\--'' ^v :;:.< ■ ' ' ' '' -'f ^ aT Z,'' yV'''^ ir ,'^. !."• •♦,■:- ^' /■;fr^ T;':>7*?::*«rf.f:- r;^. hV ^m^ J^^ I Snoy/ covered surface near sea level Scott's track Amundsen's track 0£ « 0{ '01 vX BRITISHMNTARCTIC EXPEDITION ^ %J9I0 1913 'i^ ROD o gmAq A / R T SUA 'S'^tt'^^^: 3 D H A MPSON, D.Sc. RR.S 6rrimiUee "ic^ the i 'ubiicaiicis oi ii)v: btit,ra;!ic Rd CALCirn "c» AiMJOJiAi "^ OKCi\, oriiN "&/ Jl. jufioMiM 1^; y-. M H H Sketch map of part of Europe drawn to the same scale as the map of the Ross Sea atea to which ii is attached. 13 10° =70 Scott's track - AmGndserfs track 0 BRITISH ANTARCTIC EXPEDITION 7a, ^l' 1910-1913 // ^ 9 "7^ YB . I METEOROLOGY Vol. I DISCUSSION BY G. C. SIMPSON, D.Sc. F.R.S Prepared under ihe directions of the Committee for the Pubhcation of the Scientific Results Chairman Col. H. G. Lyons, F.R.S CALCUTTA Printed by THACKER, SPINK & CO 1919 I 2 7 6 "CO MY WIFE THIS WORK Which has stolen from her many hours of companionship IS 'DEDICATED PREFACE. In November 1909 I received in India a cable from Captain Scott asking me to join liis expedition as meteorologist. Arrangements were quickly made with tbe Government of India who granted me three years' ' leave without pay ' and by the end of the month I was on my way to England. On January 1st I commenced five months of strenuous work during which instruments had to be begged, bought, designed, manufactured and tested and journeys made to Scotland, Germany, Norway and Denmark in order to discuss with the scientists in those countries the problems which we intended to investigate. I soon realized that my work would be seriously handicapped if I had to depend on the general funds of the expedition for the cost of my instruments. After discussing n\y difficulties with Captain Scott and nunierous friends in my native town, Derby, it was decided to issue an appeal to the inhabitants of Derbyshire to subscribe sufficient funds to give me a free hand in obtaining all the instruments necessary for my work. I cannot be too grateful to the late j\Ir. W. Crowther and to his daughter Miss Ethel Crowther who acted as honorary secretaries to this appeal, for in consequence of their indefatigable efiorts £500 was collected which relieved me of further financial anxieties. Subscriptions were received from all classes of town and county and I trust that these public- spirited givers will accept this work as some tangible return for their generous gifts. In addition to the instruments bought by the Derbyshire fund we received the loan of many special instruments particularly from the London Meteorological Office, the National Physical Laboratory, Mr. Mum-o, The Gutta Percha Company, the Australian Meteorological Office and the New Zealand Meteorological Office, to all of whom I wish to express the thanks both of the expedition and of myself. In consequence of this generosity all the instruments required by the expedition were ultimately obtained. I doubt whether any expedition sailed with a niore coniplete outfit of meteoro- logical and niagnetic instruments, but it is possible that other expeditions have come back with more recorded observations, for I find in the reports of other expeditions accounts of many observations which we never attempted. One of the chief reasons for this was the want of trained help. Captain Scott's ship did not stay during the winter in the Antarctic, but after landing the shore parties returned to New Zealand. Thus we were deprived of the scientific help which nearly all previous expeditions have received from the navigating officers of the ship when their own work was in abeyance during the sojourn of the ship in its winter quarters. There were probably more trained scientists at Cape Evans than have ever been together in any previous expedition, but each and iv PREFACE every one had liis own worlc to attend to and had little or no time to spare for the most hard pressed of his colleagues. As will be seen later I did receive considerable help, but this could not be depended upon and the programme of magnetic and meteorological work had to be planned for one man. Even then it was found impossible to carry out this limited programme, for difficulties with the instruments were unending and new problems connected with the effects of the cold and the blizzards on the self-recording instruments cropped up almost daily. Some of these had been anticipated and precautions taken, but frequently experience proved them unavailing. From tlie first the whole of ni,y time was taken up in keeping the record of the most important elements complete and there was no opportunity for making elaborate investigations of special problems. What is more, the whole conditions are so different from those with which one is fam,iliar that it takes at least a year to adjust one's ideas to them. At the end of the first year I could see many interesting problems calling for inves- tigation and had planned to follow them up during the second year's work, but this could not be done for I was recalled to my work in India, when the Terra Nova returned to the Antarctic in January 1912. The most obvious omission in the meteorological record is that of atmos- pheric humidity. There are two methods of measuring humidity : first by means of wet and dry bulb thermometers and secondly by means of hair hygrometers. The first however is very unsatisfactory at low temperatures and Mr. Dines in reviewing the two-hourly readings of the wet and dry bulb thermometers made on t»he Discovery Expedition wrote : — • ' The general opinion of meteorologists seems to be that the tables by which the relative humidity is obtained from the wet and dry bulb readings are open to doubt at very low temperatures and very low values of the humidity and hence it hardly seems worth while cal- culating the values of the humidity from these readings.' ' One cannot help feeling regret that the opportunity to revise the humidity tables at low temperatures could not have been utilised.' With the whole of the wet bulb readings taken on the Discovery Expedition unused I did not feel inclined to add to the accumulation. Also there was much more important work to be done in the Antarctic than revising the hum,idity tables, for this can be done in more accessible regions of the globe. I therefore decided to take no more wet bulb readings. The use of the hair hygrometer at low temperatures has been studied by Professor Mohn and his students in Norway, and a hygrometer designed which is said to be reliable in use. In the short time available for preparation before the expedition sailed I had no opportunity to obtain and famiharise myself with this instrument. The consequence was I removed direct observations of the humidity of the atmosphere from my programme hoping that I should be able to devise some indirect method of observation when in the Antarctic. The opportunity for this however never arose, so humidity is not discussed in the present work. PREFACE V Another omission is any account of meteorological optics. This is a subject in which I am particularly interested and I gave it constant study when in the South. With the exception however of iridescent clouds I came across no 23heno- menon which appeared to me to involve any new physical principles. All the halos 1 saw appeared to conform to the physical explanation given in such great detail by Pernter in his Meteorologische Optik, and I had not the time or instruments accurately to measure the halos to find small departures from the theory. My study of iridescent clouds led to the formation of a new theory to account for this beautiful natural phenomenon. As this has already been published in the Quarterly Journal of the Royal Meteorological Society it has not been included in this work. The measurement of precipitation is another omission, but this is discussed in its proper place in the book. Even with all these omissijns of observations and the strictly limited pro- gramme it would have baen impossible to keep the record complete if I had not received help from other members of the expedition. Captain Scott therefore arranged that during the winter the night watchman should record hourly observa- tions of the aurora and take the midnight and 4 a.m. meteorological observations. But in addition to this I needed help in work which required co-operation and I always found the members of the expedition, from Captain Scott to the cook, willing and ready to accord this help if their own duties would allow. For all this help I am sincerely grateful and to each and all my thanks are due, but naturally I received more help from certain members of the expedition and to them I would like to express my thanks in particular. First and foremost I must record my gratitude to Lieutenant H. Bowers of the Royal Indian Marine. Bowers was one of the hardest worked men with the expedition, yet he was always ready to help those who needed help. He could not be idle for a moment, and the whole of his recreation was taken in doing something for someone else. When he needed exercise he would take with him a pony for a walk or go up on to the foothills of Erebus to read the thermometer in one of the outlying screens. His greatest help to me was in connection with the balloon ascents, for he undertook the filling of the balloons while I prepared the instruments to be attached to them. Then, as it was dangerous for one man to go alone in search of the fallen instrument, we went together on many a long tramp over the rocks and sea ice surrounding Cape Evans following the trail of the long silk thread which had been attached to the instrument and should have led us to it. Bowers loved the Antarctic and he was the only man with us who was never heard to anathematize the weather or the discomforts of Polar life. One of the gleams of comfort accompanying the shock of the news of his death with the Polar Party was the knowledge that he would have considered the object worthy of the sacrifice. Another busy man always ready to help was G-rifiith Taylor, and it was only through his willingness to take over the meteorological work for a fortnight VI PREFACE that I was able to experience the joys and trials of sledging. But this was not the only occasions on which he showed his interest in meteorology by under- taking part of the drudgery of making observations, for he was always ready to take a set of observations if I was away from the hut at observation time. C. S. Wright was my right-hand man in regard to the magnetic work and during the first year in addition to his own work of chemistry and ice study he was always ready to relieve me of a set of absolute m.agnetic observations or to take the routine meteorological observations. On my return at the end of the first year he took over as much of nT,y work as possible and with the help of T. Gran continued the chief magnetic and meteorological observations, so making the record of these complete for two years. Gran's meteorological work during the second year was remarkable as he was faced by difficulties with which we did not have to contend during the first year and in working up the results I have frec^uently had cause to be grateful to him for the trouble he took to make the record complete. During the summ.er when nearly everyone was away from the hut and there was no night watchman E. Nelson volunteered to relieve me of the n\idnight observations and this made it possible for me to get a tolerable night's rest in spite of having to turn out at 4 a.m. to read the meteorological instrum,ents. Turning now from the work in the field to the work in the study I find that this part of m,y undertaking also requires explanation and the acknowledg- ment of help ungrudgingly given. The whole of the data was not in my hands until near the end of 1913 and before the discussion was half completed the war broke out. These were strenuous times and as the work had to be done in my spare time it was not until April 1916 that the discussion was conipleted. It then seemed best not to print the work until the end of the war, but as the war dragged on its weary way it was decided in June 1918 to make the experi- ment of publishing the book in India. I was then engaged on war work and had little time to give to work of this nature, which seemed of little account in view of the great issues then being fought out on the battlefields of Europe. The publishing house was also hard pressed and the printing of the work has been long drawn out, but at last in happier times the book is ready for issue. One of the effects of the preparation of such a work in such circumstances has been the impossibility of consulting original works, copies of which were not available in India. Frequently I have had to rely on extracts and reviews of previous work where I should have preferred to obtain the original papers, but the disorganized state of the world and the uncertainty of mails made this impossible. Many meteorologists will look in vain in this book for statistical results with which they have become familiar in similar works. I am no statistician and statistical meteorology has no attraction for me, therefore I have not loaded my discussion with statistical tables. These have all been banished to Volume III which will consist of tables only and will, I hope, prove a happy hunting PREFACE vii ground for the statistical meteorologist. On the other hand no statistical investi- gation has been too laborious when it has been undertaken to unravel some physical problem. The keynote of my work has been physics and I have attempted to find the physical explanation of each of the meteorological phenomena observed. It will probably be found that many of my physical explanations are not sound and will not bear critical examination, still to me an imperfect physical e^iplana- tion is better than none, for it acts as a thread to bind the facts together and being an object for attack may lead ultimately to a correct explanation while the mere statement of the facts might have been passed unheeded. Those who have the patience to read through this book will find many fascinating unsolved problems and these, I hope, will be incentatives to further investigation both in the study and in the field. A word must now be said about the units used in this work. I had to take with me the instruments available and as most of the ordinary meteoro- logical instruments were provided by the London Meteorological Office they were graduated in English units : the thermometers in degrees fahrenheit and the barometers in inches. Thus if I had not used these units in n\v discussion every reading would have had to be converted into another set of units. But what units should these be ? On my return I found the meteorological world vehe- mently discussing the relative values of the fahrenheit, the centigrade and the absolute scales of temperature, and tlie inch, the centimetre and the millibar units of pressure. The controversy still rages and we now have all these units in actual use by official meteorologists. In such an atmosphere of unrest it would have been difficult to justify the immense labour of converting my observations from the units in which they were taken into any other set of units. Therefore in the ordinary meteorological and climatological observations I have used the old English units of fahrenheit degrees in temperature and inches in barometric pressure. On the other hand, in special problems I have used the units which are most frecjuently met with in previous works. My upper air observations have been given in centigrade degrees and metric units, and when I have been discussing the work of continental writers I have followed these units even in ordinary climatological elements. Thus I lay myself open to criticisms of inconsistency, but if my inconsistency helps to facilitate discussion I am quite content. The book has been written in a transitional age and it bears the marks of the transition. I have received much help in writing this book, and in every case in which I have asked for data it has been willingly and promptly supplied. I cannot record all the help which I have received in this way but I must acknowledge that which I received from the directors of the meteorological services of Australia and New Zealand. Mr. Hunt in Melbourne and the Rev. D. C. Bates in Wellington extended to nie personal hospitality and placed the resources of their departments at my disposal. The data sent to me by Mr. R. C. Mossman have been of the utmost value and I am grateful to him for the help he has given. Dr. Gilbert viii PREFACE Walker and Mr. C. Norniand of the India Meteorological Department have always been ready to discuss my problems with me and to these discussions I owe many clear views of problems which at first appeared dark and hazy. The greater part of the reduction of the observations has been done by Babu Mohammad Bakhsh of the Simla Meteorological Office and his reliable work has been a great relief to me, for it is good to be able to trust your chief computer. The diagrams have all been prepared by Mr. E. M. Batt of Simla and as the excellency of his work is obvious to all readers, it is only necessary for me to record my thanks. The work, begun ten years ago, is now finished and about to be submitted to those who are interested in problems of Polar meteorology, but it leaves me with a sense of sadness for the two men to whom it would have made the strongest appeal and whose opinion I should have valued above all others are no longer with us to receive it. Over and over again as point after point was cleared up I have longed to be able to show the result to Captain Scott, for there was hardly a problem of Antarctic meteorology which we had not discussed together. His interest in every scientific problem with which the expedition was concerned was intense and I do not think that I have ever met a man who had the true scientific spirit so utterly unalloyed. To most of us who have given our lives to science our investigations are frequently tinged with an unscientific desire to increase our scientific reputations, but with him, it was the added knowledge alone which gave pleasure. He was constantly looking forward to the successful completion of the journey to the Pole, the exact value of which was perfectly clear to him, in order that he might spend his remaining time in the Antarctic in opening up new country and making new discoveries. And Professor H. Mohn has also passed away. Every meteorologist knows how much Polar meteorology owes to Professor Mohn for his discussion of the results obtained by Nansen in the Arctic, and only a few months before his death he discussed the meteorological data brought back from the Antarctic by Captain Amundsen. He gave to me liberally of his great experience on my visits to Christiania, and after my return from the Antarctic he placed the whole of Amundsen's data at my disposal, before even he himself had discussed them. He was keenly looking forward to the publication of the results of Captain Scott's expedition, and I feel confident that no one would have been able so well as he to separate the sound from the unsound in this work. Simla : G. C, S. March 1919. CONTENTS. Preface . L— INTRODUCTION The Geogi'aphical Features of the Eoss Sea Ai-ea The Stations History of the Stations Time ..... Weather .... II.— TEMPERATURE Instraments and Methods Geogi'aphical Distribution of Temperature Sea -level Temperatures Plateau Temperatures Vertical Distribution of Temperature Temperature and Winds Daily Variation of Temperature September to March. Period with Dii'ect Solai Radiation May to July. Period wth No Direct Solar Radiation Fourier Coefficients Cape Adare .... Non-periodic Changes of Temperature Mean Monthly Temperature Annual Variation of Temperature . Annual Temperature The Mean Temperature of the Antarctic III.--WIND Instruments and Methods Wind Velocity .... Mean Wind Velocity Annual Variation of Wind Velocity Annual Variation of High Winds and Calms Daily Variation of Wind Velocity Frequency of Winds of DifEerent Velocities Wind Direction Structure of the Wind . ['. Winds at Cape Adare Motion of the Upper Atmosphere Plateau Winds Wiads on the Plateau near the Magnetic Pole Winds on the Plateau to the West of Ross Island Winds on the South Polar Plateau Plateau Winds and Blizzards , Page. iii-viii 1 1 5 8 10 11 15 15 21 21 38 41 44 52 54 66 72 73 73 79 83 91 92 94 94 95 95 97 99 100 103 107 115 128 131 141 141 142 142 144 53501 CONTENTS. IV.— CLOUD AND PRECIPITATION Kind of Cloud .... Cloud Amount .... Mean Cloud Amount Cloud and A¥ind at Cape Evans Annual Variation of Cloud Daily Variation of Cloud Precipitation .... v.— PPvESSURE Instrmuents and Methods Mean Pressure and Annual Variation McMurdo Sound Framheim .... Cape Adarc .... Geographical Distribution of Pressure Daily Variation of Pressure Fourier Coefficients Tvventyfour-hourly Period Twelve-hourly Period Non-periodic Pressure Changes Pressure Waves Pressure Surges Pressure Correlation Correlation of Mean Monthly Pressure within the Antarctic Pressure Correlation between the Antarctic and Surrounding Places VI. -PRESSURE, WINDS AND WEATHER .... Pressure Waves and Pressure Distribution Pressure Waves and Winds ..... Theoretical Discussion of the Effects of Pressure Waves Comparison of Theory with Observations . Pressure, Winds and Weather at Cape Adare . Pressure, Winds and AVcather at the Gauss Station Conclusions ......... VII.— THE GENERAL AIR CIRCULATION OVER THE ANTARCTIC VIII.— UPPER AIR OBSERVATIONS MADE BY MEANS OF BALLOONS Winter Ascents ......... Summer Ascents ......... IX.— THE HEIGHT OF THE BARRIER AND THE SOUTH POLAR PLATEAU The Height of the Barrier Heights on the Beardmore Glacier Heights on the South Polar Plateau The Height of the Antarctic Continent X.— ATMOSPHERIC ELECTRICITY Potential Gradient Annual Variation Daily Variation Atmospheric Radio-activity Index Page. 147 147 M'J 149 151 152 155 159 1(53 163 166 166 171 173 175 176 17'.t 180 180 184 186 189 196 i;i(; 201 206 210 217 218 230 239 241 245 248 270 273 277 286 286 288 290 294 302 .302 312 313 319 321 ERRATA. Figure 20, page 54 and Fig. 21, page 56, for Range read Amplitude. Table 21, page 55, last column, ,> 1° >, H". Figure 53, page 158, „ Fraimheim „ Framdrift. Page 268, line 11, „ contract „ contact. Page 268, line 37, „ above „ about. CHAPTER I. INTRODUCTION. The Geographical Features oe the Koss 8ica Area. AS the greater part of this work deals with the ineteorology of the Ross Sea area of the Antarctic, it is essential to have a clear idea of the general geographical features of this region. This is particularly necessary as the several characters of the Sea, Barrier, and Plateau exert their own influences on the air overlying them, and their boundaries deter- mine to a large extent the direction of the air motion. In figure 1 (frontispiece) an attempt has been made to indicate diagrammatically the jihysical features of the Ross Sea area during .January. The region is divided naturally into three main divisions : («) the extensive plateau on the west, (6) the Ross Sea, and (c) the Barrier. The Plateau. — Geological evidence points to a great seismic catastrophe having at some time taken place in this region. The high land which once probably extended over the whole region appears to have split along a line running almost due south from near Cape Adare, and the eastern portion to have sunk below sea-level. The line of fracture is now shown by a narrow belt of broken land, which forms a fringe along the whole eastern edge of the plateau, and through which it is necessary to pass to reach the high land from the Barrier or the Ross Sea on the east. The fringe is very narrow in most places, being less than twenty miles in width, so that the descent from the plateau to sea-level is very rapid_ It is intersected at right angles by numerous valleys, through which the accumulated ice on the plateau flows down to sea-level in the form of gigantic glaciers. The plateau has been reached from the Ross Sea, side in three places. In 1903 Captain Scott made the ascent almost due west of Ross Island, and travelled on the summit to the position 78° S., 147° E. He found the height of the plateau in this region to be about 8,000 feet. In 1909 David reached the south magnetic pole (72° 25' S., 155° 16' E.), where the height of the plateau was determined to be 7,300 feet. The plateau near the South Pole has been vi.sited by Shackleton, Amundsen, and Scott and the height found to be just under 10,000 feet. The character of the plateau is the same at each of these positions. As soon as the mountains and valleys which form the fringe are left behind, the plateau becomes a level, unbroken plain, the snow surface being unrelieved by any feature to break the dull monoton}' of the c utlook. The Ross Sea. — In 1841 Ross discovered tlie open sea which now bears his name. After forcing his way through a thick belt of pack ice he emerged into open water near to Cape Adare. He then sailed south along the coast of tb.e platej.u which I'.e called South Victoria Land, until he sighted an active and an extinct volcano side by side, to which he gave the names of his ships, Erebus and Terror. As these mountains barred his way to the south he was forced to turn to the east, and after sailing a short distance he discovered the great ice-cliff which forms the southern limit of the open Ross Sea. Ross .sailed eastwards along this great cliff of ice hoping to find a way to the south, but as the ico-cliff made this (|uite impossible he gave to it the very suitable name of the ' Great Ice Barrier.' Scott, in 2 INTRODUCTION. the Biscnvenj, thirty-nine years later, was able to explore this ice barrier more thoroughly than had been done by any other explorer. He found that the ice-cliff extended from Cape Crozier almost due east to the new land which he discovered and named King Edward VII Land. The height of the Barrier varies between a few feet and 240 feet, but the general height is between 50 and 100 feet. Scott was unable to proceed to the east of King Edward VII Land owing to thick ice. With Scott's first expedition what may be called the permanent extent of the Ross Sea on its western and southern sides was fixed. The extent of the Ross Sea on the east is not known. Every explorer has had his progress in this direction stopped by thick, impassable sea ice. It is obvious that the edge of this ice varies from year to year. In 1842 Ross sailed up the east of the Ross Sea along a course almost due north-west from near King Edward VII Land, and he found ' compact hummock}^ ice ' to the east of him. In 1912 Evans in the Term Nova sailed somewhat to the east of Ross's track, and he had the greatest difficulty in forcing his way through the pack ice. There can be little doubt that the sea to the east of 165° W. longitude is permanently covered with ice, and that this meridian may be considered to limit the open water of the Ross Sea on the east. Looking now at the map we see that the Ross Sea on the north opens into the Antarctic Ocean in about latitude 70° S., and is bounded on the west by the coast of the high tableland, on the south by the Great Ice Barrier, and on the east by an undefined mass of ice or land which is probably at or very near sea-level, in any case the probabilities are against its being high land similar to that on the west. For our discussion of the meteorology of the region it is essential to have some knowledge of the varying ice conditions of the Ross Sea. Unfortunately the data available are very few, and they chiefly refer to the three summer months, December, January, and February. During January the Ross Sea is nearly free from ice between the coast of South Victoria Land and the heavy sea ice near longitude 165° W. There are, however, isolated patches of pack ice and floating bergs to be met with in all parts, and probably these occur more fre- quently in the eastern than in the western half. These floating masses of ice are indicated by lozenge-shaped areas in figure 1. North of the Ross Sea there is an almo.st constant stream of pack ice which has broken oft' from the coast of the Antarctic Continent and is slowly drifting from east to west. The Terra Nova, in 1910, entered this stream in latitude 65|-° S. and took nearly three weeks to force her way through, which she did not do until she was in latitude 71|-° S. In March the sea commences to freeze, but the wind is a great obstacle to the formation of a peimanent coat of ice. During March 1910 the sea in McMurdo Sound froze over as far as the eye could see several times, only to have every bit of ice removed by a subse- quent blizzard. In April the ice along the coast gradually became sufficiently thick not to be removed by the wind, and afterwards daily increased in thickness and so became more and moie able to withstand subsequent gales. It was not until the end of May that the Sound was completely frozen over. During the winter of 1912 the gales were so frequent that the Sound never became properly frozen over, and open water extended right up to Cape Evans for the greater part of the winter. The wind must play a similar part in the freezing of the Ross Sea, and as there is very much more wind over the western half of the sea than over the eastern half it is almost certain that the ice will form earlier and grow much thicker in the east than in the west. By the middle of the winter it is probable that the ice conditions over the Ross Sea are somewhat as follows. The greater part of the eastern half of the sea is firmly frozen THE GEOGRAPHICAL FEATURES OF THE ROSS SEA AREA. 3 over with ice which in the course of the winter attains a thickness of between ten and fifteen feet. Also along the western coast all the bays and inlets are frozen and the fixed ice probably extends several miles out from the coast. Between the fixed ice on the two sides there is an area which is sometimes frozen over and sometimes free from ice, even in the middle of the winter. That .such a region of thin variable ice does exist was made certain by the observations of the party which visited Cape. Crozier in July 1911. From Cape Crozier the Ross Sea can be seen, and the party made a record of the changes in the ice-covering of the sea during the time they spent there. When they arrived on the 15th July they found that the Ross Sea was completely frozen as far north as they could see, ' but much of the ice appeared to be young and thin, with little snow en it.' The next day they saw in the distance a cloud of frost smoke, which is a sure sign of open water. On the 17th they report ' two open leads of water, like broad irregular streets extending from the Cape Crozier cliffs away to the north- east and lying more or less parallel to one another.' These leads had disappeared the next day, but more open water was seen on the 19th. When this party descended to the sea ice they came to the conclusion that the ice had shortly before been blown out and only recently formed again, and they note ' some support is lent to this po.ssibility by the absence of all snow-drifts on to the sea ice from the ice foot.' A day or two after the party reached Cape Crozier a very severe gale occurred. For several days winds of hurricane force blew. When the wind dropped and the Ross Sea could again be seen it was found that the ice had all been blown north. Cherry-Garrard says ' in the gale the ice went right out from the Barrier as far as we could see. But we could see the ice on the liorizon from 900 feet up : just a line of blink, but it was there I feel sure. Otherwise open sea.' Everything therefore points to the centre of the Ross Sea being a region in which the ice is constantly forming and being blown away so that it can never become thick, and in which open water is constantly present either in leads or large open expanses. Very little is known about the winter ice conditions in the north of the Ross Sea, The only information is that obtained from observations made at Cape Adare. At Cape Adare the station was on the west of a lofty promontory and only the sea to the north- west was visible from the station. Young ice commenced to form during the first week of April, but here again it was constantly removed by high winds. It was not until the end of May that the ice was sufficiently firm to allow of journeys being made across it. After this large leads were opened by every gale and the presence of Antarctic petrols in July indicated that the open water was not very far away to the north. At the end of October the sea ice became too rotten for safe sledge travelling and in December the ice both cast and west of the Cape broke out with great rapidity. These observations point to the ice conditions at the north of the Ross Sea being very similar to those seen from Cape Crozier, and probably the ice here never becomes more than four or five feet thick, and is constantly, during the whole winter, opening up to produce long and wide leads of open water. These conditions probably exist to two or three hundred miles to the north of Cape Adare, whore the northern boundary of the frozen area is very hkely composed of pack sometimes free and sometimes fu'mly bound together by thin new- ice. The ice conditions in the Ross Sea are therefore in the main something as follows : — In the summer the open water is bounded on the west by the coast of South Victoria Land, on the south by the vertical face of the Great Ice Barrier, and on the east by a variable 0 4 INTRODUCTION. mass of sea ice the extent of wbicb is not known, but wbich in all probability extends in a north-nortb-westerly direction from Kin;^ Edward VII Land. A certain amount of detached ice floats on the open sea and this is most abundant in the cast and in a l)c!t across tli entrance to the sea in approximately latitude G7° S. In March new ice commences to form and by the end of the month the ice is permanently fixed in sheltered bays and during April and May this fiinge of ice extends outwards from the coast. During the winter the cast of the sea and a belt along the coast of South Victoria Land are permanently frozen over, between these two regions of fixed ice there is an area extending from the Barrier north- wards to the open ocean some hundred miles cr so north of Cape Adare in which the sea is sometimes frozen over and sometimes cleared of ice by the wind and in which the ice never attains any great thickness. The Barrier. — It has been stated that Ross gave the name ' Bairier ' to the ice-cliif which ban-ed his way to the south. Ross was not able to mount to the top of the cliff, but when subsequent explorers did so they found it to be the seaward edge of a great almost level plain of ice and snow. The name originally given by Ross to the cliff has been extended to this great snow plain, and now the original significance has been lost and the word ' Barrier ' is used almost exclusively as a place name for the level low-lying area of snow and ice which is terminated on the north by the ice-cliff to which Ross gave the name originally. On the west the Barrier is bounded by a continuation f)f th(^ escarpment which further north has formed the western coast of the Ross Sea. Starting near Ross Island and pro- ceeding due south along the level surface of the Barrier, one has on the right the lofty mountains of the escarpment between which large glaciers flow down from the surface of the plateau 8,000 to 10,000 feet above the level of the Banier. The high land continues on the right until latitude 82° S. is reached, where the escarpment turns slowly to the east and cuts across the direct way to the Pole. From 82° S., 160° E., the escarpment of the plateau runs almost south-east to near 85° S., 160° W., where according to Amundsen's observations it turns sharply to the right and continues in a more southerly direction, passing within about 120 geographical miles of the Pole. The eastern boundary of the Barrier is unknown. There are indications of land in a line to the south of King Edward VII Land, but it is almost certain that if land does exist here it is low land or a series of low islands. For meteorological purposes we may consider that the whole of the area beyond the escarpment of the plateau is low, and if it is not actual Barrier it is snow-covered low land which is similar in most respects. Neglecting for the present the character of the region to the east of longitude 160° W., we may say that the area bounded on the north by the great ice-cliff joining Ross Island and King Edward VII Land, on the west and south by the escarpment to the plateau and on the east by longitude 160° W., is occupied by the Barrier. The area of this region is ap- proximately 200,000 square miles, i.e., nearly the same as France. The distance from Ross Island to the Beardmore (jlacier is almost exactly the same as from Paris to Marseilles. The whole of this huge area is practically a level plain. Barometric determinations indicate that its surface is on the average 170 feet above sea-level and this is only slightly higher than the ice-clift"s on its northern edge. The question of the origin and the formation of the Barrier is discussed elsewhere ; it is sufficient to give here the conclusions arrived at. There is little doubt that for the greater part, if not for the whole, of its extent the Barrier is a floating sheet of ice. If this is so and assuming that there is seven times as much ice below the level of the sea as floats above, the average thickness of the sheet of ice is something like 1,400 feet. The Barrier surface is snow, often soft, but sometimes hardened by the wind, THE STATIONS. 5 so that, the feet .sink in only a fraction of an inch. The Barrier surface is nowhere com- posed of liard ice, and it is probably only at considerable depths that the snow loses its porous texture and becomes solid ice. This fact will be slioxvn to have an impoi-tant bearing on the temperatures measured on the Barrier. The Stations. Cape Evans.— The winter quarters of Captain ^^cott's second expedition were situated on a small cape jutting out from Ross Island into McMurdo Sound (see figure 2). Rost Island 170° Fig. 2. Map of the .Sunoundings of Cape Evans. is almost entirely occupied by two volcanoes, Mount Erebus and Mount Terror, the formw of which is still active. It is separated from the mainland by McMurdo Sound. The western coast of the island runs almost due south for fifty miles from Cape Bird to Hut Point and IS roughly parallel with the opposite coast of the Sound which is forty miles away. 6 INTRODUCTION. McMurdo Sound is therefore approximately rectangular in shape, the long and the short sides being fifty and forty miles respectively ; it is open to the Ross Sea on the north and closed by the Barrier to the south. The Sound has open water during .January and February, but freezes over during the winter. The winter conditions of the ice are however very varied from year to year. During some winters the whole Sound is frozen over from April to the end of December, while during other winters frecjuent gales remove the ice as soon as it is formed and only the southern half becomes firmly frozen over. The summit of Mount Erebus is about fifteen miles to the north-ea.st of Cape Evans, and is over 13,000 feet high. The side of the mountain in this region rises rapidly from the coast and is covered with one huge glacier. Cape Evans is a small triangular piece of land at the foot of the mountain. One side of the triangle is against the slopes of Erebus while the other two are washed by the waters of the Sound. During the summer a large propoi'tion of the black volcanic debris of the Cape is exposed, but in the winter it is all covered by snow except in a few places where the wind sweeps the surface clear. The slopes of Erebus to the south of Cape Evans are cpiite impas.sable, hence when the sea is not frozen over Cape Evans is entirely cut ofl' from the south. When the Sound is frozen over the sea ice forms a splendid surface over which the coast of Victoria Land can be reached or journeys made to the south. It was mentioned above that McMurdo Sound is closed to the south by the Great Ice Barrier. As a comparison of the conditions at Cape Evans Avith those on the Barrier will form an important part of this work, it is essential that the geographical features to the south of Cape Evans should be thoroughly understood. We have already become familiar with the great surfaee cf the Barrier extending southwards from the icc-clifEs which form the southern limit of the Ross Sea. The edge which the Barrier presents to McMurdo Sound is much less imposing than that to the Eoss Sea, for the latter is a cliff of ice approxi- mately 100 feet high, whi'e the end of McMurdo Sound is closed by a clifi only a few feet high and this is in many places, especially at the end of the winter, almost obliterated by snow-drifts which foim inclined planes from the Barrier surface to the sea ice. The edge of the Barrier ice is shown in figure 2. It will be seen that it extends from near Hut Point in a zigzag line to the coast of South Victoria Land just to the north of the Koettlitz Glacier. The position of Hut Point is also of considerable importance as it was here that the Discovery Expedition had its headquarters in 1902-04. The hut itself was situated on a .small cape around which open water extends during most summers. It is possible however to reach the Barrier from Hut Point even when the sea is not frozen, for a way leads over land from the hut to the po.sition where the Barrier permanently joins the land. Details of the exposure of the instruments and the lie of the land at both Cape Evans and Hut Point will be considered as occasion arises in the discussion of the observations. Cape Evans to the Pole. — A great number of meteorological observations were made on the sledging journeys undertaken in connection with the attempt to reach the Pole ; it is therefore desirable that the route taken on those journeys should be described here. The first pari of the journey was from Cape Evans to Hut Point. This could only be under- taken when McMurdo Sound was frozen over. The hut erected at Hut Point by the Dis- covery Expedition w^as used as a base for all journeys further south and here it was possible for various parties to wait while the Sound froze over or until the time arrived for con- tinuing their journey. From Hut Point the Barrier was reached either over the sea ice to Safety Camp, or through ' the Gap ' ; then, in order to miss crevasse.' formed in the Barrier ice by the Bluff, the course was laid east-south-east to Corner Camp in 77° 54' S., 167° 17' E. From Corner Camp the journey to the Beardmore Glacier was made along THE STATIONS. 7 or near to the 170° E. meridian. Ninety-.six geographical miles south of Corner Camp, at 79° 30' S., 169° 22' E., was the main depot on the Barrier called One Ton Camp. There was considerable passing to and fro between Cape Evans and One Ton Camp, but beyond this point journeys were made only during the summer of 1911-12 in connection with the main Polar journey. Framheim. — As far as one can gather from the published account of Captain Amundsen's Expedition, the situation at Framheim was ideal from a meteorological point of view. The house itself was erected on the surface of the Barrier about one and a half miles from the sea ice in Whale Bay, and there were no hills anywhere near which could affect the force or direction of the wind. The general conditions at Framheim are entirely differoit from tho.se at either of the two other stations in the Ro.ss Sea area. To the south the Barrier extends as an unbroken level plain and to the north exists the Ross Sea, the ice conditions of which have already been considered. The ice conditions in the immediate neighbourhood of Framheim are of importance, but Amundsen has given no information of the state of the sea ice at different times of the year. Looking at the map (frontispiece), however, we see that the permanent ice which bounds the Ro.'-'s Sea on the east leaves a bay on the south of which Framheim lies. It is very probable that this bay freezes over very early in the winter and that the ice attains a considerable thickness. Cape Adare. — Cape Adare is the northern point of a promontory twenty miles long which forms a kind of horn to the extreme north-easterly point of South Victoria Land. The promontory CAPE ADARE Hut' SON \Olfa:r { V irW^ SKETCH MAP ROBERTSON BAY AND CAPE ADARE ^^'>\tS-^/ Fjo. 3. Map of the Siiiroiinclings of Cape Adare. 8 INTRODUCTION. is a Ions narrow liigh ridge running approximatply nortli and soutli (see figure 3). The camp was formed on a beach on the west side of the promontory and the land rose very rapidly behind the station to the ridge some 4,000 feet above sea-level. Thus the station was shut out from all winds between south-east and north-west and its open hoiizon only extended over Robertson Bay to the west and over the Antarctic Ocean to the north-west. The situation was far from ideal from a meteorological point of view. History of the Stations. Cafe Evans. — The Terra Nova with the whole expedition on board arrived in McMurdo Sound on .January 4, 1911. and the discharge of stores commenced at Cape Evans the same dav. As soon as the general work of unloading allowed, the Stevenson screen for the thermometers was erected and the first set of meteorological observations was taken on shore on .January 13. During the next few weeks the various meteorological and magnetic instruments were set up and put into working order. With few exceptions they were all in use on the 1st March, 1911, from which date the full meteorological programme may be considered to commence. While the Terra Nora was unloading at Cajje Evans the meteoro- logical log was kept more or less completely on the shij) and these observations have been found of much use in supplementing the shore observations during the first few weeks of preparation. Until the end of February 1912 the writer remained in charge of the meteoro- logical work at Cape Evans and the record of all the chief meteorological factors is neariy unbroken. After the first year's work the ship returned to Cape Evans, to take off the expedition, but owing to the non-return of Captain Scott from the South Pole, it was found necessary to leave a party of men for another year. I had hoped to stay with this party, but letters brought bv the ship necessitated my early return to India. I therefoie left Cape Evans on the 4th March, 1912, and C. 8. Wright, assisted by T. Gran, took over the meteorological work. The lar"e meteorological and magnetic programme which had needed all my time in addition to considerable help from other members of the expedition was no longer possible. Also as soon as the second winter was over all available strength was devoted to the search for the Polar Paitv. In consequence the meteorological work was less complete during the second vear, and there is an unbroken record for two years of only a few of the most important meteorological elements. SledgitHj Journeys from Cajoe Evans. — Soon after the landing at Cape Evans a party was sent to lay a depot as far south on the Barrier as pos.sible. This party left Cape Evans on January 24, 1911, and reached 79° 30' S. on February 17th where they left one ton of stores, giving the depot the name One Ton Camp. The party returned to Hut Point and found the sea ice, which formed the road to Cape Evans, all gone out so they were compelled to remain at Hut Point until the new ice had formed to a sufficient thickness to allow of the journey back being completed. A valuable set of observations was made on this journey. During the winter Wilson, Cherry-Garrard and Bowers made a wonderful sledge journey to Cape Crozier. They encountered such weather as no other party of men has ever ex- perienced when sledging. They left Cape Evans on June 27, 1911, and returned on August 1, 1911, thus being away during 36 days. Bowers kept a most complete meteorological journal, which is of the greatest value as showing for the first time the conditions on the Barrier at midwinter. On this journey the record low temperature for the Antarctic, - 70°!''.. was observed. HISTORY OF THE STATIONS. In September 1911 Evans took a party to Corner Camp on the Barrier. The journey was short but it gave important information about the early spring conditions on the Barrier. In the same month Captain Scott made a sledge journey across McMurdo Sound to the Western Mountains. The most important result from a meteorological point of view was a series of observations the writer was able to make on fog bows and iridescent clouds, which led to the explanation of these phenomena published in the Quar. Journal Roy. Met. Soo., Vol. 38, page 291, 1912. The various parties which traversed the Barrier during the summer of 1911-12, in connection with the journey to the Pole, kept meteorological journals which have been of the greatest use in the investigation of the meteorology of the Barrier. The journal of the Polar Party was kept by Bowers, and during the last few weeks afiords remarkable evidence of his devotion to the .scientific work of the expedition. The records of three observations a day are practically complete until March 4 ; they then became irregular, but one set of observations was made each day until March 11 when they ceased. The final camp was made on March 21 so that observations were carried on to within ten days of the setting in of the blizzard which proved fatal. It is convenient to tabulate here the chief meteorological journals kept by the various parties on the Barrier. Table 1. i Name of Party. Ob.server. Olio Toil Dapot Partj' Bowers Dates, January 26. 1911 March "lO, 1911 Position. Hut Point to One Ton Depot and back. Corner Camp Party Evans September 9, 191 1 September 15, 1911 Hut Point to Corner Camp and back. Main Polar Party . Bowers November 3, 1911 March 12, 1912 Cape Evans to Pole and back to 80° S. Motor Party oui . Evans . October 27, 1911 November 21, 1911 Motor Party retvirn Day . November 25, 1911 December 20, 1911 Hut Point to 80- 32' S., 1G9° 23' E. 81° 10' S., 169° 30' E. to Hut Point. Dog Party 1st Retu'n Party Meares Wright November 5, 1911 January 4, 1912 Hut Point to Mount Hope and back. December 22, 1911 January 26, 1912 Upper Glacier Depot to Hut Point. 2nd Return Party . Evans . ' I Depot Party . Day January 4, 1912 February 9, 1912 Plateau 87° 19' S , 160° 40' E. to Barrier 79" 33' S , 169° 22' E. December 26. 1911 January 21, 1012 Hut Point to One Ton Camp and back. 1st Relief Party Cherry-Garrard February 26, 1912 Maich 16, 1912 . Hut Point to One Ton Camp and back. 2nd Relief Parov Atkinson March 27, 1912 April 1, 1912 Hut Point to seven miles beyond Corner Camp and back. 10 INTRODUCTION. During 1911 and 1912 Taylor led two parties to investigate the geology of the Western Mountains. A full meteorological journal was kept on each journey, and these have been found particularly useful in plotting weather maps. During the second 3'ear meteorological observations were made on several journeys, but as the simultaneous observations at Cape Evans were far from complete, these records are not so valuable as those made during the first year. Cajpe Adare. — The party landed on February 18, 1911, and the meteorological record was commenced by R. E. Priestly on the 27th. It was kept with great regularity until the party left Cape Adare on 3rd January, 1912. During one or two short periods when all the men were away on sledge journeys the observations are wanting, but the observations made by the sledge parties have been of great help in filling in the record. Framheim. — Meteorological observations were made at Framheim by Captain Amundsen's Expedition three times a day from 1st April, 1911, until 29th .January, 1912. In addition a valuable set of observations was made on the Barrier between 8th September and 26th January during Captain Amundsen's journey to and from the Polo. The results have been published by the Fridtjof Nansen's Fund.* Tlie Terra Nova. — Three voyages were made between New Zealand and the Antarctic by the Terra Nova and throughout a full meteorological log was kept. The voyage? were as follows : — 1 si Voyage. Left New Zealand (Port Chalmers) Arrived McMurdo Sound . Left McMurdo Sound Arrived off King Edward VII Laud Visited Bay of Whales Returned to McMurdo Sound . Left McMurdo Sound Arrived Cape Adare Left Cape Adare Discovered Gates Land . Arrived New Zealand (Stewart Island) November 30, 1910. January 3, 1911. February March 2nd Voyage. Left New Zealand (Port Lyttelton) December Arrived Cape Adare January Left Cape Adare .......... Arrived Cape Evans February Left Cape Evans *........ . March Arrived New Zealand (Akaroa) April 3rd Voyage. Left New Zealand (Port Lyttolton) December Arrived Cape Evans ......... January Left Cape Evans Arrived New Zealand (Oaraaru) Febniarv 28, 1911. 9 1911. 4, 1911. 8, 1911. 9, 1911. 18, 1911. 20, 1911. 22, 1911. 27, 1911. 15, 1911. 4, 1912. 5, 1912. 5, 1912. 4, 1912. 1, 1912. 14, 1912. 18, 1913. 22, 1913. 10. 1913. Time. Cape Evans. — Owing to the great convenience of using Greenwich time in the scientific work of the expedition, it was decided to keep our watches and clocks set to the time of the 180th meridian instead of to local time. In the following whenever times at Cape Evans or on journeys undertaken from Cape Evans are given without qualification they will be * Ro.ild Amundsen's Antarctic Expedition, Scientific Results, Meteorology by H. Mohn. TIME— WEATHER. 11 times according to this practice. Local time at Cape Evans is fifty-four minutes behind that of the 180th meridian, hence for moat meteorological purposes, when local time is required, it is sufficient to deduct one hour from the times given. As four minutes is imperceptible on most of the meteorological self-recording instruments used, the clocks were set to station time and hourly values tabulated. Conversion to local time was then obtained by deducting one from the denominator of the hours at the head of each column. Cafe Adare and FramJieim. — ^Local time only was used at these stations. Weather. It will probably be useful to conclude this introductory chapter with a short general description of the weather conditions which we are about to discuss in detail. Weather at Cape Evans. — McMurdo Sound is situated between two masses of high land : Ross Island on the east and the Western Mountains on the west. The air motion in the Sound is therefore limited, and all the winds can be classed according to whether the air motion is through the Sound from the north or through the Sound from the south. The former includes at Cape Evans all winds between E.N.E. and W., and the latter all winds between W.S.W. and E., and these two classes will throughout be specified as northerly winds and southerly winds respectively. There is a close relationship between the wind direc- tion and weather, and it is convenient to consider the weather at Cape Evans under three headings : — (a) The weather associated with northerly winds. (b) The weather associated with calms and light winds. (c) The weather associated with southerly winds. The period during which the wind blows with a greater velocity than 10 miles an houi from the north and the south is 8 per cent, and 49 per cent, respectively ; while during the remaining 43 per cent, of the time the air is calm or the wind velocity less than 10 miles an hour. Weather associated with Northerly Winds. — The chief characteristics of the type of weather associated with northerly winds are relatively high temperatures and an almost total absence of cloud. The air is nearly always clear during a northerly wind ; but the mnd sometimes raises a little surface drift from the ground if there happens to be newly-fallen snow lying. Northerly winds interfere very little with the outdoor work of the expedition and therefore they receive very scant notice in the popular descriptions of Antarctic life, but we shall see that they are very important from the meteorological point of view. Weather associated with Calms and Light Winds.— One of the interesting features of the air motion in McMurdo Sound and on the Barrier is the tendency of the wind to bo either entirely absent or to be blowing a gale. The records for twenty months show that the wind was blowing at 30 miles an hour or over during 30 per cent, of the total period, while during 22 per cen'. the velocity was 4 miles an hour or less. As far as I have been able to find there is no other place in the world at which such a large proportion of high winds is associated with such a large proportion of calms. In the summer the tem- perature during calm weather is little different from the average of the month, but during the winter it is very much lower. Except for its low temperature during the winter, calm weather has no outstanding peculiarities : the sky may be completely overcast or completely clear, and also occasionally snow falls. Weather associated with Southerly Winds. — During practically half the whole time the wind blows from the south with a greater velocity than 10 miles an hour, and of this period a 12 INTRODUCTION. half is occupied by winds of a greater velocity than 30 miles an hour. The majority of winds from the &o\ith are typical blizzards. As in every section of this work reference will be made to the blizzard, it is desirable that a short account of this phenomenon should be given here. The origin of the word blizzard is unknown, but it is now used in meteorological literature for any high wind accompanied by unui^ual cold and snowfall. In the Antarctic the members of our expedition used the term rather loosely and any high -wind was called a blizzard almost independently of its direction and of the weather conditions accompanjing it. There is, however, no doubt about the description of a typical blizzard which would be given by anyone who has spent a year in McMurdo Sound. In the most typical case light cirrus cloud first spreads over the sky and gradually gets thicker and lower, until finally the whole sky is covered with a uniform dark gray layer of cloud. For a longer or shorter period there may be a complete calm and then the wind lises. We shall see later that over the Barrier the blizzard wind blows from the south parallel to the Western Mountains, but Ross Island stands up as a great obstruction in its path and the southerly wind is deflected into two streams, one of which passes the Island on the east, and the other passes into and blows through McMurdo Sound on the west. The actual direction of the air motion in a blizzarxl at different parts of the Sound depends on the surrounding land masses. At Cape Evans the blizzards are deflected by the shoulder of Mount Erebus into easterly to south-easterly winds. The arrival of the wind is frequently very sudden and within a few mijiutes, a calm may be replaced by a wind of 30 or 40 miles an hour. Figure 46, page 128, shows three typical examples of the sudden setting in of a blizzard. This however is by no means a general rule, and on many occasions a gentle southerly wind will got stronger and stronger until it develops into full blizzard force. In a true blizzard the wind is accompanied by clouds of driven snow. The snow is in the form of exceedingly fine grains which penetrate through the smallest chink or hole in a house or tent. The whole air appears to be full of drift, so that it is impossible to see any great distance, and when it is at its worst even a tent cannot be seen for more than a few yards. Not only does the drift make it difficult to see, but anyone exposed to it seems to become bewildered and to lose all power of thinking clearly. For these reasons it is sheer folly to attempt to travel in a blizzard even when the tempsrature is relatively high and the wind at one's back. There can be no doubt that the large amount of drift in a blizzard is due to the high wind sweeping along with it all the snow which has been precipitated, for none can settle out of the air except in a few sheltered places where large snow-drifts accumulate. Thus with only a moderate rate of precipitation the lower atmosphere in time contains a great deal of snow. Opinions differed as to the relative importance of old snow blown up from the surface and new snow which is falling. M)' own experience led to the conclusion that thne can never be really bad drift unless new snow is actually falling. I have certainly seen a high wind convert newly-fallen snow from the surface into drift, but this drift was never so thick and bewildering as the drift which accompanies heavily overcast skies from which new snow is falling. If no new snow has fallen for several days the highest wind produces only a low surface drift, for the surface of the snow becomes compact in a comparatively short time. For a really bad blizzard both high wind velocity and actual precipitation are necessary. The wind during blizzards is often squally and generally gusty, and on some occasions the gustiness is very marked, see figure 38A. There is a tendency for a blizzard to become more gusty towards its end ; but this is by no means a general rule. WEATHER. 13 From beginning to end of a blizzard the wind is extremely constant in direction. Neitlier at Cape Evans nor on the Barrier is there any evidence of a regular change of wind direc- tion during a blizzard. When the blizzard comes to an end, the air motion stops entirely, or there may be a sudden change in wind direction of 180°, and the wind, sometimes of a high velocity, then blows from the north. For a typical blizzard snowfall and drift are necessary, but from a meteorological point of view there is no essential difference between a southerly wind with little or no cloud and a blizzard with drift so thick that om, cannot see more than a couple of yards ahead, for evsry grade between these two extremes occurs. In all our statistical work, therefore, and generally in our descriptions, a blizzard will really mean a high southerly wind, which may or may not be accompanied by snow and drift. The duration of a blizzard can be anything from an hour or two to several days. There are many examples of a southerly wind rising in a few minutes from a calm to 30 or 40 miles an hour and then dying away again within the hour; on the other hand the longest blizzard occurred in June 1912 when from 20 hours on the 7th until 11 hours on the 14th, i.e., for 6 days and 14 hours, the anemometer recorded more than 20 miles in every hour and the mean velocity during the period was 48 miles an hour. The temperature during southerly winds in the winter is always much higher than during calm weather. This has given rise to the idea that the temperature during blizzards is abnormally high, and often in Antarctic literature one's attention is called to ' the un- expected warmth of southerly, i.e., Polar, winds.' We shall show that the southerly wind is not a warm wind, but as would be expected it is always colder than a corresponding wind from the north. The sequence of weather changes is remarkable. It was a popular opinion amongst the members of the expedition that a blizzard was usually preceded by a northerly wind. As a matter of fact the records do not support this opinion. It frequently happens that a period of northerly wind precedes a blizzard, but it as frequently happens that a blizzard commences without a previous nort.herly wind. On the other hand many blizzards end by the wind changing completely round and blowing from the north. We shall show in the section on Pressure, Wind and Weather that the theoretical sequence of weather is a cycle in which northerly and southerly winds alternate, with the period of the southerly v\inds longer than that of the northerly winds. In reality this is what happens, but the cycle is constantly broken up by periods of calms. Blizzards are therefore associated with northerly winds, but they as frequently occur after as before. Weather at Franiheim.— The weather at Framheim is totally diii'erent from that at Cape Evans. The tiiie Antarctic blizzard does not occur at Framheim and during Amundsen's stay only 2 per cent, of the wind observations recorded velocities over 30 niiles an hour — this compares with 30 per cent, at Cape Evans. The frequency of calms at Framheim is nearly twice as great as at Cape Evans and during 42 per cent, of the whole time the wind is 4 miles an hour or less. The direction of the prevailing wind at Framheim is due east, while there is practically no air motion from W.N.W. to N.N.E. Although Framheim is only 60 miles further south than Cape Evans its temperature is very much lowei. The mean temperature at Framheim during the period April to September 1911 was 19°F. lower than at Cape Evans, and the observations during 1911 and 1912 indicate that in all probability the mean annual temperature at Framheim is 15°F. lower than at Cape Evans. Weather at Cape Adare.—Q&^e. Adare is situated on the coast of the Antarctic Ocean which is the most boisterous ocean in the worid. In a similar situation about 700 miles 14 INTRODUCTION. further west Mawson found the highest average wind velocity of any place in the world, the average velocity for a whole year being 50 miles an hour. It is therefore amazing to find that the mean wind velocity at Cape Adare is less than 10 miles an hour, and that for 72 per cent, of the whole time the wind velocity is less than 5 miles an horn-. Thus the frequency of calms is nearly twice as great as at Framheim and nearly four times as great ~as at Cape Evans, and at each of these places the frequency of calms has been considered remarkable. (At Yarmouth in England the frequency of calms — 0 to 4 miles an hour — is only 5 per cent.) On the other hand Cape Adare is visited by violent hurricanes : during 10 months 14 storms occurred in each of which the wind was recorded as 11 or 12 on the Beaufort Scale. These high winds, as a rule, lasted cnly a few hours ; but ^^hile they blew the members of the Cape Adare Party were in deadly fear that their hut and all their belongings would be blown into the sea. Although the wind direction in the hurricanes is from the south-east or south they are in no way connected with the blizzards of Cape Evans. Thus the outstanding features of the weather at Cape Adare are the remarkable fre- quency of calms and the abnormally high winds. CHAPTER II. TEMPERATURE. Instruments and Methods. Instruments and their Exposure.-The exposure of the thermometer screen at Cape Evan« could not have been better. Immediately behind the hut the land rose rapidly to a smaU hill 64 feet above sea-level. The top of the hill, called Windvane Hill, was so exposed to the wind that it was always quite free from snow even after the severest blizzards Here he screen was mounted on four stout posts well bound together and secured firmly into the kenyterock The thermometers were about five feet above the ground, but the latter sloped away on all sides so that the temperatures measured were practically free-air temperatures In the screen were (a) a mercury dry bulb thermometer, (6) a mercury maximum thermo- meter, (c) a spirit minimum thermometer; and (d) a bimetallic thermograph. Short^l rr°^"'^iJ" '^" ''"" "'' "" "'^'"'''^ commercial instrument made by Messrs. Short and Mason. The pen was actuated by means of a bimetalhc coil and marked in the :t::ZrVjT \rr' '''''■ ''^ -^-^^ ^^^-^'^^ ^'' t^^m^o^.,^. in polar climat were met. When the temperature fell below about -30°F. the clock stopped in spite of all ^01 having been cleaned out of the bearings; also the whole instrument became coked u! wi h driven snow dunng blizzards. These were not faults of the thermograph, which was a plendid instrument to work with. The bimetallic control was very strong,'the s ^s tivenes of the instrument was very uniform over the whole scale and remained constant thr ug t The heat capacity was small so that rapid changes of temperature were registered. It was realised in England that we should have difficulty with the thermograph in the screen A thermograph was therefore designed which would register inside the" hut "vhere the clock and other moving parts would not be subject to the severe outside conditions bra s tube through which air could be drawn by an electric fan. This was fixed on to hi north-east side of the hut, and a very narrow lead tube, passing through the wa 1 o ^hin ;r;2 • T' Z ''' ''''-' ''''' ''''''' ^^^^ ^ ^P-^ -™- -"Gained mthm a movable iron piston floating on mercury inside the hut. To the piston the pen bu b was lead every four hours, so that the record did little more than register the change cS^Tut ir\r : T:r ^* '--' ^^^ ^'^^^-^ '-^ -- --^ *° ^^-- ^^^ past t,: i col, but It was found that it made practically no difierence whether the fan was in Crandiiorsf^;'^ '"''r'^'' ^'^^ ^^" ^'- - '-^-^ -^^»- T^« i-trument w Sluggish and not so well exposed as the thermograph in the screen 16 TEMPERATURE. Method of taking Temperature Observations. — The screen was visited each morning between 8 and 8-30 a.m. As soon as the door had been opened the pen of the thermograph was depressed, the time noted, and the dry bulb thermometer read. Thus the time and tempera- ture recorded by the therraograi^h were controlled. The maximum and minimum thermo- meters were then read and reset, immediately afterwards all three thermometers were read again and theii readings recorded. MetJiod of Reducing the Thermograph Records. — The mean temperature at Cape Evans for nineteen months was obtained from hourly values of the thermograph records, hence it is important that these records should be correctly reduced. The method employed was different for the two instruments. (a) The Screen Thermograph. — The papers supplied by the makers on which temperature and time lines are engraved were used. If these lines had been correctly drawn and the rate of the clock and the setting of the pen adjusted once for all it would only have been necessary to read ofE the temperature at the correct time intervals. It is well known, however, that these instruments can never be set with the required accuracy, hence the necessity for some method of reduction which will allow for the errors of time and pen setting. It has been stated above that every day the pen was depre.ssed at a recorded time. Hence on every sheet there were seven time checks. Taking these checks into account it was easv to mark the trace by a series of dots, each one of which corresponded to the exact hour. At the instant the time mark was made each morning the temperature of the dry bulb thermometer was read and recorded. A comparison of the reading of the trace and these recorded temperatures ga^e the error of the thermograph. The mean error for the week was calculated from the seven observations and applied to the trace throughout the week. It will be seen that by this method the thermograph record was reduced to the scale of the dry bulb thermometer. (b) The Hut Thermograph. — This was a daily instrument, a new paper being put on to the drum each morning at 8 o'clock. These papers were blank. The pen of the thermograph was depressed every hour by an electric current controlled by the standard clock so that accurate time was shown on the record itself. The thermometer fixed near the bulb of the thermograph was read every four hours and these readings were entered on to the trace at the corresponding times. The deflection of the thermograph for a given change of tem- perature had been previously determined and a glass scale constructed. This scale was used for marking the po.sition of O^F. on the sheet at each of the eight points determined by the eye-readings and through these marks a line was drawn from which the trace at each hour was measured with the glass scale. Thus the recorded temperatures were reduced to the scale of the thermometer attached to the bulb of the thermograph. The Accuracy of the Temperatures Observed. — It has just been pointed out that the two thermograph records depended on the scales of the two thermometers. Both these thermo- meter's, when tested at Kew before the expedition sailed, had no errors over the range of temperatures to which they were exposed in the Antarctic. Unfortunately the}' were left in the Antarctic so they could not be compared on the return. Still they were compared in the Antarctic with other thermometers and there were no indications of any change having taken place. Thus the hourly values of temperature may be considered to be quite correct. The accuracy of the maximum and minimum thermometers must be considered in two periods— (a) up to the end of February, 1912, and (b) after this date. During the former of these two periods the maximum and minimum thermometers were read immediately after setting each morning and compated with the simultaneous reading of INSTRUMENTS AND METHODS. 17 tho dry bulb thermometer as already described. These daily comparisons gave a good check on the behaviour of the thermometers and the results are shown in the following table, in which the mean values found by the daily comparisons from February, 1911, to February, 1912, are collected, the numbers in brackets giving the number of comparisons involved. Table 2. Thermometer Coinjxiritions. Difference between Mercury Thermometer No. S-tT-t and Mercury Maximum Tliermo- meter No. :U3l). Difference between Mercury TKermometer No. 8474: ami Spirit Minimum Thermometer No. 3450. Temperatiirefj .\t Kew In Antarctic At Kew In .A-ntarctic > +!.'>' F. . 00 00 (89) 0 0 +0-3 (89) + 15° to -15' . +0-2 +0'1 (151) 00 +0-4 (155) < -i.r F +0-2 (pr.) -01 +0-4 (73) The mercury dry bulb thermometer No. 8474 was in use throughout the expedition and was the standard for the reduction of the screen thermograph. The above comparison shows that it and the ma.ximum tliermometer No. 3136 had retained their Kew corrections. On the other hand the spirit minimum thermometer had changed its correction by -f- 03. In practice the method of correcting the maximum and minimum thermometers was to consider the dry bull) thermometer No. 8174 to be correct, then at the end of each month the mean corrections of the other two thermometers were determined from the daily com- parisons during the month and applied to the individual readings. After February, 1911, the daily comparisons were not made and there were several changes in the maximum and minimum thermometers used. The dates of the changes were not al- ways recorded, and there are other small uncertainties. After a careful f-tudy of tho records the appropriate corrections to apply throughout have been determined, I believe, with a fair degree of certainty, but as some doubt still remains the maximum and mininuim temperatures during the second year may be in error by ± -5° F. Method of Measuring Temperatures on Sledging Journeys. — Experience on previous expedi- tions had taught that there are considerable practical difficulties when the air temperature has to be measured on sledge journeys. As it is quite impossible to transport a suitable screen some form of artificially ventilated thermometer becomes a necessity. An Assman'a ventilated psychromoter would appear at first sight to meet the case, but these instruments were tried on Captain Scott's first expedition and found to be quite unsatisfactory. The actual difficulties will be found described in 'The Voyage of the Discovery,' Vol. I, page 276; it is sufficient here to mention the impossibility of getting clockwork mechanism to work in low temperatures, and the appreciable weight of any but the smallest pattern, which are too small for accurate reading under the difficult conditions met with in the Antarctic. Some form of the sling thermometer was the only alternative. The simple method of swinging a thermometer on the end of a string becomes practically impo.ssible when one has to work in thick fur mitts, and in bulky clothes nearly as stiff as thin sheet metal owing to the frost. After considerable thought and experiment a form of sling thermometer based on a very old principle was designed, which proved to be admirable in every way when put to the severe test of continued use on sledging journeys. 18 TEMPERATURE. The instrument is shown in figure 4. The thermometer T is attached rigidly to an aluminium back 25 cm. wide and 225 cm. long. The latter is pivoted at F to a solid is^p^jaiiiRtiimgRiik^^g - -K Fiu. 4. Sling Tlierniouietei-. aluminium handle 22-5 cm. long in such a way that by gripi^ing the handle near the end the thermometer can be made to revolve rapidly around F by a slight circular motion of the wrist. The back K is hinged by means of a piece of raw hide at M so that after use the back folds over the handle, in which a groove E, D has beeji cut to take the thermometer. When closed and the back secured by the shding catch A (made very large so that it can be opened and closed with the hand in mitts), the thermometer is enclosed in a strong metal box and it is practically impossible to break no matter how roughly the instrument is handled. It was the oiiginal intention to have the handles made of wood, but as it was found that these would be very expensive and aluminium handles could be cast at a very small cost, all the instruments we took, with one exceiition, were made of metal. These were quite satisfactory in all except the lowest temperatures, but when the temperature was below - 40°F. it became painful to hold the instrument for the time necessary to swing it to obtain a correct reading, even when wearing the thickest mitts. The one instrument having a wooden handle was perfect and was used by Bowers throughout the Polar journey, until, when very near the end, it was broken, jHobably as a consequence of the weak state in which Bowers then was. It was usual during sledging to take meteorological observations each time camp was made and broken. Thus normally observations were taken three times a day : («) In the morning just before the march for the day commenced. (b) Near midday, during the lunch halt. (c) In the evening after the tent had been erected and while the evening meal was being prepared. INSTRUMENTS AND METHODS. 19 Nearly all the sledging thermometers (spirit) were provided with minimum indices, and after the sledge had been straightened for the night the open thermometer was carefully placed under the sledge in such a position that it was shielded from radiation. There is little doubt that a thermometer so placed gave minimum temperatures too low by a degree or tw'o, but it will be shown that the minimum temperatures are of great value and cannot be neglected in obtaining the mean temperatuie for the day. The daily range of the tem- perature on the Barrier is so great that the mean temperatures cannot be obtained rom simple temperature observations taken two or three times a day, for the time at which the observations are taken makes a very large difference in the temperatures recorded. The following example will illustrate this and will serve to show how the mean temperatures on the Barrier were obtained. During the main southern journey temperature observations were taken by Bowers and Meares. The former marched with the ponies and the latter travelled with the dogs. During the greater part of the journey, the ponies marched at night and rested during the day. In consequence the observations were generally taken by Bowers as follows : — («,) When the march .started at about 9 p.m. (h) At the mid-march halt at about 3 a.m. (c) When the march ended at about 10 a.m. Owing to the fact that the dogs did the same march as the ponies in a very short time and without any break, the times for observing used by Bowers were not convenient to Meares. The latter, thcrefoie, made his observations at (a) 4 a.m., (b) 8 a.m., and (c) 4 p.m. approximately. The following table gives the two sets of observations for a short period of four days: — Table 3. Bovvers. Meares. Hour;;. i-' Hours. °r November 17 3 -19 November 17 4 -H H -12 9 — 4 21 -12 15 2 ,, 18 n — 21) !8 5 -III 9J - 4 9 + 4 201 — 0 16 1- 4 1!) 3 -1.5 19 4 — 0 10 _ 2 8 0 21 J - 0 12 + s ,, 20 . H -11 20 4 - 4 m + 2 8 + B 21i - 1 Mean -9-3"F Mean — loT Each ot the observers took observations three times a day and yet the mean tenipoia- tures shown by the observations differs by nearly 8 degrees. The reason is at once apparent 20 TEMPERATURE. as soon the observations arc plotted. Tliis has been done in the upjier curve in fi'mre 5 Bowers' observations being sliown ])y a dot in a circle and those of Meaies by a circle. Fig. 5. Tempeiatuies on Barrier. It will be seen that with one or two exceptions the observations lie on a regular curve. Another set of observations made in the same region at the same time proves the correctness of the temperature curve arrived at by combining these two sets of observations and shows also the imperative necessity of taking the minimum temperature into account in calculating the mean temperature on the Barrier. A small party had gone ahead of the main party to a pre-arranged position where they waited until the main party came up. The junction was effected on the 21st November when, unfoi-tunately, the detached party ceased to record temperatures. This party had been taking observations three times a day and had used a minimum thei'mometer under a sledge to obtain a nigbt temperature. The results are plotted in the lower curve of figure 5, and it will at once be seen that the second curve GEOGRAPHICAL DISTRIBUTION OF TEMPERATURE— SEA-LEVEL. 21 is almost exactly a replica of the first. It will also be noticed that if the temperatures recorded during the day (shown by the dots on the diagram) only bad been considered the mean temperature of the period would have been much too high. Without the independent evidence of the low night temperatures shown in the upper curve which were obtained with a sling thermometer, one would have been in doubt as to the accuracy of the minimum temperatures shown in the lower diagram found by exposing the minimum thermometer under the sledge. These two curves show clearly that the daily variations of temperature arc nuich too great to be neglected, and that true mean temperature cannot be obtained without taking this variation into account. Nearly all the temperature observations made on sledge journeys have, therefore, been plotted and the mean daily temperatures determined by reading the curve at four-hourly intervals. The only exceptions are the tempeiatures of the first Return Party and of Cherry-Garrard's Relief Party. In these two cases minimum temperatures were not I'ecorded while observations were taken in the morning, in the evening and near to midday. If the midday temperatures had been taken without counterbalancing night temperatures the means would have been too high, hence in these two cases the morning and evening temperatui'es only have been used in obtaining the mean daily temperatures. Geographical Distributiox of Temperature. ; ■; Sea-level Tempmdures. Winter Temperature on the Barrier.— V\ hen the Discovery was in the Antarctic, 1902-04, the regular temperature observations were made in a Stevenson screen erected on the sea ice near to Hut Point, and other temperature observations were made at a position a little to the south of Cape Armitage (marked A on figure 2). The distance between the two positions was only one and a half miles, yet the temperatures recorded at Cape Armitage during the winter were verv much lower than those observed at the same time at Hut Point. Differences of from 10° to 20° F. in the recorded minimum temperatures were quite common, and differences of even 25°F. between the two places occurred at times. These large differences could not be understood and Mr. Dines, who discussed the Discovery temperature observations, was inclined to believe that ' the thermometer screen at the Discovery was placed too near the tide crack, and that the temperatures are unduly high on account of the influence of the sea-water in the crack.' Captain Scott strongly repudiated this explanation, and those who are familiar with ice conditions will agree with him that even if there had been a tide crack near the screen, which there was not, it could not have affected the temperatures to anything like the extent shown by the observations. The explanation \nll be apparent at once when we have considered the temperature observations made on the journey now to be described. A party led by Wilson left Cape Evans on 27th January, 1911, to visit the emperor pen- guin rookery at Cape Crozier. The path they took is shown on figure 2. They went first over the sea ice to Hut Point, and while on this part of their journey the temperatures they measured were similar but slightly higher than those measured simultaneous!