Data and code for " A Spatially Adaptive and Locally Interpretable Framework for Surface NO2 Estimation under Sparse Monitoring Networks"

This repository provides the data, source code, auxiliary datasets, and spatial weighting files associated with the manuscript entitled “A Spatially Adaptive and Locally Interpretable Framework for Surface NO₂ Estimation under Sparse Monitoring Networks”. The study develops a spatial-dependence-aware machine learning framework, Gaussian Distance-Weighted LightGBM (GDW-LGBM), for estimating near-surface NO₂ concentrations under sparse monitoring conditions. The repository contains the materials used to implement, train, validate, and interpret the GDW-LGBM model, including the cleaned training dataset, auxiliary files for spatial model training and validation, Gaussian-based spatial weighting files, model training scripts, and scripts for grid-level feature importance and SHAP-based analyses. Detailed descriptions of the files, software environment, and general workflow are provided in the README.md file.