Geographia Technica, Vol 21(2), Special Issue: Artificial Intelligence Applications in Geography, 2026, pp. 188-205
A SHEAR-STRENGTH-DERIVED LITHOLOGICAL WEAKNESS INDEX FOR GIS-BASED LANDSLIDE SUSCEPTIBILITY MODELLING IN MOUNTAINOUS ROAD CORRIDORS OF NORTHERN VIETNAM
Thanh Long NGUYEN 
, Thi Ngoc Ha TRAN , Quoc Lap KIEU , Dieu Trinh NGUYEN
ABSTRACT: Lithology is commonly incorporated into landslide susceptibility models as a categorical map variable, although this representation may not adequately express the mechanical behaviour of slope-forming materials. To improve the physical interpretability of lithological information without substantially changing the conventional GIS-based modelling workflow, this study introduces a shear-strength-informed lithological standardization approach for landslide susceptibility assessment in a mountainous road-corridor environment. Conventional lithological groups were converted into a Lithological Weakness Index (LWI) derived from laboratory-measured cohesion and internal friction angle from 58 samples, and the index was integrated into a Random Forest modelling framework. The approach was tested along National Road QL279 in Lao Cai Province, northern Vietnam, using topographic, environmental, rainfall, geological, hydrological, infrastructural and landslide-inventory data. The RF-LWI model achieved an AUC-ROC of 0.936, accuracy of 0.865 and F1-score of 0.847 on the independent testing dataset, with a mean cross-validated AUC-ROC of 0.943. High and very high susceptibility classes covered 52.71% of the study area. Compared with the model using conventional reclassified lithology, the LWI-based model produced only a modest improvement in predictive performance, but it provided a more mechanically meaningful and interpretable representation of lithological control. The results indicate that shear-strength-derived lithological standardization is useful mainly as a refinement of the geological predictor rather than as a source of large accuracy gains. This approach may support more transparent GIS-based landslide susceptibility modelling in mountainous road corridors where lithological heterogeneity and road-related slope disturbance interact.
Keywords: Landslide susceptibility; Lithological Weakness Index; Shear strength; Random Forest; GIS-based modelling.
Full article here