New Probabilistic Forecast Models for the Prediction of Tropical Cyclone Rapid Intensification

Abstract

he National Hurricane Center currently employs a skillful probabilistic rapid intensification index (RII) based on linear discriminant analysis of the environmental and satellite-derived features from the Statistical Hurricane Intensity Prediction Scheme (SHIPS) dataset. Probabilistic prediction of rapid intensity change in tropical cyclones is revisited here using two additional models: one based on logistic regression and the other on a naïve Bayesian framework. Each model incorporates data from the SHIPS dataset over both the North Atlantic and eastern North Pacific Ocean basins to provide the probability of exceeding the standard rapid intensification thresholds [25, 30, and 35 kt (24 h)?1] for 24 h into the future. The optimal SHIPS and satellite-based predictors of rapid intensification differ slightly between each probabilistic model and ocean basin, but each set of optimal predictors incorporates thermodynamic and dynamic aspects of the tropical cyclone?s environment (such as vertical wind shear) and its structure (such as departure from convective axisymmetry). Cross validation shows that both the logistic regression and Bayesian probabilistic models are skillful relative to climatology. Dependent testing indicates both models exhibit forecast skill that generally exceeds the skill of the present operational SHIPS-RII and a simple average of the probabilities provided by the logistic regression, Bayesian, and SHIPS-RII models provides greater skill than any individual model. For the rapid intensification threshold of 25 kt (24 h)?1, the three-member ensemble mean improves the Brier skill scores of the current operational SHIPS-RII by 33% in the North Atlantic and 52% in the eastern North Pacific.