30 SMITHSONIAN MISCELLANEOUS COLLECTIONS VOL. 65 



different portions is too inconvenient and fails when the radiation is 

 rapidly changing. 



On the other hand, the value k is here dependent on the accuracy 

 with which the radiation constant a is determined. Further, since 

 the emissive power of the strips, which is different for different wave 

 lengths, enters into the constant k, this constant can be applied only 

 for cases where the radiation is approximately of the same wave 

 length as in the experiment from which k is computed. In the night- 

 time this may be considered the case, the emissive power being the 

 same for all heat waves longer than about 2 p. But the instrument 

 cannot, without further adjustment, be used for determining the 

 radiation during the day, when the diffused radiation from the sky 

 of short wave length enters as an important factor. 



The constants of my three instruments, of which No. 17 and No. 18 

 were used al I'assour and California, and No. 22 in California, have 

 been determined at the Physical Institute of Upsala on two occasions, 

 before the expeditions by Dr. Lindholm of that Institute and after 

 the expeditions by myself. The two determinations of the constants 

 differ from one another only within the limits of probable error. 



No. Before After Mean 



17 10.4 IO.4 IO.4 



18 I I.I 10.7 IO.9 

 22 1 1.6 1 1.8 11.7 



For the computations from the Algeria values the first values of the 

 constants (for 17 and 18) have been used, for the California observa- 

 tions a mean value between them both. For the determination of the 

 constants, Kurlbaum's value for o- has been used 



cr = 7.68- 10- 11 



not so much because this value is at present the most probable per- 

 haps, as in order that observations with these instruments may be. 

 directly comparable with those of older ones. At any rate the rela- 

 tive values of the radiation must still be looked upon as the most 

 important question. 



The galvanometers that I have used were of the d'Arsonval type. 

 They were perfectly aperiodic, and bad a resistance of about 25 O and 

 a sensitiveness of about 2 ■ io~ 8 amp. per mm. at meter distance. They 

 generally showed a deflection of between 30 and 70 mm., when the 

 strips were exposed to a clear sky. The galvanometers and the 

 p\ 1 geometers were made by G. Rose, Upsala. 



In the use of the compensation instrument one has to be careful 

 that the instrument has had time to take the temperature of the 



