Abstract: Most existing bimodal soil-water characteristic curve (SWCC) models are derived from the empirical models for unimodal SWCC, in which the parameters have no clear physical meanings or value ranges generally. When the parameters are determined by these models, there are always multiple solutions, thus the theoretical researches on the bimodal SWCC are restricted to a large extent. According to the capillary theory, the bimodal SWCC is usually associated with the bimodal pore size distribution (PSD). On account of this, a fractal description method for the bimodal PSD is presented by using Menger sponge model, and a physical model for the bimodal SWCC is established. The established model is employed to fit twenty sets of published data of bimodal SWCC, and compared with the models proposed by other scholars. The results show that the proposed model has a good fitting performance. Furthermore, a method for pore classification is proposed from the established bimodal SWCC model. It is proved that the results obtained by this method are relatively accurate, with the advantages of simple application and clear theoretical basis.