The development of rill is the result of a complex interaction of soil properties with high spatial and temporal variability. The rill morphology may act as a determinant as well as a stochastically driven process. The description of their complex morphology requires geostatistics and nonlinear theories. A study of the morphological characteristics of rill evolution was performed via indoor soil pan rainfall simulations. Experiments were run with loessial soil (bulk density of 1.25g/cm3) under a rainfall intensity of 90mm/h with a 20° slope gradient. A detailed scan of the slope, topographic analysis, and statistical analyses were the main methods utilized in the study. Statistical analyses showed that characteristic parameters exhibited differences in the expression of the dynamic change of soil loss. According to the comparison of quantitative methods, we found that the fractal dimension can reflect the complexity of rill network, but it is not able to reflect the dynamic changes of the erosion process perfectly. In contrast, geomorphologic comentropy is sensitive to the dynamic changes of soil erosion intensity, and its variation can reflect the dynamic changes of erosion perfectly. The topological parameters can reflect the degree of stability in the rill network structure, but it is not suitable for reflecting the dynamic changes of erosion intensity. The geomorphologic comentropy was evaluated as the most appropriate characterization parameter for the rill morphology description.