82 REPORT— 1898. 



adding to the list those of the years 1889, 1895 and 1896, years that 

 appear in subsequent tables, are as follows : — 



1889. July 17, November 1. 



1890. November 8. 



1891. April 8, 12; May 14, 15, 16. 



1892. January 4, 5, 6 ; February 13, 14 ; March 6, 12 ; April 25, 26; May 1, 



2,18, 19; June 2, 3, 27; July 12, 13, 14, 16, 17; August 12; No- 

 vember 4, 5 ; December 5. 



1893. April 26 ; July 16 : August 7, 18 ; November 2. 



1894. February 23, 24, 25, 28 ; March 30, 31 ; July 20 ; August 20; Septem- 



ber 15 ; November 13. 



1895. None. 



1896. None. 



Regarding the selection of five days in each month for the quiet day 

 tabulation, it should be explained that the days selected are really days 

 of quiet magnetism, to which, until the end of the year 1893, I can per- 

 sonally testify. And further, the selection was made so as to bring the 

 mean of days in each individual month as near to the middle of the 

 month as possible. Occasionally, on account of the prevalence of mag- 

 netic disturbance, this condition could not be quite fulfilled. Still, in 

 seventy-eight of the ninety-six individual months here treated, the mean 

 of selected days comes within two days of the middle of the month ; 

 besides which, in combinations of months grouped together in different 

 ways, the influence of deviation in this respect becomes greatly diminished, 

 indeetl, usually nearly neutralised or destroyed. Further, whenever quiet 

 day values are compared, as the same quiet days are employed the question 

 of dissimilarity of epoch disappears. It is only in the comparison of the 

 Greenwich all day tabulation with the Greenwich five day tabulation in 

 Tables I., II. and III., in the comparison of the Greenwich diurnal range 

 of the all day tabulation with that of the quiet day tabulation in Table 

 XI. (under columns b-c), and in Table XII., that the slight difference of 

 epoch comes in. But the corrections required in these cases are so small, 

 as compared with the magnitude of the variations and differences of 

 elements exhibited, that it can scarcely be said to be worth attempting 

 any correction, remembering further that, excepting in Table V., the 

 figures are throughout carried one figure further than that of the actual 

 tabulation. The effect on the values of Table XII. is the more important, 

 and will be considered when speaking of that table. 



In the quiet day results of Tables I., II., and III. the non-cyclic incre- 

 ment — that is, the algebraic excess of the terminal midnight value over that 

 of the commencing midnight value — has been in each month eliminated 

 by assuming in each element a uniform change through the twenty-four 

 hours, and making correction accordingly. For the all day results no 

 similar correction was needed. In the further tabulation of the two sets 

 of numbers, they have been grouped by seasons in the way before men- 

 tioned, which represents more dii-ectly the seasonal effects than does the 

 ordinary quarterly grouping, since neither the combination of the first 

 three months of the year, nor of the last three months, gives due repre- 

 sentation of the midwinter effect, March and October being both months 

 having large diurnal inequalities. In declination, the values of hourly excess 

 of quiet day value have, in the different seasonal periods, much the same 

 character, but the Spring and Autumn numbers are, on the whole, 

 numerically the larger. The sums of the twenty-four values of excess, 



