Accuracy assessment in glacier change analysis

Abstract: This thesis assesses the accuracy of digital elevation models (DEM) generated from contour lines and LiDAR points (Light Detection and Ranging) employing several interpolation methods at different resolutions. The study area is Jostefonn glacier that is situated in Sogn og Fjordane county, Norway. There are several ways to assess accuracy of DEMs including simple ways such as visual comparison and more sophisticated methods like relative and absolute comparison. Digital elevation models of the Jostefonn glacier were created from contour lines for years 1966 and 1993. LiDAR data from year 2011 was used as a reference data set. Of all the interpolation methods tested Natural Neighbours (NN) and Triangular Irregular Network (TIN) algorithms rendered the best results and proved to be superior to other interpolation methods. Several resolutions were tested (the cell size of 5 m, 10 m, 20 m and 50 m) and the best outcome was achieved by as small cell size as possible. The digital elevation models were compared to a reference data set outside the glacier area both on a cell-by-cell basis and extracting information at test points. Both methods rendered the same results that are presented in this thesis. Several techniques were employed to assess the accuracy of digital elevation models including visualization and statistical analysis. Visualization techniques included comparison of the original contour lines with those generated from DEMs. Root mean square error, mean absolute error and other accuracy measures were statistically analysed. The greatest elevation difference between the digital elevation model of interest and the reference data set was observed in the areas of a steep terrain. The steeper the terrain, the greater the observed error. The magnitude of the errors can be reduced by using a smaller cell size but that this is offset by a larger amount of data and increased data processing time.