2 SMITHSONIAN MISCELLANEOUS COLLECTIONS VOL. 128 



3. To those who hold that the observed fluctuations in the earth's 

 supply of radiation from the sun, which I claim to be periodic, are too 

 small in percentage to affect weather appreciably, I remark that it 

 makes no difference how the family of periods was discovered, or 

 whether the effects that I ascribe to solar variation are really of solar 

 causation. These 2j periods exist in temperature and precipitation, 

 however they may be produced. But in Part 2 I shall advance sug- 

 gestions tending to explain the large weather effects from small solar 

 changes. 



4. The forms and amplitudes of these periods in precipitation and 

 temperature are determined by tabulations of monthly records, pub- 

 lished in "World Weather Records." Over 1,000 months of records, 

 ending with December 1939, are used in all tabulations. They fix 

 mean forms and amplitudes of all the periods, applicable at all times. 



5. Every precipitation or temperature value forecasted is obtained 

 by a synthesis of about 23 terms. Each term depends on more than 

 1,000 monthly records. In the study of St. Louis temperature and 

 precipitation, 1,032 months, of the years 1854 to 1939, were employed 

 to establish the basis, which, accordingly, centers on the year 1897. 



6. A determination by such a synthesis of the march of precipita- 

 tion at St. Louis for any selected year is of the nature of a prediction, 

 backward or forward, from 1897. Even if the selected year falls be- 

 tween 1854 and 1939, its own observed monthly mean precipitation or 

 temperature, though used in computing the basis of the forecast, can 



12 



have but = li- percent influence on the forecast. 



1,032 ^ 



7. Knowing that any curve, even the profile of a girl's face, can be 

 fairly represented by a Fourier's series of sufficient number of terms, 

 some critics, at first sight, may think it nothing remarkable that, with 

 23 terms, the march of precipitation or temperature can be well repre- 

 sented. But the proposition (6) has nothing in common with such a 

 Fourier series. The girl's profile would be built by a Fourier's series 

 solely on measurements taken within that profile. No satisfactory 

 representation of it could be made by a Fourier's series made from 

 measures on over 1,000 other girls. In my case 1,032 monthly 

 records govern the synthesis for each and all years, and over 1,000 of 

 them lie outside any selected year to be forecasted. 



8. Of 100 years of St. Louis precipitation forecasted, 70 seem fairly 

 satisfactory and yield high correlation coefficients with the events. 

 The failure of the other 30 is reasonably explained. 



9. As shown by Dr. W. J. Humphreys in his "Physics of the Air," 

 figure 227, great volcanic eruptions, which throw high columns of 



