REPORT OF THE SECRETARY. 29 



from Portland, Maine, in latitude 43° 54' 5, longitude GO"^ 57' 4, and 

 74 feet above high water. The observations were made at 7 a. m., 

 1 p. m. and 6 p. m., and relate to indications of the thermometer and 

 barometer, direction of the wind, state of the weather, amount of 

 rain and snow, character of clouds, occurrence of thunder-storms, 

 fogs, frost and hail, earthquakes, auroras, etc. 



From these observations the mean temperature of each da}^ of each 

 month is deduced and arranged in tables. The mean temperatures, 

 however, require a small correction in order to reduce them to the 

 mean temperature of the day which would be given from twenty-four 

 or hourly observations instead of only three observations. In the dis- 

 cussion of the temperature the correction applied on account of the 

 irregular hours was deduced from a series of observations taken at 

 every hour of the twenty-four at Toronto and Montreal, which are 

 found to have been subjected to the same fluctuation of temperature 

 as Brunswick. To understand this, perhaps the following expla- 

 nation is necessary: By adding all the tc-mperatures observed at 

 each hour of the day, for example all at o'clock, into one sum, and 

 dividing these by the whole number of observations at this hour, we 

 obtain the average or mean temperature of that hour, and by repeat- 

 ing the process for every other hour we obtain a series for each hour 

 of the twenty-four; also, by adding together all the average tempera- 

 tures of each hour of the day and dividing by twenty-four, we obtain 

 the mean temperature of the day. If the mean temperature of the 

 day be compared with the mean temperature of each hour, some of 

 the latter will be a little above and others a little below the former; 

 and as these differences are found to be the same over a large extent 

 of country, we may apply them to observations made at one, two, or 

 three hours, so as to get the same result which would be obtained had 

 the observations been made at every hour during the twenty-four. 

 Thus it has been found, from several series of hourly observations in 

 different parts of the United States, that those made at 7 a. m., 2 

 and 9 p. m., give a nearer approximation to the mean temperature of 

 the day than those made at any other hours. We dwell some- 

 what on this point because the idea has been prevalent that the best 

 times for determining the mean temperature are at sunrise, noon and 

 sunset. But since sunrise and sunset are variable hours, it is obvious 

 that corrections similar to those we have mentioned above cannot bo 

 readily applied to them. 



The observations at Brunswick, having been duly corrected in the 

 way we have mentioned, present, during a period oi Jifty-tivo years, 



