196 
DR. C. CHREE: DISCUSSION OF KEW MAGNETIC DATA 
many years before the others, and a variety of scales were employed. Then the 
curves of the 209 originally selected disturbed days were not smoothed, while those of 
the ordinary days were. Thus in a good many cases, at the midnight common to a 
disturbed and an ordinary day two readings were taken, at widely different times, 
one on the unsmoothed curve, the other on a smooth pencil trace. In individual cases 
these two midnight readings differed considerably, and this of course influenced the 
balance of the n.c. changes. The difference between the n.c. changes in H on the 
209 days of the original disturbed list and the supplementary list of 33 days may 
appear suspicious, but is easily accounted for. During a large magnetic storm H 
nearly always shows a slight rise at the start. This is usually followed by a fall, 
which goes on until the value has diminished below the normal, sometimes much 
below the normal. There is then a recovery, which may go on at a gradually 
diminishing rate for some days. The ends and beginnings of storms were represented 
by a larger proportion of the 33 than of the 209 days. One of the 33 days showed 
an n.c. change of + 140y. 
In the case of D what the absolute observations suggested was not a real 
instrumental drift, but occasional small discontinuities due probably to movements of 
iron in the building. In the case of H there is confirmatory evidence from the base 
line values that the instrumental creep is in the direction simulating a fall of force, 
but they suggest — 15y per annum as a more probable estimate than — 5y as found 
above. In the case of Y the instrumental creep in reality seems to fluctuate in 
direction. When a sensible change of sensitiveness occurred in the course of the 
year, the tendency to creep seemed more apparent. On individual ordinary days the 
n.c. change in V is mainly a temperature effect. This may in fact be recognised in 
the figures given in Table IY. The four months April to July include most of the 
summer rise of temperature in the magnetograph room, the principal part of the 
annual fall taking place in the four months September to December. The mean daily 
n.c. changes during these two groups of months are by Table IY. : 
From April to July — 2'00y. 
,, September to December +l‘32y. 
The two means will naturally include equal or approximately equal contributions from 
any regular source of drift which is independent of temperature, such for instance as 
might arise from gradual weakening of the magnet. If we ascribe the difference 
between the two four months’ means obtained above solely to temperature, and take 
the known temperature coefficient, viz., 12'5y per 1° F., then assuming the rise and 
fall of temperature in the two groups of months equal, we find for its amount 
(3-32/2) x (120/12-5) = 16"0 F. 
This is not far from the truth. The annual range in reality usually exceeded 20° F., 
but the rise usually began in February and continued throughout part of August. 
