Research On The Stochastically Perturbed Parameterization Based On GRAPES Regional Ensemble Prediction System

The ensemble prediction model stochastic physics perturbation method is an important means to describe the forecast uncertainty caused by the random error of the model physical process.It can increase the probabilistic forecasting skills of small-scale weather events,and has become a hot topic in the current ensemble prediction research field.There are three kinds of stochastic physics schemes--the stochastically perturbed parameterization tendencies(SPPT)scheme,the stochastic kinetic energy backscatter(SKEB)scheme and the stochastically perturbed parameterization(SPP)scheme.At present,although the Global and Regional Assimilation and Prediction Enhanced System-Regional Ensemble Prediction System(GRAPES-REPS)has been established in China,the ensemble prediction experiment for the uncertainty of key parameters in the physical parameterization schemes has not been carried out.SPP has become a widely-concerned stochastic parameter perturbation method in recent years,developing the SPP method in GRAPES-REPS is of great significance for a more comprehensive description of the model uncertainty and development and improvement of the current model perturbation method.A SPP scheme consisting of temporally and spatially varying perturbations of 18 key parameters in the Kain–Fritsch convection,WSM6 microphysics,Medium-Range Forecast PBL and Monin–Obukhov surface layer parameterizations is developed in the GRAPES-REPS and its probabilistic forecasting performance is evaluated.In addition,sensitivities of parameter perturbations in different physical processes,spatial and temporal decorrelation scales,as well as energy evolution characteristics and ensemble prediction performance were analyzed,and SPPT and SKEB are also applied with the SPP to evaluate various combinations of multiple stochastic physics schemes.Finally,a single-physics suite in combination with multiple stochastic physics schemes is developed to investigate if the combined scheme can be an alternative to the multiphysics scheme.A combination of the three stochastic physics schemes is compared with the multiphysics scheme in GRAPES-REPS.The stochastic perturbation ensemble experiments are performed for a summer monsoon month in 2015,and multipleverification metrics,such as the ensemble spread,RMSE,consistency(spread-error relationship),CRPS,Tala-grand diagrams and outlier scores were employed to evaluate the probabilistic forecasting performance in both precipitation forecasting and surface and upper-air verification.The most significant conclusions drawn are as follows:(1)Compared with the control experiment without SPP,the stochastically perturbed parameterization(SPP)scheme can effectively improve the probabilistic forecasting skill in the precipitation and isobar verification,decreasing the root mean square error and outlier,and generate a positive improvement to the overall performance.Furthermore,simultaneously disturbing the parameters in the cumulus convection,microphysics,boundary layer as well as surface layer parameterization schemes achieves better ensemble prediction performance than perturbing part of the parameters in any single parameterization scheme.Moreover,in general,the internal/kinetic energy change after perturbation accounts for 0.001% of the original total/kinetic energy,the energy before and after the SPP perturbation is basically the same.(2)The choice of spatial and temporal decorrelation scales of stochastic patterns has considerable impact on the ensemble prediction performance.The optimal ensemble prediction results can be obtained when choosing the temporal decorrelation coefficient of 12h(larger spatial scale)and the truncated wave number of 20n(larger spatial scale)of the stochastic perturbation field.(3)The multi-stochastic scheme which combining SPP with SKEB and SPPT illustrates that combinations of multiple stochastic physics schemes can better address model uncertainties and further improve the probabilistic skill.What’s more,the combination of all three stochastic physics schemes(SPP,SPPT and SKEB)outperforms any other combination of two schemes in precipitation forecasting and surface and upper air verification to best capture the model errors and improve the forecast skill,which has promising prospect of operational application.(4)Compared with the multi-physics ensemble forecasting scheme which is commonly used in regional ensemble forecasting operation,the verification scores show that although the improvement of the ensemble mean error is not obvious,the probabilistic forecasting skill of precipitation forecasting and isobaric variables are better improved,especially,the equal-likelihood of the ensemble precipitation forecasting skill is significantly improved.Overall,the single-physics combining with multi-stochastic ensemble forecasting schemes(SPP,SPPT and SKEB)outperforms the multi-physics ensemble forecasting scheme.