the period of the year and to the difference on any day hetween the
observed mean daily temperature for the nine years and the tempera-
ture which would be represented on that day by the first order sine
curve obtained by analysis of the twenty-five-year curve. The eight
curves so obtained will be spoken of as the temperature curves for
each wind; but it must be remembered that they do not represent
observed temperatures, but differences between mean observed temperatures and temperatures given by the first-order curve at any
given period of the year.
The term " for each wind " also requires some amplification. Owing to the marked effect of local geographical conditions at any station, the recorded direction of the wind does not necessarily give a very accurate estimate of the true quarter from which the station was obtaining its air supply, and it is of course the source of air supply which will affect the atmospheric temperature at any station. The days were, therefore, not classified according to the direction of the wind recorded as having been observed at Kew, but according to the direction of the gradient (high to low) at right angles to the isobars in the neighbourhood of the station, and a curve has been called the curve for the east wind when it really represents a curve for a southern gradient, and so on for other directions. It must be observed that the term " east " is not quite accurate, since the direction of air supply for a southern gradient will not be from due east, but from a little north of east. However, as we are only dealing with eight points of the compass, the direction of the wind when the gradient is to the south will be described with sufficient accuracy as an east wind.
In Diagram 3 the temperature curves for each wind are given. It will be seen that each curve tends to resemble in some important respects, and in kind, if not in degree, the second-order curve for the twenty-five- year means. For example, the east wind temperature curve has a minimum in the beginning of May, and maxima in August and February ; but these are not equal minima and maxima, and the curve is obviously not simply harmonic. The north-east wind has a mini- mum in spring and autumn and a maximum in winter and summer, although the dates do not exactly correspond with those of the east wind. The curve for the north wind is irregular but has a distinct absolute maximum at the end of July. The curve for the north-west wind has its maxima in February and at the end of July, and minima in spring and autumn, and so on. Thus all the winds have curves whose characteristics suggest those of the second-order curve, although they are not identical with them, and the effect represented by the second-order curve is still apparent in the temperature curves for each separate direction of air supply, and consequently is to some extent independent of the quarter from which the air is supplied.
