Accuracy analysis on automatically extracted river network based on SRTM DEM — A case study of Oujiang Watershed in Zhejiang Province

To explore the possibility to automatically update the 1:250,000 river network data, this paper chooses the Oujiang Watershed in Zhejiang Province as a study case, extracts automatically the river network based on 90m SRTM DEM, and assesses its accuracy in terms of point positional accuracy, river morphology parameters, and river-network-matching-difference. Since the accuracy of extracted river network is highly correlated with the drainage area threshold, this paper proposes the average drainage branching ratio to determine this threshold, which can improve the data accuracy by 1.24%. The results show that under any topography, the positional accuracy of automatically extracted river network is high than that of 1:250,000 data, which accounts for 21.4% on plains, 15.7% on hills, and 53.0% on mountains, respectively. Furthermore, the greater the relief, the higher the accuracy of extracted data. In terms of river morphology, the river-network-matching-difference is less than 1.2%, which implies the automatically extracted river network seems to best match 1:250,000 data. Therefore, automatically extracted waterlines based on SRTM DEM can basically meet the update requirements of 1:250,000 river network data.