mm ■ »"■'■■.■•, B Sffili mm) Jfl. n ■Hi DUDLEY KN( NAVAL POS MONTE! S3 940 «ONTER?ySJGc^THSCHO0L ,-' {4~' l NPS68-81-006 NAVAL POSTGRADUATE SCHOOL Monterey, California THESIS RAPID OCEANOGRAPHIC DATA GATHERING: SOME PROBLEMS IN USING REMOTE SENSING TO DETERMINE THE HORIZONTAL AND VERTICAL THERMAL DISTRIBUTIONS IN THE NORTHEAST PACIFIC OCEAN by Glenn W. Lundell September 1981 Thesis Advisor: G . H . Jung Approved for public release; distribution unlimited. Prepared for: Naval Ocean Systems Center Code 5 31 San Dieao, California 92152 T 2 0 2 3 0 SECURITY CLASSIFICATION OF THIS RAGE fWhon Dmim Bntorod) REPORT DOCUMENTATION PAGE 1 REPORT NUUIIR NPS 68-81-006 2. GOVT ACCESSION NO 4 TITLE mna Subtttlo) Rapid Oceancgraphic Data Gathering: Sore Problems in Using Remote Sensing to Determine the Horizontal and Vertical Thermal Distributions in tfre tfortbeasfc p^-j-f-j^ rv-&^ 7. AuTmOR,» Glenn W. Lundell • performing organization name ano aooress Naval Postgraduate School Monterey, California 9 3940 II CONTROLLING OFFICE NAME ANO AOORESS Naval Postgraduate School Monterey, California 93940 TT WONITORING AGENCY NAME * AOORESSff/ dlllotmnt from Controlling Ollleo) READ INSTRUCTIONS BEFORE COMPLETING FORM 1 RECIPIENT'S CAT »LQG KuM! S,/VP1 °f ?EPO£L* "»e.RlOO COVERED Master's Thesis; September 1981 «. PERFORMING ORG. REPORT Nu Mien • • CONTRACT OR GRANT NOMBERr.J 10. PROGRAM ELEMENT. PROJECT TASK AREA 4 WORK UNIT NUMBERS N6600181 WR00082 12 REPORT DATE September 1981 13. NUMBER OF PAGES 188 IS. SECURITY CLASS, tot tht, report; Unclassified 1S«. DECLASSIFICATION/ DOWNGRADING SCHEDULE t«. DISTRIBUTION STATEMENT oi thf Koport) Approved for public release; distribution unlimited. 17. DISTRIBUTION STATEMENT at tho omotroel ontorod In Block 20, II dltlotont frowi Roport) IS SUPPLEMENTARY NOTES 1»- KEY WORDS (Contlnum on roworoo ildo II noeooomrr mnd Idmntity *r kloek iwjmbor) Satellite; NOAA-6; Remote Sensing; GOSSTCOMP; Northeast Pacific Ocean; Thermal Structure; Sea Surface Temperature; Subsurface Thermal Structure; Fronts; Eddies; AXBT 20. ABSTRACT (Contlnum an -•»•«■•• mtdo H nocooomrr and Identity *r block .•.,< NOAA-6 satellite AVHRR data and AXBT data were collected in the Northeast Pacific Ocean in late 198 0 as part of the Naval Postgraduate School-sponsored Acoustic Storm Transfer and Response Experiment which was in turn part of the U.S . -Canadian Storm Trans- fer and Response Experiment (STREX) . Some of the problems in trans- ferring AXBT geographical positions to satellite images were solved by designing a computer program with accuracies of less than 2 pixe] DD FORM JAN 71 1473 EDITION OF < MOV •• IS OBSOLETE S/N 010 J-014- 6601 UNCLASSIFIED SECURITY CLASSIFICATION OF THIS PAOE (9hon Dot* Kntotod) UNCLASSIFIED OCUWTV CIAMI>IC«T18II Of Twit »«Q«f»— fWlj t«<*~« #2 0 - ABSTRACT - (CONTINUED) Thermal comparisons were made between AXBT, NOAA-6, and GOSSTCOMP data with the result that NOAA-6 data was on the average 2.9°C colder than AXBT data and 3.2°C colder than GOSSTCOMP data. Linear regression methods reduced to 0.3°C the difference between NOAA-6 and AXBT data. Use of this method over a period of 15 days produced a mean error of 0.5°C. Although NOAA-6 cannot sense directly the subsurface thermal structure, it is excellent for observing surface manifestations of horizontal thermal features. Further investigation into using satellite data as the basis of an empirical relationship between the surface temperature and the subsurface vertical thermal structure is warranted. DD Form 1473 31*02-014-6601 UNCLASSIFIED iicu««*v clami*» great circle distance between node and the subsatellite point great circle distance between node and subsatellite point 15 ACKNOWLEDGEMENTS The opportunity to work with experts in fields that had been only an avocation prior to my arrival at the Naval Postgradu- ate School has been a rewarding and enjoyable experience. The guidance and support of my thesis advisor, Dr. Glenn Jung, Professor, Department of Oceanography, and LCDR Cal Dunlap, Assistant Professor, Department of Oceanography, has been tremendous, especially when the proverbial "light at the end of the tunnel" was pretty dim. This thesis presented many problems that may have been left unresolved were it not for the expertise and understanding of Dr. James Mueller, Adjunct Professor, Department of Oceanography His conceptualizations of the overall geometric problems in correcting satellite imagery were extremely helpful. In addi- tion, satellite support came from Mr. Bob Wrigley, Technology Applications Branch at NASA Ames Research Center, who gave un- selfishly of his time both to ensure access to the NASA-IDIMS system and to coach me in its use. Mr. Larry Breaker, NOAA- NESS, was kind enough to supply the satellite tapes and the ephemeris data in addition to answering my many questions on NOAA-6 operations. Finally, a special thanks goes to my colleagues in the Environmental Acoustic Research Group, LCDR Thorn Holt and Mr. Gene Brown (NAVOCEANO) , who not only arranged the P-3C flights but also provided a sounding board on which to bounce ideas . 16 I. INTRODUCTION The collection of oceanographic data has always been time consuming and expensive. With the development of environmental satellites, a method for the rapid gathering of oceanographic data was available to complement that data gathered on re- search cruises. Unfortunately, there are still some problems in using satellite data, such as the effects of the atmosphere on the radiative transmission path and the effect of the geo- metric distortion of thermal features found on polar-orbiting satellite images as from NOAA-6. This thesis is part of a series of on-going studies at the Naval Postgraduate School by the Department of Oceanography Environmental Acoustic Research Group. The overall goal of the Group is to continue the development of those aspects of acoustical oceanography that have a significant effect on naval tactical applications. In pursuit of this goal, the Group was a participant in the joint U .S . -Canadian Storm Transfer and Response Experiment (STREX) held in the fall of 1980 in the Northeast Pacific Ocean. The Group is particularly interested in investigating whether or not satellites can fulfill the role presently played by ships and/or air-dropped expendable bathythermographs (AXBT) in gathering sea surface temperature data for use in forecasting the ocean's thermal structure, using procedures similar to those developed by the U.S. Naval Oceanographic Office Antisubmarine Warfare Environmental 17 Prediction (ASWEPS) Program in the early 1960 's. If direct correlations could be found between satellite-derived sea sur- face temperatures and the vertical thermal structure, then a rapid method of surveying the world's oceans could result, with numerous naval ramifications. As a part of the experiment conducted by the Group under the title Acoustic Storm Transfer and Response Experiment (ASTREX) , this thesis was directed toward the examination of some of the problems in using satellites to observe the hori- zontal and vertical thermal distributions in the waters of the Northeast Pacific Ocean. Of particular interest was the problem of locating open-ocean geographic positions (ship, AXBT, etc.) on satellite images with as much accuracy as possible. A major portion of this thesis is devoted to this subject. If comparisons are to be made between satellite- obtained thermal values and ground-truth thermal values, then an elimination of location errors between the two media makes the results that much more significant. The reasons why NOAA- 6 satellite imagery is distorted and a method to eliminate any location errors successfully are presented below. In addi- tion, once location accuracy was assured, various meaningful comparisons were made between satellite, bathythermograph, and GOSSTCOMP (Global Operational Sea Surface Temperature Computation) data. The results of these comparisons are also presented. 18 II. REMOTE SENSING IN QCEANOGRAPHIC RESEARCH In 1870 and 1879 respectively, the authors Edward Everett Hale and Jules Verne wrote about placing an artificial satellite in orbit, Hale using a huge waterpowered flywheel and Verne a gun of sufficient muzzle velocity (Corliss, 1967). Hale envisioned the satellite as an aid to both navigation and communication. Up until 19 35, the idea of launching man into space remained the ideas of small amateur organizations with some notable exceptions, such as the efforts of Robert Goddard in Auburn, Massachusetts, in the 192 0*s. With the stirrings of war in Europe in the late 19 30 's, the now infamous V-2 rockets were developed and launched from Peenemuende by a team of German scientists including Wernher von Braun. Captured en masse in 1945 by the Allies, this group of scientists and their hardware were transferred to the United States. The Army Air Force subsequently commissioned a study by Rand Corporation, who reported in 194 6 for the first time that an earth-orbiting artificial satellite could be used scientifically in the fields of meteorology, biology, and com- munications (Corliss, 1967). This launched a nine-year effort by lobbyists both inside and outside the government culminating in President Eisenhower's announcement on 29 July 1955 that the U.S. would launch an earth-orbiting scientific satellite to investigate the environment. The Soviet Union announced similar plans on the following day. October 4, 1957 saw the 19 launch of Sputnik-I by Russia followed on 31 January 1958 by the launch of the first U.S. satellite, Explorer-I . Since that time, hundreds of scientific satellites have been launched in order to investigate topics from numerous fields. One of these, NOAA-6, was the satellite whose data were used on this project. NOAA-6 is an earth-orbiting, environment-sensing spacecraft with applications in meteorology and oceanography. This section reviews the use of satellites in oceanographic studies, provides a description of the operation of NOAA-6, and briefly summarizes the normal oceanographic conditions for that region of the Northeast Pacific Ocean where the experi- ment took place. A. OCEANOGRAPHIC CHARACTERISTICS OF THE PROJECT AREA The area chosen for this project encompassed that region of the Northeast Pacific Ocean between latitudes 40 N and 50 N and between longitudes 12 6 W and 139 W. See Figure 1. The oceanographic conditions of this region have been extensively studied by Tully (1961; 1964), Tabata (1964; 1965; 1978), and Roden (1975) among others. The reader is referred to these works for more detailed information as only a brief summary of their findings is described below. As seen in Figure 1, the project area is located mainly in an oceanic water mass transitional region between the Sub- arctic Water Mass, predominantly to the north of 4 5 N, and the Pacific Equatorial Water Mass, predominantly to the south of 2 3 N. The Subarctic Current flowing eastward along 45 N 20 Figure 1. Major surface currents and water masses of the Northeast Pacific Ocean (after Kibblewhite et al. , 1977) 170* W 60' 160* 150* 140* 130* N SUBARCTIC PACIFIC WATER ! 50* V ±L>8A j ^ORTH £C/f ^ 120* W 60* N / ASTREX PROJECT ^ AREA 50* curr&nT fe 10* N 170* W EQUATORIAL PACIFIC WATER 160* 150# 140* 130* 10* N 120* W 21 latitude divides on the northwest side of the project area into the Alaska Current, which circles counterclockwise to the north along the Canadian and Alaska coastlines; and into the California Current, which flows southward along the western coast of the United States. Observations from the project area would be expected to show the physical characteristics of the Subarctic Water Mass; however, those observations in the southern portion of the project area may be somewhat tempered by the colder offshore waters of the California Current. 1 . Thermal-Salinity Structure An excess of precipitation over evaporation of approxi- mately 25 cm/year (Tabata and Giovando, 19 62) has helped to create a layer of water extending to a depth of 100 meters in the Subarctic Water Mass that is isohaline during the winter months. A permanent halocline extending from 100 to between 200 and 300 meters exists in which the salinity increases by 1 °/0O to approximately 33.8 %Q and marks the maximum limit of seasonal effects (Tully, 1964) . The top of the permanent halocline at 100 meters also marks the maximum depth of the seasonal thermocline. During the summer, the thermocline forms between 25 and 50 meters, influenced heavily by wind mixing effects alone. With the coming of the fall and winter months with their intensive storms, the surface waters begin to cool considerably and both convection and wind mixing erode the thermocline until iso- thermal conditions exist to the top of the permanent halocline. This condition usually is reached by February at which point 22 the waters continue to cool until the end of March, when the heating season begins. See Figure 2 for a general depiction of the seasonal structure. Figures 3 and 4 show the expected mean thermal structure for the project area for the months of November and December. Figure 5 shows the annual surface salinity maxima while Figures 6 and 7 show the expected layer depths also for November and December. Both the Subarctic Current and the California Current have surface speeds less than one knot with volume transport averaging 10 to 15 million cubic meters per second (Knauss, 1978) . As a result, the water in the project area is exposed to constant climatic conditions over many months and has suffi- cient time to adjust to seasonal variations. During November and December, there is an expected net heat loss in the pro- 2 ject area ranging up to 400 g-cal/cm /day (Tabata, 1961) . This heat loss is directly responsible for the winter convective mixing process mentioned above; therefore, the vertical thermal structure is due more to the area's heat budget, storm cycle, and salinity layers than to any influx of new water. 2 . Internal Waves Internal waves have been shown to cause amplitude fluctuations of 4.5 to 5 meters at the level of the seasonal thermocline in the Northeast Pacific Ocean (Tabata and Giovando, 1962) . These oscillations may cause a periodic thickening of the thermocline from 10 to 50 meters due to phase differentials between the top and bottom of the thermocline (Tully, 1964) . These periodic fluctuations vary with depth as oscillations of 23 Figure 2 . Seasonal oceanographic structure of the Northeast Pacific Ocean (from Tully, 1964) JOC«- EXCESS P*EC1PITAT!CN- ■HEATINO SEASON- : DOLING SEASON- 'III' 'I, 'i 'i 1 1 X »H< «M«!«q \ 'll Miiiiiiiliili CsflWCttv* mitlnf 50- SEASCNAL THEPUOCUHE ^ '!||||||||ui mini i mimiiirtrai .-r ■PI • in* "» 1 fiy. t'ki/r'.'.'V/V.' ■'■"■>'•■.:•'•' -'Or.r r Intwrnal warts-..-- >?>-•".