Flood Susceptibility Dataset for Southern China Based on Seven-Model SHAP Attribution

This dataset supports the study “Hydrological Proximity and Road Infrastructure Dominate Flood Susceptibility Over Rainfall: Seven-Model SHAP Attribution in Southern China.” It contains processed spatial datasets used for flood susceptibility modelling across the Pearl River Basin and adjacent coastal basins in southern China. The dataset includes historical flood observations, river networks, major rivers, road-network data, 3 km road kernel density, population density, urban extent, and other flood-related conditioning layers. These data were prepared to support seven-machine-learning-model flood susceptibility analysis, SHAP-based factor attribution, pixel-level uncertainty assessment, and basin-scale anthropogenic coupling analysis. The dataset is provided for reproducibility, validation, and further research on flood susceptibility, hydrological connectivity, infrastructure impacts, and subtropical flood-risk governance.本数据集支撑题为《水文邻近性与道路基础设施相较于降雨主导华南地区洪水易感性：基于七种模型的SHAP（SHapley Additive exPlanations）归因分析》的研究。本数据集包含经预处理的空间数据集，用于华南地区珠江流域及邻近沿海流域的洪水易感性建模工作。数据集涵盖历史洪水观测数据、河网水系、主干河流、道路网络数据、3公里道路核密度、人口密度、城市建成区范围以及其他洪水相关影响因子图层。本次数据制备旨在支撑七种机器学习模型的洪水易感性分析、基于SHAP的因子归因、像素级不确定性评估以及流域尺度人为耦合分析研究。本数据集公开发布，可为洪水易感性、水文连通性、基础设施影响以及亚热带洪水风险治理相关研究提供可重复性保障、结果验证与进一步探索的支撑。