NO. 3 SOLAR VARIATION AND WEATHER — ABBOT 2$ 



amples will show both excellent successes and disappointing partial 

 failures. 



About 10 years ago, working quite alone from the beginning with 

 "World Weather Records," and computing myself every succeeding 

 step, I began the study of the precipitation at St. Louis, Mo. This 

 research occupied me for over three years and resulted encouragingly 

 beyond expectation (see P. 4211). It was early discovered that the 

 normal monthly values, customarily published with long monthly 

 weather records, must be recalculated in two parts, one for low, the 

 other for high sunspot numbers (P. 4090). Sunspot activity makes 

 a very important difference, both in the magnitudes and in the yearly 

 run of monthly means. This obtains both for precipitation and tem- 

 perature. After inspecting the curves of Wolf sunspot numbers, 

 it seemed best to set the boundary between low and high sunspots 

 at 20 Wolf numbers. We have thus far studied 54 stations, using 

 "World Weather Records" (now published up to 1950). These rec- 

 ords that we have used have been electronically recomputed by 

 Jonathan Wexler to give normal monthly values for sunspots ^20 

 Wolf numbers. The observed monthly values are all reduced to per- 

 centages of these new sets of normals. 



In order to eliminate as far as practicable the shifting of phases 

 attending time of the year, periods less than 15-1/6 months are 

 separated into three groups. These are : January- April ; May- 

 August ; September-December. There are two series of these groups, 

 one for sunspots less than 20 Wolf numbers, and the other for sun- 

 spots above 20 Wolf numbers. As we see, exaggerated conditions as 

 "smog" in Los Angeles County, Calif., the multitudes of automobiles, 

 airplanes, factories, and forest fires which have attended the increase 

 of population have greatly altered atmospheric transparency and 

 thereby have shifted phases. We imperfectly allow for this by dividing 

 the whole interval of records, say 1870-1950, into halves. 



So, up to the period 15-1/6 months, consideration of the (5 parts 

 of the year) X (2 ranges of Wolf numbers) X (2 halves of the record 

 interval) yields 12 groups of records for each period. Dropping the 



3 divisions of the year, from 18-1/5 to 45-1/2 months there are 



4 groups for each period. For the last 3 periods, 54-3/5, 68-1/4, and 

 91 months, we drop the halving of the time interval, and retain only 

 separation of ^20 Wolf numbers. Altogether we separate the rec- 

 ord into (12xl5)-F(4x9)-|-(3x2)=222 groups. This separation 

 would restrict many groups to too few members to deserve confidence. 

 So we make the assumption that, though different individually in 



