A Framework for Probabilistic Assessment of Liquefaction Manifestation

As part of the next generation liquefaction (NGL) project, we are developing probabilistic triggering and manifestation models using laboratory data and cone penetration test (CPT) case histories in the NGL database. The case histories are used to develop probabilistic models for surface manifestation conditional on susceptibility, liquefaction triggering, soil properties, stratigraphic details, and other features. Susceptibility is interpreted as a sole function of soil composition and is expressed as a probabilistic function of soil behavior type index, I_c, obtained from CPT. A triggering model is derived based on laboratory tests on high-quality specimens from literature; this model captures mean responses and uncertainty reflective of data dispersion and is considered as a Bayesian prior that will subsequently be updated by field observation data. A manifestation model is then regressed from field case histories where surface manifestation was or was not observed, information on soil conditions that enables identification of layers likely to liquefy, and ground shaking conditions. We describe the approach applied to develop our manifestation model; for a given layer this model considers layer depth, thickness, CPT tip resistance, and I_c. The result of this process is a logistic function in which manifestation probability decreases with increasing depth, decreasing thickness, increasing tip resistance, and increasing I_c. Profile manifestation is then derived by aggregating individual layer manifestation probabilities.