Researches in Earth Sciences

Researches in Earth Sciences

Landslide susceptibility zoning using support vector machine (SVM) model (From Goujebel pass to Ahar City)

Document Type : Original Article

Authors
masoud rahimi
Seyed Reza Hosseinzadeh
Department of Geography, Faculty of Letters and Humanities, Ferdowsi University of Mashhad, Mashhad, Iran
10.48308/esrj.2026.107033
Abstract
Introduction
Landslide is one of the most frequent natural hazards, capable of threatening human lives and negatively impacting the natural ecosystem on a larger scale (Tien Bui et al, 2012; Balamurugan et al, 2016; Dey et al, 2024). In terms of severity, landslides rank as the seventh most catastrophic event among the various geohazards occurring on the Earth's crust (Nadim et al, 2006). Landslides, as common natural geological processes in mountainous areas, present serious threats to human safety, transportation infrastructure, economic growth, and ecological environments. Currently, landslides are increasing worldwide, and countries around the globe are experiencing the threat of landslide disasters. In general, landslide susceptibility evaluation can offer a basic foundation for landslide prevention and control, enabling comprehensive management with targeted emphasis (He et al, 2025). Since the 1960s, numerous methods have been developed by researchers worldwide. These methods are generally classified into two main categories: knowledge-driven models, such as expert scoring, analytic hierarchy process (AHP), fuzzy logic, and fuzzy comprehensive evaluation; and data-driven models, which include traditional approaches like information value, frequency ratio (FR), and logistic regression, as well as machine learning techniques such as artificial neural networks (ANN), support vector machines (SVM), random forests (RF), and decision tree models (He et al, 2025, Zhao et al, 2024). The Goujebel pass in East Azerbaijan Province is prone to numerous and large-scale landslides due to its specific topographic, geological, and climatic conditions. The frequent occurrence of landslides has not only caused the destruction of natural resources but has also become a serious threat to human settlements, infrastructure, and particularly the key Tabriz–Ahar transportation corridor. Therefore, accurately identifying high-risk areas and producing reliable landslide susceptibility maps can play a crucial role in reducing human and economic losses and improving the safety of transportation routes in this region. The objective of this research is to assess and map landslide susceptibility in the area from the Goujebel Pass to Ahar City using the Support Vector Machine (SVM) model and to analyze the role of influencing factors.
Materials and Methods
The study area is located in northwestern Iran, within East Azerbaijan Province. Geographically, it lies between 38°21' and 38°29' north latitude and 46°50' to 47°02' east longitude. Covering an area of approximately 193 square kilometers, it encompasses sections of the mountains surrounding the Goujebel Pass and extends to the western of the city of Ahar.The primary data used in this study include 1:50,000 topographic maps, 1:100,000 geological maps, a digital elevation model (DEM) with a resolution of 12.5 meters from the ALOS-PALSAR satellite, as well as Sentinel-2 satellite imagery and Google Earth images. For spatial analyses and data processing, ArcGIS, ENVI, and SPSS Modeler software were used. The main method for landslide modeling in this study is the use of the Support Vector Machine (SVM) model. This method was developed in the 1990s and, due to its high efficacy in various algorithms, is recognized as one of the most widely used approaches (Riaz et al, 2024). This model learns the complex relationship between effective factors and landslide occurrence by being trained on a dataset consisting of historical landslide locations and contributing factor layers. The learned relationship is then extrapolated to the entire study region to create a susceptibility map (Kavzoglu et al, 2014). A landslide inventory is recognized for providing a comprehensive record of historical landslide events currently present in the study area. The landslide inventory map (LIM) can be developed using high-resolution satellite imagery, aerial photographs, previous literature documenting landslides, and extensive field surveys (Ali et al, 2022). In this method, a historical landslide distribution map of the study area was first prepared and then divided into two datasets: 70% for training and 30% for testing, while non-landslide points were also randomly selected. Subsequently, the landslide conditioning factors including variables such as elevation, slope, aspect, lithology, distance to faults and drainage networks, the Topographic Wetness Index (TWI), land use, and vegetation cover were prepared in raster format. In the next step, the values of these layers at the locations of the training points were extracted in the ArcGIS environment to construct the data matrix. The SVM model was trained using this matrix with the application of the radial basis function (RBF) kernel, and its parameters were optimized. The performance of the final model was evaluated using the testing data and metrics such as overall accuracy and the ROC curve. Finally, the model was applied to the entire study area, and the final landslide hazard zonation map was produced in five classes: very low, low, moderate, high, and very high.
