Simplified Bayesian Ground Motion Models for Cumulative Absolute Velocity in Central and Eastern North America

Cumulative absolute velocity (CAV) has recently emerged as a useful intensity measure (IM) for predicting the occurrence of liquefaction and its consequences, including foundation settlement. However, few ground motion models for CAV exist that are applicable in Central and Eastern North America (CENA). The relative paucity of strong motion data for this region and tectonic setting, particularly for large earthquakes and short distances to rupture, is the primary challenge hindering the development of such models. This study applies a Bayesian approach to develop a ground motion model for CAV in CENA. This approach consists of first developing a model using a large database (drawn from the NGA-West2 database), then updating the coefficients in light of observations from a smaller database which is specific to the region of interest (drawn from the NGA-East database). The models developed using the Bayesian approach are compared with using the same functional form in a traditional regression strategy with the NGA-East data, as well as with the model regressed with the NGA-West2 data and with other models in the literature. The Bayesian approach prevents overfitting in the NGA-East data, where few records are available for large magnitude earthquakes at short distances. The use of NGA-West2 data to constrain development of models for CENA follows existing studies that use NGA-West2 models as a baseline and develop adjustments to make the models applicable in a different region and tectonic setting. The Bayesian approach proposed in this study is also applicable for developing other region-specific ground motion models for regions that lack data compared to regions such as California, New Zealand, and Japan that have relatively rich data available.