Uncertainty quantification for compressibility and settlement response of clays

Uncertainties in the compressibility properties of clayey subsoil are among the most significant risks related to ground-supported infrastructures. The uncertainty in settlement prediction can be dealt with by overly conservative design or by uncertainty quantification. The input needed for reliability analysis, such as the quantified inherent variability and transformation uncertainty, may be estimated using indicative literature values. It remains unclear how applicable these global summaries are to soft Finnish clays, which are marked by special characteristics, such as high sensitivity and compressibility. Consequently, Janbu method (tangent modulus method) is often preferred over the compression index method, and its cross-correlated parameters may complicate the reliability analysis. Hence, various extensively studied clay sites in Finland have been used to quantify the inherent variability in compressibility properties. In addition, a database of Finnish clay soils was compiled and used to quantify the transformation uncertainties in existing and newly derived empirical correlations for predicting the compressibility of clays. The quantified uncertainties were used to perform reliability analyses aimed at estimating the probability distribution of settlements for test embankments. The results showed that the inherent variability for Finnish clays mostly aligned with existing literature. However, applying the global summaries to Finnish clays may underestimate the mean compressibility and overestimate the variability within homogeneous clay layers. Similarly, the transformation uncertainties that were quantified using the database of Finnish clays were relatively smaller than global summary. However, since these results were based on a limited collection of sites, global summaries should also be considered when characterising the uncertainties for Finnish geotechnical conditions. The case studies of reliability analyses demonstrated that the cross-correlation between Janbu parameters needs to be considered using the copula function or approximation by modulus number modification; otherwise, the uncertainty in the settlement response is overestimated. It was also shown that using cautiously chosen characteristic values for all the settlement parameters (i.e. the semi-probabilistic method of Eurocode 7) can lead to an overly conservative design and inconsistent safety margins. Hence, reliability analysis should be preferred if possible. Regarding the Janbu method, Monte Carlo simulation with copula was found to provide the most accurate representation of uncertainty in settlement response. However, if the compression index method is used, the first-order second-moment method (FOSM) can provide an adequate estimate of the uncertainty in the settlement response, especially when applied to normally consolidated clays.