2 SMITHSONIAN MISCELLANEOUS COLLECTIONS VOL. 68 



one which in the Hght of our present knowledge might be expected to 

 respond most readily to solar changes. A station situated in the 

 center of a great continent, and in a tropical or subtropical region 

 seemed, for this reason, the most favorable. 



CORRELATION OF SOLAR RADL\TION WITH TEMPERATURE 



AT PILAR 



Pilar is a station in Central Argentina, lat. 31° 39' S., long. 63° 

 51' W., from which there were fairly complete and accurate observa- 

 tions at hand and appeared to meet these requirements. 



The part of the day when the temperature appeared most likely 

 to respond to solar changes was the afternoon ; accordingly the after- 

 noon maximum for each day was selected for the first trial. 



The method usually employed in determining the correlation factor 

 for two variables is, first to obtain the average value of each variable 

 separately, and subtracting this from the individual observations, to 

 obtain a set of residuals from which the correlation factor may be 

 computed by the formula : 



2:1-3; _ 



In the present case, the successive values of x are the departures of 

 the individvial observations from the average value of solar radiation 

 during the period under consideration, and the successive values of 3; 

 are the deviations of the temperature on each day from the mean of 30 

 days. In the observation of solar radiation of Mt. Wilson in 191 3 

 there was a change in the mean value during the latter part of 

 September, and, as this change corresponded closely with a change of 

 instrument, it seemed best to divide the observations into two parts, 

 one preceding September 23, and the other following that date. 

 Dr. Abbot believes that a real change occurred in the mean solar 

 values at that time ;TDUt as my object was to study the correlation of 

 the short period solar changes with terrestrial meteorological changes 

 of short period, it was desirable to eliminate any change in the mean 

 value, by dividing the period into two. In 1914 there was no indica- 

 tion of an abrupt change in the mean values of solar radiation during 

 the season, so that the deviations were taken from the mean of the 

 whole season. The results are given in table i under the heading x. 

 These values of x added to the mean value at the top of the column 

 give the observed values at Mt. Wilson on the dates given in column i. 



In the case of the temperatures, it is also necessary to eliminate 



