Research on Extracting Digital Drainage Network Based on DEM and Remote Sensing

Extracting digital drainage network and physical descriptions of watershed based on DEM is the precondition of research on building distributed hydrological model and hydrological simulation. There´re some traditional methods to extract digital drainage network based on DEM. Aiming at the deficiencies of those methods, many scholar provide improved methods such as extending virtual river network information by digitalizing from map or RS image and extracting the river network from RS image. But these improved methods still have some problems, such as lower automatization level and limited field of application. This paper explicates the method of extracting digital drainage network of watershed based on DEM and remote sensing data with four steps: Firstly, perform geometric correction and radiant correction on RS image data, and then extract water body information by decision tree classification. Secondly, based on the information extracted, obtain Digital River and Lake Network (DRLN) on flat area by extracting the center line of rivers using Delaunay triangular network. Thirdly, determine the flow directions on cells inside/outside the DRLN, the DRLN can transfixion parts of the flat area and lead the flow direction of adjacent cells, so the problem of parallel line in flat area can be solved to a certain extent. Simultaneously, with the controlling of DRLN, the drainage structure is close to natural river way. Finally, correct the lack of lakes when extracting digital drainage network based on DEM and DRLN through the lake information obtained by RS image. With these steps, the paper perform an experiment upon Qinhuai River watershed, the result of this experiment indicates that this method can effectively solve the problem of parallel water ways and the main channel which departs its natural position. In flat area and mountainous area, this method can extracting digital drainage network of watershed according to the natural drainage network.