A Regionalized Geospatial Liquefaction Model for California

Global geospatial liquefaction models are currently being used for rapid response mapping after earthquakes and for scenario and risk assessment. These models are simple to implement as they use globally available geospatial inputs (e.g., slope-based VS30, global water table depth, precipitation, and distance to nearest water body). Prior work has demonstrated that the global geospatial liquefaction models can be regionally biased, and recent efforts have worked to develop regionalized geospatial liquefaction models. In this work, we will present a regionalized geospatial liquefaction model for California, which uses a Bayesian logistic regression framework to update the global liquefaction model. The regionalized geospatial liquefaction model uses the global geospatial liquefaction model as the prior and then uses liquefaction and nonliquefaction observations from past earthquakes in California to update the model coefficients. The updated California geospatial liquefaction model has reduced the bias and the uncertainty in predicting liquefaction after earthquake events in California.