*' Vtf/» — imaionai ^ — * «P* MAY At* JUL AUG SEP OCT ««CV OEC -*»» FEB OCEANOGRAPHIC STRUCTURES SUBARCTIC PACIFIC OCEAN 24 u -p G O 03 0) QJ -r-\ S 0 H Q 0) Si 0) S.C > O G 2 -H en 0) M 3 Cn •H &4 o iO O lO O lO CVJ lO N O C\J (w) Hld3Q 25 U — -p r- o a\ u -p en U 3 -P id u a g e o 0) M -P IW Id O •H •P M 0) > c a) S M QJ 0) o tu Q id 4J id fd a> M -p o Q) •r-i O a 0) SZ ■P a; u 3 ■H En (tu) Hld3Q 26 Figure 5. Annual surface salinity of the project area (from Robinson, 1976) 27 Figure 6. November mean layer depth in project area (from Robinson, 1976) 28 Figure 7. December mean layer depth in project area (from Robinson, 1976) depths in feet 29 isotherms at the level of the halocline are five times greater than oscillations in the region of the thermocline. 3 . Fronts and Eddies The Subarctic Front, usually found between latitudes 40 N and 45 N, may be present in the center of this project area. This front is characterized by the lack of a density front in the upper 100 meters, by the region of the strongest surface baroclinic flow being to the south of the surface temperature and salinity fronts, and by the mixed layer depth extending to the top of the halocline at 100 meters on its northern edge (Roden, 197 5) . In the southern section of the project area, eddy formation may be present in that area in- fluenced by the California Current which has been described as a relatively shallow meandering current with alternating warm and cold tongues (Bernstein et al. , 1977). B. USE OF SATELLITES IN OCEAN THERMAL STUDIES In 1968, a study was done comparing satellite-obtained sea surface temperatures with monthly mean surface temperatures with the result that the satellite values were anywhere from 3 to 8.3 degrees C lower than the mean (LaViolette and Chabot, 1968) . The relative horizontal gradients observed in their satellite data, however, were fairly consistent with similar mean gradients from the historical data. This pattern of satellite-obtained sea surface temperatures being lower than the mean or the actually observed sea surface temperatures persists until today, except that technological advances have 30 reduced the differences in temperature between the two sets of data so they now are between 0.5 to 3.0 degrees C (Rao et al. , 1972; Brower et al . , 1975; McMillin, 1975; Cogan and Willard, 1976; Barnett et al. , 1977; Tabata and Gower, 1980). Prior to 1972, the oceanographic use of satellite-obtained sea surface temperature was severely limited by both the en- gineering characteristics of the satellite radiometers and by the environmental aspects causing atmospheric attenuation. Large instantaneous f ields-of-view (IFOV) limited the resolu- tion capacity of the satellite and the large values for the variations in the electronic signal (NEAT) caused spatial and temporal errors, making it difficult to detect the gradients associated with oceanic fronts (Legeckis, 1978). Several methods were proposed to remove the atmospheric contamination responsible for the majority of the difference between satellite and observed sea surface temperature values. The 3 to 8.3 degree difference found by LaViolette and Chabor (1968) came from satellite data that were not corrected for atmospheric attenuation, but in 1969 they developed a daily averaging method to lessen its impact (LaViolette and Chabot, 1969) . Vukovich (1971) developed a filtering technique to accomplish the same purpose while Smith et al . , (1970) used a statistical method which, when compared with ship observa- tions, had both bias and random errors of less than 1 degree C using early NIMBUS satellite data. Maul and Sidran (1973) investigated the effects of the atmosphere, nadir angle, cloud amount, cloud height, and random noise which resulted in a 31 theoretical error (2 degrees C) for the NOAA satellite series, then soon to be launched. In early 1970, NOAA launched ITOS-1 which was the first satellite in the NOAA series of satellites of which both NOAA- 6 and NOAA-7 are now in orbit. In early 19 70, NOAA-NESS be- gan working on a satellite data processing model, to include the effects of atmospheric attenuation, which was the prede- cessor to the GOSSTCOMP (Global Operational Sea Surface Tempera- ture Computation) model (Brower et al. , 1976). With the launch of NOAA-2 in 1972, a more advanced radiometer was put into use with an IFOV of about 1 kilometer and a much reduced system NEAT of less than 3.0 degrees C (Legeckis, 1978). With this improved system, sea surface temperature fronts could be detected and monitored. Among the studies done during the following few years were those of LaViolette (1974) on upwell- ings off the west coast of Africa, Stumpf and Rao (1975) on tracking eddies in the Gulf Stream, and Bernstein et al. , (1977) on the comparison of eddies in the California Current with direct observations. By the mid-1970' s NOAA-NESS had refined their satellite data processing model; however, comparisons with observed data by NOAA itself and by others found that the quality of measure- ments varied with time and geographical area and were related to the temperature gradient field; good correlation came from regions of weak gradients and marginal results came from regions of strong gradients (Brower et al., 1976). Klein (1979) found that NOAA-5 sea surface temperatures in the Northeast Pacific 32 Ocean that had been subjected to the GOSSTCOMP model were biased 3.5 to 3.9 degrees C and suggested that the error was a result of overcorrection by the model for atmospheric attenu- ation. With the launching of TIROS-N in 1978, NOAA-NESS up- dated GOSSTCOMP to take advantage of the Advanced Very High Resolution Radiometer (AVHRR) on this, and on the follow-on NOAA-6 and NOAA-7, satellites. Whereas NOAA-5 had a NEAT of 1 to 1.5 degrees C, the TIROS-N/NOAA A-G satellite series has a NEAT of 0.12 degrees C (Schwalb, 1978). The improvement in NEAT should result in a better correlation between observed and satellite-derived sea surface temperatures. Chahine (1980) suggested that an absolute accuracy of 1 degree C in these differences could be obtained by simultaneous observations of atmospheric and surface emissions with multi-channel radiometers, using spectral regions of the 3.7 ym carbon dioxide windows as the main sounding channel. An instrument to accomplish this has yet to fly on a satellite. 1 . Problems Associated with a Satellite Data Base Briefly described below are three common problems asso- ciated with using satellites in thermal studies. Atmospheric attenuation is important when interpreting satellite-derived temperatures, while location accuracy is important when ther- mal comparisons are made between ship, satellite, and AXBT data. The depth to which present-day radiometers sense the thermal structure concludes the section. 33 a. Atmospheric Attenuation As will be described in detail in Section II. C. 2, data from the 10.5 to 11.5 um infrared channel on NOAA-6 were used on this project. Radiation in this spectral region emitted from the earth's surface or from cloud tops is attenu- ated in its passage through the atmosphere to the radiometer. The major contribution to this attenuation is water vapor which can be responsible for up to a 9.0 degree C correction in the satellite data (Brower ejt al. , 1976) . The amount of water vapor in the atmosphere varies horizontally, vertically, and in time with the least amount of absorption around the 9.5 to 10.5 urn region (Fett and Mitchell, 1977). Other absorbers and their possible corrections are carbon dioxide (0.1 to 0.2 degrees), ozone (0.1 degrees), and aerosols (0.1 to 0.9 5 degrees) Details on the physics of this absorption process can be found in Roberts et al. , (1976) and Weinreb and Hill (1980). Many atmospheric correction techniques have been tried in an attempt to correct satellite data. Some of these were discussed previously. A knowledge of the vertical mois- ture field would help significantly in reducing the attenuation effects but these data are not generally available. In any case, the multispectral approach to this problem seems to offer the best chance to reduce this type of error significantly (Chahine, 1980; Deschamps and Phulpin, 1980). b. Location Accuracy A major portion of this project was devoted to locating geographic positions correctly on satellite imagery. 34 Estimates of location error vary widely. The data archived from early satellites in the NOAA series had accuracies within 20 kilometers along the orbital track but for high zenith angles the accuracy decreased to within 40 kilometers (Conlon, 1973) . Subpixel accuracies were discussed by Bernstein and Ferneyhough (197 5) on LANDSAT imagery. A satellite image con- taining land is usually much easier to correct geometrically than an image that contains open ocean, simply because control points are much easier to identify on the land image. A method of correcting VHRR imagery using a simple algorithm by Legeckis and Pritchard (197 6) had a mean accuracy of 5 kilometers. A technique whereby ship positions were transferred to satellite images by comparing cloud features yielded errors of 80 to 90 kilometers although this was not the main purpose of the study (Cogan and Willand, 1976) . Another study, using both a zoom transfer scope and NOAA-supplied photo- graphic prints of satellite imagery used in conjunction with plastic overlays containing latitude and longitude lines, had a location accuracy to within 10 to 50 kilometers depending on the distance of the feature being located from land control points (Tabata and Gower, 1980) . A full discussion of location accuracy can be found in Section IV below. c. Remote Sensing of the Vertical Structure The sea "surface" temperature that a satellite senses is a manifestation of a complex process that occurs in the top few millimeters of water called the thermal boundary 35 layer. This layer is subject to the processes of net upward heat flux, infrared and solar radiation, and turbulence with the resulting temperature difference between the top and bottom of this layer of up to 1.0 degrees C (Katsaros, 1980) . Typical radiometers sense only the radiation emitted from a depth of about 50 urn. Direct measurement by satellites of the deeper vertical thermal structure is not possible with the instruments carried onboard the satellites in orbit today. Techniques using Raman lidar systems have been developed theoretically and prototypes experimentally tested with reported accuracies within 0.2 degrees C (theoretical best value) to depths of 30 meters (Leonard et_ a_l. , 1979) . Conclusions from this study suggest that the structure to depths of 100 meters may be de- tectable. The physics of the Raman spectra used in this pro- cess can be found in Murphy and Bernstein (1972). C. NOAA-6 OPERATION The NOAA-6 satellite is the second satellite in a series of third generation, polar-orbiting satellites that began with the launch of TIROS-N on 13 October 1978 at the Air Force Western Test Range, Vandenberg Air Force Base, California. The TIROS-N/NOAA A-G satellite series, of which NOAA-A became re- designated NOAA-6 upon its successful launch, is a joint re- search effort of the United States, the United Kingdom, and France and is operated by the National Environmental Satellite Service of the National Oceanic and Atmospheric Administration 36 (NOAA-NESS) under the U.S. Department of Commerce. The United Kingdom provided one of the three sounding units onboard the satellite, France supplied the onboard data collection system (DCS) , the National Aeronautics and Space Administration (NASA) funded the development and launch of TIROS-N, and NOAA supplied the funds for the NOAA-6 satellite. The mission objective of this satellite series that directly relates to this thesis is the continuous monitoring of the environmental features in the western hemisphere which is accomplished in conjunction with a second satellite system, also operated by NOAA, the Geo- stationary Operational Environmental Satellite (GOES) System. It should be noted that TIROS-N ceased operation in late 1980. NOAA-6 was still functioning at the writing of this thesis and NOAA-7 began operating in June 1981. For the purposes of this project, only those spacecraft systems that were extensively used or are important to the understanding of the results are explained below. The reader is referred to Schwalb (1978), Hussey (1979), Lauritson, et al. , (1979) , and ITT Aerospace (undated) for a fully detailed des- cription of the many instruments onboard NOAA-6. Sections of these references, especially the works of Schwalb and Hussey, were used extensively below. 1. The Spacecraft NOAA-6 used an Atlas-F launch vehicle which is a com- paratively small rocket approximately 28 meters tall and weighing about 600,0 00 kilograms. See Figure 8. The main body of the rocket detaches after launch and a second stage solid 37 Figure 8. Atlas-F launch vehicle (from Hussey, 1979 — SPACECRAFT FAIRING ^-o NOAA-6 SPACECRAFT ATTACH FITTING BOOSTER ENGINE (2) SUSTAINER ENGINE 38 rocket motor, an integral part of the NOAA-6 satellite itself, burns until depletion putting the satellite into a nominal 833-kilometer orbit. a. Physical Structure The satellite itself, as shown in Figure 9, con- sists of three sections. The Reaction Support Structure (RSS) includes the injection motor mentioned above, the attitude control propulsion system, and an 11.6 square-meter solar cell array. The Instrument Mounting Platform (IMP) includes the attitude control sensors and the Advanced Very High Resolution Radiometer (AVHRR) . The five-sided central structure, located between the RSS and the IMP, includes twelve thermal control louvres and the earth-facing communications antennae. The satellite is 3.71 meters long and 1.88 meters in diameter. Its weight at launch was 1420 kilograms which reduced to 737 kilo- grams once established in its orbit. b. The Attitude Determination and Control Subsystem (AD ACS) When a satellite sensor, such as the AVHRR, scans the surface of the earth, the attitude of the spacecraft is extremely important in determining during data analysis just where the sensor looked. Any roll, pitch, or yaw on the sat- ellite will make the application of scan geometry extremely difficult and significant errors would result. Because of this, the ADACS system was designed to maintain the attitude of the spacecraft to within 0.2 degrees (3-sigma) of -che local geographic reference (Schwalb, 1978). This value is obtained 39 >1 Q) en B e o M 4-4 4J 4-1 id M O i Q) GO U) B g 0 i-i u-i g id u &< to •H X! 0 iH 4-1 (0 4J id T3 v£> I O 0) z o < H 2z 353 = r 5 o r-* cr O UJ a t" 13 oa x *: 11 si Q Z ui o (— — x <|- § tl 8 a. x >*• r- ? z S cc 5 5 — BE a. — rs4 , 1 1 1 1 1 f 3 D 7 (/J X 2? ? CO z a 0 -j s »: ui !/) ^ XI S K siN3wnHiSNi 3ivh viva MOT 42 earth scan data into 2048 computer data words per scan line. The pulse that is sensed at the initiation of each scan line originates when the AVHRR scan mirror, which rotates at 360 RPM producing 6 scan lines per second, reaches a precise posi- tion in its sweep just prior to scanning across the surface of the earth. The data are stored in memory and then subsequently read out at a rate suitable for the HRPT on a first-in first- out basis. Any one of the 2048 data words or samples, along with the number of the scan line on which it is located, de- fines a pixel. Throughout this project, the term pixel will be defined by the designation (scan line number , sample number) or, in short, (NL,NS) . A more comprehensive discussion of this process can be found in Section III.B below. 