154 DR. C. CHREE: ATMOSPHERIC ELECTRICITY POTENTIAL GRADIENT 



When the phase angle in a Fourier wave varies much throughout the year, 

 contributions to a diurnal inequality from different months to some extent neutralise 

 one another. In such a case the amplitude derived from the annual or seasonal 

 diurnal inequality may give a less accurate idea of the average size of the Fourier 

 wave than is supplied by the arithmetic mean of the amplitudes derived from the 

 inequalities of the individual months. For this reason arithmetic means of the 

 c-coefficients are given in Table V., in addition to the values derived from the mean 

 diurnal inequality for the year. 



The phenomena presented by the 24-hour wave in Tables V. and VI. favour the 

 division adopted of the year into three seasons. In the four winter months c : is 

 conspicuously large and the variations in a^ small. In the four summer months, on 

 the other hand, r, is conspicuously small and a, is distinctly less than in winter, i.e., the 

 hours of maximum and minimum are later. The equinoctial months represent a 

 transition from summer to winter. 



In the 12-hour wave there is much less seasonal variation either in amplitude or 

 phase angle. The amplitude of the 8-hour wave is fully as suggestive of the more 

 usual division of the year into a summer half, April to September, and a winter half; 

 but the phase angles favour the division adopted into three seasons. In the 6-hour 

 wave the four winter months resemble one another in the smallness of the phase angles, 

 but there is no marked difference in this respect between equinox and summer. 

 There is obviously a good deal that is " accidental " in the amplitudes and phase angles 

 obtained for individual months. Little significance, for instance, can be attached to 

 the smallness of the July value of c l in Table V., or of the August value of c t in 

 Table VI. 



12. There are marked differences between the results for c l and t in Tables V. and 

 VI., but as these arise from differences between the earlier and later of the 15 years 

 they are best studied by comparing data for the periods 1898 to 1904 and 1905 to 

 1912. This comparison is made in Table VIII. It is confined to arithmetic means of 

 the amplitudes, and to seasonal values of the amplitudes and phase angles, which 

 suffice to bring out the main features. The data for 1898 to 1904 have been derived 

 from E! by multiplying the amplitudes by 1'91 and altering the phase angles from 

 G.M.T. to local mean time. If any hesitation is felt in accepting the multiplier 1'91, 

 it may be pointed out that it brings the mean values of the potential gradient for the 

 two periods into agreement. Practically identical conclusions would have been reached 

 if we had expressed the Fourier amplitudes not in absolute measure, but in terms of 

 the corresponding mean potential gradients. The multiplier does not affect the phase 

 angles. 



The outstanding feature in Table VIII. is the smallness of the differences between 

 the results for the 12-hour wave from the two epochs. In the case, however, of the 

 24-hour wave the differences are obvious. The average amplitude is considerably 

 larger in the second period than in the first, though summer shows the opposite 



