1- Department of Soil Conservation and Watershed Management Research, Isfahan Agricultural and Natural Resources Research and Education Center, AREEO, Isfahan, Iran
2- Department of Water Science and Engineering, Faculty of Agriculture, University of Maragheh, Maragheh, Iran
3- Department of Soil Conservation and Watershed Management Research, Khorasan Razavi Agricultural and Natural Resources Research and Education Center, AREEO, Mashhad, Iran
2- Department of Water Science and Engineering, Faculty of Agriculture, University of Maragheh, Maragheh, Iran
3- Department of Soil Conservation and Watershed Management Research, Khorasan Razavi Agricultural and Natural Resources Research and Education Center, AREEO, Mashhad, Iran
Abstract: (507 Views)
Extended Abstract
Background: Landslides are one of the most important geomorphological hazards in Iran. The heavy human and financial losses of landslides (damage to villages, roads, infrastructure, farms, and human casualties) have doubled the importance of studying them. For this reason, accurate knowledge of landslide areas that have occurred in the past and identification of areas with high susceptibility to landslides can be of great help in reducing the damage caused by this phenomenon. In recent decades, machine learning has received serious attention as a powerful tool in earth sciences, especially in the study and prediction of landslides. Machine learning can predict which areas or zones are prone to landslides based on historical data (past events). Compared to conventional methods such as multi-criteria decision-making, the advantages of machine learning include high processing speed, higher accuracy, flexibility, and lower cost. The Chalus Road is one of the busiest and most dangerous roads in Iran. Landslide investigation on Chalus Road is of great importance. The reasons for this importance can be summarized in several main axes: life and financial safety, economic and tourism importance, specific geological and climatic conditions of the region, and environmental problems. Machine learning can identify high-risk areas, predict the probability of landslide occurrence, improve life safety, reduce economic losses, and improve crisis management. Therefore, the aim of this study is to investigate the landslide susceptibility in one of the watersheds of Chalus Road (Sira Watershed).
Methods: The Sira Watershed on the Chalus Road is located in the northwest of the large watershed of the Karaj Dam basin. There is a total of six villages with 762 households and 1904 people in the watershed. First, landslide areas at the watershed level were identified using the database available at the General Directorate of Natural Resources and Watershed Management of Alborz Province, and the interpretation of aerial photographs. Then, extensive field visits were used at the watershed level to match and verify the identified points. Finally, 56 landslide points were verified based on field visits. The final and confirmed landslide points were divided into two parts: training and validation, in a ratio of 70 to 30 percent for modeling. Next, 10 important factors affecting landslide susceptibility were used to zone landslide susceptibility, including altitude, slope, distance from stream, distance from road, curve number, soil hydrological groups, lithology, land use, and the topographic wetness index. Next, the random forest method was used in the R software to determine landslide susceptibility. The random forest method is one of the powerful supervised ensemble machine learning algorithms used for classification and regression problems. In this study, the model efficiency was evaluated in the training and validation stages using the receiver operating characteristic (ROC). Then, the landslide susceptibility map was classified into five classes with very low, low, moderate, high, and very high sensitivity based on natural breaks in ArcGIS software. Finally, the area and percentage of each landslide susceptibility class were evaluated in the entire watershed and separately for rural areas, and their prioritization was done based on the area of high and very high landslide susceptibility.
Kalvan and Kalha villages are 7% and 16% susceptible to landslides, respectively. However, Sira village, with 73.5% of the village area, is exposed to high and very high landslide susceptibility and is the most important village in the watershed in terms of landslide resilience measures. In total, 675 households with a population of 1700 in the villages in the watershed are exposed to landslide susceptibility. It should be noted that there is a total of 333 landslide zones with an area of more than 1 hectare, equivalent to 1393 hectares, in the Sira watershed.
