GIDPC.Green Infrastructure · Disaster Prevention
Research & Technology
Climate Risk ManagementUnder Review

AI-Enhanced Spatial Optimization of Permeable Pavement for Urban Stormwater Management: Surrogate Modeling, NSGA-II, and SHAP Analysis in Seocho-gu, Seoul

Under a 10-year design storm, full (100%) permeable-pavement coverage reduces runoff by 24.8% — but 73% of that reduction comes from the first 30%, marking 30–50% as the cost-effective sweet spot.

Authors
Junsuk Kang
Year
2026

Keywords

Permeable PavementSurrogate ModelXGBoostNSGA-IISHAP

Background

Rapid urbanization expands impervious surfaces and intensifies urban flood risk. Permeable pavement is a proven LID measure — but where and how much to deploy under fixed budget and area constraints is a separate multi-objective optimization problem.

Method

  • Automated geospatial processing (OSM + Copernicus DEM) builds EPA SWMM for 18 Seocho-gu sub-catchments
  • Scenario sweep across 0–100% coverage under a 10-year design storm
  • XGBoost surrogate (R² = 0.79) replaces expensive SWMM runs
  • NSGA-II multi-objective optimization delivers the Pareto front (runoff reduction vs installation area)
  • SHAP quantifies per-sub-catchment influence

Key Results

  • Baseline runoff 614,298 m³, 24.8% reduction at full coverage
  • Diminishing-return pattern: first 30% delivers 73% of full-coverage benefit
  • Cost-effective sweet spot: 30–50%
  • Pareto front: 14.7–22.7% runoff reduction / 3.8–63.5% installation area
  • Naegok-dong identified as priority investment site (largest catchment area)

Significance

This study provides a reproducible, scalable, and interpretable methodology for LID spatial optimization in dense urban watersheds — a decision-support tool that bridges quantitative recommendation and qualitative policy choice.