Rapid assessment of earthquake-induced building damage using SAR and multispectral imagery

Rapid and accurate post-earthquake building damage assessment is essential for effective emergency response and disaster mitigation. This study proposed a CatBoost-based multiclass damage model (CMDM) that integrates synthetic aperture radar (SAR) coherence, multispectral indices, texture features, and peak ground acceleration (PGA) to classify building damage levels in Antakya, Turkey. Sentinel-1 time-series SAR data, Sentinel-2 multispectral imagery, and PGA data were processed to derive six key features, followed by SHapley Additive exPlanations (SHAP)-based feature selection. The CMDM, combined with a One-vs-Rest classification strategy, was trained to categorize buildings into minor damage, major damage, and destruction classes. The model achieved an overall F1-score of 0.831, with class-wise F1-scores of 0.834, 0.776, and 0.884 for the minor damage, major damage, and destruction classes, respectively. The receiver operating characteristic area under the curve (ROC-AUC) and precision–recall area under the curve (PR-AUC) values exceeded 0.95 and 0.91 across all classes, indicating strong discriminative performance under imbalanced conditions. Spatial patterns revealed that severe damage was concentrated in the central urban plain and along fault-proximal zones, consistent with the seismic intensity distribution. These findings demonstrate that multi-source remote sensing fusion enhances the efficiency and reliability of rapid earthquake damage mapping.