}' at Cape Evans. At 1-30 p.m. on the 28th, when they were just off Hut Point, they measured - 26°F. which was the temperature at the same time at Cape Evans. They then rounded Cape Armitage and apjiroached the edge of the Barrier, which at this point was only 12 feet above the sea ice. Wilson in his report states: 'Coming down the snow slope oft' the Barrier was a stream of very cold air w-hich we felt first when we were only a few yards from the foot, and lost very soon after reaching the top.' At 7 p.m. they camped on the Barrier, and at 8 p.m., when meteorological observations were taken, the temperature was found to 22 TEMPERATURE. be _ H°p., which was 18°F. colder than the temperature they had measured off Hut Point only six and a half hours previously. That there had been no general fall of temperature was shown by the Cape Evans temperature having risen by two degrees in the same period. Assuming, therefore, that there had been little or no change at Hut Point, the temperature on the Barrier must have been nearly 20 degrees lower than the simultaneous temperature at Hut Point only five miles away. The stream of cold air encountered at the edge of the Barrier must have been the direct consequence of this large diflierence of temperature. On the return journey the experience was repeated. At 9 hours on the morning of the 31st July, seven miles from the edge of the Barrier, the temperature was — 57°F., which was 24 degrees below the corresponding temperature at Cape Evans. The party reached the Barrier edge and descended to the sea ice just before 3 p.m., and Wilson wrote: 'Here again we felt the flow of cold air pouring from the Barrier on to the sea ice, so we camped about 100 yards away to be out of it and had lunch. The temperature here was — 43°F.' They were then three miles from Hut Point and therefore not far from the position where the Cape Armitage thermometer bad been exposed in the Discovery days. They arrived at Hut Point at about six o'clock and here the temperature was —27°. Thus the temperature rose 14° on leaving the Barrier and another 16° in the thiee miles between the Barrier edge and Hut Point, while the tempera- ture at Cape Evans had only varied from -33° to —28" F. during the whole period. From these observations we see that the temperature at Hut Point was the same, or nearly the same, as at Cape Evans and that the temperature on the Barrier was between 20 and 25 degrees lower. This explains the low temperatures recorded in 1902-04 at Cape Annitage : they were the direct consequence of the cold air flowing off the Barrier from which the small bay in which the Discovery lay was entirely screened. The remaining observations of the Cape Crozier party will now be considered to illus- trate the temperature conditions on the Barrier in the depth of winter. The position of the party at the evening camp of each day is shown in figure 2. The following table gives the mean daily temperatures recorded with the corresponding tempera- tures at Cape Evans, and general weather notes are added: — * Table 4. Date. Tempebature. Barrier. Cape Evans. Weather. Differ- ence. 1 .June 29 . .,30 July 1 2 -.50-9 -59-7 -64-2 -620 op -25-7 -31-7 -361 -34-4 t °F 1 — 25-2 Clear ami calm at Cape Evans and on Barrier. — 28'0 Do. dn. do. -278 1 Do. do. do. -27-6 Do. do. do. * As stated on page 21, the individual observations on the Barrier were u.scd to plot the course o£ the temperature throughout the day, and by measuring up this curve at four-hourly intervals a much truer value of the mean temperature for each day was obtained than would have been the case if the observations alone had been considered. In the following discu.ssion, unless otherwise stated, mean temperature.^ have always been obtained in this -nay GEOGKAPHICAL DISTRIBUTION OF TEMPERATURE— SEA-LEVEL. Table 4 — conld. 23 Date. TiiMPEEATUBE. Weather. Barrier. Cape Evans. Differ- ence. 'E oj,' op July :{ . —56 7 -35-7 -21 0 Calm with little cirrus at Cape Evans and cm Barrier. 4 . —36-9 -2o-2 -11 7 High wind with snow at Cape Evan.'i and un Barrier. 5 . -5.5-4 -324 -230 Calra with cirrus at Cape Evans and on Barrier. (> . -72-9 -43-7 -29-2 Calm and nearly clear at Cape Evans and on Barrier. 7 . -04-9 -40-2 -24w In morning calra and clear ; in evening blizzard started at Cajie Evans but not on Barrier. S . -48-9 -347 -14'2 Blizzard with very high wind at Cajie Evans, calm on Barrier with light cloud. 9 . -26-7 -19-4 - 7-3 Do. do. do. HJ . -19 0 + 1-6 -20-6 Blizzard at Cape Evans and on Barrier. ,. 11 . 4- 5-3 + 6-0 - 0-7 To do. 12 . - 20 > 3-3 - 5 3 Very strong blizzard at Caj e Evans and on 13 . -23-7 - 4-0 -19-7 Barrier. Blizzard cea.sed on Barrier at 3 a.m. and at Cape Evans at 2 p.m. Sky partly cleared. ., 14 . -23-5 -7-7 -15-8 Light variable winds with cloudy sky at Cape Evans and on Barrier, fciome light snow. 15. -20-5 -11-4 - 91 .Moderate wind and partly overcast sky at Cape Evans and Cape Crozier. Party at Cape Crozi.?r. I'i . -23-9 -16-2 - 7-7 Sky nearly clear but moderate variable wind at Cope Evans and Cape Crozier. .,17. — 22-2 -18-0 - 4-2 Sky nearly overcast and light to moderate wind. ,, 18 . -270 -17-8 - 9-2 Little cloud, more wind at Cape Crozier than at Cape Evans. 19 . -30 9 -343 h3-4 Calm and clear at Cape Evans and at Cape Crozier. 20 . -25-5 -27-5 f 20 [light wind in morning and high wind in evening at both places. „ 21 . —22 2 -30 0 + 7-8 Clear at Caps Evans ; overcast at Cape Crozier, light wind. 22 . — 154 •• Heavy blizzard commenced at Cape Crozier at 3 A.M. and at Cape Evans at 7 a.m. , 23 . •• - 2-3 Blizzard of hurricane force at Cape Crozier and gusts over 60 miles per hour at Cape Evans. ., 24 . ■• - 61 •• Blizzard stopped early in the morning, after- wai-ds dull day with light wind. 25 . -16-5 — 9-1 — 7-4 Wind rose again at about middav. Over- east sky on Barrier again. 24 TEMPERATURE. Table 4 — concld. Date. Temperature Weather. Barrier. Cape 1 Evan-i. DifEer- ence. oy op •F July 2U . . -29-4 — 15! -14-3 Wind continued until near midday, then fell calm. 27 . -4(5-7 -20-2 -20-,5 Clear and calm at Cape Evans and on Barrier. ,, 28 . -43-4 — 28-(i -14-8 Clear and calm on Barrier. Clear with light wind in afternoon at Cape Evans. ,, 29 . -45-5 -18-0 -27-5 Calm on Barrier, high wind at Cape Evans until the evening.' sn . . -63-4 -33 ! — 30-0 Clear and calm at Cape Evans and on Barrier. During the first few days that the party was on the Barrier calm weather with little or no cloud prevailed over the whole region. In consequence low temperatures occurred, but while the mean temperature did not fall below -36-4°F. at Cape Evans it fell to — 64'2 F. on the Barrier. During these five days the mean temperature on the Barrier averaged 26°F. below the corresponding temperature at Cape Evans. On the next day a high wind accompanied by overcast sky and snow set in. The temperature at once rose by 10 degrees at Cape Evans and 20 degrees on the Barrier, but the Barrier temperature remained on the average of the day 11 degrees below that of Cape Evans, althougli at 9-30 .\.m. the difference was only G degrees. After the wind had died away a period of extreme cold set in on the Barrier. On July 6th (for position see figure 2) the temperatures measured on the Barrier were the following : — Table 5. On Barrier. Cape Evans. Difference. 'F. op F. Minimum during night between 5th and 6th —74 -44 -30 Temperature with sling thermometer at 9 h. 3(» m. -oa -45 -24 Do. do. do. 12 h. 00 m. — 76 -44 -32 Do. do. do. 17 h. 15 m. —76 -43 -33 Do. do. do. 24 h. 00 m. — 38 -48 -20 The Barrier temperatures were taken with extreme care. Against tiie entrv in the meteo- rological log at 17 hours 15 minutes, Bowers has noted ' Carefully checked by Wilson ' ; lie also measured the temperatures with a spare thermometer and found the two agreed within half a degree. This is the lowest temperature recorded on the Barrier — or anywhere in the Antarctic — audit is satisfactory to know that it was so carefully checked. The mean temperature on this day was ■-72-9°F. on the Barrier and — 43-7°F. at Cape Evans. Thi.s was the coldest day at Cape Evans during our stay in the Antarctic and no doubt GEOGRAPHICAL DLSTRIBUTION OF TEMPERATURE— SEA-LEVEL. 25 it was the coldest day on the Barrier.* Although the temperature was so low at Cape Evans the difference between the Barrier and Cape Evans has only been exceeded on a few occasions, most of which occurred in March, 1912, during the abnormally low temperatures which proved fatal to the Polar Party (see page 29). This cold period was succeeded at Cape Evans by the most severe weather of which I have been able to obtain a record. During the morning of the 9th. while the temperature was still — 35°F., the wind rose to 43 miles in the hour. Between 7 and 8 P.M. the wind velocity was 63 miles pei hour with a temperature of — 31°F. As the blizzard continued the temperature slowly rose until it reached +8°F. at 2 p.m. on the 11th, i.e., a little over three days after the startin" of the blizzard. The blizzard was not felt by the party on the Barrier near Terror Point until the early morning of .Tul\- 10th— twenty hours after it had commenced at Cape Evans ; but during this period the temperature had been rising rapidly and steadily with an overcast sky from which snow was falling. When the blizzard set in on the Barrier the wind was very high and accompanied by much snow and the temperature rose more rapidly until on the morning of the 11th it became the same as that at Cape Evans. The blizzard ceased on the 13th when the temperature at once fell, but more rapidly on the Barrier than at C!ape Evans, and in consequence large temperature differences again became established. On the 1.5th the party reached Cape Crozier and camped on the shoulder of Mount Terror at an elevation of over 700 feet above the Ross Sea which lay completely frozen over below them. For six days they measured the temperature at this place, during three of which it was lower than at Cape Evans (average difference —7-0) while on the remaining three it was higher (average difference +4-4). On some of these days there was a moderate amount of wind and on others the weather was calm, but no blizzard occurred to equalise the temperature over a large area. It therefore appears that the temperature at Cape Crozier is very similar to that at Cape Evans. After being six da_vs at Cape Crozier a terrific blizzard arose and lasted for three days : 22nd-24th July. During this blizzard the party was nearly lost owing to the blowing away of their tent and the unroofing of the temporary hut which they had built. During this period the weather was too bad to allow of temperature observations being taken. When the blizzard was over the party was forced to return to headquarters with all haste. They spent the 25th and 26th in getting to the true Barrier surface which they traversed on the next four days, finding again the low temperatures which they had encountered there on their way out. The mean temperature for the four days was — 49-8°F. with a corresponding temperature of — 26-6°F. at Cape Evans, a difference of — 23-2°F. The leniainder of the journey back to winter quarters, how the temperature rose on leaving the Barrier and how the temperature at Hut Point was found to be the same as that at Cape Evans, has already been described. This journey which Scott has described as ■ the hardest which has ever been done ' has revealed conditions on the Barrier in \vinter which had previously been suspected but never proved. They are the same as subsequent observations showed to hold throughout the whole year, but much accentuated. To act as a guide in our future study they will be summarised here. (ft) Temperatures on the Barrier during the winter are much lower than those found in McMurdo Sound and at Cape Crozier near to the frozen Ross Sea. * A minimum thermometer left at One Ton Camp throughout the winter recorded — 730°F. and the lowest temperature measured at Framheim was — 73\!'F. 26 TEMPERATURE. (6) The greatest difference is found on clear, calm days. (c) High winds not only raise the temperature in McMurdo Sound and on the Barrier, but also reduce the difference and occasionally remove it altogether. Spring Temperature on the Barrier. — After the return of the Cape Crozier Party the Barrier was not visited again until September when the sun had returned and spring condi- tions were becoming established. On September 9 a small party led by Evans left Cape Evans to visit Corner Camp (for position see figure 2). At Hut Point they again had temperatures similar to those at Cape Evans, and on reaching the Barrier low temperature;; were immediately experienced. The following table gives the mean daily temperatures met with during a stay of four days on the Barrier :— Table 6. Mean Temperature. Weather. Date. Barrier. Cape Evans. Differ- ence. Barriei. Cape Evans. o-p •F. °F. September 10 . -41-5 -21-0 -17-5 Light air, Ci. S. Calm or light wind, overcast some snow. 11 . -4.5-5 -30-4 -15-1 Blizzard in evening. Calm, overcast. 12 . -3.5-2 -21-7 -13-5 Blizzard Calm or light wind, overcast. i:! . -1-1-2 -23 1 -21-1 Light: to moderate southeily 'wind.s with low S. clouds. Calm. Sky cleared during day. Mean -41'6 -21-8 -16-8 On the journey a minimum temperature of — 72°F. was registered. Although much wind was met with on the Barrier and the sky was overcast, low tem- peratures prevailed on the Barrier, where the mean temperature for the period was 16-8°F. below the simultaneous temperature at Cape Evans. Summer Temperatures on the Barrier. — During the summer (from 1st November, 1911), there was a great deal of travelling backwards and forwards on the Barrier. Each party was provided with meteorological instruments and in consequence a great many data are avail- able. It would take up too much space to discuss the results of each party in detail and therefore the actual observations have been printed in Vol. Ill, and here we will confine ourselves to a general discussion of the results. The curves on figure 6 show the tempera- ture observations in a very comprehensive manner. In this figure the mean daily tempera- tures at Cape Evans are shown by the upper thick line. The observations on the Barrier have been divided into two sets, one comprising those made to the north of One Ton Camp and the other those to the south. The former are shown by a thin continuous line and the latter by a thin broken line. The observations made on the plateau are shown GEOGRAPHICAL DISTRIBUTION OF TEMPERATURE— SEA-LEVEL. 27 j3 o t^ ti a -> M O u V a o f ¥ ^ - b s r>-i "-> * < 7 1 ^ to - ~^_, ® ( c _ •« > < -< — IZ== — -J i^— - ~--.^ r-- .--- — "' c :=> < ? 10 J^oz^ < > 1 1 :_-,. -< ^ ::^ \_ ^ =7 ^^^^" r >. x- " ^ ^ s^ 1 < 'S r r^ "l^r-- s r' Bcrjiornnri. y^ , -*'i / - ^t F ?^ /- ^ ■ Cape Evans . North oF One Ton Ca South. " Beardmorc G acier Polar Plateau c "> • ^^^ "S 1 > - -^ ! — ^ 3 ? 2 "C D •= 2 'S "« o ^ s o 28 TEMPERATURE. by the lower thick line, while the observations made on tbe Beardmore Glacier are shown by the thick dotted line which joins the plateau curve to a Barrier curve at each end. At present we are not concerned with the temperatures measured on the Beardmore Glacier and the plateau. It will be noticed that between the 1st November and ith December all the Barrier temperatures are below those of Cape Evans, and that during tbe few days in November, when there were parties to the north and south of One Ton Camp, tbe southern tempera- tures were the lowest except on two days. During this period the Barrier to the south of One Ton Camp was on the average 10 degrees colder than Cape Evans. On December 5 the Barrier was affected by a deep depres.sion which appears to have caused a great inflow of moist warm air from the Ross Sea into the south of the Barrier. The main southern par-ty was at the time near the southern limit of the Barrier 12 miles from the foot of the Beardmore Glacier. Readers of Captain Scott's book will remember how serious this storm was to the Southern Party, causing four days' delay and covering the surface with three feet of soft wet snow. The general conditions which accompanied this storm will be considered later ; at present it is important to us as showing that high temperatures occur sometimes on the Barrier as far south as 83|° S. A very remarkable circumstance was that during this period the temperature was higher in the extreme south than at Cape Evans. Table 7. To show the Increase of Temperature towards the South. Cai-e Evans. MOIUK i'AKTV. tSoUTUEKN 1'aKTY. Date. LatituJu liy. Latitude 79J°. Latitude S3J°. °F. ^F. op December 5 . . . 19-2 160 28-2 „ 6 . . . 2-t-3 29-9 320 ,. 7 . . . 19-4 23-7 310 S IMS L'2!) :;i 0 Tbe probable cause of this inversion of the geographical distribution of temperature was that the air from the Ross Sea passed first into the south of the Barrier far to the east and then travelled northwards along the western edge, and had become cooled by the time it reached Cape Evans. Support is given to this explanation by the fact that the snowfall was much greater in the south than in the north. The weather was disturbed throughout December on the Barrier, and as will be seen from the curve it was not until the 4th January that the Barrier temperatures became again consistently below those of Cape Evans. Throughout January this was so, but it will be noticed that sometimes the Barrier was coldest to the north and sometimes to the south of One Ton Camp. During December the Barrier temperatures had varied on the whole very little from those of Cape Evans ; but in .January the difference steadily increased, except GEOGRAPHICAL DISTRIBUTION OF TEMPERATURE— SEA-LEVEL. 29 during a few disturbed periods. As February advanced the Barrier cooled very rapidly and by February 8 the Barrier was l.'3°F. colder than Cape Evans. On February 18 the Polar Party reached the Barrier on their return from the Pole, they found lower temperature than they had expected but after two days on the Barrier the temperatures rose to a mean daily temperature of — iO°F. This recovery was, however, very short-lived and from J'ebruary 22 the temperature fell at an amazing rate from — 4'.'3°F. on February 22 to — 43°F. on March 9. As everyone knows this low temperature, so entirely unexpected, was one of the chief causes of the great tragedy. After March 9 regular temperature observations ceased,, but from a few entries in Captain Scott's diary it appears that the tem- perature remained in the neighbourhood of -40° until March 20. During these last few days the difference between the Cape Evans and the Barrier temperatures was over 40° and exceeded the largest difference found by the Cape Crozier Party by more than 10 degrees. During the latter part of this period a second series of temperature observations was made on the Barrier. Cherry-Garrard had left Hut Point on February 26 to meet the returning party with the dogs. He reached One Ton Camp on March 3 and remained there until the 10th. The following table gives the temperatures during these days :— T.^BLE 8. Datf. Cajx; Evans. Latitude 79-2-" One Ton Camp. Latitude 81" Polar Party. >rarch3 „ 4 ... „ 5 . . „ 7 ... ,.8 ... ^F. 5-7 65 4-6 5-4 7-7 2-1 Diftorci.cc F -23-7 -21-4 (-35-0) (-35-0) -350 -39-0 1 Difference 1'. -245 -17-.5 -260 -17-5 -200 -.{3 0 F. -30-2 -24'0 -300 -229 -27-7 —351 F -29-4 — 27 'J -39 0 -40-4 -42-7 -411 Mean 5 3 -231 1 -2S-4 -31-5 — :>(is These low temperatures on the Barrier so early in the year were entirely unexpected and have no parallel in either north or south polar regions. It is therefore of importance that we should know whether they are the normal conditions in that region. Some light is thrown on the question by a short journey made by Atkinson between 27th and 31st March, 1912. when he made an unsucce.ssful attempt to reach the Polar Party whose non-appearance had given grave cause for anxiety. Atkinson was only able to go from Hut Point to Corner Camp, but this was the same tract which Cherry-Garrard had crossed in his return from One Ton Camp a few day.s previously. The teinp(U.atuios arc shown in figure 6 and the following table gives the data: — 30 TEMPERATURE. Table 9. To show Rise in Temperature over the North of the Barrier during March, 1912. Date. Cape- Evau;- Hut X'oiut to Corner Camp. Difference. "F. "F. F. March U +0-5 -ie-5 -170 Cherry-Garrard . . \ ,. !•'> — :VG -18-5 -US) 1 ,.16 —7-7 -18-0 -10-3 /* March 27 + 11-9 -10-2 -221 .,28 +11-2 - 50 -16-2 Atkinson ., 29 + 111 - 70 -181 ,.30 + Gi -13-7 -201 I .31 - 2-8 -17-2 -U-i Atkinson's temperature.s sbow a great rise on tho.se experienced by Cherry-Garrard nearly a foi-tni"ht before, and this is more easily realised by examining figure G in which the rise between Cherry-Garrard's last observations and those of Atkinson is clearly shown. These observations and the general trend of the curves in figure G can leave little doubt that the temperatures experienced by the Polar Party were abnormally low. Captain Scott was well aware of the rapid fall in Barrier temperature during February, but his statement in his wonderful Message to the Public that ' no one in the world would have expected the tem- peratures and surfaces which we encountered at this time of year ' was certainly justified. It is difficult to discuss the records of this period mingled as they are with such tragedy, but they clearly bring to light the possibility of great cold at an extremely early period in the year within a comparatively few miles of an open sea where the temperatures were over 40 degrees higher. The cause of these low temperatures will be considered later, after other temperature conditions have been discussed. Monthly Values of Barrier Temperature.— T:he actual observations of temperature on the Barrier are far too few to determine directly the mean monthly Barrier temperatures, but an indirect method allows us to form some idea of this important factor. It is well known that variations in meteorological conditions are generally similar over fairly large areas. Thus the differences in a meteorological element between two stations undergo much smaller changes than the actual element itself, and while many years may be necessary to find the mean value of the element at cither station, a much shorter time is necessary to obtain a reliable value of the difference. If, then, we know the mean value of the element from a long series of observations at one station and the difference between that and another station from a short series, the mean values at the latter can be obtained by applying the known differences to the mean values at the former. We now have temperature data for McMurdo Sound for five years, from which the mean monthly temperature can be obtained with some fair degree of accuracy, and these values are given on page 81 below. We have now to see if it is possible to determine from the few available data the difference between McMurdo Sound and the Barrier. The temperatures measured on the Barrier have been divided geographically into the regions north and south of One Ton Camp, 791" S. This geographical division is not all GEOGRAPHICAL DISTRIBUTION OF TEMPERATURE— SEA-LEVEL. 31 that could be desired, because it sometimes happened that observations which fall in the two divisions were made only a few miles apart, for example when two pai-ties weie near to One Ton Camp, one to the north, and the other to the south ; while at other times the whole expanse of the Barrier separated the parties taking the observations. Still this is the best that can be done with the data. The mean daily temperatures for the Barrier have been obtained from the ob.servations made by the sledging parties in the way already described. These have been collected according to months and the geographical positions, and the results entered in the following table : — Table 10. • Mean Temperature Difference., Barrier-Cape Evans. Month. BARRIER NORTH OF ONE TON CAMP. BARRIER SOUTH OF ONE TON CAMP. 1 Number of days. Mean day of psriod. ATeax Tempekature. Number of days. Jlean day of period. Mean Temperature. Cape Evans. Barrier. Differ- ence. Cape Evans. Barrier. Differ- ence. °F. =F. F. °F. °F. op January 26 13th +20-4 + 13-6 - 6-8 17 8th +20-3 + 13-6 — 0-7 February . 28 14th + 18!) + 3-9 — 15-0 1' 24tl) + 4-8 — 16-9 -21-7 March 27 16th + 4-4 -170 -21-4 9 4th + .5-2 -32-9 -381 July . 20 10th -23-4 -431 -19-7 September . 4 12th -24-8 -41-6 -10-8 •• ■■ October 4 29th + 30 - 4-0 — 70 November . 17 8t1i + 10-8 + 2-1 — 8-7 21 20th + 12-2 + 11 -111 Pecpmher . 18 1.5th +23 1 +20-9 o.o 28 17th +21-7 +20-9 - 0-8 T^liA f,amr\arQ MlTd fll-p aT'cinr'O.j V»Af AX'OOl 1 r'o»-io lH^.^rQ»-l« 0 11 A tKo P nmar ii m-fVi nf On.. T r\ii T'cimit are shown in figure 7 by dots enclosed in circles, while the differences between Cape Evans o 0 0 -10 0 -20 O -30 O -40 / V / ^^ s. A \ y / \ \ \ / \ N O J ^Q X 0^ — ©-Bar X.. _ > rier nort _ sout h One Tc h- "- n Camp X Nov ' Dec Jan. Feb. Mar. Apr. May. June. Jul/ Aug. Sep. Oct. Nov. Dec. Jan. _^ Fig. 7. Temperature difference between Cape Evans and Barrier. 32 TEMPERATURE. and the Barrier south of One Ton Camp are shown by small crosses. At first we will dis- regard the latter and fix our attention on the former. It will be seen at once that the circles lie fairly regularly about a curve having a maximum in December and a minimum in May or June, and there can be little doubt that this curve gives approximately the mean difference in temperature between Cape Evans and the Barrier north of One Ton Camp throughout the year. By taking the mean values of this curve for each month and applying them to the mean temperature of McMurdo Sound we obtain the mean temperature of the Barrier. Table 11. Mean Temperature of MeMurdu Sound and of the Barrier north of One Ton Camp Month. McMurdo Sound 5 years. Difference from curve. Barrier north of One Ton Camp. January .... February .... Mareli .... April .... May ..... June ..... July August .... September .... October .... November .... December .... F. +24 + 16 + -t — 9 -Id -12 -15 -15 -12 _^ 2 + 14 + 25 F. ( -15 -10 -21 -23 -23 -21 -HI -15 -11 — r. - 3 "F. + 17 + 1 — 15 -30 -33 -35 -36 -34 -27 -13 + 8 _i_o.:i Year J- 0-7 -15 — 15 Turning now to the temperatures on the Barrier south of One Ton Camp we see from the crosses in figure 7 that it is impossible to -construct a similar curve giving the annual variations of temperature for this region. It has already been explained that the high tem- perature over the south of the Barrier in December was mainly due to a period of very unsettled weather which may or may not be usual at this time of year. The only safe conclusion to be drawn from the summer observations is that during November, December, and January, there is no large consistent difference in temperature between the north and south of the Barrier. After January, however, the temperature appears to fall much more rapidly over the south than over the north of the Barrier, but for reasons already given it is more than probable that 1912 was an abnormal yeap. It is quite impossible to believe that normally there is a difference of nearly 40 degrees in March between McMurdo Sound and the south of the Barrier. In fact the position of the cross for March in figure 7 is further support, for the contention that Captain Scott experienced unusually low tempera- tures on his return from the Pole. We are therefore left mth the conclusion that the temperatures after the summer are lower over the south than over the north of the Barrier, but that the amount of the normal difference is unknown. Framheim.—FoT our study of the geographical distribution of temperature the observations made at Framheim are of the greate.st import.ance. Although Fraraheim was so near the GEOGRAPHICAL DISTRIBUTION OF TEMPERATURE— SEA-LEVEL. 33 Ross Sea it was actually on the Barrier and appears to have experienced in consequence true Barrier temperature.s as will be seen from the following table : — Table 12. Ti'mperiUnre at Framheim 1911-12. Month. Tempera- ture at Framheim. Difference Framheim — Cape Evans. Month. Tempera- ture at Framheim. Uitierertee Framheim — Cape Evans. -April May .June July August F. -17-7 -31-7 -29-9 -33-7 —ISO ^F. -16-6 -20-9 -lli-4 -12-6 -27-5 .September October November December January . 'F. -3.5-5 -11-6 + 4-1 -fl9-9 -l-U-o °F. -19-7 - 8-2 - 8-2 - 2-1 - 68 It will be shown later that Framheim shares with other parts of the Barrier a large daily range of temperature. Now at Framheim temperature observations were made three times a day at 8 hours, 14 hours and 20 hours. The 14 hours' observations fell at the time of the daily maximum temperature and as there were no corresponding observations taken at or near the minimum, the mean of the observations spaced as these are must be too high. Hence in calculating the mean temperatures given in the above table the 14 hours' observations have been rejected and the observations at 8 hours and 20 hours only consi- dered. There can be no doubt that the mean temperatures calculated in this way are nearer the true values than if all three observations had been taken into account. Temperatures at Cape Adare.— T^\i'i observations available for determining the mean tem- peratures at Cape Adare will be discussed later (page 83), we will use here the mean values which will be then derived. Table 13. Temperature at Cape Adare. Temperature. Month. Temperatuke. I Month. Cape Adare. Difference Cape Adare — McMurdo Sound. Cape Adare. Difference Cape Adare — McMurdo Sound. .(anuary . February . M.xrcli April May June -t-31-6 -f-27-0 -I-18-7 + 9-4 — 2"2 -14-5 °F. + 7-9 -t-Il-2 -f U-3 + 18-2 r 8-3 - 2-6 July August . September October . November December °F. -11-9 -136 - 7-5 - 0 0 H 18-5 -t-29-5 °F. -f2-7 -l-l-O -f4-2 + 1-5 + 4-3 +4-G Year -J- 7-0 -fC3 The most interesting and important conclusion to be drawn from this table is that during the winter from June to Augu.st there is practically no difference between the mean tem- perature at Cape Adare and at Cape Evans although they are separated by more than six degrees of latitude. The difference is greatest during the three months February, March, and April, being then about 15°F. 34 TEMPERATUEE. 170' W£3T LONC.. lei/ Fig. 8. laotherms for January. We are now in a position to discuss the probable geographical distribution of temperature in the Ross Sea area. As typical months January and July will be taken. The temperatures used are the mean temperatures obtained from all available observations. Table 14. Normal Temperatures. Cap(t Aflaro McMurdo Sound Framheim .... Barrier North of One Ton Oamp January. °F. "F. +32 -12 +24 -15 +15 -34 + 17 -36 July. These data have been entered on maps and an attempt made to construct isotherms from them. The results are shown in figures 8 and 9. The considerations which have led to the form of the isotherms arc the following. GEOGRAPHICAL DISTRIBUTION OF TEMPERATURE— SEA-LEVEL. 35 WEST LONG. 160 FlO. 9. Isotherms foi- July. Isotherms for January. — The temperature in McMurdo Sound i« 24° E. and at Cape Adare 32°F., it is therefore probable that the temperature over the open water of the Ross Sea is approximately 30°F. The floating ice and ice-covered land will probably reduce the temperature to 25°F. near the south and east of the sea. The 25°F. isotherm has therefore been shown to encircle the sea near to the Barrier and to the permanent ice in the east, with the 30°F. isotherm surrounding a warmer central area. The temperature of the whole of the Barrier i.s within a degree or so of 1.5°F.. which fixes the position of the L5°F. isotherm. The 20°F. isotherm has been drawn between those for 15°F. and 25°F., and we see that the closeness of the isotherms near to the edge of the Barrier indicates the large temperature gradient which must exist between the cold Barrier surface and the relatively warm open sea. Isotherms for July. — The temperatures at Cape Evans and at Cape Adare are practically the same during June, July, and August and they are more than 20° above those of the Barrier. We shall show later (see page 90) that during the winter the temperature of the sea ice decides the temperature of the air above it and one knows that the sea ice is much thinner 36 TEMPEKATURE. over the centre of the Ross Sea than either in McMurdo Sound or at Cape Adare. AVe also know that during the winter open leads often form in the sea to the north of Cajie Crozier. Fig. 10. Temperature distribution along ISO" E. longitmle. We are therefore quite safe in assuming that the mean temperature over the centre of the Ross Sea is somewhat liigher than at Cape Evans and at Cape Adare. The centre of the Ross Sea has therefore been encircled by the — 10°F. isotherm. The discontinuity of tem- perature at the edge of the Barrier found to exist in McMurdo Sound by the Cape Crozier Party is much more likely to exist along the main edge of the Barrier ; hence we have here a temperature discontinuity with several i.srtherms following the edge of the Barrier too close to be separated. The sea ice is probably very thick near Framheim during the winter as it forms early, and the atmosphere is still and cold in this region. Hence it is probable that the isotherms open out near Framheim in the manner shown in the map. The physical meaning of the temperature distribution which is shown by the isotherms for July will probably be made more clear and convincing by setting out the temperature distribution in another way. The character of the ice along the meridians 180° E. and 150° W. is .shown diagrammatically in figures 10 and 11. The former meridian runs down the centre of the Ross Sea and the diagram, figure 10, shows open sea north of about latitude G4° S., then a belt of loose pack gradually getting more dense until in about GEOGRAPHICAL DISTRIBUTION OF TEIMPERATURE— SEA-LEVEL. 37 latitude 67° S. the sea is completely frozen over, but the ice is broken up by a few cracks. South of latitude 78° S. we have the Barrier surface. 60 150°W.Lon| 10" 0 TEMPERATURt Flo. II. Temperature clistribiition along 150^ \V. longitude. In figure 11 the character of the ice along 150° W. is shown. Hero again we have open water north of about latitude 64° S. Then the belt of loose pack followed by sea ice inter- sected by opening and closing leads. It is almost certain that south of 70° S. on this meridian there is permanent heavy ice either in the form of a Barrier or of a snow-covered land. Therefore south of this latitude the surface conditions will be similar to those on the Barrier. On each of these diagrams there has been superposed a curve of temperature, the ordi- nates being latitude and the abscissa? temperature. Examining figure 10 we see that north of latitude 65° S. the temperature falls slowly with increase of latitude. In latitude 65° S. the loose pack is entered and the ice causes a rapid fall of temperature which continues as we proceed south until the more or less completely frozen centre of the Ross Sea is reached in latitude 67° S. From here onwards towards the south the temperature de- creases very slowly until the Barrier is reached in latitude 78° S. The discontinuity of temperature at the Barrier edge is shown by the temperature running horizontally for 15°F. From the edge of the Barrier towards the south the temperature again falls slowly. 38 TEMPERATURE. Turning now to figure 11 we again see the slow fall of temperature over the open sea and a rapid fall as the pack is reached. The rate of decrease of temperature falls off over the sea ice and when the heavy permanent ice is reached in latitude 70° S. the decrease of temperature as we proceed towards the south becomes slow and finally reaches the same rate as over the Barrier. It must be borne in mind that a great deal of what has just been written is based only on conjecture ; it would no doubt have been more accurate to have used the word ' probably ' before many of the statements, but for clearness that has not been done. As further actual observations are not likely to be obtained in the near future the above des- cription of the probable temperature conditions in the Ross Sea area must stand as best representing our present knowledge. Plateau Temperatures. Captain Amundsen reached the South Polar Plateau on 21st November, 1911, and left it on January 5, 1912, while Captain Scott reached it on 22nd December, 1911, and left it on February 6, 1912. Thus observations were made on the plateau continuously from 21st November, 1911 to February 6, 1912. As the plateau is not level the temperature varies on account of changes in height, hence it is necessary to reduce all the temperatures measured to a constant height before they can be compared. It has become an international convention to reduce the temperature to sea-level by applying a constant temperature cor- rection of +-5°C. per 100 metres of ascent (■274°F. per 100 feet).* In his discussion of the observations made on the plateau by Amundsen. Professor Mohn did not follow this convention, but from the observations concluded that the actual temperature gradient was •53° C. per 100 metres. He therefore reduced the temperature to sea-level by applying this correction. For the reasons given by Hann it is more satisfactory to retain in all parts of the world the same value of the temperature gradient and this has been done in the following discussion In reducing the plateau temperatures to sea-level it is necessary to know the height at which each observation was made. Mohn has calculated the height of each place where Amundsen's temperatures were measured and the same has been done for Scott's journey (see Chapter IX). Unfortunately the two determinations of the height of the South Pole do not agree, Mohn making it 2,454 metres while I make it 2,765 metres. As the observations must be reduced to the same standard it is necessary to take one or other of the above values. I have therefore increased all the plateau heights given by Mohn by 311 metres in order to make the height of the only common station agree in the two sets of observa- tions. Amundsen's temperature observations were taken at irregular times, on some days only two observations were made and on others three, four or six. I have taken as the mean temperature of the day the mean of two observations taken as far as possible at the same hour in the morning and evening. Thus on December 1st observations were made at 8-30 a.m., 1 P.M. and 5-30 p.m., the mean of the observations at 8-30 a.m. and 5-30 p.m. after reduction to sea-level has been taken as the mean temperature of the day. Scott's observations have been treated somewhat differently. All the temperature observa- tions were plotted and a curve drawn through the points as already described for the Barrier. This curve was measured at four-hourly intervals and the mean temperature for the 24 hours calculated for each day. The height at the lunch camp was taken as the height for the dav. * Sec Hanu Lehrbiich d. Mctcorologie, 3i(l edition, page 138. GEOGRAPHICAL DISTRIBUTION OF TEMPERATURE— PLATEAU. 39 During the period that Amundsen and Scott were on the plateau there were parties on the Barrier south of One Ton Camp (latitude 79|-° S.) except dusing the last 17 days so that observations are available for a comparison between the temperatures on the plateau and over the south of the Barrier. The observations on the plateau were made between 86° S. and the Pole, we may therefore take the mean latitude of the plateau as being 88° S. The mean latitude of the south of the Barrier may be taken as 82° S. In figure 12 the thick curve shows the mean daily temperature on the plateau reduced to sea-level, while the thin curve shows the mean daily temperature over the south of the 30"r SOTf lOV r Plateau \ {reduced to saa Jteyel SoLfth Barrier 20 25 November December January February Fig, 12. Plateau and Baiiier temperatures. Barrier. It wilt be seen that except at the end of November the reduced plateau tempera- tures were generally lower than the simultaneous temperatures over the south of the Barrier. It is interesting to notice the large correlation between the temperatures on the plateau and on the Barrier, all the larger changes affecting both regions. The high temperature on the plateau recorded by Amundsen at the end of November occurred during a blizzard, which kept him confined to his camp for four days. The wind during this blizzard was verv unusual, being from the N.E. At the same time the weather was nearly calm and cloudless on the Barrier. The high temperature both on the Barrier and plateau at the beginning of December was associated with the storm which held up Captain Scott at the foot of the Beardmore Glacier for several days. Table 15 contains the mean values of the reduced plateau temperatures, and the corre- sponding temperature on the south of the Barrier and at the base stations. Table 15. Mean Temqwrdtiire. Cape Evans. Framhcim. South Barrier. Plateau reduc- ed to sea-level. Latitude .... 77r 78i' 82° »^' December January .... 220 213 190 U-5 "F 21-0 17-3 7-4 40 TEMPERATURE. During December, as we have already pointed out, the temperature over the Barrier is practically uniform, the temperature over the south of the Barrier being the same as the mean iemperature at Cape Evans and Framheim. It seenis reasonable to assume that if the Barrier had extended right up to the Pole the temperature there would have been the same as that over the rest of the Barrier, viz., 21°F. The redured plateau temperature is however only 17-3°F., which indicates that during December the plateau is between 2 and 3°F. colder than its geographical position wariants. During January we have not sufficient observations from the south of the Barrier to determine the change of temperature with latitude on the Barrier. Plateau observations are however available for the whole of the month. From table 15 we see that during January the reduced plateau temperature was 10-3°F. lower than the mean of Cape Evans and Framheim. How much of this difference h due to change of latitude it is impossible to say, but if it were all due to latitude it would indicate a fall of temperature at the rate of rO°F. per degree of latitude. This amount seems impossibly large for this time of the year, therefore it is almost certain that the low relative temperature on the plateau both in December and January is not due to its geographical po.sition, but is mainly due to its high elevation. From this discussion it ajjpears clear (a) that the temperature on the plateau is lower than its geographical position warrants ; (b) that this relative deficiency of temperature is greater in January than in December ; (c) that the observations made on the plateau cannot tlierefore be used in determining the change of temperature with change of latitude. It is instructive to compare the reduced plateau temperatures with the sea-level tem- peratures at the corresponding latitude and time in the northern hemisphere. According to Mohn's * determinations the mean temperatures at 88° N. during June and July are 28'6°F. and 311°F. respectively. Thus we have Table 16. North and South Polar Temperatures. 88° S 88° N .Sea-level temperatures. December ■June op 17 3 28-6 January July 7 4 311 DiffiTonoe ... ... -11-3 . . . ! -23-7 Thus during the midsummer month the temperature in the neighbourhood of the South Pole, when reduced to sea-level at the rate of ■5°C. per 100 metres, is 11-3°F. (6-3''C.) lower than the corresponding temperature in the neighbourhood of the North Pole. The difference is still greater in the next month, for whereas in the north the temperature con- tinues to rise for a month after the solstice it commences to fall in the .south immediately after midsummer day.f In consequence the temperature is 23-7°F. (13-2°C.) lower in January near the South Pole than near the North Pole in July. We have so far considered only the plateau temperatures after they have been reduced to sea-level, but it is of considerable interest to know the actual temperatures which were * The Norwegian North Polar Expedition, 1893-1890, Vol. VI, Meteorology, by H. Mohn, page 575. •f See page 88. VERTICAL DISTRIBUTION OF TEMPERATURE. 41 experienced on the plateau by the two parties ; the following table has therefore been pre- pared (see also figure 6) : — • Table 17. Actual Teiiiperdtures on llie South Puhir Phtleau. Month. ObservtT Mean observed j Maximum observed j Minimum observed temperature. temperature. [ temperature. December January Ari.undsen _ 8-6'F= -22-6T \ +5o°F=-U-°C — 19-3°F=-28-r)°C Scott . . — lS-7'F = -28-2°C — 3-2°F = — 19'6°C ' — 29-7°F=— 34 S C Thus the mean actual temperature in December was — 8-6°F. It must be accounted as one of the wonders of the Antarctic that it contains a vast area of the earth's surface where the mean temperature during the warmest month is more than 8° below the Fahrenheit zero, and where throughout the month the highest temperature was only 4-5'5°F. It should also be noticed that the actual temperature on the plateau was lO^F. lower in January than in December, indicating a very rapid lowering of the temperature immediately the solar radiation commences to decrease. If this fall was not abnormal it indicates that the tem- perature on tiie South Polar Plateau is influenced by changes in insolation much more than any other place for which we have records. Vertic.vl Distribution of Temperature Having now used all the available data for discussing the geographical distribution of temperature at ground level, we will turn to the question of its distribution in the upper atmosphere.* In figure 13 are plotted all the observations made by the aid of balloons of the tem- perature of the upper air over the frozen McMurdo Sound. In this diagram heights are shown in metres and temperatures in centigrade degrees. It will be noticed that all the ascents made in November and December, summer months (Nos. 5 to 10), show temperature decreasing with height and the mean gradient for the first 2\ kilometres is 6-8°C. per 1,000 metres. It is of great interest to notice that this is practically the same gradient as tliat found in summer months in Europe and Ameiica although the actual temperatures in the Antarctic summer are similar to those of winter in those regions. The conditions are entirely different during the winter. On account of darkness upper air observations could not be made before August. In this month, however, four successful balloon ascents were made and the instruments recovered. The difficulties of the balloon work in cold weather were so great that only moderate heights could then be reached. The ascent on August 13, No. 1, was made on one of the coldest days of the year, the tem- perature on the ground being — 39°C. at the time of the ascent. The temperature record on this and on all other occasions in the winter revealed the temjjerature inversion which is often met with on calm winter days in Europe and America. The inver- sion extended higher on August 13 than on any of the other days, but it was not so pronounced as on August 16. The curve for August 30, No. 4, is interesting as it was The upper air observations are discussed in detail in Chapter VIII. 42 TEMPERATURE. obt Vc. Xcr v» \c/> \^ \ V \) / -i—^ — \ ^ 1 -50°c -45°c -40c -35c -30c -25c -20c -15c Temperature Fig. 1.5. Effect of removing inversion Layer. weather as found by the balloon observations. The inversion up to 1,000 metres is shown, and from there upwards the temperature falls 4°C. per 1,000 metres. If now a wind springs up the lower cold layers are swept away, and if it disturbs the air up to 2,000 metres a temperature gradient of about 65°C. is set up in this region. The temperature of the air above 2,000 metres will not be materially affected but from this point downwards the temperature rises 6'5°C. for each 1,000 metres. The new temperature conditions are shown by the line (a) in the diagram. Before the wind sprang up the ground temperature was — 35°C. and by the time the wind has produced the gradient the temperature has risen to — 22°C., i.e., the wind has produced a rise of 13°C. = 23°F. If the Avind had disturbed the air to 4 kilometres the temperature would be indicated by the line {h) which represents a rise of the ground temperature of 18°C. = 32°F. 46 TEMPERATURE. From these con.siderations wo should expect the temperature in winter during winds to be considerably higher than during the calms. The first effect of a wind under such conditions is to raise the temperature cjuite independently of the direction from which it comes. But when the wind has removed the cold surface layers the temperature it produces will depend on the direction from which it comes. Thus we should e.xpect both southerly and northerly winds to be warmer than calms, but the former to be themselves colder than the latter. This is clearly shown to be the case during the months from June to September by the mean temperatures entered in the first four lines of the follovring table : — ■ Table 18. Temperature and Wind at Cape Evans. I 11 III Northerly winds having velocities 11 to 30 miles per hour. I\' ' VI VII VIII Mouth. Calm, wind velocities 0 to 5 miles per hour. Southerly Winds HAVING Velocities Difference III— II. Difference IV— II. Difference IV— III. U to 30 miles per hour. Greater than 30 miles per hour. °F. -F. °F. °F. °F. F. "F. . June -20-6 -in-3 - S-2 - 3 7 + 10-3 fl2-4 +2-1 July —24-7 - 8-9 -12-4 - 4-3 + 15-8 + 12-3 -3-5 .August -2oS - 5-0 - 9 1 — 1-4 +20-8 + l(i-7 -41 September -22 4 — 10-0 — 120 - 4-(; J-12-4 -lOi — 2-(t October - .'5-7 - 2-3 — 40 — 0-2 4- 3-4 + 1-7 -1-V November + 14-7 4-17-0 + 10-4 4- S-7 + 2-3 - 4-3 -6-6 December + 21-S +23-4 +21-8 +21-8 + 1-6 0-0 -1-6 .January . . j +21-9 +24-4 + 19-7 +2.3-1 i + 2-5 — 2-2 -4-7 February + 20-3 +23-4 + 15-6 + fi-e 1 + 3-1 - 4-7 -7-8 March . + fi-.-) + 7-5 + 5-1 + 3-4 1 + 1-2 - 1-2 -2-4 April - 3-9 _ O.'t — 3-8 — 2-6 i + 1-7 J- 0-1 -1-6 May -12-n — 9'3 — 9-fi — 34 + 2-7 ! i_ 2-4 -0-3 During these four months the temperature during calms (column II) was considerably lower than during either northerly or southerly winds (columns III, IV and V). The differ- ence in temperature between northerly wind.s having velocities between 11 and oO miles per hour and calms is shown in column VI and for similar southerly winds in column VII. It will be seen that both northerly and southerly winds raise the temperature by over 10°F. It was stated above that the more violent winds as they disturb the atmosphere to a greater height should be warmer than the less violent winds [curves (6) and («) in figure 15]. This is clearly seen to be the case by comparing columns IV and V, which give data for winds from the south having velocities of 11 to 30 and greater than 30 miles an hour. The higher winds, although from the same direction, are nearly 8°F. warmer than the less violent winds. The same was found to hold for northerly winds, but they occurred too TEMPERATURE AND WIND. 47 .seldom to show the eii'eot so clearly, and have therefore not been includetl in the table. In column VIII the differences in temperature between northerly and southerly winds of velocity between ] I and 30 miles an hour aie shown. With the exception of June the southerly winds were the colder by two to four degrees. The June observations were taken in 1911 and 1912; in the latter year the southerly winds obeyed the rule and were 4-6°F. colder than the northerly winds ; it is therefore probable that the June discrepancy would disappear if more observations were available. It is to be noted that although in the wdnter the Barrier is probably on the average nearly 30°F. colder than the Ross .Sea, the southerly winds which blow from the Barrier are not quite 5°F. colder than the northerly winds which blow from the Ross Sea. This is entirely due to the fact that the temperature of the winds is governed by the temperatm-e of the upper atmosphere, which has bean sliown to be only slightly colder over the Barrier than over the Ross Sea during the winter (see page 44). This result confirms, therefore, the assumptions used in determining the probable temperature of the upper atmosphere. The layer of cold air near the ground, on the removal of which depends the sudden rise of temperature when a wind commences, cannot come into existence if suiBcient solar radiation is received to keep the temperature of the ground above that of the air in contact with it. This appears to take place in October and from then on to the corresponding month in the autumn, Febniary, the solar radiation is sufficient to cause convexion currents which maintain a sensible temperature gradient, and prevent the formation of a cold ground layer. In the absence of solar radiation in March, April, and May the same effect is pro- duced by the relatively warm water of McMurdo Sound, for it is not until the end of May that the ice is sufficiently thick to prevent an appreciable warming of the air by the under- lying water. From October to May, then, the cold layer cannot form over McMurdo Sound and the figures in columns VI and VII for these months show only a small difl"erence of temperature between calm weather and windy weather. It will be noticed, however, that the northerly winds are two or three degrees warmer than calms, while the southerly winds are colder than calms except in the extreme months of the period. This is exactly what would be expected from geographical considerations when there is no cold surface layer to be removed ; for southerly winds are cold and northerly winds are warm. It will also be noticed that with the exception of December the difference in tem- perature between noi-therly and southerly winds (column VIII) is greater in the summer months, November to February, than in the winter. May to August, although the actual tempera- tures of the Ross Sea and Barrier have the reverse relationship. This again confirins our conclusion tliat the temperature of the upper air determines the temperature of the winds, for we have already shown that the difference in temperature in the upper air over the Ross Sea and Barrier is greater in summer than in winter. The small difference of only 1'6°F. in December is a direct consequence of the uniformity of temperature found over the whole Ross Sea area in this month. Examples of the Effect of the Wind on Temjierature. — In the foregoing wo have discussed the influence of the wind on the temperature by talcing the observations in the bulk, and have found that the average conditions are explained by taking into consideration the vertical as well as the horizontal distribution of temperature. We will now examine a few typical examples which will show the process in action in individual cases. For this purpose figures 16 to 19 have been prepared from the records of the thermograph and the Dines' anemo- meter. The thin continuous curve shows the temperature according to the scale on the left of the diagram, and the lower ragged curve the wind velocity according to the scale on the right of the diagram. The wind direction is entered along the bottom of the diagrams. 48 TEMPERATURE. Typical Case of a Soutlierly Blizzard on September 16 and 17, 1911 (figure 16). — The forenoon of September 15 was cold after a clear, calm night, and therefore we may expect ♦Id" /!Oljrs 1 8 15lH. /tf iV ■f 8 16lH. 16 20 4 a 17th. September 1 Tl-id =9.2(5 -C A-UM j^ ..£.-.Jt...E.S.E- -HtNift..- CALM-.1= ESE* Fia. 16. Wind ami Temperature. that a layer of cold air had formed near the ground with an appreciable temperature iuver- .sion above. The temperature of thi.s layer is shown by the temperature curve to have been about — 3.5°F. At 8 km. the sky commenced to cloud over, at first with cirrus clouds but later with heavy low clouds. When such conditions precede a blizzard it indicates that the upper air has commenced to move from the south, but is moving over the lower layers without disturbing them. The condensation of water vapour indicated by the cloud has warmed up still further the upper air, and as radiation to the clear sky has been cut off by the clouds, the temperature of the cold air near the ground commenced to rise. This rise is shown in the temperature curve to have taken place steadily until midnight of the 15th. By this time the lower layers had become disturbed and little puffs of wind from the south commenced to be shown on the wind trace, and the temperature curve became irregular as the cold layer was invaded by warmer air from above. Between 8 hours and 18 hours on the 16th the surface layer became more and more disturbed as shown both by greater temperature irregularities and the increasing frequency of the fitful light wind gusts ; it is obvious that the moving layer was now coming lower in the atmosphere. At 18 hours the blizzard commenced, the wind suddenly appeared at the ground and swept away the remainder of the surface layer, as is .shown by the sudden rise in wind velocity and the almost instantaneous rise of temperature of l.j degrees. The wind continued to increase in velocity until after midnight, and it was accompanied by an increase of temperature. The wind then decreased and disappeared completely at mid- dav on the 17th. It will be noticed that as soon as the wind had reached its maximum the temperature commenced to fall — this is a mo,st characteristic feature of southerly blizzards. In practically all cases the temperature decreased towards the end of the blizzard. TEMPERATURE AND WIND. 49 In this case we have a cold surface layei of — 35°F., being replaced by a warm current of air of +5°F., a change of 40''F. Thus, if the conditions previous to the blizzard had been something like those shown in figm-e 15, the observed change would have necessitated the thorough mixing of the lower 5,000 metres of the atmosphere. The cirrus clouds with which the disturbance commenced were well above this height, so that the explanation given is well able to account for the observed effect. Southerly Blizzard preceded by a Northerly Wind — May 15, 16, and 17 {figure 17). — After midnight on the 14th the sky was clear and the temperature was unusually low, showing ^i/rs ■c s ISth. /e ?o •C S 16™. 16 ^o -i(J -2tf 4 8 -CALM- 7m. May 191! Fig. 17. Wind and Temperature. the formation of a cold layer in spite of the thin ice. .Tust before 8 hours on the 15th, when the temperature was — 25°F., a northerly wind sprang up which removed this layer and caused the temperature to rise by 10 degrees. An interesting feature of this northerly wind was that above it there were alto-cumxilus clouds moving from the south. The northerly wind continued until about midday on the 16th, when, after a calm hour, a southerh' blizzard sprang up accompanied by much cloud and drift. During this blizzard the temperature rose slowly until near the end, when it started as usual to decrease before the blizzard came to an end. Three things are interesting in this case^(l) the sudden rise of temperature M'hen the northerly wind swept away the cold surface layer ; (2) there was no sudden rise of tem- perature when the southerly wind commenced, for the layer had already been cleared away ; (3) the fact that the temperature rose much higher with the southerly than the northerly wind. The explanation of the latter fact is clear and supports the explanation already given of the cause of temperature changes under the influence of the wind. It has been stated that over the northerly wind there was a wind from the south ; thus the air must have been moving in sharply marked layers and therefore there could have been little or no vertical mixing, hence as soon as the northerly wind had removed the cold surface layer it could produce no further increase in temperature. On the other hand, the southerly wind must have extended much higher and the production of snow shows that there was considerable ascensional motion, thus a normal temperature gradient must have been established to a considerable height, raising the ground temperature to nearly 0°F. This explanation is further supported by the fact that after the blizzard stopped the wind returned to the noi-th and the temperature of the northerly wind was then not what it had been previously, but what it was during the southerly blow. 50 TEMPERATUEE. Removal of the Cold Surface Layers hy a Ligltl Local Wind—Septemher 22 to 25 {fujvre 18). —The thermograph trace shows many interesting examples of a sudden pufE of wind causing 12 ^15 2ZnD.2Ci 25 RD. i 24th. lb Z5ffl.SEPTEMBER 1911 Fig. 18. Wind and Temperature. an instantaneous rise of temperatui-e of many degrees. These sudden rises prove conclusively that there must be warm air very near to the cold air which is so suddenly removed. There is no reason to believe that there was warm air anywhere near in a horizontal direc- tion, hence the warm air could only have been part of a warm layer lying above the cold surface layer. The period September 22 to 25 contains several good examples, and the thermograph trace for this period is reproduced here. A detailed discussion is unnecessary, the relationship between the wind and the temperature is obvious. Cold Blizzard during the Summer— February 14, 15, and 16, 1912 (figure 19).— In this diagram also the relationship between the wind and the temperature is obvious. But we have Fig. 19. Wind and Temperature. a different effect from those previously considered. Here we see that the direction of the wind plays the chief part in deciding the temperature. The two short periods of noitheriy TEMPERATURE AND WIND. 51 wind brought with tbem comparatively high temperatures, while, when the southerly wind set in, the temperature dropped rapidly and remained low so long as the wind continued. The examples given above have had to be restricted in number, and therefore only typical cases have been chosen from the large number available. It goes \vithout saying that the direct relationship between wind and temperature was not always so clearly shown as in the cases illustrated ; but the relationship is more or less discernible throughout. The raising of the temperature by all winds in the winter and the dependence of the temperature on the wind direction in the summer are the two outstanding features of the relationship between wind and temperature at Cape Evans. The large rise in temperature caused by southerly blizzards during the winter was first observed on the Discovery Expedition. The true cause was not realised, and the explanations then given have led to two very erroneous ideas becoming firmly fixed in meteorological literature. They are both included in the following paragraph taken from page 424 of the discussion of the meteorological results of the Discovery Expedition. ' The rise of temperature in these gales from south-east and east-south-east, doubtless, is in some measure due to the fohn effect produced dynamically by the descent of air from the mountain tops, and also to the check given, by wind in motion, to cooling by radiation. But may it not, to some extent at least, be attributed to the conveyance of warmer air drawn from lower latitudes, circulating about a low pressure system and conveying to the locality the higher temperatur-es of the region whence it originated ? ' These explanations have been generally accepted and quoted in nearly every paper dealing with the meteorology of the Ross Sea area ; it is important, therefore, that they shordd be criticised in the light of our fuller knowledge. Fohn. — The fohn effect has been introduced to account for the high temperature of the winds. Air is supposed to start from the plateau surrounding the Barrier and to be dynami- cally warmed by its forced descent to sea-level. If such forced descent does take place, for which however there is no evidence, the air might be warmed, but at the same time there would be a corresponding fall in the relative humidity of the air. Now one of the chief characteristics of an Antarctic blizzard is the heavily overcast sky and large precipitation, which would be qiute impossible with descending air. Even if a satisfactory explanation had not been already given this consideration alone would be sufficient to .show that the high temperatures in blizzards are not due to a fohn effect. Air Circulation about a Low Pressure System.- — It has been usually assumed that blizzards are due to the passage eastwards of low pressure systems having their centres over the southern ocean. It will be shown later that such cyclones do not affect the Ross Sea area and that blizzards are due to entirely different causes. Even if this were not the case it is easy to show that the high temperatures in blizzards are not due to ' the conveyance of warmer air drawn from lower latitudes, circulating about a low pressure system and con- veying to the locality the higher temperature of the region whence it originated.' In such a circulation air which reached McMurdo Sound as a south-easterly wind, would have previously passed over Framheim as a northerly wind when it was nearer its place of origin. It would therefore be warmer at Framheim than when it reached McMurdo Sound after passing over the cold Barrier. To test this hypothesis the temperature at Cape Evans and Framheim during blizzards at the former station have been compared, with the result shown in the followinji; table : — 52 TEMPERATURE. Table 19. Difference in Temperature Frnmheim-Cape Evans during Blizzards at Cape Evans. Muntli. WIND AT CAl'E EVANS. Month. WIND AT CAPE EVANS. 1 1 TO 30 MILES PER HOUR FRO.M SOUTH. GREATER THAN 30 MILES PER HOUR FROM SOUTH. 1 1 TO 30 MILES PER HOUR FROM SOUTH. GREATER THAN 30 MILES PER HOUR PROM SOUTH. Difference of temperature. Difference of temperature. Difference of temperature. Difference of temperature. °F. -4-4 -0-3 —5-6 +.5-2 — S-f. ,1911. Apri 1 Jlay June July August °F. -14 0 -•23-7 -ir)-4 -14-6;' -421 "F. - (il - 9-4 - 6-7 - 8-9 -31 S 1911. September October . November December 1912. .January . -28 6 - 7-1 - 7-8 - 0-9 - 4-3 Except wth high wnds in December 1911 the temperature was lower at Framheim than at Cape Evans dming blizzards, hence the air in the blizzards at Cape Evans cannot have passed over Framheim as warm northerly winds. Because of the relatively high temperature of blizzards during the winter the explanations discussed above have been formulated, but it was forgotten that if the explanations were true for the winter they would be true for the summer also. Thus we should expect high temperatures in summer blizzards as well as in winter blizzards. But we have already shown that this is not the case' for summer blizzards are cold. We have now shown that the fohn effect and the circulation of air around low pressure systems do not explain the relationship found between the temperature and wind ; on the other hand an explanation has been given depending on the vertical distribution of tempera- ture which explains all the observed conditions ; it is therefore reasonable to conclude that the latter is the true and sufficient explanation. Daily Variation of Temperature. In this and the following section use will be made of three terms which have often been used by writers as being almost interchangeable, but to which a special significance will be attached throughout this work, therefore it is desirable that they should be carefully defined : — (a) The expression daily variation of temperature %vill be used to express the daily march of the temperature when all temperature changes other than those due to true daily periodic causes have been eliminated as far as possible. The irregular and non-peiiodic changes due to \nnd, cloud, etc., are eliminated by combining a large number of days having a similar periodic temperature variation ; for then the irregular changes more or less completely cancel one another while the true daily variation remains. In addition a correction has to be applied to remove the effect of the yearly variation of temperature. Duiing certain seasons of the year this may be large, for example during September 1911 each midnight was on the average •9°F. warmer than the previous midnight owing DAILY VARIATION OF TEMPERATURE. 53 to the large change of temperature which occurred between the beginning and end of the month. The correction applied for this purpose is the usual one + _,!_ (9- — f^^) (/i — 12) in which r^ and r.,^ are the mean readings answering to the first and second midnights of the day and n the hour interval since the first midnight. (6) When the true daily variation of temperature has been obtained in this way, the difference between the highest and lowest hourly values gives the daily amplitude of temperature, (c) The expression daily range of teviperature is strictly confined to the dift"erence between the maximum and minimum temperatures recorded during each day. The average value of the. daily range is much larger than the corresponding amplitude. In the Antarctic the unperiodic temperature changes are so large in comparison with those due to true period causes that a long series of years would be necessary to eliminate them completely. By combining the results of the observations at Cape Evans and Hut Point we have four years' observations for all the months from February to August and three years' observations for the remaining months. In the following discussions the results of all the years combined have been used and only where it is necessary in order to investigate particular points will the results for individual years be separately considered. In the volume of tables, however, the data for each year are given in such a way that they are easily available for further investigation. The daily variation for each month is shown in table 20 and the results are plotted in the left half of figure 20. As the data for Hut Point have been given only for two- hourly intervals, it was only possible to combine alternate values of the Cape Evans' tempera- tures with them. Hence two-hourlv intervals are used in the curves and tables . Table 20. Daily Variation of Temperature a.s determined from all the Observations of Captain Scott's first ami second Expeditions. Local time. Janu- ary. i9(»3 1904 1912 Febru- ary 1902 1903 1911 1912 March. 1902 1903 1911 1912 .\pril 1902 1903 1911 1912 May. 1902 1903 1911 1912 June. 1902 1903 19U 1912 July. 1902 1903 1911 1912 ^"g">*t- teraber. 1902 1902 1903' 1903 1911 ' 1911 1912 \ Octo- ber. 1902 1903 1911 Nov- ember. 1902 1903 1911 Dec- ember. 1902 1903 1911 Year. 0 2 4 0 8 10 12 14 16 18 20 22 -1-2 -2-47 —2-82 -1-33 -048 -fO-58 + 1 -26 + 1-78 +2-17 -fl-78 +0-90 -fn-28 -0-89 -1-68 -1-54 -0-91 -0-18 +0-34 +0-95 + 1-21 4- 1-27 +.0 89 +0-63 -Oil +012 —033 -0-79 —0-75 -021 +0-48 +0-64 +0-62 +0-36 +0-03 -015 —003 +0-21 +006 -1-0-52 +0-2o —006 +009 -0-.34 —0 06 —Oil -0-.33 -0-31 4-0-08 -013 —0-60 —0-40 -0-50 -0-28 +0-26 -fO-64 + 0-70 4-0-69 + 0-13 -0-34 -0-43 -0 46 -073 -0-23 +0 07 -014 —0-08 + 0-24 +0-02 4-0-20 + 0-61 +0-30 -0-10 -0-25 +0 03 +0 80 -003 -0-07 +0-10 -0-31 +01.i -004 -0-42 -014 4-0-08 +0 32 +0-08 +0-59 +0.53 +0-16 -031 -0-70 -0-39 -0-40 +0-13 +002 —0-15 -0-44 -0-18 +019 -0-71 —0-89 -0-08 +0-43 +0-87 +0-75 +0-49 -0-23 +0-14 -1-15 —0-82 — 1-18 -1-41 -0-58 + 1-13 + 1-33 + 1-50 + 1-35 +0-52 +0-05 -O-S"? -6-2 -1-79 -1-26 — 1-61 -1-22 -0-35 +068 + 1-39 +2-12 + 1-66 + 1-08 +0-14 —100 + 131 -0-63 -1-57 -1-30 -0-98 -0-72 +0-46 +0-95 + 1-51 + 1-47 + 1-09 +0-37 -063 -0-57 -0-79 -0-56 -0-58 -0-32 +0-37 +0-54 +0-84 +0-78 +0-50 +0-10 —0-23 Mean +23-2 + 14-6 +4-2 -8-3 -11-9 -131 -14-1 -14-3 1-3 -15-5 +23-8 -0-4 Amplitude 50 31 15 0-9 1-3 1-3 1-2 1-8 2-9 3-9 3-2 16 Time of maximum 16 16 12 4 i 1 15 18 4 4 14 14 14 14 14 Time of minimum 4 2 1 4 12 2 2 18 12 8 0 4 2 0 54 TEMPERATURE. From figure 20 it wll be seen that the observations are not sufficient to give smooth curves, especially in the winter months, when the daily amplitude is small and non-periodic changes large. Fig. 20. Daily variation of temperature. McMurdo Sound and I'ram Drift. The sun returns after the winter on August 21, so that September is the first month of the year in which the sun is above the horizon every day. The days lengthen at a rapid rate, and on October 26 the sun does not set again until February 17, after which the days shorten rapidly until on April 24 the sun rises for the last time. September to March — Period with direct Solar Radiation. From September to March the daily variation of temperature is well marked and ia obvioiusly the result of the daily variations in the intensity of solar radiation. DAILY VARIATION— SEPTEMBER TO MARCH. 55 The amplitude ol the daily oscillation increases steadily from iScptembcr to January and decreases frLim January to Marcii, with tiie single exception of the month of December in which it is somewhat smaller than in November and January. As December showed this peculiarity each year it is probably a real effect. Before discussing the physical meaning of the curves for this period in McMurdo Sound, it is desirable that we should compare the daily variation of temperature over the Barrier, and also at other places of high latitudes. Daily Variation of Temperature on the Barrier. — The method of obtaining the daily variation from temperature observations made during sledging journeys on the Barrier has been described on page 21. By combining the observations made on the Barrier by different parties and dividing the observations into months irrespective of the latitude in which they were made, the follow- ing results have been obtained, from which it will be seen that the temperature amplitude on the Barrier is much larger than in McMurdo Sound (table 20) : — Table 21. Daily Variation af Temperature on the Barrier. Party. Date. Latitude. No .of Mean tempera- ture. °F. Vakiation. .\nii 11- tude. days. 4 8 1 12 16 20 24 a Motor . Polar . Nov. 1 to Nov. 30. Nov. 7 to Nov. 30. 78° to SQi" 78° to 82i° 25' 24. ■49 +2-4 —7-2 +0-8 +5^9 +5-9 +0-5 -6-2 15° S g e Motor. Day's Dep6t. Dec. 1 to Dec. IS . Dec. 27 to Dec. 31 80" to 78° 78° to 78° 18 ■23 5) + 18-4 -GO -0-5 +3-5 + 4-1 +2-0 -3-3 1° Day's Dep6t. Jan. 1 to .Ian. 21 . 78° to 79J° 21 + 14-0 -6-3 +0-7 + 5-5 + 5-4 +01 -5 6 13° a Polar . F'eb. 19 to Mar. 4 . 83P to 81° 15 -19-5 -57 + 1-7 + 6-4 + 5-1 -01 -73 15' All . Nov., Dec, .Jan. + 11-6 -65 +0-3 +50 +51 + 09 -50 11J° The variations for each month have been plotted in figure 21, but the soale has had to be reduced considerably from that used for figure 20, therefore the mean summer daily variation for Cape Evans has been plotted on the same figure for comparison. For the three months November, December and January combined, the amplitude on the Barrier was ir5°F., while for the same months in McMurdo Sound it was only 4-0°F. A comparison of the simultaneous temperatures measured on the Barrier and at Cape Evans is very instructive and gives a real insight into the cause of the large difference in the daily 5G TEMPERATURE. amplitudes. For this purpose tlie nb.servatious made during tlic southern march of the Polar Party have been plotted in figures ■i'Ja and 22b. The lesults of this party have been chosen, because, as already stated (page 19), the observations made with sling thermometers are wtll spaced over the day, and as mini- mum thermometers were not used there can be no question as to the reality of the low' night temperatures. The temperature at Cape Evans is shown in figure 22 by a thick line, and that of th? BaiTiar by a thin line. On the latter curve the observations of the two observers (see page 20) are indicated by the use of dots and circles. As the observers were often separated by ten or more miles, and as the temperature changes were so rapid it is not surprising that on some occasions the observations vary by a few degrees, but whenever the temperature became fairly constant it will be found that the observa- tions agree very well. Between the 17th and 22nd November there was onl}' a slight south-westerly breeze on the Barrier and from the figure it will be seen that the daily variation of temper- ature during this period was enormous, the average amplitude being 20°F. For com- parison it may be stated that the mean daily amplitude over India is 19'2°F., while it is only in the desert areas when the sun is nearly in the zenith at midday that the amplitude goes above .39°F. Thus the Barrier with its relatively small change in daily insolation, the sun only oscillating between 10° and 35° above the horizon, has occasionally a temperature amplitude comparable with that of tropical India. Fig. 21. Daily variation of temperature. Harrier. Fig. 2'2a. Simultaneous temperature on Barrier and at Cape Evans. A comparison of the curves for the Barrier and Cape Evans reveals a most important relationship : the maximum temperatures on the Barrier are only a few degrees below the DAILY VARIATION— SEPTEMBER TO MARCH. 57 isimultamious teiuperatureB at Cape Evans, while the minimum temperatures are very umch lower. It is therefore not abnormally high temperatures, but abnormally low temperaturec which cause the large temperature amplitude. Fig. 22b. Simultaneous temperature on Barrier and at Cape Evans. Daily Tem/perature Variation at Frnmheim. — ^It has been stated that the temperatures at Framheim were similar to those of the Barrier. It would therefore have been interesting to compare the daily variations at that station with the observations made on sledging journeys. Unfortunately temperature observations are only available at Framheim for the three hours of observations, 8 a.m., 2 p.m. and 8 p.m. These observations, however, show that the amplitude at Framheim was very nearly that measured on the open Barrier. The relationship is clearly shown in the following table in which the same days have been used in each case for calculating the means :— Table 22. No. of Days used. Change FROM S A.M TO 2 P.M- Change from 2 p.m. to 8 p.m. Month. Barrier. Framlieim. Cape Evans Barrier. Framheim. Cape Evans =F. op F. F. F. F. November i December 30 23 +5-8 +50 + 4-8 H-3-4 +2-0 +2-6 -6-1 —2-5 -3-7 -11 -1-8 -1-7 January 21 + 5-3 +5-0 +2-3 -.5-9 -3-6 -0-8 Nov., Dec, Jan. +5'4 +4-4 +2-3 -4-8 -3-0 -1-2 i The large daily amplitude on the Barrier is therefore supported by the observations made at Framheim, and it is the largest amplitude yet measured in any polar region. That the daily amplitude of temperature on the Barrier should be between two and three times greater than in McMurdo Somid points to some essential difference in the thermal conditions of the two places. The effect of solar radiation will be considered later, but it may be stated here that the variations of solar energy during the day were less on the Barrier than at Cape Evans, and therefore they are unable to account for the large difference in temperature variations. Under the same conditions of solar radiation the temperature of the air vnW depend entirely on the nature of the ground upon which the radiation falls, for the radiation does not warm the air directly, but only by first warming up the ground on which it falls. Now the Barrier is composed entirely of snow, which is never closely 8 58 TEMPERATURE. packed, the low temperature and entire absence of melting preventing the snow crystals combining into anything approaching solid ice. Unfortunately no actual measurements of the density of the Barrier snow are available, but it is almo.st certain that it was only -2 or -3 of that of ice. In McMurdo Sound on the other hand the surface is sea ice, which at the end of the winter has a very thin coating of snow. On account of the high wind which blows through the Sound there are then large patches of ice with no snow at all and the snow which does exist is large grained and very compact. The thermal conductivity of ice is very much greater than that of snow, and the con- ductivity of snow is greater the greater its density. Let us now consider the temperature of the surface of the Barrier during a bright sunny day and the following night. During the day time the sun shines on the snow and al- though a large proportion is reflected, some is absorbed. As the conductivity of snow is very small this heat is not conducted downwards to any considerable depth, but is all retained near the surface ; also on account of its small density the thermal capacity of snow is small. Thus a very little heat, being retained near the suiface and acting on a substance of small heat capacity, raises the temperature of the snow near the surface appreciably. During the night the snow surface radiates heat to the clear sky and its temperature falls, but those conditions which caused its temperature to rise while the sun was shining now act in the opposite direction. On account of the small heat capacity of the snow a small loss of heat by radiation lowers the temperature con.siderably, and as the low conductivity pie- vents heat being conducted from below, the temperature of the surface may f^ll very greatly during a clear night. Let us now consider that the Barrier is changed into ice with just sufficient snow on the surface to give it the same absorbing power. The amount of heat absorbed \vi)l be the same as before, but owing to the much greater conductivity of ice — at least a hundred times as great as that of snow — changes of temperature at the surface are accompanied by a greater transference of heat into the mass of the ice. This keeps the surface temperature lower during the day, but during the night when the air temperature falls this same amount of heat can agam reach the surface, and if the radiation is particularly great still more heat which was stored up in the ice will be conducted outwards and so prevent the temperature of the surface falling so low as it would have done had the mass been snow through which heat can only pass slowly to the surface. If now instead of the Barrier we consider a sheet of ice floating on the sea, the temperature of which cannot fall much below freezing point, we have a still greater protection against an excessive fall of temperature during the night ; for the ice with its high conductivity is always conducting beat from the warm sea below and is therefore always a little warmer than the mean temperature of the air above it. Thus as soon as the surface temperature falls owing to radiation, the heat from the sea stored up in the ice becomes available to counteract any large fall of temperature of the surface due to radiation. Thus we see that the chief difference between a snow and an ice surface is that the latter prevents low radiation temperatures, while excessive radiation always produces low temperatures over the snow. We have now the explanation of the different daily variations of temperature over the Barrier and over McMurdo Sound. Very low night temperatures are possible over the former owing to the bad conductivity of the snow, while over the latter the high conductivity of the ice and the store of heat retained in it from the warm sea below prevent low tem- peratures during the night. It will be remembered that the observations showed that the day temperatures of the Barrier were nearly the same as those at Cape Evans, while the night temperatures on the former were many degrees lower than those at the latter. DAILY VARIATION— SEPTEMBER TO MARCH. 59 Comparison of the Daily Temperature Variation during September to March at McMurdo Sound and at other Stations in High Latitudes. So far as I am aware there is only one other set of temperature observations in high latitudes so complete as these at McMurdo Sound with which to make comparison. The observations made during the drift of the Fram in a mean latitude of between 82° and 83° N. give the daily temperature variation for each month from a homogeneous series of observations extending over three years. For comparison the curves given in Professor Mohn's masterly discussion of the meteorology of that expedition have been added to figure 20 as they show marked and important differences from those for McMurdo Sound. For purposes of comparison the northern observations have been plotted against those of the corresponding months in the south ; thus January in the north is plotted against July in the south and so on. The first striking difierence is in the time of year at which the daily variation has its greatest amplitude. In McMurdo Sound the amplitude increased slowly from the return of the sun on August 24:th until January (neglecting for the moment the decrease in December), while in the north the amplitude increased more rapidly and the greatest amplitude was reached as early as the second month after the return of the sun. From this month, April (corresponding to October in the south), the amplitude in the north decreased rapidly to a very small value at midsummer, after which there was a temporary rise during the last two months, August and September, before the sun set for the winter. The contrast in this respect shown between the McMurdo Sound and Fram data is found to be present when other stations are investigated. There appear to be two types of polar climate. In one, of which the Fram conditions are typical, the daily amplitude of tem- perature has its maximum in one of the spring months, while in the other the maximum daily amplitude is found in a summer month. The following two tables give examples of the two types : — Table 23. ' Fram ' Type. Month. Fram 82^" N. Gauss 66° S. Amplitude. Mean temperature. Amplitude. Mean temperature. S. September 1 N. March / S. October "\ N. April / S. November ") N. May / • S. December ■) N. June / S. January \ N. July / S. February ) N. August / S. March \ N. September / op- 1-6 5-9 3-0 2-4 1-2 1-9 1-4 -22-5 - 90 + 12-2 +28-8 +320 + 28-8 + 15-S °r. 6-6 9-3 7-5 6-2 56 4-3 * F + 0-1 + (-8 + 19-6 +29-7 +.30-4 +26-4 + 171 *TUo Gauss diily a.npl.tulo for Mirch is g;veu as 3-S2^C. =ti-9°F. in D.-ut. Su^pjlar Expedition, Meteorc logy, Vol. 1, part 1, page 53. This, however, is much too larg;. During the first .sixteen days of March the teinparature observations ware taken on b:)ard the Gauss, and during twelve of the remaining days in a Stevenson screen on the ice n3ar the ship. CilcuUtiug the amplitude for the 16 days on which the observations were made on the ship and for tha 12 on the ice wa obtain the following: Ship 4-71=C., Ice 2 COX'. As the ship's observations obviously give too large an amplitude and twelve days are not sufficient to give a reliable value, I have not included thi.s month in my discussion. 