Application of fractal techniques to the comparative evaluation of two methods of extracting channel networks from digital elevation models

The extraction of channel networks from digital elevation models was carried out using two different techniques: the 'flow accumulation method' (which is based one the now almost traditional steepest descent flow routing) and the 'multilevel skeletonization method' (which is based on curvatures in contours). Series of networks representing the same drainage basin were extracted with each of the methods using different extraction thresholds. Various stream ratios and fractal dimensions were determined from these networks based on three different hierarchical models. The study shows that for each extraction method, all ratios irrespective of hierarchical model exhibit similar variation with respect to extraction threshold. On the other hand, for the flow accumulation method of extraction, the different types of fractal dimensions vary in different ways with extraction threshold while in the case of the multilevel skeletonization method all the fractal dimensions decrease as extraction threshold increases. We attribute this difference in the behavior of fractal dimensions to the difference in approach between the two methods of extraction. Apparently the multilevel skeletonization method employs a more natural scheme.