Application of a high-resolution ensemble prediction system to the prediction of tropical cyclones over the West Indies basin.

The predictability of tropical cyclones is limited by the intrinsic chaotic nature of the atmosphere and uncertainties of numerical weather prediction models. In this context, probabilistic forecasts may be beneficial to better anticipate these events. Ensemble Prediction Systems (EPSs) are available at the global scale, but their coarse horizontal resolution often limits their relevance for the prediction of local rainfall and wind gusts for instance.

On the other hand, since convection-permitting models can significantly improve TC predictions and the associated impacts at the local scale, a prototype of high-resolution EPS, based on the French limited-area model "Arome-Overseas", has been developed. This ensemble has 12 perturbed members and is run with a 2.5km horizontal resolution. It accounts for uncertainties arising from the initial conditions, the lateral boundary conditions, the surface conditions and the model. The Arome-EPS has been evaluated on the severe TC cases that occurred in September 2017 in the West Indies (Irma, José and Maria). Positive impacts are obtained on the TC intensity and simulated rainfall. Similar improvements have been observed over the South-West Indian Ocean.

These simulations are part of the French TIREX project, whose goal is to improve t he cyclone alert, and to reinforce the capacities of adaptation and response of territories and societies in a context of climate change. How to properly use this probabilistic information for risk management is an essential question that this project will ad d ress , based on an original interdisciplinary approach involving experts in risk geography, geomorphology, spatial analysis, political sciences, information and communication, sociology, scientific mediation, atmospheric modeling and climatology . First results will be presented.