2 . NOAA-6 Onboard Sensors There are three primary environmental sensors onboard NOAA-6. The TIROS Operational Vertical Sounder (TOVS) con- sists of the High Resolution Infrared Radiation Sounder (HIRS/ 2) whose purpose is to provide data to allow calculation of vertical temperature profiles and atmospheric water and ozone concentrations, the Stratospheric Sounding Unit (SSU) , and the Microwave Sounding Unit (MSU) . The second sensor, the Space Environment Monitor (SEM) , consists of a Total Energy Detector (TED) , the Medium Energy Proton and Electron Detector (MEPED) , and the High Energy Proton and Alpha Detector (HEPAD) . The last of the three systems, the AVHRR, was the sensor system extensively used on this project and will be described in detail below. For an in-depth discussion of the first two 43 sensor systems mentioned above, the reader again is referred to Schwalb (1978) . The AVHRR aboard NOAA-6 is a four-channel scanning radiometer that is sensitive to energy in four regions of the electromagnetic spectrum. Table 1 below is a summary of NOAA-6 channelization. Table 1 NOAA-6 AVHRR Channelization (adapted from Schwalb, 1978) CHANNEL WAVELENGTH (pm) REGION PURPOSE 1 0.58-0.68 visible cloud coverage land-water bound, snow-ice extent 2 0.725 - 1.1 visible- as above near-ir 3* 3.55-3.93 mid-ir sea surface temp. cloud mapping 4 10.5-11.5 far-ir sea surface temp. cloud mapping * On NOAA-6, channel 3 is very noisy and usually not used An afocal 20.3 cm-aperture telescope, which produces a field of view of 1.3 ± 0.1 milliradians and an instantaneous field of view (IFOV) ground resolution of 1.1 kilometers at nadir (Lauritson, et al_. , 1979) , separates the radiant energy into the four spectral regions with the help of secondary op- tics. The radiant energy in each of these regions is then focused on its respective detector. The quantity of energy 44 sensed by the detector then is converted to a count value from 0 to 255 in the format used for this project. Channel 4 was the main channel from which information was gathered and it uses a mercury cadmium teluride (H C,T ) detector optimized for g d e r best sensitivity between 10.5 and 11.4 micrometers (Schwalb, 1978) . The spectral response curve for channel 4 is shown in Figure 11. The noise equivalent differential temperature (NEAT) , a measure of the random or coherent two-dimensional noise patterns superimposed on the data signal broadcast to earth, is less than 0.12 degrees Kelvin at 300 degrees Kelvin. Pre-launch AVHRR calibration is covered in a report by ITT Aerospace (undated) and post-launch thermal calibration of channel 4 is covered extensively in a report by Laurtison, et al. , (1979). For every scan line, the radiometer views deep space (0 radiance) and then a blackbody target designed into the radiometer housing and kept heated to 15 degrees centigrade. To a first order approximation, the radiometer output is linear with input energy (Schwalb, 1978) so a two-point linear cali- bration, using the above values, is done during every scan sequence. Channel 4 with its H C,T detector, however, has a not-quite-linear response due to the physical properties of the H C,T . Lauritson, et al., (1979) have generated a table g d e — — r of errors for this channel, which represents the difference between the actual target temperature and the temperature de- rived from the two-point calibration. Table 2 is a summary of these data. Note particularly the small errors around 285 degrees Kelvin, for this is the sea surface temperature range 45 Figure 11. NOAA-6 AVHRR channel 4 spectral response curve (from Kidwell, 1979) 100 ) 50 ToT___L_ Wavelength (ym) 46 determined by the AXBT drops. With this information, a table of count-value-to-temperature conversions was generated for the time of this project by NOAA-NESS and is included in Appendix A. Table 2 NOAA-6 AVHRR Channel 4 Nonlinearity Errors (from Lauritson et_ al . , 1979) TARGET TEMPERATURE ERROR (degrees K) 305 0.5 295 0.3 285 0.0 275 -0.4 265 -0.3 3. NOAA-6 Orbital Parameters Because the development of the satellite data set was so closely intertwined with the NOAA-6 orbital parameters, their discussion is included in Section III.B.3.C below. 47 III. DATA COLLECTION AND PROCESSING TECHNIQUES The use of satellite data on any research topic intro- duces extensive data processing problems, especially when one considers that a typical NOAA-6 infrared satellite image con- tains over nine million pieces of data. This section explains the procedures used on this project to collect, process, and analyze NOAA-6 satellite imagery with emphasis in the area of geographic location accuracy. Also included in this sec- tion are the procedures to collect and process the AXBT data as well as the collection of the GOSSTCOMP product. A. AXBT COLLECTION AND PROCESSING As part of the data base for this project, a series of six Navy P-3C aircraft flights were staged out of NAS Moffett Field, California, for the purpose of dropping a pattern of bathyther- mographic sonobuoys (AXBT) . The dates of these flights were 15, 17, and 19 November and 1, 3, and 5 December 198 0. These flights were scheduled as part of the Naval Postgraduate School's research effort on behalf of the joint U.S . -Canadian Storm Transfer and Response Experiment (STREX) . Eact of the nine-hour flights flew northwestward from Cape Mendocino, California and proceeded to drop a series of AXBT's along a track 133 3 kilometers (720 nm) long as shown in Figure 12. The spacing between the buoys was 55.6 kilometers (30 nm) . The first four flights flew out and back on the center track dropping AXBT's on positions 1 through 24. Flight 5 flew the 48 Figure 12. AXBT patterns for the project area 5rf Ai. 4tf 140° — I north track 54 center 13 \,55 track " south track 5CN 4 x x c* a ■ f- vi C> a i N > r- - «» — ri 13 N ■ — n c fl c i M B* N O i. 01 d) 3 Cn •H 5 *? I" o« » e» o o t? '': — — '1 1 '. i -: fl ?: i < rif l ci ri :i : l :i -I : I -| : i • I :l ri c: " ~ •i "■* a - - -, -i _ -, ^ ., _ . -^ ~ _ , a 3 3 "J 3 3 3 .3 a 3 3 :3 — '3 CO a 3 3 fl fl - -- - -1 — rt fl fl fs* tf5 f n O tv. fl !- ?! h fl — e i fl -^ w r» fl fl N fl - 3 - ■> fl T t)C- - T - - I* f 1 L? B> e» fl fl ^ .-: ■? 3 - i- • ^•TTTrf-rT'rvvT-rTTrT' 54 15 show two additional products, a depth-of-isotherm summary and a listing of the temperature at 5 meter intervals for each AXBT. The accuracy of this system is also within the accuracy limits of the AXBT. It should be noted also that the AXBT-digitizer had provisions for recording the raw received signal from the AXBT onto an analog tape recorder. These tapes then were used during the analysis phase as a direct input to the AXBT-digitizer in order to verify questionable temperature profiles . B. SATELLITE DATA SET SELECTION AND PROCESSING The decision criteria used to determine which satellite passes to examine were reviewed in the following order: (1) the satellite pass coverage had to include the ocean area where the AXBT's were dropped; (2) the time of the satellite pass should be as close as possible to the time when the AXBT's were dropped; (3) the ocean areas containing the AXBT's should be rela- tively cloud-free; and (4) there had to be at least one clearly identifiable landmark somewhere on the full satellite image. Two outside government facilities were used in addition to the facilities at the Naval Postgraduate School in order to choose satellite passes which met these decision criteria. 1. NOAA-NESS The facilities of the Satellite Field Services Station of the National Oceanic and Atmospheric Administration's National 55 Figure 15. Digitizer temperature-at-5-meter-intervals summary, an example (from Kilonsky, 1981)' date time lat. long. axbt depth temp. \ \ I f no. (m) (C) • / i 1191180 3 192014 42330O 1291200 30 4 __ 77 01481 51403 101403 151483 201481 251403 301401 351482 401482 431482: 501402 531421 001341 651217 701104 751071 001045 031032 901013 95 972 1G0 933 105 945 110 929 115 917 120 911 125 904 130 902 105 0U9 140 0(1! 145 862 150 C5 1 153 013 10O 030 105 033 170 027 175 030 109 019 105 Oil 190 797 195 794 200 793 205 700 :.M0 776 215 773 220 770 225 705 23'» 760 235 752 240 740 245 740 250 733 255 730 200 725 265 717 270 711 275 705 209 690 205 694 290 688 295 682 300*074 3C5 071 310 000 313 062 320 057 325 053 330 052 335 649 340 644 345 644 350 -j 36 335 030 .,'00 02 1 365 <,22 370 017 375 014 30!) 012 2191133 3 193010 433000 130 UOO 59 5 77 01443 51445 101413 151440 291440 251438 301440 351442 401440 451442 501446 3514 19 001311 051136 701091 751059 091049 051034 901021 93 996 100 9C9 105 959 IIO 943 115 928 120 912 125 097 130 900 133 094 140 006 145 078 150 203 155 0!3 100 029 165 022 170 010 175 Oil lOO 003 185 797 199 709 195 703 200 772 205 7(0 210 770 215 770 229 700 225 762 230 757 235 75-1 240 751 245 745 230 73 5 255 735 COO 727 205 722 270 719 275 713 209 700 205 700 290 092 295 607 SCO 076 305 071 310 003 315 052 320 044 325 034 330 025 335 614 340 609 345 604 330 500 355 209362 507 305 504 379 579 375 574 illio 571 3191130 3 145034 4 12100 131 700 55 6 77 01393 513':0 101397 151395 291395 251393 391397 351395 401394 431394 501393 551222 001120 051039 70 991 75 967 00 959 05 949 90 932 93 918 100 993 105 9C0 110 070 115 070 120 002 125 053 130 030 135 027 140 014 145 701 1-50 773 135 705 100 701 165 75:» 170 714 175 738 IO0 735 1G5 730 190 733 195 734 200 733 205 735 210 727 215 727 229 727 225 719 239 717 235 702 240 694 245 692 230 024 255 602 "09 071 265 003 270 003 275 051 289 64 0 205 044 299 039 295 636 200 023 305 617 310 006 313 000 320 590 325 593 339 500 335 502 340 574 345 571 350 971 355 305 ,'00 301 305 355 379 3-'iQ 375 547 300 545 4191 I CO 3 20 23'1 45 400 132 700 30 7 ------ ^ 01314 51311 101314 151314 201317 251319 301317 351316 401316 451314 501317 3513C0 001112 651013 70 930 75 952 00 933 05 920 90 900 95 009 ICO S70 105 05! 110 035 115 Oil 120 794 125 703 139 760 135 754 140 744 145 733 130 ?33 155 733 160 733 163 733 170 733 175 735 109 733 105 733 190 735 195 733 200 724 205 732 210 727 215 727 220 722 225 722 230 722 235 710 240 712 245 710 230 701 255 695 200 090 205 (.0 !■ 270 002 275 077 209 667 205 637 293 0 17 295 639 300 031 305 020 310 020 315 013 320 OOO 325 001 330 390 335 302 340 374 345 571 330 "03 355 557360 350 305 547 (70 543 375 530 300 536 5191120 3 201323 4340C0 133 UOO 03 8 77 013C0 51306 101306 151300 291306 251303 301303 351396 491305 451306 30I2C3 331306 001207 031029 70 904 73 922 00 906 05 003 90 070 95 859 100 '134 105 843 110 027 115 796 120 709 125 742 130 719 135 703 140 698 145 698 130 093 135 701 100 702 165 702 170 0911 175 702 109 705 1 05 700 190 706 195 703 209 702 205 70! 210 095 215 695 220 090 :>25 007 239 604 235 679 240 674 243 668 230 003 253 600 260 052 265 049 270 049 275 041 2!J9 035 205 031 290 620 295 625 3G0 029 3C5 012 310 012 313 001 320 595 325 390 339 302 335 574 340 371 345 569 330 303 353 337 MO 553 305 510 370 542 375 539 300 531 6191123 3 202220 4622C0 1341200 60 9 77 01193 51203 101204 151204 29120 1 251204 301204 351203 401203 451204 5G1204 53120 ■ 001201 051130 70 1)33 75 050 09 037 03 024 90 01 I 95 005 100 SCO 105 797 110 734 115 709 120 750 125 742 130 095 135 001 140 079 145 604 130 090 153 694 I oO 701 165 701 170 703 175 703 109 701 105 703 190 791 195 698 200 «95 2C3 606 210 679 215 002 220 002 225 074 239 671 235 067 240 065 245 663 230 000 233 053 200 049 205 041 270 039 275 020 289 623 205 017 290 612 295 606 300 .jOI 305 590 310 590 315 501 320 370 325 509 330 366 335 355 340 353 345 547 3?? ~.\~i 3vv 5,r ■]'•'} saa -m asz -ni am aza am aaa ma 7191 ICO 3 204120 46 700 1334000 04 17 77 01279 51202 101202 151202 201202 251202 301202 351202 401201 451281 5912i2 551279 001271 651 147 701931 75 948 OO 905 05 005 99 070 95 837 !00 c-: i 105 032 110 019 115 000 120 700 125 772 130 757 135 731 140 744 145 730 150 722 135 721 1 00 721 105 724 170 730 175 727 100 727 105 735 190 730 195 719 200 719 205 710 210 719 215 717 220 717 225 714 230 711 235 701 240 697 245 690 230 037 255 602 2oO 071 205 000 270 037 275 057 209 052 205 047 290 036 295 631 300 023 305 620 310 017 315 009 320 005 325 091 339 593 335 500 340 503 345 577 350 374 353 300 300 364 303 357 370 553 375 545 300 542 56 Environmental Satellite Service (NOAA-NESS) in Redwood City, California, were used in initially selecting the satellite passes. The Redwood City facility is one of three NOAA-NESS stations that monitor NOAA-6; the other two are the Command and Data Acquisition (CDA) stations in Gilmore Creek, Alaska, and Wallops Island, Virginia. Redwood City differs from the CDA stations in that Redwood City records the digital High Resolution Picture Transmission (HRPT) readout consisting of three channels of AVHRR data in the 8-bit precision field- station format. These 1600 BPI, 9-track computer-compatible magnetic tapes then are archived in Redwood City on a 9 0-day rotating basis. The CDA stations record the various other data formats broadcast from NOAA-6 as well as recording the HRPT data in 10-bit precision which then are forwarded to the NOAA Suitland, Maryland, facility where processing and archiv- ing on a more permanent basis occur. The precision loss in going from the 10-bit HRPT data to the 8-bit HRPT data is between 0.4 and 0.5 degrees when making thermal comparisons (Kidwell, 1979) . The amount of data recorded per satellite pass depends on the satellite's elevation in relation to the receiving station antenna and can be limited by the 13-minute capacity of a standard length magnetic tape. Passing directly over Redwood City's antenna, NOAA-6 would be within reception range for 15.5 minutes, depending on orbital altitude, and could provide data from a circular area 6200 kilometers in diameter centered on the antenna (Schwalb, 1978) . According to Schwalb, the satellite provides useful data only if it is 57 at least five degrees above the horizon. This reduces the contact time to 13 minutes and the circular area to 5200 kilometers . a. Field-Station Format The field-station format differs from the CDA- station format in that it is a combination ASCII-Binary format consisting of a single header record at the beginning of the tape followed by up to 15000 data records. Each of the data records is a sequential interleavening of the scan lines and the recorded channels as shown in Table 3 below: Table 3 Field-Station Format RECORD CONTENTS 1 header 2 scan line 1 — AVHRR channel A 3 scan line 1 — AVHRR channel B 4 scan line 1 — AVHRR channel C 5 scan line 2 — AVHRR channel A 14998 scan line 5000 — AVHRR channel A 14999 scan line 5000 — AVHRR channel B 15000 scan line 5000 — AVHRR channel C channel A, 3, or C = any sequence of channels 1, 2, 3, 4 