1904] Circulation of the Atmosphere. 29 



region of the clouds of intermediate height and probably does not 

 correspond with any specific discontinuity in the atmospheric layer. 

 It is accordingly remarkable that the separation of the surface, or 

 resultant distribution, into two distributions of opposite type should 

 be so complete when the level of 4000 metres is taken as the surface 

 of separation. It is not desirable to follow out the consequences of 

 small differences that might be found, because the calculation of the 

 isobaric distribution at high levels is itself as yet not susceptible of 

 very great accuracy, and indeed the distribution of pressure at sea 

 level even in the northern hemisphere is not entirely free from 

 uncertainty, owing to the uncertainty of the reduction for altitude. 



The remarks in this paper refer to the circulation in middle and 

 higher latitudes, because the determining force for steady motion is 

 assumed to be dependent upon the velocity of motion of the air and 

 rotation of the earth. The acceleration computed from these elements 

 contains sin X as a factor, and it is therefore without serious influence 

 in the equatorial regions. 



It is fortunate that this is so, because the drift of upper air over the 

 equator is generally accepted as being from east to west, and about 

 80 miles per hour may be assigned as the rate of this drift. From the 

 results of Sir J. Eliot's work on the cloud observations of India,* it 

 appears that at Simla, in latitude 31 N., the westerly current in the 

 upper air is extraordinarily steady throughout the year, whereas at 

 Madras the upper current shows considerable variation with the season. 

 Between the westerly current at Simla and the easterly equatorial 

 upper current there must be a region where the conditions in the upper 

 air are in many ways similar to the surface conditions in temperate 

 latitudes, that is to say, steady motion would involve the existence of 

 two oppositely directed streams of air at the same level, but in not 

 far distant latitudes. It would appear that for rotatory storms 

 originating in that region the gradient must depend only upon local 

 centrifugal action and the velocities for given barometric variations 

 must be correspondingly great. I am not sufficiently well Acquainted 

 with the sequence of events in a tropical hurricane to be able to follow 

 out the suggestion that those phenomena have their origin in the 

 upper air; I hope to be able on a subsequent occasion to cite 

 examples to show that storms in the region of the minimum pressure 

 in temperate latitudes may arise from special surface conditions, and 

 there is at least some evidence for the correlative origin of tropical 

 hurricanes. 



I have made no comparison of the first results of this method of 

 analysis of the average barometric distribution with the conclusions 

 arrived at by J. Thomson and Ferrel, for the general circulation of the 



* ' Indian Meteorological Memoirs.' vol. 15, Part I, 1903. 



