185 



reducing the observations, a mean value was applied for P in the formula 2mr~ 8 (l + Pr~ 2 ...), expressing 

 the deflecting force at distance r due to a magnet of moment m. In calculating this mean value, the 

 observations of June 3, 9 and 15, July 4 and 22, and December 17, 1903, and that of January 26, 1904, 

 were omitted, as the differences between the results from the two deflection distances (30 and 40 centims.) 

 on these occasions were clearly abnormal. The table gives the values of H for each observation and the 

 mean derived from the observations in each month ; it also gives the values deduced from the individual 

 observations for the magnetic moment, m, of the collimator magnet at C. The small variability in the 

 values deduced for m is a marked tribute to the care with which the observations were taken. Another 

 gratifying feature is the constancy of m from beginning to end of the observations. Though in regular 

 use for eight months, the magnet shows no certain loss of magnetic moment. 



The observations were taken at somewhat variable hours. Afternoon hours prevailed, but the mean 

 hour of observation would vary considerably for the different months. No direct observations were made 

 of the diurnal change of horizontal force, and an attempt to get a general idea of its character by grouping 

 the data according to the hour of observation led to results which were too irregular to inspire confidence. 

 If the diurnal variation had been large, the method adopted could hardly have failed to indicate it clearly. 

 The fact that the variations, whether regular or irregular, were not very large is fairly obvious from the 

 observations themselves. If we omit the observation of June 3, which presented some abnormal features, 

 individual values of H varied only from 0-25655 to 0-25754. A range of 99y from 36 observations is not 

 compatible with numerous large disturbances. We may conclude that whilst appreciable corrections, 

 differing slightly from month to month, are probably required to reduce the monthly means in Table III. 

 to the mean value for the 24 hours, the uncertainties are still comparatively small. The monthly means 

 suggest, on the whole, that the horizontal force is decreasing, but no great weight attaches to this 

 conclusion. 



5. Diurnal Variation of Declination. Hourly readings were taken of the collimator magnet of the 

 Unifilar Magnetometer, with scale erect, throughout four periods each of five days. The periods 

 commenced and ended with midnight, their dates being December, 1903, 6 to 10 and 21 to 25, and 

 January, 1904, 4 to 8 and 19 to 23. Throughout each period the setting of the azimuth circle remained 

 unaltered, the observer simply noting the scale division of the magnet. Table IV. gives the hourly 

 readings as recorded in scale divisions, 1 scale division representing an angle of l'"80. The table also 

 shows the ranges for each day, uncorrected for non-cyclic change, both in scale divisions and in minutes of 

 arc, and, finally, the hours of occurrence of the principal maximum and minimum. There is one feature in 

 Table IV. that will appeal to everyone familiar with magnetic data, and that is the remarkable absence of 

 disturbance. During the whole twenty days the daily range did not fall below 5' 9 nor rise above IT'S, 

 the arithmetic mean derived from all the days being 8' -7, and this mean value being exceeded on nine of 

 the twenty days. The hour of the principal maximum occurred once at noon and four times at 1 p.m. ; 

 on the other fifteen days it occurred either at 2 p.m. or 3 p.m. The hour of minimum was more variable, 

 this turning point being seldom very clearly marked. 



6. Table V. shows the mean diurnal inequality derived from each five-day period, the results being 

 uncorrected for non-cyclic change, and also the mean diurnal inequalities corrected for non-cyclic change 

 for the ten December, the ten January, and the whole twenty days. The results are all in minutes of arc. 

 The ranges and the hours of occurrence of the principal maximum and minimum are also recorded. The 

 inequality for the whole twenty days was derived independently of those for the two ten-day periods, and 

 the non-cyclic corrections to the three sets of figures were also applied independently. As the reductions 

 were not carried beyond O'-Ol, there is naturally at some hours a difference between the twenty-day 

 inequality and the arithmetic mean of the ten-day inequalities in the last figure retained. 



The diurnal inequality for the whole twenty days is shown graphically in fig. 1, p. 187. For a curve 

 based on only twenty days' observations it is, on the whole, extremely regular. There is clearly a little 

 irregularity between 4 a.m. and 9 a.m. This arises from the variability in the hour of the minimum, which 

 leads to the curve being abnormally flat. There is a nearly stationary part of the curve from 8 p.m. to 

 11 p.m., and possibly observations for individual months derived from a series of years might show a 

 poorly developed secondary maximum and minimum, an hour or two apart, in the late evening. 



2 B 



