Modeling Landslide Susceptibility Using Artificial Neural Network Algorithm: A Case Study of Shahid Abbaspour Dam Basin, Northeast Khuzestan

جغرافیا و مخاطرات محیطی

Articles in Press, Accepted Manuscript, Available Online from 12 June 2025   

Document Type: Research Article

DOI: 10.22067/geoeh.2025.92634.1557

Authors

Mahshid Moavi¹; Heeva Elmizadeh^{* 2}; Mojgan Entezari³

¹PhD Student in the Faculty of Geographical Sciences and Planning, University of Isfahan, Isfahan, Iran

²Khorramshahr University of Marine Science and Technology, Khorramshahr, Iran

³Faculty of Geographical Sciences and Planning, University of Isfahan, Isfahan, Iran

Abstract

Landslides are natural hazards that pose significant risks to human lives and the environment. Landslide susceptibility maps are vital tools for planning, management, and risk mitigation. Various methods exist for generating these maps, but their effectiveness varies across studies. This research aims to develop a landslide susceptibility map for the Shahid Abbaspour Dam watershed using an Artificial Neural Network (NNET). Landslide data points were randomly split into 70% training and 30% testing datasets. Fifteen influencing factors were selected as model inputs, including elevation, slope, aspect, curvature, SPI, Topographic Wetness Index (TWI), distance from roads and rivers, geology, soil texture, land use, and precipitation. Model validation was performed using accuracy metrics and statistical indices (RMSE, Kappa, MAE, R-Squared). The NNET model demonstrated strong predictive performance, achieving an accuracy of 0.8543. The study identified rivers as the most critical factor influencing landslide susceptibility in the area, followed by slope and geological formations. The highest susceptibility zones were found in the southwest and a small northern section of the watershed. The resulting map can aid in land-use planning and development strategies to mitigate landslide risks in the region.
Introduction
Landslides represent one of the most dangerous natural hazards that pose significant threats to human life, infrastructure, and the environment, frequently resulting in substantial financial damages and loss of life (Sun et al., 2020; Bao et al., 2020). These geological phenomena are characterized by the downward movement of soil and rock materials along slopes, which may occur either suddenly or gradually due to natural triggers or human-induced disturbances (Mohammadi et al., 2018). Given their destructive potential and frequent recurrence in vulnerable areas, particularly in regions with documented historical occurrences, landslides have become a major focus of scientific research and risk assessment studies globally.
Materials and Methods
The research focuses on the Shahid Abbaspour Dam watershed, geographically situated between 49°23′ to 50°21′ E longitude and 31°26′ to 32°39′ N latitude in the northeastern part of Khuzestan Province, Iran. This area constitutes an important segment of the Greater Karun watershed (Ali Bakhshi et al., 2020). The study adopted a comprehensive five-stage methodology to assess landslide susceptibility, which is defined as the probability of landslide occurrence in a specific area based on local terrain characteristics.
The first stage involved an extensive literature review and systematic data collection. In the second stage, fifteen critical influencing factors were identified and prepared as information layers. These factors included elevation, slope gradient, slope aspect, proximity to faults, distance from roads, distance from watercourses, geological formations, land use patterns, NDVI (Normalized Difference Vegetation Index), soil characteristics, seismic activity data, slope curvature, Topographic Wetness Index (TWI), Stream Power Index (SPI), and precipitation data. These spatial datasets were processed and analyzed using ArcGIS 10.8 software, with foundational data derived from multiple sources including 1:50,000 scale topographic maps and 1:100,000 scale geological maps of Khuzestan Province. The digital elevation model (DEM) with 30-meter resolution was obtained from ASTER satellite imagery, while specialized indices like TWI and SPI were calculated using SAGA-GIS 9.0.1 software. The third stage implemented the Artificial Neural Network (NNET) model for landslide susceptibility mapping. This machine learning approach, executed in the R programming environment using the nnet package, was trained with 70% of the landslide inventory data while the remaining 30% was reserved for validation purposes. The fourth stage generated the final landslide susceptibility zonation map. The fifth and final stage involved rigorous model validation using multiple performance metrics including Accuracy, Root Mean Square Error (RMSE), Kappa coefficient, Mean Absolute Error (MAE), R-Squared value, Sensitivity, Specificity, Positive Predictive Value (PPV), and Negative Predictive Value (NPV).
Results and Discussion
The comprehensive analysis revealed three primary factors exerting the most significant influence on landslide susceptibility in the study area. The foremost factor was proximity to rivers, which affects slope stability through several mechanisms including bank erosion, increased soil moisture content, and alterations in groundwater levels. This finding corroborates previous research by Zakerinejad & Amoshahi (2022), who similarly identified stream proximity as the most critical determinant of landslide occurrence, particularly within approximately 100 meters of watercourses. Slope gradient emerged as the second most influential factor, with particularly pronounced effects in the southwestern sector of the watershed encompassing the Dezful region. This area's mountainous terrain features steep slopes and dense drainage networks, creating conditions highly conducive to slope failures. These observations align with the findings of Shojaian et al. (2022) from their study of the Melaghefar watershed in northeastern Khuzestan, as well as with research by Maa et al. (2019) and Pollveck (2019). The third most significant factor was geological composition, particularly the presence of Quaternary alluvial deposits and the erosion-prone Gachsaran Formation, which contains substantial marl components. These geological units were identified as particularly susceptible to slope instability, a conclusion supported by Mohammadi et al. (2022) in their investigation of the Izeh and Deh Sheikh (Abbaspour Dam) basin. In contrast to these major factors, elevation and slope curvature demonstrated relatively minor influence on landslide susceptibility in this study, a finding that differs from Selamat et al. (2022) who reported elevation as a primary controlling factor in their research area. The susceptibility mapping results classified approximately 31.30% of the study area (75.37 km²) as high-risk zones, predominantly located in the southwestern and northern sections of the watershed. Model validation yielded excellent performance metrics, with an overall Accuracy of 0.8543 and R-Squared value of 0.9916, indicating strong predictive capability. Detailed sensitivity analysis across different susceptibility classes showed consistent performance, with values ranging from 0.8060 for Class 1 (lowest susceptibility) to 0.8717 for Class 4 (high susceptibility).
Conclusion
This research successfully identified and quantified the key factors controlling landslide susceptibility in the Shahid Abbaspour Dam watershed, demonstrating the effectiveness of the Artificial Neural Network approach for spatial prediction of landslide hazards. The findings provide valuable scientific basis for land-use planning and risk mitigation strategies in this environmentally sensitive region. The high predictive accuracy of the model suggests its potential applicability to other similar watersheds facing landslide risks.

