24 SMITHSONIAN MISCELLANEOUS COLLECTIONS VOL. I46 



Turning to Category 1, Division 2, I first remove 273/18, then 

 273/45. After removing these overriding submultiple harmonics, 

 273/18, 273/45, there is left only the nearly smooth sine curve 

 273/9. The amplitude of this final curve for Category 1, Division 2, 

 is 34 percent of normal rainfall at Buenos Aires. This percentage is 

 160 times the percentage of the normal solar-constant radiation given 

 as 0.21 percent in P. 4213 (here fig. 10) for the period 30-1/3 

 months. 



In further emphasis on the existence of subharmonics preventing 

 direct recognition of the longer periods, while disclosing shorter ones 

 otherwise hidden, I refer to eight telling cases : P. 4103, p. 4 ; P. 4352, 

 pp. 6, 8, 11;P. 4390, p. 12. 



7. LONG-RANGE FORECASTING OF WEATHER 



As yet theory fails to explain why the family of harmonic periods 

 related to 273 months, which exists in the sun's variation of radia- 

 tion, should exist also with identical periods in precipitation and 

 temperature on earth. Still more surprising, the weather amplitudes 

 have frequently a hundred times the percentage apparent importance 

 of the identical periods present in solar-constant measures. But a 

 huge body of measurements accumulated over 40 years, including 

 detailed study of weather in 54 cities, proves that it is so. The periods 

 in weather, to be sure, are hidden from casual view because though 

 invariable in solar radiation (P. 3893, p. 22, figs. 11), their phases 

 vary with place, time of the year, prevalence of sunspots, and growth 

 of population. All these things alter atmospheric transparency and 

 thereby shift times of terrestrial response to solar causes. But when 

 long-time weather records are grouped with attention to eliminating 

 as far as possible all these disturbing influences, and are cleared of 

 overriding shorter harmonic periods, the identical periods observed 

 in solar-constant measures stand out clearly in weather records all 

 over the world as sine curves. 



In view of all these facts, may it not be possible to forecast rain- 

 fall and temperature by adding together the amplitudes of these 

 weather periods ? This would require that the forms and amplitudes 

 which the periods presented in past years, and the disturbances of 

 their phases due to the causes mentioned above, would all remain 

 sufficiently constant. 



We have made at 54 stations tests of such long-range weather fore- 

 casting. On the whole the results have been very encouraging. Ex- 