60 TEMPERATURE. Table 24. ' McMurdo ' Type. Montli. ilcMcKDo Sound 77^-° S. Snow Hill 64i°S. Framheim * 78r S- Ampli- tude. Mean tempera- ture. AmpU- , ^«*" Ampli- tude. Mean tempera- ture- September October . November Deceinljer January . February March . °F. IS 2-9 3-9 3-2 5'0 3! 1-r, "F. -15-5 - 6-2 -rl31 +23-8 -f23-2 + 14-6 + 4-2 °F. 'F. 4-6 + 3-9 50 4-14-9 5-6 -rl7-4 6-1 4-28-4 4-3 4-30-4 4-9 4-25-6 4-0 4-13-5 "F. e-4 5-4 86 5-0 7-2 °F. -35-7 -11-6 4- 4-1 4-20-1 4-14-9 The two example.s under the Fram type show the maximum amplitude early in the spring after which the ampUtude decreases until midsummer. In the case of the Fram there is a temporary rise during August before the amplitude begins to decrease to its \vinter value. The amplitade in the second type increases until the summer is reached, and it is interesting to notice that in each case tho warmest of the three summer months shows a slight decrease ir the amplitude as compared with the months on either side. Both Professor Mohn and Professor Meinardus have given explanations of the Fram type, both are different and both appear not to be the full explanations. A discussion of these two explanations will be fruitful iu shov^lng some important factors connected with the tempera- ture of polar regions and therefore I propose to give their explanations, point out where they fail, and then attempt to give a true explanation. Professor Malm's Explanation of the ' Fratu ' Type of Daihj Teitiperuture Variation.— M.ohi\ fii-st examines the effect of cloud and wind on the daily variation of temperature and finds that both these factors reduce the amplitude. The effect of the cloud is to reduce the variations in solar radiation and radiation from the surface, while the wind stirs up the lower atmosphere and distributes the effect of the varying temperature of the gromid through a laige mass of the atmosphere. He also calculates the 'difference in the amount of solar energy received on each square centimetre of surface at midday and midnight, on which the temperature variation must ultimately depend. We will now quote from page 605 of Mohn's discussion : — ' In the following table I hava put together the mean monthly values of (1) the diurnal amplitude f of the radiation of the sun and sky, a ; (2) the amount of cloud, c ; The data for Framheim are only approximate. At tliis station temperature observations -were taken only at o A.M., 2 P.M. and 8 p.m. As.suming as a rough approximation that the mean of the 8 a.m. and 8 p.m. observations gives the mean of the day and tliat the 2 p.m. observation is near the maximum, the approximate amplitude has been taken as twice the excess of the 2 p.m. mean above the mean of the 8 a.m. and 2 p.m. observations. tl have taken the liberty of altering Hie -n-ord 'range' u.scd by Muhn into 'amplitude' in oitler to keep tile use of tlieae two words consistent tliroughout this work. DAILY VARIATION— SEPTEMBER TO MARCH. 61 (3) the velocity of the wind in metres per second, v ; and (4) the diurnal amplitude of the temperature of the air, A, for the months March to September. Table 25. Month. a e V A Cal. 0—10 m.p.b. °C. March 0168 5-32 4-25 0-89 April 0-515 4-84 4-05 3-29 May 0-625 7-57 4-97 1-69 June 0-633 8-67 4-56 1-32 July .... 0-639 9-06 4-40 l)-89 August 0-681 8-45 4-31 1-07 September ..... 0-273 9-10 4-74 ' 0-78 ' The radiation varies regularly from month to month, and is nearly constant from May to August. The amount of cloud varies considei ably, and is highest from May to September. The velocity of tht; wind does not vary much. The diurnal amplitude of the temperature shows the largest variation during the months from March to September. From March to April the amplitude of radiation increases rapidly, the amount of cloud decreases to a minimum, as does also the velocity of the wind ; and all three factors are working to raise the diurnal amplitude of the temperature. From April to May the amplitude of radiation rises a little, but the amount of cloud is rapidly increasing, and also the velocity of the wind. The two last-named factors bring the amplitude of the temperature down notwithstanding the eiTecfc of the radiation. From May to June the amplitude of radiation increases very little, but the amount of cloud increases one degree, and the temperature amplitude is lowered by the cloudi- ness in spite of the increasing radiation and decreasing wind velocity. From June to July the same process goes on. From July to August the amplitude of radiation rises (the mean latitude of the Fram being relatively lowered), and the amount of cloud and the velocity of the wind decreases. The temperature amplitude rises, all three factors working together. From August to September the amplitude of radiation is going down rapidly, the amount of c'.oud and the velocity of the wind are increasing, and all three factors cause the amplitude of the temperature to decrease. . We see that the maxima of the diurnal amplitude of the temperature in both April and August, correspond with a higher amplitude of the radiation, and minima of cloud and wind velocity. The low amplitude of the temperatra-e in summer, particularly in July, corresponds to a high degree of cloudiness during a season which has hardly one clear day.' There can be little doubt that Professor Mohn's explanation contains part of the truth ; but if the observations are examined in another way it will be seen that it is not the whole. Professor Mohn attributes most of the irregularities in the temperature amplitude from month to month to the changes in the cloud and wind. That this cannot be so will be seen from the table given on page 604 of Mohn's work, part of which is reproduced here. 62 TEMPERATURE. Table 26. Clodd. Wind. Month. Cloud. 0-0 A. Cloud A. Wmd. A. .Vind A. April .... 0-515 3-5G 10 2-63 30 3-52 6-0 2-16 May ... 0-625 00 1-89 10 1-44 3-0 2-28 6-3 1-72 Juns .... OGU 4-5 2-36 10 1-08 3-2 1-39 6-5 1 03 July .... 0-639 4-4 0-77 10 0-74 3-6 0-69 5-9 0-64 This table has been computed by taking the mean daily temperature amplitude on days with 0 cloud and 10 cloud (in June and July there were no clear days, so days less than half overcast were used), and on days with mean vnnd velocitj'' greater and less than 4'5 miles per second. Tha values entered in the columns headed Cloud and Wind are the mean values of these factors during the days consideied. Now it will be noticed that with one exception in spite of the increase of the amplitude of the solar radiation from April to July the daily amplitude of temperature decreases in the same period in every type of weather. This gives good reason to believe that if there had been no change in the type of weather the temperature amplitude would still have decreased. Therefore we cannot ascribe the decrease in temperature amplitude to changes in wind and cloud. Further tbe fact that the temperature amplitude was greater during April and May with completely overcast skies than it was during June and July with sides less than half overcast, gives further support to the conclusion that cloud was not the chief cause of the decrease in the amplitude. On the other hand as the wind decreased every month from May to August (see table 25), wind cannot have been the cause. Professor Meinardus's Explanation of the ' Fram' Type of Daily Temperal.iire Variation. — In his discussion of the meteorological results of the Gauss Expedition, Meinardus says : — ' The amplitude of the daily period of temperature is greatest during the spring months and in March ;* in the summer it is considerably less, probably because the midday temperature during this time is approximately that of the freezing point which cannot be passed on account of the heat during the process of melting. Also during the summer the effect of radiation on the air temperature is reduced owing to the shortening of the night and lengthening of the period of sunshine. Similar changes of the temperatm-e amplitude were found during the drift of the Belgica and the drift of the Fram. Also there the regular variations were greatest in the spring. On the contrary the greatest values occurred in January at the Discovery station where the freezing point was seldom reached.' Two probable explanations are given in this paragraph, neither of which was considered by Mohn : [a) a curtailing of the amplitude in the warmer months due to the temperature of the surface being unable to rise above the freezing point, and (6) a decrease in the amplitude due to the shortening of the night and the lengthening of the period of sunshine. The latter of these two is obviously based on a misconception ; the daily amplitude of temperature does not depend on the amount of energy received from the sun, but on the difierence between the amount of energy received at midday and at midnight. In any one latitude the difference of energy received on a square centimetre of ground at midday and midnight increases right up to the summer solstice. There is therefore no decrease in the * See (ootuote to page !iu. DAILY VARIATION— SEPTEMBER TO MARCH. 63 amplitude of solar energy between the spring and summer to account for the decrease of temperature amplitude. As the temperature of an ice surface cannot rise above the freezing point, it is obvious that the temperature amplitude will be decreased as soon as the mean temperature closely approaches that point. This is in all probability the reason why the warmest months of the McMurdo type of daily temperature variation have a smaller amplitude than those on either side. Even when the mean temperature of the month is low, as at Framheim and McMurdo Sound, there are a few days in the month when the maximum temperature might have gone higher if the melting of the ice had not prevented the complete warming up of the surface. It must be remembered that the temperature ampUtude of the surface itself is many degrees greater than that of the air, and therefore the surface might reach and strive to pass the freezing point on many days with air temperature well below that point. But this explanation does not account for the rapid decrease in amplitude which takes place in the Fram type a long time before the mean temperature gets near to the freezing point. Thus in the case of the Fram the amplitude falls from 5-9°F. in April to half its value in the following month, yet in the latter month the mean temperature was practically twenty degrees below the freezing point. The same was the case, but to a lesser extent, with the Gauss. In November the maximum air temperature only once reached the freezing point and throughout the month the mean maximum temperatiue was 7'5°F. below the freezing point, yet in spite of this tha temperaturb amplitude decreased from 9-3°F. in October to 7-5°F. in November. Suggested Explanation of the Variation in the Amplitude During the Period September to March. — There can be no doubt that the prime factor in determining the daily temperature ampli- tude is the solar energy amplitude. The first step in the investigation should, therefore, be to plot the observed temperature amplitude3 against the ca'culated energy amp'itudes to see if there is any close relationship. This has been done for two stations of the Fram type and two of the McMurdo type, namely, (a) Fram and Gauss : (6) McMurdo Sound and Snow Hill, and the observations made on the Barrier have been included for reference. Mr. Normand of the India Meteorological Department, has very kindly made the calculations of the solar energy amplitude by the method described by Mohn on page 588 of his discussion of the meteorological results of the Fram Expedition. This method takes into account both the energy received directly from the sun and also that diffused from the illuminated atmosphere. The following table contains the results- of the radiation calculations :— Table 27. Solar Radiation (gram. cols, per min. per sq. cm. of horizontal surface). Month. Snow riiLL tU"21' S. 6..i° OJ' S • Q. MCiViUKDO OOUNU 77= 3o' S. 8l°4rN. 83" 12'. Sjuth. North. 5 t Q 4 a i -a 60 6 a a; i-i a i +3 ji '5 •a 6 C SeiJiember March . 0-J_' U-iJo 0-92 0t>6 0-00 0-S6 0 30 11-00 0-30 0 ly OOj 0 17 October . April . 1-48 0-00 1-48 1-40 0 00 1-40 0-^-i 0 05 0-78 064 012 0-52 November May . 1-88 0-02 1-86 1-80 003 1-77 131 019 112 1 03 0-41 0 6- December June . 2 04 OOo 1-99 201 0-07 1-94 1-52 0-.34 1-18 1-22 0-59 o-c.:. January . July . 1-96 0-04 1-92 I -90 n-Od 1-85 1-37 0-26 1 11 114 0-50 0-6 1 February August 1-64 0-00 1-64 1-58 001 1-57 1-02 008 0-94 0-85 0-17 0-68 March September . 11 3 0-00 M3 l-Oo 0-00 105 0-48 0-00 0-48 0-32 0-05 0-27 64 TEMPERATURE. Ill figure 2o the temperature amplitudes given in tables 23 and 24 (pages 59 and 60) Lave been plotted against the solar energy amplitudes given above, the points for the different 12 II Id ( 1 ■ N D.J r •. Fram ® . Mc Mordo Sound @. Gau&s e. Snow Hrll ■ .. Barrier < JO ,,- - 1 ^^-" --'e« f es ,- ^•''' "' oe^ ^ ^ ^«U -" ' 0^ jeBK »• ,,-- »s'" re ... ®r"^ ,--'' ^, ,,-'' f«l ^, ---' ^.-' - '' ^•0 11®- ,..., '"" ^,-' --^•5 - < )S M® A* ---" • --' __-' '-''''' D 2i^ 077 pliblde a^*& ^larJtcu D -1 -2 -J -4 -5 -6 1 -8 -9 H M !•? 1-3 M 1-5 1-6 I-? 1-8 1-9 2-0 2-1 | Fig. 23. ArapIituJe of temperature plottefl against amplitude of solar radiation. stations being indicated by various signs, and the months to which they refer by entering against each the initial letters of the months. For the sake of comparison the months for the Fram have been marked both by their own initials and by the initial letters of the correspond- ing months in the south. The first striking feature of this diagram is the relatively small variations from a linear relationship shown by most of the points. Neglecting for the moment the point for the Barrier, twenty-four of the twenty-seven points lie \vithin the two dotted lines. This leads to the conclusion that except for small differences the relationship between temperature amplitude and energy amplitude is linear. The three points which do not lie within the dotted lines are the most important for our present discussion. It will be noticed that they are the spring months of the Fram type, being the Pram's observations for April and the Gauss's observations for September and October. Both Mohn and Meinardus in their discussion of the temperature amplitude tried to explain the low values of the amplitude in the summer, we now see that it was the high amplitude in the spring which needs explanation. The abnormal temperature amplitude on the Barrier is clearly shown by the position of the point for the Barrier near the top of the diagram, and it is the position of this point which supplies the solution of the whole problem. We have already shown that the difference in temperature amplitude over the Barrier and McMurdo Sound is due to the sea ice preventing low nocturnal radiation temperatures. This action of the sea ice could not exist if there were a thick layer of loose snow upon it. Now both the Fram and Gauss were surrounded by sea ice, it is therefore important to examine the varying character of the snow covering of the ice during the year. Unfortunately I cannot find in Mohn's discussion of the Fram's observations anv remarks on the character DAILY VARIATION— SEPTEMBER TO MARCH. 65 of the snow accumulations on the sea ice during the voyage, but in Meinardus's account of the Gauss observations there are a few important remaiics. We will therefore consider the latter and then see how far similar considerations can be applied to the Fram conditions. . The Gauss was frozen in on February 22. As soon as the sea was completely frozen over snow commenced to accumulate on the surface, and the depth of the snow increased throughout the winter and first months of spring. The exact depth of the snow is not given, but we are told that the snow surface rose about 60 cm. from .June to October (page 1.57 *). With the increasing intensity of solar energy the surface however quickly underwent an important change, to quote Meinardus (page 154). ' The advent of warm weather and strengthened solar radiation produced a gradual lowering of the snow surface on account of the molting together of the snow- crystals, a process, however, which was often retarded through fresh snow accumula- tions. Still the consequence of the progressive concentration of the snow was that the old snow gauges which had been snowed up in the middle of October appeared one after the other.' We see from these remarks an important change in the character of the snow-covering to the sea ice. At the end of the winter and until the end of October there was an appre- ciable depth of snow which on account of low temperatures and the absence of melting would be in a comparatively loose condition, the surface only being compacted by the wind. Such a snow layer would have a low conductivity which together with its depth would make it an effective protection against a supply of heat from the underl3ang sea ice. The surface during September and October would therefore approach the condition of the Barrier, with a consequent large daily amplitude of temperature. After October the snow, as we are informed, became more compact and its thickness decreased. This process probably continued at first rapidly and then more slowly until mid- summer, when what little snow was left would oppose only a small impediment to the transfer of heat from the sea ice to the air. Thus we should expect at the Gauss station the daily amplitude of temperature to decrease, relatively to the amplitude of solar radiation, from October to midsummer. An examination of figure 23 will show that this occurred : in spite of increasing amplitude of radiation from October to December the temperature amplitude greatly decreased and in December the point for the Gauss station is exactly in the middle between the two dotted lines. After December the points for the Gauss station do not depart largely from the linear relationship between temperature amplitude and solar radiation amplitude. Returning now to a consideration of the Fram observations, there is no reason to suppose that the sea ice in the north would not also be covered by a blanket of dry loose snow at the end of the winter. Also as the snowfall is known to be heavier in the north than the south the depth of snow was probably greater around the Fram than around the Gauss. It is interesting to notice that the two points for March and April for the Fram lie on the line joining the point for the Barrier with the origin. This is what would be expected if the surface were similar in the two legions and the temperature amplitude depended only on the radiation amplitude. By midsummer in the north as in the south the snow-covering would have become a very poor insulator and we find a rapid decrease in the temperature amplitude. The Fram point for June is in the middle between the two dotted lines, showing therefore a normal relationship between the temperature and insolation amplitudes. The only difficulty is the low * Meinardua. Deut. Sudjjolar-Exped. Meteorologie, Band I. 66 TEMPERATUKE. position of the point for May ; for one would have expected this to be abont half way between the points for April and June, as was the case at the Gauss station. Here I think Mohn's explanation has weight. Compared with April, May was a cloudy and windy month and these together would have reduced the temperature amplitude if there had been no change of surface. Therefore the effect of the cloud and wind must be added to the change of surface, thus accounting lor the low position of May relatively to April, but it is interesting to notice that in spite of the cloud and wind the point for May is higher than the mean position of the other points, showing that the surface had not got into the normal state for sea ice. The low position of the Fra7)i point for July is in all probability due to the explanation given by Meinardus ; for the mean temperature of this month was the freezing point, and the daily amplitude of temperature could not be great above a wet snow surface. From this discussion we see that the variation of the temperature amplitude in the Fram type is mainly accounted for by changes in the surface and solar radiation, these are the chief factors, small variations may be due to other causes such as' cloud, wind or local conditions. We must now examine why the McMurdo type diilers from that of the Fram type. The explanation is that owing to the local conditions there was no large change at any of the stations of this type in the surface conditions between the spring and summer. At Framheim the surface was that of the Barrier which can undergo no change. At McMurdo Sound and Snow Hill the observations were on the coast between a snow-covered land and a frozen sea. In McMurdo Sound there was little change of the snow-covering of the land throughout the year and the sea ice was almost swept clear of snow by the blizzards during the winter. In fact there was more snow on the sea ice in January than in October, and the effect is seen by the relatively high position of the point for the former month in the diagram. Similar conditions probably held at Snow Hill which was subject to very high dry winds which must have prevented any large accumulation of snow on the sea ice. Summary and Conclusions : Temperature Variation in the Period of Solar Activity, September to March. — (a) The daily temperature amplitude during this period over sea ice nearly free from snow is a linear function of the daily amplitude of solar energy. {b) In consequence the temperature amplitude increases from the return of the sun to midsummer and then decreases, except that the warmest month has generally a lower ampli- tude due to the inability of the surface to warm up above the freezing point. This gives rise to the McMurdo type of daily temperature variation. (c) A layer of loose snow increases the temperature amplitude. (d) In consequence, in situations where there is a layer of loose snow covering the sea ice in the early spring months the temperature amplitude is relatively high. As the solar energy increases the snow-covering becomes more dense and thinner, in consequence the temperature amplitude generally decreases from spring to midsummer in spite of the increasing amplitude of solar energy. This gives rise to the Fram tyj^e of daily temperature variation. (e) In places where the surface consists of a very deep layer of snow — for example the Barrier — the temperature amplitude is very large. Daily Variation of Temperature during May, June and July. Period with No Direct Solar Radiation. During May, June and July the sun does not rise above the horizon in McMiudo Sound and the amount of indirect radiation from the sky is so small that it may be entirely neglected. DAILY VAEIATION— MAY, JUNE AND JULY. 67 In these circumstances one would not expect any consistent daily variation of the temperature, and a first glance at the curves for these months given in figure 20, page 54, seems to confirm this opinion. A closer investigation, however, shows that the curves are not entirely irregular. By taking the mean of the three months, May, June and July together, we obtain curve 1 in figure 24. This curve has been derived from observations extending over four years, it is therefore the mean curve of 368 days ; hence it should be free from accidental irrf^gularities. This curve shows two remark- able features : (a) The temperature is above the aver- age from 9 A.M. to 7 p.m. and below for the remainder of the day except for (b) a sudden and pronounced rise at 4 A.M. • The question at once arises — -are these features real ? The usual method for testing the reality of a feature which is shown in the average of a large number of separate events, is to divide the data up into different blocks and investigate each one separately. In the present case our data are taken from three months' observations. May, June and July, in each of four separate years. A conve- nient method of dividing them up is to combine the three months together for each year giving four separate blocks ; again the data for May from the four years may be combined, and similarly those for June and July giving another three separate blocks. By this method we can see whether the feature looked for is shown in each year and also whether it appears separately in each of the three months. The number of observations which falls in each sub-division is small, therefore the irregularities will not neutralise one another so well as when all the observations are included in one block. We must therefore expect the curves for each sub-division to be more irrcnilar than that for the whole data shown in curve I. But if the features are real they should show up against the irregu- larities. We are looking for two features : if (a) is present, i.e., if the temperature is above the average from 9 a.m. to 7 P.M., then in each of the divisions the sum of the depar- tures from the mean between these times should be positive ; if (b) is present and there is a sudden and pronounced rise at 4 a.m., the departure at 4 a.m. should be greater than the average departure at 2 a.m. and 6 a.m.; in other words. Fig. 24. Daily variation of temperature. McMiudo Sound. May, June and July. departure at 4 a.m. — departure at 2 a.m. + departure at 6 a.m. should be positive. Thus we have two numerical relationships to look for. The data have been divided as described above and the resulting departures for every two hours from the mean of the day calculated. The curves for the mean of each month, i.e., 68 TEMPERATURE. the means for May, June and July during the four years, have already been seen on figure 20, page 54. The means for each year taken separately are shown as curves 2 to 5 on figure 24. The results of the numerical tests for (a) and (h) arc entered in the following table, where a plus sign indicates that the feature is present and a negative sign that it is absent : — Table 28. Data arraugect according to Test for (o) Mean departure for hours 10 to is inclusive Test for (b) Excess of the departure for 4 a.m. over the mean departure for 2 a.m. and G A.M. A positive sign indicates the presence of the effect A positive sign indicates ths picsence of the effect All Days .... Curve 1, Fig. 24 +019 "F. +0-35 Years May, June, July, 1902 Curve 2, Fig. 21 „ 1903 „ a „ 24 „ „ 1911 „ 4 „ 24 „ „ 1912 „ 5 „ 24 +0-18 + 0-34 +0-42 -012 + 0-18 + 0-7.5 —0 02 +0-44 Months May of all years . Curv? 9, Fig. 20 June „ „ . „ 10 ., 20 July 11 ,. 20 + 0-98 +0-20 —0-10 +0-1? +010 +0-80 Cloud < .5 througliout the day Curve 8, Fig. 24 Cloud > 5 „ „ „ 9 „ 24 -0-09 +012 +032 — n-o.) Wind< 20 miles in every Curve 6, Fig. 24 hour during the day. Wind> 20 miles in one or ,, 7 ,. 24 more heurs diiring the day. +013 +0-28 +0-39 + 027 The Reality of the Feature: Excens of Temperature During the Hours 9 «.«(. to 7 ^j.h(. during May, June and July. — From column 2 of table 28 we see that taking all the observations together the mean excess of temperature during these hours is positive. It is also positive in each of the years 1902, 1903 and 1911, each considered separately; it is absent, however, ill 1912. When the months are taken separately the feature is seen to be present in May and June but not in July. Thus out of the seven separate sub-divisions of the data five show the effect while only two do not. This must be taken as very good evidence of the reality of the effect. We can therefore conclude that in McMurdo Sound during the three winter months when there is no direct solar radiation the day is warmer than the night. In other words in spite of the absence of a var}'ing solar radiation the variations of temperature are in the same direction, but much less in intensity, than when the sun shines. DAILY VARIATION— MAY, JUNE AND JULY. 69 It would be of great interest to know if this were a feature of the whole of the Antarc- tic in high latitudes, but unfortunately all the other stations in the Antarctic from which we have data aie either very near or just outside the polar circle, hence they received some radiation during these months. They each have warmer days than nights, but this may be due to the sun and therefore cannot be taken as proof of the presence of the feature found in McMurdo Sound. The Reality of the Feature : Marked Excess of Temperature at i a.m.— The results of the test for this feature are shown in column 3 of table 28. Four a.m. is warmer than the mean of 2 A.M. and 6 A.M. in six out of the seven sub-divisions into which the data have been divided. It was absent only in the year 1911. There is thus even better evidence for the reality of this feature than was found for the days being warmer than the nights. A glance at the curves for McMurdo Sound in figure 20, page 54, -vvill show that this excess of temperature at i a.m., which is present throughout the winter months, becomes more marked at other periods. While it is feebly marked during May and June it becomes so large in July and August that it then forms the main maximum temperature of the day. It is still well marked in September, but now the returning sun has produced a larger maxi- mum later in the day. From October to March the 4 a.m. temperature shows no maximum, but it is noteworthy that of these months October, November and March show a similar effect at 2 a.m. ; may not this be the same effect but displaced by the minimum due to the solar radiation falling very near to 4 a.m. ? After March the effect reappears and is very markbd in April. We now see that in McMurdo Sound there is an excess of tempe- rature at 4 A.M. compared with 2 a.m. and 6 a.m. in every month from April to September inclnsive, and the same effect is possibly present in October, November and March although during these months the minimum of the solar temperature amplitude is a disturb- ing factor. The same effect is clearly seen in the temperature observations at the Gauss station although data for one year only are available. The Gauss wintered just outside the Antarctic Circle and therefore direct solar insolation was never entirely absent ; yet during four months, July to October, the maximum at 4 a.m. is clearly distin- guishable. As the curves given in Plate V of Meinardus's work have been smoothed the effect we are discussing is not clearly shown. I have therefore plotted the actual data from midnight to 8 A.M., and reproduce them here in figure 25. In July and August the maximum is clearly visible as an actual peak on the curves. It is visible in September and October as an interruption in the smooth course of the curve at 4 a.m. The Snow Hill observations also show the same effect in July and August as will be seen from the following table : — - Table 29. Departures from Mean Temperature at Snow Hill. •Ifl 3 1 I 3 4 S 7 ■M •rti -10 -M -JO 1 Y- T' ^ — « s n^ ~~ ^ ^ / < s k \ \ ^ :•• S / / S, / « \ \ 0 / \ -w H 0 u r$ Fio. 25. Daily variation of temperature. Gams station. Midnight to 8 A..M. Jlouth 0 1 2 3 4 5 6 7 8 A.M July . August . — OlS -010 -034 + 016 -0-47 +015 -0-34 -022 -0'17 -0-08 -052 -0-08 -0-37 -0-20 -061 —038 -0-31 -0-25 70 TEMPERATURE. Thi Cn'ite of Warm D iijs In the Absence of Insolalion and of the Excess of Temperature at 4 a.m. — Wo, have now shown that, the former of these eft'ects is real in McMurdo Sound, which is the only station at a sutRciently high latitude to allow of the effect being visible, while the latter effect has been shown by all tests to be a real factor in McMurdo Sound and to have been present at the Gauss station and Snow Hill. It is not shown in the onlv observations that we have from Cape Adare (see page 73 below). I am not at pr-sent able to offer any explanation of either of the effects. The only glimmer of light comes from the observations made on the Fram voyage in north polar regions, which will now be discussed, but up to the present this has not led towards a physical explanation. Daili/ Temperntarc Variation in North Polar Region.'' from the Records of the 'Fram.' — During the voyage of the Fram which lasted from July 1893 to August 1896, the sun was below the horizon during 12 months. From the observations Mohn deduces the daily variation of 410 -10 (2) +to +10 .Cloud 10 All Obs. Cloud O +10 -10 Wind > 4-5 m.p.s +»0 -10 0 hours 2 ID 12 14 16 18 20 22 24 Fig. 26. Daily variation of temperature during the dark months. Frani Drift. temperature during these dark months .shown in curve 1 of figure 26. It will be .seen that it is totally different from the corresponding curve for McMurdo Sound shown as curve 1 of figure 24j page 67. DAILY VARIATION— MAY, JUNE AND JULY. 71 The temperature is above the mean of the day from 9 p.m. to 8 a.m. and below during the rest of the day. No excess of temperature is shown at 4 a.m. and the day is colder than the night ; thus neither of our effects is present. Mohn concludes that the daily variation is fully accounted for by the daily variation of the wind : ' In the dark season the north component of the wind's frequency exceeds the south component uninterruptedly from 7 a.m. to 8 P.M. and the northerly winds are colder than the southerly winds. It seems reasonable to ascribe the diurnal period of the temperature found in the dark season to the effect of the wind.' But Mohn has gone further ; he has determined the daily march of temperature for the dark jnonths (a) during cloudless days ; (b) during overca.st da3's ; (c) during days with an average wind velocity less than 4-5 metres per second ; and ((/) during days with an average wind velocity greater than 4-5 metres per second. The results for each of these for the three winter months, November, December and January, are shown as curves 2, 3, 4 and 5 of figure 26. It will at once be seen that the temperature variations on cloudless and on windless days are almost mirror images of the variations on overcast and windy days. The warm period found in the Antarctic between 9 a.m. and 7 p.m. is certainly present in the north on overca.st days, and the excess of temperature at 4 a.m. is very marked in the north on days with little wind. This leads us to investigate the effect of cloud and temperature on the Antarctic temperature variations. The variation was therefore calculated for cloudless day.-' and overcast days, but it was found that the number of days was too small to give smooth curves, therefore all days were considered in which the cloud amount was less than 5 at all ob.servations, and all days in which the cloud amount was more than 5 at all observations. These curves were still irregular but smoother than (he former, they were howevei the same in general run. They have been plotted as curves 6 and 7 of figure 24, page 67. The days of May, June and July of the four years were then divided according to wind by first taking out those days in which the hourly wind velocity did not rise to 20 miles an hour in any one hour throughout the day, these formed the first class ; the remaining days on which the wind rose above 20 miles per hour during at least one hour in the day formed the second class. The temperature variations for each class were determined and plotted as curves 8 and 9 of the same figure. Examining the two latter curves we see that they are so similar to one another and to the mean of all days, curve 1, that wind can have little direct influence in McMurdo Sound on either of the features under discussion. The influence of cloud, however, although small, is quite marked. It will be seen that curves 6 and 7 in their general run are similar to those found for the north ; days with cloudy skies are warmer during the day time than during the night, while days with clear skies are on the whole warmer during the night hours than during the day hours. But a still more important relationship is shown. The excess of temperature at 4 a.m. is clearlv marked with clear skies, 4 a.m. being the warmest hour of the day ; while there is no indication of the excess at 4 a.m. on cloudy days. It therefore appears that our first effect, the day warmer than the night, is associated with cloudy days, while the second effect, excess of temperature at 4 a.m., is associated with cloudless or slightly clouded days. Except that the ma.ximum is at 2 a.m. instead of 4 a.m. in the north the same relationship holds there. Mohn has given an explanation of the average conditions in the north, but he gives no suggestions for the remarkable difference in daily variation according to cloud and wind.* * It should be. remarked that in the north cloud is so intimately related to wmd, windy daj's being nearly always overcast, that it is impossible to separate the two effects. The curves for no cloud and little wind on figure 26 are practically from the same data; similarly with the curves for overcast skies and windy days. 72' TEMPERATURE. This is obviously the key to the whole problem, the solution of which would be an important addition to our knowledge of the physics of the atmosphere. Already I have spent more time over this problem than over any other discussed in this work. As the investigation has led only to negative results, this is not the place to go into details. It is to be hoped that the problem will be taken up a^ain when more data become available in the future. Fourier Coefficients. The Fourier Coefficients for the daily variation of temperature are given in the following table. The coefficients lead to no new results. The value of a— the amplitude of the whole- day period — is greatest in January and it is slightly le.ss in December than in the two neigh- bouring months. The phase of the whole-day period does not vary to any large extent from September to March, when the daily variation is governed by the sun. During the period April to August, when the solar energy is weak, both the amplitude and phase of the whole- day period are extremely irregular, this is to be expected from the curves for this period mven in figure 20, page 54. The amplitude of the half-day period is small throughout, and the irregularity of the phases for the different months shows that this jjeriod has no real physical basis. Table 30. Fourier Coefficients of the Ditihj Variation of Temperature. No. of months a A b B °F. O °P. ■> January . 2 2-22 219 0-2G 191 February 4 1-41 223 0-21 195 March 4 0-52 229 0-36 95 April 4 0-24 35 0-05 338 May 4 0-61 238 0-30 45 June 4 0-34 215 0-24 245 July 4 0-21 31 0-10 345 August . 4 0-40 54 0-27 276 September 3 0-54 219 0-29 9 October . 3 1-37 235 0-30 57 November 2 201 229 0-12 30 December 2 1-56 224 0-10 156 Nov. 1 Summer Dec. \ Jan. 1 6 1-90 225 0-04 159 El'^i"- .... ^'e-,'. O^ef 1 1* 0-57 227 0-20 53 May ] Winter June ^ July 1 12 0-27 238 0'06 359 Year 40 0-81 228 0-03 32 DAILY VARIATION— d" APE AD ARE. 73 Daily Variation of Temperature at Cape Adare. — Observations of temperature were taken at Cape Adare every two hours in June and July ; during otber months observations were not taken during the night hours. As it is impossible to correct the latter for non-periodic change they will not be dLscussed here, but the data will be found in the book of tables. Table 31. Daily Variation o f Temferature at Cape Adare. Month. 2 4 6 8 10 +0-6 -0-3 12 14 +(l-i' O-ll 10 IS + 0 1 20 9-> 24 Mp.an June 1911 . July ., -0-.5 -01 0 0 -on 0-0 -1- 0 .5 +0-O +0-1 +0-3 — O-O -O-i +0-3 -0-7 +0-4 +01 +CS3 -0-7 +0-4 -161°F. — 14-9 F. No excess of temperatures at 4 a.m. is shown in either of these months. The day hours were warmer than the night hours in June, but in July the period 16 hours to 24 hours was above the average. Non-periodic Changes of Temperature. In the previous section the object was to determine the true daily variation of tempera- ture unaffected by irregular changes. With this end in view, data from as large a number of days as possible, having a similar periodic variation, were combined in order that irregular changes which are independent of the time of day would cancel one another, the residual variation was assumed to be mainly due to true periodic causes. The irregular variations which were removed in this process are, however, of considerable meteorological importance and \vill be studied in this section. We have now therefore to employ a process which will eliminate the true daily variation and leave the non-periodic changes. Many methods have been used by other writers for this purpose, but most of them lead only to statistical results. As the essential purpose of the investigation is to find the physical laws underiying the meteorological observations, I have chosen two methods which appear to me to be the most suitable for the available data. The first method which depends on daily maximum and minimum temperatures will be used in discussing the non-periodic temperature variations in McMurdo Sound, because maximum and minimum temperatures are available for a longer period than temperatures recorded in any other way. On the other hand, no maximum and minimum temperatures were recorded at Cape Adare and Framheim, hence to compare the irregular temperature changes at these stations with those at McMurdo Sound another method will be adopted. Maximum ami Minimum Temperatures in Relation to N on- periodic Temperature Changes.— U there were no irregular temperature changes the maximum and minimum temperature of the day would be determined by the daily periodic temperature variatioiis. In this case the daily temperature range would be the same as the daily temperature amphtude.* This is practically the case in tropical countries. When, however, there are irregular temperature changes, these increase the difference between the range and the amplitude. In the Antarctic where irregular changes are very large compared with the regular changes, the mean range is considerably greater than the amplitude, and the difference, range - amplitude, affords a qualitative measure of the non-periodic changes. The difference between the range and the amplitude which we shall now study will for convenience be called the ' reduced range.' * See the definition of these terms, page 53. 10 74 TEMPERATURE. In the following table are given the mean temperature, the mean range, the amplitude, and the reduced range for each month. The values for the amplitude are those given in table 20, page 53, and the temperatures have been obtained from data for eaoh month from four years. Table 32. Unperiodic Temperature Variations in McMurdo Sound. Month. Mean temp<,Tature. Mean max. Mean mm. Mean range. Amplitude. Reduced range . °J? "F o-p "P 'F 'F Janaary + 23 + 28-7 + 16-9 11-S .5 0 0 8 Febru.-iry + ir. + 20-0 + 8-7 ii;i 31 8-2 March . + 4 + 9-4 2*2 11 0 ]-5 10 1 April . — S - 2-0 -15-5 i:j5 0 9 120 May -12 - 5-2 -20-7 loo 13 U2 June . -i:i - 4-9 — 22-4 17 o 13 102 July . -14 - (il -23-6 17 0 12 103 August, -14 - 6-5 -238 17:! 13 10 0 September . -!.-> - 5-2 22'7 17 5 1-8 Io-7 October - fi + 2-2 -12'1 14."! 2-9 11-4 November . + 13 + 18-5 + 7-4 ill 3-9 7-2 December +24 + 28-6 + 18-3 103 3-2 7 1 From the column for reduced range we see that non-periodic temperature changes are smallest in January and largest in July. Also by comparing the first and last columns we see that the reduced range has almost the same yearly variation as the mean temperature. This close relationship between the non-periodic temperature changes and the mean tem- perature, leads at once to the conclusion that cold weather is favourable to unsteady tem- peratures while warm weather is unfavourable. If, however, we examine the reduced range for any one month in difEerent years, we do not find that it follows very closely the mean temperature. Thus, in the month of June the years arranged in order of increasing mean temperature were 1902, 1903, 1911 and 1912, but in order of decreasing reduced range they are 1903, 1902, 1912, 1911, the data are as follows:— Table 33. June. 1902. 1903. 1911. 1912. Mean temperature Reduced range .... -100 17-4 -13-8 21 3 -135 110 -9 2F 145 NON-PERIODIC CHANGES. 75 If, however, we had used the average minimum temperature instead of the mean tempera- ture for the comparison, the years would have been arranged thus : — Table 34. Juiif. l'.:Mi:j. HHI.'. 11»11. 19i2. Mean miiiinuiin .... Reduced ranuo .... -i'7d 21:! -l'(i-;! 17-1: - 1 a-i 11 (! -ISCF 14.-) in which only one month of the lour is out of the sequence. The investigation has been carried further by plotting the reduced range for the cold months against the average maximum and average minimum temperature of the month. The points for the months May to August are plotted in figure 27. A glance shows at once that whereas there is only a small tendency for the points to lie on a straight line in the upper -40 30 1 1 70 ® ® ® MAXIMUM ® I* g il ^ « ® (!) ® ® ® 0 1 -15 -1 1 0 b ( ) + 5 20 C i--^^ W ® MINIMUM ■®^®- ® r - ® «)® 10 w ® ® n -55 -30 -25 -20 -15 Fig. 27. Reduced temperature ratiRe plotted against maximum and minimum temperatures, diagram, there is a considerable tendency in the lower diagram. This means that during the period May to August the reduced range, i.e., the unperiodic temperature changes, depends much more on the minimum than on the maximum temperatures. The relationship, expressed in statistical terms, is represented by the correlation coefficients — 7 and —-2 respectively. 76 TEMPEEATURE. During the warm months of the year there is a similar tendency, but much less marked. For the months November to February the correlation coefficients between the reduced range and the minimum and maximum temperatures are - 4 and -0 respectively. Thus we see that during the warm months neither the maximum nor the minimum tempera- ture plays a large part in determining the unperiodic temperature change,-. In order to understand these results, the temperature conditions in McMurdo Sound must be kept in mind. We will first confine our attention to the cold months May to August. There is little difference in mean temperature during this period, hence they may all be consi- dered together as having similar general temperature conditions. We have already shown that during the cold weather the direction from which the wind blows has little effect on the temperature, both northerly and southerly winds having higher temperatures than hold during calm weather. We have also shown that low temperatures ar^ associated mth a cold layer of air just above the surface, while high temperatures during the winter are caused by the removal of this cold layer and the establishment of a normal temperature gradient through- out the lower atmosphere. A little consideration will show that minimum temperatures are much more largely governed by the cold layer than maximum temperatures, while maximum temperatures are more affected than minimum temperatures by the relative warmth produced by the winds. For every time a cold layer forms even for a few hours it is recorded by the minimum thermometer, but it must exist undisturbed throughout the whole 24 hours if it is to affect the maximum temperature. Similarly a wind which only lasts a few minutes may raise the temperature and give a high maximum temperature, but if it does not last 24 hours it may not affect the ininimum temperature. A month during which cold layers frequently form will have a low average minimum temperature. But the cold layer is very unstable, for at a comparatively small distance above it, there is warm air ; hence during a period when the minimum temperature every day is low, there may be a large proportion of days with a relatively high maximum temperature. On the other hand when windy conditions exist both the maximum and minimum temperatures are high for the cessation of the wind for a few minutes does not at once lead to a cold surface layer which takes time to form. Thus frequently when maximum temperatures are high, minimum temperatures are high, but much less frequently are maximum temperatures low when minimum temperatures are low. In other words where the temperature is largely governed by the formation and removal of cold layers, average maximum temperatures cannot undergo such large variations as average minimum temperatures. That these conditions are fulfilled in McMurdo Sound during the winter is clearly seen from figure 27 in which the points for the minimum temperatures are spread over a much larger range than those for the maximum temperatures. In fact the average deviatio)i of the maximum temperatures from their common mean is only a little more than half that of the minimum temperatures, being 3-8°r. and 6-2°F. respectively. The tendency then during this period is for temperature changes to take place between two limits, the lower being the temperature of the cold layers, which vary largely from month to month both in intensity and duration, and the upper being the approximately uniform temperature which exists when the normal temperature gradient has been established in the lower atmosphere. It is now clear why the non-periodic changes during the winter are much more nearly related to the average minimum than the average maximum temperature, for the latter is an indication of the prevalence and intensity of the cold layer, the formation and sweeping away of which give rise to the majority of the temperature changes. In the summer months the cold siu-face layers do not form, while the temperature is now affected by the wind direction. Hence we see that not only are the temperature variations NON-PERIODIC CHANGES. 77 much smaller than in tlio winter, but tbey depend now much less on the minimum tempera- ture ; the correlation factor between reduced range and minimum temperature in the summer being only half what it was in the winter. Cotnparison of Non-periodic Temperature Changes at Cape Evans, Cap)eAdare and Framheim. — Unfortunately no maximum and minimum observations are available at Cape Adare and Fram- heim. Hence it is impossible to compare the aperiodic changes of temperature at these stations with Cape Evans by the method just used. The comparison may be made, however, by the second method mentioned above. The change in temperature from day to day depends almost entirely on aperiodic changes. The difference has therefoie been taken between the temperature at the same hour on consecutive days. At Framheim temperature observations were taken at 8 a.m., 2 p.m. and 8 p.m., the mean difference between the readings at each of these observations and the corresponding one on the following day have been u,sed. This gives 90 determinations of the change in a month of thirty days. The changes have been grouped according to whether the temperature rose or fell during the interval and the mean for the former entered in table 35 in the column headed + and the latter in the column headed -•. (The occasions when the changes were zero have been divided equally between the positive and negative groups.) In the right hand half of the table the number of times the changes were positive and negative have been entered. In order to make the comparison complete the same procedure, using the same hours, have been applied to the observations made at Cape Evans and Cape Adare, the results being entered in the same table. Unfortunately no observations are available at Cape Adare for January or February or at Framheim for February and March. Table 35. Interd lURNAL VARIABILIiy. Number of ocobrrences. Montn, ( 1 Cape Adare. Cape Evans. Framheim. Cape Aclaro. Cape Evans. Framheim. 1911 + — + — + — + — -f — + — February • • 2-8 3-6 40 41 March . 2-2 2-,') 4-3 3-9 39 .54 47 •16 April 5-3 3-9 4-3 io 10-7 !l-6 30 54 40 .50 43 44 May 4-5 1-4 7(1 .5 3 112 11-2 39 54 46 47 49 44 Jiiae 73 6-5 7-7 1110 134 14 3 44 46 34 56 44 46 .July . S-4 71 8-r. 70 161 14.5 38 55 43 50 43 .50 August . 101 81 9-4 7-0 17 0 13 7 47 40 41 52 44 49 September IIS OS 111 70 106 106 20 40 40 50 50 40 October 41 4-1 4-2 4-2 12 0 Oo 29 31 40 47 40 53 November 31 21 46 3-8 5-4 e-7 44 40 44 46 54 30 December 30 2-n 3-4 2-8 4.5 40 50 43 44 49 46 47 1912 January 28 3fi 45 4-2 50 43 40 47 To facilitate discussion the mean values of the interdiurnal variability irrespective of sign have been entered in table 36 together with the mean temperature. 78 TEMPERATURE. Table 3G. Month. Intekdiurnal vakiability lURBSPECTlVE OF SIGN. Mean temvekature. Caj)e Adare. Cape Evans. Fram- heim. Cape Adare. Cape Kvans. Fram- heim. mil February • - ■Ai + 18-7 Marcli . ■t-1 .. + 19-8 -f 7-2 April 4(i 4-4 Ill + 80 - 11 -17 8 May +•4 Gl 11-3 — ()•() -10-8 -31-7 .June 6 9 S-0 13-S -Iiil -135 -29-9 July 7-7 81 l.i-3 -14-9 -21 1 -32-8 August . 9 1 S-2 1.5-3 -13-7 -211 -47-7 September 9-3 !»-3 10-.^ - 2fi -l.'-r8 -3.V6 October . 4 1 \-l iri-s + 0(5 - 3-4 -llO Noveinhtr 2fi 4-2 (ii + 18-9 + 12-4 + 4 2 December . 2-7 3-1 4-3 + 27-4 + 22-0 + 2(1 (J 1912 January . 3-2 4-4 + 21-3 + 1.V1) The data used in tbe above table are only for one year, but the interdiurnal variation shows, as did the reduced range, that at Cape Evans the non-periodic temperature changes are largest in the winter and smallest in the summer. So far as the observations go the same is the case at Cape Adare and Framheim. Thus in the yearly variation the non-periodic changes follow closely the mean temperature. In every month the mean temperature at Framheim is lower than either of the other two stations, and the interdiurnal variability is greater. At Cape Evans and Cape Adare the same relationship does not hold so closely, but in six months out of ten Cape Evans has both the lower temperature and the greater variation. Cape Adare was very much nearer to permanent open water than Cape Evans, therefore the horizontal temperature gradient must have been much larger at the former than at the latter station, this was particularly the case in June when the mean temperature at Cape Adare was actually lower than at Cape Evans and the open ocean very much nearer. In spite of this. Capo Evans had on the whole the greater temperature variation, which shows that the horizontal temperature gradient is much le.ss efficient in producing temperature clianges than the vertical gradient. The large non-periodic variability at Framheim is particularly instructive. The low temperatures on the Barrier have already been ascribed to the great radiation which is possible over its surface of snow. The large variability of temperature signifies that tbe low Barrier temperature is confined to a low layer above which the air has a much higher relative temperature. In other words the large interdiurnal variation of temperature at Fram- heim strongly supjiorts the view already expressed that there is on the average a greater temperature inversion over the Barrier than over McMurdo Sound, so that the temperature of the upper air over the whole region is much more uniform than the temperature near the ground. If we examine the figures in table 35 for the period April to October we find that with few exceptions at all stations the rises of temperature are greater than the falls, but their MEAN MONTHLY TEMPERATURE. •79 frequency is less. Both these conditions are consistent with rapid rises of temperature and slow falls. This is exactly what would occur with the removal and formation of cold layers, for such a layer can be removed almost instantaneously by the setting in of a wind from any direction, but the fall of temperature would take place much more slowly, as the air cools under the abnormal radiation when calm weather sets in again. Mean Monthly Temperature. Description of the Data Available. McMurdo Sound. 'Discovery' Expedition, Hut Point 1902-04.— On pages 373 to 384 of the volume of meteorological results of the National Antarctic Expedition 1901-04 are given a complete set of two-hourly temperature observations made at Hut Point from February 9, 1902, to the end of January, 1904. Thus with the exception of eight days in February, 1902, the record is complete for two whole years. The mean values given on these pages are used in the following and not those given on page 464 of the same work which appear to have been computed from defective data ; the difference, however, is small. ' Nimrod ' Expedition, Cape Royds, 1908-09. — THe meteorological observations made on Shackleton's expedition have not yet been published, but mean monthly values of the tem- perature recorded at Cape Royds were given to Professor Hann who has published them on page 690 of volume III of his ' C'limatologie.' No information has been given as to the method of obtaining the monthly values, but probably they are based on two-hourly readings of the thermometers. 'Terra Nova' Expedition, Cape Evans, 1911-12. — There is a complete set of hourly data for the period February, 1911, to August, 1912. In addition, observations of daily maximum and minimum temperatures are available for 18 days in January, 1911, and for the remainder of 1912 after the end of August. The maximum and minimum temperatures can be used for finding the true mean temperature if the corrections to be applied are known. Fortunately it is possible to determine a close approximation to these corrections by comparing the observations of mean maximum and mininuim \\ith the true mean from observations made in three years when the observations are complete. In the following table are entered the differences between the true mean temperature based on either hourly or two-hourly observations, and the mean of the maximum and minimum temperatures, from which the corrections to be applied are found. Table 37. Corrections to he Applied to the Mean of the Maximum and Minimum Temperatures in Order to Obtain the True Mean Temperature. Year. September. October. Nuvember. December. January. "F -F °J? °F "e 1002 + i-:> 4-1-4 4-0:5 -o-;{ -0-5 llio:j + 1-2 4-1-4 4-1-9 4-0-7 4-0-7 mil -0-4 4-0-0 -01 4-0-5 4-0-6 Mean +0-7 4-1-1 4-0-7 4-0-3 4-0-3 80 TEMPERATURE. The values for 1902 and 1903 are taken from page 459 of the book quoted above, except that the value for January, 1902, has been reduced to -05, because the value given, —1-0, is ob\^ou.sly too large, and is based on defective observations as is pointed out by Dr. Chree on page 460. It is obvious from the above table that if the mean corrections are applied, the temperature derived from the mean maximum and minimum will not depart by one dcn-ee from the true mean. They have therefore been applied to the observations for January, 1911, and September to December, 1912, thus giving two years' satisfactory mean temperatures for 1911 and 1912. The following table summarises the above. Table 38. Data used fur Mean Monthly Temperature in McMurdo Sound. Station. IViiod. Data available. Hut Point Do. Cape Roycls . Cape Evans Do. Do. February, 1902 March, 1902. to January, 1 9(1-1 .March. 190S, to Ft-bruary, 19(19 , January, 1911 February, 1911, to August, 1912 . . . . September, 1912, to December, 1912 ... Two-hourly value.-i lor 20 days. Two-hourly values. Probably two-hourly \-alue.s. Eighteen days, ma.Kimum and minimun temperatures corrected by +0-3. Hourly values. Maxiuuiiu and minimum tcra])eratures corrected by -f 0-7, +11, +0-7 and +03 respectively. Thus for McMurdo Sound we have temperature data for five years from three stations. It will be shown later that Hut Point is probably slightly colder than Cape Evans which is a^ain slightly colder than Cape Royds, but by combining all the observations we obtain a better knowledge of the general temperature conditions in McMurdo Sound than could be obtained by considering separately the observations made at each station. Cape Adare. — The observations taken at Cape Adare by the Southern Cross Expedition under the leadership of Captain Borchgrevink have been published by the Royal Society in the volume ' Southern Cross Antarctic Expedition, Magnetic and Meteorological Observa- tions, 1902.' From March 4, 1899, to January 29, 1900, temperature observations were taken every two hours between 9 a.m. and 9 p.m. As the mean of these observations would not give a true mean temperature, the observations at 9 a.m. and 9 p.m. only have been used in obtaining the mean monthly temperature. In this series no observations for February are available, an approximate value has, however, been obtained by completing the curve of yearly variation by freehand. The members of Captain Scott's Northern Party were at Cape Adare during part of 1911. They took temperature observations as a rule every two hours from 8 a.m. to 8 p.m. In June and July, however, the observations were taken every two hours throughout the day and night. Thermograph records are not available, hence the mean monthly temperatures have been calculated from the means of 8 a.m. and 8 p.m. observations. The observations are complete from March to December inclusive. The values for the missing months January and February have again been supplied by completing the curve for the remaining months by freehand. MEAN MONTHLY TEMPERATURE. The above is summarised in the following table : — Table 39. Data used for Mean Monthly Temperature at Cape Adare. 81 Fi-liruai'v, ISyW .... Intcrpolatioir. March, 1899 ' Mean of 9 .i.M. and 9 p.m. for 28 days. April, 1S99, to Docemler. 1899 . Mean .of 9 a.m. and 9 p.m. .JiiniMry, I9ilii . Mean of 9 a. a. and 9 p.m. for 28 days. Jaiuuirv and February, 1911 Interpolation. Mareli, 1911. to Dcci-mbw, 1911 Mean of S a.x. and 8 p. .v.. Framheim. — Temperature observations were taken at Framheim at 8 a.m., 2 p.m. and 8 P.M. Professor Mohn has reduced the mean monthly temperatures for the period April, 1911, to January, 1912, from the 8 a.m. and 8 p.m. observations. Values for the months February and March have been obtained by completing the curve for yearly variation by freehand. Barrier. — The method of obtaining mean monthly temperatures on the Barrier have been described above (page 32). The values used refer to the Barrier north of One Ton Camp 79° 30' S., and south of Ross Island. Mean AlontMi/ Temperatures in McMurdo Sound. — Table 40 contains the mean temperature observed each month during five years in McMurdo Sound. Table 40. Mean Monthly Temperature. {°F.) Station. Year. January. P J3 < >1 c 3 1-^ 3 ijj 3 < a b4 o o -a s iz; s Year. Hut Point . 1902 . +22'5* + 159 +8-0 - 71 -12-5 -160 - 8 1- -165 -120 -8-5 +12-0 +231 -01 Do. 1903 . 4-2(il + 11-2 -0-8 -loy -160 -13-8 -211 — 165 -186 -6-8 + 15-4 +25-7 -2-7 Cape- Royds . 1908-09 +26-1 +20-5 +4-8 -10 8 - 5-4 - 71 -170 -157 - 5-6 +4-5 + 17-1 +300 + 3-4 Cape Kvans . 1911 +22-4 + 18-7 +7-2 ^ 11 -10-8 -135 -211 -211 -15-8 -34 + 12-3 +220 -0-4 Do. 1912 . +21-3 + 12-9 + 2G - 80 - 70 - 9-2 - 5-6 - 3 1 - 6-7 +3-5 + 143 + 23'8 +3-2 Mean 5 years . +23-7 + 15-8 + 4-4 - 8-8 -10-5 -11-9 -140 -146 -11-7 -21 + 14-2 +24-9 +0-7 In the previous section we have shown that daily non-periodic temperature changes are greater in the winter than in the summer, and it is easy to see from the above table that the monthly temperatures are moie irregular in the wiiater than in the summer. Thus the difference between the warmest and coldest July is 15-5°F., while the corresponding difference for January is only 4-8°F. This is still more clearly seen from the following table in which the departures of each month from the mean temperature of the month ba.sed on the five years are given. 1904. 11 82 TEMPERATURE. Table 41. McMurdo Temperature. Monthly De])artures from the Average. Year. Jan. Feb. Mar. April. May. June. July. Aug. Sept. Oct. Nov. Dec. Year. 1902 -1-2 +01 + 3-6 + 1-7 -20 -M + 6-5 - 1-9 -0-3 -6-4 2-2 -1-8 -0-8 1903 . +2-4 -4-6 -3-6 —81 _5-5 -1-9 -6-.5 - 1-9 -69 -4-7 — 1-2 + 0-8 -3-4 1908 09 + 2-4 +4-7 +0-4 -20 +51 +4'8 -2-4 - 11 + 6-1 +2-4 +2-9 +5-1 + 2-7 1911 -1-3 + 2-9 +2-8 + 7-7 -0-3 -16 -e-5 - (i 5 -4 1 -1-3 -1-9 -29 -11 1912 . -2-4 -2-9 -1-8 +0-8 + 2-9 +2-7 +9-0 + 11 -5 40 +5-7 + 1-4 +0-1 -11 + 2-5 Mean 1-9 3-0 2-4 41 3-2 30 6-2 4-6 3-2 1-7 2-3 3- 3/2-1 Taking the mean values for the season.s we have — Table 42. Mean Monthly Departures of Tempe.ratxire from Average in McMurdo Sound. F. Summer Nov.. Dec, Jan . ... I'O Autumn Feb., Mar,, April . ■ . 3-2 Winter May. 3m\c, July ..... 41 Spring Aug.. Sept., Oct. .... 4-1 Year 3-3 From the above table we see that the mean variability is twice as great in the winter as in the summer. In order to compare the above values with those for other places the following data are taken from Hann's Meteorologie, 3rd edition, page IIU : — Table 43. Mean Monthly Departure of Temperature from Average. North Russia JLd Russia. North Germany. Italy. Kngland- T. op 'F. °F. "'F. Summer . 2 9 2 •.5 1-6 1-8 1 8 Winter . 61 56 3-6 2-5 2-5 Year 41 3-8 2 3 2-2 2-3 The monthly variations in MciMurdo (Sound, as far as can be judged from five years' observations, are in both summer and winter less than in Russia. In the winter they are greater than in Italy and England, but of about the same value in the summer. On the mean of the year the monthly variations in McMnrdo Sound are less than in Russia, but greater than in England, Italy or Germany. ANNUAL VARIATION. 83 The great dependency of temperature on wind during the winter has already been pointed out, and this is again shown in the mean monthly values of temperature and wind. The following are typical cases : — Table 44. Tempemture and Wind : McMnrdo Sound. Mean temperature. Mean wind. i i!Mi:{ April \ { Mill °F. -16-9 - 11 Miles/hour 9 16 j 1902 June < I 1912 -160 - 9 2 12 32 1 1903 July ( 1912 -21-0 - 5-6 9 29 j I'JII August < I 1912 -21 1 - 3 1 17 25 Annual Variation op Temperature. The following table contains the mean monthly temperatures in the Ross Sea area derived from the data described on pp. 79-81. The values are plotted on figure 28. Table 45. ; Mean Monthly Temperatures. (°F.) Mnn til . Mc.VlUKDO SOUN'D. Cape Adakb. Framheim. North Barrier. 5 years. 2 years. 1 year. Computed. Januar Februai March April May June July August Septem Octobei Novem Deceml V •y ber bar )er +23 7 . +158 + 44 - 88 -10-5 -11-9 -146 -14-6 -117 - 21 + 14-2 +24-9 (+31(>) (+270) + 187 + 9-4 — 22 -145 -11-9 -n-fi — 7-5 — 0-6 + 18-5 +29-5 + 14 5 (+ 4-2) (- 6 7) -17 7 -31 7 -29-9 -337 -486 -35-5 -11 6 + 4-1 + 19-9 + 17 + 1 -15 -30 -33 -35 -36 -34 -27 -13 + 8 +22 Year + 0-7 + 7-0 -14-4 -15 84 TEMPERATURE. The curves on figure 28 show a remarkable similarity in the yearly temperature varia- tions at the four positions. At each station the fall from the summer maximum continues Dec Jan. Feb Maf April. May. June. July. Aug. Sepl. Oot Nov Deo. Jan. Feb. +30F + 20F + 10 F /• OF -ZQTf -30 F — 50F \ TEMPERATURE Cape Adore — ..— . Frcanheun Bfjurier i 'X ^ Fig. 2^. Annual vaiiation of temperature. at a nearly constant rate for three or four months, and then is suddenly checked ; and there are several months during the winter when the temperature is approximately constant. The physical meaning of these curves will best be investigated by comparing the temperature variations on the Barrier with the variations in a similar northerly latitude and also with the changes in solar energy. Mmn Temperature at 78° iV.— The mean latitude of McMurdo Sound, Framheim and the north of the Barrier is approximately 78° S. The mean temperature at the correspondiu.s ANNUAL VARIATION. 85 iatitudc in the north can be found from the values given by Mohn * for each 5° of latitude between 60" N. and the North Pole. Interpolating from these values we find the following as the mean monthly temperatures for 78^ N. Table 46. Mean Monthh/ Temperature at 78" N. ^F.) Janu- ary- Fcbrii ary. March. April. May. .Jime. July. August - Sep- tember. Octo- ber. No- vember. Decem- ber. NVar- -234 -24 0 -20-6 -62 I-1.51 +W-H +36-5 +338 4-20-3 +0-9 -8-5 -16-6 +32 Annual Variation of Insolation. — Angot has calculated the total heat received durin" each month at all ten degrees of latitude from the equator to each pole.f In his calculations he takes as unity ' the quantity of heat which falls on unit horizontal surface at the equator during an equinoxial day, supposing the sun to be at its mean distance from the earth and neglecting atmospheric absorption.' He then calculates the heat received assuming difierent values of atmospheric absorption, from which we choose the results obtained by taking the coefficient of transmission to be 0-7. The following table shows the total insolation received in each month at 78° N. and 78" S. Table 47. Total Insolation Received at 78° N. and 78° S. Janu- ary. Febru- ary. March. April. May. June. July. August. Sep- tember. Octn. ber. No- vember. Decem- ber. 0 158 78' N- 78° S. 0 13-2 0 61 0-8 1-2 50 0 11-5 0 14-9 0 !2-7 0 6 0 0 15 Oo 0 4-5 0 11 5 Comparison of Temperature with Insolation in the North and South Polar Begions.—The curves on figure 29 represent the following : — The mean monthly temperature at 78° N. The mean monthly temperature on the north of the Barrier. The total monthly insolation at 78° S. The difference between the insolation for 78° N. and 78° S. is too small to affect the following discussion ; therefore only the curve for 78° S. has been shown. For convenience of reference the curves for the north have been plotted against the corresponding months for the south, .i.e., the value of January in the north has been plotted against July in the south and so on. The curves for the Barrier and 78° N. show several most important dift'erences : (a) Whereas the temperature curve for the Barrier follows that of the insolation very closely, the curve for 78° N. has a lag of a month. (b) During the months of no insolation the temperature on the Barrier undergoes very little change, while in the north the temperature continues to fall after the insolation ceases in April nearly as rapidly as it had fallen previously. * Norwegian Nortli Pole Expedition, 1893-1896, page 575. + Ansot. Annales du Bureau Cent. Met. de France, 1883, page B 121. 86 TEMPERATUEE. +vrF South Jan. Feb Mar April May June Jury Aug Sep Oct. Nov. Dec Jan. Feb Mar April May June July Aug. Sep Oct NoflTM July Aug Sep Oct No^. Dec. Jan. Feb. Mar April May June July Aug Sep. Oct. Nov- Dec. Jan. Feb. Mar. April Fig. 29. Temperature and Insolation. (r) The temperature in the north oscillates about a mean temperature 18°r. higher than that over the Barrier, and as the amplitude of the oscillation is greater in the north than the south, the maximum temperature in the north is not only higher than in the south but is several degrees above the freezing point. Before we can explain these differences, it is necessary to understand the local conditions to which the observations refer. We have already described the surroundings of the Barrier. The temperature curve refers to that part of the Barrier between One Ton Camp in 79|° S. and the southern coast of Ross Island. The mean latitude is somewhat greater than 78° to which the other curves refer, but the difference is so small as to be immaterial for our discussion. The temperature curve for the north does not refer to any one locality, it is based on the mean temperature of latitude 78° N. A glance at the map of the north polar region will show that this latitude, except where it crosses Greenland, runs almost entirely over the Polar Sea, approximatsly 80 per cent, lies over water and 20 per cent, over snow-covered land. An important point is that within this circle of latitude practically the whole area is a permanently frozen sea, while the great continental masses of Asia and North America surround it on the outside for nearly three-quarters of its circumference. During the greater part of the year the sea along latitude 78° N. is itself frozen, but during the summer there are occasionally breaks in the ice through which the open water comes to the surface. It is a well-know-n fact that owing to the large heat capacity of water it warms up slowly during spring and summer and cools slowly during the autumn and winter. Haun has ANNUAL VARIATION. 87 shown this very clearly by analysing into harmonic terms the mean annual temperature varia- tions of a number of stations in the same latitude having land and sea climates. The results for 60° N. are shown in the following * :- Table 48. Yearly Variation of Temperature n.nd Insolation in 60° N. Latitude. Sea climate .... 7-9 .sin (2o3-6+r) + l-4 .sin (59-4 +2^;). Land climate .... 43-6 sin (271-3 + .i-) + 2-4 nin (:U1 ■3 + 2^-). Insolation .... 9-02 sin (29(vo + j:) + 1-2S sin (144-2+2.r). It will be noticed that in each case the amplitude of the fii'st term is much larger than the second term, therefore its phase gives very approximately the phase of the whole wave. Along 60° N. latitude the temperature variation lags behind the insolation by 15° at land stations and 43° at sea stations. As very approximately 1° lag represents 1 dav, the lag is half a montji at land stations and nearly one and a half months at sea stations. Analysing the curves on figure 29 in the same way we obtain the following ;— Table 49. Yearly Variation of Temperature and Insolation. Insolation 78' N. . (i-9 sin {117 + .r)t+3-2 sin (I43 + 23r). Temperature 78° N. 31-0 .sin ( 82+.r)t+4-9 sin (l(l9+2.r). Temperature Barrier 28-0 sin (108+j;) +78 sin (133+2x). Insulation 78° S. . . . 7-1 sin (116 + 0,-) +?0 sin (142 + 2x). Barrier temperature lags only 8° (i.e., 8 days) behind the insolation, while the temperature in 78° N. lags 35°. Therefore as regards lag, the Barrier behaves like a land station and 78° N. like a sea station. The amplitude, however, is a difficulty. The lag at a sea station compared with a land station is due to the absorption of heat bv the water during the season of rising temperature and the giving up of heat during the season of falling tempera- ture. This process of necessity reduces the amplitude of the temperature variation, and the greater the lag, the greater the reduction in amplitude. But we find that the average station in 78° N. has a larger temperature amplitude than the Barrier, in spite of a slightly less insolation amplitude. It is, therefore, impossible to explain the lag of temperature in the north as being due to its sea climate, for its greater amplitude would then be unexplained. In order to study more closely the corresponding changes in temperature in the north and over the Barrier, the following table has been prepared giving the change in temperature from each month to the next. * Hann Lehrb. d. Meteorologie, 3rd edition, page 98. In the following discussion each month cones!ponds to 30° and the mean value fur January has the zero angle, Februaiy 30'. Marc-h 00'. and so on. t For convenience of comparison 180° has been added to the phase of the harmonic terms in the north. 88 TEMPERATURE. Table 50, Change of -Temperature from Month to Month. Barrier - Aus. to Sept. Sept. to Oct. Oct. to Nov. Nov. to Dec. Dec. to .Tan . .Jan. to Feb. Feb. to March March to Apri 1 April to May May to June June to July July to Aug. l- + 7 + 14 + 21 + 14 - .'5 -18 - 14 - 1,5 - 3 2 - 1 + 2 +36 + U-4 +21 ■;! + I3-.T + 59 -2-7 -13.5 -19-4 -n-i -8-1 -0-8 -on 78° N. ] Feb. to March March to April April to May May to .June .Tnnf to Juiy .Julv to Aug. Aug. to Sept. Sept. to Oct. Oct. to Nov. Nov. to Dec. Dec. to Jan. Jlh. to Feb. The minimum temperature occurs over the Barrier in July, at 78° N. in February, which corre.sponds to August in the south. During the period of increasing insolation (September to December in the south, and March to June in the north), the temperature rises at approxi- mately the same rate in the north and south, 51°, and 49° respectively. The temperature, however, continues to rise in the north after midsummer, and July is 5-9°F. warmer than June, while over the Barrier the temperature commences to fall almost as soon as the insola- tion decreases, and January is 5°F. colder than December. The continued rise of tem- perature in the north after the solstice, coupled with the fact that the mean temperature during the two warmest months in the year, July and August, is well above the freezing point, is the key to the problem. Even during July and August, the greater part of the sea in latitude 78° N. is completely frozen over, and where it is not frozen over, the temperature of the sea is between 2 and 3°F. below the freezing point. Hence as the temperature of the surface does not rise above the freezing point, the air during these months is warmer than the surface, and therefore cannot obtain its heat locally. During July and August, the north polar ocean is surrounded by snow-free land, the temperature of which is well above the freezing point, and the winds carry warm air from the laud over the polar ocean, so causinif a higher temperature than would be produced by the insolation received locally. The course of the temperature in the north between the winter minimum and the summer maxi- mum is now obvious. At the end of the winter, the north polar ocean is frozen over, and surrounded by snow-covered land, which is locally colder than the (tcean further north. Thus the first sunshine of the year falls on a snow surface at a temperature much below the freez- ing point, hence the conditions are similar to those over the Barrier. The rise in temperature is, therefore, similar in both north and south. But over the laud of North Asia and America, the snow commences to melt exposing land and rock on which the insolation acts with "reater intensity. The snow covering rapidly decreases, and by the end of May has almost completely disappeared from the land, the temperature of which rises well above the freezing point. The pfessure distribution is then favourable for a considerable interchange of air between the warmer land and the colder ocean. The prevailing wind direction over the whole area mav not be from the south, but there are sufficient southerly winds to carry a considerable amount of warm air northwards with a consequent raising of the mean temperature over the ocean. Thus after May insolation is not the only source of heat in 78° N., but to it must be added the winds carrying warm air from the relatively warm continents in the south. The summer temperature is, therefore, well above the freezing point, and the maximum occurs a month after the insolation has commenced to decrease. ANNUAL VARIATION. 89 The coaclitions are entirely diltereiit in the south. Practically the whole region within the Antarctic Circle is snow-covered land, and completely surrounding this continent is an open ocean, the temperature of which in high latitudes even in midsimimer does not rise above the freezing point. On the return of the sun, the air over the Barrier warms up in propor- tion to the insolation received. At mid.summer the temperature is still 10°F. below the freez- ing point. Compared with the north, there is little inflow of heat from lower latitudes because the prevailing wind is almost entirely from the south, also the few winds which do blow from the north come from a cold ocean instead of from a warm land surface. Thus local insolation is practically the only source of heat even at midsummer, hence when it decreases after the solstice, the temperature decreases with it. The low summer temperatures in the Antarctic have been a source of difficulty to meteoro- logists. Mr. Dines, in discussing the results of the Discovery Expedition, wrote * : — • ' The low mean temperature of the summer is also strange, and very difficult to explain The insolation in the Antarctic in summer is greater than the insolation on any other part of the earth at any other time, this being the date at which the sun is nearest to the earth, and if temperature depended only on solar radiation we should have the highest terrestrial temperature occurring in December in the neighbourhood of the South Pole.f A large mass of ice pre- vents the temperature from rising above the freezing point, because the air, being nearly pervious to the radiation, takes its temperature chiefly from the surface with which it is in contact, and if that surface be ice or snow, it cannot be above 32° F. But ice, except in its power of evaporation, is in no way more efficacious than any other rock in checking a rise in temperature up to its own melting point. Why then does not the mean summer temperature at least reach the freezing point as indeed it does in the north polar regions, where the insolation is less intense ? ' The answer to the (piestion is now quite clear. Of the solar energy which falls within the Antarctic Circle, such a large proportion is lost by direct reflection from the snow that the remainder is not sufficient to raise the temperature of the aiik to the freezing point before the solstice is reached, and the energy commences to decrease. The Arctic Circle on the other hand is during the summer surrounded by snow-free continents, the temperature of which rises well above the freezing point, and warm air is carried from them across the whole polar region. Insolation, as Mr. Dines points out, is less in the north than the south, but insolation plus warm winds is much greater in the north than the south ; hence the tlifferenco in the respective maximum temperatures. During the period of decreasing insolation, the temperature over the Barrier falls slightly less rapidly than it rose giving a small lag behind the insolation, so that the mean temperature in April, when the insolation ends, is 4°F. higher than in August, when the sun appeared after the winter. This, however, is to be expected, for tho snow sui-face and the air above it have some slight heat capacity. The lag, however, is very much greater in tho north. At the end of the summer, the ice in the north is thinner than at the end of the winter, ther<; is some open water, and the surrounding land surfaces are still free from snow, and, therefore, have a relatively high temperature. All these effects supply heat, and keep the temperature high, thus when the sun sets for the last time in October, the temperature in the north is 24-9°F. higher than when it rose in February. * National Antarctic Expedition, 1901-1904, Meteorology, Part I, page 465. t In tliia sentence Mr. Dines neglects the atmosphere. If the atmospliere is consijercd, the region of maximum insolation is 35^ S., while tlie South I^lc receives less heat than the Equator. See Plate B 15 of Angot's work. 12 90 TEMPEKATURE. We now come to one of the most characteristic differences between the temperature in the north and the south. From April to August in the south, and from October to February in the north, no insolation is received. The temperature in the south during these months is nearly stationary, the fall being only 4°F. on the Barrier, while in the north the tempera- ture continues to fall, and February is 24 9°F. colder than October. Here again we have an effect of the different local surroundings. The Barrier is only a few miles from a sea, the temperature of which, during the winter, is, as we have already shown, 25°F. warmer. Over the Ross Sea, there are vertical convexion current.'', while over the Barrier there are none, therefore, the temperature of the sea governs the temperature of the upper atmosphere. Every blizzard which blows removes the cold air from the Barrier surface, and raises the temperature to that in adiabatic equilibrium with the warm upper air. The fall of temperature during the winter over the Barrier is, therefore, retarded, and the temperature remains nearly constant. In the north, the land which was warm in the summer is relatively cold in the winter, therefore, there is no great diti'erence in temperature over a short horizontal distance, and the upper air is likely to be much colder than the upper air over the Barrier. The temperature over the greater part of the north polar regions falls at the same rate, and, therefore, there is no local source of heat after insolation ceases, hence until insolation again commences after the winter, the temperature continues to fall. The curves for the yearly variation of temperature at the few places in the Ross Sea area given on figure 28, page 84, become of great interest in the light of the previous discussion. December is the warmest month at the Barrier, Framheim and McMurdo Sound, showing that at these stations the temperature follows the insolation very closely. At Cape Adare, however, the maximum temperature occurs in January, but Gape Adare is much more affected by open water than the other stations, and, therefore, its temperature cannot follow the insola- tion so closely. The effect of open water is clearly seen in the rates at which the tempera- ture falls after the summer maximum. The Barrier is furthest away from open water, and its temperature falls the most rapidly. Framheim is so situated in the south-east angle of the Ross Sea that it has the Barrier on three sides and open sea only on the fourth ; the tem- perature fall there is, therefore, not so rapid as on the Barrier, but much more rapid than in the McMurdo Sound, where the sea is open in most years until the end of March. Cape Adare is almost entirely surrounded by open water until May or .June, and the lag of tem- perature is greatest at that station. The character of the curves during the winter months is most interesting. The tempera- ture at both Cape Adare and McMurdo Sound continues falling until it is between — 10°F. and — 15°F. The simplest explanation is that this is the temperature at which the partially frozen Ross Sea gives up sufficient heat to counterbalance the radiation to the clear winter sky, and, therefore, it is the lowest mean temperature at Cape Adare and McMurdo Sound which are so much under its influence. If we neglect for the moment the low temperature of August at Framheim, we see that the winter temperatures are very similar at Framheim and on the Barrier ; this, as explained above, is due to the blizzards which remove the cold air over the Barrier and replace it by air in temperature equilibrium with the upper air whose temperature is governed by the relatively warm Ross Sea. The mean temperature of August 1911 at Framheim may .safely be judged to be abnormal ; a longer series of observations would, doubtless, show that during the months May to August the mean temperature at Framheim would be nearly stationary, as it is at other stations in the Ross Sea area. ANNUAL TEMPERATURE. 