Keywords

Landslide; Artificial Neural Network Algorithm; Sensitivity; Shahid Abbaspour Dam Watershed

References

Ali Bakhshi, T., Azizi, Z., Vafaeinejad, A., & Aghamohammadi Zanjirabadi, H. (2020). Survey of area changes in water basins of Shahid Abbaspour Dam caused by 2019 floods using Google Earth Engine. Ecohydrology, 7(2), 345–357. [In Persian] https://doi.org/10.22059/ije.2020.295785.1272

Bao, Y., Zhai, S., Chen, J., Xu, P., Sun, X., Zhan, J., ... & Zhou, X. (2020). The evolution of the Samaoding paleolandslide river blocking event at the upstream reaches of the Jinsha River, Tibetan Plateau. Geomorphology, 351, 106970. https://doi.org/10.1016/j.geomorph.2020.106970

Benmakhlouf, M., El Kharim, Y., Galindo-Zaldivar, J., & Sahrane, R. (2023). Landslide susceptibility assessment in Western External Rif Chain using machine learning methods. Civil Engineering Journal, 9(12), 3045–3060. https://doi.org/10.28991/CEJ-2023-09-12-018

Esfandiary Darabad, F., Rahimi, M., Navidfar, A., & Arsalan, M. (2020). Assessment of landslide sensitivity by neural network method and vector machine algorithm (Case study: Heyran Road, Ardabil Province). Quantitative Geomorphological Research, 9(3), 18–33. [In Persian] https://doi.org/10.22034/gmpj.2020.122210

Geological and Mineral Exploration Organization of Iran. (2018). Identification of landslide susceptibility in the city of Masjed-e-Suleiman, General Directorate of the South Western Region (Ahvaz). [In Persian]

Ghaedi, S., Amouzegar, S., & Shojaiean, A. (2022). Landslide microzonation using fuzzy grey correlation analysis (case study: Mollaghafar drainage basin, northeast of Khuzestan Province). Advanced Applied Geology, 12(2), 337-350. [In Persian] https://doi.org/10.22055/AAG.2021.36387.2195

He, Q., Shahabi, H., Shirzadi, A., Li, S., Chen, W., Wang, N., ... & Ahmad, B. B. (2019). Landslide spatial modelling using novel bivariate statistical based Naïve Bayes, RBF Classifier, and RBF Network machine learning algorithms. Science of the Total Environment, 663, 1-15. https://doi.org/10.1016/j.scitotenv.2019.01.329

Hejazi, A., Rezaeimoghaddam, M., & Naseri, A. (2020). Landslide hazard zoning using artificial neural network models and TOPSIS downstream of Sanandaj Dam. Hydrogeomorphology, 7(24), 65–82. [In Persian] https://doi.org/10.22034/hyd.2020.11060

Hungr, O., Leroueil, S., & Picarelli, L. (2024). The Varnes classification of landslide types, an update. Landslides, 11, 167–194 .https://doi.org/10.1007/s10346-013-0436-y

Ma, S., Chen, J., Wu, S., & Li, Y. (2023). Landslide susceptibility prediction using machine learning methods: A case study of landslides in the Yinghu Lake Basin in Shaanxi. Sustainability, 15, 15836. https://doi.org/10.3390/su152215836

