Application of geomorphons for geomorphological mapping of the Soutpansberg mountain range, South Africa

Accurate geomorphological maps are essential for landscape evolution studies and hazard assessments, yet reliable classification methods remain scarce in complex mountainous regions. This paper presents a workflow that combines geomorphons with an expert-based hierarchical geomorphological classification system to map the geomorphology of the Soutpansberg mountain range in South Africa. Three mapping phases are recognized: (1) data collection and preprocessing, (2) parameter optimization and feature extraction, and (3) validation and analysis, each comprising multiple processing steps. A 30m resolution digital elevation model is used to manually map geomorphological features in a representative area of interest, applying an expert-based geomorphological legend supported by several land-surface parameters and high-resolution aerial imagery. In parallel, ten geomorphon classes were extracted from the digital elevation model and linked to eight common geomorphological features in the Soutpansberg mountain range. Using training and evaluation samples, the look-up distance and flatness threshold parameters for geomorphon extraction were optimized and geomorphological features were automatically extracted for the total study area. The results show that the workflow successfully captures the major geomorphological environments and common geomorphological features in the Soutpansberg mountain range. The classification accuracies of abundant, flat, and homogeneous features, such as plateaus and river terraces, are relatively high, whereas those of less abundant landforms, such as foot slopes and riverbeds, are lower. The findings emphasize that further exploration of automated methods in data-poor, relatively complex mountainous terrains, such as the Soutpansberg mountain range, is promising. Improvements in each phase of the workflow – for example, using higher-resolution elevation data, incorporating field-based training and validation information in selected areas, and fine-tuning model parameters - should be investigated to enhance accuracy and support landscape interpretation in regions where traditional mapping is limited. • Geomorphological Mapping of the Soutpansberg Mountain Range • Objective : to develop a workflow combining geomorphons and expert-based classification to map the geomorphology of the Soutpansberg mountain range in South Africa. • Methodology : •Used a 30m SRTM DEM and calculated land surface parameters (slope, TRI, hillshade, topographic openness). •Manual expert classification was performed in a representative Area of Interest (AoI). •Geomorphons were extracted and optimized using look-up distance (120m) and flatness threshold (<5°). • Geomorphological Features Identified : •Eight features: plateau, gently sloping surface, steep sloping surface, footslope, riverbed, river terrace, river valley, alluvial fan. •Three environments: gravitational, fluvial, hydrology. • Accuracy Assessment : •Overall accuracy: 26.5% , Cohen’s Kappa: 0.15. •Best classified: River terrace (93% user accuracy). •Poorly classified: Riverbed and alluvial fan (0% accuracy). • Key Findings : •Flatter, homogeneous features (e.g., plateaus, terraces) are well captured. •Complex or transitional features (e.g., footslopes, riverbeds) show lower accuracy. •The geomorphons method is promising for data-poor, mountainous terrains. • Outlook : •Future improvements include higher-resolution DEMs, field validation, and deep learning integration. •Potential for hybrid methods combining geomorphons with OBIA and expert rules.