1S75.] Variation of the Sim's Heat. 23 



have been obliged to restrict my comparison to those of consecutive years 

 that have been recorded with the same instrument, and wherever an instru- 

 ment has lasted over a single twelvemonth only or less, its register has 

 been totally set aside. 



The next precaution necessary is to eliminate as far as possible from 

 the individual registers, those irregularities which are due to variations in 

 the state of the sky. This, however, can be done but very imperfectly, 

 otherwise thau on the mean of a very large number of observations. It 

 results from the actinometric observations of Pouillet, Kamtz, Quetelet and 

 Althaus, that with a vertical sun, and a sky free from all visible cloud or 

 haze, the proportion of solar heat that penetrates the whole thickness of 

 the atmosphere, and is therefore effective at sea-level, does not amount to 

 more than two-thirds or at the utmost three- fourths of that which reaches 

 the exterior of our atmosphere. Herschell estimates it at the former quan- 

 tity. But in India, the atmosphere, when cloudless to the eye, is by no 

 means so diathermanous as is here assumed. Sometimes for many days 

 together, with settled weather and a cloudless sky, the sun thermometer 

 gives steady maximum readings, not differing more than one or two degrees. 

 A day follows on which there is a good deal of cloud, and perhaps some rain, 

 and the diathermancy of the atmosphere is so increased in the intervals of 

 the clouds, that the sun- thermometer registers 10° or 15" above any of its 

 previous readings. Such cases occur frequently in all the registers. It is 

 probable therefore that on days registered as cloudless, not less than half 

 the solar radiation and frequently much more is absorbed by the atmo- 

 sphere. In order to obtain data that shall be fairly comparable, I have in 

 most cases selected those daj 7 s on which the sky was either cloudless at 10 

 A. M. and 4 p. M., or had on the average not more than one-fifth of cloud. 

 In the case of the two comparatively cloudy stations Silchar and Port 

 Blair, I have been obliged to extend these limits ; in the former case to three 

 tenths, in the latter to one half. The monsoon months, June to September, 

 are omitted in these tables. 



Another method of proceeding which I have adopted in order to verify 

 these results is to take the two highest readings recorded in each month 

 (including the monsoon months) as the data for comparison. 



The four following tables give the results. In Tables I. and II. the 

 comparison is restricted to the registers of those stations and years in 

 which the same instrument has been read continuously for at least two 

 consecutive calendar years. The differences of each pair of years are given 

 separately for each station, and the means of the whole. This method of 

 comparison, however, admits of a very small portion only of the data being 

 utilized, since it excludes all broken years, and therefore in Tables III. and 

 IV. I have adopted a modified course of proceeding, which admits these. 



