Abstract: Due to the scale effect of changing resolution for grid-based digital terrain analysis, it is important to determine an appropriate resolution (or a range of appropriate resolutions) for gridded digital elevation model (DEM) in practice. The commonly-used approach to determining appropriate resolutions is based on a scale effect curve characterized by certain types of statistics (the mean of local variance is commonly used), which is calculated on the multi-resolution dataset of a specific topographic attribute derived from the gridded DEMs with a series of spatial resolutions. The resolution with the highest value of the mean of local variance is considered to be an appropriate DEM resolution. Although several topographic attributes (such as slope gradient and curvature) have been used in this approach, there are few researches on the effects of different topographic attributes on the results from this approach. In this paper, we used an experiment to compare the effects of three types of topographic attributes (i.e., slope gradient, profile curvature, and horizontal curvature) applied to determining appropriate DEM resolutions. The experiment was conducted in three study areas with different terrain conditions, i.e. Xuancheng area with low relief, Coweeta area with high relief, and Kaixian area with more complex terrain conditions. The tested topographic attributes were calculated from the DEMs with a series of spatial resolutions (i.e., 5, 10, 15, 20, 25, 30, 50, 100, 150, 200, 300 m). For each topographic attribute, two widely-used algorithms were tested. One is proposed by Wood (1996), and the other is the one implemented in Arc-GIS software. The appropriate DEM resolution for a study area had the highest value of the mean of local variance derived from each tested topographic attribute, respectively. Experimental results showed that the effects of slope gradient and curvature are different on the results of the appropriate DEM resolution, while there is little difference between the results from profile curvature and horizontal curvature. The appropriate resolutions based on slope gradient information (i.e., 15, 50, and 100 m for Xuancheng area, Kaixian area, and Coweeta area, respectively) are coarser than those based on curvature information (i.e., 5, 10~15, and 5 m for Xuancheng area, Kaixian area, and Coweeta area, respectively), while the difference between them is less for the study areas with low relief. When slope gradient information was used, the appropriate resolution results for the study area with high relief are coarser than those for the area with low relief. When curvature information was used, the appropriate resolution results for the study areas with simple terrain conditions (such as Xuancheng area with low relief, and Coweeta area with high relief) are finer than those for the study area with complex terrain conditions. The range of appropriate resolutions from each topographic attribute for the study areas with simple terrain conditions is much wider than that for the study areas with complex terrain conditions. For each specific topographic attribute there is no difference between the effects of the tested algorithms of topographic attribute on the results of appropriate resolutions.

Key words: appropriate resolution, digital elevation models (DEM), digital terrain analysis (DTA), scale, topographic attribute