Revisiting of London Clay Simple Correlations Using a Bayesian Approach

The London Clay (LC) is an Eocene stiff clay which underlies a large proportion of London with thickness up to 130 meters. As such it has been extensively studied, and there exists many empirical relationships that estimateits stiffness and strength properties based on some simpler parameters (e.g. depth, plasticity index, etc.). Such relationships are routinely used in practice, especially on small construction projects where ground investigation is often too limited for establishing site-specific correlations. However, there is an increasing trend for geotechnical practices, codes and standards to move towards reliability-based methods and wider use of statistical evaluation of the variations in the geotechnical parameters. It requires rigorous methods to quantify both the total uncertainty of the measured property (including inherent soil variability, measurement errors and statistical uncertainties) and the uncertainty of the transformation model.
This paper investigates the derivation of one of these simple correlations, the variation of undrained strength with depth, using a Bayesian statistical approach tobetter understand these uncertainties. The study relies on a large database gathered during the site investigation for the construction of a majorrail infrastructure project running East-West across London. The analyses explore the predictive capacity of different models including hierarchical models to investigate the variability between different LCsub-units. In practice, the obtained correlations may then be combined with site-specific data from an appropriate site investigation using a Bayesian approach.

Keywords: Stiff clay, London Clay, Undrained strength, Unconfined undrained, Bayesian analyses, Correlation, Uncertainties, Soil properties.