Based on the random finite element method (RFEM) and Monte-Carlo simulations, a non-stationary random field model is established for the simulation of slopes with depth-dependent soil strength parameters or slopes with weak interlayers. The most likely failure surface is obtained by the radiation scanning method, and the sliding depth and volume of the non-stationary random field slope are simulated. Then the sliding depth and volume obtained by this method are compared with those of the homogeneous slope and the stationary random field slope, so as to study the influence of the non-stationary random field slope with the depth-dependent soil strength parameters on the sliding depth and volume under different slope angles and anisotropy conditions. The results indicate that the study of non-stationary random field sliding depth and volume based on RFEM provides a referential analysis angle for the geological hazard risk and consequence evaluation. The higher the depth dependence of the parameter (the greater the change rate b), the slower the average sliding depth and volume of the slope gradually decreases. In the process of evaluating the risk consequences of a large slope landslide, a preliminary estimation can be made based on the results of the homogeneous slope analysis. In a non-stationary random field, as the anisotropy increases, the variation coefficient of the sliding volume at all angles increase. Under different slope angles, the greater the correlation of the soil slope in the horizontal direction, the greater the uncertainty of the weak interlayer in the soil slope, and the larger the variation coefficient of the sliding volume.