Probabilistic forecasting of rainfall from radar nowcasting and hybrid systems

摘要

The use of Quantitative Precipitation Forecasts (QPFs) from either Numerical Weather Prediction (NWP) or radar nowcasting models in flood forecasting systems extends the time available to issue warnings and take actions. However, uncertainty in the rainfall input affects the accuracy of flow predictions. Radar nowcasts have a higher skill at short lead times, whereas NWP models produce more accurate forecasts at longer lead times. Hybrid systems, merging NWP and radar-based forecasts, have been developed to produce more skilful forecasts than either independent component (i.e. NWP/radar nowcasting). This study aims at assessing radar nowcasts and hybrid forecasts provided by the state-of-the-art model STEPS. The forecasts were run on a 1000 km × 1000 km domain covering the UK, at 2-km spatial and 15-min temporal resolutions. Results show that the forecasting system benefits from the blending with the NWP forecasts.