,S5G DR. C. CHREE: ANALYSIS OF RESULTS FROM THE KEW MAGNETOGRAPHS 
hav 3 in all probability a lunar as well as a solar diurnal variation. The incidence 
of the selected 5 quiet days in the lunar month is largely a matter of accident, and 
consequently the way in which the lunar variation affects hourly values based on the 
selected 5 quiet days must vary from month to month. The lunar influence appears 
however to Ije so small that any uncertainty on this ground must be trifling, even 
in the case of data from a single year, and when we are dealing with data from 
5 or G years, the uncertainty should be no greater than in the case of data derived 
from all days in a single month. 
A further point that is more fully dealt with in § 31 need only be mentioned now. 
The hours to which the curve measurements refer are G.M.T., whereas a true solar 
diurnal variation is connected ])resumably with the true local solar time. The fact 
that noon at Kew is 1-|- minutes later than noon at Greenwich is comparativelv 
unimportant, as the difference is small and affects all seasons of the year alike, but 
the considerable annual range—some 31 minutes—in the “equation of time” must 
be borne in mind. 
§ 19. A really serious difficulty remains to be mentioned. The Kew magneto¬ 
graphs are placed in a room partly underground in the basement of the Observatorv, 
where tlie annual variation of temperature, though mucli less of course than in an 
unprotected upstairs room, still amounts to some 20'’ F. At same seasons of the 
year this implies a somewhat troublesome change of temperature from dav to day, 
but speaking generally there is seldom a difference of more than 1° F., and usuallv 
only a difference of a few tenths of a degree between the mean temperature on 
successive days. It was thus possilfle to allow pretty satisfactorily for the difference 
between the mean temperature of the room on tlie five c[uiet days a month and on 
the days of the absolute observations, from eye readings of a mercury thermometer 
placed under the glass shade containing the vertical-force magnet. Up to the end of 
189G the readings of this thermometer were the only direct source of information as 
to tlie temperature of the magnetographs. During 189G the thermometer had been 
read usually thrice a day, at 10 A.M., 4 P.M., and 10 r.ir. Prior to that it had l)een 
read twice a day, but in the earlier years of the period only once a day. In the end 
of 1896 the late Kew Committee, on my initiative, introduced a thermograph. It 
was so situated and protected as to give readings in close accordance with those of 
tlie mercury thermometer—still read thrice daily—and temperature con'ections based 
on its readings have been ap})lieil from 1897 onwards to the II and Y curi es. In the 
case of the H magnetograph the temperature coefficient is little over ly for 1° F., 
and the temperature correction is very small. In the case of Y the correction is 
distinctly appreciable. An examination however of the thermograph results obtained 
early in 1897—at the season, as it has turned out, when the diurnal variation of 
temperature is least—and a comparison of the annual and semi-annual magnetic 
inequalities for 1896 and 1897, led me unfortunately to the conclusion that the 
neglect of the diurnal variation of temperature in previous years was less serious 
