational forces in (12) are, therefore, decreased. In the model, therefore, this tends to produce — < 0, and 0 t radial inflow is established.2 Qualitatively, through the neglect of inflow, we have increased the effect of surface drag on the angular momentum tendency, and this then allows inflow to redevelop quickly. Experiments G3 and H3 reveal that slight alterations of the poten- tial temperature do not result in significant modifications of the development of the hurricane (fig. k) . 3 The implication is that the changes in temperature configurations during the organi zational period are not particularly important for the onset of rapid development. Yanai (196*0 observed that development of a strong warm-core in the uppei troposphere occurred during the period of rapid intensification. Experi- ments G3 and H3 appear to be consistent with this observation. iThe vertical motion, w, is equal to zero at the lower boundary. 2The reader is referred to Rosenthal (1970) for an explanation of the computation of the surface pressure gradient force in the model. 3The initial vortex is generally a few degrees cooler than is the vortex when these experiments are started. 10 50 I 1 1 1 1 1 1 T " • — . 1 1 O UJ C/> / 7 1 J 1 I 1 I ^40 _ G3 AND H 3 ' r 1 k - o UJ UJ a 1 /» / /• ' 1° ' P O 30 z " 1 to SURFACE o - 1 /o ' / ° //: '/ o // • • 2 z> 2 10 X < 2 ft 1 1 1 • 1 1 48 96 144 192 240 TIME (HOURS) 288 336 384 Figure 4. Maximum surface wind as a function of time. Control data are solid lines, experiment G3 dashed lines, and experiment HZ circles, Although the temperatures in the upper troposphere are generally slightly decreased at the start of experiments G3 and H3, the partition- ing of released latent heat between the lower and upper troposphere is also changed (see Rosenthal (1970) for details); therefore, relatively more heat is applied to the upper troposphere. This then tends to counteract the alterations in temperature imposed by the experimental procedure. 11 232 3. EXPERIMENTS ALTERING SPECIFIC HUMIDITY Careful inspection of the control experiment during the organiza- tional period shows that the changes in humidity are considerably more impressive than are the changes of other dependent variables. Figures 5, 6, and 7 illustrate the evolution of the distribution of relative humidity HOUR -0 .0 RELATIVE HUMIDITY(PCT) CONTROL o o "T 1 1 i I I QQ § ^^ CM - - « — - E 300 - C£ ZD CO o 40 40 m . 40__ 40 — CO s LU C£ Q_ § Rn Rn rr Rn Rn fin Rn fin An fin o CD - in o L_l_ l_._ L 1 1 L 1_ _L . J_ 1 1 1 25 45 65 85 105 125 145 165 185 205 225 RADIUS (KM) Figure 5. Cross section of relative humidities at 0 hours of the control. Isohumes are in percent. (Ordinate values of -pressure are assumed to correspond to the heights given in table lj and are so scaled. This assumption is approximately valid during organization because the pres- sure changes at a constant height are minimal.) 12 233 HOUR =120.0 RELATIVE HUM I 0 I TY ( PCT ) CONTROL CD C£ ZD CO CO l_U C£ Q_ 25 45 65 85 105 125 145 165 185 205 225 RADIUS (KM) Figure 6. Cross section of relative humidities at 120 hours of the control. Isohumes are in percent. (Ordinate labeling same as fig. 5 J for the control. Initially, the humidity profiles are nearly horizontally uniform.1* At 120 hours, the depth of the moist layer has increased and ■♦The initial ("as compared with base state) relati ve humi di t ies are calcu- lated from the ratio of the horizontally uniform specific humidities to the saturation specific humidities derived from (1). This explains the small deviations from horizontal uniformity. 13 234 saturation (to be interpreted as stratus-type cloud) is evident in some regions of the middle troposphere. By 192 hours, the moisture shows further increases, and the middle tropsopheric region of saturation is more widespread. This sequence is consistent with Riehl's (195*0 hypothesis that the principal role of the organizational period in real storms is to form a deep moist layer from an initial moisture distribu- tion that is relatively dry aloft. HOUR =192.0 RELATIVE HUM I D I T Y ( PCT CONTROL ^\oO- M 1 IMX \ / nut \ o \ -ScP- \ ma I — '00-"" IMX MI mu KM too in rix niN 100 .100 100, 80_ .100 25 45 65 85 105 125 145 165 185 205 225 RADIUS (KM) Figure 7. Cross section of relative humidities at 192 hours of the control, Isohumes are in percent. (Ordinate labeling same as fig. 6.) 14 When the specific humidity is restored to its initial value in experiments Gh and H*t, there results a significant modification in the storm development (fig. 8). Experiments G^A and H^A (where the specific humidity is only altered above the boundary layer) reveal that the response of experiments G*t and Hk is due mainly to the deletion of moisture above the boundary layer (cf. figs. 8 and 9). Rosenthal (1970) found that when the initial relative humidity was 90 percent everywhere, the model storm developed more rapidly with only a very short organiza- tional period. 50 1 — 1 -|— -1 1 1 — 1 1 ,_^ O e UJ o CO 0 J 40 G4 AND H 4 / / ° / ° o o o o - O / • UJ / ° UJ / ° / ° 0. / ° en / ° O 30 / ° / ° - z / ° £ i e / ° UJ / ° o / ° if 20 / ° / ° - (T / ° O / ° V) / ° / ° 2 / ° / ° 3 / ° 1 10 1 o - X •-. < Oo„oooo 3E ^ — 1 1 1 1 J -1 ' x 48 96 144 192 240 TIME(HOURS) 288 336 384 Figure 8. Maximum surface wind as a function of time. Control data are solid lines, experiment G4 dashed lines 3 and experiment H4 circles. 15 236 In the real atmosphere, the establishment of a deep moist layer during organization implies that cumulus clouds may entrain relatively moist air making their net condensation greater and their growth to cumulonimbus more likely. Although model clouds are undilute, the partitioning of condensate between precipitation and reevaporat ion in a convective column depends on the difference between the cloud and environmental humidities (Rosenthal, 1970). With a dry upper level S 50 to O 40 UJ UJ Q. V) 30- UJ O £ 20 CK 3 CO iio 2 x < 2 G4A and H4A » » « ,MM""«»"MM«»I«I«.«„„„H1|„ 48 96 144 192 240 288 TIME (HOURS) 336 H40o UJ UJ a. to - 30Q z 50 20 UJ a < u. cr CO 10 1 x < 2 384 Figure 9. Maximum surface wind as a function of time. Control data are solid lines3 experiment G4A dashed lines, and experiment H4A circles. 16 environment, a greater proportion of the condensate is reevaporated and less of the condensate is realized as rain (latent heat). The rainfall rates for the nondevelopi ng experiments are considerably less than the control (figures not shown), confirming that less latent heat is avail- able to the model vortex in the absence of a deep moist layer. At the start of experiment G4 , a deep moist layer is lacking (fig. 10; note the similarity to fig. 5), and the tangential winds are HOUR =120-0 RELATIVE HUM I D I TY ( PCT ) EXPT G4 105 125 145 165 185 205 225 RADIUS (KM) Figure 10^. Cross section of relative humidities at 192 hovers of experiment G4. Isohumes are in percent. (Ordinate labeling same as fig. 5.) 17 238 smaller than those of the control at the initial instant. Hence, the failure of the storm to develop during the time span of the experiment (6 days) is not surprising. A deep moist layer is also lacking at the start of experiment H^ (fig. 11). However, in this case, rapid development does occur. Here, the tangential wind is greater and fric- tional inflow is quickly produced. Root-mean-square (RMS) deviations of HOUR =192.0 RELATIVE HUM I D I TY ( PCT ) EXPT H4 r 25 45 65 85 105 125 US 165 165 205 225 RRDIUS (KM) Figure 11. Cross seaticn of relative humidities at 192 hours of experiment H4. Isohumes are in percent. (Ordinate labeling same as fig. 5.) 18 239 the relative humidity from the control at 192 hours and the maximum surface winds for experiment Hk are depicted in figure 12. 5 Rapid development roughly corresponds to a minimum in the humidity deviations, indicating that such development is coincident with the formation of a moisture field compatible with that of the control (at 192 hours). 204 216 228 240 252 TIME (HOURS) 264 276 Figure 12. Maximum surface wind and RMS relative humidity deviations from 192 hours of the control as a function of time for experiment H4. Maximum surface winds are shown by solid lines 3 and RMS devia- tions by dashed lines. 5The initial boundary layer humidities are relatively moist Ctable 1). Therefore, the percent of humidity changes is very small throughout the storm's 1 i f e cycle . 19 240 k. SUMMARY AND CONCLUSIONS During the "organizational period" of the model hurricane, friction reduces the gradient winds at the surface to subgradient values. This induces low-level inflow that results in the horizontal convergence of water vapor and, therefore, to cumulus convection. The condensate in these cumuli is largely evaporated into the dry upper level environment. There is little rain, but the macroscale humidity increases. Eventually, the troposphere becomes extremely moist and large amounts of latent heat are imparted to the upper troposphere, thus inducing the development of a warm core. Surface pressures are decreased, especially towards the storm center where the upper level heating is most pronounced. The horizontal convergence of water vapor is concomitantly increased, implying more vigorous convection that ultimately results in rapid development. The numerical experiments discussed here indicate that rapid development can only occur after the deep moist layer has been produced, and that the changes that occur to other meteorological variables during the organizational period are important only to the degree that they feed back on the moisture field. A more intense vortex results in more rapid development, because the horizontal convergence of water vapor develops more rapidly; therefore, less time is required to create large moisture contents to great heights. 20 241 5. ACKNOWLEDGMENTS Computations were performed at the NOAA computer complex, Suitland, Maryland. Access to the computer facility is via a terminal located at NHRL in Miami , Florida. 21 242 6. REFERENCES Anthes, R. A., S. L. Rosenthal, and J. W. Trout (1970, Preliminary results from an asymmetric model of the tropical cyclone, Monthly Weather Review, 99 No. 10, Ikk-Jid . Hebert, P. J. and C. L. Jordan (1959), Mean soundings for the Gulf of Mexico area, National Hurricane Research Report No. 30, (U. S. Depart- ment of Commerce, National Hurricane Research Laboratory, Miami, Flori da) , 10 pp. Riehl, H. (195*0, Tropical Meteorology (McGraw-Hill, New York) 392 pp. Rosenthal, S. L. (1970), A survey of experimental results obtained from a numerical model designed to simulate tropical cyclone development, ESSA Technical Memorandum, ERLTM-NHRL 88, (U. S. Department of Commerce, National Hurricane Research Laboratory, Miami, Florida), 78 pp. Rosenthal, S. L. (1971), The response of a tropical cyclone model to variations in boundary layer parameters, initial conditions, lateral boundary conditions, and domain size, Monthly Weather Review, 99 , No. 10, 767-777. Rosenthal, S. L. and M. S. Moss (1971), The response of a tropical cyclone model to radical changes in data fields during the mature stage, N0AA Technical Memorandum, ERLTM-NHRL 96, (U. S. Department of Commerce, National Hurricane Research Laboratory, Miami, Florida), 18 pp. Yanai , M. (196A), Formation of tropical cyclones, Reviews of Geophysics, 2, No. 2, 367-it1*t. 22 243 24 Reprinted from Second Symposium on Meteorological Observations and Instruments, March 27-30, 1972, San Diego, California, 205-216. AN ASSESSMENT OF THE PRESENT INSTRUMENTATION FOR THE MEASUREMENT OF CLOUD ELEMENTS AND OUR NEEDS William D. Scott National Hurricane Research Laboratory, NOAA Miami , Florida il. INTRODUCTION Instrumentation for the measurement of cloud elements has been and is inadequate, and has undergone only slight improvement in the past fif- teen years. This is primarily because the instru- ments have been mounted on aircraft, and there are many difficulties in acquiring data on a moving platform. At this time, our numerical modeling of cloud dynamics and microphysics has developed to the point where further improvement is, in part, limited by our instrumentation and data retrieval techn iques . Fortunately, instruments are now emerging from the developmental stage which offer hope for upgrading our instrumentation and improving the data retrieval problem. These instruments are generally capable of measuring and displaying cloud properties in real time with some degree of in-flight data processing. Generally, the instru- ments rely on optical techniques in which the laser plays a role, use state-of-the-art solid- state electronics, or incorporate recent innova- tions. In addition, the instruments are designed to acquire samples with reasonable statistical quality and to avoid tedious data reduction. Unfortunately, it appears that the cloud hydrometeors that are of the most importance in modeling of the precipitation process and the as- sessment of our weather modification attempts have generally been neglected in the instrument devel- opment. The two areas of immediate concern are 1) the measurement of the "tail" on the cloud droplet distributions and raindrop concentrations and 2) the measurement of the number and character- istics of ice crystals. At present, it appears that several instruments of reasonable quality can measure cloud droplet concentrations and their distributions, but there are no operationally re- liable instruments for the real time measurement of ice crystals or drops with sizes above 50 u. However, other parameters of perhaps more importance have been neglected in our studies. In fact, it appears that well documented case studies for the evaluations of our modeling efforts are almost non-existent, despite the num- erous field operations that are underway each year. Specifically, I refer to qualitative and semiquantitative information such as photographs, cloud base height and updraft velocity, presence of precipitation below cloud, and radar informa- tion. In most cases an appropriate thermodynamic sounding for the cloud environment is unavailable. In the following development, I hope the basis for the above statements will become apparent . §2. CLOUD MODELING Using data outout from the cloud model developed at NCAR by Danielsen et a). (1972) we can see the necessity for accurate measurements of cloud elements. Fig. la shows the actual radar return from a northwest Colorado hail storm in June, 1970. Fig. lb shows the calculated radar ref lecti vi ties . — 1 r RADAR ECHO BOUNDARY x 10 12 14 TIME (mm) Fig. 1. Measured (a) and calculated (b) Radar Reflectivities (Courtesy of E. Danielsen) . Note that Fig. ]a is a spacial profile whereas Fig. lb is a temporal profile. But, nonetheless, the simulated storm shows realistic qualities. From these data and the internal structure of the 205 244 model, Danielsen et a), were able to conclude that the radar return was primarily from the ice or hail in the storm. But the radar reflectivity was calculated from the particle sizes using Mie scattering theory. For smaller sizes and larger wavelengths, the radar reflectivity is dependent on the sixth power of the particle size. Presen- tly, our most accurate, though tedious, techni- ques cannot measure hydrometeor sizes with an accuracy better than about 20% ami, even so, the sample is only marginally significant. The model appears to be producing a more realistic result that can be justified experimentally. The model can be grossly altered by relatively small changes in the cloud droplet spectrum. Fig. 2a illustrates this, in terms of the precipitation development, for the cloud drop- let distributions shown in Fig. 2b. Fig. 2 Precipi tation Development (a) for initial droplet distribution (b) (Courtesy of E. Danielsen) . CATEGORY 0 5 10 13 20 25 30 55 40 9 TiK B iSi s 7 ■{< >»< 6 i hS 5^ ~\y \^ 1 * ^ k ■S Tu ' ' f ». «---. : i V l ^ v' ' p - s ^ £ o V V1" ^^ / .. *.L - .j v_, ;.___._2 -?> , '! x~ i"" :~ij£".:"Ai. -*■-- + , >;- -•} | "V "! i "::2:::::::::::j ; i:.as.st< 5 ..ILlLlLLIiiiilLJllLlLLlI 50 100 250 500 lif DROP RAOII rfrrr m ^i rr-rr i 1 1 1 C25 5 4 5 6 75910 IS 20 50 25 3 4 5(78910 15 20 ,11 I I I Mil I I I I I II II I I I I fill I II 1 1 I I I II fill I III 2.5) 4 3678910 13 20 30 25 3 4 3 678910 IS 20 30 RADIUS [yuml The times after the initial cloud formation are respectively, 22.5, 2k, 32, and 39 minutes for distributions Dl , D2 , D3 , and M. Since the or- dinate is logarithmic, several orders of magni- tude more precipitation particles are developed when D*4 is the initial cloud droplet distribution than when Dl is the initial cloud droplet distribu- tion. This is reasonable since the initial pre- cipitation embryos are already present in D^t , but in Dl these embryos must form by condensation and coalescence before precipitation-sized particles can develop. The effect of a larger cloud base updraft velocity is similar to removing the tail on the distribution. The air is simply whisked up through the area of warm cloud development be- fore precipitation can develop. The ice phase, however, and its develop- ment are of even more significance. Whether a given cloud will glaciate, or turn to ice, cannot be determined a priori at this time. In fact, the concept of particles that initiate the formation of ice in a real cloud is presently being contes- ted. In most cases we cannot predict within three or four orders of magnitude how many ice crystals will form in a given air parcel. As a result, the development of the colder regions of clouds re- mains unspecified and our present attempts at weather modification are, for the most part, guesses. We need more measurements of ice crystal concentrations, types and sizes as well as meas- urements of the aerosol structure. In summary, measurements needed to up- grade our modeling efforts are listed in Table 1. TABLE 1 I I BASIC MEASUREMENTS NEEDED FOR CASE STUDIES OF CUMULUS CLOUDS: A. Complete thermodynamic sounding, in near space and time B. Cloud base height and cloud base up- draft radius C. Cloud top height and tower radius as a function of ti me D. Radar echo RH I profiles as a function of time . BASIC MEASUREMENTS NEEDED FOR INPUT TO 2-D AND 3-D MODELS: A. Basic data set of I (above) Surface temperature and moisture field data below cloud base Meso or subsynoptic analysis of the kinematic properties of the wind field (flow, convergences, etc.) Cloud sizes, cloud number distributions, detailed cloud photographs Roughness characteristics for boundary layer parameterization Aircraft or Doppler wind field measure- ments PPI echo displays Dense rain gauge measurements. DETAILED MICROPHYS ICAL MODELING REQUIRES EVEN MORE DETAIL: A. Detail of the aerosol structure, CCN , Aitkin particle concentrations. Chemical composition, sizes, and physical shape Cloud base droplet concentrations and distributions or some indication of the distribution form and tail C. Multi-level hydrometeor distribution D. Multi-level ice particle distributions and types . (Courtesy W. Cotton) II B. C. D. E. F. G. H . B. §3. INSTRUMENT REQUIREMENTS Table 2 below gives the characteristics expected of assorted cloud elements, the expected sampling requirements, the ranges of present in- strumentstand the investigators presently develop- ing instruments. 206 245 Table 2 Sampling Requirements in Different Ranges Sma I I Intermediate Large water drops cloud droplets ice crystals rain- drops cloud ice snow- flakes s ize(diameter) 1-50 y 50-200 u >200 \i expected concentration !02-103/cm3 10-103/m2 1-102/m3 vol . samp led in 1 km of f 1 ight path 10 cm3 10 liters 1 m3 samp I ing area 10" cm* 1 cm2 10 cm2 present instru- slide formvar foil ments used impactor replicator impactor J-W investigators: Dye Ryan Knollenberg Cannon Cannon Ketcham L. Nelson Rusk in L. Nelson Cunningham Radke Knollenberg Sutherland Scott Kyle Mach Kyle Joss Radke L. Nelson Cannon R. Nelson ( radar) in the raindrop region, the variations of the sampl- ing requirements with liquid water are illustrated in Fig. 3- This curve was calculated by Cunningham (1951) using the Laws and Parsons distribution for raindrops and an aircraft traveling at 100 m/sec. Note that, generally, the maximum requirement is for a larger sample size for smaller liquid water content. This reflects the fact that, for a given raindrop distribution, smaller liquid water contents require larger volumes to provide a statistical num- ber of the largest drops. This illustrates the basic problem of measuring the "tail" of a IWHUW MWU Uftt distribution. The curve of maximum sample sit is determined by the requirement that not more than one particle exist in the sample volume at a given time. §*l. A CRITIQUE OF PRESENT INSTRUMENTATION The formvar replicator manufactured by Meteorological Research, Inc., is an instrument which collects and duplicates particles on. 16 mm Mylar film with a plastic solution, about 5% formvar in chloroform. The instrument came into being through considerable effort, and today, even though it has been much improved, it requires nearly continuous care to maintain less than a desired operational ability. Because of necessary trade-offs, the in- strument design has several undesirable features, any one of which may grossly alter the quality of the data. Tha instrument was designed for measur- ing cloud particles below 500 p and has a sampling area which is about 0.1 cm2. This means that it samples about 0.01 m3 in a 1 km path, a volume insignificant for the sampling of precipitation. In practice, so much information is contained on the film that it is necessary to take spot samples of, at most, 1 in 10 frames. Hence, the practical sample volume is about 11 in a 1 km path, a volume only significant for gathering information on cloud droplets and small ice particles of high concentrations. In the measurement of cloud droplets, there is a back- ground of broken drops, broken crystals, and air bubbles that tend to be produced when the film is coated with plastic solution. The problem is enhanced when consideration is given to the col- lection efficiency and particle trajectories around the boom. Fig. ^ shows theoretical concentration factors for sample collection about a C-130 aircraft based on particle trajectories calculated by the ilACA (see report of Allied Research Associates, Inc., prepared for Air Force Cambridge Research Laboratories under AF 1 9 (CO'y- 5500, I960). .1 .» « j ■ • « t LOW MTtft MBTCMT «/•* Fia. 3. Effect of samplinq area on accuracy (courtesy of AFCRL) . mOPLIT DUHETCd tWCMOMI Fl 9- 4. CONCENTRATION FACTORS FOR PARTICLES RESULTIM FROM AIRFLOW ABOUT THE C-1SO HULL. (COURTESY OF AFCRL.) The curves are for a level aircraft. Noting the large shadow zone and concentration factors, we see that, when operated on the fuselage of a C-130 with the boom about 3 feet into the air- stream, the instrument can be grossly in error depending on the angle of attack of the aircraft and the presence of turbulence or updrafts, vari- ables in a constant state of change. Of course, since the sample slit is located at the stagna- tion point on the airfoil (the boom), a slight 207 246 change in the angle of attack could affect the aerodynamic collection efficiency, which is some- what less than one, especially for the smaller part icles . Still another problem plagues pressur- ized aircraft. That is the problem of drying the replicated crystals. The design generally uses warmed, dry, outside air which passes over the film in a heated chamber. The pressure in- side the inner portion of the instrument inside the aircraft may become slightly less than the dynamic pressure outside due to leakage from the cabin or an imbalance of the flow of drying air. But, if this happens, air will flow out through the sample slit and eliminate the sample. Another basic problem lies in the oper- ation of the instrument close to 0°C. If the in- strument is attached to an aircraft flying at 100 m/sec, there is about a 5°C warming of the leading edge due to aerodynamic heating. This tends to cause a melting of ice particles that are collected at temperatures between 0°C and -5°C. Also, as normally constructed, the instru- ment contains a large heater near the sampling port to eliminate the possibility of rime ice covering the sample hole. In the normal opera- tion, with the de-ice heater operating, it is possible to have melting of ice samples in a very cold environment. The saving point is the evap- oration of the chloroform whi ch cools the film and tends to compensate for the warming. Of course, this cooling itself may cause formation of ice on the film surface or the freezing of freshly collected water drops. Also, improper drying of the replaced crystals can melt the ice crystals prematurely. Apart from these con- siderations, the rolling up of the film and sub- sequent handling and viewing (optical) effects can distort spherical shapes so that they have sharp features, so that an ice- like character can be associated with a water sample. In summary, assorted problems with the instrument can give an ambiguous measurement of particle numbers as well as a quite ambiguous measure of the character of the particles. So, unless extreme caution is exercised in the opera- tion of the instrument, it is not even able to give a qualitative assessment as to the presence of ice and water in a cloud. Another popular instrument is the Johnson-Williams (J-W) hot wire liquid water meter. The principle of operation relies on a hot wire which is stretched perpendicular to the air stream and the temperature of the wire measured using the thermal variations of its electrical resistance. Water droplets colliding with the wire cause cooling and the temperature of the wire is used as a measure of liquid water, in grams/m . The device has been used operationally for several years and comparisons generally show that it has an accuracy better than 50%. It has but three fai 1 ings : 1) It does not respond to droplets greater than about 30 p and so is only sensi- tive to cloud droplets (see Fig. 11b). 2) The deterioration of the wire, elec- tronic noise, and drift as well as temp- erature, velocity, and pressure fluctua- tions make the readings meaningless at low liquid water contents (below 0.1 gm/m3). 3) It Hoes not respond to the presence of ice (see Fig. 13). Also, as normally supplied by the manu- facturer, the instrument has a slow response (about 1 sec) and so cannot measure maxima in the water distribution. It also cannot count droplet numbers or measure their size, quantities of importance in cloud development. However, Ruskin has developed a technique using the t ransmi ss i vi ty of the cloudy air over a 60 ft pathlength which, coupled with a measurement of liquid water, gives a measure of both droplet numbers and the Sauter mean radius , 3/2\ of the droplet distribution. The instrument of Levine ( 1 965) is a variation of the J-W instrument: it consists of two parts, each intended for measuring droplets in one of the two size classes, cloud water and rain water. One portion, resembling a banjo, is approximately the equivalent of the Johnson- Williams instrument with a larger wire; it is sensitive to cloud droplets. The other portion, resembling a cone, is primarily sensitive to the larger drops. The ratio of the output gives a measure of the volume mean drop size as well as cloud water and rain water. A third instrument In wide use is the aluminum foil impactor, manufactured by Meteor- ology Research, Inc. This device is very oper- ational and reasonably accurate, Is sensitive to raindrops (particles above 200 y) and can measure the character of the particles to determine ice crystal concentrations and type, provided the particles are above about 750 y in size. Basic- ally, the instrument exposes a strip of aluminum foil to the airstream. The sampling area is about 20 cm so that approximately a 2 m3 sample is taken in 1 km of cloud. Though there have been several attempts to calibrate the instrument, the correspondence between impression size and drop size remains in error by at least \0%. This is because of a difficulty in interpreting the impressions. The impressions generally are not totally well defined so that there are "hard" and "soft" impressions. Also, photographs of the original foil data are usually taken for image enhancement, storage, and convenience in the reduction of the data. The angle of lighting affects the data and the measured sizes of the Itopress ions. Let us not omit the problem with data retrieval using the formvar replicator and the foi 1 impactor. The data necessari ly are not ob- tained in real time, and often (with the formvar replicator) data is not collected due to a mal- function of the instrument. But after collecting the data, it requires at least a minute to obtain only qualitative judgments from a single frame. This means that 6 hours are required to reduce 10 seconds of data from one cloud traverse. The te- dious nature of the work and the number of clouds in a single hurricane makes this an insuperable job. Presently, the analysis of data from the foil impactor for one experiment in a tropical storm, measuring each impression, requires 208 approximately six months time for data reduction. Fortunately, optical analyzers now on the market show some promise of lessening this burden. Another common measurement made of cloud elements is the number of ice nuclei. At this time it appears that the concept of ice nuclei is not clearly defined and it is likely that there are at least four types of particles that are effective in the creation of Ice in clouds. These are: a) freezing nuclei b) sublimation nuclei c) contact nuclei d) processed nuclei . Also, it is not known what role secondary |ce pro- duction has in the ice budget of clouds. That is drop freezing and breakup, ice crystal fragmenta- tion and other such phenomena. Hence, it appears that present measurements of "ice nuclei" are am- biguous to such a degree that the basis of the measurement itself needs reappraisal. The more basic measurement appears to be the measurement of actual ice concentrations in cloud. §5. INSTRUMENTS UNDER DEVELOPMENT There are instruments now under develop- ment that are capable of making measurements in nearly real time. Specifically, in the measurement of raindrops, a raindrop disdrometer has been deve- loped by Sutherland and Booker (1970) and is pre- sently being sold as a commercial unit by Weather Sciences, Inc. The device uses an impact sensor which detects the momentum of particles hitting a surface pointed into the airstream (see Fig. 5) • NINE INPUT AMPS UNO LOGIC ELEMENTS Count.rt -0 Fig. 5. Raindrop impaction disdro- meter (courtesy J. Sutherland). From its inception, the instrument has been plagued with the acoustical and electronic noise familiar to designers of aircraft instruments. The noise due to aircraft vibrations and turbulence in the air couple with the natural resonance frequen- cy of the probe to produce a combined noise prob- lem that limits the instrument to the measurement of particles greater than 500 u in light or mod- erate rainfall with aircraft speeds below 70 m/s. A comparison of the data collected with this instrument to data from a similar instrument mounted on the ground (one highly successful and developed by Joss and Waldvogel, 1967) is shown in Fig. £. Since the aircraft was located 300 feet above the ground-based sampler, the possibi- lity exists that both instruments were not in the same rain shaft. Nonetheless, the data do show some correspondence except that the airborne in- strument indicates too many smaller raindrops — perhaps due to electronic noise. In the cloud droplet area, several in- struments are under development. One, first manufactured by Keiley in I960, uses an electro- static technique, employing induction charging. c 100- 10 E o ■5 E E E u 0.1 RR40: 073315- 073325 EDT. RD69: 073400-073530 EOT T 1 1 1 AIRBORNE W.S.I. DISDROMETER (RR40) R • 5.2 mm/kr (0.367 gm/m3) JOSS" ground based DISDROMETER (R0 69) R' 2.69 mm/hr (0.168 «m /»') (RECONCILED) J I L 0.5 30 10 1.5 2.0 2.5 DROP SIZE (mm) Fig. 6. Comparison of airborne and ground based disdrometers (courtesy J. Sutherland). The present version of the instrument is being developed at NCAR by Dye, Abbott, and Sartor (see Sartor and Abbott, 1970). The drop- lets are drawn into a hole (250 v in radius) and collide with a probe held at a high potential (500 v) . An electrical pulse, produced during the impact of a droplet, Is a measure of the droplet's size. The pulses are recorded to produce a histogram of the droplet distribution. Comparisons of the output of the in- strument with the J-W liquid water meter and the slide impactor (an instrument that gathers the cloud droplets in a dye or powder on a small glass slide) are shown in Fig. 7a and 7b. T -r -3.0+r- io 2.0 K Ul h- < 1.0 i ii' i — i — r — rn — r— i — t—r _ Electrostatic Oisirometer — — J-W Hot Wir« ■ ■ ■ _l- 1008 40 100850 TIME (EST) Fig. 7a. Comparison of electrostatic disdro- meter with slide Impactor (courtesy J. Dye). 248 FLIGHT NO 5 SLIDE NO 6 FEBRUARY 26,1971 MIAMI 80 00 FT Fig. 7b. Comp- arison of electro- static disdrometer wi th si ide impac- tor (courtesy of J. Dye). — i 1 r Elictrottotic Di*4ronttttr .... Slidt 10 w -•> i 1 \ 10 20 30 DIAMETER (/i) Despite a general difficulty in com- paring the data due to different time and space scales, the comparisons are good. The deviations may merely reflect the effects of our meager knowledge of droplet collection efficiencies in this region. Of importance is the fact that the high resolution of the instrument allows one to discern detail in the cloud heretofore unseen. The data quite often show small regions within the cloud where the distribution is considerably wider than the surroundings. A similar electrostatic principle was adapted to the measurement of ice crystals by Mach at the University of Washington (1969) and McTaggart-Cowan et al. (1970) at the State University of New York. The individual charges acquired by a wire stretched perpendicular to the air stream are measured and used as an indication of the presence of individual ice crystals. Both naturally acquired charges and charges produced by induction charging are measured depending upon the potential of the wire. At this time the technique shows promise but it is still overcome with electronic noise and data recording prob- lems . The most promising instruments in almost all fields use optical techniques, usually employing the laser. The devices use three tech- niques, light scattering, extinction and imaging. Ryan et al. (1972) have developed an instrument for measuring cloud droplets based on the scat- tering principle. A diagram of the device is s hown in Fig. 8 . Fig. S. Cloud droplet disdro- meter (courtesy R. Pyan). The instrument uses a 55° scattering angle and over- comes the physical limitation of the aircraft mount by using a small mirror in a cowl outside the air- craft. Laboratory calibrations are used to convert the output pulses into drop sizes which are tallied by incrementing electronic memory elements. This gives an output once every 0.5 seconds for recording. The instrument is sensitive to cloud drop- lets between 1 and 30 u and has been used success- fully during the past two years in flights' aboard the NASA Corvair 990. Typical particle distributions in various cloud forms are shown in Fig. 9a, b, c, d, e. In the case of the stratoform cloud, Fig. 9a, a passive microwave radiometer was present on board the aircraft which measured the total liquid water over a column from the thermal microwave emis- ASCENT IN STRATUS OFF CALIF. 1350- 2440 C J. lOOr- 55 22 DROPLET DIAMETER \fi.) (B) _ ■ ■ l ■ .CUMULUS IN GULF Ba B ioor 4 22 DROPLET DIAMETER^ r) ALTO Cu NORTHERN CALIFORNIA 16.000FT (-4.1°C) Sac. into cloud 4 13 DROPLET DIAMETER (fi) PARTLY GLACIATED CUMULUS. NEVADA 24.000FT (-22°C) (E) CIRRUS ANCHORAGE. Alaska lr 31.700FT (-43°C) 55 85 EQUIVALENT DIMENSION Fig. 9. Droplet distributions (courtesy R. Ryan) 210 249 sions by the cloud water. The liquid water meas- ured with the disdrometer agreed in an approxi- mate way with the water content indicated by the radiometer. Note the broad distribution of large droplets in the maritime cloud (Fig. 9b) and the narrow distribution of droplets in the continent- al cloud (Fig. 9c). Also, as ice is formed in the cloud, t lie spectrum becomes diffuse (Fig. ?d) and in a cirrus cloud layer the distribution becomes quite diffuse (Fig. 9e) . It appears that the instrument can detect the presence of ice indirectly by recording the appearance of a broad distribution. Cunningham of AFCRL has another unit of interest that is intended for the measurement of raindrops. It uses 90° scattering by individual raindrops for detection. But, most important, the crossection of the sample is about 100 cm2, indicating perhaps the largest sample volume of any such instrument. (A ra i ndrop-sonde developed by R. Nelson of the State University of New York has a sample area of about 50 cm . It uses water-sensitive oscillographic paper to measure the drops through their impressions.) This is within the safe operating range of Fig. A. The device is intended for measuring raindrops and at present it appears that its lower limit is around 0.5 mm. Assorted problems, including electronic noise and the background of scattered light from cloud or fog droplets, have plagued the device since its inception, many years ago. The 60 ft. path length instrument of Ruskin operates on the extinction principle over an aggregation of particles. A light source is located on the wing of the NRL Superconste 1 lat i on and a detector is placed in the upper radome. The measured t ransmi ss i v i ty can be converted into a measure of both the number of particles and the size using a simultaneous measure of liquid water. The liquid water content can be obtained from the J -W hot wire or Ruskin's total water meter. This is an instrument that measures the attenuation of a Hydrogen-Lyman a, UV , emission line as it passes through the sample volume. The instrument evaporates all the liquid or solid i.ydrometeors in the air parcel and measures the total vaporous water in the air parcel; it is operational and is presently being manufactured by Cambridge Instruments. Its major faults are a sensitivity to variations In C02 and 02 in the air so that it must be calibrated systematically against a standard dew point indicator. Its re- sponse is fast--so fast in fact, that the melting of individual precipitation particles can be seen in the electrical output. Qualitatively, it appears that the area of the individual pulses indicates the particle mass. Also raindrops tend to give relatively sharp pulses while ice crys- tals give broad pulses, possibly due to their slow evaporation. (Similar pulses appear when the output of the J-W hot wire is recorded at high speed.) Extinction measurements are perhaps the most physically sound as they give a direct mea- sure of the physical size of the particle from a measure of its shadow. The basic shadowgraphi c measuring technique was used by Levine (1965) with photographic recording. Recently, however, Knoll- enberg (1971) at the University of Chicago has ex- tended the method to the smaller cloud droplets and aerosol sizes using lasers and electronic light detectors. Knollenberg tunes the laser to operate in the 01* lasing mode in the 1 ase.r source to pro- duce a donut shaped intensity distribution in the viewing volume. This defines the sampling volume and increases the s ignal -to'-noise ratio by the use of the double pulses that occur in the detected light intensity as a droplet passes through the donut intensity profile. This unique feature cou- pled with improved optics allows the use of an ex- tinction system to measure cloud droplets on- board aircraft. Presently underway at the Univer- sity of Chicago is the development of a three- beam instrument using this technique for measur- ing particles in the Venusian atmosphere. The pod is only about 6" diameter by 2V long and weighs 30 lb., yet it contains the capabi 1 i ty of measuring particles from 1 to 32 V with the neces- sary electronics. Knollenberg developed this system in conjunction with a unique optical array imaging system which consists of a series of optical fibers which are shadowed when a particle passes through a field of light (see Fig. 10). COLUMATED SOURCE FIBER OPTICAL TO ARRAY PHOTO x DETECTOR! Fig. 10. Optical Array Technique. (Courtesy of R. Knollenberg) 20 30 40 50 60 10 20 30 40 sec Fig. II. Data from the optical array (a) com- pared with J-W (b) the cutoff of the J-W. (Courtesy of R. Knollenberg). 211 250 A complex Set of comparators and logic elements record a size spectrum of particles entering the field of light. The system as a whole has a re- solution down to about S p, a resolution somewhat less than the above mentioned extinction system. Data recorded in scattered cumulus clouds with the device are compared with the J-W hot wire on Fig. lla and lib. In Figure I la the comparison Is quite good. This is probably because the errors in the system are sms! I as a result of its uniform response throughout the sampling range as well as the relative unimportance of the smaller droplets in determining the cloud water. (The practical lower limit of the instrument is about 8 ]i.\ Fig. lib shows the response of the J-W liquid water meter explicitly. It appears that all the drops below about 35 to 40 y are measured by the instrument. Also, crystal di- mensions have been measured in cirrus clouds, at 18,000 to 27,000 feet and at temperatures be- tween -10 and -35°C. Fig. 12 presents a sample distribution. Fig. 13 shows a comparison be- tween the liquid water content of cirrus crystals and the liquid water indicated by the J-W hot wire instrument. The crystals were columnar as JULY 22,1970 COLUMNAR CIRRUS CRYSTALS Fig. 12. Size distri- bution for cirrus ice crystals (courtesy of R. Knol lenberg) . Fiq. 13- Re- sponse of the J-W to ice (courtesy R. Knol lenberq) . 4 i.o E JULY 25,1970 ■ / CIRRUS CRYSTALS COLUMNAR J-W 0948 LIQUID L WATER 09 49 09 50 TIME IN MINUTES determined with a formvar replicator. For all practical purposes the hot wire instrument did not respond. Note the relatively high equivalent water content in these clouds (about 0.1 gms/m3). These results indicate the great importance cirrus clouds have in the global water budget. The attractiveness of extinction mea- surements is also reflected in the instrument presently being designed for the National Hail Re- search Experiment by Kyle. Extinction measure- ments are also being considered in a design by Radke at the University of Washington. Radke, however, hopes to manufacture an instrument which considers the optical properties as well. The intent is to use' partially polarized light and ob- serve the scattered light in extinction with a sensor covered with a crossed polarizing filter. The result is that the beam is attenuated in the presence of raindrops while the presence of ice rotates, depolarizes, or otherwise alters the po- larized light so that there is enhanced trans- mission. The output is a negative pulse from the photo-multiplier tube when an ice crystal enters the sample volume, and a similar positive pulse with a raindrop. The Ice Particle Counter developed by Mee Industries attempts to capitalize on three optical properties of ice to detect and count ice particles: a) The tendency for ice to rotate the plane of polarization of incident light. b) Reflections of light from the specular faces of the crystals. c) A preferred scattering angle by ice crystals . The source light to the device is polarized and the detector (a tube) is covered by a polarizing filter oriented so that normally there is nearly (01 UNP0LANIZE0 LI0HT AND DETECTOR *Pl*Tt O -I0*C /«COUJWi Q -10'C m«ULU -40*C (bl POLARIZED DETECTOR J. SCATTER!*© PLANE JO'C Fiq. 11). Measured intensity of scattered light from a population of ice crystals (see Huffman, 1970). complete extinction. Scattered light is measured at an angle of about 125° from the light source. Fig. 14 indicates that this angle is indeed an optimum angle for the detection of scattering characteristics specific to ice. The output is an electrical pulse cor- responding to the presence of a particle. Some response is observed when water drops are present. It appears that a pulse of equivalent height is produced by a 1mm drop and an ice crystal 100 p in size. If, however, the sharpness of the peaks is considered, it appears that there is only about a 25°» chance of ambiguity. Ice crystals appear to create sharp pulses whereas water drops create rounded pulses. The electrical output of the instrument is apparently a function of the square of the size of the particle, a feature of most of the optical instruments and desirable to maintain a uniform response over a large particle range . 