Estimation of Minimal Error During the Process
of DEM Creation with the Help of R software

The whole thesis is concerning about the estimation of error while computing the digital elevation model, which is represented by a grid. The aim of the thesis was to introduce basic principles of the spatial interpolation, to provide a classification of fundamental interpolation procedures and to tell to the audience about several usual methods.

The methods described in the work are linear triangulation, inverse distance weighted, minimal curvature Spline method, geostatistical Kriging method and global Trend method.

The main part of the work is the own project which aim is to provide a classification of incorporated methods referring to a minimal absolute error in modelling of the selected surfaces. The strategy was to focus on data distribution (regular or irregular), on the data density, and to incorporate random point selection when picking up a certain part from the set of reference points.

One imaginary and four real surfaces were used to analyse the performance of six methods: IDW Nearest Neighbors, IDW Fixed Radius, Regularized Spline, Spline with Tension, Kriging and Trend.

The quantification was made using the ordinal classification based on the evaluation of mean, maximum and third quartile of the absolute difference surface, of range between the maximum positive and negative deviation from the reference surface and of the results of T-test and F-test of so-called average surface against the reference surface.

The results showed that the most accurate method is the Regularized Spline, followed by the Spline with Tension. Linear kriging also achieved good results. While both IDW methods were the most inaccurate.

The diagnostics of the relation between detected error and morfometric parameters of the reference surface was also made. It is not fully cogent, but there is some sort of positive relation between difference surface and planar curvature of the reference surface, while the opposite relation can be observed between profile curvature and the difference surface.

Author’s name	Bc. David KONEČNÝ
Title	Estimation of Minimal Error During the Process of DEM Creation with the Help of R software
Thesis type	Diploma
Department	Dept. of Geoinformatics, Faculty of Science, Palacký University in Olomouc, Czech Republic
Supervisor	Mgr. Pavel TUČEK
Publication year	2008
Keywords	grid, DEM, spatial interpolation, IDW, Spline, Kriging, linear triangulation, R, reference surface, error estimation
Number of pages	32 + 31
Number of appendices	24
Language	czech