Landform-element recognition through machine learning and computer vision

Geomorphologists increasingly turn to automated machine learning approaches in order to characterize the land surface. This chapter reviews the concepts underpinning such approaches. We describe the ways in which machine learning is applied to geomorphometry problems, including methods of supervised, unsupervised, and self-supervised learning, pixel- and object-based analyses, and deep learning. We additionally explore the strengths, limitations, and opportunities presented by machine learning geomorphometry approaches. Unsupervised and self-supervised case studies are presented demonstrating the classification of topographic landforms without ground truth data.