Evaluation of the WRF Physics Ensemble using Multivariable Integrated Evaluation Approach over Haihe river basin in north China

  A crucial step of the application of WRF in regional climate research is selection of the proper combinations of physical parameterizations. In this study, we performed experiments in WRF to assess the predict skill of various parametrization schemes sets in simulating precipitation, temperature over the Haihe river basin. The experiments driven by ERA-INTERIM reanalysis data are performed for a period of summer (1 June to 31 August, 2016) in this domain with 13 km grid spacing. Fifty-eight members of physics combinations thoroughly covering five types of physics options are assessed against the available observational data by utilizing the multivariable integrated evaluation (MVIE) method. It is deduced that the best performing setup consists of CAM5.1 microphysics, MRF PBL, BMJ Cumulus, CAM Longwave/Shortwave radiation, and Noah Land Surface schemes. To identify the robustness of the optimal scheme set, the vector field evaluation (VFE) diagram for displaying all simulations reveals that the optimal one is distinguished from others by higher vector field similarity coefficient(Rν), smaller root mean square vector deviation(RMSVD). The model deviations spatially for the precipitation show a promising tendency that a strong overestimation about 5 mm/day for the default configuration evolves small biases of the optimal setup with a range between -1 and 1 mm/day, and the surface temperature forecasts have improved to some extent although not significant as that of precipitation. The temporally analysis of the spatial average of all simulations exhibits that for temperature the optimal setup is more approaching to the observational data, but for precipitation no remarkable difference between all simulation and the observations. Further analysis of the sensitivities of model output to different types of physics option suggests that, microphysics, PBL, and Cumulus schemes have more significant impact on the model performances measured by a multivariable integrated evaluation index (MIEI) than radiation scheme and Land Surface schemes.