A Technical Evaluation of Landslide Vulnerability Using Analytical Hierarchy Process ( AHP ) and the Frequency Ratio ( FR ) Methods for Chamoli Region of Uttarakhand, India

• Published in: Geological journal (Chichester, England)
• Main Authors: Singh, Shubham, Nayak, Nirlipta Priyadarshani, Aggarwal, Ashish, Verma, Harsh Kumar
• Format: Journal Article
• Language: English
• Published: 28.07.2025
• ISSN: 0072-1050; 1099-1034
• DOI: 10.1002/gj.70045

The Chamoli region of Uttarakhand, India, is vulnerable to landslides because of its varied terrain, intense monsoonal rains and growing infrastructure. Landslides in this region pose a significant threat to life, infrastructure, and the local economy, urging the need for better landslide susceptibility maps that will be beneficial for disaster preparedness and management. To evaluate landslide vulnerability throughout the region, the study combines GIS (Geographic Information System) tools with the AHP (Analytical Hierarchy Process) and the FR (Frequency Ratio) models. The analysis included 11 landslide causative factors, namely geology, slope, plan curvature, profile curvature, aspect, rainfall, NDVI, LULC, proximity to roads, drainage and faults. These causative factors are chosen based on their relevance to the regional terrain conditions, literature survey and correlation with landslide occurrences. The results from the AHP and FR models classified the area into five susceptibility zones. 21.47% of the region is classed as having very high susceptibility under the AHP model, whereas 12.63% is classified as such by the FR model. These findings indicate that a major portion of the Chamoli region faces significant landslide risks, highlighting the need for focused risk reduction strategies. Model validation using AUC (Area Under Curve) analysis revealed success rates of 0.829 (95% Confidence Interval (CI): 0.805–0.845) for AHP and 0.820 (95% CI: 0.806–0.834) for FR, while the prediction rates for AHP and FR were 0.834 (95% CI: 0.806–0.857) and 0.818 (95% CI: 0.796–0.839). The close alignment between success and prediction rates in both models indicated minimal overfitting, suggesting the models' consistent performance on both known and new data. Overall, the results demonstrate that both AHP and FR are effective and reliable approaches for landslide susceptibility mapping in the study area.