Post-processing ensemble rainfall forecasts for short-term streamflow forecasting purposes over the Kôn river basin

Abstract

Post-processing of ensemble rainfall forecasts from global numerical weather prediction models (NWPs) is often required before they can be used for ensemble flood forecasting purposes. This paper evaluated the effectiveness of post-processing NWP’s ensemble rainfall forecasts through a quantile
mapping technique (QM) in improving rainfall forecasts over the Kôn river basin. Ensemble rainfall forecasts of the ECMWF system from THORPEX (The Observing System Research and Predictability Experiment) Interactive from September to December over the period 2014 to 2019 were used for this
study. The evaluations were done by comparing the QM-processed rainfall forecasts against the raw forecasts and the ones processed with a simple mean-bias removal technique (called DMB) using multiple verification metrics. The results showed that the performance of QM varied with the selected
verification metrics. The QM-processed rainfall forecasts were shown to be more skill full than the raw and mean-bias removal ones in term of categorical and probabilistic metrics, but deterministic metrics.