Frank, KarinLang, UlrichBartz, Dirk2015-11-192015-11-1919983-211-83209-2-https://doi.org/10.2312/vissym19981000In Scientific Visualization, surfaces have often attached data, e. g. cutting surfaces or isosurfaces in numerical simulations with multiple data components. These surfaces can be e. g. the output of a marching cubes algorithm which produces a large number of very small triangles. Existing triangle decimation algorithms use purely geometric criteria to simplify oversampled surfaces. This approach can lead to coarse representations of the surface in areas with high data gradients, thus loosing important information. In this paper, a data-dependent reduction algorithm for arbitrary triangulated surfaces is presented using besides geometric criteria a gradient approximation of the data to de ne the order of geometric elements to be removed. Examples show that the algorithm works su ciently fast to be used interactively in a VR environment and allows relatively high reduction rates keeping a good quality representation of the surface.10.2312/vissym19981000