212 251 Thus far we have omitted optical im- aging instruments that take photographs of the cloud elements. A basic instrument for photo- graphing cloud hydrometeors in flight has been made operational by Lavoie et al . (1970) of Pennsylvania State University. The instrument uses a motion picture camera and a strobotak or pulsed light source. At this time the instru- ment has been used successfully to photograph raindrops and snowf lakes in rlight. A similar instrument is presently installed on the NCAR Sailplane. It employs a special photographic technique developed by Cannon (1970) which uses a double light source to produce two images of the cloud par- ticles. The distance between the images is a measure of the droplet size; only droplets within the depth of field show the double imane so that the sampling volume is well defined. The aircraft system uses an airfoil above the fuselage to produce the appropriate back light- ing. Droplets in the volume act like little lenses and produce the double images whereas ice crystals in the volume are opaque scatters which are lighted from the front. Hence, the system is capable of measuring and sizing both water drops and ice crystals. Perhaps the most interesting devices being developed are those which use holographic techniques. Ketcham of the Univer- sity of Utah has developed a character recog- nition technique which shows promise of giving a real-time readout of particle numbers and sizes in the cloud droplet ranges. The tech- nique is illustrated in Fig. 15. Parallel light from a coherent source illuminates the particles, passes through the lens and converges at the focal point. The placing of an appro- priate filter (holographic picture) at the focal Step 1 . Product ion of f i Iter. \ PHOTOGRAPHIC PLATE (PRODUCES FILTER) RECONSTRUCTED REFERENCE BEAM < ■ pTnWrTm^\/-rTrrTTTTTTM DROPLET FILTER point eliminates the source of light, and imaging by another lens produces a bright spot, the Image of the reference beam, which corresponds to the presence of a drop in the viewing volume. A photo cell placed at the bright spot detects the presence of the droplet. The filter is a holographic dif- fraction pattern for a specific particle size, so that the Instrument counts particles of a given size. A series of filters than gives a measure of the entire droplet distribution. Of course, there is overlap. This is illustrated by the numerical simulation shown in Fig. 16. In this case a 7 V mask (filter) 5 Fig. 16. Relative response to a 7 U mask (courtesy W. Ketcham) . J 5 7 9 11 1W was used and droplets of the sizes shown on the abscissa were introduced into the volume. We see that errors of '0% and 30% are accepted when, respectively, 5 and 13 p droplets are in the sample volume. The technique has not yet entered the development stage but it shows promise of giving a real-time readout of sizes and concentrations for large sample volumes. L. Nelson and Kunkel (see Kunkel, 1970) at the Air Force Cambridge Research Laboratories have developed a complete holographic optical system. The instrument takes a picture of the sample volume and preserves the total optical character of the sample. Later, in the labora- tory, the image is restored and the particles analyzed. Then, the volume is scanned and the sizes measured. It is not generally possible to see the whole particle but a good cross section can be measured. The resolution of the instrument is 5 U to 25 U depending on the position of the sample volume. The instrument has been mounted in a van to accommodate some two tons of auxiliary equipment. Other limitations such as a large power requirement and vibration will probably limit the instrument to ground use for some time. A comparison with the J-W liquid water instrument is presented in Fig. 17 and a sample droplet histogram is shown in Fig. 18. The com- parisons do indicate the feasibility of this instrumental technique. Unfortunately, the data reduction problem with the instrument is insuperable so that it will probably never be more than a calibration standard. Step 2. Reproduction of reference beam spot. DIFFRACTED LIGHT RECONSTRUCTED RIGINAL BEAM Fiq. 15. Illustration of character recognition technique (courtesy W. Ketcham) . Fig. 17. Temporal comparison between holographic camera and J-W (courtesy L. Nelson). JD 19 JUL* 1965 OH' 0304 Tlue (EOT > 213 252 (0)1} JUNE 65 0700- 0729 EOT LWC- 096 ( b I 4 JUIV 65 0444-0435 EOT IWC- 066 « » w n Fig. Ift. Droplet distributions from the holoaraphic camera (courtesy of L. Nelson). §6. THE MOST PROMISING INSTRUMENTS Basically, our instruments fit into three classes. The first class consists of instruments with qualities as absolute standards that, perhaps, require tedious data analysis and have little or no ability for real-time analysis (i. e. photographic instruments). The second class consists of instruments capable of high resolution and response in nearly real time (i.e. instruments with electronic readout). In our operational programs, the second class of instru- ments should be used in conjunction with the first class so that the second class supplies the bulk of the data at high resolution and the first class is used for spot checks and calibration. The older instruments, the formvar replicator, foil impactor, etc. fit into the first class, as do the photographic instruments including the holographic camera. Hence, it is wrong to say that these instruments will be re- placed with the new generation of instruments. They supply invaluable supplementary information for the second generation. The third class of instruments, not heretofore mentioned, consists of those instru- ments that give an overall view of the cloud. Data supplied by these instruments include photo- graphs of the clouds, radar PPI and RH I returns, Doppler shifts, infrared and microwave emissions, and perhaps laser and acoustical reflectivities. This class virtually cannot be used for the ac- quisition of quantitative data without some infor- mation from the instruments in classes one and two. An example is the measured radar reflec- tivity. It is affected by the character of the particles (i.e. whether spherical or hexagonal, water or ice), the number of particles, but, mostly, the s ize of the cloud elements. Its dependence on the 6th power of the size makes it necessary that some in-cloud measurements be available to interpret the data. Of the instruments in the second class, Ryan's cloud droplet disdrometer and the J-W hot wire appear to give the most realistic measurements of cloud droplet properties and they both appear to be highly operational. Note that there was consistent agreement between several of the above instruments and the data from the J-W hot wire. In the measurement of larger cloud drops, raindrops and ice crystals, however, there are no instruments of reasonable opera- ti onal cal iber. In the measurement of raindrops the raindrop djsdrometer manufactured by Weather Sciences, Inc. and Cunningham's light scattering Instrument (with the large sampling volume) show promise. But, in the measurement of Ice crystal concentrations, there are no truly operation instruments in Class Two, although the Mee Ice Particle Counter may be able to measure Ice crystal numbers. Also, there are no instruments that measure cloud elements In the intermediate size range. Every attempt should be made to extend the ranges of the cloud droplet instruments to cover this range. §7. INSTRUMENT CALIBRATION It is contrary to the basic philosophy of science to put together an instrument and pro- ceed to use that instrument to measure an unknown quantity. Yet, we in the atmospheric sciences commit this type of heresy without second thoughts. Before data can be said to have any quality it must be calibrated with knowns . In particle measurements, this means supplying the instruments with particles of known size and character at a known velocity. The ideal procedure is to supply the instrument with a stagnant sample in the labora- tory and check the response. Then the instrument is moved to a natural environment and tested, say, on a mountain top with real hydrometeors . Finally, if the instrument shows the desired qualities, it is tested on-board aircraft in real clouds. Complete testing of an instrument in a simulated environment is nearly impossible with our present facilities. For instance, Cornell Aeronautical Laboratories, Buffalo, New York, or the Fog Facility at the State University of New York, can supply a fog with known characteristics but they cannot supply aircraft velocities. Natick Laboratories in Massachusetts can supply temperatures down to -70 °C and air velocities to '♦O mph, but they cannot supply a fog of known character and their chamber is too small to ac- commodate crystals or raindrops with appreciable fall velocity. Wind tunnels generally are too small or cannot supply an appropriate sample. The Air Force Arnold Test Center in Tenneessee has the most capabilities. It can supply fog, rain, and snow at speeds up to sonic velocities in a chamber as large as 16 feet across. However, the facility is intended for the testing of aircraft engines, is in frequent use, and costs $1,000 an hour. There is a real need for a National facility for the testing of instruments and this should be one of our priorities for the future. It is hard to imagine the ideal facility, but it would have to have a size comparable to that of a small cloud to accommodate the residence times of the larger particles and simulate the turbu- lence and air motion. It would have to contain a high speed vehicle to move the instrument through the cloud. Plus, it would have to be in a location with extremely cold and warm environ- mental conditions to simulate the thermal condi- tions and minimize the power requirements. These conditions could perhaps be met by building a high-speed trolley within one of the underground cavities formed by a nuclear explosion. 214 253 In terms of size, the best approxima- tion to this facility is the huge chamber at Eglin Air Force Base. It has room for several aircraft and will accommodate the C5A. Ice and fog can be supplied with an airflow produced by fans. But any final tests will have to be made using the clouds themselves as test environments; this re- quires that we be content with instrument compar i sons . A practical solution to the instrument testing might be to test Instruments operated on ground together with instruments of the second type that can operate in the air. Then, the instruments of the second type could form second- ary standards and further comparisons could be made on-board aircraft. Using this "bootstrap" technique some estimation of accuracy could be placed on the data from each instrument. §8. FUTURE TESTS PLANNED Practically, the logistical and politi- cal problems that must be solved in order to conduct instrument testing and comparisons on board aircraft are almost without bound. For instance, suppose several organizations agree to conduct tests of their instruments under deve- lopment. Then, even though the specific investigators agree to the goal, they are likely to disagree as to the means. Supposing agreement is obtained and an aircraft facility agrees to conduct on-board tests of the instru- ments, arrangements for installation of the instruments on the aircraft or in a pod must be made including inspection by the FAA. Then, for the larger aircraft, at least $500 must be avail- able for each hour of operation. However, there presently are so many operational programs that the time would have to be scheduled at least six months in advance. Of course, if only two weeks were allocated for the tests, there could well be no clouds available at that time. In the testing program, the problem of instrument reliability arises. At this time few of the operational instruments for the measurement of cloud particles have attained an operational reliability better than 501. The instruments in development are likely to have a practical reliability somewhat less than this. This means that, if three instruments are being tested, there is only a 10% chance that all three will function simultaneously. Obviously the testing of more than three of these instruments is ludicrous. Nonetheless, there are tentative plans to proceed with limited instrument comparisons early in 1973- These will most likely be only for the comparison of techniques for measuring cloud droplets and probably will be conducted in a laboratory, in a wind tunnel, or on-board aircraft. The results of these tests and others should give our measurements of cloud elements more credence and bring forth the fail- ing points in our present models of cloud micro- physical and dynamical processes. The result should be an upgrading of our understanding of cloud development. Acknowledgements The material presented herein was for the most part contributed, by the members of the Symposium on the Measurement of Cloud Elements, held June 10 and 11, 1971, in Chicago. The principle contributors were: R. R. Braham, Jr., University of Chicago; W. R. Cotton, Experi- mental Meteorology Laboratory, N0AA, Miami; D. E. Culnan, Bureau of Reclamation, Denver; R. M. Cunningham, Air Force Cambridge Research Laboratory, Bedford; E. F. Danielsen, National Center for Atmospheric Research, Boulder; J. E. Dye, National Center for Atmospheric Research, Boulder; H . A. Friedman, Research Flight Faci- lity, N0AA , Miami; U. Katz, Cornell Aeronautical Laboratory, Buffalo, N. Y.; W. M. Ketcham, Uni- versity of Utah; R. G. Knollenberg, University of Chicago; T. G. Kyle, National Center for At- mospheric Research, Boulder; L. D. Nelson, SUNY, New York; R. E. Ruskin, Naval Research Labora- tory, Washington, D. C; R. T. Ryan, A. D. Little, Inc., Cambridge; R.L. Smith, Eglin AFB , Florida; J. T. Sutherland, Weather Sciences, Inc., Norman, Oklahoma; D. M. Takeuchi, Meteorology Research, Inc., Alta- dena , Cal i forni a . REFERENCES Cannon, T. W., 1970: A camera for photographing airborne atmospheric particles. Image Technology, April/May. . Cunningham, R. M., 1951: Airborne raindrop size measurement and instrumental techniques. Proceedings of the Conference on Water Resources, Illinois State Water Survey, Urbana, Illinois, 1-3 October. Danielsen, E. F. , R. Bleck, and D. A. Morris, 1972: Hail growth by stochastic collection in a cumulus model. J. Atmos . Sci . , to be published in March. Huffman, P. J., 1970: Polarization of light scattered by ice crystals. J. Atmos. Sci., 27, 1207-1208. Joss, J. and Waldvogel , A., 1967= Ein spektro- graph fUr neiderschlagstropfen mit auto- matischer auswertung. Pure and Applied Geophys. , 68, 2't0-2'*6. Knollenberg, R. G. , 1971: Particle size measurements from aircraft using electro- optical techniques. Proceedings of the Electlro-opt ical Systems Design Conference- 1971, West Anaheim, California, 18-20 May, 218-233. Kunkel , B. A., 197' •' Fogdrop - size distribu- tions measured with a laser hologram camera. J. Appl. Met. , 10, i*82-^86. Lavoie, R. L., J. A. Pena, R. dePena, R. L. Ruth, R. P. Greiner, D. A. Corkum, J. L. Lee, and C. L. Hosier, 1970: Studies of the microphys ics of clouds. Final Report 16 for NSF Grant GA-3956 , Department of Meteorology, Pennsylvania State University. 215 254 Levine, J., 1965: The dynamics of cumulus convection in the trades-a combined obser- vational and theoretical study. Ph.D. thesis, Dept. of Meteorology, Woods Hole Oceanograph ic Institution, Massachusetts. Mach, W. H., and P. V. Hobbs , I969 : The elec- trical particle counter. Contributions from the Cloud Physics Laboratory, Research Report III, App. A. McTagga rt-Cowan , J. D., G. G. Lala, and B. Vonnegut, 1970: The design, construction, and use of an ice crystal counter for ice crystal cloud studies by aircraft. J. Appl. Met . , 9, 294-299- Ryan, R. T., H. Blau, Jr., P. C. von Thu'na, and M. L. Cohen, 1972: Cloud mi c ros t ructure as determined by an optical cloud particle spectrometer. J. Appl. Met., submitted for publ i cat ion . Sartor, J. D., and C. E. Abbott, 1970: An elec- trostatic cloud droplet probe. Preprints of the Conference on Cloud Phys i cs , 24-27 August, Fort Collins, Colorado, 97"98. Sutherland, J. L., and D. R. Booker, 1970: An airborne momentum sensing raindrop spectrometer. Preprints of the Conference on Cloud Physics, 24-27 August, Fort Col 1 ins , Colorado, 101 -102. NOTICE The National Hurricane Research Labor- atory and the Department of Commerce do not approve, recommend, or endorse any proprietary product or proprietary material mentioned in this publication. No reference shall be made to either organization or to this publication in any advertising or sales promotion which would indicate or imply that the National Hurricane Research Laboratory or the Department of Commerce approves, recommends, or endorses any proprietary product or material mentioned herein, or which has as its purpose an intent to cause directly or indirectly the advertised product to be used or purchased because of th is publ icat ion . 216 255 25 Reprinted from The Review of Scientific Instruments 43, No. 1, 152-153. Details for Constructing a Miniature Solid State Electrometer Probe William D. Scott National Hurricane Research Laboratory, NOAA, Miami, Florida 33124 (Received 19 August 1971) A general interest in an electrometer designed by the author in 1966 has prompted this note detailing its con- struction and calibration. The device was described in some detail by Scott1 and Scott and Hobbs2 and schematic dia- grams for a modified version of the device in Scott and Levin.3 Figures 1 and 2 show the basic schematic diagram and the assembled device. The MOSFET and bipolar transistors used should be determined by the particular application, but the M103, a protected MOSFET manu- factured by Siliconix has given good reliability. The sensi- tivity of the device is primarily determined by the com- ponent input capacities since the device is voltage sensitive and V = Q/C; hence, the use of a device with lower capaci- tance will increase the sensitivity. The device (with the X10 amplifier described by Scott and Levin3) can detect charge transfers that occur in times less than 1 msec at io"n Fig. 1. Schematic diagram of electrometer. levels as small as 10-6 esu (~ 10-15 C). The input impedence of the device as shown is 10" $2, which can be varied by changing the high megohm resistor (available from I.R.C., Philadelphia, or Eltec Instruments, Lancaster, N. Y.). The device is constructed from a short length (2.5-5 cm) of 0.635 cm o.d. polyethylene tubing. The M103, (TO-18), with the leads wrapped in aluminum foil to prevent destruction of the MOSFET by electrostatic charges, and the 2N3904 (plastic case the same size as the TO-18 case) are pushed into either end of the tubing. Then assorted holes for the resistors and capacitors and their leads are drilled in the central portion of the tubing. These com- ponents are then placed in the holes and their leads bent and clipped so that they lie along the outside of the tubing wall and the connecting wires lie parallel to one another with as much overlap as possible. They are soldered in place partially melting the polyethylene sheath and forming a rigid assembly. The input power leads and signal output connections are made with a length of 3- or 4-wire shielded cable through a rubber stopper to the back end (bipolar transistor end) of the device. The high meg resistor is soldered directly to the gate lead of the MOSFET transistor together with, perhaps, a standard 1.59-mm pin. Of course the whole assembly is mounted parallel for insertion into the quartz envelope. Then any jagged solder joints are smoothed and the entire assembly (except the front end and the MOSFET) wrapped with two layers of standard Teflon tape. A piece of Co-netic and a piece of Netic foil (Perfection Mica Co., 256 NOTES 153 Illinois) are cut and wound on an appropriate mandrel so that each forms a sheath that doubly shields the entire device, including a portion of the power and signal cable but excluding the MOSFET and the front end of the device. (If the shield covers the 10" ohm resistor, the capacitance will be increased.) At this point the input leads are attached to a batten and output indicator through a junction and control box. The aluminum foil is removed from the MOSFET and the input impedance and voltage gain checked for appropriate response. If there are no shorts and the device is functional, a small piece (or two) of colored Drierite (CaS04) is placed on the inside of the rubber cork using a small amount of varnish. A small amount of epoxy resin is put on the sensitive tip, and the whole assembly is inserted inside a prepared 5-8 cm length of 12 mm o.d. quartz tubing. Care should be taken to assure that the tip is guarded the whole time since the quartz will generally have large charges on its surface. After the resin has set, the electrometer probe may be calibrated for input capacity, voltage gain, and resistance as follows: first, by attaching a small capacitor of known value (1 — 10 pF) in series with the input and applying a 1-kHz signal and noting the attenuation. Care should be taken that the input leads do not contribute significantly to the capacity. Second, by placing a charge on the input and noting the decay time (t = RC) of the output. The voltage FlG. 2. The electrometer and control box assembly. gain and input resistance are established by merely noting the output voltage with various input direct voltages and series resistances. The device should have a dynamic range of ±1 V and a frequency response in the megacycles. 1 W. D. Scott, 1968: "Single Charging Events Due to Collisions in Natural Snowfall." Presented at the 4th International Conference on the Universal Aspects of Atmospheric Electricity, Tokyo, 1968, in Planetary lilectrodynamics, edited by S. C. Coroniti and J. Hughes (Gordon and Breach, New York, 1969), pp. 85-99. 2 W. D. Scott and P. V. Hobbs, Quart. J. Roy. Met. Soc. 94, 510 (1968). J W. D. Scott and Zev Levin, J. Atm. Sci. 27, 463 (1970). 257 26 Reprinted from Science 177, *?25-**26 Reprinted from 4 August 1972. volume 177. pages 425-426 Open Channels in Sea Ice (Leads) as Ion Sources Abstract. Open channels In sea ice may be acting as sources of atmospheric ions. In 1970 an experiment was con- ducted at Point Barrow, Alaska, to measure the background aerosol (/). The measurements were made at a time when the ice was starting to break up offshore. One of the aerosols that was monitored consisted of sodium- containing particles (2); the number of these particles seemed to increase when- ever a freshly opened lead appeared upwind of the station. The conductivities of the positive and negative ions were also measured, and they seemed to increase in magnitude with the appearance of the leads. To check this observation, a Gerdien-type conductivity instrument (3) was flown to an area where freshly opened leads were observed. The instrument was positioned there in such a way that it sampled air which had just passed over a freshly opened lead (Fig. la, site A). Measurements were carried out at three different heights above the level of the ice (30, 70, and 120 cm) and are presented as a profile of both positive and negative conductivities (Fig. lb). Baseline measurements were also made near the lee of the lead, 30 cm off the ice (Fig. la, site B; see arrows). Errors in the measurements may have been large because a calibration could not be done on the site. However, the measurements of conductivities of both the positive and negative ions were made by simply reversing the polarity of the battery and hence reversing the radial electric field in the sampling tube of the instrument, so the data should have a high relative accuracy and the trends in the profiles should be ac- curately portrayed even though the absolute conductivities may be in error. The results in the lee of the lead in- dicated that the conductivities of ions of both polarities were somewhat larger than the baseline values. Also, at 70 cm above the ice, the conductivity of the negative ions was greater than that of the positive ions. The conductivity values decreased rapidly with height and became comparable to the baseline values at about 120 cm. The fluid in the leads was an ice slush composed of water and pieces of ice with no visible bubble activity. This observation suggests a mechanism of ion formation at the leads which is dif- * Site A 120 r SiteB a. E 100 80 60 40 (b) Fig. 1. (a) Schematic diagram of the site of the experiment (b) Conductivities of positive ( + ) and negative (— ) ions. 20 J Site B baseline values 4 4 j. 4 6 e 10 Conductivity (mhc-eirf' xlO"1*) ferent from the mechanism of surf elec- trification (4), or is composed of a number of mechanisms, some produc- ing negative ions and some positive ones. Surf electrification originates from the bursting of bubbles at the sea surface and results in the production of a predominantly positive charge, al- though a similar mechanism in fresh water often results in the production of a negative space charge (5). It is pos- sible that some very small bubbles do burst in the ocean-water portion of the open leads and produce excess positive ions. The production of the excess nega- tive ions may possibly be a result of the breakup of microbubbles and the re- lease of gas either during the melting process or during the freezing of new water (6). Most significant, however, are the indications that the leads are acting as sources of atmospheric ions. William D. Scott Sea-Air Interaction Laboratory, National Oceanic and Atmospheric Administration, U.S. Department of Commerce, Miami, Florida 33130 Zev Levin Department of Environmental Sciences, Tel- Aviv University, Ramat Aviv, Israel Reference* and Notts 1. L. F. Radke and J. Plnnons, report of the Cloud Physics Group, Department of Atmo- spheric Sciences, University of Washington 2. L. F. Radke and P. V. Hobbs. A Amos. Set. M. 281 (1969). 3. W. E. Cobb and B B. Phillips, U.S. Weather Bur. Tech. Pap. No. 46 (1962), p. 18. 4. D. C. Blanchard, Progr. Oceanogr. 1 17 (1963). 5. S. O. Cathman and W. A. Hoppel / Ceophyj. Res. IS, 1041 (1970). 6. J. C. Drake, Quart. J. Roy. Meteorol. Soc. «4, 176 (1968). 7. The logistic support for this work was pro- vided by the Office of Naval Research (Arctic Program) under contract with the Department of Atmospheric Sciences, University of Wash- ington, principal investigator, Norbert Unter- steiner; grant NOO14-67-A-0103-007. We thank D. Shell for assistance in the collection of data at the lead and the staff of the Arctic Research Laboratory (ONR; University of Alaska), directed by Dr. M. Brewer, for in- valuable backup support. 28 April 1972; revised 16 June 1972 258 27 U.S. DEPARTMENT OF COMMERCE National Oceanic and Atmospheric Administration Environmental Research Laboratories NOAA Technical Memorandum ERL NHRL-100 ON THE INTENSIFICATION OF HURRICANE CELIA (1970) Clark L. Smith National Hurricane Research Laboratory ^ \$ * Coral Gables, Florida f Ogn August 1972 V Pj\i < 259 CONTENTS Page Abstract 1 1. Introduction 1 2. History 2 3. Data and analysis 5 3.1 Data 5 3.2 Analysis 11 4. Diagnostic analysis 14 5. Computations of radial and tangential velocities, absolute angular momentum flux, relative vorticity, and divergence 15 5.1 Average radial velocity— table 1 17 5.2 Average tangential velocity— table 1 19 5.3 Average absolute angular momentum flux— table 1 21 5.4 Average vorticity and divergence— table 2 21 6. Dynamic instability in the outflow layer 24 7. Oceanic influences relative to intensity variations of Hurricane Celia 26 8. Summary and conclusions 33 9. Acknowledgments 33 10. References 34 260 ON THE INTENSIFICATION OF HURRICANE CELIA (1970) Clark L. Smith1 Low and high cloud motions derived from ATS— I I I satel- lite pictures, wind data from surface ships, vertical tem- perature profiles through the upper layer of the ocean, and sea-surface temperatures were used to examine intensity variations of Hurricane Celia. Fields of radial and tangential wind components, ab- solute angular momentum flux, and absolute vorticity were computed from analyzed streamlines and isotachs. Radial and tangential velocities and absolute angular momentum flux were averaged around circles of several radii from the storm center. Average values of relative vorticity and di- vergence within these circular areas were also computed. High-cloud motion computations were successful in identifying the intensification process in the outflow layer. The low cloud motions, unfortunately, were not representative of the storm inflow layer. An investigation of dynamic instability at the outflow level was made. The results did not indicate strongly that dynamic instability contributed to the intensification of the storm. Final deepening followed a major change in direction of the surface inflow over warm shelf waters of shallow depths. 1. INTRODUCTION One measure of the intensity of a hurricane is given by its central barometric pressure. This study examines two outstanding pressure fluctu- ation events of Hurricane Celia (1970). These events were the filling which occurred shortly after initial deepening to hurricane strength and the rapid deepening (43 mb) during the 15-hour period before landfall. The air-sea interaction in the hurricane for both events was investigated Only for the second event was an attempt made to describe the hurricane motion field. This description was confined to the active inflow and out- flow layers of the hurricane. ^ow with the National Center for Atmospheric Research, Boulder, Colo. 261 For convenience, studies of the hurricane motion field may be clas- sified into two categories. The first contains those studies for which data extend outward from the center to the region of maximum winds. The second involves those studies dealing with motions on a larger scale for which data extend outward from the center to a radial distance of 10° of latitude; such data are generally sparse from the center to a radial dis- tance of 2°. Because no research aircraft reconnaissance was made into the central region of Hurricane Celia, this study is of the second cate- gory. The hurricane motion field was determined by cloud motions, ship winds, and the conventional surface and upper air networks. Previous hurricane motion field descriptions of the second category involved data composites in time and space; namely, works by Hughes (1952), Jordon (1952), Miller (1958), and Krueger (1959). Miller (1958) remarked on the limitations of composite investigations and on the necessity for individual studies. However, for most storm situations, there is a lack of observational data in the wide region between the hurricane-produced maximum winds and the outer envelope where the wind structure becomes totally controlled by the environment. One purpose of this study was to determine whether this deficiency could be overcome through the use of cloud motions derived from satellite pictures. With varying success, a number of researchers have related the sea- surface temperature to the intensity or path of a hurricane; to cite a few, Miller (1957), Fisher (1958), Tisdale and Clapp (1963), and Jordan (1964). However, Jordan and Frank (1964), Leipper (1967), and Perlroth (1967) recognize that the oceanic influence on a hurricane also depends on the vertical temperature profile in the top layer which determines the amount of energy available for transfer to the disturbance. This study employs data relating both sea-surface temperature and vertical tempera- ture distribution to the observed intensity variations of Hurricane Celia. 2. HISTORY For an excellent detailed account of the 1970 African disturbance that entered the Gulf of Mexico and became Hurricane Celia, see the article by Simpson and Pelissier (1971). For this study, only a brief resume cover- ing hurricane development to landfall is presented below. At approximately 0000 Greenwich Mean Time (GMT) on August 1, a low- level vortex, classified as a tropical depression, entered the Gulf of Mexico from the western tip of Cuba and rapidly reached hurricane inten- sity (figs. 1 and 2a). The initial deepening to a central pressure of 965 mb abruptly ended at 0000 GMT on August 2. During the next 12 hours, the storm filled to 986 mb. Celia's central pressure then remained rela- tively constant for about 18 hours. Final deepening, which was continuous AIR FORCE GULL 9 CELIA 14 EYE FIX AT OJ20 Z. WEST MOVE- MENT 300*19 KTS. AT 20901 69 KT. ISOTACH EXTENDS 19 NM EAST FROM CENTER AS OPPOSED TO 99 NM AT 1890 Z. 99 KTS. EXTENDS 49 NM EAST OF STORM. CELIA 23 VORTEX DATA PROFILE ALFA AZIMUTH 310* TIME 0928Z TO 0619 Z. LEFT REAR QUAD I WALL DARK, RIGHT REAR QUAD 1 WALL DARK, WELL DEFINED FEEDER BANDS ALL QUADS DOPPLER INOP NGP SO SPL EYE 27*41' N 96* 94' W AT 1919 Z. DIA 16 NM, FIX GOOD EYE CIRCULAR AIR FORCE a Q3. 13092 G 963 MB H. 700 MB 2783 M I NOT OBSERVED K. 140* 70 KTS. 0. ELLIPTICAL - MAJOR AXIS 130V310*. LENGTH 23 NM. LENGTH MINOR AXIS 19 NM R. CLOSED WALL S. FLIGHT LEVEL EYE DISPLACED NORTH OF SURFACE EYE BY WINOS AND TEMPERATURE. i£££££ 3T0RM POSITIONS CENTRAL PRESSURE (MB) • QAY TIME (GMT) DETAILED AIRCRAFT EYE/CENTER MESSAGE B. DATE AND TIME OF FIX (GMT) G. MIN. SEA LEVEL PRESSURE (MB) H. MIN. HEIGHT AT STANDARD LEVEL (MB/M) I. ESTIMATE OF MAX. SURFACE WINDS (KTSJ K. MAX. FLIGHT LEVEL WINDS NEAR CENTER (DEB. KTS.) 0. EYE SHAPE /ORIENTATION /DIAMETER (NMaiTMl) R. EYE CHARACTER & REMARKS Figure 1. Hurricane Celia 1970 storm positions (NHC best track) and excerpts from aircraft and land radar reports. 263 1010 1000 CO — 990 CC = 980 CO bJ Q. 970 < OC 960 - Z Id 0 950 - 940 SATELLITE MOTIONS 6/2 13261 - 16522 _L ,.\ ... ~H~ SATELLITE MOTIONS 8/3 13J0J -1657 2 _L _L 00 H 12 Z OOZ 12 Z 00 Z 12 Z 002 a/i a/2 a/3 a/4 TIME (GMT) 12 Z Figure 2a. Hurricane Celia 1970 central pressure by time, Figure 2b. Hurricane Celia 1970 observing stations in analysis area. 264 to landfall (Aransas Pass, Tex., 1610-1645 Central Daylight Time-CDT), be- gan approximately at 0600 GMT on August 3 when the eye of Celia was about 200 n mi east-southeast of Corpus Christi, Tex. 3. DATA AND ANALYSIS 3.1 Data On two occasions before Hurricane Celia's landfall, satellite winds- representing low and high cloud motions-were obtained from film loops composed of Applications Technology Satellite (ATS-III) photographs pro- vided by the National Hurricane Center (NHC); namely, the time intervals involved were from 1526 to 1652 GMT on August 2 and from 1530 to 1657 GMT on August 3 (fig. 2a). These two occasions were chosen to compare the storm circulation in a comparatively steady state with one of rapid deep- ening. (Conventional data were also obtained from observing stations within the analysis area; these stations are shown in fig. 2b.) Figures 3 and 4 are ATS-III photographs of Celia, selected to correspond with the mid-interval of each satellite film loop-1614 GMT of August 2 and 1619 GMT of August 3, respectively. Figure 1, designed primarily to indicate the storm positions of Celia as determined by the NHC, also shows excerpts from aircraft and land-based radar reports. Two of these reports, Air Force Gull 3 Celia 7 at 1752 GMT on August 2 and Air Force Gull 5 Celia 13 at 1856 GMT on August 3, summa- rize observed conditions that were prevailing near the satellite-motion time intervals. A comparison of the two reports reveals a difference in central pressure of 31 mb, in estimated surface wind of 20 kt, and in max- imum wind speed at the flight level (700 mb) of 50 kt and shows the change from a circular eye shape on the 2nd to one strongly elliptical on the 3rd, From the low and high cloud motions, streamline and isotach fields were analyzed. Additionally, for comparative purposes, a surface level was also analyzed similarly with data derived primarily from ship wind reports. Figures 5, 6, 7, 8, 9, and 10 represent the streamline and iso- tach analyses of the surface-level, low-cloud, and high-cloud motion data. 265 Figure 3. ATS-III photograph of Hurricane Celia 1970, 1614 GMT> August 23 1970. Ship wind reports on the surface-level motion analyses (figs. 5 and 6) were supplemented with 1000-mb land-based rawin data at 1200 GMT. Available data at 1800 GMT are represented by a dashed shaft placed, when feasible, on the reporting land station or near it. Numerals in paren- theses indicate height of winds in thousands of feet. Isotachs are given in knots. On August 2, ship wind reports at 1200, 1500, 1800, and 2100 GMT were composited around the storm position at 1600 GMT to correspond closely with the time of the satellite motions. For composition purposes, ship displacements from actual positions were, in most instances, not large because the reporting times used were predominantly 1200 and 1800 GMT. On August 3, it was possible to obtain a more synoptic representa- tion by using ship wind reports at 1800 GMT, along with a few observations at 1500 GMT without displacement. 266 Figure 4. ATS-III photograph of Hurricane Celia 1970, 1619 GMT, August 3, 1970. Low cloud motions (shaft with arrowhead) derived from a series of satellite photographs, figures 7 and 8, were supplemented with 850-mb rawin data at 1200 GMT or with rawins at a level above the surface in the High Plains of western Texas and Mexico. With but slight adjustment, the 700-mb rawins may have been used in conjunction with the low cloud motions, Winds at 500 mb, however, differed considerably from the cloud velocities except in central and southern Texas on August 3. A comparison of 1000-mb winds (figs. 5 and 6) with the low cloud motions (figs. 7 and 8) indicates a general agreement in direction but not in wind speed. On figures 7 and 8, available 850-mb rawins at 1800 GMT are represented in the same manner as described for the surface-level analyses. 267 High cloud motions (shaft with arrowhead) derived from satellite data, figures 9 and 10, were supplemented by 150-mb rawins at 1200 GMT, plotted on-station. For comparison, the 200-mb observations represented by a broken shaft were placed on the reporting station, when feasible, or near it. > /' ia J "\(2) (2) (D> m 4fr 4 -±t> 5 \ 10 ^ 12) (if J K^tOz X lV y^i 20- -^ ^ vy 't S>N \80° 20 ^l v% *s tSjj^S i "s ^ :s 10 Figure 5. Surface -level streamline and isotaoh (kt) analyses, Hurricane Celia 1970 composite, 1200, 1500, 1800, and 2100 GMT, August 2, 1970. 268 The true level of maximum outflow for Hurricane Celia was not deter- minable. For a mature severe hurricane, this level is generally above the 200-mb pressure surface. The choice of the 150-mb level was made from the standpoint of best fit of satellite observations with rawin in- formation covering the entire analyzed region. V \ 10 \ 5 5U 10 / lit ' io\ \ / \\(4) V \ ■\\ A OCX / \ /\ JI \ Jr \ 1 1 - / /P 1 ' / Jc\ \ ^s y ^^^\ (ir^^^ \ /■ \i \ : if ~^^^r^ Hv J >20 ' \ Vi— ~-io 4 c ' /a. a I i i (^ iJWjY ji v v \j 7 1 1 v\ j v> - xa Vi \"\ aA°v ^v \^ \ \ J' *~^ ^ X*— y-- .- ■"-25 '--IS / ' \ i ' \ i * \ i V 11 > \ I \ \ ^ \ \ \ N \ \ - \ x J> \\ \ \ \ \ \ \ \ \iof x xN X. — ^X^" 1(1) / ^ — -^< \ \ \ \ \ X / \ / "v. ""* "* "* ^*^***^^^ \ V\ \J \\h ^-^^rrr~- ^ ~-io V \ s XX ' /N. ' \ \ \ \l ! / \ ' ^-20 \\5/ \ y -^e ^ Figure 6. Surface-level streamline and isotaoh (kt) analyses, Hurricane Celia 1970, 1800 GMT, August Z, 1970. 269 Figure 7. Low -cloud-lev el streamline and isotach (kt) analyses > Hurricane Celia 1970, 1526 to 1652 GMT, August 2, 1970. A clear example of high-cloud motion incompatibility with 200-mb rawins occurred on August 2 (fig. 9). Merida, on the Yucatan Peninsula of Mexico, reported a light west-northwest wind, indicating that at 200- mb a trough extended westward from a vigorous cold low in the western Caribbean, across the Yucatan Peninsula, to the Gulf coast of Mexico. This wind was in marked contrast with nearby cloud motions (in the vi- cinity) of some strength from the northeast that agreed with the 150-mb wind at Merida. Further, on August 3 (fig. 10), the high cloud motions over the Southeastern United States were more compatible with the 150-mb rawins. 270 20 \\ ! i 'i , \ 1 , \ x J i r ! »< - X I 20 \ i" 1 ' ~*\ ■»?» v—- 1 ' A -' \ It ,' \ \ ' ' ' \ \ ' T to / &/X y^±5 :-^J±o r ^~ ,10 3° j Si \ X yL L s t A V-J i \ \ A ki l V v ^6)\^^ \ // 1 < M C^^r 1 \ \ \ V s^^\vJ X\'/' 1 j \ \ / / \ 1 \ ; \ / ^i 1 \ \ — ^^5 ^^ \ * \ \v \ \ i \ Y^-—^ i \ \ B>/ ' 1 \ \ '\ \ ' \ \T ' \ \ 1 \ \ ' \ v * i \ \ \^ \ v rVsA v 1 ^5 \ \ re \ V \ \ — 1 \ ?K \r*\ Jk \ V J \ N X -20 Figure 8. Low-aloud-level streamline and isotaeh (kt) analyses, Hurricane Celia 1970, 1530 to 1657 GMT, August 3, 1970. 3.2 Analysis On August 2, similarities between the surface and low cloud analyses, figures 5 and 7, respectively, were the anticyclonic ridge which extended in an arc across the Gulf coast of the United States, the diffluent asymp- tote located northeast of the hurricane center, and the broad southerly- to-southeasterly confluent flow which extended from the western Caribbean across the Yucatan Channel and Peninsula. By contrast, at the surface level, the winds were 5 to 10 knots slower, the cyclonic storm circula- tion was smaller, the col was farther north, and the broad confluent flow was more easterly than at the low cloud level. 271 On August 3, the large-scale features of agreement between the sur- face and low cloud analyses, figures 6 and 8, respectively, remained un- changed except that the diffluent asymptote northeast of the center was absent on the surface analysis. In the latter analysis, a confluent asymptote in the right front quadrant was prominent. Another dissimilar MAX / - jo^-^^20 - - - -<*/,. 30/ : jo j v A y^\y ^^^40'' _—- — N \ 20 L) >/ / H V •* — f y 1 - $ 30^>^ \ ^sy^\. \N«\^ x yy \ t$K$o^ y/ ^\ — --5K >v\ T*\ v \ • hTv \ ^^ / ' 1 y^ \ / ' , fry \ it u fT 20\\ \ 1*&S\\\ ft ¥o) III 1 h' 4 V6-] r f*30 ~ y^ i' s'' io^yy\\ I / ( Tl \ \ \ ,1 1 1 \ v»- ' 1/ f"> 1 \yj> ' i TV \ \ 15 ---^yy / ^y^ ^^ $yy/* / x y / /' jf / fS A **N / ' / / i / — *^n! J / y } J ^a y f~ * \S '/ ) / ' t / V / * l ¥ / / / / toh^ r Figure 9. High- cloud-lev el streamline and ieotaah (kt) analyseSj Hurri- cane Celia 1970, 1526 to 1652 GMT, August 2, 1970. 272 feature was the angle between the surface and low-cloud wind directions, particularly in the broad southerly flow. This angle had increased in the 24-hour period. At the low cloud level, a major part of the southerly flow did not enter the storm circulation, but instead diverged around the anti cyclonic ridge. Figure 10. High- aloud-level streamline and isotach (kt) analyses, Hurri- cane Celia 1970, 1530 to 1657 GMT, August 3, 1970. 273 For a better understanding of the differences between the surface and low cloud motions, the shear vector for these two levels was computed at grid points on a 16 by 20 array. Details concerning the array are de- scribed in section 4. Of primary interest is the turning angle with height, shown as field quantities in figures 14e and 14f. In regions of inflow for both August 2 and 3, this angle varied between 20° and 40°. The high-cloud-motion streamline and isotach analyses for the 2nd and 3rd are shown in figures 9 and 10, respectively. The hurricane out- flow, which was confined to a region primarily northwest of the storm center on August 2, extended to the southwest on August 3, and the speeds significantly increased. 4. DIAGNOSTIC ANALYSIS Fields of radial and tangential wind components, absolute angular momentum flux, and absolute vorticity, shown in figures 11, 12, 13, and 14, respectively, were calculated from the surface-level, low-cloud, and high-cloud streamline and isotach analyses. A 16 by 20 grid array was used to compute the basic u- and v-compo- nents. The distance between two grid points represented 100.7 kilometers on the earth's surface. At grid points, radial and tangential components were calculated relative to the storm centers at 1600 GMT. At the two lower levels, this computation was made relative to the moving center, that is, by subtraction of the storm velocity at each grid point to elimi- nate asymmetries arising from superposition of vortex and steering cur- rent. At the high cloud level, this procedure was not adopted. No ma- chine smoothing of the initial u- and v-components nor of the subsequently derived fields was employed. A variable map factor was used in all field computations. On August 2, the bottom grid row was positioned at latitude 18°N and the beginning column at longitude 100°W. On August 3, the two positioning points were latitude 20°N and longitude 105°W, respectively. 274 At each grid point, the absolute vorticity and absolute angular mo- mentum flux were obtained from the following relations: absolute vorticity = |~ - |y + f (1) fR2 absolute angular momentum flux = V (V.R + —*— ) (2) r y c. where V is the tangential wind component, V is the radial wind component, R is the radial distance from the storm center to a grid point, and f is a variable Coriolis parameter. The diagnostic analyses illustrated by figures 11, 12, 13, and 14 are mainly self-explanatory; however, reference to certain figures will be made in subsequent sections. 5. COMPUTATIONS OF RADIAL AND TANGENTIAL VELOCITIES, ABSOLUTE ANGULAR MOMENTUM FLUX, RELATIVE VORTICITY, AND DIVERGENCE Using grid point data for the array described in section 4, average values of radial and tangential velocities and absolute angular momentum flux were computed around the storm center at radii of 2°, 3°, 4°, 5°, and 6°. Results of these calculations for the values at 1600 GMT on August 2 and 3 are shown in table 1. In addition, computations were made of average values of relative vorticity and divergence within the above circles defined by the follow- ing relations: average relative vorticity = -i— / V ds (3) iA* average divergence = -=— / V ds (4) where VQ, V , and R are defined as in equations (1) and (2) above, and ds is an element along the circumference. Results of these calculations are shown in table 2. 275 Table 1. Mean Radial and Tangential Velocities and Mean Flux of Absolute Angular Momentum Radius Meai ve i radial 1 ocity Mean tangential velocity Flux of mean absolute angular momentum (deg) (m sec-1) (m sec"1 ) (1013 m3 sec"2) Surface level 2 -2.0 15.0 -1.0 3 -1.2 9.7 -0.6 4 -1.2 6.5 -1.0 5 -1.3 4.7 -1.6 6 -1.4 3.6 Low cloud level -2.3 2 -2.0 14.2 -0.5 3 -1.6 11.8 -0.7 4 -2.0 10.0 -1.5 5 -1.7 8.1 -1.6 6 -1.3 6.0 High cloud level -1.5 2 7.4 -7.0 0.9 3 6.2 -6.0 1.7 4 5.1 -6.0 2.6 5 5.0 -5.6 3.9 6 3.6 -4.6 3.6 August 3 Surface level 2 -4.3 8.7 -1.3 3 -2.5 6.3 -1.2 4 -1.9 4.8 -1.4 5 -1.3 3.3 -1.4 6 -1.0 1.9 Low cloud level -1.2 2 -0.5 14.0 -0.1 3 +0.0 9.0 0.2 4 +0.0 7.4 0.3 5 0.4 6.2 0.8 6 0.9 3.7 High cloud level 1.7 2 10.3 -9.0 0.2 3 7.8 -7.9 1.1 4 6.7 -7.3 2.2 5 5.6 -7.1 2.7 6 4.5 -6.0 3.4 - Infl ow + Outflow 276 Table 2. Mean Are at Relative Vortioity and Mean Areal Divergence August 2 Augi 1st 3 Radius Areal rel . Areal Areal rel. Areal vorticity divergence vorticity divergence (cleg) (lO"5) do-5) Surface 1 evel (lO"5) (10-5) 2 13.5 -1.8 7.8 -3.9 3 5.8 -0.7 3.8 -1.5 4 2.9 -0.5 2.1 -0.8 5 1.7 -0.5 1.2 -0.5 6 1.1 -0.4 Low cloud level 0.6 -0.3 2 12.7 -1.8 12.6 -0.4 3 7.1 -1.0 5.4 +0.0 4 4.5 -0.9 3.3 +0.0 5 2.9 -0.6 2.2 0.1 6 1.8 -0.4 1.1 2.7 High cloud leve' i 2 -6.2 6.6 -8.1 9.2 3 -3.6 3.7 -4.7 4.6 4 -2.7 2.3 -3.3 3.0 5 -2.0 1.8 -2.6 2.0 6 -1.4 1.1 -1.8 1.3 The tabular values may be regarded as samples at unknown heights nthin the layers of low-level inflow (surface and low cloud motions) and ngh-level outflow (high cloud motion). It was not possible to determine /nether, for example, the high-cloud-level radial velocities (table 1) :orrespond to maximum values or layer means. Further, it can be stated /ith certainty only for the surface level that the sampling took place it the same height level at each computation time. 5.1 Average Radial Velocity— Table 1 At all radii, the radial velocities for the high cloud level were larger on August 3 than on August 2. The percent of velocity increase from the 2nd to the 3rd was almost the same at each radius, averaging 277 Figure 11. Radial wind component field, by level, 1600 GMT. 11a. Surface level, August 2, 1970; lib. Surface level, August 3, 1970; lie. Low aloud level, August 2, 1970; lid. Low aloud level, August 3, 1970; lie. High aloud level, August 2, 1970; and llf. High aloud level, August 3, 1970. 