Results and Discussion
This study aimed to produce a landslide susceptibility map of the Goujebel pass using a Support Vector Machine (SVM) model. Analysis of geomorphological indices showed that more than 98% of the landslides occurred at elevations between 1,450 and 1,750 m, with 42% of them specifically concentrated in the 1,640–1,740 m elevation class. Approximately 88% of the landslides occurred on slopes of less than 30%, and the north aspect, accounting for 63%, was the most significant slope direction for landslide occurrence. The study of geological indicators showed the determining role of lithology, such that the PLQ-c unit (discontinuous conglomerate with marly interlayers) covers only 44% of the area but hosts 95% of the landslides. From a hydrological perspective, a significant overlap was observed between high values of the Topographic Wetness Index (TWI) and the spatial distribution of landslides, with 66% of the landslides occurring within 100 m of drainage networks. Regarding land cover, approximately 70% of the landslides occurred in rangeland areas; however, the NDVI did not exhibit a clear pattern. Evaluation of the Support Vector Machine (SVM) model with a radial basis function (RBF) kernel demonstrated very good performance, with an AUC value of 0.933 for the testing dataset. Based on the results of this model, lithology, with an importance coefficient of 0.268, was identified as the most influential factor, followed by elevation (0.141) and slope (0.138) as the next most significant factors controlling landslide occurrence in the study area. The final hazard zonation map showed that the majority of the area (approximately 54%) is classified as very low and low hazard, about 21.3% of the area is classified as moderate hazard, and 24.4% falls within the high hazard class. High-hazard zones are primarily located on the sensitive PLQ-c formation on moderate slopes (10–30%), which represent the main centers of large and active landslides in the area. Therefore, it can be concluded that a strong and significant relationship exists between the spatial distribution of the PLQ-c formation and the occurrence of major landslides in the region. Water infiltration into conglomerate layers with marl and silt interbeds leads to an increase in pore water pressure, a reduction in internal friction, and ultimately weakens the cohesion of the layers. Under such conditions, the slope gradient plays a significant role as a facilitating factor. For example, some of the large and prominent landslides in the study area, including the one located east of Pireh-Yousefian village, have occurred precisely in areas where the PLQ-c Formation has a significant outcrop and where favorable hydrological conditions have prevailed. Notably, landslides within this formation are not only more frequent but also larger in area and volume. In general, zones with high and very high landslide susceptibility can be identified as
the primary centers of instability within the study area. The accurate identification of these hazard zones can serve as a basis for restricting construction development, monitoring slopes, and implementing preventive measures to reduce future landslide damage.
Conclusion
The present study employed the SVM model with an RBF kernel to perform landslide hazard zonation in the Goujebel area, demonstrating high performance (training AUC = 0.943 and testing AUC = 0.933). Lithology, with a coefficient of 0.268, was the most important factor, and the PLQ-c formation accounted for over 95% of the landslides. Next, elevation and slope were influential factors, with coefficients of 0.141 and 0.138, respectively. Landslides mainly occurred at mid-elevations (1450–1750 m) and on moderate slopes (10–30%). Land cover and land use factors played a lesser role. The landslide hazard zonation of the Goujebel area indicates that structural and lithological controls have a dominant influence on slope instability. The zonation map of the area showed five hazard classes, with approximately 54% of the area classified as low and very low hazard, and 24% as high hazard. High hazard areas are primarily located on the PLQ-c formation and on moderate slopes, requiring continuous monitoring and management. The results demonstrated that combining SVM with spatial data is an effective tool for producing landslide susceptibility maps and supporting disaster management, environmental planning, and infrastructure design. Special attention to susceptible formations (particularly the PLQ-c formation), control of hydrological factors, and monitoring of areas with moderate slopes are among the key strategies for reducing landslide risk in this region. The findings of this research can serve as a basis for environmental planning, disaster risk management, infrastructure design, and land-use planning in the study area. Although the study demonstrates strong performance, future research is encouraged to explore alternative models to further enhance landslide susceptibility predictions.
