Repository associated with "Two-Dimensional Deep Learning Scheme with One Predicator Only to Parametrize Global Lightning"

This is the source code of the two-dimensional ResUNet model, as well as scripts to reproduce evaluation figures in "A Two-Dimensional Deep Learning Scheme with One Predicator Only to Parametrize Global Lightning". training_resunet_lightnig_proxy_sensitivity.ipynb is the model training code. evaluate_reunet_results.ipynb is the model evaluation code. feature_importance_analysis.ipynb is the feature importance evaluation code. evaluate_reunet_results_extreme.ipynb is the model interpretation code. dir_resunet_archi directory contains the ResUNet model used in the paper. dir_elmtree_archi directory contains the Elmtree model used in the paper, from Verjans & Franzke, 2025 (https://zenodo.org/records/13920004). input_data directory contains the input features and lightning data products used for computing the resunet, ELM-tree, PR92, and F14 predictions, and for their evaluation. Reference: Yin, M., Wang, Y., Fang, Y., Li, F., Qie, X., Wang, Y., et al. (2026). A two‐dimensional deep learning scheme with one predicator only to parametrize global lightning. Geophysical Research Letters, 53, e2025GL120502. https://doi. org/10.1029/2025GL120502 Verjans, V., & Franzke, C. L. E. (2025). Development of a Data‐Driven Lightning Scheme for Implementation in Global Climate Models. Journal of Advances in Modeling Earth Systems, 17(2). https://doi.org/10.1029/2024ms004464