• An efficient approach is proposed for risk analysis of spatially variable slopes. • An empirical method is developed to identify representative slip surfaces. • A two-layered soil slope is studied to illustrate the proposed approach. • The risk of slope failure can be accurately evaluated with much less efforts. • The contribution of each failure mode to the risk can be effectively quantified. Quantitative risk assessment of slope failure is an important prerequisite for formulating rational strategies for landslide risk mitigation and developing a landslide risk-based warning system. Efficient risk assessment of slope failure in two-dimensional (2-D) spatially variable soils is a challenging problem. This paper proposes an efficient approach for quantitative risk assessment of slope failure considering the 2-D spatial variation of soil properties in the limit equilibrium analysis framework. To facilitate the risk assessment of slope failure, an empirical method is developed for the identification of slope key failure modes. With the proposed approach, the key failure modes of the spatially variable slope can be well identified and their contributions to the risk of slope failure can be effectively quantified. Finally, a two-layered soil slope is studied to illustrate the proposed approach. It is found that the prominent advantages of the proposed approach include being computationally efficient and accounting for the effect of the correlations among the failure modes.