Application of GIS techniques for landslide susceptibility mapping in the mountainous regions of Fetakgomo-Tubatse Municipality

Abstract

Landslides constitute a significant geohazard in mountainous regions where complex interactions between terrain morphology, geological structures, hydrological conditions, and anthropogenic activities contribute to slope instability. Despite increasing concerns regarding slope failures along critical transportation corridors in South Africa, regional-scale comparative landslide susceptibility assessments remain limited, particularly within the structurally complex terrains of the Bushveld Igneous Complex. This study presents a GIS-based landslide susceptibility assessment of the Fetakgomo-Tubatse Municipality, Limpopo Province, South Africa, with particular emphasis on engineered slopes along the R37 and R555 transport corridors. Ten landslide conditioning factors, including slope angle, aspect, elevation, lithology, soil type, rainfall, land use/land cover, and proximity to roads, rivers, and geological lineaments, were integrated using three susceptibility modelling approaches: the Analytical Hierarchy Process (AHP), Fuzzy Logic, and the Extreme Gradient Boosting (XGBoost) machine-learning algorithm. The AHP model employed expert-derived pairwise comparisons and achieved an acceptable consistency ratio of 0.074, while the Fuzzy Logic model incorporated membership functions and a gamma operator to represent environmental uncertainty and gradual susceptibility transitions. Model performance was evaluated using Receiver Operating Characteristic Area Under the Curve (ROC-AUC), confusion matrix analysis, overall accuracy, balanced accuracy, McNemar’s test, and no-information rate statistics. The results indicate that the Fuzzy Logic model achieved the highest predictive performance (ROC-AUC = 0.769), followed by the AHP model (ROC-AUC = 0.757) and the XGBoost model (ROC-AUC = 0.739). Confusion matrix evaluation further confirmed the superior classification performance of the Fuzzy Logic model through higher overall and balanced accuracy values. High-susceptibility zones were concentrated along steep slopes, weathered lithological units, structurally controlled terrains, and road-cut sections where geological discontinuities interact with anthropogenic slope modifications. The findings demonstrate that uncertainty-based susceptibility modelling provides improved predictive capability within geologically heterogeneous terrains and highlight the value of integrating expert knowledge, spatial analysis, and machine-learning techniques for landslide hazard assessment. This study contributes one of the first comparative susceptibility assessments for the Eastern Limb of the Bushveld Igneous Complex and provides a transferable framework for infrastructure planning, disaster risk reduction, and slope management in mountainous environments worldwide.