Terrain simplification from grid DEMs based on spatial autocorrelation

It is very necessary to construct multi-scale DEMs. And terrain simplification algorithms from grid DEMs of high resolution offer an effective solution to approach multi-scale DEMs. This paper presents a simple method for simplifying terrain from dense grid DEMs based on grid cell´s spatial autocorrelation. By calculating local Moran´s I index for DEM cells, the simplified DEMs is reconstructed with terrain feature points gotten by proper Moran´s I threshold value. Choosing an area in loess hilly-gully region as test site with 5m DEMs, the effect of reconstructed DEMs is evaluated. Experiment results from a comparison with Very Important Points (VIP) method show that the proposed algorithm is with clear principle, easy procedure, high efficiency, and perfect precision especially in keeping geo-morphological skeleton by analyzing spatial distribution of `surface-specific´ structural points used to reconstruct new DEMs, the value of elevation statistical parameters, the structure of terrain feature lines and slope values gotten from the DEMs reconstructed by the local Moran´s I index and VIP algorithm.