58 The 40-byte header record, all in ASCII, contains the ground station identification (SFO for Redwood City) , the channel numbers identifying which three of the four available AVHRR channels were recorded, the time (GMT) of the first scan line, the duration of the pass, and the orbit number. See Figure 16 for an example of the header record. Each of the remaining 15000 or so data records have identical 2138-byte formats beginning with a 14-byte ASCII "mini-header" consisting of an identification sequence, the specific AVHRR channel num- ber from which the data in the record originated, the Julian date of the scan line, and the time (GMT) of the scan line. Following the "mini-header" are 10 bytes of telemetry data, 6 bytes of back scan data, 10 bytes of space view data, and 50 bytes of space data, all in binary format. The remaining 2048 bytes, also in binary, are the video data from which esti- mates of the sea-surface temperature are derived. See Figure 17 for an example of one of these data records. b. Ephemeris Data Set A set of ephemeris data for NOAA-6 also is main- tained at Redwood City. An ephemeris data set consists of tracking information so that the field station can capture the satellite's data stream as the satellite rises above the hori- zon and passes overhead to the opposite horizon. See Figure 18 and Figure 19 for examples of an ephemeris data set. More importantly to this project, the ephemeris also contains the subsatellite points for the pass calculated at one minute intervals. A subsatellite point is that point on the earth's 59 Figure 16 Header record — field-station format (adapted from Kidwell, 1979) WORD BYTE CONTENTS BYTE CONTENTS TYPE 1 1 station ID 2 station ID ASCII 2 3 station ID 4 blank ASCII 3 5 blank 6 channel A ASCII 4 7 channel B 8 channel C ASCII 5 9 hours 10 hours ASCII 6 11 minutes 12 minutes ASCII 7 13 seconds 14 seconds ASCII 8 15 duration i-min 16 duration -min ASCII 9 17 duration i-sec 18 duration-sec ASCII 10 19 orbit 20 orbit ASCII 11 21 orbit 22 orbit ASCII 12 23 orbit 24 blank ASCII 13-20 25-40 blank channel A, B, or C = channel 1, 2, 3, or 4 As an example, a pass selected for the project may have a header record as follows: SFO 1340333481300 7244 indicating a Redwood City tape (SFO) containing AVHRR channels 1, 3, and 4. Time of the first scan line was 03 hours 33 minutes and 4 8 seconds (GxMT) while the duration of the pass recorded was 13 minutes and 00 seconds. The orbit number was 7244. 60 Figure 17 Data record--f ield-station format (adapted from Kidwell, 1979) WORD BYTE CONTENTS BYTE CONTENTS TYPE 1 1 ID 2 ID ASCII 2 3 ID 4 ID ASCII 3 5 channel no. 6 day ASCII 4 7 day 8 day ASCII 5 9 hours 10 hours ASCII 6 11 minutes 12 minutes ASCII 7 13 seconds 14 seconds ASCII 8-12 15-24 telemetry data Binary (average) 13- -15 25-30 back scan data (average) Binary 16- ■20 31-40 space view data (average) Binary 21- ■4 5 41-90 space data (raw) Binary 46- 91-2138 video Binary 1078 61 o CO u 0 aj u CQ I U M— I 0) iH a e X vO m *J" CV oc c in fO «-t o CO >o n r- — < 0 o- •o t— i <\i o t >T ro t>< •—> o •o in CO in • • • • • • »-■ o en r-4 c\» ci O 0' ac _l II z o Z Ui UJ *-» u > • CC X _j t— i II o II ■u O. o < X II z a. Ui H- «. > UI > «*> M r- Z u -J u. »- u. 0» CO c t— i O c z: a < * o "> »-4 K o K * uu >-< UJ -J « z ^ < O z u o *t < o a. z UI o z • UJ or CO r- io UJ — a £ «t ac •* < -j i- r-> OC Q. O a. X at Ui oc »~ z o rj — 1 CO o iTi CO CO CM co «n <* >!• »r >- »-« o m -o u> on «/> h- fO i- «n o •* •r» f- rvj o ro O o z ^ o» o • << o» <© i/> er fM 1-4 in •0 iM «M IT r- f^ *t •-4 o o x\ o rj O z 2 • • • • • • ** < o o oo f» m OJ > ~ o 0- !«-> o «t a —• l-« fM z UJ -J ^ 2 a I at - z UJ UJ > • n * •-• K o a Ui —1 /_-» X ■— t t-4 o o -r ex «J o »- z ^^ r *-« ►-« < Of o *- Of ey z o u z C t~ •-• i/i a «I -} Z -1 < «« ID o a. 2 £ u z a. Q <. . 1 w. u a- hi C' CJ •t O a. ■a in 2T i/> u t- •SI r> UI z Ui — • UC x c lu a a >r co (M o U^ o ■C ^-4 •o • • — « o •-I o f^ cu f* ro CO • • ■ ■ * • Uj • s: • ^« • K in X £ O •— • o t- 0. UJ I O r- c «t c a. c Ui CC p- UJ or r- CD in «» r CL ^ <^i z i_» Ui -^ hi z h- a? Q vo «* ^4 •— ■ »- -J u. 3 «l o _J r O c *• > i- ^ UJ •» o ac 62 c 1) ' Q O CO u 0) 0) u CQ s o u 4-1 0) rH Q4 o C ■P 0) CO >o x» -r ^ ci i'i »j ^ -^ h —• p- r 1 » u ■ \ 3s "J" -ft <*■> rO to i\i it. CO «-* 3^ ir» m f- ro r-4 **■ O) O .1 l"» n >}• J C'l D ^ H O Ij. H (O u i-< iv rvj eg 00 >j- ^ •-* >r -o i"" 1- r- co >o r-i .-H ,-*) ^- ,-h <— • r-» .-* 1— I ■ 1. > «*"■ u" O O »T -0 6 aj O C. .T\ 4\J >0 J> O ►- '-^ —« »-< <\< (V> Ufl IT. >f (*1 i\ H ,.. < O O v_y ^ t— u_ o ,-0 i, l/° u,» (_ > »-i xr ,y-< ,~j p«. ;?» ,-* p. <7-(*-«*ia'v'-«r< vr ^J" .M n h OO-t -O 03 O CJ r*i T •Of*-Xv0v o .— 1 rvj t\i r^ h r U^ (* IT.' -J1 J'OOOO*"*'-*'- H H H IB »»4 1% OJ (VJ CVJ <"^ CL - t-ZZZZZZZZZZZZZZZZZZZZZZ Zi -t r -r x) 0s cc n r»s 1 vr c ^ — • >o p- lp ' > 'r j-- >r ?• />_j»»»**w* . . * *..*.. (- o rr: .p rJ 'j* -o r^ 'J -£ <\ >j» o «v 0 ir «-j o; 4 • <-« r-~ >r o ic 1- r N >o -«.> *0 u"> »T> if. ■+ I" ■-" r*l v t*\ <\- .-•• i\» — < —• »~ ir. 2 p- ►— c >j r o r- co c O — ' "N <«"> -t -»" -o r- co a j h«\j ro>r ?" -t -i ><■ >i* >r -r vi »■ j^ j1 ^ in '/ t in n u"i O U u\ tf1 if i/\ IT iT 1/' <■> J^ a"i .^ m J iTi^^vl' ^i^i^O^O 22 T «-4 r— I H H r •—•!—!<-■— Hi— ft— 4 .■■ i— i ■— <■—(»- •—«—•»—<«— 4 .— «.— 4 • >>- >4j-O-O>0^O>4j,4i^O^O^J^0^>vU^J-O^0^J'J'>OO«J^0 a.- «* 63 surface directly beneath the spacecraft and represents the middle of the scan line for the AVHRR. All the subsatellite points for each scan line taken together represent the ground track that the satellite followed in its orbit. It should be noted that any alignment errors made when the AVHRR module was attached to the spacecraft during construction may result in the subsatellite point not being the center of the AVHRR scan line. A summary of alignment data may be found in an undated report prepared by ITT Aerospace for NASA. For the purposes of this project it was decided that any alignment errors were so slight as to be negligible and therefore that the sub- satellite point would represent the center of each AVHRR scan line. Other information in the ephemeris data set important to this project were the orbital elements listed in the preface to each ephemeris including orbital period, semi-major axis, eccentricity, and inclination. This information was vital to the orbital calculations made further on in this project and will be explained there. c. Initial Satellite Pass Selection Each AVHRR scan line is approximately 284 0 kilom- eters long with 142 0 kilometers on each side of the subsatellite point. With this information as well as the ephemeris data sets for the dates of the P3-C flights and a chart of the Northeast Pacific Ocean, it was relatively easy to determine specific passes which viewed the ocean areas where the AXBT ' s were dropped, thus satisfying the first decision criterion 64 mentioned above. Twenty-two satellite passes were thus selected for further screening. The selection from these 22 passes of orbits whose time matched as closely as possible the time of the AXBT drops was done in conjunction with the investigation of cloud cover- age over the ocean area of interest. As the project relied solely on the use of the infrared channels of the AVHRR and because cloud cover effectively prevents AVHRR scan coverage of the ocean surface, the absence of cloud cover in the ocean area of interest was a major factor in pass selection. Also, as will be discussed below, NOAA-6 is a sun-synchronous satellite that circles the earth 14.2 times every 24 hours. As a result, NOAA-6 views the same earth location at the same local sun time each day. This translated to our ocean area of interest as between 0330 and 0400 (GMT) for ascending passes and between 1650 and 1720 (GMT) for descending passes. Since Redwood City maintains hourly pictures taken from the visual channels of the geostationary GOES-WEST satellite, examination of these pictures for cloud coverage in the ocean area of interest re- sulted in the selection of one pass for each of the six flight dates that represented the best compromise between matching times and cloud coverage. The six passes chosen for further examination are listed in Table 4 below. 2 . NASA-Ames Research Center The last criterion to be satisfied, identification of a landmark on each image, was done on the Interactive Digital Image Manipulation System (IDIMS) located at the Technology 65 15 Nov 80 1 17 Nov 80 2 19 Nov 80 3 01 Dec 80 4 03 Dec 80 5 05 Dec 80 6 Table 4 Selected Satellite Passes AIRCRAFT PROJ. AXBT DROP ORBIT ORBIT TYPE FLIGHT NO. TIMES (DTG) NO. TIMES (DTG) PASS 151905-160049 7209 151658-151711 D 171932-180236 7244 180331-180 346 A 191843-192232 7266 191712-191716 D 011826-020031 7429 0103 44-010358 A 031826-040127 7465 0 31656-031710 D 051816-060013 7486 050355-050408 A where D = descending and A = ascending Applications Branch of the Airborne Missions and Applications Division under the Director of Astronautics, NASA Ames Re- search Center, Moffet Field, California, a. IDIMS The IDIMS system is a software package that inter- acts with a minicomputer (HP-3000) , a display terminal, a 25- inch COMTAL display screen, a Dunn Instruments color camera recorder, and a high-speed printer, and is used extensively to work with satellite data, especially LANDSAT imagery. Op- tions are available that allow the user to manipulate inter- actively satellite imagery so that specific topics of interest may be investigated like land use with LANDSAT data, sea surface temperature, cloud cover, or ice pack coverage as examples from TIROS-NOAA imagery or the many other applications available from NIMBUS-7 imagery. The mechanics behind specific options 66 are proprietary data owned by Electromagnetic Systems Labora- tory, Inc. (ESL) of Sunnyvale, California, who developed I DIMS and to whom the reader is referred for more detailed information (ESL, Inc., 1978). A second IDIMS system is located at the Scripps Institution of Oceanography in La Jo 11a, California, where initial training was done by the author in the use of the IDIMS system. A third IDIMS system, also operated by NASA Ames, is located in a mobile van that tours the Western United States; its facilities were used upon one of its stops at the Naval Postgraduate School. b. Landmark Identification The basic procedure for landmark identification used by this project on the IDIMS system was as follows: (1) run a short tape routine on each of the se- lected pass's magnetic tapes to identify the number of data records, hence the number of scan lines per pass (number of data records minus 1 header record, all divided by 3) ; (2) read the 3-channel AVHRR data from the magnetic tape into computer memory and initiate IDIMS processing; (3) recall that data comprising the infrared channel from memory and display it on the COMTAL display screen; (4) enhance the displayed infrared imagery for temperature using false colors; (5) use the Dunn color camera recorder to produce an 8 by 10-inch color Polaroid photograph of the enhanced image ; 67 (6) use the ZOOM option of IDIMS to enlarge se- lected sections of the displayed image in order to locate landmark pixels by scan line number and sample number; and finally, (7) use the PICPRINT option of IDIMS to dump to the high-speed printer the count values of all pixels within a specified area surrounding the landmark. An explanation of certain aspects of this procedure is explained in the sections below. (1) Count Values. In order to display a satellite image on the COMTAL display screen, the IDIMS system sequen- tially unpacks the video data read into memory from the magnetic tape. These data consist of count values between 0 and 255 which represent the difference in detected energy be- tween a look at deep space and a look at a radiating surface such as the earth or clouds (Schwalb, 1978). These count values are used by the IDIMS system to produce an image with a grey-scale intensity range from 0 to 255 in order to match the same range of count values. Options available on the IDIMS system allow various color assignments based on these count values including an automatic full-spectrum false color assign- ment where red is "hot" and blue is "cold" or vice versa. Also available are options allowing a single color to vary in its saturation over the full count range of 256 values or the assignment of a specific color to an individual count as "blue=count 125" or to a range of counts as "blue=counts 125 through 130". Assigning colors in this manner tends to produce 68 a confusing image if many hues are used indiscriminantly . For purposes of this project, the automatic full-spectrum false color assignment of red "cold" and blue "hot" was used so that the oceans were blue and the cloud tops, being much colder, were red in all the photographs, c. Pixel Identification The identification of a landmark pixel and the subsequent assignment of a scan line number and sample number are made easy by the IDIMS system but the principles behind their assignment had to be understood so that other landmarks and buoy positions could be located as needed later on in the project. As seen in Table 4 above, three of the six selected NOAA-6 passes were ascending passes and three were descending passes. An ascending pass is one where the satellite in its orbit crosses the earth's equator heading northwards while a descending pass is one where the satellite crosses the equator heading southwards. As the satellite is moving, the AVHRR scan mirror sweeps from right to left perpendicularly across the satellite's velocity vector six times per second with each sweep defining a single scan line. Although the scanning mirror rotates in a complete circle, only the data located 55.4 degrees either side of nadir is retained. Nadir is a term similar in meaning to the subsatellite point in that nadir represents that point on the sweep of the scanning mirror when the mirror is pointed at the spot on the earth's surface directly underneath the spacecraft. The radiometer data 69 stream from this 110.8-degree arc is electronically divided by the MIRP into 2048 equal samples each representing 0.054 degrees of the total arc (110.8 divided by 2048). Therefore, when the satellite transmits its data stream to the earth receiving station, the first 2048 bytes of video data taken together is intrinsically labeled scan line number 1 while the first byte is intrinsically labeled sample number 1 or, as used throughout this project, (1,1). The second byte of the first scan line is labeled (1,2). Sample number 1505 from scan line number 3520 would be labeled (3520,1505). Sample number 1 through 1024 are located to the right of the sub- satellite point when looking in the direction of the satellite's velocity vector while samples 1025 through 2048 are to the left. Because of this right-to-left pixel numbering system, when an image is displayed on the COMTAL unit by the IDIMS system, an ascending pass image looks reversed and upside down while a descending