Post-Flood Ecosystem Recovery and Policy Zoning in the Chi River Basin, Thailand

Understanding ecosystem recovery following flood events is critical for sustainable floodplain management, yet causal evidence linking flooding to vegetation recovery remains limited. This study investigated the causal impact of the 2022 flood on ecosystem recovery in the Chi River Basin, Thailand, using 232,172 spatial points with multi-temporal VCI data (2020–2024). We employed Propensity Score Matching (PSM) to control for confounding factors and Difference-in-Differences (DiD) to estimate the causal effect. XGBoost with SHAP analysis identified predictors of Slow Recovery Hotspots, and a four-tier policy zoning framework was developed for sustainable land use planning. The DiD estimate of −0.0734 (95% CI: −0.0802, −0.0666) in standardized VCI units (equivalent to a raw VCI reduction of 2.47 units on the 0–100 scale) suggests that flooding caused a significant reduction in vegetation recovery, with effects persisting for at least two years. Slow Recovery Hotspots—defined as locations where two-year post-flood VCI remained below pre-flood levels—comprised 51.1% of the study area (118,643 points), with estimated annual economic losses of 226.2 Million Baht (95% CI: 203.6–248.8 Million Baht). SHAP analysis revealed that pre-flood VCI (48.6%), elevation (21.6%), and pre-flood NDVI (13.5%) were the dominant predictors. A critical elevation threshold of 145–155 m was identified, with flood frequency showing cumulative effects (≥2 events out of 5 observation years significantly increasing risk). A four-tier policy zoning framework (Red: 4.5%, Yellow: 46.6%, Orange: 10.0%, Green: 38.9%) was developed, providing actionable recommendations including no-build zones, land buyouts, building codes, early warning systems, and drainage improvements. Scenario-based analysis suggests that implementation could reduce economic losses by 30–40% (67.9–90.5 Million Baht/year). The approach demonstrates the value of combining causal inference with machine learning for evidence-based floodplain management. This framework contributes to SDGs 2, 6, 11, 13, and 15 and is transferable to other flood-prone regions in Southeast Asia.