NO. 3 RADIATION OF THE ATMOSPHERE ANGSTROM 3 1 



surroundings before measurements are made. If the instrument is 

 brought from a room out into the open air, one can be perfectly safe 

 after ten minutes exposure. When measurements are made on the 

 tops of mountains or at other places where the wind is liable to be 

 strong, I have found it advantageous to place the galvanometer as 

 near the ground as possible. By reading in a reclining posture one 

 can very well employ the instrument box itself for the galvanometer 

 support. Some heavy stones placed upon, at the sides, and at the 

 back of the box will keep the whole arrangement as steady as in 

 a good laboratory, even when the wind is blowing hard. 



For the measurements of the current used for compensation 

 milliammeters from Siemens and Halske were employed. 



The measurements of the humidity, as well as of the temperature, 

 were carried out with aid of sling psychrometers made by Green 

 of Brooklyn. The thermometers were tested for zero, and agreed 

 perfectly with one another. 



In order to compute the humidity from the readings of the wet 

 and dry thermometers I have used the tables given by Fowle in the 

 Fifth Revised Edition of the " Smithsonian Physical Tables " 1910.' 



B. ERRORS 



The systematic error to which the constants of all the electric 

 pyrgeometers are subject has already been discussed. There are 

 however some sources of accidental errors in the observations, and 

 I shall mention them briefly. The observer at the galvanometer will 

 sometimes find — especially if there are strong and sudden wind gusts 

 blowing upon the instrument — that the galvanometer does not keep 

 quite steady at zero, but swings out from the zero position, to which 

 it has been brought by compensation, and returns to it after some 

 seconds. The reason for this is probably that the two strips are 

 not quite at the temperature of the surroundings. From measure- 

 ments on the reflection of gold, it appears that the bright strip must 

 radiate about 3 per cent of the radiation of a black body, consequently 

 it will remain at a temperature slightly lower than that of the sur- 

 roundings, which will sometimes cause a slight disturbance due to 

 convection, the convection being not perfectly equal for the two strips. 

 Another cause of the same effect is the fact that the strips are covered 



1 These tables are calculated from the formula 



p--pi — o.ooo665 (t — L) (1 + 0.00115k) 

 (Ferrel, Annual Report, U. S. Chief Signal Officer, 1886, App., 24). 



