Dual-Polarization Radar–Hydrodynamic Coupling and Uncertainty Assessment in Urban Pluvial Flood Simulations

Accurate urban-pluvial flood simulation demands rainfall inputs whose resolution matches the scales of hydrological response. Dual‑polarization weather radar, a cutting‑edge remote‑sensing platform, provides volumetric rainfall fields at sub‑kilometer resolution, yet its integration into process‑based flood models is hindered by scale mismatches, coupling uncertainties, and scarce validation. This study presents a multi-scale, radar-driven urban flood modeling framework that couples a scalable space–time rainfall cube with a bidirectionally linked 1D–2D hydrodynamic model, enabling real-time feedback between surface and drainage systems. A spatial projection and temporal interpolation method is developed to translate raw polar-coordinate radar echoes into rainfall grids that are dynamically aligned with hydrological simulation units. Using a factorial experiment across five spatial (100–2000 m) and five temporal (7.5–60 min) resolutions, we evaluate model performance under two severe storm events in Nanjing, China. Results show that simulation accuracy degrades markedly when spatial resolution exceeds 500 m or temporal resolution exceeds 15 minutes, reflecting a threshold beyond which radar inputs fail to resolve critical hydrological drivers. However, further refining the resolution to 100 m and 7.5 min yields only marginal improvement, suggesting diminishing returns under urban hydrological complexity and radar retrieval uncertainty. Station-specific sensitivity analysis reveals that terrain variability, drainage density, and land use heterogeneity jointly modulate model responsiveness to rainfall input resolution. This work provides a transferable methodology for integrating radar remote sensing with urban flood models and offers practical guidance for selecting input resolutions that balance computational efficiency with hydrological realism, and is applicable to data-rich cities facing flash-flood risks.