|
Assessment of gully erosion-prone areas in the Aleshtar basin using machine learning models
|
Ali Davoudi *  , Amir Karam  , Parviz Zeaiean Firouzabadi , Ali Ahmadabadi |
| Department of Geomorphology, Faculty of Geography, Kharazmi University, Tehran, Iran. |
|
|
Abstract: (48 Views) |
Objective: Gully erosion is one of the most prominent forms of water erosion, with significant impacts on land degradation, reduced soil fertility, and increased sedimentation in water resources.
Method: In this study, to identify and map areas susceptible to gully erosion in the Aleshtar watershed, Generalized Linear Models (GLM), Boosted Regression Trees (BRT), and Boosted Linear Models (BLM) were used, along with spatial-environmental parameters influencing the occurrence of gully erosion, and the R software. The variables examined included topographic, climatic, hydrological, vegetation cover, land use, and physical and chemical soil characteristics, which were extracted using Geographic Information Systems (GIS) and remote sensing data.
Results: Spatial distribution assessment of gully-prone areas by all three models showed that the marginal areas of the Aleshtar plain, including the central, southern, and western parts of the watershed, have the highest susceptibility to gully erosion. Model accuracy evaluation using the ROC curve indicated that the BLM model (AUC = 0.84) performed better at identifying high-risk gully erosion areas than the GLM and BRT models. Based on the BLM model, 39% of the watershed falls into the very high and high susceptibility classes, 22% into the moderate class, and 39% into the low and very low susceptibility classes. Analysis of the importance of variables used in modeling showed that distance from rivers, Topographic Wetness Index (TWI), geological units, Terrain Ruggedness Index (TRI), and precipitation are the most influential factors in the occurrence of this phenomenon.
Conclusion: The results of this study can serve as an effective tool in land management and erosion mitigation planning. It is recommended that future studies use advanced hybrid models and higher-resolution data to improve prediction accuracy and provide more optimal solutions.
|
|
| Keywords: Gully erosion, Machine learning model, Aleshtar watershed, Geographic Information, System GIS, Land management, BLM model. |
|
|
Full-Text [PDF 2173 kb]
(13 Downloads)
|
Type of Study: Applicable |
Subject:
Special
Received: 2025/09/8 | Accepted: 2025/09/21 | Published: 2025/09/21
|
|
|
|
|
|
|
| Send email to the article author |
|
|
|
| Add your comments about this article |
|
|
|
