8 



Mr. Finlay, on the Variations of 



[Sept. 3, 



montli, we get the following differences between the computed values 

 and the observed values of the table above : — 



Jan. 1 . 



-4-0-01 



Apr. 1 



„ 15 . 



+0-01 



„ 15 



Feb. 1 . 



—0-02 



May 1 



„ 15 . 



. —0-01 



,,'15 



Mar. 1 . 



. —0-01 



June 1 



„ 15 . 



. —0-02 



,, 15 



July 1 . 



„ 15 . 

 Aug. I . 



,, 15 . 



. —0-03 



. +0-01 



0-00 



0-00 



Sept. 1 . 

 „ 15 . 



. +0-01 

 0.00 



Oct. 1 . 



. —0.01 



., 15 . 



. —001 



Nov. 1 . 



. —001 



M 15 . 



. —0-02 



Dec. 1 . 



. -0-01 



„ 15 . 



. —0-02 



0-00 



0-00 



0-00 

 -fO-02 

 -fO-02 



0-00 



These residuals are so small that we may say the formula represents 

 the observed values perfectly. 



The question now comes ** To what cause are these changes due ? 

 Will rainfall or temperature account for them?" The following table 

 gives the mean monthly amount of rainfall and the mean monthly 

 temperature of the air for the same years : — 



Table II. 



Rainfall. 



Temp. 



o 



Rainfall. 





Temp. 





in. 





in. 



o 



January 



.. 0-55 



68-86 



July 



3-79 



54-24 



February 



.. 0-61 



68-46 



August 



3-40 



55-36 



March 



.. 1-03 



66-03 



September . . 



2-00 



57-24 



April 



.. 1-84 



62-20 



October 



1-78 



60-33 



May 



.. 4-13 



57-78 



November . . 



1-13 



63-52 



June 



.. 4-53 



55-17 



December . . 



0-95 



66-67 



Mean monthly rainfall = 2*14 in., mean temperature = 61 -32°. 



The temperatures are taken from the readings of DoUond's ther- 

 mometer in the window-crib close to the transit-circle, and though 

 not representing perfectly the shade temperature, on account of the 

 position of the crib, yet they have the advantage of being a continuous 

 series by the same thermometer in the same position. 



In considering the rainfall I am at a loss ho^ to represent its 

 accumulated effect, whether simply according to the amount of rain 

 or according to some more complex law. The sudden increase of 

 rain in May and June agrees well with the fall of the level curve 

 towards the end of June, but while the rainfall has reached its 

 average amount in September, the level curve still continues to fall 

 till the end of November. This, however, may be due to accumulation 

 of rain. Again, the level curve mounts steadily through March and 

 April, although the rainfall in April is three times as much as in 

 January or February. On the whole, I do not think the rainfall is 

 a perfectly satisfactory determining cause. 



The dotted curve in fig. 1 represents the variation of the monthly 

 temperatures from the mean of the year laid down to scale so that 

 the amplitudes of the temperature curve and the level curve shall be 

 nearly the same. The perfect similarity of the two curves is striking, 

 and it is impossible to withstand the conclusion that the two curves 

 are intimately connected, and that the changes of level are in some 

 way due to the changes of temperature throughout the year. But 

 now a curious point arises — the times of maxima and minima of the 

 two curves do not agree ; the level follows the temperature by about 

 four months. It seems, therefore, that the changes of level are not due 

 to the effects of heat directly on the instrument itself, but to an 

 actual movement of some part of the Observatory Hill, in such a posi- 



