• A radar coherence-based method enables individual-building damage mapping. • The method achieves 86% accuracy and outperforms existing rapid damage products. • An XGBoost model predicts building damage density under data-limited conditions. • Extends medium-resolution radar from area-based to building-level damage mapping. On 6 February 2023, two major earthquakes (Mw 7.8 and Mw 7.5) struck southern Türkiye, causing extensive building damage and substantial human casualties. Rapid post-earthquake response requires reliable information on damaged buildings at the individual-building level, yet most existing rapid damage products remain aggregated or area-based. This study presents an integrated framework that enables large-scale individual-building-level damage identification using medium-resolution Sentinel-1 SAR data and complements regional-scale damage prediction under data-limited conditions. In the first stage, a temporal coherence-based approach incorporating homogeneous pixel selection, non-local filtering, and multi-indicator fusion was developed to identify earthquake-damaged buildings across dense urban areas. The resulting Building Damage Proxy Map (BDPM) captures damage patterns at the scale of individual structures and achieves an average identification accuracy of 86%, outperforming existing rapid damage products by over 9%. Comparative analyses with independent SAR- and optical-based datasets demonstrate that the proposed method more reliably detects damaged buildings, including structures missed by optical imagery due to cloud cover, acquisition timing, or visually inconspicuous damage. In the second stage, an XGBoost-based model is applied to predict building damage density using seismic, topographic, geological, and building-related factors, providing regional-scale situational awareness when post-event observations are unavailable or delayed. Overall, this study advances the operational use of medium-resolution SAR data for building-level earthquake damage identification and offers a scale-aware framework for rapid damage assessment to support emergency response and recovery planning.
Liu et al. (Fri,) studied this question.