248 COMPARISON OF ARCTIC AND ANTARCTIC DISTURBANCES. 



curves of .adjacent days, if quiet, or to a regular diurnal variation derived from a number of days. The 

 latter alternative is, of course, the more satisfactory theoretically ; but there is this difficulty, which I have 

 dealt witk elsewhere, that the regular diurnal .inequality which one gets depends on the nature of the days 

 from which it is derived. For instance, taking means from 11-year results at Kew, the departure from the 

 mean Declination for the day at 2 p.m. on a representative January day is : 



+ 2' - 21 on the Astronomer Royal's selected quiet days, 



+ 2' 66 on ordinary undisturbed days, 



+ 4' 88 on the average highly disturbed day. 



For simplicity, we may suppose the secular change and sun-spot influence non-existent, though in reality 

 these are complications which have to be reckoned with. Let us suppose that on a certain disturbed 

 day in January the Declination at 2 p.m. departs from the mean for the month by + 3' 66. Is the 

 "disturbance" + 1''45, +1'-00, or -r-22 1 ! The +4'-88 departure from the daily mean on the 

 representative disturbed day in January, be it noted, represents a regular diurnal inequality, or at all 

 events something which we have no present means of distinguishing from a regular inequality. Its excess 

 over the ordinary day value may, of course, indicate a tendency for a particular phase of disturbance to 

 occur at a particular hour of the day, but it may mean that the causes operative in producing the regular 

 diurnal inequality are for some reason e.g., increased conductivity in the upper atmosphere more 

 effective on disturbed days than others. 



This source of uncertainty is equally present when the normal curve is derived, as BIRKELAND seems to 

 have derived it, by reference to a day or days adjacent to the disturbed one. The adjacent days may be 

 of the very quietest type, or may themselves be considerably disturbed. 



There is a further complication in that disturbance has sometimes a tendency to be associated with a 

 temporary alteration in the value of magnetic elements, which disappears gradually like after-strain in a 

 metal. Horizontal Force, for instance, is sometimes very considerably depressed for days after a really 

 large disturbance. Thus it may make all the difference in the world to one's decision as to whether at a 

 particular hour the element is above or below the normal value, if one happens to take for comparison the 

 day after a disturbance rather than the day before. In one instance I observe that Prof. BIRKELAND 

 noticed the occurrence of this precise source of uncertainty. 



In the case of the Antarctic curves, the uncertainties which exist elsewhere are mostly much enhanced. 

 We have seen in Chapter III, Tables XII and XIV, that disturbance exerts an unusually marked influence 

 on the amplitude of the regular diurnal inequality. The Antarctic D and H curves were always sensibly 

 disturbed, and the disturbance was usually sufficient to obscure, if not to totally conceal from the eye, the 

 trend of the natural diurnal variation. The diurnal variation of V would very likely have been readily 

 recognisable if the scale-value had been as low as 5y throughout, and the temperature coefficient had been 

 small ; but, as matters stand, mere optical inspection of the curve tells merely whether disturbance has 

 been specially active or not. The days of Prof. BIRKELAND'S choice lay entirely in the months October, 

 1902, to March, 1903, and so came in the Antarctic Summer, when the magnetic variations and the 

 temperature changes in the hut were both at their maximum. 



4. There is a rather different kind of uncertainty to which BIRKELAND refers more than once, e.g. p. 64, 

 viz., the uncertainty in the estimate of the hour at which a particular movement took place. When one 

 is dealing with stations not too far apart, especially when furnished with magnetographs of the same type 

 such, for instance, as Stonyhurst, Falmouth and Kew this source of uncertainty may usually be largely 

 eliminated, provided there are any rapid movements which are shown at all the stations. The curves 

 from the different stations are sufficiently alike to render identification of corresponding points easy. The 

 accuracy of one's identification of any one peak can be verified by reference to other peaks. Thus while 

 one cannot decide what the error actually is in the times shown on the curves of any one of the stations, 

 one can usually determine approximately the difference between the errors at the different stations, and 

 so make sure that the comparisons made refer to one and the same absolute time within a minute or so. 



When we have rounded curves like that on p. 247 representing a slowly varying disturbance, a mistake 

 even of several minutes is unlikely to make much difference in one's estimate of a disturbance, provided 



