Variations in Building-Resolving Simulations of Tsunami Inundation in a Coastal Urban Area

Abstract

The direct simulation of inundation in developed urban areas presents a much greater challenge than the more common bare earth simulations that use roughness, which are used in many tsunami studies. This study intercompares the performance of four longwave models for tsunami inundation on a detailed topographical model of Kainan, Wakayama, Japan, with laboratory results. All simulations include buildings, which have a large impact on overland flood propagation. Inter-model comparisons yield several apparent characteristics: (1) variations between models were small in areas that are always wet; (2) wetting, drying, and overland propagation increased inter-model variation in the inundation front arrival time, maximum water surface elevation, and overland flow velocities; (3) inundated areas and maximum water surface elevations show lower inter-model variation (V) than inundation front velocity and maximum current velocities. Sources for V appeared to occur from differences in wetting, drying, and detailed code implementation rather than major differences in model physics. Using published tsunami fragility models, V led to significant differences in the predicted damage. Differences were largest for fragilities that used velocity and lower for fragilities that only used maximum inundation depths. Based on these results, inundated areas and water levels from building-resolving simulations might be assigned relatively higher confidence, and all the predicted velocities should be considered to have a greater error and potentially should be considered only when using ensembles.