Quantification of model uncertainty and variability in Newmark displacement analysis

Newmark displacement model has been extensively used to evaluate earthquake-induced displacement in earth systems. In this paper, model uncertainty and variability associated with the Newmark displacement analysis are systematically studied. Fourteen Newmark displacement models using scalar or vector intensity measures (IMs) as predictors are compared in this study. In general, model uncertainty for the vector-IM models is found smaller than that of the scalar-IM models, and remains consistent over different earthquake magnitude, distance and site conditions. Yet, the model uncertainty of these Newmark displacement models is still much larger than that of the ground-motion prediction equations (GMPEs) for IMs, indicating further development of the models is much needed. Considering the variabilities contributed from both GMPEs and Newmark displacement models, the total variability of the predicted Newmark displacements is rather consistent among the scalar- and vector-IM displacement models, due to extra sources of variability introduced by incorporating additional IMs. Finally, a logic tree scheme is implemented in the fully probabilistic Newmark displacement analysis to account for the model uncertainty and variability. Sensitivity analysis shows that specific weights would not significantly influence the displacement hazard curves as the results may be dominated by outlier models. Instead, selecting appropriate GMPEs and Newmark displacement models is more important in using the logic-tree framework.