is a considerable deviation from it in the upper levels. In the temperate zones the normal vertical temperature gradient is only about 5.40° Centigrade, though it may be considerably more or considerably less according to the circumstances. It may be generally said that, except in restricted regions, the air does not cool as fast in going upwards as it should if it were caused by mere vertical expansion. The upper levels of the air are too warm; warmer than they should be if that law prevailed. In the temperate zones they are very much too warm, and that is why the vertical gradient is less than it should be according to that law. The fact is that the warm masses of air which flow from the tropics towards the poles retain their heat above what they should have for the given latitude, and in that way the upper levels of the atmosphere are maintained at a considerably higher heat than would be expected. When the air has once cooled to about 70° below zero, Centigrade, it seems disinclined to cool much further, and in the levels from 12,000 to 16,000 meters high there has been discovered a tendency for the air to be somewhat warmer than it is in the levels below, say from 8,000 to 12,000 meters high. It is generally thought that this phenomenon is due to radiation in some of its effects, but it is still a subject of discussion. If we should assume as the average vertical gradient for the entire atmosphere a rate of about 7° Centigrade per 1,000 meters then we should find that the temperatures in the tropics fall off too fast, and in the temperate zones too slow to conform to this average gradient. Now the mathematical law shows that if the lower levels of the atmosphere are relatively too warm for the upper levels there will be a westward drift as in the tropics, and if the upper levels are too warm relatively for the lower levels there will be an eastward drift as in the temperate zones. Speaking a little more broadly still, in order to avoid discontinuity, that is to say, changes by jumps in the atmosphere as regards the barometric pressure at the different levels, since the warm air has less density than the cold air, it follows that the warm air must move faster over the surface of the earth than does the cold air. Hence it is that in the tropics the air is too warm for its altitude, and it must move off faster than it otherwise would in the tropics. The westward drift in the lower levels compensates for this excessive temperature, and in the upper levels of the temperate zones the excess of motion compensates for the higher temperature. We find exactly the same principle working in the formation of hurricanes and tornadoes. Hurricanes develop in the northern hemisphere in the late summer and early autumn, and this is the season when the cool air of the northern latitudes begins to spread southward towards the equator as the sun begins its southward march into the southern hemisphere. At first the cool air flows over the warm air in the higher levels. This in a general way increases the vertical tern-
