Geng et al. (2025) Analyzing Pluvial Flooding Influenced by Urban Road Network Metrics Based on Hydrodynamic Simulation and SHAP Values

Identification

• Journal: Water Resources Management
• Year: 2025
• Date: 2025-12-26
• Authors: Yanfen Geng, Peng Liu, Xiao Huang
• DOI: 10.1007/s11269-025-04410-4

Research Groups

• Department of Port & Waterway Engineering, Southeast University, Nanjing, China

Short Summary

This study quantitatively investigates the influence of urban road network structure metrics on pluvial flooding using 2D hydrodynamic simulations, statistical analysis, and interpretable machine learning. It reveals that regular grid networks enhance drainage efficiency, while cul-de-sac types exacerbate flooding, with topography and network connectivity metrics exhibiting spatially heterogeneous impacts on flood depth and velocity.

Objective

• To analyze how the spatial distribution of flooding is influenced by different community road network forms.
• To identify the intrinsic relationships between road network structure metrics and flooding, and quantitatively represent them.
• To elucidate how the flood-inducing effects of these structure metrics exhibit spatial heterogeneity across various geographic units.

Study Configuration

• Spatial Scale: Community-scale (approximately 1.41 km²) in Jiangning District, Nanjing, China. Simulations were also conducted on seven typical road network forms.
• Temporal Scale: Hydrodynamic simulations were performed for four rainfall return periods: 10 years, 20 years, 50 years, and 100 years, using the Chicago rainfall pattern.

Methodology and Data

• Models used:
  • Hydrodynamic Simulation: Two-dimensional shallow-water equations (depth-averaged) discretized using the finite volume method on unstructured grids, with the HLLC approximate Riemann solver.
  • Rainfall Model: Chicago rainfall pattern, parameterized for Nanjing.
  • Statistical Analysis: Spearman correlation analysis, Mann-Whitney U test.
  • Machine Learning: Random Forest (RF), Geographically Weighted Random Forest (GWRF), SHAP (Shapley Additive exPlanations) values for interpretability analysis.
• Data sources:
  • Elevation Data: NASA Earth dataset (12.5 m resolution Digital Elevation Model - DEM).
  • Landcover Data: Jiangsu Province dataset (2021).
  • Road Network Models: Generated through CityEngine for typical forms; high-resolution satellite imagery for actual urban community road networks.
  • Rainfall Data: Nanjing storm intensity formula.

Main Results

• Road Network Forms: Regular grid networks (e.g., x5) demonstrate optimal drainage performance, with peak flood depths 21.67% lower than the average maximum of other forms and minimal severe inundation (0%). Cul-de-sac-type networks are more flood-prone due to limited drainage outlets.
• Flood Distribution Patterns: Grid-type networks exhibit a bipolar distribution with higher proportions of both light and severe inundation, while cul-de-sac networks show lower extremes but larger proportions of moderate inundation.
• Key Influential Metrics:
  • Topography-driven metrics: Elevation (EV) and Depression Degree (DD) are identified as the most influential drivers for flooding, with EV showing a strong negative correlation with flood depth (r up to -0.537).
  • Topological connectivity metrics: Edge Betweenness Centrality (EBC) and Depth Value (DV) significantly affect runoff pathways, with DV showing a negative correlation with flood depth (r up to -0.421).
• Rainfall Return Period Impact: Correlations between metrics and flood depth generally strengthen with increasing rainfall return periods. Topography-driven metrics' correlations with maximum flow velocity also strengthen, while topological connectivity metrics' correlations with maximum flow velocity decline.
• Elevation Impact: The highest correlation coefficients for metrics are observed in higher elevation groups (E3 and E4), indicating a more pronounced influence on flooding in these areas.
• Local Structure Metrics: Dead-end roads (DER) are significantly associated with flood depth (rank-biserial correlation coefficient, rbc = 0.559) and maximum flow velocity (rbc = 0.865), prominently guiding runoff. Four-way intersections (FWI) show a statistically significant but small effect size.
• Spatial Heterogeneity: SHAP-based interpretability analysis reveals pronounced spatial heterogeneity in the influence of key metrics (e.g., local SHAP_EV ranging from −0.072 to 0.675), with topography-driven metrics being robust in real terrain, while topological connectivity metrics become crucial in idealized uniform slope conditions.

Contributions

• Developed a novel quantitative framework integrating 2D hydrodynamic simulation, statistical analysis, and geographically weighted SHAP values to systematically analyze the influence of road network structure metrics on urban pluvial flooding.
• Provided a comprehensive evaluation of flood depth and velocity distributions across seven representative road network forms, identifying optimal and flood-prone configurations.
• Proposed and examined ten structure metrics, offering a multi-level characterization of road network features (topological connectivity, topographic drivers, geometric complexity, local connectivity/disconnection).
• Quantified the intrinsic relationships between these structure metrics and urban flooding, and elucidated their spatial heterogeneity, enhancing the understanding of structure-flooding coupling mechanisms.
• Offered critical scientific evidence and practical guidance for urban road network planning and flood risk mitigation strategies, particularly emphasizing the often-overlooked role of road network connectivity.

Funding

• The National Natural Science Foundation of China (No. 51979040).

Citation

@article{Geng2025Analyzing,
  author = {Geng, Yanfen and Liu, Peng and Huang, Xiao},
  title = {Analyzing Pluvial Flooding Influenced by Urban Road Network Metrics Based on Hydrodynamic Simulation and SHAP Values},
  journal = {Water Resources Management},
  year = {2025},
  doi = {10.1007/s11269-025-04410-4},
  url = {https://doi.org/10.1007/s11269-025-04410-4}
}