Epistemic Uncertainty Treatment in Seismically Induced Slope Displacements Using Polynomial Chaos

Abstract

Performance-based probabilistic approaches (PBPAs) for estimating seismically-induced slope displacements (D) provide hazard-consistent estimates through the evaluation of D hazard curves (DHCs), and hence, its use in practice is appealing. Importantly, the epistemic uncertainty to develop a DHC needs to be considered, which is typically done using a logic tree approach with discrete branches for the system properties and seismic displacement models (SDM). However, the few existing SDMs, do not allow one to accurately capture the full epistemic uncertainty range. This study uses the polynomial chaos (PC) theory to develop a computationally efficient framework for propagating the epistemic uncertainty in the median D, associated with alternative SDM models. PC expansions allow one to account for the epistemic uncertainty distribution in DHCs in a computationally efficient manner, which cannot be possible using the traditional logic tree approach. The necessary steps for the implementation of the proposed framework are discussed, and an illustrative example of its application in engineering practice is presented.