Probabilistic framework for characterising analytical model uncertainty in geosynthetic-reinforced soil walls

Quantifying the main sources of uncertainty associated with analytical models for predicting reinforcement loads is essential for ensuring safety, supporting reliability analyses, and enabling reliability-based optimization in geotechnical design. This paper presents a probabilistic framework to characterise uncertainty in load prediction models for geosynthetic-reinforced soil walls. Model uncertainty is assessed through a Bayesian updating framework that compares analytical predictions with monitoring data from well-instrumented walls, while explicitly accounting for the uncertainties associated with model parameter estimation. The random variables considered include the soil unit weight and the angle of internal friction. The applicability of the framework is illustrated using two simple analytical methods: the American Association of State Highway and Transportation Officials (AASHTO) Simplified Method and the K-Stiffness Method. This framework also allows model uncertainty to be updated as new monitoring data become available. In addition, it can be extended to other analytical or numerical models commonly employed for the design of geosynthetic-reinforced soil walls. This flexibility enhances its potential for practical adoption and for guiding future reliability-based studies.