FORECASTING FROM HARMONIC PERIODS 

 IN PRECIPITATION 



By C. G. Abbot 

 Research Associate, Smithsonian Institution 



This paper presents evidence showing that the identical family of 

 harmonic periods found in solar variation is also present strongly in 

 terrestrial precipitation, and may be used for long-range forecasting. 

 I wish to direct attention of those interested in water supply to fig- 

 ures 7 and 8, and to the Conclusion of this article. Figures 7 and 8 

 show in A the march of yearly precipitation at two stations for 36 

 years. The curves A in figures 7 and 8 are made from recorded 

 observations after it happened. The curves B, which are for practical 

 purposes nearly identical with curves A, could be predicted and were 

 actually predicted from records of observations made long ago before 

 the events happened. 



1. PERIODS IN SOLAR VARIATION 



Volumes 5, 6, and 7 of the Annals, Smithsonian Astrophysical 

 Observatory, tabulate over 9,000 days' measures of the solar constant 

 of radiation. At page 13 and figure 1, Smithsonian Publication 4545, 

 it is computed that the daily accidental error of a measure of the solar 

 constant from one solar station is 0.007 calorie. Considering the num- 

 ber of stations responding, and the loss of days by clouds or other 

 causes, the accidental probable error of mean monthly values is esti- 

 mated at 0.05 percent of the solar constant. 



Mr. Jon. Wexler has electronically smoothed the monthly solar 

 constant values, 1921 through 1952, by the formula c^ = l/10(a + 2& + 

 4c + 2d + e) which gives the central month, c, | as much weight as 

 the total weight of the other four. His results are given by table 1 and 

 figure 1. 



In the year 1922 an exceptional depression of about 8 percent in 

 figure 1 is seen. This may possibly indicate the existence of a solar 

 period of long duration. The results from 1923 to 1955 display many 

 cases of long trends with amplitudes of solar variation up to 3 per- 

 cent. Figure 8a of Publication 4545 shows that the monthly values of 

 the solar constant repeated themselves, approximately, in great detail 

 with an interval of 273 months. 



SMITHSONIAN MISCELLANEOUS COLLECTIONS, VOL. 148, NO. 8 



