PRESENT STATE OF OUR KNOWLEDGE OF THE UPPER ATMOSPHERE. 91 
Bosch tube thermometer is free from this defect, but has other faults, 
pointed out by Schmidt and Gold, which prevent the possibility of its 
giving quite accurate results.' The possible calibration errors of the 
Dines instrument are somewhat larger, as the scale is microscopic, the 
calibration curves of the barograph are not linear, and the temperature 
correction is not a simple function. 
The average error found from independent calibrations is 0°1 C. for 
Continental instruments and 0°°8 C. for the Dines instrument at the lowest 
temperatures. For higher temperatures the possible error is of the same 
order of magnitude. The principal error in the temperature traces probably 
arises purely from faults of construction in the instrument, such as in- 
sufficient rigidity and sticking of the pens. The best way of eliminating 
these defects is probably to increase the rigidity by simplifying the con- 
struction and diminishing as far as possible the number of levers, joints, 
and bearings, so that the working parts operate as directly as possible on 
the recording pens. 
The heights indicated by the barometers have many times been com- 
pared with the heights calculated from trigonometrical observations, and 
the maximum heights have been found correct to +2 per cent. up to 
10km.? The intermediate heights show greater errors with some 
instruments, owing to lag. An example given by Teisserenc de Bort 
showed large differences *, e.g. :— 
7 ‘ | 
Ascent. 
| 
Height Difference kms. between Barometric Wee | 
and Trigonometric Heights | 
pone | 
Km. Km, Km. 
: —0°6 | 
8 0-4 —(°4 
Max. height 0.1 01 | 
This instrument was specially chosen to show the effect. 
The average error in pressure of the Continental instruments amounts 
to 5 mm. of mercury and that of the Dines instrument to8mm. The 
errors to which these give rise in the final results are as follows :— 
Heights, km. 5 10 15 20 | 
Difference corresponding to : Km. Km. Km. Km. | 
5 mms. 01 0-2 0-4 0°6 
0-2 03 05 1:0 
8 mms. | 
| 
The errors introduced in working up the traces are very small in 
Continental instruments owing to their large scale, and for the Dines 
instrument are of the same order as those occurring in the calibration. 
The principal error in the final results arises from the uncertainty of the 
temperature correction of the barometer. Teisserenc de Bort estimates 
the correction due to this cause to be 0:5 km. at 14 km., and 2 km. or 
even more at 20km. Hergesell, however, claims to have shown that 
1 Quart. Journ. Roy. Met. Soc., 1909, Met. Zeit., 1909. 
2 L. T. de Bort, and Etude de lV Atmosphére, Observatoire Constantin, Fascicule ii. 
p. 76. 
® Compt. rend., July 10, 1905, p. 153. 
