Abstract:Identifying the optimal flow accumulation threshold (threshold) based on digital elevation model (DEM) is a key process to define the drainage of a river basin. Due to lack of underlying surface information, it’s common to establish a quantitative calculation model for threshold relying on the watershed topography and real river network. Except topography, however, vegetation, precipitation and other underlying surface factors were also correlated with the characteristics of river network. Furthermore, the extraction of river network based on grid DEM had two different flow direction allocation strategies, i.e., the single flow algorithm and multiple flow algorithm. Since the procedure of multiple flow algorithm was complicated, the present study mainly focused on selection of threshold under single flow algorithm and lack of the research of threshold determination of multiple flow algorithm. Aim at above problem, the climate and underlying surface conditions were considered to explore the relevance between influence factors (i.e., topography, vegetation and precipitation) and threshold, and a calculation model was put forward to compute threshold based on multiple linear regression model with different algorithms in Hubei Province. The results showed that the optimal thresholds had a positive correlation with local slope, average precipitation and vegetation coverage rate. The flow accumulation threshold was increased with the increase of terrain steepness, rainfall and vegetation coverage. The calculation model of flow accumulation threshold can effectively integrate multi-source influence factors in different regions and obtain reasonable threshold. The significance level of rejecting the null hypothesis were both lower than 0.05 in different flow direction allocation strategies and the coefficients of determination (R2) were higher than 0.9. Good results of extraction can also be got by applying this method to other river basin. It meant the result of simulated river network by calculation model of flow accumulation threshold had high similarity with the real river network, it also provided rational hydrological information for agroforestry planning, agricultural disaster prediction and other industrial applications.