A novel heuristic flow direction algorithm for stream modeling in flat areas using digital elevation models

Accurate stream modeling from DEM depends on hydrologically consistent flow direction assignment. This remains challenging in flat areas where traditional algorithms often produce unrealistic streams. We propose a novel heuristic flow direction (NHFD) algorithm that reinterprets flat areas not as topographic dead zones, but as morphologically informative units. Unlike existing algorithms that rely on external terrain context or prior river networks, the NHFD algorithm automatically infers backbone flow paths within flat areas and integrates them into the calculation of flow direction. These backbone flow paths serve as the most probable flow conduits and form a guiding skeleton for flow direction assignment, ensuring hydrological consistency. The algorithm requires no external datasets or auxiliary inputs. Evaluated across three globally distributed small test sites and the Yangtze River Basin, NHFD produces stream networks with substantially fewer parallel streams and better alignment with actual river centerlines, outperforming the D8, PF, THFD, MAZE, and DINF algorithms. By transforming flat terrain into an active source of hydrological inference, NHFD enhances the accuracy and robustness of large-scale, automated stream mapping in low-relief environments.