3'"0 SMITHSOXIAX MISCELLANEOUS COLLECTIONS VOL. 51 



The fact that meanwhile the same results have been confirmed by 

 observations made elsewhere, can only be incidentally mentioned, 

 since in this present summary, as already stated, for want of time 

 I must confine myself to the material submitted in this present 

 publication. 



When heretofore one refused to entertain the idea that such large 

 temperature gradients existed in the highest strata, the reason lay 

 in the consideration that the temperature could not diminish to 

 infinity. But we must not forget that the term temperature of the 

 air becomes less applicable in proportion to the increasing distance 

 from the earth, and that for the extremest rarefaction the ordi- 

 nary considerations must be replaced by an entirely different series 

 of ideas. 



The median portion of the curve shown in fig. 40, which is based 

 simply on the consideration of currents ascending and descending 

 adiabatically, already shows in its general course a certain agreement 

 with facts that will hereafter be more exactly described, but with 

 certain appreciable limitations. 



For instance, it follows from considerations based on the above- 

 given assumptions that the diminution of temperature with altitude 

 in the median atmosphere strata must be less than in adiabatically 

 ascending and descending dry air, and especially so in those layers 

 in which the condensation is most frequent and most considerable, 

 namely, between the altitudes 1000 and 4000 meters. This is in 

 fact the case qualitatively; but the diminution actually observed is 

 much smaller than would result from the above-described method of 

 formation of averages. 



For saturated air at temperatures between + 26 C. and — 30 C. 

 ascending adiabatically the temperature gradient at 1000 meters 

 altitude varies between — 0.37 and — 0.88 . Under the assump- 

 tion that such ascending currents interchange during equal intervals 

 of time with adiabatically descending currents whose gradient is 

 always — 0.99 there should result average gradients that lie 

 between — 0.68 and — 0.93. 



For adiabatically ascending air that leaves sea-level with the 

 temperature + io° C. and attains its dew-point at altitude 1000 

 meters, the gradient at this elevation is — 0.59 and the average 

 of this value and that of adiabatically descending air is — 0.79 

 or around — 0.80 C. per 100 meters. 



But for this altitude the observations give an average value of 

 — 0.50, or 0.58 if we exclude those cases in which large temperature 

 reversals were observed, or values that are far smaller than those 



