DIURNAL VARIATIONS OF TERRESTRIAL MAGNETISM. 
29 
solar cycle. A longer period than seven years would, of course, have been 
advantageous, but the labour of computation was already great. 
The method of computation adopted, and in particular the method of calculation of 
the non-semi-diurnal harmonic components, of changing phase, has been described in 
an earlier paper, and a reference to this must suffice here. # In that paper are given 
the data, so obtained, from observations made at Pavlovsk and Pola. These are 
two of the five observatories chosen for consideration in this research ; the other three 
are Zi-Ka-Wei, Manila and Batavia. The results obtained for the three latter have 
not hitherto been published ; they are to be found in Tables IV. and V. The first of 
these contains the Fourier coefficients corresponding to the different phases of the 
moon, reduced to the epoch of new moon ; these data are subject to certain 
corrections to amplitude and phase (cf. § 6 of the paper cited) which for convenience 
have been applied only to the mean results. The latter, transformed in terms of the 
geographical components of force, are given in Table V. The results for all the five 
observatories are collected in Tables VI. (a) to ( d ). 
The method of treating the seasonal changes is slightly different from that adopted 
for the solar diurnal variations. Instead of dividing the year into four quarters it 
was divided into three equal parts, November to February representing the winter 
solstice, May to August the summer solstice, and the intervening four months the 
equinoxes. It would have been better, for purpose of comparison, if this method of 
sub-division had been adopted also for the solar diurnal variations ; and, as Dr. Chree 
points out, this sub-division of the year corresponds more closely than the one adopted 
with the actual seasonal changes in the solar diurnal variation. 
As regards the solstitial data, since the semi-sum and semi-difference form the 
basis of analysis ( cf. Tables VI. ( c ) and (cl), corresponding to the solar diurnal 
Tables II. (/3) and (y)), the mean solstitial and the seasonal harmonics of Table G 
result from eight months’ material, while the equinoctial material is based on 
only four months of the year. Less weight must accordingly be attached to the 
latter than to the former. It may also be remarked that the mean solstitial and 
equinoctial epochs are slightly different for the two sub-divisions of the year, and 
that therefore some allowance must be made for this when comparing the solar and 
diurnal results. 
The analysis of the “ observed data of Tables IV. (a) to ( d ) is similar to that 
explained in § 8; owing to the small number of observatories dealt with, however, no 
* ‘ Phil. Trans.,’ A, 214, p. 295, 1914; cf. also A, 213, p. 279, 1913. It should be noted that in § 6 of 
the former paper a phase correction is given with the wrong sign, viz., -2L/29 degrees instead of (as it 
should be) + 2L/29 degrees. In applying the correction it is to be understood that the time of lunar 
transit at Greenwich has been used as the local time of lunar transit on the same civil day at the other 
stations, otherwise 360 degrees would have to be added to or subtracted from L degrees. The phase angles 
given in Table VI. (a), p. 316 of the former paper, need to be diminished by 4 • 2 degrees (Pavlovsk) and 
2‘0 degrees (Pola) on the above account. 