pass image looks normal where normal is defined as having Alaska to the north or top of the image, Hawaii to the west or left of the image, and California to the east or right of the image. An ascending pass image, when displayed on a display screen or when stored into computer memory for further processing, has Alaska on the bottom, Hawaii on the right, and California on the left of the image. This occurs because IDIMS always displays pixel (1,1) in the upper left corner of the COMTAL unit. Although landmark identification can be easily done on either type of pass, the geometry involved further on in this project rests heavily on a clear understanding of which 70 type of pass you are analyzing and on which side of the sub- satellite point is the landmark or buoy. Figure 2 0 is an example of an ascending pass while Figure 21 is an example of a descending pass. In each figure, the satellite would travel directly up or down the center of the image respectively. The IDIMS system automatically keeps track of this numbering system and conveniently displays the scan line num- ber, sample number, and count value for the pixel you have identified on the COMTAL unit using a movable cursor. By use of the ZOOM feature on IDIMS and with reference to a chart of the local area, it is usually easy to identify landmarks. In most of the passes used on this project, the San Francisco Bay Area was identified readily and its (scan line number, sample number) determined. For purposes described later on, up to 20 landmarks per satellite pass along the west coast of the United States from Glacier Bay, Alaska, south to Mexico and from San Francisco east to Pyramid Lake, Nevada, were identi- fied. Only one of these landmarks is needed to navigate the image as will be explained later on. The PICPRINT feature of IDIMS allowed the dumping to a high-speed printer of the count values of the pixels sur- rounding the landmark. This was done as a method of verifying the accurate position of the feature chosen for landmark iden- tification. Figure 22 is an example of a PICPRINT output where by using a variation of the game of connecting the dots one can connect count values in order to recognize features. This method of verification will not work, or is made more difficult, 71 Figure 20. NOAA-6 ascending pass, from IDIMS processing 72 en a •H CQ I o 2 CN Q) U en •H Cm 73 Figure 22. PICPRINT output of count values, San Francisco Bay entrance; cloud-free image 112 110 112 114 113 114 117 L24 126 123 1 17] 108 IC7 111 114 114 114 113 112 113 116 11 7 I 14 f~] T \ 137 106 112 112 113 114 1 L5 1 13 1 14 115 JJ-^ 1 13 i i n 106 106 113 112 112 112 113 114 1 16 113 /Tl UJJ 107 119 117 119 116 121 117 119 117 1 16 114 115 1 11 113 ' 108 \i i r 109 N 10 9 JJ-5- 138 103 116 1 10 ~TU9 108 -UiL 118 119 117 118 118 1 14 v\T\ 109 1C8 109 1 11 1 11 rto> V 113 1 15 1 18 122 119 I 16 u l j 10 fi 103 103 139 108 109 115 1 18 Ul 118 n5- -ttr 10 8 108 108 103 103 109 li0\ .111 ^TTK 11 I 1 L2x 110 108 ica 108 103 108 109 109 109' 1C9 ISfi " 10-9 108 An- 113 114 108 ice 1C3 108 103 - -_ IF1C 1C9 139 109 1C5 LC9 109 109 1 10 TlZ 11- ULX- I 1ft- 109 139 L09 PAP 112 113 i 12 1 13 ■ii2- 112 112 /PlO (1 ia "Al/ /111 -I Li I 14 OCEA 1 L0 / 1 12 113 113 1 1? 114 114 1 13 113 1 11 * A w * A. •* 1 \ ft „ ' 112 113 113 1 13 112 113 1 13 113 112 1 12 109 110 /111 1 13 113 114 1 13 113 113 114 114 1 13 1 13 110 110 /111 1 12 113 1 13 1 13 112 113 1 13 113 112 1 12 109 110 111 1 12 113 113 113 11? 114 1 13 112 113 1 13 110 no 1 11 111 113 I 13 1 13 112 113 1 13 112 1 12 1 13 110 110 112 113 114 113 113 113 1 13 110 110 112 114 115 SAN 1 13 112 112 i 11 110 110 112 114 115 114 113 111 , 'Tuf 110 no 112 115 115 I 14 114 111 106 109 1JLJL/ 113 114 115 115 115 1 L5 116 115 115 1 15 114 1 14 112 113 I 13 112 113 113 106 110 , U2 \ 108 110 / 111 1 14 115 1 15 115 I 14 1 14 113 112 112 111 J 107 110/ 112 1 15 I 16 115 115 1 14 I 14 11 3 113 113 108 110/ 113 115 116 1 15 1 14 1 14 112 114 113 UL flta. _JJD6_I if the temperature of the land is the same as the water, if any low-lying clouds (fog) have similar temperatures as either the water or land, or if clouds obscure sections of the land. Normally, the visual channels of the spacecraft are used but since many of the passes occurred at night, only the infrared channels were usable. Figure 2 3 is an example of a case where clouds interfered with the landmark identification process . Because an average pass contained 4 680 scan lines each with 2048 samples or about 9.6 million pixels per pass, a convenient method of matching pixel numbers (hence count values) to AXBT positions was necessary; thus a system was needed to "navigate" the image. 3 . Satellite Image Navigation As mentioned in the introduction, there are many sources of error when one wants to compare satellite-derived sea sur- face temperatures and AXBT-derived sea surface temperatures. It was decided at the beginning of this project that an at- tempt would be made to reduce as much as possible one of these, that being the earth location errors associated with transferring AXBT drop positions to a satellite image, so that temperature comparisons could be made. Several methods were tried and three of those methods that produced the smallest errors are described below. For the purposes of developing procedures for locating AXBT ' s on the satellite image only, an assumption was made that the geographical location of the AXBT's was accurate; hence any aircraft navigation errors, 75 J Figure 23. PICPRINT output of count values, San Francisco Bay entrance; cloud-covered image. (Note: This image is from an ascending pass hence it appears upside down . ) 131 123 125 124 125 124 121 121 l ?2 123 125 125 ,12 0 130 125 1 SAN FRANCISCO t 2 0 ; i 9 120 1 13 1 19 1 13 i 2 0 i iar" \%fir'\ 1 9 '1*1 7 114 J28J 126 124 122 121 1 19 1 1 3 1 13 1 13 1 1 7 i/7 1 16 1 13 ? 27* 1.26 I 25 1*23*127" 124 1*2 "T" 124 122 12*7^127- " 1^2.1 . I~2 4 N-9- i 22 -4' f 3- - 1 19 11-7- 1 1 6 'i'j 6 1 14 1 14 1 1 3 1 12 1 12 130 134 129 127 1 34 13-2 125 123 125 125 125 f2"f * t2J0L 1 26 .Lt-SL- 123 1 17 115 l 2V0— rrT 1 M "i 1*6* 1 14 123 130 13 0 13 0 123 1 27 1126 f ""? 7 t 26 1 26 1 w 1 127 125 1J3. 132 1 3 n 123 127 r?3^. 1 "*> *? 126 126 130 130 129 125 t3&. 135 134 uv M3 131; '>3g--' 1 3 0 1 3 0 1 29 129 129 ' 131 132 134 131 134 128 131 141 143 1 - yQ*jZ 134 1 3 0 J 34 129 133 ■ 130 134 130 133 1 29 29 2 3 i 32 1 32 131 136 131 134 146 146 144 141 1 39 1 36 1 -■ I 36 132 ■ 7 O 1 35 1 36 i 33 149 154 147 144 152 149 142-* f 4/3 136 ' 1 35 ' si V 7 0 33 (32 MARI N COUNTY ] "7 1 156 1 52— 4-4-?— f44 143 1 4 0 •il o 132 1 32 33 134 t 37 1 35 154 K52 >5- W'l 141 1 43 1 42 141 I 33 137 138 137 149 146JI42 t33 130 128 1 33 ' 40 146 1 4 6 144 141 i 33 143 1<42 140 T37 130 ' 30 33 139 1 43 143 143 1 44 135 136 136 t37 1 O »' 135 ! 33 1 32 132 i 34 t 36 133 139 127 123 '129 130 131 134 ! 35 1 34 134 1 •j •!• 1 33 133 132 76 ballistic errors on the falling AXBT ' s once launched from the aircraft, drift errors on the floating AXBT, or human errors in transcribing positional data from aircraft displays to logs were ignored. These sources of error will be discussed later, a. Zoom Transfer Scope A Bausch and Lomb Zoom Transfer Scope was used initially in an attempt to transfer the AXBT positions to the satellite image. A zoom transfer scope allows one opti- cally to overlay a chart, on which the AXBT positions have been plotted, onto a satellite image where enough distinguish- ing features (landmarks) are evident so that by optically stretching, condensing, or rotating the chart, landmarks on both chart and image coincide. Once the landmarks coincide, the operator manually marks with a pencil the AXBT positions onto the satellite image. While the system works fine with small area images consisting mostly of land, it could not be used satisfactorily on this project for a number of reasons. First, each of the selected NOAA-6 passes covered an area ex- tending from Northern Mexico to Alaska and from mid-Pacific to the western United States. Reducing a chart of this area to a size suitable for use on a zoom transfer scope (about 10 by 10 inches) necessarily requires reduction in the accuracy of plotting geographical coordinates. Second, on each of the NOAA-6 passes, approximately 80 to 90 percent of the coverage area was open ocean with any visible landmass only on the edge of the image. Landmasses on the edges of these images are much more distorted than landmasses near the subsatellite 77 points due to a combination of satellite scan geometry, earth curvature, and the transfer of these images to a flat medium like a photograph or chart. Third, the AXBT ' s were dropped along a line over 1300 kilometers long stretching northwestard from Cape Mendocino, California. There are no landmarks in the Northeast Pacific Ocean between Cape Mendocino and the Aleutians; therefore location accuracy decreased the farther away from the coast the AXBT's were dropped. Fourth, marking a chart manually with a pencil necessarily involves inaccura- cies especially when one is trying to locate geographically an item as small as an AXBT. Last, and the hardest to over- come, is that once the buoy position is marked on the satellite image, some method must be found to determine the pixel number and hence the count values of the AXBT's position. Remember- ing that there are 9.6 million pixels per image, determining the exact pixel to choose for a count value would involve some guesswork and possibly even large-scale pixel averaging. Satellite images unfortunately do not come marked with latitude and longitudes, nor do charts contain scan line numbers and sample numbers. b. IDIM's TRNSFORM A second method of trying to locate an AXBT on the satellite image involved the use of an ESL, Inc-developed IDIMS function called TRNSFORM. TRNSFORM is used mainly in registering LANDSAT imagery and involves the calculation of a transformation matrix between matching sets of control points using a least-squares fit method (ESL, Inc., 1978). A first, second, or third order transformation is possible. TRNSFORM was not designed to navigate NOAA-6 imagery mainly because TRNSFORM requires 15 to 20 landmarks spread over the entire image in order to obtain a small pixel error. Therefore, use of this function also failed to provide the accuracy desired for this project for one of the same reasons that the zoom transfer scope failed, in that landmasses were present only on the edges of the images . The method finally used, and from which a location accuracy of less than 2 pixels resulted, was the development of a computer program that determined a satellite's orbit referenced to a single landmark in the image and from this, when given an AXBT latitude and longitude, could determine the scan line number and sample number of the AXBT. The development of the program required a basic understanding of the orbital dynamics of NOAA-6 as well as a working knowledge of spheri- cal geometry. c. NOAA-6 Orbital Dynamics For orbital information in this section, the work by Stewart (1979) and Schwalb (1978) was used extensively. The NOAA-6 satellite is a sun-synchronous satellite which means that its orbital plane rotates at the same rate as the rotation of the earth about the sun. As a result, the satellite views a point on the Earth's surface at the same local sun time each day. Table 4 above listed those times that NOAA-6 viewed the AXBT drop area. According to Schwalb (1973) , the orbital plane precession rate is approximately 79 equal to 0.000000199 radians per second or 0.986 degrees per day eastwards. This rate is achieved by placing the satellite in an orbit with a suitable inclination. In the case of the NOAA-6 satellite, the inclination was determined prior to launch to be 98.739 ± 0.15 degrees, where inclination (i) is defined as the angle the satellite's orbital plane makes with the earth's equatorial plane measured counterclockwise from east. A retrograde inclination (i_) is the supplement of the inclination. See Figure 24. Figure 24 NOAA-6 orbital plane inclination north pole earth s equatorial plane tellite's bitai ane i = inclination (from ephemeris) i = retrograde inclination The period of the satellite, obtainable from the ephemeris data set (as is the inclination) , is the amount of time it takes the satellite to make one orbit of the Earth. 80 The predetermined launch value for the NOAA-6 period was 101.58 minutes; therefore, NOAA-6 orbits the Earth 14.18 times per 24 hours. For each orbit the earth rotates 25.40 degrees east- ward. See Table 5 for a summary of NOAA-6 orbital parameters. Table 5 NOAA-6 Orbital Parameters orbital plane precession rate inclination (i) retrograde inclination (i_) period orbits per day earth rotation per orbit orbital altitude 0.986 deg/day east 98.739 ± 0.15 deg 81.261 ± 0.15 deg 101.58 minutes 14.18 25.40 degrees east 833 ± 18.5 3cm The predetermined launch altitude of the satellite was 833 : 18.5 kilometers. The orbit of NOAA-6, to a first approximation, is an ellipse. From the ephemeris data set the semi -major axis of the ellipse and its eccentricity can be found, thus making the satellite's altitude on any pass simple to calculate as shown in Figure 2 5 below. The mean satellite altitude (H) is that distance measured from the center of the Earth and can be derived using Equation (1) . Using the above orbital information, the data in Table 6 below was derived for the six selected NOAA-6 passes used in this project. 