Abstract: There is a correlation between each loess strength parameters. In the traditional slope stability calculation, the loess strength parameters c and φ are often seen as two independent random variables. The neglect of their correlation leads to calculation deviation. In this paper, the two-dimensional joint normal distribution is used as the distribution profile of the loess strength parameters, and the Bayes formula is to optimize the parameters. Taking as an example of the potential landslide in the sewage outlet of Xi'an Xianyang Airport, the reliability index of the slope is calculated. The results show that Bayes makes it possible to reduce the variance of the estimated results, to get more reliable estimated results of loess strength parameter and what’s more, to avoid errors caused by small sample size. Correlation coefficient nearly has no effect on stability coefficient, but has much effect on the reliability index and the failure probability. The stability of the slope will be over-estimated without considering the positive correlation coefficient of c and φ value. The stability of the slope will be seriously under-estimated when the c and φ negative correlation is not taken into account. The estimated strength parameters from Bayes of two-dimensional joint normal distribution evaluate accurately the stability of loess slope.