Machine Learning‐Based Postprocessing Ensemble of AI and NWP Models for Short‐Term Air Temperature Forecasting Over Beijing

Accurate short‐term air temperature forecasts are critical for societal and economic activities, particularly for energy management and urban planning. This study evaluates 6–72 h air temperature forecasts over Beijing from January 2022 to December 2023 using both numerical weather prediction (NWP) and AI‐based models. The AI models, GraphCast and Pangu, were initialized with the ECMWF Integrated Forecasting System (IFS) and NOAA Global Forecast System (GFS), resulting in GraphCast‐IFS, Pangu‐IFS, GraphCast‐GFS, and Pangu‐GFS. Forecast skill exhibits pronounced seasonal and diurnal variability, with AI models consistently outperforming their driving NWP models, particularly when initialized with IFS. Among them, Pangu‐IFS exhibits superior predictive skill. To further enhance forecast performance, outputs from AI and NWP models are postprocessed using a multimodel ensemble of three machine learning (ML) algorithms: multiple linear regression (MLR), gradient boosting regression tree (GBRT), and extreme gradient boosting (XGBoost). The XGBoost ensemble achieves the largest improvements, reducing the root‐mean‐square error (RMSE) by 24.1% and increasing the temperature forecast accuracy (TFA) by 22.0% ( p < 0.01) relative to the simple ensemble mean (SEM), with particularly pronounced enhancements at stations in the mountainous regions of western and northern Beijing. Error decomposition indicates that these improvements primarily arise from reductions in systematic bias and phase errors, with bias correction being the dominant factor. These results indicate that integrating AI, NWP, and ML‐based postprocessing can improve short‐term temperature forecasts, particularly over complex terrain, and may benefit operational and energy‐related applications.