Incorporating inherent uncertainties in seismic liquefaction assessments

Abstract

Liquefaction assessments for tailings facilities have significant uncertainty, yet their results are typically presented as single-value deterministic factors of safety (FoSs) to compare against standard practice guidance. Some of the uncertainty is due to inherent material variability or measurement error associated with in situ testing techniques. There is also a considerable – and often ignored – contribution from the transformation models used to convert in situ measurements to design parameters. For seismic liquefaction assessments, the key sources of transformation model uncertainty are from the empirical relationships relating standard penetration testing or cone penetration testing (CPT) results to cyclic resistance and residual undrained strength. While the uncertainties associated with the material variability and measurement error will vary by site and can usually be reduced through additional investigation, the model uncertainties are constant and irreducible without further work. This paper shows how the transformation model uncertainties associated with CPT-based liquefaction assessments can be incorporated into design practice to estimate probabilities of failure and lead to very different outcomes than simply considering FoS alone.