Fragility Functions for Liquefaction-Induced Ground Failure

Abstract

The predicted severity of liquefaction manifested at the ground surface is a popular and pragmatic proxy of damage potential for infrastructure. Toward this end, the liquefaction potential index (LPI) and similar models are commonly used, and often codified, to predict surface manifestations on level ground. These predictions typically use deterministic thresholds from the literature—obtained via calibration on case-history data—to classify the expected manifestation. While widely adopted, such thresholds obscure the uncertainty of expected outcomes and are incompatible with probabilistic frameworks. Proposed thresholds are also intimately tied to the liquefaction analytics used to compute them and to the methodology used to select them, each of which can conflict with forward applications, leading to erroneous predictions. Accordingly, using 15,223 case histories from 24 earthquakes, this study develops fragility functions that probabilistically predict surficial manifestations of liquefaction using triggering and manifestation models popular in practice. Deterministic workflows are easily extended by selecting appropriate fragility coefficients; options are provided for six cone penetration test (CPT)–based triggering models, one CPT-inversion filter, three manifestation models, and three manifestation severities. The model application is demonstrated by predicting (1) liquefaction manifestations in Christchurch, New Zealand, resulting from an Alpine Fault earthquake, wherein a logic-tree is used to ensemble predictions from 18 models, and (2) the return period of liquefaction manifestations in the South-of-Downtown (SODO) district of Seattle, wherein predictions are compared to historical observations.