WEATHER GOVERNED BY THE SUN — ABBOT 95 



in the 10-day mean values. There are, indeed, quicker solar varia- 

 tions which run their courses in a few days, and these are believed 

 to have important effects on weather, but the study of them must 

 be deferred until steps now being taken somewhat increase the ac- 

 curacy of our daily observations. In mean values covering 10 days, 

 or better still 1 month, the daily errors, some being plus, some minus, 

 are largely smoothed away. 



Figure 2 shows the monthly mean solar-constant values since 1920 

 analyzed to yield the 12 periodicities above referred to. In making 

 the analysis the data are treated in several separate parcels for all 

 the periodicities of 25 months and less, so as to see if the earlier and 

 later years agree in presenting similarly the periodicity in question. 

 In illustration, I call attention to the periodicity of 11 months, for 

 which the results of three partial analyses of 5 years each are first 

 shown. At the bottom of that series a heavier line gives the general 

 mean for 15 years. A fair agreement between the three 5-year 

 intervals is apparent. However, as all the periodicities, and the 

 accidental errors besides, are confused in the original data, it is not 

 possible to separate perfectly and determine accurately the indi- 

 vidual periodicities as well as one would like to do. Especially for 

 the longer periodicities, of 34 months and over, the determinations 

 of the curves are imperfect because in 15 years there are so few 

 repetitions of them. 



Curve B is the summation of the 12 periodicities indicated by the 

 heavj'^ lines at the bottom of figure 2. When compared with the 

 original curve of observation A, it is apparent that a very good fit 

 has been obtained. Not only major changes such as that of 1922, but 

 minor details in the curve of observation A are closely repeated in the 

 synthetic curve B. In fact, the average departure between curves A 

 and B over the 180 months covered by curve A, figure 2, is less than 

 one-fifth of 1 percent. This is really a surprisingly good result and 

 lies, indeed, within the accidental error of the observations. 



It seems apparent, therefore, that the emission of the sun varies, 

 and that its variation is the complex result of the simultaneous exist- 

 ence of at least a dozen periodic terms. But what adds decidedly to 

 the interest of this conclusion is a fact that happened to be noticed 

 after many of the periodicities had been found. It is this: They 

 are all approximately integral submultiples, or, as we should say in 

 arithmetic, aliquot parts of 23 years. For 23 years is 276 months, 

 which divided by the numbers 1, 3, 4, 6, 7, 8, 11, 13, 25, 28, 34, and 

 39, gives, respectively: 23 years, 92, 68, 46, 39%, 34, 25Ki, 21^(3, llKe, 

 9%, 8K7, and 7K3 months. 



If 274 months rather than 276 had been chosen as the least common 

 denominator, the series of numbers just given would certainly have 



