Variation of the Solar Radiation. 83 
connexion the reader’s attention is invited to the values of 
August, October, and December, 1903. 
Fifth, as the great bands of the infra-red are presumed to 
be solely atmospheric, they should disappear in the energy- 
curve outside the atmosphere. Accordingly the same result 
should be reached by the extrapolation in these regions, 
whether we determine a transmission coefficient at the bottom 
of the bands themselves, or from the smoothed curve above 
them. It appears from studies of Mr. Fowle, of the Astro- 
physical Observatory, that such is the case, except for the very 
deepest bands, where the determination is subject to many 
kinds of error all known to tend in the same direction. 
The coincidence of al! these kinds of evidence does not, 
however, in the writer’s belief, give absolute assurance that 
the values of the solar radiation may not be underestimated, 
a defect inherent in the method of observation at a single 
station. 
We cannot go outside the atmosphere and prove that our 
determinations of its absorption are correct, but at all events 
the results are the best that can be reached at present from 
our single and low station. We may next consider their 
numerical values and what they appear to show. 
TT: 
Ezperimental Results. 
(a) Atmospheric Transmission.—In the following Table are 
given the atmospheric transmission coefficients for rays of 
certain wave-lengths for separate days of 1903-4, and the 
mean of numerous similar observations of 1901-2. It is to 
be remembered that these determinations are made only on 
the most cloudless days, so that they represent a trans- 
parency probably above the normal, though they donot include 
data of days when cumulus clouds were frequent, such as 
are often associated with a very deep blue intervening sky. 
Attention has already been drawn by the writer* to the 
diminished transparency during the early months of 1903 as 
compared with 1901-2. The reader will now observe that 
this condition appears not to have continued so marked during 
the last three months of 1903, although the transparency did 
not recover fully its maximum value of 1901-2. 
It seems probable that such long-continued alterations of 
atmospheric transparency, if general over considerable areas 
of the earth’s surface, might directly influence observed 
* ‘Nature,’ November 5, 1903. 
G2 
