HOURLY OBSERVATIONS OF AIR TEMPERATURE AND PRESSURE. 
625 
In the third component, from April to August, /Xg is always in the first quadrant, 
or between midnight and 2 a.m. In September and October it changes rapidly to 
the third quadrant, in the last of-these months 40 per cent, of the values being found 
in the second quadrant, and in November the value is established in the third quad¬ 
rant, that is between 4 a.m. and 6 a.m., and there it remains till the end of February. 
In March /Xg falls back to the first quadrant, 7 5 per cent, of the values being in the 
second quadrant and the rest in the first quadrant, with the exception of a single 
case which falls back to the fourth quadrant, or before midnight. The variation in 
any month of the five-years mean from the mean of 20 years in no case exceeds 5°, or 
20 minutes of time, and the mean variation for all months is 2°T, or 8|- minutes. 
The fourth component shows double maxima and minima values of /x^, the former ' 
occurring at the two equinoxes, the latter at the two solstices. In the winter 
months, November to January, the greatest frequency is in the first quadrant, 
between midnight and 1.30 a.m., with a considerable number before midnight or in 
the fourth quadrant, and a few in the second quadrant, or after 1.30 A.M. Again 
in May, June, and July the values of /x^ are chiefly in the first and second quadrants, 
with a few cases in the fourth quadrant, wdiile the values in June invariably fall in 
the first quadrant. From February to April, and again from August to October, the 
greatest frequency of /Xj, is in the third quadrant, or between 3 a.m. and 4.30 A.M., 
with deviations in the one direction to the second quadrant, and in the other to the 
fourth quadrant. In March and September 95 per cent, of the values fall in the 
third quadrant; February and April on the one side, and August and October on 
the other, being months of transition, between the equinoctial maxima and solstitial 
minima. The largest variation of the five-years mean from that for 20 years, for 
any month, is 10°, or 40 minutes of time, and the average for all months 4°'3, or 
17 minutes. The absolute values of the coefficients being very small, the 
maximum being less than j°, and the average being less than iwth degree Fahrenheit, 
the computed values of are necessarily far more liable to error than those for the 
other components, and it is rather a matter for surprise that the variations should be 
so small, than that they should reach the actual quantities mentioned. 
The values of p in all cases depending on that of the angle whose tangent is qojq, 
‘ the consistency of the value of p implies consistency in the ratio of to q. At the 
same time as P \/{p^ + q^) a considerable variation of the absolute value of P is 
quite compatible wdth consistent or invariable values of p, and such variation in P 
will not tend to discredit the evidence in favour of such values of p being trust¬ 
worthy. 
It may here also be noticed that as the value of P may obviously be expressed as 
' p/sin X, where tan x = pjq, it follows that if the value of x, and consequently the 
I epoch of maximum phase be regarded as fixed, for any time or place, then P wdll 
simply vary as p. 
On the whole it may be regarded as sufficiently established that the computed 
MDCCCXCIII.—A. 4 L 
