Research On Ensemble Forecast Improvement Of GRAPES Model Precipitation And Radar Echo Reflectivity

When the numerical model forecasts future weather,the results usually have some uncertainties(errors)because of the non-linear chaos characteristic of the atmosphere motion,the initial conditions errors and the model itself errors.Especially for some forecast variables with strong local characteristics,such as precipitation and radar reflectivity,their forecast uncertainties are more obvious.The ensemble forecast is an effective method to solve this uncertainty.This paper aims at the uncertainty of heavy precipitation and radar reflectivity in the numerical model,based on GRAPES-REPS(Global and Regional Assimilation and Prediction Enhanced System-Regional Ensemble Prediction System)regional meso-scale ensemble forecast model and convection-permitting ensemble forecast model,try to improve the probabilistic forecast skills of convective weather process in Chinese operational ensemble forecast,and conduct the following three researches:(1)with the in-depth analysis of the impact of applying Stochastically Perturbed Parameterization(SPP)on winter precipitation ensemble forecast in China,estimate the applicability of Stochastically Perturbed Parameterization in winter precipitation ensemble forecast;(2)develop a precipitation-based stochastic perturbation function generation method and conduct a series of ensemble forecast experiment for convective heavy rain in the establishment periods of the EASM on southern China,and estimate the effect of convection-permitting precipitation ensemble prediction;(3)design a new method for calculating radar reflectivity of sub-grid precipitation,improve the radar reflectivity probabilistic forecast skills.Through these researches,get a further understanding of the forecast uncertainty of convective weather process,provide a scientific basis for further improving the forecast ability for the convective weather process in the GRAPES-REPS regional ensemble prediction model.The results and conclusions are as follows:(1)Based on the GRAPES-REPS regional ensemble prediction model with10 km horizontal resolution,select 16 sensitive parameters in physics process parameterization schemes,construct the SPP scheme.Conduct total 31 days ensemble forecast experiments from December 12,2018 to January 12,2019,and analyze the SPP scheme effect for precipitation ensemble forecast.It finds that the probabilistic forecast skills for isobaric surface elements can be improved,but the probabilistic forecast skills for precipitation have not changed much when applying the SPP scheme in winter.The reason is maybe the perturbed sensitive parameters in the SPP scheme are mostly related to the precipitation of sub-grid physics processes.However,the winter rainfall precess in China is mainly related to the uncertainty of the baroclinic dynamic mechanism,the model forecast precipitation is mainly large-scale grid precipitation,and there is less convective sub-grid precipitation.Therefore,the SPP scheme does not improve the winter precipitation ensemble forecast significantly.(2)Aiming at the problem that most stochastic perturbation function are not closely related to the precipitation process,based on the GRAPES convection-permitting ensemble prediction model with 3km horizontal resolution,aim at the characteristic of heavy rainfall process,we design a precipitation-based stochastic perturbation function generation method.This method is that combining the first-order auto-regressive stochastic perturbation function with some environmental factors which affect precipitation,constructing a Precipitation-related Stochastic Forcing Function(Pre SF)which related to the spatial-temporal scale and uncertainty of precipitation environmental factors.Though two ensemble forecast experiment cases(1.April 11,2019 and 2.April 24,2019)which is warm heavy rain in South China,we analyze the effect of applying a new stochastic function on precipitation and other factors.The results show that using the Pre SF can improve the precipitation probabilistic forecast skills significantly,especially for heavy rain to torrential rain,the improvement is very significant.It also improves the forecast skills of isobaric surface elements to a certain extent.(3)Aiming at the problem that the simulated radar reflectivity cannot reflect the sub-grid precipitation information of the sub-grid physics process(Kain-Fritsch scheme)in the GRAPES-REPS regional ensemble prediction model with 10 km horizontal resolution,we develop a new radar reflectivity calculation method for the sub-grid precipitation.This method is to subtract the sub-grid precipitation rate from the downdraft evaporation rate,according to the Z-R relationship for estimating precipitation based on radar,estimate the radar reflectivity on each layer,and combine the simulated the radar reflectivity of the microphysics process,obtain a new radar reflectivity.Though two 15-day forecast experiments for two different time periods(April 11–25,2019 and August 1–15,2019),it shows that the reflectivity products calculated by the new method clearly indicate the precipitation generated by the cumulus parameterization scheme in the model,especially for events with numerous subgrid-scale precipitation processes,it can better simulate the radar reflectivity related to subgrid-scale precipitation.It means that applying this new method in models can improve the forecast skills of radar reflectivity.