EUVIMEDEUVIMED entdecken
Crossref

An Integrated GIS and Machine Learning-based Landslide Susceptibility Mapping in Wayanad District, Kerala, India

Sarath Babu Duggirala, Rupesh Kumar Mishra, Rama Krishnaiah Kondragunta, Akhil Gurijala, Venkateswara Rao Jampani

Disaster Advances · 31.10.2026 · 32

DOI: 10.25303/1912da032048 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Disaster Advances, online erschienen am 31.10.2026. Verfasst von Sarath Babu Duggirala und 4 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Landslides are among the most destructive natural hazards in the Western Ghats, frequently causing substantial loss of life, infrastructure damage and environmental degradation. Wayanad District, Kerala, is particularly vulnerable because of its rugged topography, intense monsoon rainfall and complex geological conditions. This study presents an integrated Geographic Information System (GIS) and machine learning framework for landslide susceptibility mapping using Random Forest (RF), Extreme Gradient Boosting (XGBoost) and Light Gradient Boosting Machine (LightGBM) models. A total of 800 historical landslide inventory points and eleven landslide conditioning factors including elevation, slope, aspect, rainfall, Topographic Wetness Index (TWI), geology, geomorphology, land use/land cover (LULC), lineament density, distance to roads and NDVI, were employed to develop the predictive models. The dataset was divided into a 70:30 training-testing ratio and model performance was evaluated using fivefold cross-validation, confusion matrices, Receiver Operating Characteristic (ROC) curves, Accuracy, Precision, Recall, F1-score, Cohen's Kappa and Matthews Correlation Coefficient (MCC). Among the evaluated models, LightGBM achieved the highest predictive performance, with an Accuracy of 96.11%, Precision of 94.62%, Recall of 97.78%, F1-score of 96.17%, ROC-AUC of 0.961, Cohen's Kappa of 92.22% and MCC of 92.30%, outperforming XGBoost and Random Forest. Feature importance and SHAP analyses identified rainfall (23.5%), slope (20.8%) and TWI (15.4%) as the most influential conditioning factors governing landslide occurrence. The susceptibility outputs of the three models were integrated to generate a final landslide susceptibility map, classified into Very Low, Low, Moderate, High and Very High susceptibility zones. The integrated model effectively reduced prediction uncertainty and showed strong agreement with historical landslide locations. The proposed GIS-integrated ensemble machine learning framework provides a reliable and transferable approach for landslide hazard assessment, supporting disaster risk reduction, landuse planning and sustainable infrastructure management in mountainous regions.

Zitieren

APA

Duggirala, S. B., Mishra, R. K., Kondragunta, R. K., Gurijala, A., & Jampani, V. R. (2026). An Integrated GIS and Machine Learning-based Landslide Susceptibility Mapping in Wayanad District, Kerala, India. Disaster Advances, 32. https://doi.org/10.25303/1912da032048

Vancouver

Duggirala SB, Mishra RK, Kondragunta RK, Gurijala A, Jampani VR. An Integrated GIS and Machine Learning-based Landslide Susceptibility Mapping in Wayanad District, Kerala, India. Disaster Advances. 2026:32. doi:10.25303/1912da032048

BibTeX
@article{Duggirala2026,
  author = {Duggirala, Sarath Babu and Mishra, Rupesh Kumar and Kondragunta, Rama Krishnaiah and Gurijala, Akhil and Jampani, Venkateswara Rao},
  title = {An Integrated GIS and Machine Learning-based Landslide Susceptibility Mapping in Wayanad District, Kerala, India},
  journal = {Disaster Advances},
  year = {2026},
  pages = {32},
  doi = {10.25303/1912da032048}
}
Kostenlos starten →