A Bayesian ground-motion model for volcano-tectonic earthquakes in southwest Iceland

Victor Moises Hernández Aguirre , Rajesh Rupakhety

Volcano-tectonic earthquakes in Iceland and other volcanic regions produce ground motions that differ systematically from those of ordinary shallow crustal tectonic earthquakes. The recent unrest on the Reykjanes Peninsula has provided an exceptional opportunity to quantify these differences, with multiple intense swarms preceding and accompanying eruptive episodes. Building on Hernández-Aguirre et al. (2023; HR23), we develop an expanded empirical ground-motion model (GMM) for southwest Iceland using strong-motion recordings from 2021–2024. The dataset comprises >270 earthquakes (Mw ≥ 3.5) and >4000 three-component records within Repi ≤ 120 km. We model PGA, PGV and 5%-damped pseudo-spectral acceleration over a broad period range for both horizontal and vertical components. The GMM is formulated as a Bayesian hierarchical mixed-effects regression with event and station random effects and calibrated using integrated nested Laplace approximation (INLA). Informative priors and a two-step calibration are used to reduce station–distance confounding under uneven station–distance coverage. The updated model confirms robust volcano-tectonic scaling characterized by elevated near-field amplitudes and rapid attenuation relative to tectonic models. Approximately 20% of records with Mw ≥ 4.2 and Repi ≤ 40 km exhibit a PGA ratio V/H > 1, indicating that vertical seismic demand can be significant in near-source conditions. Compared with HR23, the total standard deviation is reduced, consistent with improved median predictions from added path/depth terms and site classification. The resulting model provides a statistically robust basis for PSHA and scenario-based risk assessment in Icelandic volcanic environments within the calibrated domain.