This study proposed a random Smoothed Particle Hydrodynamics method for analyzing the post-failure behavior of landslides, which is based on the Karhunen–Lo(e)ve (K-L) expansion, the non-Newtonian fluid model, and the OpenMP parallel framework. Then, the applicability of this method was validated by comparing the generated random field with theoretical result and by simulating the post-failure process of an actual landslide. Thereafter, an illustrative landslide example was created and simulated to obtain the spatial variability effect of intal friction angle on the post-failure behavior of landslides under different coefficients of variation (COVs) and correlation lengths (CLs). As a conclusion, the reinforcement with materials of a larger friction angle can reduce the runout distance and impact the force of a landslide. As the increase of COV, the distribution range of influence zones also increases, which indicates that the deviation of influence zones becomes large. In addition, the cor-relation length in Monte Carlo simulations should not be too small, otherwise the variation range of influence zones will be underestimated.