DIURNAL VARIATIONS OF TERRESTRIAL MAGNETISM. 
of the field. The notation is the same as that explained in §§ 9 and 1 1 , only the 
seasonal divisions and the unit of force (here O'Oly in place of OTy) being different. 
In Tables VI. ( b ) to ( d ) the calculated values of a n and b n , corresponding to these 
harmonic functions, are placed for comparison beside the values computed from the 
observational data. 
The agreement between observation and calculation is naturally less good than for 
the solar diurnal variation, both because the accidental error in the data is greater in 
the present case (the whole effect being smaller) and because single observatories are 
here used in place of groups of observatories, so that the local irregularities are 
larger. The agreement is better, on the whole, in Tables VI. (c), ( d ) than in 
Table VI. ( b ), which rests on only half as much observational material as the two 
former ; it is also better for the West than for the North component of force, as in 
Part II. The agreement is surprisingly good in Table VI. ( d ) for the horizontal force 
components. The vertical force variations are smaller than the horizontal force 
variations, and some of the values of A m n and B„”, determined from the former, 
are very uncertain. On the whole, however, while the present data might be very 
considerably improved upon, the results prove more satisfactory than I had 
expected, at any rate for the horizontal force potential A m n and B /rt ". In judging the 
success of the analysis, regard may be had to the agreement of phase between the 
harmonic components of different periods and the reproduction of other features of 
the analysis of the solar diurnal magnetic variation, which show the close parallelism 
of the two phenomena. Although some of the tables of observed and calculated data 
in Table VI. do not seem to show much correspondence between the two, the results 
in Table G suggest that the assumptions underlying the analysis (i.e., that the 
variations can be represented by the functions Q B n+1 or Q„") are sound, and that the 
discordances from the results of calculation arise from a relatively large amount of 
accidental error in the observational data. 
The use of so small a number of observations is, of course, a fit ground for criticism, 
and calls for a repetition of this part of the investigation on a larger scale. The 
present is to be considered as merely a pioneer attempt. For this reason the analysis 
has been narrowly restricted, and the results must be discussed with due recognition 
that the percentage error is not small. 
§ 14. Comparison with van Bemmelen’s Data. 
As bearing upon the question of the accidental error of the initial data of 
Tables VI. (a) to (d), it is interesting to compare the present values of a 2 , b 2 , 
with those calculated by van Bemmelen for the same observatories (only the a 2 , b 2 
coefficients are given by the latter author). It should be borne in mind, howmver, 
that neither the epoch, the amount of observational material dealt with, nor the 
method of computation, was the same in the two cases. Dr. van Bemmelen divided 
