Comparison of drainage-constrained methods for DEM generalization

In multi-scale digital terrain analysis, the main goal is to preserve the basic ‘skeleton’ with changing scales and to deliver more consistent measurements of terrain parameters at different scales. The drainage lines serve the basic morphology features and ‘skeleton’ in a basin, and therefore play an important role for most applications. Many drainage-constrained methods for DEM generalization have been proposed over the last few decades. This article compares three drainage-constrained methods: a Stream Burning algorithm, the ANUDEM algorithm as an example of a surface fitting approach, and the Compound method as an example of a constrained-TIN approach. All of these methods can be used to build coarser-scale DEMs while taking drainage features into account. The accuracy of the elevations and several terrain derivatives (slope, surface roughness) in the new digital terrain models along with the geometry or shape of key terrain features (streamline matching rate, streamline matching error) is then compared with each other to analyze the efficacy of these methods. The results show that the Compound algorithm offers the best performance over a series of generalization experiments.