SELECTION OF APPROPRIATE INTERPOLATION METHODS FOR CREATION DEMs OF VARIOUS TYPES OF RELIEF BY COMPLEX APPROACH TO ASSESSMENT OF DEMs

The main aim of a research presented in this paper is selection of the most suitable interpolation method and its parameter settings for creation of digital elevation models (DEMs) of base types of relief in the Czech Republic (flatlands, hilly areas, uplands and highlands) from the DMÚ 25 database. The suitability of interpolation methods and their settings was explored using the assessment of metric accuracy of resulting DEMs, i.e. difference between estimated data and reference data. The non-spatial indicators (root mean square error – RMSE, total absolute error – AE and hammock index – H) were used to assess the quality of DEM in terms of metric accuracy. However, complex approach to the assessment of DEMs must contain also evaluation of spatial distribution of errors. For these reason, the spatial cluster analysis LISA has been used. Its sense consists in a possibility to identify the spatial clusters of statistically significant high error values. The next aim of the research was thus using of spatial assessment of metric accuracy of DEMs to verify the results of non-spatial assessment and also to find the differences in spatial behaviour (distribution) of errors in DEMs rated using weighted order as high-quality and low-quality DEMs. It is evident from the results that if input elevation data are dense enough, non-spatial characteristics of metric accuracy can be viewed as decisive for assessing quality of DEMs and differentiation the high-quality DEMs form low-quality DEMs. The results from local cluster analysis LISA can be than used for verification of non-spatial assessment. However, when using spatial analysis LISA for assessment of DEM quality it is necessary to consider also selected non-spatial indicators in order to avoid misinterpretation.