Data driven analysis on changing patterns of reactive nitrogen deposition in East and Southeast Asia

East Asia (EA) and Southeast Asia (SEA) are global hotspots for reactive nitrogen (N_r) deposition, accounting for 38% of the world's nitrogen inputs, yet sparse ground observations and biases in chemical transport models (CTMs) have hindered accurate impact assessments. Here, we leveraged an ensemble machine learning framework, integrating random forest, gradient boosting trees, and neural networks, to develop data-driven analyses for wet and dry N_r depositions. Rigorous evaluation using 10-fold cross-validation and independent testing shows that our machine learning estimates outperform CTM-based results. Based on independent testing (R = 0.74 for wet NOᵧ, 0.76 for wet NH_x, 0.85 for dry NOᵧ, 0.75 for dry NH_x), our model improves prediction accuracy by 54.0% for wet NOᵧ, 65.0% for wet NH_x, and 26.0% for dry NH_x compared with CTM-based independent test results. Gridded maps (0.25° × 0.25° resolution) show that total N_r depositions in EA and SEA were 36.3–39.0 Tg N yr⁻¹ during 2000–2020, with wet deposition accounting for 58% and reduced nitrogen (NH_x) comprising 62% of total loads. Analyses of spatiotemporal distributions reveal that EA's depositions align with anthropogenic emission changes, with its wet deposition ammonium-to-nitrate ratio exhibited a downward trend over the period, while SEA's trends are strongly modulated by precipitation. These findings provide a robust scientific foundation for integrating N_r management strategies with climate resilience planning via deposition, enabling targeted interventions to address dual challenges of air quality protection and carbon neutrality goals in these rapidly developing regions.