68 Lieut.-General K Strachey. Harmonic Analysis [Mar. 2, 



while it returns in the opposite direction in the change hack to 

 winter. 



For the first component the variation of the five years' mean of fi 

 from the twenty years is in no month more than or ten minutes 

 of time, and. the average for all months is less than half that amount. 



In the second component the epoch of maximum, which during the 

 winter months is always after midnight, falls back in the summer, 

 when it is at times before that hour. The variation of the five-year 

 mean from the twenty-year mean is in no month more than 6°, and 

 the average is only 2°*3, or nine minutes of time. 



In the winter months the maximum of the third component is always 

 between 4 a.m. and 5 a.m. ; in March it changes rapidly, in the 

 summer being found invariably between midnight and 1 A.M., while 

 after September it returns to its winter position. The variation of 

 the five-year from the twenty-year mean in no month exceeds 5°, and 

 the average in all months is only 2°1, or 8| minutes of time. 



The fourth component shows double maxima and minima, the* 

 former at the equinoxes, the latter at the solstices. The largest varia- 

 tion of the five-year mean of any month from the twenty-year mean 

 is 10°, and the average for all months is 4°'3, or seventeen minutes.. 

 Considering how small are the absolute values of the coefficients p±, 

 and q it on which the value of ^ 4 depends, the average being a little 

 less than xoth of a degree Fahrenheit, it is rather a matter of surprise 

 that the variations should be so small than that they should reach 

 their actual amounts. 



It may be noticed that the total amplitude of the components being 

 y/ (p 2 + q 2 ) , a considerable variation of its value is quite consistent 

 with invariable or slightly varying values of which depend on the 

 ratio pjq. 



The component of the first order, which in the winter is more than 

 double the magnitude of any of the others, and in summer more than 

 ten times as great, gives the dominant character to the daily curves 

 of temperature. In the series of twenty years variations in different 

 years of as much as 100 per cent, are to be found for almost every 

 month, but for the most part even these irregularities disappear in 

 the mean of a series of five years, and the monthly means for the twenty 

 years are remarkably consistent. 



The progression of the value of P, in the course of the year, follows 

 approximately the sine of the sun's meridional altitude, and the 

 empirical formula 



P = 10 cos z-0-91 



gives a close approximation to the values shown in the tables, if a 

 " lagging " of eight or ten days is allowed in reckoning the sun's 

 place. 



