A spatial-explicit analysis of influencing factors of observed floods in the Yangtze River Delta, China

Flooding is not only shaped by hydrological and hydraulic mechanisms, but also substantially influenced by landscape configuration. Elucidating the relationships between historical floods and environment determinants is crucial for advancing both flood resilience strategies and landscape sustainability practices. The objectives of this study were: (1) to spatial-explicitly quantify flood-environment associations at a large scale, and (2) to characterize potential scale effects governing these relationships. An integrated methodological framework of spatial autocorrelation analysis, hotspot detection, and comparative modeling using Ordinary Least Squares (OLS), Geographically Weighted Regression (GWR), and Multi-scale GWR (MGWR) was employed to investigate historical floods observed from satellite images, taking China's Yangtze River Delta (YRD) for an empirical study. The results showed that floods occurred 23 times in the YRD from 2000 to 2020 and demonstrated significant spatial autocorrelation and hotspots. The MGWR model outperformed other models in establishing the flood-environment relationships, reaching an overall R2 of 0.68 (locally ranging 0.572–0.829). Notably, the landscape shape index (MGWR coefficients − 2.08 – − 0.38) and shannon's diversity index (0.03–0.19) emerged as key influencing factors, indicating that landscape patterns matter in influencing flood occurrence. Additionally, spatial non-stationarity and scale effects were revealed, which distinguished global, medium, and local factors affecting flood occurrence. These findings implied for spatially explicit strategy to managing flood risk and promote landscape sustainability. Also, the methodology based on flood observation and spatial statistics offered a tool to investigate the mechanisms of large-scale flood occurrence in other regions.