Slip-Informed Ground-Motion Models in Subduction Zones

ABSTRACT Ground-motion models (GMMs) are fundamental tools for seismic hazard analysis but typically rely on simplified distance metrics that fail to capture the spatial heterogeneity of energy radiated during earthquake ruptures. In this study, we evaluate the performance of GMMs that incorporate rupture heterogeneity through slip distributions and interseismic coupling, comparing them to conventional formulations based on the minimum-distance metrics. We selected all interface events from the Next Generation Attenuation for subduction zone regions project database for which at least one coseismic slip model was available, resulting in 19 earthquakes (Mw>6.7) and approximately 1400 ground-motion records. From this dataset, we developed both traditional RRUP-derived GMMs and alternative GMMs that account for rupture heterogeneity through mean distance metrics (RP). Residual analyses demonstrate that slip-informed GMMs yield systematically lower interevent variability and reduced total uncertainty, with the interevent standard deviation (τ) decreasing by up to 30% globally. These improvements are most pronounced at short-spectral periods (T ≤ 1.0 s) and for megathrust earthquakes (Mw>8.1), in which τ drops by up to 50% relative to RRUP-based models. Strikingly, these localized performance peaks are particularly relevant for seismic engineering and hazard assessment. Furthermore, we evaluated the use of pre-event interseismic locking models as forward-modeling tools, finding remarkable consistency with coseismic slip-informed predictions. Although physically informed approaches introduce additional epistemic uncertainty due to source model parameter variability, they represent a critical step toward enhancing the physical realism and predictive capability of subduction-zone GMMs.