Modulation of Soil Initial State on WRF Model Performance Over China

The soil state (e.g., temperature and moisture) in a mesoscale numerical prediction model is typically initialized by reanalysis or analysis data that may be subject to large bias. Such bias may lead to unrealistic land‐atmosphere interactions. This study shows that the Climate Forecast System Reanalysis (CFSR) dramatically underestimates soil temperature and overestimates soil moisture over most parts of China in the first (0–10 cm) and second (10–25 cm) soil layers compared to in situ observations in July 2013. A correction based on the global optimal dual kriging is employed to correct CFSR bias in soil temperature and moisture using in situ observations. To investigate the impacts of the corrected soil state on model forecasts, two numerical model simulations—a control run with CFSR soil state and a disturbed run with the corrected soil state—were conducted using the Weather Research and Forecasting model. All the simulations are initiated 4 times per day and run 48 h. Model results show that the corrected soil state, for example, warmer and drier surface over the most parts of China, can enhance evaporation over wet regions, which changes the overlying atmospheric temperature and moisture. The changes of the lifting condensation level, level of free convection, and water transport due to corrected soil state favor precipitation over wet regions, while prohibiting precipitation over dry regions. Moreover, diagnoses indicate that the remote moisture flux convergence plays a dominant role in the precipitation changes over the wet regions. Key Points Both soil temperature and soil moisture are corrected to investigate the impacts of the corrected soil initial state on the atmospheric processes Multiple processes associated with the East Asian summer monsoon precipitation are diagnosed, showing the higher rain possibility over the wet regions and lower possibility over the other regions due to the soil initial state's impacts The simple soil initial state correction delivered in this paper has shown some practical value over regions of dense observations