278 about 25 percent. The surface -lev el rates at radial distances of 2°, 3°, and 4° were also larger on the 3rd than on the 2nd. Because radial veloc- ities for both the surface and high cloud levels were higher at the second computation time, it is concluded that they reflected storm inten- sification. On August 2, radial velocities for the low aloud level were nearly the same as those for the surface level, but they were markedly differ- ent on August 3. On that latter date, except at the radius of 2°, small values of outflow occurred. In section 3, descriptive differences be- tween the two stream levels were discussed. Tabulated results indicate that the heights of the low-cloud motion levels were not identical on the 2 days. On the 3rd, the sampling may have occurred above the low-level inflow layer which, because of increased storm intensity, was probably thinner than on the 2nd. Figure 11 shows the radial component fields for the surface, low cloud, and high cloud levels. The surface and low-cloud motion fields on August 2 indicate that the principal inflow region was located southeast of the storm center along an axis through the Yucatan Channel. On August 3, the same motion fields indicate that this inflow region was located northwest of the storm center. At the high cloud level on the 2nd, figure lie, the outflow was ori- ented along an axis southeast to northwest, symmetrically opposite to the inflow orientation at the two low levels. On the 3rd, figure llf, this symmetry weakened as outflow spread to the southwest. 5.2 Average Tangential Velocity— Table 1 At the surface level, tangential velocities at all computed radii were smaller on August 3 than on August 2. Except at the radius of 2°, the tangential velocities for the low-aloud- level motions show the same decreasing trend on the second day. Figure 12 shows the tangential component fields for the surface, low cloud, and high cloud levels. From near circular symmetry at both low levels on the 2nd, figures 12a and 12c, respectively, the tangential 279 Figure 12. Tangential wind component field, by level, IS 00 GMT. 12a. Surface level, August 2, 1970; 12b. Surface level, August 3, 1970; 12c. Low cloud level, August 2, 1970; 12d. Low cloud level, August 3, 1970; 12e. High cloud level, August 2, 1970; and 12 f. High cloud level, August 3, 1970. 280 motion field became appreciably asymmetric on the 3rd, caused partly by the storm's close proximity to land. At all computed radii, the tangential velocities for the high cloud level were higher on the 3rd than on the 2nd, another indication of storm intensification. 5.3 Average Absolute Angular Momentum Flux— Table 1 The absolute angular momentum flux values for the surface level on both days were nearly the same. This may be partially explained by an examination of the tabular radial and tanqential velocities for radii of 2°, 3°, and 4° which enter into equation (2). On August 3, tangential velocities decreased at those computed radii while radial velocities in- creased. Also, on either day, average values of absolute angular momen- tum flux among the radii showed little variation. At the low-cloud-mtior\ level, the average absolute angular momen- tum flux values were nearly the same as those at the surface level on August 2. On August 3, however, the low-cloud-flux values were positive outward beyond the radius of 2°. This anomaly is attributed to the sign of the radial velocity component. The diagnostic fields of radial veloc- ity (fig. 11) and of absolute angular momentum flux (fig. 13) illustrate the importance of this sign. At the high-cloud-mt\or\ level, the average absolute angular momen- tum flux values show a consistent increased outward flux with radial distance from the center and decreased values at all computed radii on the 3rd over those on the 2nd. 5.4 Average Vorticity and Divergence— Table 2 Values of relative vorticity and divergence, representing averages for the circular areas of specified radii, are shown in table 2. Field analyses of absolute vorticity for motions at the surface, low cloud, and high cloud levels are shown in figure 14a, 14b; 14c, 14d; and 15a, 15b; respectively. 281 ^5 -K^ '!i 1 5 16 26 I1 v figure 13. Abeolute angular momentum flux field, by level, 1600 GMT. 13a. Surface level, August 2, 1970; 13b. Surface level, August 3, 1970; 13c. Lew cloud level, August 2, 1970; 13d. Low cloud level, August 3, 1970; 13e. High cloud level, August 2, 1970; and 13 f. High cloud level, August 3, 1970. 282 Figure 14. Absolute vortieity field, by level, 1600 GMT, and field of turning angle of low-cloud-level and surface-level shear vector, 1600 GMT. 14a. Absolute vortieity field, surface level, August 2, 1970; 14b. Absolute vortieity field, surface level, August 3, 1970; 14c. Absolute vortieity field, low cloud level, August 2, 1970; 14d. Absolute vortieity field, low cloud level, August 3, 1970; 14e. Field of turning angle of low-cloud-level and surface-level shear vector, August 2, 1970; and 14f. Field of turning angle of low-cloud-level and surface-level shear vector, August 3, 1970. 283 6. DYNAMIC INSTABILITY IN THE OUTFLOW LAYER As a possible contributor to the intensification of Hurricane Celia, the dynamic instability of the upper tropospheric outflow on August 2 and 3 was investigated. According to Alaka (1963), for a symmetrical station- ary vortex when baroclinic effects are negligible, one criterion for dy- namic instability is Ca(jp+f)<0 (5) where V is the wind speed, R. is the radius of curvature of the air tra- jectories, and c is the absolute vorticity given by equation (1). a Following Alaka (1961), an approximate criterion for dynamic insta- bility is Ca(j^-+f)<0 (6) where R is the radius of curvature of the streamlines. The radius of streamline curvature, R , was evaluated by a method used by Alaka (1962) that assumes local time derivatives are negligible compared to advective terms. The total wind field was used to calculate the instability criterion, equation (6), in that the eddies associated with the upper tropospheric outflow were of the same order of magnitude in horizontal extent as the mean outflow. Figures 15a and 15b show the fields of absolute vorticity at the high- cloud-motion level for August 2 and 3, respectively. On the 2nd, there were two isolated areas of negative absolute vorticity of small magnitude located southeast and southwest of the storm center. On the 3rd, a siz- able region of negative absolute vorticity was located northwest of the storm center. Figures 15c and 15d show the fields of anomalous winds at the high- cloud-motion level for August 2 and 3, respectively. On the 3rd, an ex- tensive anomalous wind area was located approximately 6° of latitude west of the storm center. 284 Figure 15. Absolute vortieity , high, cloud motion, 1600 GMT; anomalous winds, 1600 GMT; and product of absolute vortieity and anomalous winds, 1600 GMT. 15a. Absolute vortieity, August 2, 1970; 15b. Absolute vortieity, August 3, 1970; 15c. Anomalous winds, August 2, 1970; 15d. Anomalous winds, August 3, 1970; 15e. Product of absolute vortieity and anomalous winds, August 2, 1970; and 15f. Product of absolute vortieity and anomalous winds, August 3, 1970. 285 Figures 15e and 15f show the product field of absolute vorticity and anomalous wind for August 2 and 3, respectively. On the 2nd, there were areas where this product was negative, indicating dynamic instability; however, the negative areas may be considered insignificant in compari- son to surrounding regions of large positive values. On the 3rd, there was one region of instability west of the storm center slightly more in- tense than the small areas present on the 2nd. On both days, areas of positive and negative absolute vorticity were well correlated with normal and anomalous winds. This condition is in agreement with the results of a study involving the hurricane outflow layer by Black and Anthes (1971). 7. OCEANIC INFLUENCES RELATIVE TO INTENSITY VARIATIONS OF HURRICANE CELIA An important oceanic influence favorable to disturbance intensifica- tion is an adequate heat source. While researchers have long recognized the significance of the sea-surface temperature, they have also given another major oceanic influence, the depth of the upper mixed layer, in- creased attention within the past decade. According to a study by Jordan and Frank (1964), a mixed layer depth of about 30 meters is required to pre- vent cold water mixing from below. In the case of Hurricane Celia, an attempt was made to relate both of these oceanic influences to the ob- served intensity variations illustrated by figure 2a. Fortunately, near the time of interest, the Research Vessel (R/V) Alaminos of Texas A&M Uni- versity traversed the Gulf of Mexico and the Caribbean Sea, making nu- merous surface and bathythermograph observations- Figures 16a and 16b represent sea-surface temperature analyses (in °F for the former; in °C for the latter) for the period before Hurricane Celia's entrance into the Gulf of Mexico approximately 0000 GMT on August 1. Figure 16a is a 7-day (July 24-30) composite analysis transposed to a Mercator projection, made on a operational basis by the Fleet Weather Cen- tral, Norfolk, Va. Figure 16b is a 4-day (July 29-August 1) composite analysis, enlarged to include the entire Gulf of Mexico, made on a research Figure 16a. Seven-day sea-surface temperature (°F) composite analysis (July 24 through July 303 1970) by Fleet Weather Central 3 Norfolk, Va, TEMPERATURE Figure 16b. Four-day sea-surface temperature (°C) composite analysis (July 29 through August ls 1970) by author. 287 basis by the author from data supplied by the National Climatic Center of NOAA. The same assumption underlies both analyses; namely, that at a point on the ocean surface, the temperature did not vary appreciably for the period involved. Figures 16a and 16b show a cold trough of sea-surface temperature, oriented approximately southeast-northwest, extending from the Caribbean, through the Yucatan Channel, to the central Gulf of Mexico. The track of the developing hurricane, figures 1 and 17b, was initially east of the trough over waters that were 3° to 4°F warmer than the trough. Coinci- dental with the end of the first deepening phase, at 0000 GMT on August 2, the storm was centered over the cold trough. At 0600 GMT on August 3, corresponding to the final explosive deepening phase, the center of the storm was once more over a warm sea surface. Figure 17a shows the track of the R/V Alaminos to and from the Yuca- tan Channel, with the sea-surface temperature (in °C) determined by bucket measurement plotted at each vessel station. The correspondence between the bucket measurements of the R/V Alaminos and the sea-surface tempera- ture distributions of figure 16a and 16b, where comparable, is considered acceptable. Figure 17b shows the depth of the upper mixed layer (in m) measured from the bathythermograph traces recorded by the R/V Alaminos. The mixed layer depths on the inbound and outbound passages averaged about 40 and 15 meters, respectively. The inbound observations represent in-depth conditions shortly before Hurricane Celia's entrance into the Gulf of Mexico. Measurements on the outbound passage, although taken about 3 weeks before those of the inbound passage, may also be considered repre- sentative during Celia's passage. This assumption considers that the depth and stability of the upper mixed layer were determined by the ex- tent and character of the eastern Gulf of Mexico loop current. On its inbound passage to a position at about latitude 26°N, the vessel reported large mixed layer depths characteristic of the eastern side of the anti- cyclonic current. It is assumed that the recorded shallow measurements on the outbound passage, particularly those near the storm path, reflected in-depth conditions outside the western side of the current. 288 BUCKET TEMP CCI OUTBOUND BUCKET TEMP PC) INBOUNO Exocnple - [7V4 17 1) XBT CONSECUTIVE SLIDE NO : DATE ANO TIME Of OBSERVATION (GMT) STATION; BUCKET TEMP CCI Figure 17a. Bucket temperatures recorded by Texas ASM University Research Vessel Alaminos on Cruise No. 70-A-10. Figure 17b. Depth of mixed layer measured from expendable bathythermo- graph trace recorded by Texas ASM University Research Vessel Al aminos 289 Figure 17b shows that between 1200 and 1800 GMT on August 1, during the initial deepening phase of Celia, the center of the storm passed over an ocean area of large mixed depths, and that at 1800 GMT on August 2, during the steady-state phase of the hurricane, the storm center was near a region of very shallow mixed depths. The relation between hurricane intensity and mixed layer depths in the final deepening phase, which began at about 0600 GMT on August 3, must be pursued indirectly because the observational records of the R/V Alaminos ended in the mid-Gulf. On figure 17b, the 10-, 20-, and 50- fathom isobaths for the Texas and Louisiana Shelf are shown. Histori- cally, in the upper 30 meters, the average temperature difference between the upper and lower boundaries is 2° to 3°C for this coastal region (Leipper and Volgenau, 1970). Figure lib shows that at 1600 GMT on Au- gust 3 at the surface level, the storm inflow was principally from the north through southwest directions over shallow shelf waters. Figures 18a and 18b represent analyses of surface air temperature (solid lines) and of dew-point temperature (dashed lines), both given in °C at approximately 1600 GMT on August 2 and 3, respectively. All land- based reports were obtained from 6-hourly station collectives at 1800 GMT. Data contained in figures 18a and 18b were obtained from same vessel ob- servations as those in figures 5 and 6. However, the reported air tem- peratures were plotted at actual ship positions. Figures 19c and 19d represent sea-surface temperature analyses, given in °C, at approximately 1600 GMT on August 2 and 3, respectively. The same vessel observations as used in figures 5 and 6 were employed without ship position displacement. In sparse-reporting areas, these analyses were extrapolated on the basis of composites, figures 16a and 16b. The dashed isolines indicate where use was made of this information. Figures 19a and 19b represent analyses of the surface air temperature minus the sea-surface temperature, both given in °C, obtained from the sub- traction of the analyses of figures 16a and 19c and of figures 16b and 19d, The outstanding feature of both figures 19a and 19b is the location of the region ahead of the storm path which offered the greatest potential for evaporation. 290 Figure 18a. Air and dew-point temperature (°C) at surface level, ^1600 GMT, August 2, 1970. 8/3/70 Figure 18b. Air and dew-point temperature (°C) at surface level &1600 GMT, August 3, 1970. 291 8/2/70 Figure 19. Air minus sea-surface temperature analysis, ^1600 GMT, and sea-surface temperature analysis, ~1600 GMT. Temperatures in °C. 19a. Air minus sea-surface temperature analysis, August 2, 1970; 19b. Air minus sea-surface temperature analysis, August 3, 1970; 19c. Sea-surface temperature analysis, August 2, 1970; and 19d. Sea-surface temperature analysis, August 3, 1970. In summary, a probable contributing cause of filling after the ini- tial deepening to 965 mb was Hurricane Celia's movement over a preexist- ing trough of relatively cold sea-surface temperature and shallow mixed depths. Figure 11a shows that at 1600 GMT on August 2, the time of the first satellite motion analysis, inflow at the surface level was through the Yucatan Channel over relatively cold waters. Subsequently, from evi- dence of additional satellite film loops (Simpson and Pelissier, 1971), the southerly inflow was displaced westward over the Yucatan Peninsula. 292 Presumably, this later surface inflow came in contact with warmer Gulf waters. From an unknown time, inferred to be near 1303 GMT on August 3, when an elliptical eye shape was first detected by reconnaissance air- craft, the surface inflow changed direction. Final deepening is ascribed to energy transferred to the inflowing air while passing over the warm shelf waters of shallow depths. 8. SUMMARY AND CONCLUSIONS Factors considered in this study which may have influenced the central pressure variations of Hurricane Celia were the sea-surface temperature, the depth of the upper ocean mixed layer, and the air-sea temperature dif- ference. The horizontal gradient of these parameters in conjunction with the region of low-level inflow was emphasized rather than their absolute values. Intensification resulting from dynamic instability in the outflow layer in connection with the final rapid deepening before landfall was not indicated by this study, but may not be ruled out. More dense and accurate data than were available and more stringent assumptions than were used in the criteria may be necessary for a valid evaluation. The surface-level and high-cloud motion radial velocities and the high-cloud motion tangential velocities verified the rapid fall in central pressure observed between August 2 and 3. However, computations of the low cloud motions indicated that they did not apply to the inflow layer on August 3. Cloud motions derived from satellite pictures may be of considerably more value, provided that the height levels at which they apply are known. An arbitrary assumption of cloud motion heights is ap- parently not warranted because of the appreciable vertical wind shear and divergent flow known to exist in the hurricane motion field. 9. ACKNOWLEDGMENTS The author thanks Banner I. Miller and Brian Jarvinen for the use of their averaging program to compute the guantities shown in tables 1 and 2. Also, words of thanks are extended to Willis Pequegnat, Department of 293 Oceanography, Texas A&M University, College Station, Tex., for the tem- perature and bathythermograph data on the R/V Alaminos Cruise 70-A-10 and to the U.S. Naval Oceanographic Office, Washington, D.C., for the sea- surface temperature analysis shown in figure 16a. In addition, my appreciation is extended to R. Cecil Gentry for his critical review and comments, to Harry Hawkins for his numerous suggestions and guidance, to Robert Carrodus and staff for their drafting, to Charles True for his photography and reductions, to Arnold Recht for his assistance in reducing the satellite photographs to film loops, and to Cathy Delgado for her typing. 10. REFERENCES Alaka, M. A. (1961): The occurrence of anomalous winds and their signifi- cance, Monthly Weather Review, 89, 482-494. Alaka, M. A. (1962): On the occurrence of dynamic instability in incipient and developing hurricanes, Monthly Weather Review, 90, 49-58. Alaka, M. A. (1963): Instability aspects of hurricane genesis, National Hurricane Research Project, Report No. 64, 23 pp. Black, P. G., and R. A. Anthes (1971): On the asymmetric structure of the tropical cyclone outflow layer, Journal of the Atmospheric Sciences, 28, 1348-1366. Hughes, L. A. (1952): On the low-level wind structure of tropical storms, Journal of Meteorology , 9, 422-428. Fisher, E. L. (1958): Hurricanes and the sea-surface temperature field, Journal of Meteorology, 15, 328-333. Jordan, E. S. (1952): An observational study of the upper wind-circulation around tropical storms, Journal of Meteorology, 9, 340-346. Jordan, C. L. (1964): On the influence of tropical cyclones on the sea surface temperature field, Proceedings Symposium on Tropical Meteor- ology (Rotorua, Nov. 5-13, 1963), ed., J. W. Hutchings, New Zealand Meteorological Service, Wellington, 614-622. Jordan, C. L., and N. L. Frank (1964): On the influence of tropical cy- clones on the sea surface temperature field, NSF Grant GP-621, Florida State University, Tallahassee, Nov. 15, 1964, 31 pp. plus illustrations and tables. Krueger, D. W. (1959): A relation between the mass circulation through hurricanes and their intensity, Bulletin of the American Meteorologi- cal Society, 40, 182-189. 294 Leipper, D. F. (1967): Observed ocean conditions and Hurricane Hilda, Journal of the Atmospheric Sciences, 24 , 182-196. Leipper, D. F., and D. Volgenau (1970): Hurricane heat potential of the Gulf of Mexico, talk presented at April 22, 1970, meeting of the American Geophysical Union, Washington, D.C., release for use by science writers . Miller, B. I. (1957): On the maximum intensity of hurricanes, National Hurricane Research Project, Report No. 14, 19 pp. Miller, B. I. (1958): The three-dimensional wind structure around a trop- ical cyclone, National Hurricane Research Project, Report No. 15, 41 PP. Perlroth, I. (1967): Hurricane behavior as related to oceanographic en- vironmental conditions, Tellus, 19, 258-268. Simpson, R. H., and J. M. Pelissier (1971): Atlantic hurricane season of 1970, Monthly Weather Review, 99, 269-277. Tisdale, C. F., and P. F. Clapp (1963): Origin and paths of hurricanes and tropical storms related to certain physical parameters at the air-sea interface, Journal of Applied Meteorolory, 2, 358-367. 295 28 U.S. DEPARTMENT OF THE NAVY J.W. WARNER, Secretary Naval Weather Service Command R.M. CASSIDY. Captain, USN, Commander **:°'c% \ U.S. DEPARTMENT OF COMMERCE P.G. PETERSON, Secretary National Oceanic and Atmospheric Administration R.M. WHITE, Administrator **r«<*,r PROJECT STORM FURY ANNUAL REPORT 1971 MIAMI. FLORIDA JUNE 1972 296 Project STORMFURY was established by an Interdepartmental agreement between the Department of Commerce and the Department of Defense, signed July 30, 1962. Additional support has been provided by the National Sci- ence Foundation under Grant NSF-G-17993. This report is the tenth of a series of annual reports to be pre- pared by the Office of the Director in accordance with the Project STORM- FURY interdepartmental agreement. Additional copies of this report may be obtained from: U.S. Naval Weather Service Command Department of the Navy Washington Navy Yard Washington, D.C. 20390 or National Hurricane Research Laboratory P.O. Box 8265, University of Miami Branch Coral Gables, Florida 33124 NOTICE The National Hurricane Research Laboratory and the Naval Weather Service Command do not approve, recommend, or endorse any proprietary product or proprietary material mentioned in this publication. No reference shall be made to either organ- ization or to this publication in any advertising or sales pro- motion which would indicate or imply that the National Hurricane Research Laboratory or the Naval Weather Service Command approves, recommends, or endorses any proprietary product or material men- tioned herein, or which has as its purpose an intent to cause directly or indirectly the advertised product to be used or pur- chased because of this publication. 297 TABLE OF CONTENTS INTRODUCTION 1 HISTORY AND ORGANIZATION ! PROJECT STORMFURY ADVISORY PANEL 3 PUBLIC AFFAIRS 3 PYROTECHNIC DEVICES - SILVER IODIDE 4 AREAS OF OPERATIONS 4 PLAN FOR FIELD OPERATIONS - 1971 4 FIELD OPERATIONS - Dry Runs and Cloudline 6 FIELD OPERATIONS - Hurricane Ginger 8 RESEARCH ACTIVITIES 9 OPERATIONAL AND RESEARCH DATA COLLECTION 11 OUTLOOK FOR 1972 H REFERENCES AND SPECIAL REPORTS 12 APPENDIX A: PROJECT STORMFURY ADVISORY PANEL RECOMMENDATIONS 3-4 December 1971, Washington, D.C. 17 APPENDIX B: REPORT ON SEEDING OF HURRICANE GINGER 19 Harry F. Hawkins, Kenneth H. Bergman, and R. Cecil Gentry APPENDIX C: HURRICANE MODELING AT THE NATIONAL HURRICANE RESEARCH LABORATORY (1971) 55 Stanley L. Rosenthal APPENDIX D: THE MUTUAL INTERACTION OF HURRICANE GINGER AND THE UPPER MIXED LAYER OF THE OCEAN 63 Peter G. Black and William D. Mallinger APPENDIX E: MODELING THE SEEDING EFFECT IN HURRICANE GINGER 89 W. D. Scott and C. K. Dossett APPENDIX F: THE TYPHOON RESEARCH PROGRAM AT THE ENVIRONMENTAL PREDICTION RESEARCH FACILITY AS AN AID TO PROJECT STORMFURY 107 Samson Brand APPENDIX G: DIURNAL VARIATION IN HURRICANES 121 Robert C. Sheets APPENDIX H: NAVAL WEAPONS CENTER CONTRIBUTIONS TO STORMFURY DRY-RUN AND CLOUDLINE OPERATIONS 2 to 13 August 1971 127 Shelden D. Elliott, Edward E. Hindman, II, and William G. Finnegan APPENDIX I: SOME STATISTICAL CHARACTERISTICS OF THE HURRICANE EYE AND MINIMUM SEA-LEVEL PRESSURE 143 Robert C. Sheets 298 APPENDIX J: TYPHOON SEEDING ELIGIBILITY IN THE WESTERN NORTH PACIFIC 157 Wi 1 1 iam D. Mai 1 inger APPENDIX K: COMPARISONS OF RAINFALL AND RADAR MEASUREMENTS IN TROPICAL STORM FELICE AND HURRICANE DEBBIE 165 W. D. Scott and C. K. Dossett 299 PROJECT STORMFURY ANNUAL REPORT 1971 INTRODUCTION This was a year of excitement and progress for Project STORMFURY. For the first time extensive sea-temperature and bathythermograph obser- vations were obtained in a hurricane. The hurricane models were improved, and great progress made in the interpretation of results from model ex- periments. In addition, a modification experiment was conducted on a hurricane. The 1971 hurricane season produced only one eligible storm (Hurri- cane Ginger) upon which to experiment. Although first priority of the Project was to repeat the eyemod experiment conducted in Hurricane Debbie in 1969, "Ginger" was not the type of storm for this experiment. In fact, Ginger was so large and diffuse, had such flat gradients of pressure and winds, and so few clouds suitable for seeding, that she was a poor sub- ject for even the rainsector type experiment that was accomplished on 26 and 28 September. In spite of "Ginger's" lack of suitability for our planned experiments, extremely valuable data were collected that will fa- cilitate research on this type of storm. In addition, a great deal of experience was gained in conducting the "Rainsector Experiment." This was the first time it was used in the STORMFURY operations. In addition to the Ginger experiments, STORMFURY forces operated to- gether in dry-run exercises and in cloudline experiments. The dry-run exercises were conducted from the Naval Station Roosevelt Roads and Ramey Air Force Base, Puerto Rico, on 3 to 6 August followed by cloudline ex- periments on 8 to 14 August at Barbados. Forces remained on alert but did not actually deploy for STORMFURY again until 25 September for Hurri- cane Ginger. Research efforts were again intensified as evidenced by the appen- dices in this report. During the year, much effort was expended in making plans for the move of the Project to the Pacific for part of the 1972 typhoon season. This move was not approved, and the project is currently planning to op- erate only in the Atlantic area during 1972. HISTORY AND ORGANIZATION Project STORMFURY is a joint Department of Commerce (NOAA) -Department of Defense (Navy) program of scientific experiments designed to explore the structure and dynamics of tropical cyclones and their potential for 300 modification. The Project, which was formally established in 1962, has as its principal objective experimentation to reduce the intensity of hurricanes. The experiments involve strategic seeding of hurricanes with silver iodide crystals dispersed from aircraft with pyrotechnic devices developed by the U.S. Navy. Navy and NOAA scientists and aircraft, sup- plemented by those of the U.S. Air Force, have cooperated in STORMFURY ex- perimental operations since 1961, when the first informal agreement was proposed. To date, the experiments conducted by the Project consist of Hurricane Esther - seeded in 1961 - Single seeding on each of 2 days Hurricane Beulah - seeded in 1963 - Single seeding on each of 2 days Tropical Cumulus Cloud Seedings - 1963 Tropical Cumulus Cloud Seedings - 1965 Tropical Cloudline Seedings - 1968 Tropical Cloudline Seedings - 1969 Hurricane Debbie Seedings - 1969 - Multiple seedings (eyewall experi- ment) on each of 2 days Tropical Cloudline Seedings - 1970 Hurricane Ginger Seedings - 1971 - Multiple seedings (rainsector ex- periment) on each of 2 days Tropical Cloudline Seedings - 1971 Since 1962, only three hurricanes have been seeded. The results of the Hurricane Debbie multiple seeding experiments conducted on 18 and 20 August 1969 were extremely encouraging in that a decrease in the maximum wind velocity of the hurricane was observed on both days. Although not conclusive, time sequences of the wind, radar, and other data strongly suggest that a modification to Hurricane Debbie was achieved. The Hurricane Ginger rainsector seeding experiments on 26 and 28 September 1971 were performed on a poorly defined, diffuse storm. This makes direct comparison with results obtained in Hurricane Debbie impos- sible. Some modification to the clouds in Ginger did occur as a result of seeding, but little effect on the hurricane as a whole was expected and any that occurred is difficult to assess (see appendix B). Dr. Robert M. White, Administrator of NOAA, and Rear Admiral W. J. Kotsch, U.S. Navy, Commander Naval Weather Service Command, who was suc- ceeded by Captain R. M. Cassidy, U.S. Navy, in September 1971, had over-all responsibility for the cooperatively administered project. The Project Director in 1971 was Dr. R. Cecil Gentry, Director of the National Hurricane Research Laboratory (NHRL), Miami, Florida. The Alter- nate Director was Dr. Harry F. Hawkins, also of NHRL. The Assistant Pro- ject Director and Navy Project Coordinator was Captain L. J. Underwood, U.S. Navy, Commanding Officer of the Fleet Weather Facility, Jacksonville, Florida (FLEWEAFAC JAX). The Alternate to the Assistant Project Director was Lt. Comdr. G. Oakes, U.S. Navy, also of FLEWEAFAC JAX. Mr. Jerome W. Nickerson, Environmental Prediction Research Facility, Monterey, California, was Technical Advisor to the Navy; Dr. S. D. Elliott, Jr., Naval Weapons Center, China Lake, California, was NWC Project Officer; Mr. Max Edelstein, Naval Weather Service Command Headquarters, Washington, D.C., was assigned liaison duties representing the Navy; and Mr. William D. Mai linger (NHRL) 301 was assigned liaison duties for the Project Director and NOAA and acted as Data Quality Control Coordinator. PROJECT STORMFURY ADVISORY PANEL The Advisory Panel of five members represents the scientific commun- ity and provides guidance concerning the various scientific and technical problems of the Project. Their recommendations have proved of great value to the Project since its inception. The Panel reviews results from previous experiments, proposals for new experiments and their priorities, the effectiveness of data collec- tion systems, evaluation of data collected, eligibility criteria for storms to be seeded, and other items as applicable. During 1971, the Advisory Panel consisted of the following prominent scientists: Professor Noel E. LaSeur, Chairman (Florida State University), Dr. Roscoe R. Braham, Jr. (University of Chicago), Professor Charles L. Hosier (The Pennsylvania State University), Professor Norman A. Phillips (Massachusetts Institute of Technology), and Professor Jerome Spar (New York University). The Panel met in Washington, D.C., on 3 and 4 December 1971, to dis- cuss various problems and to be briefed on the research results obtained from the seeding of Hurricane Ginger. Some of the items discussed were the status of hurricane modeling research, operational plans for calendar year 1972, proposal to move the experimental program to the Pacific in CY 1972, and pyrotechnics and data collection problems. The untimely death of Professor James E. McDonald in June 1971 was a blow to the Advisory Panel, but he has been ably replaced by Professor Roscoe R. Braham, Jr., of the University of Chicago. Recommendations from the December 1971 Panel Meeting are included in this report as appendix A. PUBLIC AFFAIRS A coordinated press release and fact sheet for STORMFURY were dis- tributed to the media before the experimental season. The TV film clip, which had been disseminated to various television stations in the hurri- cane belt, was not updated because it was considered essentially current. During the cloudline experiments in Barbados, several newspaper arti- cles evolved and one taped interview with the Project Director was played on local radio. During the seeding of Hurricane Ginger, considerable media interest was generated. Seats in the Project aircraft were made available and several reporters and photographers made flights into the storm. Even though the aircraft operated from six different bases, no serious problems arose and all requests for information and interviews were fulfilled. 302 PYROTECHNIC DEVICES - SILVER IODIDE The pyrotechnics prepared for the 1971 season were similar to the STORMFURY I unit used in the 1970 seeding experiments. This unit, developed under the leadership of Dr. Pierre St. Amand of the Naval Weapons Center, China Lake, California, was provisionally designated WMU-2(XCL-1)/B. The pyrotechnic grain of this unit is similar in composition and per- formance to that of earlier ones, but the unit incorporates a pressure re- lief, bore-safety, and time delay function that permit it to be certified for general use in all appropriate racks and aircraft without special sup- ervision. More details of the pyrotechnics used can be found in appendix D of the 1969 Project STORMFURY Annual Report and appendix H of this report. AREAS OF OPERATIONS Eligible areas for experimentation in 1971 were the Gulf of Mexico, the Caribbean Sea, and the southwestern North Atlantic. Operations in these areas were limited by the following guideline: A tropical cyclone was considered eligible for seeding as long as there was only a small probability (10 percent or less) of the hurricane center coming within 50 n miles of a populated land area within 18 hours after seeding. There are two primary reasons for not seeding a storm near land. (1) A storm seeded farther at sea will have reverted to its natural state be- fore affecting and possibly damaging a land area. (2) Large changes in the hurricane structure occur even when portions of its circulation pass over land. These land-induced modifications would obscure the effects ex- pected to be produced by the seeding experiments and greatly complicate the scientific evaluation of the results. PLAN FOR FIELD OPERATIONS - 1971 The period 2 August to 31 October was established for STORMFURY op- erations in 1971. Aircraft from the following units were planned as STORMFURY forces during the season: Navy Weather Reconnaissance Squadron FOUR Navy Attack Squadron EIGHTY-FIVE NOAA Research Flight Facility Air Force 53rd Weather Reconnaissance Squadron Air Force 55th Weather Reconnaissance Squadron Air Force 58th Weather Reconnaissance Squadron Naval Weapons Center Operations Plan No. 1-71 was provided to participants. It covered flight operations, communications, instrument calibration and use, data collection and distribution, logistic and administrative procedures, airspace reservations agreements, and public affairs. 303 The plan also provided for a series of fall -back research missions to be used when no eligible hurricane was available for seeding after deployment of project forces. These research missions are primarily data gathering or storm monitoring missions in unseeded cloud systems or storms. As recommended by the STORMFURY Advisory Panel, first priority was given to the eyemod experiment in order to gain additional data that could be correlated with those collected during the 1969 "Debbie" seeding exper- iments. This multiple seeding of the clouds in the annulus radially outward from the maximum hurricane winds calls for five seedings at 2-hour inter- vals. Each seeding consists of dropping 208 pyrotechnic units along a radially outward flight path, starting just outside the radius of maximum winds. The hypothesis in 1969 and early 1970 stated that the introduction of freezing nuclei (silver iodide crystals produced by the pyrotechnics) into the clouds in and around the eyewall should cause a chain of events that includes the release of latent heat, development of a new eyewall at a greater radius than the original one, changes in temperature and pressure gradients, and a reduction in maximum winds. Data from several experiments and individual cases are needed before definite conclusions regarding the validity of this hypothesis can be assumed. Because the magnitude of natural variations in hurricanes is some- times as large as the hypothesized artificially induced changes, it is frequently difficult to distinguish between the two. Second priority was given to the rainsector, and third to the rain- band experiments. The rainsector experiment tests whether some of the latent energy in the air flowing toward the center of the hurricane can be intercepted and released while it is still between 50 and 100 n miles from the center. If successful, this experiment should result in the dispersal of the energy over a larger area, rather than concentration near the center. Clouds in a 45° sector between about 50 and 75 n mile radii are seeded to stimulate growth. This sector is selected where an abundance of warm moist tropical air is being carried by the low-level winds toward the clouds nearer the center of the storm. If stimulated cloud growth in the seeded sector causes moist air to ascend to the outflow layer at a relatively large radius, some of the energy normally released near the center of the storm would be released at greater radii and could result in a reduction in the storm's maximum winds. All suitable clouds in the designated sector are seeded. The seed- ings are made in four periods of 50 minutes each, separated by nonseeding periods of 50 minutes. During the latter periods, the monitoring aircraft collect data in the seeding area for documenting changes in the cloud structure. These data supplement the other data collected to measure changes of the storm's intensity. The rainband experiment has the same objective as does the rainsec- tor experiment and, in addition, should permit opportunities to study the interaction of seeded clouds with other clouds in the same and nearby 304 rainbands. Clouds are seeded along a rainband (a line of clouds spiral- ing around and toward the center of the storm) at 50 to 150 n miles from the storm center. Seeding such a rainband may produce a dispersion of the energy of the hurricane over a larger area and should provide informa- tion and data needed to improve the design of other modification experi- ments. The rainband experiment provides data needed for studies of cloud interactions. To facilitate visual observation, a rainband can be selected that is well removed from the central vortex area and not obscured by the main cloud system of the hurricane. The Advisory Panel has recommended that cloudline experiments contin- ue to be conducted to collect data for studying the dynamics of clouds organized into systems such as rainbands. These experiments can be con- ducted when there are no hurricanes and should provide additional oppor- tunities for evaluation of seeding effects. During these experiments, various seeding agents and dispersing techniques can also be tested. Cloudline experiments were scheduled for 8 to 13 August from Barbados. Project STORMFURY field experiments are extremely complex operations that require extensive planning and effective coordination. During the multiple seeding experiments, there are as many as 12 aircraft simultane- ously operating in the hurricane circulation. Safety of the aircraft and personnel is paramount throughout the experiment. It is obvious that training, professionalism, and dedication are vital to safe and success- ful operations in the weather extremes encountered. Aircraft, communica- tion systems, data collection systems, and data recording systems must be in peak operating condition. The seeder aircraft must be carefully and accurately vectored by radar and voice communications for the seeding runs. Teamwork is mandatory. For these reasons, dry-run exercises must be conducted before operations in a hurricane environment. This dry-run also provides opportunities for testing equipment and procedures and for crew training. FIELD OPERATIONS Dry Runs and Cloudline Dry runs were conducted from the Naval Station Roosevelt Roads, Puerto Rico, on 3 to 6 August following a general briefing on 2 August. Partici- pating in the dry runs were aircraft from the Navy Weather Reconnaissance Squadron FOUR (VW-4), NAS Jacksonville, Florida; Navy Attack Squadron EIGHTY-FIVE (VA-85), NAS Oceana, Virginia; NOAA's Research Flight Facility, Miami, Florida; Air Force 53rd Weather Reconnaissance Squadron, Ramey AFB, Puerto Rico; Air Force 55th Weather Reconnaissance Squadron, McClellan AFB, Sacramento, California; and the Air Force 58th Weather Reconnaissance Squadron, Kirtland AFB, New Mexico. A contractor-owned and operated light aircraft was also furnished by the Naval Weapons Center, China Lake, California. In addition, scientists participated from the Naval Weather Service Command Headquarters, Washington, D.C.; Naval Weapons Center, China Lake, California; Fleet Weather Facility, Jacksonville, Florida; University of 305 Miami, Coral Gables, Florida; and NOAA's National Hurricane Research Lab- oratory, Coral Gables, Florida. Dry-run exercises for the STORMFURY eyemod experiment were conducted on 3 August, and for the rainsector/rainband experiment on 6 August. Comprehensive individual debriefs of each flight were made. A general critique with all units present was held after the eyemod experiments. During these training missions, flare tests were conducted for Air Force safety certification. Twenty silver iodide cannisters were dropped to determine if there were ejection or ignition problems. All flares ejected and burned properly. This further confirmed that the flares were still dependable after a year of storage including months in a moist, tropic environment. Following the dry-run exercises, selected STORMFURY forces deployed to Barbados to conduct the cloudline experiments and to test various seed- ing agents. The primary objective of these experiments was to train STORMFURY personnel in cloud seeding operations. A secondary objective was to collect data to learn how seeded clouds interact with one another. Such studies may contribute to the design of better hurricane modifica- tion experiments. Forces participating were: Navy, two WC-121N's, and one Cessna 401; Air Force, two WC-130's, and one WB-57F; NOAA, two DC-6's, one C-130, and one B-57. Experiments were conducted from Barbados on 8, 9, 10, 12, and 13 August. During the first 2 days, the experiments were curtailed because of a shortage of certified hi-octane fuel. By the 10th, the problem had been resolved and the forces divided into two groups of four aircraft each, with two additional high flying aircraft working with both groups. The Air Force WC-130 and the Navy P-3 aircraft were evaluated for the first time as auxilliary seeders for Project STORMFURY operations. Both appeared to be excellent for this purpose, except they cannot operate at the higher altitudes flown by the jet seeders in the eyemod experiment. On 12 and 13 August, two groups of aircraft tested new seeding solu- tions disseminated from airborne burners and new experimental flares de- veloped by the Navy. These tests were to determine if the flares effi- ciently nucleated supercooled water drops and if these flares and burners might be useful in STORMFURY operations in future years. Tests were conducted on 8, 10, and 13 August to determine the relia- bility of smaller pyrotechnic units (output of 110 gm Agl) with the same mixture as that used in STORMFURY hurricane experiments. From observations of cloud characteristics before, during, and after these seedings, it ap- peared that the smaller units performed satisfactorily. During the dry-runs and cloudline experiments, the normal problems with aircraft instrumentation occurred. The cloud physics instruments a- board the NOAA aircraft worked only about two-thirds of the time and had to be repaired at night. While the Air Force had minor problems, their problems did not affect the operations to any significant degree. The 306 Navy had radar problems but worked harder than ever before to make their equipment operate. The spirit and enthusiasm of everyone connected with the project were excellent during the exercises. The move to Barbados was an excellent choice both as a base from which to operate and from a standpoint of avail- ability of suitable clouds for the experiments. Aircraft traffic conges- tion and lack of suitable targets have been encountered when operating from other areas in the past. FIELD OPERATIONS Hurricane Ginger Hurricane Ginger formed in the Atlantic, southwest of Bermuda on 9 September 1972. Moving somewhat erratically, it progressed eastward and northeastward across the Atlantic to beyond 49°W longitude before making an anticyclonic loop back toward the west on 15 September. Ginger moved within range of STORMFURY aircraft by 24 September, and deployment of forces was ordered calling for first seeding (Tango-time) at 1200 GMT, 26 September. Project aircraft were already widely dispersed because of reconnais- sance missions on tropical cyclones Ginger and Janice. Assembling all forces at a common base before starting the operation was impracticable because of shortage of time. Thus, for the first time, Project STORMFURY conducted a major experiment with aircraft operating from several differ- ent airfields. These included NAS Jacksonville, Florida, the major cen- ter of operations; Naval Station Roosevelt Roads and Ramey AFB, Puerto Rico; Miami, Florida; Bermuda; and Naval Air Station, Oceana, Virginia. Extensive use of radio and telephones helped alleviate many of the brief- ing problems that arose because of the dispersion of the forces. The option of executing either an eyemod or a rainsector experiment was held open until the morning of each experiment. This was possible because the flight patterns in these experiments are quite similar ex- cept for those of the seeding aircraft. Hurricane Ginger was not suitable for an eyemod experiment on either 26 or 28 September (see appendix B). Ginger was a broad, diffuse storm with poorly defined and flat wind and pressure profiles. An eyemod ex- periment should be conducted on a mature hurricane with central pressures less than 980 mb, a well-defined eyewall , an abundance of cumuli form clouds surrounding the eyewall, and maximum winds close to the center of the storm with well-defined peaks. Although the eyemod experiment was impractical, conditions were be- lieved at the time to be marginally suitable for the rainsector experi- ments. They were conducted on 26 and 28 September. This was the first time that the rainsector experiment had been attempted, and even though the storm was not ideal, much was learned about this type of storm and whether storms such as Ginger could be beneficially modified. 