1823.] the Barometer to the Measurement of Heights. 169 



In the temperate regions, the case is widely different — the 

 frequency and extent of the accidental variations, disturb and 

 disguise the effects of the diurnal. It nevertheless does exists 

 and attentive observation may soon detect it. We must, there- 

 fore, take it into account, and when it becomes a question to 

 determine the mean pressure of the atmosphere, we must no 

 more neglect the horary variations in the compensation between 

 the accidental, than the accidental in the compensation between 

 the horary. This certainly makes the problem a little more 

 comphcated. The difficulties increase, the task becomes longer; 

 but the method of proceeding is not altered in its nature. 



It is necessary, therefore, that in each series the observations 

 should belong to the same hour ; for every hour having its par- 

 ticular variation, a series composed of observations made at 

 different times contains the diurnal variation as an indetermi- 

 nate quantity, and under an irreducible form. In the next 

 place, the hour of each series must coincide with one of epochs 

 of the diurnal variation ; for the comparison of series which 

 belong to intermediate hours does but imperfectly compensate 

 the deviations of the diurnal oscillation, 



I have determined for our climates the progress of these 

 horary variations. In summer the maximum is at 8, a. m. and 

 10 p.m.: the minimum at 4, a.m. and 4, p. m. In winter there 

 is an hour's retardation in one of the four epochs, and an hour's 

 advance in the others ; they are respectively 3 and 9, a.m. and 3 

 and 9, p. m. In the spring and autumn, 3i- and 8^, a. m. ; 3-|- and 

 9-^f p. m. We have then only to make four observations in the 

 day at these four epochs ; and to continue the series for a suffi- 

 cient length of time to compensate accidental variations, and 

 then to take the mean of each series, from which again we may 

 take the mean for the day. But it may be asked, what must be 

 the length of time necessary to obtain a compensation of the 

 effects of accidental variations ? To judge from the general 

 practice we might suppose that the series must be continued for 

 a period altogether indefinite ; but in fact, however capricious 

 the phenomena of accidental variations may seem, they yet 

 recognize certain laws. Each epoch has its pecuhar share of 

 these variations, the result of which, after every compensation 

 has been made, constitutes its character. This is the case with 

 each different season. The system of observation must proceed 

 by years ; because the result of the year compensates the cha- 

 racteristic accidental variations of the different seasons ; and a 

 barometric mean ought not to include fractions of years, because 

 it will then incline to the side of that season which is doubly 

 represented. Equal probability can only be attained in periods 

 of the same nature and extent. The mean of a complete year is 

 only to be corrected by the mean of another year ; and each 

 enters into the common result for the half of the difference ; a 

 third year, for a third, &c. and as the differences are small in 



