Abstract: Bayes’ theory provides an effective tool to properly characterize the spatial variability of soil properties and quantify the effect of site investigation data (e.g., undrained shear strength data) on reliability of slope stability. However, site investigation data sequentially appears at different spatial locations of a slope, and the model to characterize the spatially varying soil properties (e.g., random field model) usually involves a great number of uncertain parameters. These pose a great computational challenge for Bayesian updating of slope reliability considering spatially varying soils. This paper proposes a collaborative reliability updating approach of slope stability with spatially varying soil properties considering changing site investigation data. It first makes use of Bayesian updating with structural reliability methods (BUS) to simulate random fields and perform slope stability analyses, and then employs the rejection sampling principle and collaborative analysis to characterize the spatially varying soil properties and update the reliability of slope stability considering different test data. As site investigation data spatially appears within a slope, repeated simulations of conditional random fields and a significant number of slope stability analyses are avoided. Moreover, the combination of BUS makes it possible for efficient slope reliability updating using Bayesian analysis that involves high-dimensional model parameters. A single-layered soil slope with a non-stationary random field is employed to demonstrate the effectiveness and validity of the proposed approach. It shows that the proposed approach provides an effective tool for dynamic characterization of soil spatial variability and real-time reliability updating of slope stability under changing site investigation data.