91 Annual Temperature. From table 40, page 81, it will be seen that the mean temperature of two j-ears at Hut Point was — 1-4°F., of two years at Cape Evans +1-4°F., and of one year at Cape Royds +3-4°F. Now Hut Point is near the end of McMurdo Sound, and, therefore, very near to the Barrier ; Cape Royds is near where McMurdo Sound opens into the Ross Sea and Cape Evans is between the two. It is, therefore, likely that these different temperatures represent a real geographical variation of temperature which decreases as one proceeds south in McMurdo Sound from the Ross Sea to the Barrier. A few simultaneous temperature observations made during 1911 and 1912 at Cape Evans and Hut Point contirm this conclusion, as will be seen from the following table: — Table 51. Difference hetircen SimidUnimiis Ohserralions of Temperature at Cape Evans and Hut Point. Month. No. of day.s of observations. Temperature difference. Cape Evans — Hut Point. February, 1911 March, 1911 March, 1912 April, 1911 April. 1912 May. 1911 September. 1911 October. 1911 November, 1911 i:t \l) 1.-. :i(t 30 12 S 26 4 — 1-7 _5-2 -2-4 -50 -5-5 -20 -12 -0-7 + 1-4 ■ The average variation of the mean temperature in individual years from the five-year mean is only 2-l°F. (see table 41, page 82) ; if this is partly due to geographical position, the coldest year 1911 being observed at Hut Point, and the warmest at Cape Royds, it shows a remarkably small variation from year to year, for average variations of this amount are met with at much lower latitudes. In table 52 are entered the mean temperatures of the stations in the Ross Sea area. Table 52. Mean Annual Temperature. station. Latitude. Longitude. Mean annual temperature. Period of observation. °F. Cape Aclare . 71° 18' 170° 9'E. + 70 Two years (three months interpolated). McMurdo Sound . 77° .38' 106° 24' E. + 0 7 Five years. Fraraheim . 78° 38' 19.5° 30' E -14 4 One year (two months interpolated). Barrier . ' . 79° 00' 170°00'E -15 See ijage 30. 92 TEMPERATURE. The Mean Temperature of the Antarctic. From a olimatologioal point of view, it is of f:;reat importanco to know the mean tem- perature of all parts of the globe. Several attempts have been made to estimate the mean temperature of the Antarctic ; but on account of the paucity of data, the results are open to a considerable amount of uncertainty. The latest attempt was made by Professor Meinardus in 1909.* His investigation gave the following values for the mean temperature between 60'' S. and the South Pole. Table .5.3. Memi Temperature of the Antarctic (Meinardus). Latitude ..... Tompt-raturi- .... GO S. 2.i-"F. 7(1 S. llii'F. SO" y. ^r, IF. 00 s, -13 0 F. These values have been plotted in figure 30, and the new data for Cape Adare, McMurdo Sound, Framheim, and the Barrier added for comparison. SOUTH LATITUDE Fig. ."iO. Mean temperature of the Antarctic. 9o°S The value for Cape Adare falls almost exactly on the curve, but that for McMurdo Sound is too high, while those for Framheim and the Barrier fall far too low. Meinardus's curve is afn estimate of the average temperature over the whole of various latitudes, and it is necessary to review in the light ot the new data whether Meinardus's values are satisfac- tory. Framheim is only one degree further south than McMurdo Sound, yet they differ in mean temperature by 15°F. We have already shown that this difference is due to McMurdo Sound • Scobel's Geographischea Handbuch, page 74. MEAN TEMPERATURE OF THE ANTARCTIC. 93 being so much under the influence of the sea. There can be little doubt that the temperature of all places on latitude 78" S. at any considerable distance from open water have temperatures nearer that of Framheim than that of McMurdo Sound, in fact the tem- perature of Framheim is probably higher than that of places on the same latitude but further from the sea. So far as we know, the only other part of the Antarctic at which the sea is ever open so far south a.s 78" S. is in the Weddel Sea, where Filchner penetrated in 1912 to 77° 4-)' S. The rest of the 78" S. parallel, in all probability, passes over ice as thick as that of the Barrier. It is thus reasonable to conclude that the average temperature of all places near sea level along latitude 78° S. is near to that found for Framheim and the Barrier, and therefore must be below the value given by Meinardus. .4s all the temperature data used by Meinardus have been obtained at stations either on the coast or over the sea itself, they are open to the same objection when used to give the average temperature over the Antarctic Continent. Cape Adare is surrounded on three sides by open water during seven or eight months of the year, and during the remainder open water is never far away. It is, therefore, almost certain that Cape Adare has a higher temperature than large tracts at the same latitude which are far away from the sea, and have a snow surface similar to that of the Barrier. But the temperature at Cape Adare has been given by Meinardus as the mean of the latitude on which Cape Adare is situated. Thus we see that at latitudes 71° S. and 78° S. Meinardus's mean temperatures for the whole latitude are too high, and at the latter, the evidence points to an error of about 11°F. It is, therefore, safe to conclude that all Meinardus's values south of the Antarctic Continent are too high, because he has based them on coast temperatures, and not on inland temperatures. Considering how little we know of the inland temperatures of the .Antarctic, and of the height of the continent, it seems useless to carry the discussion further. We can, however, say definitely that the temperatures H the various latitudes south of the Antarctic Circle are certainly lower than those given by Meinardus. This conclusion is of importance when we compare the mean temperatures of the Arctic and Antarctic for Meinardus's values are already considerably lower than those determined by Mohn for corresponding latitudes in the north. Table 54. I/atitude. Mcinardu.s"s mean temperature.? (or south latitudes. .Mohn".s values for north latitudes. Differenee. fin' 70" . . . . SO" 90' "F. +2.5-7 + 96 - 51 -130 + :!00 + 127 - n fi — .so 'F. -43 -3-7 -4-5 -41 According to the above table the temperatures in north polar regions are between 4°F. and 5''F. higher than at corresponding positions in the south and as Meinardus's values are too high the real difference is considerably greater. C H A P T K R III. WIND. Instruments and Methods. Anemometers. — Three anemometers were used at Cape Evans. [a] A small Robinson anemometer having .3-inch cups on 7g-inch arms and calibrated to the factor 2-73. This instrument was considered to be the standard instrument. (6) An anemometer also having the same size cups and arms, but so aiTanged that after a certain number of revolutions an electrical circuit was closed for a short time. Wires were taken from the anemometer to a recording instrument within the hut, by meaii.'-' of which a pen, \\Titing on a drum revolved by clockwork, was raised a short distance every time the circuit was closed. Thus a series of steps was drawn on the record each one of which corresponded to a defbiite amount of wind. A time mark was made on the actual record every hour by means of an electric current from the standard clock. The value of the \\ind amount corre- sponding to a step of the record was determined at frequei\t intervals, by counting the steps on the record corresponding to a day's run of the standard anemometer. Both of these instruments (n) and (h) were mounted side by side on a wooden frame on the top of Wind Vane Hill. (c) A Dines pressure tube. This instrument had been specially designed to meet the difficult conditions of the Antarctic. The head had been so arranged that it could be easily removed and cleared of snow, and the suction part of the head had been erected separately from the vane. Also a reservoir had been inserted between the head and the fioat in order to prevent any snow which entered the nozzle accumulating in the connecting pipes. The instrument as designed in London was specially made by Mr. Munro of Cornwall Road, London, and very kindly lent by him to the expedition. The head was erected at the east end of the hut within which was the recording pai-t. The temperature withrn the hut was generally above £i-eezing point, but it was found advisable to use petroleum as the liquid in the cistern, as on one or two occasions the water and glycerine supplied froze. The change in design of the head proved a great success in practice. During blizzards a cert.ain amount of snow accumulated just within the nozzle ; but every four hours, when meteorological observations were made, this was cleared out and it was only in the worst blizzards that any part of the record was lost omng to the nozzle becoming choked up with snow. The exposure of this anemometer on the roof of the hut would not have been satis- factory if it had been desired to use its record for obtaining true wind velocities, because the hut was in the lee of Wind Vane Hill and therefore not exposed to the full velocity of the wind. The records however have not been used for obtaining actual velocities, but only for studying the ' structure of the wind.' The velocities recorded by the cup anemometers INSTRUMENTS AND METHODS. 95 on Wind Vane Hill varied between 1'25 and 1'50 times the velocities recorded by the Dines anemometer at the Hut, the ratio varying from time to time as small changes in wind direction made appreciable differences in the amount of shielding. On a few occasions the Dines record has been used for supplying missing velocities when the cup anemometer was out of order. From February i, 1911, to September 30, 1912, the electrical cup anemometer gave a practically unbroken record of the hourly velocity of the wind. It has been found possible to fill in the few lacunae either from the records of the Dines anemometer or bv reasonable interpolation. Wind Vane.— A self-recording wind vane was installed dming February, 1911, on the top of Wind Vane Hill. From the firet trouble was experienced in this position. It was exceed- ingly difficult to change the paper during high winds, and the first five days' record was lost owing to the blowing away of the paper when it was taken oft' the drum. The exposure however was so good on the top of this hill that I was very reluctant to move the instru- ment, but when low temperatures set in, the stopping of the clock became so frequent that it became necessar}' on May 16, 1911, to remove it to the hut, where it was possible to have the clockwork inside. The hut was however in tli6 lee of Wind Vane Hill and the wind direction was affected to some extent by this obstacle. The change in direction was really very small ; but it was just sufficient to cause the predominant wind to be changed from S.E. to E.S.E., and a certain proportion of the E.S.E. winds just crossed the dividing line into the E. division. This small break in the contiimity of the series was unfortunate, but practically it has no large significance for, as will be shown later, the wind direction was so largely affected by Mount Erebus, that for all piactical pui-poses the winds may be divided simply into nort.herly and southerly, the former blowing from the Ross Sea to the Barrier, and the latter fi'om the Barrier to the Ross Sea. The self-registering wind vane was nf)t entirely satisfactory as it did not turn very freely and the time scale was so small that it was difficult to fix with accuracy the time when the wind clianged. Also it gave other troubles so that there was a disappointingly large number of missing records. Owing however to the great preponderance of winds from the directions about E.S.E. and N.N.W. it was possible to fill in a large proportion of the missing record with tolerable certainty. There is a more or less complete hourly record from February 7, 1911, to the end of August, 1912, and in addition readings of a small vane exposed on AVind Vane Hill were taken each morning from January 13, 1911, to December .30, 1912. Wind Velocity. Mean Wind Velocity. The members of our expedition who had been previously in th.e Antarctic with Scott or Shackleton, constantly remarked on the great amount of wind experienced in 1911 as compared with that of their previous visits. The following year, however, was verj' much worse, and the mean wind during 1912 at Cape Evans was the highest recorded up to that time during twelve consecutive months in any part of the Antarctic. The large variations from year to year and from month to month are shown in the following table in which the mean velocities for each month are given for the four years 1902, 1903, 1911 and 1912. 96 WIND. Table 55. Mean Wiiul Velocitij (miles per hour). Hut Point. Cape Evans Mouth. Mea 11 four 1902. l9o:j. 1911. 1912. years. Jauuary .... 7 * u 12 5 10-9 91 Fcljiuary .... 12 t 10 22-7 20- o 1U3 .March .... lot 11 2o(i 24-8 17 9 .\]jril ..... 12 t 9 15-9 20-2 14 3 .May 17 lu 120 250 lUO • luuo ..... 12 11 i:i-2 318 17(1 .July 12 9 18o 28 8 17 1 August .... 14 11 10-7 29 3 IfS Septembur .... 1(1 11 U-5 22-5 14-5 Outoljer . . 10 12 17-9 17 3 14 3 NoTember .... « 11 Hi-1 18-9 13-5 Uecembor .... 8 8 U7 111 104 Mean 11-0 9-9 lfi-7 ■21 i 14-8 It is very probable that tlie exposure of tlie anemumeter at Hut Point was not so free as at Cape Evans, but that would not account for the whole difference between the two periods 1902-03 and 1911-12. We have the opinion of the men who were in the Antarctic during both periods and they were unanimous that the wind conditions in the second period were much worse than in the first. Then again there is the large variation in the two years in the same place. June, 1912, had nearly 2}^ times as much wind as June, 1911, and May, 1912, had over twice as much as May, 1911, it is therefore not improbable that the whole year 1912 had over twice as much wind as 1903. The abnormal amount of wind in the six months April to September, 1912, is remarkable. The mean velocity during these months was 25' C miles per hour as compariid with the corre- sponding velocities of 138, 110 and 15' 1 in the other years respectively. The greatest amount of wind was recorded in June, 1911, during which month , the mean velocity was over 40 miles an hour during 41' 1 per cent, of the hours in the month and over 60 miles an hour during G per cent. The mean velocity experienced during this month, 31'8 miles per hour, is the highest yet recorded in the Antarctic, the neaiest being 30G miles 1904. t From Beaufort estiinatea. VELOCITY— ANNUAL VAKIATION. 97 an hour during March, 1903, at Snow Hill in the west Antarctic* The highest velocity of the two years occurred on May i, 1912, when a wind of 81 miles in the hour was recorded. Annual Variation of Wind Velocity. Examining the numbers in the last column of table 55 which give the mean monthly values for the four years (plotted on figure 31), we see that the annual variation has a double period with two maxima and two minima. It would probably be more correct to say Fig. 31. Annual variation of wind velocity. that the annual variation has a main period with a minimum in the summer and a maximum in the winter upon which is superposed a pronounced secondary maximum in February and March. With large variations in velocity from year to year which we have already noted it ia not surprising to find that the annual variations are not identical from year to year. Thus the absolute maximum of monthly wind velocity occurred in May in 1902, in October in 1903, in March in 1911, and in June in 1912. Nevertheless by examining the whole trend of each curve we see that the wind in each year does conform more or le.ss closely to the mean of the four years, i.e., in each year there is a marked minimum in January, a maximum in February or March and then a second maximum in one or other of the winter months. We may therefore accept the mean of the four yeats as representing fairly closely the normal annual variation of wind strength. The physical explanation of this annual variation would be interesting, but it cannot be given with any certainty. The following considerations may however be useful in indicating the lines along which an explanation may be found. * Since this was written Mawson haa reported the tremendou-s wind velocities found in Adelie Land, where the mean velocity for a year was 50 miles an hour. 13 98 WIND. Taking the annual variation to consist of — (a) a .siraj)le period with a miiiimiun in the summer and a maxinmm in the wintei', (6) a pronounced secondaiy maximum in Febmaiy and March, it will be found tliat the former of these is present at all stations near the Antarctic Con- tinent for which we have da,ta except Framlieim and Cape Adare, while the latter is present at all except the Belgica station. Table 56. Annual Variation of Wind at Antarctic Stations {miles per hour). Station. Framheim Hut Point Gajje Evans . Cape Adare . Belgica * Snow Hill * . Gauss Station * Petermann Island . 10-7 7-8 !J-2 80 6.5 7-2 129 117 12-; i 8-.3 10 0 91 7-8 81 110 210 ■- — T* >. ^ >. -fj cS <; s -^ — V /^ ^ V- _/ / 30 0 1 \ V 70 ^ J \ V •^ /^ N s Summer 1^ -^ \ \ N ove nbe r - Jan uar ' / y 10 V -N S, ^ / 20 S V "^ ^ / 0 10 Au uir n -./ /N s F ebr jar ftp- il N>- . V ^ - --. ^ ^ J /" 0 30 20 to 0 A / '\ V V\ inte r ^ / r -/' \ \ Mi y- Jul ' r "^ \ V ^ \, J r "^ 7 V V \ \/ /' / /^ .^ ^ y^ -A \ \ s, Yt ar /- ^ ^ ^ — -/ ^ y n 0 I 1 3 4 b 6 7 8 3 10 II 12 13 14 15 16 17 W 19 20 ?l Zl li ?4 Hours Fig. 33. Daily variation of calms. VELOCITY— DAILY VARIATION. 103 The variation during the ^vinter was not the same in the different years. In 1912 when there was an abnormal amount of wind during the winter months the daily variation was very large and quite different from that during the winter of any other of the four years. It is therefore very doubtful whether the variation shown during the winter has any real signi- ficance. Daily Variation of Calms.— The frequency with which calms (0—1 nnle per liour) occurred is shown by the curves on figure 33. The data used are for the twelve months March, 1911, to February, 1912. As is to be expected these show the same effect as the mean wind velocity, the calms being most frequent in the night and least frequent in the daytime. There is however one important difference. The character of the curve for the three winter months, May-Julv, is practically the same as during the other seasons of the year. It is easy to see why there should be more calms during the night and early morning than in the afternoon during seasons of the year with marked difference of insolation at these times. But why this characteristic should remain during months when the sun was entirely absent is not clear. The effect is most marked and is in all probability quite real. Frequency of Whuls of Different Velocities. The frequency with which winds of different velocities occur at any place has been strangely neglected in meteorological investigations; although, as we shall show, it is capable of givin" important information with respect to the forces at work producing the air motion. If from a rec(jrd of hourly wind velocities at any place we count the number of times each individual wdnd velocity has been reported it is possible to plot a curve showing the relative frequency with which each velocity occurs. In practice it is sufHcient to choose groups of velocities and in the following the groups will be 0 to 4, 5 to 9, 10 to 14 n to n +4 miles per hour. The data for three years' hourly observations of wind velocity at Yarmouth on the east coast of England and for the winter months December, January, and February from two years at Jubbulpore in the centre of India have been treated in this way with the following result : — Table 58. Frequency of Winds of Different Velocities, expressed as a percentage of the whole, at Yarmouth and Jubbulpore. YariMouth — year .... .hil)l)iil]H)rc — Uoc, Jan . Fvh. 0-4 5—9 10—14 15—19 20—24 25—29 30—34 35—39 40—44 15 miles 45- 49 pta- hour. 5-2 02-8 2:}o 2S-G 28-4 6-6 190 15 11 5 0-5 .5-7 111 30 21 0-7 These values have been plotted on figure 34. It will be seen at once that the curves foi' the two places are very different. At Jubbulpore the first group is the greatest, in fact more than 60 per cent, of all the observations fall within the group 0—4 miles an hour. At Yarmouth on the other hand this group contains only 5-2 percent, of the observations, while it is the third group which contains the greatest number of observations. From the curves it is clear that while at Jubbulpore calms occur more frequently than any wmd velo- city, at Yarmouth the most frequent wind is one having a velocity of about 8 miles an hour. 104 WIND. Yarmouth (Eng/and ). Jubfeulpors ( India ). The frequency curves for Yarmouth and Jubbmlpore are ^ood examples of two distinct types of frequency curves, and before pro- ceeding further we must consider what is their meaning. There can be little doubt that they indicate two different types of weather, namely, the types associated with anticyclonic and cyclonic pressure distribution. It is well known that while air motion is essential to a cyclone it is destructive of an anticyclone. It is therefore clear that a place generally under anti- cyclonic conditions will have more calms and light winds than high winds, on the other hand a place which is frequently visited by cyclones can have very few calms, and there must, there- fore, be some wind velocity which has a maxi- mum frequency. The interior of India during the winter is subject to strongly anticyclonic conditions, and cahas are the most frequent wind values experienced. On the other hand England throughout the year is subjected to a succession Fig. .S4 Frequencie.s of wind velocities. of cyclones, and calms are seldom experienced. Thus the pressure types in these two cases agree with the type of wind frequency. That the frequency type changes with the pressure type is shown by the summer observations at Jubbulpore. In June, July, and August the mterior of India becomes an area of low pressure. This is a cyclone which although of great extension is of little depth and the gradients on the average are very small. Nevertheless the most frequent wind is no longer in the first group but falls in the second group, the values being :^ Table .59. Wiwl frequency at Jubhdpore During the Monsoon. Jubbulpore — June, July, August . 0-4 5—9 10—14 15—19 20—24 miles 25—29 per hour. 225 27-0 18-5 4-7 00 01 Thus during these months Jubbulpore has the same type of frequency curve as Yarmouth and both are subject to a cyclonic distribution of pressure. We have thus been introduced to two types of frequency curves. In one the relative frequency increases as the velocity decreases right down to calnis ; this type is associated with anticyclonic pressure conditions and therefore will be called the anticyclonic type. In the other the frequency increases as the wind decreases down to a certain velocity after which the frequency decreases as the velocity decreases and calms may have a very small frequency ; this type is associated with cyclonic pressure distribution and therefore will be called the cyclonic type. The v\ind observations at Cape Evans, Framheim, Cape Adare, Gauss Station and Snow Hill have been investigated to find the type of their frequency curves, with the result that all prove to be of the anticyclonic type. The data are contained in table 60 and the curves VELOCITY— FREQUENCY. 105 have been pl(jtted in figure 35. In view of Meinardus's opinion that the stations on the edge of the Antarctic Continent are under cyclonic influence this is an important result and one which will prove useful when we come to discuss the pressure and wind conditions of tlie Antarctic as a whole. To make the investigation complete the observations taken Flo. 35. Frequency of wind velocities. during two years on the drift of the Fram (1894 and 1895 between latitudes 79° and %b\'^ N.) were analysed in the same way. The first group, 0 — 4 miles per hour, contained only 8'7 per cent, of the observations, while the second group was the maximum with 41-4 per cent., thus the frequency curve in north polar regions is of the cyclonic type. 14 106 WIND. Table 60. Frequency of Winds of Different Velocities, per cent. 0 5 10 15 . 2(1 25 30 35 40 45 50 .55 > (JO mile.s to to to to to to to to to to to to per 4 9 14 19 24 29 34 39 44 49 54 59 hour. Station, Period investigated. 0 1-9 41 6-4 80 l()-8 131 153 17-5 19-8 22-0 24-2 > 20 4 metri to to to to to to to to to to to to per .se- 1-8 4-0 63 8-5 10-7 130 15-2 17-4 19-7 21-9 24 1 26-4 cond. Franilicim . 10 months. 42-2 25-9 101 80 30 1-9 1-4 (10 0-3 Capo Evans 20 months. 22-4 15-3 8-0 7-0 7-7 8-2 7-7 0-7 0-2 4-8 2-6 1-3 1(1 Cape Evans nortli- 20 months. 220 19S 14,s 143 12 1 8-2 4-5 2 0 1.2 0-3 01 01 erly winds only. Cape Ailaie 10 months. 72-0 7-. 5-(l 3 0 2-,-) 2-(l 1.5 1-0 10 0-5 0-5 0-5 30 Snow Hill . 19 months. 20-3 , 14-4 1 11 7 1(1-7 9-i» 7-S 0-8 5-2 51 3-9 2-2 13 OS Gauss Station 11 months. 27 19 13 12 10 7 3 3 3 2 Kerguflen . 12 months. 11 SO 178 224 2.50 l-r2 0-0 2-9 O-fi (1-2 01 North Polar 2 years 8-7 41-4 31-4 12-3 4-2 1-4 (t-5 01 For comparison' with th(3 curves Eor the stations near the Antarctic Continent, the wind observations made at Kerguelen were examined. This .station, although near to the Antarctic, is certainly under the influence of frequent and deep cyclones. The frequency curve for this station is also shown on figure 3-5. It is most strongly of the cyclonic type, the first group containing only 11 per cent, of the observations, and the maximum frequency does not occur before the fifth group wdth wand velocities from 20 to 24 miles an hour. The frequencv curves have not been calculated for other stations than those discussed in this section, but it is very unlikely that two curves could be more widely different than those for Cape Adare and Kerguelen, which are both marine stations on the same ocean. Although all the curves for the true Antarctic Stations are of the anticvelonic type they show most instructive differences. The curves for Cape Adare and Framheim are .similar to those for Jubbulpore during the winter, they all decrease rapidly and regularly from their maximum. The curve for Snow Hill is somewhat different : it has the maxinnim in the first .group, but the maximum is only 20 per cent, of the whole and the curve decreases very slowly although regularly. A curve of the simple anticyclonic type with 4.3 per cent, of the wincfe of a greater velocity than 20 miles an hour mu.st be very unusual. The curve for the Gauss Station shows a depai-t.ure from the simple anticyclonic type. In.stead of decreasing regularlv from the maximum the curve decreases at fir.st fairly rapidly and then after the third group much more slowly so that the curve has an upward bulge between the third and the seventh groups showing that the frequency of the winds between 15 and 30 miles an hour is greater than they would be at a place having the same mean velocity and the anticyclonic type of frequency. This anomaly is much more marked at Cape Evans in fact after the fourth group, the frequency actually increases to the sixth group after w'hich it decreases even more slowly than the curve for Snow Hill. Framlieim is about three hundred miles to the east and Cape Adare is nearly the same distance to the north of Cape Evans. Yet neither of these stations shows the anomaly which is so strongly marked at Cape Evans. There is obviously at Cape Evans and to a lesser extent at the Gauss Station some factor affecting the winds other than the simple anticyclonic distribution of pressure. DIRECTION. 107 As to what this factor is at Cape Evans there can be little doiiljt. 'When the winds are investiffated vnth reference to their direction it is seen that the winds from the north do not show the abnormality, their frequency decreasing quite regularly from light winds to high winds. Tlae southerly winds— the blizzards— are alone responsible for the large excess of winds of an average velocity of about 30 miles an hour. We shall show in chapter VI that blizzards are neither cyclonic nor anticyclonic winds, but owe their origin to changes of pressure brought about by waves of pressure which travel from the south-east and afiect the pressure over the whole Ross Sea area. Thus the presence of blizzards superposed on an anticyclonic pressure distribution produces the peculiar form of the frequency curve at Cape Evans. The similarity of the curves for Cape Evans and the Gau.ss Station indicates similar condi- tions in the two localities, and we shall use this evidence later when discussing the conditions at the latter station (see page 247). This discussion of the frequency curve has taught us the following : — (a) McMurdo Sound, Framheim, th.e Gauss Station, Cape Adare and Snow Hill are all under the influence of an anticyclonic pressure distribution. {h) Kergulen and the north polar regions are under the influence of a cyclonic pressure distribution. (c) The shape of the curve at McMurdo Sound and to a lesser extent at the Gauss Station, shows the presence of some factor neither cyclonic nor anticylonic which factor at McMurdo Sound is clearly the blizzard, it is therefore likely that bliz- zards having a similar origin exist at the Gauss Station also. As the investigation of the frequency curves is worthy of considerably more attention than it is possible to give them here, especially as to certain periods of high winds and seasonal changes, the data for each month used in tabic 60 are given in full in the volume of tables where the method used in treating wind velocities measured on the Beaufort Scale is also described. Wind Direction. Hourly observations of wind velocity and direction at Cape Evans are available for the period February 6, 1911, to August 31, 1912. The following table summarises all the data : — Table 61. Wind Directions at Cape Evans. ^ ^ ^ ^ 8 f4 H p4 H JC H ■J. ^ J riablc and doubtful. ^ s: ^ ^ IZ 2: fe5 Z H W H CO OQ 01 a: n-o CO o ^ Percontafje of 0-0 ni 0-5 2-3 4-0 2-5 0-2 01 on 170 47-2 20. 2 2-2 n-0 00 3-6 total wind . Percentage of 0-0 0-2 0-7 2-8 4-2 3-0 0-3 0-2 0-0 9-2 34-8 171 25 0-3 0-0 0-1 10-4 13-7 froquency. Moan velocity 0 9 13 Ifi 19 14 14 11 8 37 27 23 17 7 9 8 0 5 mile.s per hour. 108 WIND. A glance at the upper diagram in figure 36 wliicli exhibits the total wind from each direc- tion shows at once how insignificant the air motion was from all directions except E., E.S.E. and Total Wind Whole period Feb 1911 -Au|191^ inclysive Mean VelociLy whole period Vi miles per hour Fig. 36. Total wind and mean velocity. S.E. which together had 84-4 per cent, of the total. Of the remaining 12 per cent, (neglecting variable and doubtful) 88 per cent, came from the three directions N.W., N.N.W. and N. The mean wind velocity for each direction is shown on the lower diagi-am of figm-e 36. There is little difference in the velocity of the thi'ee noi-therly directions which have any appreciable air motion, but N.N.W. which has the gi-eatest air motion has also the gi-eatest velocity. The velocities of the three important winds in the S.E. quadrant are instructive. The E.S.E. direction has the greatest air motion, but the E. has the greatest velocity. This is not only true in the mean, but it was frequently noticed in individual storms. During bliz- zards the wnd vane generally pointed to the E.S.E., the percentage fi-equency in this direc- tion being very much higher than for the directions on either side, but as the wind increased to its most violent effort.s the vane swung slightly to the left and the direction became nearer E. than E.S.E. although it practically never reached due E. On the other hand during lulls the vane swung slightly to the right, so that S.E. and very occasionally S.S.E. directions were recorded. From figure 36 we see that all the directions from S.S.E. to E. have high velocities, but that tie mean velocity increases as the wind becomes more easterly. We shall find the reason for this in our fur-ther discussion. The predommance of the winds from the S.E. quadrant was so great tliroughout the year that the resultant direction varied very little from month to month. A good measm-e of the steadiness of the wind direction is given by the ratio of the velocity of the resultant wind to the mean velocity independent of direction. It is obvious that with winds all from the same quarter this ratio would be 1, while if the rnnds were uniformly spread over the whole compass the resultant velocity would be zero and the ratio would fall to nothing. DIRECTION. Table 62. Cape Evans Wind. 109 Month- Resultant direction. Resultant velocity. Mean velocity. Ratio. Maximum velocity during one hour. Mean direction- 1911. Miles per hour. Mile.s per hour. Felijuary S. 4RE. (S. 61 E.) 22-.^ 22-7 10 61 S. 46 E. March . 8. 49 E- (S. 64E.) 22-3 25-6 0-9 57 S. 51 E. April S. 6?, E. (S. 78 E.) 9-2 13-9 0-6 55 S. 73- E. May S. 8(1 E. 4-9 120 0-4 54 S. 88 E. •Tune S. 67 E. 9-7 13-2 0-7 56 S. 67 E. July S. 78 E. 16-1 1 S-.5 n-9 66 S. 72 E. August . S. 78 E. 12-.5 16-7 n-8 66 S. 74 E. September S. 77 E. 11-8 14 5 0-8 57 S. 76 E. October . S. 85 E. 13 0 17-9 0-7 59 S. 85 E. November 8. 78 E. 13-5 161 0-8 43 K. 74 E. December N. 89 E. 8-2 14-7 0-6 48 N. 88 E. 1912. January . S. 77 E. 8-6 100 0-8 54 8. 70 E. February S. 68 E. 18-2 20-5 0-9 63 8. 68 E. March S. SO E. 19 0 248 0-8 68 S. 79 E. April S. 72 E. 130 20-2 0-7 60 S. 78 E. May S. 78 E. 188 2;)-n j 0-8 81 8. SO E. .June! S. 68 E. 20-3 31 -8 OS 73 S. 67 E. July S. 66 E. 24-7 28-7 1 0-9 79 S. 68 E. August . f.Tune 1911 ) Year -j tn } . iMav 1912 1 S. 63 E. 21-4 25-3 : 1 0-8 56 S. 64 E. S. 77 E. 13-7 - 1 0-8 81 8. 76 E. It has already been explained that in May, 1911, the position of the wind vane was changed in consequence of which the wind directions recorded were shifted somewhat to the E. To allow for this change and to make the series homogeneous, about 15° needs to be added to the resultant direction for February', March, and April, 1911 ; this has been done and shown by the figures in brackets. Throughout the period February, 1911, to August, 1911, the resultant direction only varied from 8. 61 E. in February, 1911, to N. 89 E. in December^ 1911, i.e., a change of only 30° (a little over one compass point). The average ratio of the resultant velocity to the mean velocity was '8 showing the concentration of the wind into a few directions about the resultant direction. Taking the twelve months from June, 1911, to May, 1912, the resultant velocity was 13-7 miles per hour and the resultant direction S. 77 E. 110 WIND. The resultant, velocity and direction for the year 1903 recorded at Hut Point were 6'7 miles per hour and N. 80 E. respectively. Thus the resultant direction was 23° more to the north at Hut Point than at Cape Evans. As the places ai'e so near together this can be only the result of local deflections of the wind. We must therefore discuss the effect of the land masses which influence the direction of the x^-ind in the McMurdo Sound region. Fig. .37. Wind directions. The general distribution of land can be seen on figure 37. The great mass of Ross Island is separated from the Western Plateau by the McMurdo Sound. Erebus is 13,000 feet high and its summit is only 12 miles from Cape Evans and between 40 and 50 miles from the table- land 9,000 feet high just across the Sound. It is quite clear therefore that these two masses which rise well above the lower mnds must affect the flow of air in their neighbourhood. DIRECTION. Ill We are chiefly concerned witli blizzard winds, we will therefore examine first the effect of the land distribution on these winds. If there had been a large number of observers taking simultaneous observatio)is at many different points it would have been possible to show the actual motion of the air at any given time during a blizzard. As these observers were not present the next best thing is to examine observations taken at different times in different places during blizzards. This it is possible to do from the meteorological records kept by the various sledging parties. Also the directions of the sastrugi give valuable inform- ation as to the directon of high winds. Strenm Lines during Blizzards. — In figure 37 the results of such an iuvestio-ation are shown. The stream lines indicated on the diagram are what one would expect from the con- figuration of the land, but in support the following evidence for several of the mo.st important positions is added. Position A. — -Cape Evans. — Wind E.S.E. During blizzards the wind was constantly from S.E., E.S.E. and E., of these E.S.E. was the most frequent asid this direction is shown. Position B. — ^The prevailing winds over McMurdo Sound were clearlv indicated by the direction of the sastrugi. There were two distinct sets of sastrugi on the line joinin" Cape Evans and Butter Point. One set. the most marked, indicated a wind from somewhat south of east and the other set a wind from between S.E. and S.S.E. There can be little doubt that the former were formed during the violent winds which were deflected by the shoulder of Erebus ; while the latter were formed bv tlie moderate south-easterly winds which constantly blew through McMurdo Sound from the Barrier. Both sets of sastrugi became more southerly as the Sound was crossed from east to west ; the actual observations were : Table 63. Sastrugi in McMurdo Sound. ~>h miles from Cape Evans ..... 12 ,. ., „ 25 Near C'apb JJemai-ihi on the western euast 1st set. 2ud set. E.byS. E.S.E. 8.E. E.S.E. S.S.E. No definite sastrug'. Position C.—Hut Point.—Wind E. to N.E. Both tlie records ni the Discovery Expedition and of the men of our expedition who remained at Hut Point show winds between E. and N.E. during the most violent blizzards. There can be little doubt that the wind is here deflected by the high ridge running N.N.E. fi'ora Hut Point. Position Z).— Wind S.E. On March 12, 13 and U, 1911, a part.y was at the point marked D during a very violent blizzard. S.E. winds were recorded throughout. Position £.— Variable winds. On July 4, 1911, the Cape Crozier party was at the point indicated by E. Throughout this day an east-south-easterly blizzard blew at Cape Evans with velocities up to 5'2 miles an hour, but the party recorded : ' Overcast all day with steadily falling snow. Wind 3 to 4 with occasional gusts from E.N.E. to S.E. and light breeze from S.S.E.' thus showing variable winds during a blizzard which was giving steady high winds from the E.S.E. at Cape Evans. Further within the large bay in the south of Ross Island the wind is always light and variable. The bay was explored right up to the coast by members of Captain Scott's first expedition and it was found to be full of light snow showing little or no signs of heavy winds. Dr. Wilson called this bay Tranquil Bay. 112 WIND. Position i''.— Wind S.W. Winh the Cape Crozier Party was at F. on July 10 a)Kl 11, 1911, there was a blizzard at Cape Evans throughout which a high wind blew from the E.S.E. and E. The wind experii^iced by the party was just as high and as steady but it was from the S.W. and S.S.W. Position G. — Cape CVo'/er.^Wind S.S.W. As is well known the Cape Crozier Party ex- perienced an e.Yceedingly violent blizzard on .July 22 and 23 while at Cape Crozier. During the height of this blizzard the wind at Cape Crozier was S.S.W. while at Cape Evans it was steadily from the E. Position H. — Corner Camp.~ln the autunni of 1911 a small wind vane was erected at Corner Camp which was so arranged that a steel point comiected to the vane pressed on to the aluminium ba.se plate, the idea being to deduce from the position of the maximum scratching the direction of the prevailing wind. Wh.en Captain Scott saw this vane in November, 1911, after it had been exposed throughout the winter he foimd that the maximum scratching' indicated 'a predominance of wind from the S.W. quarter." Position I. — ^The winds at I and to the south are clearly indicated by the sastrugi which all point to the high winds coming from the S. or S.S.W. From figure 37 it appears safe to conclude that during blizzards the wind streams along the w^st of the Barrier parallel to the edge of the high land. When this southerly stream impinges on Ross Island it breaks up into two branches one of which passes Cape Crozier as a S.W. or S.S.W. wind, and the other enters McMurdo Sound as a S.E. wind. Thus the easterly components of the wind at Cape Evans and Hut Point are the result of the air being compelled to move round the south-west shoulder of Erebus. During the most violent blizzards there is a great crowding of the stream lines on to the slopes of Erebus and then the wind becomes due E. at Cape Evans, while at Hut Point there is a local ridge which still furt.her deflects the wind, and the wind which is E. at Cape Evans becomes E.N.E. or even N.E. at Hut Point. The fact that a high easterly wind at Cape Evans occurs when there is a violent southerly wind along the west of the Barrier, will be of great importance when we come to construct isobaric charts. The conditions which give S. to S.W. winds over the Barrier are isobars running more or less from south to north with the low pressure in the east ; when isobars are drawn in this way on a small scale map it often appears as though the easterly wind at Cape Evans is blowing from low to high pressure. We have now seen that the majority of winds at Cape Evans are the deflected winds of the air stream which flows from the south during the blizzards. It is quite clear that there can be no true winds blowing across the Sound for the Western Mountains on the one side and Erebus on the other prevent such air motion. As seen from Cape Evans Mount Erebus occupies the horizon from E. to N.N.E.. hence winds from the directions between these points are practically impossible. The three directions E.N.E., N.E., and N.N.E. have only "5 per cent, of wind frequency. The horizon becomes open from N. to N.W. and the frequency of the winds from these directions increases, forming as we have already seen the only winds of any importance beyond those fi'om the S. to E. quadrant — the three directions N., N.N.W., and N.W. have 122 per cent. From the W.N.W., thougli W. to S. the hoiizon is again closed by the Western Mountains and the six wind directions from W.N.W. to S. have together only r4 per cent, of the total winds. Thus we see that the only motion possible in the McMurdo Sound is from the Ross Sea to the Barrier and from the Barrier to the Ross Sea. From the position of Cape Evans it happens that there are two directions which sharply divide off .the two kinds of air motion. During nineteen months not a single wind from the W.S.W. was recorded, and from the DIRECTION. U3 opposite direction E.N.B. only six. These two directions divide tlie winds into the two classes, and all winda between E.N.E. and W. tlirough X. blow from the Ross Sea to the Barrier, and all winds between S.S.W. and E.X.E. through S. blow from the Barrier to the Ross Sea. It is clear from the above discussion that the wind observations at Cape Evans give no indications beyond the fact that the pressure is such that the air flows in one or other of the directions through the Sound. Beyond this there can be no close connexion between the wind direction and the pressure gradient or any other meteorological factor. It has there, fore been found very useful to group together all the winds of one class and neglect the small variations within that class. Thus all the winds enumerated above which blow from the Ross Sea through the Sound to the Barrier have been grouped together and will in future be referred to as noi-t.herly winds. Similarly all the winds which blow in the opposite direction have been grouped together and will be referred to as southerly winds. Tlie following table shows the result of combining the winds into these two main classes. Table 64.* * In this tablo the wincl for which the direction was vaiialile or unknown has heen divided amongst N. and S._ in jiroportion to tlie recorded fn-queiioy. tlius if the frequency of the S. winds was twice that of the jj. winds one-third of tlie doulitfiil whids has been included under N. ' and Iwo-tjiirds under S. 15 114 WIND. During three-quarters of the wliole period the wind was blowing from the Barrier through McMurdo Sound, during a little more than half of the remaining period the wind was blowing from a northerly direction, and during the rest of the time the air was entirely still. The average veloeit\' of the wind from the south was nearly twice that of the wind from the north. The total flow of air from the south was nearly nine times that from the north. All of which shows the predominant part played by the winds fioin the south. Later on we shall have occasion to study in greater detail the weather conditions which accompanied the two types of wind and also the causes which give rise to them. For this purpose it is useful to group the observations not only according to direction but also accord- ing to velocity. As already stated the recording wind vane was not very satisfactory for low velocities, but above a velocity of 10 miles an hour the direction record during nineteen months is practically complete. It has therefore been decided to tlivide the winds having velocities up to 10 miles an hour into two groups irrespective of direction, these being 0 to 5 miles an hour and 6 to 10 miles an hour. Above 10 mile,s an hour the winds of each direction have been grouped into two classes of («) II to 30 miles an hour and (b) above 30 miles an hour. As these groups will be used very frequently it is convenient to give here a table showing the number of observations which fall under each head. Table 65. The number of Observations in the six Chief (rroups of Wind. North. Irrespective OF niREPTION. South. Month. >30 1 1 to 30 0 to 5 6 to 10 11 to 30 > 30 miles per hour. February . 1911. 0 13 114 31 226 16u M f - trr- ^-T^ J ^ Til nt-if Tttfrf null III r 'ri 1 ri'Miiii ,',)t JBiBTIinl 1 iiiiimiiuikii 1 '- H 3 — iiililli. 1 III 1 i i;.i:iilMliUL" r!;;ilj/i 1 1 ;5« ii 1 I'jijillt lii: ."rj II.!.;..' Ul ill' -III UUJlVl 1: '^" i':iiiiS [« i 1 m ■ nrt ■«-HI [^J-^ ^^ (JH "airmip' Rii'mmfFT'iir" if ' — 1 ■'■ 1 t H- LLJ — i — 1 ■Hf>^ " ' - |)f|||Tlfll nlnh:lp n J'- 1 »H J f 1 a D" •" s 1 1 0 !•- CAPE EVANS OBSERVATORr Pressdre Tdbe Anemometer Record. " - 1 H9 on I 1 1 5 7 0 0" i 3 « s ^ « 9 ro ^ — no ^^^^ wo 2S — — ^^ =^ i=t^ b=^ -^ ^g r^ UeJ F«^ t^ F=^ fc±: A-^ ^r^ 1 ^ pst J. EK F= "TCSg ^ -s-^ +- g : =rrJ 1 N= S^ Nf k^ ■1- - =H ^ .- 1 — ■ >- so 3 > ir ■^ ^ flfif W Ill W i 1 m ^ -4, 1 — «! 2-- &6 - ■ mimm — r 1 t.: J ™ "__T- - 1 1 1 ll i 1 P OOS- ^7 1 HO OH E E = -i:::-! I— z:::^ Q — 1 = — ■iiliihiin "t'lji^^ 9 .. 34s«;e»iOii>' Fig. 38. Anemosiams, southerly winds. STRUCTURE OF THE WIND. CAPe CVAN5 OBStRVATORV PRESSUBE TtoE ANEWOIffiTER RECORD. 1 ■ 10 W»- II NOON 23«5e7«9IO OTi"z»49e7e 9 „.. »>-^ — no •^^ -S- •M i '"t: UL^ "Nfl u^ miL^ Hsz:^ JM-iY a«B-^ 2*5=S iM_^ bmc jWM, {am. /iiiw. ^™^ .Mitf. Htfw : .t-^ H^ H=- *r-C_ h-^ Jt-"^ /, — X i'^ "i 9 g,n- ^ I 1? - — ^ . >■ I — aH Ol „ ^ fFH lii #nrJ pffl ^^ _lll pw ^ ^ fe .. _ .0 1 lag -^ 1= — i^t Npt^ m W kj ^ ,. < — =J B - 1 HI — J- l^* E 3 -+ 5 I ] 1 itij"" i 0 X MW.MWTM TOTTtXMAB.lQIMtna M ■ t^. .^i,-. >'^ , -fM- ':i* «pt ewAM^ — OBSERVKTORY PRESSURE ToBE AnsaioketerRccokd ■o^ ' WW" * t » • ^s _ B r a » w Fig. 39, Aremogiams, northeilySwinds. 118 WIND. Ill r.oiineotioii w itl^ an investigation into the velocity equivalents of the Beaufort Scale * mafle by the writer in 190-J, a tlctcnnination of gustiiiess was inoidcntally made for two coast stations in the British Isles, viz., Scilly Isles and Holyhead. The "ustincss was only investigated with reference to the average velocity and no account was taken ol wind direction or season. The mean value was also obtained from all hours hence it is soni(-.what greater than the value which would have been obtained if the method used in this discussion had been followed. The result showed that at both stations the gusli- ness was the same and was indepciuh^nt of wind velocity. The mean gustiness f(.r all velo- cities was found to be (38. In other words the average difference between the highest gust in an hour and the lowest lull was -08 times the mean velocity. As this is the only determin- ation of this nature of which I am aware we must take it as the standard with which to compare our Antarctic results. The Antarctic records have been worked up for the twelve months, Maich, 11)11, to February, lUTJ, inclusive. The mean gustiness for the jMirind was 1-U4. Thus the conclusion reached by an inspection of the traces is confirmed : the winds at Cape Evans are more gusty than those on the coast of tlreat Britain in the ratio of ru4 to "08 which ratio would have been still further increased if the British records had been reduced by the method used for the Antarctic records. Wiile in the Antarctic we soon noticed the characteristic difference in the gustiness of northerly anil southerly winds shown on figures 38 and 39, and we could as a rule determine the direction of the wind from an examination of the anemometer trace without going outside the hut. Whenever the trace was uniform without marked gusts and lulls we knew that the wind was from the north while a gusty wind was nearly always from the south. By examining the mean gustiness of northerly and southerly winds separately we find that southerly winds are more gusty than northerly winds, and the higher the mean velocity the more the "ustiness of the southerly winds exceeds that of the northerly winds, this is shown by the last line of table OG which gives the ratio of the gustiness of winds for different velocities. Table 6(3. Gudiness of W Ind (tccurdiiuj to Mean Velocihj. Miaii vul i(.:ity iiiiU-s piT liuur. 10— 2U :.'0— 3(1 :io— 4(j 4U- -oO Diii-clion ..... Gu.stiness ..... N. S. ■99 lit; N. S. •81 106 N. S. •67 -99 N. s. •90 Hatiu ..... I-.") 1-2 i-:j The gustiness of the wind at Scilly and Holyhead was foimd to be the same for all velocities ; this however was not so at Cape Evans for table 66 shows that with an increase of velocity for both northerly and southerly winds the gustiness decreases. This does not mean that for high \\'inds the actual difference in the velocity of gusts and lulls decreased, but that the ratio of this difference to the mean velocity decreased. Thus for a wind of 15 miles an hour the average difference between gust and lull was 1.5x 1'16=17'4 while for a velocity of 4.J miles an hour the corresponding difference was 4.JX'90=40 5, hence while the actual difference in the wind changes increased from 17-4 to 40") the gustiness fell from I'Ki to -flO. The gustiness of the wind varied considerably during the year. * a. ('. Siin])son : " TIk; Beaufort Sctie of Wind Furce," M. I). iNo. ISO. STRUCTURE OF THE WIND. 119 Table 67. Yearlij Varialiuii, af (Jasiina 1911 I!Ml'. Mar. Apl. May. ■Tune. 99 •Taly. Aug. 125 Sept. Oot. 126 Nov. nee. •Ian. FeV.. All winds ... 1-06 ■98 ■82 ro2 1^3fi ■92 •97 1-02 •92 These values are plotted in the upper curve of figure 40, from which it will be seen Mar Apr May Oun July Aug Sep Oct Nov Dec Jan Feb 1-3 1-2 11 < 10 •9 •8 / k 1 / \ \ / \ k / Gustir ofWi ess nd \ >^ V s \\ / \ ^ N \ V r / \ y / \ } ^^^ y^ > \ ^ / Tempe Max.- •'ature Min. \ y \ + 20°F + 10°F > < V > / N ( OT - 10°P -20T K / y: \ ^ Temp lean erature A / "^^ N ,y^ / ■f^ 20'F 15»F l(yF Fig, 40. (iiiHtine.vs of wind. 120 WIND. CAPE EVAMe ., OBSERVATORY Pressdhe Tdbe Anemometer Recobd . • ^Mta H ■ AM — 19 1 9 10 IO-- 11 NOOH 2»*5B»09iOii mo- 1" i s • • i B 7 •* ~ h 1 M- 1 , «0 ZIZ^ — ' ^ =J oe^ ER— jf?— — =; OCT^ «"- ^ 2i . S 10 — ■ 1-JW. c ^ iJ -"^ _tafc. I ^ i_tfli. i "■ h b" 10 1 M C-' '- Z 3 "^ S 6 t > "\ 8^ H^ CAPE EVANS OBStRVATORV Pressure Tube Anemometer Record - - 1— tO*» II HOOK 123*567 09 10 II imD" i"- Z 3 « 5 « 7 a 9 10 1 HO -R- 7 VT-;,- — -r — ^_„^ — ■. ^ •0 M 20 1 _, _^ , ^^ - gu. — r«^- 1-WBW ^^ c a *■ _lIE_i -^ ZDE^ jji^ lEE^ paid f*-^ jat; _U1V ,4-aE: ^^csc LEE s™ *r^-^ ^EBE ii^ s°^ +^™ _L-J. _^ ■i^ -^ ijnE ijni 1^ ' m ^ H'H '"5 5 5 ,0— f SD o H 3 h--— li +1 ■ ^1 — t l^p dl'ifl \\ 1 M. 2 r — Nn-i 1 1 HIIBIW 1 ls± ^fe Ijlt^ijm 1 i DJH^nHBj IJlLIi/ ;o -A — Mi nt^ Ift IBB [JM HE »s - -Lk-4 1 "Bl D" pp IIHJ ^ W« t^ &( _ 005- 1 P -4 » ^ — 1 ^ f^ ^ pj e k*i ri ^'^^J7 — ^ I 1 0 1 0 I** _ CAPE EVAH8 omehvatory PRESSURE ItWE ANEMOMETER RECORD. - -.. aMT. •O" tl IKK)PI t 2 3 ♦ 5 6 T e 9 10 n m!>- .-- Z 3 • . . AJl 19_-. _ B 9 » 11^ SB? B - • - 5 — - > a 5 3 • -'■ — 1 ■111 1 fi 1 I^ ^ m — 1 — — H 3 3 J - H 4 1 fW 1 « ffflPI i ffl A il f 1 I 1 h ri 90 20 o< _ V— ^r^ — — mV B sr to II MIO" K-- 2 3 Fig. 41. Anomogranis, squalls. WIND. 121 that the gustiness rose rapidly from May to September, remained high in October and then decreased rapidly and remained nearly constant during November, December, January, and February. This variation from month to month is just as marked with winds from the north as with winds from the south. Also if we fix our attention on any given wind velocity the variation fi-om month to month is exactly the same, thus the variation is not connected with the increased or decreased violence of the wind. It is pmely a seasonal change which affects equally all winds. Cause of gustiness. — It is known that the gustiness of the wind depends largely on the character of the country over which the wind blows. For instance the gustiness is much greater in a town than over flat marshy country,* the reason being that obstacles to the wind's motion set up eddies which show themselves by increasing and decreasing the wind's steady motion. At first one might be inclined to say that this is the cause of the high gustiness of the wind in McMurdo Sound, for there the wind is sweeping round the great obstacle of Mount Erebus at the foot of which the anemometer was installed. But this explanation fails entirely when we try to apply it to the variations in the gustiness which our investigation has revealed. If the gustiness is due to the violence with which the wind strives to pass obstacles why should it decrease as the \vind velocity increases ? Also the obstacles were exactly the same during August, September, and October as during November, December, and January, yet during the former months the gustiness was 33 per cent, greater than during the latter months. The following appears to be the most satisfactory explanation. We have frequently referred to the layer of abnormally cold air which forma near the ground during the cold months. In the next section we shall see how this cold layer is able to underrun a wind and raise the moving air above the ground so that at ground level the air is calm while above the wind continues to blow. The efiect of such a cold layer of air is easily seen. \Vhen the pressure gradient tends to set the whole mass of air into motion, the lower layer of cold air is in contact with the ground and does not readily move while the warm air above slides over the cold layer with little or no friction. Thus the upper warm air is soon moving more rapidly than the lower cold air. It is the interaction of these two layers of air which causes the unsteadiness in the resulting wind. One can picture a gusty wind as being analogous to a dirty stream of water entering a clear but stagnant lake. An observer situated at the bottom of the lake would be surrounded at one moment by dirty water and the next by clear, and he would find that the masses of dii-ty water were generally moving more rapidly than the clear water. The dirty water corresponds with the warm upper current, which drags along with it masses of the cold stagnant air like the dirty stream drags along the clear water of the lake. In the way that the observer on the bottom of the lake is able to distinguish the origin of the water which pa.sses him by its colour, so we are able to distinguish the successive masses of air by their temperature. The thermographs at Cape Evans were not particularly sensitive nor rapid in action, but their records give conclusive evidence of rapid changes in temperature during winds. Pai-ticularly during blizzards the thermograph trace became very thick, and frequently the trace was thickened over more than five degrees showing that the pen was constantly oscillating by this amoimt, which must have been much less than the * Shaw : Reports and Memoranda No. U of the Advisory Committee f c r Aeronautics, page 4. 16 122 WIND. CAPE EVANS OMERVATOflY PRESSURE TtreE ANEMOMETER RECORD. WOOH r_«."- AM 19- •a K>— ^ a : ^=^ H- Ir^ ^ fr-^ +^ 1 i)r-^ Ti-" iH^ ,^ ESE 1 Lir-^ 1 iS-^ *^ ...., -1-^ 1 — 1 4 E -^ -k— S3 no un •0 »i s - in OS -i CI — 0«- »( — .. ";'Jt«"T ^ ^ =ft ™ ■ jP| — f i iz: ~rl ^ J >- 50 o a «f 30 20 10 0 ° h D-- 1 R MM I 3 9 ^0 1 UID" '" 2 i 7 B 0 1*- D.S»UTH TOtTC>iH«M.t«M»OHa CAPE EVANS_ 0B3ERVAT0RV Pressure TtJBEANEMOMETERRECORD. 10- - II WOOH I L. MOM). SOUTH im CAPE EVANS Pressure Tube At4euometer Record. . H. ■■ la (C- II nCOM I25*S6 7B9I0IIII 0" 1" I 3 « a > 7 e 9 10 f- M K S :ac: jJUl^ pac ^-liL_ fr-^ psc k-^^ Ir'^ f^^UiM lir^ p= Ar^ #r^ ^-^ M"^ ;*t-^ 1— ^ *— ^ ;H^ +-^ ir-^ v-^ 1'^ >• ' §„ - S ■ £0 3 1 3 ♦- S 3 -^J -da t — ~ -IB J U— *- 53 — mtiiBi^H^i -1 JLO M - 91 ^ 3 ^^- 1^ 1* siF =1=S — f- 10 O-06- e — **¥*" cnmmu KUt »UTMm TCHMUI ^?^ — T* » H > 1 •10" h4 -L^ s e .—u. ra* 1 CHAM 1 T. 9 Fig. 42, AnemOKiaiiis, leciiiclescenres. STRUCTURE OF THE AVIND. 123 actual changes in air temperature owing to the sluggishness of the instrument. Eigure 43 shows a very typical example (September 19 — ^22). hours /s 19th. 8 20 th. '6 20 4 21 ST. 12 '6 20 22 NO Sep. 1911. -CALM Fig. 43. Wind and tempeiatme. During the calm weather of the 19th the cold layer formed near the ground. During the 20th the temperature slowly rose probably due to the presence of the southerly current which was sliding over the cold layer, occasionally it dragged some of the stagnant air vnth. it as shown by the fitful gusts of wind recorded by the anemometer and the rapid changes of temperature. At 16 hours on the 20th the upper current descended to the ground removing the cold layer and causing a rapid rise of temperature of over 20° F. After this the blizzard blew for many hours, and the thermograph trace is very thick — compare the thermograph trace during the 19th and 21st— showing that air masses of very different temperatures were moving past the thermograph. The gustiness of the wind is clearly shomi on the reproduc- tion of the anemograph record. During the 22nd the thermograph trace was particularly dis- turbed and shows very large variations of the temperature. We have shown above that the gustineis was much greater during August, September, and October than during the remaining nine months. This lends very great support to the present explanation. On page 76 we discussed the unperiodic temperature changes as shown by the average difference between the readings of the maximum and minimum thermometers and it was shown that these changes were mainly due to the formation and removal of the cold layers which we are now considering. A large average difference of the maximum and minimum temperatures was shown to indicate the prevalence of the cold layer. The values of this difference have been plotte^i on the middle curve of figure 40 for the same months as the values of the gustiness are shown on the to pcurve. It will be seen that the curves are very similar, thus indicating that both phenomena are due to the same cause. The significance of the similarity between these two curves is increased by the fact that no other meteorological factor has a marked maximum or minimum in September, 1911. The curve of mean temperature is added to figure 40 for comparison. If our explanation is correct we should expect that conditions which remove the surlace layer would reduce the gustiness. Now no factor is so efficient for removing this layer as a 121 AVIND. CAPE EVANS NOON _OB3ERVATo«Y Pressure Tube Anemometea Record. e 9 lO • CAPE EVANS 065EBVAT0HY PRESSURE TUBE ANEMOMETER RECORD. 10*- II NOON l23«Se7B9IOli Ult^ . . M « ifc.i»^ I » -S 6 7 8 9 0 '^x- M M 6 « > fe ^- 1 *~ z — -n^- -3^- — lit 05 -: — i tt i»]i' it4— =1 i ■ P fe =1 P 1 fe ^ i [g)ii''i , k m bi !• OOS- e— ; s^ NC 1 Ml m ^ rf ^ » 1 Ml lu! rr: z =4 4 n al 3 53 ^ 1 __QAPE. EVANS _ H)*- tl NOON OBSERVATORY 2 » PRESsiniE TOBE Anemometer Record'' * 5 6 7* a 9 10 11 mo" 2 « >• s 7 AM_ _I9_ 9 •0 ' f* -Vp— roo * " - WJKt 3^" APRIL so 2 io~ ir^ !!r^ lir-^ fr-^ iH^ tV^ i^ tV^ Tf!^ ir^ ^ff^ -A-" l4r^ ^^^ ^^ Vt^ '-^ 4^ t ^ V 'H^ Jl| V ■rt^ w "Ht- «=^ s : 5 - — L * 1 r ^ : s) 5 3 * — u, • ■ — « • &5 r is bd li 1 [pi n dkitt i 20 01 _ oos- >•- 1 NC i 1 wf^' i c y m i ) 1 } B"' ^ ^== m d ; U W ii^ KiG. 44. Aiieinouraraa. STRUCTURE OF THE WIND. 125 high wind, and we have on several occasions refeired to the efficient mixing of the air brought about by high winds. This is without doubt the cause of the decrease of gustiness with increase in wind velocity. Light winds only disturb the layer and do not remove it, hence they are abnormally gusty, while during high winds the layer has disappeared and the winds are leas gusty. Other features of the structure of the wind.— hi addition to gustiness the Cape Evans» winds show a considerable amount of squallines.i. If one examines the trace for October 15 — 16, figure 41A, p. 120, a series of waves of wind intensity will be noticed ; between 12 hours and 20 hours four distinct periods of maximum wind strength were recorded. For the rest of the day recurrences of similar waves appear. Each one of the increases in wund strength was accompanied by a swing of the wind vane slightly towards the 8., the change in direction only being small, from about E. by S. to E.S.E. The change in wind direction indicates that the waves were really of the nature of squalls. Squalls of this nature are recognisable in nearly all the wind records but with considerable variations in intensity and period. In the example just discussed the period of the squalls was about two hours, similar squalls but with a period of considerably less than an hour will be clearly seen between 1 a.m. and 3 a.m. on the morning of October 21st — figure 41B. Squalls occurred in winds from the north and the south. A typical example of squalls with northerly mnds is shown in figure 41C. An interesting development of squalls is shown in the three charts reproduced on page 122. The first two of these are charts of consecutive days, so that the second is the con- tinuation of the first. The chart for the first diy shows fully developed squalls of great intensity, these died out and for the first eight hours on the second chart the wind was not squally. Just, however, as the wind was dying away a squall was recorded after which there was an almost complete calm broken by recrudescences of wind for short intervals. There appears to be little doubt that the calm was due to a layer of stagnant air above which the southerly wind contiaued to blow, and the recrudescences of wind were the result of squalls which broke through the surface layer for a few minutes at a time. This explanation is supported by the thermograph trace which shows a rapid fall of temperature when the wind ceased at 18 hours on the 27th, and each of the recrudescences of wind was accom- panied by a large rise of temperature, showing that the cold layer was temporarily removed each time the wind reached the surface. The third chart on this diagram shows the same effect, the irruptions of wind being obviously due to an upper moving layer of aii- extending down to the ground for short periods. The wind traces for April 1, 2, and 3 shown on the opposite page are very remarkable. It will be noticed that at 8 hours on the 1st a fairly high south-easterly wind was blowing, this died down to a calm and at about H hours a northerly wind commenced to blow, and blew until just after 22 hours, this was followed by a calm until at about 3 hours on the 2nd the northerly wind again sprang up and lasted until nearly noon. After this another calm until nearly 16 hours when a south-easterly wind commenced. The variations in the south-easterly wind are interesting, three periods being shown with calms in between. The changes in temperature with the wind are instructive. The temperature during the south- easterly wind at 8 hours on the 1st was -f4°F., during the northerly wind between 14 and 20 hours +11°F., thus showing a rise of 7°F. With the next calm the temperature fell again to its previous value and only rose slightly with the succeeding northerly wind. During the calm from 12 to 15 hours on the 2nd the temperature was about+ 7°F. ; with the setting in of the southerly wind at 15 hours there was a sudden drop of temperature to -2°F., 126 WIND. *~ ' _5ape; EVANS ^ToTocRVATOHY r, Pressure Tdee Anemometer Record •-"lO** II NOON I 2 a^* 5 e 7 0 9 10 s e 7 a 9 10 ■ _ CAPE EVANS 0B8ERVAT0HY PRE&SURE TUDE ANEMOMETER RECORD. 10- '- M WOOM " AM To _ ^IG. 45. AneniograiiiH. STRUCTURE OF THE WIND. 127 i.e., a fall of 9°F. The temperature rose again as the wind fell, only to go still lower in the following peroid of wind. Tlius the northerly winds were warm and the southerly winds cold. There is little diffi- culty in explaining these wind and temperature changes. On these days there was little pres- sure difierence between the Ross Sea and the Barrier and the wind passed backwards and forwards through McMurdo Sound under local differences of pressure. The cold southerly wind and the warm northerly wind simply brought the temperatures from the Barrier and Ross Sea respectively with them. The trace for the next day, figure 440 shows a continued wind from the south but with varying intensities and two periods of calm. After the latter, at 6 hours on April 4, the northerly wind again appeared accompanied by a rise of temperature. The continuation of this wind is shown on figure 45A. This trace is most remarkable, the sudden break in the northerly wind between 11 hours and 12 hours 30 minutes has all the appearance of the instrument having gone out of order. But this was not so, the break in the curve was the result of a real interruption in the northerly wind current and during the interval there were a few gusts of wind from the S.S.E. When the northerly wind commenced again at 12 hours 30 minutes it blew exactly as if it had not been interrupted, and the wind trace on each side of the break obviously forms a single continuous curve. Coincident with the break in the wind trace there was a similar break in the temperature curve, during which the temperature fell three degrees. The temperature curve before and after the break is continuous in the same way as the wind trace. It is impossible to believe that a northerly wind of fifteen miles an hour could be stopped and restarted in the sudden way shown on the curve, with a slight wind from the south in the interval. There can be no doubt that the northerly wind was continuous but that during the interval cold air frojn the south was able to force its way under the northerly current and raise the latter above the recording instruments. An interesting example of a similar nature but with the break occurring in a southerly wind is showTi in figure 45B. There can be little doubt that the part of the curve between 23 and 24 hours is a continuation of the main curve which was broken at 21 hours. The wind at the commencement of the blow was from the E.S.E. and gradually worked to the B. as the wind increased. During the break the wind vane swung first to the south then back through E. to N. and then back to the south. It was then unsteady between S. and B. until the main current again set in soon after 23 hours when it remained steady at E.S.E. until the calm. The temperature curve showed a sudden rise of about 3°F. when the break in the wind curve occurred and the temperature returned to its previous value when the main air current was re-established. There can therefore be little doubt that in this case a flow of air from the north lifted the air current from the south above the instruments, thus it was an exact reversal of the previous case. Figure 45C is another example of the same phenomenon. The temperature record is interesting in this case also. At 10 hours a sudden drop took place in the wind velocitv with no appreciable change in wind direction, at the same time the temperature commenced to rise until the calm occurred. Wlien the wind rose again after the calm the temperature fell and reached its original value at the same time that the wind returned to its full strength. During the interval shown as a calm on the wind record, the wind vane appears to have made a single swing to the E.N.E. and back again ; otherwise the wind direction re- mained practically constant throughout. The lifting of the air current again is the only possible explanation. 128 WIND. A remarkable characteristic of southerly winds was the suddenness witii which they often commenced. Three typical examples are shown in figure 46. One would expect that such I 1 -c — A a < PSEH 3 7 -SE- 2 ■> E> • 7 5 6 7 8 9 10 MO 100 90 ao TO 60 50 dO 30 20 10 0 K 3 13W^ CATOr ^CS^ f£st; X If 13 z z « > i. J n ? m t b. e V. 1 J 10 J 5 6 7 8 9 ^ :ese- 7 Est TESE TESE1 •|CEf fH Ih lifl ii ! ii,,.iIu1b Hram ■I muH ■ HHUi^i fflP^MIH JLj — H ' li" Irt i 1 3 ■■■' 8 3 t 70 W J I c 60 - > so ? 40 z i April 2Qth. m May S/st. m. Fig. 46. Sudden setting in of blizzards. September 1st. tSll a sudden change in wind conditions would be connected with a sudden change in the baro- meter. The relationship between the barometer and blizzards will be discussed in the chapter on blizzards, but it may be mentioned here that there is no obvious connection between the sudden commencement of a blizzard and the barometer. In fact it would be quite impossible to say when a sudden commencement of a blizzard occurred from the barograph trace although in some cases there was an almost instantaneous rise m the wind velocity of over 30 miles an hour.* Cape Adare wind.s. The winds at Cape Adare are one of the paradoxes of the Antarctic. We have already seen that the frequency of light winds is very high. The Beaufort Scale was used for estimating the wind strength and from March to September, 1911, 693 per cent, of all the observations were recorded as either 0 or 1 on this scale. In other words, during considerably more than half the time there was practically no wind. On the other hand Cape Adare was visited by winds of hurricane strength. The observers were experienced men and were not inclined to overestimate the wind strength, yet there are many estimates of wind strength of 11 and 12 on the Beaufort Scale. WTjen forwarding the meteorological records of Cape Adare, Priestly sent to me a personal letter. This was not intended for publication, nor was it intended as a contribution to science, yet it would be impossible to give a clearer idea of the conditions during these storms than is contained in this colloquial description, and this must be my excuse for quoting from it here. * In a few instances a very slight discontinuity in the barograph trace accompaniod a sudden change in the wind velufitv. but in most cases even this was absent. CAPE AD ARE WIND8. 129 Priestly wrote : — ' The anemometer has been a fertile source . of troiiblroximately be taken to represent the atmosphere at heights of 10,000—13,000 feet, 13,000—16,000 feet and above 16,000 feet. The motion of Erebus smoke and of the clouds could only be observed when they were visible. It often happened that the motion could be recorded every time observations were made on one day and only tnice or twice on another day. If during the whole of the first day the motion remained constantly from say the north and on the second day con- stantly from the south there would be six entries of the iiorth motion and only one or two of south motion, although the south motion continued just as long as the north motion. In order to reduce as much as jjossible this source of error whenever a sequence of the same direction was recorded on any one dav it was only entered once in the reduction of the data. Thus 544 observations of Erebus smoke were reduced to 364, 166 of medium cloud to 133, 73 of hi,gh cloud to 67. The same procedure was followed in reducing the observations of cloud motion at Cape Adare ; so that out of 405 observations of the direction of low clouds at Cape Adare only 229 were used and out of 45 of high cloud only 41 were used. MOTION OF THE UPPER ATMOSPHERE. 133 The following table gives the results of the observations : — Table 72. Percen'wje frequence of air motion at different heiyhl'i in the atmosphere over Cape Emns and Cape Adare. Cape Evans. Cape Adare. VViiitl. Medium clouiLs. Calms . N. N.NE. N.E. . EN E. E. E S E. SE. . S.S.E. . S. s.s.w. s.w. . \V.9..\\. w. W.N.W. ' N.W. . N.N.W. Variable and nn observations Mean direction 1(1-4 3.5 113 (l-i> l(-(l 9-2 34-8 17 1 2o (1-3 0-0 0-1 0 0 0-2 0-7 2-8 4-2 13-7 S. 77 E 8-3 8-3 38 0-8 0 II 2-3 3-8 11-3 9-8 6-8 0-8 1-.5 0-8 4-5 2-3 14-3 211 Erebu.i .smoke. 11 9-9 High clouds- Wind. 3-3 13-. 121 14-8 20 -G 223 N. 26 W. S. 82 W. 3-0 13 4 6-0 4-5 li-0 9-0 10-5 e-0 4-5 3 0 1-5 0-0 1 .-> 9-0 7-5 4-5 10-5 N. 27 E. 38 4 2 4 0-2 MS 0-3 17 9-0 10-G 4-0 15-7 1-2 2-8 0-.5 3-6 18 6-3 0-6 Low clouds. S. 17 E. 0-0 61 2-6 6-6 0-9 3-9 11 8 240 31 14-8 0-4 31 0-0 3-9 1-7 14.0 31 S. .54 E. High cloud.:. 0-0 122 0-0 12 2 0-0 2 -.5 49 147 0-0 9-8 0-0 0-0 O-O 147 0-0 26-9 N. 23 W. Before proceeding to discuss the directions it may be pointed out that calms were sur- prisingly frequent in all layers of the atmosphere. This can be seen from the numbers entered in the line of calms in the above table. It should be remarked however that the air motion described as calm was entirely different in each of the regions discussed. Thus at the ground at Cape Evans, calm means a wind velocity of less than 1 mile an hour, and at Cape Adare, an estimate of 0 on the Beaufort Scale. In the case of Erebus smoke a calm was recorded when the smoke rose so vertically that no direction could be assigned to it. This means that the air through a great thickness was for all practical purposes actually still and yet this phenomenon was observed four times in twelve months. Of the observations of medium clotids 83 per cent, showed a motion so small that it could not be detected by means of the nephoscope described above, and the same was true of 3 per cent, of the observations of high clouds. On November 19, 1911, a balloon was sent up with an instrument which was arranged to be detached by a time fuse. After about 25 minutes the instrument was seen to fall 134 WIND. All observations. MC MURDO SOUND Upper motion when wind at surTace was 20 miles per hour or more from the Sooth CIRRUS CLOUDS EREBUS SMORE MEDIUM CLOUDS Upper motion when wind at surface was other than 20 miles per hour or nnore from the South Surface Winds WESTERN PLATEAU SOUTH POLAR PLATEAU 10% 20/i. :to% r,o% ^-l^.^t-ln^-^^-l^-^^-^^-l^-^^-^l-lUJ^u-J^-^^-l^-ll-^^-^^-i^-l^-^^-^^-^l /Vos 4 5i6 plol-ted fo y^ scale Fig. 47. Motion of atmosphere, Percentage fieciuency. :\IOTION OF THE UPPER ATMOSPHERE. 135 within 100 metres of its starting point. The instnmient record showed that it had risen 4,000 metres, hence up to this height the atmosphere was absolutely calm (see page 278). The directions of motion in the different layeis over Cape Evans are best discussed with the aid of figure 47. In this figure the diagrams down the left liand side, Nos. i. 4, 7 and 10, represent the percentage freriuence with vvhich air motion from each direction was recorded from the high clouds down to the ground, these diagi-ams will first be considered. Air motion at the grotnid.— -The lowest diagram. No. 10, shows the winds. It has already been explained that the wind motion at Cape Evans is limited by the Western Mountains on the west and Monnt Erebus on the east, so that there is practically no motion except through McMurdo Sound : from the Barrier to the Ro.ss Sea, represented bv the winds between south and east, and motion from the Sea to the Barrier, represented by winds between north- west and north. Air motion, at the height of the medium clouds, 10,000 to 1.3,000 feet. The diagram for the medium clouds 7, shows that the motion at this height is also mainly limited to two directions more or less parallel to tho.se of the wind at the surface, which might be expected from the statement made above that the medium clouds were mainly at a lower level than the top of Alonnt Erebus. A comparison of the diagram for the surface winds with the one for the medium clouds shows an important difference, namely that the motion from northerly direction is much more frequent at the level of the medium clouds than at the snrface Even when periods only are considered during which the snrface winds are over 20 miles an hour from the south the medium clouds move nearly a.« frequently from northerly as from southerly direction {see figure 47(8)). Air motion at the height of Erebus smoke, 13,000 to 16,000 feel.—The sunnnit of Erebus is above all the surrounding land masses and the direction of its snK)ke gives the first clue to the unimpeded air motion over the Ross Sea area. It will be seen that the diagram for the Erebus smoke, No. 4, differs from that for the medium clouds only in one particular^ namely in the addition of considerable motion from the west and south-west; there is still little or no motion from the east and north-east, while the lelative amounts fiom the Jiorth-west and south-east are practically the same. The mean direction of the motion of the atmosphere at the height of Erebus is from S. 82° W. which is slightly to the south of we.st. It will be shown later that the frequence of south-west motion of Erebus smoke has probably been somewhat under-estimated at Cape Evans, if this possible error were corrected it would bring the mean direction slightly further south, at the same time it will be shown that the mean direction of Erebus smoke observed by the Discovery Expedition, S. 56° W., was probably too nnich to the south. The true direction is therefore probably between S. 82° W. and S. 56° W. and so mav be taken as S. 70° W. or W.S.W. The direction of the motion of the air shown by Erebus smoke is in marked contrast to the direction of the wind at ground level. The surface winds are almost entirely from an easterly direction while those at the height of Erebus smoke are nearly as exclusively from a westerly direction. This means that between the ground and about 15,000 feet the pressure gradient is reversed, the low jiressure being over the Barrier and the high over the Ross Sea. There can be little doubt that this is a direct consequence of the large difference of temijera- ture between the Barrier and the Ross Sea. Owing to the higher temperature over the Ross Sea than the Barrier the decrease in pressure with height is less over the former area than the latter, thus at a certain height the low pressure over the Ross Sea is compensated for by the less dense air colunm and the pressure difference between the Barrier and the Sea 136 WIND. disappears. It is a simj)le matter to calculate the height at whii^h this occurs. On the mean of the year the following value.s are probably coiTect : — Table 73. ao3s Sea tiirrier .Mran tuinperaturi-. Mi-au pressure. 7 F. = -i4=C. -lOT. = -23T. 2fl.n.-)' = 738 mm. 29c'5" = 743 mm. With these values the pressure over the Sea and Barrier are equal at -5,500 feet = 1,680 metres, which height is oon.siderably below the height of Erebus whicli is approximately 13,000 feet. At the top of Erebus the gradient from the south to the north is almost exactly the same as that from the north to the south at sea level. The isobars will almost certainly run nearly parallel to the edge of the Barrier and vAW therefore be almost due east and west. Hence the gradient wind at the height of Erebus will be due west and this is the cause of the great preponderance of wind from the west shown by the Erebus smoke. At first sight it would seem strange that at the height of the medium clouds the \nnd blows so much more frequently from northerly directions than it does either below or above as shown by the surface winds and Erebus smoke respectively. The gradient on the ground is from south to north and the gradient is reversed well below the height of the medium clouds where it becomes from north to south. In the free air above Erebus a nort.h to south gradient produces west to east motion, but in the confined space below Erebus, where the medium clouds form, the air cannot move to the east and therefore moves along the gradient, i.e., from north to south. Thus mo.st of the westerly winds shown by the Erebus smoke are represented by northerly windi at the height of the medium clouds. Thus northerly motion is much more frequently shown by the medium clouds than by Erebus smoke. Air motion at the height of the cirrus clouds, above 16,000 feet. — ^Tbe cirrus clouds form at a distinctly higher altitude than that normally reached by Erebus smoke and figure 47(1) shows that the air motion revealed by them is essentially different. Reasons will be given in the next section for believing that the winds in the south-east quadrant both at the height of Erebus smoke and the cirrus clouds are connected with the southerly blizzards and if we provisionally neglect these, the diagram for the cirrus cloud is almost the same as that for the Erebus smoke rotated through a right angle. Thus, in so far as the motion is not affected by the local blizzards Erebus smoke shows an almost exclusive motion from the west to the east while the cirrus clouds show an ecpially exclusive motion from the nortli to the south. Without the south-easterly %\-inds the mean motion at the height of the cirrus clouds would be from somewhat to the west of north, Including these winds brings the mean motion to N. 27° E. which is approximately N.N.E. Blizzards and upper air motion. — ^An examination of the four diagrams in the first coluirm of figure 47 shows that at all heights in the atmosphere the frequency of the winds from the south-east quadrant is greater than would be expected fifun the frequenc}' in the south-we.'.t and north-east quadrants. The question at once arises as to the connexion between this relatively large frequency in all layers of the atmosphere and the surface blizzards. To investigate this question the upper currents were classified according to the simultaneous winds at the surface. Owing to the small number of observations of the cirrus clouds it was not possible to classify the cloud motion according to all wind diiections and velocities on the gi'ound ; it was therefore MOTION OF THE UPPER ATMOSPHERE. 137 decided to separate out the upper motions when the wnd at the surface was 20 niiles per liour or more from the soutli, and compare these with the roniainins observations. The former class will then represent blizzard \\-inds and a comparison with the remaining observations will show if the blizzards d(i affect the motion of the upper atmosphere. The following table gives the numerical results and they have been plotted in the second and third columns of figure 47. Table 74. Upper air motion, according lo /rimls at the surface. Percentage frequency. Medium clouds. Ekebus smoke. High clouds. Southerly winds > 20 miles per hour. Other wind.s. Southerly winds > 20 miles per hour. Other winds. Southerly wmds> 20 mik-s per hour. Other winds. N 8 7 5 10 6 21 N.N.E V 3 12 5 N.E. . .... 2 0 2 2 C 2 E.N.E 0 2 9 2 E 2 2 1 3 9 11 E.S.E 3 3 ' 18 5 S.E 20 6 22 9 3 7 S.S.E 17 6 3 5 S 5 6 6 10 0 7 s.s.w 0 1 3 0 s.w 0 2 12 15 (I n W.S.W 2 1 3 0 w 3 6 25 27 9 7 W.N.W 0 4 6 9 N.VV 8 1.5 26 22 0 7 N.N.W U 28 12 11 No motion .... 