the dates just mentioned. The cold periods all require the presence of the northerly or easterly types; the warm periods, either of the southerly type in winter, or of anticyclones in summer; while wet or broken periods indicate the recurrence of intense cyclones of any type. Returning to our old illustration of a globe surrounded by a circulating atmosphere, we can readily suppose that at the same date every year, when the sun is in the same place, the motion of the air will tend to reproduce the same kind of eddies in the same localities. VALUE IN FORECASTING. But now we have to consider how the knowledge of these recurrent periods can be utilized in forecasting. In our last chapter on Diurnal Variations we called attention to the nature of the daily period of heat. In this, the most obvious of all meteorological phenomena, we found that though there is a powerful heating influence present every day, still that other causes are sometimes so powerful as to obliterate or invert the action of the sun. As a consequence of this we cannot affirm absolutely that the night will be colder than the day, though, of course, such is generally the case. If we were to attempt to forcast the heat at any hour by reference to the mean curve of diurnal range, we should sometimes give most erroneous forecasts; if, on the contrary, we looked at the chart for 8 a.m., we could often safely predict that the ensuing night would be warmer than the day. From the temperature diagrams which were there given, we also drew the important inference August 6 to 11.-A cold period; Buchan's fifth. August 12 to 15.-A hot period; Buchan's second. No prognostics are associated with these two latter periods. September. The easterly and northerly types are rare during this month; the gales or broken weather at the equinox are almost invariably of the westerly type. About the 30th a fine period is experienced for a few days-the "Indian Summer" of North America. October. About the second or third week a spell of the easterly type of moderate intensity is common. October 18.-A fine quiet period about this time"St. Luke's Summer." This and the other summers which occur at this season have sometimes been stated to be due to the liberation of heat during the condensation of vapour, and formation of ice, which begins to take place on a large scale in the polar regions soon after the autumnal equinox. According to this theory, the opposite phenomenon of cold in April and May is supposed to be caused by the absorption of heat due to the melting of ice. Both ideas are purely fanciful. The spring cold we have already explained; the autumn summers are due to the recurrence of tranquil periods at that season. November 6 to 12.-A cold spell; Buchan's sixth, associated with the northerly type. The 11th is " St. Martin's Little Summer," popularly considered in the Mediterranean to be a period of warm, quiet weather. December 3 to 9.—A warm period; Buchan's third. The general explanation of all these periodicities is identical. They all depend for their origin on a tendency of certain types of pressure-distribution to recur about the dates just mentioned. The cold periods all require the presence of the northerly or easterly types; the warm periods, either of the southerly type in winter, or of anticyclones in summer; while wet or broken periods indicate the recurrence of intense cyclones of any type. Returning to our old illustration of a globe surrounded by a circulating atmosphere, we can readily suppose that at the same date every year, when the sun is in the same place, the motion of the air will tend to reproduce the same kind of eddies in the same localities. VALUE IN FORECASTING. But now we have to consider how the knowledge of these recurrent periods can be utilized in forecasting. In our last chapter on Diurnal Variations we called attention to the nature of the daily period of heat. In this, the most obvious of all meteorological phenomena, we found that though there is a powerful heating influence present every day, still that other causes are sometimes so powerful as to obliterate or invert the action of the sun. consequence of this we cannot affirm absolutely that the night will be colder than the day, though, of course, such is generally the case. As a If we were to attempt to forcast the heat at any hour by reference to the mean curve of diurnal range, we should sometimes give most erroneous forecasts; if, on the contrary, we looked at the chart for 8 a.m., we could often safely predict that the ensuing night would be warmer than the day. From the temperature diagrams which were there given, we also drew the important inference that, because we do not see a diurnal range every day, we must not infer that there is no such thing as a diurnal solar influence. If, then, such a powerful influence as the direct rays of the sun can be so easily masked, we can readily understand that a weaker influence, such as the declination of the sun on any particular day, can readily be obliterated. We can safely say that the change in the altitude of the sun is of secondary importance, because we see every day great changes in the distribution of pressure, which are certainly in no way related to the seasonal change in the declination of the sun. We need not, then, be surprised that the types of heat or cold do not recur absolutely every year, only that there is an undoubted tendency to do so. When once we have realized this, we can easily understand the following statement of the use of a knowledge of recurrent annual types in forecasting. A forecaster is not justified in saying that any period will occur absolutely; still, when about the time of its usual recurrence the synoptic charts show sigus of the expected type, then the forecasts for a few days ahead can be issued with greater confidence. For instance, suppose that about the 6th of November the charts begin to show traces of the northerly type, then-but not before-there would be good grounds for saying that a period of cold weather, which usually occurs at this season, has already set in, and may be expected to last for five or six days. The forecaster is thus enabled to issue a much longer forecast than he can as a rule safely attempt. CYCLICAL PERIODS. By these we mean periods which run through their whole course in any time other than a day or a year. Many investigators have thought that they have detected traces of periodicities of temperature of about twelve days and of 25.74 days, the latter apparently connected with the time of the sun's rotation. Others have endeavoured to detect periodicities of rain or heat for longer epochs, especially one of 11.1 years, which would coincide with a period of sun-spots. As this is the one of most importance and greater interest, and as a discussion of it will serve to illustrate the whole nature of periodicities, we shall confine our attention to a short notice of this cyclic period only. SUN-SPOTS AND WEATHER. Ever since the year 1775, we have more or less complete records of the relative extent of black spots on the sun's surface. These records show a most unequivocal recurrence of sun-spot maxima at intervals of about eleven years; but the actual amount of surface covered at each maximum is very irregular. In the lower part of Fig. 63 we give a reduction of diagram which shows the relative extent of black spots on the sun as plotted by Professor Balfour Stewart. If we were to draw over this curve another which showed the mean daily range of magnetic declination for the same year, we should find that there was an unmistakable similarity between the |