| Abstract Submission No. | ABS-3.1-0061 |
| Title of Abstract | Machine Learning¿Based Landslide Susceptibility Assessment of Wayanad Plateau in Southern Peninsula, India |
| Authors | Girish Gopinath* |
| Organisation | Kerala University of Fisheries and Ocean Studies (KUFOS) |
| Address | Kerala University of Fisheries and Ocean Studies (KUFOS) Kochi, Kerala, India Pincode: 682 506 E-mail: gg@kufos.ac.in |
| Country | India |
| Presentation | Oral |
| Abstract | The Western Ghats, a globally recognized biodiversity hotspot, has recently experienced escalating environmental stress due to urban expansion, terrain modifications, and intensifying rainfall variability. The eastern escarpment of Wayanad district in Kerala has emerged as a critical zone, where recurrent landslides between 2018 and 2024 culminated in a catastrophic event in 2024 that triggered a Landslide Lake Outburst Flood (LLOF), causing severe loss of life and property. To address the urgent need for systematic risk evaluation, this study applies a machine-learning framework to generate a detailed landslide susceptibility map for the region. A spatial inventory of over 260 past landslide occurrences was integrated with twelve conditioning factors, including lithology, soil characteristics, topography, hydrological indices, land use/land cover, and proximity to linear features. The Random Forest (RF) algorithm, optimized through grid search and validated with five-fold cross-validation, was employed to capture the complex relationships between landslides and geo-environmental drivers. Model performance was evaluated using multiple statistical indicators, including root mean square error (RMSE), log loss, receiver operating characteristic (ROC) curves, and confusion matrix metrics. The RF model demonstrated strong predictive capability, with an AUC of 0.911 for training and 0.85 for independent validation, though minor overfitting was observed. Importantly, the susceptibility map successfully delineated the recent Wayanad landslide site as a high-risk zone, highlighting the model¿s practical relevance. These findings underscore the potential of data-driven approaches for guiding hazard preparedness, land-use regulation, and long-term disaster resilience in vulnerable mountain environments. |
| Keywords | Landslide Susceptibility Assessment, Machine Learning, Wayanad Plateau, Southern Peninsula, India |