8 8 I 3 2 Examining first the diagrams for medium clouds we see that with blizzard winds the frequency from the S.S.E. and S.E. has grown at the expense of the winds from the N.N.W. and N.W. Rememljoring that these clouds are contained in the channel over McMurdo Sound, this is the result we should expect, if the pressure conditions causing the blizzards extended so high as the medium clouds. It is interesting to notice that there is a not inconsiderable number of winds at the lieiglit of the medium clouds which move against the high southerly winds at the surface, wliicli only emphasises the fact pointed out several times previously that large dift"crences of air motion exist separated by very small vertical distances. The diagrams for Erebus smoke are interesting, for the southerly wind diagram is very similar to that for the remaining winds, except that the frequency of the south-east winds is more than doubled. This seems to point to blizzard conditions being superposed on the normal conditions, and not that the normal conditions undergo a change which results in la 138 WIND. blizzard conditions. In other woids the blizzards at the surface do not as a rule affect the normal conditions at the height of Erebus smoke, but every now and then the blizzards reach as high as Erebus and then give south-easterly winds there. The number of observations of high clouds is too small to insist on small differences in the two diagrams, but the greatly increased frequence of motion from the E.8.E. during blizzards at the surface almost certainly points to the fact that occasionally during blizzards the south-easterly motion does extend right uji to the cirrus cloud layer. It must be remembered that the upper aii' motion could not be observed owing to low clouds when the most violent blizzards occurred. The occasional observations of cloud and smoke motion with high southerly winds were made when the sky was more or less free from lower clouds, hence the diagrams do not represent true blizzard conditions. We may therefore conclude that the increase of southerly motion at the different cloud levels would have been more marked if the motion could have been observed during the more violent blizzards. It will be noticed that there is still a considerable amount of wind from the south-east quadrant throughout the upper air when -the blizzards were not blowing at the surface. This liowever is not surprising, for blizzard conditions in the upper air may start before and remain after they are active at the surface. With the data available it is impossible to prove that all the winds from the south-east quadrant in the upper air are connected with blizzards, but we have shown a sufficiently close relationship to make this probable. We can now sum up the above discussion and state the general outlines of the mean air motion over Cape Evans from the ground upwards. Ground level. — The air motion is entirely backwards and forwards through McMurdo Soimd and the motion through the Sound from the south far outbalances the motion from the noith, the relative frequency in the two cases being 64 per cent, and 12 per cent. (24 per cent. being accounted for by calms, variable winds and no observations). Height of the medium clouds (10,000 to 13,000 Jeel and therefore below the level of the toj) oj Erehiis). — At this height the motion is still backwards and forwards through the .Sound, but the motion from the north is now more frequent than the motion from the south, the relative frequencies being 55 per cent, and 37 per coit. (8 per cent, being accounted for by no motion of the clouds). HeigJit of Erebus smoke (13,000 feet to 16,000 feet).— At tliis level the motion of the atmosphere is no longer affected by the surrounding land masses. There is now considerable motion from the west and south-west — directions which were not possible below the siuumit of Erebus. If we assume that the wi-nds from the south-east are mainly due to blizzards and therefore of local origin, the observations of Erebus smoke show that Ijetween 13,000 and 16,000 feet the general motion of the atmosp^here is almost entirely from the western half of the horizon with north and south added. The mean direction from all the observa- tions corrected by observations made on the Discovery Expedition is probably from the west-south-west, and if the local blizzard winds are neglected the mean motion is practically from due west. Height of the high clouds {above 16,000 feet). — The high clouds show a certani amount of motion from the south-east quadrant which in all probability is due to the local blizzards. If these winds are neglected practically all the motion at this height is from the northern half of the horizon with east and west motion added. The mean motion excluding the winds in the south-east quadrant is slightly from the west of north and including these winds N. 27° E. JIOTION OF THE UPPER ATMOSPHERE. 139 Thus at the ground levul the traiisfercuco of air in almost entirely from the south, at the height of Erebus smoke the motion is from the west and at th.o greatest height observed the motion is from the north. Motion of the Upper Atmosphere at Cape Adare. — The above conchisions are borne out by the observations made f)f cloud motion at Cape Adare, the results of which have been included in table 72, page 133. In figure 18 three diagrauLS are shown, for the winds, the SURFACE WIND LOWER CLOUDS \ UPPER CLOUDS ^ / ^ / 1 Z 3 \ II 10°/, 20% 30% 40%, 50% Fig. -IS. Motion of atmosphere. Percentage frequency at Cape Adare. low clouds and the high clouds. At the level of the low clouds the chief motion is from the S.E. and the mean motion is 8. 51° E. showing an excess of motion from the south over the north. The clouds described as high at Cape Adare were in all probability lower than the high clouds observed by means of the camera obscura at Cape Evans, and therefore their mean height was probably between the high clouds and Erebus smoke as observed at Cape Evans. It will be noticed that the diagram is intermediate between those for Erebus smoke and high clouds at Cape Evans, and the mean direction N. 23° W. is also inter- mediate. Thus at Cape Adare also the transference of air at the ground level is from south to north and in the upper atmosphere from north to south. Rui Point Observations. — The observations of upper air motion made^ at Hut Point (within 15 miles of Cape Evans) by the Di.scovery Expedition need to be reviewed in the light of the new observations. For this purpose the diagram from page 496 of the Meteorological Results of the National Antarctic Expedition 1901-01 has been reproduced here, figure 49. With regard to the surface winds we have already explained that the east and north-east winds at Hut Point were really the southerly winds of the Barrier deflected by the foot- hills of Erebus and therefore correspond to the south-east winds at Cape Evans. Comparing the lower clouds with the surface winds we see nothing to correspond with the prevailing direction of the surface winds. Now if the clouds were really lower clouds their directions should not depart largely from the directions of the surface winds for they are largely carried along in the lower current. Further there can be no doubt that the frequency with which the lower clouds moved from the south must have been much greater than the frequency with which they move from the north, for all over the west of the Ross Sea area the lower atmosphere moves mainly from the south, yet on this diagram the frequency is ^jracti- cally the same from the north as from the south. On looking up the actual meteorolo- gical register the reason is obvious. The direction of the motion of the clouds was never entered when the sky was completely overcast with Itnv clouds ! Hence all the cloud motion during blizzards which is exclusively from southerly directions has been neglected, thus the MO WIND. diagram of frequence of motion of lower clouds is useless for giving the mean motion of the lower atmosphere. The difference in the diagrams for the Erebus smoke as observed from Hut Point and Cape Evans is large, the chief discrepancies being in the north-west and south-west directions. SURFACE tV/NDS. LOWER CLOUDS. UPPER CLOUDS. M^ EREBUS SMOKE. 52 % 10% rt ?o% —I — 30% Fio. 49. Motion of atmosplieie. Percentage fieiiuency, Hut Point It is probable that the cause is the difference in the point of view of the mountain from the two stations. It is quite possible that the observations of south-west direction at Cape Evans were ailiectcd for the reason already stated, namel_y, that the mountain itself, being to the north-east, hid much smoke travelling away in that direction, hence the south- west directions in the Cape Evans diagram very likely need to be increased. It is not clear why the north-westerly directions were so few as scon from Hut Point and so many as seen from Cape Evans, but that this was due also to the point of view is supported by the observations made on sledge journeys from Hut Point which gave 10 per cent, of PLATEAU WINDS. 141 observations from the north-west as compared with only 3 per cent, as observed from Hut Point itself. This indicates that tlie position at Hut Point was not good for observing motions of Erebus smoke from the north-west, while as one got into a better position during the sledge journeys the relative number of observations from the north-west increased. A perfectly free view would no doubt have increased the relative frequency still further. It therefore appears safe to conclude that the frequency from the north-west as observed at Hut Point should be increased If the south-west frequency at Cape Evans and the north-west at Hut Point were in- creased as here suggested, the two diagrams would become the same in all essentials. On the Discovery Expedition cloud motion was determined without instrumental aid and therefore the high clouds, the direction of which could be determined, were lower than those observed at Cape Evans. From internal evidence I am inclined to believe that the high clouds observed at Hut Point were those described as medium at Cape Evans, with a few of the lower cirrus clouds included. As Hut Point is not within McMurdo Sound westerly and south-westerly motion is possible at the height of the medium clouds above that station, therefore the motion of these clouds over Hut Point was diiierent from their motion over Cape Evans. It \\all be noticed that the diagram for high clouds at Hut Point is almost exactly the same as the diagram for Erebus smoke as observed at Cape Evans, especially if the latter is corrected as suggested above by increasing the observations from the south- west. Thus the motion of the so-called high clouds over Hut Point confirms the mean motion of the atmosphere deduced from the observations of Erebus smoke at both stations, leaving the motion of the upper atmosphere to be indicated by the true cirrus clouds observed by aid of the camera obscura at Cape Evans. Plateau Winds. The plateau which exists behind the Western Mountains and extends from north of Cape Adare to well beyond the South Pole is at an elevation of between 6,000 and 10,000 feet. The wnds on its surface are therefore important in discussing the motion of the atmosphere. This plateau has been reached six times : (1) by Captain Scott in November 1903 when he ascended the Ferrar Glacier and travelled westwards over the We.stern Plateau to 78° S. and 146^° E. ; (2) by Shackleton in 1908-09 when he ascended the Beardmore Glacier and reached the Polar Plateau for the first time ; (3) by David who in the same year ascended by means of the Drygalski Glacier in his successful attempt to reach the south magnetic pole ; (4) by Captain Amundsen who reached the South Pole on December 17, 1911 ; (5) by Captain Scott in 1911-12 on his ill-fated journey to the Pole ; and (6) by members of Mawson's Australian Expedition who in 1913 attempted to reach the south magnetic pole from their base in Adelie Land. Of these six visits to the plateau detailed meteorological 'results are only available at present for Captain Scott's two journeys and for Amundsen's. It is only possible therefore to discuss the wind records in any detail for these three journeys, although from the descrip- tive accounts of the other journeys more or less fragmentary information about the winds can be obtained. Winds on the Plateau near the vxagnetic pole. Members of Mawson's expedition led by R. Bage * left their main base in Adelie Land on November 10, 1913, and proceeded towards the magnetic pole. Their journey took them * The Home of the Blizzard, Vol. 1, page 274. U2 WIND. in a S.S.E. direction, and tlic.y reached a height uf 3,6.JU feet in a little over :30 miles from the coast, after which they descended somewhat owing to crossing the Mertz Glacier valley. After passing the valley they regained their previous height and then slowly ascended to their furthest point in 70° 36' 8. and 148° 10' E. where the heiglit was just over 5.900 feet. In the description of their journey a great deal is said about the strength of the wind, but very little about the direction. This is not to be wondered at considering the strength of the wind which was so great as to be a constant handicap to their efiorts. While crossing the Meitz Glacier valley they experienced a wind from the 8.\V. which was obviously blomng down the valley. On November 24, while still in the depression due to the valley and moving to the S.E., the remark is made 'we were marching a little to the east of the wind,' so I gather that here the wind was between S.S.E. and S. They continued to encounter head winds, but there is only one more entry giving a definite direction. On December 5 ' for the first time on the trip the wind veered round to the south-east.' From this remark I conclude that the general trend of the wind was from the south. Until we have the meteoro- logical observations taken on this impor-tant journey we can only say that on the plateau between the magnetic pole and Adelie Land the wind is very strong and from some direction near the south. Professor David's Party, which a-cended the plateau to the magnetic pole from the Ross Sea side in December 1908, experienced on first reaching the plateau winds from tlie west, or slightly north of west. These were apparently winds blowing down the declivity which continued the valley up which they had ascended. As far as I can gather from Professor David's account as soon as the eiiects of the glacier valley were left behind the «ind gene- rally blew from some direction between south and south-east, and the sastrugi confirmed this to be the prevailing direction. Thus both David's and Bage's observations point to a southerl}' wind over the plateau near the magnetic pole, but that near the heads of the large glaciers the air motion is towards and down the glacier valleys. Winds on the Plateau to the west of Ross Islatul. Captain Scott was on the Western Plateau for 30 days reaching 78° S., 146° 30' E. The height of the plateau increased slowly from about 7,500 feet to a little over 7,700 feet at his most westerly point. The frequency of the winds observed is. shown diagrammatically in figure 47(11), page 134. It will be seen that the winds blew almost entirely from direc- tions between S.S.E. and W.S.W., the mean direction being S. 31° W. It is impossible to say whether the observations made on these 30 days can be taken as typical of the general wind conditions at this position, but as the sastrugi were mainly from the S.W. and occa- sionally from the W.S.W. it is probable that they are fairly typical. The chief result is that the winds on the plateau are from a more southerly direction than those revealed by the motion of Erebus smoke. When near to the edge of the plateau Captain Scott encountered both on going and returning stormy winds from the W.S.W., i.e., from the plateau into the head of the glacier valley. Winds on the South Polar Plateau. Amundsen reached the plateau on November 21, 1911, in 85° 36' S., 167° 42' W. and arrived at the Pole on December 17, he returned along the same track and left the plateau on January 5, 1912. Scott reached the plateau on December 22, 1911, in 85° 10' S., 159° 30' E. and the Pole on January 17, he left the plateau from the same position on February 4. Thus Ammidsen made wind observations on the plateau durii^ 44 days (November 22 to PLATEAU WINDS. 143 January 4 inclusive) and. Scott durinc; 4.5 days (December •22 to February 4 inclusive, observations are wanting on tlie .5th and 6th February). As a rule each observer recorded the wind direction and force three times each day, but occasionally they observed more frequently and occasionally less frequently. In order to obtain a regular series of observa- tions, when more than three observations were taken on one day only three of the recorded observations have been used, while on the few occasions on which less than three observa- tions were recorded probable values have been inserted when possible. In this way the series is homogeneous and complete except for two directions in Amundsen's series which it was impossible to supply. Scott's Party recorded wind forces on the Beaufort Scale, these have been reduced to velocities by the equivalent values given on page 130. Amundsen estimated the wind strength in metres per second, which have been converted into miles per hour. It is necessary to say a few words with regard to the directions used by the two observers. Scott approached the Pok^ almost e.vactly along the 160'' E. meridian while Amundsen approached it along a mean meridian of approximately 170° W. Now each observer naturally recorded a mnd as south when it blew along the meridian on which he was situated, thus for Scott a south wind blew along the 160° E. meridian and for Amund^en along the 170° W. meridian, thus the winds recorded as south by the two observers were inclined to one airother by 30°, and other directioias differed by the same amount. In other words a flow of air which was south for Amundsen was 30° to the west of south for Scott. In order to correct for this cliiference of designation of the same wind all the observations made on the plateau by Amundsen have been rotated clockwise through one compass point (27|°), thus making them agree very nearly with the direction which Scott would have experienced. This may be stated by defining a south wind on the plateau as air motion parallel to the 160° E. meridian with other directions related in the normal way to this main direction. The following tables summarise the chief results of the wind observations. Table 75. Wind on South Polur Plateau fS.=parallel to 160° E. merulmn). Z H Z H Z H a CO 0-7 30 11-30 U— 30 >30 Percentage frequency . Do. do. Do. do. Mean oloiul amount Cloud amount. ()4 21 34 30 36 37 27 36 21 30 49 S 18 74 0—1 2—8 9—10 3-0 -r(\ 4-9 (i-4 S-4 It is a noteworthy fact that during calms and light winds, 0— .5 miles per hour, clear skies and overcast skies occur with equal frequence and the mean cloud amount is practically 5. Moderate winds from the north have similar cloud conditions as those during calms. On the other hand, the effect of strong northerly winds is remarkable, for during 64 per cent, of the time they blow, the sky is practically cloudless, and the mean cloud amount is as low as 30. The effect of southerly winds is quite clear. Even with moderate winds the sky is overcast during half the time and clear only during one-fifth. With increasing southerly wind strength, the contrast becomes greater, and when the wind strength is greater than 30 miles an hour, the sky is overcast during 74 per cent, of the time and clear only during 8 per cent., the mean cloud amount reaching the high figure of 8-4. Annual Variation of Cloud. For reasons already given (page 149), the cloud amounts recorded on the Discovery Expedition were too low and therefore have not been used in obtaining the mean cloud amount for McMurdo Sound. There seems no reason, however, to doubt that the yearly variation is more or less accurately given by them and therefore they will be used to obtain this variation. Besides recording the actual amount of cloud, the observers recorded in their weather remarks the state of the sky according to the four grades b, be, c, and o. The former of these corresponds to the c'oud amount 0 and 1 and the latter to 9 and 10, there- fore they can be used in combination with the estimates of the frequence of these grades of cloud made at Cape Evans. Table 84. Monthly variation of Cloud Amount (0 — 10) from mean of the year. a u .a J3 1 =3 S3 t 4^ s I S a 1 a a 1 1902 .... 1903 . . . . 1911 .... 1912 . . . Mean .... +1-5* +0-9 -0-6 -11 00 +0-6 + 1-3 + 1-8 -01 +0-9 + 1-6 + 1-2 +0-5 -M + 11 -0-1 +0-3 -0-9 -0-6 -21 2-2 -0-4 -1-6 -0-7 -1-3 -2-6 -13 + 10 + 0-1 -0-6 -0-8 +0-9 +0-4 0-0 00 00 + l-;> + 1.7 + 1-1 + 1-3 0-0 +2-0 0-0 -1-3 -0-4 -0-8 -01 -1-3 5-6 45 6-2 64 + 0-2 + 0-9 + 0-9 + 0.1 -0-8 -1-2 -10 -01 + 0-1 + 1-2 +0-2 -06 1 904 ANNUAL VARIATION OF CLOUD. Table 85. 153 Monthly variation of percentage frequency of Clear Skies (b in 1902 and 1903 and 0 — 1 in 1911 and 1912) from mean for each two years. >^ b 1 ^ c £ a c8 i i >, 1 6 a a CD -C o s o a o 1902 and 1903 . -16 - 6 - 1 + 8 + 7 +20 + 17 +4 -5 -18 -17 + 4 mil and 1912 . . . Mean ..... + 4 - 6 -17 -15 -G + 13 + 19 + 7 +3 +2 -12 - 1 + 1 -12 - 8 + 1 + 10 + 19 + 12 +i -2 -15 - 9 +2 Table 86. Monthly variation of percentage frequency of Overcast Skies (0 (overcast) in 1902 and 1903 and 9 — 10 in 1911 and 1912) //-oH! wiertw o/ each turn years. >i u .13 0 1 fa a. v a 3 "5 "-5 1 3 a % o u o a o a 8 1902 and 1903 + 15 + 2 + 4 -10 2 -9 -14 -1 4-1 + 10 + 10 - 6 1911 and 1912 Muan -15 + 17 + 12 + 5 -12 -7 + 3 - 5 +5 — 1 +11 - 9 -12 0 + 10 + 8 _ 2 n -8 + 2 0 + 13 0 - 9 These values have been plotted on fignre 50, the curve for clear skies being inverted so that its variations may be compared directly with the other two. It will be seen that the annual variation of cloud amount is well marked, while its variations are followed very closely by the frequency of clear and overcast skies. Also an examination of the values for the individual periods given in the tables clearly indicates that the main variations are repeated in each. The curve of variation is a simple one : cloud is at a maximum near the equinoxes and at a minimum in both summer and winter. There is little difficulty in explaining the summer minimum. November, December and January are the fine weather months of the Antarctic. There are then relatively few blizzards and in their absence the Antarctic anticyclone produces clear unclouded skies. Ihe frequency of blizzards in February and March, to which attention has already been drawn in the section on wind, is directly responsible for the large increase then in cloud above the summer value. Winds of more than 30 miles an hour from the south are from three to four times as frequent in February and March as in November and December, and as these winds have on the average a cloud amount of 8'4, it is quite understandable that the average cloud for these months is high. Coming now to the winter months we at once find difficulties. During May, June and July blizzards are somewhat more frequent than during February and March, and they are accompanied by nearly as much cloud — the mean cloud during winds > 30 miles an hour from the south was 8-9 in February and March and 8-0 in May, June and July — but the average cloud amount is less than during the three summer months. We therefore have the paradox that the months with the maximum blizzards have the minimum cloud amount. It is clear that we must seek for some action which clears away the clouds as soon as the blizzards have ceased to produce them. Such an action was described by Sir Napier Shaw 3Q 154 CLOUD AND PRECIPITATION. in 1902.* Shaw pointed out tliat when thin clouds lose their lieat by radiation evaporation takes place and they disappear. The more rapid the radiation the more rapidly does the cloud disperse. Now during the \\-inter in the Antarctic the conditions are eminently suitable ff>r large and rapid radiation from the clouds. In the first place the clouds are very thin and Fig. 5(1. Annual variation of cloixl. therefore every part is able to radiate to the clear sky above, secondly they receive no heat from the sun and practically none from the ground. Thus the radiation from the clouds during the winter months is greater in polar regions than in any other part of the world. This explanation is strongly supported by figure 51. In this figure the mean cloud amount has been shown for each month during (a) winds from the south greater than 30 miles an hour, (6) winds from the south between 11 and 30 miles an hour, and (c) winds from calms to 5 miles an hour. The data used are September, 1911, to February, 1912, and March to August, 1911 and 1912, thus the most important part of each curve is based on observations from two years. Except in January the cloud amoimt with high southerly winds is nearly constant from month to month, showing that blizzards are the cause of cloudy skies through- out the year. The cloud amounts during moderate winds fiom the south and during calma * W. N. Shaw. ' La luno mange les miages.' Quarterly Journal of the Royal Meteorological Society, Vol. XXVin, page 9.5, 1902. DAILY VARIATION OF CLOUD. 155 ;iu(l light wiuds are iu all luoutlis less tliau during blizzards, but the diti'erence is very much greater from May to July than from October to March. During the former of these periods there is no direct sunlight and radiation is great, hence as soon as the wind drops the clouds disappear and the average amount of cloud during the intervals between the blizzards is only 10 miles/hour Jan. Feb. Mar.- Apl. May Jun. Jul. Aug. Sep. Oct. Nov. Dec. Jan Fig. 51. Annual variation of cloud according to wind, half that during the blizzards. On the other hand, during the period of nearly constant sun- light this force does not act, but the reverse holds, the radiation received by the clouds from the sun tends to increase rather than diminish them. Hence during these months there is only a small decrease in cloud amount in the intervals between the blizzards. We now see that the yearly variation of the cloud amount is a complex phenomenon, depending on two variables. (a) The frequency of blizzards which if acting alone would give a maximum of cloud in the winter and a minimum in the summer. (b) The amount of radiation received and emitted by the cloud particles which if it acted alone would give a maximum of cloud in the summer and a minimum in the winter. The combination of these two efEects is that minima of cloud occur in summer and winter and maxima iu the months near to the equinoxes. Daily Variation of Cloud. Observations of the amount of cloud were made at Cape Evans every four hours at midnight, 4 a.m., 8 a.m., midday, 4 p.m. and 8 p.m. As however station time was used, these observations must be put one hour earlier in local time. The record is not quite complete, 156 CLOUD AND PRECIPITATION. but it has been possible to interpolate the missing observations between 1st March, 1911, and the end of October, 1912, so that the daily variation has been obtained for twenty months, with a fair degree of accuracy. A casual glance at the results for individual months shows that one or even two years' observations are quite insufficient for giving a true value of the variation. The data for the individual months have, therefore, been given in the volume of tables, and we shall consider here only the results obtained by combining the new data with those obtained at Hut Point during the Discovery Expedition. Unfortunately in the discussion of the Discovery results the monthly mean cloud amounts at the different hours are only given in whole nunibsrs. As the daily variation in most months is of the order of one unit, this rounding makes it impDssible to combine the results there given with the new series. It was therefore necessary to recalculate the whole of the cloud data for the two years' observations at Hut Point. These new values will also be found in the volume of tables. On the Discovery Expedition observations were taken at the even hours of local time, therefore to combine the results with those of Cape Evans, which were taken at the odd hours of local time, it has been necessary to take the mean value between the successive observations at Hut Point. Thus the values for 7 a.m. local time at Cape Evans have been combined with the mean values for 6 and 8 a.m. at Hut Point, and so on. The following table gives the departure from mean for the mouths and seasons based on all the available data. Table 87. Daily variation of Clotid Amount [Hut Point and Cape Evans). Month. No. of years. 3 hours. 0-0 7 iours. 11 hours. 15 hours. 19 hours. 23 hours. Ampli- tude. 0-5 January 2 +0-1 -0-1 -0 2 -0-1 +0-3 February 3 -01 + 0-2 0-0 +0-2 -0-5 +0-1 0-7 March i 0-0 00 -0-2 -01 +0-3 -01 0-5 April 4 -0-2 + 1-0 + 0-6 +0-2 -0-6 -11 2-1 May 4 -0-4 + 0-2 +0-6 + 0-4 -0-3 -0-4 10 June 4 -0-2 + 0-2 0-0 +0-1 -0-2 00 0-4 July 4 -0-2 +0-1 +0-4 +0-4 -0-1 -0-6 10 August 4 -0-7 +0-3 +0-2 +0-2 0-0 -0-1 10 September 4 -0-4 +0-2 +0-1 +0-9 0-0 -0-8 1-7 October . 4 0-0 0-0 -0-2 +0'1 +0-2 0-0 0-4 November 3 -0-1 -0-4 0-0 +0^0 +0-2 -0-3 1-0 December 2 + 0-6 +0-7 +0-1 -0-4 -0-5 -0-5 . 1-2 Sea sons. November-January . +0-2 +0-1 0-0 0-0 -0-1 -0-2 0-4 February-April • • -0-1 +0-4 +0-1 +0-1 -0-2 -0-3 0-7 May-July .... -0-3 +0-2 +0-3 +0-3 ^ -0-2 -0 3 0-6 Auguot^October •• -0-4 +0-2 +0-1 +0-4 +0-1 -0-3 0-8 0-.5 Year • . •• -0-1 +0-2 +0-1 +0-2 -0-1 -0-3 Daily variation of cloud. 157 Autumn Tlie values for the ueasoiis and year have been plotted on figure 52. In all seasons except summer the maximum cloud amount is recorded at either 7 hours, 11 hours or 15 hours and the minimum at either midnight or 3 a.m. In other words, except in summer, the maximum occurs during the day hours and the minimum during the night hours. Unfortunately the measurement of cloud amount is ■ not made by instruments and tlie estimates depend partly on subjective factors. Any one who has tried to estimate the amount of cloud at night is aware of the difficulty of coming to a light decision. Even the casual observer must have often been surprised to see clouds clearly visible on what ho thought to be a cloudless sky when a distant lightning flash has lit up the hori- zon. A long and careful survey of the siiy is necessary to detect clouds on a moonless night, and even then an experienced observer may miss some. Now in the Antarctic all the observers were not experienced, and also the conditions on a cold windy night are not conducive to a deliberate survey of the sky to detect small masses of cloud. On the other hand, when there is day- light even an inexperienced ob- server taking only a rapid glance at the sky can form a very good estimate of the amount of cloud. Thus the amount of daylight plays an important part in the estimate of the amount of cloud. If there was the same average amount of cloud throughout the day and night, the eflect of the daylight would be to produce a fictitious daily variation having a maximum during the hours of diylight and a minimum during the hours of dark- ness, I.e., exactly the same variation as is found during all seasons in the Antarctic except summer, and we must therefore attempt to decide in how far the observed variation is affected by the variation of dayli"-ht. There are three periods to be considered. {a) When there is sufficient daylight throughout the twenty-four hours for the clouds to be clearly seen. For our purpose this may be taken as the whole period between which the sun does not sink more than 10° below the horizon, i.e., in McMurdo Sound th'i period from September 30 to March 14. Thus the months October to February are included in this period, and March may also be added as observations were taken at 1 p.m. local time which extends the jjeriod slightly. 23 23 Fig. 52. Daily variation of cloud. 158 CLOUD AND PKECIPITATION. {!)) The period during which the sky is too dark for clouds to be seen by the aid of the sunliglit. AVe may take this period as lasting during the time the sun does not rise within 10° of the horizon, i.e., in McMurdo Sound the period from June 5 to July 8. Thus the month of June is the only montli which is included in this period. (c) The remaining period, during which there is daylight during part of the twenty-four hours and darkness during the remainder, i.e., in McMurdo Sound the periods from March 14 to June 5, and from July 8 to September 30. Thus April and May, July, August and September may be considered as belonging to this period. Now during the periods (a) and (b) daylight cannot affect to any appreciable extent the observations of cloud at any time of the day, while during the period (c) it affects the observations to its maximum extent. In figure 53 the amplitude of the daily variation of cloud has been plotted for each month, and we see that during the period (((), October to March, the amphtude is very small in five of the seven months. Also in the period (6), June, the amplitude is very small. On the other hand each of the two periods into which (c) is divided has a month of maximum amplitude. This alone would make us strongly suspect that the day- light plays a largo part in deter- mining the apparent daily varia- tion of cloud amount ; but when we find that almost exactly the same variation of the amplitude during the year is shown by the observations made on the Fram drift when the variation of day- light were nearly the same, the suspicion becomes almost a cer- tainty. Curve (6) of figure 53 shows the amplitude of the daily variation of the cloud for each month as observed on the Fram, the months being arranged so that the same seasons in the north and south correspond. The similarity of the curves for McMurdo Sound and the Fram drift is strilving. In the north the months of constant daylight, August and September, have very small amplitudes and also the winter months November * and December. Between these two minima there are large maxima during the period when each day has a period of daylight and total darkness. It is therefore very questionable whether the curves in any month except those of complete daylight and complete darkness give any information about the yearly variation of 20 I \ 1-0 / \ \ / \ r 7 / \ 1 / V 0-5' A \ / ^ L \ 1 \ 00 J M'=M urdo an. r sb. M ar. A pi. M ay. Ji jn. Ji >\. A u^. S ep. Oc t. N M. 0 EC. Jan 15 10 0-5 00 J V / K A \ \ y V \ / r \ / \ / ^ 1 [ rami leim \ \ Fig. 53> Amplitude of daily variation of cloud. PRECIPITATION. I59 cloud, and in these months the amplitudes are so small that no reliance can be placed on them. In fact the summer variation based on the Hut Point observations is almost the exact reverse of that based on the Cape Evans observations, showing that the small amplitude is mainly the result of chance. We are therefore reluctantly compelled to conclude that the daily variation of cloud amount teaches nothing reliable as to the true changes in the cloud amount durino- the day except that it is very small, so small in fact that during the gi-eater part of the year it is masked by the subjective effect of the varying amount of daylio-ht. Precipitation. No rain fell during the whole of our stay at Cape Evans, and up to the present there is no record of rain at either Cape Adare or in McMurdo Sound. On the other hand, a well- developed rainbow was seen to the north-north-east of Cape Evans on February 14 1911. Every one who has taken meteorological observations in the Antarctic has been faced with the difficulty of finding some way to record the amount of snowfall. Up to the present this problem has not been solved. The difficulty is that practically all the snowfall occurs during high winds. The snow is carried along in the air and does not settle except in the lee of any obstacle to the winds motion. Any ordinary snowgauge would become full of drift snow in a very short time, while the surrounding ground might not have received any increase to Its snow-covenng. Also it very frequently happens that with a perfectly clear sky from which no snow is falling the wind raises clouds of loose snow from the ground which enters the gauge and so gives a false record. I had many ideas as to measuring the effective snowfall on my arrival in the Antarctic, but none of them proved of any practical use. It was no use trying to measure the depth of snow on the ground, nor to keep a space clear and measure the snow on it each morning. For no snow accumulated on ground subject to the full force of the wind, while in sheltered places the snow would accumulate to a depth determined entirely by the size and shape of the obstacle, and be entirely unrelated to the quantity of precipitation. My experience is that it is totally impossible to form even an approximate idea of the quantity of snow precipitated from the atmosphere. Even on a perfectly level plain the snow accumulation would be no gauge of the precipitation, for it would receive snow which had fallen miles away and been driven forward by the wind. All that is possible in this section is to give a general statement of the snow-covering existing around the station at different times of the year, and to give some idea of the frequence with which snow fell. When we arrived at Cape Evans, the small promontory on which the station was built was— as future experience showed— remarkably free from snow. The beach on which the hut was built had no snow-covering at all, so that there were several hundred square yards of black volcanic debris fully exposed to the warm sun. When the expedition left twenty-four months later, the beach was buried in several feet of snow and the hut itself was almost invisible under a great snow-drift which had formed around it. After our arrival in January, 1911, there was a slow but steady diminution in the snow and ice accumulations on the Cape, so that the size of the small glacierettes obviously de- creased. This was due mainly to ablation, for it continued throughout the first winter when thawing was quite imposible. A cave had been dug for the magnetic instruments in a small glacierette— in reality the permanent drift which had consolidated into firm ice in the lee of 160 CLOUD AND PRECIPITATION. Windvane Hill — and the roof of this cave became so thin that towards the end of the summer ot 19 12 I began to have serious doubts as to its safety. Thus from January, 1911, to the end of February, 1912, there was a marked diminution in the snow and ice accumulations on the Cape. When the sea froze in March, 1911, a tb.in deposit of snow accumulated on the sea ice to a depth of one or two inches, this never increased all through the winter and when in September we made a sledge journey across the Sound the only places where we could find sufhcient snow for erecting the tent were the drifts in the lee of broken ice. The blizzard at the end of September (2.5th to 28th) left large drifts about the hut and hills, but no accumulation of snow on the sea ice. During October and November there was a great loss of snow. The drifts near to the hut diminished and all the snow disappeared from large tracts of the sea ice. From Cape Evans to Cape Barne on November 29, 1911, the floe was entirely clear of snow exposing the smooth blue sea ice. Thus in the neighbourhood of Cape Evans the total snowfall for eleven months had produced no addition to the snow accumulations, but on the contrary the permanent snow- drifts had become visibly smaller. On Decomber 6, 1911, the first heavy snowfall unaccompanied by high wind was ex. perienced. This left 18 inches of snow on the floe which, in consequence of a high northerly wind on the 8th, settled down to a depth of about i inches over all the frozen sea in our neighbourhood. At the end of this snow-storm the accumulation of snow about the hut was the greatest we had experienced ; large drifts covered most of the stores around the camp, and large tracts of the hill which had always been clear before were now buried under snow. From this time on successive blizzards added to the accumulation and 1912 was just as marked for the increase of snow as 1911 had been for its disappearance. Onlv during the summer months was there anjiihing equivalent to the snowfall with which we are familiar in temperate regions. The snow fell on December 6 in ordinary flakes, but at other times of the year flakes were never observed. The snow was chiefly in the form of small grains showing little crystalline structure, but occasionally crystal stars would fall. In the absence of wind the snowfall was always light, and from March to October the around never received more than a mere sprinkling of snow in the absence of blizzards. Durin" blizzards the air became filled wtli snow literally in the form of dust. It was then practicaDy impossible to say whether new snow wa,s falling or whether the whole mass of snow in the air was simply fallen snow carried along by the wind. In the meteorological log we have three kinds of entries referring to snow and drift. First the meteorological symbol for snow is entered, to which is almost always attached the remark ' slight snow ' or ' very little snow falling.' Secondly, the two symbols for snow and drift are entered together, but very frequently the sign for snow is queried, or the remark is added ' probably snow,' thus showing the great difficulty of deciding whether new snow was fallin" or not. Finally, there is the symbol for drift alone, and to this the remark ' surface drift ' is sometimes added. In the latter case, there would certainly be no new snow fallinc, but it is impossible to say how often the symbol for drift refers to drift with and without snow, hence the records of drift include drift with and without snowfall. From March, 1911, to October, 1912, observations were taken every four hours and the number of time the symbols for snow, snow with drift and drift were recorded have been counted. In the follo\ving table these have been expressed for each month reduced to one hundred observations. The numbers therefore in the table give for each numth the average frequency with which each sign was recorded in one hundred hours. PRECIPITATION. Table 88. Number of times Snow and Drift recorded in 100 hours. 161 Moutii. Snow. Snow and Drift. Drift. Total. 1911. March ....... 10 ■t 9 23 April 22 6 7 35 May 9 10 li 25 Juae ....... 6 4 2 12 July C, 7 9 22 August ....... 5 2 13 20 September ...... 4 2 11 17 October 9 11 15 35 November ...... 7 5 1 13 December ...... 8 4 12 24 Ifil2. January 11 1 o 14 February 11 12 11 34 Marcli . 13 7 ir, 30 April 8 2 12 22 May 11 8 8 27 June 2 24 24 50 July ....... 10 28 22 Oil August ....... 10 18 20 54 September 1 1.5 2fi 42 October 8 10 18 i 30 From the purely human point of view this table shows the disagreeableness of the climate in McMurdo Sound. Out of twenty months there were only four in which there were less than 20 per cent, of hours with snow or drift. In twelve months the percentage of snow and drift was more than 25 per cent, and in three it was over .50 per cent. 21 162 CLOUD AND PRECIPITATION. Tlie great excess of bad weather during 1912 over 1911 is clearly indicated. From March to October 1911 there wore 2-5 hours out of every hundred with snow or drift ; in the same months of 1912 the percentage was 41. Collecting the results according to seasons we have the following : Table 89. Number of times Snow and Drift were recorded in 100 hours Season. Snow. Snow and Drift. Drift. Total. Spring — August to October Summer — November to January Autumn — February to April Winter — May to July .... Year 6 9 13 7 10 3 6 14 18 5 11 12 34 17 3(1 33 9 8 12 2!l The summer has the least frequency of snow and drift, while the other seasons have practically the same. For the summer we have only observations for one year, but the ex- pedition was at Cape Evans for another, and there is little doubt that the relative freedom from bad weather found during November, 1911, to January, 1912, was repeated during these months in other vears. CHAPTER V. PRESSUEE. Instruments and methods. Barometers. — The following table shows the barometers used on the expedition and their corrections as determined at Kew before leaving England. Table 90. Barometers. Where u.sed. Pattern. Number. Kew correction- Ship Kew 1103 -•003" Cape Evans (general u.se) Kew 1137 -•010" Capo Evans (for comparison) Fucss 1667 -010" Cape Adare (general use) Kew 1156 -■012" Cape Adare (for comparison) . Fuess lOOS ^■010" Owing to the danger of air entering the barometers during the journey and before they could be set up at the land stations, the two Fuess barometers were taken, because with these instruments it is possible to detect and correct for any air which might enter. These barometers are not, however, convenient for general use, it was therefore intended to use them only for checking the Kew barometers, which were to be used for taking the actual observations. .Just before the Terra Nova left Lyttleton, in New Zealand, all the barometers were set up in a room on shore and compared. About twelve readings of each barometer were taken, and the results were compared after each reading had been corrected for temperature and instru- mental error as determined at Kew. As it is impossible to say that any one of the instru- ments had undergone no change, the difference is given in table 91 between the barometer shown at the end of each line and the one shown at the head of the column. Table 91. BaronKter Comparison in Lyttleton. 1163. 1157. 1667. 1156. 1668. 1163 , , + •015 + •019 + ■(120 + 017 1137 -015 .. + •004 + ■005 + -002 1607 -010 -•00.J: .- + 001 — 002 um -•020 -005 — ■OOl • • -■003 1668 -•017 — •002 + ■002 + ■oos •• 164 PRESSURE. From this table it will be seen that at Lyttletoii all the barometers agreed with one another except the ship's barometer, No. 11G3, which appears to have been reading about + 018" too high. After this it was impossible to make another general comparison, but ' it is possible to show that the instruments when in the positions in which they were used were in good order. The following is the evidence for the different barometers. Cape Evans' Barometers, No. 11 D7 and No. 1667. — The two instruments were compared on five occasions during the two years, mth the following result, after their Kew corrections have been applied. Table 92. Barometer Comparisons at Cape Evans. Date of comparison. 11.37-11)07. August 21, 1911 •ooa January 11, 1912 IKKI May 13, 1912 — •001 October 1, 1912 ■000 ■January 1, 1913 -003 From these comparisons it is safe to say that both barometers were in good order after landing at Cape Evans. Cape Adare Barometer, No. 1156. — Unfortunately the Fuess barometer which was taken to Cape Adare for purposes of comparison was broken in landing. Therefore no comparisons could be made after setting up the Kew barometer, No. 11.56. On the return of this barometer to Lyttleton it was compared with the standard barometer in the observatory there (No. 1121, made by Hughes, London) and found to be exactly correct. It is tolerably safe therefore to assume that it was correct while in use at Cape Adare. This conclusion is made quite certain by the following observations. On Jamiary 3, 1912, the Terra Nova reached Cape Adare at nine o'clock in the morning to take away the party. The observation made on the ship at 8 A.M. was 29-54", while that made at the same time at Cape Adare was 29-537", thus showing that the Cape Adare barometer was in good order. The Ship's Barometer, No. 1163.^ — This barometer was compared at Kew before and after the expedition, the comparisons giving — -005 and 000 as the corrections, respectively. From a comparison, however, with the other barometers in Lyttleton and with the Cape Evans' barometer in the Antarctic, this barometer appeared to read -020" too high while used in the Antarctic. The barometer, however, was of the ship's pattern and only graduated to 010", therefore a very high degree of accuracy is not to be expected. In reducing the readings of this barometer I have used a correction of — -010 as being the mean of the corrections found in England and in the Antarctic. Framheim Barometer. — When the results of the pressure observations at Cape Evans and Framheim were first examined, it caused great surprise to find that the pressure at Framheim was so much lower than at Cape Evans. Framheim is further south than Cape Evans and has a much lower average temperature, therefore for both these reasons one expected the pressure at Framheim to be the higher. Although there was no reason to doubt the accuracy of the barometers at either Cape Evans or Framheim, it was felt that a comparison of the INSTRUMENTS AND METHODS. 165 barometers at the two stations was desirable. Unfortunately no direct comparLson was then possible, but that there was no appreciable instrumental error could be determined indirectly, as follows. At the end of January 1911 the Terra Nova left McMurdo Sound to proceed to King Edward VII Land, where it was intended to establish the eastern pai-ty. When she was in McMurdo Sound her barometer was compared with the one in use at Cape Evans. As she proceeded eastwards the ship's barometer fell continuall}' compared with tha readings at Cape Evans, thus confirming the fall in pressure over the east of the Ross Sea. As is well Icnown the Terra Nova encountered tbe Frain in the B-iy of Whales and during Febraary 4, 191], the two ships were near to one another. From thiir meteorological logs the following com- parison of the barometer readings is possible. Table 93. Barometer Comparison. Ciipj Evans. Terra Xoca. Fraiit. February 4tli, 1911, i hours . 29 29 28-91 28-90 8 hours . 12 hours . 29-29 29-00 2900 16 hours . 29-28 29- 00 28-99 20 hours . 2926 28-90 28-99 Mean 29-28 28-97 28-98 From this it will be seen that the two ships' barometers were in satisfactory agreement while the Cape Evans' pressure was -30" higher than the simultaneous pressure in the Bay of A\Tiales (Framheim). This result, so far as it goes, confirms the relative low pressure near Framheim, but as the barometer on the Fram which is shown to agree with the one on the Terra Nova was not the one which was subsequently used at Framheim, the desired comparison of the Cape Evans and Framheim barometers was not effected. This final comparison, however, is possible from the log of the Fram and the records of Framheim during the period in January 1912 when the Fram was waiting in or near the Bay of Whales for the return of Captain Amundsen's party from the South Pole. From January 14 to January 27, 1912 (inclusive), there were simultaneous observations on the Fram and at Framheim and these with all corrections applied agree within -2 mm. of each other, the Framheim barometer reading being the higher by this amount. Thus the Framheim barometer, through the intermediary of the barometers on the Terra Nova and the Fram, was compared with the Cape Evans' barometer, and the comparison shows that, if anything, the Framheim barometer read slightly higher than the Cape Evans' barometer, which would tend to reduce and not increase the large pressure difference foimd between the two stations. Method of Reduction of the Observations. — At Cape Evans twp Richard barographs were in use from January 27, 1911, to the end of December, 1912. On January 24, 1911, the barometer was set up, but owing to the large amount of work entailed in setting up the magnetic and meteorological instruments it was impossible at first to take more than one set of meteorological observations each day. On March 2, 166 PRESSURE. 1911, however, barometer observations were commenced at intervals of four hours. The.se were continued ahnost without break until the end of October 1912. In November 1912, owing to nearly all the members of the exjDedition taking part in the different search parties, the barometer observations became irregular, but on every day the barometer was read at least once and on most days three or four times. The eye readings of the barometers were reduced to sea-level and constant gravity. The barograph traces were measured at each hour and the readings tabulated. Against the corresponding barograph measurement each reduced barometer reading was entered, and the difference taken. Ihese differences were then plotted on the actual barograph sheet to a scale 10 times larger than the barograph scale. As most of the changes in the difference were due to lag in the barograph, it was found that the curve of differences was very similar to the actual barograph curve. This similarity made it very easy to draw the curve of differences and is a great advantage of this method of reducing a barograph trace. When the curve of dift'erences had been drawn, it was read off at every hour and the value entered against the corresponding value from the barograph curve. The barograph and difference values were then added together, giving the correct barometer reading at each hour. These measurements were all made to jjjyth of an inch, which is the greatest accuracy possible with a barograph. The mean values of the month from the barograph reductions for each of the hours at which the barometer was read were found to agree to -fj^jjy-th of an inch with the mean of the barometer readings. During the period when barometer observations were not taken every four hours the reduced barograph records are correspondingly less accurate, but there is no doubt that even in these periods the mean monthly value is as accurate as the barometer correction, and the mean daily variation for the month is probably as accurate as could be obtained with the full set of observations, for this element is more affected by the non-periodic pressure changes than by the method of reduction. A similar method was used in reducing the Cupe Adare barograph record, but as during the greater part of the year eye observations were not taken during the night, the reduction is not so perfect. It was therefore considered sufficient to reduce the Cape Adare barograph trace only at every second hour. During a few periods all the members of the Cape Adare Party were away from the station for several days. The values of the pressure for these days have been obtained whenever possible from the barometer observations taken in the field- Mean pressure and annual variation. McMurdo Sound* The following table contains the monthly values of pressure at McMurdo Sound for the four years for which data are now available. • * Since this chapter was finished ii,nd ready for the press, Mr. Mossman has very kindly drawn my attention to the fact that while the pressure data contained in tables 1 and 3 of chapter XI, pp. 476-482 of the volume of results of the Discovery Exi)edition have been reduced to gravity at latitude 45°, those in table 2 have not. As there is nothing to point this out in the chapter itself, I had used the mean monthly pressux'es given in table 2 under the impression that they were fully reduced. Thus half the data used in this chapter was wrong by -070 inch. This necessitated recalculating most of the tables of pressure and also altering many of the diagrams. All the tables have been corrected, but it was too late to alter all the diagrams. Tho.se diagrams which were materially affected have been corrected, but the remainder have been left unchanged. Thus in a few diagrams the plotted points do not agree with the figures in the tables, but in none of these cases is the character of the diagram affected- One cannot too strongly deprecate the use of pressure data, some of which are fully and the remainder only partially corrected. MONTHLY MEANS. Table 9i. Monthly pressure in McMurdo Sound. Reduced to 32°F., sea-level and gravity at 45°. (Inches. IG7 Year. $ 1 >> 3 j3 >> 6 a 1-5 1-5 +3 3 < a a 6 g s & 1902 . 1903 . 1904 . 1911 . 1912 . Mvan Smootlied 9-461 9131 9-296 9-431 9 327 9-298 9-308 9-. 527 9399 9-488 9-211 9-171 9-478 9-359 9-317 9-354 9-344 9-227 9-227 9-151 9-315 9-. 502 9-110 S-S82 9398 9-169 9-078 8- 998 8-911 9215 9188 9-132 9384 9 059 9-156 9-442 9-078 8995 8-825 9-090 9-634 8 947 9-630 9-201 9-505 9-294 9-7.52 9188 9-353 (a) 9-229 (b) 9-258 9-219 9-330 9-365 9-317 9-377 9-237 9-217 9161 9-112 9-260 8-997 9-368 9-435 9-265 9-365 9344 9-344 9-327 9-267 9-208 9163 9161 9-157 9-155 9-292 9-392 9-265 (a) Feb. 1902 to Jan. 1903. (b) Feb. 1903 to Jan. 1904. The above values have been plotted in figure 54 and the months of the same year connected up by a characteristic line. The resulting diagram is exceedingly instructive showing as it does the tremendous variations of the pressure for the same months from year to year compared with the normal yearly variation. It also shows the impossibility of forming even an approximate idea of the yearly variation from one year's observations. Thus the year 1903 showed the maximum pressure in June while in 1912 the minimum pressure occurred in this month. Similarly November was the month of maximum pressure in 1912 and the month of minimum pressure in 1903. When the mean values of each month for the four years are taken the irregular changes from month to month are still large (thin curve in figure bHa) and it is clear that many more years would be required before a smooth curve showing the true yearly variation would be obtained. When one has to deal with a curve subject to such large accidental variations it is legitimate to endeavour to eliminate them by smoothing the curve. I have tried several methods of smoothing this curve and find that the most satisfactory result is obtained by applying the formula b=-^— | — -■ The numerical results are given in the last line of table 94 and the thick line in figure b5a shows the resulting curve. In its main features this curve is probably correct. The numbers given in the last line of table 94 have therefore been taken throughout this work as the normal values of the monthly pressure in McMurdo Sound. The result that the pressure is lowest during the cold winter months and highest during the warm summer months is very surprising, but there can be little doubt of the reality of the phenomenon. 168 PRESSURE. 28-80 29-70 fA^M URDO Sound Pres SURE " — ---19C ....19C IQI 12 \ 1 2 \ 191 29-60 \ I \ o c / 1 1 \ \ 29-50 / \ \ / \ \ / 1 \ 29-40 ^ I i > s ': •. / •/ \ \ \ \ / \ 29-30 / : r . / / \ A 1 • \ ^ /: / ■ 1 / ; \\ 1 \, . \ \ / / . \ / \ ' \ / / ^ \ 1/ \ 29-20 •^\ \ \ / f \ \ 1 / \. \ \ 1 1 / 1 / / 1 \ \ 1 ^ '- \ \ 1 / . \/y'i 1 M i ■. 2910 \ 4 29-60 Cape Evans ■•/ IT; "ram he 'm w I / vt 1 i (J c I I 1 •f / , J I-' •1 1 1 \ \ f — \^- \ 1 1 \ \ ./ ■f ■! 1 \ W / / ; ii V. r '■\^1 / J \ 1 \ \. / \ \ 1 s ^ \ \ / \ 1 (1! ■'!'■ 28-90 \ ' — / % \ \ 1 \ \ 1 1 28-60 . March April May June July Aug. Sep. Oct. Nov. Dec. Jan. Fig. 56. Mean monthly pressiue, lyU-lL*. MONTHLY MEANS. 171 Coiiiparisoa of the fresmrc at Cafe Evans, Framheim and Cape Adare.-^The following table gives the simultaneous mean monthly pressure reduced to sea-level and gravity at latitude 45° at Cape Evans, Framheim and Cape Adare. Table 95. Mean monthly pressure. Reduced to 32°F., sea-level and gravity at 45°. (Inches.) Munth. Cape Evan?. Framheim. Cape Adare. Caps Evaus -Fram- heim. Cap-: Evans -Cap'j Adarc. 1911. March 29-21 2912 •- ■09 April 29-32 29-13 29-25 -19 -07 May 29-23 29-07 29-06 -16 ■17 •Tuni' 29-11 28-92 2911 -19 ■110 July 2908 28-91 29-01 -17 ■(18 August . 29-19 28-98 2907 -21 ■12 Septi-mbor 29-16 28-94 28-98 -22 ■18 October . 28-82 28-65 28-76 -17 •06 Novemlicr 29-63 29-53 29.56 -10 ■07 December 29-75 29-70 .';9-73 -05 •02 1912. Janimry . 29-43 29-40 -03 These values are plotted on figure 56 and it will be seen at once how closely related are the pressure changes at all the three stations. The series contains only ten months' observations for Framheim and Cape Adare and these, extending over less than a year, are not sufficient to give either the mean pressure for a complete year nor the yearly variation, but by the aid of the close relationship between the pressure at the three stations it is possible to amplify the data, and in the following two sections approximate values of the normal pressure and the yearly variation will be obtained for Framheim and Cape Adare by the aid of the more complete data for McMurdo Sound. Framheim. The pressure differences between Cape Evans and Framheim for each month are entered in column 5 of table 95. There is a large yearly variation in the pressure difference, the difference being largest in September and least in January. When the differences are plotted (curve I, figure 57 ) it is seen that they lie approxiinately on a curve which is much more regular than the curves of the pressures themselves. This curve is so regular, being almost a sine curve, that it is reasonable to conclude that the two missing values for February and March if they had been available- would have been on or near the curve. We may therefore use the values given by the curve for supply- ing the missing data. According to the curve the diSerence between the pressure at the two stations in February and March is 03" and 05" respectively. Applying these differences to the pressure for February and March 1911 at Cape Evans, we obtain the following values for a complete year's observations at Framheim. 172 PRESSURE. April May June July Aug. Sep. Oct. Nov. Dec. Jan. Feb. March April May June July Aug Fig. 37. Pressure cUffeiences. Table 9G. Pressure at Fr 1 Difference ■aa •03 •10 •15 •19 •20 ■2u •IS -15 -10 -05 -12 Framheim 29-33 29-31 29-29 29-23 29-12 29^02 28^ 96 2S^96 28-98 29- 00 29-19 29-34 29- U The yearly pressure variation for Framheim given in table 9G and tabic 97 are shown in figure 55c. It will be seen that the variation for the single year, thin curve, has the same general characteristics as the variation obtained from the four years' observations in McMurdo Sound, thick curve, the chief difference being in the high values of the pressure during the months of November, December 1911 and January 1912. Either curve shows that the yearly variation is very similar at both McMurdo Sound and Framheim, but that the latter has the larger amplitude. From a physical point of view the moet interesting feature of the pressure relationship between McMurdo Sound and Framheim is that the pressure difference is not that which would be expected from the temjierature relationship. At both t-tations the pressure is least in the coldest months. Further Framheim is colder than McMurdo Sound and its mean pressure is less, also in the months in which the temperature difference between the two stations is the greatest the pressure difference is the greatest. Such a relationship would be impossible if the density of the air only controlled the pressure ; we shall show later that the solution of the apparent paradox is to be found in the motion of the air as constrained by the Western Mountains (see page 238). Canpe Adare. In addition to the pressure values for Cape Adare given in table 95 further data are available from the observations made in 1899-1900 by the Southern Cross Expedition. For the eleven months, March 1899 to January 1900, we have pressure observations made twice a day, at 9 a.m. and 9 p.m. ; although these are not so good as observations taken more frequently they give values for the mean monthly jwessure which are not far from the true values. We have then two separate series of pressure data, the first consisting of eleven monthly values and the second of ten monthly values. It is possible to extend these two series into complete years with a considerable amount of certainty. Consider first the second period which commences with March and ends with December 1911. From figure 56, page 170, it will be seen that during the period when simultaneous observations were taken at Cape Evans, Cape Adare and Framheim the mean monthly pres- sure at Cape Adare, with one alight exception, lay always between the values for Cape Evans and Framheim. It therefore seems reasonable to conclude that the pressure for January 1912 at Cape Adare lay somewhere between the values which wo have for that month at Cape Evans and Framheim. As it happened these values were near together, being 29-43" and 29'40" respectively. It is almost certain therefore that the pressure at Cape Adare during 174 PRESSURE. that month could not be far from 29-42". Taking this value \vc have now two complete ■sets of mean monthly pressure from March to January. February is therefore the onlv missing month, and we cannot do better tlian take for this jnontli the mean of t-lu^ values for .Tanuarv and March. Completed in this way we have the following two years' pressure data for Cape Adare, the values not actually observed being entered in brackets. Table 98. Mean monthly 'pressure at Cape Adare. Reduced to 32°F., sea-level and gravity at 4.5°. (Inches.) s 1 5 3 p. < 1 3 4 3 be < S c c o c c o > ll o 1899-191)0 . 29-26 (29-24) 29 14 29-35 29-08 29-02 29-46 29-07 28-69 28-94 29-40 2931 29-16 1911-1012 . ;29-42 (29-24) 2912 29-25 29-00 29-11 29-01 2907 28-98 28-76 29-. 50 29-73 29-19 Mean .... 29-34 29-24 29-13 29-30 29-07 29-06 29-24 29-07 28-84 28-85 29 48 29.52 29-18 Smoothed 29-36 29-24 29-20 29-20 2912 29- 11 29-15 29-06 28-90 29-00 29-33 29-40 2918 In addition to this determination of the yearly pressure variation at Cape Adare based on two incomplete years of observation we are able to reduce the four years' observations at McMurdo Sound to Cape Adare by means of the ten months' simultaneous observations in 1911. The monthly differences Cape Evans — Cape Adare are given in the last column of table 95, page 171, and they are plotted in curve II of figure 57.* It will be seen that the points lie on a curve similar to the one found for Framheim, but that the amplitude is much less. The first line of the following table gives the values of the smoothed difference measured from the curve and in the second line these differences have been applied to the normal pressure of McMurdo Sound to obtain the normal pressure at Cape Adare. Table 99. Normal pressure at Ca-pe Adare as determined fro)n four years' observations in McMurdo Sound and I lie smoothed difference found in 1911. Reduced to 32°F., sea-level and gravity at 45°. (Inches.) >. o a; c3 1^ 1 < 1 2 >, < p o o c G 1 s rt > Difference . •03 -03 -05 ■07 -09 -10 •10 •10 ■09 -07 -06 ■04 -07 Cape Adare . 29-33 29-31 29-29 29-26 29-18 2911 29-06 29-06 29 07 2909 2923 29-35 2919 The values for the yearly variation found by the two methods, table 98 and table 99, are shown in the thin and thick curves, figure 55(6). The two curves are similar, varying appreciably only in July and tlie summer months. The liigh value on the thin curve for .* The differences for May and June have been combined as individually they lie ratlier far froui tlie c-uive, but comliini'd fit -well witli the other points. GEOGRAPHICAL DISTRIBUTION. 175 July is due to the abnormally high July pressure fouud in 1899, and the high summer values are due, as in the case of Framheim, to the unusually high pressure during the summer of 1911-12. Summary. — From this discussion of the monthly valu,iS of pressure at the three Ross Sea stations we see that whether we take only the actual observations made at Framheim and Cape Adare or the values for those stations deduced from the four years' observa- tions at McMurdo Sound we arrive at the same general type of annual pressure variation at all three stations. There seems little doubt therefore that in the Ross Sea area pressure is highest in December and lowest in September or October. The pre.ssure falls fairly regularly from December to July, remains more or less constant until October and then rises very rapidly to its maximum in December. The annual variation of pressure has been determined in other parts of the Antarctic, but except at Snow Hill wh.ere observations were made for twenty months, the data are only for twelve months or less at each station. It has already been pointed out that the yearly variation based on a single year's observations in the Antarctic can lead to very erroneous conclusions. It would therefore be unprofitable to compare the individual results from other stations, and a general comparison can best be made when the meteorology of the Antarctic as a whole comes to be reviewed. Geographical distribution of pressure. We have now determined the mean annual pressure at the throe stations in the Ross Sea area to be : — Cape Evans 29-26" Cape Adare 29- 19" Framheim 29-14" The winds at Cape Evans are so affected by the surrounding mountains that they give us little information of the actual pressure gradient in the neighbourhood of the station. There can however be little doubt that the air motion over the west of the Barrier is mainly from the south, therefore we must assume that the isobars over that part of the Barrier run more or less parallel to the Western Mountains with the highest pressure close to the mountain range. The resultant wind direction at Framheim during the ten months — April 1911 to January 1912 — was from S. 74° E. The exposure for winds at Framheim was almost perfect, we must therefore conclude that the isobars run very aiiproximately in this direction with the low pressure to the north. If the wind velocity is proportional to the gradient the isobars must be closer together near Cape Evans than near Framheim, for the resultant wind velocity during the ten months mentioned above was 10-7 miles an hour at Cape Evans and only 3-1 miles an hour at Framheim. The isobars over the Barrier and the south, of the Ross Sea shown on figure 58 have been drawn to conform with these condi- tions. The distribution of the pressure near Cape Adare cannot be judged from the wind direction and pressure at that station. The pressure distribution shown over the north of the Ross Sea in figure 58 has been deduced from the general pressure distribution shown in the weather maps contained in Volume II. The pressure distribution shown in figure 58 is probably correct in its main outlines, but the distance between the isobars is only to be taken as indicating the gradient, for in the absence of free wind ob.servations at Cape Evans and Cape Adare it is impossible accu- rately to determine the gradient. The diagram shows that the mean pressure over the Barrier is higher than over the Ross Sea, and there is every indication that the pressure is iower 176 PRESSURE. over the south of the Ross Sea than over the north. Wliether there is on the mean of the year a depression over the Ross Sea with closed i.sobars must be left an open question. EAST LONa (70 WCST LONG. if.o" Kio. .08. Pressure distribution year. Daily variation of pressure. It has already been explained how hourly values of pressure were obtained at Cape Evans from the barographs corrected every four hours by means of the readings of the mercury barometer. This series of observations extends from January 1911 to December 1912, i.e., for 23 months. There are also available the two hourly readings of the barometer made during 24 months, from February 1902 to January 1904, at Hut Point. The secular barometer changes are so great in McMurdo Sound, that a single month's observations are practically useless for determining the true daily variation. It has therefore been decided to give the data for each month separately in Volume III and to discuss here only the final result obtained by combining the four years' observations for each month, which give fairly regular and constant values for the daily variation. As only twc-hourly barometer readings were made on the Discovery Expedition the nioan values for the fouj- DAILY VARIATION. 177 years can only be given at intervals of two hours, but these are sufficient for showing the trend of the daily variation. 1 o ; 1 1 * 6 1 II IG 1 12 It 16 I 18 20 .t 1 ?4 — V ■^ \ s N ^ / Decembor 1 1 1 - \ / S ~>; y 1 1 ' Sun continua/(i/ above horizon / / s -- ^ /-- — ^ "---- ^ ^ / , aixuf iry ^ ■-- -— --^ \ / "^ November 1 1 / ^ \ 1 1 1 Sun sets for ,/ "^ \ "~-- 7^ 1 ■~~-- "^ short period only ~f / ^ ^ + 01 4 1 f-0 ' •="' "»' ) 1 1 ' 1 05" ! i 6 6 10 1 X 1 : ic \ H > II ) : 3 ? 2 2 % c 2 1 6 0 1 12 14 16 IB 2 1 ! D 2 2 2 < / / \ f^ ....^ --^ -0- -0- Day and night ^^5»L_J ^. \ ^ / March -"■^ ne ar/y equc »/ -^ Septoraber r" 1 Ob ' ^ P^ y \ •*. — T^ J ^ S — ow" / -- ^ -^ \ £u T ris es ^ /■ / / \ \ / T'H — AprU ^hor ■tpe '•/oo'c /7/tf ^ \ / August X ____ / r' — \ k / / / -\ \ \^ ,i "^ -— 7- Us :r^ — ^ -^ "" Jl Jy "x / hehiv horizon **■... ^ / ^ y ■■ JCIES ■^^ ~ ~~^ 3 I > { 1 3 ! 4 r J 3 2 0 -c 2 - >4 / \. \ / y s N umr nv.-. r7e/- ran. +■0 4 t+-0 3" +• no'^ ,p 2 I 5 D 2 4 6 3 ' r 1 0 22 i 4 1 1 I . s 0 2 4 ) 8 0 ; 2 1. ♦H 0- / -^ ^ —- ^ —^ .^ ^ |X - 0- —0 ^ ^ H— Autumn -7 / / 1 ^p^ Auff. Oct 35' —i 305" 1 — 0 L' — ' 010 -^ --- ■~,^ ^ ^ ■^ ivini ff/- Tuly "^ ~ i»> . / -^ '^ ^ y yt>« R ' t - 30 r r ix-^s^T^-— ^ 1w (^ yr -20- J in Feb. -1 — A3T \pl. May. jS^-~-!jiL_Mi^£fr-t fk Nov w. on -10- Year! Li iron art ion of- Oz 1^1 1 1 1 Fig 59. Drily variation of pressure. McMurdo Sound, 4 years. In order to bring out the dependence of the daily variation of the barometer on the times of rising and setting of the .sun, the curves have been plotted in pairs on figure 59 so that months having similar solar conditions are brought into juxtaposition. 178 PRESSURE. In every month throughout the year there is more or less marked evidence of two maxima and two minima. On the mean of the year (bottom but one curve) there i? a minimum near 4 hours and a maximum at 10 hours, these will be referred to as the morning minimum and maximum. The second minimum occurs at 16 hours and tl»e second maximum at 20 hours, these will be called the evening minimum and maximum. During the months November, December and January the sun was above the horizon the whole time. It will be noticed that the curves for January and December are very similar, but that the curve for November is somewhat different. This difference is due to 1911 and 1912, for in 1902 and 1903 the November curve was almost exactly the same type as that for January and December. It is very probable that with a .sufficient number of years of observations the variation for November would prove to be similar to that of January and December. The combined curve for these three months, shown as the summer curve in the lower half of the figure, iii probably typical of the period during which the .sun is above the horizon the whole time. In this curve the morning minimum and maximum are highly develojied while the evoning minimum and maximum are almost absent. During February and October the sun only sets for a short period during a few of the days, and these two months have remarkably similar daily pressure variations wh.ich are not very different from those of March and September, when the days and nights are of approxi- mately equal lengths. All the four curves, except for September, show the double daily varia- tion in which the morning maximum is larger than the evening one. The sun rises only for a short time on some of the days in April and August, and is below the horizon on the remaining days. These two months have almost identical variations, whicli are different from those of the months on either side. In both months the morning minimum and the evening maximum are abnormally developed, so that, except for a slight irregularity at midday, the c^lrves are simple with a minimum at 6 hours and a maximum at 20 hours, the amplitude being abnormally large. It is difficult to see any reason why the appearance of the sun for a short time at midday should produce this effect on the pressure variation. The sun is entirely below the horizon during May, June and July. It is therefore sur- prising to fuid that while the curves for May and July are similar, the curve for June is different from them. The following points appear of importance in this survey : — (a) The great similarity of the curves placed together in pairs in figure fjQ shows that the daily barometer variation depends largely on the relative length of day and night. (h) During the months that the sun is more above the horizon than below, October to February, the morning maximum is more developed than the evening one — the chief feature of all the curves (the November curve qualified as stated above) being the pronounced maximum at 10 hours, (c) Three of the five months during which the sun is more below than above the horizon, April, June and August, show an almost total absence of the morning maximum ; but, on the other hand, the intermediate months. May and July, have the morning maximum well developed and larger than the evening one. It is difficult to combine {b) and (c) into a single statement giving the dependence of the type of variation on the position of the sun. If it were not for May and July, it would appear that the morning maximum was developed with a high sun and absent when the sun was low, or entirely below the horizon ; but as the morning maximum is better developed than the evening one during May and July, this generalisation does not hold. DAILY VARIATION. 179 Fourier Coefficients. The Fourier coefficients have been obtained as foilowa. The coefficients have been cal- culated from the actual departures from the mean for (a) each month and season in 1911 and 1912 separately, using 24-hourly values ; (&) each month and season in 1911 and 1912 combined, using 24-hourly values ; (c) each month and season in 1902 and 1903 combined, using 12-hourly values. In order to combine the two periods 1902-03 and 1911-12, one of which has 24-hourly values and the other only 12, two methods were adopted. First, the actual departures for every two hours were combined from the four years' observations giving twelve values in the day (those plotted in figure 59) which were then analysed in the usual way; and secondly, the coefficients obtained by (&) and (c) were combined algebraically. The latter method is preferable, as it uses all the observations made, while the former discards all alternate hours in the 1911 and 1912 period. As a matter of fact the results of the two methods varied very little, but the double calculation was a useful check. The results for the four years combined, published and discussed here, are those obtained by the second method. The series used is the ordinary one : dB = «i sin (Ai-t-x) -f- a2 sin (A2 + ix) -f as sin (As-j-Sa;). Table 100. Fourier coefficients of the d%ilij v.iriation of the baroinster at McMurdo Souiul based on the -periods February 1902 to January 1904 and February 1911 to December 1912. «! Ai a2 A2 03 As .January •0032" 278° •002h" 140° •0006" 69° February . ■0017" 291° •002J" 150° ■0005" 282° March •0029" 248° •0034" 149° ■O007' 49° April . •0071" 179° •0026' 14->° •0007' 260° May . •0005' 50° •0021" 120° •0011' 324° June . •0046" 124° •0(l2(i" 71°' ■0002" 269° July . •0006" 342° •0010" 141° •0008" 286° August •0086" 170° ■0011" 142° •0006" 3.38° September •0010" 106° ■0018" 146° •0004' 326° October •0024', 272° ■0018" 149° ■0003" 348° November ■0032' 220° 0023" 162°- (1009" 160° December •0030" 2.32° , 0027" 146° ■0003" 359° I November 1 Summer < December S 1 January 1 •0035" 248° •0025' 148° ■0003' 110° ( February j Autumn < March } ( AprU ) •0029" 208° ■0027" 147° ■0003" 298° May 1 Winter • Juno > Uuly 1 •0014" 113° ■oniT" 11G° •0007" 304° ( August Spring < September ( October •0029" 179° •0015" 146° 141° ■0004" 336° Year . ■0020" 204° ■0(120" ' -OOOS* 3211° 180 PRESSURE. Twenty four-hourlij jjeriod. This term reflects the characteristics which we have already noticed from our survey of the actual curvas. The months with the largest whole day amplitudes are April and August. The strange difference between June and the months on either side of it is also shown by the large anip'itude of June and the small amplitudes of May and July. It will a'so be noticed that the amplitudes and phases for the whole day period of months forming pairs in figure 59 are very similar. Naturally the phases depend largely on whether the morning or evening maximum is the most developed. Thus the phase for the months having the morning maximum highly deve- loped— October to February — have phases varying about 250°, while the phase for those with the eveniiig maximum devebped — April, June and August — is nearer 150°. For the year the amplitude of the whole day wave is only -002" and the phase 204°, which brings the maximum of this wave at 16 hours 24 minutes, and the minimum at 4 hours 24 minutes. Twelve-hourly period. From the point of view of the physics of the atmosphere this is by far the most important term of the harmonic series. Examining first the amplitude this shows a marked yearly variation with a maximum in March and a minimum in August. The values of an are plotted as the bottom curve of figure 59 and it will be seen that they lie very nearly on a sine curve having its maxi- mum in February and its minimmn in August, the variations in cio therefore do not coincide with the seasons, the maximimi occurring two months after midsuimner and the minimum two months after midwinter. The phase of the half daily period is remarkable. During nine out of the twelve months the phase lies between 140° and 1-50°, i.e., during these months the time of the maxima and minima does not vary by twenty minutes.* Of the three remaining months, May, June and November, the latter really departs from the above rule by accident due to a very abnormal November in 1912. The three years 1902, 190.3 and 1911 combined give a phase for Novem- ber of 148° and therefore we are justified' in saying that only two of the months. May and June, have phases varying appreciably from the other months. Looking at the seasons this result comes out even more forcibly, for the pha.se in the summer, autumn and spring only varies by 1 degree from 147°, i.e., the phase in these seasons does not vary by more than two minutes from the mean. On the other hand the phase for the winter months is only 116°, i.e., it varies from the mean phase in the other seasons by a whole hour. That this is not an accidental result is seen from the following statement of the phase in the individual winter months. Table 101. Values of ^2- May. June- July. Season. l'Jii2-ii;i .... 131 = 61° 164° 133° 1911-12 .... 106° 74° 118° 93° 4 years .... 129° 71° 141 116 * A variation in A^ of 30° is equivalent to a variation of one hour iu the time of the maxima and minima. DAILY VARIATION. isi Witli the exception of July 1902-03 each of these sub-divisions shows the reduced phase, the jTreatest reduction talcing place in Jiuie — ^the midwinter mouth. If we now examine the curve for the winter in figure 59, we see that it varies from the curves for the other aeries in one important particular. The fall from the evening maximum to the morning minimum is irregular. After commencing to fall from 22 hours to 24 hours the fall is arrested and it rises shghtly from 0 to 2 hours and then falls to a minimum at G hours. On the analogy of the other curves the fall shou'd have been something like that shown by the dotted line in the figure. The whole effect has the appearance as if an increase of pressure had been imposed on the normal daily variation between midnight and 6 a.m. If we examine the curves for May, June and July, this same abnormality is clearly visible on each. It is also showm on the curves for each of these months in each of the separate periods 1902-03 and 1911-12. There can be no doubt that this abnormal rise of pressure between midnight and 6 a.m. is a real effect during the winter months. If we eliminate the effect to some extent by correcting the curve as shown by the dotted line in figure 59 the value of A2 for the winter season becomes 136°, instead of 116°, i.e., the phase has become much more nearly equal to the phase in other months. It appears reasonable to onclude that the half daily period has the same phase 146° throughout the year, but in the winter, the abnormal rise in pressure between 0 hours and 6 A.M. alters the apparent phase by nearly 30°. This abnormal rise in the pressure is exceedingly interesting. As stated above, an examin- ation of the curves of the individual months leaves little doubt of the reality of the pheno menon. In our discussion of the daily variation of the temperature a similar abnormality of the temperature at 4 a.m. was noted. It is impossible not to connect the two phenomena. It was shown that the temperature effect was visible in the records of Snow Hill and the Gauss Station. The pressure effect is shown in the Snow Hill records, but not in those for the Gauss Station. No simple explanation of either of these two effects is obvious, and they are worthy of a much fuller investigation than it has been possible to give them here. Cause of the twelve-lwvrly haromeler oscillation. — In low latitudes the twelve-hourly baro- meter oscillation is very strongly marked and at all places within 50° latitude of the equator the maxima and minima occur at every place at approximately the same local time. The conclusion that this oscillation is due to a wave travelling around the world parallel to the equator is therefore irresistible. In 1882 Lord Kelvin * suggested that this wave was due to a natural oscillation of the atmosphere. Since then a great deal of research has been undertaken to calculate the ampli- tude and phase of this oscillation by applying mechanical laws to the known constants of the atmosphere. Theory shows that such a wave should have a constant phase at all places on the same latitude when local time is used and that the amplitude should be largest at the equafor and zero at the poles. When the results of meteorological observations made in north polar regions came to be examined it was found that there the oscillation conformed with neither of these conditions. In the first place the amplitude of the oscillation was found to be much larger than it ought to have been according to the theory and that the phase was nothing like constant in local time. When discussing the results of the Lady Franklin Bay Expedition, 1881-83, Greely f remarked that in high latitudes the phase of the half daily barometer oscillation was much more nearly constant if Greenwich time was used than if local time was used. In other Thomsou. p. R. S. Ediu. 11, 1882. f See Haun. Met. Zoit. 7, p. 8, 1800. i82 PRESSURi;. words instead of the maxima and minima occurring at each place at the sa e local time, they occurred at each place at the same absolute time. Thus in north polar regions the barometer oscillation apjiears not to be due to a wave travelling round the circles of latitude, but the whole cap of air appears to oscillate outwards from the Pole. In table 102 are collected together all the observations of the twelve-hourly barometer oscillations available from high southern latitudes.* Table 102. Constants of (he half daily harometer oscillation in the Antarctic. Station- Latitude. Longitude. a.. A., Time of Maxima. No. of years. mm. Local time. Green- ■vvioli time. hr.s. lirs. MoMurdo Sound 77 3S' S. 16C 30' E. ■051 141 loo 109 4 Belgica 70 35' S. 86= 24' W. ■038 307" 47 104 Gauss 66 02' S. 89 38' E. ■043 221" "■7 17 Snow Hfll 64^ 22' S. 57" 00' W. ■036 104 8 5 123 H Laurie Island t 60° 4t' S. 44° 39' W. . ^081 183° 8^9 11 9 South Georgia . 60° 30' S 36° 00' W. ■214 161° 9^7 12-1 ' It is clear from this table that the times of maxima are not constant in either local time or Greenwich time. The oscillations cannot therefore be duo to waves travelling round the earth parallel to the circles of latitude, which would cause the maxima to occur at the same local time at all stations ; nor to the oscillation of the polar cap which would cause the maxima to occur everywhere at the same Greenwich time. It is possible, however, that the oscillation observed at each station might be due to the sum of two oscillations, one round the earth and the other to and from the Pole, for the sum of two oscillations of the same period is another oscillation of the same period but with a different amplitude and phase. In order to investigate this solution I was led to review the work done previously on the twelve-hourly barometer oscillation in the northern hemisphere and the results of this work have been published in the Quarterly Journal of the Royal Meteorological Society, Volume XLIV, No. 185, p. 1, 1918. It was shown in that work that the observed amphtude and phase of the twelve-hourly barometer oscillation in all parts of the world can be explained by the existence of two atmospheric vibrations — (a) a vibration around the world parallel to the circles of latitude, and • (6) a vibration between the equator and the poles parallel to the circles of longitude. The former is expressed by h sin (2a;+B) and the latter by c sin (2X+C-2X) in which x is the local time, B and C are constants, and h and c are functions of the latitude. * Observations are also available for Port Charcot, but they are so widely different from those obtained at the neighbourmg station of Snow Hill that they are clearly wrong ; they have therefore not been used in this discussion. t Anales do la Ofiicina Meteorologica Argentin.i. In this publication the value.s given on page 63 for the Fourier coefficients are wrong, the values used here have been calculated from the data given in table 44, page 62. DAILY VARIATION. 183 It was also found from observations made in the northern hemisphere that b =0-937 cos^.^ B=154:° c =-137 (sin2 0 -1) C =105" Thus the amphtude and phase at any place can be obtained at once from the relationship 02 sin (2a; -I- Aa) =0-937 cos^ sin (2x + 154°)+0-137 (sin2

