Periodic Seeding -58- 



"T'he value of CC(28) at these nine points of intersection 

 ranged from 0.50 to 0.85. The area represented is 1.5 million 

 square miles. 



"Recently we have extended this grid of regularly spaced 

 stations to include the intersections of the 45° N parallel with 

 the 70° W and 110° W meridians, these points giving CC values 

 of 0.66 and 0.65 respectively. The 30° N, 80° W intersection 

 just off Jacksonville, Florida, also gave a correlation of 0.65. 

 We thus have an area of two million square miles or 2/3 of the 

 area of the United States in which CC(28) exceeds 0.50 with a 

 mean value of CC(28) =0.67. 



"We have also examined these periodicities at corresponding 

 points for preceding and for following periods. The 28-day per- 

 iod in May showed low correlations. On the other hand, the two 

 preceding periods gave highly significant values. Apparently 

 the high periodicity in the upper air temperatures started about 

 January 25, 1950 and continued on until about May 1, 1950, covering 

 an average area of about half of the United States. 



"For the nine points of intersection during a 28 -day period 

 in April, 1950 the total variance of the temperature was deter- 

 mined by taking the total sums of the squares of the deviations 

 of these temperatures from their mean and dividing by 27, the 

 number of degrees of freedom. The data obtained in this way 

 are called the 'total variance'. Ey multiplying these values for 

 each of the nine stations by the corresponding square of the cor- 

 relation coefficient CC(28), one obtains the 'periodic component 

 of the variance'. 



"Exactly similar calculations were made for a 28-day period 

 in April, 1949 when there was no periodic seeding. The results 

 are given in Table 3. At each point the upper figure is the 

 'periodic component of variance' for the April, 1950 period, 

 and the lower figure is the corresponding value for April, 1949. 

 The average values for all these nine points show that the 

 'periodic variance' in 1950 was 18 times as great as in 1949. 



"Table 4 gives the corresponding values of the 'residual 

 component of variance' obtained by subtracting the 'period vari- 

 ance' from the 'total variance'. These data then indicate how 

 all the other kinds of periodicities, beside the seven-day peri- 

 odicity, compared v/ith one another in the two years. It will be 

 seen that there is only about 10 per cent difference between the 

 average variance of this type for 1950 and 1949. 



