NO. 3 SOLAR RADIATION AND THE WEATHER CLAYTON 25 



These values show periods of about ^;^ days alternating with 

 periods of about 22 days during the first half of the year 191 5 and a 

 persistent period of 32 to 33 days during the second half. These 

 curious phenomena gave rise to numerous hypotheses and a great 

 deal of work was done in testing them with the hope of finding a 

 satisfactory solution. 



The most plausible hypothesis seems to be that the true period is 

 about 30 to 33 days, due to outbreaks of heated gases on the solar 

 surface in the region of the prominences. These gases produce an 

 excess of heat for a certain time after eruption, then becoming 

 cooled produce the reverse effect by absorbing the radiated heat of 

 the solar surface beneath them. This hypothesis is illustrated by the 

 curve a, b, c, d, etc., at the bottom of figure 6. The maximum a repre- 

 sents the heating effects of gases which, by the time of the next solar 

 rotation, have become so cooled as to produce a cooling effect. 



With a semi-rotation, the temperature rises to normal at c and is 

 followed by a still deeper minimum owing to further cooling, the 

 temperature again rising to normal at d and, if new outbursts begin, 

 it might easily rise above normal again at / and g and continue 

 emitting heat for several revolutions. With such conditions, the 

 maxima h and c would occur at intervals of 22 to 24 days followed 

 by an interval of 30 days between c and d and by two further intervals 

 of 22 to 24 days between d and f, and they by a succession of periods 

 of about 30 days. Something like this succession would explain the 

 correlation obtained during the early part of 191 5 ; but to verify such 

 an hypothesis, visual observation of the solar surface and the co- 

 operation of the solar physicist will be necessary. 



The results of this investigation strengthen the conclusion pre- 

 viously arrived at ; namely, that periodicities of different length are 

 caused by different rates of rotation of the solar surface which origi- 

 nate complex changes in solar radiation ; but in the eff"ect on terrestrial 

 temperature there appear to be two dominant periods, one corre- 

 sponding with a synodical equatorial rotation and another corre- 

 sponding with a rotation in the latiti!^(e of the solar prominences. 

 This conclusion is especially well illustrated by the correlations of 

 the first 26 days of January, 191 5, with the following days. The 

 correlation of the means of three and one-half days minus the means 

 of seven days is shown by curve p in fig-ure 6 and shows a very dis- 

 tinct maximum at about 26^ days, while the correlation for the means 

 of 14 days minus the means of 28 days shows by curve i a distinct 

 period of 33 days. In the first case the value of the maximum corre- 



