342 



A STUDV OF CORRELATIONS AMONfJ TERRESTRIAL TEMPERATURES. 



grade scale and correcting the departui-es fur the sun-spot fluctuation. They are given 

 in some cases for the individual stations, and in others for entire regions. The column 

 "India" is the weighted mean of the four Indian stations alone, which has been 

 separately formed for a reason which will hereafter !je shown. 



In combining the departures into a general mean it is advisable to assign different 

 weights to different stations, on account of the diversity of the mean fluctuation, as 

 shown in the several columns. If we could regard each departure as independent of 

 all the others, and free from any source of sj^stematic error, the weights would be pro- 

 portional to the inverse square of the mean fluctuations, as given in each column. 

 But this course would result in giving too great a relative weight to the stations of 

 small fluctuation. Actually, in the first combination, the weights used are those at 



the bottom of the several columns. 



Table VI. 



Treatment of Annual Departures. 



'J'he process of applying the criterion for correlation is shown in Table YJ. To 