307 Two types of aircraft were used in seeding Ginger. An Air Force WC- 130 from the 53rd Weather Reconnaissance Squadron did two seedings on 26 September, and the fourth seeding on the 28th. The Navy A-6 aircraft of Attack Squadron EIGHTY-FIVE did the first three seedings on 28 September. For a full discussion of the Hurricane Ginger experiments see appen- dix B. RESEARCH ACTIVITIES In addition to the successful cloudline experiments and 2-day seeding of Hurricane Ginger, progress in hurricane modification research contin- ued at a rapid pace. This progress was due to the efforts of research workers of the National Hurricane Research Laboratory, Coral Gables, Florida; the Naval Weapons Center, China Lake, California; and at cooper- ating universities. The research results are providing a firmer and broader base to support the future work of the project. Appendix B, "Report on seeding of Hurricane Ginger," by Drs. H. F. Hawkins, K. H. Bergman, and R. Cecil Gentry, discusses the Hurricane Ginger experiment. Included are a brief history of Hurricane Ginger, a synoptic analysis of Ginger for 24 to 30 September 1971, seeding decisions and choice of experiments on Hurricane Ginger, and a description of the rainsector experiments actually conducted on 26 and 28 September 1971. The analyses show that Hurricane Ginger had little potential for benefi- cial modification, but it did provide an opportunity for conducting the rainsector experiments and acquiring a great deal of knowledge concerning this type of large and diffuse tropical circulation. Appendix C, "Hurricane modeling at the National Hurricane Research Laboratory (1971)," by Dr. S. L. Rosenthal, summarizes the substantial improvements made in the NHRL asymmetric hurricane model and significant progress made in physically interpreting results from various hurricane models during 1971. "Hurricane seeding" simulations were performed with the NHRL circularly symmetric model and are summarized in this report. Appendix D, "The mutual interaction of Hurricane Ginger and the upper mixed layer of the ocean," by P. G. Black and W. D. Mallinger, examines the criteria for the existence of intense sea-surface cooling beneath the hurricane due to upwelling, mixing and/or evaporation. Data from various hurricanes, including Hurricane Ginger, show that intense cooling occurred only when the ratio of the storm speed to the baroclinic wave speed at the thermocline was less than three. Airborne expendable bathythermographs and airborne infrared thermometers were used to collect data during traverses of Hurricane Ginger. These data showed that sea-surface temperature de- creases up to 4°C occurred on one day when the storm was nearly stationary, and decreases of 2°C occurred on another day when the storm was moving faster. The effect of decreases of these magnitudes on storm development may be as large as those caused by seeding. The need to differentiate be- tween these two effects is emphasized. Fluxes of latent and sensible heat were calculated on a once-per-day basis and compared with storm intensity measurements. This calculation needs to be repeated more often for the comparisons to be truly meaningful. 308 Appendix E, "Modeling the seeding effect in Hurricane Ginger," by Dr. W. D. Scott and C. K. Dossett, attempts to calculate changes in cloud structure that should be expected from seeding in Hurricane Ginger. The cloud model used for the calculations is the one-dimensional Lagrangian model of Cotton (1970), called the PSU-71 model, and the input data were the temperature and humidities measured in the storm. Some of the cloud environments in Ginger appeared to be conducive to the development of clouds that were seeded. The limitations of the techniques and data used for this determination and recommendations for future investigations of seedability are also discussed. Appendix F, "The typhoon research program at the Environmental Pre- diction Research Facility as an aide to Project STORMFURY," by S. Brand, describes the research conducted by this unit which is considered to be of value to Project STORMFURY. Among those items discussed are general statistics of western North Pacific tropical cyclones, intensity fore- casting research, movement forecast research, and future research plans. This research will be of value to Project STORMFURY particularly if the project moves its operations to the Pacific Ocean in the future. Appendix G, "Diurnal variation in hurricanes," by R. C. Sheets, stud- ies significant diurnal variations in the structure or intensity of hurri- canes to determine if there is variation of the maximum wind speeds. This study indicates that no significant diurnal variation of hurricane inten- sity exists on the average, as reflected by maximum wind speeds or mini- mum sea-level pressures. Appendix H, "NWC contributions to STORMFURY dry-run and cloudline op- erations 2 to 12 August 1971," by Dr. S. D. Elliott, E. E. Hindman and Dr. W. G. Finnegan, outlines the Naval Weapons Center's major contribu- tions to Project STORMFURY. These contributions include the development and supply of pyrotechnics, silver iodide, ice nuclei generators, the dis- pensing systems acquired for their use on aircraft, and the development of new ice nuclei generating systems for potential use in the STORMFURY ex- periments. The characteristics of the various pyrotechnic units employed in the 1971 STORMFURY program are summarized along with a description of the pyrotechnic testing during the cloudline experiments. Appendix I, "Some statistical characteristics of the hurricane eye and minimum sea-level pressure," by R. C. Sheets, examines the hurricane radar eye and the minimum sea-level pressures and their correlations with other parameters, such as location of the storm, minimum sea-level pres- sure, and maximum wind speed. Frequency distributions of the changes of minimum sea-level pressure are also shown. This study shows that little correlation exists between the size of the hurricane eye and selected in- tensity and location parameters, except that the correlation was good with the latitude of the storm center. Appendix J, "Typhoon seeding eligibility in the western North Pacific," by W. D. Mai linger, examines earlier studies of typhoon eligibility for seeding, and investigates new combinations of stations from which to con- duct seedings (Guam/Okinawa and Guam/Phillippines). These studies use new, more stringent eligibility rules than past studies in order to provide 309 more safeguards for some of the countries west of the proposed experimental areas. The study showed that by judicial use of the forces, even with ad- ditional restrictions as to eligible operating areas and seeding eligibil- ity rules, opportunities for experimentation in the western Pacific would make such a move of the Project most desirable. Appendix K, "Comparisons of rainfall and radar measurements in Tropi- cal Storm Felice and Hurricane Debbie," by Dr. W. D. Scott and C. K. Dossett, examines the cloud activity of tropical storms through the radar signatures for radar PPI displays. This paper presents Z-R relationships derived from in situ measurements of raindrop number and sizes and pre- sents data that show the structure of the rainwater content in the storms. It also emphasizes that the radar echoes may be more closely associated with downdrafts rather than updrafts. This is important since the seeding is presumably made more effective when done in the updrafts. OPERATIONAL AND RESEARCH DATA COLLECTION Data collection procedures are considered adequate for the Project. Continuous effort, however, is needed to maintain data quality control due to ole and often inadequate equipment, shortages of repair parts, and rotation of new personnel into the data collection jobs. Polaroid (CU-5) cameras were again furnished to the Air Force for use on the WC-130 aircraft radars. One camera was also furnished to the Re- search Flight Facility for use on their C-130. Automatic time-lapse cameras are desperately needed for these radars, but the Polaroids pro- vide extremely useful data. The Research Flight Facility conducted several research missions in tropical circulations; however, again in 1971, the hurricane season was one with few good data collection opportunities. A one-plane mission was flown into Tropical Storm Fern in the central Gulf of Mexico on 8 September. On 9 September, another one-plane mission was flown in Hurricane Edith while the storm was south of Jamaica. A three-plane monitoring mission was flown on 11 September, followed by a two-plane mission on the 12th, and a single-plane mission on 13 Sep- tember while Hurricane Ginger was moving from southeast to east of Bermuda. Processing of STORMFURY films was again accomplished by a commercial firm in Miami. No major difficulties were encountered. OUTLOOK FOR 1972 Project STORMFURY will become a joint Department of Commerce/Depart- ment of Defense project in 1972, instead of between the Departments of Commerce and Navy. Project STORMFURY will continue to operate in 1972. Dry-run exer- cises and cloud! ine experiments will again be scheduled. 310 Primary emphasis will continue to be placed on repeating the Debbie type of experiment and conducting monitoring missions in unseeded storms for comparison purposes. The WC-121N Navy aircraft have been replaced by the WP-3A in the Navy Weather Reconnaissance Squadron FOUR (VW-4). REFERENCES AND SPECIAL REPORTS Anthes, R. A. (1971): Iterative solutions to the steady-state axisymmet- ric boundary layer equations under an intense pressure gradient. Monthly Weather Review, 99, No. 4, pp. 261-268. Anthes, R. A. (1971): A numerical model of the slowly varying tropical cyclone in isentropic coordinates. Monthly Weather Review, 99, No. 8, pp. 617-635. ~ Anthes, R. A. (1971): Numerical experiments with a slowly varying model of the tropical cyclone. Monthly Weather Review, 99, No. 8, pp. 636- 643. Anthes, R. A. (1971): The development of asymmetries in a three-dimen- sional numerical model of the tropical cyclone, ERLTM-NHRL Technical Memorandum No. 94, NOAA, U.S. Department of Commerce, NHRL, Miami, Florida, 55 pp. Anthes, R. A. (1971): The response of a three-level axi symmetric hurri- cane model to artificial redistribution of convective heat release, ERLTM-NHRL Technical Memorandum No. 92, NOAA, U.S. Department of Com- merce, NHRL, Miami, Florida, 14 pp. Anthes, R. A. (1972): Non-developing experiments with a three-level axi- symmetric hurricane model, ERLTM-NHRL Technical Memorandum No. 97, NOAA, U.S. Department of Commerce, NHRL, Miami, Florida, 18 pp. Anthes, R. A., S. L. Rosenthal, and J. W. Trout (1971): Preliminary re- sults from an asymmetric model of the tropical cyclone. Monthly Weather Review, 99 3 No. 10, pp. 744-758. Anthes, R. A., J. W. Trout, and S. S. Ostland (1971): Three dimensional particle trajectories in a model hurricane. Weatherwise, 24, No. 4, pp. 174-178. Anthes, R. A., J. W. Trout, and S. L. Rosenthal (1971): Comparisons of tropical cyclone simulations with and without the assumption of circ- ular symmetry. Monthly Weather Review, 99,, No. 10, pp. 759-766. Black, P. 6. (1971): Use of satellite cloud motion data to infer the out- flow structure of hurricanes. Presented at Goddard Space Flight Center, Beltsville, Md., 25 May, 1971. Black, P. G. (1971): The asymmetric structure of the hurricane outflow layer. Presented at the Seventh Technical Conference on Hurricanes and Tropical Meteorology, Barbados, W.I., 7 December 1971. 311 Black, P. G., and R. A. Anthes (1971): On the asymmetric structure of The tropical cyclone outflow layer. Journal of Atmospheric Sciences, 28, pp. 1348-1366. Black, P. G. (1972): Interaction of Hurricane Debbie with the mid- Atlantic trough. Seminar presented at the National Center for At- mospheric Research, Boulder, Colorado, 18 February 1972. Black, P. G. (1972): Use of satellite and airborne remote sensing tech- niques to infer the mutual interaction of hurricanes and the upper mixed layer of the ocean. Seminar presented at ERL Headquarters, Boulder, Colorado, 18 February 1972. Black, P. G. (1972): Use of remote sensing techniques to infer the mutual interaction of Hurricanes Ginger and Debbie and the upper mixed layer of the ocean. Seminar presented at the National Hurricane Research Laboratory, Miami, Florida, 8 March 1972. Black, P. G. (1972): Interaction of Hurricane Debbie (1969) with the mid-Atlantic trough as revealed by Nimbus III radiation data and ATS-III cloud motion data. Seminar presented at the National Hurri- cane Research Laboratory, Miami, Florida, 28 March 1972. Black, P. G., H. V. Senn and C. L. Courtright (1972): Airborne radar observations of eye configuration changes, bright band distribution and precipitation tilt during the 1969 multiple seeding experiments in Hurricane Debbie. Monthly Weather Review, 100» PP« 208-217. Carlson, T. N. (1971): Outbreaks of Saharan air and dust over tropical Atlantic. Seminar presented at the University of Miami Meteorologi- cal Department, Coral Gables, Florida, 20 November 1971. Carlson, T. N., and J. M. Prospero (1971): Outbreaks of Saharan air and dust over tropical Atlantic. Seventh Technical Conference on Hurri- canes and Tropical Meteorology, Barbados, W.I., 7 December 1971. Gentry, R. C. (1970): Hurricane modification -experiments and prospects. Presented at the Hurricane Foresight Conference, New Orleans, La., 30 April 1970. Gentry, R. C. (1970): Progress on hurricane modification research October 1969 to October 1970. Presented at the Twelfth Interagency Weather Modification Conference, Virginia Beach, Va., 27-30 October 1970. Gentry, R. C. (1970): Project ST0RMFURY. Presented at the Atlantic Oceanographic and Meteorological Laboratories, Miami, Florida, 5 November 1970. Gentry, R. C. (1970): Modifying the greatest storm on earth. Presented at the Council for the Advancement of Science Writing, Raleigh, North Carolina, Eighth Annual Briefing, New Horizons in Science, 17 Novem- ber 1970. Gentry, R. C. (1970): Hurricane modification research. An abbreviated version presented at N.W.S. Hurricane Conference, 1 December 1970. 312 Gentry, R. C. (1971): Hurricane modification research. Presented at the Interagency Hurricane Conference, 13 January 1971. Gentry, R. C. (1971): A review of Project STORMFURY. Presented at the President's Science Advisory Committee, Washington, D.C., 18 January 1971. Gentry, R. C. (1971): Hurricane modification experiments and hypothesis. Presented at the U.S. —Japan Conference on Cumulonimbus Modification of Tropical Nature, Miami, Florida, 17 February 1971. Gentry, R. C. (1971): Project STORMFURY, 1971. Presented at Interdepart- mental Conference on Weather Modification, Skyland, Va., 14 October 1971. Gentry, R. C. (1971): Prospects for hurricane reduction and precipitation management in the southeast. Presented at the Second National Biolog- ical Congress of the American Institute of Biological Sciences, Miami Beach, Florida, 23-26 October 1971. Gentry, R. C. (1971): Project STORMFURY. Presented at the First United- Methodist Men, Miami, Florida, 26 October 1971. Gentry, R. C. (1971): A lady called Camille (movie). Presented at the Allapattah Lions Club, 2 November 1971. Gentry, R. C. (1971): Project STORMFURY, 1971. Presented to Advisory Panel of Project STORMFURY, Washington, D.C., 3 December 1971. Gentry, R. C. (1971): Project STORMFURY - 1971. Presented at the Seventh Technical Conference on Hurricanes and Tropical Meteorology, Barbados, W. I., 6-9 December 1971. Gentry, R. C. (1971): The modification experiment on Hurricane Ginger, September 1971. Presented at the Seventh Technical Conference on Hurricanes and Tropical Meteorology, Barbados, W. I., 6-9 December 1971. Gentry, R. C, T. T. Fujita and R. H. Simpson (1971): Hurricane Celia 1970 -a case study of unusual structure revealed by damage survey, radar, satellite and other meteorological analyses. Presented at the Seventh Technical Conference on Hurricanes and Tropical Meteo- rology, Barbados, W. I., 6-9 December 1971. Gentry, R. C. (1972): Hurricane modification research. Climatological Re- search, Hermann Flohn 60th Birthday Volume, Bonner Meteorologische Abhandlungen 17 (In press). Gentry, R. C. (1972): Project STORMFURY - 1971. Interdepartmental Hurri- cane Conference, Miami, Florida, 19 January 1972. Gentry, R. C. (1972): Hurricanes. Seminar presented at Colorado Univer- sity Department of Aerospace Engineering, 27 January 1972. Gentry, R. C. (1972): Project STORMFURY. Presented at Typhoon Conference of U.S. Forces, Sanno Hotel, Tokyo, Japan, 8 February 1972. 313 Gentry, R. C. (1972): Hurricane Modification. Seminar presented at Royal Observatory at Hong Kong, Japan, 18 February 1972. Gentry, R. C. (1972): STORMFURY 1971. Presented at ICAS (Interdepart- mental Committee on Atmospheric Science), Washington, D.C., 10 March 1972. Gentry, R. C. (1972): Work of National Hurricane Research Laboratory with special emphasis on problems related to the sea. Presented at NAS-NAE Advisory Committee to NOAA on Marine Science, Miami, Florida, 22 March 1972. " Gentry, R. C. (1972): Hurricanes. Presented at Goddard Space Flight Center Scientific Colloquia, Beltsville, Md., 24 March 1972. Gentry, R. C. (1972): Hurricanes. Presented at Bellaire Research Center, Shell Development Company, Houston, Texas, 25 May 1972. Gentry, R. C. (1972): Project STORMFURY. Seminar presented for Meteo- rologists in Mexico, Mexico City, D.V., 26 May 1972. Hawkins, H. F. (1971): Structure of hurricanes. Presented to representa- tives of Republic of China, Miami, Florida, 2 December 1971. Hawkins, H. F., T. W. Huck and P. G. Black (1971): Tropical Storm Felice. Presented at the Seventh Technical Conference on Hurricanes and Trop- ical Meteorology, Barbados, W.I., 7 December 1971. Hawkins, H. F. (1972): Experiments in hurricane modification. Science Honor Society, 14 February 1972. Hawkins, H. F. (1972): Briefing on Project STORMFURY. Presented to rep- resentatives of Government and/or Weather Service of Nassau on 19 May 1972; Jamaica, 22 May 1972; Tegucigalpa, 25 May 1972; Tampico, 29 May 1972; Meyida, 30 May 1972; and Belize City, 31 May 1972. Koss, W. J. (1971): Numerical integration experiments with variable res- olution two-dimensional Cartesian grids using the box method. Monthly Weather Review, 99 j No. 10, pp. 725-738. Rosenthal, S. L. (1971): A circularly symmetric, primitive equation model of tropical cyclones and its response to artificial enhancement of the convective heating functions. Monthly Weather Review, 99* No. 5, pp. 414-426. Rosenthal, S. L. (1971): The response of a tropical cyclone model to var- iations in boundary layer parameters, initial conditions, lateral boundary conditions and domain size. Monthly Weather Review, 99 3 No. 10, pp. 767-777. Rosenthal, S. L., and M. S. Moss (1971): Numerical experiments of rele- vance to Project STORMFURY. ERLTM-NHRL Technical Memorandum No. 95, NOAA, U.S. Department of Commerce, NHRL, Miami, Florida, 52 pp. 314 Rosenthal, S. L., and M. S. Moss (1971): The response of a tropical cy- clone model to radical changes in data fields during the mature stage. ERLTM-MRL Technical Memorandum No. 96, NOAA, U.S. Department of Com- merce, NHRL, Miami, Florida, 18 pp. Scott, W. D. (1972): An assessment of the present instrumentation for the measurement of cloud elements and out needs. Preprint of the MS Second Symposium on Meteorological Observations and Instrumentation, March 1972, 205-216. Scott, W. D. (1972): An assessment of the present instrumentation for the measurement of cloud elements and our needs. Presented at the Second Symposium on Meteorological Observations and Instrumentation, San Diego, California, 30 March 1972. Scott, W. D. (1972): Modeling the seeding effect in Hurricane Ginger. Seminar presented at the Naval Post Graduate School, Monterey, Cali- fornia, 13 April 1972. Trout, J. W., and R. A. Anthes (1971): Horizontal asymmetries in a numer- ical model of a hurricane. EELTM-NHRL Technical Memorandum No. 93 3 NOAA, U.S. Department of Commerce, NHRL, Miami, Florida, 18 pp. 315 APPENDIX A PROJECT STORMFURY ADVISORY PANEL RECOMMENDATIONS 3-4 December 1971 Washington, D.C. Recommendation ONE: Seeding Capability Project STORMFURY must maintain the capability of multiple seeding operations with pyrotechnics at an altitude where temperatures are -25°C or colder. We recommend sufficient pyrotechnics be acquired to execute eight to ten "eyemod" experiments if Project STORMFURY undertakes operations in the Pacific. Recommendation TWO: Aircraft Location and Tracking a. Seeding aircraft must be located on a real-time basis with an accuracy of 1 to 2 n miles relative to the designated area being seeded. b. Post analyses of the seeding experiments should be able to docu- ment the location of all aircraft with a relative accuracy of 1 n mile. Recommendation THREE: Communications Air-to-air communications should be capable of locating, identifying, and controlling all aircraft involved in the experiment to ensure proper execution of seeding and monitoring operations and to ensure aircraft safety. Recommendation FOUR: Monitoring Capability a. The Project should have sufficient aircraft to monitor the hur- ricane from 3 hours prior to the first seeding to 6 hours after the last seeding, a total of approximately 17 hours. b. Instrumentation on the aircraft should be improved, especially with regard to radar capability and cloud physics measurements. Recommendation FIVE: Radar Capability a. The Panel noted with concern the deterioration of the RHI radar capability. Every effort should be made to redevelop this capa- bility. 316 b. Radar capability in terms of PPI presentation should also be im- proved including, if possible, at least qualitative measurements and comparisons of radar reflectivity. Recommendation SIX: Cloud Physics Capability a. Cloud physics instrumentation should be improved to measure liquid water content, for example, with the use of a Johnson- Williams or equivalent system with continuous recording. Icing- rate measurements are also recommended. b. We recommend the installation of instrumentation to measure drop size and number density of supercooled water droplets and ice crystals. c. The above measurements should be acquired before and after seed- ing at altitudes and along azimuths sufficient to determine whether seeding has changed the distribution and type of the cloud particles. Recommendation SEVEN: Sea Surface Temperatures We recommend that instrumentation be further developed to enable measurement of the temperature of the sea surface and the tempera- ture distribution through the thermocline layer. Recommendation EIGHT: The Panel reiterates its previous recommendation that a season's op- eration of Project STORMFURY in the Pacific be carried out, but we believe that unless recommendations 1 through 6, above, can be imple- mented, the results of such an operation would be seriously jeopar- dized. 317 APPENDIX B REPORT ON SEEDING OF HURRICANE GINGER Harry F. Hawkins, Kenneth H. Bergman, and R. Cecil Gentry National Hurricane Research Laboratory A BRIEF HISTORY OF HURRICANE GINGER The future Hurricane Ginger was first designated a tropical depres- sion on 5 September 1971, after a cold upper level low extended its influ- ence downward and warmed while moving slowly southeastward some 600 n miles south-southwest of Bermuda. It curved rather sharply northeastward on 7 September and turned eastward about 300 n miles south of Bermuda on 10 September (fig. 1). On this same day, it reached tropical storm strength and continued moving eastward and northeastward. Research flights into Ginger were conducted on 11, 12, and 13 September, and during this period the storm became a full hurricane. The yery first flight into tropical storm Ginger revealed a character- istic that seemed to be maintained throughout most of its life. Profile A, figure 2, shows the wind speed relative to the moving storm (as a function of radial distance) and also reveals the broad plateau of winds that sur- rounded the eye at 1630 GMT. The maximum winds are not distinguished by sharp peaks with strong horizontal shears on either side. However, it may be pointed out that this latter type of structure should not be expected at this premature stage of the storm's development. These early passes similarily revealed a minimum D-value of about -250 ft, followed by fur- ther deepening to slightly more than -300 ft, before the end of this series of passes by the first flight in the storm. All of these penetra- tions, which were at 5,000 ft altitude (as were subsequential flights), indicated temperatures between 19 and 20°C in the eye with environmental temperatures between 16 and 17°C. Succeeding flights were arranged to provide an almost continuous monitoring of the storm as it developed. Profile B, figure 2, shows the relative winds some 6 hours later at 2200 GMT, 11 September. The peaks were only a little better defined as the wind speeds reached minimal hur- ricane strength, at least momentarily. The preferred radius of maximum winds appeared to be between 20 and 25 n miles from the center. Succes- sive penetrations revealed that temperatures in the eye reached 20 to 21°C on every pass, and minimum D-values were reaching -500 to -525 ft. The third flight (Profile A, fig. 3) revealed that by 0203 GMT, 12 September, 318 Figure 1. Track of Hurricane Ginger with reported central pressures in- dicated along path. The approximate seeding areas are also indicated. winds were indeed of hurricane strength, although only over a restricted area. The radius of maximum winds was again not too well defined. It appeared to be established around 25 n miles from the center. Tempera- tures in the eye were now reaching 21 to 22°C, and the minimum D-values recorded were in the range -525 to -550 ft. The fourth flight (Profile B, fig. 3) revealed that, by 1330 GMT of the 12th, wind speeds on both sides of the storm were of hurricane strength, Again the peaks were not too well defined, and there was a rather broad plateau of wind speeds greater than 50 knots on either side of the storm. At no time on this flight was a temperature as warm as 21°C recorded in the eye. Most of the minimum D-values fell in the range between -550 and -575 ft. The succeeding plane found winds (Profile A, fig. 4) of 70 knots on either side of the storm at 2200 GMT. These winds appeared to peak about 319 HURRICANE GINGER RELATIVE WINDS SEPTEMBER 11, 1971 REF. LVL. 4,7 80 FT. 30 20 RADIAL 10 DISTANCE 0 10 20 (NAUTICAL MILES) 30 40 50 Figure 2. Horizontal wind profiles through center of Hurricane Ginger. Wind speeds are relative to the moving storm center. 20 n miles from the center of the eye. Temperatures as warm as 23°C at flight level were recorded at this time, and D-values as low. as -850 ft were noted in the eye. The last flight in this sequence at 0200 GMT, 13 September, showed relative winds (Profile B. fig. 4) of about 80 knots on both sides of the eye. The storm was about as well structured here as at any time for which research flight data are available. Maximum wind speeds were about 15 n miles from the center of the eye and the peaks were fairly well defined, although between 25 and 70 n miles radial distance there was a fairly broad annul us of winds averaging about 50 knots. The temperatures in the storm at this time were not notably different from those of the preceeding pene- trations, namely, the maximum temperature at flight level appeared to be about 22 C, although on one of the passes through the eye, a transitory peak of 24 C was recorded. On this same pass, D-values for the first time fell below -1,000 ft, reaching a minimum of -1,020 ft. Thus we have a picture of a tropical storm slowly developing into a hurricane. This was certainly not a case of explosive deepening; never- theless, the organization was continually inproving, maximum winds were increasing in strength, and the shears on either side of the peak winds seemed to be growing stronger with time. Unfortunately, the storm was also becoming more distant from the staging base of Bermuda, so the mon- itoring flights were terminated for this reason plus the possibility of interesting developments occurring elsewhere. 320 70 60- HURRICANE GINGER RELATIVE WINDS SEPTEMBER 12.1971 REF. LVL. 4,780 FT. »- o z a UJ £ Q Z 50- 40- 30- 20- 10- _L J_ ,U» 13452 02482- _L J_ 60 50 40 30 20 10 0 10 20 30 40 RADIAL DISTANCE (NAUTICAL MILES) 50 Figure 3. Wind profiles for Ginger. Note that the eye remains relatively broad and diffuse. Ginger reached its all-time maximum strength somewhat later on 13 September with a central pressure of 950 mb and sustained winds of 90 knots. As shown by figure 1, she was moving almost eastward at this time, but on 14 September further eastward movement was blocked by a ridge of high pressure to the north and east. This forced the storm to make a large, slow anticyclonic loop that was accompanied by some filling and by weakening of the maximum winds. On 15 September, it began a slow drift towards the west, upon which was superimposed a very small cyclonic loop, but the net result was that on 22 September, the storm was about 50 n miles northeast of its position 10 days previously on 12 September. The storm proceeded towards the west-southwest with central pressures slightly above 990 mb. On 24 September, it again came to almost a complete halt and began a slow southwestward movement accompanied by further deepening to about 976 mb. During 27 September, Ginger turned toward the northwest and set a fairly steady course for the North Carolina coast, where it moved inland near Moorehead City in the evening of 30 September. Seeding experiments in Hurricane Ginger were carried out on 26 and 28 September; the locations of the seeding areas relative to the storm center are shown in figure 1. 321 80 70 HURRICANE GINGER RELATIVE WINDS SEPTEMBER12-13,197i REF. LVL. 4,780 FT. 60- H O ^401 a Hi Hi a. W30 '20 10- 2 205Z 1402 * K. (A) SEPTEMBER 12 (B) SEPTEMBER 13 0215Z 60 50 40 30 20 10 0 10 20 30 40 50 RADIAL DISTANCE (NAUTICAL MILES) Figure 4. Wind profiles for Ginger. The storm is more intense s radius of maximum winds is smaller > and eye is better defined than on -previous days, SYNOPTIC ANALYSIS OF HURRICANE GINGER FOR 24 THROUGH 30 SEPTEMBER 1971 Synoptic map analyses covering a portion of the North Atlantic Ocean and adjacent areas and centering on Hurricane Ginger have been prepared for 0000 GMT and 1200 GMT, 24 through 30 September 1971. These consist of conventional isobaric surface and contoured 500 mb analyses along with 200 mb streamline analyses. Operationally produced analyses prepared by the staff of the National Hurricane Center were used as reference. The analyses of upper level wind fields were aided by cloud pictures trans- mitted from the ATS-3 synchronous satellite and film-loops made from these pictures; the pictures and film-loops were provided by the National Hurri- cane Center. In the following discussion of the analyses, attention is directed primarily to factors of a synoptic nature that may have resulted in changes in the intensity of the storm or in its direction and speed of movement. Since Ginger was seeded on 26 September and again on 28 September, it is 322 desirable to identify those changes in Ginger brought about by the synop- tic situation as distinguished from any changes that may have resulted from the effects of the seeding missions. Although such identification of synoptic factors influencing Ginger has been attempted below, it is usually difficult or impossible to draw definite conclusions about these factors. In part, this is because sufficient observational data near the storm are often lacking, especially at the upper tropospheric levels. As indicated in the previous section, Ginger was a large hurricane in areal extent and rather diffuse in structure. The overall radius of the storm circulation averaged about 400 n miles, and the radius of winds exceeding 30 knots, as determined from ship observations, averaged about 250 n miles during the period of study; these dimensions place Ginger among the largest of Atlantic hurricanes in terms of physical size. The distribution of wind speeds about the center was fairly symmetrical in this slow-moving storm, but there was a slightly greater areal extent of gales in the northern semicircle as compared with the southern one. The ATS-3 cloud pictures and film-loops show a relative lack of cirriform clouds in the outflow layer above Ginger as compared with other well- developed hurricanes. The eye region, usually visible in the satellite pictures, was larger than normal and at times, exceeded 60 n miles in diameter, but did not appear to be sharply delineated by a well -developed wall cloud. Since Ginger was of unusually large horizontal extent and was located outside of the tropics proper during most of its lifetime, there has been some question about whether it was truly a tropical cyclone; there is the possibility that the storm may have had extra tropical or "hybrid" charac- teristics. However, the analyses show that Ginger possessed the warm-core structure at upper levels, which is considered typical of tropical cy- clones. During most of the period under consideration, the center of Ginger was located approximately 300 to 400 n miles south of an east-west baroclinic zone that extended across the western North Atlantic at this time, but the available data indicate that the temperature distribution in the inner part of Ginger remained fairly symmetrical about the center, and that baroclinic effects, while not altogether absent, were compara- tively weak and largely restricted to the northern periphery of the storm circulation. From 1200 GMT, 24 September until 0000 GMT, 27 September, Hurricane Ginger moved slowly southwest; thereafter it moved northwest (fig. 1). The average speed of movement was about 3.8 knots through 1200 GMT, 28 September, after which the storm accelerated to 10.4 knots on 29 Septem- ber. Some deceleration to 7.5 knots occurred on 30 September as Ginger approached the North Carolina coast. Throughout the period under consideration, the surface synoptic situ- ation in the northwest Atlantic area was unusual in that the Bermuda- Azores subtropical anticyclone was weaker and located farther eastward than climatology indicates. Additionally, an anticyclone of polar origin persisted over the North Atlantic north of about 35 N during most of the period, affecting the course of the storm and possibly also changes in its intensity. 323 At the beginning of the per zone with waves was positioned a tropical air associated with the cal anticyclone (located near 35 north (fig. 5). A second front GMT, 24 September. Behind this by an anticyclone of polar origi GMT, 25 September (fig. 6), the had moved off the Middle Atlanti the existing frontal zone in the iod, a relatively weak east-west frontal t approximately 40°N latitude, separating hurricane circulation and the subtropi- from modified polar air farther Texas to New York at 0000 the surface was dominated located in Iowa. By 1200 portion of this second front N 50°W) extended from second front, n with center northeastern c and New England coasts and merged with Atlantic, strengthening this zone and Figure 5. Ginger was drifting slowly west, while Tropical Storm Janice was moving northwest just east of the Lesser Antilles. Cold front and polar anticyclone over Midwest were rapidly approaching the east coast. 324 Figure 6. Ginger has intensified and started south, as cold front and polar anticyclone move off east coast. Remains of weakening Janice con- tinue to move towards the northwest. moving it somewhat farther south to about 35°N across that portion of the Atlantic just north and northeast of Ginger. At the same time, the polar anticyclone originally located to the north of the old frontal zone was replaced by tne new polar anticyclone that moved eastward off the New England coast on 26 September (fig. 7). This anticyclone subsequently tended to stagnate just south of the Maritime Provinces (in the vicinity of 45 N 65°W) until 0000 GMT, 29 September, while maintaining a central pressure of 1025 to 1030 mb. The displacement of Ginger towards the southwest, which commenced late on 24 September and continued through 26 September, was probably associated with the movement of this anticyclone to a position north of the hurricane circulation. 325 1020 ^JlfynT&CE CHART y S&T 26, j.971 NL200 Figure 7 . Intrusion of modified polar air into western side of Ginger's circulation is complete. Remains of Janice begin to be absorbed in the dominant Ginger circulation. After 0000 GMT, 27 September, the approach of a weather disturbance from the central United States weakened the anticyclone's western limb. And after 0000 GMT, 29 September, the anticyclone commenced moving east- southeastward from its position near the Maritime Provinces. The slow drift of Ginger towards the northwest, which commenced on 27 September continued on 28 September (fig. 8), and the more rapid movement of the storm towards the northwest on 29 and 30 September appear to be corre- lated with these events. 326 Between 0000 GMT, 25 September and 0000 GMT, 27 September, a portion of the polar front then moving slowly southeastward off the Atlantic coast of the United States was swept more rapidly southward around the western side of the Ginger circulation, eventually merging with a pre-existing shear line extending southwestward from the Ginger vortex towards its as- sociated neutral point. The evidence for this movement from surface syn- optic reports is rather scant, but film loops made from ATS-3 cloud pic- tures clearly show the movement of a band of lower clouds associated with the southward surge of the front. The flow at 500 mb at this time indi- cated that this cool air penetration on the western flank of Ginger was Figure 8. Ginger accelerating towards the North Carolina coast as western limb of -polar anticyclone weakens. Mac or seeding effort on 28 September commenced shortly after this time chart. 327 probably quite shallow, and the surface observations indicate that the air mass was rapidly modified as it crossed the warm waters of the Gulf Stream, whose axis lay approximately 300 n miles to the northwest of the center of Ginger at this time. Surface temperature and humidity gradients across this portion of the frontal zone rapidly became negligible south of about 35°N. Throughout the remainder of the period considered, a flow of modified polar air continued to feed into the hurricane circulation from the anti- cyclone positioned to the north or northeast, but Ginger appears to have been "protected" from any massive penetration of unmodified polar air into its circulation by the presence of the Gulf Stream axis and its continua- tion, the North Atlantic Drift, between it and the polar anticyclone, and by the circumstance that the upper tropospheric circulation, with a west- erly jet at or polewards of 40 N latitude, favored only relatively shallow intrusions of air from the northern anticyclone to extend as far south as the latitude of Ginger. Therefore, the effect of the modified polar air penetrations into Ginger's circulation on changes in intensity of the storm is not likely to have been very pronounced. But it is interesting to note that inten- sification of Ginger, as indicated by reported central pressure values, commenced late on 24 September, when the fresh surge of polar air from the continent arrived at the extreme periphery of Ginger's circulation, and continued through 25 September. Central pressure decreased from 993 mb at 1200 GMT, 24 September to 976 mb at 0000 GMT, 26 September, followed by a rise to 982 mb at 1200 GMT, 26 September. It is possible that the advection of modified polar air into the storm circulation was responsible for the temporary intensification by adding a baroclinic energy source to the existing thermodynamic one. However, the fact that the core of Ginger moved into an area of appreciably warmer sea-surface temperatures during this time may have been a more important factor in the intensification noted above. The role of sea-surface temperatures on changes in Ginger's intensity is discussed in detail in appendix D. The central pressure of Ginger decreased from 980 mb at 0000 GMT, 27 September, to 975 mb at 1200 GMT, 27 September; central pressure was still 969 mb but subsequently increased to 981 mb by 1200 GMT, 28 Septem- ber (fig. 8). The ATS-3 pictures indicate that Ginger became a better organized storm, with a more sharply defined eye and greater cirrus out- flow (fig. 8) during the daylight hours of 27 September as compared with either 26 September or 28 September. It is possible that an influx of tropical air with high moisture content, associated with the remains of Janice, into Ginger's circulation was responsible, at least in part, for the temporary intensification noted on 27 September. Another synoptic event that may have resulted in a change in Ginger's intensity was the absorption of the remnants of dissipating Tropical Storm Janice in the dominant Ginger circulation on 27 September. By 25 September, Janice had deteriorated into an area of disturbed weather, characterized by intense convective activity but little or no remaining circulation, while moving on a northwestward course that passed just northeast of the 328 Lesser Antilles. This area of disturbed weather continued to move towards the northwest, approaching the Ginger circulation from the southeast on 26 September. By 1900 GMT, 26 September, ATS-3 cloud pictures indicated that the area of heaviest cumulonimbus activity associated with the remnants of Janice was located in the vicinity of 24°N 64°W, or about 420 n miles south- east of the center of Ginger (fig. 9, top). At 1200 GMT, 27 September, the heavy convective activity was concentrated in a triangular area of 2° or 3° latitude per side centered on 23°N 67°W, or about 360 n miles south- southeast of Ginger (fig. 9, middle). This represents a westward shift of 180 n miles, possibly related to easterly flow that prevailed at the 200 mb level at this time. Subsequently, this area of convective activity moved slightly to the northwest while becoming elongated along a north- northeast to south-southwest axis and eventually formed one of the eastern spirals of Hurricane Ginger. The intensity of the convective activity appeared to diminish as the area became organized into a spiral band of the hur'ricane. The ATS-3 pictures indicate that integration into the spiral band system of Ginger was completed by 1700 GMT, 27 September (fig. 9, bottom). Ginger's central pressure remained at about 981 mb, with some minor fluctuations, until 0700 GMT, 29 September, after which an increase to 987 mb took place by 0000 GMT, 30 September. At the same time, Ginger moved towards the North Carolina coast at an accelerated pace. This move- ment of the storm may have permitted further weakening as Ginger's unusu- ally large circulation gradually extended over the Atlantic coastal plain, allowing cooler, drier air from the interior to be advected into the storm. In addition, the northwestward movement of Ginger allowed the storm's circulation to extend over the cooler waters north of Cape Hat- teras, thus progressively reducing the protective and modifying effect of the Gulf Stream on importations of polar air into the storm. An examination of the 500, mb charts for this period indicates that the westerly wind maximum at this level was located at or poleward of 40°N latitude in the western North Atlantic area and thus too far north to have an appreciable effect on Ginger. A basic pattern of long-wave trough over the west-central United States, flat ridge over the eastern United States, and zonal flow across the western North Atlantic persisted throughout the period. A cutoff low was near 30°N 35°W at the beginning of the period; this feature gradually drifted southward and opened into the tropical easterlies at 500 mb later in the period. A long wave trough in the westerly flow was developing near 40°W longitude at the end of the period. The 500 mb analyses for 1200 GMT, 26 September and 1200 GMT, 28 September are shown as figures 10 and 11. Ginger was bounded by two anticyclones at 500 mb for most of the period. One anticyclone was over northern Florida and the adjacent north- eastern corner of the Gulf of Mexico during the first part of the period (fig. 10). This anticyclone did not extend down to the surface nor up to 200 mb. After 0000 GMT, 27 September, this anticyclonic circulation moved southward to a position over western Cuba while weakening in intensity. This movement lowered 500 mb heights to the west and northwest of Ginger 1900 GMT, 26 SEPT. 1971 "N 1248 GMT, 27 SEPT. 1971 1650 GMT, 27 SEPT. 1971 &• --*'" Figure 9 , Cloud photographs of Ginger showing absorption of the remains of Janice in the Ginger circulation. 