Evaluating Predictive Performance of Global and Regional Ground Motion Models Using a New Caribbean Strong-Motion Dataset
Description:
Most of the existing probabilistic seismic hazard analysis (PSHA) studies for the Caribbean region adopted global and regional ground motion models (GMMs). However, there is a lack of systematic attempts to assess the predictive performance of these GMMs using local strong-motion recordings. We present such a study by first compiling a local strong-motion dataset consisting of 1639 recordings retrieved from the Center for Engineering Strong Motion Data (CESMD) and the Engineering Strong-Motion (ESM) Database. Among them, 729 CESMD recordings are from the 2019-2020 Southwest Puerto Rico earthquake sequence with moment magnitudes 4.5≤MW≤6.4, whereas 910 ESM recordings are from 197 events in Lesser Antilles with moment magnitudes 3.4≤MW≤7.4. We also pre-select global and regional GMMs applicable to the Caribbean region’s various tectonic regime characteristics, including the 2014 Next Generation Attenuation-West2 (NGA-West2) and the 2020 Next Generation Attenuation-Subduction (NGA-Sub) models. To facilitate the model comparison, multi-dimensional ground motion trends (e.g., with respect to spectral period, moment magnitude, rupture distance, VS30) are investigated using trellis plots. The performance and ranking of selected global and regional GMMs using the Caribbean dataset are assessed through residual analysis and probabilistic scoring methods such as the log-likelihood (LLH) by Scherbaum et al. (2004). The results show that global models such as the NGA-West2 and Cauzzi et al. (2015) outperform other crustal models, whereas NGA-Sub models perform reasonably well using the subduction dataset. The outcome of this study is used to construct data-driven and objective logic trees of GMMs for seismic hazard and risk assessment in the Caribbean region.
Session: 2025 Puerto Rico and the U.S. Virgin Islands National Seismic Hazard Model Update
Type: Oral
Date: 4/18/2023
Presentation Time: 05:15 PM (local time)
Presenting Author: Yenan Cao
Student Presenter: No
Invited Presentation:
Authors
Yenan Cao Presenting Author Corresponding Author Yenan.Cao@rms.com Moody's RMS |
Emel Seyhan Emel.Seyhan@rms.com Moody's RMS |
|
|
|
|
|
|
|
Evaluating Predictive Performance of Global and Regional Ground Motion Models Using a New Caribbean Strong-Motion Dataset
Category
2025 Puerto Rico and the U.S. Virgin Islands National Seismic Hazard Model Update