The research, led by Man Wang of Wuhan Business University and Zihan Guo of China University of Geosciences, tackles a long-standing weakness in urban flood modeling. Traditional approaches tend to compute the underground one-dimensional pipe network and the two-dimensional surface flow field in isolation. A pipe-network model alone can flag which manholes overflow, but it cannot show where that water spreads once it reaches the street. A surface model alone, conversely, has no way of knowing how much water the buried network can actually convey, and so risks dramatically over- or underestimating local hazard. The mismatch, the authors argue, has hindered deterministic hydraulic support for flood-control engineering in exactly the places that need it most.

To close that gap, the team built an integrated 1D–2D framework linking the U.S. Environmental Protection Agency’s Storm Water Management Model (SWMM) with LISFLOOD-FP, a raster-based surface inundation model. SWMM handled the first half of the problem: transforming rainfall into runoff across 845 sub-catchments and routing it through a reconstructed drainage network of 869 nodes and 890 conduits, solved with the Dynamic Wave scheme that captures pressurized flow, backwater propagation, and even flow reversal when downstream river levels push back. LISFLOOD-FP then took over, receiving time-varying overflow hydrographs from each surcharged node as dynamic water sources and simulating how the escaping water redistributed across a 12.5-meter digital elevation model of the district.

The coupling itself is a technical feat. Sub-catchment boundaries were delineated using Thiessen polygons around each drainage manhole, then cross-checked against terrain slopes derived from the 12.5-meter DEM to weed out geometrically plausible but hydrologically inconsistent assignments. When SWMM detected that a node was overflowing, the overflow volume during each coupling interval was transferred to the corresponding LISFLOOD-FP grid cell according to node coordinates, becoming a time-varying source term in the two-dimensional calculation. The result is a closed loop that connects pipe surcharge to street ponding in a single, physically consistent simulation.

The design storm was no ordinary rain. The team adopted a 120-minute Chicago synthetic hyetograph with a 20-year return period, discretized at 5-minute intervals into 24 rainfall pulses. A peak-position coefficient of r = 0.405 placed the storm’s maximum intensity roughly 48.6 minutes after onset, preserving both the pre-peak accumulation that saturates surfaces and the post-peak recession that reveals how long water lingers. That temporal structure matters: a uniform rainfall input would smooth away exactly the transient surcharge dynamics the researchers wanted to capture.

Under this scenario, the numbers are stark. Ninety-six of the 869 modeled drainage nodes—11.05 percent of the network—overflowed, releasing a cumulative 128,400 cubic meters of water onto the surface. The longest continuous overflow lasted 2.18 hours, and the average was 0.92 hours. But the overflow was far from uniform. Twenty-four overflowing nodes clustered in the Linkonggang Avenue–Wuhuan Road area and twenty-one in the Taxi Road–Fushan Avenue area, together accounting for nearly half of all overflow. Simulated maximum inundation depths at representative road sections ranged from 0.32 to 0.48 meters, with the Taxi Road–Fushan Avenue section reaching 0.48 meters and persisting for an average of 5.5 hours before receding at just 0.07 meters per hour.

What explains this spatial concentration? The answer, the study shows, lies in the coincidence of three factors: runoff generation, local conveyance limits, and downstream hydraulic constraints. Dongxihu’s comprehensive runoff coefficient is 0.56, but highly hardened industrial and road surfaces—industrial and warehousing land alone covers 28.11 percent of the district—push local coefficients to 0.70, compared with just 0.25 for green spaces and squares. When the receiving-water level at the terminal outfall rises, the hydraulic gradient available for gravity drainage shrinks, backwater propagates upstream through the connected pipe network, and surcharge intensifies. In the Taxi Road–Fushan Avenue area, low-lying terrain, a restricted stormwater box-culvert outlet, and pipe-diameter bottlenecks compounded the problem. By contrast, the Huanhu Road section, buffered by surrounding green space and storage areas, produced only seven overflow nodes—7.29 percent of the total—and its inundation receded fastest, at 0.28 meters per hour.

Because continuous observations of pipe discharge, nodal hydraulic head, and surface water depth were unavailable for the study area, conventional calibration was impossible. The researchers instead relied on two complementary lines of evidence. First, they tuned SWMM parameters—surface-flow resistance on impervious areas, depression storage, and the maximum Horton infiltration rate—over five rounds until the simulated comprehensive runoff coefficient reached 0.561, within 0.18 percent of the area-weighted benchmark of 0.56. Two observed rainfall events in June and August 2025, with totals of 130.22 and 101.10 millimeters, yielded event-based runoff coefficients of 0.512 and 0.488, deviations of −8.73 and −13.01 percent from the calibrated reference. Second, the simulated high-inundation zones were cross-checked against historical waterlogging records from the Dongxihu District Emergency Management Bureau and field investigations. Areas with depths exceeding 0.4 meters aligned closely with officially designated high-risk sites along Taxi Road, Jinshan Avenue, and Huanhu Middle Road.

The authors are careful about what this agreement does and does not prove. The runoff coefficient constrains total runoff magnitude but says nothing about where inundation occurs; the spatial check confirms the principal distribution of hotspots but not the accuracy of grid-scale depths or durations. The 12.5-meter terrain resolution, while adequate for district-scale hotspot identification, cannot resolve individual curbs, building thresholds, or stormwater inlets—features that can materially alter local hydraulic connectivity. Operational degradation such as inlet blockage, sediment accumulation, and structural deterioration was not modeled dynamically, meaning local inundation may be underestimated where the real network performs below its assumed capacity. The results, the team emphasizes, should be read primarily as scenario-based identification of waterlogging hotspots rather than fully field-validated hydraulic predictions.

The implications for urban flood policy are concrete. The study argues that fragmented fixes—simply enlarging pipes or building embankments—ignore the subsurface–surface hydraulic connectivity that turns pressurized conduits under river backwater into drivers of street overflow. Instead, mitigation should be spatially prioritized and mechanism-driven: in upstream industrial zones constrained by legacy infrastructure, source-control measures such as green roofs, permeable pavements, and urban wetlands should cut peak runoff before it reaches the network; in low-lying bottleneck sections like Wangan Avenue and Taxi Road, targeted pumping stations and detention facilities are needed to force water out when gravity drainage fails. Future work, the authors suggest, should incorporate drainage-system deterioration explicitly, use field inspection data to define reductions in inlet capture and conduit capacity, and add time-dependent gate and pumping operations—turning a scenario tool into a genuinely predictive instrument for the cities that will need it most.

Subject of Research: Coupled 1D–2D hydrodynamic simulation of urban waterlogging in Dongxihu District, Wuhan, using integrated SWMM and LISFLOOD-FP modeling under a 20-year design storm

Article Title: Hydrodynamic simulation and risk evolution of urban waterlogging based on integrated SWMM–LISFLOOD modeling: a case study of Dongxihu District, Wuhan

Article References: Wang, M., Guo, Z., Zhang, Q., Chen, L., Bao, X., & Zou, J. (2026). Hydrodynamic simulation and risk evolution of urban waterlogging based on integrated SWMM–LISFLOOD modeling: a case study of Dongxihu District, Wuhan. Frontiers in Earth Science, 14, Article 1976240. https://doi.org/10.3389/feart.2026.1976240

Image Credits: AI Generated

DOI: 10.3389/feart.2026.1976240

Keywords: urban waterlogging, SWMM, LISFLOOD-FP, hydrodynamic modeling, drainage network, Wuhan, flood risk, design storm, node overflow, surface inundation, runoff coefficient, 1D-2D coupling