Parallel algorithm of watershed delineation by considering the topological relationship into information organisation

Abstract

Watershed boundaries are vital in ecology, hydrology, and geomorphology. With advances in Earth observation, larger and more accurate datasets demand efficient parallel methods for watershed delineation that minimize memory usage and boost efficiency. Existing algorithms often suffer from high memory consumption or low performance. Currently, the Sub-First-Round (SFR) watershed is an essential component of data-partitioning-based parallel watershed delineation. Our approach revises SFR encoding rules within tiles and adopts a new data organization strategy, reducing memory usage while improving efficiency. Compared to traditional methods (WDG, MESHED, Hiep-Thuan), our algorithm saves 6.13 %–59.68 % in memory usage. Though slower than WDG and MESHED, this trade-off is necessary due to data partitioning. The significant reduction in memory usage makes it feasible to perform watershed delineation on much larger datasets. Finally, our proposed algorithm outperforms the Hiep-Thuan algorithm by 11.84 % in efficiency.