60 SMITHSONIAN MISCELLANEOUS COLLECTIONS VOL. I46 



formerly, and in a good location, and with a tower telescope, perhaps 

 Dr. Langley's hope might be at least partly realized. 



13. FINAL EVIDENCE 



Notwithstanding the evidences contained in the references cited 

 below some meteorologists may still be reluctant to accept forecasts 

 many years in advance. In the absence of conclusive theoretical 

 demonstration that the small percentage changes in solar radiation 

 can cause changes of identical periods of many times larger per- 

 centage in weather, and that these are hidden by phase changes from 

 direct disclosure, they may still withhold belief. Therefore I present 

 an additional observation which is so striking that some have con- 

 sidered it conclusive. 



If it is true that the 273-month family of regular harmonic periods 

 exists in weather, with such amplitudes that by their summation a 

 controlling influence is exerted, then it follows that the weather 

 should tend strongly to repeat its features at intervals of 22 years 

 9 months. I showed such a tendency in the precipitation of Peoria, 

 111., in 1934 by figure 33 of P. 3339, reproduced as figure 1 of 

 P. 4095, 1952. But now I will present a much more telling evidence 

 from the records of precipitation at Nashville, Tenn. 



Taking from our files the computations on Nashville prepared for 

 P. 4390 in 1958, I lengthened my forecast for Nashville through 

 1970. Considering only the 6 years 1965 through 1970, I looked 

 back 22 years and 9 months to the interval April 1942 to March 

 1948, 6 years. 



Figure 53 gives a graphical comparison of my forecast, from 1965 

 through 1970, with the observed precipitation at Nashville from 

 April 1942 through March 1948. The values plotted are, as stated 

 in P. 4390, smoothed by 3-month consecutive means and are depar- 

 tures from the normals given in table 9, P. 4390. I have computed 

 the correlation coefficient for the 6 years between the two curves of 

 figure 53, and also the correlation between the two curves of figure 2, 

 page 3, of P. 4390, for the 6 years 1950 through 1955, all from 

 Nashville precipitation. The two correlation coefficients are, respec- 

 tively, -f 0.469 ±0.061, and 4- 0.737 ±0.024. 



So the correlation coefficient between the direct forecast and the 

 event, 1950 through 1955, is 30 times its probable error, and the 

 correlation coefficient between the forecast, 1965 through 1970, and 

 the observed precipitation at Nashville, April 1942 through March 

 1948 (22 years 9 months previous) is 7.6 times its probable error. 



