Abstract: The accuracy of excavation response prediction models is generally influenced by various uncertainties, including those related to soil parameters, model uncertainties, measurement errors. Bayesian methods provide a novel way to reduce and/or quantify these uncertainties, and is a natural framework for improving model predictions by systematically integrating prior knowledge with observational data. However, existing Bayesian updating methods typically addressed the uncertainties with soil parameters or/and model biases, while the measurement errors are ignored.. Besides, correlations between different excavation stages are also overlooked for mathematical convenience. These simplifications may lead to unreliable predictions in practice. In this study, a novel Bayesian updating method is proposed, which simultaneously incorporates uncertainties in soil parameters, model bias, observational errors, and stage correlations. Two case studies are used to illustrate and validate the method. The results demonstrate that the proposed approach significantly enhances the accuracy of semi-empirical models in predicting excavation responses across different soil types.