Abstract: Because of the heterogeneity and uncertainty of soils， soil parameters are anisotropic in space. In order to study the influence of soil parameter anisotropy on slope reliability， the two-dimension random field is discretized by the local average method with C + + language to obtain the covariance matrix. With the mean value of the parameters and the standard deviation input， the calculation model of slope random field was established. With combination of the finite difference strength reduction， the response surface method and the Monte-Carlo method， the influencing rule of the correlation distance in the model and the soil cohesion coefficient on slope reliability was analyzed. The results showed that with the calculation model of the random field， the slope failure probability increases with the increase of the correlation distance under the same variation coefficient of soil cohesive force in the model. The slope failure probability shows an upward trend along with the increase of the variation coefficient of the soil cohesive force under the same correlation distance in the calculation model of random field. Compared with the horizontal direction， the influence of the correlation distance variation on the slope reliability of the calculation model is more obvious in the vertical direction.

Key words: Slope reliability, Random field model, Local average method, Variation coefficient, Correlation distance