158 DR. C. CHREE: AN ENQUIRY INTO THE NATURE OF THE RELATIONSHIP 



7. In the case of D, Table II. , we see on the whole a close resemblance between 

 the results from the two longest series of data. The later period shows, however, a 

 slightly increased value of b and a slightly decreased value of a in winter, equinox, 

 and the year as a whole. The last period, 1889-96, in the " all" day data shows a 

 further increase in b and decrease in a ; and the corresponding " quiet " day results 

 give a still smaller value of a, especially in the winter months. The "all" and 

 " quiet " day mean values of b for the year, from the 1889-96 period at Greenwich, are 

 practically identical and very close to the corresponding Kew value. The values of 

 b/a at Greenwich are in each season slightly larger for the " quiet" days than for the 

 corresponding "all" days; and comparing the "all" day data amongst themselves 

 we have an increase in b/a in passing from the longest to the mean period, and in 

 passing from the mean period to the shortest period. This would imply that b/a has 

 increased of late years. 



At Greenwich, as at Kew, b is conspicuously lowest in " winter" ; but no one of 

 the four columns of Greenwich results gives so distinct an excess in the equinoctial 

 over the summer value as appears at Kew, and, on the whole, we should infer that 

 the equinoctial and summer values of b at Greenwich are practically equal. 



At Greenwich, as at Kew, b/a is distinctly smaller in summer than in the other 

 seasons ; but the " winter" value of b/a at Greenwich, instead of markedly exceeding 

 the values for the other seasons, as at Kew, would seem to be if anything slightly 

 smaller than the equinoctial. 



On the whole, the variation of b/a throughout the year at Greenwich is surprisingly 

 small. 



8. The H ranges in Mr. ELLIS' tables are expressed in terms of the value of H at 

 Greenwich. To make the results comparable with those for other stations, I have 

 expressed the ranges in terms of ly as unit. In doing so, I treated H as constant for 

 each period, and as possessing the following values : 



Not knowing the exact procedure followed at Greenwich, I may not unlikely differ 

 slightly from the exact values adopted there ; but, for the purpose of the present 

 enquiry, such small uncertainty as may exist is immaterial. The values of b/a are, of 

 course, independent of the unit adopted. 



The 1841-96 and 1865-96 data in Table III. present some conspicuous differences; 

 a is larger in the former series than in the latter in every single month of the year, 

 but b shows exactly the opposite phenomenon in 9 months out of the 12. This 



