Geomorphometry-Automatic Landform Classification

Abstract

In the past, landforms were represented in physiographic and morphometric maps by hand drawing. With developments in digital elevation models (DEM), geographic information systems (GIS) and image analyses, automatic extraction of landforms from morphological parameters and data storage in databases is now possible and are actively utilized in various fields, such as geomorphology, soil science, and ecology. In the above scopes, DEM data forms the database of morphometric parameters, such as, relief, slope, curvatures, and topographic openness. Presently, calculation of parameters, implementation of relationships with landforms, scaling, classification methods, topological relations, homogeneity, and generalizations during the transformation of 3D components of landforms, such as mountains, peaks, slopes, valleys, and plain, into 2D geometric elements in computers are being investigated. In this study, the methods and applications for the automatic extraction of the landforms developed by researchers across different disciplines were reviewed. Classification methods were grouped as combined parameters method and unsupervised/supervised classification methods based on pixel/object. This paper emphasizes the importance of adopting machine learning to implement new models applicable to all terrains.