Evaluation of a spatially adaptive approach for land surface classification from digital elevation models

ABSTRACT

Classification of land surface to landforms is fundamental to interpretation of various environmental processes. The heterogeneous landform descriptions and classification approaches, in combination with the scale dependence of digital elevation models (DEMs) and their products, defy the development of an interoperable and transferable automated landform classification approach. A theoretical framework has proposed that land surface should be regionalised to morphologic meaningful objects, delimited by discontinuities (i.e. slope breaks and inflections) and subsequently classified with morphometric and contextual criteria. However, an automated methodology meeting these conditions is still lacking. This study is an attempt to automate this framework through the investigation of a modified version of a spatially adaptive pattern-based approach and its potential to produce morphologic meaningful objects of various shapes and sizes, present at the given DEM resolution. These objects were classified to 15 landform element classes based on semantic descriptions, including criteria of morphometry, relative topographic position and topological relations. Results were visually analysed by draping them over DEMs and contours and quantitatively assessed with fuzzy classification tools. The modified pattern-based approach was proven to be efficient for delineation of morphologic meaningful objects in DEMs. The classification approach was transferable to various landscapes and DEM resolutions, given that it uses spatially flexible fuzzy criteria.

KEYWORDS:

• Landform elements
• OBIA
• regionalisation
• scale
• topography

Disclosure statement

No potential conflict of interest was reported by the authors.

Supplemental data

Supplemental data for this article can be accessed here.