The destructive impact tornadoes have on communities has sparked interest in predicting the risk of impacts on seasonal time scales. Here, the authors demonstrate how to build statistical models for predicting tornado rates. They test the models with tornado counts accumulated over a 45-year period aggregated to counties in the State of Oklahoma and to cells in a latitude/longitude grid across a large portion of south central United States. The spatial model provides a fit to the counts, which includes terms for the spatial correlation and the population effect. A space-time model not only provides a similar fit to annual counts but also includes a term for a time-varying climate factor. This work contributes to methods for forecasting severe convective storms on the seasonal time scale. Keywords: climate, risk prediction, space-time model, statistical model, tornadoes. 1 INTRODUCTION Seasonal climate forecasts are now a matter of routine. Predictions of how much rain and heat can be expected during the summer are issued during spring by weather agencies across the globe, by region and by country. Even single seasonal predictions of, say how many hurricanes can be expected along a coastline are available and accurate enough to warrant attention by the property insurance industry. However, what is missing from the suite of seasonal forecasting products are long-range forecasts of severe convective storm activity. Potentially useful skills (accuracy above random guess) at predicting tornado activity prior to the start of the season has been noted recently [1,2]. However, given the large gaps in our knowledge of how climate influences severe weather and the dearth of methods to forecast it on the seasonal scale, basic and applied research is needed, which focuses on statistical modeling, diagnostic understanding, and methods to predict. A major impediment to issuing seasonal convective storm forecasts is that the events of interest are too small (e.g. tornado) to be resolved within the current dynamical forecast models. Long-lead predictions of severe weather environments can be made with dynamical models but the necessary conditions do not sufficiently distinguish between days with and without tornadoes.An alternative approach is to fit a statistical model to a historical tornado database. Climate patterns related to active and inactive seasons provide the essential information to make predictions. However, population growth and changes to procedures for rating tornadoes result in a heterogeneous database. Various methods for dealing with data artifacts have been proposed [3][4][5] with most assuming a uniform region of activity and estimating occurrence rates within a subset of the region likely to be most accurate. For example, tornado reports are