A Stochastic Tropical Cyclone Intensity Model for Wind Hazard Assessment Using the Geographically Weighted Summary Statistic Method

Abstract

In this study, a stochastic tropical cyclone (TC) intensity model based on a deterministic fast physics-based TC intensity model is developed for potential applications in the TC hazard assessment. To account for the effect of randomness, an error term is introduced to the ordinary differential equation for the TC intensity. The error term is assumed to be the summation of a mean part and a white noise. The geographically weighted summary statistics method is adopted to estimate the geographically varying mean and white noise intensity of the error term using the 1,010 historical tropical cyclone records from the western North Pacific. The effects of the mean and random parts of the error term on the model’s performance are investigated in terms of the individual samples and the probabilistic characteristics of TC intensity evolution. The results show that, by introducing the error term, the difference of the 25-, 50-, and 100-year recurrence interval values between the model and observation can be largely reduced from the level of 10% to less than 4%, validating the adequacy of the proposed model in the TC hazard assessment.