280 DE. CHAELES CHEEE: SOME PHENOMENA OF 



disturbed day we should obviously have an inequality of a kind with a very large 

 range. 



In the present instance mean monthly inequalities are based on a comparatively 

 small number of days, in no case more than 30 and in one case only six. 



There is thus a risk in all the months of a considerable pseudo-element in the 

 apparent diurnal inequality, and the risk increases when the days of a particular 

 month are specially few but include one or two of outstandingly large absolute ran we. 

 Another possibility that requires to be taken into account is that a particular month 

 of the year may accidentally contain an undue proportion of the most highly disturbed 

 days i.e., a proportion much in excess of what it would have contained if we had 

 been dealing with 100 years instead of 11 and consequently differences may appear 

 between different months which are not really due to true seasonal variation. 



Table V. is intended to elucidate the real chai'acter of the material utilised. It 

 includes data for D as well as for H and V, expressing all in terms of force so as to 

 facilitate comparison. 



8. The first column of Table V. gives the mean absolute daily range of D in 

 terms of ly as unit derived from the quiet days of the 1 1 years, D being the only 

 element for which these ranges have as yet been got out. The second column gives 

 the corresponding mean values from the 209 disturbed days. The next two columns 

 give the largest and least of the daily absolute ranges from the disturbed days. Of 

 the two sets of results for H the first is also derived from the 209 days ; the second 

 gives for the 4 winter and 4 equinoctial months the results derived from the ten days 

 of largest H range in each month. The maxima of range in this case are, of course, 

 the same as for the whole 209 days, and so are not repeated. The first two sets of 

 results for V are exactly analogous to those for H ; the last set was obtained from 

 the years 1892 to 1895 only, these being the years of largest sunspot frequency. In 

 this case seasons only were considered, the means attached to, say, winter allowing 

 equal weight to all individual days of the 4 months November, December, January, 

 and February. Except in this case, seasonal means allow equal weight to the 

 individual months, so that individual days are not equally potent. 



One reason for deriving data from 10 days a month for the winter and equinoctial 

 months, and for these only, was that only 42 of the whole 209 days fell in the four 

 summer months May to August. There was thus room for suspicion that the average 

 summer day might represent more highly disturbed conditions than the average 

 winter or equinoctial day. The regular diurnal inequality is much reduced in 

 amplitude in winter and so more easily obscured by disturbance, and thus a lesser 

 degree of disturbance might suffice to get a winter day classified amongst the 

 disturbed. 



The quiet day D results in Table V. show a nearly constant range from April to 

 August, a small but decided reduction in September and March, with a sharp fall to 

 a minimum in December, and a more gradual recovery in January and February, 



