TEE.EESTEIAL MAGNETISM AT KEW OBSEEVATOEY. 
99 
effort. The selection of the original 5 clays a month from consideration of the H 
ranges was due to its greater simplicity. If there had been in each month 5 and only 
5 days of magnetic character “ 2,” one would naturally have preferred to take them. 
But some months had no days of character “ 2 ” at all, while others had considerably 
over 5, and there was no obvious simple way of selecting 5 days from each month as 
the most disturbed. One might, of course, have selected every day of character “ 2,” 
but this would have given enormously more weight to the highly disturbed than to 
the quiet years. 
Before discussing the results of the enquiry it is desirable to consider the nature of 
the data. In a large disturbance, or what is usually called a “ magnetic storm,” all the 
elements are as a rule so disturbed that one unhesitatingly assigns a “ 2 ” on inspection 
of the trace of any one of the elements, whether D, H, or even V. But there are many 
days to which I have assigned a “2” which would get only a “ 1 ” if the record of the 
most disturbed element—-whether D or H—were left out of account. In the present 
investigation, as explained above, the selection of days depended entirely on the H 
range. It was thus certain a priori that in some months one or more of the selected 
days would only have a character figure “ 1,” while days not selected would have 
character “ 2.” Still it was abundantly clear that the character figure of the 
representative day n would be much above the average in every month, and that when 
a large number of months were included, the mean character figure from column 7i 
would be much larger than any other. The method was thus certain to give a well- 
marked pulse with its crest in column n. 
Again, there is a tendency for disturbed days (or days of large H range) to occur 
in groups of 2 or more rather than singly. Suppose we have 3 consecutive days of 
character “ 2 ” amongst the 5 selected for one month, all the other days in their 
neighbourhood being of character “ 1 ” or “ 0.” Obviously the character figures from 
these 3 days will occur in each of the 5 columns headed n —2 ton^-2. Thus the pulse 
will not be a wall-sided one confined to day n, but will extend to adjacent days. 
The tendency to occur in groups is however less marked in the case of high character 
figures than in that of large spot areas. Thus we know in advance that we are certain to 
have a pulse, with crest at day n, somewhat resembling that of sunspot areas in 
Table IV., but probably less wide. Five or six days after the crest we may expect 
the pulse to die out, and thereafter, if no period exists shorter than 35 days, we may 
expect the mean character figures derived from the successive columns to show only 
irregular accidental departures from a dead-level value. If a period of m days exists, 
then we may expect a second pulse with its crest in column n-\-m, the height of the 
crest above the surrounding level being small unless the period is a well-marked one. 
§ 22. The results of the investigation appear in Table XI. It includes data from 
the whole eleven years, and from three specified sub-groups of years. The figures in 
the columns headed n — b to n-l-35 represent the corresponding mean values of the 
character figure, the days contributing to each mean being 660 for the 11-year and 
O 2 
