DISTRIBUTION OF SURFACE TEMPERATURE i99 



comparatively few records for July and August, but these months are evidently not much warmer than 



September. 



Before any isotherms were drawn the Antarctic convergence was drawn in its mean position on 

 each of the temperature charts, and the mean position of the pack-ice edge for the month in question 

 was added. The latter was traced from the original charts reproduced in the report on the pack-ice 

 (Mackintosh and Herdman, 1940, plates Ixxi-xcv), and no alterations were made except for one small 

 adjustment. Inspection of the sea temperatures in the Falkland sector in November and re-examination 

 of the ice records suggested that the mean position of the ice-edge here should be drawn a little farther 

 south. Accordingly the line of the ice-edge in November between 10 and 70° W has been moved 

 about 20-40 miles farther south. 



The lines of the ice-edge and the convergence form very useful starting points in the drawing of the 

 isotherms. The line of the ice-edge can be taken as coinciding approximately with the isotherm of 

 - 1-5° C. (It does not of course always coincide for the sea temperature at the ice-edge is sometimes 

 above -1° in summer or below -1-5° in winter, but some allowance can be made for this.) From 

 Tables 5 and 7 (pp. 189 and 192) we can calculate which isotherms should lie on the line of the con- 

 vergence. If these steps are accepted then the intermediate isotherms must lie at intervals withm a 



limited belt. 



The isotherms are drawn at intervals of 1° C, and in the first place were sketched in pencil on the 

 temperature charts. Those falling on the convergence were entered first, and they were derived from 

 Tables 5 and 7. For example, in April in 60° W (see Plate VIII) the middle temperature by Table 5 

 is 4-0^^ and the range by Table 7 is 2-6°. This gives a gradient from 27 to 5-3° which includes the 

 isotherms of 3, 4 and 5°. In 50" W, however, the gradient is from 3-2 to 5-0° which includes only the 

 4° and 5° isotherms. The f isotherm must therefore diverge away from the line of the convergence 

 between 60 and 50'' W. In this way the isotherms were drawn along the whole length of the con- 

 vergence for each month. 1 At some points (e.g. 10° E) the average range of the gradient is less than 

 1° C and the line of the convergence cannot accommodate more than one isotherm. Elsewhere the 

 convergence can be distinguished by the concentration of two or more isotherms. 



The isotherms between the ice-edge and the convergence, and one or two also in sub-Antarctic 

 water north of the convergence, were then sketched in on the temperature charts. With data from several 

 different years included in a chart for one month, it was at once obvious that any attempt to join all 

 points of a given observed temperature would lead to absurdities. However, such points of equal 

 temperature were connected as often as possible provided this did not involve any improbable deviation 

 from the general trend of the adjacent isotherms and of the lines of the ice-edge and convergence. 

 If in any locality observations were available from only one year it was assumed that they represented 

 average conditions, unless observations from that and other years in adjacent regions indicated any 

 departure from the average. Where observations from different years suggested different latitudes for 

 the isotherm the line was drawn as nearly as possible in the mean position. At many points of course 

 it was difficult to decide where to place an isotherm, and even where observations were plentiful the 

 isotherms could be drawn with much more confidence at some points than at others. However, they 

 were drawn so as to make as good a fit as possible with the observed temperatures. Fig. 8 B (p. 197) 

 is an example of a fairly good fit. In this area the October temperatures for 1930 and 1938 corresponded 

 almost exactly In 1925 and 1926 in October and in 1936 in September the water was a little colder, 

 at least in the northern part of the area. September and October temperatures in 1938 were very 



''whereas the majority of isotherms were in this way drawn by eye on the charts of observed 

 1 But see below with regard to special adjustments at one or two points. 



