Landslide susceptibility assessment in the Western Ghats adjoining Idukki township, Kerala, India, using integrated geomorphological and geotechnical parameters with AHP and machine learning
- 1College of Engineering Trivandrum
- 2Geological Survey of India
- 3Department of Geology, University of Kerala
Abstract
The Idukki district, Kerala, located in the Western Ghats is highly prone to landslides as the highly rugged terrain responds in complex ways during intense rainfall events. Landslide susceptibility of the region is being assessed in this study by applying Random Forest (RF) and Analytic Hierarchy Process (AHP) by utilising geomorphological and geotechnical parameters. The methodology involves a systematic workflow that includes data preparation, model development, and validation. Results suggest that while 28.43% and 37.78% of the recorded landslides falls well within the high and very high susceptibility zones, respectively in the AHP model, the RF model shows a more balanced distribution with the smallest area under the very low susceptibility class (13.56%; 9.48 km²) and the largest under the medium susceptibility class (24.60%; 17.18 km²). Model validation using AUC, precision, recall, accuracy, and F1-score attribute a high reliability for the RF model in comparison with the AHP model. Although AHP produced comparatively lower validation scores, all metrics exceeded 0.70, indicating good predictive performance. Also, the ROC–AUC values of 0.99 for RF and 0.85 for AHP further demonstrate the superior accuracy and reliability of the RF model for landslide susceptibility mapping. This demonstrates that by integrating the geomorphological and geotechnical factors based on their relative importance, landslide susceptibility maps can effectively capture the complex relationships between conditioning factors and landslide occurrences.
Related
- An integrated framework to assess social inequity and flood susceptibility: a case study of Harris County, Texas
- Two-stage assessment of regional geological hazard susceptibility under mining disturbance
- Adaptive Inversion Optimization of the Static Factor of Safety for Newmark-Based Earthquake-Induced Landslide Susceptibility Assessment
- Geomorphological and geophysical characterization of the MARNOK region (Mid-Atlantic Ridge north of the Kane Fracture Zone: 23°50′N-25°15′N)
- Geological and Geomorphological Controls on Rocky Shore Platforms: Insights from the Basque Coast, France
BibTeX
@article{sandeep2026landslide,
title = {{Landslide susceptibility assessment in the Western Ghats adjoining Idukki township, Kerala, India, using integrated geomorphological and geotechnical parameters with AHP and machine learning}},
author = {Arjun Sandeep P and Anoop S and Anju S S and Jaikiran K P},
journal = {Discover Geoscience},
year = {2026},
month = oct,
doi = {10.1007/s44288-026-00778-x},
url = {https://doi.org/10.1007/s44288-026-00778-x},
note = {CC BY},
}
RIS
TY - JOUR
TI - Landslide susceptibility assessment in the Western Ghats adjoining Idukki township, Kerala, India, using integrated geomorphological and geotechnical parameters with AHP and machine learning
AU - Arjun Sandeep P
AU - Anoop S
AU - Anju S S
AU - Jaikiran K P
PY - 2026
DA - 2026/10/25
JO - Discover Geoscience
DO - 10.1007/s44288-026-00778-x
UR - https://doi.org/10.1007/s44288-026-00778-x
AB - The Idukki district, Kerala, located in the Western Ghats is highly prone to landslides as the highly rugged terrain responds in complex ways during intense rainfall events. Landslide susceptibility of the region is being assessed in this study by applying Random Forest (RF) and Analytic Hierarchy Process (AHP) by utilising geomorphological and geotechnical parameters. The methodology involves a systematic workflow that includes data preparation, model development, and validation. Results suggest that while 28.43% and 37.78% of the recorded landslides falls well within the high and very high susceptibility zones, respectively in the AHP model, the RF model shows a more balanced distribution with the smallest area under the very low susceptibility class (13.56%; 9.48 km²) and the largest under the medium susceptibility class (24.60%; 17.18 km²). Model validation using AUC, precision, recall, accuracy, and F1-score attribute a high reliability for the RF model in comparison with the AHP model. Although AHP produced comparatively lower validation scores, all metrics exceeded 0.70, indicating good predictive performance. Also, the ROC–AUC values of 0.99 for RF and 0.85 for AHP further demonstrate the superior accuracy and reliability of the RF model for landslide susceptibility mapping. This demonstrates that by integrating the geomorphological and geotechnical factors based on their relative importance, landslide susceptibility maps can effectively capture the complex relationships between conditioning factors and landslide occurrences.
SN - 2948-1589
ER -