Abstract
U.S. National Hurricane Center (NHC) forecasts for tropical cyclone tracks and wind speeds are extended in time to produce spatially disaggregated probability forecasts for landfall location and intensity, using a weighted bootstrap procedure. Historical analogs, with respect to the forecast characteristics (location, heading, and wind speed) of a current storm, are selected. These are resampled by translating their locations to random positions consistent with the current forecast, and recent NHC forecast accuracy statistics. The result is a large number of plausible Monte Carlo realizations that jointly approximate a probability distribution for the future track and intensity of the storm. Performance of the resulting forecasts is assessed for U.S. tropical cyclone landfall probabilities during 1998–2006, and the forecasts are shown to be skillful and exhibit excellent reliability, even beyond the 120-h forecast horizon of the NHC advisory forecasts upon which they are based.
Corresponding author address: Daniel S. Wilks, Dept. of Earth and Atmospheric Sciences, Cornell University, Ithaca, NY 14853. Email: dsw5@cornell.edu