Though seasonal hurricane predictions have been issued for several years, predictions of net activity across the entire Atlantic fail to provide enough information for public officials, disaster preparedness considerations, and the general public-at-large. For example, suppose a seasonal forecast were issued for the Atlantic indicating a general below average season. Although this forecast may verify, coastal dwellers may be lured into a false sense of security due to a misinterpretation of this forecast as a decreased risk for their area. In such a situation, it only requires one major hurricane to cause catastrophic losses and impinge on an otherwise unsuspecting public; Hurricane Andrew of 1992 provided a perfect example of such an occurrence, as this was an intense, landfalling hurricane in
an otherwise inactive year.
Hence, we believe that the failure to specify location seriously reduces the usefulness of seasonal forecast models, since the accurate prediction of an active or inactive season may not prove beneficial to coastal residents if the location of the activity cannot be identified. As proof, we note that many active seasons have occurred this century with few, if any, landfalling storms (e.g., 1981). Conversely, an inactive season (e.g., 1983 and 1992) could prove to be quite damaging if the path of just one of the storms crosses a vulnerable area.
Recently, Lehmiller et al. (1997) used multivariate discriminant
analysis techniques to develop statistically significant and skillful models for making extended-range forecasts of hurricane activity within specific locations of the North Atlantic basin. These sub-basins include the Caribbean Sea, the Gulf of Mexico, and the Southeast U.S. coast. For the Gulf of Mexico and the Caribbean Sea, the model predicts the presence or absence of (intense) hurricane activity and not the actual number of storms that will occur in a region; landfall is predicted for the Southeast U.S. coast. The approach used in these forecasts expresses the prediction of activity within a sub-basin by a single probability produced by the
statistical models. These, in turn, can then be compared to climatologolical values to determine the relative risk of each event occurring.
Here, we debut the results of our early June prediction model for hurricane location and update both the December and August prediction schemes. Like the August model, the June model predicts hurricane and intense hurricane activity for the Caribbean basin (an update to our early December forecast) and intense hurricane activity for the Gulf of Mexico. Both the December and August models are now more robust with the inclusion of new parameters which identify North Atlantic sea surface temperatures (SSTs). Cross-validated (hindcast) forecast accuracies range from 67% to 88% for the regions where successful models can be developed.