Automatic Extraction of Drainage Networks from Digital Elevation Model and Remotely Sensed...

Automatic Extraction of Drainage Networks from Digital Elevation Model and Remotely Sensed Data


Asadollah najafi
gis manager
Transportation & Terminals Organization Iran
Email: [email protected]

Abbas Alimohammadi

Mapping of drainage network is one of the basic tasks needed in many geographical and environmental studies.Usually,these networks are mapped through ground surveying ,visual interpretation of aerial photos and sattelite images and or delineation from the existing topographic maps.By developing Geographical Information System(GIS),drainage networks are usually digitized from the existing topographic maps.Unfortunately ,this method is expensive and time consuming.

Digital Elevation Model(DEM) and remotely sensed data have high capability for extraction of drainage networks.By doing so,slope is an important problem for proper delineation of drainage networks data.Because in higher slopes streams are relatively small and therefore they donot show considerable contrast in the sattelite images ,whereas in DEM because of precise definition of flow directions in high slopes ,the accuracy of network streams delineation is considerably high.Stream networks as defined from DEM in lowlands are very poor,at the other hand,because of difficulties and fuzziness in definition of flow directions.

In this research,a new approach for delineation of drainage networks using the two data sources including DEM and landsat TM data and combination of their results has been developed.The study site is located in north of iran.Flow accumulation method has been used for extraction of drainage networks from DEM.In this approach flow direction and flow accumulation of each cell has been calculated and by definition of an appropriate threshold,drainage networks have been extracted.Classification methods,band rationing and linear and directional filtering have been used to define drainage networks from sattelite data.As compared to others, classification has showed the best performance.

Visual evaluation and qualitative evaluation using Kappa statistic ,have been used for evaluation of the quality of the extracted networks.Digitized networks from topographic map have been used as reffrence.Because of digitizing errors,accuracy of Kappa index base on pixel by pixel camparison , is very low (about 39 percent). Therefore ,some drainage networks have been screen digitized and used as reffrence.The results show about 80 percent agreement between.The role of slope in extraction of drainage networks from DEM and TM data has been studied and 15 percent slope has been defined to be an appropriate threshold for optimal combination of the two sources.