., 'Oct. } j Nov. J Summers Dec. \ „ „ |jan. j 152 1.31 145 194 •4S2 •675 •614 •469 •0064 0104 •0085 •0049 The mean yearly values based on four years' observations are : average depth of wave from crest to trough =-572 inches, average length =152 hours whicn gives an average change per hour of -0075 inches. There is an appreciable yearly variation in the character of the waves, the amplitude being greatest and the period shortest in the winter, thus giving in that season rapid barometer changes. The summer is the period with the greatest steadiness of the barometer. Pressure waves at Cajje Adare and Framheim. — At Cape Adare and Framheim we have not a complete year's observations, but fortunately wc are able to obtain approximate values by a comparison with the McMurdo Sound results. Table 108 contains the data at Cape Adare, Cape Evans and Framheim for the period April to December, 1911. 188 PRESSURE. Table 108. Pressure waves in the Ross Sea area. April to December 1911. Station. Latitude. Moan length of wave. Mean variation in wave. Mean hourly change. 'S. Hours. Inches. Inches. Cape Adare . 71 Ill ■580 ■0110 Cape Evans 78 142 •605 •0082 Framheim 79 152 •66U ■0087 During the mis.sing months the waves are longer and not so deep as during the rest of the year, therefore the above results are dificrent from what would have been obtained if a complete year's observations were available. It is reasonable to suppose that by applying the factor necessary to convert the above values for Cape Evans to the mean value found for four years to the data for the other stations an approximate yearly mean wdll be found. Thus in McMurdo Sound the mean length of the waves for the four years was 152 hours, while during the period considered in table 108 it was only 142, hence the factor necessary to convert the mean for the broken period into the four-yearly mean is ^/ Similarly the factor for converting the amplitude of the waves is .572. 6U5 Ajiplying the same factors to the observations at Cape Adare and Framheim we obtain the mean yearly values given table 109. m Comparison of pressure ivaves at Antarctic stations.— In table 109 the mean length of the waves, the mean pressure variation in the waves and the mean hourly change are given for eight stations in or near the Antarctic. Table 109. Pressure waves greater than -2 inch (5 m.m.) in the Antarctic (yearly average). Station. Latitude. Mean length of wave. Mean variation in wave. Mean hourly change. °S. Hours. Inches. Inches. Kerguelen 49 03 ■634 ■0184 South Orkneys 61 91 ■642 ■0141 Snow Hill . 64 1(17 ■567 ■0106 Gauss Station 66 12-2 ■642 ■OlOG Belgioa 70 124 ■6:iO ■0094 Cape Adare . 71 110 ■.549 ■0092 McMurdo Sound . 78 152 ■572 ■0075 Framheim 79 16,} •624 ■0077 In this table the stations are arranged in order of latitude and three important results are at once apparent : — • (a) The length of the waves increases steadily as the latitude increases. (6) The amplitude of the waves is nearly constant. SURGES. 189 (c) The result of (a) and (/;) is a steady decrease in the rapidity with which the pres- sure changes take place as high latitudes arc attained. The physical explanation of those relationships will be taken up in chapter VI. Pressure surges. The pressure changes which we have studied so far have been actual changes of the barometer from hour to hour and we have shown that the pressure variations are equivalent to waves having a period of five or six days. When the pressure changes are plotted on a much smaller time scale, it is seen that the waves we have so far considered are themselves superposed on waves of a much longer period. These long waves, which I propose to call pressure surges, are clearly seen on plate I. On this plate the mean daily pressures at Cape Adare, Cape Evans and Framheim have been plotted and the points connected by a thin line. The irregularities on the thin line are due to the pressure waves which we have already considered, and as we wish to study the longer pressure waves we must eliminate them. This is most conveniently done by arithmetically smoothing the curves. If instead of plotting against each day the mean pressure of the day we plot the mean pressure of ten days about it, small irregularities are smoothed out while the resulting curve follows closely the larger waves. Thus the thick curves on plate I have been obtained by plotting the mean pressure from the 1st to the 10th of any month against the 5th, the mean pressure for the ■2nd to the 11th against the 6th and so on. Th.e waves on this thick line are the surges which we are now- about to investigate. Comparison of pressure surges in different parts of the world. A glance at the plate shows that the same surges affect all three stations in the Ross Sea area and later on we shall discuss the relative intensity of the surges at each station. Before we do this, however, it seems desirable to enquire whether surges of this nature are confined to the Antarctic or are recognisable in other parts of the world. As far as I know a similar investigation has not been made previously, therefore it is necessary to calculate the ten-day means for the stations at which the comparison is to be made. If the pressure surges are to be compared at several stations this necessitates calcu- lating 365 ten-day means for each year for each station. If data for several years at several stations are to be used, and several years' observations would be necessary for a complete discussion, this will entail a very large amount of computing. I was unable to undertake such a large amount of work, but the important r6!e these pressure surges play in the pressure changes in the Antarctic constrained me to do the necessary calculations to compare the surges for one year at least at several typical stations in all parts of the world. The year chosen for this comparison was from March 1902 to February 1903, as during the greater part of this period there were observations in the Antarctic at the following widely separated stations : Hut Point, Snow Hill, the Gauss Station and on Kerguelen. The stations outside the Antarctic chosen to compare with these were the following ; — Wellington and Adelaide, which with Kerguelen are typical of the Southern Ocean. Seychelles as t3'pical of a tropical ocean. Bombay as typical of a monsoon climate. Greenwich as typical of the northern temperate zone. Irkoutski as typical of the centre of a large continent. Stykkisholm as typical of the great Icelandic centre of action. Vardo as typical of noith polar regions, data from stations further north during the period considered not being available. 190 PRESSURE. in order to show the curves for twelve months and for all twelve stations on plate II the scale used for plate I has had to be diminished, but the relationship between pressure and time remains the same on the two plates. One or two striking results are at once apparent. The curves for Seychelles and Bombay show that for all practical purposes surges do not exist in the tropics. Also the curves for Adelaide, Greenwich and Irkoutski are obviously intermediate between those for the tropica and the circumpolar stations. Further discussion of the curves can only be undertaken by means of a similar statistical investigation to the one used for the pressure waves. As with the waves, surges of greater variation than -2" (.5 m.m.) will be selected and their average length and depth calculated. It is quite obvious that one year is far too short a period to give satisfactory results, but a seneral oversight can be obtained from the following table : — Table 110. Pressure surges. March 1902-February 1903. Station. Latitude. Longitude. Mean 'length of surge. Mean variation in surge. Me.m daily change. Days. Inches. Inch/day. Hut Point. IT 31' S. 166° 45' E. 35 ■628 •036 (McMurdo Sound, 4 yearb) . (46) (•336) (•024) Gauss .... 66° 2' S. 89° 38' E. 42 •449 ■021 Snow Hill. 6V 22' S. 57° 0' VV. 39 ■534 ■028 Kerguelen. 49 25' S. 69° 53' E. 37 •559 •030 Wellington W 16' S. 174° 27' E. 25 •453 ■036 Adelaide .... 34° 56' S. 138° 33' E. 73 ■363 ■010 Seychelles. 4° 45' S. 55° 45' E. No surges. Bombay .... 18° 54' N. 72° 49' E. No surges. Greenwich. 51° 28' N. 0° 0' 37 ■507 •027 Irkoutski .... 52° 16' N. 1(14° 19' E. 51 •413 ■016 Stykkisholm 65° 5' N. 22° 46' W. 28 •625 •045 Vardo .... 70° 22' N. 31° 8' E. 34 •535 •031 The surges at Hut Point were very much larger during the year chosen for the above table than in any other of the four years for which we have observations. I have therefore added in brackets the mean values obtained from all the observations and these should be taken into account in comparing the stations. It is not easy to draw any very simple relationship from the numbers, and it is question- able whether one year's observations are sufficient to give satisfactory results, but the following inferences may probably be correct. In the southern hemisphere the surges are deeper in high than in low latitudes, also the length of the surges is, on the whole, greater near the continents than over the open ocean. SURGES. 191 Similarly in the northern hemisphere Stykkisholm and Vardo have deeper surges than Greenwich and Irkoutski because they are further north. Also the surges are shortest at Stykkisholm which is oceanic and longest at Irkoutski which is continental. The above indicates a relationship which has been found for pressure instability by other methods ; namely, that unperiodic pressure changes increase from the equator to the poles and from the centres of the continents to the centres of the ocean. Annual variation of the surges. — A glance at the curves on plate II shows that in the northern hemisphere the surges are the most developed in the winter months, the curves being much more disturbed during November, December and January, than during June, July and August. Such a seasonal variation is not clear in the Antarctic. Judging from the curves for the Gauss Station and Snow Hill one would be inclined to put the period of least disturbance from June to September, i.e., in the winter, but the Hut Point curve has the least variation in December, January and February. One year's observations, however, are not sufficient to settle this point with such irregular curves. Taking the four years' observations for McMurdo Sound and dividing the year into two halves at the equinoxes, we have the following result : — Table 111. Pressure surges in McMurdo Sound (Jour years' data). Mean length of surge. Mean variation in surge. Mean daily change. Days. Inches. Inch/day Year 46 ■5?.6 ■024 Summer half-year 39 ■473 ■024 Winter half-year .... 52 •590 •023 This indicates that the surges are longer and deeper in the winter than in the summer, the resulting mean daily change being approximately the same. We have now to consider three most interesting and important questions with regard to surges : — (a) The extent of country aSected by any one surge. (b) The relative intensity of a given surge at the different places affected by it. (c) The origin and cause of the surges. We shall first consider the surges in the Ross Sea area, where the stations are so near together that the surges are easily recognisable at each station, and then turn to the more widely separated stations in other parts of the Antarctic. Comparison of surges in the Ross Sea area.— A glance at plate I will show at once that the same surges affect all the stations in the Ross Sea area, but this is not sufficient ; we must find some method of comparing the intensity at the three stations. This can only be done by recognising the maximum and minimum of the same surge at each station, and then calculating the average value of a number of surges thus recognized. One gets into diffi- culties, however, as soon as one attempts to do this, because there is a certain amount of latitude in deciding whether or not a maximum on a curve should be considered as being the maximum of a surge. Thus there is a well-marked maximum on the Framheim curve 192 PRESSURE. on April 15, -nhich, if we had this curve alone, we misht take as the maximum of a sur -1--51 20 hours 13-1-12 29-05 ■ -f-40 2 hours 13-1-12 IV Maximum 20 03 6 hours 15-1-12 29-52 •' 20 hours 15-1-12 29-45 9 hours 16-1-12 From this table we see that the waves are not so large on the plateau as at Framheim, but the nunrber of observations is too small to give a mean of real value. It is rmportant to find how much earlier the phase of the waves is on the plateau than at the other PRESSURE WAVES ON THE PLATEAU. 215 stations and we notice that the four epochs occurred earlier on the plateau by the following amounts : — Table 124. Time on PLATEAU BEFORE TIME AT Wave number. Framheim. Cape Evans. I 14 hours 28 hours II 11 „ 18 „ III 24(?), 6(?) „ ,v 14 27 „ Neglecting the minimum III, the epoch of which is not clearly marked at any of the stations, we see that the phase on the plateau is about twelve hours ahead of Framheim and twenty-four ahead of Cape Evans. These values can only be approximate for three waves cannot possibly give an accurate mean (they give the difference Cape Evans- Framheim, 12 hours, while the true mean is nine hours), but it is important to know that these well-marked waves as well as several minor ones which have not been discussed here prove that the waves arrive on the Polar Plateau before they affect Framheim and Cape Evans and that they take something like twelve hours to advance from the plateau to Framheim. The pressure observations made by Amundsen when on the Polar Plateau have also been examined. The route taken by Amundsen was nothing like so level as that followed by Scott and as he travelled more rapidly I find it impossible to rid his pressure curves from the effect of changing height. It happens, however, that from December 15th to 18th he was at or near the Pole and therefore his height did not vary appreciably during this period. During these days a very small pressure wave passed over the sea-level stations which is clearly shown by the Polar observations, the maximum occurring at 6 a.m. on the 16th at the Pole and fourteen hours later at Framheim. Thus this single wave confirms the results made by Captain Scott's Party. The pressure curve at the Pole during this period has been added on plate 22, Volume II, for comparison with the curves at the sea-level stations. Turning now to the observations made on the plateau behind the Western Mountains near to McMurdo Sound which are plotted on plate V, we see that here also there is no doubt that the waves of pressure which affect McMurdo Sound are recognisable on the plateau Unfortunately during this period the pressure waves were less in evidence even than during the Polar journey. There is only one clear point of inflexion, the one recorded on the 23rd November, but from this and a few of the smaller irregularities one can conclude with some certainty that the waves affect the plateau, and the phase is either equal to or slightly behind that at McMurdo Sound. The waves are also slightly smaller than at sea- level. We have now shown that the pressure waves can be detected at sea-level in the Ross Sea area and also on the surrounding plateau. They appear first on the Polar Plateau, then about twelve hours later at Framheim, nine hours later at Cape Evans, and finally nine hours later still at Cape Adare. If we assume as a first approximation that the waves travel with a linear wave front, these observations are ^iufficient to fix the direction in which they travel. 216 PRESSURE, WINDS AND WEATHER. Figure 63 is a rough sketch map of the region under consideration. The South Pole is marked S.P., the mean position of the Polar Plateau P.P., Framheim F., Cape Evans E. and Cape Adare A. EAST LOka 170° WEST LONG- 160' Fio. 63. Diagram of pressure wave front. Now the wave takes twice as long to travel from Framheim to Cape Adare as from Framheim to Cape Evans, hence if we bisect the line A F in Y the wave must reach Y at the .same time as Cape Evans, hence E Y is the position of the wave front at the moment it reaches Cape Evans. As it happens E Y is practically perpendicular to the line joining Framheim and Cape Adare, hence this latter line gives the direction of movement of the waves. Now if we draw lines F X and P.P. Z, parallel to E Y they give the position of the wave front as the wave arrives at each station, and the perpendicular distance between them is proportional to the time taken for the wave to travel from place to place. It will be seen that the distance F Z is similar to F Y, hence the wave would take approximately the same time to travel from the Polar Plateau to Framheim as from Framheim to Cape Evans, and we found this to be the case. We also see that the wave would affect the Western Plateau, W.P., very shortly after it arrived at Cape Evans, which also was shown to be so by the observations. PRESSURE WAVES AND WINDS. 217 There can therefore be little doubt that we have to do with waves of pressure travelling outwards from the centre of the Antarctic Continent along directions parallel to a line joining Cape Adare and Framheim. It is interesting to notice that this line produced passes very near to the position 80° S., 120° W., about which the pressure surges were found to have their maximum intensity (see page 195). Pressure Waves and Winds. b It becomes necessary now to examine the pressure waves to see if there is any intimate relationship between them and the winds at Cape Evans, Framheim and on the Barrier. The conditions at Cape Adare will be considered separately later. If one examines the wind arrows placed along the pressure curves for Cape Evans the periods of blizzards, calms and northerly winds are clearly seen. If, however, one examines the pressure curve for Cape Evans alone it is difficult to find any certain relationship between the pressure changes at that station and the accompanying wind conditions. To fix our attention we will examine the pres-sure and winds on plate 17. The fir.st blizzard on this sheet occurred on October 10th with a nearly steady barometer at Cape Evans. The second commenced on the evening of the 12th just when the barometer commenced to rise sharply ; but the blizzard stopped, for no apparent reason, while the barometer was still rising rapidly. The third occurred on the night of the 15th-16th while the barometer was rapidly falling. It ended at midday on the 16th by the wind suddenly turning completely round and blowing strongly from the north. When this occurred there was no apparent change in the pressure curve such as is usually associated with a complete reversal of the wind. The two remaining blizzards on this plate are associated with the crests of two small pressure waves, while the trough between the two waves passed during calm weather with an occasional light wind from the north. This is exactly the reverse of what happened when the large wave passed on the 14:th, for then the crest of the wave was accompanied by a calm and the blizzards occurred before and after the crest passed. The same will be found on all the plates, blizzards start and stop at all phases of the actual pressure waves and for no obvious reason so far as the pressure alone is concerned. Take as another example plate 7. Here we see a large pressure wave between a trough on .Tune 16th and a trough on the 22nd. During the passage of this wave there is first a period of calm, then a period of high northerly winds which suddenly changed into a southerly blizzard. Then a period of calm with an occasional wind from the north followed by a blizzard, then another high wind from the north. All these changes took place while the crest of a single wave passed over the station. On such evidence as this one is tempted to say that the winds are entirely independent of the pressure waves and occur under the influence of some other motive force. The ca.se, however, takes on a different light when instead of the actual pressure at Cape Evans one examines the difference in pressure between Cape Evans and Framheim. The bottom curve on each plate shows the pressure difierence Cape Evans — Framheim, plotted for convenience to twice the scale used for the pressure curves. On this curve are repeated the winds at Cape Evans. Looking now at the lower curve corresponding to the last example, June 16th to 21st, we see that the wind changes which occurred so capriciously when compared with the actual pressure at Cape Evans are very closely related to the changes in the pressure difierence between Cape Evans and Framheim. The blizzards all occur when the pressure difference curve is rising, and the calms and northerly winds when the curve is falling. The wind changes direction almost exactly at the moment the lower curve starts to rise or fall. Turning back to plate 17 we see that there was no blizzard when the large crest passed on the 14th because this wave produced little or no effect on the pressure difference, on the other hand each of the small waves on the 19th and 21st was accompanied b> large rises 28 218 PRESSURE, WINDS AND WEATHER. on the bottom curve and therefore blizzards occurred while they passed. This relationship which is so clear on these two plates will be found to be the general rule throughout all the plates. There are certainly exceptions, but compared with other meteorological rules this is a very rigid one. We can therefore conclude that a blizzard will occur when the pressure difference between Cape Evans and Framheim increases and high northerly winds, calms or light southerly winds will occur while the pressure difference decreases. We have now found that it is difficult to see any close relationship between the winds at Cape Evans and the actual pressure waves, as would be the case if the pressure waves were due to the passage of high and low pressure systems which are maintained by the winds circulating romid them — good examples of which are to be seen in the curves for Melbourne at the top of each plate. On the other hand there is a close relationship between the winds at Cape Evans and the changes in pressure difference between that station and Framheim, 300 miles to the east. Theoretical Discussion of the Effects of Pressure Waves. We will now investigate what effect the passage of pressure waves would have on the pres.sure distribution and winds supposing that the waves are imposed on the Ross Sea area from without ; this is reasonable in view of the fact that we have been able to identify the same pressure waves on the surrounding plateau and at the sea-level stations. These waves will be supposed to travel over the area adding their pressure to the existing sea-level pressure and in consequence changing the pressure distribution and air motion. We have therefore first to determine the pressure distribution without the waves, and then to impose a series of waves on this distribution and examine the result. For simplicity in the following discussion we will assume that the three stations Cape Adare, Cape Evans and Framheim are situated at the corners of an isosceles right angle triangle of which the line joining Cape Adare and Framheim is the hypotenuse. In figure %4m the three stations are represented by the points A, E and F, the line A E B then represents the line of the Western Mountains and the line E F the edge of the Barrier. We will first consider what would be the pressure distribution if all air motion could be restrained. In the absence of a difference of temperature between the Ross Sea and the Barrier, pressure would increase slowly from the north to the south owing to the general increase of pressure with latitude which is a common feature of the Antarctic. This would cause the pressure to be higher over the Barrier than over the Sea to the north of it. There is, however, a large temperature difference between the Barrier and the Sea and the chief pressure difference would be due to it. Both effects then give high pressure over the Barrier and low over the Sea. If there were no air motion the isobars would run something like those shown on figure 64a. That is, they would run approximately east and west and parallel to the edge of the Barrier. We have now to examine how this simple pressure distribution will be affected when motion of the air takes place. In the southern hemisphere, owing to the influence of the earth's rotation, the air motion takes place along the isobars in such a direction that the high pressure is on the left of the direction of flow and the low pressure on the right. Under the pressure distribution shown on figure 64a the whole air over the Ross Sea area would tend to flow to the west as an easterly wind. This, however, is not possible owing to the Western Mountains which act as a wall running north and south at right angles to the easterly wind induced by the pressure distribution. It is easy to see that the air from the eastern half of the Barrier would flow over to the west and would then be forced, along with the air from the western half, to travel to the northwards paraflel to the line of the Western Mountains. THEORETICAL PRESSURE DISTRIBUTION. 219 There would therefore be an easterly wind over the east of the Barrier and a southerly wind alonii the line of the mountains. It is also clear that the intensity of the southerly wind LOW SE.A BARRIER HIGH HIGH ia) (b) Fig. 61. Theoretical pressure distribution, must increase as it flows towards the north because it is constantly receiving air both from the south and from the east of the Barrier. In fact practically the whole of the air set in motion on the Barrier is concentrated in the current which flows across the north-west corner of the Barrier near to Cape Evans. This is shown diagrammatically by the arrows in figure 646. Now it does not matter how the motion of air is deflected or restrained the pressure and wind velocity will always adjust themselves in time until (neglecting the effect of friction) the air is travelling along the isobars with the high j^ressure on the left and the low pressure on the right and the distance between the isobars will be inversely proportional to the wind velocity. When the wind flow shown in figure 646 has become established the isobars will have adjusted themselves to be parallel to the flow and near together where the wind is strong. We therefore get a system of isobars like that shown on the same figure. We see that over the east of the Barrier where the air can move from east to west the original isobars are not afiected, but near the Western Moimtains the isobars turn sharply to the north and crowd together near Cape Evans where the air current is strong. Whereas before the air motion started the pressure at Framhoim and Cape Evans was the same, we see from figure 646 that when motion occurs it is necessary to cross two isobars in going from Framheim to Cape Evans, hence the pressure is now appreciably higher at Cape Evans than at Framheim. This increase of pressure is due to the concentration of the air motion over the north-west of the Barrier, but it is brought about in two ways, first the stoppage of the easterly wind over the west of Barrier causes the piling up of the air there and a consequent increase of pressure, also the air moving rapidly from south to north near the 220 PRESSURE, WINDS AND WEATHER. Western Mountains is pressed to the west by the force due to the earth's rotation and this produces a pressure gradient away from the mountains. Both effects raise the pressure and cause the isobars to adjust themselves to the direction of the air movement. Going further we see that an increase of the pressure difference between the Barrier and the Ross Sea increases air flow over the west of the Barrier with a greater concentration of the motion near to Cape Evans where the pressure rises. Thus a uniform lowering of the pressure over the Ross Sea or a raising of the pressure over the Barrier increases the southerly wind at Cape Evans and causes the pressure at Cape Evans to rise relatively to the pressure of Framheim. We have now to investigate the result of impressing travelling pressure waves on this general distribution. So far we have not considered the magnitude of the pressure differences, the isobars shown on figures 6ia and 646 being only diagi'ammatic. The required values are found and the theory substantially supported by noticing that in all essentials the pressure distribution shown on figure 646 which has been derived from first principles is the same as that of the average pressure distribution found from the actual observations and already given in figure 58, page 176. In the latter we have numerical values for the isobars and therefore in the future discussion we \vill use it in place of the approximate diagram of figure 646 On figure 65a the isobars from the average pressure distribution have been entered to fit our diagrammatic representation of the geography of the Ross Ssa area.* The whole area 45° 270* 45 (a) Fio. 65 (a) Actual pressure distribution. (6) (6) Imposed pressure waves. * Figures 65 to 73 were all prepared before the error described in the footnote to page 166 was detected. Hence the pressures used in these diagrams are all -OS" too low. This, however, is quite immaterial to the present discussion as the pressure distribution is unaffected. THEORETICAL EFFECTS OF PRESSURE WAVES. 221 has been divided into squares, and in each square the pressure has been entered in agreement with the isobars. The waves which are to be impressed on this pressure distribution travel from Framheim to Cape Adare, i.e., in the direction of the line of squares a, b, c, ... n, o. The wave fronts are at right angles to this line. Values will be given to the letters a, b, c, etc., according to the pressure of the wave which is travelling from a to o. As the pressure due to the wave is the same along the whole wave front the values of a, b, c, etc., must be added to the pressures in the line of squares to which the letter is attached. Thus the value of i determined by the position of the wave has to be added to the pressure entered in each of the squares along the line E to i. The wave to be investigated may vary in two particulars, (a) in amplitude and (6) in the difference in phase %^ath which it arrives at the three stations. A large number of waves having different amplitudes and phases have been investigated, and it is found that the result is not affected in principle, but only in detail by changing these variables. From these waves the one has been chosen for discussion here which gives the best general result. This wave is given at Cape Evans by dp =-3" cos 9 and it travels from Framheim to Cape Adare in a quarter of its period. So that the pressure changes at Framheim are dp=-3" cos (61+45°) and at Cape Adare dp=-3" cos (6-45°). In figure 6.56 the wave to be investigated is shown in the same way as the pressure curves on the plates in Volume II. The abscissa represent equal intervals of time, but as we are not concerned with absolute time, which involves the rate of travel of the waves, but only with time in so far as it affects the phase of the wave when it arrives at the different stations, the phase of the wave at each interval at Cape Evans is given instead of the time. The wave arrives at each station an eighth of a period after it left the last. Thus if the phase at Cape Evans is e, it is 6+45° at Framheim and 9-45° at Cape Adare. The most important curve on the plates in Volume II was found to be the one giving the difference in pressure between Cape Evans and Framheim, a similar curve therefore has been added to figure 656. We shall consider the effect of the wave at eight intervals during its passage commencing from the epoch when the maximum of the wave has just reached Cape Adare. The posi- tions of the wave at each of the epochs considered are shown in figure 656 by the vertical lines at e = io", 93°, 135°, 180°, 225°, 270°, 315° and 360°. For each position of the wave two diagrams are given. In the first the pressure which results by the simple addition of the pressure wave to the normal pressure is shown. In the second an attempt has been made to allow for the alteration in the pressure which the air motion would produce, taking into account the inertia of the air and the obstacles to its motion. Thus for each position of the wave there will be two diagrams similar to figures 64a and 646, which are the corresponding diagrams for the conditions in the absence of the pressure waves. 222 Pressure, winds and weather. (a) Fig. 66. Modified pressure distribution <^=<15°, 9 = 45°, figure 66. — Starting with the crest of the pressure wave at Cape Adare the change of pressure due to the inii>ressed wave has been shown in the upper line of bold figures. These values have been added to the lines of figures against each in the normal diagram, figure 65a, with the resulting pressure over the whole area shown by the figures in the squares. Isobars through these figures show the resulting pressure distribution. The added pressure is sufficient to reverse the normal pressure distribution and the highest pressure is near Cape Adare and the lowest over the Barriei to the south-east of Framheim. The simple addition of the pressure makes the isobars as shown on figure 66a start on the Western Mountains and run to the north-north-east. The air motion would be in this direction except that it obviously cannot flow awaj' from the mountains along their whole length. The air motion near the mountains will be from high to low pressure but parallel to the range, i.e., the wind will be from the north to the south near the mountains. Over the Ross Sea at some distance from the mountains the air motion will be along the isobars, i.e., from the south-west. These directions have been shown by arrows in figure 666 in which the isobars have been changed to be in accordance with the air motion. This adjustment of the isobars to the air motion lowers the pressure at Cape Evans by about '15", as will be seen by comparing tht; isobars in the two halves of the diagram. The probable wind strength has been shown by the number of feathers on the arrows, which, however, are only qualitative as no calculation seems feasible. THEORETICAL EFFECTS OF PRESSURE WAVES. 223 e= 90° (a) (b) Fig. 67. Modified pressure distribution. 0 = 90°. e = 90°, figure 67. — The wave has travelled on an eighth of its length, but the change in pressure has not been sufficient to affect nxaterially the pressure distribution. The wind is still northerly at Cape Evans, but has become south-easterly at Framheim. 224 PRESSURE, WINDS AND WEATHER. («) {b) Fig. 68. Modified pressure distribution. 0 = 135°. e = 135°, figure 68.— D^^'ing the interval the pressure distribution has undergone an important change, the lowest pressure is now over the south-east corner of the Ross Sea near to Framheim. There is still an appreciable gradient for south-west winds over the east of the Ross Sea and the pressure decreases from north to south along the Western Mountains so that the west of the Barrier has a lower pressure than the west of the Ross Sea. Northerly winds will therefore continue near the mountains. The pressure is now nearly uniform over the Barrier where there can be little or no air movement. THEORETICAL EFFECTS OF PRESSURE WAVES. 223 Vv^^\ (9 = 180'" A / sf^X/is X^ jbN. / \ %\/\.'^\ 1 "^^ " /\ %^x'^'v\. / y '\ry^'^/\ "/\"^\ / / 'xryf 9" /^ " /K^ 74^^3oNv / / K "^^"/\ "/\ '^/X ^V\ / / °/\ / /\ " /S. " /\ " ^x ^^\. / / «5 Py " /\ ^ /\ " /\ '^ /V" " /S. / / LOW X' W\ 'VX ''JX 'Vv^^x' v\ V ^^-"""^^ E :a so xC/) V''^ ')\/ ;o\^^*irv sj\c^ ?X E '^V 3jV/ ffiX / m \| / m \ 7'' »\ / (X /jo^ ■ iX . 3i\/8; >r *X^eiX^'y\»/\ *Xr '5/ 3i\/ ^y^w\(^ ^V\V\ v\ '^X^°^^ i — . ■"^"""'^ \ " X\ ® X\ ** /\ '* /\ * /vJ'^'v^'X^^ SO<^ SsX^ 35 X^ *Xr '°^^<'''0^ 03^X • — ^ ^^^ \ '* X^ * y\ "X\ 5^\ "X^"*/^ L ^/^ *X\"/\"^^^ 30 miles per hour. 11-30 mUes per hour. 11-30 miles per hour. >30 miles per hour. Barometer change in four hours. -•017 —015 -•003 -F-004 + •014 We thus see that the theoretical relationship between the winds and pressure does agree in the main with the actual observations.* The pressure curves on the plates show that the actual waves which travel across the area vary greatly in amplitude and phase and we have considered only one in which a = -3" and

10 miles per hour. 1 0 8 3 42 6 7 1 7 2 4 1 0 0 1 0 17 10 south. >10 miles per hour, 1 0 6 0 14 1 4 n 15 6 15 4 4 0 3 2 23 7 north. The first line shows the preponderance of winds from the direction between N.E. through S. to S.W. The remaining directions from W.S.W. to N.N.E. have only 8 per cent, of the wind between them. COMPARISON OF THEORY WITH OBSERVATIONS. 233 . The remaining two lines show how the winds at Framheim vary during blizzards and northerly winds at Cape Evans. The relationship comes out more clearly if we group the Framheim winds as in the following table :— Table 127. Winds at FramJieim. Winds at Cape Evana. Percentage frequency. RE.SULTANT. NE, ENE, E, ESE, SE. s, ssw, sw. Du-eotion. Velocity milies per hour. All winds .... > 10 miles per hour, south > 10 miles per hour, north 47 66 5 25 13 36 S. 74° E. S. 88° E. S. 8° W. 3-7 7-8 2-3 During bhzzards N.E. to S.E. mnds at Framheim are nearly five times as frequent as S. to S.W. winds, while during northerly winds at Cape Evans the S. to S.W. are actual more frequent than the N.E. to S.E. winds. The relationship comes out still more clearly by taking the resultant direction and velocity of the winds at Framheim during blizzards and northerly winds at Cape Evans. From table 127 we see that during blizzards the resultant wind at Framheim is nearly due east, while during northerly winds it is from slightly to the west of south. Also the resultant velocity at Framheim is nearly three times as great during blizzards as during northerly winds. Thus the theoretical winds and the actual winds are in excellent agreement. We have now to examine the maps showing the isobars based on the actual observations to see if they agree with the theoretical diagrams which we have constructed. A rapid glance through the maps is sufficient to show that the two main types of which figures 71 and 66 are representatives are constantly repeated, while the great majority of the maps vary only in detail from one or other of the theoretical diagrams. It would be too tedious to describe in detail many of the maps and to point out where they agree and where they differ from the theoretical diagrams. The reader will be able to see these points for himself if he is sufficiently interested to spend some time in com- paring together the maps, plates and diagrams. It will therefore be sufficient if one or two typical examples are shortly discussed here to indicate what should be looked for in a further study of the data. The first plate contains a very typical example of the effect of a moving pressure wave on the pressure distribution. The actual pressure wave between the 6th and 10th was not very large, but it produced a normal difference wave (bottom curve). The two maps for April 6th are similar to the two figures 66 and 67 which represent the crest of a wave at Cape Adare with the difference wave decreasing. During April 7th the pressure distribution was changing and the two maps on this day are similar to figures 68 and 69. The evening map shows that the easterly wind has started near to Framheim, but it has not yet reached Cape Evans. At midnight on the 7th the blizzard started at Cape Evans and the maps for April 8th are similar to the theoretical diagrams for the bfizzard, figures 71 and 72. The map for 8 a.m. on the 9th is similar to figure 73 with the high pressure to the south of Cape Evans. In the next map, 8 p.m. on the same day, the high pressure is shown to have passed Cape Evans and the distribution is now becoming similar to that with which we started, figure 66. The map for 8 a.m. on April 10th is almost exactly the same as figures 67 and 68. 30 234 PRESSURE, WINDS AND WEATHER. It will be seen that a complete wave has passed during this period and the maps have gone through the complete sequence shown in the theoretical diagrams. A similar complete sequence is seen if the maps from April 13th, 8 p.m., to April 17th, 8 A.M., are compared with the theoretical diagrams. In fact for every well-marked wave on the lowest cirrve of the plates the maps will be seen to go through the same sequence in a more or less complete manner. It will be felt that the paucity of observations over such a large area leaves a great deal of freedom in drawing the isobars, and that if there had been more stations their observations might not have fitted in with the simple scheme of pressure changes here dis- cussed. From October to the end of the year there were parties on the Barrier which pro- vided so many more stations, and one therefore turns with interest to the maps for this period to see whether the sledging observations do or do not confirm the theory. Unfortun- ately during this period the pressure waves are not well marked and for days together the pressure was nearly constant over the whole area (see plates 19 to 23). Thus while the pressure distribution will be found to conform to the main types there are few occasions when the passage of waves makes it possible for us to follow the complete sequence of changes. There are, however, one or two examples which are worth studying. During the period October 23rd to October 29th the lowest curve of plate 18 shows that two complete waves passed over the area, and as at this time Captain Amundsen was taking observations on the Barrier the period is suitable for our discussion. Both maps for October 23rd show a distribution similar to that of figure 66 or 67. During the next day, 24:th, the pressure changes to the blizzard type and is therefore on the whole similar to figures 70 and 71. At 8 a.m. on October 25th the distribution is the same as that shown on figure 73 except that the gradient is stronger. The high pressure is just south of Cape Evans and is passing northward so that with the next map 8 p.m., October 25th, we have returned to the starting point with the high pressure over the north-west of the Ross Sea, figure 66. A similar sequence is repeated during the next four days so that at the end of the sequence, 8 a.m. on October 29th, the pressure distribution is the same as at the beginning, 8 P.M. on the 25th. These and other examples, which the reader will see on turning over the following maps, on all of which there are observations from the Barrier, show that when there are large changes of pressure the Barrier observations fit in as well as can be expected with the theoretical distributions of pressure which we have deduced on the assumption of travelhng pressure waves. There can be no doubt now that the passage of the pressure waves alters the pressure distribution making it at one time favourable for blizzards and at another favourable for northerly winds, but it must not bo forgotten that the first step in our theoretical discussion was to show that in the absence of pressure waves the conditions are favourable for blizzards. We will now examine one or two interesting cases of blizzards in the absence of well-marked pressure waves. A good example is shown on plate 12. On and after August 18th there are no large pressure waves shown on this plate. On August 19th, however, a blizzard commenced at Cape Evans. The map for 8 p.m. on August 17th shows a nearly complete calm at all stations, the north-westerly wind at Framheim being only a light local breeze of 1'4 miles an hour. From the actual pressure at the three stations it is obvious that the isobars must run from the east to the west as in figure 64o, the Barrier being at a higher pressure than the Ross Sea. Throughout the 18th a light easterly wind was blowing over the ea.st of the Barrier as shown by the wind at Fram- heim, while the air was calm at Cape Evans. It will be noticed from the lowest curve COMPARISON OF THEORY WITH OBSERVATIONS. 235 on the plate that during the whole of the 18th the pressure at Cape Evans was slightly rising relatively to Framheim. This has been explained as due to the piUng up of air over the west of the Barrier along the Western Mountains. At 8 a.m. on the 19th a blizzard has started at Cape Evans and now v\ath the increased motion the pressure difference between the west and east of the Barrier rapidly increases, and the pressure difference Cape Evans - Framheim rises by nearly half an inch. Another important and interesting example is the blizzard which commenced at midnight of November 15th. At 8 a.m. on November 15th a very shallow low pressm-e area appears over the south of the Barrier. During the day this disappears and by the evening it is replaced by a high pressure area. Although the wind at Framheim is from the south-east, it is almost certain that it is from the east over the greater part of the north of the Barrier. The pressure now rises in the west relative to the east and by the next morning typical blizzard winds are reported from all parts of the Barrier. It is interesting to notice that there has been practically no change in the height of the barometer at any of the five stations between 8 p.m. on the 15th and 8 a.m. on the 16th and yet there is a very large change in the wind conditions which has resulted in a new distribution of the isobars. Similar large changes in wind conditions with little, if any, change in pressure are quite common and many examples will be found if the maps are examined in detail. In drawing the maps many interesting facts attracted one's attention, but it would not be wise to discuss the maps in detail as one feels that it is very questionable whether the run of the isobars is anything more than a rough approximation to the actual conditions. There can, however, be little doubt that the general pressure distribution and the sequence of pressure changes is that shown by the maps. One point, however, must be discussed before we review our conclusions. We have repeated- ly shown that the consequence of the flow of air from the east to the west of the Barrier, when its motion is stopped and the air is forced by the Western Mountains to flow north- wards, is a rise of pressure at Cape Evans relative to Framlieim. We have just examined two cases of blizzards which were accompanied by these relative changes of pressure in the absence of pressure waves, and we found that the pressure differ- ence Cape Evans - Framheim was nearly as large as in the case of blizzards accompanying large pressure waves. The question at once arises may not the bUzzards be the cause of the pressure waves and not the pressure waves the cause of blizzards ? A study of the curves on the plates will soon dispose of this idea. On plate 17 there is a large pressure wave with its crest at Framheim and Cape Evans on the 14th. There was a short blizzard just before, and a longer one just after, the crest passed Cape Evans. Now it will be noticed that both these blizzards were accompanied by a rise in the difference curve although the absolute pressure at both Cape Evans and Framheim was rising throughout one blizzard and falling throughout the other. In this case the pressure wave affected Cape Evans and Framheim almost simultaneously, therefore for practical purposes the pressure wave did not affect the pressure distribution. The blizzards occurred exactly as they would have done if the pressure curve had been a straight line, and the pressure differences shown on the lowest curve were due to the blizzard. The way that the induced pressure difference is impressed on the pressure wave is clearly seen in the case of the second blizzard, for the fall of pressure is decreased at Cape Evans and increased at Framheim. In this case it is obvious that the pressure difference is due to the blizzard and is superposed upon the pressure wave. It is therefore clear that the large pressure wave itself cannot be due to the blizzard. Granted that pressure waves exist independently of blizzards the fact that they arrive at Framheim before Cape Evans and at Cape Evans before Cape Adare proves that they 236 PRESSURE, WINDS AND WEATHER. travel. If they travel they must cause pressure diSerence between the stations. We have therefore two causes at work producing pressure difierence between Cape Evans and Framheini, first the passage of the pressure waves and second the dynamical effects of the blizzards. On account of the coincidence that the line joining Cape Evans and Framheim is at right- angles to the Western Mountains and at the same time cuts the wave fronts at an angle of 45°, the two causes increase the pressure diSerence at the same time. Some pressure waves do not produce blizzards, the difference curve in this case is due to the travelling wave alone, examples are seen on May 22 and 23, June 27 to 29, July 30 to August 1, etc. We have just discussed a typical case, in which changes of pressure due to the blizzards are super- posed on the pressure waves and we have examined several cases of pressure difierences produced by blizzards in the total absence of pressure waves. Thus there are examples of all the possible effects to be expected according to the theory. To these considerations we also have to add the fact that the pressure waves are recog- nisable on the surrounding plateau and appear on the Polar Plateau before they reach the Barrier, hence the waves cannot be caused by the blizzards. There is no escape from the conclusion that the pressure waves are imposed on the Ross Sea area from outside. During the study of the records, which led finally to the theory sketched above, a large number of statistical investigations was made to find the relationship of the pressure condi- tions and changes at the three stations which accompanied the blizzards, northerly winds and calms at Cape Evans. The results of these investigations are contained in table 128. Six classes of wind were taken, five of which are shown in the table ; the class of wind 6 to 10 miles an hour has been left out because it adds nothing to the discussion. Table 128. Pressure and Winds. Winds AT Cape Evaks. North. Calm. 0-5 South. >30 11-30 11-30 >30 (1) Presauro difference. Cape Evana-Fram- heim. + 03-2 + ■189 + 193 + 191 + ■243 (2) Pressure difEerenoe, Cape Evans-Cape Adare. -Oil -012 + •071 + ■105 + 143 (3) Change of preasure difference in 4 hours, Cape Evans-Framheim. -038 -023 + ■007 + •007 + •021 (4) Change of pressure in 4 hours at Cape Evans. -017 -016 -005 + ■004 + 015 (5) Change of pressure in 4 hours at Fram- heim. -I--021 + •007 + •002 -003 -006 (6) Preasure at Cape Evara, departure from mean of month. -■201 -112 + •028 + ■008 -■005 (7) Preasure at Framheim, departure from moan of month. -■054 -•107 + 031 -■008 -040 (8) Preasure at Cape Adare, departure from mean of month. -091 -020 + 045 -■029 -■060 COMPARISON OF THEORY WITH OBSERVATIONS. 237 (1) Pressure difference, Cape Ecans-Framkeim,—lt is interesting to notice that the pressure difference remains positive with all kinds of air motion. In other words the pressure on the average is higher at Cape Evans than Framheim in all winds. This result is obvious from the lowest curve on the plates which is practically never below the zero line. The difference, however, is least with high northerly winds and greatest with high southerly winds. This is in a12 >12 35 51 2" 105 >12 88 0" 44 9 180 Barometer falls and rises ^' days. in two 6° 102 77 >12 >12 65 88 2° 54 11 125 0" 34 8 180 Barometer falls and rises i" days. in three 6° 4° 58 47 12 10 88 111 2° 36 8 142 0° 26 6 180 * Scale given on page 99 of Meinardua. 244 PRESSURE, WINDS AND WEATHER. From the first line of tlie tabic we see that a cyclone travelling from west to east at 9' 5° of longitude a day with its centre 6° of latitude to the north of the station and causing a fall and rise of the barometer of half an inch in one day would produce a mnd velocity of 251 miles an hour, and the wind direction would change 35° between the commencement and ending of the change of half an inch in the pressure. Now Beaufort force 12 was never recorded at the Gauss Station, hence the cases which give the wind force as 12 or over have to be excluded. This leaves us with one case in the first group, two cases in the second, and three in the third. It will be noticed that the change in the wind direction in these six remaining cases is always above 100°. In other words no cj^clone travelliHg with the given velocity could cause a fall and rise of the barometer of half an inch in less than three days without the wind changing direction at the station by more than 100°. The impossibility of Lockyer's cyclones which have their centres 6° to the north of the Gauss Station producing the observed effect is particularly obvious. For such a cyclone to produce a fall and rise of half an inch at the Gauss Station, even in three days, would necessitate a hurricane wind and a change of direction of 88° there, so what the conditions would be nearer the centre can best be left to the imagination. The case is not improved by considering other rates of travel or other shapes of isobars, for a little consideration will show that any change which decreases the angle through which the wind should change will increase the wind strength. Thus any wind change of less than 100° would always be accompanied by hurricane winds. Thus at the latitude of the Gauss Station, no cyclone, no matter what its rate of travel or distance of its centre from the station, could produce a fall and rise of the barometer of half an inch in three days or less without either changing the wnd direction by more than 100° as it passed, or causing winds of more than hurricane force. Now at the Gauss Station we see barometer changes of much more than half an inch in C3usidei-ebly less thin three days without anij appreciable change of wind direction and the only conclusion to be drawn is that these pressure changes are not produced by moving cyclones. There seems no alternative but to assume that the Gauss Station is subjected to travelling pressure waves similar to those in the Ross Sea area. If such waves travel outwards from the continent in all directions their wave fronts will be approximately parallel to the coast. Now along the coast there must ba a steep barometric gradient similar to the one between the Barrier and the Ross Sea and the waves will modify this pressure, increasing and de- creasing it. Thus we see that the pressure waves will increase and decrease the gradient and with it the wind velocity, but as the isobars and wave fronts are parallel there will be little change in the wind direction. Just as we found it impossible to say a priori what relationship should exist between the pressure waves at Cape Evans and the wind changes there, so it is impossible to say whether at the Gauss Station the easterly winds should be stronger or weaker with high or low barometer or should increase or decrease with rising and falling pressure. It is very significant that at the Gauss Station the wind frequently changed direction suddenly from east to west, and next to the nearly constant winds from the east, winds from the west had the greatest frequency. This would be the natural consequence of pressure waves, but cannot be explained by cyclones unless one makes the impossible assumption that the centres of the cyclones passed frequently exactly over the station when travelling from north to south or from south to north and never passed over the station when travelling from west to east. CONCLUSIONS. 245 The weather characteristics which Meinardus found to accompany high easterly winds at the Gauss Station are as easily explained by the travelling wave theory as they are on the cyclonie theory. All these characteristics were found at McMurdo Sound to accompany blizzards and therefore may be expected to occur during the high winds at the Gauss Station. In fact they should be more pronounced at the latter station for the high winds there were from slightly north of east and so indicate an indraft of air from the open ocean which would on this account be warm and humid. It is true that there are considerable difficulties in explaining all the relationships between wind and pressure at the Gauss Station on the wave theory, but there are none so great as the cyclone theory encounters in the large pressure changes with a nearly constant wind direction. Before the problem can be finally solved we must hava simultaneous observations from neighbouring places as was the case in the Ross Sea area. It is possible that the results of the Australian Expedition, which recently wintered near to the Gauss Station, will help in this desired direction. The data for Kerguelen and Snow Hill have also been examined to see if the conditions there are governed by pressure waves or cyclones. The observations at Snow Hill show large pressure changes with little or no change in wind direction, but there are many cases in which the wind changes as the barometer rises and falls. There can be little doubt that cyclones pass over this station, but it is equally true that all the pressure changes cannot be ascribed to them. At Kerguelen there is a much more intimate relationship between the changes of wind direction and changes of pressure than at any of the stations on or near the Antarctic Continent. But even at this station, not infrequently, there are large barometer changes without appreciable changes of wind direction, which may or may not be due to similar pressure waves to those recognised over the Antarctic Continent. Turning now to the two top curves on the plates in Volume II giving the pressure and winds at Melbourne in Australia and The Bluff in New Zealand we see a most close relation- ship between the winds and the pressure changes. This is most clearly seen in the case of Melbourne because on that curve the wind observations are more complete. The relationship is extremely close : the falling barometer is accompanied by winds from some northerly direc- tion, at the instant the barometer ceases to fall and commences to rise the wnd changes and then blows from somj southerly direction while the barometer is rising. We know from the daily weather charts of Australia that the weather at Melbourne and The Bluff is entirely dominated by travelling cyclones and anticyclones. Conclusions. Cape Evans and Framheim. — At these stations the evidence is almost conclusive that the barometer changes are not due to travelling cyclones and anticyclones but to real waves of pressure which travel outwards from the Antarctic Continent. Gauss Station. — According to Meinardus the barometer changes at this station are due entirely to travelling cyclones. Evidence has been brought forward to show that this is impossible and reasons have been given for believing that at this station also pressure waves are mainly responsible for the barometer changes. Cape Adare. — The evidence indicates very clearly that at Cape Adare the main barometer changes are due to pressure waves while secondary changes are occasionally due to travelliug cyclones and anticyclones. 246 PRESSURE, WINDS AND WEATHER. Kerguelen and Snow Hill. — The records for these stations show both pressure waves and cyclones, but the latter predominate at Kerguelen and the former at Snow Hill. The Bluff and Melbourne. — At these stations there is little if any evidence of pressure waves, the barometer changes being almost entirely due to travelling cyclones and anticyclones. From these results we are led to recognise long and deep pressure waves radiating out from the Antarctic Continent and extending over the Southern Ocean to some unknown distance from the continent, traces of which can still be seen in the pressure curves for Kerguelen. Over the Southern Ocean there are cyclones and anticyclones which travel on the whole from west to east. These systems are nothing like so large as those described by Lockyer, but in all probability are in all parts of the ocean of a similar size to those shown on the Australian Daily Weather Reports, and their centres may pass anywhere between the coast of Australia and the coast of the Antarctic Continent. This result explains completely the results obtained in the statistical investigation of the non-periodic pressure changes. The waves over the Antarctic are long, regular and deep as can best be seen on the curves for Framheim. The barometer changes due to passing cyclones over the Southern Ocean are deep but short. Hence as "we pass from the region of cyclones over the ocean to the pressure waves over the continent the length of the waves steadily increases, but there is little change in the amplitude, which is the conclusion already reached on pag3 188 from the observations. Returning now to the theories of Lockyer and Meinardus with which we commenced this discussion, we see that there is no evidence of the large cyclones having their centres on 60° S. and covering the whole area from the south coast of Australia to the ice barrier aromid the Antarctic, which are the chief features of Lockyer's work. We have also shown that the pressure changes at the Gauss Station, Cape Adare and Hut Point used by Lockyer in his discussion are not due to cyclones over the Southern Ocean and therefore give no indications of the frequent y, intensity or rate of motion of such cyclones. Further we have seen that the centres of small intense cyclones pass quite near to Cape Adare and we know from the daily weather charts of Australia that the centres of cyclones often pass near to the southern coast of that continent. It is therefore reasonable to suppose with Meinardus that the centres of cyclones occur in all parts cf the Southern Ocean, but probably more frequently near 60° S. than in any other latitude. Thus Lockyer's scheme of southern hemisphere air circulation, simple as it is, does not fit the facts and must be abandoned. On the other hand t^ere seems to be no justification for Meinardus's conclusion that the weather at the Gauss Station is governed by these Southern Ocean cyclones, and not by the Antarctic anti- cyclone. He admits that along the whole coast of the Antarctic Continent and for some distance over the surrounding sea the mean isobars are concave towards the high pressure in the south, but holds that the mean pressure distribution is not the governing factor in the weather. The mean pressure distribution, he says, is the result of the sum of the instantaneous pressure distributions and the latter are mainly due to the cyclones which pass to the north of the station. Thus the weather at the Gauss Station in spite of the shape of the mean isobars is not under the influence of the Antarctic anticyclone but mider the influence of the cyclones which give rise to the easterly winds. Now we have shown that the easterly winds at the Gauss Station and the pressure change there are not due to cyclones, therefore the whole of Meinardus's scheme falls. In place of it we consider that the Antarctic anticyclone extends outwards from the continent to some distance over the sea, and that the pressure and wind changes are due to pressure waves radiating outwards from CONCLUSIONS. 247 the continent as they do in the Ross Sea area. That the Gauss Station is under the influence of anticyclonic conditions is made very probable by the large percentage of clear skies which it has (see table 79) while it is made almost certain by the shape of its wind frequency curve (see figure 35). A station which has snch a large proportion of calms and light winds could not possibly be dominated by a cyclonic pressure distribution. We have also shown (page 107) that the shapa of the wind frequency curve indicates pressure conditions similar to those of McMurdo Sound, which, although so near to the low pressure area over the Ross Sea, is governed by the Antarctic anticyclone and pressure waves. CHAPTER VII. THE GENERAL AIR CIRCULATION OVER THE ANTARCTIC. As stated at the commencement of the last chapter it was originally thought that atmos- pheric pressure decreases from the belt of high pressure near latitude 35° S. right up to the South Pole. The discovery of the high easterly winds south of latitude 40°, however, shows that this is impossible. In 1893 * the significance of these easterly winds was clearly recognised and the existence of an anticyclone over the Antarctic was postulated to account for them. Later expeditions confirmed an increase of pressure with increasing latitude south of about latitude 6()° S. and the idea of an anticyclone over the region within the Antarctic Circle was confirmed. The Antarctic anticyclone now became as firmly fixed in the minds of meteorologists as the idea of the Polar cyclone had been previously. We have already seen that it occupies a very prominent part in Lockyer's scheme of the air circulation over the southern hemisphere, and Hepworth and others speak of it as if it were an undoubted feature of the Antarctic. Hobbs goes still further and contends that an anticyclone exists over every extensive snow-covered land, and takes the Antarctic and Greenland as the two most pronounced examples. To the anticyclones which owe their origin to a snow-covered land Hobbs has given the name ' glacial anticyclone ' and he has worked out at considerable length the meteorological featiu-es of such anticyclones. His conclusions as to the conditions over the Antarctic are so important that they must be considered in detail here. Hnhhs' Theory of the Glacial Anticyclone.^ — Hobbs summarizes the evidence for fixed glacial anticyclones under the following eight headsj : — • (1) Centrifugal flow of surface air currents above inland-ice masses. (2) Outward (centrifugal) sweeping of surface snow largely derived from the central areas, and its deposition and accumulation as a marginal fringe about the inland- ice. (3) Snow in large part wind-driven above the sloping portions of the ice mass. (4) Sudden warming of the air at the end of the blizzard — fohn effect in descending currents. (5) Behaviour of upper air currents and movements of the cirri. (6) The evolution of the Antarctic blizzard and its termination. (7) Areas of relative calm corresponding to the flat central bosses of the ice domes. * Frioker. Journal of the Royal Geographical Society, Volume II, page 254, 1893. Murray. Journal of the Royal Geographical Society, Volume III, pa^e 17, 1894. ■]• Tho Rolo of the Glacial Anticyclone in the Air Circulation of ths Globe, by W. H. Hobbs. Proo. American Pliilosophical Society, Volume LIV, No. 218, August 1915. ^ J Ibid; page 689. HOBBS' THEORY. 249 (8) Air highly charged with moisture wnthin the flat central area of calms and precipita- tion of snow or ice near the glacier surface. We will now consider this evidence and Hobbs' conclusions in so far as they apply to the Antarctic. We shall find it more convenient not to follow the' order in which the heads are given above, but at the beginning of each of the follomng paragraphs the numbers of the heads which are considered in the paragi'aph are given. (1). At every place where prevaiHng wind directions have been observed at the edge of an inland-ice mass there has been found a larger or smaller component in the outward direction. This is not only true at sea-level, but the air is found to be flowing outwards from the high plateaux both in Greenland and the Antarctic. Special attention is drawn to the winds observed by David, Shackleton, Scott and Amundsen, all of whom on their ascents to the Antarctic plateau encountered winds blowing down the valleys. (7) & (8). This outward motion necessitates anticyclonic pressure distribution, with descending air currents in the centre, where the air will be relatively calm. The deposition of ice to supply the permanent snow under these anticyclonic circumstances is described as follows. The frequency with which cirrus clouds are observed over Greenland and the Antarctic is pointed out. It is then considered that the ice of these clouds is brought down with the descending air current and deposited as ice crystals on the surface of the plateau. Hobbs says : — ' There is however the probability that in general this snow and ice is adiabatically melted and vaporized during its descent to the plateau, and subsequently frozen as it mixes with the cold air above the plateau surface.'* This method of explaining Antarctic snow especially the largo quantities associated with blizzards is quite unsatisfactory and the following remarks by Meinardus are very much to the point : — ' I will consider next a possibility, which however has very little probability. It is that in the central area of an anticyclone the dynamical warming of the descending air is reduced by radiation, which in the dry air of high latitudes may be taken to be very great, and in the regions near the ground the cooUng by radiation may be so great that the air is cooled below its dew point. In this case the small amount of aqueous vapour in the air would be in part condensed as rime, which would replace falling snow as the cause of the ice sheet over the Polar region. Does this process take place to any appreciable extent ? In my opinion it hardly does, for the descending air in the anticyclone is very poor in vapour so that in order to reach the desired end very large masses of air would have to be brought into contact with the sm-face and in addition would have to be cooled very much in order to bring it under its exceedingly low dew point. Now in the central part of an anticyclone the air is usually very calm, hence the renewal of air necessary for appreciable condensation can only take place very slowly and only small quantities of water can be given to the surface. If on the other hand one assumes strong air motion, then the action of radiation and with it the cooling below the dew point ceases to act, and evaporation of rime previously deposited might even occur. On these grounds I greatly doubt whether the ice-covering of the central Antarctic can be due to the formation of rime to any extent. 'f * hoc. cit., page 208. t Deutsche Sudpolar-Expedition, page 327. 32 250 GENERAL AIR CIRCULATION. One must agree mth Meinardus in this matter and there can be little doubt that Hobbs has left unsolved what we shall see in the next section is the greatest problem of the Antarctic anticyclone, namely, the origin of the precipitation within the anticyclone. (2) & (3). Hobbs shows from the records obtained in Greenland and the Antarctic that all alon" the coast and at the outlets from the plateaux there are great accumulations of snow which have been deposited by the \vind and he says : — ' What may be characterised as the centrifugal snow broom which sweeps out snow deposits from the central areas and collects them upon and about the margins of continental glaciers, is a necessary consequence of strong anticyclonic conditions and its work is in evidence within all areas where inland ice has been extensively explored.'* Hobbs is not very happy in his line of reasonmg here, for the snow accumulations do not themselves prove anticyclonic conditions, they only show that the winds blow outwards, of which th^. actual wind observations have given conclusive evidence. The snow accumulations do however prove that the conditions are not cyclonic, for if they were the greatest accu- mulations would be in the interior where the forced ascending currents would be accompanied by gi-eat precipitation. (4) & (6). The blizzard is considered by Hobbs to be one of the strongest factors in support of the general theory of the glacial anticyclone. He therefore goes somewhat fully into the mechanism of the blizzard. As similar explanations have been given elsewhere it is worth our while to examine this theory in some detail. According to it the blizzard is the result of the cooling of the lower air layers which is supposed to proceed until the dense cold air over the inland ice becomes unstable when there is a great outrush of cold air towards the surrounding warm air over the ocean. I will o-ive Hobbs' explanation of blizzards in his own words,! and add a series of remarks to point out weak places in the theory. 'The sequence of events during a bUzzard begins with gentle northerly winds which continue for a day or two during which temperatures are low.' Except that northerly winds naturally occur only in the interval between bhzzards, there is nothing to indicate their connexion. As a matter of fact northerly winds most often occur immediately on the termination of a blizzard and might therefore more justly be connected with the blizzard just passed than with the blizzard to come. We have already shown in the discussion of temperature that the temperature during northerly winds is higher than during similar winds from the south. ' David has suggested that during this time air is flowing south to take the place of air whose volume has been reduced as a result of the heat abstracted from it on the ice sirrface.' This sucfcestion is obviously impossible. Any loss of volume due to coohng would be made good by air flowing inwards in the upper atmosphere and not along the surface where it would encomrter the increased pressure due to the dense air. Also a very Uttle consideration of the chauf^e of volume with temperature would have showm the fallacy of the suggestion even assuming that all the motion takes place along the surface. ' Then there follow two or three days of absolute calm, during which the temperature continues to fall. Still further cooled upon the ice surface, the air, a week or * hoc. cit., page 198. ■)• Loc. cit., page 208. HOBBS' THEORY. 251 more after the calm begins, starts to move outwards in all directions and so develops (on the edge of the Barrier) a south-easterly blizzard.' It would be interesting to know what has been holding the cold heavy air in place on the ice surface during ' the week or more ' that it has been cooling. All theories similar to this neglect the fact that the air will start to move as soon as it commences to coal, ths cons3quenc3 of which is that the cooling might produce a flow of air, but it could njver produca a typical blizzard with its sudden commencement and its violent air motion. ' Simultaneously with this movement the steam cap over the volcano of Erebus, which normally indicates an upper current from the south-west, swings round to the north and takes on an accelerated movement as though it were being drawn from that direction to supply air to the void resulting from the violent surface current towards that direction.' No such action of blizzards on Erebus smoke as that mentioned here was observed by us. We have already discussed th? motion of Erebus smoke during blizzards and have found that the only change is a considerable increase in the frequency of motion from the south- east, so that the relative frequency with which the Erebus smoke moved from the north was less during blizzards than at other times. ' Corresponding to the increased velocity, the normal f ohn effect near the Pole must be much increased as it is also on the descent of the surface current from the plateau. As soon as the warming of the Polar air from this cause has become general, the high air pressure of the central area is automatically reduced, and thus the blizzard gradually brings about its own extinction. To the warming efiect of the descending air current there is rather suddenly added the latent heat of condensation of the moisture when it is precipitated in the form of fine ice crystals nithin the air layer just above the snow-ice surface. The rather sudden termination of the blizzard may be thus in part explained.' Unfortunately for this very ingenious explanation of the end of a blizzard, the temperature observations do not support it in any way. The main rise in temperatm'e which is such a well-marked feature of winter blizzards occurs at the instant the wind rises. The highest temperature is generally reached when the blizzard is at its height, and in nearly all cases the temperature falls appreciably during the last few hours of the blizzard's duration (see the diagrams showing simultaneous wind and temperature during typical blizzards given on pages 48 and 49 above). If Professor Hobbs' theory of blizzard action were correct, we should expect a steady rise of temperature from the beginning to the end of the blizzard and a sudden fall of temperature as soon as the wind dropped. On considering the whole of Hobbs' paper one cannot helj) feeling that in sjjite of his failing to explain the origin of the precipitation and the mechanism of blizzards he has made out a very strong case for the existence of an anticyclone over all extensive masses of inland ice and over the Antarctic in particular. One would therefore be inclined to agree with the generally accepted idea that there is an intense anticyclone concentric with the Pole and covering the whole of the Antarctic Continent. On the other hand, however, Meinardus in his discussion of the results of the Gauss Expedition attacks the theory of the Antarctic anticyclone with gi-eat vigour and one must admit with most convincing success. We will therefore now examine the problem from Meinardus's point of view. 252 GENERAL AIR CIRCULATION. Meimrduss Theory of the Air Circulation over the Antarctic .*—MemaTdviii starts his dis- cussion from the generally accepted statements: (1) In regions of relatively low pressure there exists an ascending air current which, if the consequent cooling is great enough, leads to the condensation of water vapour first in the form of clouds and then of precipitation; (2) in regions of high pressure, on the contrary, there is a descending current in which on account of dynamical warming, not only is the condensation of water vapour prevented, but the air becomes abnormally dry. His reasoning then proceeds as follows : Further one can lay down the law that, excluding very comphcated pressure systems, the precipitation mil exceed the evaporation in regions having a cyclonic air circulation while in regions with anticyclonic circulation the reverse will be the case. In regions where there are changing cyclonic and anticyclonic systems the pressure type which has on the yearly average the greatest frequency, intensity, or duration will decide whether precipitation or evaporation will predominate. It is also possible to proceed in the reverse direction and from the observed conditions of precipitation and evaporation deduce the prevailing type of weather. Now it is known that there is a constant transfer of water chiefly in the form of ice from the interior to the surrounding sea along the whole circumference of the Antarctic and this necessitates, if the climate is not undergoing change, that more water in the form of vapour must enter the Antarctic than leaves it. This consideration carries with it the conclusion that the Antarctic, considered as a whole, resembles a region in which cyclonic conditions predominate over anticyclonic. Having thus proved the necessity for cyclonic conditions over the Aaitarctic, Meinardus proceeds to show how in spite of the increase of pressm-e at sea-level in high latitudes cyclonic conditions prevail over a large area. His method is as follows : With the values of pressiu'e and temperature found by the different expeditions, mean sea-level values for the whole of the Antarctic are calculated for January, July and the year. With these average values at sea-level and assuming a constant fall of temperature with height it is a matter of arith- metic to calculate the pressui'e at different heights. The results of such a calcidation are given in the following table : — Table 131. From Meinardus. S. Latitude. Mean tempekatdre °C. Sea-level year. Mean pkessuke (mm.)- January. July. Year. 2,000 m. 4,000 in. January. July. Year. January. July. Year. 60° 2-8 -10-6 - 3-5 740 575 568 572 443 431 437 70° -1-3 -22-0 -12 8 743 575 563 569 441 423 431 80° -4-3 -28-7 -20-6 748 577 563 567 441 420 427 90° -60 -33 3 -250 750 678 562 666 441 416 424 Diff. 90°— 60° -8-8 — 22-7 -21-5 4-10 +3 -6 -. -2 -15 -13 Colunms 2 to 5 give the assumed values of temperature and pressm-e at sea-level, the sea- level pressure being considered constant throughout the year and to increase by 10 mm. between 60 S. and the Pole The remaining columns contain the calculated pressure at 2,000 * Jleinardus. Deutsche Sudpolar-Expedition, page 326. MEINARDUS'S THEORY. 253 and 4,000 metres for January, July and the year at the difierent latitudes. The last line contains the difference of the factors between 60° S. and the Pole. The conclusions to be drawn from this table are in Meinardus's ow^l words * : — ' The table shows the following : The Antarctic anticyclone under the assumptions made is no longer present at 2,000 metres in the winter and on the average of the year, it has clearly given place to the Polar cyclone at this level. In January on the other hand there is still a slight increase of pressure towards the Pole at the 2,000 metre level, but even in this month there iS a small decrease at the 4,000 metre level. In other words the anticyclone in January reaches up to about 3,000 metres or a little higher, above this the surfaces of equal pressure are nearly horizontal. The conditions deduced for January probably hold for December also ; in November and February the gradient will be similar in sign to that for the winter because the temperature of these months is considerably lower than that for December and January.' The calculations have been based on the assumption amongst others that the anti- cyclone is as fully developed in January as in July. In view of the yearly varia- tion of wind strength this is hardly likely. If the January pressure is decreased relatively to the July pressure the anticyclonic gradient in 2,000 metres is naturally decreased also, and it is possible that it might then even disappear at this height, in other words the anticyclone might not reach 2,000 metres even in January. In spite of the uncertainty of the main assumptions, the important result may be accepted that the anticyclone is a phenomenon of the lower atmosphere in all months and only in the summer rises above the 2,000 metre layer.' Meinardus then refers to his calculation of the height of the Antarctic Continent f by which he has shown that if one-third of the area ■ftdthin the Polar circle is at sea-level the mean height of the remainder must be 2,000 metres. And the conclusion is dr-awn that the surface of the high land within the Antarctic is above the anticyclone and therefore subjected to cyclonic conditions. In this way he relegates the Antarctic anticyclone to a relatively small fringe of the whole region within the Antarctic Circle, and gives the greater part of the area over to the dominance of the Polar cyclone which he calculates exists above 2,000 metres. The consequence of Meinardus's reasoning is most clearly shown by a series of three diagrams which he pubUshed in the April 1914 number of the Washington Monthly Weather Review and which are reproduced here. The diagram reproduced as figure 75 represents ' Diagrammatic cross section of the south Polar regions to show the position of the isobaric siurfaces and directions of the winds. Fig. 75. Meinardus's diagram of Antarctic pressure, vertical. * Loc. cit., page 331. t See page 294 below. 