Liang et al. (2026) Urban flood risk and the structural importance of cities in a spatial interaction network: evidence from the Yangtze River Delta Urban Agglomeration

Identification

• Journal: Geomatics Natural Hazards and Risk
• Year: 2026
• Date: 2026-09-23
• Authors: Yiyin Liang, Haipeng Lu, Hangling Ma, Yanmin Wang, Yanwen Song, Bo Wu, Shuliang Zhang
• DOI: 10.1080/19475705.2026.2736941

Research Groups

• State Key Laboratory of Climate System Prediction and Risk Management, Nanjing Normal University, Nanjing, China
• Key Laboratory of VGE of Ministry of Education, Nanjing Normal University, Nanjing, China
• Jiangsu Center for Collaborative Innovation in Geographical Information Resource Development and Application, Nanjing, China
• Jiangsu Province Surveying and Mapping Engineering Institute, Nanjing, China

Short Summary

This study assesses urban flood risk in the Yangtze River Delta Urban Agglomeration from 1990 to 2020, integrating a spatial interaction network to account for structural dependencies among cities. It finds that network information significantly improves flood risk inference beyond local attributes, identifying distinct high-risk and structurally critical cities for differentiated management.

Objective

• To characterize the spatiotemporal evolution of urban flood risk in the Yangtze River Delta Urban Agglomeration (YRDUA) from 1990 to 2020 using a multidimensional H-E-V-R framework.
• To construct a flood risk spatial interaction network (FRSI-Network) based on spatial adjacency and directed hydrological connectivity.
• To identify structurally critical nodes within the FRSI-Network using centrality metrics (degree, connection strength, betweenness centrality).
• To evaluate the incremental predictive value of network information for urban flood risk inference compared to local city attributes using a staged XGBoost framework.

Study Configuration

• Spatial Scale: Yangtze River Delta Urban Agglomeration (YRDUA), comprising 27 cities.
• Temporal Scale: 1990 to 2020 (data points for 1990, 2000, 2010, and 2020).

Methodology and Data

• Models used:
  • Hazard-Exposure-Vulnerability-Resilience (H-E-V-R) framework for urban flood risk assessment.
  • Analytic Hierarchy Process (AHP) and linear weighted sum model for calculating the comprehensive flood risk index.
  • Network structural analysis (in-degree, out-degree, in-strength, out-strength, betweenness centrality).
  • XGBoost-based progressive inference framework, including:
    • Baseline model (local attributes only).
    • Network model (local attributes + network context).
    • Interaction model (baseline predictions + network-based residual correction).
• Data sources:
  • Geospatial: Administrative Boundary Data (Resource and Environment Science and Data Centre - RESDC, CAS), Digital Elevation Model (DEM) (USGS EarthExplorer), River Network Density (RESDC), Land Use/Land Cover (LULC) (CLCD Dataset, Wuhan University), Normalized Difference Vegetation Index (NDVI) (National Ecosystem Science Data Centre - NESDC), Building Footprints (OpenStreetMap - OSM), Sub-basin Data (HydroBASINS v1c, HydroSHEDS).
  • Meteorology: Daily Precipitation Data (National Meteorological Information Centre - NMIC, CMA).
  • Socio-economy: Population Statistics (China City Statistical Yearbook), Gross Domestic Product (GDP) Statistics, Unemployment Rate, Medical Resources.
  • Disaster: Historical Inundation Maps (Global Flood Database - GFDB), Historical Flood Records (EM-DAT).

Main Results

• Flood risk in the YRDUA exhibited significant spatial heterogeneity and temporal changes from 1990 to 2020, with high-risk cities concentrated in highly urbanized, riverine, and coastal areas of the eastern YRDUA. In 2020, Shanghai, Suzhou, Wuxi, and Nantong formed a contiguous high-risk cluster.
• Hangzhou, Xuancheng, and Wuhu were identified as the main structurally critical nodes due to their high betweenness centrality (0.314, 0.271, 0.218 respectively for Hangzhou, Xuancheng, Wuhu) and out-strength (2.840 for Hangzhou), indicating their prominent positions in cross-regional connectivity. Shanghai and Suzhou were identified as high-risk nodes requiring priority intervention, highlighting a distinction between local risk and structural importance.
• The staged XGBoost models demonstrated the incremental predictive value of network information for flood risk inference. The network model reduced the Mean Absolute Error (MAE) by approximately 5.7% (from 0.0240 to 0.0226) and Root Mean Square Error (RMSE) by approximately 6.8% (from 0.0309 to 0.0288) compared to the baseline model. The interaction model achieved the best overall performance, further reducing MAE by approximately 9.9% (to 0.0216) and RMSE by approximately 12.9% (to 0.0269) compared to the baseline, and improved risk classification accuracy to 0.8519.

Contributions

• Introduces a novel framework that integrates multidimensional flood risk assessment with spatial interaction network analysis, explicitly addressing structural dependencies (administrative adjacency and hydrological connectivity) among cities in urban agglomerations, a limitation in previous city-independent risk assessments.
• Identifies and distinguishes between high-risk cities and structurally critical nodes within the urban agglomeration network, providing a basis for differentiated flood risk management strategies that consider both local risk conditions and regional structural importance.
• Quantifies the incremental predictive value of network information for urban flood risk inference using a staged XGBoost approach, demonstrating that incorporating network context significantly improves prediction accuracy beyond local city attributes, without requiring event-scale causal flood propagation modeling.

Funding

• Key Laboratory of Land Satellite Remote Sensing Application, Ministry of Natural Resources of the People's Republic of China (Grant No. KLSMNR-K202501)
• National Key R&D Programme of China (Grant No. 2024YFC3808903)
• National Natural Science Foundation of China (Grant No. 42271483)

Citation

@article{Liang2026Urban,
  author = {Liang, Yiyin and Lu, Haipeng and Ma, Hangling and Wang, Yanmin and Song, Yanwen and Wu, Bo and Zhang, Shuliang},
  title = {Urban flood risk and the structural importance of cities in a spatial interaction network: evidence from the Yangtze River Delta Urban Agglomeration},
  journal = {Geomatics Natural Hazards and Risk},
  year = {2026},
  doi = {10.1080/19475705.2026.2736941},
  url = {https://doi.org/10.1080/19475705.2026.2736941}
}