Significance of quantifying uncertainties in probabilistic modeling and a possible approach to select the best: A study using SPT- and CPT-based liquefaction case histories

There have been several statistical methods developed to evaluate the probability of seismically induced soil liquefaction. Among these methods, the logistic regression has been the most widely used method to model the probability of liquefaction using the Standard Penetration Test (SPT)- and Cone Penetration Test (CPT)-based case histories. However, the predicted probabilities using logistic regression can be different based on the uncertainties related to distribution of explanatory variables, significance of coefficient of explanatory variables, and distribution of liquefaction to non-liquefaction case histories. In this paper, a possible approach has been developed to select the logistic regression model that best addresses these uncertainties for SPT- and CPT-based case histories using various statistical tests. This approach is developed for the most updated CPT-based case histories from Ku et al. (2012), and SPT-based case histories from Idriss et al. (2010). Further, the logistic regression model is compared with the existing probabilistic models for CPT- and SPT-based case histories. This study shows that without considering the aforementioned uncertainties, the probability values can be significantly different and can give false sense of risk of liquefaction for a particular site of interest. © 2014 American Society of Civil Engineers