Uncertainties in Cyclonic Hazard in Tropical Islands

Tropical islands, due to their location, are highly exposed to climate-ocean related hazards. Among these hazards, tropical cyclones (TCs), which generate extreme weather conditions, storm surges and marine submersions, are particularly devastating. Their extreme intensity, relatively rare occurrence, and sparse spatial distribution complicate their observation, making numerical modelling an essential tool for characterising TC events, and improving our understanding of their dynamics. However, accurately modelling TCs remains very challenging, and significant uncertainties persist in the modelled hazard even after the event. Hence, this study aims to quantify uncertainties in the modelled atmospheric hazard arising from three main sources: physical-process uncertainties associated with current limitations in our understanding and representation of key mechanisms (e.g. turbulent fluxes at the air–sea interface, planetary boundary layer physics, convection) ; numerical uncertainties, linked to model design and computational constraints (resolutions, numerical schemes) ; and forcing uncertainties (initial and boundary conditions, land interactions). We focus on the cyclonic hazard in two contrasted tropical territories: the mountainous island of La Réunion located in the south-west Indian Ocean, and the small volcanic islands of the Caribbean arc in the North Atlantic basin, taking into account the specific characteristics of both territories (small islands with steep bathymetry, orographic effects), and the ocean basins (synoptic conditions, availability of observations). Using the Weather Research and Forecasting (WRF) model with three two-way nested domains (9 km, 3 km, 1 km), we conduct an ensemble of high-resolution retrospective simulations of some of the major events impacting these territories since the 1980s to quantify the respective contribution of the different sources of uncertainties. This work will help to design effective TC risk management and adaptation strategies.