Recent work by the National Severe Storms Laboratory and the Storm Prediction Center has focused on improving our understanding of the spatial and temporal variations in the climatology of severe weather hazards in the United States. This project has heretofore emphasized the climatological aspects of severe weather that are relevant to short-term probabilistic prediction, disaster mitigation, and to the annual cycle of the hazards that are important for public awareness efforts.

The current work expands the scope of the project by looking at the bulk properties of the distribution of tornadoes within the United States for the period 1955-1999. Such a study may prove to be important for a wide variety of users including weather forecasters, the insurance industry, risk managers, and emergency management officials.

The short period of record over which tornado observations have been reliably collected causes numerous problems when trying to make estimations of the statistical properties of the tornado distribution. Perhaps the most important, and difficult, question that has evolved from this study is 'what is the variability of tornado occurrence?' Raw numbers of reported tornadoes have increased from approximately 500 in 1950 to 1400 in 1999. Such large changes make it extremely difficult to assess what value to use as the normal number of expected tornadoes in a given year.

If one can develop meaningful definition of a normal tornado year then the cumulative distribution function can be used to determine how far ahead of, or behind, normal the year is in terms of number of tornadoes.

Another question one might like to answer is 'Is it possible to make seasonal forecasts of tornado occurrence?' An initial attempt would be to make forecasts of the number of tornadoes that are likely to occur over some period of several months. Preliminary results suggest that if years start off relatively slowly (few tornadoes) that they rarely recover to be 'normal' years and never turn out to have large numbers of tornadoes whereas if years start out with large numbers of tornadoes early in the year that this does not guarantee that the total number of tornadoes for the year will also be large.

An attempt is made to characterize the spatial climatology of tornadoes for the current year through a given date relative to the climatological probability of tornado occurrence. An approach such as this provides a more physically meaningful interpretation of the current year's tornado activity than simply counting the number of tornadoes that have occurred in a given state in a given year. Clearly tornado activity is not tied to political boundaries and therefore should not be linked such arbitrary limits.