Review of statistical back analysis methods for road embankments built on soft clay soils: progression to Bayesian

Abstract This study presents a systematic review of statistical back analysis methods applied to embankments built on soft soil. Soft clay soils present a significant challenge to engineers due to their low shear strength, low hydraulic conductivity, large void ratios and high compressibility characteristics making overlying infrastructure susceptible to excessive deformation and instability. The inherent variability of these deposits when compared with other geomaterials makes it difficult to characterise their behaviour and predict future performance. Preloading the ground using embankments to induce consolidation is one of the oldest and most established techniques in geomechanics for improving poor ground conditions. To confirm that the embankment performance meets the prescribed deformation design criteria, consolidation behaviour is routinely monitored through field instrumentation within the framework of the Observational Method. This process of observational calibration and validation relies on back analysis where model parameters are refined based on measured field behaviour. Traditionally, back analysis has been performed by manually adjusting key parameters to fit a settlement curve or by using semi-empirical graphical methods. Statistical methods of back analysis, while conceptually well established, remain limited in practice. With the widespread availability of high-performance computers and recent advances in machine learning leveraging these capabilities, statistical approaches have become increasingly practical for solving complex geotechnical problems. This review provides a comprehensive summary of recent developments in statistical and Bayesian inference methods for back analysis with an emphasis on their application to road embankments constructed on soft clay foundations.