31 330 Figure 10. This chart corresponds in time to figures 7 and 12. Note that Ginger is well removed from the influence of the westerlies at this level, and may have played a role in the movement of Ginger towards the north- west, which began at this time. However, 500 mb heights along the mid- dle Atlantic coast rose again somewhat after 0000 GMT", 28 September as a new anticyclonic circulation began to develop over the central Appala- chians (fig. 11). Throughout the period, the other anticyclonic persisted in the area east of Ginger. Initially this anticyclone was two-lobed and Figure 11. Synoptic in time to figures 8 and 13. Outflow is primarily to the north and east of the storm. oriented along a north-south axis near 55°W 26 September, the anti cyclonic circulation orientation near latitude 30°N. The 500 mb proximately in location with a surface anti anti cyclonic circulation at 200 mb through deep anticyclonic circulation of subtropica September, the coupling between the surface ened, although the 500 to 200 mb portion of longitude. After 0000 GMT, gradually assumed an east-west anticyclone corresponded ap- cyclone as well as with an 27 September, indicating a 1 type. After 0000 GMT, 28 and 500 mb anticyclones weak- the anticyclonic circulation 332 remained fairly intact. A westward movement, or building, of this upper level anticyclone occurred as Ginger moved towards the west. As the 500 mb anticyclone over northern Florida moved southward and weakened, a trough and associated low center developed over South Carolina and Georgia at 0000 GMT, 28 September. This feature subsequently drifted south to latitude 30°N, over northern Florida (fig. 11), and persisted there while lengthening into an east-northeast west-southwest trough across the entire northern Gulf of Mexico at 0000 GMT, 30 September. The trough does not appear to have had any significant influence on the move- ment of Ginger, other than lowering 500 mb heights west of the storm. South of Ginger, several cyclonic and an ti cyclonic eddies existed in a very weak easterly flow at 500 mb, but none of these appear to have had any significant influence on Ginger. The 200 mb analyses for 0000 GMT, 24 September through 1200 GMT, 30 September show a general pattern characterized by westerly flow poleward of about 35°N, a persistent anticyclonic circulation in the vicinity of 30°N 60°W, a mid-Atlantic cyclonic trough extending northeastward from about 20°N 55°W, and a trough axis oriented nearly east-west near latitude 20°N and extending from Puerto Rico westward into the Gulf of Mexico. The latter trough tended to break up into separate cyclonic circulations later in the period. ATS-3 pictures and film loops of cirriform clouds were used, along with conventional and aircraft data, to delineate the stream- line flow pattern at, or near, the 200 mb level. The analyses for 1200 GMT, 26 September, and 1200 GMT, 28 September are shown as figures 12 and 13. The high-level outflow from Hurricane Ginger was readily identifiable on most of the analyses, though in some cases it tended to merge with the outflow associated with the warm anticyclone located farther east (fig. 13). It is believed that the inner part of the high-level outflow from Ginger should show cyclonic curvature, and this feature has been shown on the 200 mb analyses, but confirming observations were lacking in most cases. The radius at which the outflow streamlines first take on anticy- clonic curvature is in doubt, but is believed to be fairly large -roughly 150 n miles or more - since Ginger was an unusually large diameter storm at the lower levels. The region of troughing between Puerto Rico and the Gulf of Mexico spawned three distinct upper-level cyclonic vortices during the period under discussion. The first two of these cyclonic circulations are over Haiti and Central America in figures 12 and 13; a third subsequently formed west of the Florida penninsula. None of these vortices appeared to be associated with synoptic features at either the surface or the 500 mb level. Much of the western Caribbean and Gulf of Mexico areas experienced episodes of showery weather during this period; these episodes were gen- erally associated with easterly disturbances on the surface weather analy- ses but were apparently unrelated to the analyzed flow patterns at 200 mb. The analyses indicate that the bulk of the 200 mb outflow from Ginger was carried off towards the north and east, merging with the westerly flow 333 located near and polewards of 35 N. This appears to have been especially true after 0000 GMT, 26 September. Earlier, the analyses also indicated considerable movement of outflow air from Ginger into the east-west trough axis over the Greater Antilles. The limited data at 200 mb precludes meaningful calculations of divergence associated with the Ginger outflow Figure 12. 200 mb streamlines corresponding in time to figures 7 and 10. Outflow is primarily to the north and east of the storm. Figure 13. Compare with figures 8 and 11. Outflow is now almost entirely to the north and east of the storm. at this level. There does not appear to be much correlation between Gin- ger's direction of movement and the flow pattern at 200 mb, but the in- creased amount of outflow to the north and east that started on 26 Septem- ber may possibly be associated with the subsequent turn of Ginger towards a northerly direction on 27 September. 335 In summary, Ginger during the last week of September was a large but rather diffusely organized storm of minimal hurricane intensity. It ap- pears to have had the warm core structure of the true tropical cyclone, although its proximity to a baroclinic frontal zone throughout the period suggests that the storm may have derived a relatively minor portion of its energy from this source. A temporary intensification on 25 September might have been due to either baroclinic energy input or movement of the storm over warmer waters, or both. Another temporary intensification on 27 September appears to have been related to Ginger's absorbing the re- mains of deceased Tropical Storm Janice. The Ginger circulation was large enough so that it was not readily "steered" in its course by adjacent circulation systems. The westward component of Ginger's movement might be partly attributable to the lati- tude effect in the relationship for conservation of potential vorticity; however, the persistence of intensified easterly flow on the poleward side of the storm circulation, which was associated with a surface anti- cyclone to the north, was undoubtedly associated with the westward drift of the storm. The southward component of movement on 24 through 26 Sep- tember appears to have been associated with the reinforcement of an ti cy- clonic conditions north of the storm; whereas, the change to a northward component of motion on 27 September and thereafter is explainable in terms of the weakening of the northern anticyclone in its western limb and its eventual displacement to the east. Decreasing 500 mb heights to the west and north of the storm on 27 September and later also probably favored movement of the storm in this direction. An outflow layer associated with Hurricane Ginger was usually detectable at the 200 mb level, but data were not sufficient to calculate divergence amounts at this level. SEEDING DECISION AND CHOICE OF EXPERIMENTS As Hurricane Ginger drifted slowly westward on 24 September, a re- quest.for deployment of forces was sent to all STORMFURY contingents and called for an experiment on 26 September. The operational reconnaissance reports consistently portrayed a storm of minimal hurricane strength with maximum winds in the range of 60 to 70 knots; these were fairly distant from the center of the storm — varying between 40 and 60 n miles. Ginger was also characterized by a lack of strong vertical development and by maximum winds that were not sharply peaked, so that the horizontal shears on both sides of the maximum winds were quite weak. Turbulence was noted to be seldom stronger than moderate, and rainfall did not appear to be excessive. Reports also indicated that the eyewall was of relatively modest vertical extent and did not completely encircle the center of the storm. The highest priority experiment on the STORMFURY roster was to dup- licate, as closely as possible, the Hurricane Debbie eyemod experiment of 1969. This type of experiment calls for seeding just outside the eyewall in a region of fairly tall cumulus towers, but it requires that the peak winds be well defined with strong horizontal shears on either side, which was certainly not the case in Hurricane Ginger. Figure 14 shows the wind 336 speed profiles for Hurricane Debbie on 18 August 1969, before and after the seeding experiment. Superimposed on this is a wind speed profile from Hurricane Ginger on 26 September. It is quite obvious that the sharply peaked profiles of Hurricane Debbie are not duplicated by the pro- file of Hurricane Ginger. For this and other reasons having to do with the vertical extent of the wall cloud, and that the peaks should be close enough to the center so that a seeding area located radially outward from the belt of maximum winds should still be well within the annulus of cyclonic circulation at altitudes of 15,000 to 33,000 ft, Ginger was not i 1 i 1 1 1 1 i r~ ~r~ RELATIVE WINDS ■ HURRICANE "GINGER" SEPT. 26,1971 ■ HURRICANE "DEBBIE" AUG. 18, 1969 (Before Seeding) ■ HURRICANE "DEBBIE" AUG. 19, 1969 (After Seeding) 100 90 80 0 - X X X X X X 50 40 30 20 10 0 10 20 30 RADIAL DISTANCE (NAUTICAL MILES) 40 50 Figure 14. Winds in (relative to moving storm center) Hurricane Ginger compared with those in Hurricane Debbie. Note the broad3 diffuse struc- ture of Ginger's eye compared with that of Debbie3 even after seeding. 337 considered to be a suitable subject for the eyemod experiment. Evidence of the broad, diffuse nature of Ginger is even better illustrated by fig- ure 15, where the pressure profiles may be compared with those for Hurri- cane Debbie. It was well recognized beforehand, therefore, that an eye- mod experiment on Ginger would not be likely to result in any dramatic flattening of the peak winds or gross changes in the wind and pressure fields, because they were already so flat or poorly defined. 50- 40 30 20 I- UJ uj 10 u. UJ 2-10 I Q -20 -30 -40 -50 -SO "D" VALUES 1 1 r ADJUSTED HURRICANE "GINGER" SEPT. 26,1971 HURRICANE "DEBBIE" AUG. 18, 1969 (Before Seeding) HURRICANE "DEBBIE" AUG. 19, 1969 (After Seeding ) 20 10 0 10 20 30 RADIAL DISTANCE (NAUTICAL MILES) Figure 15. Comparison of adjusted D-values for the same times as in fig- ure 14. The steep profiles indicate that Debbie was more concentrated, as well as more intense than Ginger. Bote that the Debbie D-values were obtained at 12,000 ft and the Ginger ones at 5,000 ft; hence, the shapes of the profiles, not the absolute values, are pertinent to this compari- son. 338 Two other preplanned experiments were available, either one of which could be used on Hurricane Ginger. In the first, the rainband experiment, both the seeding and monitoring are concentrated on a single band with the objective of finding out whether seeding in various fashions will make the clouds in the band grow or dissipate. Monitoring determines whether changes in the clouds of the seeded band will interact with, and cause changes in, cloud structure of other nearby bands or cause local changes in the wind field that might possibly have more far-reaching effects on the wind structure of the storm itself. The rainsector experiment, ultimately selected for Ginger, involves seeding all the clouds in rainbands located in a specified area of the storm. For example, in one of the seedings of Ginger, the aircraft was told to seed all appropriate clouds between 70 and 90 n miles from the center in the sector bounded by the 10° and 60° (measured from true north) azimuths from the storm center. In practice, this usually meant seeding the clouds in one or more rainbands. Each seeding consisted of dropping one or more silver-iodide flares into any suitable cloud that could be found in the prescribed area. The seeding continued for 50 minutes, and the same clouds might be seeded more than once. Suitable clouds are ideally defined as those having cores of liquid water in a reasonably strong updraft with temperatures colder than -5°C at the seeding level. In Ginger, the seeder aircraft flew at 22,000 ft, where temperatures were about -12°C. The objective of the rainsector experiment is to modify the clouds in the rainbands, presumably through stimulating them to greater growth. If they are encouraged to greater growth by this process, it is hypothe- sized that increasing amounts of low-level air will be inducted into the cloud base and thus be prevented from continuing on an inward spiral to the wall cloud. Consequently, if the clouds are highly seedable and are seeded correctly, there should be some apparent effect on the local wind field. This broadly stated objective of the rainsector experiment has been based on qualitative reasoning only and has not yet been tested by simulation in a dynamic model. The three-dimensional model currently being developed at the National Hurricane Research Laboratory is now reach- ing the point where a numerical simulation of this experiment can be per- formed. Despite the stated objective, note that from what we already knew about the broad, poorly defined wind field and the lack of vertical development in Ginger, it seemed rather unlikely that there would be any significant change in the Ginger wind field in response to seeding. Now, experience indicates that many clouds in the rainbands located within the rainsector area of a typical intense hurricane will be ex- pected to have seedability. They should grow more when seeded than under natural conditions. If this stimulated growth taps the stream of low- level inflowing air and causes it to ascend until it reaches the outflow layers of the storm, this air will never spiral into the old eyewall , ac- celerating as it moves to lower pressures and giving up its latent energy in the wall cloud. Some proponents of this experiment have argued that, under optimum conditions, one might even create a secondary vorticity cen- ter or possibly a secondary wind center near the area of seeding, and that 339 this second center might compete with the main circulation for the avail- able energy supply. For Ginger, there was such a scarcity of clouds with suitable up- drafts and liquid water content that STORMFURY personnel had to select the sector to be seeded on the basis of where such clouds could be found at radii greater than the radius of maximum winds. The limited number of suitable clouds strongly restricted the choice of a sector. THE RAINSECTOR EXPERIMENT, 26 SEPTEMBER 1971 The A-6 aircraft (Navy Attack Squadron EIGHTY-FIVE (VA-85), Oceana, Virginia) were scheduled to make the first three seedings with the WC-130 (from the Air Weather Service, 53rd Weather Reconnaissance Squadron, Ramey AFB, Puerto Rico) as a back-up for the third seeding and primary for the fourth and fifth seedings. Because of radar problems on the command and control aircraft, it was neither practical to direct the seeders to as- signed areas on the morning of 26 September nor possible to record where the seedings would be relative to the center of the storm. Therefore, the first three seedings were cancelled. By the time the Air Force WC-130 ar- rived in the storm, a back-up WV-121 (Navy Weather Reconnaissance Squadron, FOUR (VW-4), Jacksonville, Florida) had arrived with better working radar, and the WC-130 was better able to record its own track relative to the hurricane center. Consequently, the WC-130 was authorized to seed, and two "highly selective" seedings were accomplished before the aircraft had to return to base. It turned out that the aircraft could top most of the storm clouds at 23,000 ft. The seeding was carried out at 22,000 ft in the southern sector of the storm; they are shown in figure 16. Thirty-one pyrotechnic devices (each producing 160 grams of silver iodide) were used on the first seeding, which commenced about 1700 GMT, 26 September. These devices were placed in updrafts showing some evidence of the presence of liquid water —as indicated by icing on the cockpit wind screen. It was the judgment of those aboard the seeding aircraft that these updrafts were rather weak and the liquid water content moderate or less. The return flight along the previously seeded line seemed to indicate even weaker updrafts and downdrafts, fine ice particles, and little or no liquid water. No startling cloud growth was evident, and it appears unlikely that any significant contribution to the cirrus shield resulted from the seeding. Furthermore, it was the opinion of those a- board that, in view of the state of the rainband encountered, the area had probably been overseeded. The second target group was seeded with 15 flares between 1900 GMT and 1913 GMT using considerably more restraint. Here again the rainbands consisted of small (1 to 3 n miles across), rel- atively weak and diffuse echoes 5 to 15 n miles apart. These cells con- tained only mild to moderate updrafts. The earlier seeding was performed over a 45° sector south of the storm center at a mean radial distance of 90 n miles from the center. The second seeding was over a more limited sector located south-southwest of the storm center at about 135 n miles radial distance from the center of the storm. A considerable portion of these passes were flown under IFR conditions, and the targets were selected by their appearance on the APN-59 X-band radar. 340 AREAl SEEDING TIME AIRCRAFT SEPTEMBER 26. 19 E ALTITUDE FLIGHT NO. I 1701-1717 S 1734-1754 WC-130 22.000FT 710926V 1901-1914 WC-130 22.000FT. 710926 V SEPTEMBER 28.197 ■ 1402-1404 ft 1407-1410 A- 6 22,000 FT. 710928L 1550-1553 A-6 22.000FT 710928M 710928M ~ 1556-1558 A-6 22.000FT 1602-1607 A-6 22.000FT 710928M 1611-1616 1718-1725 M. 22.000FT 22.000FT 719??8M ^ A-6 710928N 1727-1732 A-6 22.000FT 710928N 1735-1736 A-6 22.000FT 710928N N'T* 1742-1747 A-6 22.000FT 710928N .. "GINGER" HURRICANE 150 Figure 16. Approximate locations of seeded areas on 26 and 28 September. Seeding on the 26th was south of the storm center; whereas, the more massive seedings of the 28th were in the northeastern quadrant. 341 Seedability of the clouds was calculated using data from vertical soundings made with dropsondes released from the Air Force planes. By seedability is meant the amount of additional vertical growth that might be expected from a seeded cloud as compared with growth in the unseeded clouds. These computations are reported in appendix E, "Modeling the Seeding Effect in Hurricane Ginger," by Scott and Dossett. Computations using data from several of the soundings indicated no seedability, but those made from the dropsonde sounding closest to the seeded clouds did indicate considerable seedability under certain assumptions (see appen- dix E). Unfortunately, no sounding was made close to the seeded clouds -- neither close in time nor close in space. Measurements of the liquid water content at the 5,000 ft level (cal- culated from samples taken by the foil impactor) showed the presence of precipitation size particles over a radial distance of 30 n miles in the southeast quadrant. The maximum liquid water per cloud averaged approxi- mately 1 gm/m3 over this region with one or two isolated clouds with val- ues of 2 gm/m3. In the northern quadrants the average and areal coverage was much less, and only isolated clouds contained significant precipita- tion sized particles. These are very low water contents for hurricane clouds. No recorded measurements are available for the layers above the 0°C level on 26 September, but the water content should be much smaller at those higher levels. This conclusion was supported by the visual ob- servations of the crew in the seeder aircraft. In fact, few of the "water* clouds (those containing little ice) extended far above the zero isotherm. If one considers the relatively small amounts of liquid water in the clouds, one should not expect dramatic changes following the seedings. The results of the seedability computations suggested some growth should occur, but enough inconsistency existed in the results of these computa- tions (see appendix E) that their indications relative to the seeded clouds are not clear. RESULTS The X-band radar available aboard the seeder plane was not equipped to take time-lapse photography. Consequently, no consistent comparison is possible between the radar echoes before and after the seedings. On some of the other aircraft, it was subjectively noted that some brightening of the seeded echoes occurred following the seeding itself; however, since none of these other aircraft were equipped for time-lapse photography either, the evaluation remains a subjective one. The ATS-3 satellite photographed Ginger repeatedly during this period. Those who examined the echo brightness in the area of the seeding thought that some brighten- ing of the echo and increase in the cirrus blow-off occurred; however, this was far from a conspicuous change. Dr. Fujita of the University of Chicago used individual frames selected from the morning and late after- noon of 26 September, and he composited these frames so that they could be compared at length. An analogous set of frames was also prepared from Ginger on 27 September. He found that unbiased evaluators seemed to be able to distinguish the day (26 September) on which the seeding occurred 342 from the day (27 September) on which no seeding occurred; however, he was unable to make any objective evaluation of this difference. (It might al- so be pointed out that the storm was more intense on the 27th than on the 2 adjacent days.) The expectation that such limited seeding over a relatively restricted area in both time and space should have any perceptable effect on the large- scale aspects of such a big, sprawling storm seems highly questionable. Figure 17 shows the trend on minimum sea-level central pressure for 25 through 30 September with the seeding times also indicated in the figure. The minimum sea level pressure curve suggests that the storm began deepen- ing about 1200 GMT, 26 September, and that during most of the seeding per- iod the deepening trend continued, although a slight filling was apparent immediately following these limited seedings. It is deemed exceedingly unlikely that the seedings had any cause-and-effect relationship to the small rise noted about 0000 GMT, 27 September. Figure 18 shows wind speed profile envelopes from southwest to northeast for the before-seeding and after-seeding traverses of Ginger on 26 September and early 27 September. The two solid lines indicate the range of the maximum and minimum winds (relative to the moving storm) recorded at the various radial distances from the center of the storm for 0940 GMT through 1732 GMT, 26 September, before the seeding took place. The two broken lines indicate the range of wind speeds measured on the several passes made by the aircraft between 2050 GMT, 26 September through 0410 GMT, 27 September after the seeding. There is wery little suggestion of any systematic change in the southwest profile, although a small decrease of the peak winds after the seeding HURRICANE GINGER SEPTEMBER 1971 GQ l±J K980 Z) w LJ tr a. 970- < I- Z LJ O 960 ' ' 1 1 1 1 1 1 1 1 1 1 --*~ f / / \ . MINIMUM^ /'\ / \ ^^SEA LEVElX,-~x / ^ \ 1 / / / 1 1 ^ N PRESSURE / \J \ \ i II \ / III 1 TIME OF SEEDING i i i i i i i OOZ 12Z 09/25 12Z 09/26 12Z 09/27 TIME 12Z 12Z OOZ 09/28 09/29 09/30 Figure 17. Variation of central sea-level pressure in Hurricane Ginger. Profile has been slightly smoothed. The times of seeding runs are also indicated. 343 HURRICANE GINGER RELATIVE WINDS SEPTEMBER 26,1971 80 T — BEFORE SEEDING (0940-1732Z) — AFTER SEEDING (2050-0410Z) A />'; : s, », » f, >' v 20 0 20 RADIAL DISTANCE (NAUTICAL MILES) 40 60 80 Figure 18. Composite envelopes of relative winds in Hurricane Ginger be- fore and after seeding. Profiles are on a southwest-northeast axis pass- ing through the center of the storm. about 60 n miles from the center is apparent. There is also a suggestion of an increasing peak about 35 n miles out. In the northwest sector of the storm, the general level of the winds seems to show an increase from about 35 n miles to possibly 75 n miles out, and what appears to be a minor maximum at 30 n miles radial distance may have weakened some in the time period. More significant changes are apparent on the southeast-northwest pro- files (fig. 19), which suggest that after the seeding period wind speeds were somewhat stronger in the southeast quadrant of the storm and signifi- cantly stronger in the northwest sector. It seems reasonable to attribute these changes to the slow deepening or intensifying trend indicated by the central pressures rather than to associate them with the seedings. A sim- ilar comparison was made of the temperature profiles before and after the seedings. They show little consistent change in the southwest and north- east sectors, but they also suggest that there may have been appreciable warming in the southeast sector of the storm out to about 60 n miles. In view of the radius at which the seeding was conducted, it seems rather difficult to associate this warming of 1 to 2°C with the seeding activity. 344 HURRICANE GINGER RELATIVE WINDS SEPTEMBER 26,1971 80- 70 5J60- H O z w50 o Id Ul 0. 40- 30 20 10 ft f I! r 1 1 1 - BEFORE SEEDING (1025-1130Z) - AFTER SEEDING (0001-0225Z) Hlfo 0 60 SE V'/VW V\ '.'.; MW 20 0 20 RADIAL DISTANCE (NAUTICAL MILES) 40 80 80 Figure 19. Composite envelopes of relative winds in Hurrioane Ginger be- fore and after seeding. Profiles are on a southeast-northwest axis pass- ing through the center of the storm. SEEDING EXPERIMENT OF 28 SEPTEMBER 1971 Hurricane Ginger reached a temporary minimum pressure just before 0000 GMT, 28 September, and thereafter began a long trend of filling, although some minor dips occurred in this weakening trend. On 28 September, the radars of the VW-4 Squadron from Jacksonville were working better, and the three seedings scheduled by the A-6 attack bombers of the Navy VA-85 Squad- ron from Oceana, Virginia, were executed as scheduled. In these cases, each seeding consisted of the disposition of about 200 pyrotechnic devices. They were fired at about 5 second intervals from the aircraft, which flew under IFR conditions at 22,000 ft; no special effort was made to place them solely in significant updrafts associated with abundant supplies of super-cooled liquid water. The first seeding passes took place in a 15° sector at about 60° azimuth between 1402 and 1410 GMT (fig. 16). The second seedings were between 1550 and 1616 GMT. The first seeding was at a radial distance of about 100 n miles, and the second seeding was cen- tered about 60 n miles from the center of the storm. The third group of seeding passes was about 100 n miles out over a 45° sector, centered at an azimuth of about 40°, and lasted from 1718 through 1747 GMT. 345 The computations made with the dropsonde data indicated that poten- tial for clouds to have seedability (appendix E) was very poor in the vicinity of the first seedings (1402 to 1410 GMT), but was quite good* for the areas of the second and third seedings (1550 to 1616 GMT and 1713 to 1747 GMT). The measurements of liquid water content at 5,000 ft (cal- culated from samples taken by the foil impactor) showed several clouds in the eastern quadrant and a few clouds in the northern quadrant that had 1 to 2 gm/m3. No recorded measurements were made of the water content at levels where the temperatures were less than 0°C, but samples taken with the Formvar replicator at 19,000 ft indicated that the water content was slightly greater than had been observed from the Air Force aircraft while seeding 26 September. An observer in the RFF airplane flying in the areas of the first three seedings at 19,000 ft reported that yery little turbu- lence existed and that the clouds were somewhat stratiform in structure mixed with weak vertically suppressed cumuli. The liquid water content of the clouds was certainly less than one would normally expect in moderate to intense hurricanes. Again, as on 26 September, the combination of lack of vertical cur- rents in the clouds and lack of liquid water at subfreezing temperatures suggest that the clouds should not be expected to grow dramatically after the seeding. The computations of seedability (appendix E) did, however, suggest the potential for growth in areas of the second and third seedings. The Air Force WC-130 from the 53rd Weather Reconnaissance Squadron at Ramey was scheduled for the fourth and fifth seedings. It approached Ginger from the south-southeast. Aside from some patchy cirrus overhead and a few patches of altocumulus, the storm was easily topped at 22,000 to 23,000 ft. The southern eyewall came close to flight altitude, and the eyewall on the northern side was estimated to reach some 27,000 to 29,000 ft but ap- peared to be quite fuzzy with soft glaciated tops. The sky over the eye was clear of all cirrus. In the previously seeded northeast sector of tbe storm, outside the eyewall the clouds reached just about to flight altitude, so that the flight was conducted in and out of the tops of these clouds. On the whole they appeared to contain s/ery little liquid water. For the most part there seemed to be a more than adequate supply of ice crystals available. The updrafts were rather weak and widely scattered. Conse- quently, this was deemed an unsuitable area for further seeding activity. Taller and higher cells were visible in the distance, but when these were checked, they were found to be too far out radially to make any reasonable contribution to the designed seeding plan. Radar difficulties aboard the command and control aircraft, and its backup, further delayed the resump- tion of any coordinated seeding activity. Seedable clouds were finally located well east of the storm center and some 190 to 200 n miles out. After clearance to seed was obtained, five pyrotechnic units were dropped ^Appendix E discusses the limitations of these calculations and the assump- tions used in making them. 346 into selected cells that showed a definite updraft but only a barely ac- ceptable level of liquid water content. Visual monitoring failed to sug- gest that the seeding contributed greatly to the cloud growth. The clouds were already growing when seeded; they appeared to become glaciated a short time after seeding and failed to rise much above 26,000 ft, as estimated from the aircraft. In summary, the convective cloud tops observed during this flight seldom reached or exceeded 23,000 ft, except in the two or three cumulus lines located well south of the main storm envelope which were traversed en route to Ginger. The northern half of the eyewall contained fuzzy glaciated tops that extended to an estimated 27,000 to 29,000 ft but con- tained no active growth at flight altitude of 22,000 to 23,000 ft. There were a few isolated cells located 75 to 150 n miles to the northeast, east, and southeast of the storm, but even the best of updrafts encountered on penetrating these cells were relatively weak, and the liquid water con- tents were relatively light, producing only very little visible streaming on the wind screen. Most of the precipitation encountered was either ice granules, which did not adhere to the screen, or small ice particles, ap- parently graupel . Considerable aircraft icing was encountered, both cloudy and clear, and at one time, around 2040 GMT, the WC-130 was iced to the point of sinking at 300 ft/minute. The characteristics of all clouds en- countered indicated that seeding had already occurred, either naturally or artificially. The appearance of cloud bases in the large stretches of cloud visible outside of the eye, especially in the south and southeast sectors, suggested relatively dry conditions at lower levels. This lack of low-level cloudiness may well be associated with the cooler sea-surface temperatures, which are described in appendix D. RESULTS Once again some of the observers watching the radar scopes at the time of the seedings reported that echoes in the seeded area became brighter during the first three seedings. And, once again none of the radars close to the seeded areas had time-lapse photography capabilities, so that no quantitative analysis can be made. The ATS-3 satellite photo- graphs of NASA were studied in a similar fashion to those of 26 and 27 September. Here again the satellite pictures suggested brightening to some of the subjective evaluators, and Dr. Fujita's composite photographs of the storm "before" and "after" were again subjected to unbiased evalua- tion. Most of those who studied the photographs and compared them with those of an unseeded day came to the conclusion that they could tell the seeded day (28 September) from the unseeded day (27 September). No objec- tive measurement of this difference was possible. The wind speed envelopes for the experiment of 28 September are pre- sented in figures 20 and 21. In the south-southwest sector of the storm, there is a strong suggestion that wind speed values are being eroded with time, and that the "before" profile may be significantly different from the "after" seeding wind speed envelope (fig. 20). In the north-northeast quadrant, there is some erosion of values in the inner 30 to 35 n miles, 347 HURRICANE GINGER RELATIVE WINDS SEPTEMBER 28,1971 80 T BEFORE SEEDING (0740-0916) AFTER SEEDING (2336-0215Z) 20 0 20 RADIAL DISTANCE (NAUTICAL MILES} 80 Figure 20. Composite envelopes of relative winds observed in Hurricane Ginger before and after seeding. Profiles are on a south- southwest north-northeast axis passing through the center of the storm. There was only one before-seeding research flight in the south-southwest sector. but there appears to be little significant change beyond 40 n miles from the center. Similarily, in the west-northwest sector (fig. 21), the inner values of the envelopes seem to be eroding with time up to a radial dis- tance of about 70 n miles from the center. On the east-southeast side of the storm, there appears to have been little significant change, other than the emergence of the peak about 50 n miles out. These changes may be compa value profiles shown in figures were gathered between 0740 and D-value in the center is about between 2336 GMT, 28 September mum D-value has risen to about noted at most radii, they were the south-southwest sector, the red with the 22 and 23. 0916 GMT, 28 -680 ft. Th and 0215 GMT -530 ft. Al greatest in weaker wind envelopes of the smoothed D- The "before" profiles, which September, suggest the minimum e "after" envelopes gathered 29 September suggest the mini- though rises in D-values were the center of the storm. In speeds seem to be associated 348 HURRICANE GINGER RELATIVE WINDS SEPTEMBER 28, 1971 r 1 r BEFORE i SEEDING i (0920-1144Z) i i 80 AFTER SEEDING (2135-2336Z) - 70 J\ i x » ;yvV >4« t 55 6° t- o z £50 \A v »1*^v \AhyiA.- o UJ tK a. lr\( u to HU o *30 \ \ \ Z'V'aT /iff' Jy 20 // tr^fVf - 10 V \ *» M_ \f n5v Ja UJ -30 — ^S>^^ \ ' J^ s wm u. x^ NN N / ^r * * 40 * V / ^r ~* o ^^ X C-40 X. VN Nv '* J**"*' "" OT Xv N N 0 UJ >X ^""n. ^N + ' sr^S 3 ^\v ^V X>. *' •yy -J > -50 - N. ^-- „"*~'*~-. ~~'„~S-''' '' /j Y Q >v ~~~: // -60 - -70 - ssw 1 l_ ... 1 i NNE 80 60 40 20 0 20 RADIAL DISTANCE (NAUTICAL MILES) 40 60 80 Figure 22. Composite envelopes of adjusted D-values observed in Eurrioane Ginger before and after seeding. Profiles are on a south- southwest north-northeast axis passing through the center of the storm. SUMMARY OF THE HURRICANE GINGER EXPERIMENTS As we have seen, the seeding of 26 September consisted of dropping 46 pyrotechnic flares in the southerly sector of the storm. Some reported echo brightening on their radar scopes following the seeding, but no docu- mentation of such changes is available. A return run down the seeded band suggested to those aboard the seeding plane that there was less liquid water and that the updrafts were weaker than before the seeding. In view of the loose organization of Hurricane Ginger and its marked lack of verti- cal development, there is considerable reason to doubt that any signifi- cant updraft was stimulated by the seeding. Certainly no outstanding cumulonimbus activity appeared to develop at the site. The wind speed profiles suggest that only minor changes did occur (except for some in- crease in wind speed in the northwest sector), but there seems consider- able reason to doubt the small significance of the variations if we were inclined to attribute them to the seeding. 350 HURRICANE GINGER ADJUSTED D -VALUES SEPTEMBER 28, 1971 o - -10 - -20 - ui UJ £ -30 < T -50 -60 - -70 - 80 nrrnnr **rrniNr fnndT 1^7^ 7] — DtrUnt OttUlfNo \\JjH! IljjlJ AFTER SEEDING (1338-2141Z) " - n^V S*' /l/ ~ "^V^V ,'''' /y' VnN%^s. •• ^^s ^V ^^7 s^i^w 7'V /^ ^Sv •'•'' S/ >^^ ^s'S j// ~ N^^X^ — ""^^ jTS ^^0>y WNW ESE 60 40 20 0 20 RADIAL DISTANCE (NAUTICAL MILES) k!0 60 80 Figure 2d. Composite envelopes of adjusted D-values observed in Hurrioane Ginger before and after seeding. Profiles are on a west-northwest east- southeast axis. Both of these indicate a rise in central pressure at this level (43780 ft). On 28 September the seeding was much more extensive, in that over 600 of the pyrotechnic flares were dropped in the north-northeastern sector of the storm within about 3^ hours. Once again a brightening of the radar echoes was noted at least temporarily. However, there was no evidence of extensive build-ups either visually or by radar when the fourth seeder plane scanned the area some 2 hours later. The crew had considerable difficulty finding suitable seeding targets in the area that had already been seeded. It may be reasoned that the seeded air mass had been swept on around the storm and the effects, if any, could well be 50 to 100 n miles removed from the seeding site. There was, it is true, some suggestion that the profile of wind speed in the west-northwest and south-southwest sectors of the storm tended to show some flattening out, i.e., the inner peak winds tended to be suppressed. And, to a lesser extent this was also true in the north-northeast quadrant. However, these effects are noted inside of the 60 n mile radius and only a small fraction of the seeding took place within the 50 to 60 n mile annulus. It seems more reasonable to associate these decreases with the longer term filling of the storm, which continued right on into 29 September. 351 Direct evidence that any significant cumulonimbus activity was stim- ulated by the seeding is totally lacking. In view of the hurricane's very limited ability to engender its own conspicuous hot towers, this hardly seems very surprising. We are forced to the tentative conclusion that Hurricane Ginger was a sprawling old storm that (on 28 September) was filling and losing strength. It did not provide the environment necessary for any vigorous new growth, and under these conditions the areas were undoubtedly over-seeded to such an extent that significant growth was highly unlikely. APPENDIX C HURRICANE MODELING AT THE NATIONAL HURRICANE RESEARCH LABORATORY (1971) Stanley L. Rosenthal National Hurricane Research Laboratory INTRODUCTION The past year saw substantial improvement of the National Hurricane Research Laboratory (NHRL) asymmetric hurricane model, as well as signifi- cant progress in the physical interpretation of its results. A summary of this work and an account of current efforts are presented in the next section. A comprehensive summary of "hurricane seeding" simulations was prepared and published. These calculations were performed with the NHRL circularly symmetric model. Major results are briefly summarized in "Numerical Guidance for Project STORMFURY." Attempts were made to design a hurricane model with an explicit rep- resentation of some cloud physical process. Such a model would allow a more rigorous simulation of cloud seeding experiments. Unfortunately, because of other duties, this research was not pursued as vigorously as planned. However, enough was done to make it clear that some basic re- search (theoretical) is essential before the problem can be efficiently carried further. These matters are expanded upon in "Cloud Physical Ef- fects and Hurricane Models." The Theoretical Studies Group (NHRL) had substantial changes of per- sonnel during 1971. This, in part, explains why our progress has been less rapid than in preceeding years. Dr. Richard A. Anthes left the group in August to take a faculty position at Pennsylvania State Univer- sity. Mr. Walter J. Koss left in August to spend a year in residence at Florida State University on a NOAA scholarship. Because of a rearrange- ment of assignments (related to the possibility of Project STORMFURY op- erations in the Pacific), Mr. James W. Trout has taken on supervisory responsibility for the NOAA/Miami computer terminals, and this now con- sumes approximately 50 percent of his time. Mr. Michael S. Moss (M.S., meteorology, Florida State University) joined the group and now has prime responsibility for experimentation with our circularly symmetric model. Dr. Robert W. Jones (Ph.D., meteo- rology, Oklahoma University) is now a temporary member of the group and is responsible for further development of the asymmetric model. Dr. Burt Morse (Ph.D., applied mathematics, The Courant Institute, N.Y. University) 353 joined the group and is responsible for the mathematical problems con- cerned with variable grids, nested grids, and moving grids. THE ASYMMETRIC MODEL Papers by Anthes et al . (1971a, 1971b) describe the status of this model at the time of preparation of last year's annual report. Detailed discussion of the improvements during the year is given by Anthes (1971d). Trout and Anthes (1971) give the results of detailed statistical analysis of the asymmetric features generated by the model. A brief survey of these papers follows. The accuracy of finite difference estimates of space derivatives has been improved through the introduction of a staggered horizontal grid. An explicit water-vapor cycle is now included; hence, the assumptions about boundary layer humidity required by Anthes et al . (1971a) are no longer necessary. The formulations of the Austausch coefficients have been im- proved. In particular, the coefficients for lateral mixing are calculated from a nonlinear form similar to that employed in the general circulation models of the Geophysical Fluid Dynamics Laboratory, NOAA. Lateral bound- ary conditions have also been substantially improved. Despite these revisions, experimental results are qualitatively simi- lar to those obtained from the original version of the model, initial con- ditions consist of a circularly symmetric vortex in gradient balance. With symmetric initial conditions, the solutions to the differential equa- tions are symmetric for all time. However, truncation errors, round-off errors, and the lateral boundaries produce weak asymmetries (on the order of 10-10 percent of the symmetric component) after the first time step. These perturbations grow with time and eventually become a significant part of the circulation. Experiments show an early period with near cir- cular symmetry followed by a later period with important asymmetric fea- tures that are similar to those observed with real hurricanes (spiral rainbands, anticyclonic eddies in the outflow layer, concentration of the outflow into two jets). Since it is more appealing to deliberately introduce perturbations of known amplitude and variance rather than to rely on truncation and round-off to produce the initial asymmetries, Anthes (1971d) performed an experiment in which asymmetries of 10"1 percent (10 orders of magni- tude greater than the perturbations produced by truncation and round-off) were introduced by adding random numbers to the initial velocity compo- nents. This shortened the symmetric phase of the life-cycle and allowed the asymmetries to become evident earlier. However, the structure and intensity of the model storm were essentially the same as that of the earlier experiment. The model produces spiral rainbands, with rainfall rates that average about 2 cm/day. The bands are about 90 km wide at large distances from the storm center and merge with the eyewall in the storm core. These bands rotate cyclonically and propagate outward at about 24 knots. The temperatures in the bands are not appreciably different from those of 354 their environment. Although the bands appear to be internal gravity waves modified by latent heat release, the mechanism that excites them has as yet to be explained. There does seem to be an interesting (al- though obscure) relationship between the appearance of the rainbands and the breakdown of symmetry in the outflow layer. During the symmetric phase (when circular variance is small), wave number four accounts for virtually all of the little variance that is present. The orientation of wave number four with respect to four rather sharp kinks on the lateral boundary of the grid indicates that this early asymmetry is forced by the shape of the grid. However, when the asymmet- ric stage is reached, wave numbers one and two become dominant and ac- count for most of the variance. According to Trout and Anthes (1971), during the symmetric stage, the standard deviations of the wind compo- nents are about 0.3 m/sec, while that of the temperature is about 0.05°C at a radius of about 100 km in the outflow layer. During the asymmetric stage, in the same portion of the storm, the standard deviations of the wind components are about eaual to the circular averages. The tempera- ture field, however, remains quite symmetric with its standard deviation rarely over 1°C. Anthes (1971d) established dynamic instability as the mechanism that allows the asymmetric component of the outflow layer to grow. He also showed that dynamic instability was generated as a natural consequence of the model storm's internal dynamics and that the tipping term in the vorticity equation played the major role in this process. The storm's kinetic energy budget revealed the following energy cycle. The kinetic energy of the low-level, circularly averaged flow is maintained by cross isobaric flow of the symmetric component of the transverse circulation. In the outflow layer, the symmetric component of the motion is maintained by vertical transports of kinetic energy by the symmetric component of the transverse circulation. The kinetic energy of the asymmetric compo- nent of the motion in the outflow layer is derived from the kinetic en- ergy of the symmetric motion. The asymmetric component loses kinetic energy through friction and baroclinic effects. With the departure of Dr. Anthes (August 1971), this research was temporarily suspended. Dr. Jones has now resumed the work and is pres- ently coding the model to include a computational domain of 4000 x 4000 km. This involves using expanding grids similar to those discussed by Koss (1971). There will also be input from the work of Dr. Morse, who has been investigating both the distortion of the dynamics that occurs at the interface of grids with different resolution and techniques for minimizing these effects. In this first attempt, the fine mesh portion of the domain will remain stationary. After we obtain some reasonable results and gain some experience with this version of the model, the next (and most difficult) aspect of the problem will involve a fine mesh that moves with the vortex center. 