Extraction of Drainage Network from Grid Terrain Datasets Using Parallel Computing

Drainage-network extraction from grid terrain datasets has a marked influence on many applications such as hydrologic analysis, soil erosion and geomorphology. However, the drainage-network extraction is normally very time-consuming using the sequential program, especially for grid terrain datasets with high resolutions and large scopes. This paper proposed a set of parallel algorithms to preprocess DEM, determine flow routing and delineate drainage network in an MPI programming model. Both depressions and flat areas can be processed by the DEM preprocessing algorithm. For using the number of upslope-dependence neighbors of a cell, the flow-routing algorithm makes the flow-accumulation calculation similar to a local algorithm. The experiment results show that the proposed parallel approach to extract the drainage network performs much better than the sequential algorithm and has a better parallel efficiency.