Uncertainties Related to Scale and Sampling Window Size in Defining Macro Landforms

Abstract

This study focuses on the statistical significance of sampling window sizes, which are used to define macro landforms and the differences they cause in definitions. In the automatic classification of landforms, the problem of determining the optimum scale remains important. Therefore, the relations between the scale factor and the window size constitute the first step, thus classifying landforms. The evaluations were carried out using GMTED2010 and MERIT DEM at different resolutions. The differences in the definitions of different scales and analysis windows caused by the border uncertainties between mountainplateau and mountain-plain that are specific to Türkiye were discussed using the UNEP-WCMC 2000 classification algorithm. Data matrices were created using DEM derivatives such as elevation, slope, and topographic relief for these areas and their descriptive statistics. The test results, which include the combinations of scale and window sizes that best represent the area in selected fields, indicate that the defined macro landform units can result in a more different map as the generalization capacity increases with the changes made in the window size. More meaningful results emerged with the upper limit of the 2.5 km NAW size determined in our study’s window size tests performed at varying rates. In landform classification, mountain boundary relationships were more sensitive to NAW size than DEM resolution.