Uncertainty In Feature-based Generalization Of Dem

Digital elevation model is among the most important data of spatial information and terrain analysis in GIS.Scale effect which exists in geographical space always leads to different terrain models and different terrain representations in different scales.Generally,coarser elevation model can be produced from finer ones with information eliminating and terrain rebuilding.This process is called terrain generalization.Many factors have influences on terrain accuracy such as terrain morphology,sample density,interpolation methods and so on.Being aware of these factors brings a lot of benefits on digital elevation modeling and terrain analysis.So does in terrain generalization.Researchers have revealed the effects on terrain accuracy of terrain morphology,sample density,interpolation methods and spatial resolution.This thesis focuses on the impact of uncertainty in terrain generalization.Two types of uncertainty exist in terrain generalization.The uncertainty of original data will create an impression on generalized terrain on the one hand,on the other hand,different methods in simplification always bring about different results.In this thesis,uncertainty of original data is modeled first using sequential Gaussian simulation.Simplified terrain are generated by the maximum z-tolerance,the compound method and the skeleton method separately.Contours,RMSE and skeleton retention are compared and the probability of uncertainty is calculated.The result shows that both two types of uncertainty have influences on terrain accuracy.The influence of the uncertainty of simplification methods is larger than the uncertainty of original data's,especially in mountain areas.Terrain morphology is always the most important factor in terrain generalization.Sample density is a significant factor.Sometimes,the uncertainty of simplification methods can even cause more influence on generalized terrain than sample density.Interpolation method has the least effect like that of the uncertainty of original data.