the Barometric Formulce, (^-c. 667 



reduced to I'l metre, being an error very small indeed, about one in two 

 thousand parts very nearly. 



With respect to errors of heights of the forenoon (from two hours to two 

 hours and a half after sunrise), when the horary correction vanishes, it appears 

 by inspection of Table XIII. that such errors are more variable and unsteady 

 than the corresponding errors near sunset. For example, the error for the 

 month of July is as much as 7-48 metres, that of June being 5-40 metres, and 

 that of April 4-54 metres, although the mean error for the seven months of the 

 forenoon, from March to September (both months inclusive), is only 1-84 

 metres, being but the one-hundredth of a metre different from the corresponding 

 mean of the afternoon (viz., 1'85 metres). 



These facts clearly indicate sunset as the moment of the day most favoiu'- 

 able for barometric observations, and the comparison of mean sunsets for the 

 nine months (which have two moments, that the horary corrections vanish), 

 with the mean moments of adjoining column, shows the same thing; for by 

 Table XIIL, the mean of sunsets is 6 h. 34 m., being different only by three 

 minutes of time from the mean of moments of observation of adjoining column, 

 whereas the mean of sunrises differs about two hours and twenty minutes from 

 the mean of adjoining column. In my paper on the Constants of the Baro- 

 metric Forraulaj, to which this, my present paper, is but a supplement, I have 

 recommended (in absence of a reliable table of horary correction) sunset and 

 two hours after sunrise, as the moments of the day most favourable for baro- 

 metric work. I now feel disposed to modify my opinion, as to observations 

 of the forenoon, and to recommend, as the best moment, the moment equally 

 distant with sunset, from the moment of highest temperature of the day, gene- 

 rally between one and two hours, p. m. 



As to other periods of the day, I should not be afraid to make calculations 

 for height by my own formula (in absence of a sound local table of corrections), 

 making use of the large Table X. of page 650, for the seven months from 

 Mar<jh to September (both months included), particularly if the arithmetic 

 mean temperatures, as given by the detached tliermometers, be not widely 

 different from the corresponding arithmetic means of temperature of Table X. 

 I do not, however, feel myself authorized to recommend such practice to others ; 

 in this matter let every one judge and act for himself, and modify his practice 

 according to his experience of its workings. 



