On the Methods of obtaining Mean Temperatures. 51 



ment ; and instead of having it completely protected from reflec- 

 tion, and in almost complete obscurity, taking care at the same 

 time to secure a free circulation of air about it, we find them often 

 in such situation as to receive from walls or the earth a considera- 

 ble portion of reflected heat ; often in closed apartments or against 

 walls, the temperature of which is influenced throughout the day 

 by the full force of the sun. Galleries fronting the north are fa- 

 vorite places for suspending thermometers, under the impression 

 that no heat is reflected from that side. In this way we often 

 find, in this climate, that when a thermometer properly placed 

 gives a temperature of 90° or 91°, ail the others in the vicinity 

 will stand at 95°, 100°, or even 110°; so that the annual mean 

 derived would be very greatly above the true one. There is, 

 however, it appears to me, a source of error much more general : 

 I speak of the methods of calculating the mean after the obser- 

 vations have been made, and of the adoption into general use of 

 methods which have been found to give correct results at particu- 

 lar places. The methods most in use are the following, viz. to 

 take three observations, and from these calculate the mean di- 

 rectly ; some fix upon sunrise, 2 p. m. and sunset ; others, as in 

 the Army Meteorological Register, and in those of most of the 

 meteorological societies, upon 7 a. m., 2 p. m. and 9 p.m. I have 

 ascertained, by taking the mean of hourly observations made 

 here by myself, that any of these methods give means which, 

 during every season of the year, are too high, and that the error 

 in excess is greater, in proportion as the diurnal exceeds the noc- 

 turnal temperature. The last method, which is now in most 

 general use, has probably been adopted in colder climates, as 

 agreeing experimentally in its results with those derived from 

 more frequently repeated observations during the same period ; 

 or possibly under the impression, that the 9 p. m. observation 

 would give an approximation to the nocturnal mean, while the 

 diurnal temperature would result as a mean from those at 7 a. m. 

 and 2 p. m. In either case this method might be the best adapted 

 for many climates, but as the relative value of the temperatures 

 at the times fixed for observation are not constant, but will vary 

 in passing from one climate to another, there is nothing in the 

 method to suit it for universal use. The fact is, it appears to me 

 almost impossible to fix upon any method, except that of hourly 

 observations, which will in every place afford even an approxi- 



