NO. 3 



EFFECT OF SOLAR RADIATION CLAYTON 



17 



solar surface as seen from the earth at intervals of about 27 days, 

 the length of a solar rotation. 



This is a true solar period but would not be indicated by the method 

 of averages used by Stewart and Schuster because spots are con- 

 tinually disappearing, and new spots appearing in other parts of the 

 sun. 



For periodicities of this class the method of correlation offers a 

 method of research which I have tried successfully in studying 

 meteorological periods. Even if a period reverses its phase from time 

 to time this method would still bring out the period. 



The method consists in obtaining the correlation factor for intervals 

 of successively greater lengths. For example, all the solar radiation 

 measurements so far as the observations permit are correlated with 

 those made one day later, two days later, three days later, and so on 

 to any successive maxima of correlation indicating periods of a 

 length corresponding to the interval shown by the maxima. 



The unsmoothed measurements of Abbot, Fowle, and Aldrich in 

 1913 were treated in this manner with the following results : 



Table 7. — Correlation Factors for Diifercnt Periods 



These results are plotted in figure 3. They show chief maxima at 

 o, 12, 23 to 24, and 35 days indicating a period of about iif days. 

 There are secondary maxima at 5, 10, 15, 20 days indicating a less 

 marked period of five days. The probable error of the maximum on 

 the I2th day is ±.094 or about ^ of the computed correlation value. 



The means of each consecutive five values of the correlations are 

 shown by the broken curve. These show two maxima, a chief one at 

 22 days and a secondary maximum at 1 1 to 14 days. 



Underneath this solar curve is plotted a curve of correlation factors 

 computed from the 8 a. m. and 8 p. m. temperatures observed at 



