Integrated statistical post-processing methods for categorical and quantitative errors correction of numerical precipitation forecasts
The utilization of statistical post-processed numerical precipitation forecasts is a significant approach to extend the effective forecast period of hydrological forecasting.Existed statistical post-processing methods struggle to simultaneously correct dichotomous and quantitative errors,and their impact on the effective forecast lead time for precipitation forecasting is frequently overlooked.In this study,we introduce a novel post-processing scheme called EQM-BMGD,which combines the Empirical Quantile Mapping model(EQM)and the Bernoulli-meta-Gaussian Distribution(BMGD).Additionally,we establish a comprehensive accuracy metric for evaluating the effective forecast period.Using the Han River Basin as a case study,comparative outcomes showed that EQM-BMGD integrated the strengths of the two individual methods,achieving precipitation forecasts with superior accuracy.The forecast accuracy(OP)and mean absolute error(EMA)of the post-processed average-basin forecasts increased by more than 10％,the OP of the forecast period 222-228 h was still close to 0.7,and EMA was less than 0.7 mm/(6h),and the EFPs were extended by 18-66 h.On a grid scale,the gains of OP and EMA for the 96-102 h forecast period exceeded 10％and 20％respectively for all grids.Except for a few grids in the southwest,the OP surpassed 0.8 while the EMA remained below 1.0 mm/(6 h).In addition,the EFPs of the grids in the northern part were lengthened by 18-54 h.It is demonstrated that the EQM-BMGD can effectively correct both categorical and quantitative errors,thereby enriching the available methodologies for statistical post-processing of numerical precipitation forecasts.
numerical precipitation forecaststatistical post-processingempirical quantile mappingbernoulli-meta-gaussianeffective forecast period
国家科技期刊平台
NSTL
万方数据