355 NUMERICAL GUIDANCE FOR PROJECT STORMFURY Subject to the reservations that were thoroughly discussed in last year's Annual Report (Rosenthal, 1971c), Rosenthal and Moss (1971a) pre- sented the results of a variety of "hurricane seeding" simulations per- formed with the model described by Rosenthal (1970; 1971b). From these calculations, it was concluded that the original STORMFURY hypothesis, which called for seeding of the eyewall alone, should not be employed in future field experiments. All calculations that attempted to simulate this tactic resulted in storm intensification. Simulations with seeding conducted from the eyewall outward, or com- pletely outward of the eyewall, indicated that a seeding operation just radially outward from the eyewall is optimum. This is primarily reflected in the rapid response to the artificial heating rather than in the reduc- tion of the maximum winds. The latter were more or less the same for all calculations in which a new eyewall was developed at a larger radius. The experiments indicate that seeding must be continued until the new eyewall is well formed, if reductions of the maximum winds are to be achieved. On the other hand, continued seeding for a prolonged period at the same radii beyond the time that the new eyewall is established does not appear to be desirable, because this tends to reduce the effect of the earlier seeding. In calculations where the peak storm winds were reduced, the ultimate effect was to displace the eyewall 10 km (one grid point) radially outward and to reduce the wind maximum at sea level by 3 to 4 m/sec. Variations in the seeded radii and in the rate of artificial heating altered only the amount of time required for these changes to take place. Their re- sponse times also varied with the storm's age and were shorter with older storms. The increased transverse circulation that results from applying the artificial heating tends to increase winds at relatively large radii. This is not a particularly desirable feature of a modification experiment, since it may well produce adverse storm-surge effects. However, Rosenthal and Moss (1971a) showed that these increased winds are vital if the eye- wall is to be displaced to a larger radius, and if the peak winds are to be reduced. It is the increased centrifugal and Coriolis forces, arising from these stronger winds, that prevent inflowing air from reaching the original eyewall and thus force ascent at larger radii. All tactical variations of a seeding experiment that add heat out- side the center of the natural eyewall resulted in storm modification similar to that described above. However, the tactics with longer re- sponse times are less likely to be effective in a real hurricane. It thus appears that field programs conducted with older storms, seeded just beyond the eyewall, are most likely to result in beneficial modification. The calculations also indicate that the seeding operation should be mas- sive, since larger heating rates clearly lead to a more rapid response of the model storm. Since many of the numerical experiments show tempo- rary increases in the maximum wind during the early phases of the opera- tion, it appears that the current policy of not seeding storms whose early landfall is predicted should be continued. 356 CLOUD PHYSICAL EFFECTS AND HURRICANE MODELS The methods employed by NHRL for the simulation of cloud seeding experiments with hurricanes (Rosenthal, 1971a; Rosenthal and Moss, 1971a) are clearly less rigorous than is desired. These procedures are, however, dictated by the cumulus parameterization (Rosenthal, 1970) that contains no explicit representation of cloud physical effects. If we are to strengthen this aspect of our modeling program, a technique for predict- ing the cloud-radius spectrum and the areal coverage of cumulus at each grid point must be derived. The cloud-size spectrum along with a suitable one-dimensional cloud model (e.g., Simpson and Wiggert, 1969) would allow computation of the vertical structure of clouds including their liquid water content and tops. This information, together with the areal cover- age, would allow computation of the total supercooled water content of a given grid module and, therefore, the amount available for freezing by artificial nucleation. Techniques developed for simulation of silver- iodide seeding (e.g., Simpson and Wiggert, 1969) could then be employed in a rather straightforward fashion. A crude start in this direction was made during the past year. The cumulus parameterization (Rosenthal, 1970) was removed from the code for the circularly symmetric model. It was then assumed that all convective clouds at a given grid point were of the same radius. The cloud radius was then specified by a hypothetical relationship between cloud radius and boundary-layer convergence. This is based on synoptic studies (e.g., Malkus, 1960) that indicate a marked correlation between low-level con- vergence and cloud radius. From the cloud size, a simple one-dimensional model was used to compute the cloud structure including rainfall rates and cloud water content. With the rainfall rate per cloud, and a further hypothetical relationship between macroscale water- vapor budget properties and space averaged rainfall at a grid point, the areal coverage of con- vective clouds could be computed. The clouds were then treated as "turbulence elements" and their in- teractions with the macroscale were computed by direct evaluation of the Reynolds correlation terms (see Kasahara and Asai , 1967, for a philosoph- ically similar approach) that appear when the governing equations are averaged over a grid module. Two types of calculations were carried out. In one set, initial con- ditions were taken from the mature stage of a storm generated with the original version of the model. In the second set, the initial conditions were those used with the original model (Rosenthal, 1971a). The first set of calculations behaved reasonably well. After an initial shock that re- sulted from the change in the parameterization of the cumulus, the model storm reached a new (more intense) steady state, whose macroscale struc- ture was not very different from that given by the original model, and was fairly realistic. The clouds, however, tended to show unrealistic buoyancies and excessive updrafts. The second set of experiments yielded results that were quite unac- ceptable. In these, eyewall-like features formed at radii of about 200 km with a wind maximum closeby. The clouds were much too vigorous and, at this stage, were only minimally under the control of the macroscale. The "eyewall," once formed, showed no tendency to migrate inward, as occurs with the original version of the model (Rosenthal, 1970). Attempts to alter these features, by changing the hypothetical relationship between cloud radius and boundary-layer convergence, only delayed their appearance without affecting their basic characteristics. A few experiments were conducted in which cloud bouyancy and updraft were arbitrarily limited to maxima that are empirically realistic. This resulted in substantial improvements but not to the degree desired. The "eyewall," once formed, was able to move inward to some extent and the vortex structure that was generated had, at least, some resemblance to a hurricane. Careful examination of these results seems to indicate that cloud radius is probably not a function of boundary-layer convergence alone. If the usual entrainment law (entrainment proportional to the reciprocal of cloud radius) is basically correct, and if we recognize thaj observa- tions indicate that even large tropical cumuli are only 2 to 3 C warmer than their environment, it would then appear that cloud radius must some- how be inversely related to the degree of conditional instability in the atmosphere. It is recognized that this conclusion directly conflicts with certain theoretical analyses of conditionally unstable disturbances. These theoretical studies are, however, based on models that are suffi- ciently different from clouds to allow at least some room for doubt and further theoretical analysis. It appears that further efforts with cloud models should be made in an attempt to provide some insight into these matters before attempting to engineer hurricane models with trial and error approaches, such as those described earlier in this section. It is hoped that investigations of this type will be performed during the next year. REFERENCES The following papers were published during the past year by members of the Theoretical Studies Group (NHRL): Anthes, R. A. (1971a): Iterative solutions to the steady-state axi sym- metric boundary layer equations under an intense pressure gradient. Monthly Weather Review, 99, No. 4, pp. 261-268. Anthes, R. A. (1971b): A numerical model of the slowly varying tropical cyclone in isentropic coordinates. Monthly Weather Review , 99, No. 8, pp. 617-635. Anthes, R. A. (1971c): Numerical experiments with a slowly varying model of the tropical cyclone. Monthly Weather Review , 99, No. 8, pp. 636- 643. Anthes, R. A. (1971d): The development of asymmetries in a three-dimen- sional numerical model of the tropical cyclone, ERLTM-NHRL Technical Memorandum No. 94 (NOAA/NHRL, Miami, Florida) 55 pp. 358 Anthes, R. A. (1971e): The response of a three-level axi symmetric hurri- cane model to artificial redistribution of convective heat release, ERLTM-NHRL Technical Memorandum No. 92 (NOAA/NHRL, Miami, Florida, 14 pp. Anthes, R. A. (1972): Non-developing experiments with a three-level axi- symmetric hurricane model, ERLTM-NHRL Technical Memorandum No. 97 (NOAA/NHRL, Miami, Florida) 18 pp. Anthes, R. A., S. L. Rosenthal, and J. W. Trout (1971a): Preliminary re- sults from an asymmetric model of the tropical cyclone. Monthly Weather Review, 99, No. 10, pp. 744-758. Anthes, R. A., J. W. Trout, and S. L. Rosenthal (1971b): Comparisons of tropical cyclone simulations with and without the assumption of circular symmetry. Monthly Weather Review, 99, No. 10, pp. 759-766. Anthes, R. A., J. W. Trout, and S. S. Ostland (1971c): Three dimensional particle trajectories in a model hurricane. Weatherwise, 24, No. 4, pp. 174-17a. Black, P. G., and R. A. Anthes (1971): On the asymmetric structure of the tropical cyclone outflow layer. Journal of the Atmospheric Sciences, 28, No. 8, pp. 1348-1366. Koss, W. J. (1971): Numerical integration experiments with variable reso- lution two-dimensional Cartesian grids using the box method. Monthly Weather Review, 99, No. 10, pp. 725-738. Rosenthal, S. L. (1971a): A circularly symmetric, primitive equation model of tropical cyclones and its response to artificial enhance- ment of the convective heating functions. Monthly Weather Review, 99, No. 5, pp. 414-426. Rosenthal, S. L. (1971b): The response of a tropical cyclone model to variations in boundary layer parameters, initial conditions, lateral boundary conditions and domain size. Monthly Weather Review, 99, No. 10, pp. 767-777. Rosenthal, S. L., and M. S. Moss (1971a): Numerical experiments of rele- vance to Project STORMFURY. ERLTM-NHRL Technical Memorandum No. 95, (NOAA/NHRL, Miami, Florida) 52 pp. Rosenthal, S. L., and M. S. Moss (1971b): The response of a tropical cy- clone model to radical changes in data fields during the mature stage. ERLTM-NHRL Technical Memorandum No. 96 (NOAA/NHRL, Miami, Florida) 18 pp. Trout, J. W., and R. A. Anthes (1971): Horizontal asymmetries in a num- erical model Of a hurricane. ERLTM-NHRL Technical Memorandum No. 93 (NOAA/NHRL, Miami, Florida) 18 pp. Other papers cited in the text: Kasahara, A., and T. Asai (1967): Effects of an ensemble of convective elements on the large-scale motions of the atmosphere. Journal of the Meteorological Society of Japan, 45, No. 4, pp. 280-290. 359 Malkus, J. S. (1960): Recent developments in studies of penetrative con- vection and its application to hurricane cumulonimbus towers. Cumu- lus Dynamics (The Pergamon Press, New York) pp. 65-84. Rosenthal, S. L. (1970): A circularly symmetric primitive equation model of tropical cyclone development containing an explicit water vapor cycle. Monthly Weather Review, 98, No. 9, pp. 643-663. Rosenthal, S. L. (1971c): Hurricane modeling at the National Hurricane Research Laboratory (1970), Project STORMFURY Annual Report , 1970, U.S. Department of Navy and U.S. Department of Commerce, Appendix C. Simpson, J., and V. Wiggert (1969): Models of precipitating cumulus towers. Monthly Weather Review, 97, No. 7, pp. 471-489. 360 APPENDIX D THE MUTUAL INTERACTION OF HURRICANE GINGER AND THE UPPER MIXED LAYER OF THE OCEAN Peter G. Black and William D. Mai linger National Hurricane Research Laboratory ABSTRACT Criteria for the existence of intense sea-surface cooling beneath a hurricane due to upwelling and mixing are examined. Data from various hurricanes, including Hurricane Ginger, show that intense cooling occurs only when the ratio of the storm speed to the baroclinic wave speed at the thermocl ine is less than three. Airborne expendable bathythermograph data and airborne infrared thermometer data collected during traverses of Hurri- cane Ginger indicate that a surface cooling of k C occurred on one day when the storm was nearly stationary and of 2.5 C on another day when the storm was moving faster. On both days, these data indicated an elevated thermocl ine within the eye region, which was surmounted by relatively warm water and a thermocl ine nearly 100 ft deeper than the undisturbed value within the region of maximum winds. Decreased fluxes of latent and sensible heat, associ- ated with the more intense cooling, correlated reasonably well with a decreased storm intensity as determined from central pressure and maximum wind values. Total oceanic heat flux into Hurricane Ginger is compared with that into other storms and found to be considerably less. The need to be able to separate storm intensity changes due to hurricane induced cooler sea-surface temperatures from intensity changes due to artificial modification experiments by Project STORMFURY is emphasized. INTRODUCTION It has been known for some time that the sea-surface temperature structure plays an important role in the development, maintenance, and de- cay of tropical storms and hurricanes. Various authors (Palmen, 1948, 1956; Tisdale and Clapp, 1963; Gray, 1967; Carlson, 1971) have shown that a sea-surface temperature of at least 26.5°C is a necessary condition for 361 formation of tropical storms. Others (Fisher, 1958; Perl roth, 1962, 1967, 1969; Brand, 1971) have argued that variations in sea-surface temperatures ahead of a storm's path may lead to changes in the storm's intensity and direction of motion. Still others (Riehl and Malkus, 1961; Miller, 1962, 1964; Hawkins and Rubsam, 1968) have shown that a considerable fraction of the effective energy flux into the lower layers of the hurricane comes from the ocean, and that this fraction is critically dependent on the sea-surface temperatures beneath the storm. Various authors (Ooyama, 1969; Rosenthal, 1972; Sundquist, 1972), in their experiments with numerical models of tropical storms, have discov- ered that the above-mentioned dependency can be rather dramatic. They have found that when the average sea temperature under the storm area is decreased by as little as 2°C, a 50 percent decrease in the maximum winds can occur. The most likely reason for this is the strong influence that sea-surface temperature and moisture flux has upon the initial conditions of ascending parcels within convective clouds, which make up the "heat engine" of the hurricane. However, other researchers (Jordan and Frank, 1964; Taylor, 1966; Whi taker, 1967; Leipper, 1967; McFadden, 1967; Hazel worth, 1968; Landis and Leipper, 1968; Jensen, 1970; Volgenau, 1970; Molinari and Franceschini , 1972; Revesz, 1971) have indicated that the hurricane-ocean interaction problem is more complicated than it may at first seem. Several of these authors (see table 1), using shipboard sea-surface temperature (SST) data (in one case infrared SST data) and bathythermograph observations obtained well before and after the storm, have reported marked cooling of the sur- face water in the wake of some storms as well as an elevated thermocline along the storm track. Perhaps an equal number of studies have shown very little modification of the pre-existing oceanic conditions. There- fore, the problem is to find out when substantially cooler waters will be brought to the surface beneath a hurricane and when they will not. This problem has been approached theoretically and will be discussed in a later section; however, it has not been previously observed during a storm. In fact, note that virtually no detailed SST measurements have been obtained during observations of the hurricanes in any of the aforemen- tioned studies. Nearly all the SST data used in the flux studies men- tioned earlier were obtained well before or well after the storm and were assumed to apply to the area of the storm while it was in progress. Some of these studies had to assume a uniform SST under the whole storm area. For hurricane modification experiments carried out by Project STORM- FURY, this data deficiency is especially serious. If the dependency of storm intensity on sea-surface temperature structure is as crucial as these previous studies have indicated, it then becomes necessary to be able to separate storm intensity changes caused by sea-surface tempera- ture changes from those caused by seeding, especially since both factors affect scales of energy from the cumulus scale on up. 362 en "O CL ~CJ *X> "TJi — t/1 O c ■>- cud u_ >> c r-. fOQJ 03 C"> O 03 CD 0"> > Xr- zc «— uo <=E :e .— IE r— _i .— CM r— f— >— o LT) LD CM r— I— CM CM r— .— i— CM CM CM CM t^> LD CM CO CM KO f— KO t— CM CO .— CO O r**. > — r— o co en i — o o at ■r- -t-> CT> 03 03 +-> i— CM t0 • r— **- C OQ CD •— o^ +j o ID kO CT> ■a- -o UD CT> Ol'— . — ' — ■> ID t- CTt i — ^^ i— o ra CTt i- i — CTv I — •"O r— O •» « QJ 3 . — -. s (O +-> to +•» C E *U_ rtj 4- r— Q. r— D- aj s_ o "O i — i. 01 s- a; i— QJ Oi . i — 13 to 00 to tn S- CO, i— oo ■f- C3 >,-»-> >^4-> >M- i — CD 03 i/> i — •r- cn +-> 3 F 1— 13 CD CJ3 IS ^ CD <=C CTt r— *— » i cn o. cn cl C QJ C CD (/) -Q ro S- 1 3 +J e in o •r— r- T3 cn e 3. It is interest- ing that a rather abrubt decrease in the area and magnitude of upwelling is predicted by Geisler's theory, but for a U/C value of approximately 1.4. Undoubtedly, not all of the cooling observed in the wake of the hur- ricanes used in figure 3 is a result of upwelling alone, but is more likely a result of several other factors in addition to upwelling, such as mixing and evaporation. Nevertheless, it is interesting that a critical U/C ap- pears to exist in the real world and that it also is suggested in Geisler's work. Figure 3 (right) shows that for slow moving storms (U/C < 3), the intensity of upwelling (and surface cooling) is a nearly linear function of storm intensity, as indicated by the Froude number, F. As F increases, intensity of upwelling increases. An increase in F is correlated with an increase in Vmax or a decrease in Rmax (which are themselves generally inversely correlated). The right panel of figure 3 also shows that for fast moving storms (U/C < 3), there is still a linear correlation between the amount of sur- face cooling and storm intensity. However, the amount of cooling is much ^""^ * * i 1 1 1 1 1 1 ■ 1 1 i e I T 1 T - _ T • I I - 5 • ■ g« U/C<3 ^ ■ ^^^ <5 * ^^* 0 2 I / U/C >3 __-- °. _■ -— ---'" .—-**"" . 0 _._----',""' 1 J " 1 • 16 18 20 Figure 3. Maximum sea-surface temperature decrease (LSSTmax) as a func- tion of U/C (left) . Thin vertical line indicates maximum U/C ratio for which intense sea-surface cooling will occur. Right panel shows depen- dence of maximum sea-surface cooling upon Froude number for U/C > 3 (open circles and dashed line) and for U/C < 3 (solid circles and solid line) . 367 less and the slope of the line is less. This result is not unreasonable, since sea-surface cooling by evaporation is probably the dominant mecha- nism operating for fast moving storms, and according to the transfer for- mulas, cooling by evaporation is directly proportional to wind speed and the total time the wind acts on a patch of the ocean. Also, according to Jordon and Frank (1964), the magnitude of ASSTmax shown for the fast mov- ing storms is within a factor of 2 of that to be expected due to evapora- tion. A further comparison of the response of the real ocean due to actual storms is shown in figure 4, which shows temperature profiles along the tracks of Hurricanes Betsy and Hilda. The pre-storm, along-track SST pro- file for Hurricane Betsy (1965) is shown in figure 4 (left) by the dotted line (Taylor, 1966). The post-storm SST profile reported by Taylor from shipboard measurements taken 1 week after storm passage is shown by the 00/10 00/09 u r ' I c~ 6 5 ?\ \ 7 i -J \ I / i 1 4 3 i \\ i _ \f 2 BETSY ' — ... 8 SEPT. '65 — — 10-11 SEPT. '65 i _ — — 15 SEPT. '65 • -11-25 SEPT '65 1 1 -30 29 28 -27V°0 -26 -25 - 24 00A34 00/03 00/02 o^EL r- 30 T J I j i ^ j HILDA ——24-30 SEPT'64_ — 1-13 OCT. '64 600 400 200 D (N.M.) 600 29 28 27V°C> 400 200 D (N.M.) Figure 4. Sea-surface temperature and U/C (dashed) profiles along the track of Hurricanes Hilda and Betsy. Distance along the tracks begin- ning at an arbitrary point, is on the bottom and times/dates are on the top (short tick marks are 1200 GMT positions) . 368 dash-double-dot line. Also shown are SST profiles taken with an airborne infrared thermometer 1 day (dash-dot line) and 5 days (solid line) after storm passage (McFadden, 1967). Both curves apparently suffer from a calibration error or atmospheric attenuation, since they were obtained nearly coincident with SST observations taken by the oceanographic re- search ship Alaminos. These are shown by the dash-double-dot line. If a +2°C correction were made to the infrared temperatures, they would closely coincide with the Alaminos temperatures measured after Betsy passed. When this correction is made, both sets of data show that only slight cooling (^1°C) of the surface water occurred after storm passage despite the intensity of the storm (see table 1). This result agrees well with the predictions of Geisler's theory, since the U/C ratio (indicated by the dashed line) oscillates about a value of 7. No cooling by upwelling should occur in this case. The dash-dot line for Hilda indicates the along-track profile of SST obtained by Leipper (1967) just before the passage of the storm, the solid line indicates the profile about 10 days after passage; and the dashed line indicates the U/C ratio along the track. Note that the SST has dropped nearly 7°C from just before to just after storm passage and that U/C oscillates about a value of 2. This indicates that strong upwelling would be predicted by the theory in this case. Note from AH in table 1 that strong upwelling was indeed responsible for the cooling. Therefore, in view of these data it was concluded that Geisler's sug- gestion of two different responses of the ocean to a hurricane, depending upon the U/C ratio, is valid in the real world. However, when using this idea to anticipate cooling in the wake of a storm, the data indicated that a U/C of about 3 was the critical value, above which little cooling would occur, rather than a U/C of 1.4, as suggested by theory. Based on the daily analyses made from 21 September through 29 Septem- ber 1971, SST profiles along the track of Ginger were constructed and are shown in figure 5, together with the U/C ratio. The circles on the pro- files indicate the position of the storm center relative to the profile. Based on a knowledge of the U/C ratio (fig. 5) and the model predic- tions, one would expect little effect upon the SST structure in the wake of the storm until shortly after 1200 GMT on 23 September (500 n miles along the track). Here, U/C decreases from a value of about 4 to less than 1, oscillates between 1 and 2 until about 0000 GMT on 28 September and then increases to about 3. Just as the theory predicts, the SST pro- file shows little change until 24 September, when nearly a 4°C decrease in SST occurs. The decrease is somewhat less on 25 and 26 September (*v/2.5°C) as U/C increases. On 27 and early on 28 September, there is a larger de- crease surrounding the storm (<3°C), as U/C again decreases, but with warmer waters in the center. This detailed structure will be discussed later. Later on 28 and 29 September, as U/C increases, the SST decrease becomes smaller (^2.0°C). It is interesting that the most intense de- crease that occurred on 24 September and late on 26 September show per- sistent aftereffects, with yery little temperature rise in those areas on subsequent days. 369 HURRICANE GINGER C=3 KTS Figure 5. SST and U/C profiles 29° for Hurricane Ginger. The thin line through eaoh group of pro- files is the 26°C isotherm. 2T Labeling follows the same con- vention as figure 4. -26 2000 1800 1600 1400 1200 1000 800 600 400 200 DISTANCE (NAUTICAL MILES) This figure indicates that some effect upon the intensity of the storm might be anticipated on 24 September and late on 26 September because of upwelling and mixing, transporting colder water upward from beneath the sea surface. Before discussing influences on storm intensity, we present de- tailed measurements confirming the presence of upwelled cooler water at the surface on 27 September and to a lesser extent on 28 September. USE OF REMOTE SENSING TECHNIQUES TO INFER DETAILED OCEAN STRUCTURE BENEATH HURRICANE GINGER As mentioned earlier, detailed SST measurements were made on 5 days, bracketing the Hurricane Ginger seeding operations on 26 and 28 September 1971. The days from 26 to 29 September have been analyzed so far. Sea- surface temperature measurements were obtained by Navy WP-3 aircraft at 1,000 ft, except for 500 ft on 28 September. The observations were one 370 per minute along the flight track using the Barnes Engineering airborne infrared radiometer, model PRT-4A. This instrument measures radiation emitted from 8 to 14 urn. Atmospheric water vapor can cause attenuation of emitted radiation from the sea surface which introduces some error in the measurements in this wavelength region. Variations in relative humidity of about 50 percent can cause errors in measured SST values up to 1.5°C in hurricane conditions (Maul and Hansen, 1972). This calibration has not yet been applied but will be shortly. A rough calibration for the infrared SST measurements was determined from an average temperature difference for the whole flight between air- borne expendable bathythermograph (AXBT) surface observations and the in- frared SST observations at the AXBT observation time. On the days of no AXBT observations, average differences were computed between nearby ship observations of SST and the infrared observations of SST. For the 4 days analyzed, the average correction to the infrared SST's was about -2°C. The analyses for the above data plus available ship reports for 26 to 29 September are shown in figures 6 through 9. Figure 6 shows a pool of cold water centered on the track near the 1200 GMT storm position for Figure 6. SST analysis for the area near Hurricane Ginger on 26 Septem- ber 1971. The hurricane track is superimposed and every 12-hour posi- tion shown by an open circle. The open box indicates the storm's posi- tion during the airborne SST measurements . The dashed lines indicate data sparse areas. 371 Figure 7 . SST analysis for the area near Hurricane Ginger on 27 Septem- ber 1971. The hurricane track is superimposed and every 12-hour posi- tion shown by an open circle. The open box indicates the storm's posi- tion during the airborne SST measurements. The dashed lines indicate data sparse areas. 24 September, the region where the storm slowed and cooler water was brought to the surface. Only slight cooling is indicated near the storm position on 25 and 26 September. The data on 26 September were mostly peripheral, and hence the dashed isotherms were used to indicate sparse data near the center on this day. The warm region (>27°C) north of the storm, which persists in the following figures, might be an area of warm water convergence and downwelling caused by warmer surface water being transported away from the storm region. Figure 7 shows the persistence of the cold pool near the storm's 24 September position and a new region of cooler water surrounding the storm center on this day. Of particular interest is the presence of a warm region within the eye of the storm. It is speculated that existence of an exceptionally large eye (nearly 125 n miles in diameter) allowed time for solar radiation to warm the region within the eye, since a given point on the ocean within the eye would be nearly cloud free for almost 24 hours. There are undoubtedly other possible explanations for this observation, but more research is needed before they can be clearly formulated. The observation, however, is well documented, as will be shown later, and is 372 persistent in each of the analyses shown in figures 6 to 9, most promi- nently in figures 7 and 8. The warm tongue to the north and northwest of the storm, mentioned earlier, is again present. Figure 8 shows the SST analysis for 28 September, which shows that the cold pool has now drifted westward slightly and become even cooler, probably because it has continued to be under the influence of Ginger's large circulation and further cooling by evaporation has resulted. The cooling observed on 27 September around the center of Ginger does not appear to have persisted. This could possibly be because the sparse data in that area on 28 September caused it not to be detected, or that some kind of gross measurement error occurred on 27 September. The AXBT data, to be discussed shortly, suggest that the former is the case. Some cooling (^2.5°C) was evident beneath the maximum wind region on 28 September, as was the case on 27 September. The warm eye region was also evident again. A somewhat more well-defined warm region surrounded the periphery of the cooler waters to the west, northwest, and north of the storm center. HURRCANE GINGER SEPT 28,1971 2000Z SEA SURFACE TEMPERATURE Figure 8. SST analysis for the area near Hurricane Ginger on 28 Septem- ber 1971. The hurricane track is superimposed and every 12-hour posi- tion shown by an open circle. The open box indicates the storm's posi- tion during the airborne SST measurements. The dashed lines indicate data sparse areas. 373 Figure 9. SST analysis for the area near Hurricane Ginger on 29 Septem- ber 1971. The hurrioane track is superimposed and every 12-hour posi- tion shown by an open circle. The open box indicates the storm's posi- tion during the airborne SST measurements. The dashed lines indicate data sparse areas. The analysis for 29 September, in a region displaced to the northwest of the previous analysis areas, shows a region cooled by about 2°C along the storm track in the wake of the storm. Again a region of warmer SST values is evident within the eye of the storm. Note also the presence of the Gulf Stream now just ahead of the storm, as well as a cold eddy between the storm center and the Gulf Stream. The effect of the storm on the ocean SST structure was markedly less on 29 September than on 28 or 27 September; this was due primarily to its faster forward motion. The AXBT data acquired within Hurricane Ginger are perhaps the most unique aspect of this study. To the authors' knowledge no such data have ever been gathered before. These data have supported a much higher degree of confidence for the SST analyses just discussed than would otherwise have been justified. Two days of AXBT data, comprising about 40 observa- tions on 27 and 28 September, have been analyzed so far. The data for 30 September, when Hurricane Ginger was over the Gulf Stream, have not yet been analyzed. 374 Four temperature vs. depth cross sections were constructed through and ahead of the storm on 27 September. Sections A-A' and D-D' are shown in figures 10 and 11 (see fig. 7 for exact location). Section D-D' is considered as representative of the undistrubed state, since it is nearly 200 n miles ahead of the storm center in a region of approximately 30 knot winds. This section shows an undisturbed mixed layer depth of about 100 ft. Section A-A' shows the changes in the subsurface structure brought about by the storm on 27 September when it had become almost stationary. Note that within 50 n miles of the storm center (the approximate radius of maximum winds), the thermocline bulged upward (relative to the undisturbed thermocline shown in fig. 11) by about 15 ft, indicating upwelling in this region. Upwelling is also indicated by the upward bulging and vertical displacement of sub-thermocline isotherms. However, the cooler subsurface water had not reached the surface in this region and instead was surmounted by relatively warm water. Only in the region 50 to 70 n miles from the center, especially in the right (NE) quadrant (the region of highest winds), did cooler subsurface water reach the surface, with a gradual warming rad- ially outward from this region. However, the depth of the mixed layer in this region became much deeper than that of the undisturbed state, and the thermocline became more diffuse and spread out. The mixed layer depth in- creased by about 100 ft, and thermocline gradient decreased from 20 C/100 ft to 5°C/100 ft. Figure 8 shows the three cross sections constructed through Hurricane Ginger on 28 September. Section F-F1 (fig. 12) shows that on 28 September very little upwelling was detected. The thermocline beneath the storm center was about the same depth as before the storm traversed the area. The mixing process appears to be less efficient on this day also, with only about a 50 ft decrease in the thermocline occurring beneath the coldest surface waters at the radius of maximum winds (^50 n miles). Note that, as on 27 September, warmer waters (perhaps due to surface conver- gence) were present at the storm's periphery (the outer radius of 50 knot winds). Of special interest is the extremely deep thermocline on the right side of figure 12 (140 n miles to the east-northeast of the storm center). Here the thermocline is nearly 150 ft deeper than in the undisturbed water, but with a fairly tight gradient of about 10°C per 100 ft. This is part of the region associated with the intense cooling of the sea surface that occurred on 24 September. This is the only part of this particular region for which AXBT data were obtained. Until more AXBT data are obtained, it is difficult to see whether the aftereffects of a hurricane will be a per- sistent trough in the thermocline produced by mixing, or a persistent ridge produced by upwelling. Perhaps both could exist: the trough to the right and left of the storm track and the ridge along the storm track. Additional experiments are needed to resolve this. 375 IOOQLtt r1 100 80 60 40 20 0 20 40 60 80 100 120 RADIAL DISTANCE Figure 10. Temperature-depth cross section beneath Hurricane Ginger along section A-A'. The thick vertical line is the center position of the hur- ricane and the thin vertical lines are the position of the AXBT's pro- jected onto the section. 376 80 60 40 20 0 RADIAL DISTANCE (NM> Figure 11. Temperature-depth cross section beneath Hurricane Ginger along section D-D'* the undisturbed cross section. The thick vertical line is the future hurricane track position. The thin vertical lines are the position of the AXBT's projected onto this section. 377 F Or- 100 - 200 300 400 ±500 x h- Q. uJ600 Q 700 800 900 1000 HURRICJANE GINGER SEPT 23 1971 120 100 80 60 40 20 0 20 40 Radial distance cnm> 60 80 100 120 140 Figure 12. Temperature -depth cross section beneath Hurricane Ginger along section F-F'. The thick vertical line is the center position of the hurricane and the thin vertical lines are the position of the AXBT's projected onto the section. RESPONSE OF HURRICANE GINGER TO CHANGES IN OCEANIC SST STRUCTURE The question that may now be asked is: What changes in intensity, if any, were brought about by the cooling of the surface water beneath and behind Hurricane Ginger? The answer is by no means clear-cut. However, to estimate possible effects, fluxes of latent and sensible heat from the ocean to the storm were calculated using the usual transfer formulas: Fs = >ACpCDV(tS -T) and CD = (1.1 + 0.04V) x 10"3 378 where Fs is the sensible heat flux, Fl is the latent heat flux, Cp is the specific heat of water at constant pressure, L is the latent heat of evap- oration, Ts is the sea temperature, T is the air temperature, qs is the specific humidity at Ts, q is the specific humidity at the dew point temp- erature, V is the wind speed in m/sec, Cp is the drag coefficient, and p^ is the average surface air density. The exchange coefficient is assumed to be the same for momentum, heat, and moisture. The surface air temperatures used in the above calculation were extra- polated from air temperatures measured by the low flying aircraft at 1,000 ft (500 ft on 28 September) using a lapse rate of 2.5°C per 1,000 ft. The same calculation was applied to the measured dew point temperature, from which q was calculated. The patterns of latent and sensible heat were extremely asymmetrical. Large fluxes were observed consistently ahead and just to the left of the storm's direction of motion on the 4 days studied, 26 to 29 September. The least fluxes were observed in the wake of the storm. In fact, the 24°C isotherm outlined the region of zero flux from the ocean to the air. Within these regions, the air was actually being cooled by the water. Bowen ratios (ratio of sensible to latent heat) of about 25 percent were calculated over most of the area of 50 knot winds. Small regions of anom- alously large Bowen ratios (^100 percent) were present and associated with weak fluxes (especially regions where F|_+0). Figure 13 (top) shows the total vertical fluxes (neglecting radia- tion) integrated over the area covered by each quadrant of the storm, as well as the total integrated vertical flux into the whole storm (within 200 n miles of the center) for the days studied so far. Note that the front quadrant (45° either side of the storm's direction of motion) re- ceived the least. The total storm heat input rate decreased during the day on 26 September and early on 27 September, increased during the day on 27 September and early on 28 September, then decreased late on 28 Sep- tember and during 29 September. Figure 13 (bottom) shows the variation of maximum wind and central pressure during the time studied. Between 1200 GMT on 26 and 27 Septem- ber, the maximum wind increased about 10 knots and the central pressure decreased about 5 mb while the total storm heat input rate decreased. Between 1200 GMT on 27 and 28 September, the maximum wind remained about the same and the central pressure rose about 5 mb while the total storm heat input rate increased. And, finally, between 1200 GMT on 28 and 29 September, the maximum wind increased by about 10 knots and the central pressure rose by another 5 mb, while the total storm heat input rate de- creased slightly. A direct comparison between storm intensity and total storm heat input rate is somewhat questionable in this case. The problem is twofold. First, even though sea temperature within a hurricane was measured better than ever before, this was done only once eyery 24 hours. For a more meaningful comparison with storm intensity, SST measurements and heat input rate calculations should be obtained at least once every 12 hours. Second, in a storm with such a large eye, it may have been diffi- cult to obtain the exact location of the minimum pressure and the maximum 379 1j6 O "H 14 12 - 1 1 r HURRICANE r GINGER § i.0 _ -1 u. K - < UJ 1 8 _l < P o K t- 6 z < tc a ■ < z> O .4 1 2 SEPT 25 TOTAL HEAT FLUX b» Quadrant within 200nm of c«nt«r TOTAL STORM HEAT FLUX within 200nm of c«nter T HURRICANE GINGER T 1 I I I SEPTEMBER 25-29, 1971 100 &- S60 50 12 Z 12 Z 12Z 122 12Z SEPT. 25 SEPT 26 SEPT. 27 SEPT. 28 SEPT. 29 Figure 13. Total heat input rates vs. time within 200 n miles of Hurri- cane Ginger is shown by the thick solid line in the top panel as well as total heat by quadrant. The profiles of maximum wind and central pressure with the time are shown in the bottom panel. 380 wind (there were several wind maxima evident in the isotach analyses (not shown). Thus, exact determination of storm intensity may have also been a problem. A plot of the radial distribution of radiation) for the 4 days studied is shown flux calculations for other storms. Radia radius of maximum wind. One can see that the effect of the cold water near the radi heat flux low in this region. However, th region early on 28 September (curve 9), as ward motion, contributed to a larger heat ber, the maximum heat flux per unit area i of maximum winds, which is coincident with the total heat flux (neglecting in figure 14 together with heat 1 distance is normalized to the early on 27 September (curve 8), us of maximum winds kept the e warmer temperatures in this the storm began its faster for- flux. Note that on 29 Septem- s located at twice the radius the position of the Gulf Stream. 2000 < 400 UJ x R/R, Figure 14. Profiles of heat flux from the sea to the air for Hurricane Ginger and other storms. Curves (1) and (5) are from the data of Riehl and Malkus (1961) 3 curve (3) is from Miller (1962) > curves (4) and (6) are from Miller (1964) t and curve (2) is from Hawkins and Rubsam (1968), 381 The heat flux values for Ginger are considerably smaller than those for other storms, especially in the region of maximum winds. The Ginger calculations show that the maximum heat flux per unit area is generally some distance radially outward from the radius of maximum winds. It is believed, therefore, that the other heat flux values shown in figure 14 may be too high, especially for the slower moving storms (i.e., Hilda), because cooling of the sea surface by upwelling and mixing was not con- sidered in the calculation. CONCLUSIONS Use of airborne infrared thermometers and airborne expendable bathy- thermographs has made it possible to document the presence of cold water beneath Hurricane Ginger. The data have indicated that this cooling is more evident when the storm was nearly stationary than when it was moving at a moderate speed. Furthermore, the data have suggested that the causes of this cooling may be attributed mainly to mixing and secondly to up- welling. It was more difficult to document the changes in storm intensity brought about by cooling of the sea surface, but preliminary indications suggest that intensification is associated with increased heat flux from the sea and weakening is associated with decreased heat flux from the sea. This result must be considered when evaluating effects of artificial modification of tropical storms and hurricanes. Continued gathering of SST and AXBT data is therefore considered essential to Project STORMFURY. ACKNOWLEDGEMENTS The authors thank Weather Reconnaissance Squadron Four, USN, for collecting and making available, the data used in this study. Special thanks is due to the WP-3 crew for collecting the AXBT data, especially Lieutenant Commanders Clark, aircraft commander, and Hennessey, flight meteorologist. Miss Rita Sherrill and Miss Lorraine Kelly are greatfully acknowledged for plotting and tabulating the data. REFERENCES Brand, S. (1971): The effects on a tropical cyclone of cooler surface waters due to upwelling and mixing produced by a prior tropical cy- clone, Journal of Applied Meteorology , 10, pp. 865-874. Carlson, T. N. (1971): An apparent relationship between the sea-surface temperature of the tropical Atlantic and the development of African disturbances into tropical storms, Monthly Weather Review, 99, No. 4, pp. 309-310. Fisher, E. L. (1958): Hurricanes and the sea-surface temperature field, Journal of Meteorology* 15, pp. 328-333. 382 Geisler, J. E. (1970): Linear theory of the response of a two layer ocean to a moving hurricane, Journal of Geophysical Fluid Dynamics, 1, Nos. 1 and 2, pp. 248-272. Gray, W. M. (1967): Global view of the origin of tropical disturbances and storms, Colorado State University Atmospheric Science Paper No. 114 (Colorado State University, Ft. Collins, Colo.), 105 pp. Hawkins, H. F., and D. T. Rubsam (1968): Hurricane Hilda, 1964: II. Structure and budgets of the hurricane on October 1, 1964, Monthly Weather Review, 96, No. 9, pp. 617-636. Hazelworth, J. B. (1968): Water temperature variations resulting from hurricanes, Journal of Geophysical Research, 73, No. 16, pp. 5105- 5123. Jensen, J. J. (1970): Calculated and observed changes in sea-surface temperature associated with hurricane passage [U.S. Naval Postgrad- uate School, Masters Thesis) September 1970, 55 pp. Jordan, C. L., and N. L. Frank (1964): On the influence of tropical cy- clones on the sea-surface temperature field [Florida State Univer- sity, Technical Department) NSF Grant GP-621, 31 pp. Landis, Robert C. (1966): Synoptic analysis of near surface and subsur- face temperatures in the Atlantic Ocean following Hurricane Betsy (Texas A&M Univ., Dept. of Oceanography) 37 pp. Landis, Robert C, and D. F. Leipper (1968): Effects of Hurricane Betsy upon Atlantic ocean temperature, based on radio-transmitted data, Journal of Applied Meteorology , 7, pp. 554-562. Leipper, D. F. (1967): Observed ocean conditions and Hurricane Hilda, 1964, Journal of Atmospheric Sciences, 2A, pp. 182-196. McFadden, J. D. (1967): Sea-surface temperatures in the wake of Hurricane Betsy (1965), Monthly Weather Review, 95, No. 5, pp. 299-302. Maul, G. A., and D. V. Hansen (1972): An observation of the Gulf Stream surface front structure by ship, aircraft and satellite, Remote Sens- ing of Environment, 2, pp. 109-116. Miller, B. I. (1962): On the Momentum and Energy Balance of Hurricane Helene (1958), National Hurricane Research Project Report No. 53 (U.S. Weather Bureau, Washington, D.C.), 19 pp. Miller, B. I. (1964): A study of the filling of Hurricane Donna (1960) over land, Monthly Weather Review, 92, No. 9, pp. 389-406. Molinari, R. L., and G. 0. Franceschini (1972): Bathythermograph sections across the path of Hurricane Celia {Texas A and M, unpublished report) 4 pp. O'Brien, J. J. (1967): The non-linear response of a two-layer baroclinic ocean to a stationary, axially symmetric hurricane: Part II. Up- welling and mixing by momentum transfer, Journal of Atmospheric Sciences, 24, pp. 208-215. 