Study On Stochastically Perturbed Parameterization Approaches In The Ensemble Forecasts Of Cumulus Precipitation

As one of the hotspots in the field of ensemble forecasting,stochastically perturbed parameterization approaches play a great role in estimating and characterizing model errors of ensemble prediction systems.The cumulus convection parameterization schemes remain errors in the forecasts of precipitation,and it is necessary to conduct specific research and design corresponding stochastically perturbed parameterization approaches for them.The land surface process is closely related to cumulus convective precipitation,which means surface meteorological factors should be invovled in the perturbations.The BMJ(Betts-Miller-Janji(?))scheme is widely used and based on convection adjustment.It is quite different from other mass-flux schemes and there are few literatures on this topic.Analyzing the characteristics of precipitation forecasts with the BMJ scheme and developing reasonable stochastically perturbed parameterization approaches are of great significance to improving the results of ensemble forecast when using BMJ.The impacts of stochastically perturbed parameterization tendencies and stochastic perturbed parameterization on the precipitation ensemble forecasts with the BMJ scheme were investigated.Both stochastic perturbation approaches were implemented on the ARPS(Advanced Regional Prediction System)model.Precipitation ensemble forecasts were evaluated by 20 cases occurred in eastern China in June and July of 2018 and 2019.The tendencies perturbations involved the temperature and specific humidity tendencies,while the parameters perturbations were applied in the temperature and specific humidity reference profiles.In addition,the surface temperature and moisture fluxes were perturbed with the stochastic perturbation approaches.The impacts of the surface fluxes perturbations on the BMJ precipitation forecasts were evaluated.The joint-perturbations of BMJ and surface fluxes were also investigated.The most significant conclusions drawn are as follows:(1)The precipitation forecasts with the BMJ scheme were featured by wet bias at examined thresholds,meaning more forecast precipitation events than the corresponding observations.The bias issue remained after using wetter reference profiles.The responses of the BMJ scheme to stochastic approaches substantially differ.The tendencies perturbations had little impact on the BMJ precipitation forecasts as the forecast precipitation frequency did not significantly changed,compared with the unperturbed BMJ scheme.The ensemble spread is low in the case of using tendencies perturbations.By contrast,perturbing the reference profiles had great impacts on the precipitation ensemble forecasts.Symmetric perturbations of reference profiles produced large ensemble spread,but this approach also increased the wet bias at light rain thresholds and yield small improvement on the forecast skill scores.Compared to tendencies perturbation approach,asymmetric perturbations(the perturbation mean > 1.0)of reference profiles were conducive to larger ensemble spread,higher forecast skill scores,and smaller wet bias,although the wet bias at large precipitation thresholds became larger.Moreover,the asymmetric perturbations substantially improved the precipitation spatial distribution in early forecast stage(0?3 h)and the nocturnal precipitation intensity.The large area of spurious precipitation yielded by the BMJ scheme in early stage of forecast substantially dried the air,which suppressed the precipitation intensity in subsequent forecast.The asymmetric perturbations effectively reduced the spurious precipitation and improved the precipitation intensity forecast.The quantities of temperature and specific humidity tendencies were small,which was likely the cause of small impact of tendency perturbations on the BMJ scheme.(2)The surface fluxes perturbations had small but non-ignorable influences on the BMJ precipitation forecasts.Symmetrically perturbing surface temperature and moisture fluxes increased the spatial distribution forecast skill of precipitation with large precipitation percentile,but this approach degraded the skill score for precipitation with smaller precipitation percentile.Asymmetric perturbations(perturbation mean < 1.0)of the above surface fluxes generally produced positive impacts on the BMJ precipitation forecasts.Moreover,larger perturbation scale of time and space contributes to increasing the ensemble spread.(3)The results of the joint-perturbations showed that the BMJ perturbations had larger impacts on the precipitation forecasts than the surface fluxes perturbations.The combination of the asymmetric perturbations of BMJ and symmetric perturbations(large amplitude)of surface fluxes was conducive to higher precipitation forecast skill scores.