31 Figure 25 Satellite Altitude Determination apogeev -perigee ^satellite a = semi-major axis (from ephemeris) in nm e = eccentricity (from ephemeris) b H = mean satellite altitude = (2b+a)/3 (Eq. 1) 2 1/2 semi-minor axis = a[(l-e )] PARAMETER Table 6 Satellite Data Set Orbital Parameters ORBIT 7209 7244 7266 7429 7465 7486 inclination ( degrees) retrograde inclination period (minutes) semi-major axis (km) eccentricity 98.69708 81.30292 98.69708 81.30292 98.69708 81.30292 98.69123 81.30877 98.69123 81.30877 98.69123 81.30877 101.13285 101.13285 101.13285 101.13084 101.13084 101.13084 7185.4875 7185.4875 7185.4875 7199.1856 7199.1856 7199.1856 0.001187 0.001187 0.001187 0.000603 0.000603 0.000603 mean satellite altitude 7185.4841 7185.4841 7185.4841 7199.1847 7199.1847 7199.1847 82 The reason that the orbital parameters are not constant for each pass is that the satellite is subject to many forces that tend to cause its orbit to vary. The largest of these forces is the fact that the earth is not a perfect sphere but an oblate spheroid. King-Hele (1958) and Brouwer (1959) developed mathematical solutions to describe this per- turbation whose primary effects on the orbit include changing the orbital plane precession rate and changing the period. A secondary influencing factor is the effect of atmospheric drag on the satellite which acts to change the eccentricity and is a function of the satellite's altitude. A third influ- ence is the effect of solar wind and radiation. It should be noted that during 1980, the International Solar Maximum Year, solar flare and sunspot activity reached some of the highest levels recorded (Ponte, 1981) . Lesser influences in- clude the gravitational effects of the sun and the moon on the satellite. d. Computer Navigation Program The main computer navigation program was developed under the following premise: given the orbital parameters of NOAA-6, the latitude and longitude of an AXBT, and a satellite image upon which one landmark has been identified as to (scan line number, sample number) and latitude and longitude, calcu- late the (scan line number, sample number) of the AXBT so that the count value, hence the sea surface temperature, of that pixel can be identified readily either on IDIMS or any other 83 computer system. Procedural methods were outlined by Mueller (1981) . (1) Preliminary Programs. Three preliminary com- puter programs were designed to be run on the IBM 3033. The first program, SCANLINE, was a simple block counter that counted the number of data records on each magnetic tape. The number of data records minus the header record divided by 3 gives the total number of scan lines per pass. The program listing can be found in Appendix B. The second program, TAPEDUMP, was a routine designed to dump from the magnetic tape any number of bytes per data record and to translate their ASCII-Binary for- mats into decimal notation. This program was used to verify that there were indeed six scan lines per second and its listing can be found in Appendix C. The third program, AREAMAP , was designed to function in a manner similar to the IDIMS ' PICPRINT function in that it would extract from the magnetic tape the count values of a selected grouping of pixels around the landmark or AXBT pixel. This program was used to verify landmark locations and to determine the surface thermal struc- ture around the position of the AXBT. Its listing can be found in Appendix D. (2) Common Case Geometry. In the development of the main computer program, it was necessary to consider four cases in the process of predicting an AXBT pixel. These four cases are: (a) an ascending pass where the landmark has a sample number greater than 1024 (Case 1) ; 84 (b) an ascending pass where the landmark has a sample number less than or equal to 1024 (Case 2) ; (c) a descending pass where the landmark has a sample number greater than 1024 (Case 3) ; and (d) a descending pass where the landmark has a sample number less than or equal to 1024 (Case 4) . Common to each of these four cases was the assumption of a spherical earth with a radius equal to the earth's radius at the landmark latitude. This local radius can be calculated using Equation (2) 2 2 cos (L ) sin (L ) -\/2 R = C(3443.925) + (3432.381) ] (Eq* 2) where: R = local earth radius in nm L = landmark latitude in degrees, o Also common to all four cases were the calcu- lations to determine the great circle distance between the subsatellite point and the pixel containing the landmark. These calculations refer to Figure 26 below. These calculations are made possible under the assumptions that the subsatellite point is directly be- neath the satellite on the earth's surface, that the scanning mirror of the AVHRR forms a scan line perpendicular to the satellite's velocity vector, and that the earth is a perfect sphere . 85 Figure 2 6 Determination of great circle distances subsatellite point landmark pixel earth's where: H R e = *- = mean satellite altitude from Eq. 1 local earth radius from Eq. 2 scan angle great circle distance geocentric angle zenith angle Determination of the scan angle (6 ) in degrees assumes an equal division of the arc viewed by the radiometer (110.8 degrees) into 2048 samples, thus s = (landmark sample-1024) ( 110 . 8/2048) if sample > 1024 (1025-landmark sample) (110 . 8/2048) if sample 1024 The zenith angle (e ) in degrees now can be found (R+H) (sin 8 ) 9 = sin-1[ 5 — ] . p K 86 The geocentric angle (6 ) in degrees is found from Equation (3) . i =6-9 (Eq. 3 g p s H If desired, the geocentric angle in degrees can be expressed as the great circle distance in nautical miles from Equation (4) . d> = 60 6 (Eq. 4) g g (3) Units and Notation. Because the IBM 3033 and the Fortran computer language were used heavily during this project, all angles were converted to or used in radians Table 7 lists the common conversion factors used. Notations on all figures included in this project were designed to have the same definition whenever possible so comparisons could be made between the four cases. Table 7 Program Conversions 45 degrees = J- radians = (8) [(tan (1.0 radians)] , , . (same angle in degrees) , , „ . any angle m radians = 2-j-1= (tt/4) any geocentric angle in radians expressed as a = _45 entric angie)(60) great circle distance t/4 r in nautical miles 1 nautical mile = 1.835 kilometers 87 (4) Nodal and Subsatellite Point Calculations. The main program, LOCATE, is divided into two sections. The first section calculates the orbital characteristics referenced to the previously-identified landmark. The desired output is the time and longitude of the ascending or descending node. Their calculation is dependent on the four cases enumerated above and described in detail below. Once the time and longi- tude are known, the second part of the program can proceed to calculate the pixel number for an AXBT. To begin, the calcu- lation of the ascending or descending node and time follows for each of the four cases. Case 1. Ascending pass with landmark sample number greater than 1024. The derivation of orbital characteristics referenced to a single landmark in Case 1 made use of Figure 2 7 below. At time equal 0, the satellite is directly over the subsatellite point that has an unknown latitude (L ) and longitude (X ) . The only known quantities are that the scan line that includes the subsatellite point also includes the landmark with known latitude (L ) ; longitude (XQ) ; and from IDIiMS, a known scan line number (NL) ; and sample number (NS) . From the ephemeris data set for this pass, the inclina- tion (and hence retrograde inclination (i_) ) and the period are known. From Equation (3) or (4) , the great circle dis- tance t is known. The goal of this orbital set of calcula- g tions is to find the latitude and longitude of the subsatellite 88 Figure 27 Case l--orbital characteristics satellite-* ground trac north pole equator point, the longitude of the ascending node (X ) and the time of the ascending node. By using similar triangles and the Law of Sines, the angle (s) can be determined as follows: from triangle I sin z sin L o sin p from triangle II sin (i_ - z) sin cj> sin

d>. = COS [ — ] Yt COS Finally the latitude of the subsatellite point (L ) can be determined using Equation (5) and triangle III L = sin~ [ sin(i_) sin ( ) ] . (Eq. 5) For the moment, the rotation of the Earth is ignored, so the change in longitude at the equator between the landmark longi- tude (A ) and the fixed ascending node longitude (,\ — where o an the "f" indicates a fixed or non-rotating earth derived term) can be found by solving triangle I as follows: r- , cos i> AAf = cos"1! J* an cos L The change in longitude at the equator between the subsatellite point and the fixed ascending node longitude can also be found from triangle III , cos i> AA = cos" [EoTT- s 90 Now, the longitude of the subsatellite point (\ ) can be calculated easily using Equation (6) A = A (AA - AA) . (Eq. 6) s o an The fixed ascending node longitude is now found by an o an To account for the earth's rotation during the time the satellite traveled from the ascending node to the subsatellite point, the longitude change due to rotation (AA ) is calculated and subtracted from the fixed ascending node longitude as follows: At (seconds) = =— (period in seconds) , (Eq. 7) where At is the time the satellite took to travel between the ascending node and the subsatellite point. Continuing, A,\ = 2T d-002738) (At) (Eq. 8) where 1.002738 is the sidereal day correction factor. Finally, X = \f - AA . (Eq. 9 an an r The time of the ascending node can be found by subtracting the At in seconds from the landmark time in seconds. Landmark time can be read from the magnetic tape using the tape dump program mentioned above to dump the data record containing the scan line of the landmark. Case 2 . Ascending pass with landmark sample number less than or equal to 1024. The orbital calculations for this case are very similar to those of Case 1 and have the same goals. Figure 28 shows the geometry applicable in this case. As in Case 1, the known values are the great circle distance $ from Equa- tions 3 or 4 and the landmark's latitude (L ), longitude (A ), o * o and the scan line number and sample number (NL,NS) . Figure 28 Case 2 — orbital characteristics satellite n, ground track north pole subsatellite point (LS,AS) an 92 Using spherical triangles I and II, the angle (e) can be determined as follows: from triangle I sin (i- + e) sin L sin d> from triangle II sin s sin q sin

) also can be found by -1 tan 'a = cos [-. — i \-jt~ r t (sin e ) (tan

AA = cos" [EHs-lT s The longitude of the subsatellite point (X ) can be determined from Equation 14 xs " "o + (AXdn " iA) ' (Eq- 14) and the fixed earth descending node longitude now also can be determined xdn = Xo + AAdn • Using Equations (7) and (8) , the degrees of longitude through which the earth turns while the satellite travels between the subsatellite point and the descending node can be determined (AX ) , and from this the rotating earth descending node 96 longitude can be found using Equation (15) dn dn r (Eq. 15) The time of the descending node now can be determined by adding the time calculated in Equation (7) to the landmark time . Case 4. Descending pass with landmark sample number less than or equal to 1024. In this final case, the goals and the known quantities are the same as in the other cases described above Figure 30 is used to describe the geometry associated with this case. Figure 30 Case 4 — orbital characteristics north pole satellite ground track subsatellite point equator- 97 The calculations for this case are exactly the same as those for Case 2 with some exceptions as noted below. The distance the satellite travels from the subsatellite point to the descending node, i , becomes , COS ip -1 r O, d>. = COS [ ] Yt COS 4 The latitude of the subsatellite point, L , now can be found using Equation (16) as follows: L = sin [ (sin i_) (sin <$>.) ] . (Eq. 16; s ' t The fixed-earth change in longitude between the landmark f dn longitude and the descending node longitude, AA , , now can be found by f -lrcos *o. AXdn ■ cos [^nr] ' and the change in longitude between the subsatellite point longitude and the descending pass longitude, AA , can be found from triangle III , cos $ AA = COS [ T COS L s The longitude for the subsatellite point, \ , is now found using Equation 17 \ = X - (AA - AA-,„) , (Eq. 17) s o an 98 and the fixed-earth descending node longitude becomes xdn ■ Xo + "In Again, using Equations (7) and (8), earth rotation is con- sidered now, with the rotating earth descending node longitude described by Equation (18) A, = A, + AA . (Eq. 18) dn dn r The time of the descending pass is determined by adding the time calculated in Equation 7 to the landmark time. (5) Buoy Pixel Identification. With the calcu- lation of the time and longitude of the ascending or descend- ing node, the second part of the program can proceed to calculate the (NL,NS) of the AXBT. The previous set of calculations referenced the orbital characteristics to the landmark pixel (remember pixel = (NL,NS) ) . The geographical relationship between the latitudes and longitudes of the landmark and the AXBT are known so the purpose of the remaining part of the program is to transform this geographical relationship into satellite image coordinates of (NL,NS) . Besides the quanti- ties calculated above, the only other known quantity is that the AXBT has a unique (NL,NS) . As a first guess, any arbitrary scan line can be chosen to be the "true" scan line containing the AXBT. One of the 2 048 samples along the "true" scan line could be the 99 sample number of the AXBT. The aim of the calculations below is to prove or disprove, geometrically, that the arbitrary scan line is the true AXBT scan line and, once the true scan line is selected correctly, to calculate the correct sample number. The procedure to determine the time that the satellite recorded the scan line containing the landmark was described above. Since the time of the ascending or descend- ing node was calculated in the earlier part of the main pro- gram, subtracting the two times describes the satellite flight time between the particular node and the subsatellite point of the landmark scan line. The assumption was made earlier that the satellite's orbit can be considered circular with a mean altitude (H) and that the period of the satellite was the amount of time it takes the satellite to complete one orbit; then the flight time between the node and the landmark scan line can be described as a great circle distance in radians by = 2, (flight time) 19) rs period ^ Another assumption was made earlier, which was proved by using one of the preliminary computer programs described above, that the satellite records six scan lines per second. Combining these factors, it is simple to determine the difference in scan lines between the landmark scan line and the first-guess, arbitrarily-chosen scan line; divide by 6 to get the time difference between the two scan lines; either add to (ascending passes) or subtract (descending passes) this time difference 100 from the flight time; and use Equation (19) to determine the distance between the particular node and the subsatellite point of the arbitrarily-chosen scan line. The latitude and longi- tude of this subsatellite point now can be determined using spherical triangles. Assuming that the arbitrary scan line is very close to the true AXBT scan line, the next step is to find the correct sample number along this line. Beginning with sample number 1, and then taking every 89th sample (1,90,179, ...,1959,2048) the distance between the subsatellite point and the center of the pixel can be found using Equations (1) through (4) for each of the 24 sample numbers. With this distance it is possible to calculate the latitude and longitude of the center of each of these 24 samples by using one of the four cases described below. Case 1. Ascending pass with sample number greater than 1024. This case uses the geometry of Figure 31. The angle (e) can be found easily by , cos i_ e = cos"1[— — =r-] ; (Eq. 20) COS L s from which i = cos"1 [(cos L ) (sin e) ] . (Eq. 21) The distance, in radians, of d can be found by using 101 Figure 31 Case 1— pixel determination satellite ground trjcK north pole subsatellite point possible AXBT Position (Lp A equator. n sin L -1 r s-, i = sin [ . . ] s sin l (Eq. 22) hence the latitude of the possible AXBT position (L ) can P be calculated L = sin [ (sin (d> --)>)) (sin i ) ] p s g s The angles (y) and (AA ) now can be determined by ir Y = sin , cos 1 "I r S cos L ] / 102 and &\ = sin [■ P (sin y) (sin <|> ) cos L 3-] ; thus the longitude of the possible AXBT position (A ) is determined simply by A = a + AX p s p Case 2. Ascending pass with sample number less than or equal to 1024. Case 2 geometry uses Figure 32 . Figure 32 Case 2 — pixel determination satellite ground track north pole possible AXBT position (Lp, Ap subsatellite point 0-..A.) equator. 