383 O'Brien, J. J. (1969): The response of the ocean to a slowly-moving cy- clone, Annalen der Meteorologie , 4, pp. 60-66. O'Brien, J. J., and R. 0. Reid (1967): The non-linear response of a two- layer baroclinic ocean to a stationary, axially symmetric hurricane: Part I. Upwelling induced by momentum transfer, Journal of Atmos- pheric Sciences, 24, pp. 197-207. Ooyama, K. (1969): Numerical simulation of the life cycle of tropical cyclones, Journal of Atmospheric Sciences, 26, pp. 3-40. Palmen, E. (1948): On the formation and structure of tropical cyclones, Geophysica, 3, pp. 26-28. Palmen, E. (1956): A review of knowledge on the formation and development of tropical cyclones, Proc. of the Tropical Cyclone Symposium, Bris- bane, December 1956, pp. 213-231, (Issued by Director of Meteorology, Melbourne, December 1956.) Perlroth, I. (1962): Relationship of central pressure of Hurricane Esther (1961) and the sea-surface temperature field, Tellus, 14, pp. 403-408. Perlroth, I. (1967): Hurricane behavior as related to oceanographic en- vironmental conditions, Tellus, 19, pp. 258-268. Perlroth, I. (1969): Effects of oceanographic media on equatorial Atlantic hurricanes, Tellus, 21, pp. 230-244. Revesz, W. (1971): Comparison of effects of various tropical storms on the vertical temperature structure of the ocean using pictorial rep- resentation (U.S. Naval Postgraduate School, Masters Thesis), Sep- tember 1971, 52 pp. Riehl, H., and J. Malkus (1961): Some Aspects of Hurricane Daisy, 1958, Tellus, 13, No. 2, pp. 181-213. Rosenthal, S. L. (1972): The response of a tropical cyclone model to var- iations in boundary layer parameters, initial conditions, lateral boundary conditions, and domain size, Monthly Weather Review (to be publ ished) . Stevenson, R. E., and R. S. Armstrong (1965): Heat loss from the waters of the northwest Gulf of Mexico during Hurricane Carla, Geofis. In- tern. , 5, pp. 49-57. Sundquist, H. (1972): Model tropical cyclone behavior in experiments re- lated to modification attempts, Tellus, 24, No. 1, pp. 6-12. Taylor, J. G. (1966): An approach to the analysis of sea-surface temper- ature data for utilization in hurricane forecasting in the Gulf of Mexico {Texas A and M) May 1966, 107 pp. Tisdale, C. F., and P. F. Clapp (1963): Origin and paths of hurricanes and tropical storms related to certain physical parameters at the air-sea interface, Journal of Applied Meteorology , 2, pp. 358-367. Volgenau, D. (1970): Hurricane heat potential of the Gulf of Mexico (U.S. Naval Postgraduate School), 58 pp. Whitaker, W. D. (1967): Quantitative determination of heat transfer from sea to air during passage of Hurricane Betsy, Masters Thesis (Texas A and M University, College Station, Texas). 384 APPENDIX E MODELING THE SEEDING EFFECT IN HURRICANE GINGER W. D. Scott and C. K. Dossett National Hurricane Research Laboratory INTRODUCTION Pyrotechnic silver iodide flares were dropped into active updrafts within chosen sectors of Hurricane Ginger on 26 and 28 September 1971. At approximately the same time, several dropsondes were released by the Air Force from about 30,000 ft, as assorted Navy, Air Force, and NOAA aircraft monitored the storm. This paper attempts to assess the changes in cloud structure that might be expected from the seeding through an analysis of the indicated temperature and humidity structure of the storm. THE CLOUD MODEL The cloud model used is the one-dimensional Lagrangian model of Cotton (1970), called the PSU-71 model. The model is run with a constant radius, R, an entrainment rate of 0.18/R, and a virtual mass coefficient of 0.5. The rain fallout scheme is similar to that of the Experimental Meteorology Laboratory (NOAA) and considers all drops larger than 100y as rainwater with a water content weighted mean terminal velocity V. The proportion of rainwater that falls from the circulating vortex between height steps az is determined by the ratio AzV/RW, where W is the ascent rate of the vor- tex center. Mixing of environmental air into the warm portion of the cloud is accompanied by evaporation of sufficient cloud droplets to bring the air to water saturation and, similarly, mixing in the glaciated cloud, by suf- ficient evaporation to maintain ice saturation. If the cloud's liquid water content is less than 0.1 gm/m3, in the portion of the cloud with temperatures below 0°C, it is considered to be glaciated. The dynamic equation follows that of Squires and Turner (1962), based on an approximate integration of the force balance equation for the accel- eration in terms of the buoyancy and water loading. The cloud is initial- ized by a perturbation updraft of 1 m/sec with no temperature perturbation. The model considers raindrop freezing due to containment of active freezing nuclei within the drops and raindrop freezing due to the collec- tion of cloud ice. Cloud droplets grow and become raindrops by vapor con- densation, auto-conversion, and accretion. Ice crystals grow by vapor 385 deposition and riming. Phenomena involving the ice phase are considered as a function of crystal habit, using 21 distinct categories of ice crys- tals depending on the temperature. ICE CRYSTAL NUCLEATION The ice crystal nucleation is prescribed according to the experimental curve of Fletcher (1962), the effect of ice multiplication (Hobbs, 1969), or the expected spectrum of the pyrotechnic nucleant, LW-83 (see Elliott et al., 1969). The cumulative spectrum of these quantities is shown in figure 1. Note that the effect of ice multiplication introduces an un- certainty of 103 to 10^ in the number of crystals that form at the warmer due to assorted effects, including cyrstal or simply our ignorance regarding the phenome- cloud. The seeded concentrations suffer a sim- ilar discrepancy in that it is not known if ice multiplication is present during the artificial seeding. Also, airflow past the falling projectile alters the aerosol production and may enhance the nucleation efficiency. The effect of coagulation in the nucleation tests could make the experi- mental results low as well. These latter two effects alone can enhance the nucleation efficiency by two orders of magnitude (see Elliott et al., 1969). temperatures. This may be breakup, drop splintering, non of nucleation within a WARM CLOUD DEVELOPMENT The warm cloud development follows the procedure of Cotton (1970), except that specific features are incorporated that are appropriate for hurricane clouds (appendix K, presents some of the expected rainwater characteristics). In the development, it is assumed that the composite of drop coalescence and drop breakup produces a Marshall -Palmer distribu- tion in the raindrop sizes. The conversion rate of cloud water to rainwater lOOOO 1000 100 10 1 .1 .ou .001 LW-83 (See Elliott et 01,1969) ./ •xp .6T / (See Fletcher, 1962) "^4 -8 «12 -16 -20 -24 -28 °C Figure 1. The cumulative ice nuc- lei spectra. 386 is then assumed to be approximated by the difference between the stochas- tic conversion rate formulated by Berry (1965) and the Kessler (1969) ac- cretion rate, determined by the Marshall -Palmer characteristics of the rainfall. For the hurricane clouds, we assume that 100 droplets per cen- timeter3 are formed at cloud base with a radius dispersion of 0.28. The output of the numerical calculation (CDC-6600) is shown in figure 2. T — " — I — " — r 2.5 gm /m CLOUD WATER i — i — i — i — i — i — i — i — i — i — i — r n0 = 10* N0 =±Or DOTTED LINES INDICATE VALUES CALCULATED FROM PARAMETERIZA- TION (Eqns. 1,2 , 3) 100 DROPLETS / CM3 -y = RADIUS DESPERSION = 0.28 0.8 gm/m* 0 200 400 600 800 1000 1200 1400 1600 1800 2000 2200 Figure 2. Calculated autooonversion rates, For the Marshall -Palmer N0 = 105 m"\ the curves are approximated by the parametric form, dM It L J auto (m-0.27)2 440 exp -2.3 Mooo 1060 - 100 m 1.7 |+1.5- (1) where M and m are, respectively, the rainwater and cloudwater contents in gm/m3, and t is the age of the cloud parcel in seconds. In the calculation, the difference term containing m is not allowed to be negative. Similarly, for N0 = 106 m_l+, [dM] Hau to for N0 = 107 m"^ (m-0.32)2 480 exp ■2.3 54000 1050 - 100 m 1.7 m + 1.5' (2) dM dt L J auto (m-0.45)2 530 exp -2.3 52000 1030 m 1.7 - 100 £♦1.5- (3) The curves are derived using a nonlinear least squares criteria based on the relative magnitudes of the deviation. The parameterizations are shown as dotted lines in figure 2. Note that the fit is best at the shorter times and larger liquid water contents. THE SOUNDINGS The environmental soundings were composited from three sources: air- craft measurements, dropsondes, and the mean hurricane soundings of Sheets .(1969). On 28 September, a rough horizontal analysis of the aircraft data taken in Hurricane Ginger was completed and, based on the results of the analysis, soundings appropriate to the proposed seeding areas were pre- pared. They spanned the areas between the dashed lines on the tephi grams (see fig. 3). On 26 September, seven soundings were made with dropsondes (see table 1). Four of these soundings showed adequate detail (see fig. 4). The specific locations of the dropsondes are shown in figure 5. Note that the radar PPI echoes shown on figure 5 are for a mean time and do not pre- cisely represent the storm structure when the dropsondes were dropped or the seedings took place. A comparison of temperature data from dropsonde No. 6 was made with the temperatures measured by the N0AA aircraft. At 850 mb (DC-6), the temperatures agreed to within 1°C; at 200 mb (B-57), the extrapolated temperature was about 2°c warmer than that indicated by the aircraft; at 500 mb (C-130) the aircraft temperatures were about 1.5°C warmer than the indicated dropsonde temperature. These comparisons lend 388 -60 -50 -40 -30 -20 -10 10 20 -60 -50 -40 -30 -20 -10 10 20 Figure 3. Tephigrams of Hurricane Ginger on 28 September 1971. -60 -50 -40 -30 -20 -10 10 20 credence to the quality of the dropsonde data and further substantiate the existence of the stable layer around 400 mb; a layer of importance for mod- ification. The upper level sounding points were obtained by selecting the mean hurricane sounding with characteristics similar to those observed at the lower levels; this turned out to be the mean soundings for central Table 1. Dropsonde Locations and Times for September 1971 Repor 'ted Flight Latitude Longitude Day Time No. No. No. (°N) (°w) (GMT) I1 II2 7109261 27.5 69. Z 26 0943 2 12 7109261 27.2 70.5 26 1013 3 14 7109261 27.8 67.8 26 1101 4 15 7109261 29.2 69.8 26 1131 5 16 7109261 28.1 69.4 26 1223 6 Jl 710926J 27.6 71.2 26 2155 7 J2 710926J 27.5 69.2 26 2220 8 J3 710928J 29.2 70.8 28 1020 dropsonde number as shown on figure 5 and in the text. 2Number of dropsonde as used by aircraft. 389 GINGER SOUNDING #2 FROM DROPSOND x SHEETS (1969) (IOOO- 1004MB) -60 -50 -40 -30 -20 -10 10 20 GINGER SOUNDING #6 FROM DROPSON x SHEETS (1969) (1000- 1004MB ) —I I I I -60 -50 -40 -30 -20 -10 0 10 20 -60 -50 -40 -30 -20 -10 0 10 20 Figure 4. Tephigrams of Hurricane Ginger on 26 September 1971. pressures between 1000 and 1004 mb. The lower level air temperatures and humidities indicated by the dropsondes were at variance with the surface air temperatures indicated by the low flying aircraft and the ship data. Temperatures indicated by the dropsondes were as much as 3° or 4°C low, humidities, 20 percent low. Hence, a mean humidity of 90 percent, a temp- erature of 23.5 C at a pressure of 950 mb were used to characterize the environment at cloud base; these values were based on data taken by the low level WC-121. However, the conditions at cloud base were found to be so important that in the final tabulation they were selectively changed to check the overall sensitivity of the soundings in terms of seedability. The proposed seeding areas (I, II, and III) relative to the storm center are shown in figure 6. The radar data of figure 6 were acquired approximately 5 hours before the seedings and should be used only for qualitative judgments about the presence of clouds. However, a composite of radar data (APN-59, 3 cm) from the C-130 did corroborate in a general way the presence of the features shown in figure 6. The locations of the actual seeding areas relative to the storm center for the 26th and 28th are shown in appendix B, figure 16. Note that they generally coincide with 390 // iOON.M. • DROPSONDE LOCATIONS NAVY WC-121 NO. 892 APS-20 RADAR 1840Z RANGE MARKS ARE RELATIVE TO THE AIRCRAFT POSITION Figure 5. Radar pictures of Hurricane Ginger 26 September 1971 391 H • DROPSONDE LOCATION C> PROPOSED SEEDING SECTOR NAVY WC-121 NO. 892 APS-20 RADAR 12172 RANGE MARKS ARE RELATIVE TO THE AIRCRAFT POSITION FIR Figure 6. Radar pictures of Hurrioane Ginger 28 September 1971, 392 preliminary locations I through III shown in figure 6, and that the drop- sondes closest to the seeding areas on 26 September were numbers 1, 3, and 6. SENSITIVITY OF THE MODEL Results of small perturbations in the cloudy environment are shown in figures 7, 8, and 9. These are deviations that could easily arise due to errors in the measurements or errors in our knowledge regarding the micro- physical development of the clouds. Figure 7 shows the effect of warm SOUNDING #1 GINGER ON 26th 10,000 9,000 8,000 £ 7,000 LJ u 6,000 2 ~ 5,000 x 4,000 ^ 3,000 2,000 h 1,000 -, I I I l I i I > I i I * I < CONTINENTAL CLOUD 300 DROPLETS/CM3 0 25 DISP. Nn = 107 M"4 - S = 400m BASE ■ ' I LEGEND SAME AS FIGURE 8. I i I I I I I I I I I I I I I HURRICANE TROPICAL MARITIME CLOUD 100 DROPLETS/CM3 0.28 DISP S=600m LEGEND SAME AS FIGURE 8. L_L 8 10 12 14 0 2 4 6 METERS/SEC or 6M/KG 8 10 12 14 Figure 7 . The effect of warm aloud microphysios, cloud development (see Cotton 1970, Eq. 1). We see that the seedability1 of the cloud with continental warm cloud microphysics is only slightly less than the seedability of the cloud with the parameterization of (1) Seedability is defined here as the difference between the calculated depth of the seeded cloud and the natural cloud. The seeding refers to the stan- dard seeding curve (fig. 1) and assumed one flare per cloud and a burning rate that produces twice the nuclei at the lower levels than at the upper levels. Seeding is from the 30,000 ft down to 15,000 ft, and, unless otherwise noted, the cloud radius was 1.5 km. 393 I I I I I I I I I I I I I I I I I I ' I ' I I I I SOUNDING # 1 GINGER ON 26' CRYSTAL DEPOSITIONAL GROWTH ENHANCED 5 TIMES i i 1 FREEHING LEVEL .' / UPDRAFT / 120 knots (Brand, 1970b). The number of occurrences of yery intense typhoons almost doubles from August to September. Maximum surface wind is used as a measure of intensity. The possibility of using sea level pressure (SLP) as a measure of intensity (or converting SLP to wind) was considered but rejected because the SLP historical data were incomplete. 404 CO o •cr CNI — CO I— LU O .Z -XL < — ' X o < 13 < a: > MAY JUN JUL AUG SEP OCT NOV DEC (26) (38) (86) (122) (123X89) (68) (31) Figure 2. Monthly distribution of the average maximum 24-hour change in intensity of intensifying tropical cyclones that3 during their lifetime 3 have reached tropical storm or typhoon intensity (1945 to' 1969). The number of tropical cyclones on which the averages were based is in parentheses below each month (Brand, 1972a). The monthly values are given progression throu of total typhoon son extends from to total typhoon through August, creases significa ratio of intense to total typhoon occurrences (percent at the top of the graph in fig. 3) shows an interesting ghout the year; for example (neglecting the small number occurrences in February and considering the typhoon sea- March through the following January), the ratio of intense occurrences is a nearly constant 15 percent from March However, from September through January, the ratio in- ntly except for a pronounced dip in October. 5 4 31 < 3 14 13 RATIO (%) 16 17 15 14 20 14 24 23 1 j I i J 3 1 — , p yS - I i I 2 j "-" [ ~" i i i ™ - "t - 1 O JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC JAN Figure 3. Monthly frequency distribution of very intense typhoon occur- rences (>120 knots) from 1945 to 1968. Data are based on 6-hourly re- ports or observations. The ratio (percent) of very intense to total typhoon occurrences for each month is the number at the top of the graph (Br and 3 1970b). 405 Although September has the greatest total number of typhoon occur- rences as well as the most occurrences of intense typhoons, it does not have the largest size typhoons (Brand, 1970b). The distribution of average monthly size of typhoons (from 1945 to 1968) shows a gradual in- crease in size from June through October, with a decrease occurring in November and December (fig. 4). To determine the monthly size variation of typhoons, the monthly frequency distribution of very large typhoon intensity occurrences was plotted (fig. 5). \lery large typhoons are defined as those tropical cy- clones of typhoon intensity having a mean radius to the outer closed sur- face isobar equal to or greater than 10° of latitude. Also presented in figure 5 (in parentheses below each month) is the number of individual typhoons associated with these 6-hourly occurrences; very large typhoons are quite prevalent in October. The monthly ratio of the number of large to total typhoon occurrences (percent values are shown at the top of the graph in fig. 5) indicates a distinct prevalence of very large typhoons from August through November, with an extreme peak in October. This October maximum in the percentage of s/ery large typhoon occurrences contrasts with the dip during October in the percentage of intense typhoons. Similarly, the monthly frequency distribution of small typhoons (fig. 6) was examined. Very small typhoons are defined as those tropical cy- clones of typhoon intensity having a mean radius to the outer closed sur- face isobar equal to or less than 2° of latitude. A definite peak for very small typhoons exists in August. On an individual monthly basis, July has almost as many very small typhoons (24 versus 27) as August, but uj 7 1 !' ! j - 1— - _ j ; to 6 5 * ^.^ H < \ , ! r* - i \ 4 / i \ /1 1 \~—\ — — i i i +■- 3 o 1 I Li jlj j/ IN Fl iB Mi &R Al 3R Mi XY Jl JN Jl JL Al JG SE :p o< :t NC )V OE ■c Figure 4. Distribution of average monthly size of typhoons for typhoon occurrences from 1945 to 1968. The measure of typhoon size is the mean radius to the outer closed surface isobar in degrees of latitude. Av- erage monthly size is the sum of the size of each typhoon occurrence divided by the number of typhoon occurrences for a given month (Brandy 1970b). 406 100 0 0 RATIO (%) 0 .5 .3 0 0 1 .7 1.4 7.2 1.6 0 ' i ! I 1 I ! ' ■ | | PHOON OCCURRENCES > W ^ U 0) N CD U > o o o o o o c i ! i- l r t i III/ ; | / ! i i I tr i i i/ i 1 i 1 1" i i_ . i ! ! / t ' 1 i i 1 /; i ! I l L 1 1 1 / v ;" 1 1 / \i i K 10 0 > V ' 1 1 K| i J/ « kN )) FE (C » Mi (( 1XR )) AF (1 >R ) Mi ( XY ) JUN (0) JUL (0) AUG (6) SEP (9) OCT (20) NOV DEC (6) (0) Figure 5. Monthly frequency distribution of very large typhoon occur- rences. All occurrences, which are based on 6-hourly reports or obser- vations (1945 to 1968), have at least typhoon intensity (>64 knots), and have a mean radius to the outer closed surface isobar of >10° latitude. The number in parentheses below each month is the number of individual typhoons associated with these 6-hourly occurrences. The ratio (%) of very large to total typhoon occurrences for each month is the number at the top of the graph (Brand, 1970b). the latter month has many more 6-hourly occurrences than July (119 versus 70). During August, these very small typhoons have a longer duration. The monthly ratio of the number of yery small to total typhoon oc- currences (percent values are shown at the top of the graph in fig. 6) shows a peak in August; an even higher peak occurs in June. The extremely large value for February should be discounted because of the small sample number. It is evident that occurrences of both wery large and very small ty- phoons have seasonal variations. The question then arises about whether a geographic variation exists as well. To determine this, all positions of the very large typhoon occurrences for October (when they occurred most frequently) were plotted on a map of the western North Pacific Ocean. These points, when joined, give those segments of the typhoon tracks along which a mature typhoon existed with a mean radius to the outer closed sur- face isobar equal to or greater than 10° of latitude. These very large storm track segments or points (if there was only one 6-hourly report fit- ting the criteria) for occurrences from 1945 to 1968 are presented in figure 7. 407 hi u z D o o o o o a. > 120 110 0 67 3.3 4.4 4.8 RATIO 13.8 (%) 9.5 10.9 5.3 3.4 4.7 6.0 i i — 1 1- — 100 —f- — l — — 1 90 80 70 \ i \ i 1 , 1 i 1 1 r 60 -h 1_ — i 1 50 --* --U-j — [ — - 40 i— on 20 1 10 0 — — — ■ f^l - - "NOV (14) T" J/! (C IN )) FE (1 :b ) (2 XR i) Af >R Mi (! vr >) JUN (15) JUL (24) AUG SEP (27) (19) OCT (ID DEC (9) Figure 6. Monthly frequency distribution of very small typhoon oaaur- renoeSj which are based on 6-hourly reports or observations (1945 to 1968) 3 have at least typhoon intensity (>64 knots) and a mean radius to the outer closed surface isobar of <2° latitude. The number in parentheses below each month is the number of individual typhoons as- sociated with these 6-hourly occurrences . The ratio (%) of very small to total number typhoon occurrences for each month is the number at the top of the graph (Brand, 1970b). The rectangular area in figure 7 represents the apparent geographic location for most of these very large typhoons. This area is approxi- mately between latitudes 15° to 25°N and longitudes 140° to 145 E. Similarly, a position of all very small typhoon track segments or points for August —when the frequency of very small typhoon occurrences was greatest -were plotted on a map (fig. 8). Again, although not as geographically limited as the very large typhoons, very small typhoons are found in the extreme western Pacific Ocean in a rectangular area ex- tending from Luzon, the Philippines, to Tokyo, Japan. In addition to these studies, an updated Western Pacific (WESTPAC) climatology of tropical cyclones and disturbances was published in Novem- ber 1970 (Gray, 1970). 408 f: ■1* ^«l SO»N0 _,_^< +^-. H- + ./■•)::■■•■*. -t i: 1 Ft^ure 7. 2»*>ac& segments for October typhoons (1945 to 1968) existing wzth an average radius to the outer closed surface isobar of >10° lati- tude and a maximum wind of >64 knots. The single points indicate only one 6-hourly observation fitting the criteria. The rectangular block between latitude 15° to 25°N and longitude 140° to 145°E depicts the prevalent area for very small typhoons (Brand, 1970b). INTENSITY FORECASTING RESEARCH In the past, there has been a void in intensity forecasting research that has left the forecaster with very little information on which to rely, apart from extrapolation which is difficult in a 72-hour time frame. Be- cause of this void, EPRF has been delving into intensity forecasting re- search, and some of the output could be of use to Project STORMFURY For example, 25 years (1945 to 1969) of tropical storms and typhoons were ex- amined to provide the forecaster with statistics on how the storms have changed intensity as a function of month, area, and initial intensity (Brand and Gaya, 1971b). Table 1 shows the 12-hour intensity changes in terms of percentage frequency for the observations available for Septem- ber for theQ10 latitude/longitude that extends from 10° to 20°N and from The distributions are separated into five categories according to initial intensity of the tropical cyclone. For example, for category A (initial intensity of less than 34 knots), the 75 observations show an average intensity change in 12 hours of +3 knots: 1 percent of the 409 Figure 8. Typhoon track segments for August typhoons (1945 to 1968) ex- isting with an average radius to the outer closed surface isobar of <2° latitude and a maximum wind of >64 knots. The single points indi- cate only one 6-hourly observation fitting the criteria. The rectangle encloses the prevalent area for very small typhoons (Brand, 1970b). observations showed a decrease of 10 knots; 4 percent showed a decrease of 5 knots; 60 percent showed no change; 21 percent showed an increase of 5 knots; and so forth. Tables were also published for the 24- and 48-hour intensity changes as well as for the 12 hours discussed in this example for each month and for each 10° by 10° area of WESTPAC. The geographic variations of these intensity changes (fig. 9) can be plotted (Brand, 1972a). The 12-hour changes were computed more for Project STORMFURY per- sonnel than for the forecasters who are more interested in the 24- to 48-hour time frame. For Project STORMFURY, intensity changes following seedings must be compared with climatological intensity changes for trop- ical cyclones in the area and month in question. Another area of study being examined at EPRF is rapid intensification and low-latitude weakening of tropical cyclones in the western North Pa- cific (Brand, 1972a). As indicated previously, the annual average 24-hour maximum change in intensity is 30.7 knots for intensifying tropical cy- clones. Thus, a storm exhibiting an increase in intensity of 50 knots in 24 hours, which we define as a rapid intensifier, is quite unusual but 410 Table 1. Percentage Frequency of 12-hour Change in Intensity of Tropical Cyclonic Circulations Which, During Their Cycle, have Reached Tropical Storm or Typhoon Intensity (1945 to 1969). Change in Intensity (the numbers across the top of the table) is in knots per 12-hours Centered on the Values Indicated (Brand and Gaya, 1971b). SEP LAT 10.0-19 .9N long mo.o- -149. 9E C , TOT AVG CHG A i MINI ■f + ♦ + ♦ + •f ♦ > + I 0BS 12H 42 40 35 30 25 20 15 10 05 0 05 10 15 20 25 30 35 40 42 A 75 + 3 1 4 60 21 5 1 1 4 1 B 83 + 8 2 4 2 24 25 13 11 11 2 2 2 C j 82 + 12 1 ^ 22 18 21 17 5 6 5 4 1 Dj 33 + 11 3 3 6 27 27 9 12 6 3 3 EJ273 + 8 1 2 2 31 22 15 10 7 3 4 1 1 CATEGORY - A INITIAL INTFNSITY < 34 KNOTS CATEGORY - B INITIAL INTENSITY 34 - 63 KNOTS CATEGORY - C INITIAL INTENSITY 64 - 100 KNOTS CATEGORY - D INITIAL INTENSITY > 100 KNOTS CATEGORY - E SUM OF A» B» C» AND D does occur often enough to give the forecaster a severe problem. An ex- amination of the monthly distribution of these rapid intensifies reveals that August exhibits a minimum in the ratio of rapid intensification to total tropical cyclone occurrences. This minimum could be related to the low intensification rate for August tropical cyclones. The question then arises -Why is August different? Could it be that August has the most coldcore origin tropical cyclones or the most midget typhoons? An answer to this question may follow from detailed case studies of these selected phenomena which are part of the longer range EPRF program of tropical cyclone research. The forecaster is not only interested in when this rapid intensifica- tion occurs, but also interested in where it occurs. Thus, if those track segments of tropical cyclones that intensified by 50 knots or more in 24 hours are plotted for each month, the geographic variations of this rapid intensification can be seen. For example, figure 10 (top) shows the track segments of September tropical cyclones from 1945 to 1969 which fit the rapid intensification criteria. Also shown are the tracks of all Septem- ber tropical storms and typhoons, separated into those storms having their midpoint in time in the first 15 days of the month (center) and those in the latter 15 days (bottom). The geographic area for rapid intensification is apparent. These areas are also present for other months as well. 411 100°E 110° 120° 130 140° 150° 160° 170°E 180° 1' io°n Figure 9. Geographic variation of the 24-hour intensity change (knots/ 24 hours) for September 1949 to 1969 (Brand, 1972a). In the same manner in which tropical storms and typhoons were examined for rapid intensification, an examination was also made of those storms that tend to weaken at low latitudes. The criteria used for a weakening storm were a 20 knot or greater decrease in maximum wind over 24 hours, with the decrease occurring south of latitude 25°N. As an example, fig- ure 11 shows the track segments of tropical cyclones from 1945 to 1969 fitting these criteria for July, August, and September. Ideally, these segments should be compared with the storm tracks, but for brevity only these "weakening" segments are examined. Most tropical cyclone weakening segments are associated with those storms reaching the Asian mainland or passing over the Philippines or Taiwan; however, exceptions do exist. In July, a number of northwestward- moving storms weaken over the open ocean in an area bounded approximately by latitudes 15° to 25°N and longitudes 125° to 130°E. In August, which has a tremendous amount of storm activity, surprisingly few storms weaken over the open ocean at low latitudes. In September, weakening storms are more apparent, not only in the western North Pacific but also in the South China Sea. 412 TRACK SEGMENTS FOR SEPTEMBER TROPICAL! CYCLONES (1945-1969) EXPERIENCING A 50 KNO OR GREATER INTENSITY INCREASE IN 24 HOURS SEPTEMBER (16-30) TROPICAL & TYPHOONS 1945-1969 FigvLre 10. September (1945 to 1969) track segments of rapidly intensify- ing tropical cyclones (top). Tracks of tropical storms and typhoons (1945 to 1969) having their midpoint in time in the first 15 days of September (center) 3 and those having their midpoint in time during the last 15 days of September (bottom). 413 AUGUST TRACK SEGMENTS FOR TROPICAL, CYCL0NES0945 1969) EXPERIENCING A 20 KNOT OR GREATER INTENSITY DECREASE IN 24 HOURS OCCURRING SOUTH OF 25 N. ^ SEPT EMBEF -T J A )J .-■ /■^~*^ :-' "* '-• < < frr~" ? ^ M \ -V*c T~ ^ f Figure 11. Track segments (south of latitude 25°N) for weakening tropical cyclones for July, August 3 and September (1945 to 1969). 414 Other intensity forecasting areas that NWRF and EPRF have examined include the effect the Philippines' landmass has upon tropical cyclones (Brand, 1972b), the intensity of recurving typhoons (Riehl, 1971), and the effect on the intensity of storms crossing the cold-water wakes of previous tropical cyclones (Brand, 1971a). MOVEMENT FORECAST RESEARCH The general movement forecast problem is important to Project STORM- FURY because information concerning storm movement is essential for logis- tical purposes. A significant step in obtaining such information was ac- complished with the development of the Typhoon Analog Program (Jarrell and Somervell, 1970; and Brand, 1970c). For the past 2 years, results of the Ty- phoon Analog Program developed at NWRF indicate that it is the best ob- jective forecast technique available to the Joint Typhoon Warning Center, Guam, and that it has aided in the reduction of forecast errors, especially in the 48- and 72-hour time frame. The Program in its present form is still just a first step, because a number of modifications or additions could improve it still further. The EPRF will be developing these improve- ments based on 2 years of operational experience. Because of the large numbers of tropical cyclones in the western North Pacific, multiple storm occurrence is quite common. When storms are close to each other, interaction occurs; this is a problem to the forecaster. This interaction, the "Fujiwhara" effect, was examined, and definitive re- sults concerning this interaction were found (Brand, 1970a). A separate study examined the movement effect on the path of a second storm that fol- lowed the cold-water wake of a previous storm (Brand, 1971a). Additionally, movement and development information can be found in the updated WESTPAC tropical cyclone climatology for the western Pacific (Gray, 1970) and in a publication of South China Sea tropical cyclones (Brody and Jarrell, 1969). FUTURE RESEARCH Future tropical cyclone research topics at EPRF, which could be ap- plicable to Project STORMFURY, include the following areas: 1. Speed variations of tropical cyclones, 2. Stationary tropical cyclones, 3. Looping tropical cyclones, 4. Development of forecast rules for characteristic types of tropical cyclones (i.e., midget typhoons, yery large ty- phoons, rapid intensifiers, and low-latitude weakeners), 5. The tropical upper tropospheric trough and its relation to tropical cyclones in WESTPAC, 6. The thermal structure of the ocean and its relation to tropical cyclones, 7. Shallow tropical cyclones, and 8. Longer range forecasting. 415 REFERENCES Brand, S. (1970a): Interaction of binary tropical cyclones of the western North Pacific. Journal of Applied Meteorology , 9, (3), pp. 433-441. Brand, S. (1970b): Geographic and monthly variations of yery large and very small typhoons of the western North Pacific Ocean. NAVWEARSCHFAC Technical Paper No. 13-70. Brand, S. (1970c): Space-time inventory of western North Pacific tropical storm and typhoon frequencies. NAVWEARSCHFAC Technical Paper No. 23- 70. Brand, S. (1971a): The effects on a tropical cyclone of cooler surface waters due to upwelling and mixing produced by a prior tropical cy- clone. Journal of Applied Meteorology , 10, (5), pp. 865-874. Brand, S., and R. F. Gaya (1971b): Intensity changes of tropical storms and typhoons of the western North Pacific Ocean. NAVWEARSCHFAC Tech- nical Paper No. 5-71. Brand, S. (1972a): Geographic and monthly variation of rapid intensifica- tion and low-latitude weakening of tropical cyclones of the western North Pacific Ocean. EPRF Technical Paper (to be published). Brand, S. (1972b): The effects of the Philippines on typhoons. EPRF Tech- nical Paper (to be published). Brody, L. R., and J. D. Jarrell (1969): A technique for predicting South China Sea tropical cyclones. NAVWEARSCHFAC Technical Paper No. 8-69. Gray, W. M. (1970): A climatology of tropical cyclones and disturbances of the western Pacific with a suggested theory for their genesis/ maintenance. NAVWEARSCHFAC Technical Paper No. 19-70. Jarrell, J. D., and W. L. Somervell, Jr. (1970): A computer technique for using typhoon analogs as a forecast aid. NAVWEARSCHFAC Technical Paper No. 6-70. Riehl, H. (1971): Intensity of recurving typhoons. NAVWEARSCHFAC Tech- nical Paper No. 3-71. 416 APPENDIX G DIURNAL VARIATION IN HURRICANES Robert C. Sheets National Hurricane Research Laboratory INTRODUCTION One question that arises quite often in the study of hurricane struc- ture is whether a significant diurnal variation occurs in the structure or intensity of the hurricane. The author found in an earlier study (Sheets, 1969) that some differences are apparent in the middle to upper levels (500 to 150 mb) where daytime temperatures are slightly warmer than respective nighttime temperatures. Also, instability seemed to in- crease slightly at night, but no conclusive evidence was found to substan- tiate any diurnal variation in the intensity of hurricanes. This present study was undertaken to determine if a diurnal variation occurs in the intensity of the hurricane as reflected by the maximum wind speeds. The minimum sea-level pressures (SLP) were plotted by time of day for all hurricanes in the Atlantic Ocean, Gulf of Mexico, and Caribbean Sea from 1961 through 1968. Some bias in the data exists in that storms were monitored frequently during the deepening and mature stages, but a com- parable dissipating stage was often missed because the storms recurved or struck land. The maximum wind speeds were then computed from these mini- mum sea-level pressure graphs at 6-hour intervals (0000, 0600, 1200, and 1800 GMT), using a relation presented by Holliday (1969). The minimum sea-level pressure measurements were used to define the intensity of the storm rather than using the reported maximum wind speeds, because wind speeds are often estimated. Also, a large variation occurs in the actual wind speed on relatively small scales in time and space, while the minimum sea-level pressure is considered to be much more conservative. The means and standard deviations of the computed maximum wind speed values were determined for the 6-hour intervals. The correlation coeffi- cients between these values were then determined for each time interval. For example, correlation coefficients were computed for all combinations of 6-, 12-, and 18-hour periods using data collected at 0000, 0600, 1200, 417 and 1800 GMT. The results (table 1) show a maximum difference of 1.1 knots in the mean maximum wind speeds for the selected time periods and a range of standard deviations from 18.8 to 20.3 knots. These values indicate that no significant differences are present in the samples for the selected time periods, and that the storms studied were generally of moderate hurricane intensity. A very high correlation exists between the values for the various time periods, with the correlation coefficients being greater than 0.99 in all cases and the differences occurring in the third significant digit. These correlations indicate that, on the aver- age, large changes in hurricane intensity do not occur in a set pattern for these time scales, as would be true if there were a diurnal varia- tion. The next step was to determine if a set pattern of small-scale changes in hurricane intensity occurred for the time scales previously listed. A few strategically oriented 24-hour trends could mask the effects of small- scale changes. Therefore, the 24-hour trend was removed along with the large-scale contribution in this portion of the study. The procedure used was to compute the 24-hour trends from 0000 until 0000 GMT on the following day for each set of data available. This trend was assumed to be linear over these 24-hour periods. The computed trend was then used to extrapolate the intermediate time-period values to the end of the 24- hour period. The deviation for the extrapolated value from the end point was then determined. For example, the deviation value obtained for the observations on 18 September, where the 24-hour trend is determined be- tween 1800/0000 GMT and 1900/0000 GMT, would be computed as indicated in figure I. The hypotehtical case shows a distinct 24-hour cycle similar to that which should be present if a distinct diurnal variation actually exists. Of course, the maximum amplitudes of this variation might occur at times other than 0600 and 1800 GMT. Therefore, the procedure described above was repeated for 24-hour trends computed over time intervals of 0600 to 0600 GMT, of 1200 to 1200 GMT, and of 1800 to 1800 GMT. This procedure results in considerable overlap of data, but should accentuate any syste- matic change detectable at any of the observation times. Table 1. Maximum Wind Speed Data, 1961 to 1968, Time (GMT) Mean (kt) Std. Dev No. Cases Time (GMT) Correlation Coefficient 0000 97.1 20.3 135 0000 vs. 0600 0000 vs. 1200 0.998 0.994 0600 96.1 19.8 135 0000 vs. 1800 0600 vs. 1200 0.990 0.997 1200 97.1 18.8 136 0600 vs. 1800 1200 vs. 1800 0.994 0.997 1800 97.2 19.9 131 418 o c _^ "O UJ < UJ (0 3 1010 - 1000 990 980 970 960 950 940 930 920 • t • — ** j • • t r • . t r .»• •» • t • * • • • • • • . • • • • • • • ■ .« • • •i t : • •. • •• •t .• 4. • • • • • .• • •• • • •• . t • • • -L ± -L 00 02 04 06 08 10 12 14 TIME (GMT) 16 18 20 22 00 Figure 2. Data collected by reconnaissance aircraft in Atlantic, Carib- beans and Gulf of Mexico hurricanes (1961 to 1968). 421 SUMMARY AND CONCLUSIONS Data collected in this study indicate that no significant diurnal variation of hurricane intensity exists, on the average, as reflected by the maximum wind speeds or minimum sea-level pressures. Table 1, disre- garding any sampling bias that may have occurred, shows wind speeds were very highly correlated at each of the four selected time periods and that the means differ by less than 2 percent. If a large diurnal variation exists, it would be reflected in a significant difference in the mean values and a lower correlation between values separated by 12 hours. If any diurnal variation in the maximum wind speeds actually occurs, it should be indicated by a negative or low positive correlation coeffi- cient for daytime versus nighttime deviations of the maximum wind speeds that have been adjusted for the 24-hour trend. These values are shown in table 2. Correlations for the data that are 6 hours apart are generally higher than for those that are 12 hours apart. One point that seems to be most significant is that the correlation coefficient for the 0000 ver- sus the 1200 GMT data is approximately one-half that of the 0600 versus the 1800 GMT data. From the means of these deviations, we find that the maximum wind speeds are generally less at 1200 than those at 0000 GMT, but in all cases this difference is shown to be less than 1 percent, while the standard deviations are about 6 percent. We can conclude then that al- though some differences apparently do exist, they are not very large and are well within the measurement-error range. Table 3 shows that there is some bias in the data because more data were collected during intensification periods than during the dissipation stage, with the mean 24-hour trends being approximately 7 to 8 percent. Mery low correlations were obtained for the trend versus the maximum wind speed deviations (adjusted for 24-hour linear trend) for all the selected time periods. If a diurnal variation should exist, dependent upon whether a storm was deepening or filling, then significant correlation coefficient values would have been obtained. Because only small correlations were ob- tained, one can conclude that if a significant diurnal variation exists, it is not dependent upon the intensity change of 24-hours or longer. REFERENCES Hoi li day, C. (1969): On the maximum sustained winds occurring in Atlantic hurricanes. NOAA, wbtm-SR-45, 6 pp. Sheets, R. C. (1969): Some mean hurricane soundings, Journal of Applied Meteorology, 8, No. 1, pp. 134-146. 422 APPENDIX H NAVAL WEAPONS CENTER CONTRIBUTIONS TO STORMFURY DRY-RUN AND CLOUDLINE OPERATIONS 2 to 13 August 1971 She! den D. Elliott, Edward E. Hindman, II, and William G. Finnegan Naval Weapons Center, China Lake, California INTRODUCTION The Naval Weapons Center's (NWC) major contributions to project STORMFURY have been the development and supply of pyrotechnic silver- iodide ice nuclei generators and the systems required for dispensing them from Naval and Air Force aircraft, as well as scientific and ord- nance technician staff for hurricane dry-run exercises and operational developments. The Center regards its current responsibility in the STORMFURY pro- gram to be the development of new ice nuclei generating systems for po- tential use in rainband, rainsector, and cloudline experiments and the cloud seeding technology most effective for these applications. The cur- rent STORMFURY pyrotechnic generator (the WMU-2) was designed for exclu- sive use in eyemod experiments and would be wasteful of silver iodide if used otherwise. Assuming that the STORMFURY program progresses to operational status, requirements for wery large numbers of ice nuclei generators would follow. Development of new, more efficient silver-iodide pyrotechnics or of non- silver bearing ice nuclei generators is considered to be a necessity in this event. AIRCRAFT For the 1971 STORMFURY dry-run and cloudline operations, NWC again made available the contractor-owned and -operated Cessna 401, N3221Q, pro- vided in 1969 and 1970 (Elliott and Finnegan, 1971). For this season, the MINILAB meteorological data collection and digital recording system was provided with the latest array of sensors, as described in table 1. The data sampling rate, in continuous mode, is 2.4 scans per second. 423 Table 1. MINILAB Characteristics Parameter Sensor System Range Time Altitude Indicated Airspeed Temperature Dew Point Liquid Water Content Vertical Velocity Rain Rate Digital Clock CIC Mdl. 7000 Altitude Transducer (from air- craft pi tot-static sys- tem) CIC Mdl. 7000 Airspeed Transducer (from air- craft pi tot-static sys- tem) WSI TS-22 Temperature System (thermistor bridge) Cambridge Mdl. 137-C3 Dew Pointer (cooled mirror) Johnson-Williams LWC Indicator (hot-wire bridge) Ball Mdl. 101-B Vari- ometer (capillary leak) WSI RR-40 Rain Rate Meter (acoustical impact) 24 hours (by sec- onds) 0 to 30,000 feet 0 to 350 knots •30° to +30°C 50° to +50°C 0 to 2 grams/m3 (low) 0 to 6 grams/m3 (high) -3000 to +3000 feet/mi n 0.35 to 7.25 mm dia 2 x 10"5 to 2 x 10"1 grams mass (nine size classes) The aircraft was also equipped, as before, with a Bendix RDR-100 K-band weather avoidance radar. The 52 stations (26 beneath each nacelle) for air-dropped pyrotechnic flares were augmented by a pair of booms mounted aft of and parallel to the wing trailing edges, each carrying 14 burn-in-place fusees for hori- zontal dispersion of freezing nuclei (fig. 1). A total of 70.2 flight hours was logged by 210 during 29 July to 16 August, including transit time from Arcatia, California, to Roosevelt Roads; Roosevelt Roads to Seawell Airport, Barbados; and from Barbados back to the contractor's homebase at Norman, Oklahoma. Three orientation and training flights for NWC and VA85 personnel, totaling 6.3 hours, were made at Roosevelt Roads during the dry-run exercise, 2 to 6 August. 424 Figure 1. Burn-in-plaoe fusee in operation on wing boom aboard Navy- leased Cessna 401 (21Q) during cloudline experiments. During the cloudline experiments, 8 to 13 August, five missions were flown, totaling 17.3 hours, with NWC and NHRL personnel aboard. Three of these were devoted to comparative tests of standard and experi- mental free-fall pyrotechnic seeding devices, with 21Q operating inde- pendently of the remainder of the STORMFURY forces, while the remaining two were coordinated tests in which the NWC aircraft was the seeder. Thus, as in previous years, the NWC aircraft contributed materially to the suc- cess of the program. TRAINING One important component of NWC support to Project STORMFURY is the provision of training in seeding techniques, based upon experience gained in other weather modification programs. NWC and NHRL personnel conducted a lecture and discussion session attended by VA-85 (A-6), VW-4 (WP-3 and P-3A), and 53-WRS (WC-130) aircrews following the rainsector dry-run gen- eral briefing on 4 August. Concurrently, NWC ordnance and avionics per- sonnel trained ground crews in ordnance and dispenser handline procedures. 129 425 On 3 and 4 August, two training flights were conducted. NWC personnel coordinated and demonstrated refined seeding techniques, operating on iso- lated clouds near the east and south coasts of Puerto Rico. A major train- ing session involving 21Q and two VA-85 A-6's was then conducted on 5 August in a massive cloudline 50 to 60 n miles southeast of Roosevelt Roads. This mission also yielded air-to-air documentary coverage of A-6 operations. Training was continued during the cloudline experiments in Barbados on 8 to 13 August, when NWC personnel flew aboard the P-3A, WC-130, and DC-6 aircraft and assisted the air crews in selecting and seeding the experi- mental targets. The effectiveness of this training was demonstrated by the efficient performance of Navy and USAF seeder aircraft crews during the Hurricane Ginger experiments. PYROTECHNICS The STORMFURY I eyemod seeding rounds used through 1969 (Elliott, 1970) were replaced in 1970 by the WMU-2 unit, similar in configuration and nucleant composition, but modified to meet military fabrication, storage, and safety standards. The 5,616 WMU-2 units procured in 1970 remained on hand at NAS Jacksonville and Naval Station Roosevelt Roads at the beginning of the 1971 season. For training, and for cloudline ex- periments in which seeding is performed from altitudes of 16,000 to 23,000 feet, two less-expensive, shorter burning units of similar design and construction were provided: WMU-9, containing slightly more than half the charge of the WMU-2; and WMU-1, a replacement for the former STORM- FURY III and Ex 1 Mod 0 rounds used in previous training and cloudline operations. WMU-2 and -9 employ the high-Agl output TB-1 (former LW 83) nucleant composition, suitable for moist, strongly convective clouds, while WMU-1 uses the TB-2 (former EW-20) formulation, preferred for weaker clouds (table 2). Table 2. STORMFURY Seeding Nueleant Formulations - 1971 td i* Weight TD 0 Weight TIM (%) ^± In AgI03 78.2 AgI03 24.09 Mg 10.8 Mg 8.13 Al 5.2 Al 18.04 KN03 43.74 Binder 5.8 100.0% Binder 6.00 100.0% TB-1 and TB-2 are adaptations, respectively, of the former LW-83 and EW-20 formulations (Elliott et al., 1969) , as produced by NAD Crane. 426 NWC is continuing to develop more efficient and effective seeding materials and techniques; therefore, a part of the cloudline experiment was devoted to testing three new pyrotechnic compositions and configura- tions. The seeding flare ejected from the WMU-2 cartridge contains a cylinder of pressed pyrotechnic compound 1.3 inches in diameter and 4.2 inches long; burning takes place on only one of the circular end faces. Experimental and theoretical combustion studies suggest that the most effective nuclei are produced near the outer periphery of the flame zone, because coagulation and other effects reduce the efficiency of the par- ticles produced nearer the center. According to this hypothesis, the number of effective nuclei produced by a flare of smaller diameter would be reduced approximately proportionally to the first power of the diameter, whereas the mass of compound expended would decrease as the square of the diameter, with consequent improvement in the number of effective nuclei per gram. A quantity of units carrying the experimental designation EPF- 0.5-TB1-8-6.0-1 was made up, consisting of the standard external case fitted with a subcaliber barrel from which an end-burning TB-1 flare 0.375 inches in diameter by 6 inches long could be ejected. With a length over twice that of the WMU-9 and a diameter approximately one-third as great, this unit was expected to produce one-half to two-thirds as many effective nuclei as the WMU-9 from less than one-quarter as much material. A similar unit was fabricated using a reduced diameter grain of TB 14.1, a nonsilver-bearing experimental composition developed jointly by NWC and NAD Crane, for comparison with the reduced-diameter TB-1 units. Finally, to compare the effectiveness of nuclei dispersed horizon- tally from an aircraft operating in updraft zones below cloud base, or at low levels within cloud, a number of STORMFURY I rounds were reworked into burn-in-place fusees, designated EWF-1.4-LW83-5.4T-4.5. The characteristics and use of the various pyrotechnic units employed in the 1971 STORMFURY program during dry-run/cloudline operations are sum- marized in table 3. SPECIAL CLOUDLINE EXPERIMENTS Testing Pyrotechnics The testing of novel seeding materials and techniques by Elliott and Finnegan (1971) during the 1970 STORMFURY operations was continued during the 1971 operations. These tests are important because empirical evidence compiled by St.-Amand et al . (1971) and St.-Amand and Elliott (1972) indi- cates that too much Agl introduced into a supercooled cloudline cumulus may inhibit precipitation. They suggest that too rapid a release of the latent heat of fusion may cause the cumulus to "pinch off" at its center, breaking the natural convective circulation pattern and destroying the cloud. Observations by Hindman et al . (1971) confirm this behavior and suggest that releases of smaller and more widely dispersed amounts of Agl may be more efficient than the amount of Agl (110 grams) in the present cloudline pyrotechnics (WMU-9). This has been confirmed in other NWC pro- grams by comparative tests using the WMU-1 (22 grams Agl) pyrotechnics. 427 Cv CO 0) d K CO CO K % 4i C r~ is; I I o to +5 CO +i g t*3 |— •!— I— I 4-> r— r-- C O.CF1 (O Q.I— I 3 3 — - Crco -a OJ >> o J- 3 o a: ai u _i O) CO to ai oo 0) r— 00 C1J u c <0 +J <4- U_ c/> Q +-> a; +-> U- 0J i — cc en K=c 3' o en i- en CV- Cn C C 1- aj — . _i s- ai e E •!- (O • a ai ai 4- 1^3 O I LO o o o o o o o o o o o o I— » 00 CO o 00 o o o o o o o o co o t-H OJ o o o o o o C\J ■3- 3 o o Q_ S- 3 CL o co o KG I CM o o o o o o cr> o o o o o o CO CO OvJ o o o o o o o o o o o o CO CO CNJ en a: o o o o o o c S- 3 o -Q o i-H O) o V) crv (O -Q ■a 3 O o ^— o o i— i , o 1 — en CO C\J O CO o o o o X) ro S- ro CL qj o E S_ O 1 i— § CO co 3 _l «* • o i—i en 1 C\J 0) oo •— Q. T3 -C O oo <0 T3 ai CO i— C s- OOfO • r— •r- Q. <4- -•-> E S- X 4-> 00 o o o ai 4-> 4-> C +J x: O J- 3 en (O O M- >> •r— S- +j •■- (O QJ eo T3 CO "O 4J E 3 3 O •(- i— LL. i — lO-P u< O X CO C CO c h cm fo j- m 428 It has been argued (the previous section) that more efficient use of nuclei may be obtained by seeding units with a smaller diameter. Accord- ingly, reduced-diameter revisions of the WMU-9 pyrotechnics were developed which contained 21 grams of Agl. These pyrotechnics were compatible with the racks used with the present WMU-9 pyrotechnic. The free-fall reduced-diameter TB-1 unit was tested on 8 August 1971. The test clouds were 108 km west of Martinique and 270 km northwest of Barbados. The 1200 GMT sounding from Fort-de-France, Martinique, is shown in figure 2. The atmosphere was extremely dry at the flight altitude of 4,780 m (-2°C). The test clouds were growing to a maximum of 5,490 m (-4°C) and cycling every 11 minutes. A growing cloud was penetrated at 1920 GTM; one flare was fired into the region with the largest vertical motion and updraft. Figure 3 illustrates the vertical motion and liquid water content conditions encountered during the seeding penetration. The 400 7 / ' s ld \r \ N s \ \ / / 7590m 7 / 500 V \ \ \ 5880m /> f"» / r Xd / \ £ "# - FLIGHT ALT — / uj 600 DC 7 \"~\ \ =) — \ \ \ \ — CO CO V ^-> ) \ / uj 700 DC Q_ 7 \ > \ 3175m M 800 I> ^ ^ V • 7 900 7 — \N 1547m >'z 1000 7 / / / / / /^ \ 142m 7 7 7 /^- V s> N?

