Probabilistic forecasts of tropical cyclone tracks and intensities in the southwest Indian Ocean basin

Under the responsibility of the Regional Specialized Meteorological Center (RSMC) of La Reunion, the southwest Indian Ocean (SWIO) has tropical cyclone activity close to that of the North Atlantic. Like most territories of the SWIO basin, La Reunion island is highly vulnerable to cyclone‐induced hazards and the potential impact of nearby storms is closely related to their track and intensity evolution. Although storm track and intensity forecasts have been steadily improving in the last decades, a great amount of uncertainty remains. Operational centers have therefore developed probabilistic products such as uncertainty cones for the prediction of storm track and intensity over the different cyclone basins. Unfortunately, the cone approach does not fully match the end‐user needs for efficient decision‐making. This paper provides a method to generate alternate probabilistic scenarios of tropical system track and intensity forecasts around an official forecast. The method has been calibrated and evaluated to answer the needs of the SPICy (Systeme de Prevision des Inondations en contexte Cyclonique) project that aimed to explore new probabilistic forecast products for tropical system induced hazards such as coastal inundations. A hybrid method has been developed to benefit from both climatological and dynamical existing approaches. A first set of climatology‐built scenarios is generated using the statistical distribution of RSMC La Reunion forecast errors. This initial set is then modulated using real‐time information provided by the ensemble prediction system of the European Center for Medium‐range Weather Forecasts (ECMWF). The final product is a set of about 20 scenarios that are built around the official deterministic forecast with some associated probabilities. Performance scores demonstrate the efficiency of the method against other evaluated systems especially in the first two days of the forecast. The reasonable number of defined scenarios is cost efficient and makes it possible to perform further impact‐oriented applications such as wave and storm surge simulations.