DEVIATION BETWEEN HILL SLOPE ESTIMATION METHODS AND THE ROLE OF DEM SPATIAL RESOLUTION. COMPARATIVE STUDY IN A GIS ENVIRONMENT

Abstract

Three methods are descibed between the most frequently used for the estimation of hill slope from digital elevation models (DEM), the 'd8' method (Travis et al. 1975), Horn's method (1981) and Zevenbergen and Thorne method (1987), are described. Also a new approach was examined introduced by Tarboton (1997) called 'd-infinite' method. In order to investigate the spatial distribution of deviations between the examined methods, a statistical analysis was performed in a GIS environment. The magnitude of the differences between the methods was determined by calculating the RMSEr (Root Mean Square Error), while the degree of their equivalence was expressed by the linear correlation factor (r). Most deviations were located in areas of high relief, whereas the most interesting fact was that each method overestimates or underestimates the slope in very specific morphological structures, e.g. along watersheds, streams and generally in some types of morphological discontinuities. Finally, the DEM spatial resolution effect was explored on the estimated slope values using the selected methods. Regardless the method used, a general tendency of underestimating slope values by decreasing the DEM resolution was observed. The degree of divergences between these was also highly depended on the change of spatial resolution. Thus, the importance of defining a suitable resolution for each application should be addressed.