Spatial variabilities of soil properties are usually modeled by stationary or weakly stationary random fields in slope reliability analysis. However

abundant in-situ data demonstrates that the means and standard deviations of soil properties such as undrained shear strength change with the soil depth. Thus

it is important to develop non-stationary random field models and approaches to model spatial variations of soil properties. To tackle the uncertainties of the trend component and fluctuation of component in existing models

this paper proposes an effective non-stationary random field model of undrained shear strength. The approach for modeling the two-dimensional non-stationary random fields of soil properties is also presented. Thereafter

the effectiveness of the proposed model is illustrated by comparisons with the existing non-stationary random field models and stationary random field models of soil properties systematically. A clay slope under undrained conditions is investigated to explore the effect of the non-stationary distribution characteristics of undrained shear strength along the depth on the slope reliability. The results indicate that the proposed model can simulate the uncertainties of the trend component and fluctuating component separately. Both the means and standard deviations of soil properties increase with the depth. It provides an effective means for characterizing the non-stationary distribution characteristics of soil properties. Compared with the slope reliability analysis results underlying the non-stationary random fields

it will lead to an underestimation of the probability of failure and dangerous slope engineering designs when the commonly-used stationary random fields are utilized to model the spatial variability of soil properties.