Abstract: A Bayesian framework is developed to probabilistically identify the underground stratigraphy based on I_c data. The proposed Bayesian framework identifies the most probable soil layer boundaries with the consideration of spatial variability of I_c and quantifies the uncertainties in the underground stratigraphy, which provides valuable information for making future site investigation plans and geotechnical designs. A subset simulation-based Bayesian updating algorithm (CBUS) is used to generate posterior samples of soil layer thicknesses and to calculate the model evidence for determining the most probable number of soil layers and the most probable soil layer boundaries, and the standard deviations of boundaries are calculated to quantify the uncertainty in soil layer boundaries. Finally, the proposed approach is illustrated and verified using the real I_c data obtained from a deep excavation site at Yili station of Shanghai No. 10 subway line and simulated I_c data from a virtual site. The results show that the underground stratigraphy identified by the proposed approach is based on the statistical similarity of I_c data. With the increase of statistical difference in I_c data within two adjacent soil layers, the standard deviation of the soil layer boundary between them decreases, and the soil layer boundary identified by the proposed approach is more reliable, and vice versa.