The accuracy and efficiency of the simulations in distributed hydrological models must depend on the proper estimation of flow directions and paths. Numerous studies have been carried out to delineate the drainage patterns based on gridded digital elevation models (DEMs). The triangulated irregular network (TIN) has been increasingly applied in hydrological applications due to the advantages of high storage efficiency and multi‐scale adaptive performance. Much of the previous literature focuses mainly on filling the depressions on gridded DEMs rather than treating the special cases in TIN structures, which has hampered its applications to hydrological models. This study proposes a triangulation‐based solution for the removal of flat areas and pits to enhance the simulation of flow routing on triangulated facet networks. Based on the drainage‐constrained TIN generated from only a gridded DEM by the compound point extraction (CPE) method, the inconsistent situations including flat triangles, V‐shape flat edges and sink nodes are respectively identified and rectified. The optimization algorithm is an iterative process of TIN reconstruction, in which the flat areas are generalized into their center points and the pits are rectified by embedding break lines. To verify the proposed algorithm and investigate the potential for flow routing, flow paths of steepest descent are derived by the vector‐based tracking algorithm based on the optimized TIN. A case study of TIN optimization and flow path tracking was performed on a real‐world DEM. The outcomes indicate that the proposed approach can effectively solve the problem of inconsistencies without a significant loss in accuracy of the terrain model.