An Analysis of Liquefaction-Induced Free-Field Ground Settlement Using 1,000+ Case Histories: Observations vs. State-of-Practice Predictions

Liquefaction-induced free-field settlements can be attributed to several mechanisms, including volumetric reconsolidation strain, ground loss due to ejecta, and vertical components of lateral spreads and flow slides. In practice, the volumetric-strain settlement is typically predicted using one of several semi-empirical methods, but there is no consensus of which, if any, is most accurate. Moreover, no method has been proposed to predict the magnitude of settlement owed to ejecta, even though this component could be larger than any other. The lack of a complete and trusted model for predicting free-field settlement is due, in part, to the scarcity of case-history data needed to apportion and model its individual components. However, the 2010–2011 Canterbury, New Zealand, earthquake sequence (CES) resulted in a liquefaction dataset of unprecedented size and quality, presenting a unique opportunity to assess and advance liquefaction analytics. Accordingly, this study compiles and analyzes 1,013 high-quality settlement case-histories from the CES, such that the contributing components of settlement can be reasonably accounted for and quantified. Presented herein are preliminary findings of this effort, summarized as follows: (1) popular CPT-based models for predicting volumetric-strain settlement exhibit strong bias, such that smaller total settlements (<~6 cm) are underpredicted while larger total settlements (>~6 cm) are significantly overpredicted; (2) these poor predictions cannot be explained by the uncertainty of CPT proxies of relative density, which are central to predictions of volumetric strain; (3) introduction of a depth-weighting function lessens the prediction bias, showing promise for future study; and (4) prediction-performance is related to the observed volume of liquefaction ejecta, illustrating the need to account for and model this mechanism of settlement. Collectively, the findings suggest popular methods for predicting free-field settlement have significant room for improvement.