Abstract: DEM,resently has been widely used in different fields,especially in photogrammetry for topographic map production.One of the problems of DEM is how to collect data accurately,efficiently and comprehensively.Progressive sampling conceived by Makarovic is considered a more promising approach for DEM data acquisition.The basic principle of progressive sampling is progressive grid densification starting with sampling in a coarse homogenous grid and proceeding to a finer sampling in partial grids.Data collection and analysis are performed interactively.The result is a square grid in variable density,locally adapted to the roughness of terrain relief.Progressive sampling allows for semi-automatic operation,i.e.,computer controlled Ⅹ.Y-positioning of the measuring device of a photogrammetric stereaplotter and manual height setting(Z-positioning) by the operator.Progressive sampling can be done either on an analytical plotter or on an analogue stereoplotter interfaced with a computer by a feedback loop.Progressive sampling has the potential of meeting the requirements on sampling,i.e.,to collect terrain relief data accurately,comprehensively and store them in a favourable structure for later efficient process.Based on the principle of method,corresponding parameters of progressive sampling are experimetally investigated in this paper.Through the analysis of testing results,the influences of the sampling parameters on the accuracy of DEMs and the accuracy of contour lines derived from DEM,and the interrelationship between different parameters are proved.According to this,some conclusions are drawn with respect to the experiments performed in the paper,and a tentitive procedure is proposed in which the values of sampling parameters are determined by considering the accuracy specifications of topographic map and the measuring error of height,that is:under the assumption of accuracy specification for derived contour lines requiring that the standard error of checked contour points should not be larger than 1/3 of the contour interval,a proper minimum grid spacing of progressive sampling can firstly be defined by analyzing the comprehensiveness of equispaced sampling for different regular spacings yielding a standard error of DEM of 1/5 of the contour interval.A threshold value of second difference equal to the contour interval can be chosen if the contour interval is not smaller than 5 times the standard error of measuremnet.