236 CHEISTCHUECH TERM-DAY OBSERVATIONS. 



14. The diurnal variation in tin- amplitude of the hourly ranges (R) in 1) and H as deduced from 

 Tables VA and VIA is given in Tables VIIL\ and IX,\, p. i' t.">, fur the year and the three seasons ; maxima and 

 minima are in heavy type. For comparison, the tallies also show the corresponding hourly increments (I), 

 as derived from the mean hourly readings for the year and the seasons, the largest positive and negative 

 increments being in heavy type. These last data differ from the increments derivable from the diurnal 

 inequalities in Table IA only in containing the non-cyclic change which was eliminated in forming the 

 diurnal inequalities. The presence of the non-cyclic element makes only a trifling difference. For 

 example, Table VIIlA gives as the increment in I), for the year as a whole, for the hour ending 

 at 2 the value + l'-22, whereas from the diurnal inequality in Table IA we have + 3'-02 - l'-81, 

 or +1'-21. 



Before discussing Tables VIIlA and IXA, two anomalous results should be explained, viz., the excess 

 in the I over the R figure for hour 24 for Year and Equinox in Table VIIlA. The excess obviously 

 cannot represent a physical fact, but only an imperfection in the data. It arises from the absence of 

 observational data during one or both of the two last hours of the 24. The hourly readings and the 

 hourly ranges missing were replaced independently by interpolation methods already described. The 

 inconsistency could no doubt have been removed by other methods of interpolation, but it cb'd not seem 

 worth while to take the trouble, as the results would not necessarily have made any nearer approach 

 to real accuracy. 



The R and I data in Tables VIIlA and IXA would agree if there were nothing but a regular diurnal 

 inequality with maxima and minima occurring at exact (Greenwich) . hours. As a matter of fact, the 

 difference between the two sets of figures in Table VIIlA is very small near the middle of the day, when 

 the Declination inequality change is most rapid. This is due, I think, to more than one cause. During 

 an hour when the inequality change is very rapid any uninll irregular disturbance has no chance of 

 influencing the hourly range, unless it occur near one end of the hour. If, for instance, the normal D 

 change during the hour is a rise of 2', the minimum value will occur very near the commencement, and 

 the maximum very near the end of the hour, unless there is an irregular disturbance amounting to at 

 least several tenths of a minute. It is thus obvious that the hourly range in D during one of the mid-day 

 hours when the needle is moving rapidly to the east cannot on an ordinary undisturbed day be much 

 in excess of the range derived from the hourly readings. This is unquestionably accountable in part 

 for the observed phenomenon, but it cannot, I think, wholly explain it, especially in the Midsummer 

 months, and the conclusion seems warranted that the tendency to an oscillatory, or ebb-and-flow, type 

 of variation in D was at its minimum during the day hours when the regular inequality changes were 

 most rapid. 



In the night hours the R and I values differ greatly in both the Tables VIIlA and IXA, in fact there 

 is little relationship between them. It would, however, be incorrect to assume that the R changes owe 

 nothing to the diurnal inequality even during an hour when the I change vanishes. The difference between 

 solar and mean time is not wholly negligible in any one of the seasons, still less so in the course of the 

 year, also the times of maxima and minima in the diurnal magnetic inequality vary throughout the year 

 whether we refer them to solar or to mean time. A low I value may mean that the diurnal inequality 

 changes are very slow throughout the whole hour. It may signify, however, that the mean time of 

 occurrence of a maximum or minimum falls about the middle of the hour, the actual time of occurrence 

 varying considerably in different months comprised in the same season. The contributions from different 

 portions of the season to I will then differ in sign, and may largely neutralise one another, but this 

 of course is not the case with their contributions to R. This is unquestionably very largely the cause 

 of the great difference between the R and I figures for some of the day hours, e.g. the hours ending 

 at 22 and 3 (i.e. at 9.30 a.m. and 2.30 p.m.) in Equinox and Year in Table VIIlA. In most of the 

 night hours, however, the difference between the R and I figures must be otherwise explained. In 

 Table IXA there are some day hours, e.g. in Midsummer those ending at 1, 21, and 22, when the 

 R and I figures are very close, but this is exceptional, the differences being on the whole much more 

 conspicuous for H than for D. This is in strict accordance with the greater variability already observed 

 in II, but is not, I think, due exclusively to this cause. If we divide the arithmetic mean of the R by 



