Sensitivity analysis of tsunami heights to shallow bathymetric resolution

Abstract

Both retrospective tsunami analyses and assessments of future tsunami hazards depend on accurate modeling of how tsunami waves generated offshore propagate through shallow waters near the coast. Accurate models of tsunami propagation in shallow water require high-resolution bathymetric maps, but these are often inaccessible because of the time and cost required to acquire them. In addition, tsunami models based on high resolution bathymetry have high computational processing requirements. Hence, it has been common to use globally available datasets with coarser resolutions, such as the GEBCO dataset, in modeling.Here, we examine how variations in bathymetric resolution, from 5 m to ∼455 m (GEBCO), affect simulated coastal tsunamis. Our case study includes four study sites with available LiDAR bathymetry datasets (1 m resolution). At each site 30 sets of points were randomly extracted from the LiDAR bathymetry datasets and used to generate bathymetric grids with resolutions of 5, 10, 20, 30, 40, 50, 100, 200, and 300 m at each site. These were also compared to a bathymetry based purely on the GEBCO dataset for that region (∼455 m resolution), that we modified to match the coastlines of the other bathymetry models. Tsunami waves offshore were generated by setting up an instantaneous rupture sourced from a hypothetical fault model and we used the commonly used COMCOT software to model tsunami propagation towards the coast.Using the model run with 5 m resolution bathymetry as a high resolution reference model, we observed that bathymetric grids with resolutions of 10 – 50 m can reproduce coastal wave heights reasonably well, with the maximum wave height overestimated by ≤5% or underestimated by ≤10%. For coarser bathymetric grids, however (≥100 m resolution), there is an increasing trend of underestimation. Wave heights are underestimated by at least 10% and with up to 30%, 40% and 60% underestimation for bathymetric resolutions of 100, 200, and 300 m, respectively. Notably, the commonly used GEBCO model underestimated coastal tsunami heights by as much as 70%. We also examined the impact on tsunami arrival time: and found that resolutions of 10 – 50 m exhibited a first wave arriving ∼10% earlier than expected, while coarser resolutions showed more variability, with the first wave arriving either ≤20% later or ≤10% earlier. For GEBCO-based models, the  arrival time estimate tends to be underestimated by 10 – 30% or overestimated by 20 – 50%. Our study demonstrates that using GEBCO bathymetry in numerical modeling of tsunami wave propagation in the coastal region likely leads to a significant underestimation of the wave height, with the wave also predicted to arrive too early. However, a reasonably accurate result can be achieved using a bathymetric resolution in the 10 m – 50 m range, and is achievable with reasonable computational efficiency. This study highlights the importance of shallow bathymetry in the numerical modeling of tsunami propagation.