F. Gulgen, T. Gokgoz

Yildiz Technical University, Geodetic and Photogrammetric Engineering, Istanbul, Turkey


Many scientists have studied on extracting drainage network. It is an important topic for GIS and cartographic applications in hydrology, geology, generalization, determination of political borders, hiking and mountaineering and relief representation etc. Drainage network information can be provided from land surveying by direct acquisition, and also from maps, digital images or digital terrain models by information extraction. It is usually benefited from line extraction algorithms to automatically extract it from maps and spatial databases. The basic criterion of an automatic line extraction algorithm is that extracted lines must reflect to terrain characteristics precisely. Digital elevation models (DEMs) are vital data in hydrological modeling and analysis such as automated catchments and watershed delineation processes that typically use the drainage lines as a network. Its simple matrix form makes the most efficient data type for processing with programming languages in GIS environment. There are several methods for extraction of drainage network from DEMs. The primary objective of this study is to compare the results obtained by two different methods. First, it is the most common approach that commercial GIS softwares offer. Drainage network is extracted from raster DEM based on flow accumulation values which is obtained from flow routing/direction grid. In other words, these values of the grids are fundamental parameters of this process. Second, it is software based on the Profile recognition and Polygon breaking Algorithms (PPA) to automatically extract drainage network. The program starts with profile recognition process to determine drainage targets. After connecting all drainage targets, a repeated procedure called polygon breaking is applied to eliminate the least important drainage line segments until no closed polygon remains. The results of two different methods were compared with drainage lines drawn by the operator in the stereo models of a test area. For comparison, cartographic and geomorphologic properties were used as the measurements i.e. the numbers, locations and lengths of lines, and relationships between the contours and extracted lines. Paper concludes with evaluation of results, addressing future works.