Improving Short Range Ensemble Precipitation Forecast using an Ensemble Analog Method

Improving short range precipitation forecast (SRPF) is an essential issue for operational prediction centers. The SRPF becomes especially challenging for coastal regions during flood seasons. Ensemble forecast is an important approach to deal with the inherent uncertainty of SRPF. However, ensemble forecast systems (EPSs) are often biased or under-dispersive due to the imperfect initialization and model configuration. In this study, an ensemble analog method is used to calibrate SRPF using the ECMWF 24h ensemble precipitation forecast in Jiangsu during June-August 2016. The method uses observation set of similar ensembles during a moving training period. To define the similarity between current ensemble forecast and historical ensemble forecast, both the root-mean-square-error (RMSE) method and the minimization of ensemble mean and ensemble spread difference (EM+SPREAD) method are tested. Results indicate these methods can obviously improve the reliability of probabilistic precipitation forecast of different levels. The improved continuous ranked probability skill score (CRPSS) also shows the potential application value of these methods in improving the SRPF during summertime.