METEOROLOGY. 519 



greater than in the former. Therefore, also, «D will be larger for high 

 than for low pressures ; that is to say, a must be greater for high press- 

 ures than for low. 



Br 



But from this it follows that — =^— ^ is -smaller for high and greater 



CI -LJ fJio 



for low pressures; that therefore A increases with altitude above the 

 ocean surface. And this is precisely what my observations give for the 

 summit of the Obir. 

 According to the preceding, it is to be expected that a stands in a 



simple direct ratio to the air-pressure, so that if «i is its value for 



p 

 760 mni , then for any other pressure a = „ - a-\. 



For the investigation of this dependence there are some experiments 

 at my disposal. Chistoni, Angot, and Blanford (see the previous refer- 

 ences) have submitted a largo quantity of material for high pressures.* 

 I have taken the mean of their determinations, and find for A the fol- 

 lowing values : 



Blanford 0.000827 

 Angot .000851 



Chistoni .000851 



Mean 0.000843 

 Since these comparisons, especially those of Blanford, were made 

 with large differences (t—t'),I have not considered it necessary to apply 

 the correction 



v 

 t-t'+l 

 Since now [the preceding formula and reasonings give] 



A=0.000630+°^ ) ^ 

 a 



where I have assumed for Rr a mean value of 0.000055 [which, as before 



shown, is especially applicable to the thermometers ot the Austrian 



observers], it follows that for the above three series of observations 



^ =3.0. 



T * a = 0.000630 "I 



L' ' ai ~ 0.0008 13 -0.000030 J 

 My own comparisons on the summit of the Obir, after applying the 



correction % — - give A=0.000943, whence a 2 =2.0. If a' is smaller 



t— V +1 fe ' 



in the ratio of the diminution of pressure, then should this latter have 



given 02=3.0^=2.3 and A would have been found 0.000901 instead 

 (00 



of 0.000913. 



* Since Regnault's time the factor 0.000800 has been generally retained. Still, all 

 later investigators have from their comparisons deduced even larger factors. Only 

 Sworikin finds 0.0t072"> and Mac6 de Lepinay with a swing-thermometer (tnermo- 

 metre & fronde) even 0.000G93. This latter determiuation seems to differ too far from 

 t?he others, especially the long series of Chistoni and Angot, for me to consider it quite 

 reliable. 



