200 DISCOVERY REPORTS 



temperatures, the isotherms on the Hne of the convergence were derived from Tables 5 and 7. Ahhough 

 both are dependent for the most part on the same body of data this might have resuhed in some 

 inconsistencies, for Table 5 is no more than approximately correct, and as noted on p. 191 it is not 

 certain that it is equally applicable to all longitudes. Furthermore, the position of the isotherms not on 

 the convergence are in places influenced by extraneous data. However, the relative positions of 

 isotherms on the convergence and of those to the south of it were generally found to be quite con- 

 sistent, and only at one or two points was a compromise necessary. Thus in 150-170° E the 2° isotherm 

 in December and the 3° isotherm in January are placed a little south of the convergence, although they 

 should just be on it according to Tables 5 and 7, and a similar adjustment has been made in 60° W 

 in January. 



Tentative isotherms were thus drawn for all months and for all regions except for the major areas 

 in which observations were lacking, and the next step was to compare single isotherms month by 

 month. For this purpose the isotherms of one temperature, for example all the 0° isotherms, for each 

 of the months October to February (i.e. coldest to warmest months) were traced on to one chart, as in 

 Plate XIII. It is to be supposed that during the whole of this period the sea temperature is on the 

 average rising, and that therefore the 0° isotherm (or any other) will lie farther south in each successive 

 month. This was found to be so, the 0° isotherms forming a succession of concentric rings which did 

 not touch or intersect each other except at one or two points. Such instances were assumed to be 

 due to abnormal conditions or annual variations, and the positions of the isotherms were adjusted 

 accordingly. For example, it was found at one point that the November isotherm, depending mainly 

 on temperatures in 1926, ran for a short distance a little north of the October isotherm which depended 

 on temperatures in 1936. Further inspection of the temperature charts indicated that in an adjacent 

 region the spring temperatures in 1926 were rather lower than in several other years. Therefore where 

 the overlap occurred the November isotherm was adjusted so as to lie a little south of the October 

 isotherm. The fact that such adjustments were few, and involved only small changes in the position 

 of the isotherms, suggested that they had been drawn not far from the correct positions. Some other 

 adjustments were made so that isotherms of adjacent months should be approximately parallel and 

 show similar features ; but this was only done where it involved small alterations which were not too 

 inconsistent with the observed temperatures. Sometimes the form of an isotherm changes considerably 

 from one month to the next (cf. the 0° isotherms for December and January between 30° W and 30° E, 

 Plate XIII). 



One isotherm having been adjusted in this way, it was traced back in its new form on to the tem- 

 perature charts and the process was repeated for the other isotherms for September-February, and 

 then for all isotherms for February-September (the period of falling temperatures). It amounts of 

 course to a means of supplementing the data for one month with the data for other months on the 

 assumption that there is an unbroken rise of temperature from October to February and an unbroken 

 fall from February to September (see p. 198). 



The revised isotherms were checked against the temperature charts in case of any unjustified 

 departures from the observed temperatures, and were then transferred from the original semicircular 

 charts to the circumpolar charts of a rather smaller scale which are reproduced in Plates II-XII. It 

 remained to fill in the isotherms in the regions in which no observations had been made (mainly in 

 the winter months). These were sketched on the circumpolar charts in what appeared to be the probable 

 positions by analogy with the isotherms already drawn. Each isotherm was then traced as before on 

 to a separate chart, adjusted with the same isotherms for other months, and traced back again. These 

 isotherms are shown as pecked lines in Plates II-XII. Often it was a matter of simple interpolation. 

 For example, between 80 and 110° W the 0° isotherms for November and January are based on 