Keywords
Slope hazard
Landslide
Support vector machine (SVM) model
Goujebel pass
Subjects
Adhikari, M., 2011. Bivariate statistical analysis of landslide susceptibility in western Nepal. Master thesis in geosciences. University of Oslo, p. 1-88. http://hdl.handle.net/10852/12557
Aleotti, P. and Chowdhury, R., 1999. Landslide hazard assessment: summary review and new perspectives. Bulletin of Engineering Geology and the environment, v. 58(1), p. 21-44. https://doi.org/10.1007/s100640050066
Ali, S.A., Parvin, F., Pham, Q.B., Khedher, K.M., Dehbozorgi, M., Rabby, Y.W. and Nguyen, D.H., 2022. An ensemble random forest tree with SVM, ANN, NBT, and LMT for landslide susceptibility mapping in the Rangit River watershed, India. Natural Hazards, v. 113(3), p. 1601-1633. https://doi.org/10.1007/s11069-022-05360-5
Ayalew, L. and Yamagishi, H., 2005. The application of GIS-based logistic regression for landslide susceptibility mapping in the Kakuda-Yahiko Mountains, Central Japan. Geomorphology, v. 65(1-2), p. 15-31. https://doi.org/10.1016/j.geomorph.2004.06.010
Balamurugan, G., Ramesh, V. and Touthang, M., 2016. Landslide suscepti bility zonation mapping using frequency ratio and fuzzy gamma operator models in part of NH-39, Manipur. India Natural Hazards, v. 84(1), p. 465-488. https:// doi. org/ 10. 1007/ s11069- 016- 2434-6
Dey, S., Das, S. and Roy, S.K., 2024. Landslide susceptibility assessment in Eastern Himalayas, India: a comprehensive exploration of four novel hybrid ensemble data driven techniques integrating explainable artificial intelligence approach. Environmental Earth Sciences, v. 83(22), 641 p. https://doi.org/10.1007/s12665-024-11945-z
Esfandyari Darabad, F., Rostami, G., Mostafazadeh, R. and Abedini, M., 2024. Spatial assessment and zoning of landslide risk in Zamkan watershed using support vector machine and logistic regression. Journal of Hydrogeomorphology, v. 11(40), p. 102-123. https://doi: 10.22034/hyd.2024.61467.1737 (In Persian).
Gasemyan, B., Abedini, M., Roostai, S. and Shirzadi, A., 2021. Landslide susceptibility assessment using a novel ensemble algorithm-based model (Case Study: Kamyaran city, Kurdistan province). Quantitative Geomorphological Research, v. 9(4), p. 130-146. https://doi: 10.22034/gmpj.2021.131015 (In Persian).
