Assessing consistency of ground-motion models with recorded data – a case study from Puerto Rico and the U.S. Virgin Islands

Brad T. Aagaard, Morgan P. Moschetti, & Kyle B. Withers

Submitted September 7, 2025, SCEC Contribution #14521, 2025 SCEC Annual Meeting Poster #TBD

For the U.S. Geological Survey (USGS) National Seismic Hazard Model, we require a suite of ground-motion models that span the geographic extent of the National Seismic Hazard Model, which includes 50 states and five territories. In each region, we want ground-motion models that are consistent with local data. We present a case study from Puerto Rico and the U.S. Virgin Islands that demonstrates how we adjust ground-motion models from the Next Generation Attenuation West2 and Subduction Projects, as well as a global ground-motion model for active crustal earthquakes, to better fit the data in this region. We assess the consistency of the ground-motion models against recorded peak horizontal acceleration, velocity, and spectral acceleration using linear mixed-effects regression and determine which components require adjustment. In this case, region-specific modifications to the linear site response terms substantially reduce systematic over-predictions of ground motions, particularly for spectral acceleration at periods ranging from 0.3 to 3 seconds.

For our analysis, we used the USGS software gmprocess to compile a dataset comprising nearly 5,000 ground-motion records since 2006 at 67 stations from 366 earthquakes with magnitudes ranging from 4.0 to 6.5 and depths reaching up to 200 km. More than half of the ground motions were recorded after January 2020. We classify earthquakes as active crustal, subduction interface, or subduction intraslab events based on the orientation of the moment tensor and hypocenter relative to the Slab 2 subduction zone geometry model. We consider both site surveys and proxy values for the time-averaged shear wave speed in the top 30 m (Vs30). For each ground-motion model and ground-motion metric, we examine the overall bias and trends with predictive variables, including earthquake magnitude, depth, rupture distance, and Vs30.

Key Words
linear mixed effects, site response

Citation
Aagaard, B. T., Moschetti, M. P., & Withers, K. B. (2025, 09). Assessing consistency of ground-motion models with recorded data – a case study from Puerto Rico and the U.S. Virgin Islands. Poster Presentation at 2025 SCEC Annual Meeting.

Related Projects & Working Groups
Ground Motions (GM)