center, a hurricane of more limited extent, and a tornado with a path only one thousand feet in width—all these must be properly classified before any system for the prediction of tornadoes could reach even approximately verified results. The number of tornadoes occurring in the region east of the Rocky Mountains was one hundred and seventy-two in 1884. The average is more than a hundred yearly. Whenever a tornado occurred, blanks specifying the questions to be asked and the method of investigation to be followed were at once sent to the postmaster or to the principal city or town officers in the vicinity of the scene of destruction, asking for facts, and requesting that the observer would kindly volunteer to continue a series of easy meteorological observations for the Signal Service, involving no more outlay of time than the reporter could conveniently spare. The result of this clever device was that the Signal-Office now has upon its books the addresses of more than fifteen hundred reporters in all parts of the country east of the Rocky Mountains, and especially in those States most frequently stricken with destructive tornadoes. Having thus organized a system whereby something like verification of tornado predictions could be attained. Lieutenant Finley then divided the territory east of the Rocky Mountains into eighteen districts. lie at once began making predictions in these districts, thus reaching a series of recorded averages whereby the results became more definite and local until the close of 1884 and the summer of 1885. The predictions in the latter year have become so effective that during the summer of 1886 it is hoped that, by means of signals, hundreds of lives and much valuable property will be saved.
The most striking examples of the knowledge attained regarding tornadoes are found by reference to the following statistics: During 1884, 3,228 predictions unfavorable to tornadoes were made, and of these, 3,201 were verified by reports sent in by tornado-reporters, who are instructed to carefully scan the newspapers as well as note the meteorological phenomena in a given neighborhood. When the conditions are unfavorable for the development of tornadoes, there are no unusual contrasts of temperature, the areas of warm and cold air are neither great nor well defined northward and southward, the winds are variable and not very strong, and the distribution of pressure is about normal.
When we consider that more than a hundred tornadoes occur annually, it follows that the successful prediction of safety for eighteen districts is a very satisfactory indication of the advance made by this science. The result is certainly practical and valuable, as with proper signals shown at telegraph-stations the inhabitants of Kansas, Missouri, and other States, can go to their work free from anxiety and not subject to false alarms on the appearance of every harmless thunder-storm. The predictions of safety are therefore particularly valuable in States in which tornadoes are frequent. When, however, we examine the
