Global Shear Wave Velocity Database for Probabilistic Assessment of the Initiation of Seismic-Soil Liquefaction

Engineering practitioners commonly use penetration-based methods (SPT & CPT) for assessment of seismic liquefaction triggering hazard. On the horizon, shear wave velocity (Vs) may offer engineers a third tool that is lower cost and provides more physically meaningful measurements. Development of the shear wave velocity liquefaction method has been hampered by a paucity of published velocity profiles; particularly in deeper soil deposits (>10m) and deposits subjected to high cyclic stress ratios (CSR > 0.3). A review of the literature reveals that most historic liquefaction sites fitting this depth and CSR criteria are located in Asia, though most of these sites remain untested for Vs. To remedy this scarcity of data, we set out to assemble a global Vs dataset by acquiring new data in Japan, Taiwan, China, India, and the United States (US). These data are merged with the exiting catalog of published velocity data. To acquire new field data, we use the recently developed continuous swept-sine wave spectral analysis of surface waves test (CSS-SASW). The CSS-SASW test has proven to be extremely reliable at rapidly gathering high signal-to-noise dispersion data sufficient to invert 20-40 meter Vs profiles. So far, we have acquired new velocity profiles at nearly 300 liquefaction-evaluation sites throughout Asia and the US, mostly at sites previously tested by conventional penetration methods. This new dataset represents the majority of the worlds documented sites of liquefaction occurrence since instrumental recording. To correlate the global shear wave velocity data set with likelihood of initiation of seismic-soil liquefaction, we utilize high-order probabilistic tools (Bayesian updating) developed for structural reliability. A multi-parameter limit-state function for liquefaction triggering is modeled and evaluated based on the means, distributions and uncertainties of each modelvariable. Each case history is then sub-divided into ‘quality’-ranking categories based on the conjugate-uncertainties of CSR and Vs1. A lowpass cut-off of the coefficient of variation is used filter-out poorly constrained sites. Finally for the probabilistic analysis, the Bayesian updating procedure is used to iteratively compute coefficients for the limit-state function that minimize model error. The intended outcome of this effort is a new evaluation of the Vsliquefaction-triggering boundary in light of a global data set and modern limit-state probabilistic tools.