Accurate calculation of land surface heat fluxes is scientifically significant and practically valuable for understanding land-atmosphere interaction and improving capabilities of weather forecast. The methods commonly used for calculating land surface heat fluxes in current numerical weather forecast models are based on the Monin-Obukhov Similarity Theory (MOST). These methods have some limitations and it is hard to further improve them. Meanwhile, a method based on maximum entropy production (MEP) model is proposed to calculate land surface heat fluxes in recent years. This method demonstrates great advantages and yields good results and has been increasingly applied in the study of land-atmosphere exchange. This study introduces the MEP model into the operational weather forecast numerical model (CMA-BJ model) in Beijing Meteorological Service to replace the existing MOST-based method in the model for the calculation of land surface sensible heat flux, latent heat flux and ground heat flux. Simulation experiments over the period from June to August 2022 are conducted to evaluate the performance of the operational numerical model system with the MEP model on predicting land surface and atmospheric meteorological elements and precipitation. Results show that the use of MEP model in CMA-BJ model can significantly improve the simulation of surface energy balance and thermodynamic processes of land surface and boundary layer. It also enhances the ability of the model for the simulation of air temperature, humidity and wind in the boundary layer, and ultimately increases the accuracy of precipitation prediction, especially for heavy rainfall prediction. The TS (threat scores) of rainstorm forecasts in North China and Yangtze river basin have increased by 20% and 10%, respectively. The predicted diurnal precipitation variation is also more consistent with observations with higher correlation coefficient. These results show that it is feasible to use MEP model in numerical weather forecast models and operational systems.

Surface sensible and latent heat fluxes at the Xiaotangshan, Shunyi and Wuqing sites in the Beijing-Tianjin-Hebei region are calculated using a Maximum Entropy Production model (MEP) and observation data. The input data of the MEP model are net radiation, temperature and humidity. Among them, surface temperature and specific humidity are estimated respectively by surface upward and downward long wave radiation and relative humidity. The results of the MEP model are compared with Eddy Covariance (EC) observations to evaluate the model performance. In addition, daily averaged sensible and latent heat fluxes calculated by the MEP model at the three sites are also compared with the results from a simplified method (M20) and seven reanalysis datasets (NCEP1, NCEP2, MERRA2, ERA-5, ERA-Interim, JRA55, CRA) to analyze the accuracy of these methods for acquiring the surface sensible and latent heat fluxes in the Beijing-Tianjin-Hebei region. The results show that the fluxes calculated by the MEP model are generally consistent with observations and show similar diurnal, monthly and seasonal characteristics with those observed. Compared with results of the M20 method and reanalysis datasets, the results of the MEP model agree better with observations with smaller RMSEs (Root Mean Square Error) and average deviations. Therefore, the MEP model has high accuracy and reliability in calculating surface heat fluxes in the Beijing-Tianjin-Hebei region, and shows a good applicability in this region.