6 
ASTRONOMY: C. G. ABBOT 
Proc. N. a. S. 
when the air masses were 2 and 3, respectively, that is to say, when the 
zenith distance of the sun is about 60° and about 70°. 
The function p/psc is determined in this way: A spectrum-energy 
curve taken through the solar spectrum shows deep depressions in the 
infra-red spectrum, one of which was termed by Langley par. Measure- 
ment is made of the ordinate of the smooth curve drawn across the top 
of the depression, and another measurement of the ordinate at the minimum 
of the depression. The ratio of these two ordinates forms the function 
p/psc It is a measure, as Mr. Fowle has shown, of the quantity of hu- 
midity prevailing in the atmosphere between the observer and the sun. 
Mr. Fowle indeed has worked out a method of determining the actual 
amount of precipitable water in the vertical column of the atmosphere 
by a mere measurement of the value p/ p^c- 
Since the haziness of the atmosphere is closely related to its humidity, 
but since the haziness also depends upon the amount of dust taken up 
from the soil by the wind, or such as is occasionally distributed in the 
upper air by volcanic eruptions, we hoped to get a complete indicator of 
the haziness prevailing by a combination of the value p/psc with the value 
of the sky brightness, determined by the pyranometer, in the immediate 
neighborhood of the sun. We formed the function "sky brightness di- 
vided by p/psa' which we will call "F." After a number of unsuccessful 
investigations I was led by a suggestion of Mr. Moore's to find that the 
values of the atmospheric transmission coefficients for the individual wave- 
lengths in the spectrum could be represented in terms of the function 
"F" by smooth curves in which atmospheric transmission coefficients 
were plotted as ordinates and the function "F"- as abscissae. All of the 
data for the sixty days available were plotted in this manner for each of 
the 40 wave-lengths above mentioned. The atmospheric transmission 
coefficient varies by only 2 or 3, or at most 5% in any part of the spectrum, 
while the function "F" varies through a range of from 100 to 1000 or 
more on a certain scale. Hence, it is clear that great errors in determining 
the function "F" would produce very small errors in determining the 
atmospheric transmission. 
The measurements required to determine the function "F" may be all 
completed within ten minutes. Entering the plots above mentioned with 
the values of the function "F" as thus determined, we take out from the 
plots the atmospheric transmission coefficients for all wave-lengths in the 
spectrum. With these data we may immediately determine the solar 
constant of radiation by the usual method of computation. 
The new method has been in application at Calama, Chile, since about 
July 1. Measurements by the older method, on which it is primarily 
based, have been taken on about one-third of the days, and measurements 
by the new method on practically all days since July 1. It is found that 
the agreement between the results of the older method and of the new 
