262 Af. Ramon d's Instructions for the Application of [Oct. 



In the former case barometrical measurements tend to err in 

 excess ; in the latter, in defect. 



If the winds which prevail at the two stations be different, 

 the measurement is too great when the more dense wind occu- 

 pies the lower stratum, and too small when the upper. 



Lastly, the errors augment or diminish, cecteris paribus, with 

 the horizontal distance of the two stations, and with the height 

 to be measured. 



Among the modifications of the atmosphere there is one of 

 the most hidden description, biit nevertheless most regular, 

 which has been investigated with difficulty, but which once 

 known can occasion no errors which we have it not perfectly in 

 our power to prevent. It is long since horary oscillations have 

 been perceived in the barometer. It is long since Deluc observed 

 that the different hours of the day are not equally proper for the 

 measurement of heights. 



I have observed that any formula can only be really applicable 

 at the precise hour at which we may have made the experiments 

 necessary for the determination of our coefficient; and that, 

 because the coefficient is alv»^ays affected by a quantity which 

 represents the mean ratio of the weight of the air and its pres- 

 sure, a ratio essentially variable, and different at every instant 

 of the day. 



The coefficient of the formula of M. de Laplace is adapted to 

 the hour of noon. We must, therefore, make observations for 

 the measurements of heights at the hour of noon only. This 

 precept is important, for the errors which result from the appli- 

 cation of a coefficient to hours for which it was not calculated, 

 are among the most considerable that we can make. Yet this 

 consideration will by no means prevent us from prolonging a 

 little the time devoted to the operations. The interval between 

 eleven and one o'clock does not exceed the limits which it is 

 reasonable to prescribe to ourselves ; but then if we would be 

 exact, we must operate in such a manner as to effect a compen- 

 sation between the opposite errors which may arise from this 

 source. Before noon the measurements err in defect ; after 

 noon, in excess. My practice is, therefore, to make, besides the 

 observation at noon, one or two others before, and as many 

 after, at intervals respectively equal. This method possesses 

 many peculiar advantages : we have time to examine the pro- 

 gress of the instruments; each observation serves as a point of 

 comparison to judge of the others ; and the mean term taken 

 between them is in a manner the observation of noon itself, free 

 from those errors which might be introduced by that accidental 

 state of the atmosphere which should predominate at the real 

 moment of that observation. ^ 



Lastly, in order to gain from this combination all the advan- 

 tag^es Qf which it is susceptible, it will be advisable that corres- 



