Capturing the Epistemic Uncertainty in Subduction Earthquake Rupture Parameters

ABSTRACT Subduction earthquakes are highly destructive and can trigger devastating tsunamis, highlighting the need for a consistent probabilistic seismic-tsunami hazard assessment. However, current approaches treat both hazards separately: tsunami models use static coseismic slip distributions, whereas seismic models rely mostly on empirical ground-motion models. This creates inconsistencies because both depend on different representations of the source. Here, we address this inconsistency by systematically quantifying epistemic uncertainties in ground-motion simulations derived from kinematic rupture models based on slip distributions of large subduction earthquakes. This provides a physically consistent source representation and establishes a basis for future integrated seismic-tsunami hazard assessment. We use ground-motion data recorded during the 2015 Mw 8.3 Illapel earthquake in Central Chile as a test case. We test different rupture velocities and rise-time parameterizations, as well as hypocenter locations, using two 1D velocity models to capture the epistemic uncertainty in these kinematic rupture parameters and to evaluate their impact on ground motion. We compare the resulting simulations with the observed ground motion at eight stations. We find that Bias and root mean square error decrease when adopting depth-dependent rupture velocities, with 0.3 of shear-wave velocity (VS) in the shallow domain (<10 km) and 0.5VS at depths greater than 15 km. Rupture velocity is then identified as the key parameter controlling ground-motion variability in kinematic source models. For this event, the simulations fit the observations better when the regional velocity model is used, and they more accurately reproduce ground motion in the 0.5–3 Hz frequency band.