Rill erosion is the main form of water erosion, and is an important link on the transition from sheet erosion to gully erosion, and is an important component of water erosion physical process on hillslope. As the rill erosion process with dynamic, random and temporal variation of the three major characteristics of nonlinear dynamics, therefore, traditional erosion model is difficult to accurately dynamic simulate and reproduce rill initiation and development process on slope. Most of the existing erosion and sediment yield models focus only on the general purpose of soil erosion prediction, while the lack of space-time simulation of dynamic rill morphology evolution processes. Combination of soil erosion and hydraulics knowledge, the cellular automata theory and methods into the process simulation of rill erosion, the establishment of a loess slope CA model rill erosion, rill slope during the development of a numerical simulation, and simulation results and experimental results are compared and analyzed, the model simulation results and experimental rill produce similar patterns, to verify reasonableness of the model design; In addition, by comparing the parameters of the rill erosion model to simulate the error analysis.