A New Pseudospectral Acceleration Ground-Motion Model for Subduction Zone Earthquakes in Chile

This study develops new ground-motion models (GMMs) for pseudospectral accelerations (PSAs) in Chilean subduction zones (interface and intraslab) using an extensive strong-motion data set that combines ground motions from different sources and spans a broad range of magnitudes, distances, and site conditions. The GMMs are developed for interface and intraslab tectonic settings. The study proposes a data–physics hybrid regression process that balances empirical fit with physical constraints by bounding large-magnitude, short-distance scaling, where Chilean data are scarce. Two GMM versions are derived, constrained by Next Generation Attenuation (NGA)-Sub South American and NGA-Sub global bounds, representing alternative branches of epistemic uncertainty. The developed models are also discussed in the context of recently developed data-driven GMMs for Chile, highlighting the inherent limitations of purely data-driven approaches for extrapolation to design scenarios. Compared with existing Chilean and South American GMMs, the new models better capture ground-motion scaling for small-to-moderate events and are constrained at engineering design scenarios. A Chile site amplification term consistent with other regions also emerges, in contrast to the results reported by Abrahamson and Gülerce (2020) for South America. Short-period aleatory variability is attributed to small-magnitude records and kappa effects; the final standard deviation model is developed using M ≥ 5.5 data and incorporates distance dependence, resulting in variability levels comparable to existing ergodic models for Chile. A single-station-sigma model is also developed to isolate site-specific effects, further reducing the aleatory variability to about 0.45 (ln units). An inspection of spatial site-term patterns suggests complex effects potentially linked to near-surface and deep geologic features, opening the door for future nonergodic models. Interperiod correlation residual analyses reveal stronger correlations between short- and long-period Chilean PSA residuals compared to global models. This is likely due to ergodic GMMs in Chile not capturing some site, source, or path effects, especially at long periods. The proposed GMMs and regression framework provide a regionally calibrated, physically consistent, and statistically robust basis for seismic hazard assessment, performance-based seismic design, and seismic risk assessment in Chile, while offering a transferable methodology for other regions with limited large-magnitude, short-distance data.