TEMPERATURE (°C) Figure 2. 1200 GMT sounding, 8 August 1971 from Fort-de-France, Martinque, The temperature profile is given by Yfo the dew point profile by Ys; Tj and Ys are the dry- and moist-adiabatic lapse rates, respectively . 429 CLOUD ENTRY CLOUD EXIT CLOUD ENTRY CLOUD EXIT Figure 3. Conditions during the seeding penetration at 192011 GMT on 8 August 1971. The penetration altitude was 4780 m (-2°C). maximum updraft (8.5 m/sec) was located by observing a Ball variometer3 and the region of maximum liquid water (1.2 grams/m3) was located by ob- serving a Johnson and Williams liquid water content meter4. The observed maximum water content and updraft values indicated suitable conditions 3The Ball variometer is a sensitive rate-of-cl imb meter for use in sail planes. It is manufactured by Ball Engineering Co., 2 1 40 Kohler, Dr., Boulder, Colo., 80303 (see table 1). ^Manufactured by Johnson-Williams Products, Bacharach Instrument Co., Mountain View., Ca . (see table 1). 430 for promoting additional cloud growth through seeding; however, no addi- tional growth was observed. The acceleration in growth caused by the seeding probably was offset by accelerated entrainment of dry air because the cloud was so narrow (0.5 km wide). Hence, this test of the reduced- diameter unit was inconclusive. Another cloud was seeded on 8 August 1971 at 1951 and 2000 GMT with a total of four reduced-diameter pyrotechnics. No additional growth above the neighboring clouds was observed. Apparently, large amounts of seeding material cannot promote additional growth in clouds that have small re- gions of supercooled water. Additional testing of the reduced-diameter unit took place south of Trinidad (360 km southwest of Barbados) on 9 August 1971. No atmospheric sounding was available from Port of Spain, Trinidad, although the atmos- pheric conditions appeared similar to those of 8 August 1971. One cloud was seeded at 1832, 1844, and 1909 GMT with three, three, and two reduced- diameter pyrotechnics, respectively. No additional growth was observed that could be attributed to seeding. The extremely dry conditions at -5°C caused the latent heat released through seeding to be overwhelmed by evap- orative cooling from the cloud's liquid water. The reduced-diameter (TB-1) and regular-diameter (WMU-9) pyrotechnics were tested on 11 August 1971. The test clouds, a line of towering cumulus and cumulonimbus clouds, were 270 km east of Barbados. The target cloud did not appear to be growing significantly before the first seeding run at 1808 GMT. The atmospheric conditions at seeding level (-5°C) were more suitable than on the previous test days (dew point -7 C). The cloud contained a maximum of 1 gram/m3 of liquid water and 5 m/sec updraft; it was seeded at 1808 and 1811 GMT with two reduced-diameter units. By 1824 GMT, numerous turrets were building from the seeded mass well above the -5°C flight level. A second distinct cell was building adjacent to the seeded cell at this time. This adjacent cell was seeded at 1833 GMT with six regular diameter WMU-9' s. The second seeded cell contained a maximum of 2 grams/m3 and 10 m/sec updraft on the seeding penetration. At 1837 GMT, a nearby but separate cloud exhibited explosive growth and was there- fore considered as a control cloud. By 1844 GMT, both seeded cells were growing into Cb's and were joined at the base. The control had also reached Cb stage by this time. By 1858 GMT the two seeded cells reached Cb stage, but their tops were still separated. By 1915 GMT, the two seeded cells merged into one enormous Cb and the control cloud had dissipated. After we landed at Seawell Airport, Barbados, at 1940 GMT, the seeded cloud was visible on the horizon and was the largest Cb in the region. The 11 August 1971 test demonstrated that two seeded cloudline cumu- lus grew to the same final proportions. One was seeded with 600 grams of Agl from full -diameter units and the other with 42 grams from reduced- diameter units. Seeding a tropical cloudline cumulus at its base was investigated on 12 August 1971. The N0AA-RFF DC-6 (40C) monitored the upper portions 431 of the cloud while its base was seeded with 8 to 190 grams Agl burn-in- place fusees. The cloud was vigorous during the seeding but never grew above the freezing level to allow the nucleant to promote additional growth. Two other clouds were seeded at the +7°C level. One with six fusees grew above the freezing level and exhibited spectacular growth thereafter. The other cloud seeded with two fusees did not grow past the freezing level . The attempt at seeding at cloud base indicated that it is difficult to find an organized updraft in which to release the seeding agent. The release of seeding material at cloud base and its subsequent transport to the supercooled portion of the cloud has been done in continental cumulus clouds (Todd, 1965; Dennis et al., 1971). Further flights at cloud base in the tropics should establish the potential for seeding at this location. The reduced-diameter (TB-1) pyrotechnics and a similar nonsilver- bearing (TB-14.1) pyrotechnic were tested on 13 August 1971. The test clouds were 140 km south of Barbados, immediately south of an extensive east-west oriented storm system. The first cloud was seeded with five of the TB-14.1 pyrotechnics and exhibited little growth following seeding but glaciated completely. The second cloud was seeded at -5°C with 12 reduced-diameter TB-1 pyrotechnics. The cloud contained heavy precipita- tion and peak updrafts of 10 to 15 m/sec and exhibited significant growth following seeding. The third cloud was seeded at -5°C with four of the TB-14.1 pyrotechnics. No significant growth was noted following seeding, although the cloud appeared to broaden. The test area on 13 August 1971 was surrounded by fully developed cloudlines. The first and third clouds were in a clearing between the lines and probably had their growth suppressed by subsidence from the nearby cloudlines. The second cloud was on the flank of a cloudline and probably had its growth enhanced by the line. Therefore, it is not pos- sible to draw conclusions about the effectiveness of the reduced-diameter TB-1 and TB-14.1 pyrotechnics from this experiment. However, the new nuc- leant appeared to cause significant glaciation in one out of two test clouds, and clearly warrants further testing. Testing Agl-NH^I and Ag I -Na I -Acetone Solutions During the 1970 STORMFURY cloudline experiments, seeding was conducted by burning a solution of silver iodide, ammonium iodide, and acetone in a solution-combustion generator (Patten et al., 1917). The seeding was im- portant because recent evidence (Finnegan et al . , 1971) suggests that this solution, which produces Agl of high purity, is the optimum for seeding clouds with a solution-combustion generator. The results of this test (Hindman et al . , 1971) suggested that the solution was superior to the traditional silver iodide-sodium iodide-acetone solution, which yields an Agl-Nal complex, although the results were not conclusive. Therefore one objective of the NWC participation in the 1971 cloudline experiments was to continue testing the solutions. 432 While the NWC Cessna 401 and NOAA-RFF DC-6 (40C) conducted the cloud- base seeding tests on 12 August 1971, the NOAA-RFF DC-6 (39C), C-130 (41C), B-57, and the USAF WC-130 operated together in another area. The DC-6 (39C) was equipped with a solution-combustion generator on each wingtip; one generator contained a solution of silver iodide, ammonium iodide, and ace- tone and the other a solution of silver iodide, sodium iodide, and acetone (see fig. 4). The composition of the solutions is given in table 4. The atmospheric sounding on 12 August 1971 was deduced from tempera- ture and dew point data taken by 41C during a rapid climb from cloud base (1016 mb) to the -5°C level (516 mb). The portion of the sounding above 500 mb was interpolated from the Seawell Airport, Barbados, soundings taken at 1200 GMT on 11 and 13 August 1971. The reconstructed sounding is shown in figure 5. The sounding illustrates that the ST0RMFURY operating area can support significant convective activity. Figure 4. Solution-combustion ioe nuclei generator aboard NOAA-RFF DC-6 (39C). 137 433 Table 4. STORMFURY Seeding Solution Formulations - - 1971 2 Agl-Nal Solution Weight Agl (%) Solution Weight (%) Agl Nal H20 Acetone 5.0 1.6 3.0 90.4 10070% Agl 5.0 NH^ 1.6 H20 3.0 Acetone 90.4 100.0% The test sequence was as follows: (1) 39C flew at the +5°C level and 41C at the -5°C level; (2) 41C picked the target clouds and 39C decided, on a random basis, whether the cloud would be seeded with the Agl-NHi+I-acetone burning generator, the Agl-Nal-acetone burning generator, or not seeded; (3) 39C penetrated the cloud five times, seeding on the first three, while 41C made five monitoring penetrations; and (4) the first cloud was seeded 400 500 -7 -Q E w600 cc D CO uj 700 DC Q_ 800 900 1000 / / / / / \ / / / / *\£> U \ \ s / / \ \ \ \ \ y \ s N \ ^ \ \ — — / \ s V \ —MONITOR ALT — "* \*o 7 /■ \ \ 1 ^^v \ \ / \ —SEED ALT — — \ 1 k\ 7* S. \ f \ ^ N N. V V i -<£ \1\ T \\ \ HI 2 /■ V\J ~~ / / / / / / / V / / 7 *$> <£ <£ * £ TEMPERATURE (°C) Figure 5. 1700 GMT sounding, on 12 August 1971 3 from STORMFURY oloudline operating area. The temperature profile is given by Yd> the dew point profile by Ys; Tg and Ts are the dry- and moist-adiabatic lapse rates 3 respectively . 434 at 1940 GMT with the complex 2 AgI«NaI ice nuclei, the second cloud was seeded at 1955 GMT with uncomplexed Agl ice nuclei, and the third cloud was "not seeded" at 2051 GMT. The WP-101 radar data from 39C were analyzed by planimetry of the echoes of the three test clouds. The area of the echo immediately before each penetration was estimated. Then the change in area relative to that observed on the first penetration (pre-seed) was calculated. The results of these calculations are in figure 6. The liquid water content data from 39C was also analyzed. The average water content for each pass was esti- mated and plotted against the time from the initial penetration (see fig. 7). 0\l E 4.0 h- < DC H LU < O a: < O o LU (J < X CJ 3.0- 2.0 1.0 CLOUD III (NO SEED) CLOUD I (2 Agl-Nal) - r\/ / / / / CLOUD II (Agl! 0 10 20 30 40 TIME FROM INITIAL PENETRATION (MIN) 50 Figure 6. The growth of echo area in target clouds on 12 August as esti- mated from the NOAA-RFF DC-6 (39C) WP-101 Radar. 435 0.60 00 I E LU 0.50 o o cc LU H < a LU (J < CC LU > < 0.40 0.30 0.20 CLOUD III (NO SEED) 'O * CLOUD II (Agi; / CLOUD I (2 Agl-Nal) 10 20 30 40 TIME FROM INITIAL PENETRATION (MIN) 50 Figure 7. Average liquid water content in target clouds on 12 August 3 as observed from the N0AA-RFF DC-6 (39C) as a function of time after ini- tial penetration. The echo growth results in figure 6 show that the Agi seeded cloud grew the least. Figure 7 shows that the liquid water content increased most significantly in the Agi seeded cloud. Furthermore, the liquid water was increasing when 39C left this cloud, but was decreasing when 39C left the nonseeded and the cloud seeded with 2 AgI«NaI. The small echo growth and large increase in liquid water of the Agi seeded cloud indicate more vertically oriented growth than in the other two clouds. The results (fig. 7) indicate that both seeded clouds contained more liquid water throughout the observations than did the nonseeded cloud. This suggests that the two seeded clouds produced more precipitation than the nonseeded cloud. 436 Significant observations were made by Finnegan (personal communica- tion) from 41C. He observed that the first cloud did not change appreci- ably, the second cloud showed vertical tower development, and the third cloud exhibited good growth, but neighboring clouds grew and dissipated as well. From these observations, we deduced that (1) the third cloud was not seeded, (2) the second cloud had more vertical growth; therefore, it had been seeded with Agl, and (3) by elimination, the first cloud was seeded with 2 Agl-Nal. These deductions were made before the seeding sequence from 39C was disclosed. CONCLUSIONS The following preliminary conclusions can be drawn from the NWC spe- cial cloudline experiments during the 1971 STORMFURY cloudline operations, 1. The reduced-diameter TB-1 pyrotechnics were unable to in- vigorate marginally supercooled cloudline cumulus. Previ- ous experience has shown that full -diameter pyrotechnics also do not work under these conditions. A warm-phase seeding technique must be developed to grow the margin- ally supercooled clouds into clouds that are seedable with ice-phase pyrotechnics. 2. One test, on vigorous cumulus clouds growing well above the freezing level, compared the reduced-diameter TB-1 pyrotech- nics with full -diameter pyrotechnics of the same composition and showed that two of the reduced-diameter pyrotechnics pro- duced the same results as six of the full -diameter pyrotech- nics having 15 times the nominal Agl output. 3. Cloud-base and midlevel horizontal seeding were difficult to execute and need further investigation. 4. Tests between pyrotechnically generated Agl and complex ice nuclei from a new nonsilver based pyrotechnic were inconclu- sive and need further investigation. 5. A test between Agl and 2 Agl-Nal ice nuclei indicated the Agl to be superior when both nuclei are released at +5°C. REFERENCES Dennis, A. S., R. A. Schleusener, A. Koscielski, and M. R. Schock (1971): Modification of precipitation from convective clouds in the northern plains of the United States. Proo. Int. Conf. Wea. Mod., 6-16 Sep- tember 1971, Canberra, Australia, pp. 103-110. Elliott, S. D., Jr., R. Steele, and W. D. Mallinger (1969): STORMFURY Pyrotechnics. Project STORMFURY Annual Report 19683 Appendix B. Elliott, S. D., Jr. (1970): STORMFURY Seeding Pyrotechnics - 1969. Pro- ject STORMFURY Annual Report 1969, Appendix B. 437 Elliott, S. D., Jr., and W. G. Finnegan (1971): Use of light aircraft in STORMFURY activities. Project STORMFURY Annual Report 1970, Appendix I Finnegan, W. G., P. St.-Amand, and L. A. Burkardt (1971): An evaluation of ice nuclei generator systems. Nature, 232, pp. 113-114. Hindman, E. E., II, S. D. Elliott, Jr., W. G. Finnegan, and B. T. Patten (1971): Response of STORMFURY cloudline cumulus to Agl and 2 Agl-Nal ice nuclei from a solution-combustion generator. Proo. Int. Conf. Wea. Mod., 6-16 September 1971, Canberra, Australia, pp. 130-134. Patten, B. T., J. D. McFadden, and H. A. Friedman (1971): Airborne cloud seeding systems developed and utilized by the NOAA Research Flight Facility. Proo. Int. Conf. Wea. Mod., 6-16 September 1971, Canberra, Australia, pp. 358-360. St.-Amand, P., D. W. Reed, T. L. Wright, and S. D. Elliott, Jr. (1971): GROMET II - Rainfall augmentation in the Philippine Islands. NWC TP 5097 (Naval Weapons Center, China Lake, Ca.), 110 pp. St.-Amand, P., and S. D. Elliott, Jr. (1972): How to seed cumulus clouds. Submitted to J. Wea. Mod. February 1972. Todd, C. J. (1965): Ice crystal development in a seeded cumulus cloud. J. Atmos. Soi., 4, pp. 70-78. 438 APPENDIX I SOME STATISTICAL CHARACTERISTICS OF THE HURRICANE EYE AND MINIMUM SEA-LEVEL PRESSURE Robert C. Sheets National Hurricane Research Laboratory INTRODUCTION This study is basically concerned with the hurricane radar eye and the minimum sea-level pressure (SLP) and their correlations with other parameters. These include the correlation of the size of the radar eye with the location of the storm, the minimum SLP, and the maximum wind speed. Frequency distributions of the changes of minimum SLP are also given. The data used in this study were obtained from aircraft reconnaissance and land-based reports received at the National Hurricane Center during 1961 through 1968. Since more reports were received for some storms than others, some bias was probably introduced. However, computations and graphs were made using all cases reported for 1961 to 1968, regardless of the time of the observation, and then a separate set of computations was made using data recorded at a maximum of every 6 hours. If more than one observation was made during a given 6-hour period, only the most complete one obtained during that period was used. The results showed that only very small differences existed between the two sets of data, except for a decrease in the number of cases for the 6-hour data. In most cases, the sample size was relatively large and the bias introduced probably does not appreciably affect the final results. In addition to the abovementioned portion of this study, six storms (GINNY '63, DORA '64, GLADYS '64, BETSY '65, FAITH '66, and INEZ '66) were chosen for special attention where nearly continuous data were available for an extended period. Graphs and scatter diagrams of the radar eye diameter versus minimum SLP and maximum wind speeds, and minimum SLP versus maximum wind speeds for the individual storms, as well as a composite of the data were constructed. In addition, graphs were constructed of the eye diameter, minimum SLP, and maximum wind speeds versus time for the individual storms, and of the 6-, 12-, and 24- hour eye diameter changes for a composite of the six storms. Frequency graphs of the eye diameter for filling and deepening cases were also con- structed, first from the actual reports, and then from the time-versus-eye diameter graphs, with values taken at 3-hour intervals. 439 FREQUENCY DISTRIBUTIONS FOR 1961 to 1968 The frequency distribution of the radar eye diameter for all the re- corded cases studied from 1961 through 1968 is shown in figure 1. A total of 535 observations are included. The eye diameters ranged from 5 to 120 n miles for these cases with a mean of 28.425 n miles and a standard de- viation of 17.2 n miles. The frequency distributions of eye diameters for selected minimum SLP categories are shown in figure 2. There appears to be little correla- tion between the minimum SLP and the eye diameter, because the computed correlation coefficient was only 0.012. The mean value was 26.20 n miles, with a standard deviation of 13.76 n miles, for the 109 recorded cases with a minimum SLP greater than 990 mb. A peak frequency of 25 to 30 n miles is shown, and the majority of the cases are concentrated within 10 to 45 n miles. For the cases with recorded minimum SLP greater than 970 mb but less than 990 mb, the mean eye diameter was 29.25 n miles, with a standard deviation of 18.347 n miles. No distinct peak frequency existed and a few large eye diameters are indicated. The mean eye diameter for those storms with minimum SLP greater than or equal to 950 mb, but less 100 Z90 2 80 ikj 70 "SO Ld ^50 40 30 20 10 . 1 I T FREQUENCY DISTRIBUTION OF "EYE" DIAMETERS 535 CASES (1961-1968) m m m MEAN=28425 STD. DEV.= 17.203 m I m m I 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 105 NUMBER OF CASES Figure 1. Frequency distribution of "eye" diameters -535 cases (1961 - 1968). 440 50 aiiiffZZ -i 1 1 1 1 1 1 1 r SEA LEVEL PRESSURE i 990 MB 109 CASES MEAN 26.2 NM STANDARD DEVIATION 13.6 J&&ZZES23 " ' * f.vu J L. <10 IS 20 25 30 36 40 45 50 35 60 65 70 75 80 65 90 95 100 105 110 115 120 125 EYE DIAMETER (NM.) °15 Lu -i — i — i — i — i — i — i — i — i — r o:57 m ;^;:Z yA i — i — i — " — i — " — i — i — r 970 MB. < SEA LEVEL PRESSURE < 990 MB. 134 CASES MEAN 29.2 NM. STANDARD DEVIATION 16.3 yhm//v-^ tmttTI <10 15 20 25 30 35 40 45 50 55 60 65 70 75 60 85 90 95 100 105 110 115 120 125 EYE DIAMETER (NM.) -r T T T I m^mm m ^ML 950 MB. 990 MB 206 CASES MEAN 52.2 KT STANDARD DEVIATION 17.1 35 40 45 50 55 60 65 70 75 80 85 90 95 100 105 110 115 120 125 130 135 140 145 150 MAXIMUM WIND SPEED (KT.) 30 1 ' 1 r "> 1 r- -i 1 r '20- W25H UJ co . o 15 Ol0- 5 — I — ' — I — ' — I — ' — T 950 MB S SEA LEVEL PRESSURES < 970 MB 113 CASES MEAN 95.1 KT STANDARD DEVIATION 17.4 35 40 45 50 55 60 65 70 75 80 85 90 95 100 105 110 115 120 125 130 135 140 145 150 MAXIMUM WIND SPEED (KT.) T ' — I — ' — I — ' — r 20 CASES MEAN 110.3 KT STANDARD DEVIATION 16 6 J I I I L ■< — r i — ' — i — • — i — « — i — ' — SEA LEVEL PRESSURES < 950 MB n r ^w^^t.^: ¥//A^F77a zM 35 40 45 50 55 60 65 70 75 80 85 90 95 100 105 110 115 120 125 130 135 140 145 150 MAXIMUM WIND SPEED (KT.) Figure 3. Frequency distribution of maximum wind speeds (stratified by sea level pressures; 1961 - 1968) correlation coefficient -0.789. several investigators often use the minimum SLP to measure storm intensity, because it is generally more conservative and less dependent upon the ob- server (Holliday, 1969). However, even with these errors, a high negative correlation exists between the minimum SLP and the maximum wind speed. Thus, the large sample tends to minimize the effects of random estimation errors and offers further evidence of the validity of using the more con- servative pressure measurements to evaluate storm intensity and associated maximum wind speeds. Figures 4 and 5 show a frequency distribution of latitudes and longi- tudes of the storm centers stratified by surface pressure. The correla- tion between the latitude and the surface pressure is quite small (-0.09112), 442 40 1 — 35 - 30 - 25 - 20 - 15 - 10 - 5 - 0 1 — i — i — r 1 1 1 1 1 r "i 1 r -i 1 1 1 1 1 r SEA LEVEL PRESSURES 2 990 MB 268 CASES MEAN 24.7 CN) STANDARD DEVIATION 7.69 -I 1 l_ ZL W/, Y'A VA X'\ \A \~TY~n 25 50 7.5 10.0 125 150 175 200 225 25.0 27.5 3O0 325 350 37.5 40.0 425 45j0 47.5 5QO 525 550 575 60.0 LATITUDE (°N) 30 1 — 25- 20 - 15- co10" 111 V) 5 " < o o "i i i i i i i r ~i — r i — r i — r V7X...:X/A...-V// y/aky/a 970MB £ SEA LEVEL PRESSURES < 990 MB 204 CASES MEAN 27.4 CN) STANDARD DEVIATION 6 75 J I l_ 25 SO 75 lOO 125 150 17.5 200 225 250 275 300 325 350 375 400 425 450 475 500 525 550 575 6Q0 LATITUDE CN) o or HJ55 CD 5 50 i — ' — r t — i — r t — ' — r i — i — r 1 r ' 45 40 35 30 ■ 25 20 15 10 - 5 0 A 950 MB £ SEA LEVEL PRESSURES < 970 MB" 172 CASES MEAN 26 6 CN) STANDARD DEVIATION 5.39 ifc. TtTTT, 25 50 75 100 125 15D 175 200 225 250 27.5 30O 325 350 375 400 425 450 475 500 525 55D 575 60© LATITUDE CN) -I — I — I — I- SEA LEVEL PRESSURES O50MB _l I I I I 1 r- 1 r T 1 \ " 1 1 1 r 32 CASES MEAN 24.3 CN) STANDARD DEVIATION 5 32 _l 1- _l l_ 25 50 75 100 125 150 175 200 225 250 275 30O 325 350 375 400 425 450 475 500 525 550 575 600 LATITUDE (°N) Figure 4. Frequency distribution of storm latitudes (stratified by sea level pressures: 1961 to 1968) correlation coefficient -0.091. 443 45 40 35 50 25 20 15 10 5 o ( 40 35 0->3° UJ CO 25 < O20 i. 15 cr 5 UJ CD 0 -i 1 r -i 1 1 1 r n i i i i i i i i r SEA LEVEL PRESSURES > 990 MB 268 CASES MEAN 65.9 (°W) STANDARD DEVIATION 13 9 _i I r,v,-i t-W/A _ii^ 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 105 110 115 LONGITUDE (°W) 1 1 1 1 1 1 1 \ 1 ' 1 ' . w Vs Vs _970MB 110 KTS 37 CASES MEAN 22 49 STD DEV 14 02 Figure 6. Frequency distribution of "eye" diameters (stratified by max- imum wind speeds 1961 - 1968) cor- relation coefficient -0.076. <10 30 40 50 60 70 EYE DIAMETER (NM ) 80 90 >100 445 15 10 5 0 20 15 w < ° 5 UJ (D25 LATITUDE <20eN 53 CASES MEAN 17.28 STD. DEV 8.45 [Hfescaasa. ™ •«». **...- p— ■ '.': ■Ti-'- ■ ■- . t> "Xv •'.; '"- K(', - & >-' '/.' v : - ■, /■ •5' 27.5° N*- LATITUDE £ 20°N 117 CASES MEAN 28.01 STD. DEV 12.88 EBB 1 J* ■J-C ^,-;. jivg -;-'•/ ^ £j i Jv ah 35° N » LATITUDE a 27.5° N 92 CASES MEAN 34.28 STD. DEV. 19 12 frlYK as.re:B'l _HS LATITUDE i 35° N 42 CASES MEAN 32.00 STD. DEV 25.92 Baca "^ < 10 20 30 40 50 60 70 80 90 >100 EYE DIAMETER (NM) Figure 7. Frequency distribution of "eye" diameters (stratified by latitude 1961 - 1968) cor- relation coefficient 0.250. LONGITUDE > 60° W 128 CASES MEAN 27.99 STD. DEV. 18.93 ill ESaaai w 90°W> LONGITUDE > 75*W 79 CASES MEAN 31.27 STD. DEV. 17.73 rc?ra LONGITUDE > 90* W 27 CASES MEAN 23.24 STD. DEV. 11.16 <10 _L ' ' ' ■ ■ 40 50 60 70 80 EYE DIAMETER (NM) 90 >100 446 with the eye diameter (0.25007) indicates that smaller ones are more pre- valent at lower latitudes. This condition is indicated by the mean values for the first three latitude categories. It is also interesting to note the increase in the eye diameter scatter (standard deviation) that occurs at higher latitudes. In fact, the standard deviation of eye diameter at latitudes higher than 35°N is three times as large as at latitudes lower than 20°N with an equivalent number of cases in both categories. Almost no correlation (0.01009) is shown for the storm longitude versus the eye diameter, and the standard deviations range from 11.16 to 18.93 n miles. Figures 6 through 8 were constructed using a maximum of one observation during a 6-hour period. Another set of computations and graphs was made using data recorded at a maximum of eyery 6 hours, as discussed in the introduction; however, this set was also limited to those cases having a minimum surface pres- sure less than 1000 mb. Again, only small changes were observed when com- paring this set df data with that illustrated in figures 2 through 8. The correlation coefficients (-0.31438 to -0.28537) showed a slight decrease for longitude with surface pressure, a slight increase (-0.07564 to -0.12237) for eye diameter with maximum wind speed, and similar small changes in other values. SELECTED STORMS Six storms were selected for special consideration. The reason for the particular selections were the availability of considerable and nearly continuous data from reasonably well -developed storms. Scatter diagrams were constructed of the eye diameter versus the central pressure, the eye diameter versus the maximum wind speeds, and the central pressure versus the maximum wind speeds for the individual storms as well as a composite of the cases. Only the composites (figs. 9, 10, and 11) are presented here, but the values for each storm are indicated by a different symbol. Some congestion is present, but not only do the composites show a large scatter of data points, but also the data points for each individual storm are scattered as observed. Figure 9 is a scatter diagram of the eye diameter versus the surface pressure. Note that little correlation exists between these two parameters as was indicated for a composite of all the storm data from 1961 through 1968 (fig. 2). A scatter diagram of the eye diameter versus the maximum wind speeds is shown in figure 10. Again little correlation exists between these two parameters as was indicated earlier. A large scatter of data points occurs for the composite as well as for the surface pressure versus the maximum wind speeds (fig. 11), which shows a relatively high correlation between these two parameters. However, some degree of scatter is observed for both the composite and the individual storms. Again, the problems associated with maximum wind speed measurements that were discussed earlier also apply here. 447 1012 1000 ao988 2 UJ DC ^976 CO UJ OC a. —J 964 UJ > UJ S952 940 928. ■ o D —t — I — r • — GINNY ■ -DORA A — GLADYS X — BETSY o — FAITH D — INEZ B 0*4 0 ■ 0 ■ * t X • n ° I X X X A U 0 0 ' ■ I : °s s i . " -J , 0 n ? ■ * " X 0 lb V D i g - ;5 D 4 a x x x x x ■ a I °in I i I ■ l.i | | | | | | 1 L 10 20 30 40 50 60 70 80 90 100 EYE DIAMETER (NM) Figure 9. Eye diameters versus minimum sea level pressures for six selected storms. Graphs of central pressures, maximum wind speed, and eye diameters were also plotted versus time for the individual storms. The central pressure graphs (fig. 12) were then used to determine when the storm was deepening or filling. Trends were taken into account and small fluctua- tions over short time periods were smoothed. To be included, the pres- sure trend had to persist for 6 hours or more. One case that met these requirements was ignored, because it was based on one observation that was inconsistent with the distinctive trend shown by other observations for this period. A frequency graph of the eye diameters was then con- structed for the filling and deepening cases for a composite of the data collected from the six storms mentioned above (fig. 13). The data values were obtained from the eye diameter versus time graph at 3-hour intervals. No significant difference was observed between the two cases with mean eye diameters of 29.4 and 29.6 n miles for the filling and deepening cases, respectively. 448 140 120 100 o UJ bJ a. • — ■ Total Guam Philippines 1961 4 2 6 1962 5 1 6 1963 5 1 6 1964 4 0 4 1965 5 1 6 1966 6 0 6 1967 7 0 7 1968 6 2 8 1969 3 0 3 1970 2 1 3 47 (4.71) 8 (0.81) 55 (5.51) Average number of storms per year. Guam-Philippines combination than in the Guam-Okinawa combination, because of the seeding restrictions applied for storms approaching the islands of Japan and Okinawa. Note that 2 of the 10 years would have had only three eligible typhoons. Of the 55 typhoons considered eligible for seeding in the Guam/Phil- ippines study, 29 later moved on to enter countries on the western bound- ary of the Pacific. The average estimated time after seeding until land- fall ranges from 52 to 91 hours (see table 4). Current data indicate that effects of modification are over by 12 hours after seeding. Table 4. Typhoons Eligible for Seeding that had Later Landfall First Mnmhav Total Eligible Average time After Seeding Landfall INUmper (%) Until First Landfall (hrs) 77 76 84 91 52 Japan 8 14.5 Ryuku Islands 6 11.0 China 1 2.0 Taiwan 8 14.5 Philippine Islands 6 29* 11.0 53.0 *19 of these had later landfall in China. 459 Table 5 shows the increase of eligible storms achieved by shortening the alert/response time from 48 to 24 hours. This increase would be even larger under conditions when storm data coverage in an area is limited. SUMMARY The application of more stringent eligibility rules for Japan and Ok- inawa does not significantly alter the desirability of attempting typhoon modification experiments in the Western Pacific if certain operating re- strictions can be relaxed. It does require that longer range aircraft be available, or that air- craft time on station in the storm be decreased. The ability of STORMFURY groups to react within 24 hours after alert for an experiment probably re- quires a different type of seeder aircraft from the jet attack aircraft used in the Atlantic experiments. At least, it requires that the seeder aircraft be at the operating base routinely. Use of larger reconnaissance or research aircraft as seeders may offer many advantages. These aircraft have increased endurance, better navigation and data collection capabili- ties, and can carry Project officials to direct the seeding runs. In addi- tion, they do not require the sophisticated airborne radar vector and moni- tor control required for small jet aircraft in a typhoon environment. Ad- ditional seeder aircraft, such as the jets previously used, would no longer be required if aircraft already committed to typhoon reconnaissance and research could be used. Table 5. Comparison of Seeding Opportunities with Shortened Alert/Response Time Alert/Response Operation Bases Seedin?, Opportunities Time K Number 9* hm.«c Guam and Okinawa 5.3 ^ nours Guam and Philippines 5.5 ,n hnil~c Guam and Okinawa 4.7 ^° nours Guam and Philippines 4.9 460 APPENDIX K COMPARISONS OF RAINFALL AND RADAR MEASUREMENTS IN TROPICAL STORM FELICE AND HURRICANE DEBBIE W. D. Scott and C. K. Dossett National Hurricane Research Laboratory INTRODUCTION Cloud activity is a primary manifestation of the activity of a tropi- cal storm. This cloud activity is recognized, in part, through radar signatures or radar PPI displays. The meaning of these pictures is pri- marily qualitative, but a semiquantitative estimate of cloud rainwater can be obtained using the classical Z-R relationships. This paper pre- sents such Z-R relationships derived from in situ measurements of rain- drop numbers and sizes. Also, data are presented which show the structure of the rainwater content in the storms. RAINDROP MEASUREMENTS Raindrops were measured using the aluminum foil impactor manufac- tured by Meteorology Research, Inc. (see Hindman, 1970; or Scott and Dossett, 1971). The device simply records the impressions made when drops import on an aluminum foil held into the airstream; it is sensitive to drops above 200y in size.5 The basic data are histograms of raindrop sizes; figures la, lb, and lc show the tallied drop distributions for data acquired in Tropical Storm Felice on 15 September 1970, and Hurricane Deb- bie on 20 August 1969. The data from Tropical Storm Felice (fig. la and lb) definitely are fit by the Marshall -Palmer relationship (i.e., N = N0 e-Ad)s6 but the data from Hurricane Debbie (fig. lc) show large deviations The reasons for this result are not clear, but may result from inadequate counting of the small drops during the analysis of the data from Debbie because of imperfections and corrosion on the foil. Generally, however, the data from both storms give volume mean drop sizes of about 1 mm, total drop concentrations of about 1000/m3, and (slopes) of about 15/cm, though the characteristics vary through the storm (note the difference in slope between figs, la and lb). 5 In this paper, drops over 200y in diameter are considered to be raindrops. 6The exponential distribution is the form used classically by Marshall and Palmer (1948) and has been referred to throughout the paper as the Marshall Palmer distribution. By this usage we do not intend to imply that any re- lationship exists between the parameters of the distribution as observed in tropical clouds and rainfall rates observed at the surface. 461 < > cc. UJ LU FLIGHT NO. 700915-A TIME 144418 Z-144454Z MARSHALL- PALMER REGRESSION LINE < Q E E .? cc UJ Q. 100- 441 TOTAL DROPS < > UJ or UJ FLIGHT NO. 700915-A TIME 1434592 -143528 Z MARSHALL — PALMER REGRESSION LINE 295 TOTAL DROPS 1 2 3 4 (a) DROP DIAMETER (mm) 1 2 3 4 (b) DROP DIAMETER (mm) DEBBIE FLIGHT NO. 690820 Bl TIME 1U303Z-111310Z (c) 239 TOTAL DROPS 12 3 4 DROP DIAMETER (mm) Figure 1. Histogram of drop sizes in Tropical Storm Felice (a) and (b)> and in Hurricane Debbie (c). 462 The basic drop measurements were used to calculate the rainfall rates and radar reflectivity factors shown in figures 2 and 3. The rainfall rates were calculated using the indicated drop sizes and corresponding terminal velocities. The radar reflectivity factors, Z, were calculated using the classical Raleigh scattering formula, with scattering propor- tional to the sixth power of the drop size. The least-squares line is a good fit and, in a statistical sense, lends credance to the use of air- borne radars for the measurements of rainfall in hurricanes. The data in figure 2 were acquired at a low level (1300 to 7500 ft). Similar data at approximately 14,000 ft show exactly the same multipli- cative coefficient (300), but a slightly larger exponent 01.3). Note that the line for Hurricane Debbie (fig. 3) has a coefficient of 260 and an ex- ponent of 1.35. The Z-R correlation for summertime cumulus in South Flor- ida derived by Gerrish and Hiser (see Woodley, 1970) is Z = 300 R1'1* (1) The relationship is based on measurements made with a raindrop camera for the Illinois State Water Survey. This correlation (Eq. 1) is \/ery similar to the present study. 106 1 1 1 1 TROPICAL STORM "FELICE" FLIGHT NO. 700915-B / "e ^ iO5 Jr ~ E E 7k JJJ )i/» or Z = 16, 200 R1 24 j£x o 2 _ or Z =~30 0 Ri24(R in mm/hr.) xjr* — >- /& >- >».■/ > XX X i- 103 '- Ld * J* I L_ "*/* LlI CE K iO2 < . y • - o < cr y • 7500FT 153550- 161711 G MT Y x 1600FT 174210 -180456G MT Z A 1600FT 183720 -185640GMT io1 - */ ■ 1300FT * 1 1 1 182200- 162728 GMT- 1 10: < io4 > 10s I- o 102 DC < Q < 1 ! TROPICAL 1 T STORM "DEBBIE" FLIGHT NO 690820 Bl , / — MS — Z= 20,400 R13s •M* or 0 s^ Z= 260 R135 (mm/hr.) °-j{'1 " :W + ~ *'P$ ' ~ •*j6*»* * s /m3 ) STORM / CENTER 25 30 NM 50 NM RHI ECHO CONTOUR -FIRST ISOECHO 50NM 25 '25^20 SW QUADRANT OF STORM Figure 8. RHI echo height contours in thousands of feet. to the storm center. These data show the basic trends already presented and also the life history already inferred. That is, sample A (fig. 12, solid line) shows an early stage of cloud development when the rainwater resides in the updraft. In a growing cloud, the presence of relatively large amounts of liquid water in a supercooled state suggests the possi- bility of a marked effect from seeding. Sample D (dotted line) shows the same location at a later time (^20 minutes later) when only a downdraft loaded by the rainwater remains. The cells of figure 13 are also in a young state as indicated by the 4 m/sec updrafts when rainwater samples H (dashed) and I (dotted) were taken; but the cells were "dying" a half-hour later when sample J (solid line) was taken. However, at this time, a new 469 NEW ORLEANS RADAR SEPTEMBER 15,1970 16002 200HM Figure 9. Flight track and sampling locations (Al, A2} A3) in Tropical Storm Felice (700915A1). cell with a large updraft appeared 14 n miles from the storm center with no associated rainwater (see lower curve). Through the temporal and spa- cial coherence of the data in figure 14 (G, K) are not as good, similar 470 400 200 0 £-200 -400 10 8 6 4 2 O 0 1.5 i.O 0.5 5.0 4.0 3.0 2.0 1.0 " VOLUME MEAN DROP SIZE RAINWATER ^^J J L _L 65 55 50 45 40 35 30 25 20 15 DISTANCE (N.M.) FROM STORM CENTER 12.5 Figure 10. Variations of updraft velocities and rainwater characteristics along a high level (14,000 ft) radial leg in Tropical Storm Felice (700915A1). 471 HURRICANE "DEBBIE" NAVY APS- 20 N FOR 1050Z- 11202 \ AUGUST 20, 1969 RADAR DATA Direction of Storm Movement Figure 11. Sampling locations in Hurricane Debbie (690820B1) 472 NAUTICAL MILES FROM STORM CENTER A 0917-0918 RAIN WATER (GMS/ Wfl) B 0922-0924 C 0952-0951 D 0956-0954 1 :V Figure 12. Rainwater and updraft velocities in Hurricane Debbie (690820B1) northwest quadrant. trends are present. In the double eyewall structure, note the remarkable persistence of the rainwater "cells" in all three figures (12, 13, and 14) at 10 to 13 and 24 to 27 n miles from Debbie's center. All in all, this effect is quite surprising and points to the dominance of mesoscale convergence in prescribing the development of the clouds. Figure 13. Rainwater and updraft velocities in Hurricane Debbie (690820B1) southwest quadrant. UPDRAFT VELOCITY (M/SEC) 473 -1056 -1101 1148-1144 Figure 14. Rainwater and updraft velocities in Hurricane Debbie (690820B1) northeast quadrant. CONCLUSIONS In accepting the above results, remember that for safety the air- craft tend to avoid the hard cells in the storm, so that the data cannot be considered as statistically meaningful. Also, the data are from only two storms and should not be considered of general validity until they are corroborated. Nonetheless, the results indicate: 1 The data rainfall Marshall- tered. can be used on a statistical basis to indicate Even when the raindrop spectra are far from Palmer, the Z-R relation is only slightly al- 2. The rainwater structure of the storms shows deterministic trends that are relatively stationary on a larger scale. That is, there are areas in the hurricane approximately 5 n miles in horizontal extent that show cloud develop- ment with a remarkable persistence. In terms of STORMFURY operations, the observations of a separation between the updrafts and the rainwater are significant. In this respect, inferring the location of an individual updraft from the radar data can be quite misleading. The modification of storm clouds should take ac- count of the updraft structure and the rainwater structure when seeding decisions are made. 474 REFERENCES Black, P. G., H. V. Senn, and C. L. Courtright (1972): Airborne radar ob- servations of eye configuration changes, bright band distribution, and precipitation tilt during the 1969 multiple seeding experiments in Hurricane Debbie, Monthly Weather Review, 100, pp. 208-217. Carlson, T. N., and R. C. Sheets (1971): Comparison of draft scale ver- tical velocities computed from gust probe and conventional data col- lected by a DC-6 aircraft, NOAA Tech. Memo. ERL NHRL-91. Hawkins, H. F., and D. I. Rubsam (1968): Hurricane Hilda, 1964: II. Struc- ture and budgets of the hurricane on October 1, 1964, Monthly Weather Review , 96, pp. 617-636. Hindman, E. E., II (1970): Cloud particle samples and water contents from a 1969 STORMFURY cloud! ine cumulus, Project STORMFURY Annual Report, 1969, Appendix F. Marshall, J. S., and W. McK. Palmer (1948): The distribution of raindrops with size, J. Met., 5, pp. 165-166. Scott, W. D., and C. K. Dossett (1971): An estimate of the fraction ice in tropical storms, Vrojeot STORMFURY Annual Report, 1970, Appendix G. Woodley, W. L. (1970): Precipitation results from a pyrotechnic cumulus seeding experiment, Journal of Applied Meteorology , 9, pp. 242-257. & U.S. GOVERNMENT PRINTING OFFICE: 1973—780-381 475 PENN STATE UNIVERSITY LIBRARIES i mini ii ii A0DDD75DqTbSS