Method of Treating Kew Observations.
§ 16. In discussing the Kew observations I have in general employed a method differing from the grouping system of Exner and Elster and Geitel, and have also treated the several series separately. It is clear from data mentioned by Exner that the potential gradients for individual members of his groups varied in some instances largely from the mean; and it was soon obvious that the same phenomenon would present itself if any similar treatment were applied to the Kew results. This is undesirable, because by varying the limits of the groups the accordance of the results with a particular formula may be much improved, or the reverse. However impartially, so to speak, the lines may be drawn, there is undeniably a risk of introducing some fictitious result; and no critic can feel that he is in a position to judge of the results until he has examined for himself the circumstances of the grouping, a labour he naturally shrinks from. Again a wide range of such an element as vapour density can be obtained at a particular place only by combining results from all seasons of the year. This brings us to a second question. Electrical potential gradient has like vapour density, sunshine, and temperature, a large annual variation, only, unlike these elements, it is highest in winter. It is thus obvious that when observations from all seasons of the year are treated promiscuously, there is almost sure to be a marked association of high potential with low vapour density, little sunshine, and low temperature; and a judiciously selected formula which makes potential diminish as any one of these elements increases is certain to show some approach to agreement with observation. It is thus desirable to compare together observations from a limited portion of a year, or, even better, from the same season of a series of years. Similar considerations show an advantage in treating separately results from different hours of the day. The isolation of particular seasons and hours has the disadvantage of reducing the number of observations compared together. This is, however, partly compensated for by the greater homogeneity of the material. It also enables one in sonie cases to compare readily the mean potential gradients which answer at different seasons or hours to like values of some one meteorological element (see § 23).
Anticipation of some Criticisms.
§ 17. The Kew observations were not limited to quiet, comparatively cloudless days, in the same way as the observations of Exner and Elster and Geitel seem to have been. It may thus not unlikely be supposed that the Kew results are affected by a variety of disturbing causes, which diminish their intrinsic value and their suit-
