connection between barometric pressure and atmospheric temperature,
it seems not unlikely that the one effect plays some part in the causa-
tion of the other.
The striking correspondence between the period of occurrence of the maximum of the second-order component of the gradient for westerly winds (London Aberdeen) and the corresponding components for tem- perature, leads naturally to the suggestion that the periods of maximum and minimum of the second-order component of temperature are related to special barometric conditions. They may, and indeed must, be connected with some stage in the process of annual migration of the centres of high and low barometric areas, which is the result of the unequal distribution in longitude of the land areas in the northern hemisphere. A cursory survey of the mean monthly distribution of pressures in Dr. Buchan's volume of ' Bartholomew's Atlas,' shows that April is an exceptional month of transition, but up to the present we have not found any means of representing the successive stages of the transition in such a way as to enable a comparison to be made between them and the second-order curve of temperature.
The data used in this paper have been compiled in the Meteorological Office, or extracted from the various publications in the library of the office. The harmonic analysis has also been carried out there under the supervision of Mr. 11. H. Curtis.
Summary.
The harmonic analysis of the curve of mean atmospheric temperature at stations in the British Isles shows that there is, compounded with the primary solar variation, a secondary half-yearly meteorological variation.
The effect of this variation is to moderate and lengthen the winter, and to intensify and shorten the summer.
The second-order curve which represents it, has maxima which oscil- late, in different years, over the first ten days of February and August respectively, and minima which oscillate over the first ten days of May and November.
With some exceptions, the effect is generally apparent in a single year's observations, and, owing to its varying position, its curve ap- pears to have a larger amplitude in the analysis of the temperature curve of a single year than in that of the mean curve of a number of years.
It is independent of the relative frequency of occurrence of cyclonic and anticyclonic weather, and of the relative temperature of the air during the prevalence of these different types of weather.
It is partly due to a periodic variation in the relative frequency of occurrence of the "cold," "warm," and "temperate" winds, but is
