Document Type : Original Article
Landslides are among the most significant geomorphological hazards in mountainous and semi-arid regions, causing substantial damage to infrastructure, natural resources, and human life every year. Applying scientific methods to identify landslide-prone areas can play a vital role in hazard management and environmental planning. Conducting such analyses using traditional methods is extremely time-consuming and costly, necessitating the use of more advanced tools. In this regard, Geographic Information Systems (GIS) play a unique role as a powerful tool for the collection, management, analysis, and visualization of spatial data, particularly in modeling natural hazards such as floods, soil erosion, and landslides. GIS enables the integration of diverse spatial layers such as slope, elevation, land use, proximity to faults, roads, and rivers, providing a suitable platform for implementing statistical and probabilistic models and generating susceptibility maps. The application of statistical and probabilistic models, especially within GIS environments, allows for quantitative analysis of the relationship between landslide occurrences and environmental variables, shifting the modeling process from a qualitative to a quantitative level. These approaches, by utilizing spatial data and leveraging the analytical capabilities of GIS, can identify complex spatial structures and potential dependencies among phenomena, contributing to more accurate predictions of high-risk areas. Using such models not only optimizes time and resource allocation but also enhances the accuracy of analyses, playing a significant role in reducing damages and improving environmental resilience. Among these models, the Weight of Evidence (WoE) model based on Bayesian theory is widely used to quantify the statistical relationship between landslide distribution and environmental factors. This study aims to delineate landslide hazard zones in the Qarnaveh watershed, located in eastern Golestan Province, using the probabilistic Weight of Evidence (WoE) model within a Geographic Information System (GIS) environment.
In this study, 179 landslide occurrences were identified using field surveys and available databases. Seventy percent of these events were used for modeling, while the remaining thirty percent were reserved for model validation. Fourteen conditioning factors, including elevation, slope degree, slope aspect, distance to faults, fault density, distance to roads, distance to rivers, drainage density, slope curvature, stream curvature, geology, land use, LS index, and topographic wetness index, were extracted and analyzed using ArcGIS and SAGA-GIS software. The significance of each factor was assessed through the WoE model, which was then used to generate a landslide susceptibility map for the watershed. Model accuracy was evaluated using the Receiver Operating Characteristic (ROC) curve.
Results indicated that drainage density had the strongest influence, while fault density had the least impact on landslide occurrence in the study area. The southern and southwestern parts of the watershed were identified as zones with very high landslide susceptibility. The WoE model demonstrated an accuracy of 79.7% in predicting landslide-prone areas. The study identified key environmental factors linked to landslide susceptibility in the Qarnaveh watershed, Golestan Province. Areas with elevations of 148-389 m, gentle slopes (0–11.49%), and aspects facing west, south, southeast, and northeast were most prone to landslides. High-risk zones also included regions with TWI values of 14.91-22.74, flat slope curvature, concave stream curvature, forest cover, Qal geological formations, and proximity to rivers (0–600 m) and roads (0–853 m). Other influential factors were high drainage density (3.55–6.45), LS values (44-111), and distances up to 3042 m from faults. These factors contribute to slope instability by reducing soil strength, increasing moisture, and accelerating erosion. The most influential variables were drainage density, distance to river and road, elevation, TWI, and stream curvature. Southern and southeastern slopes, exposed to higher solar radiation, showed the greatest sensitivity. The WoE model, with 79.7% accuracy (ROC), successfully identified the southern and southwestern parts of the watershed as highly susceptible to landslides.
Based on the results, the Weight of Evidence model shows a reasonably high capability in identifying landslide-prone areas with acceptable accuracy. It can serve as an effective tool in natural resource management, watershed planning, disaster risk reduction, spatial planning, and reducing vulnerability in high-risk areas. The findings of this study offer a practical tool for natural resource managers and local planners to prioritize protective actions, monitor critical areas, and guide infrastructure development and agricultural expansion in landslide-prone zones. Despite its effectiveness, the WoE model relies heavily on historical landslide data, and limited temporal precision or spatial data accuracy may affect results. Future research should consider integrating WoE with machine learning models like Random Forest or SVM, and using multitemporal satellite imagery to better capture landslide dynamics and land use changes.
5. Acknowledgment & Funding
The manuscript did not receive a grant from any organization.
6. Conflict of Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
