DISTRIBUTION OF ATMOSPHERIC PRESSURE VON BEZOLD 371 



less than in the rear and correspondingly the gradient-acceleration 

 is greater than would correspond to the average gradient." 



However, even for equidistant successive isobars in the front of a 

 depression one must expect greater accelerations and correspond- 

 ingly greater .wind velocities than in the rear. 



Hence the isobaric charts directly allow a conclusion as to the 

 gradients in that according to formula (ic) these are always propor- 

 tional to the reciprocal of the distance between neighboring isobars, 

 but not any conclusion, or at least only a crude approximation, as 

 to the gradient-acceleration, which still depends to a large degree on 

 the density of the air. 



Therefore the isobars can in no wise be compared with altitude 

 lines or isohypsen. For whilst we can from the reciprocal of the 

 horizontal distance of the isohypsen conclude directly as to the 

 gravity gradient, i. e., the tangent of the angle of inclination, and 

 thence as to the acceleration which a heavy point experiences when 

 it moves without friction on the given surface, we cannot do this 

 from the isobars. Such a conclusion would only be allowable when 

 the density of the air is uniform over the whole area under consider- 



• • 1 T ■ 



ation, 1. e., when — is constant. 



P 



But even in summer, when it is generally cooler at the base of the 

 cyclone than in the anticyclone, this condition is only seldom satis- 

 fied. For instance, in a region of maximum pressure 775 mm and 

 another of minimum 745 mm the temperatures must be respectively 

 2 7 and i6°C. if the densities are to be the same in both. But more 

 than this there is also the condition, which is almost never satisfied, 

 that the temperature be constant along each isobar. 



If the temperature were uniform over the whole region covered 

 by a chart of isobars, then certainly we would be in a position to 

 draw a system of lines whose reciprocal distance would certainly be 

 proportional to the gradient-acceleration. The formula (id) can 

 be written 



dB T 



r = J-di- Rg 



or 



From this formula by passing to differences we can deduce the fol- 

 lowing: 



A log 8 