Majd-Bavi, A., & Mumipour, M. (2022). Landslide susceptibility zonation in Shahid Abbaspour Dam district. Journal of Geography and Environmental Hazards, 10(1), 65-80. [In Persian] https://doi.org/10.22067/geoeh.2021.67029.0

Masruroh, H., Leksono, A. S., & Kurniawan, S. (2023). Developing landslide susceptibility map using Artificial Neural Network (ANN) method for mitigation of land degradation. Journal of Degraded & Mining Lands Management, 10(3), 4479–4494. https://doi.org/10.15243/jdmlm.2023.103.4479

Mohammadi, A., Shahabi, H., & Bin Ahmad, B. (2018). Integration of InSAR technique, Google Earth images and extensive field survey for landslide inventory in a part of Cameron Highlands, Pahang, Malaysia. Applied Ecology & Environmental Research, 16(6), 8075-8091. https://dx.doi.org/10.15666/aeer/1606_80758091

Mohammadi, M., Afifi, M. A., & Ghanbari, A. R. (2023). Landslide hazard zoning using a fuzzy inference system in the Izeh River basin. Geographical Sciences, 19(42), 156-176. [In Persian] https://sanad.iau.ir/fa/Journal/geographic/Article/919283

Mostofi, N. (2013). MATLAB User Guide. Tehran Publications. [In Persian]

Mousavi Nadushan, S. S. (2012). Introduction to the R computing language. Tehran, Iran: Shahid Abbaspour University of Water and Electricity Industry. [In Persian]

Pollock, W., Grant, A., Wartman, J., & Abou-Jaoude, G. (2019). Multimodal method for landslide risk analysis. MethodsX, 6, 827-836.

Rahaman, A., Dondapati, A., Gupta, S., & Raj, R. (2024). Leveraging artificial neural networks for robust landslide susceptibility mapping: A geospatial modeling approach in the ecologically sensitive Nilgiri District, Tamil Nadu. Geohazard Mechanics, 2(4), 258-269. https://doi.org/10.1016/j.ghm.2024.07.001

Rajabi, M., Rezaeimoghadam, M., & Takzare, A. (2020). Landslide hazard potential zoning using the neural network method (Case study: Alamut watershed in Qazvin Province). Quantitative Geomorphological Research, 9(3), 185-171. [In Persian] https://doi.org/10.22034/gmpj.2020.122223

Sadati, S. H., Mousavi, S. R., Vahabzadeh Kebria, G., & 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, 1-1. [In Persian] https://doi.org/10.22111/jneh.2025.50031.2071

Sadeghi Balochi, M., & Alian, S., (2025), Landslide Hazard Assessment and Visualization Using Artificial Neural Network Method (Case Study of Lahijan County). Paper presented at the Proceedings of the 16th International Conference of the Iranian Society for Operations Research, Ramsar. [In Persian] https://civilica.com/doc/1920698

Selamat, S. N., Majid, N. A., Taha, M. R., & Osman, A. (2022). Landslide Susceptibility Model Using Artificial Neural Network (ANN) Approach in Langat River Basin, Selangor, Malaysia. Land, 11, 833. https://doi.org/10.3390/ land11060833

Sun, D., Ding, Y., Zhang, J., Wen, H., Wang, Y., Xu, J., ... & Liu, R. (2022). Essential insights into decision mechanism of landslide susceptibility mapping based on different machine learning models. Geocarto International, 1-29. https://doi.org/10.1080/10106049.2022.2146763

Sun, X., Chen, J., Han, X., Bao, Y., Zhan, J., & Peng, W. (2020). Application of a GIS-based slope unit method for landslide susceptibility mapping along the rapidly uplifting section of the upper Jinsha River, South-Western China. Engineering Geology and the Environment, 79, 533–549. ttps://doi.org/10.1007/s10064-019-01572-5

Talaei, R., & Shadfar, S. (2023). Landslide susceptibility modeling using artificial neural network and logistic regression methods at the Saqezchay Basin, south of Ardabil Province. Watershed Engineering and Management, 15(3), 481-503. [In Persian] https://doi.org/10.22092/ijwmse.2022.360475.1996

Tayebi far, A. (2024). Preparing Landslide Hazard Sensitivity Maps Using Machine Learning Methods (Case Study: Kermanshah). (Master's Thesis). University of Isfahan. [In Persian]

Wahba, M., Essam, R., El-Rawy, M., Al-Arifi, N., Abdalla, F., & Elsadek, W. M. (2024). Forecasting of flash flood susceptibility mapping using random forest regression model and geographic information systems. Heliyon, 10(13), e33982. https://doi.org/10.1016/j.heliyon.2024.e33982

Zakerinejad, R., & Amoshahi, N. (2022). Assessment of Landslide Hazard Using Remote sensing data and the Maximum Entropy Model (Case Study: Kome watershed, in south of Isfahan Province). Quantitative Geomorphological Research, 11(2), 128-149. [In Persian] https://doi.org/10.22034/gmpj.2022.340900.1349