storms. The annual death casualties from tornadoes have averaged, in the last 18 years, 102 annually; but it is believed that the death rate from lightning is greater than that from tornadoes, since during March to August, 1890, the names of 110 are on record who have lost their lives by lightning, although the data are incomplete, especially as regards the southern states. These statistics cannot be passed by lightly, however, and it is doubtful if in the main they are much in error. By them it appears from five years' record that the average annual death rate by lightning in the United States is 3.8 per million of inhabitants, or 0.2 above the average. In Sweden, for sixty years, the average has been 3.0; in France, for forty-nine years, 3.1; in Baden, for seventeen years, 3.8; and in Prussia, for fifteen years, 4.4 per million. Other figures, given by a life-insurance agent in St. Louis, which the author claims to have compiled with great care, place the average annual rate of death from lightning in the United States at 206, being more than double the deaths from tornadoes. It must be understood that these figures are not vouched for, and must be very cautiously received, as originating with companies interested pecuniarily in the statistics. On the whole, therefore, it may be safely assumed that tornadoes are not so destructive to life as thunder-storms. * * * * * Professor Thomas Russell has formulated a method for prediction of cold waves. They always occur after "lows" and before "highs," and different cold waves vary in extent from three "units" to sixty. A "unit" of temperature-fall is taken as a fall of twenty degrees over an area of 50,000 square miles. The temperature-fall curves in the United States are approximately elliptical in shape. Perfect ellipses represent actual temperature falls with an error not exceeding six degrees in {52} most cases. These fall lines are intersections of planes with a cone which graphically represents the totality of temperature-fall, the contents of the cone being equal to the area of its base multiplied by its altitude, which is the greatest fall in temperature at the center of the cold wave. A formula has been devised, based on 127 special cases, representing the amount of fall in terms of the amount of barometric depression in a "low," and the amount of excess if a "high," and the density of the isothermal lines in the region. From proper consideration of the type of low area, shape of isobars, and position of the long axis, definite conclusions can be drawn as to the subsequent shape of the elliptical twenty-degree temperature-fall area and its position.