NO. 4 SUN S VARIATION AND TEMPERATURES — ABBOT I5 



2. How are its phases related to January 3.0000, 1924? 



3. Are its amplitudes variable, and, if so, have their variations a 

 correlation with variations of amplitudes in temperature effects? 



4. Are the displacements of phases in periodic curves of tempera- 

 ture departures, above referred to, found to be duplicated in direct 

 solar measurements? 



Before giving the results, a few remarks are due regarding the 

 accuracy of the solar-constant observations, in relation to the ampli- 

 tude of the expected solar variations. In my papers of 1936, I esti- 

 mated the average amplitude of the short-interval rises and falls of 

 solar radiation as about 0.7 percent. Only the larger variations could 

 possibly be discerned individually there, owing to accidental errors 

 and incompleteness of the solar-constant record. Furthermore, when- 

 ever accidental errors occurred of a sign tending to increase the ampli- 

 tudes of the apparent solar changes, they of course made the changes 

 more conspicuous, and the inclusion of these spuriously excessive 

 changes tended to raise my estimate of the average amplitude of 

 variation. Hence the estimate, 0.7 percent, is far above what must 

 now be expected, when every possible occasion of solar change is 

 included. It must be realized that all recurrences of the cycle have 

 been computed in the present research, and the 310 cases included in 

 the mean are in no way selected for large or small amplitudes of 

 solar variation. 



As shown at page 183, volume 6, Annals of the Astrophysical Ob- 

 servatory, the probable error of a solar-constant value, determined 

 from a single day of observation at one station, is 0.16 percent. Hence 

 if the average change to be expected is of the order of 0.2 percent, a 

 single month of my tabulations for the 21 years, containing only about 

 30 observed values in a column (neglecting those interpolated for sym- 

 metry's sake) cannot be expected to give a fair representation of the 

 true form of the solar variation curve. For some of its points will 

 probably be too low by 0.09 percent,'' and others too high by this 

 amount. Hence i month alone will be very unlikely to give a con- 

 vincing answer to our first query. But if we combine the average 

 results of all 9 available months, there will be the mean of nearly 300 

 observed values for each point of the mean curve, and these means 

 should be three times more conclusive than results of one month only. 



Remarks are also in order regarding queries numbered 3 and 4. 

 Regarding query 3, as we have seen in figure 3, there is a great dif- 



3 X 0.16 



^ Thus : =^—=0.09. Errors three times the probable error, as is well 



. V30 



known, occasionally occur. 



