Landslide susceptibility mapping of the Muş-Bingöl region: a comparative analysis and optimization of machine learning models

Date

2026-2

Author

Erdoğan, Nil

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

106
views

0
downloads

Cite This

Landslides are considered one of the most significant natural hazards in eastern Türkiye, where active tectonics, rugged topography and diverse geological conditions strongly control slope instability. The Muş-Bingöl region is located near the Karlıova Triple Junction and is particularly prone to landslides. In this study, landslide susceptibility mapping (LSM) was performed for the Muş-Bingöl region using three machine learning processes: Logistic Regression (LR), Random Forest (RF), and Convolutional Neural Network 1D (CNN-1D). A comprehensive landslide inventory and thirteen landslide conditioning factors, including topographic, geological, hydrological, and environmental variables, were used as model inputs. To ensure methodological consistency, all models were trained and tested using identical datasets and systematically evaluated under multiple train test split ratios ranging from 0.1 to 0.5. Model performance was determined using standard statistical metrics, including Receiver Operating Characteristic (ROC) - Area Under the Curve (AUC), Average Precision (AP), F1-Score and overall accuracy. The results indicate that the RF model performed the best, showing a better ability to distinguish high-susceptibility zones where the majority of landslides were concentrated. Although the LR model produced smoother spatial patterns, it presented limited ability to capture localized instability. While the CNN-1D model outperformed LR to some extent, it was restricted by its one-dimensional design. Overall, the RF-based susceptibility map provided the most reliable representation of landslide-prone areas and offered valuable insights for regional land-use planning and hazard mitigation.

Subject Keywords

Landslide susceptibility mapping, Muş-Bingöl, Logistic regression, Convolutional neural network, Random forest

URI

https://hdl.handle.net/11511/118470

Collections

Graduate School of Natural and Applied Sciences, Thesis