232 Meteorology of New Ha'Cen. 



If we combine any morning observation with an observation made 

 about the time of greatest heat, a\is obtain a resulting temperature 

 too great. 



Twice during each day, the height of the thermometer coincides 

 with the mean temperature of the day. This coincidence occurs 

 about 15 minutes before nine in the morning, and also about 12 min- 

 utes before eight in the evening. We should therefore obtain the 

 mean temperature by a single observation at either of these instants ; 

 but a iiniform system of observation is more likely to be adhered to, 

 wben the time of observation is an even hour. If we combine 9 

 o'clock in the morning with 8 o'clock in the evening, the resulting 

 temperature will be about a degree too high in summer, and about 

 three-fourths of a degree too low in winter. The mean for the year 

 will be only one-sixth of a degree too great ; and this is the nearest 

 approach to the mean temperature which can be obtained by com- 

 bining any two hoxxrs of observation, unless we employ inconvenient 

 hours of the night. In other Avords, there are no two convenient 

 hours whose mean corresponds with the mean temperature of New 

 Haven with the precision which is desirable in such observations. 



If the mean temperature is to be deduced from observations at two 

 hours of the same name, the hours which best combine accuracy with 

 convenience are 10 a. m, and 10 p. m. ; but these give a result about 

 one-thJi-d of a degree too great. 



If observations are made at three hours daily, we shall obtain the 

 mean temperature very accurately by observations at 6 a. m., 2 and 

 9 p, M. ; but during the winter months 6 a. m. is an inconvenient hour, 

 and not likely to be faithfully adhered to. The mean of observations 

 at V A. M., 2 and 9 p. m,, is too great ; but if we add twice the nine 

 o'clock observation to the sum of the other two observations, and 

 divide the result by 4, the error of the result for the separate months 

 in only one instance exceeds a quarter of a degi'ee ; and for the en- 

 tire year differs but one-hundredth of a degree from the true mean 

 temperature. These then are the three hours to be generally recom- 

 mended for observation, and fortunately they are extremely well 

 adapted to the convenience of most observers. 



The annual variation of temperature may be represented by Bes- 

 sel's well-known formula 



T:, = (T)4-A,sin(a;4-BJ-f A2sin(2a;+B2)+A3sin(3a'-j-B3)+ etc., 

 where T^ represents the mean temperature of a day corresponding to 

 an epoch x of the year, x being counted in degrees from some as- 



