To improve computational efficiency without accuracy losses for surface flow simulation, a dynamic bidirectional coupling model based on multigrid was applied. The flood-prone areas, where the inundation conditions were needed to detail, were discretized using finer grids, while the rest of the domains were assigned coarser grids to reduce computation time. Different time steps were adopted in coarse and fine grids. This paper presented an approach to quantitatively assess the computational efficiency under different domain discretization and grid generation conditions. Three test cases were considered to demonstrate the performance of the applied model. The results showed that the model had high computational efficiency while maintaining satisfactory simulation accuracy. The computational efficiency of the model was not only related to the size ratio of coarse to fine grids, but the area ratio of coarse grids to the entire computational domain. If the area ratio of coarse grids to the computational domain was constant, the computational efficiency improved exponentially with the increasing size ratio. If the size ratio of coarse to fine grids was constant, the computational efficiency improved linearly with the increasing area ratio. The proposed method offers a promising option for quantitatively assessing the computational efficiency of models.