NO. 4 PERIODIC SOLAR VARIATION — ABBOT I7 



much weight. Indeed I regard the dotted one as best for those years, 

 since the dotted curve fits so well elsewhere. 



Obtaining differences between synthesis and observation, the fol- 

 lowing are mean values of deviations in percentages of the solar- 

 constant result: 



Epoch 1923 Sept. to 1927 Apr. 1927 May to 1929 Nov. 1923 Sept. to 1929 Nov. 



Mean percent ... ±0.18 ±0.11 ±0.15 



Epoch 1929 Dec. to 1932 Sept. 1932 Oct. to 1935 Dec. 1929 Dec. to 1935 Dec. 



Mean percent . . . ±10.09 — o.iS ±0.12 



To reach this close agreement between synthesis and observation, 

 not only are the main trends in solar variation followed in unison, 

 but most of the short-period variable features of observation are seen 

 also in the synthesis. Such good agreement over a span of 12 years 

 seems to warrant the belief that a synthesis over the 23-year span, 

 1900 to 1923, will give a trustworthy view of the solar variation dur- 

 ing that interval. Accordingly I have made such a synthesis, and give 

 it graphically in figure 5. 



Some critics might be inclined to say that the agreement of syn- 

 thesis with observation from 1923 to 1935 is no warrant for the value 

 of a synthesis from 1900 to 1923, for, they may say, the features of the 

 31 periods used were based in considerable part on the observations 

 of 1923 to 1935. Hence the synthesis within those years should agree 

 with those observations. But if one takes any three years, as July 

 1929 to July 1932, the observations of those years could have had 

 but 10 percent influence on the features of the 31 periods derived 

 from 30 years of observation. Yet the average difference between 

 synthesis and event in that interval is but 0.062 percent of the solar 

 constant. The coefficient of correlation between synthesis and event 

 during that interval is 90±5' percent. Is not this satisfactory indi- 

 cation that a synthesis of the 31 periodicities at any epoch is apt to 

 be nearly true to fact ? 



Holding this conviction, I offer figure 5 as probably a good repre- 

 sentation of the variation of the solar radiation for most of the first 

 two decades of the twentieth century. Of course, if there occurred 

 some other extraordinary variations of the sun, such as depressed 

 the solar-constant measures by 5 percent in 1922, there may have been 

 other changes outside of those represented in figure 5. 



In figure 4 the solar-constant values began with September 1923. 

 Counting back 273 months brings us to December 1900. All the pe- 

 riods being exact submultiples of 273 months, it follows that, so far as 

 these periods comprise all the elements of change in solar radia- 



