This paper develops a multiple response surface method for probabilistic analysis and reliability sensitivity analysis of slopes considering spatially varying soil properties.The proposed method is applied to probabilistic analysis and reliability sensitivity analysis of slopes to explore the effect of spatial variability on the slope reliability.The results indicate that the proposed method achieves high computational efficiency in reliability sensitivity analysis.It can be used to effectively evaluate the slope reliability considering the spatial variability of soil properties with reasonable accuracy.The autocorrelation functions might have some influences on the slope reliability.The failure probability of slope stability obtained from the single exponential autocorrelation function is relatively small compared with the values of failure probability obtained by using other four autocorrelation functions.On the other hand

using cosine exponential autocorrelation function gives a failure probability that is slightly larger.In addition

the square exponential

second-order autoregressive and binary noise autocorrelation functions almost yield similar slope reliability analysis results.In addition

the differences in the slope reliability using different autocorrelation functions become relatively obvious when the coefficient of variation is relatively small and the scale of fluctuation of soil properties is relatively large.