DIURNAL VARIATION OF TERRESTRIAL MAGNETISM. 
179 
conductivity, though helping towards a better agreement between the diurnal and 
semidiurnal terms, are insufficient to account completely for the large excess of the 
summer variation over that observed in winter. This inequality is expressed by fl 3 2 
for the semidiurnal variation and Of for the diurnal variation, and its relative 
magnitude is indicated by the ratios f2 3 2 /fl 2 2 and D.f/flf respectively. The calculated 
value of both ratios is shown by the tables in § 10 to be y — sin 8, where 8 is the 
sun’s declination. If we compare the variations during the six summer months with 
those during the six winter months, we must substitute for sin 8 its average value, 
which is about 0 - 26. On the other hand, the results of my previous communication 
allow us to deduce the ratios L> 3 2 /n 2 2 and from the observations, and we find in 
this Off and Off respectively, or values between two and three times as great as those 
calculated from the assumed law of conductivity. 
To explain the difference we might imagine some cumulative effect, so that in 
midsummer the conduction would be greater than in winter even for the same elevation 
of the sun, but our present knowledge does not justify us in assuming this to be the 
case. I am inclined, therefore, to consider that the cause of the discrepancy lies 
in the fact that, as already suggested in the introductory remarks, the oscillations 
responsible for the barometric and magnetic phenomena are to some extent independent 
of each other, affecting different layers of the atmosphere. There are theoretical 
reasons why this should be so. It is now, I think, generally recognised that the 
importance of the semidiurnal variation of the barometer is due to the fact that the 
free period of the atmospheric oscillation, dependent on the velocity potential xJj 2 2 , is 
very nearly equal to 12 hours. But it is to be remarked that if concentric layers of 
the atmosphere be considered separately, there must be a considerable variation in 
the free periods owing to differences of temperature, and in the highest regions, in 
which alone electric currents of sufficient intensity can circulate, the temperature is 
probably so low that the free periods would be more than doubled. If we take these 
highest layers to oscillate to some extent independently, we should not therefore find 
the semidiurnal variation stand out in the same way as it does for the lower layers. 
Further, the inequalities of solar radiation in the two hemispheres near solstice ought 
to cause an appreciable oscillation dependent on the velocity potentials r// 2 and xf/f. 
The barometric variation due to xfjf is unimportant compared with that due to xfjf, 
because although the forced period is 12 hours, the free period corresponding to the 
motion involved in it has now a different value ; but in the upper layers the relative 
importance of i fjf would be increased, or, as it would be more correct to say, the 
relative importance of ip 2 2 disappears. This would account for the magnitude of the 
seasonal term in the magnetic variation. 
The suggested partial independence of the oscillations of the upper and lower 
atmospheres may also explain the discrepancy of phase, which we found to be If hours, 
but is in reality somewhat greater, owing to the fact that self-induction has been 
neglected in calculating the phase. With the calculated conductivity, self-induction 
