2 SMITHSONIAN MISCELLANEOUS COLLECTIONS VOL. 134 
What is important for the proof of variation is the accidental error 
of Smithsonian daily observations of the solar constant. There are 
several determinations of it. From volume 6 of the Annals of the 
Smithsonian Observatory, page 163, I quote the differences, in thou- 
sandths of a calorie, between daily values of the solar constant ob- 
tained at Smithsonian stations in the Northern and the Southern 
Hemispheres, respectively. This comparison covered all days observed 
in both hemispheres for the years 1932 to 1936, numbering 616. 
Taste 1.—Numbers of daily differences in solar-constant measures, 1932 to 1936, 
having different amplitudes. 
Amplitudes in thousandths of a calorie. 
Amplitudes .... 22-28 20-22 18-20 16-18 15 14 12 vi 10 
No. of days.... 17 12 10 35 13 15 20 22 2 
PEOGUCE os oneness 201 .262. 100 505. 105 210 240 242..270 
Amplitudes .... 9 8 Gi 6 5 4 3 2 I te) 
Ne Of idays oo. 3d 30. S85 AB BE 55. 550,437 nity Sena 
Prodgct a. a 300 240 -245 (258.5255 .220 , 105 74 48 fe) 
Total days 616. Total of products 4,682. 
Weighted mean daily difference, 7.60 thousandths of a calorie. 
Mrs. Hill has made for me similar tables for more recent differences 
between daily solar-constant results, where measures at Montezuma, 
Chile, are compared to those of Table Mountain and Tyrone in the 
United States, and those at Table Mountain, Calif., to those at Tyrone, 
N. Mex. For these three cases the weighted mean differences are 7.68, 
7.96, and 7.79 thousandths of a calorie. These results cover all days 
measured at both stations from 1940 to 1952. They number, respec- 
tively, 891, 283, and 202 days. The weighted mean difference is 7.75 
thousandths calorie. As the results of these recent years differ but 
slightly from those of 1932 to 1936 shown in table 1, we may adopt 
7.7 thousandths of a calorie as the weighted mean average daily acci- 
dental difference between results of widely separated stations observ- 
ing the solar constant. Assuming the stations to be of equal accuracy, 
that gives for the percentage accidental error of a solar-constant 
measure of one day at one station: 
100 X 0.00385 Xx 0.84 + 1.946=0.166, or % percent of the solar constant. 
I use 10-day and monthly solar-constant values in my investigations. 
For these the percentage probable error (if all days of these intervals 
were observed) becomes %+ V10 and %+ 30, or Mo and %s per- 
cent of the solar constant. 
