NO. 3 SOLAR VARIATION AND WEATHER — ABBOT 57 



and 17 for days preceding the beginning of sequences of observed rising 

 solar radiation. The other type comprises 17 values for days following and 

 17 values for days preceding the beginning of sequences of observed falling 

 solar radiation. Two correlation coefficients are to be computed, one including 

 the 204 values of the two contrasted types following the supposed critical 

 dates, the other for the 204 values of the two types preceding them. 



In order to avoid diluting the correlations by including extraneous influences 

 due to previous conditions, each line was first reduced to the level of zero 

 temperature departure, by adding to all 17 values in that line a constant quan- 

 tity such as to make the average temperature departure for that line zero. 



Having thus arranged the values, correlation coefficients were computed 

 between the two types for the two divisions. They resulted as follows : 



After appearance of solar change, r = — 54.3 ± 4.9 per cent, which is sig- 

 nificant. 



Before appearance of solar change, r:=ll.l ±6.0 per cent, which is mean- 

 ingless. 



The inference is obvious that the 320 dates, above described, were dates of 

 real significance, since no other consideration was used in selecting them, and 

 it is difficult to avoid the conclusion that they were dates when real solar 

 changes began. 



In vi^hat follows I shall show that the ionospheric Fe and the 

 areas of calcium flocculi are both as effective as are solar constants 

 for this observation. (See figs. 45, 46, 47.) 



Graphs from other publications will show plainly how important 

 this matter is. My latest paper on the subject, P, 4462 (see figs, 48, 

 49), shows how a modern satellite could obtain first-rate solar con- 

 stant values daily, and the results radioed to earth would give means 

 to predict temperature changes all over the world for 16 days in ad- 

 vance. Furthermore, if such satellite continued indefinitely to circle 

 the earth we would have means to discover great changes in the solar 

 radiation like that of 1922 and 1923 if they occur. This might lead 

 in course of a century or more to discovery of long-period variations 

 of solar radiation of high importance, 



12. DRIFT OBSERVATIONS 



As stated in the introduction, and illustrated in plate 1, our Mount 

 Wilson observing station was equipped with a tower telescope. The 

 solar image, about 8 inches in diameter, fell upon the slit of the 

 spectrobolometer in such a way that when the telescope clock was 

 stopped the solar image drifted centrally over the slit. The spectro- 

 bolometer was set for any desired wavelength, and the intensity of 

 that wavelength along the sun's central diameter was recorded as the 

 solar image drifted. In 1908 we began to make these drift records 



