The added value of high-resolution downscaling of the ECMWF-EPS for extreme precipitation forecasting

Abstract

We have considered the most severe flood events that affected Liguria region (northwestern Italy) in the last decade, between October 2010 and October 2019. High-resolution numerical simulations of six events have been performed with the state-of-the-art non–hydrostatic Weather Research and Forecasting (WRF) mesoscale model. For each selected case study, 50 WRF simulations have been performed corresponding to the downscaling of the 50 EPS-ECMWF members. The main aim of the present work has been to answer the question on whether high-resolution ensemble forecasts bring an added value with respect to the original EPS forecasts. To answer this question, suitable statistical indices have been considered in order to assess model performance both in terms of mean properties of the ensemble and in terms of its spread. As a result of our study, the added value of WRF-based downscaling clearly emerged in relation both to the forecast of the intensity of the extreme precipitation considered in the present study and to their spatial localization. For example, a significant added value of the downscaling process emerges being the Continuous Ranked Probability Score (CRPS) skill score for the precipitation maxima well above 20 % for all considered events, with the exception of the event belonging to the year 2019 for which the improvement, although presents, is considerably lower. Nevertheless, the downscaled ensemble still remains underdisperive.