CHRISTCHURCH TERM-DAY OBSERVATIONS. 



235 



divide the number of observed changes in H by the corresponding number in I) multiplied by 0'66 we 

 shall probably overestimate the relative variability in H, while wo shall certainly underestimate it if 

 we divide the number for H by the number for D. The results of these two operations are as follows : 



If we put the variability in II as 40 per cent, greater than the variability in D, we shall probably not bo 

 very far wrong. The figures suggest a seasonal variation in the ratio, but it does not seem to be large. 

 A variation in the ratio would suggest, of course, a seasonal variation in the dominant direction of the 

 disturbing forces. 



The tables of hourly ranges call for less comment. If the diurnal variation consisted of a regular single 

 oscillation, the numbers in the penultimate columns of Tables VA and VIA would be approximately double 

 the corresponding numbers in the last columns. The ratio in reality varies within wide limits. In 

 Table VA the largest value of the ratio, 3-6, is met with on the two June term days ; in Table VIA the 

 largest value also occurs on June 15 and is no less than 5 '4. The largest hourly ranges in D and H 

 occurred respectively on December 15, 1902, and January 1, 1903, being 3' -9 for D and 24y for H. The 

 smallness of these maxima is eloquent testimony to the remarkable quietness of the term days. 



13. To ascertain more exactly the character of the diurnal variation in the frequency of the changes of 

 reading given by Tables HA to IVA means were derived for each hour from the first 24 term clays, 

 and, in the case of D and H, also from combinations of these days forming the same three seasons as in 

 Table IA. The two term days of March, 1903, were not taken into account. 



The results are given in Table VIlA, p. 244, the absolutely largest and least of the hourly values being 

 in heavy type. The penultimate column gives the mean of the 24 hourly values, and so represents the 

 average number of changes of reading per hour for the whole year and the seasons. The last column 

 gives the average number of changes per diem. The number of changes in the case of both D and H is 

 considerably larger at Midsummer than at Midwinter, but the difference between the seasons is much less 

 conspicuous than it is in the case of the range of the diurnal inequality. 



In the case of D changes of reading are most numerous close to noon, whilst the magnet is swinging 

 over to the early afternoon maximum of easterly Declination. In the case of H changes of reading are 

 most numerous in the afternoon, either an hour or two before or else an hour or two after the afternoon 

 maximum. In all three elements changes appear to be least numerous in the early morning in the 

 neighbourhood of 4 or 5 a.m. Whilst the day changes are decidedly more numerous than the night, the 

 difference is much less than one would have anticipated from a consideration of the diurnal inequality. 

 Thus if we define " day " as the twelve hours commencing 6 a.m., and assume that half of the entries under 

 hours 7 (6.30 p.m.) and 19 (6.30 a.m.) in Table VIlA belong to the day, while the inequality figures 

 appropriate to 6 a.m. and 6 p.m. are the arithmetic means of those for 18 and 19 and for 6 and 7 

 respectively, we obtain for the ratios of the day to the night number of changes, and the day to the night 

 inequality ranges, the following values : 



2 H 2 



