Volume 2, Issue 2 — Year 2025 — Article e100039

Morteza Ebadati¹; Feng-Hu Sun²; Yu Lee³; Meysam Jabbari Moghadam⁴

1. Departmen of Geographic Information Science, Chengdu University of Information Technology, Chengdu 610225, China
2. Departmen of Geographic Information Science, Chengdu University of Information Technology, Chengdu 610225, China
3. Departmen of Geographic Information Science, Chengdu University of Information Technology, Chengdu 610225, China
4. Faculty of Basic Science, Maragheh University of Technology, Maragheh 5518183111, Iran

Landslides are among the most critical natural hazards, causing significant environmental and economic damage. This study focuses on landslide susceptibility analysis in Chalus County, Iran, using Artificial Neural Networks (ANNs) to predict high-risk areas. A dataset of 77 recorded historical landslides was compiled through field surveys, remote sensing, and satellite imagery analysis. Various conditioning factors, including elevation, geology, slope angle, rainfall, temperature, aspect, NDVI (normalized difference vegetation index), weathering, distance to cities, distance to landslides, distance to rivers, distance to roads, and distance to faults, were integrated into a GIS-based modeling framework. The ANN model was trained using a dataset divided into training and validation subsets, ensuring robust predictive performance. The results demonstrated that ANNs effectively identify landslide-prone regions, with high accuracy in distinguishing between stable and unstable areas. The susceptibility map produced highlights that northern and mountainous regions of Chalus County are highly vulnerable to landslides, primarily due to steep slopes, heavy precipitation, and geological instability. Model validation using statistical accuracy measures, including the Area Under the Curve (AUC), precision, and recall, confirmed the reliability of the predictions. The findings of this study provide valuable insights for urban planners, geologists, and policymakers in risk assessment and land-use planning. Implementing these results can support early warning systems and mitigation strategies to minimize landslide-related hazards in the region. Future research should explore hybrid AI models and additional geospatial datasets to enhance predictive capabilities further.

Landslide susceptibility, Artificial intelligence, Neural networks, Machine learning, ArcGIS