254 GENERAL AIR CIRCULATION. Arrows show the meridiaual components of the wind, shaded areas show the region of the prevailing easterlies. Altitudes in metres.' This diagram is based on the values of the pressure at different heights given in table 131 above. The conditions which would hold if there were no land mass have been repre- sented diagrammatically, and the outline of the continent superposed without altering the position of the isobaric surfaces, the isobaric surfaces eliminated by the land mass being shown by dotted Unes. This diagram shows clearly how according to the theory the greater part of the land surface is subjected to cyclonic air motion. The air streams in from all sides and therefore is compelled to rise over the land surface producing cloud and precipitation to supply the water which flows northwards in the great ice streams. The edges of the land mass only are subject to anticyclonic conditions with descending air currents. c09l ,„ "??> o-'' 0-1 ^^' 3 u„u if — Sw / ^ ^^^\ rN £L. -^./^ ^^^>N v^ ^ i V7V,> Fig. 76. Meinardus's diagram of Antai'Ctic pressure, sea-level. The next diagram, figure 76,* shows the sea-level isobars, it is explained in the original as follows : — ' Sketch of the course of the isobars at sea-level within the south Polar regions. The isobars are intentionally unnumbered except the curve for 740 mm. whose position can * Figures 76 aad 77 have been inverted in order that they may be easily compared with other diagrams in this volume, in which the Rosa Sea area is always shown above the South Pole. MEINARDUS'S THEORY. 255 be plotted with some certainty by aid of past observations. The drawing is planned to present only the jrwbable form of the isobars, considering tbe observed sea-level winds. H, high pressure ; L, low pressiu-e ; arrows, average wind direction.' This diagram then represents what the sea-level pressure would be if the existing sea-level pressure could be extrapolated under the high continent which is supposed to cover two-thirds of the area. The anticyclonic distribution is clearly slio\vn with the maximum of the high pressure near to the South Pole. Figure 77 is a reproduction of Meinardus's remaining diagram, it is described as ' Sketch of the isobars at the 4,000 metre level within the south Polar region. Arrows show the 1^ °^"^- ^ Fio. 77. Meiiiaidus's diagram of Antarctic pressure, 4,U00 metres. average direction of the upper clouds and the prevailing winds on the plateau.' As stated in the text, this diagram has been drawn by making the isobars run parallel to the winds keeping the low pressure on the right of the air motion. The diagram is convincing and the 256 GENERAL AIR CIRCULATION. winds shown * do fit in fairly well with a low pressure area somewhere near where it has been drawn in the diagram. Meinardus's reasoning and conclusions may be summed up as follows : — • As there is an excess of precipitation over evaporation in the Antarctic, as revealed by the outward flow of great ice masses, the conditions cannot be anticyclone. There is certainly an anticyclone over the Antarctic at sea-level, but a calculation shows that at 2,000 metre altitude this has already given place to a large cyclone which is more or less central about the South Pole. Other calculations indicate that the avi?rage height of the Antarctic Continent is 2,000 metres or more, hence the smface of the high land is -svithin the region of the cyclone. Thus the surface over the greater part of the Antarctic is subject to cyclonic conditions with the associated excess of precipitation over evaporation. Thus the theories of Hobbs and Meinardus are totally opposed, the latter deducing that the greater p.irt of the Antarctic is subject to a cyclonic pressure distribution while the former declares that over the whole Antarctic there is a strong anticyclone. In spite of the repetition entailed we must set out clearly the crucial points on which each theory is based. Hobbs' paper practically reduces to proving that whenever observations have been made in the Antarctic, both at ssa-level and on the plateau, the sm-face winds blow outwards from the centre ; this is only possible if the air descends from above in a central calm area and flows outward under an anticyclonic distribution of pressure. To this Meinardus objects that in such a pressure distribution and air circulation the large precipitation necessary to feed the huge glaciers and produce the constant supply of icebergs is quite impossible. On the other hand Meinardus deduces from the few observations of pressure and tem- perature available a probable distribution of these two elements at sea-level over the whole Antarctic. With these values he calculates the pressm-e distribution in the upper air and shows that at 2,000 metres it becomes cyclonic. As the mean height of the continent is probably more than 2,000 metres, its surface is under the influence of this cyclone in which unlimited precipitation is possible. Hobbs' criticism of this theory is hardly worthy of its importance, he shortly states it and then dismisses it in the following curt paragraph. ' Referring to the observations by Captain Scott and others upon the plateau back of the Admiralty Range in South Victoria I-and, Meinardus is cjiiick to seize upon the westerly winds which there prevail as evidence that the anticyclone has at these levels given place to the supposed overlying cyclone ; failing utterly to note that the winds are here blo\ving directly down slope from the ice plateau — that is, radially, other statements in the report are Ukewiss strikingly at variance with facts either known at the time or revealed by later exploration. 'f When two eminent scientists, each of whom knows what he is talking about, come to such diametrically opposite conclusions on the same evidence it is generally safe to conclude that both have some of the truth and neither all the truth. We will therefore now examine the whole question anew and seek a solution, which I believe is not very deeply hidden. * On page 264 below it wUl be shown that strong objection can be taken to some of the \iincl directiona shown by Memardus in this diagram. When these are corrected the pressure distribution must be radically changed to fit them. f Loc. cii; page 212. TTTEORETIOAL DISCUSSION. 257 Wliatever may be the actual distribution of high and low land within the Antarctic, there are two extremes between which it must lie: (a) the whole area at sea-level, and (b) the whole, or the greater part, at some unknown elevation. I therefore propose to show by means of diagrams similar to the one prepared by Meinardus and reproduced as figure 75 above, the vertical distribution of pressure in each of these two extreme cases, and then combine them along a section of the Antarctic the configuration of which we laiow to some extent. Figures 78, 79 and 80 represent these hypothetical vertical sections of the Antarctic south of latitude 50° S. In these diagrams I have chosen a much more contracted vertical scale than the one used by Meinardus. This is in order to reduce as far as possible a fallacious impression whi(h one receives on examining Meinardus's drawing. Any one looking at figure 75 cannot help associating the great mass of the continent, as there represented, with an isolated mountain rising from a level plain. In the latter case the moimtain would have very little effect on the pressure distribution, which would be very nearly the same near to the mountain as at some distance from it in the free air. Hence one uncon- sciously accepts Meinardus's drawing without reaUsing that the great Antarctic highlands will affect the pressure distribution in a very different manner fi'om that of an isolated mountain peak. If, as we have good reason to believe, the Antarctic Continent is an elevated tableland many hundreds of miles across, it is necessary to construct our diagram to give this impression and this can best be done by keeping the vertical scale as small as possible. When this is done one is no longer tempted to run isobars up to and over the surface without pausing to consider how the tableland will affect the pressure. In these diagrams an attempt has been made to represent the pressure changes in the atmosphere by means of Unes. If one calculated the air pressure at each point of the diagrams and joined all pomts at a given pressure by means of a line, we should have a series of isobars similar to those with which we are so famiUar on weather charts, except that they would represent pressures in a vertical instead of a horizontal plane. Such lines would i-ise and fall as one passed through regions of high and low pressure. Unfortunately, however, the actual change in height of such lines is far too small to be represented on our diagrams. For instance, if the sea-level pressure at the Pole were 10 mm. higher than at 60° S., the isobar which touches the sea-level at 60° S. would be raised less than 90 metres at the Pole. This amount is much too small to be shown on any diagram, which extends to 8,000 metres in the vertical direction. In order, therefore, to let the eye easily take in the changes in pressure, the vertical variation of the isobars is greatly exaggerated. This has the disadvantage that the lines drawn to represent the pressure changes are no longer true isobars, for they do not give the true pressure at each height. All that one can say is that the lines on these diagrams show by their rise and fall how the true isobars rise and fall. It must also be clearly understood that the position of the lines in these diagrams is not calculated, they are simply sketched and made to rise and fall according to the conditions which they are drawn to represent. The vertical distance between the lines is constant over each place, but it varies from place to place according to the mean tempera- ture of the air over that place. Thus the vertical distance between the lines in all the diagrams is greatest at the edges on account of the relatively high temperature at 50° S. latitude, and least over the Pole where the temperature is supposed to be least. Again the vertical separation of the lines has not been calculated, but convenient 33 258 GENERAL AIR CIRCULATION. distances taken in order to bring out clearly tlie general effect of raising and lowering the temperature. Figure 78 has been drawn to represent what would be the pressure distribution if the whole area within the Antarctic were at sea-level. The lowest pressure at sea-level is shown at 60° S. and the pressure rises from this point to the Pole. The vertical distance between the isobars depends on the temperature, hence this vertical distance decreases from the edges of the diagram to the centre. The consequence is that the rise of pressure south of 60° S. at sea-level is compensated in the upper atmosphere where the pres.sure decreases all the way from 50° S. to the Pole. In accordance with Meinardus's calculations I have made the height at which anticyclonic conditions over the Polar area give place to cyclonic to be between 2,000 and 3,000 metres. The pressure distribution in this diagram is practically the same as that shown in figure 75 prepared by Meinardus and that it represents the condi- tions which would hold if the whole area withLir the Antarctic were at sea-level no one will doubt. But what does it signify ? Simply that the ice surface causes the sea-level pressure at its centre to be higher than at its edges. This comes about from two causes : (o) the geographical position of the ice surface, which being concentric with the Pole has a lower temperature at its centre than at its margin ; (6) the formation of a glacial anticyclone due to excessive radiation which Hobbs' work shows forms over any large ice surface whether it is concentric with a pole or not. These two causes acting together would produce a relatively intense anticyclone over the Antarctic, if it were at sea-level. We will now turn to the second extreme case and consider the consequences of a high tableland within the Polar area. Figure 79 has been constructed to represent this case. The tableland has been represented as occupying the whole area within 70° S., and to be at a uniform height of 2,000 metres. What is the pressure distribution likely to be over such a vast tableland ? Meinardus assumes that it would be the same as if there were no tableland there at all. With this con- clusion I cannot agree. The pressure distribution in the upper air without the tableland shown in figure 78 depends on the presence of heavy cold air in the lowest atmosphere, when this is removed by the land mass the upper air pressure distribution must be altered. I can see no reason why the general pressure distribution over a level surface at 2,000 metres should be different from that of a similar surface at sea-level, so long as the raised surface is of sufficient extent to be the governing factor in the pressm'e distribution. The causes which produce the high pressm-e over the Antarctic specified in (a) and (b) above both come into play on this raised tableland and it is therefore reasonable to conclude that the pressure will be higher over the centre of the tableland than over its margin. In other words that an anticyclone will exist over the tableland. This has been represented in figure 79. Comparing the isobars sho\vn in figures 78 and 79 we see that the lowest is similar in both except that in the latter it only extends as far as the edge of the tableland near to which it takes a shghtly greater upward turn. The second Hne in figure 78 is nearly horizontal within the Polar region. In figure 79, however, as it approaches the upper margin of the tableland it rises and then passes over the surface indicating the increased pressure towards the centre. If we assume that the free-air temperatures are approximately the same whether the tableland is present or not, the vertical distance between the isobars will be the same at each latitude in figures 78 and 79. The lines have been drawn on this assump- tion and it is seen that at about 2,000 metres over the tableland the isobars are again nearly horizontal and above this height cyclonic conditions agaiji hold. The difference between THEORETICAL DISCUSSION. 259 this fliagram and figure 75 di'awn by Meinardus is fundamental : while he neglects the effect of the high land on the pressure distribution I simply raise the pressure distribution by the height of the land, so that instead of piercing the Polar anticyclone the tableland raises the anticyclone with it and so anticyclonic conditions are retained over its whole surface. Before we can apply the conclusions thus arrived at to the actual area within the Antarctic, we must have some idea of the general distribution of the high and low lands. Figs. 78, 79, 80. Probable Antarctic pressure, vertical cross section. I do not propose to attempt to solve the problem of the geogi-aphical features within the Antarctic, but it is possible to form some idea of which parts are likely to be high and which parts low, and we may take thes^e as giving the general configuration of the land. We know that the plateau at the Pole is about 3,000 metres high, at the position reached by Scott west of Ross Island about 2,500 metres, at the mag:ietic pole about 2,200 metres. 260 GENERAL AIR CIRCULATION. Behind Adelie Land the land rises in 30 miles to 1,000 metres ; and then in a series of undulations to the magnetic pole. Near the Gauss Station Wild found that the land rises to 1,000 metres in about 30 miles and Filchner concluded that a similar rise occurs at Prinz Leopold Land. It is reasonable to assume therefore that at all these points on the coast there is a rapid rise to the plateau. I therefore conclude that the plateau covers the whole segment of the Antarctic between Cape Adare and the Weddol Sea, decreasing in height from the Pole outwards, at first slowly and then fairly rapidly near the coast. We kiaow that the edge of the plateau on the Ross Sea side is very steep and edged by a range of high mountains. Amundsen has shown that this range of mountains is continued beyond the Pole ill such a direction that it probably connects up with Graham Land or with Prinz Leopold Land. As either is equally probable and makes little difference to our discussion I have assumed that the great escarpment which we know runs from Cape Adare nearly to the Pole continues towards Prinz Leopold Land. Thus the Antarctic is supposed to consist of two parts divided by this escarpment, as indicated in figures 81 and 82. The area on the Pacific side of the _gi-eat escarpment, single hatched in the figures 81 and 82, is supposed to be at sea-level or nearly so, while the area on the Atlantic and Indian sides, shown in the same figures by cross hatching, is high tableland, the highest region being near the South Pole. In figure 80 a section of the continent is shown along the 90" E. meridian which is continued beyond the Pole as the 90° W. meridian. This section is shown in figiu'e 82 by the thick line AB. From 50° S. latitude on the left of the diagram we have open water to the Antarctic Circle just north of which the Gauss wintered. From the Circle the land rises to the plateau, at first rapidly and then more slowly, the exact contour of the surface is not known nor is it material to our qualitative discussion. The high land continues a little way beyond the Pole when the rapid descent through the Queen Maud range of moim- tains occurs. From this point to nearly 70° S. there is either barrier or land ice of no great elevation. Near to 70° S. on this meridian the Belgica wintered and therefore sea is known to extend fi'om about 70° S. towards the north. These features are reproduced in elevation on figure 80. The diagram shows clearly how the pressure variation is supposed to be affected by this distribution of land, but the following points should be noticed. On the Gauss side of the plateau the pressure is lowest near to 60° S. in accordance mth the observations made by the Gauss Expedition. From 60° S. towards the plateau the pressure at sea-level rises. Over the plateau the pressme also rises, the highest pressm-e occurring just before the Pole is reached.* Over the area at sea- level between the plateau and the Belgica'' s position there is a well-developed anticyclone in the lower atmosphere and a marked cyclone in the upper atmosphere. Thus in this diagram we have been able to combine the pressure distribution shown in figures 78 and 79. We are now in a position to indicate the probable distribution of pressure at sea-level and at 3,000 metres over the whole Antarctic, if the distribution of high and low land is that assumed. * The remarks made above about the Imes in these diagrams which represent the pressure variation being "reatly exaggerated must be borne in mind. Tlius the fact that one of the pressure lines runs nearly parallel to the surface of the plateau does not mean that the pressure on the surface is nearly constant. The best inter- pretation of the lines is that at any given UUjU the pressure varies in the same sense that the nearest pressure line rises or falls. THEORETICAL DISCUSSION. 26] Figure 81 shows the sea-level pressure. The greater part of the isobars shown are based on actual observations and may be taken as correct, these have been represented by full Fig. 81. Probable Antarctic pressure, sea-level. lines. The dotted lines represent the hypothetical pressure distribution, over regions from which we have no observations. The latter are confined to the area at sea-level between the Ross Sea and Graham Land where an anticyclone is shown in accordance with our conclusion that the sea-level pressure increases towards the Pole. No attempt has been made to indicate the sea-level pressure over the region occupied by high land, for such a result would have no It will be noticed that in its main features this diagram does not depart largely from the corresponding one prepared by Meinardus and reproduced in figure 76. We have in each the bending southwards of the isobars over the Ross and Weddel Seas and the high pressure over the central parts of the area. 262 GENERAL AIR CIRCULATION. In our next diagram, figure 82, an attempt has been made to show the probable pressure at 3,000 metres. This height has been chosen instead of the 4,000 metres used by Memardus Fig. 82. Probable Antarctic pressure, 3,000 metres. for reasons which will be given in the next section. It is possible to outline the pressure distribution at this height from the principles used in preparing figure 80. Isobars * are shown and the method followed in drawing them will be their justification. We know from the pressure and temperature conditions over the Ross Sea and the Barrier that the gradient between these two areas is reversed above about 1,700 metres, so that at 3,000 metres the pressure over the Ross Sea is distinctly higher than over the * It may be as well to point out here that the actual pressure to which the isobars refer is not known and therefore they have simply been numbered 1, 2, 3, etc. The number of isobars drawn has been determined solely with the idea of giving a clear picture and they may be increased or decreased at pleasure, over either the low or high pressure systems, if it is considered that the gradients in each are not in keeping _^with the wind velooitioa. THEORETICAL DISCUSSION. 263 Barrier (see page 136). Tlie starting point was therefore a series of parallel lines over the Ross Sea running approximately parallel to the Barrier edge. As the lowest pressure is doubt- less over the coldest region, it has been shown over the low-lying area on the Pacific side of the Pole. Round this area the isobars already drawn over the Ross Sea must pass. Leav- ing for the moment the ends of the isobars over the Ross Sea, the other ends were continued round the low pressure area imtil they came to 40° W. longitude. Returning then to the ends over the Ross Sea, near the Western Mountains, it was realised that the pressure near the edge of the plateau must be in equilibrium with the pressure at the same height over the sea. The lines were therefore continued on to the plateau. But according to our assumption the pressure of the plateau increases inland, this necessitated that the lines should bend away from the centre and pass around the edges of the plateau. A few trials showed that these conditions were best met by carrying isobars Nos. 2 and 3 to the south and isobar No. i to the north. Isobars Nos. 2 and 3 after turning to the south run nearly parallel with the edge of the plateau and join up with their other ends, completing the closed cyclonic system over the low-lying land. A high pressure system was then indicated over the plateau by means of the closed isobars 4 and 5. Round these two high and low pressure .systems encircling isobars were drawn indicatmg that the pressure increases northwards over the whole region outside the Antarctic. If this pressure distribution is correct, it will be in accordance with the isobars shown in the vertical section of figure 80. The line along which this section has been made is indicated in figure 82 by the thick line AB. We will pass along this line from A to B and compare the pressure changes shown on it with the corresponding changes shown at 3,000 metres on figure 80. Starting on the left of each diagram we see that from 50° S. to about latitude 64 S. the pressure lines at 3,000 metres fall in figure 80, and we cross isobars 4 and 3 to a shallow trough of low pressure in figure 82. From 64° S. to 80° S. the pressure lines at 3,000 metres rise on figure 80 and on figm-e 82 we cross isobars 3, 4 and 5 to the centre of the high pressure area over the plateau. From 80° S. to 75° S. on the other side of the Pole the isobars fall on figure 80 and we pass across isobars 5, 4, 3, 2 and 1 to an area of low pressure on figure 82. From 75° S. to the right hand edge of figure 80 the pressure lines at 3,000 metres rise, indicating increasing pressure, while in the same distance in figure 82 we cross isobars 1, 2, 3, 4 and 5 in the direction of increasing pressure. Thus the horizontal pressure distribution shown in figure 82 agrees with the vertical pressure distribution represented in figure 80. We have now, from considerations of pressure and temperature alone, drawn a system of isobars which enclose a high pressiu-e area over the plateau and a low pressui-e area over the part of the Antarctic which is supposed to be at or near sea-level. The cnicial test of this pressure distribution is whether the observations of wind direction agree with the general rmi of the isobars. The pressure distribution at 3,000 metres has been chosen because this is the height in the atmo.sphere from which we have the most information about the air motion. (a) Erebus smoke gives the air motion over the Ross Sea area at a height somewhat greater than 3,000 metres. (6) The height of the plateau varies between 2,000 and 3,000 meters, hence wind obser- vations made on its sm-face give useful information. 264 GENEEAL AIR CIRCULATION. (c) Observations made in McMurdo Houiid showed that the height of the alto-cirrus and alto-stratus clouds was a little lower than the top of Erebus, so that their mean height may be taken as about 3,000 metres. This enables us to use the direction of these clouds observed at the Gauss Station, and at the South Orkneys. The motion of these clouds over McMurdo Sound was affected by the surrounding land and therefore cannot be used for this purpose. The following table contains all the available data : — Table 132. ! Air motion af approximaieh/ 3.000 metres. Position. Motion ^[^ motion from dcterramed by Reference. McMurdo Sound .... Magnetic Pole Plateau . Western Plateau .... Polar Plateau .... Gauss Station .... South Orkneys .... Belgica Erebus smoke . 1 W.S.W. Wind . . S. to S.E. Wind . . .Is. 31° W. 1 Wind . . . Parallel to 160^ E. meridian. Medium cloud . N. 74° E. Medium cloud . S. 66° W. High cloud * . . S. 66° W. Page 135 above. Page 142 above. Page 142 above. Page 144 above. Deut. Sudpolar-Exped., page 143. Anales d.l. Officina Met. Argentina, Volume XVII, page 124. Mossman. Tran. Roy. Soc. of Edinburgh, Volume XLVII, page 127. These directions of air motion have been shown on figure 82 by means of arrows flying with the wind. It ^vill be seen at once that all the arrows are approximately parallel to the isobars. At the three last stations, however, the direction of motion is inclined somewhat towards the high instead of towards the low pressure. It must, however, be remembered that each of these directions has been found from cloud observations and the presence of cloud itself indicates that the weather is disturbed and if clouds are generally associated with one type of weather their motion cannot be accepted as shomng the mean direction. If allowance .is made for these considerations, it must be admitted that the air motion is in excellent a \ \ \ V ) \ J \ -40 -30 -20 -10 Fig. 85. Upper air temperatures, summer and winter. The numbers given in table 138 have been plotted in figTire 85. Summary of winter conditions. (a) Durmg cold calm weather in the winter there is a temperature inversion in the lower atmosphere which, however, does not appear to extend much higher than 1 or 1| kilometres. (h) The mean temperature gradient above the inversion is between 0'40° and 0-50° C. per 100 metres. (c) There is generally a discontinuity in the wind direction at the place where the gradient changes sign. SUMMER ASCENTS. ■277 These conditions can best be explained by considering that the air in the upper atmos- phere is being cooled by radiation and in consequence sinking. The fact that the temperature gradient is much smaller than the adiabatic gradient indicates that the downward motion is so small that a large amoimt of the heat generated by compression is lost by radiation. Near the ground the radiation is so great and the downward motion of necessity so small that the air is abnormally cooled. Thus a layer of cold dense air forms near to the ground which the descendmg current cannot move and over which it must flow, thus accounting for the difierent winds fouiad at the upper surface of the lower cold layer. Summer Ascents. Table 139. Record Y. November 12, 1911. Time 11-l.j. Erebus smoke from the W. Height. Pres.sure. ^Temperature. Gradient. Height. Pressure. Temijorature. Gradient. Metres. mm. T. '('. per 100 m. Metres. mm. °C. °C. per 100 m. 0 746 - 9 2 1,641 600 -20-8 220 725 -110 1,954 575 -22-5 487 700 -130 2,000 574 -22-8 ■56 500 699 -130 •76 2,279 550 -240 763 675 -150 2,500 537 -25-5 . •54 1,000 656 -16-1 •62 2,617 525 -26-5 1,047 650 -16-5 2,7.57 515 -280 1,339 625 -190 3,000 498 1,500 613 -20-0 •78 A blizzard occurred throughout the 10th and continued until the morning of the 11th. The wind dropped during the afternoon, and from 17 hours on the 11th mitil the time of the ascent (11-15 on the r2th) a breeze between TS and 2-5 metres a second continued from the S.E. During this period the temperature was fairly constant, being about +12°C. Three hours before the ascent the temperatm-e commenced to rise, and when the ascent took place it was -9''C. The sky was nearly overcast with alto-cumulus clouds, and it is probable that the instrument went into, if not through, these. At first the balloon travelled in the surface S.E. wind until the alto-cumulus clouds were reached, and then it moved with them nearly due west. Soon after the balloon had ascended the surface wind increased and th? weather became thick. Some snow fell and it appeared as if a bhzzard were startmg but although the wind continued at about 20 miles an hour the sky cleared, and the blizzard did not become marked until the next day. The mean temperature gradient up to 1,500 m. was •72°C. par 100 metres. At this height the gradient became smaller for the next 1,000 metres, being only -55" from 1,500 metres to 2,500 metres. This probably was the region of movement fi-om the west and of the alto-cumulus clouds. At 2,250 metres the temperature commenced to fall more rapidly again, and from 2,250 to 2,750 metres the fall was at the rate of •84°C. per 100 metres. It will be found that in most of the summer ascents there is a similar region in which there is a smaller gradient than in the regions above and below. 278 UPPER AIR OBSERVATIONS. Table 140. Record VI. November 15, 1911. Time 10-30. A cumulus cloud iiKivini; from N. 10'' K. Height. Pres ,ure. Temperature. Gradieat. Height. Pressure. Temperature. Gradient. Metres. mm. "C. °C. per 100 m. Metres. mm. V. °C. per 100 m. U 750 -10-5 1,683 600 -200 261 725 -110 2,000 575 -21-0 •36 500 703 -12-6 •42 1,9'J7 575 -21 0 528 700. -12-8 2,324 550 -22-8 804 675 -14-8 2,500 536 -24-7 •74 1,000 658 -16-5 •78 2,663 525 -26-2 1,088 650 -17-2 2,831 513 -27-2 1,380 625 -18-7 3,000 500 -28-2 •70 1,500 617 -19-2 •54 The weather conditions during this ascent were very similar to those during the previous one. The ascent was made during a few calm hours between two blizzards. As before, the temperature had been fairly steady until the calm weather set in when it commenced to rise. The sky was 8/lOths overcast with alto-stratus and alto-cumulus clouds. The balloon moved very slowly towards the S.E. and did not appear to change its direction ; evidently the current of air which was moving the clouds from N. 10° E. extended to the ground with only a small change of direction. The gradient in the fii-st 500 metres was smaller than in the previous record, being -42° and •78" C. par 100 mstres respectively. Above 1,000 metres the gradient became much smaller, and between 1,500 and 2,000 metres it was only •36°C. per 100 metres. This reduced gradient obviously belonged only to a layer, for at 2,000 metres the rate of fall of temperature again increased, and from 2,000 to 3,000 metres the mean gradient was •72°C. per 100 metres. The characteristics of the vertical temperature distribution during this ascent were as before — a region in which the gradient was considerably smaller than in the regions above and below. Record VII a, h cmd c. November 19, 1911. The three balloon ascents made on this day are very interesting. For the first time balloons without silk thread were used, the ascent of the large balloon and the descent of the instrument with the small marking balloon attached being followed through a theodolite. At 11-30 the first balloon was sent up. The fuse used for liberating the instrument was timed to burn through in fifteen minutes. The balloon rose ahnost vertically, and a few minutes before the instrument should have been detached, it was right overhead. In this position a small motion of the balloon required a very large motion of the theodohte in azimuth to keep it in the field of view, and it was found impossible to make this rapid motion and the balloon was lost to sight. After waiting until the instrument was estimated to have had time to fall, the theodohte was used to scan the surrounding ice floe to see if the red balloon attached to the instmment was in sight. No trace of it could, however, be seen, and the instrument was given up for lost, when it was seen to fall less than one hundred metres away. The fuse had evidently burnt longer than was expected, and when the record was worked out it was foimd that it had been up nearly 4,000 metres. That after sucb an ascent the instrument should fall so near to the starting point showed that the lower atmosphere was remarkably calm, and I at once determined to make a more serious attempt to get a high ascent. SUMMER ASCENTS. 279 A new balloon and iiistmment were prepared and a longer fuse attached. TL,e second balloon ascended at 12-30 ; but unfortunately the fuse became wrapped romid the string attaching the instrument to the balloon and burnt it through ^\athin fifteen minutes of the start. The mstrument was picked up j kilometre to the N.W., the height reached proved to have been only about 3,000 metres. Later in the day (at 4-30 p.m.) another attempt was made to get a high ascent and this time all went well, but the height reached was only a little over 4J kilometres, thus only surpassing the fii'st by a little over half a kilometre. The mstrument was found on the slopes of Erebus about | kilometre to the north-east of the station. Throughout the day the sky was cloudless. Table 141. Record VII a, b and c. Ascent 1 (a) EkEBUS smoke FROJl 11 H. 30 M. w. Ascent 2 (6) Ekebus smoke fkom 12 H. 30 M. W. Ascent 3 (c) Erebus smoke fkom W. 16 H. 30 M. Height. Pressure. Temper- ature. Gradient. Height. Pressure. Temper- ature. Gradient. Height. Pressure. Temper- ature. Gradient. Metres. mm. °C. °C. per 100 m. Metrep. mm. °C. "C. per 100 ni. Metres. ram. "C. °C. per 100 m. 0 754 -11 5 0 754 -10-5 0 7.54 -110 301 725 -12-6 :U2 725 -12-8 500 706 -13-5 •40 500 706 -14^0 •70 500 708 -13^6 1 •52 567 7(J0 -13-8 577 700 -14-6 567 700 —14-0 842 675 -15-7 851 675 -16-6 840 675 , -16-0 1,000 660 -170 •70 1,000 661 -17-7 •74 1,000 662 -17^0 •68 1,125 650 -18-1 1,133 6.50 — 18-8 1,123 650 -18 0 1,416 625 -20 0 1 ,423 625 -20-7 1,414 625 ' -20 0 1,500 617 -20-5 •70 1,500 619 -21-1 •68 1,500 618 -20^4 •68 1,717 600 -22 0 1 ,723 600 -22 0 1,714 600 -21 5 2,000 578 -23-9 •68 2,000 579 -23^8 •54 2,000 577 -23-6 •64 2,029 575 -24-0 2,035 575 -24 0 2,026 575 -23-7 2,351 550 -270 2,358 550 -26-3 2,350 5.50 -25-0 2,500 540 -27-9 •80 2,500 540 -27-1 •66 2,500 539 -259 •46 2,685 525 -2S-3 2,693 525 -27 4 2,686 525 -26-7 3,000 503 -28-8 •18 3,000 502 -278 •14 3,000 502 -278 •38 3,034 500 -28-8 3,040 500 -27-8 3,037 500 -27-8 3,400 475 -30 0 3,405 475 -28-7 3,500 469 -30-5 •34 3,500 470 -29^0 •24 3,783 450 -32- 1 3.790 450 -30-5 4,000 436 -33-6 •62 4,000 4,193 4,500 4,617 438 425 410 -3V7 -32-7 -34-4 -350 •54 •54 280 UPPER AIR OBSERVATIONS. Taking the mean temperature at sea-level as — 11'0°C. during the three ascents and applying to this tempsrature the mean gradients, we obtain the following as the mean temperature at the different heights throughout the dav : — Table 142. November 19. 1911. Height. Temperature. Mean gradient. Metres. "C. "C. per 100 metres. 0 -110 500 -13-7 •54 1,000 -17-2 ■71 1,500 -20-7 •69 2,000 -23-8 •62 2,500 -27-0 •64 3,000 -281 •23 3,500 -29-6 ■29 4,000 -32-5 ■58 4,500 -35-2 •54 From the figures in table 141, it will be seen that from 500 to 2,000 metres the tem- perature remained practically the same from 11 hours 30 minutes until 16 hours 30 minutes. Above this height the temperature changed in the course of the day by about two degrees, the later ascent showing the higher temperature. It will also be noticed that at about 2,250 metres a different layer of air was met with, for at about this height there is a distinct change in the gradient. The change in the gradient was most marked during the morning ascent, the gradient being more uniform through- out at the afternoon ascent. Thus the gradient from 500 metres to 2,000 was in the three ascents, '69°, "65° and '67° C. per 100 metres respectively, while the gradient from 2,500 to 3,500 metres was -26° at the first ascent and '31° at the third ascent. The first and third ascents again reveal a region of the atmosphere in which the gradient is less than in the regions above and below, and no doubt the same would have been found to hold if the second ascent had gone high enough. On the mean of the three ascents we have the following gradients : — Table 143. Mean gradient on November 19, 1911. 1,500 m. to 2,500 m. 2,500 m. to 3,500 m. 3,500 m. to 4,500 m. •63 ■26 •56°C. per 100 m. The weather preceding the ascent was as follows : On the 17th a southerly blizzard had existed, which, as was usually the case in the summer, was cold, the mean temperature SUMMER ASCENTS. 281 when the blizzard was strongest being about -ig^'C. During the 18th the wind gradually dropped and the temperature rose. The air became calm at 17 hours on the 18th, so that a calm had existed for about 10 hours before the first ascent. Table 144. Record VIII. November 29, 1911. Time 12-30. Erebus smoke from N.W. Height. Pressure. Temperature. Gradient. Height. Pressure. Temperature. Gradient. Metres. mm. "C. "C. per 100 m. Metres. mm. °C. °C. per 100 m. 0 757 - 7-0 2,500 543 -27-7 •92 336 723 -100 2,729 525 -290 500 711 -11-0 80 3,000 507 -302 ■50 604 700 -120 3,076 500 -30-5 881 675 -140 3,440 475 -320 1,000 664 -15-1 ■82 3,500 471 -32-1 ■38 1,166 650 -16-3 3,822 450 -32 8 1,458 625 -190 4,000 438 -33 0 ■18 1,500 621 -19-4 ■86 4,223 425 -33-3 1,760 600 -21-2 4,500 408 -35-3 ■46 2,000 580 -23-1 74 4,645 403 -363 :;.07:i 575 -23-6 4,865 388 -37 0 2,396 550 -26-7 5,000 381 -37-6 ■46 The balloon ascended practically vertically moving first somewhat to the N.W. and then to the S.E. The instrument fell about 2J kilometres to the south-east. At the time of the ascent Erebus smoke was travelling from the N.W., so that the balloon entered the current of air indicated by the movement of the smoke. There were a few cirrus clouds and during the afternoon these thickened, and at 16 hours 7/lOths of the sky were covered with thick cirrus or alto-cumulus clouds which showed fine iridescent colours. Over all the mountains to the west there were heavy cumulus clouds and Erebus became obscured at about 14 hours 30 minutes. There had been very Uttle wind for the preceding forty hours, and during this time the temperature had been fairly steady, varying only a degree or so on either side of — 8°C. The temperature gradient was miusually high, the mean gradient in the first 2,500 metres of the ascent being ■83°C. per 100 metres. Near to 2,500 metres the gradient became less, and from this point the gradient decreased steadily until between 3,500 and 4,000 metres it was only ■18°C. per 100 metres. At about 4,250 metres the gradient again became greater and from 4,223 to 4,865 metres the temperature fell 3'7°C., this being a gradient of •50°C. per 100 metres. In this ascent again the region of small gTadient is clearly shown with greater gradients above and below, but the height at which it occurred was somewhat higher than on the previous occasions. 36 282 UPPER AIR OBSERVATIONS. Table 145. Record IX. December 24, 1911. Time 12-15. Erebus smoke from N.E. Height. Pressure. Temperature. Gradient. Height. Pressure. Temi)erature. Gradient. Metres. mm. °C. °C. per 100 m. Metres. mm. °C. °C. per 100 m. 0 761 - 3 3 2,157 575 -19-2 110 750 - 3-7 2 480 550 -20-9 383 725 - 5-3 2,500 548 -210 .62 500 714 - 6-1 •56 2,822 525 —22-8 650 700 - 7 3 3,000 513 -24 0 ■60 938 675 - 9-5 3,179 500 -253 1,000 671 -100 •78 3,500 479 -27-5 ■70 1,227 650 -11-9 3,547 475 -280 1,500 628 -140 •80 3,933 450 -29-9 1,525 625 -U-2 4,000 447 -30-1 ■52 1,833 000 -16-6 4,159 436 -30-8 2,000 587 -17-9 •78 The weather had been cabn for 18 hours before this ascent took place, and the tempera- ture had shown a well-marked daily variation. The balloon rose almost vertically for 18 minutes when the instrument was detached and fell about 1| kilometres to the W.N.W. The gradient in the lower atmosphere (neglecting the part near the ground where the record is doubtful) up to 2,000 metres was about •78°C. per 100 metres. Above this the gradient decreased, being only -60 C. per 100 metres between 2,500 and 3,000 metres. In the next 500 metres the gradient increased to •70°C., but above this it again decreased. The direction of Erebus smoke was very abnormal and it is possible that the second decrease in the gradient was caused by this upper current. Table 146. Record X. December 25, 1911. Time 13 hours. Erebus smoke from E. Height. Pressure. Temperature. Gradient. Height. Pressure. Temperature. Gradient. Metres. mm. "C. °C. per 100 m. Metres. mm. °C. °C. per 100 m. 0 759 - 30 3,167 500 -22-9 96 750 - 3 9 3,500 478 -256 ■70 363 725 - 5-4 3,541 475 -25-8 500 711 - 6^0 •60 3,931 450 -28-6 630 700 - 6-7 4,000 448 -29-1 ■70 918 675 - 8-9 4,337 425 -31 U 1,000 668 - 9^5 ■70 4,500 417 -32^2 ■62 1,209 650 -11-5 4,764 400 -340 1,500 626 -13-5 ■80 5,000 387 -355 ■66 1,508 625 -13-6 5,214 375 -37 0 1,816 600 -160 5,500 358 -390 ■70 2,000 585 -17^1 •72 5,685 350 -400 2,135 575 -17-6 6,000 335 -42^1 ■82 2,465 550 -20 0 6,208 325 -43-2 2,500 548 -20^2 •62 6,500 314 -44^8 ■54 2,808 625 -21-2 6,743 300 -46-2 3,000 513 -22^1 •38 SUMMER ASCENTS. 283 On the morning of December 25th, 1911, it was obvious from the smoke of Erebus that the lower atmosphere was remarkably calm, and therefore an attempt was made to "et a very high ascent. A fuse burning for over 40 minutes was used, and silver paper was attached to the instrument to make it more visible when far away. The ascent has already been described (page 217) and it was so successful that the instrument was recovered after it had been nearly 7,000 metres high. The record revealed four regions in the atmosphere : (a) From th's ground to 2,000 metres the gradient was moderately high with a mean value of -71 C. per 100 metres. (b) From 2,000 to 3,000 metres the gradient was irregular, indicating various layers. In this region the gradient was small, being on the average -SO^C. per 100 metres, but from 2,-500 to 3,000 metres only SS^C. (c) From 3,000 to 6,000 metres the gradient was again higher, being on the avera"-e 70°C. per 100 metres. (d) Above 6,000 metres the gradient fell to ■Si'C. per 100 metres; the ascent, however, ceased at 6,743 metres. In (a), (b) and (c) we recognise the three layers which have been such a marked feature in all the previous ascents in the summer. The decrease of the gradient at 6,000 metres is significant, but the ascent was not sufficiently high to make sure that this was caused by an approach to the stratosphere. It is exceedingly disappointing that this ascent did not continue for another 1,000 metres higher, as it proved to be the last one which we were able to make. The mean giadient from sea-level to 6,750 metres shown by this ascent was •64" C. per 100 metres. Combined Summer Results. Ascents were niade on six days during the summer months of November and December. It has been shown that in each one of these ascents a region was met with in which the gradient was smaller than in the regions above and below. The following table gives the approximate height of this region : — Table 147. Approximate height of the beginning and end of the region of small gradient. Date. Lower limit. Upper limit. ra. m. November 12th 1,.500 2,300 November 15th 1,100 2,000 November 19th 2,500 3,500 November 29th 2,.500 4,200 December 24th 2,200 3,000 December 25th 2,000 3,000 Mean 2,000 3,000 284 UPPER AIR OBSERVATIONS. In the following tabic the gradient measured in each ascent is shown :- Table 148. Mean gradient during the summer. °G. per 100 metres. Date. Om. to 500 m. 500 m. to 1,000 m. 1,000 m. to 1,500 m. 1,500 m. to 2,000 m. 2,000 m. to 2,500 m. 2,500 m. to 3,000 m. 3,000 m. to 3,500 m. 3,500 m. to 4,000 m. 4,000 m. to 4,500 m. 4,500 ill to 5,000 m. 5,000 m.^ 5,500 m. to to 5,500 m. 6,000 m. 6,000 m. to 6,500 m. November 12th . •76 ■62 •78 ■56 ■54 November 15th . •42 •78 •54 •36 ■74 •70 November 19th . •54 •71 •69 •62 •64 •23 •29 •58 ■54 November 29tli . •80 •82 •86 •74 •92 •50 ■38 ■18 ■46 ■46 December 24th . •56 •78 •80 •78 •62 •60 ■70 ■52 December 25th . Mean •60 •70 ■80 •72 •62 •38 ■70 ■70 •62 ■66 ■70 •82 •54 •61 •73 •74 ■63 •68 •48 •52 •50 ■54 ■se ■70 •82 •54 ■67 •68 •58 ■51 ■55 ■76 By combining the results in this way the region of least gradient appears to be between 2,500 and 4,000 metres, but this is due to the irregularity in the number of ascents which reached higher than 2,500 metres. Taking the mean temperature on the ground at the times of the ascents and applying the mean gradients as shown in the above table we get the following as the mean state of the atmosphere during the summer months as shown by these few ascents : — ■ Table 149. Mean vertical temperatures during six days in the summer. Height. Temperature. Height. Temperature. m. "C. m. 'C. 0 - 73 3,500 -292 500 -103 4,000 -31-8 1,000 -140 4,500 -34 4 1,500 -177 5,000 -37^2 2,000 -208. 5,500 -40^8 2,500 -242 6,000 -44-8 3,000 -26-6 6,500 -47^6 These numbers have been plotted in figure 85. Summary of Summer Conditions, (a) The mean gradient from the ground to 2,500 metres is approximately •eS'C. per 100 metres, but individual ascents showed gradients of between '8° and '9° in parts of this region. SUMMER ASCENTS. 285 (6) Between 2,000 to 4,000 metres every ascent showed a region in which the gradient was less than in the regions above and below. The mean gradient in this region was •54°C. per 100 metres. (c) Above the region described in (b) there was an increase in the gradient. {(I) In the highest ascent the gradient again decreased above 6,000 metres. These results show a gradient in the lower atmosphere as large as that found over land in temperate regions during summer, and occasionally approached the adiabatic gradient in dry air. Such a gradient indicates considerable forced motion in the atmosphere and as at the times of the ascents there was practically no wind this motion could only have been caused by convexion currents resulting from the sun shining on the snow surface. The decrease in temperature gradient which occurs between 2,000 and 4,000 metres is probably due to the fact that in this region the air has to adjust itself to the different conditions of pressure and temperature which exist at this height over the Ross Sea area and over the high tableland to the west. The height of this tableland as found by Captain Scott in November 1903 is approximately 2,500 metres and as the temperature gradient over the high snow-covered plateau must be different from that in the free air at the same height over a region at sea-level, some adjustment must take place which is shown by the reduced temperature gradient. Above this region of adjustment normal temperature gradients are re- established CHAPTER IX THE HEIGHT OF THE BARRIER AND THE SOUTH POLAR PLATEAU. The height of the Barrier. Nearly all the sledging parties which visited the Barrier had aneroid barometers and from their records it is possible to determine the mean height of the Barrier with some certainty. The barometers were compared with the standard at Caps Evans when each party left and when it returned. From these comparisons the corrections to be applied were determined. The corrected readings have been compared with the simultaneous pressure at Cape Evans and the differences tabulated. The track of all the parties on the Barrier was practically the same and has been shown on the map which forms the frontispiece to this volume. After leaving Hut Point the Barrier was reached at Safety Camp from which point the track went almost due east to Corner Camp in 78" 3' S., 168" 59' E. At this camp, as its name impUes, the track turned sharply to the south and the whole of the remainder of the journey to the foot of the Beardmore Glacier was made on or near to the 169" E. meridian. The observations have been collected into convenient geographical groups and the mean difference between the simultaneous pressure on the Barrier and at Cape Evans determined for each group. The results are contained in the following table : — Table 150. Mean pressure difference between Cape Evans and the Barrier during the period November 1911 to January 1912. Position. NumV.er of observa- tions. Mean pressure difference. Position. Numle. of observa- tions. Mean pressure difference. Inches . Inches. Hut Point . 9 -001 80° 10' to 80° 30' 21 -fO-20 Hut Point to Comer Camp. 38 +008 80' 30' to 80" 50' 40 0-22 Corner Camp 32 0 09 80° 50' to 81° 10' 16 0-17 Corner Camp to 78° 30' S. 41 015 81° 10' to 81° 30' 15 0-20 78° 30' to 78° 50' 32 018 81° 30' to 81° 50' in 0-21 78" 50' to 79° 10' 36 0-20 81 50' to 82° 10' 14 n-25 79° 10' to 79° 30' 44 019 82° 10' to 82" 30' 12 019 79° 30' to 79° 50' 28 019 82° 30' to 82° 50' 13 0-19 79° 50' to 80° 10' 24 0-20 82° 50' to 83° 10' 15 0'15 THE HEIGHT OF THE BARRIER. 287 The numbers in this table have been plotted on a curve in figure 86, from which it vviU be seen that the pressure relative to Cape Evans falls to 79° S. and then remains constant to 83° S. •30 •20 CO 0) o c ■10 •00 ■" ) o O y^ "o" O O O 0 / / 'q 7 "■f- , , 1 1 1 1 1 1 1 1 $ 7b' 79° 80°' Latitude 81' 82° ^ FlQ. 86. Dillerence of pressure betweeu Cape Evans and positions on the Barrier. The first question we have to decide is how much of this pressure difference is due to pressure gradient and how much to difference in height ? It is quite impossible from the barometer observations alone to answer this question, but the air motion throws considerable light on the subject. Both the wind observations and the sastrugi show that the prevailing wind over the west of the Barrier is from the south. Thus over this region the isobars must run almost parallel to the track followed by the sledging parties. This leads us to the conclusion that there can be little or no pressure gradient along the track, if there is any change at all it will be a slight increase towards the south. Thus the whole fall of pressure observed must be due to change of height, but there may be a slight additional change of height compensated for by a slight increase of pressure. The latter, however, cannot be determined beyond the fact that it is small and therefore may safely be neglected. We will therefore assume that the whole pressure difference between Cape Evans and the Barrier is due to difference of height ; the height of the Barrier can then be determined from the pressure and temperature observa- tions. The average pressure and temperature on the Barrier during the period over which the observations extend was 29'50" and 15°F. respectively. South of 79° S. the pressure on the Barrier is uniformly ^20" below the pressure at Cape Evans. These data give 170 feet = 52 metres as the mean height for the Barrier south of 79° S. At Corner Camp the pressure is ^09 inches below that at Cape Evans which corresponds to an elevation of 80 feet. It is interesting to notice that Mohn calculates from Amundsen's observations that the mean height of the east of the Barrier is 60 metres, it therefore seems very probable that the true height of the Banier is between 50 and 60 metres above sea-level over the greater part of its immense expanse. Thus the chief heights along the track from Cape Evans to the Beardinore are :— Cape Evans — Sea-level, Corner Camp — 80 feet. One Ton Camp (79i° S.) to the Beardmore— 170 feet. 288 THE HEIGHT OF THE BARRIER AND THE SOUTH POLAR PLATEAU. Heights on the Beardmore Glacier. The barometer used by the Polar Party was compared during the ascent of the Beardmore Glacier 24 times with a hypsometer, the mean correction found was --05". As the hypso- raeter was broken before the top of the Beardmore was reached, the comparisons could not be continued throughout the journey. The barometer had been set to be correct when leaving Cape Evans, and as it did not return for recomparison it has been assumed that the small change of -05" found by the hypsometer tests was real and this value has been used throughout. The barometer thus corrected agrees well with other barometers which were being read by other parties in its neighbourhood. For purposes of the calculation of heights the journey from the Barrier to the Pole and back will be divided into four parts, namely : — (1) From Pony Depot on the Barrier to Mid-Glacier Depot. (2) From Mid-Glacier Depot to Upper Glacier Depot. (3) From Upper Glacier Depot to 3° Depot. (4) From 3" Depot to the Pole.* In the first three intervals the method of determining the height was as follows : When going from A to B the last barometer reading at A and the first at B were extracted from the meteorological register and the mean temperature in the interval determmed. From these the difference of height was calculated as though the pressure observations had been made simultaneously. As there was usually several days between the actual observations it was necessary to correct for any general change of pressure during the interval. To do this the pressure and temperature at Cape Evans were taken for the same times and from these the equivalent change in height was calculated. It was then assumed that the whole of the atmosphere had been raised or depressed by this amount so that this value was apphed as a correction to the observations made on the Beardmore. Luckily there were no large pressure changes durin'^ the peiriod the Polar Party was on the Glacier so that this correction is always small and therefore any error in the assumption is of subordinate importance. An example will indicate the method. Detailed Calculation of the heigJit between the Barrier and Mid-Glacier Depot. Polar Party Outward. Date. December 9th. 6 hrs. December 17th, 19 hrs. Fony Depot Mid-Glacier Depot Pressure. 29-65 1 •JU'lsf Change. 3-47 Ditference Temperature correction Cape Evans pressure. December 9th. 6 hrs. December 17th, 19 hrs. 29-96 1 20-76/ Mean temperature. 23°F. Approximate height table 20. - 55 + 128 Temperature correction ... Error due to pressure change 182 9 173 Mean temperature •20'F. Approximate height from table 20 Smithsonian Tables 228 3,6U 3,3S3 - 207 3,176 - 173 Correction for pressure change. 3,003 Ditference height. Temperature correction from table 21 Smithsonian Tables 3,0U0 feet - 184 300 „ - 18 80 „ - 5 207 * For geographical positions, see table 152, page 291, HEIGHTS ON THE BEARDMORE GLACIER. 289 No correction has been applied for the humidity of the air and for latitude as both of these are much too small to be considered. Difference in height between Pony Depot and Mid-Glacier Depot. Feet. (a) Polar Party Outward 3,003 (6) Polar Party Returu 3,203 (c) First Return Party 3,231 Mean . 3,147 (d) Simultaneous Observations . . . . . . . . .3,117 Mean . 3,132 (a), (b) and (c) were obtained by the method described above. The same barometer was used for (a) and (b) and a different one for (c), it is therefore gratifying to see that the values for (b) and (c) are so nearly alike. The mean value from the three sets of observations («), (b) and (c) is 3,147 feet. When the Polar Party was at Mid-Glacier Depot on their outward journey Meares was on the Barrier only a short distance from the foot of the Beardmore (82° 21'). Thus we have simultaneous observations with the two stations not very far apart. The observations reduced ia a similar manner give the value (d) which is for all practical purposes identical with the mean of (a), (b) and (c). As a simultaneous observation is of more value than observations taken by the method used for (a), (b) and (c) ; {d) has been given the same weight as the mean of the other three observations and the difference in height between the Pony Depot and Mid-Glacier Depot is thus determined to be 3,132 feet. This determination must be considered highly satisfactory as the values obtained by the two methods and with three different barometers agree so well together. Difference in height between Mid-Glacier Depot and Upper Glacier Depot. Feet. (a) Polar Party Outward 3,870 (6) Polar Party Return 3,839 (c) First Return Party 3,838 Mean . 3,849 These three determinations agree so well that their mean may be accepted with great confidence, (c) was again obtained with a different barometer from (a) and (6). Difference in height between Upper Glacier Depot and 3° Depot. Feet. (a) Polar Party Outward 2,353 (6) Polar Party Return 2,177 Mean . 2,265 For this interval observations made by the Polar Party only are available, and the two determinations do not agree with one another so well as the previous ones. Fortunately there is a simultaneous observation which checks the whole determination for the Barrier to 3° Depot. 37 290 THE HEIGHT OF THE BARRIER AND THE SOUTH POLAR PLATEAU. The First Return Party left the main Polar Party at the Upper Glacier Depot and by the time the Polar Party was at 3° Depot the First Return Party was on the Barrier in 80 54' S. The simultaneous observations give the difference in height between the Barrier and 3° Depot as 9,198 feet ; while adding the three values for the stages we have 3,132+3,849+2,265=9,240 ; these two values agree better than might have been expected their difference being only 48 feet, we therefore can take their mean 9,222 feet as being the true difference in height between the Barrier and 3° Depot. Height of the South Pol.a.r Plateau between 3° Depot and THE South Pole. The changes of height on the plateau are relatively small and the method used above is no longer applicable. The secular changes of pressure are now of the same order as the changes of pressure due to change of height and therefore cannot be eliminated by the changes of pressure at the base station which also is now nmch further away. The new method was the following : The daily routine of the Polar Party was very regular, each day the march occupied about 14 hours and the night rest 10 hours. The barometer was read always, at the bsginning of the march, at the lunch camp and at the end of the march. Thus before and at the end of each march the barometer was stationary for 10 hours. The change in barometer during each of tbe periods of rest was determined and the mean rate of change before and after each march calculated. This mean rate was then assumed to have remained constant diiring the period of march and the observed difference of pressure corrected accordingly. The following example will make the method clear : — Calculation of change in jpressure due to change in height during a march on the plateau. 1 2 3 4 5 6 i 7 8 9 10 11 12 Date. Time. Difference of time. Bar. Difference of bar. Rate of change of bar. during rest per hour. Mean. Change of bar. during march due to pressure change (8)X(4) Change of bar. due to change of height (0)-(9) Tempera- ture. Mean tempera- ture. Rest . . \ March . . } Rest . . i January. 1 2 2 3 9 P.M. 1 C A.M.' 8 P.M.]" G A.M. I 9 14 10 2014 20 12 19-96 19-91 > -02 \- I -05 -0022 1 -0050 J -■003G -05 -1 -11-5) -13 In this example we see that during the first night rest the barometer changed at the rate of —-0022" per hour and during the second at the rate of —-0050" per hour. From this we conclude that the barometer changed, due to pressure change, at the mean rate of -■003G" per hour during the march. As the march lasted 14 hours the total barometer change due to pressure change was —0036x14=— "05", while the actual observed change was --16" thus leaving — •16+-05=--H" to be caused by change of height. A change of -11" with a mean barometer of 20" and mean temperature of -13°F. gives a change of height of 131 feet. HEIGHTS ON THE SOUTH POLAR PLATEAU. 291 By this method we find that the change of height in each half degree of latitude between 87" S. (3" Depot) and the South Pole to be as follows :— Table 15L Okmje of ImgU in each half degree of latitude on the Polar Plateau along the 160" E. meridian. Ascent -j-) Descent -. Outward Return . Mean 87= 87° 30' 88 88° 30' 89' 89= 30' 1^1*'' +232 + 160 + 100 -320 -270 -210 feet. +287 + 160 +000 -151 -370 -260 feet. +260 + 160 + 50 -235 -320 -235 feet. From this table we see that the height of the plateau increases to about 88|° S. and then decreases to the Pole. The mean values have only to be added successively to the height of 3° Depot, determined above, to give the height of the various positions on the plateau. The following table gives the heights of the various points from Pony Depot to the Pole. Table 152. Distance station. Latitude. Longitude. Height above Barrier. Height above sea -level. from Pony Depot along track. S. W. Feet. Pe.- 1. Geo. miles. Pony Dej)6t 82= 47' 170= 45' 0 170 0 Lower Glacier Depot 83' 30' 171° 30' 355 525 45 Mid-Glacier Depot . 84° 24' 170° 3,132 3,302 99 Upper Glacier Depot 85° 07' 163° 6,981 7,151 l.'ii 3° Depot .... 86° 56' 163° 9 222 9,392 273 Plateau . 87° 30' 160° 9,482 9,652 307 Do 88° 00' 160° 9,642 9,812 337 Do. 1J° Depot , 88° 30' 160° 9,692 9,862 307 Do. .... . 89° 00' 160° 9,457 9,627 397 Do 89° 30' 160° 9,137 0,307 427 South Pole .... 90° 00' — 8,902 9,072 457 292 THE HEIGHT OF THE BARRIER AND THE SOUTH POLAR PLATEAU. The height of the stations have been plotted against the distance along the track in figure 87. ( in nn n d/Ve 900( J — cr" 3*D8pOl />*? /« .j.^ ■V. ■V. y >- r- X 7i I0( ^ IpfM. Depot I i. "/■' 3 U >^« M*< / / 4 3 2 00 ' / f / acie pot •^ )- / / 1^ / / ►^ r- ^ >i 1/ act wD f" epo /•3 ;■/■ o o o OPOOPOOOOOOQOOOOO OOOOQOQOOQOOOOQOOOOO OOOO &«). Miles from Pony Depot along tracli Fig. 87. Heights on the Beardmore Glaciei- and South Polar Plateau. This determination of the height of the various depots from the Barrier to the Pole leads to the conclusion that the height of the South Pole is 9,072 feet, or 2,765 metres, above sea-level. Mohu in his discussion of the meteorological observations made on Amundsen's journey to the South Pole has also determined the height of the different points on Amundsen's track and has found the height of the South Pole to be 2,454 metres or 8,051 feet. It will be seen that the difference between the two determinations — 311 metres or 1,021 feet — is extreme and much beyond the limits of accuracy to be expected. Mohn calculates that his determination of the height of the Pole is accurate to ±196 metres. Mohn treated Amundsen's barometric observations rather differently from the way Scott's observations have been treated here. Instead of dividing the journay from the Barrier to the Pole and back into a small number of distinct stages, he calculated the change in height during each day's march. This method appears to me to be objectionable in that the un- avoidable errors in reading the barometer enter into the determination of the height each day and the errors may or may not cancel out. This, however, is not the chief error, for a much more serious one is that Mohn has not made any allowance for the general change in barometer pressure. As it happens this is rather serious. Amundsen left, the Barrier on November 15th, was at the Pole on December 17th and reached the Barrier again THE HEIGHT OF THE SOUTH POLE. 233 on January 9th. Thj mean pressure at Cape Evans and Framheim on these days was the following ; — Table 153. Pressure charuje at sea-level. Date. Cape Evans. November 15th December 17th January 9th . 29-58 29-80 29-12 Framheim. 29-54 29-76 29-20 Mean. 29-56 29-78 29-16 Change. + ■22 It will be noticed that the general pressure rose as Amundsen went from the Barrier to the Pole and fell as he returned, both these changes would make the difference in height b^ween the Barrier and the Pole appe'ar less than it really was. The mean change in pressure was -42 inches. This is equivalent to a change of height of approximately 400 feet, by which amount the height of the Pole would be underestimated. Even with this correction, Mohn's determination of the height of the Pole would only be about 8,500 feet, leaving another 500 feet to be accounted for. I have tried to apply the method used in reducing Scott's observa- tions to those taken by Amundsen but have not been successful, as Amimdsen's marches on the plateau were more irregular than those of Scott and his track on the plateau was much less level. I have, however, been able to use Amundsen's observations to show that the height of the Pole determined from Scott's observations is not in error by 200 feet. °it seemed very desirable in the first place to be assured that there was no radical error in either of the barometers. This seemed very unlikely for Amundsen's barometer was fre- quently checked by two hypsometers while on the plateau and Scott's barometer was also checked by a hypsometer a short time before the plateau was reached. Luckily a direct comparison is possible which shows that the barometers were reading with the accuracy to be expected under such conditions. On December 31st, 1911, Amundsen was at 87° 9' S., 167" 7' W. and Scott was at 86" 56' S., 165" 6' E., thus they were approximately only 87 geographical miles apart. According to Mohn's calculation Amundsen was then 1 metre below the elevation of the Pole; while Scott was, according to the above calculations, 320 feet above the Pole. As the positions were so near to the Pole, both these differences may be accepted quite independently of the actual height assigned to the Pole. Thus on this day Amundsen was 323 feet below Scott. The pressure and temperature recorded were the following: Amundsen 6 p.m., barometer 20-67;", temperature -2°F. ; Scott 2 p M barometer 20-28", temperature - 10-5°F. If the difference of height is correct Scott's barometer should have read -27" below Amundsen's while we see that it was -39" lower. Thus the barometers differ by only -12" from their true relative readings. Part of this difference may be due to the fact that the barometers were not read simultaneously, part to the distance between the two observers, and part to errors m the determmation of their relative^ heights. In any case it is quite obvious that there was no difference in the barometers sufficient to account for the discordant determinations of the height of the Pole. Delermmcition of the height of the Pole hy simultaneous observations of Amundsen's Barometer ai the Pole and or^e on the Barrier near to the foot of the hills. -Amm^dsen was at, or within six miles of, the Pole throughout January 16th, 17th and 18th. The mean P-ssure and temperature observed by him during these three days (three observations each day were 20-48 and -7"F At the s^me time Meares mth Captain Scott's dogs was near to the toot ot the Beardmore in 82" S. His mean pressure and temperature for the same days (also three 294 THE HEIGHT OF THE BARRIER AND THE SOUTH POLAR PLATEAU. observations each day) were 29- 62" and 24''F. These observations give the difference in height batween the Pole and the Barrier as 9,170 feet. Taking the height of the Barrier to be 170 feet as before, this gives the height of the Pole as 9,340 feet above sea-level ; which is only 238 feet higher than the value determined above. Thus it strongly supports our determination, but is still further away from Mohn's determination of 8,051 feet. It will be noticed that this is a determination quite indepjiident of Scott's barometer, and depends only on Amundsen's observations and those of a barometer which was compared with the standard at Cape Evans only a few days later. Determination of the height of tJie Pole by simultaneous ohservations of Scott's and Amundsen's Barometers. — On January 12th Scott was in 83" 52' S., i.e., within 68 miles of the Pole, while Amundsen was on the Barrier in 84^ S. This day has been chosen for the comparison because the pressure was very steady both on the plateau and on the Barrier, also Amundsen was then travelling rapidly northwards and if we delay the comparison imtil Scott was at the Pole, Amundsen would then be so far north (in 82^ S.) as to make the comparison less valuable. Scott's position was then 660 feet above the Pole as determined by concordant observations made on the journey to and from the Pole. The observations were : Amundsen 2-30 P.M., barometer 28-77", temperature +200''F. ; Scott 1-30 p.m., barometer 19-36*, temper- ature -22°F. These reduced in the usual way give the difference in height between the two positions to be 9,649 feet, deducting the 660 feet which Scott was above the Pole and adding 150 feet for Amundsen's height on the Barrier (according to Mohn) we find the height of the Pole to be 9,104 feet above sea-level. This value is only 32 feet different from our main determination and may be considered to support it completely. Summing up we have the following independent estimates of the height of the South Pole above sea- level : — (1) Mohn's determination from Amundsen's observations (2) Our chief determination mainly from Scott's observa- tions *......... (3) Comparison of Amundsen's barometer at the Pole and a barometer on the Barrier ..... (4) Comparison of Scott's barometer near the Pole and Amundsen's on Barrier ..... The mean of the last three determinations is 9,172 feet = 2,796 metres which is probably very near to the true height of the South Pole above mean sea-level. The height of the Antarctic Continent. Our knowledge of the land masses within the Antarctic is very limited, it is therefore exceedingly interesting to find that an estimate of their extent can be made from purely meteorological considerations. This has been done by Meinardus who concludes that if land occupies two-thirds of the whole area within the Antarctic Circle its average height must be 2,000 ± 200 metres. Such a mass of land is so great that it raises the mean height of the earth's whole surface from 205 to 240 metres, or the mean height of all known land is raised from 700 to 825 metres. Such a result is of great scientific importance. Now that Shackleton's attempt to cross the Antarctic Continent has failed, we are unlikely to obtain confirmation of Meinardus's conclusion from actual exploration for many years to come, hence in the interests of both geodesy and meteorology Meinardus's conclusions need critical dis- cussion. This must b3 my excuse for the following pages. * la Volume III will be found the detailed data on which this determination is based, and the height of all the camps. Feet. Metres. 8,051 2,454 9,072 2,765 9,340 2,847 9,104 2.775 THE HEIGHT OF THE ANTARCTIC CONTINENT, 295 Meinai'dus's work * may be summarised as follows :— The barometric pressure at any place is a direct measurement of the mass of air lyin^ over that place. If therefore we integrate the pressure over the whole surface of the ulobe (actual pressure, not pressure reduced to sea-level) we obtain the total mass of the earth's atmosphere. This mass must be constant, hence the integral of the pressure over the whole world must be the same at all seasons of the year. The pressure conditions are sufficiently well known over the globe outside the Antarctic Circle to make it possible to effect this integration, and Meinardus has made the necessary calculations for the two months January and July from charts of the normal pressure for these two months. The result of the calculations shows that outside the Antarctic Circle the total mass of air is greater in July than in January. Thus in order to have the total mass of air the same all over the world in these two months there must be less air over the Antarctic in July than in January. In other words the integral of the pressure over the area within the. Antarctic Circle must be greater in January than in July and the mean difference in pressure January-July must be U mm. This seems contrary to what one would expect, for in July the temperature of the air is much lower than in January, hence we should expect the air to be denser and there- fore the pressure higher in July than in January. Meinardus shows that this apparent paradox can be explained by assuming that the land surface within the Antarctic Circle is at a relatively great height above sea-level. To do this he first discusses the mean pressure and temperature over the Antarctic in the months of January and July respectively, using all the data available in 1909. This leads him to believe that the mean pressure over the Antarctic at sea-level is the same in these two months and equal to 745 mm. The mean temperature which he deduces are January -S^C, July -26°C. He assumes further that the temperature decreases by 5''C. for each 1,000 metres of ascent during both months. Meinardus's reasoning then proceeds as follows : — • Let A, figure 88, represent a column of air over the Antarctic Continent in January. The pressure at the base (sea-level) is then 745 mm. and the temiJerature -3°C. The a. 4 = /"-/s /. - / t'-Sc - '- y bi a) ^4 in 1 ± b, f>, — 6 =745 A January B July Fig. 88. Diagram of vertical pressure, aumiuer and winter. * Meinardus, ' Die mutmassliclie mittle Hohe des antarktiachen Kontinents, ' Petormanns Mitteil, 1909 Heft 11 and 12. 296 THE HEIGHT OF THE HARRIER AND THE SOUTH POLAR PLATEAU. vertical temperature gradient being fixed at 5°C. per 1,000 metres the temperature at all points in the atmosphere is known. It is therefore a simple matter to calculate the height at which the pressure has any given value. In figure 88 horizontal lines are drawn in column A to represent the height at which the pressure has the consecutive values bj, b2, h^, h^, etc. In the same figure column B represents the same column of air in July. The pressure at sea- level is again 745 mm., but the temperature is — 26''C. Lines have again been drawn to represent the height at Avhich the pressure has the same values bj, 1)2, h^, h^, etc. It wall be noticed that in B each of these lines is lower than in A and the difference is greater in the upper than the lower part of the figure. This is because the pressure decreases more rapidly in a cold air column than in a warm one. At the height h represented by the dotted line in the figure the pressure has changed from bx in January to by in Julyi and it is a simple matter to calculate the value bx — by . Conversely it is as easy to calculate the height at which b^^ —by has any given value. Under the conditions assumed Meinardus calculates that at 1,350 lupties above sea-level the pressure is 11 mm. less in July than in January. If therefore the air below this height over the Antarctic did not exist the total pressure over the Antarctic would be that required to make the total pressure over the whole globe the same in January as in July. The obvious way to account for the absence of this air is to assume that the average height of the laud within the Antarctic Circle is 1,350 metres. From this he concludes that there does actually exist within the Antarctic Circle a continent so high that its mass spread uniformly over the Polar cap T^-ithin the Antarctic Circle would '^ive an average level of 1,350±150 metres. Assuming that one-third of the area in question is at sea-level, the average height of the continent over the remaining two- thirds works out as stated above to be 2,000±200 metres. We do know that there is high land within the Antarctic, the Pole itself being on a vast plateau over 2,.500 metres above sea-level, and there is no reason to doubt that this hio-h land does act as suggested by Meinardus. There can therefore be little doubt that qualitatively his theory is correct and that it is a most important contribution to our know- ledo'e of meteorology and geodesy ; but whether reliance can be placed on the accuracy of his numerical results which he assumes by the ±150 metres attached to the estimated average height remains to be examined. The accuracy of the numerical calculation depends on two main considerations: (a) The physical principles on which the relation between meteorological data and height is calculated, and (b) the accuracy of the meteorological data on which the calculation is based. The 'physical jJrinciples.— The physical principles underlying Meinardus's calculation have already been stated, but it is advisable to restate them in the form in which they are actually applied. The area considered is that within the Antarctic Circle and is therefore concentric with the South Pole. Part of this area is at sea-level and part is occupied by high land, the relative sizes of the two parts are unknown. Meteorological data, all of which have been obtained at sea-level, are available from cei-tain stations in or near the area under consideration. By making the assumption that these observations are typical of the whole Antarctic at sea-level and that meteorological conditions depend on latitude only, the average pressure and temperature at sea- level over the whole area are determined.* The pressme at different heights in the free atmosphere over the area at sea-level is then calculated for January and July, and the height is determined at which the pressure difference between * In this section we are not considering tiie nature or accuracy of the data used by Meinardus, the argument proceeds independently of the actual values used. THE HEIGHT OF THE ANTARCTIC CONTINENT. 297 these two months has a definite value. Finally it is assumed that the pressure changes on the surface of a high land at this height will be the same as those fomid for the free air. It is this latter assumption which it is necessary to investigate. Why should the pressure over the high continent bear any relationship to the pressure in the free atmosphere over the low-lying area ? There is no air under the continent to expand and contract with changes of the season and so to regulate the pressure on the surface. Probably the idea which underlies the assumption is that the pressure at the ed"e of the high land must ba in equilibrium with that in the surrounding free air, and that this equilibrium decides the pressure over the whole of the adjacent continent. But we know quite well that the pressure conditions within any area cannot be deter- mined from a knowledge of the pressiure conditions around the boundary of that area. The interior pressure conditions can ba changed indefinitely without altering the boundary pressure conditions provided that the air is free to move. For example, the 760 mm. isobar of an anticyclone may occupy at one time exactly the same position as was occupied at another time by the 760 mm. isobar of a cyclone. In this case the boundary pressure conditions are identical but the interior pressure conditions are entirely different. Thus the pressure at the boundary of the high land might always be in equilibrium with the pressiure of the adjacent free atmosphere without giving any indication of the pressure over the interior of the high land. We must therefore conclude that the average pressure on the siurface of the high land may baar no relationship to the average pressure existing at corresponding heights in the free atmosphere. It may be contended, however, that the present problem is not concerned with the actual pressure, but with the change in pressure between January and July, and if the pressure in the free air is 11 mm. higher in January than in July it is probable that the mean pressure over the high land will be higher by the same amount. If this were the case it would mean that the pressure over every element of the land surface had changed by the same amount as the boundary, the consequence of which would be that the pressure gradients would remain the same over the high land in January as in July.* Now it is almost inconceivable that the pressure gradients, whatever they may be, are the same over the Antarctic high land in the winter as in the summer. Let the pressure distribution be cyclonic as supposed by Meinardus or anticyclonic as is more generally supposed, the system is bound to be more intense in the wmter than in the summer, which means that the gradients are steeper in winter than summer and the change in mean pressure greater or less, according to which system is chosen, than the change of pressure at the boundary. The physical principles underlying Meinardus's theory thus indicate that the changes in pressure over the high land from January to July may bear no close relationship to the changes in pressure at corresponding heights in the surrounding free atmosphere. Hence cal- culations based on the pressure changes in the free atmosphere may give an entirely wrong value for the average height of the Antarctic Continent. We will, however, for the sake of a thorough investigation of this important and interesting problem, now consider the meteorological data used by Meinardus to see the accuracy of his calculations irrespective of the above conclusions. The pressure data. — ^The actual sea-level pressure is of only secondary importance in the calculation, the result depending chiefly on the difference in pressure between January and * It 13 of course possible to conoeivo an entirely now pressure distribution which would give the correct change in the moan pressure, but this would bo a matter of pure chance and therefore cannot be taken into axount. 38 298 THE HEIGHT OF THE BARRIER AND THE SOUTH POLAR PLATEAU. July. To realise this it is only necessary to remember that there would be no problem if the pressure actually found by observation at sea-level were 11 mm. higher in January than in July, for then the pressure over the whole globe would be equal in these two months without the presence of the continent. The high continent is only necessary to account for the fact that sea-level observations do not give this diSerence in pressure between January and July. From his discussion of the pressure data Meinardus concluded that the average pressure at sea-level over the whole Antarctic was the same in January as in July. Since Meinardus \vrote the following additional data have become available : two years' observations in McMurdo Sound, and one year's observations at Framheim, Cape Adare, L'ile Petermann and in ths Weddel Sea. All these new observations show a lower pressure in July than in January. In the following table all the available data tcom Antarctic stations have been collected.* The average difference of pressure is given for each station as determined from the actuals for each month, and then again after the data have been smoothed in the usual way by applying the formula b (smoothed) = — j to the monthly values. With the large unperiodic changes of pressure from month to month which are such a featiu-e of the Antarctic conditions there can be Uttle doubt that the smoothed values are the better, for they eliminate to some extent large irregularities and therefore more nearly approach what would ultimately ba obtained from a longer series of observations. In obtaining the means of the groups of stations in columns 5, 6, 8 and 9 the differences have been weighted according to the number of months from which the individual means have been derived. Table 154. Pressure differences Jamuiry-Jithj in mm. 1 2 3 4 5 6 - 7 8 9 10 U Place. Latitude. Longitude. Actual. Smoothed. Number op MONTHS January -July. Mean. Mean. January -July. Mean. Mean. January. July. Snow Hill Port Charcot . L'ile Petermann Gauss Station . Belgioa Drift . Cape Adare McMurdo Sound Framheim Weddel Sea t . 64° S. 65" 65" 70° 71" 78° 79° 57 °W. 63 °W. 63 °W. 90 °E. 86 °W. 170 °E. 166 °E. 193 °E. + 4-91 - 6-50 + 2-39 - 018 - 0-80 -1- 2-54 -f 4-30 -f-12-50 + -t- 0-50 ■\ -\- 4-97 ■+ 300 -f- 2-63 - 2-50 -f 2-67 4- 2-37 - 3 30 -f- 5-27 + 515 -fl300 -f . -1- 0-58 + 0-31 .+ 3-87 1 1 1 t 1 1 2 4 1 2 1 1 1 1 2 4 1 * For actual values used and references see Volume' III. t Five days' observations for December, thirteen for January and twenty-eight for February only are available. In obtaining the smoothed values these have been weighted according to the number of days, double weight being ^,5 December -|-2xl3 January -\-2& February, given to January thus ^^ ^ % See figure 1 and table 1 of Veroffen des K. Preuss. Met. Institutes, No. 265, Bd. IV, No. U THE HEIGHT OF THE ANTARCTIC CONTINENT. 