Higher than this, the temperature decreases at a fairly uniform rate, but more slowly than the adiabatic rate. Although no clouds were visible, yet the relative humidity increased greatly, both during the ascent and descent, near 1500 and 2700 metres, these being about the heights at which cumulus and alto-cumulus clouds usually form. During September 1898 four kite-flights were made on four successive days when an anti-cyclone and a cyclone passed nearly over Blue Hill. This is a rare occurrence, and the mechanism of these phenomena was accordingly studied by Mr. Clayton, some of whose deductions will now be given, illustrated by Plate XIII. Figs. 1 and 2 give the temperature plotted according to height on September 21 in the anti-cyclone, and on September 22, when the barometric pressure was falling, the full lines, as in previous diagrams, indicating observations during the ascents, and the broken lines observations during the descents. It is seen that from the ground the lines all incline upward to the left, indicating a fall of temperature, to a certain height when the lines bend to the right sharply, showing a sudden rise of temperature. Above this, the temperature again falls, but more slowly than at lower levels. The general prevalence of this phenomenon was noted by Welsh in his balloon ascents in England in 1854, and the high kite-flights at Blue Hill show it to be very frequent below 2000 metres. The plane of increased temperature usually determines the height of the tops of cumulus and strato-cumulus clouds. Above 2000 metres other sudden rises of temperature are found during the highest kite-flights. [Illustration: PLATE XIII.--RESULTS OF KITE-FLIGHTS AT BLUE HILL DURING AN ANTI-CYCLONE AND A CYCLONE.] Figs. 3 to 6 show the changes in the various elements during the four days at some of the following levels, viz. near sea-level, 200, 1000, 2000, and 3000 metres. Fig. 3 shows the changes in the barometer at the four levels, from which it is evident that the fall of pressure was greatest near sea-level. Fig. 4 shows temperature changes at the different levels, and indicates that the changes were of the same nature up to 3000 metres. The greatest non-diurnal range of temperature is seen to be at 1000 metres, and it diminishes both at higher and at lower levels. Fig. 5 shows changes in relative humidity at 200, 1000, and 2000 metres. The curves show that the greatest range of humidity was at 2000 metres. There the relative humidity rose from almost zero, in the anti-cyclone on the twenty-first, to saturation at the same level in the cyclone. At 200 metres the change is similar to that at 2000, but is less in amount. At 1000 metres the relative humidity fell until the twenty-second, but then rose rapidly, showing the very dry air at 2000 metres on the twenty-first had descended as low as 1000 metres on the twenty-second.