Performance Comparison of Probabilistic and Deterministic Liquefaction Triggering Models for Damage Assessment in 23 Global Earthquakes

Abstract

Analyses of an extensive database resulting from the 2010-2011 Canterbury, New Zealand, earthquakes have shown that existing liquefaction damage indices (e.g., the Liquefaction Potential Index: LPI) have significant shortcomings when used with deterministic liquefaction triggering models. These findings have raised questions as to whether probabilistic triggering models would be more efficacious. Accordingly, this study compiles nearly 10,000 liquefaction case histories from 23 global earthquakes to compare the performance of probabilistic and deterministic triggering models, used within the LPI framework, for predicting the surficial manifestation of liquefaction. In doing so, LPI is modified to accept the probability of liquefaction triggering (PL), in lieu of the factor of safety against liquefaction triggering (FSliq). It is shown that the modified LPI performs worse, suggesting that PL either provides no statistically distinguishable benefits in terms of prediction accuracy, or is presently accounted for in such a way that is not optimal. In conjunction with previous studies, these findings support the need for an alternative liquefaction damage index that more adequately accounts for the mechanics of liquefaction manifestation.