299 From this table it will be seen that out of the nine sets of observations seven show an excess of pressui'e in January when smoothed values are considered and six when the actual values are considered. The stations have been divided into two groups, one containing the stations in the Ross Sea area and the other the remainder. The latter are all situated on the northern edge of the continent and within a few miles of the open Southern Ocean. With the exception of Cape Adare, the stations in the Ross Sea area are much further south than the remaining stations. Now from these observations it is necessary to decide what is the difference of pressure between January and July over the whole region within the Antarctic Circle. Personally I feel that the data are far too few to give anything like the required information, but we can obtain some conclusion to compared with the one used by Meinardus in his calcula- tion. Baschin has found that over the Southern Ocean between 50° and 60° S. latitude the pressure difierenc3 January-July is — 0'73 mm., and this result may be accepted with consi- derable confidence. From fewer observations he determines the mean difference within 60° and 66^° S. (the Polar Circle) to be zero. The group of stations on the edge of the Antarctic Continent as shown in the above table have a mean difference of 4-0'58. This indicates that the difference is changing from negative values over the Southern Ocean to positive values over the Antarctic Continent. It is therefore significant that the three most southerly stations all show a high positive value for the difference. Unfortunately these three stations are all in the Ross Sea area and therefore it may be argued that the positive values are local and are due to the Ross Sea and do not really represeiit tne conditions in high latitudes. In this connexion the observations at Framheim are of great importance. Meteorologically Framheim is more under the influence of the Antarctic Continent than any other station, as is seen from its low temperature and absence of wind. The observations in July 1911 and January 1912 show a difference of pressure for January-July of +12-5 mm., which is by far the largest pressure difference between these two months recorded anywhere in the Antarctic. Taken alone this great difference — larger than the 11 mm. required by Meinardus — might be considered to be accidental and no doubt to some extent it is. We have remarked above * that the difference in pressure between two stations undergoes much smaller changes than the actual pressure itself, and further we concluded that the yearly march of the pres- sure difference between Cape Evans and Framheim shown in figure 57, page 172, was so regular that it probably represents very nearly the true normal difference of pressure between these two stations. By applying these differences to the normal pressure of McMurdo Sound based on four years' observations we found that the normal pressure at Framheim in January and July is 7450 and 735- 6 mm. respectively (see table 97, page 173). This gives a probable normal difference of pressure January- July at Framheim of 9-4 mm. as compared with 5-15 mm. in McMurdo Sound. If such large and similar variations exist between coast and con- tinental stations elsewhere, it shows the futility of attempting to determine even approximately the average pressure difference between January and July for the whole Antarctic from the few available observations, all of which have been made at coast stations. As the considerations we have just been discussing are difficult to apply I see no other way to treat the observations available than to take their arithmetic mean. It must not be * Page 172. 300 THE HEIGHT OF THE BARRIER AND THE (SOUTH POLAR PLATEAU. forgotten, however, that the mean obtained from coast stations in all probabihty gives a value too low by several millimeters. Table 154 shows that the mean of all the observations taken in the Antarctic is a difierence for January-July of 3- 00 mm. or 3- 87 mm. according as actual or smoothed values are considered. From this I think we are justified in saying that the average pressure difference at sea-level over the Antarctic from January-July is as likely to be +3 mm. as it is to ba 0, which is the value accepted by Meinardus. If wa apply the value +3 mn. instead of 0 mm. in Mainardus's formula, leaving all other assumptions unchanged, we find that the average height of the land within the Antarctic Circle works out to be 966 metres, which is very different from Meinardus's value of 1,350± 150 metres. The tempsmture data.— It is as nscsssary to know the average temperature in January and July over th3 whole Antarctic as it is to know the pressure, and again it is the differ- ence, much more than the actual values, which affects the result. After a careful discussion of all the data available in 1909, Meinardus concluded that the average temperature at sea-level over the whole Antarctic in January is — 2-7°C. and in July -25-2''C., giving a difference of 22-5°C. Again Meinardus neglected the fact that all the observations used by him were obtained at coast stations. Who would think of determining the average range of temperature over any other continsnt from a fiw scattered observations at coast observatories ? Here our discussion of the Barrier temparatures comes to our help. In table 4-5, page 83, it is shown that the difference in temparature between January and July is 21°C. at Cape Evans and 29 "C. on the Barrier. This indicates that the temperature difference January- July may be 8 C. more at a land station than at a coast station in the same latitude only a few miles distant. It therefore does not seem unreasonable to conclude that, considering how much of the Antarctic is far removed from the ssa, Meinardus's value of 22-5"'C. for the difierence in temperature between January and July obtained entirely from coast stations is much too small. If Meinardus's value for the July temperature is lowered by only 5°C., leaving all other of his assumptions unchanged, the average height of the surface within the Antarctic Circle is reduced to 1,050 metres. The vsrtical temperature gradient.— Meina,rdus assumes that the vertical temperature gradient is -•5°C. per 100 metres both in January and July. Now the balloon observations made in McMurdo Sound and by Barkow in the Weddel Sea have proved that during the winter there are large temperature inversions in the lower atmosphere. There are not sufficient balloon observations to give an average value of the temperature gradient during July, but there cannot be the slightest doubt that it is much less in July than in January. In fact it is not improbable that over inland areas where the temperature is very low and where the wind is much less than near the coast there is a nearly permanent temperature inversion up to heights of at least 2,000 metres. It therefore appears to me that the average vertical temperature gradient over the whole Antarctic in July is as likely to be 0 as ■5°G. per 100 metres. Putting this value in Meinardus's formula and leaving all other values unchanged the average height of the surface within the Antarctic Cii'cle is raised to 1,610 metres. It is of interest to see what is the effect on the calculation of height if all three of the above new values are introduced into the formula. In the following table the new and THE HEIGHT OF THE ANTARCTIC CONTINENT. 301 old values ane shown, and in the last line the average heights of surface within the Antarctic Circle calculated from them by Meinardus's formula have been entered : — Table 155. At sea -level. Meinardus's values. Possible new values. January pressure .... 745 mm. Same. July pressure 745 mm. 742 January temperature .... - 3°C. Same. July temperature .... --26°C. -31 °C. January vertical temperature gradient . •5°C. per 100 m. Same. July vortical temperature gradient •5°C. per 100 m. 0 Average height 1,350 metres. 852 metres. 'flna'a vpliip nf 1 !^.^n inpfrpsl i« rpf liif^pfl fn R!\9. inpfr PK wn mil J change. This review of the meteorological data on which Meinardus has based his calculation of the average height of the Antarctic Continent has shown that by making assumptions which have as much probability as those of Meinardus, the resulting height can be very much changed. My own opinion is that considering the vast extent of the continent and the few observations, none of which have been taken under true continental conditions, but only on the sea • coast, it is quite hopeless to get even an approximate solution of the problem. For this reason I have been careful in the above discussion not to claim that my values are nearer the truth than those of Meinardus, I have simply claimed that they are at least as probable as his, and have contented myself by pointing out that the alternative data alter Meinardus's result so much that the accuracy he has claimed is far too great. Conclusions. (a) If it is true that the average pressure over the Antarctic must be 11 mm. higher in January than in .July in order to keep the mass of air over the whole globe the same in these two months, then it is probable that qualitatively Meinardus has given the correct explanation. (b) The method used, however, to calculate the height of the land is open to serious objection, being based as it is on the assumption that the pressure over the surface of a great continent can be determined by the pressure at the boundary. (c) Further the meteorological data used are far from being certain. By using other data which at least are as probable as those of Meinardus, the mean height is reduced from 1,350 metres to 852 metres. But in view of conclusion (6), both these estimates may be far from the truth. CHAPTER X. ATMOSPHERIC ELECTRICITY. Owing to the large amount of work entailed by the ordinary meteorological and magnetic investigations undertaken by the expedition, it was found impossible to do more in atmospheric electricity than to measure the potential gradient and the radio-active contents of the atmos- phere. Both of these unfortunately came to an end after the first year's work, as the few men who remained in the south during the second year had their hands full with other duties. Potential gradient. Only thos3 who have attempted maasurements of atmospheric electricity under Polar condi- tions can have any idea of the great difficulty of the work and the constant attention which the apparatus requires. Th> difficulties met with in the ordinary meteorological observations have not been described in the above discussion as they can be overcome with a little fore- thought, and each observer must work out the bast means to meet his own difficulties, but with atmospheric electricity so few observers have had the necessary experience that a few words here of the difficulties met with and the way they were overcome may not be out of place. The method of recording the potential gradient is well known. We have first the collector which is exposed in a suitable position in the open air. The function of the collector is to bring itself and the apparatus attached to it to the potential of the air in its immediate neighbourhood. Until the discovery of the radio-active substances only two practical collectors were known, (a) the Kelvin ' water dropper ' and (b) the flame. The first of these is quite unusable in Polar regions owing to freezing, and the difficulties connected with the second make its continuous use with self-recording instruments practically impossible. The radio- active collector does not freeze and does not blow out, but in its usual form it can only dispose of such a small amount of electricity that the question of insulation becomes very important. Unless the insulation is practically perfect, the collector cannot dispose of the electricity which leaks to the insulated system rapidly enough, and the apparatus charges up to a potential more or less below the potential of the air near to the collector and the apparatus then records a mixture of changes of potential and changes of insulation. The collector used was a small copper rod about 3 mm. in diameter and 5 cm. long. This had been coated with polonium which has the great advantage of emitting only a rays, so that only the air in the immediate neighbourhood of the collector is ionised. The collector was mounted on a thin wooden rod which in turn was fixed to an iron pole, so that the collector itself was about ten feet above the ground. The wooden rod was insulated from the iron pole by means of a plug of ebonite covered in suliihur. An inverted tin can was placed over the insulator in the method shown in figure 89. This was found neces- sary in order to protect the insulator from hoar frost which in certain conditions of the atmosphere was found to be deposited on every thing exposed to the sky. INSTRUMENTS AND iMETHODS. 303 During blizzards fine driven snow entered this can and settled over the insulator and destroyed the insulation. There is no means of getting over this diffi- culty, for the drift during blizzards enters every small crevice and no matter how the protecting device is arranged it soon becomes full of snow which connects the insulated parts to earth. This, however, was not a serious difficulty for one is chiefly concerned with the potential gradient only during fine weather, so the loss of part of the record during blizzards did not affect the main results of the work. Also during blizzards the potential was far above the limits of the recording electrometer and so could not have been measured if the insulator had remained perfect. The chief practical effect was the necessity to clean the insulator at the end of each blizzard and to put it in good order for the subsequent fine weather. The Benndoif self-recording electrometer had of necessity to be within the living hut, the temperature of which was kept, except occasionally, well above the freezing point. A very liberal share of the hut was apportioned to the physical laboratory but unfortunately it was in the coldest corner of the hut. In the h\it twenty-five men lived in one large room in which all the cooking was done. In consequence the hiimidity was always high and water vapour literally distilled into the physical laboratory. This vapour condensed on to the walls which were generally covered with a coating of ice or water. It was through this wall that the wire to the potential gradient instrument had to pass, and the method of doing this was the first difficulty which had to be solved. In ordinary observatories, the connecting wire is usually passed through a hole in the wall sufficiently large for the wire not to touch it when slightly displaced by the wind. Such a hole was quite impos- sible in our hut, for it would have let in cold air, but much more serious it would have been the entrance for drift snow during the blizzards. The wall of the hut consisted of two thicknesses of wood, then an air space of three inches, and then two more thicfaiesses of wood. The first attempt made was to pass a metal tube through the wall, plug it up at each end with sulphur and pass a thin wire through the centre. The sulphm' then not only closed the tube but acted as supporting insulators for the wire. This method soon proved unusable, for the water vapour which was deposited on the well-insulated wooden wall was deposited still more rapidly both inside and outside the metal tube. The water was in itself sufficient to destroy all insulation, but on account of the metal tube projecting into the cold air outside the hut, the end within the hut was so cold that the water froze and a large mass of ice crystals formed upon it which were pretty to look at but totally destructive to the insulating properties of the sulphur. Some method had to be found which would not only protect the insulator from the water deposited all over the wall but would also not conduct cold from tho outside, and so cause an intensified deposition on the insulator. There is no need to go into all the experiments made ; but the final solution of the problem will be given. As is usually the case, the solution is so simple that one wonders why it was not the first thought of. Figure 90 shows the completed insulator. A and B are the two separate parts of the wall through which the insulator was to convey the conductor. C C is a brass rod about one-eighth inch thick. Around this rod a coating of sulphur D was cast Fig. 89. Insulator for collector rod. 304 ATMOSPHERIC ELECTRICITY. to a diameter of about one inch, and so long that it projected two inches within the hut and one outside. E was a round tin box which had contained a Kodak fihn, the lid was removed and a small brass tube, F, soldered to the bottom so that it could be pushed along the rod C and completely surround the end of the sulphur. G was an Outside i Inside • n;'-^>:-C', r-^.'-r". v-^-'^; "--■•■' ■-•'•'"' ■■^:^'.<■^J'^;^^■:y:':^^/x^^f H J Fig. 90. Inanlator through wall. ordinary tin canister. The lid had a hole made in it of the size of the sulphur rod and it was tacked on to the wall at H. In the bottom of the canister a small hole, just larger than the rod C, was pierced ; then when the canister was pushed home into its lid H it completely surrounded the whole end of the insulator and the rod C passed freely into the hut. It will bs noticed that the tin canister was not connected with the outside and therefore did not b3Com9 excessively cold, on the contrary it exposed so much of its surface to the air of the hut that it quickly took the tempsrature of the air, and so water vapour was not deposited on it excessively. The metal rod which connected with the outside air had the inner cylinder attached to it and this tended to keep its end from remaining very cold. Any moist air which entered the outer tin canister was deposited on the outside of the inner tin cylinder and very little found its way within the latter to the insulator. A few pieces of calcium chloride placed in the canister at J absorbed moisture and kept the air within fairly dry. This insulator worked splendidly and gave practically no difficulty for several months, but then the sulphur cracked and the insulation failed. When this happened a new insulator on the same principle was constructed except that the rod of sulphur was replaced by a rod of ebonite, and thereafter no further difficulty was experienced, so long as the sur- face of the ebonite was periodically cleaned with a piece of emery paper. The end of the insulator outside the hut did not need the elaborate protection which was necessary within the hut, so a simple tin can was placed around it to keep oil the deposit of hoar frost. A wire was then taken in a single stretch from the end of the insulator outside the hut to the collector ten yards away. During the summer months this wire was a constant source of trouble, for it was continually being broken down by skua-gulls flying into it. These gulls hving in a country without trees, wires or other obstructions in the free air never looked where they were going when flying about the hut and not only flew into the wire but on one or two occasions actually collided with the collector rod itself. It was dis- tinctly annoying to lose many hours of record during fine weather owing to the clumsy habits of these birds. INSTRUMENTS AND METHODS. 305 The self-recording Benndorf electrometer inside the but was also not free from its difS- culties. It has already been explained how the physical laboratory was the coldest comer of the hut. The instruments in it were therefore always colder than the general temperature of the room, especially in the morning when the breakfast was cooked while the instruments were particularly cold after the night. Under such circumstances it was practically impossible to maintain high insulation when the air had free access to the insulators. The insulators on the Benndorf self-recording electrometer are peculiarly badly situated from this point of view, for they are under the instrument and exposed to the air. The insulators holding the quadrants are not important, for the battery which charges the quadrants is quite capable of supplying the electricity for a relatively large leak ; but not so the insulator through which the wire from the collector passes, for unless this insulator is practically perfect the potential to be measured may be greatly reduced. On the Benndorf electrometer there are three insula- tors connected with the system which is charged by the collector : (a) an insulator between the needle and the suspension, (h) an insulator supporting the metal plate on which the acid pnt rests, and (c) the above mentioned insulator through which the connecting wire passes, (a) and (b) are completely within the case while (c) is half within and half without. It was the latter which gave all the trouble and therefore it was done away with entirely. A stout wire was soldered on to the plate carrying the acid pot and a narrow sUt cut in one of the panels of the body of the electrometer through which it freely passed. All the essential insulators were now within the body of the electrometer which was supposed to be kept diy by the acid. Still the insulation was far from perfect. Every time the electrometer was opened to adjust or examine the instrument, warm damp air entered the cold interior and the insulation was immediately destroyed. Also the body of the electro- meter was not airtight and even when closed damp air entered. It was at once realised that the only hope was to keep the interior of the electrometer at a higher tempera- ture than the air of the hut. This was done by inserting a small heating coil which was supplied with a current of 'll ampere from the accumulators. After this had been done there was comparatively little difficulty with the insulation of the electrometer and the case could be opened at any time for examination without the insulation materially suffering. There is one more cause of difficulty and worry to be mentioned. The polonium collector could not be protected from the formation of hoar frost over its whole surface. This coating of ice reduced its activity greatly and caused an appreciable reduction of the potential regis- tered. This difficulty could not be entirely removed but it was reduced as follows: We happened to have with us two identical polonium collectors, these were so arranged that they could be rapidly fLxed to and removed from the top of the collector rod. Whenever meteorological observations were made, every four hours, the polonium rod in use was examined and if it had a coating of frost it was removed and replaced by the other one. The frost-coated rod was then suitably suspended in a canister containing calcium chloride, which dried it and kept it always ready for immediate use. When all these changes had been made as the result of bitter experience it was only by constant attention to the electrometer, insulators and collector that satisfactory records could be obtained. Two or three times each week the sensativeness of the electrometer and the state of the insulation were tested. To do this the polonium collector was removed, then the electrometer to which the whole insulated system was attached, was connected to a Wulf electroscope and charged. 39 306 ATMOSPHEEIC ELECTEICITY. The electroscope reading was noted when three successive points were registered by the electrometer The electrometer registered every two minutes and it was easy to see whether there was an appreciable leak from the change in the electroscope reading during the interval. The whole system was then connected to earth for ten minutes or so to mark the zero line. Thus on the trace itself were frequently recorded, the state of the insulation, the position of the zero, and the deflection con-esponding to a known potential. It was necessary to find the reduction factor for converting the voltages recorded by the electrometer to the potential gradient in volts per metre over a level surface. The frozen sea within a few yards of the collector made this an easy matter. The necessary observations were made on five days during November, December and January, ] 911-12. The reduction factor proved to be linear over the whole scale which was a great convenience in working up the records.* Potential Gradient and Drift.— When the air was full of drift snow the potential giadient was as a rule very high. Also a very little drift would increase the potential above the extreme range of the electrometer, i.e., 367 volts per metre. This is most clearly seen when the average potential during different wind strengths is examined. When the observa- tions for the whole year are arranged according to wind strength we get the following result : — Table 156. Potential Gradient and Wind. Wind, miles per hour. 0—10 11—20 21—30 31—40 >40 Potential gradient, volts per metro. 104 >159 >212 >260 >304 Frequency > maximum •7% 10% 24% 45% 65% Number of observations 3,240 948 848 626 302 It will be seen that with winds of 0 to 10 miles an hour the potential gradient was 104 volts per metre, while it steadily increased to > 304 volts per metre with winds over forty miles an hour. The range of the electrometer would only record 367 volts per metre, hence the value of the higher potentials coidd not be recorded. ^Vhen the needle reached the extreme for only a shoi-t time in an hour, one could form some idea from the shape of the existing curve what would have been the shape of the missing portion, and in this way a rough estimate of the mean potential in the hour could be determined. When the needle was against the extreme stop for an appreciable time, it was only possible to say that the mean potential during the hour was > 367 volts per metre. In taldng the mean value of the potential the sign > had to be neglected, hence all the values of the potential gradient for wind strengths over 11 miles an hour are smaller than they would have been if the true high potentials had been measured. Thus the sign > has been prefixed to the values in table 156. It is clear that the highest mean potential which could be recorded in this way is 367 volts per metre, and it will be noticed that this value is nearly reached * A description of the method of obtaining the reduction factor and of many other details of tlie apparatus used in the atmospheric electricity work will be found in my paper entitled ' Instrumente zur Beobachtung der atmospharischen Elektrizitat ' published in Physikalische Zeitschrift, Volume 14, page 41, 1913. NEGATIVE POTENTIAL GRADIENT. 307 with winds above forty miles an hour. The numbers in the first line of table 1.56 are there- fore only qualitative, but they clearly show that the higher the wind, i.e., the more the drift the higher the potential. The same result is shown in the second line. The numbers in this line were obtained as follows : Th? number of timss that the mean hourly potential gradient exceeded th? hmits of the electrom3ter was counted for each group of winds. This number was then divided by the total number of observations in that group and multiplied by one hundred, the result b^ing the percentage frequency with which the mean hourly potential gradient exceeded the maximum range of tho electrometer during winds of the specified velocity. We see that with winds of 0 to 10 miles an hour the maidmum was exceeded in only •7 per cent, of the cases, while with winds greater than 40 miles an hour the maximum was exceeded in 65 per cent, of the cases. In other words excessively high potentials were seldom recorded during light winds, when there was little, if any, drift, while with high winds which were practically always accompanied by heavy drift the potential was as a rule excessively high. Drift occurred in two forms : there was first the drift which appeared to fill the atmos- phere darkening the sky and making it almost impossible to see more than a few yards away ; and there was the low surface drift which only extended a short distance from the ground and above which the air was clear and the sky frequently cloudless. Both kinds of drift usually affected the potential in the same way, for as soon as the slightest drift com- menced the potential became verj'' high and remained so until the drift ceased. Negative potential gradient. — ^During the whole year the potential gi-adient was negative only on nine days ; as these are important in the following theoretical discussion each will be described in detail : — • (1) April 20, 1911. — Throughout this day there was a high S.E. wind of between 30 and 40 miles an hour. The sky was thickly clouded, and there was either no drift or a very little low surface drift. From midnight to 9 a.m. the potential was positive and high, then it decreased rapidly, and just after 10 o'clock it became negative. For several hoiu's afterwards the needle oscillated fi-om one to the other side of the zero. The potential was never negative for long and high negative values were not recorded. At this time the Sound was not fi'ozen over and there was some raised fog over the open water. The temperature was about -10°F. (2) June 1, 1911. — During the early morning there was some new snow and excessive drift ; during this period the potential was positive and high. By midday, although the wind was still between 40 and 50 miles an hour-, the drift had entirely ceased. From 13 hours to 16 hours the potential was frequently negative, but only for short intervals, the mean potential being just on the positive side. Later in the day the potential became high on the positive side, but without drift. ■ The sky was overcast throughout the day. (3) Septetnber i, 1911.— There was little or no wind on this day. Until midday the sky was practically cloudless and the potential was normal. At about 1 p.m. heavy cbuds came up from the north and by 4 p.m. a thick mist lay over the station. With the approach of the clouds the potential decreased and between 2 P.M. and 3 p.m. it crossed the zero and was negative for about fifteen minutes. After this the potential rose again and was fairly normal for the rest of the day. There can be Uttle doubt that the cloud was the cause of the reduction and 3U8 ATMOSPHERIC ELECTRICITY. reversal of the potential gi'adient. In this case the changes in the potential were comparatively regular, there being few of the rapid changes which generally accompanied the reversed potential, (i) October 2, 1911. — There was a heavy blizzard on this day with much new snow and drift. It was impossible to prevent the insulators becoming covered with snow, therefore the record is doubtful ; but for several hours between 4 A.M. and 9 A.M. the needle was frequently on the negative side of the zero, therefore it is almost certain that duiing this period the potential was negative. (5) November 6, 1911. — This was another day with a high southerly wind between 30 and 40 miles an hour. Until 9 a.m. there was some surface drift and the potential as usual was positive and high. The drift then ceased, but the wind continued, and the potential steadily decreased to zero at about 3 p. m. and then became negative for about 40 minutes, after which it became positive again. The clouds were heavy and low throughout the day. (G) December 8, 1911. — This is the most instructive case. Two days previously there had been the greatest snowfall of the year, which had been in the unusual form of large flakes. The whole of the frozen Sound was covered with about 18 inches of light loose siaow. At 1 p.m. a high wind from the north .sprang up driving the loose snow before it. The potential at first was high and positive, then at 4 p.m. it commenced to fall rapidly, at 6 p.m. it had become negative and the needle of the electrometer had swung to its extreme negative position where it remained with a few short breaks until 11 p.m. There is no doubt that during this period high negative potential occurred and it was the longest period of continuous reversed gradient recorded. Throughout the whole period of negative gradient the sky was clear, but there was a very heavy surface drift of light snow caused by the high wind. The drift, however, was not higher than about 3 or 4 feet, above which the air was quite clear. The whole of the drifting snow was therefore well below the collector. (7) December 9, 1911. — The high wind of the previous day continued, but when it had removed the upper layer of soft loose snow the drift decreased and none was reported after 8 a.m. On the whole the potential was high and positive, but it was most irregular, the needle of the electrometer constantly swinging to, and occasionally across, the zero. Between 10 a.m. and 1 p.m. the conditions were reversed, for then the potential was negative and high with occasional swings to the positive side. Afterwards the potential returned to the positive side and remained high until the wind dropped, when it became normal. In this case we have an unsteady potential and a long period of negative potential with little or no drift, but again a high wind. (8) December 30, 1911. — Throughout this day there was a moderate wind 20 — 30 miles an hour from the E.S.E. with snow and drift. The potential was very low all day and was sometimes positive and sometimes negative. The record is almost exactly like one obtaiired on a rainy day in temperate regions. The temperature was between 20° and 25" F. (9) Febriuiry 10, 1912.— Between 3 pm. and 4 p.m. a little snow fell and the potential gradient became negative for about half an hour. There was a light wind POTENTIAL GRADIENT AND DRIFT. 309 19 miles an hour, from the N.N.W. At this time the Sound was nearly free from ice. Of these nine records of negative potential gradient four (1), (2), (5) and (7) occurred with a high wind but no drift ; three (4), (8) and (9) occurred during snowfall ; one (6) occurred with a clear sky but very heavy low surface drift ; and one (3) occuried during the passage of heavy low cloud but without wind, drift or snow. The explanation of the high potentials during drift proves to be very difficult. I did not realise this at first and considered that the whole phenomenon was due to the ice particles becoming charged positively by collision amongst themselves and with the snow- covered ground. Rudge has experimented on the well-known electrical effects accompanying the raising of clouds of dust, and shown that when dust is blown into the air or even let fall through the air it becomes highly charged. Treating air full of snow-drift as a dust cloud of great extent, it is reasonable to suppose that the snow would, like the dust, become highly charged. Thus, when the whole lower atmosphere is full of positively charged snow, there would be a high positive potential gradient. It follows as a consequence that if the positively charged snow was all or mainly below the level of the collector of the potential gradient apparatus the registered potential gradient would be reversed. This is easily seen, for the normal positive potential gradient is caused by the negative charge on the earth, and if this charge is replaced temporarily by a la3'er of drifting snow highly charged with positive electricity, the gradient must be reversed. By far the heaviest surface drift I observed in the Antarctic occurred on December 8, 1911, when a high northerly wind carried along the recently fallen loose snow. This drift was not only very heavy, but it was also very low) and all well under the collector ; it seemed, therefore, very strong support to the theory of positively charged snow to find that during this period the potential gradient was negative and high. The first difiiculty encountered was, when preparing this chapter, I searched for other cases of surface drift accompanied by negative potential gradient. I then found that instead of being accompanied by negative potential gradient surface di'ift was nearly always accom- panied by very high positive potential gradient. At first it occurred to me that this might be due to the clouded sky which usually accompanied the drift, surface or otherwise, while the surface drift of December 8th occurred under a perfectly clear sky. I therefore searched the records for cases of surface drift with a clear sky and found seven cases beyond that of December 8th. In all these new cases the potential during the surface drift was positive and very high. This disposes at once of the explanation that the drift snow has a liigh positive charge, for if it had, so long as it was mainly below the collector, the potential should be reversed and only when it extended more above than below the collector should the positive potential gradient be increased. Stated in other words it is quite impossible to explain the observed facts by considering only the charge on the driven snow, for charged snow would cause the sign of the potential gradient to be reversed when the distribution of the snow changed from being mainly below the collector to being mainly above ; while the observations show that the potential gradient is nearly always positive and high both when there is only a little surface drift and when the drift is so great and extends so high in the atmosphere that there must be more drift above than below the collector. One is forced, therefore, to seek the electrification elsewhere than on the driven snow. It is incon- ceivable that the surface of the earth can become electrified, for this would necessitate the surface being highly insulating, otherwise the charge would be neutralised at once. There is no doubt that a snow surface is a bad conductor of electricity, for we used for our telephone from Cape Evans to Hut Point, a distance of 15 miles, a bare aluminium wire laid over 3i0 ATMOSPHERIC ELECTKICITY. the snow surface while the return was the sea. During the winter the ' speaking ' over the wire was perfect, showing that the insulation was sufficient, but after the naiddle of November no electrical signals could be sent through the wire showing that then the snow had lost its insulating power. If the snow could not insulate for the low voltages and high current density used in the telephone circuit during the summer, it certainly could not insulate for the high potentials and small charge produced by friction. As the positive potential gradient was as high during blizzards in the summer as in the winter, it is obvious that the charge concerned is not on the ground. The only remaining place to look for the charge is the air itself, and I beheve that the charge carried by the air is the solution of our problem. There can be no doubt from Rudge's work quoted above and from experiments made by the writer that the cause of the electrification of dust when blowoi into the air is not friction electricity as commonly understood, but an effect similar to that found when water is splashed or water drops broken, in which the solid or liquid particles retain one kind of electricity while the opposite electricity is given to the air, probably in the form of slowly moving ions. Let us imagine an isolated cloud of ice particles which on account of the turbulent motion of the air are constantly colliding with one another. The result would be that the ice particles would be charged with one kind of electricity (the sign of which will be discussed later) and the air would receive the opposite charge. There would, however, be no exterior field produced, as the two charges would neutralise one another at an appre- ciable distance from the cloud. If, however, the snow particles slowly settled in the air, the charge on the snow would become concentrated in the lower half of the cloud and that on the air in the upper haU:. This would produce an electrical field in the cloud itself. Now imagine that the cloud comes into contact with, the ground, then every time an ice crystal touches the ground its charge is lost, but the opposite charge remains in the air above. The cloud in this way would soon have an excess of the electricity associated with the air and in course of time might become very highly charged. It is now only necessary to assume that the snow in the process becomes negatively charged and the air positively charged to have a complete explanation of the normal relationship foimd between the drift and the potential gradient. Let us consider first the case of sui-face drift. The term surface drift is never used when the driven snow rises more than a few feet above the ground. The ice particles in this drift are constantly colliding with one another and ^vith the fixed snow on the ground. According to the theory every colUsion is accompanied by a separation of electricity, the ice particles becoming negatively charged and the aii- receiving a positive charge. The snow, however, is constantly coming into contact with the gromid when it gives up any charge it has, it will probably become charged again at the instant of separation, but every contact ^\'ith the ground while adding more positive electricity to the air adds no further charge to the snow. Thus the air in which the drift is carried along becomes more and more highly charged with positive electricity, without the snow retaining the corresponding quantity of negative electricity. Owing to the irregularities of the air motion the charged air near the groimd mixes with the air above and in a short time the whole of the air above the ground pro- bably to several hundred feet becomes more or less highly charged with positive electricity. Between this positively charged air and the ground an intense electrical field may be set up. The direction of the field is the same as the normal field of the atmosphere and therefore the efiect of the surface drift is to produce a high positive potential gradient. Let us now consider that the whole of the lower atmosphere becomes full of cbifting snow to a great height. The collisions" of the snow particles produce the same separation of electricity, but owing to the constant downward motion of the snow relatively to the air POTENTIAL GRADIENT AND DRIFT. 311 containing it, there tends always to be an accumulation of positive electricity in the air of the upper part of the cloud. The field set up in this way is again parallel to the earth's normal field and therefore increases it, thus there is no reversal of the field when the drift snow extends from the ground to heights well above the collector. In other words the potential gradient -will be positive and high whenever there is drift quite independently of whether the drift is suiface drift or true drift throughout the lower atmosphere. We have now explained the vast majority of the observations and it is only necessary to prove the rule by explaining also the exceptions. The most important case is the high negative potential observed during heavy surface drift on December 8th. It has been found that when dust is raised in the air the resulting charge is affected by the character of the dust. Fine dust has been found to be charged with the opposite kind of electricity from that associated with coarser dust of the same material. Thus the loose snow in the drift of December 8th might have the opposite charge from the hard grains of ice of which surface drift in th^. vast majority of cases consisted. This assumption seems, however, imnecessary. On December 8th the surface drift was unusually heavy on account of the large amount of loose snow lying on the ground after the recent heavy fall of large flake snow. In such a drift the separation of electricity would be excessive, therefore the air immediately above the drift would be very highly charged with positive electricity. If at the same time there was little tendency to the formation of ascending air currents, the concentration of positive electricity would remain in the air near the upper hmit of the drift. It is not difficult to conceive that under such conditions there would be more positive electricity below the collec- tor, which was 10 feet above the ground, than above it. This would fully account for the reversed potential gradient. It also explains why later on the potential became positive, but very unsteady, with occasional returns to the inverted direction ; for then the positively charged air extended higher into the atmosphere and the collector had generally more positive electricity above it than below it, but occasionally the original conditions were re-established for short p?riods. That the potential should be occasionally reversed during periods of snowfall, as in cases (4), (8) and (9) above, is not difficult to explain. Snowfall is a sign of ascending currents, for they are necessary to supply the water vapour necessary to the production of snow The air of these ascending currents must constantly pass out of the top of the cloud and take with it positive electricity. If this air then enters an air cirrrent from a different direction from the lower wind, it will be removed from the region of the cloud. The whole cloud will then have a residual charge of negative electricity associated with the snow. This being above the collector will reverse the field and a negative potential gradient will be recorded. The five cases of negative potential gradient without appreciable drift cannot be ex-plained by the above consideration which are based on the separation of electricity associated with drift, and as no explanation of the reversed gradient is obvious it will serve no useful purpose to make assumptions, which cannot be tested, to provide an explanation. The above discussion leads to the conclusions — (a) The electricity which affects the recorded potential gradient dming drift is not associated with the driven snow ; (6) but with, the air in and above the drift ; (c) the separation of electricity takes place when ice crystals collide, the ice becoming negatively charged and the air positively charged. These conclusions are based entirely on the observed potential gradient, it is ob\-ious that the only satisfactory test would be to examine the sign of the charge on the drift 312 ATMOSPHERIC ELECTRICITY. snow itself. If this were found to be negative, the conclusions would be proved beyond doubt. The need to make such an investigation was realised while we were in the Antarctic, and I had devised the method to be used, but my unexpected recall at the end of the first year prevented the experiments being actually made. It is to be hoped that the simple experi- ments will be made by the first observer who has the opportunity. Yearly variation of the potential gradient. It is customary always to exclude periods of disturbed weather when obtaining the normal potential gradient. This practice has been followed in the following discussion, with the further limitation that all periods when the wind velocity was greater than 5 miles an hour have also been excluded. This was necessary in order to rid the observations of the effect of drift. The procedure adopted was to tabulate the values of the mean hourly poten- tial gradient during periods in which the wind velocity was 0 to 5 miles an hour. To rid the observations still further of the effect of recent or approaching high wind and drift, the first and last hours of each period were also rejected. Thus every value used in determining the normal potential gradient was not only obtained during periods in which the wind was 5 miles an hour or less, but also at least one complete hour separated it from past or approaching bad weather. Needless to say all hours during which snow fell during calm weather were also excluded. Table 157 contains the mean potential gradient for each month * and the number of hours used in the determination. Table 157. Mean monthly potential gradient at Cape Evans. Volts per metre. Potential gradient Number of hours used . Jan. Feb. Mar. April. May. June. July. Aug. Sept. Oct. Nov. Dec. Year. 96 89 154 ! 94 1 75 21 79 43 67 80 221 112 85 179 78 233 105 213 90 105 88 173 107 148 87 These values are plotted on figure 91, from which it will be seen that the yearly variation is fairly regular with a minimum in May and a maximum in December. In other words the potential in the summer is distinctly higher than in the winter. A similar variation of the potential gradient has been found wherever observations have been made in the Antarctic. These are the Belgica Expedition, 71° S., 87° W. ; Hut Point, 78° S., 38 E. ; Port Charcot, 65° S., 64° W. ; and Petermann Island, 65° S., 64° W. It can therefore be accepted without question that the potential gradient in the Antarctic is higher in the summer than the winter. This is a most important result, for it is the exact opposite of the conditions in the northern hemisphere, where numberless observations have shown that the potential is higher in the winter than the sununer. The question at once arises are the observations made in the Antarctic typical of the conditions over the whole of the southern * I very much regret that the values of yearly and daily potential gradient given in my paper 'Chief Results of the Meteorological Observations made on Captain Scott's Antarctic Expedition ' published in the Quarterly Journal of the Royal Meteorological Society , Volume XL, page 221, 1914, are not quite accurate. This was due to the care, lessness of one of my clerks who made many mistakes in computing the tables of potential gradient. The er - are not important and cannot affect any conclusion whicli may have been d^a^^^^ from the published values. errors YEARLY VARIATION OF POTENTIAL GRADIENT. 313 hemisphere ? If so, then we have that the potential gradient is at its maximum and minimum at the same times over the whole world, which would be a most important conclusion. It would mean that the total charge on the earth's surface is highest when the earth is near the sun and lowest when furthest away, thus connecting the electrical state of the earth and its atmosphere with cosmical rather than terrestrial factors. This important and interest- ing conclusion cannot, however, in the present state of our knowledge be accepted. It is a remark- able fact that we know more about the electrical state of the atmosphere in the Antarctic than in the rest of the southern hemisphere put together. So far as I can find there are only five series of observations in the southern hemisphere outside the Antarctic ; these are the observations made at (a) Melbourne in Australia by Neumayer during 1858-63, (b) Cape Jan. Feb. Mar. April May June July Aug. Sep. Oct. Nov. Dec.| 110 100 i 90 V S.80 tA ♦- o >70 60 / V / 1 k / \ / > \ ^ \ / f N r^ k / > f s / Flu. 91. Yearly variation of potential gradient. Horn by the French Mission in 1882-83, (c) Batavia in 1890-95, (d) Samoa in 1906-08, and (e) Buenos Aires in 1911-12. Assuming that all these determinations are satisfactory we find that two, at Caps Horn and Batavia, show the highest potential in the southern summer and so are in agi-eement with the Antarctic result while the remaining three, at Melbourne, Buenos Aires and Samoa, show the reverse, the highest potential being in the winter and the lowest in the summer. With such conflicting evidence, some of which does not come up to the modem standard of reliability, we cannot do more than wait for more and better ob- servations ; but in the meantime with the observations at Melbourne, Buenos Aires and Samoa before us we cannot accept the conclusion that the potential gradient over the whole world is at its maximum and minimum at the same times. The harmonic analysis of the yearly variation of the potential gradient at Cape Evans gives P=86-6+12-5 .sin {132° o^'+x)+3-7 sin (48° 2.5' -f 2a:), in which x=:15 for January, 45° for February, and so on. Daily variation of potential gniient. The mean hourly potential gradient during fine weather as defined in the last section was investigated to find the daily variation for each of the four seasons and for the year. In the following table the results are entered and against each of the mean values the number of hours used to determme it is given, 40 314 ATMOSPHERIC ELECTEICITY. Table 158. Daily variation of potential gradient. Fine weather and wind 0 — 5 miles an hour. Aug .-Oct. Nov -Jan. Feb.- -April. May -July. Y EAR. Local time. P.G. No. of hours. P.G. No. of hours. P.G. No. of hours. P.G. No. of hours. P.G No. of hours. 0—1 94 21 116 20 105 7 75 25 95 73 1—2 87 21 115 21 129 7 77 28 95 77 2—3 89 24 106 22 102 7 81 29 92 82 3—4 93 23 110 24 105 7 86 30 97 84 4—5 93 24 116 21 95 8 84 27 96 80 5—6 90 27 110 21 107 9 86 24 96 81 6—7 99 29 109 25 96 11 92 20 100 85 7—8 121 28 101 24 92 10 88 20 104 «2 8-9 109 25 93 24 87 9 91 20 97 78 9—10 96 22 81 20 70 10 84 18 85 70 10—11 98 20 76 18 63 9 77 14 81 01 11—12 84 22 81 19 68 9 78 15 78 65 12—13 90 18 75 21 48 5 09 19 75 63 13—14 92 20 72 21 56 1 66 21 76 63 14—15 79 21 67 10 60 1 57 22 67 00 15—16 76 22 77 15 67 1 62 18 72 56 16—17 83 20 82 17 71 3 71 16 78 56 17—18 83 19 87 16 52 5 65 21 75 61 18—19 83 19 100 17 67 6 65 21 80 63 19—20 86 21 90 19 78 7 71 22 82 09 20—21 93 22 110 19 91 8 77 21 92 70 21—22 87 26 107 22 72 6 71 22 87 76 22—23 79 ' 20 109 16 67 6 72 20 83 71 23—24 85 27 103 17 86 7 73 19 86 70 Tbe curves based on this table are the thick ones shown in the upper part of figure 92. It will be seen at once that the daily variation is very similar in all seasons, the maximum occurring during the forenoon and the minimum during the afternoon. From the few observa- tions made by Bsrnacchi on Captain Scott's first expedition C. T. R. Wilson drew the same conclusion, so there can fie little doubt that this is the normal course of the potential gradient in McMurdo Sound. It is a surprising result for it is very abnormal. At all other places in the world where reliable observations of the potential gradient have been made the main minima and maxima occur near 4 a.m. and 8 p.m. respectively. The question at once arises is the abnormal daily course of the potential gradient in McMurdo Sound due to the local meteorological conditions or is it a true electrical effect ? If the two last columns of table 158 are compared, it will be seen that there is a re- markable parallelism between the value of the potential gradient and the number of hours DAILY VARIATION OF POTENTIAL GRADIENT. 315 0 I 2 3 i b i 7 8 3 in ii_r " " '■' ■'■ " 'f '? f ? ]^ Y 7 Fl«. 92. Daily variation cf potential gradient. 316 ATMOSPHERIC ELECTRICITY. used in the determination. Thus during the early morning from .5 a.m. to 10 a.m. the potential is high and the observations are numerous. Daring the period of minimum potential from 1 P.M. to 7 p.m. there are relatively few observations. If the numbers are examined in more detail, it will be seen that the paralleHsm is very close. It will also be seen that the same relationship is repeated in each of the four seasons, the period of high potential coincides with the period of numerous observations and vice versa. That there are more observations in the early morning and fewer in the early afternoon is not chance, it is due to the fact that the air is more often calm in the early morning than in the afternoon. This is clearly brought out in the curves on figure 33, page 102, which show the frequency with which calms (0—1 miles per hour) occurred during the same period as that for which we have the potential gradient observations. The similarity between the curves showing the daily variation of calms and the daily variation of potential gradient (figures 33 and 92) is most remarkable and points to some close relationship between the two phenomena. Reasons for such a relationship do not appear difficult to find. It is a well-known fact that, generally speaking, when the electrical conductivity of the air is low the potential gradient is high, also that a high humidity of the air causes a low3ring of the conductivity. Thus one would expect p3riods of high humidity to coincide with high potential gradient. Unfortunately no humidity measurements were made at Cape Evans, but during calm weather one might expect the humidity to depend largely on the temperature. Now during the greater part of the year the temperature is lowest in the early morning and highest in the afternoon, we might therefore conclude with a certain amount of justifica- tion that the humidity is highest in the early morning and lowest in the early afternoon. Thus we see that the high potential might be accounted for in this way during the greater part of the year ; but what about the remaining part ? We have already shown (page 67) that during the three winter months, May, June, and July, the temperature variation is very small, and what there is of it makes the afternoons warmer than the night and early morn- ing. It must, however, be pointed out that all hours were used in this determination. Would the effect be the same if only the calm hours had been considered ? The fact that there are more calms during these months in the early morning than in the afternoon would lead us to expect that the afternoons would be the warmer ; for we have repeatedly shown that the more the wind the higher the temperature. To test whether or not this is so the temperature of every hour during May, June, and July, for which the potential gradient has been used, has been tabulated and the daily variation of the temperature of these hours determined. The following table contains the results and shows that the temperature is practically constant throughout the twenty-four hours and that there is no temperature variation at all similar to that of the potential gradient variation : — Table 159. Mean temperature of the ho'irs from which the potential gradient variation has been determined in May, June and July. °F. A.M. P.M. 1 2 3 4 5 6 7 -21 -20 8 9 10 11 12 -22 -22 -23 -22 -22 -21 -22 -23 -22 -23 -22 -22 -21 -21 -21 -23 -20 -23 -19 -22 -20 -23 DAILY VARIATION OF TOTENTIAL GRADIENT. 317 We thus see that during May, June, and July the potential gradient variation cannot be ascribed to changes in the conductivity of the air caused by the variation of temperature and humidity. It is therefore reasonable to assume that the potential changes in other seasons of the year also are not due to this cause. The next step was to investigate whether the daily variation of potential gradient which shows itself in the mean of a large number of broken periods exists when the weather re- mains calm throughout twenty-four hours. There are not many periods during which the wind did not rise above five miles an hour during twenty-four consecutive hours. There were none during February, March, and April ; but six ware found during May, June, and July, eight during August, Saptembsr, and Ocfcobsr, and four during November, December, and January. Eich of thss3 periods generally extended over a few more than twenty-four hours, but in each case the best series of twenty-four consecutive hours was chosen. From these few observations the daily variation was obtained and the result has been shown by the thin curves in figure 92. It will be seen at once that the character of the variation is the same and the differences between the thick and thin curves are to be expected on account of the few observations used to obtain the latter. This has confirmed the character of the variation for calm weather, but there is still the possibility that meteorological conditions are the determining factor. For example during Miy, June, and July, in spite of the absence of any appreciable temperature variation there is a pronounced tendency for the air to be calmer in the morning than iu the afternoon. This tendency for which no physical explanation can be given probably acts during all periods of calm and may affect the potential gradient as well as the air motion. If the potential gradient variation is not due to this meteoro- logical condition, whatever it may be, the variation should be the same when the air is not calm, i.e., it should be similar during winds as during calms. Unfortunately the presence of drift so affects the potential gradient during appreciable air motion that its daily variation is completely masked. The drift, however, does not become serious until the wind rises above 10 miles an hour. If therefore we investigate the potential gradient duiing i^eriods in which the wind velocity was between 6 and 10 miles an hour the daily variation of the potential, if it exits, should be recognisable. The hourly values of the potential gradient during winds of 6 to 10 miles an hour were therefore tabulated and it was found that in all seasons they showed the same general daily variation as during calms. The number of observations, however, in each season is very small, so that the curves are irregular. On the mean of the whole year, however, the observations are sufficient to give a rehable result ; numerical values are contained in table 160 and the curve is shown at the bottom of figure 92. Table 160. Potential gradient during winds of 6 to 10 tniles an hour. Volts per metre (Number of observations in brackets). A.M. P.M. 0—1 1—2 2-3 3-4 4—5 5—6 6—7 7—8 8—9 9—10 10—11 11—12 102 (31) 82 (38) 89 (28) 73(42) 98 (28) 72 (47) 93 (30) 85 (36) 103 (28) 79 (38) 98 (15) 85 (43) 100 (21) 90 (33) 103 (28) 98 (34) 91 (23) 99 (36) 95 (35) 103 (30) 91 (44) 104 (33) 85 (49) 104 (26) The curve for the potential gradient during winds of 6 to 10 miles an hoar is practically the same as that during calms; it is therefore quite clear that the variation of the potential gradient is not caused by the peculiar meteorological conditions existing during calm weather 318 ATMOSPHERIC ELECTRICITY. and as it is difficult to conceive of any meteorological condition which is the same during calms and winds and during summer and winter, we are forced to the conclusion that the variations in the potential gradient are real and not due to local meteorological condi- tions. It has already been remarked that the daily variation of the potential gradient found in McMurdo Sound is very abnormal. It was also shown above that the yearly variation in the Antarctic app3ars to bs different from that of the rest of the world ; are the abnormal yearly and daily variations in tho Antarctic related ? The only other observations of the daily variation of the potential gradient made in the Antarctic (other than those made by Bernacchi in McMurdo Sound) are two series made by members of Charcot's two expeditions on the coast of Graham's Land.* Both of these show a variation very similar to the normal variation found in other parts of the world: the main minimum is near tod a.m., but the main maximum is early, occurring at 3 p.m. instead of at 8 p.m. as is more usual. We are thus led to conclude that the daily variation of the potential gradient in McMurdo Sound is not only different from that in other parts of the world but is different from that of other places in the Antarctic. In other words not only is the daily variation abnormal but so far as observations go it is unique. The question at once arises is there anything in the geographical position of McMurdo Sound which also is miique ? McMurdo Sound possesses one feature which is different from all other places for which atmospheric electricity obs3rvations are available. On account of its position batween the magnetic and the geographical poles the direction of the earth's magnetic field relative to the direction of rotation of the earth is practically reversed from what it is at all other stations. | In other words in this part of the world if an observer stands with his face looking along the lines of magnetic force he moves on account of the earth's rotation from right to left instead of from left to right. In view of our total ignorance of any relationship between terrestrial magnetism and atmospheric electricity it is impossible to say what effect one would expect from this reversal, but a comparison between the daily variation of the potential gradient in McMurdo Sound with that of the most northerly station for which we have a complete set of data is instructive and sugges- tive. During 1903-04 the present writer spent fourteen months in the Lai:)p village of Karasjok (69° 17' N., 25° 35' E., 129 metres above sea-level) with the object of investigating the electrical conditions of the atmosphere there.J The results of the daily variation of the potential gradient for a whole year are shown in the upper curve of figure 93. As the amplitude of the daily variation in Karasjok was much larger than in McMurdo Soimd, the scale has been reduced in the ratio of 25=^10. A comparison between the curve for Karasjok and the curve for Caps Evans given in figure 92 shows a great similarity in the general shape of the two curves, but the times at which the maxima and minima occur are different. The chief minimum occurs at 14| hours at Cape Evans and at 5 hours at Karasjok, a difference of 9J hours ; while the chief maximum occurs at 7J hours at Cape Evans and at 21 hours at Karasjok, a difference of 10^ hours. It is obvious, therefore, that if the time at Cape Evans were put back 10 hours, the curves would agree not only in shape but also in phase. To bring this out more clearly the curve for Cape Evans has been smoothed by the formula — ^ — and plotted under the curve for Karasjok with its time values put back 10 hours. The similarity of the two curves is now very apparent. * I. I. Key. E.xpeditioa antaiotique francjaise, 1903-05. Hydrographic, Physique du Globe, Paris, 1911, page 191. I. Rouch. Deu.xiome Expedition antarotique fran(;ai3e, 1908-10. Sciences physiques, Paris, 1913, page 14. t The declination at Cape Evans is 153^ E. t G. C. Simpson. Philosophical Transactions, A, 205, pagea 4-97, 1905. DAILY VARIATION OF POTENTIAL GRADIENT. 319 200 190 ISO 170 leo 150 MO ' ,^ k 1 IIO lOO 90 so 70 ^ s \ / \ > Y \ __ ^ ^ ^ \ ,^ r r \ 130 I20 IIO lOO 90 80 ^ Y \ V karasjok ^ ^ J r N / N ^ y V 0 ) M K ' ( r ^ ' i 1 1 1 1 1 1 ) 11 12 13 14 15 16 17 1 8 19 20 21 2? 23 24 '\. r — - ^ ^- \ \ V \. / / c jpe Eva ns \ ^ ^s >^ N \^ y r .^ y K ) II _ I? 13 14 " 15 16 17 18 19 20 ?1 22 23 24 1 2 3 4 5 6 7 8 9 10 1 Hours 1 Fig. 93. Daily variation of potential gradient. The same conclusion is reached if we analyse the two curves into Fourier's saries. The equations to the two curves are Karasjok P=L39 + 39 sin (9 + 177'') + 23 sin (28 + 1-58 ) Cape Evans P= 86 + 12 sin (e+ .33')+ .5 sin (29 + 228") in which in each case 9 is measured from local midnight and equals 1-5" for each hour. Now the diSerence in phase between the two first terms is 177°-33''=144°, which is equivalent to 9-6 hours and the difference in phase between the two second terms is 1.58°-228''=290"', which is equivalent to 9-7 hours. Hence if we change the time at Cape Evans by 9-6 hours, the phase of both terms becomes practically identical at both stations. Now is it only coincidence that this value of the chaiage in phase is almost the same as the diiierence in magnetic declination at the two places ? The declination at Karasjok is approximately 0" and at Cape Evans 153" E., i.e., a difference of 153", which is very near the phase corresponding to 9-6 and 9'7 hours (144" or 145"). This result is very suggestive, but we are not justified in discussing it in detail, for many more observations at places having different values of declination are necessary before we shall be able to say with certainty that there is any relationship between the direction of the earth's magnetic field and the daily variation of the potential gradient. Atmospheric radio-activity. Forty-six measurements of the radio-activity were made during the winter months, May to August, and twenty-eight during the summer months, December to February. The method used was the same as that employed on the Terra Nova on her voyage from England to New Zealand.* The object of the work was to compare the atmospheric radio-activity in Simpson and Wright. Proceedings of Royal Society, A, Volume 85, page 175, 1911. 320 ATMOSPHERIC ELECTRICITY. the Antarctic where there is a permanent covering of snow and ice with that of ordinary land and the open ocean. On the arbitrary scale used (approximately that defined by El.ster and Geitel and denoted by A) * we found during our voyage the following results : — Tabi/E 161. Radio-activity. Over land (Matjesfontein) . Over ocean A mean. A max. A min. 124 (17) G (102) 256 21 52 1 Exactly the same instruments and methods were used in the Antarctic with the following results : — Table 162. Rculio-asiivity at Cape Evans. May to August . . December to Febraary A mean. A max. A min. 20 (46) 16 (28) 60 30 6 6 We see that the radio-activity in the Antarctic is higher than over the open ocean and lower than over the land of South Africa. This is exactly what we should expect if the origin of the emanation is the radium contained in the rocks of the earth's crust. AVhen the effect of the meteorological conditions are considered, we find here as in every other part of the world that the meteorological conditions which favour stagnation of the air near the ground result in relatively high radio-activity. Thus we should expect the radio- activity to decrease with increasing wind force and to increase with decrease of temperature and this is found to be the case. The mean values of the radio-activity for various wind strengths and temperatures are shown in the following table : — Table 16.3. Rculio-activity. wind and temperature. A No. of observations . Wind. Tempeeatuee. 0—10 11—30 >30 <-15 —15 to 0 0—15 >15 20 (40) 17 (27) 14 (6) 22 (20) 17 (14) 15 (5) 16 (28) * Elster and Geitel. Phys. Zeit., 1902, Volume 3, page 305. INDEX. A Air movement in layers, 125, 127. Amplitude, daily : definition of, 53. Amundsen, Captain R. : visit to the South Polar Plateau, 1, 141 ; observations taken on the journey to the South Pole, 10, 38, 39, 111-115, 215, 231 ; observations to determine the height of the South Pole, 292-291. Anemometers : in u.se at Cape Evans, 94. Anticyclone, Antarctic : cloud, 151 ; Gauss Station, 246 ; Hobbs' theory, 248 ; Meiuardus's theory, 252 ; theoretical discussion, 2-57-268 ; wind frequency, 104. Atkinson, E. L. : observations taken by, 9, 29, 30. Atmospheric electricity : chapter X, .302 ; instruments and methods, 302. B Bage, R. : visit to the south magnetic pole, 141. Balloons : description of, 270. Balloon instruments : methods of retrieving, 271. Balloon observations : summer ascents, 277 ; winter ascents, 273. Barkow, E. : temperature gradient over the Weddel Sea, 300. Barometers, 163. Barrier: extent of, 4; geographica 1 features of, 4; surface of, 4. Baschin, 0. : pressure over the Southern Ocean, 299. Bates, Rbv. D. C, vii. Batt, R. M., viii. Beaufort scale of wind velocity, 130. Bernacchi, L. C. : observations of potential gradient, 316. BLtzzARD(s) : cloud and drift during, 148 ; cold during summer, 47, 50, 52 ; erroneous explanation of temperature during, 51 ; fohn, 51, 251 ; general description of, 12 ; Hobbs' theory of, 250 ; motion of upper atmosphere during, 136 ; origin of precipitation during, 266 ; pressure waves and, 217, 226, 234, 235 ; sudden comniencement of, 128, 238 ; temperature at Cape Evans and Framheim during, 52 ; wind and temperature during, 48 ; wind direction during, 108 ; winds on plateau during, 144. BoDMAN, G. : Antarctic winds, 98. BoRCHGREViNK, C. E. : observations at Cape Adare, 80. Bowers, H. R. : v ; journey to Cape Crozier, 8 ; journal kept on Polar journey b}^, 9 ; obser vations taken by, 19, 20, 24. c Calms : annual variation of, 99 ; daily variation of, 103 ; temperature during, 46 ; weather asso- ciated with, 11. Cape Adare : description of the station at, 7 ; history of the station at, 10. Cape Crozier : winter journey to, 21. Cape Evans : description of the station at, 5 ; history of the station at, 8. Cherry-Garrard, A. : journey to Cape Crozier, 8 ; observations made by, 3, 9, 21, 29, .30. 41 322 INDEX. Cloud(s) : amount 149 ; annual variation of, 152 ; daily variation of, 155 ; during blizzards, 148 ; height of, 132; kind of, 147; motion of, 132 ; radiation and, 154; wind and, 151, Cloud and precipitation : chapter IV, 147. Cold surface layer of air : effect of, on maximum and minimum temperatures, 76 ; formation of, 43 ; gustiness and, 121 ; non-periodic changes of temperature and, 76 ; removal of, 45, 50 ; sudden commencement of blizzards and, 238 ; wind and, 100, 125. CoNVEXiON CURRENTS : clouds and, 149 ; vertical temperature gradient and, 43 ; wind velocity and, 100. Corner Camp: position of, 6; wind vane at, 112. Crowther, Miss E., iii. Crowther, W., iii. Cyclones : at Cape Adare, 239 ; at the Gauss Station, 241-247 ; over the Southern Ocean, 206- 209, 245, 246. David, T. W. E. : joiirney to the south magnetic pole, 1. 141, 142. Day, B. C. : observations taken by, 0. Derbyshire fund, iii. Dines, W. H. : iv ; discussion of temperature observations taken on the Discovery Expedition, 21, 89 ; hydrogen generator and balloon instruments, 270. Drift snow : general description of, 12 ; during blizzards, 148 ; frequency of, 161. E Elster and Geitel : unit of atmospheric radio-activity, 320. Erebus, Mount : features of, 6. Erebus smoke : motion of, 131. Evans, E. R. G. R. : observations taken by, 9, 26 ; voyage in the Ross Sea, 2. FiLCHNER, W. : exploration in the Weddel Sea, 93. Fog : 147 ; cause of, on the Barrier, 268. Fog bows, 9. FoHN AND blizzards, 51, 251. Fram type op daily temperature variation : Meinardus's explanation of, 62 ; Mohu's explana- tion of, 60. Framheim : description of the station at, 7 ; history of the station at, 10. G Gauss Station : position of, 242 ; pressure, winds and weather at, 242-247. General air circulation over the Antarctic : chapter VII, 248 ; Hobbs' theory of, 248 ; Meinardus's theory of, 252 ; Hobbs' and Meinardus's theories compared, 256 ; theoretical dis cussion of, 257-268. Gold, E. : barometric gradient and wind, 242. Gran, T. : vi ; meteorological work, 8. Greely, a. W. : daily variation of pressure in the Arctic, 181. Gustiness : definition of, 115 ; annual variation of, 119 ; cause of, 121 ; temperature and, 123. Gutta Percha Company, iii. H Hann, J.: land and sea climates, 86; temperature observations at Cape Royds, 79; vertical tem- perature gradient, 38. Height of the Barrier and South Polar Plateau : chapter IX, 286. INDEX. • 323 Height(s) : of the Antarctic Coutiiiont, 294 ; of the Barrier, 286 ; on the Beardmore Glacier, 288 ; of the South Polar Plateau, 290 ; of the South Pole, 294. Hepworth, M. W. : air circulation over the Antarctic, 206 ; depressions in the Ross Sea area, 209. Hertz's : thermodynamic diagi-am, 266. floBBS : theory of blizzards, 250 : tlieory of glacial and Antarctic anticyclones, 248. Holyhead : gustincss at, 118. Humidity, iv. Hunt, H. A., vii. Hut Point : situation of, 6. Hydrogen generator, 270. I Ice conditions : near Cape Crozier, 3 ; in McMurdo Sound, 2, 6 ; in the Ross Sea, 2, .3. Insolation : annual variation of, at 78° N. and 78° S., 8-5 ; annual variation of temperature, 84-90 ; temperature on the plateau, 41. Introduction : chapter I, 1. Iridescent clouds, 9. Isotherms: .January, 34, 3-5; July, .35. J Jubbulpore : frequency of winds at, 103. K Kelvin, Lord : daily variation of pressure, 181. Koppen, W. : non-periodic pressure changes, 184. LocKYER, J. S. : depressions in the Ross Sea, 209 ; theory of southern hemisphere surface air circulation, 206, 208, 246. M Mawson, Sir D. : expedition, 141, 144 ; wind in Adelie Land, 14, 97. McMurdo Sound : general description of, 6. McMurdo type of daily temperature variation, 60. Meares, C. H. ; observations taken by, 9, 19, 20. Meinardus, W. : depressions at the Gauss Station, 242 ; depressions in the Ross Sea, 209 ; determination of the height of the Antarctic Continent, 294 ; explanation of the ' Fram ' type of daily tem- perature variation, 62, 64 ; general pressure distribution over the Southern Ocean, 208, 246 ; mean temperature of the Antarctic, 92, 93 ; snow-covering at the Gauss Station, 65 ; theory of Antarctic air circulation, 252. Meteorological Office: Australian, iii ; London, iii ; New Zealand, lii. Mohammad Bakhsh, viii. Mohn, H. : viii ; calculation of solar radiation, 63 ; daily temperature variation in the Arctic, 70, 71 ; determination of the height of the South Pole, 293 ; explanation of the ' Fi-am ' type of daily temperature variation, 60, 64, 66 ; mean temperatures in the Arctic, 93 ; ' Norwegian North Polar Expedition,' 40, 59, 63 ; temperature on the South Polar Plateau, 38. Mossman, R. C. : vii ; large pressure changes in Iceland, 186 ; pressure data of the Discovery Expedition, 166 ; supply of pressure data, 200. MuNRO, R. W. : iii ; pressure tube anemometer, 94. N Nelson, E. W., vi. Normand, C. W. : viii ; solar energy amplitude, 63. 324 INDEX. National Physical Laboratory, iii. Northerly winds : frequency of, 107, 113 ; mean velocity of, 113 ; temperature during, 46 ; total motion of, 113 ; weather associated with, 11. One Ton Camp : position of, 7. Optics, meteorological, v. P Pack ice, 2. Plateau : geogi'aphical features of, 1 ; visits to, 1, 141, 142. Potential gradient : annual variation of, 312 ; daily variation of, 314 ; during drift, .306 ; mag- netic declination and, 320 ; instruments and methods, .302 ; negative, .307. Precipitation : at Cape Evans, 1.59 ; origin of, within Polar anticyclone, 266. Pressure : instrunjents and methods, 163 ; large change from October to November, 186 ; large changes in Iceland, 186 ; non-periodic changes, 184 ; rise due to cyclones near Cape Adare, 241. Pressure, annual variation : at Cape Adare, 173 ; at Framheim, 171 : at McMurdo Sound, 160 ; at Gauss Station and Snow Hill from correlation, 199 ; summary, 17.5. Pressure changes and winds, 236. Pressure correlation : between the Antarctic and surrounding places, 201 ; within the Antarctic, 196. Pressure, daily variation : 176 ; abnormal values near 4 a.m., 181 ; at Antarctic stations, 182 ; cause of twelve-hourly oscillation, 181 ; Fourier coefficients, 170 ; twelve-hourly period, 180 ; twentyfour-hourly period, 180. Pressure difference(s) : between Cape Evans and Framheim and winds, 236 ; betweer January and July over the Antarctic, 298. Pressure distribution in Ross Sea area : actual, 175 ; theoretical, 219 ; temperature and, 173, 237 ; weather and, 238. Pressure distribution over the Antarctic : at sea-level, 261 ; at 3,000 metres, 262 ; vertical, 258. Pressure gradient : in lower and upper atmosphere, 136. Pressure, mean monthly : at Cape Adare, 173 ; at Framheim, 171 ; in McMurdo Sound, 167. Pressure surges : 189 ; annual variation of, 191 ; affect the whole Antarctic, 193 ; cause of, 196 ; centre of system of, 196 ; comparison in Ro.ss Sea area, 191. Pressure, winds and weather : chapter VI, 206 ; at Cape Adare, 239 ; at the Gauss Station, 241 ; in Ross Sea area, 210. PRESSUR.E WAVES : 186 ; comparison at Antarctic stations, 188 ; direction of motion in Ross Sea area, 216 ; effect of, comparison of theory with observations, 230 ; at the Gauss Station, 244 ; moving pressure systems and, 212 ; not caused by blizzards, 235 ; on the plateau, 213 ; theore- tical effect of, 218 ; over the Southern Ocean, 245 ; pressure distribution and, 210 ; winds and, 217. Priestly, R. E. : meteorological record at Cape Adare, 10 ; winds at Cape Adare, 128. B Radiation and cloud, 154. Radio-activity : atmospheric, 320. Range, daily : definition of, 53. Reduced range : definition of, 73 ; correlated to maximum and minimum temperatures, 74. Ross, Captain Sir J. C. : discovery of the Ross Sea, 1, 2. Ross Sea : geographical features of, 1 . RuDGE, W. A. D. : electrification of dust, 309. s Sastrugi : in McMurdo Sound, 111. SciLLY Isles : gustiness at, 118. , INDEX. 325 Scott, Captain R. F. : viii ; ascents to, ami observatious made on, the plateau, 1, 38, lll-lii 214, 215, 293, 294 ; joiiraey to the western mountains, 9 ; low temperatures on the southern journey, 29, 30 ; remarks on the journey to Cape Crozier, 25 ; storm at the foot of the Beardmorc Glacier, 28, 39 ; temperature measurements on the Discovery Expedition, 21 ; use of Assmanu's psychrometer, 17; voyage along the Barrier in 1902, 1, 2; wind vane at Corner Camp, 112. Shackleton, Sir E. : temperatures at Cape Royds, 79 ; visit to the plateau, 1, 141, 14(j. Shaw, StR N. ; disappearance of clouds due to radiation, 153 ; gustiness, 121. Sledgixg journeys : 8 ; method of measuring temperatures on, 17 ; use of minimum thermometers on, 19, 20. Sledging thermometers, 17. Snowfall : 159 ; frequency of, 161. Solar radiation : during September to March, 63 ; amplitude of daily variation of temj)erature and, 64. Southerly winds : frequency of, 113 ; mean velocity of, 113 ; temperature during, 46 ; total motion of, 113 ; weather associated with, 11. Southern Ocean : pressure conditions over, 246. South Pole : height of, 294 ; routes from Cape Evans to, 6. Structure of the wind, 115. Squalls, 125. Sun : periods above and below horizon, 54. Surges : see • pressure surges '. Taylor, Q. T. : v ; journeys to the western mountains, 10. Te'MPERAiURe(s) : chapter II, 15 ; accuracy of the observations, 16 ; Arctic and Antarctic, com- pared, 40, 93 ; Cape Adare, 33 ; Cape Armitage, 21, 22 ; Cape Crozier, 25 ; Framheim, 32 ; geographical distribution of, at sea-level, 21 ; gustine.ss and, 123 ; in.struments and methods, 15 ; interdiurnal varial)ility of, 77 ; lowest recorded in the Antarctic (— 76°F.), 8, 24 ; mean annual, 91 ; mean of the Antarctic, 92 ; method of ob.serving at Cape Evans, 16 ; method of observing on sledging journeys, 17 ; pressure distribution and, 173, 237 ; on the South Polar Plateau, 38; wind and, 44, 83. Temperature, annual variation : 83 ; Arctic and Antarctic, compared, 84 ; explained, 88 ; insolation and, 85 ; land and sea climates and, 87 ; at Cape Adare, 83 ; at Framheim, 83 in McMurdo Sound, 81 ; on tlip North Barrier, 32, 83 ; at 78° N., 85. Temperature on the Barrier : daily variation of, 55 ; method of reducing, 19, 20, 21 ; mean monthly, 32 ; spring, 26 ; summer, 26 ; winter, 21. Temperature, daily variation : 52 ; at Cape Adare, 73 ; Fourier coefficients, 72 ; on the Fram drift, 54, 59, 70 ; at Framheim, 57, 60 ; ' Fram ' type of, 59 ; at the Gauss Station, 59 ; mean range of, 74 ; mean reduced range of, 74 ; in McMurdo Sound, 54 ; ' McMurdo ' type of, 60 ; at Snow Hill, 60 ; state of surface and, 58, 64, 65 ; solar radiation and, 64. Temperature, daily variation, May, June and July : 66 ; marked excess of temperature at 4 A.M., 69 ; excess of temperature from 9 a.m. to 7 p.m., 68. Temperature, daily variation, September to March : 54 ; explanation of, 63 ; summary and conclu.sions, 66. Temperature differences : Barrier-Cape Evans, 31 ; Hut Point-Cape Evans, 91 ; January to July over Antarctic, 300. Temperature inversion : 41 ; see also ' cold surface layer of air.' Temperatures, maximum and minimum : mean monthly values of, 74 ; corrections to convert to true mean, 79 ; correlation with reduced range, 74. Temperature, mean monthly : data available, 79-81 ; departures from average in McMurdo Sound, 82; at Cape Adare, 83; at Framheim, 83; in McMurdo. Sound, 81: on the North Barrier, 3%, 83 ; at 78° N., 85. Temperature, non-periodic changes : 73 ; cold surface layer and, 76 ; comparison between Cape Evans, Cape Adare and Framheim, 77 ; maximum and minimum temperatures and, 73. 326 INDEX. Temperature, vertical distribution : 41 ; convexioD currents and, 43 ; see also ' Upper air observations.' ' Terra Nova,' voyages of, 10. Thermograph : in the Hut, 15, 16 ; in the screen, 15, 16. Thermograph records : method of reducing, 17. Thermometers : comparison of, 17 ; sledging, 17 ; in use at Cape Evans, 16, 17 ; minimum used on sledging journeys, 19, 20. Time : used at the stations, 10. u Units, vii. Upper air observations : chapter VIII, 270 ; winter ascents, 273 ; summer ascents, 277. Upper atmosphere, motion of : at Cape Adaie, 133, 139 ; at Cape Evans, 131 ; at Hut Point, 139 ; over Antarctic at approximately 3,000 metres, 264 ; during blizzards, 136. V Variation, daily : definition of, 52. Vertical distribution of temperature : 41. Volume II : description of the maps in, 211 ; description of the plates in, 210. w Walker, G. T., viii. Weather: associated with calms, northerly winds and southerly winds, 11 ; at Cape Adare, 13; at Cape Evans, 11 ; at Framheim, 13 ; pressure distribution and, 238. Weather maps : method of construction, 211. Wilson, C. T. R. : potential gi-adient on Scott's first expedition, 314. Wilson, E. A.: journey to Cape Crozier, 8, 21, 22, 24; Tranquil Bay, 111. Wind : chajrter III, 94 ; Cape Adare, 128 ; cloud and, 151 ; cold surface layers of air and, 125 ; comparison of simultaneous winds at Cape Evans and Framheim, 232 ; gustiness, see ' gusti- ness ' ; instruments and methods, 94; structure of, 115; sudden breaks in velocity of, 127; sudden commencements of, 128 ; at various heights over Cape Evans, 138 ; pressure changes and, 236 ; pressure differences and, 236 ; temperature and, 44, 83. Wind direction : 107 ; during blizzards, 108 ; northerly and southerly groups, 113. Winds, plateau : 141-146. Wind steadiness : definition of, 108. Wind vane : in use at Cape Evans, 95 ; self-recording at Corner Camp, 112. Wind velocity : annual variation of, 97 ; annual variation of high winds and calms, 99 ; cold surface layers of air and, 100 ; daily variation of, 100 ; daily variation cf calms and, 103 ; daily variation of, and convesion currents, 100 ; frequency of, 103 ; highest recorded, 97 ; mean monthly values of, 96. Wright, C. S. : vi ; observations made by, 9 ; took charge of meteorological work, 8. Y Yarmouth : frequency of winds at, 103. 3000, March 1912 lb ^o a 30 28-00 30 OO 28 OO March April May 10 20 3r 10 20 30 10 20 31 June jojy AuguBt September October November December January February J" ^2 3° 10 ;0 31 10 ?0 31 10 20 30 10 ?0 31 K) 20 30 10 20 31 10 20 31 IQ ;o 23 ^OOOf 28- OO 3B-00 GAUSS PLvn: i\- HiATE V ■i