Multifractal analysis applied to the study of the accuracy of DEM-based stream derivation

A correct description of river network morphology is very important when it is used to study different features of a river's morphology as well as phenomena related to it, such as erosion, nitrogen retention or sediment pollution. In recent years, different algorithms have been developed to extract drainage networks directly from Digital Elevation Models. In this paper, the suitability of ArcHydro extension, developed for ArcGIS Desktop and based on the D8 algorithm to generate river networks has been studied by using multifractal analysis. The river networks generated by ArcHydro tools were compared to those provided by photogrammetric restitution for different flow accumulation threshold values. One limitation of the D8 algorithm in order to generate the most appropriate river networks is the correct choice of the flow accumulation threshold value. Multifractal analysis has shown itself to be an efficient approach to determine the most suitable flow accumulation threshold values. Moreover, this approach has allowed characterizing the morphology of river networks and testing the quality of ArcHydro results. According to the multifractal spectra, the main difference between the networks considered here is the lesser stream density with a low channel order detected for ArcHydro results compared to photogrammetric restitution. In addition, the relationship between ruggedness number for the largest contributing area and fractal dimensions in each study area has been explored. The use of multifractal analysis has been extended to drainage network simulation going beyond the descriptive aim of previous works.