Key points are not available for this paper at this time.
Dual-grid models address the computational bottlenecks of large-scale (>103 km2) urban flood modeling with solution updates on a coarse grid that are informed by topographic data on a fine grid. However, dual-grid models may poorly resolve levees, leading to loss of accuracy. Here we present a grid edge classification method whereby specific edges of the coarse grid are flagged to gather nearby topographic data from the fine grid and create a contiguous physical barrier. The method relies on levee location data in a polyline format, and does not require levee height data since that information is stored on the fine grid. Using a 6804 km2 model of the Los Angeles Metropolitan Region with 3 m topographic data and 987 km of levees, the proposed method is implemented and evaluated. Simulations using coarse grids of 15, 30 and 60 m capture flood extent consistent with fine-grid models based on a critical success index (CSI) of 90, 87 and 82%, respectively. Edge classification improves CSI up to 7 percentage points over a model with unclassified coarse grid edges, and reduces the false alarm ratio up to 10 percentage points. Differences in model performance across the study area are noted, including lower accuracy on urbanized alluvial fans. With compute costs that scale with the coarse grid, dual-grid models can efficiently realize more accurate large-scale models of urban flood hazards.
Kahl et al. (Fri,) studied this question.