103 The angles (e) and (i ) and the distance ( ) are found using Equations (20) , (21) , and (22) respectively. The latitude of the possible AXBT position becomes L = sin" [sinU + 6) sin(ig)] The angle (AX ) now is found by P , (sin e) (sin ) AX = sin" [ = 2_] , p cos L and therefore the longitude of the possible position is X \ - A,\ p S p Case 3. Descending pass with sample number greater than 1024. Case 3, although using Figure 33 for its geometry, uses exactly the same equations as in Case 1 with, the exception that the longitude of the possible AXBT position is found by X = X -AX p s p 104 Figure 33 Case 3 — pixel determination north pole satellite ground track (LS.AS) subsateilite point possible AXBT position (Lp, A„ equator. Case 4. Descending pass with sample number less than or equal to 1024. Case 4, although Figure 34 represents its geometry, uses exactly the same equations as in Case 2 with the exception that the longitude of the possible AXBT posi- tion is found by X + AX s p In addition to calculating latitude and longitude for each of these 24 samples along the arbitrary scan line, it also is necessary to calculate the great circle 105 Figure 34 Case 4 — pixel determination north pole satellite ground track position (Lp, Ap subsatellite point 0-..AJ equator distance (d) between the AXBT geographical position and those of the 24 samples using Equations 23 d = cos" [ (sin L, sin L ) + (cos L, cos L ) cos (AD-:vvJ 1 ' ^Eg* 23^ where: true AXBT latitude true AXBT longitude (Eq. 2 3) By taking the sample number that has the smallest of these great circle distances, creating a bracket 106 89 samples wide on either side of this center sample, and pro- ceeding as before through the appropriate case number for each of the 180 samples in this bracket, the sample that has the shortest great circle distance between itself and the AXBT position can be selected. The reason for the wide bracket is to allow for earth rotation and curvature whose effects are especially noticeable on the edges of the satellite image. If, as before, we assume that we had a scan line very close to the true AXBT scan line, a 10 by 10 pixel "square" box is created around the sample with the smallest great circle dis- tance. In reality, this box is not perfectly square due to the curvature of the earth and the motion of the satellite during the scan sequence. Those boxes near the subsatellite point would be more perfectly "square" than those boxes on the edges of the image. The scan lines on the top and bottom of the box as well as the samples on each of the four corners are subjected to the same calculations described above to find the latitude and longitude of each of the samples on the four corners. The geometry of this box is shown in Figure 35. The calculations to find the (NL,NS) of the AXBT begins with finding the change in the latitudes and longitudes with respect to the changes in the sample and scan line numbers. 3L lrL3~Ll L4"L2, — -> L in + in J 3 (NL) 2 L 10 10 107 («-,,* Figure 35 Box geometry L4,A4) 0-2^ 2 ) (LrV where L. = l X . l L ,\ o o L ,A P P L , L, c d latitude of sample on i-th corner longitude of sample on i-th corner latitude and longitude of the sample having the smallest great circle dist AXBT latitude, longitude check procedure latitudes 3L (NS) 1^2^! + L4'L3 2 L 10 10 3 X (nl; lrX3"Xl VX2 2 L 10 .0 (NS) 1 X2"M :v4""3 2 L 10 10 108 Continuing, two intermediate equations are required A = (Xp-X1) + jj^y (sample A) + -|A_ (scan line x) . B = (lp"li) + msT (samPle A) +TW (scan line x) Finally, determination of the (NL,NS) of the AXBT can be made A B 3A/3(NS) 3L/3(NS) 9 A/3 (NL) 3L/3 (NL) 3 A/3 (NS) 3L/3 (NS) v,c _ A ,3 A/3 (NL) , , , NS " 3A/3(NS) L3\/3(NS) (NL,J Rarely will the arbitrary scan line chosen as a first guess be close to the true scan line of the AXBT. In this case, after the great circle distances have been calcu- lated between the initial 24 samples and the AXBT position, and the "square" box has been set up, a check is made to see how "small" is the smallest great circle distance. As described earlier, the satellite follows a ground track as it travels poleward that cuts across meridians of latitude at an angle set up by its inclination. This means that scan lines are not oriented east-west along degrees of longitude but are oriented northwest-southeast (descending pass) or northeast-southwest (ascending pass) crossing many degrees of longitude and lati- tude. The check involves comparing the latitude of the AXBT 109 with the latitudes L and L, as shown on Figure 35 above. If c d the arbitrary scan line is very close to the true AXBT scan line, for ascending passes, latitude L will be south of the AXBT latitude while latitude L. will be north of the AXBT latitude. The reverse is true for descending passes. If the arbitrary scan line is far away from the true AXBT scan line, both L and L, will be either north or south of the AXBT c d latitude. In this case, the smallest great circle distance is converted to an integral number of scan lines which are added to or subtracted from the first-guess arbitrary scan line number depending on whether L and L, were both south or north of the AXBT ' s latitude respectively. The jump to a new scan line initiates the entire procedure again beginning with the steps necessary to calculate Equation (19). This jump process terminates when the number of scan lines to be jumped is 5 or less at which time the boxing procedure begins with the eventual determination of the AXBT ' s (NL,NS) . The main program, whose listing may be found in Appendix E, was initially set up to be run interactively on a display terminal. The program was used to locate all the AXBT's that were dropped from the P-3C. Verification of the accuracy of the program was done by selecting 15 to 2 0 land- marks per image and asking the program to predict the (NL,NS) even though they were known already from the I DIMS system. Results of this verification will be discussed below in Section IV. A. 110 C. GOSSTCOMP The GOSSTCOMP sea surface temperature charts were obtained for the period of this project from NOAA-NESS. These charts are produced on a weekly basis by NOAA-NESS using procedures outlined by Brower et al. , (1976) and since updated to take advantage of the AVHRR on NOAA-6. Ill IV. RESULTS A. NAVIGATION ACCURACY A major effort was made on this project in an attempt to reduce the effects of geometric distortion associated with locating landmarks or open-ocean positions on satellite imagery. Previously published works with earth location errors in ex- cess of 10 kilometers at nadir were suitable for regional location and analysis of mesoscale features; however, it was believed that accurate comparisons of thermal data were signi- ficant only if the products being compared were co-located in the same geographical position. The location of thermal features is especially important in naval tactical applications. If a submarine were taking advantage of the unique acoustical properties of an eddy or ocean front, then acoustical prosecu- tion by the opposing forces would be more successful if the sensors employed by this group were located so as to take advantage of the thermal feature also. If pre-mission informa- tion mis located the edge of the front or eddy due to the geometric distortion inherent in satellite imagery, then the results could be disastrous to one of the parties. Ultimately, any error in sensor placement could prove disastrous whether caused by satellite imagery or not; part of the purpose of this project was to make the location error as small as possible Using the main computer program, the (NL,NS) of each AXBT was predicted. The error in this prediction was determined to be within 2 pixels of the true AXBT (NL,NS) . The procedure 112 to verify this accuracy began with the identification of the (NL,NS) of up to 20 landmarks on each satellite image. Each of the landmarks then was treated like an AXBT and its geo- graphical coordinates were input to the computer to see what the program would predict for each landmark's (NL,NS) . These predictions then were compared to the IDIMS-determined (NL,NS) , and the separations in pixels were determined. The results are summarized in Table 8 below. Table 8 Statistical Summary of Navigation Accuracy SCAN LINE ERROR mean 1.32 scan lines standard deviation 0.9 3 scan lines 99% confidence level 0.88-1.90 scan lines SAMPLE ERROR mean 1.35 samples standard deviation 1.17 samples 99% confidence level 0.31-1.90 samples The conclusion drawn from this statistical summary was that the predicted AXBT (NL,NS) is within 2 pixels of the true AXBT (NL,NS) . Because of earth curvature, pixels close to nadir are not as wide as those out on the edges of the image. Sample number 1 and 2048, on -che right and left edge of the image respectively, are 4.3 kilometers wide whereas sample number 1024 and 1025, located to the right and left of nadir 113 respectively, are only 0.77 kilometers wide. As a result, the 2-pixel error can be as small as 1.9 kilometers, if the predicted (NL,NS) is at nadir; or as large as 10.7 kilometers, if the predicted (NL,NS) is at the edge of the image. Table 9 lists the navigation error associated with selected sample numbers. Notice that the error is not linear with distance from nadir but is less than 5 kilometers over 8 0% of the image and less than 3 kilometers over 5 0% of the image. Only on the outer 10% of the image does the error balloon from 4.8 to 10.7 kilometers. Table 9 Navigation Errors Associated with a 2-Pixel Error SAMPLE NUMBER ERROR (km) 1 10.7 200 4.8 400 3.0 600 2.5 800 2.0 1024 1.9 1200 2.0 1400 2.3 1600 3.0 1800 4.3 2048 10.7 114 There are some other sources of navigation error that should be kept in mind when using data developed by this method. Alignment of the AVHRR module during its assembly prior to launch could be the source of constant offset error. This type of error was described previously in this paper; analy- sis of the navigation results did not show any consistent off- set bias that could be attributed to module alignment errors. Through the process of verifying the 2-pixel accuracy, many landmarks were identified on the IDIMS system as explained above. The data from NOAA-6 infrared channel number 4 were used for landmark identification and their use could introduce errors in assignment of the (NL,NS) . These errors arise from trying to identify ' landmarks whose surface temperature may not be very different from the surrounding surfaces. This effect would become even more pronounced if ground fog were present. When selecting these landmarks, the best contrast was effected by land-water boundaries, examples of which were Point Lobos west of Monterey, the San Francisco Bay entrance, Alcatraz Island, Point Reyes, the Columbia River mouth, Lake Tahoe, and Glacier Bay among others. Many possible landmarks were not considered if there were insufficient contrast to identify the feature. A good example of this was the Seattle- Tacoma-Olympia area where the numerous bays and tributaries had surface temperatures close to land temperatures, thus making it very difficult to distinguish a specific pixel as being some peninsula or promontory. 115 A third source of error involved the number of signifi- cant figures used in the mathematics of this project. Single- bit precision was used during computer processing. Although this may have had an effect after numerous computations (the average program run to calculate 2 4 AXBT positions per image executed 475,000 statements) , it was felt that the number of significant figures was more critical. An example of this was the determination of decimal geographical coordinates. The navigation system on the P-3C supplied the computer-calculated coordinates of the AXBT's to seconds of latitude or longitude. One second of latitude error is equal to 0.1 kilometers, which in itself is not so large; however, most landmarks were identified using charts with scales of 1:2,000,000. After determination of the coordinates, a decimal conversion to three decimal places was completed. If errors in this procedure were compounded by weak land-water contrast on the infrared image used for landmark identification, it could contribute significantly to the 2-pixel error. A fourth source of error has to do with the resolution of the AVHRR itself. As discussed earlier, the 1.1 kilometer resolution would necessarily make it difficult to identify something like the Transamerica Building in downtown San Francisco. An example of where this could be a contributing factor to the 2-pixel error would be in using the most western point of Point Reyes. If the scan sequence is such that the radiometer does not resolve this point, then the first pixel it does identify as being land would be to the east of the 116 point. The user of the satellite image would have a very difficult time in trying to determine whether or not this has occurred. As a result, the user would assign the geographi- cal coordinates of the most western point, introducing a 1- pixel error immediately before any computer processing begins. It is felt that if landmark's could be identified using sharper land-water contrast or using a visible channel vice the infra- red channel if it is available, and if geographical coordinates could be assigned with greater accuracy, the majority of the 2-pixel bias would be eliminated. Another error to be considered is that the satellite may not be perfectly stable in its orbit. It is likely that small amounts of pitch, roll, or yaw occur from time to time although the ADACS system was designed to keep these attitudes to a minimum. The last error to be discussed is the use of several assump- tions made during this project. The Earth was assumed to be spherical and although the radius was calculated to be that radius at the latitude of the landmark, a small error will be introduced in calculations involving the earth radius term at the latitude of the buoys. Similar small errors arise with the assumptions made concerning the satellite orbit during scan line calculations, and with the calculation of the mean altitude of the satellite above the earth's surface. In conclusion, it is believed that the 2-pixel error found on this project could be reduced further to sub-pixel accura- cies if some of the errors described above were eliminated 117 or refined. Since the development of LOCATE, a program using most of the same techniques as LOCATE has been developed to predict the geographical coordinates of open ocean images with similar accuracies (Mueller, 1981) . B. THERMAL COMPARISONS 1. Horizontal Distribution As usually can be expected when satellite-derived sea surface temperatures that are uncorrected for atmospheric attenuation are compared with AXBT values of sea surface temperature, the satellite temperatures were colder than the AXBT data by a mean difference of 2.9 degrees C. Table 10 lists the mean temperature difference values and the corres- ponding standard deviations for this and following comparisons, Figures 36, 37, 38, and 39 show sea surface temperature com- parisons of this and other methods to be described below along the buoy line. The majority of the 2.9-degree error can be attributed to the effects of the intervening atmosphere be- tween the ocean's surface and the satellite radiometer. Cloud contamination of the satellite values was not considered a major factor due to the screening process that went into se- lecting the data. Out of the six satellite passes selected for study at the beginning of this project, only three met the full requirements that were required for processing. Two of the passes were not considered due to scattered clouds over enough buoy positions to make any comparisons useless and one pass was not considered because it contained no clearly 118 Table 10 Temperature Comparison Statistics COMPARISONS satellite vs. AXBT 17 November 01 December 05 December overall MEAN (C) STANDARD DEVIATION -3.0 0.5 -2.6 0.4 -3.0 0.6 -2.9 0.5 satellite vs. GOSSTCOMP 17 November 01 December 05 December overall GOSSTCOMP vs . AXBT 17 November 01 December 05 December overall •2.0 •3.2 •4.3 •3.2 •0.7 0.5 1.2 0. 