Multi-source SAR data-driven flood extent mapping and flood risk assessment in the coastal area of east China: A case study of Donghai County, Lianyungang

Floods are among the most frequent and destructive natural hazards globally, yet research on small-to-medium, county-scale basins in eastern China remains scarce. This study focuses on Donghai County, a rapidly urbanizing and river-network-dense region in Lianyungang (Jiangsu, China), to explore multi-source Synthetic Aperture Radar (SAR) data for flood extent mapping and risk assessment. Four flood extraction methods (Random Forest, SVM, object-based segmentation, K-means) were compared, with Random Forest achieving the highest accuracy (overall accuracy = 94.03%, F1-score = 0.93). Using this method, multi-source SAR data (Sentinel-1A, GF-3, Chaohu-1) captured the July 2024 flood's spatiotemporal dynamics: inundated area expanded from 100.5 to 291.9 km 2 , with severe flooding in the eastern and southeastern lowlands. Land-use analysis showed cropland was most affected (87.13% of inundated area), followed by built-up land (11.90%). Long-term precipitation data (1901–2023) analyzed via EOF-based variability analysis and nonparametric trend tests revealed a weak, locally heterogeneous upward trend in flood-season rainfall. Height Above Nearest Drainage (HAND) model-based risk assessment indicated 87% of inundated areas were in very high/high-risk zones. The findings demonstrate the utility of integrating multi-source SAR, machine learning, hydro-climatic context, and topographic analysis for flood monitoring, supporting flood mitigation and risk management in eastern China's coastal plain regions.