Abstract: A Bayesian theory based model is developed for the assessment of regional terrain stability. This model combines the statistical model with the infinite slope stability model. The uncertainty of parameters in the assessment of regional terrain stability is eliminated because the parameters can be adjusted according to the real historical events. The adjustment of parameters are based on the assumption that the failed slope is stable before it rains and unstable after it rains. First, the initial distribution of parameters is set up. Then, the final distribution of parameters is calculated through MCMC(Markov Chain Monte Carlo) methods. This model is applied to the landslide events in a watershed, in Caiyuan Village, Fujian Province, China. The parameters of the model are adjusted using the proposed approach in this study case, where the adjusted dimensionless cohesion C of the area is 0.028,the adjusted effective friction angle is 16.7, and the adjusted ratio of hydraulic conductivity to rainfall is 529.026m. The unstable areas take up to 76.0% of the whole area. This model, which automatically adjusts parameters according to historical landslide events, would be a useful tool for landslide hazard assessment.