Source Code and Sample Data for: Momentum Decoupling Drives Low-Wind Events During Heatwaves in China

1. Explainable Machine Learning Scripts (Python): 01_train_xgboost.py: Script for data preprocessing, hyperparameter optimization, and training the XGBoost model to predict surface wind speeds under extreme thermal stress. 02_shap_analysis.py: Script for SHAP (SHapley Additive exPlanations) value computation and generating publication-quality visualizations (e.g., summary and dependence plots) to quantify the marginal contributions of meteorological predictors. 03_cross_regime_sensitivity.py: Script executing the cross-regime sensitivity experiment, applying the heatwave-trained model to non-heatwave baseline data to dynamically verify the wind suppression mechanism. 2. Dynamical Modeling Configurations: Model_configuration.zip: Contains the specific namelist.input and namelist.wps files used for configuring the high-resolution Weather Research and Forecasting (WRF version 4.2.1) model simulations for the historical representative cases. 3. Sample Data: heatwave_2020_CMA_sample_data.zip: Includes a de-identified, minimal working sample of the meteorological dataset to demonstrate the functionality of the provided Python scripts. (Note: The full underlying ERA5 reanalysis fields and CMA observational records are publicly accessible via their respective official portals, as detailed in the manuscript's Open Research). Usage: Python scripts require xgboost, shap, scikit-learn, pandas, and matplotlib. Please ensure the sample dataset is placed in the ./data/ directory before execution.