lO SMITHSONIAN MISCELLANEOUS COLLECTIONS VOL. 94 



Figure 5 illustrates the mean result of all the evidence from 1920 

 to 1934, inclusive. It depends on observations at Calama and Monte- 

 zuma, Chile ; Mount Harqua Hala and Table Mountain in the United 

 States, and Mount Brukkaros, Africa." 



The range of variation of solar radiation as indicated by the lo-day 

 mean values of the solar constant is given by table 2. 



4. Periodicities in Solar Variation 



To casual inspection the solar variation is irregular. More careful 

 inspection discloses an 8-month periodicity. Eleven other periodicities 

 have also been found and evaluated. As successively discovered, they 

 have been removed by subtraction, one by one, from the numerical 

 record so as to simplify the search for other solar periodicities. The 

 process of evaluating and removing periodicities is illustrated for an 

 1 1 -month period by table 3 and figure 6. Plate 2 shows a machine 

 capable of doing the same thing.'" * 



The reader will note that this computation of the ii-month solar 

 period is separated into several parts nearly similar to each other whose 

 mean result is to be repeated consecutively and added to consecutive 

 repetitions of other periodicities to produce the second curve in fig- 

 ure 7. The partial mean curves computed in table 3 are seen to differ 

 somewhat in form and amplitude, but to agree fairly closely as to the 

 phases of maximum and minimum values of solar radiation. These 

 independent determinations at different epochs, all yielding 11 -month 

 periodicities in nearly the same phase, seem to strongly support the 

 veridity of the 11 -month solar period. The third group, indeed (1930- 

 1934) shows about 3 months lag in phase. As will be shown in sections 

 14B and 25 below, there is some reason to anticipate a change of phase 

 of some of the periodicities about January 1934. Possibly this is the 

 cause of the observed phase-shift. Later observations will settle it. 



° The values given in fig. 5 and table 2 are provisional for the years 1931-1934 

 and may be altered in revision. 



' See The Periodometer, Smithsonian Misc. Coll., vol. 87, no. 4, 1932. 



' In the analysis of curves, most investigators employ developments of Fourier's 

 methods. That is, they represent the observed curve as a summation of a number 

 of arbitrary harmonic curves of integral periodic relationships. These constituent 

 harmonic curves have the periods i, J, i, . . . i/n of the entire unit length of the 

 curve analyzed. In such a case as that of a harmonic analysis of the sun-spot 

 numbers, none of the constituent harmonics have any independent physical 

 sigi-iificance whatever. Nor is it to be supposed that the harmonic form itself 

 represents at all closely the march of any physical quantity connected with the 

 phenomenon. It has seemed to me preferable to discard this tedious and arbitrary 

 procedure, and to compute the actual mean forms of the solar periodicities as 

 illustrated by table 3 and fig. 6. 



