Abstract: A two-dimensional empirical eigenfunction model is proposed to analyze and predict the variations of riverbed profile. This model consists of two along and cross-river eigenfunctions and one temporal eigenfunction. These eigenfunctions can be generated based on a series of measured riverbed profiles. The proposed model is applied to analyze the riverbed variations of the downstream of Chou-Shui River in Taiwan. The prediction of riverbed variations is also conducted by using the generated eigenfunctions with the help of the Markov process model.A dimensionless effective flow discharge is assigned to be the forcing parameter in the predicting model, which controls the riverbed changes. The results show that the proposed method can reasonably analyze and predict the riverbed variations.