Application of geomorphons and Object-Based Image Analaysis for geomorphological mapping

This is the thesis project for the Bachelor in Future Planet Studies at the University of Amsterdam.
The files include the written thesis and the associated ArcGIS Pro project package. A link to the GitHub repository is provided in the details section.

Access to the full dataset is available upon request.

Accurate geomorphological maps are essential for landscape evolution studies, geoconservation efforts, and hazard assessments, but reliable classification methods are still lacking in complex mountainous regions. This study evaluates two digital approaches for geomorphological classification; Object-Based Image Analysis (OBIA) and geomorphons. The aim is to assess the efficiency, accuracy, and suitability of these methods in mapping landforms across South Africa’s Soutpansberg mountain range. OBIA groups pixels with similar characteristics into objects, emphasizing contextual analysis, while geomorphons rely on pixel-based patterns of terrain geometry.  A 30m DEM is used for direct geomorphons classification, while OBIA uses derived LSPs. Parameters were optimized using expert-classified training samples, and accuracy assessments and spatial analyses were conducted on the resulting landform maps. Results show that OBIA achieves higher overall accuracy and better agreement (accuracy: 41.8%, k: 0.30) with expert classifications than geomorphons (accuracy: 26.5%, k: 0.15) in the Soutpansberg, particularly for complex landforms, due to its integration of multiple surface parameters and contextual information. Geomorphons performs best for flatter, more homogenous classes such as Plateau and River terrace. Spatially, OBIA produces fewer, more coherent landforms, while geomorphons results in fragmented outputs. Geomorphons offers a fast, automated, and reproducible method suitable for large-scale mapping, but its performance is sensitive to parameters and terrain. OBIA, though more labor-intensive, provides greater adaptability and precision, making it valuable for diverse regions. The choice between OBIA and geomorphons depends on mapping objectives, terrain complexity, and available resources. Future work should explore the synergy of OBIA’s and geomorphons respective strengths, higher-resolution DEMs, and testing in similar regions. Field observations and exploring tailored rule sets could further improve classification accuracy and reliability.