THE DEM GENERALIZATION BASED ON THE FILLING OF VALLEY COVERAGE

J. Li, T. Ai

School of Resource and Environment Sciences, Wuhan University, 430072, China

lilideyx@126.com, tinghua_ai@tom.com

As an important terrain representation, the DEM model usually needs different resolution to meet requirements at different levels. So the generalization of DEM converting the data from fine to course plays a significant role in GIS domain. The traditional generalization of DEM is based on image processing via average smooth or other methods without considering the distribution of landform features such as valley, ridge, peat, peak etc. The generalization result will destroy the original main structure. Map generalization is such a process using reduced data to represent the main information with the important feature remained. This paper presents a method to generalize terrain data based on filling valleys of DEM. The decision is made on the analysis of landform from the perspective of geographic level and the concrete operation is made on the geometric level. The first step is to judge one valley is important or not. Then let the unimportant valley coverage to be filled. We extract the valley branch rather than the complete valley channel and organize the structured relationship of each drainage system by a hierarchical tree. which can be used to evaluate each valley’s absolute importance from the point of view of macro level and the capability of accumulation of each valley can evaluate it’s relative importance from the point of view of micro level. The second step is to fill the deleted valley coverage through three methods to fill valleys on DEM in this study, namely the plane filling (simulation), the convexity filling (simulation) and the concave filling (simulation). Compared with traditional terrain generalizations which is based on geometric scale transformation or image processing, this method focuses on the decision of the importance of valley, which can remain the primary geographic characteristics and discard the fragmental ones.