Do physics-based models improve predicted ground motion variability? Insights from dynamic rupture simulations of the 2023 Turkey Earthquake Sequence

One of the key challenges of empirical ground-motion models is the ability to capture ground-motion variability, which may stem from different source, path and site effects. This challenge may be addressed using simulated data from physics-based, non-ergodic earthquake simulations. Dynamic rupture models capture the nonlinear interaction of source, path and site effects in a self-consistent way and once integrated with observations, reproduce a variety of geodetic and seismic data well to first order (e.g., Taufiqurrahman et al., 2022; Jia et al., 2023; Gabriel et al., 2023). However, these models may not fully account for the variability in ground motions, particularly in the orientation, periods (Tp), and amplitudes of long-period pulses (Yen et al., 2024). In addition, the physical mechanisms underlying high-frequency radiation remain debated (Graves & Pitarka, 2016; Ben-Zion et al., 2024). 

In this study, we investigate the effects of incorporating both on-fault and structural small-scale heterogeneities within 3D dynamic rupture models of the 2023 Turkey earthquake doublet. Specifically, we focus on how these heterogeneities influence rupture dynamics, together with the spectral content and variability of the modelled ground motions. We analyse the impact of small- and large-scale fractal on-fault roughness, a heterogeneous distribution of fracture energy (Dc) and the dynamic friction coefficient (𝜇d), and initial supershear rupture speed compared to sub-shear earthquake initiation.

Our findings reveal that rupture dynamics are most significantly influenced by the introduction of an initial supershear rupture speed, which results in an expected larger seismic moment along the nucleating Nurdaği-Pazarcik splay fault and an earlier triggering of the East Anatolian Fault (EAF) compared to the other models. Although this leads to a greater overall energy release, the release pattern along the EAF remains fairly consistent across all models, suggesting that the added ingredients primarily act to amplify seismic moment rather than drastically alter rupture dynamics. Additionally, Dc heterogeneities have the most significant influence on long-period pulse properties. In contrast, small-scale roughness and 𝜇d heterogeneities exhibit a damping effect on pulse period (Tp) as they mostly influence high-frequency radiation. However, these modifications fail to translate into significant changes in the overall spectral content across the different models. Notably, despite the added heterogeneities, the pulse orientations remain predominantly fault normal and are only minimally impacted by Dc heterogeneities, supershear rupture speeds, and large-scale roughness. 

This study demonstrates that incorporating a heterogeneous distribution of fracture energy has one of the strongest impacts on both the rupture dynamics and frequency content of 3D dynamic rupture simulations, further contributing to a better understanding of how different dynamic rupture heterogeneities influence ground shaking, a critical step towards comprehensively capturing ground-motion variability and enhancing physics-based seismic hazard assessment.