The data contains geospatial data, model scripts, and figure generation code used in the study. The full pipeline used two modular Python scripts: lsm_pipeline_final.py (end-to-end pipeline: Phases 0–6, including CRS verification, feature extraction, ML training, spatial mapping, Jenks classification, and manuscript generation) and generate_all_manuscript_figures_final_1.py. The analysis was conducted in Python 3.10 using scikit-learn, LightGBM, XGBoost, SHAP, rasterio, geopandas, mapclassify, and scipy.