Ghiasi, V., Mohd Pauzi, N.I., Karimi, S. and Yousefi, M., 2023. Geomechanics and Engineering, v. 34(3), p. 267-284. https://doi.org/10.12989/gae.2023.34.3.267
Gokceoglu, C. and Sezer, E., 2009. A statistical assessment on international landslide literature (1945–2008). Landslides, v. 6(4), p. 345-351. https://doi.org/10.1007/s10346-009-0166-3
Gupta, R.P. and Joshi, B.C., 1990. Landslide hazard zoning using the GIS approach—a case study from the Ramganga catchment, Himalayas. Engineering geology, v. 28(1-2), p. 119-131. https://doi.org/10.1016/0013-7952(90)90037-2
Guzzetti, F., Carrara, A., Cardinali, M. and Reichenbach, P., 1999. Landslide hazard evaluation: a review of current techniques and their application in a multi-scale study, Central Italy. Geomorphology, v. 31, p. 181-216. https://doi.org/10.1016/S0169-555X(99)00078-1
Guzzetti, F., Carrara, A., Cardinali, M. and Reichenbach, P., 1999. Landslide hazard evaluation: a review of current techniques and their application in a multi-scale study, Central Italy. Geomorphology, v. 31(1-4), p. 181-216. https://doi.org/10.1016/S0169-555X(99)00078-1
Hong, H., Pradhan, B., Xu, C. and Bui, D.T., 2015. Spatial prediction of landslide hazard at the Yihuang area (China) using two-class kernel logistic regression, alternating decision tree and support vector machines. Catena, v. 133, p. 266-281. https://doi.org/10.1016/j.catena.2015.05.019
Huang, Y. and Zhao, L., 2018. Review on landslide susceptibility mapping using support vector machines. Catena, v. 165, p. 520-529. https://doi.org/10.1016/j.catena.2018.03.003
Kanungo, D.P., Arora, M.K., Sakar, S. and Gupta, R.P., 2009. Landslide susceptibility zonation (LSZ) mapping – a review. Journal of south Asia disaster studies, v. 2(1), p. 81-105. http://hdl.handle.net/123456789/786
Karami, F., Bayati Khatibi, M., Kheirizadeh, M. and Mokhtari Asl, A., 2020. Evaluation of Performance of Support Vector Machine Algorithm in Landslide Susceptibility Zoning in Ahar-chai Basin. Journal of Geography and Environmental Hazards, p. 8(4), v. 1-17. https://doi: 10.22067/geo. v8i4.83263 (In Persian).
Karimi Sangchini, E., Arami, S.H. and Dastranj, A., 2024. Accuracy assessment of GLM and SVM models in preparing a landslide susceptibility map (Case study: Karganeh watershed, Lorestan province). Researches in Earth Sciences, v. 15(3), p. 119-136. https://doi: 10.48308/esrj.2024.104720 (In Persian).
Kavzoglu, T., Sahin, E.K. and Colkesen, I., 2014. Landslide susceptibility mapping using GIS-based multi-criteria decision analysis, support vector machines, and logistic regression. Landslides, v. 11(3), p. 425-439. https://doi.org/10.1007/s10346-013-0391-7
Kumar, D., Thakur, M., Dubey, C.S. and Shukla, D.P., 2017. Landslide susceptibility mapping & prediction using support vector machine for Mandakini River Basin, Garhwal Himalaya, India. Geomorphology, v. 295, p. 115-125. https://doi.org/10.1016/j.geomorph.2017.06.013
Lee, S., Hong, S.M. and Jung, H.S., 2017. A Support Vector Machine for Landslide Susceptibility Mapping in Gangwon Province, Korea. Sustainability, v. 9(1). https://doi.org/10.3390/su9010048
Malamud, B.D., Turcotte, D.L., Guzzetti, F. and Reichenbach, P., 2004. Landslide inventories and their statistical properties. Earth Surface Processes and Landforms, v. 29(6), p. 687-711. https://doi.org/10.1002/esp.1064
Manzo, G., Tofani, V., Segoni, S., Battistini, A. and Catani, F., 2013. GIS techniques for regional-scale landslide susceptibility assessment: the Sicily (Italy) case study. International Journal of Geographical Information Science, v. 27(7), p. 1433-1452. https://doi.org/10.1080/13658816.2012.693614
