EARTH’S MAGNETISM AT PAVLOVSK AND POL A (1897-1903). 
299 
lunar hour 0 for all observatories ; the necessary correction for this is indicated 
later.* 
The length of a lunar day is not quite constant, but its mean value is approximately 
24 h 50 m in solar time; for the present purpose it will be assumed to equal 25 h of 
solar time (which seems much preferable to the adoption of 24 h , as has often been 
donef). As the hourly differences to be copied out are given at intervals of one solar 
hour, the lunar day will, for the first stage of the work, be regarded as subdivided into 
25 equal periods of duration of one solar hour, and the last 25 columns of the “ lunar ” 
sheets should be headed 0 h to 24 h . In copying out the hourly differences, therefore, 
the first value written down on any row (each of which relates to a single lunar day) 
is that one which corresponds to the local civil time written in the second column of 
the row; after this the next 24 consecutive hourly differences are written in order. 
The times of lunar hour 0 in the second column generally differ by successive amounts 
of 25 h , and where this is so the hourly differences run on consecutively from one row 
to the next; occasionally the difference is only 24 h , however, and in this case one 
hourly value has to be set down twice, once under column 24 h of the former row, and 
again under 0 h of the following row. To ensure that the right set of hourly differences 
is being copied out, the time of the last entry in each row of the lunar sheets should 
be checked. In this re-arrangement it is important, for convenience and accuracy, 
that the columns of the lunar sheets should be wide enough to allow of the preserva¬ 
tion of the separation of the positive and negative values of the hourly differences, 
the former on the left and the latter on the right hand of the columns; this facilitates 
the formation of monthly and group sums, and obviates the use of signs. Each lunar 
sheet should contain the lunar days of a single calendar month, occupying about 30 
rows; a few hours at the beginning or end of the month may be taken from the 
preceding or following calendar month in order to make each lunar day complete. 
§ 4. When all the hourly differences for the period under investigation have been 
copied out on to the lunar sheets these should be carefully scanned for the purpose of 
picking out any very disturbed days for rejection. This assists also in the detection 
of any mistakes which have been made in the previous part of the work, as serious 
mistakes in forming the hourly differences in copying is generally apparent from the 
“ run ” of the numbers. As regards the rejection of disturbed days, I have adopted a 
very conservative policy; the proceeding must, in any case, be of an arbitrary nature, 
and a criterion which will mark off a given proportion of days must vary in different 
years, at different observatories, and with different elements (even at the same time 
and place). The period covered by this investigation (1897-1903) was intentionally 
chosen during a sunspot minimum, when quiet magnetic conditions prevailed, so that 
in a more disturbed period it would be necessary to reject a rather larger proportion 
* P. 302. It is assumed that the hourly values are tabulated, as is usual, according to the local civil 
time of the observatory. 
f As, for instance, at Batavia, Bombay, and in all Sabine’s work on the lunar magnetic variation. 
2 Q 2 