Conclusion: This study has provided a valuable tool for planners and crisis managers by providing high-precision landslide susceptibility maps. The findings show that combining machine learning with spatial data can revolutionize the prediction of natural hazards, especially landslides. To achieve practical results, however, cooperation between government institutions, academia, and local communities is essential to implement sustainable solutions. This research is an effective step toward reducing natural hazards by combining modern technologies and spatial analysis, and can be used as a model for other similar areas in Iran.
Background: Landslides are one of the most important geomorphological hazards in Iran. The heavy human and financial losses of landslides (damage to villages, roads, infrastructure, farms, and human casualties) have doubled the importance of studying them. For this reason, accurate knowledge of landslide areas that have occurred in the past and identification of areas with high susceptibility to landslides can be of great help in reducing the damage caused by this phenomenon. In recent decades, machine learning has received serious attention as a powerful tool in earth sciences, especially in the study and prediction of landslides. Machine learning can predict which areas or zones are prone to landslides based on historical data (past events). Compared to conventional methods such as multi-criteria decision-making, the advantages of machine learning include high processing speed, higher accuracy, flexibility, and lower cost. The Chalus Road is one of the busiest and most dangerous roads in Iran. Landslide investigation on Chalus Road is of great importance. The reasons for this importance can be summarized in several main axes: life and financial safety, economic and tourism importance, specific geological and climatic conditions of the region, and environmental problems. Machine learning can identify high-risk areas, predict the probability of landslide occurrence, improve life safety, reduce economic losses, and improve crisis management. Therefore, the aim of this study is to investigate the landslide susceptibility in one of the watersheds of Chalus Road (Sira Watershed).
Methods: The Sira Watershed on the Chalus Road is located in the northwest of the large watershed of the Karaj Dam basin. There is a total of six villages with 762 households and 1904 people in the watershed. First, landslide areas at the watershed level were identified using the database available at the General Directorate of Natural Resources and Watershed Management of Alborz Province, and the interpretation of aerial photographs. Then, extensive field visits were used at the watershed level to match and verify the identified points. Finally, 56 landslide points were verified based on field visits. The final and confirmed landslide points were divided into two parts: training and validation, in a ratio of 70 to 30 percent for modeling. Next, 10 important factors affecting landslide susceptibility were used to zone landslide susceptibility, including altitude, slope, distance from stream, distance from road, curve number, soil hydrological groups, lithology, land use, and the topographic wetness index. Next, the random forest method was used in the R software to determine landslide susceptibility. The random forest method is one of the powerful supervised ensemble machine learning algorithms used for classification and regression problems. In this study, the model efficiency was evaluated in the training and validation stages using the receiver operating characteristic (ROC). Then, the landslide susceptibility map was classified into five classes with very low, low, moderate, high, and very high sensitivity based on natural breaks in ArcGIS software. Finally, the area and percentage of each landslide susceptibility class were evaluated in the entire watershed and separately for rural areas, and their prioritization was done based on the area of high and very high landslide susceptibility.
Kalvan and Kalha villages are 7% and 16% susceptible to landslides, respectively. However, Sira village, with 73.5% of the village area, is exposed to high and very high landslide susceptibility and is the most important village in the watershed in terms of landslide resilience measures. In total, 675 households with a population of 1700 in the villages in the watershed are exposed to landslide susceptibility. It should be noted that there is a total of 333 landslide zones with an area of more than 1 hectare, equivalent to 1393 hectares, in the Sira watershed.
Conclusion: This study has provided a valuable tool for planners and crisis managers by providing high-precision landslide susceptibility maps. The findings show that combining machine learning with spatial data can revolutionize the prediction of natural hazards, especially landslides. To achieve practical results, however, cooperation between government institutions, academia, and local communities is essential to implement sustainable solutions. This research is an effective step toward reducing natural hazards by combining modern technologies and spatial analysis, and can be used as a model for other similar areas in Iran.
Type of Study: Research |
Subject:
بلايای طبيعی (سيل، خشکسالی و حرکت های توده ای)
Received: 2025/07/14 | Accepted: 2025/12/1
Received: 2025/07/14 | Accepted: 2025/12/1
Send email to the article author
| Rights and permissions | |
 |
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. |
