Statistical Downscaling Precipitation Forecast for Hydropower Industry and Its Calibration Using Frequency Matching Method

Abstract

This study deals with high-resolution precipitation forecasts for hydropower industry using a statistical downscaling method based on the linear regression of the categorized daily precipitation forecasts taken from the European Centre for Medium-Range Weather Forecasts (ECMWF), Japan Meteorological Agency (JMA), the US National Centers for Environmental Prediction (NCEP), and the United Kingdom Met Office (UKMO), in the TIGGE archive as well as the quality-controlled precipitation data from China Merged Precipitation Analysis (CMPA). To further improve the forecast skill of the daily precipitation, the calibration of the precipitation forecast has been performed by using a statistical postprocessing approach called the frequency matching method (FMM). The results show that the statistical downscaling forecast skill using categorized precipitation scheme is much larger than that of bilinear interpolation and that using uncategorized precipitation scheme in terms of the equitable threat score (ETS), anomaly correlation coefficient (ACC) and root-mean-square error (RMSE), no matter the precipitation is light rain, moderate rain, or heavy rain and the above. The calibration of the precipitation forecasts using FMM can significantly reduce the false alarm of the light rain and the missing rate of the heavy rain and the above. Hence, it can improve the inflow forecast skill in the hydrological models which make use of observed and predicted precipitation as input variables.