Optimizing convection‐permitting ensemble via selection of the coarse ensemble driving members

• DOI: 10.1002/met.2137
• Publication Date: 2023-08-29T03:36:46Z
• Authors: Pavel Khain; Alon Shtivelman; Yoav Levi; Anat Baharad; Eyal Amitai; Yizhak Carmona; Elyakom Vadislavsky; Amit Savir; Nir Stav
• Keywords: Ensemble forecasting; Ensemble Learning

Abstract Nowadays, several global ensembles (GEs) which consist of tens members are being run operationally. In order to locally improve the probabilistic forecasts, various forecasting centers and research institutes utilize GEs as initial boundary conditions drive regional convection permitting (RCPEs). RCPEs demand significant computer resources often a limited number ensemble is affordable, smaller than size driving GE. Since each RCPE member obtains from specific GE member, there many options select members. The study uses European Centre for Medium‐Range Weather Forecasts (ECMWF) consisting 50 members, 20 COSMO model over Eastern Mediterranean. We compare approaches automatic selection propose optimal methods, including random selection, consistently lead better performance driven RCPE. comparison includes verification near surface variables precipitation using metrics. results validated methods physics perturbation. Besides configurations, we show that at high intensities spatial up‐scaling recommended in obtain useful forecasts.