Quantification of Tsunami Bathymetry Effect on Finite Fault Slip Inversion

The strong development of tsunami instrumentation in the past decade now provides observations of tsunami wave propagation in most ocean basins. This evolution has led to the wide use of tsunami data to image the complexity of earthquake sources. In particular, the 2011 M(w)9.0 Tohoku-Oki earthquake is the first mega-event for which such a tsunami instrumentation network was available with an almost complete azimuthal coverage. Source inversion studies have taken advantage of these observations which add a lot of constrain on the solutions, especially in the shallow part of the fault models where other standard data sets tend to lack resolution: while on-land data are quite insensitive to slip on the often-distant shallow part of a subduction fault interface, tsunami observations are directly sensitive to the shallowest slip. And it is in this shallow portion that steep bathymetry combined with horizontal motion, the so-called bathymetry effect, can contribute to the tsunami excitation, in addition to the direct vertical sea-bottom deformation. In this study, we carefully investigate the different steps involved in the calculation of this bathymetry effect, from the initial sea-floor deformation to the prediction of the tsunami records, and evaluate its contribution across the main subduction zones of the world. We find that the bathymetry effect locally exceeds 10 % of the tsunami excitation in all subduction zones and 25 % in those known to produce the largest tsunami, either from mega- or tsunami- earthquakes. We then show how the bathymetry effect can modify the tsunami wave predictions, with time shifts of the wavefront and amplitudes sometimes varying by a factor of two. If the bathymetry effect can have a strong impact on the simulated tsunami, it will also affect the solution of the finite-fault slip inversion. We illustrate this later aspect in the case of the Tohoku-Oki earthquake. We find that not accounting for the bathymetry effect will not necessarily cause strong variations in the spatial extent of the inferred coseismic rupture but can severely distort the solution. We also find that the bathymetry effect improves the consistency of the slip model inverted from tsunami data with seafloor geodesy observations, implying that taking the bathymetry effect into account reduces the epistemic uncertainties on tsunami modeling. Implementing this easily quantifiable effect in the tsunami early warning system could thus lead to improved estimates of the tsunami impact across ocean basins.