Improved short-term hurricane intensity forecasting using regression on core measurements

While tropical cyclone track forecasting has improved noticeably over the last 20 years, intensity forecasting has remained a bit of an enigma to forecasters. Despite increased computing capabilities and more sophisticated dynamical models, statistical models, such as SHIPS, still often outperform their dynamical counterparts.

This research seeks to develop a new statistical-climatological forecasting scheme to improve short-term intensity forecasts for well-developed (having a defined eye) tropical cyclones in the Atlantic basin. Using Vortex Data Messages gathered by Hurricane Hunter reconnaissance flights and stored in the National Hurricane Center's ATCF archives, a Vortex Data Message climatology from 1991-2008 is developed and exploited. This climatology includes storm-scale thermodynamic parameters to aid in TC prediction. Finally, stepwise multiple regression is performed to create a SHIPS-style intensity forecast model.

Cross validation results show that this new regression scheme is superior to SHIPS at all forecast times (12-48 hours) and for all storm intensities, potentially indicating that a new benchmark for TC intensity forecasting has been reached. Other implications, such as the ability to produce probabilistic intensity range forecasts, will also be discussed.

Psuedo-real-time results from a web page that displays the forecast produce will be presented and evaluated.