Abstract:Watershed subdivision is an important basic work of distributed hydrological model, which is related to the accuracy of basin hydrological simulation and prediction. In the process of watershed subdivision in large-scale and complicated terrain area, the traditional method has some shortcomings in accurately describing the spatial extent of the study area, determining the outlet of the basin, and extracting the river network of the inflow area, which commonly uses a single catchment area threshold to extract the river network and divide the watersheds based on surface runoff and flow model. Consequently, a watershed subdivision method, using stem-branch topological codification, was proposed based on automatic recognition of the outlets of basin and the fusion of river network with variable catchment area thresholds. The method mainly includes the determination of basin outlets, the determination of the maximum and minimum catchment area thresholds, the multi-threshold virtual river network integration and watershed subdivision and coding. Taking China as an example, 25 174 watersheds were divided, with an average area of 381 km², a maximum area of 5 512 km² and a minimum area of 4 km². The coincidence degree between the simulated area and the actual area reached 99.92%. While using the traditional method, the number of watersheds was 21 768, and the average area was 434 km², and the coincidence degree was only 98.5%. Moreover, the results were compared with the actual water systems and the current basin of China to verify the applicability of the method. The program is realized through programming that is fast and efficient. It takes into account the computational efficiency of the model and the accuracy of watershed subdivision, which lays a good foundation for developing distributed hydrological model in large scale region and helps to better serve the management of hydrology and water resources in the region, especially at the national macro level.