Accuracy Analysis Of The Extraction Of Drainage Networks Of Sihu Basin Based On DEM

Drainage networks are not only an important natural geographical factors, but also of vital significance on the development of resource and the flood control and disaster mitigation. The extraction of river networks from DEM data is a key to hydrological analysis and its relative studies. The accuracy of the extraction of drainage Networks is also directly affects the precision of the related research. Digital elevation model (DEM) is a widely used in watershed feature extraction. The way of extracting information directly of the network and watershed from digital elevation model is the foundation of the development and application of distributed hydrological model. The basic elements of DEM resolution and sampling methods have a certain influence on the extraction of water related factors.The present study takes Sihu basin which is in Hubei Province as the research area. A test method of positioning accuracy, the test of channel offset (TCO), is developed to test the error of extracted river networks in this study. Referring to the interpretation data of Landsat OLI images (with the resolution of 15 m), we test the extracted accuracies of river network from ASTER DEM (30 m) and SRTM DEM (90 m) by the TCO in Sihu basin respectively. The results show:(1)Drainage water threshold The results showed that the drainage water threshold was the key parameter for the extraction accuracy of water system, and the greater of drainage water threshold, the fewer and thinner of extracted water system; the smaller of drainage water threshold, the more concentrated of extracted water system vice versa. The number of river source, density of water system, class of river and the total river length decreased as drainage water threshold increased, but the main river length changed slightly. When threshold was 3.29 km2, water system extracted was relative coincident with the actual condition.(2) River extraction The result indicates that there are great differences in river extraction by using diverse data sources and using the different ways with the same data source. The results extracted from different data (SRTM DEM, ASTER DEM, Landsat OLI) are vary in channel length and stream number. The maximum channel density is between computer automatic interpretation and artificial visual interpretation, the channel density is 0.10km/km2 and 0.36km/km2.(3) Channel offset Referring to the interpretation data of Landsat OLI images (with the resolution of 30 m), we test the extracted accuracies of river network from ASTER DEM (30 m) and SRTM DEM (90 m) by the TCO in Sihu basin respectively. The results show that the channel offsets extracted from SRTM and ASTER DEM are about 0.05 km to 2.84 km and 0.06 km to 5.46 km. The result also shows that channel offset changes with the channel level in hierarchy of river network:higher the level of channels is, larger the channel offsets. We believe that it is an accumulative effect of the error from D8-algorithm.These errors lie on the range of permissible error in hydrology and water resources management at watershed scale. The extracted accuracy of river networks from SRTM DEM (90 m) is higher than SRTM DEM (90 m), meaning that it not only depends on the resolution of DEM, but also on the data resources.