Abstract: Establishing a reasonable rainfall infiltration model is an important prerequisite for revealing the rainfall-induced slope failure mechanism and disaster prevention and control. The traditional Green-Ampt model does not consider the distribution of soil stratification and transition layer formed by rainwater infiltration. This results in a large deviation on the calculated infiltration rate. Thus, the traditional model is difficult to apply to the spatially varying slopes. A method is proposed for calculating the infiltration rate of an arbitrary soil layer. The thickness of transition layer is estimated based on the relationship among the infiltration rates underlying different soil layers. Based on this, an improved Green-Ampt model is proposed to analyze the rainfall infiltration process in the slope considering the spatial variability of saturated hydraulic conductivity of soil. The improved Green-Ampt model is further applied to an infinite slope to analyze its seepage and stability for both homogeneous and heterogeneous soils under the rainfall infiltration. The results obtained from the improved model are systematically compared with those obtained from a traditional Green-Ampt model and the numerical solutions of Richards equation. The results indicate that the distribution of water content and factor of stability calculated from the proposed improved model are more consistent with the numerical solutions of Richards equation than those of the traditional Green-Ampt model. The proposed improved model can lay a solid theoretical foundation for analyzing the rainfall infiltration processes in the heterogeneous slopes, and formulating effective measures for the prevention and control of rainfall-induced landslide disasters. Additionally, it is found that there is a dependence between the thickness of transition layer and the saturated hydraulic conductivity as well as the infiltration rate and volumetric water content at its top, while it is not directly related to the total depth of rainwater infiltration.