Partitioning of Terrain Features Based on Roughness

Abstract

Surface roughness is a key parameter that reflects topographic characteristics and influences surface processes, and characterization of surface roughness is a fundamental problem in geoscience. In recent years, although there have been basic studies on roughness, few studies have compared the concept and quantification of roughness, and there have been few studies that have evaluated the ability of partition terrain features. Based on 1″ resolution Shuttle Radar Topography Mission (SRTM) data and previous studies, we selected the Qinba Mountain region of China and its adjacent areas as our study area, and used 13 different roughness algorithms to extract roughness in this study. Using spatial patterns and statistical distributions, the results were analyzed, and the best algorithm suited to partitioning terrain features was selected. We then evaluated the ability of the algorithm to distinguish the terrain morphology. The results showed the following: (1) The 13 algorithms were able to be classified into four types, that is, gradient (SLOPE), relief (root mean squared height, RMSH), local vector (directional cosine eigenvalue, DCE) and power-spectral (two-dimensional continuous wavelet transform, 2D CWT). (2) The SLOPE and RMSH algorithms were better able to express and distinguish terrain, as they were able to macroscopically distinguish between four types of terrain in the study areas. Based on power-spectral methods, 2D CWT had the same discrimination ability as the first two methods following a normalization transform, whereas the DCE method had a general effect and could only distinguish two types of terrain. (3) Different roughness algorithms had their own applicability for different terrain areas and application directions.