Simplifying a hydrological ensemble prediction system with a backward greedy selection of members – Part 2: Generalization in time and space

摘要

An uncertainty cascade model applied to stream flow forecasting seeks toevaluate the different sources of uncertainty of the complex rainfall-runoffprocess. The current trend focuses on the combination of MeteorologicalEnsemble Prediction Systems (MEPS) and hydrological model(s). However, thenumber of members of such a HEPS may rapidly increase to a level that may notbe operationally sustainable. This paper evaluates the generalization abilityof a simplification scheme of a 800-member HEPS formed by the combination of16 lumped rainfall-runoff models with the 50 perturbed members from theEuropean Centre for Medium-range Weather Forecasts (ECMWF) EPS. Tests aremade at two levels. At the local level, the transferability of the 9th dayhydrological member selection for the other 8 forecast horizons exhibits an82% success rate. The other evaluation is made at the regional or clusterlevel, the transferability from one catchment to another from within acluster of watersheds also leads to a good performance (85% success rate),especially for forecast time horizons above 3 days and when the basins thatformed the cluster presented themselves a good performance on an individualbasis. Diversity, defined as hydrological model complementarity addressingdifferent aspects of a forecast, was identified as the critical factor forproper selection applications.