.png){kind=icon}
.png){kind=icon}
Return
An integrated approach for landslide susceptibility mapping: a case study of Idukki District, South-West India
.png){kind=icon}
PRE
AI
A
S
K
P
DOI：10.1016/j.asr.2026.02.025.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Recent monsoon season landslides in Kerala, which have led to repeated loss of life and large-scale damage in several districts, underline the growing severity of the problem. This study focuses on the Idukki district, a region historically prone to recurrent debris flows, with an increasing frequency of landslides observed in recent monsoon seasons. The primary aim of this study is to generate detailed landslide susceptibility maps for the region by applying statistical (FR), machine learning (RF, CART), and deep learning (CNN-2D) techniques, along with an integrated ensemble model to enhance predictive performances. This study uses a comprehensive set of 28 conditioning factors to improve the reliability of the landslide susceptibility assessment. Among the applied models, the Random Forest (RF) achieved an AUC of 93.4%, followed by the Convolutional Neural Network (CNN) with an AUC of 89.97% and the Classification and Regression Tree (CART) with an AUC of 84.8%. The integrated CNN-RF ensemble further improved the predictive capability, attaining an AUC of 92.3%, with field verification confirming a close correspondence between the predicted high susceptibility zones and historically documented landslide locations. Furthermore, the research evaluates the impact of sparse vegetation using Radar-based analyses across different models, offering additional insights into landslide susceptibility. In summary, this investigation is crucial in identifying high-risk landslide areas within the Idukki district and assessing the effects of diverse terrain characteristics on landslide occurrences.
Keywords:
Landslide susceptibility
Random Forest
Convolutional Neural Network
Classification and Regression Tree
Radar
Journal
IF:
2.8
Papers:
1.3K
Citations:
2.0W
Organization
Cited Papers