NO. 10 SOLAR RADIATION AND WEATHER STUDIES ABBOT 4I 



computed from curves i, 3, 4, 6, 8 and 9, is as follows and is illustrated 

 at A of figure 16. The mean values which follow are expressed as 

 usual in units of o.°i Centigrade/"" 



1.3 2-3 2-8 5-4 5-2 4.6 3.4 2.3 0.6 —0.3 —1.8 



—2.8 —1.9 —0.2 1.5 3.4 5.1 3.8 2.1 1.2 0.4 



The range of the general mean is o.°8 Centigrade. This mean curve 

 represents the tabulation of 39 lines in each of its 21 columns or 819 

 months in all. The contradictory results found in the remaining 24 

 lines, representing 504 months, themselves somewhat approach a 

 common type. Its mean form, shown at B of figure 16 is as follows: 



—4.4 -4-2 -4-4 —5-1 —5-0 — S-2 —2.8 —0.7 0.6 0.2 o.g 



0.8 —0.5 1.6 2.3 1.9 2.6 0.5 —1.3 —2.9 —4.3 



Being plainly associated with periodic changes in the sun, as the dates 

 of the appearance and disappearance of contrasting phases in these 

 curves 2, 5, 7, and 10 appear to be, the existence of these curves of 

 a contradictory type does not, in my judgment, reasonably require 

 us to doubt the evidence of the other 70 years or of their own 40 years 

 that 21 months is a veridical period in terrestrial temperature, pro- 

 duced by a periodic solar variation. 



c. Progressive Removal of Determined Periodicities 



Acting on the conclusion just expressed, I have felt it justifiable 

 to remove, one after another, the mean evaluations of the various peri- 

 odicities, and to remove them in parcels of 11^ or 2^ years at a time, 

 so as to eliminate them to the highest degree possible despite changes 

 of both phase and amplitude. As I am aware that this course will 

 be criticized and perhaps disowned by meteorologists and statisticians, 

 I pause at this point to refer to the 12-month periodicity, as computed 

 from the residuals of the 5-month smoothed Berlin temperatures, 

 after removing in the way just indicated, and in the following order, 

 the 7, 8, II, 13.6, 21, 25, 34, 46, 68, and 9I month periodicities. Mean 

 values for each ii| years from 1819 to 1929 are given by the dotted 

 lines in figure 15. These results may be compared with the closely 

 juxtaposed curves for the 12-month periodicity, as previously com- 

 puted directly from the original data, and already referred to under 

 14-A. The very great similarity in general between the two sets of 

 curves indicates that the removal of all of those many periodicities in 

 ii|-year or 23-year parcels has not ruined the residuals for the purpose 

 of the 12-month analysis. Figure 15 also includes a similar pair of 

 juxtaposed analyses of 12-month periodicities for Copenhagen, and 



^"''The lines of these two tables (too long for page width) are to be read 

 consecutively like two lines of text, not staggered as might be thought. 