3 0.9 0.6 0.9 1.1 0.7 0.6 0.6 1.0 119 Figure 36. Sea surface temperature comparisons, 17 November 1980, center track 20 o CC15 5 LU Q. UJ1Q- Hi o £ < 11! AXBT' raw N0AA-6 data 0 I—*— I ■ ■ ' I I I I I ■ I I I 1 1 1 1 11 191817162524232221 3241506 7 8 9 0 1 2 3 4 BUOY NUMBER 120 Figure 37. Sea surface temperature comparisons, 1 December 198 0, center track 20i o HI §15 < 111 CL UJ t- LU o < U. CC 3 CO < UJ CO 10- 5- »■ i i GOSSTCOMP^. y ■ f» p , | ti | i ■ | 1 11 1 1 1191817162524232221 3241506 7 8 9 0 1 2 3 4 BUOY NUMBER 121 Figure 38. Sea surface temperature comparisons, 5 December 1980, center track 201 c Ui §15 5 X Ui 0. Ill 10 o 2 - CO < UJ CO 10 G0SSTC0MP *Vraw NOAA-6 data 555555666666 456789012345 BUOY NUMBER 123 identifiable landmarks. Passes selected for complete pro- cessing were 17 November, 1 and 5 December 1980. From personal observations onboard the P-3C during AXBT deployment, it was noted also that there was no ground fog to interfere with satellite measurements of the AXBT positions. The time difference of one, three, and three hours between the last buoy drop and the satellite flyover for 17 November, 1 and 5 December, respectively, probably is not a factor as the lowest and highest mean error values were found on 1 and 5 December with 17 November having an intermediate value. If there were a correlation, one would expect 17 November to have the smallest error but this was not the case. The transient warming of the surface waters during the afternoon, the so- called afternoon effect, did not occur due to the weather conditions during the three-week project period; hence, this process also was ruled out as a source of the error. Con- stant wind speeds in excess of 20 knots from the south on 17 November, in excess of 25 knots from the northeast on 1 Decem- ber, and in excess of 15 knots from the westnorthwest on 5 December (National Weather Service, 1980a) along the buoy line kept the surface waters under constant wind-mixing condi- tions. In addition to the winds, ship observations of the sea state at the northern end of the buoy pattern found four to ten foot swell and two to six foot waves. These turbulent mixing conditions are diametrically opposed to the formation of the afternoon effect (James, 1966). 124 A method of "field-calibrating" the satellite data to eliminate the effects of the atmosphere was suggested by Tabata and Gower (1980) . Using a simple linear regression technique, they plotted numerous ship-obtained surface tempera- tures versus satellite count values and found that over a limited area and a limited time period between satellite and ship observations (1.5 days), the error between satellite and ship values could be reduced to 0.5 degrees C. This tech- nique was tried using the data from this project. An absolute mean difference of 0 . 3 degrees (s = 0.2 degrees) was found between satellite and AXBT values. The time period between satellite and AXBT observations was three to eight hours. Flight crews usually will not have the luxury of ex- pending 24 AXBT's on a tactical mission however, so the linear regression technique was tried using only two buoys. The rationale behind using two buoys was that this is the number of AXBT's usually carried on both S-3A and P-3 aircraft. Addi- tionally, a scenario could exist whereby a satellite photo obtained prior to the flight could pinpoint two sections of the tactical operating area where thermal differences exist and those two locations could be designated for AXBT deployment The point is to try and get a spread in temperature between the two AXBT's. Using the two-buoy method, the data from AXBT positions 1 and 13 on 1 December were used for the linear re- gression. Predictions of temperatures from count values found a mean difference error of 0.3 degrees (s = 0.3 degrees), the same value found using the 2 4-buoy method. When this same 125 regression formula with constants calculated from 1 December data was used to predict 5 December (4 days later) and 17 November (15 days earlier) temperatures, mean errors of 0.5 degrees (s = 0.3 degrees) and 0.4 degrees (s = 0.3 degrees), respectively, resulted. Although examination of all possible cases would be necessary before conclusive results could be stated, these preliminary estimates indicate that it should be possible to use two AXBT's and an infrared satellite image to predict sea surface temperatures within 0.5 degrees for at least two days after the original satellite pass, a prediction tool particularly helpful if clouds obscure the sea surface during those two days or if AXBT assets are in short supply. In addition, these procedures can be used in near real-time processing of current satellite images and do not rely on any atmospheric model processing. In any case, these thermal predictions that are very accurate in location and fairly accurate in temperature would be tactically significant, in place of mean values or best-guess values, when doing sound velocity calculations near meander, eddy, or frontal regions. Relative temperature gradient analysis displayed the expected correlation between satellite and AXBT values. Both the AXBT and satellite gradients were 0.6 degrees per 60 nm on 17 November and 1 December while on 5 December the AXBT gradient was 0.56 degrees per 60 nm and the satellite gradient was 0.52 degrees per 60 nm. An interesting thermal feature whose horizontal sur- face manifestation was detected by both the satellite and 126 the AXBT was a meander in the final stages of closing off from its parent body of water to form an eddy. See the darker region transected by the buoy line through buoy positions 23, 3, 22, and 4 in Figure 40. This warm-core meander had an approximate 100 nm diameter with the exception of an open arm extending southward into its parental water mass. The diameter was verified by the buoys dropped on the northern and southern tracks. These buoys were 6 0 nm away from the center rack and the warm meander did not show up on any of the thermal traces. The satellite indication of this diameter resulted in a slightly larger radius, a fact attributed to the thermal resolution limi- tations of the satellite data in determining weak temperature boundaries. In the satellite images, this meander is surrounded on the west, north, and east by the Subarctic Current-California Current confluence. The center of the meander had a surface temperature of 15.8, 14.9, and 14.4 degrees C on 17 November, 1 and 5 December, respectively. A chart of the monthly mean surface temperature for November 1980 (Renner, 1981) clearly shows the intrusion of a large tongue of warm water from the area between San Francisco and Hawaii northward along the west coast of the United States. The decrease in the surface temperature of the meander over the project time period was reflected by the decrease in the surface temperatures all along the buoy line. Both the satellite and the AXBT ' s recorded mean changes of 0.9 degrees between 17 November and 1 December and 0.7 degrees between 1 December and 5 December. This drop in temperature 127 128 is to be expected considering the weather conditions as men- tioned above. On 17 November, a low pressure system was firmly entrenched over the Aleutians while a high pressure system was anchored off of Southern California. This pressure pattern is typical of the Northeast Pacific in early winter. A cold front extending southward from the Aleutian low moved across the buoy line during the evening of 17 November and crossed the U.S. coastline during the morning of 18 November. Winds before passage were southerly at 25 knots while after passage the wind shifted northerly at 30 knots; hence the condition for considerable wind-mixing existed. A series of cold fronts on 20-22 November and 25 November also passed through the project area continuing to lower the sea surface temperature and drive the mixed layer depth deeper. On 1 December, low pressure cells were established west of the coast of Washington and about 500 nm west of the central California coast. The Washington low strengthened and cen- tered near buoy positions 7 on 2 December. This low was accompanied by winds in excess of 3 5 knots on 3 December over the entire buoy area while a slow-moving cold front hugged the coastline. On 4 December, a high pressure area, previously established in the Gulf of Alaska, moved into the project area from the north pushing the cold front well inland although the low remained off the Washington coast. The high moved southerly on 5 December, influencing the weather over the entire buoy area (National Weather Service, 1980b) . See 129 Figures 41, 42, and 43 for the surface weather depiction charts for 17 November, 1 and 5 December, respectively. Because the linear regression model mentioned earlier was used to remove the effects of the atmosphere with fairly good results, a comparison was made between the satellite data and the GOSSTCOMP product. Sea surface temperature pro- ducts from GOSSTCOMP have been subjected to an atmospheric correction model and are issued on a weekly basis. On all three days, the satellite data from this project were colder than the GOSSTCOMP values. The mean difference for 17 November, 1 and 5 December were 2.0 degrees (s = 0.9 degrees), 3.2 degrees (s = 0.6 degrees), and 4.3 degrees (s = 0.9 degrees) respectively with the overall mean of 3.2 degrees (s = 1.1 degrees). This overall mean agrees fairly well with the 3.5 to 3.9 degree bias enumerated by Klein (1979). The reason for the 3.2 degree bias can be attributed directly to the effects of the atmosphere, exactly the same situation as seen in the AXBT versus satellite comparisons mentioned previously. An interesting point to be made is that the 3.2 degree bias of GOSSTCOMP versus satellite data is higher than the 2.9 degree bias of AXBT versus satellite data. This led to a comparison between GOSSTCOMP and AXBT data with the result that GOSSTCOMP values were 0.3 degrees (s = 1.04 degrees) warmer overall than the AXBT values. Because the project area was never totally cloud-free during the period of observations, it is felt that the overcorrection for atmospheric effects described by Klein (1979) is still a factor in the warmer GOSSTCOMP values. It 130 o o o o o oo y X >7 .<• 131 132 IVJ O o o o o CO UJ «3 c 133 should be mentioned that the GOSSTCOMP product did not indicate the warm meander that the AXBT and satellite data located, probably due to the large grid structure used by GOSSTCOMP. All three methods of sea surface temperature determin- ation, AXBT, satellite, and GOSSTCOMP, were compared to the 20-year mean surface temperature values of Robinson (1976). AXBT values versus climatology resulted in the November AXBT ' s being 0.3 degrees colder than the mean while the December values were about the same as the mean. The reason for the cooler surface waters in November is probably a result of the high incidence of weather frontal passage with accompany- ing high winds through the project area. As was expected, climatology did not show the warm meander. Comparison of satellite versus monthly mean data found the satellite data averaging 3.0 to 2.7 degrees colder than the mean for November and December. Comparison of GOSSTCOMP versus the mean resulted in GOSSTCOMP being 0.4 degrees warmer than the mean for November and 2 . 0 degrees colder than the mean for December. 2 . Vertical Distribution There is no known way at present to sense remotely the vertical thermal structure in the ocean; however, if one combines knowledge of the horizontal gradients with clima- tology, a fairly accurate synopsis of the upper ocean thermal structure is possible. A more accurate picture can be 134 formulated if the satellite data are augmented with well- placed AXBT drops. From climatology, the expected mean surface tempera- ture and the mean layer depth for November were 12.9 degrees and 50 meters (s = 5 meters) respectively over the project area. The AXBT mean surface temperature and mean layer depth for 17 November were 12.6 degrees and 58 meters (s = 6 meters). For December/ climatology means were 10.8 degrees and 67 meters (s = 6 meters) and the AXBT means were 10.8 degrees and 71 meters (s = 7 meters). Figures 44, 45, 46, and 47 show the vertical structure along the buoy line on 17 November, 1 and 5 December (center track and north track) respectively. From the numerous oceanographic studies in the area (Tully, 1961; Tabata, 1961; etc.) , it is known that during this period of the year, the layer depth is deepening towards the maximum limit of the top of the permanent halocline at 100 meters. The 100-meter depth is not reached usually until February. The deepening of the layer is directly attributable to the turbulent mixing conditions caused by the sustained high wind speeds and by the convective mixing caused by the surface cooling during the calmer periods. The weather pattern for late-November and early-December was discussed previously. A general rule of thumb is that warm surface waters generally exhibit shallow layer depths while colder surface waters ex- hibit deeper layer depths. This pattern held true throughout the project area. Although definitive sea surface temperature- layer depth relationships were not within the scope of this 135 «-f> EE«.:.8 8 S g S 8 ? °o « 8 g 8 S 8 g ~r::":::::: :::_:::.:::.::.:: (iu) Hidaa :: ... :_:^r;- 136 HI M +J +J c it) o O 2T Q jfi o in b ip c\j u> r* o (N u) "!i « S &j 8 s w g