Analysis on the Approach to Determine an Appropriate Window Size for Grid-Based Digital Terrain

Due to the scalar effects of changing window sizes in grid-based digital terrain analysis, it is important to determine an appropriate window size for calculating local topographic attributes in practice. A reasonable appropriate window size must be spatial-variant but constant, such as the often-used window size of 3×3. Currently, a potentially-available approach for determining an appropriate window size characterizes a scale-effect curve derived from a specific kind of local topographical attribute calculated for different window sizes, as proposed by Schmidt and Andrew (2005). However, there has ben little evaluation of this approach, such as research on the effects of different topographic attributes used, applicability in areas with different terrain conditions under different grid sizes. In this paper we conduct an experiment to evaluate this approach. We test two kinds of local topographic attributes (i.e., slope gradient and profile curvature) in three study areas with different terrain conditions under different grid sizes. Experimental results show that the results for appropriate window size achieved based on different topographical attributes were totally different. However, further analysis shows that slope calculation in real applications, the results of spatially-variant appropriate window size by the tested approach perform almost same as the constant window size of 3×3, which is the traditional option. So the current approach to determining spatially-variant appropriate window size might not be effective. Further study on a design for a new and effective approach is needed to determine the spatially-variant appropriate window size for calculating local topographic attributes