Marjanović, M., Kovačević, M., Bajat, B. and Voženílek, V., 2011. Landslide susceptibility assessment using SVM machine learning algorithm. Engineering geology, v. 123(3), p. 225-234. https://doi.org/10.1016/j.enggeo.2011.09.006
Melchiorre, C., Matteucci, M., Azzoni, A. and Zanchi, A., 2008. Artificial neural networks and cluster analysis in landslide susceptibility zonation. Geomorphology, v. 94(3-4), p. 379-400. https://doi.org/10.1016/j.geomorph.2006.10.035
Nadim F., Kjekstad, O., Peduzzi, P., Herold, C. and Jaedicke, C., 2006. Global landslide and avalanche hotspots. Landslides, v. 3(2), p. 159-173. https:// doi. org/ 10. 1007/ s10346- 006- 0036-1
Nadim, F., Kjekstad, O., Peduzzi, P., Herold, C. and Jaedicke, C., 2006. Global landslide and avalanche hotspots. Landslides, v. 3(2), p. 159-173. https://doi.org/10.1007/s10346-006-0036-1
Riaz, M.T., Riaz, M.T., Rehman, A., Bindajam, A.A., Mallick, J. and Abdo, H.G., 2024. An integrated approach of support vector machine (SVM) and weight of evidence (WOE) techniques to map groundwater potential and assess water quality. Scientific reports, v. 14(1), 26186. https://doi.org/10.1038/s41598-024-76607-3
Rostamizad, G. and Dastranj, A., 2024. Evaluating the sensitivity of the landslide event using the support vector machine algorithm. Water and Soil Management and Modelling, v. 4(4), p. 299-312. https://doi: 10.22098/mmws.2023.13934.1379 (In Persian).
Sadati, S.H., Mousavi, S.R., Vahabzadeh Kebria, G. and Roshun, S.H., 2025. Evaluation of Random Forest and Support Vector Machine Models in Landslide Risk Mapping (Case study: Tajan Basin, Mazandaran Province). Journal of Natural Environmental Hazards, v. 14(45), p. 133-154. https://doi: 10.22111/jneh.2025.50031.2071 (In Persian).
Sassa, K. and Canuti, P., 2008. Landslides-Disaster Risk Reduction. Springer. https://doi: 10.1007/978-3-540-69970-5
Sepahvand, A. and Beiranvand, N., 2024. Landslide susceptibility mapping using various soft computing techniques (Case study: A part of Haraz Watershed). Water and Soil Management and Modelling, v. 4(2), p. 261-278. https://doi: 10.22098/mmws.2023.12678.1263 (In Persian).
Tien Bui, D., Pradhan, B., Lofman, O., Revhaug, I. and Dick, O.B., 2012. Land slide susceptibility mapping at Hoa Binh province (Vietnam) using an adaptive neuro-fuzzy inference system and GIS. Comput Geosci, v. 45, p. 199-211. https:// doi. org/ 10. 1016/j. cageo. 2011. 10. 031
Bui, D.T., Pradhan, B., Lofman, O., Revhaug, I. and Dick, O.B., 2012. Landslide susceptibility mapping at Hoa Binh province (Vietnam) using an adaptive neuro-fuzzy inference system and GIS. Computers & Geosciences, v. 45, p. 199-211. https:// doi. org/ 10. 1016/j. cageo. 2011. 10. 031
Van Westen, C.J. and Soeters, R., 1998. Geographic information systems in slope instability zonation (GISSIZ). ITC. 156 p.
Xie, W., Nie, W., Saffari, P., Robledo, L.F., Descote, P.Y. and Jian, W., 2021. Landslide hazard assessment based on Bayesian optimization–support vector machine in Nanping City, China. Natural Hazards, v. 109(1), p. 931-948. https://doi.org/10.1007/s11069-021-04862-y
Xu, Z., Che, A. and Zhou, H., 2024. Seismic landslide susceptibility assessment using principal component analysis and support vector machine. Scientific reports, v. 14(1), 3734 p. https://doi.org/10.1038/s41598-023-48196-0
Zhou, S. and Fang, L., 2015. Support vector machine modeling of earthquake-induced landslides susceptibility in central part of Sichuan province, China. Geoenvironmental disasters, v. 2(1). https://doi.org/10.1186/s40677-014-0006-1