Abstract: The soil on the ablation zone is degraded in shear strength parameters due to repeated wet and dry cycles, and there is spatial variability in the parameters, both of which are key factors affecting slope stability, while most of the existing studies only consider one of them. For this reason, a new method for analyzing the stability and reliability of slopes that considers both factors is proposed. In this, a parameter random field is simulated using the Karhunen-Loève method, and dimensionality reduction is performed using sliced inverse regression, which in turn leads to the construction of an extreme gradient boosting surrogate model based on the augmented whale optimization algorithm. The Three Gorges Reservoir Area Baishuihe landslide is analyzed as an example to explore the effects of degradation of shear strength parameters and spatial variability of the ablation zone on the probability of landslide failure. The results show that: the proposed method can greatly improve the computational efficiency and accurately estimate the probability of landslide failure (P_f); landslide P_f increases with the number of degradation times of the parameters of the fallout zone and tends to stabilize after the fourth time; the spatial variability of saturated permeability coefficients has a small effect on the reliability results when water level changes are not taken into account, while the spatial variability of the effective internal friction angle has a higher effect on the distribution of the factor of safety than that of the effective cohesive force.