ABSTRACT
Landslides, as destructive mass movements, pose significant risks to human life, infrastructure, and the environment. Accurate susceptibility zonation is crucial for land-use planning and risk mitigation. This study aims to identify the optimal model for landslide susceptibility mapping in the Sadat Hosseini area of Khuzestan Province, Iran, a region characterized by sensitive lithology, steep slopes, and increasing human activity. Three widely used models—the Anbalagan, Haeri-Samiei, and Analytical Hierarchy Process (AHP)—were implemented and compared. Key influencing factors, including lithology, slope gradient and aspect, land use/land cover, distance from rivers and roads, seismicity, and elevation, were analyzed within a Geographic Information System (GIS) framework. Thematic layers for each factor were created, weighted according to each model's methodology, and integrated to produce final susceptibility maps. The results categorized the area into five susceptibility classes: very low, low, moderate, high, and very high. The Anbalagan model allocated 35.25% and 17.90% of the area to the high and very high susceptibility classes, respectively. Validation against an inventory of existing landslides revealed that most recorded events occur in high and very high susceptibility zones, particularly within young alluvial deposits and the Aghajari Formation on south-facing slopes. Comparative analysis demonstrated that the Anbalagan model yielded the highest conformity with the actual distribution of landslides in the region. The study concludes that the Anbalagan method is the most suitable model for landslide susceptibility zonation in areas with geomorphological and climatic conditions similar to the study area. Its data requirements, typically derived from topographic and geological maps and satellite imagery, make it a practical and widely applicable tool for preliminary hazard assessment in Iran.

1. INTRODUCTION
Landslides are among the most prevalent and damaging natural hazards, often triggered by a complex interplay of geological, geomorphological, climatic, and anthropogenic factors. Iran, with its vast mountainous terrain, diverse geology, high seismicity, and variable climate, is highly prone to slope instability. Effective risk management and sustainable development planning in susceptible regions necessitate reliable identification of areas vulnerable to landslides. The Sadat Hosseini area in northeastern Khuzestan Province has experienced numerous landslide events, threatening local communities and infrastructure. While various quantitative models exist for landslide susceptibility mapping, their performance can vary significantly depending on local environmental conditions and data availability. This study addresses the critical need to evaluate and identify the most accurate predictive model for this specific region. Previous research has applied individual models in different parts of Iran, but a comparative assessment of the Anbalagan, Haeri-Samiei, and AHP methods in the geomorphic context of the Zagros foothills remains limited. This research distinguishes itself by conducting a direct comparative analysis of these three established models to determine which one best predicts the spatial pattern of landslides in the Sadat Hosseini area, thereby providing a validated tool for local planners and disaster managers.
2. MATERIALS AND METHODS
The methodology was based on a comparative analysis of three susceptibility models: the semi-quantitative Anbalagan and Haeri-Samiei methods, and the multi-criteria decision-making Analytical Hierarchy Process (AHP). The primary data sources included a 30-meter resolution Digital Elevation Model (DEM) for deriving topographic parameters (slope, aspect, elevation), a 1:100,000 scale geological map for lithological classification, Landsat 8 OLI imagery for generating land use/land cover and vegetation indices, as well as datasets for roads, rivers, and seismicity. An inventory of 42 historical and existing landslide locations was compiled through field surveys using GPS and interpretation of satellite imagery. In a GIS environment, eight causative factors were standardized into thematic raster layers: lithology, slope angle, slope aspect, land use/land cover, distance to rivers, distance to roads, earthquake intensity, and elevation. For the Anbalagan and Haeri-Samiei models, each class within a factor layer was assigned a fixed score based on its presumed influence on slope stability, as defined by the respective model's guidelines. For the AHP model, pairwise comparison matrices were constructed to derive the relative weights of each factor and their internal classes based on expert judgment, with consistency checks performed using Expert Choice software. All weighted layers were then aggregated using a weighted linear combination in GIS to generate susceptibility index maps for each model. These continuous index maps were classified into five susceptibility categories using the natural breaks (Jenks) method for consistent comparison.

3. RESULTS
The susceptibility zonation maps produced by the three models revealed distinct spatial patterns and area distributions across the five susceptibility classes. The Anbalagan model resulted in the most extensive area classified as high and very high susceptibility, combining to cover approximately 53.15% of the total study area. Specifically, 35.25% was classified as high and 17.90% as very high susceptibility. In contrast, the Haeri-Samiei model was the most conservative, with only 2.58% of the area in the high susceptibility class and no area classified as very high. The AHP model produced an intermediate result, assigning 7.86% of the area to the very high susceptibility class. Spatially, all models consistently identified south and southwest-facing slopes, areas underlain by the Gachsaran Formation (marl and gypsum) and young alluvial terraces, and regions with moderate slope gradients (approximately 15-30 degrees) as being more susceptible. The validation phase, comparing the model outputs with the inventory of 42 known landslides, provided a key performance metric. A significantly higher percentage of the inventoried landslides (over 78%) fell within the high and very high susceptibility zones of the Anbalagan map. This overlap was less pronounced for the AHP model and considerably lower for the Haeri-Samiei model.

4. DISCUSSION AND CONCLUSION
The comparative analysis clearly indicates that the Anbalagan model outperforms both the Haeri-Samiei and AHP models in accurately predicting landslide-prone areas in the Sadat Hosseini region. The superior performance of the Anbalagan method can be attributed to its scoring system, which is particularly sensitive to the lithological and geomorphic conditions dominant in the study area, such as erodible marls and alternating hard-soft rock sequences. The AHP model's reliance on expert judgment, while flexible, may introduce subjectivity that may not fully capture local specificities. The Haeri-Samiei model appears overly general for the high-resolution, local-scale analysis required here. The findings confirm that lithology (especially the Gachsaran and Aghajari Formations) and slope aspect (south and southwest) are the dominant controlling factors, with moderate slopes being more critical than very steep ones, likely due to the greater accumulation of weathered materials. The uniformity of the seismic factor across the small study area minimized its differentiating role in this particular zonation. The primary conclusion is that the Anbalagan model provides the most reliable and practical tool for preliminary landslide susceptibility assessment in the Sadat Hosseini area and similar environments within the Zagros structural zone. Its input data are readily available, and its methodology is straightforward to implement in GIS. For future research, it is recommended to enhance the model by incorporating more detailed hydrological factors, such as rainfall intensity and groundwater flow, and to validate its application in other physiographic regions of Iran to further test its robustness. Furthermore, integrating the resulting susceptibility maps with data on elements at risk would constitute a vital next step towards comprehensive landslide risk assessment.