Abstract: To determine which DEM data are more suitable for river network extraction in mountainous river basins, this paper selects mountainous river basins in the upper reaches of Yangtze River and Pearl River, including the Liqiu River Basin, the Zagunao River Basin, the Heshui River Basin, the Gantangjiang River Basin, and the Lijiang River Basin above the Darongjiang River Hydrological Station, and uses ALOS 12.5 m DEM, ASTER 30 m GDEM and SRTM 90 m DEM as the base data. The extraction of the river network in the study area is based on Arcpy, and the extraction results are corrected and evaluated by the river network ensemble difference and Arc Hydro Tools to explore the influence of the selection of the optimal catchment area threshold and topographic features on the extraction of the digital river network in mountainous watersheds. The results show that:（1） ALOS DEM with 12.5 m resolution has the smallest river network ensemble difference, which can reflect the degree of water system development in the mountainous watersheds more realistically, and the river network ensemble differences of ALOS DEM in the Liqiu River Basin, the Zagunao River Basin, the Heshui River Basin, the Gantangjiang River Basin and the Lijiang River Basin above the Darongjiang River Hydrological Station are 0.380%, 0.275%, 0.312%, 1.374% and 0.873%.（2） By correcting the DEM data with Arc Hydro Tools, the river network ensemble differences of the three DEMs are reduced, which can significantly improve the extraction accuracy of the river network, and the extraction effect of ALOS DEM before and after correction is optimal and applicable to different river basins. The SRTM 90 m DEM is still better than the ASTER 30 m GDEM in the upper Li River Basin and the Li River Basin after correction, while the accuracy of the ASTER 30 m GDEM is higher in the Zaguzhong River Basin, the Heshui River Basin and the Gantang River Basin after correction；（3） the coefficient of determination of the SRTM DEM is 0.930 2 in the upper Li River（the Li River basin above the Da Sol River Hydrological Station）, and the coefficient of determination of the ALOS DEM in the Zaguzhong River Basin is 0.985 4. The coefficient of determination for the rest of the conditions is more than 0.99, which indicates that the correlation is good.Based on the mean-variable point analysis method to determine the best catchment area threshold for the study area, the best catchment area threshold is 20 000 for ALOS DEM and 500 for ASTER GDEM and SRTM DEM；（4） The large undulating hills are favorable for river network extraction, and the DEM correction converts some watersheds into large undulating hills favorable for river network extraction, and the ALOS DEM has the best extraction effect. Therefore, the modified ALOS DEM extracts the watershed water system with the best effect, and the results can provide some reference for the extraction of water system in mountainous watersheds.