Predicting Atlantic Tropical Cyclone Seasonal Activity in April

Abstract

Seasonal hurricane forecasts are continuing to develop skill, although they are still subject to large uncertainties. This study uses a new methodology of cross-correlating variables against empirical orthogonal functions (EOFs) of the hurricane track density function (HTDF) to select predictors. These predictors are used in a regression model for forecasting seasonal named storm, hurricane, and major hurricane activity in the entire Atlantic, the Caribbean Sea, and the Gulf of Mexico. In addition, a scheme for predicting landfalling tropical systems along the U.S. Gulf of Mexico, southeastern, and northeastern coastlines is developed, but predicting landfalling storms adds an extra layer of uncertainty to an already complex problem, and on the whole these predictions do not perform as well. The model performs well in the basin-wide predictions over the entire Atlantic and Caribbean, with the predictions showing an improvement over climatology and random chance at a 95% confidence level. Over the Gulf of Mexico, only named storms showed that level of predictability. Predicting landfalls proves more difficult, and only the prediction of named storms along the U.S. southeastern and Gulf coasts shows an improvement over random chance at the 95% confidence level. Tropical cyclone activity along the U.S. northeastern coast is found to be unpredictable in this model; with the rarity of events, the model is unstable.