An analysis of scaling eects is performed to evaluate whether data aggregation is a useful regionalization tool or whether it leads to an unacceptable loss of information. One issue concerns the appropriate resolution of digital elevation models (DEMs) used to derive geomorphological input parameters for hydrological models. In particular, the scale problem of watershed division by a channel network and smaller sub-basins is addressed. The investigation involved commercially available data sets with dierent horizontal and vertical resolutions and systematically aggregated DEMs. A stream network and the contributing subareas were derived from a DEM with a distinct critical support area. By varying this threshold area various watershed con®gurations were obtained. The sensitivity of surface runo simulations to all watershed con®gurations was studied with synthetic storms and by means of an in®ltration excess runo model.
The study revealed that elevation data with dierent resolutions diverge enormously in landscape representation and in the derived parameters such as slopes, ¯ow directions and channel networks. Coarse DEMs show a smoother terrain and shorter ¯ow paths than highly resolved data. The contributing threshold area controls the extent of the watershed con®guration and therefore determines the drainage density. These topographic and geomorphological features were used to explain dierences in the runo simulation results. When watershed con®gurations with a varying extent of the channel network were derived from a distinct DEM and then used to simulate surface runo, the drainage densities of the con®gurations correlated with the simulated runo volume. A distinct drainage density, however, did not necessarily lead to similar simulation results when dierent DEMs were used. Since the hydrological model permits rein®ltration, the runo volume depends directly on the lengths of the overland ¯ow. Therefore, the mean length of the overland ¯ow paths might to a certain degree be considered as a scaling factor.