NO. 7 MONTEZUMA SOLAR-CONSTANT VALUES ABBOT 7 



Nevertheless, I began this present research without assuming a 

 273-month master period. First of all I removed the yearly varia- 

 tion from the values in column 2, table i, as noted above. I then 

 plotted the residual values and found that by far the most prominent 

 periodic variation displayed in a large-scale plot of the residuals was 

 of about 39 months. Seeking to fix its length as accurately as pos- 

 sible, by careful inspection of the large-scale plot, I finally decided 

 on 39-V months. I am not sure that the period may not be 39 months, 

 which is exactly 1/7 of 273 months; for the presence in the data of 

 many other periodicities, and of accidental errors of observation, 

 makes fixing of the exact length of a long period doubtful. Never- 

 theless, a table was prepared of seven columns, alternately of 39 and 

 of 40 months in length. The mean of these columns is plotted in 

 figure 3, c. As the reader will see, the march of this 39^-month peri- 

 odicity is nearly a regular sine curve, and its amplitude is 0.0069 

 calorie, more than one-third of i percent of the solar constant. 



The 39^-month periodicity was removed by subtraction to give a 

 second list of monthly residuals. These also were plotted on a very 

 large scale. There showed then a periodicity of considerable ampli- 

 tude, approximately 91 months in length. A table 91 months long of 

 three columns was made from the second residuals. With so few 

 columns entering into the mean it seemed best to smooth the mean 

 values by 5-month running means of them. The smoothed values 

 being plotted, the 91 -month periodicity appeared plainly, but super- 

 posed thereon there appeared a period of ^ of 91 months. As it 

 would be preferable to determine this curve of about 15 months by 

 itself at a later stage, a smooth curve was drawn of 91 -months period, 

 cutting symmetrically through the 15-month superposed excrescences. 

 The 91 -month periodicity had the amplitude 0.0054 calorie. It is 

 not of sine form, but rises rapidly to maximum, and falls slowly to 

 minimum, like the well-known sunspot frequency curve of 11 years. 

 This 9 1 -month periodicity was removed from the data, leaving a 

 third list of residuals, which were plotted on a large scale. 



The third list, when plotted, showed clearly a strong periodic fluc- 

 tuation of about 68 months. This was determined by forming a table 

 of four columns, taking their mean, smoothing it by 5-month run- 

 ning means, and plotting the smoothed means in a curve given in 

 figure 3, h. Very clearly there is a period of 1/7 of 68 months super- 

 posed on the principal curve. Not wishing to evaluate a 9|-month 

 periodicity until a later stage, I drew a smoothed curve as shown in 

 figure 3, h. It is nearly of sine form, and has an amplitude of 0.0053 

 calorie, slightly under one-third of i percent of the solar constant. 



I 



