Do 3D Dynamic Rupture Models Capture the Variability in Long-Period Velocity Pulses? Insights from the 2023 Mw 7.8 Kahramanmaraş Earthquake

Capturing ground motion variability, especially in near-fault long-period velocity pulses, is a key challenge for seismic hazard assessment. Empirical methods often rely on simplified assumptions and may not fully capture the non-linear interplay of source, path, and site effects. Physics-based dynamic rupture simulations offer a self-consistent alternative, but their ability to reproduce variability in near-fault ground motions, such as velocity pulse orientation, period, and amplitude, remains uncertain. We systematically investigate the effect of fault geometry and on-fault heterogeneities in a suite of 3D dynamic rupture simulations of the 2023 Kahramanmaraş, Türkiye, earthquake. We compare dynamic rupture scenarios that separately incorporate large-scale fault waviness, fractal fault roughness, heterogeneous critical slip-weakening distance, heterogeneous dynamic friction, and supershear versus subshear initiation, each resolving up to at least ~1 Hz. We systematically analyse the influence on rupture dynamics, frequency content, and long-period pulse variability, while ensuring all models have comparable seismic moment rate release. While all models capture near-fault pulse amplitude variability, supershear initiation and fracture energy heterogeneity exert the strongest influence on pulse period and orientation. Despite added complexity, most modelled pulses remain predominantly fault-normal, contrasting observed broader ranges of orientations, but supershear rupture speed locally increases variability in pulse orientation. We discuss a simpler main fault model incorporating >700 off-fault fractures, which increases variability in both pulse amplitude and orientation, highlighting the importance of fault zone complexity. Incorporating both heterogeneous on-fault frictional properties and off-fault complexity is promising for advancing realistic, non-ergodic ground motion models and physics-based seismic hazard assessment.