Proceedings of the Open source GIS - GRASS users conference 2002 - Trento, Italy, 11-13 September 2002 Processing digital terrain models by regularized spline with tension: tuning interpolation parameters for different input datasets Tomas Cebecauer* , **, Jaroslav Hofierka* , ***, Marcel œri* , ** * GeoModel s.r.o., HlavÆčikovÆ 26, 841 05 Bratislava, Slovak Republic, tel. +421 905 441 409, fax +421 2 6531 5915, e-mail cebecauer@geomodel.sk, suri@geomodel.sk , hofierka@geomodel.sk ** Institute of Geography, Slovak Academy of Sciences, tefÆnikova 49, 841 73 Bratislava, Slovak Republic, tel. +421 2 5249 2751, fax +421 2 5249 1340 *** Department of Geography and Geoecology, Faculty of Humanities and Natural Science, University of Preov, 17. novembra 1, 081 16 Preov, Slovak Republic 1 Introduction Digital terrain models (DTM) are widely used in geographical information systems (GIS) for different purposes, ranging from simple data analysis and visualization to sophisticated and complex modeling. The measured terrain data are often processed to DTM using spatial interpolation. Different methods of interpolation are currently available in GIS, however, most commonly used are: inverse distance weighted average interpolation, interpolation based on a triangulated irregular network (TIN), variational methods and geostatistical methods (kriging). Details about these methods can be found in literature, e.g. [1]. The quality of DTMs depends mainly on the quality of input data and interpolation method. The selection of appropriate interpolation method is important because for a particular type of data one method performs better than the other [2] (e.g. contours vs. GPS data). Hence, the quality of interpolation method is also determined by its capabilities to process different types of data. The regularized spline with tension (RST) interpolation method is one of the best interpolation methods available in current GIS [3, 4, 5]. RST is a variational method based on the assumption that the interpolation function should pass through (or close to) the input data points and should be at the same time as smooth as possible [6]. The specific feature of RST is the existence of regular derivatives of all orders needed for surface geometry analysis [6]. Tension and smoothing parameters are useful to change the shape and smoothness of surface according to the modelled phenomenon. The accuracy of interpolation by RST can be assessed by various statistical methods evaluating interpolation errors (residuals) in input points or predictive error of interpolation in areas outside input points using cross-validation technique [5, 6, 7]. Unfortunately, these quantitative measures usually describe only a statistical aspect of interpolation quality, but applicability of the interpolated DTM for a specific purpose can be still limited (e.g. hydrologically correct DTM, correct representation of undersampled features) [c.f. 7 page 317, 8]. The paper identifies common problems in application of s.surf.rst on different types of terrain data. A careful selection of RST parameters and, if necessary, pre-processing of input data may substantially improve interpolation results. Application examples include elevation data taken from contours and elevation points, photogrammetric measurements at various scales and volume of datasets. The results of interpolation using different