TEMPERATURE AT HALIFAX, NOVA SCOTIA. 31 



In Table IV are shown the errors in the approximate monthly means 

 when the daily mean is considered as the simple arithmetic means of 

 the temperatures observed at 7 a.m., 2 p.m. and 9 p.m., and of those 

 observed at 6 a.m., 2 p.m. and 10 p.m. ; and also when the daily mean 

 is taken as equal to the fourth part of the sum of the temperatures at 

 7 a.m., 2 p.m., and twice the temperature at 9 p.m. 



It is seen that the arithmetic mean of observations'at 7 a.m., 2 p.m. 

 and 9 p.m., gives a result too high by 0.° 56 on the average of all 

 months, and nearly eight-tenths too high from April to July. 



The arithmetic mean of observations at 6 a.m., 2 p.m. and 10 p.m., 

 is in no case more than half a degree in error; it is too high from 

 November to February, and too low during the rest of the year, the 

 average error irrespective of sign being a quarter of a degree, and the 

 error in the annual mean less than one-tenth in defect. 



When the observation at 9 p.m. is reckoned twice, the greatest error 

 which occurs in any month is slightly greater, but the average error of 

 the twelve monthly means, the signs of the errors being disregarded, 

 is 0.° 18, and the error in the annual mean the same as in the 

 preceding case. 



These results accord in their general character with the experience 

 of other places at which the observations have been sufficiently frequent 

 for determining the diuroal variations. 



As regards suitability for yielding daily means, 7, 2, 9, 9, and 6, % 

 10, may be regarded as of nearly equal merit; and as 7, 2, 9, are in 

 many respects more convenient to observers than the other combina- 

 tion, and less liable therefore to interruptions, these hours are recom- 

 mended to those observers in Canada who read their instruments three 

 times each day.* 



The numbers in Table III, as the title states, are the monthly means 

 of the diurnal variations, and are only adapted therefore for the reduc- 

 tion of monthly means at single hours to monthly means for all hours, 

 and for the converse reduction. Comparison of like hours in contiguous 

 months will show a considerable difference in the analogous variations. 

 To correct daily means, therefore, it is necessary to possess tables in 

 which diurnal variations are given at much shorter intervals. As 

 three years is scarcely sufficient for the elimination of accidental irre- 

 gularities, the computation of the diurnal variations for every fifth day 



* 7 a.m., 2 p.m. and 9 p.m., have teen adopted for many years ty the Smithsonian Institution. 



