ABSTRACT
Probabilistic models for the pairwise correlations between undrained shear strengths and the chosen test indices are necessary for the Bayesian analysis. Therefore, efforts were made in compiling a correlation database, e.g. the Su vs. SPT-N data points shown in Figure 1, from the literature to estimate the probabilistic models for the pairwise correlations. For the database, it is always a concern whether the compiled data points are enough to cover a wide range of possible scenarios. The following guidelines are followed to mitigate this concern: (a) Unless mentioned explicitly, data points in the database do not include those from special clays, e.g. fissured and organic clays. Therefore, the corresponding correlations should not be applied to those soils; (b) In the case that all data points of a particular correlation are from the same geographical region, that correlation may be only applicable to that region. In general, the applicability to other regions is questionable. Bayesian updating will not reduce the physical limitations inherent in existing pairwise correlations. It merely provides a more rational and systematic method for combining information that will serve as a useful complement to engineering judgment.
