نوع مقاله : مقاله پژوهشی
نویسنده
پژوهشکده حفاظت خاک و آبخیزداری، سازمان تحقیقات، آموزش و ترویج کشاورزی، تهران، ایران.
چکیده
کلیدواژهها
موضوعات
عنوان مقاله [English]
نویسنده [English]
Soil erosion prediction at the watershed scale remains a critical challenge in sustainable land and water resource management. Although many studies have applied various empirical and statistical models, most have not adequately integrated long-term sediment records with a regional approach that accounts for both environmental and hydrological variability.The aim of this study is to develop regional models for predicting soil erosion rates through the integration of long-term sediment monitoring data and advanced statistical techniques, including Factor Analysis (FA), Principal Component Analysis (PCA), and Clustering methods, across the Urmia Lake watershed. Sediment data from 30 selected gauging stations within the major sub-basins were collected over a 20-year period and correlated with hydrological, topographic, geological, climatic, and land-use parameters of the study area. Initially, PCA was applied to reduce multicollinearity among input variables and to identify the principal factors controlling sediment transport. Subsequently, using hierarchical clustering based on hydrological behavior and sediment load, the sub-basins were divided into two distinct groups. Finally, the results of the factor analysis were employed as optimized inputs for multivariate regression modeling, leading to the development of regional predictive models. The validation results demonstrated that the developed models possess a high explanatory capacity, with an average coefficient of determination (R²) of 0.89 at the regional scale. The findings indicated that three primary factors — agricultural land area, erosion-prone lithological formations, and mean annual discharge — exerted the greatest influence on variations of the Sediment Delivery Ratio (SDR) across the sub-basins. The developed regional models provide strong potential for rapid assessment of erosion risk in newly monitored sub-basins without the need for extensive field measurements, offering an effective decision-support tool for sustainable land management, soil erosion control, and watershed policy planning within the Urmia Lake basin.
کلیدواژهها [English]
Urmia Lake basin, the largest inland water body in Iran, represents a vital socio-environmental resource and is currently facing rapid decline due to both climatic and anthropogenic stressors. Water-induced soil erosion and subsequent sedimentation processes in its tributaries drastically reduce soil fertility, degrade water quality, and lead to siltation that impairs lake water inflows and increases environmental risks for the entire basin. Identifying the key drivers of sediment yield and mapping spatial erosion potential are crucial for effective watershed management. This study addresses the challenge of limited sediment data in the region by integrating statistical and GIS-based techniques to develop robust, spatially-explicit prediction models for erosion and sediment delivery.
The Urmia Lake watershed covers approximately 52,000 km² and comprises several major sub-basins, including Zarrineh Rood, Simineh Rood, Aji Chay, and Lesser Zab. Thirty sediment gauging stations with consistent records of annual streamflow and suspended sediment (covering 1991–2021) were selected for analysis.
Thirty predictor variables covering five thematic categories—physiography, climate, geology/geomorphology, land use/vegetation, and soil physical properties—were extracted using advanced GIS platforms and remote sensing (Landsat NDVI).
- Physiography: sub-basin area, mean slope, shape index, elevation.
- Climate: mean annual precipitation, maximum rainfall intensity, aridity index.
- Geology/Geomorphology: percentage of erodible formations, lithology, drainage density.
- Land Use/Vegetation: agricultural land and rangeland area, NDVI.
- Soil: dominant soil texture, infiltration rate.
To address multicollinearity and reduce model complexity, Principal Component Analysis (PCA) was applied, transforming the suite of 30 interrelated variables into a set of orthogonal principal components. Hierarchical Cluster Analysis was then used to stratify sub-basins into two homogeneous regional groups based on physiographic and hydrological features. Within each cluster, stepwise multiple regression models were established to predict the natural logarithm of the Sediment Delivery Ratio (SDR), the main indicator of basin-scale erosion risk. Model performance was validated using the coefficient of determination (R²), Nash-Sutcliffe Efficiency (NSE), and Relative Root Mean Square Error (RRMSE). Seven widely-used empirical SDR equations (e.g., Roehl, Fournier, Garbrecht) were evaluated to select the most spatially-accurate formulation for final erosion mapping.
PCA revealed that just three principal components (agricultural land area, abundance of erodible lithology, and mean annual discharge) explained over 85% of the SDR variance across sub-basins. This demonstrates that spatial SDR variability can be reliably captured using a compact set of landscape, land use, and hydrologic indicators.
Regression models developed for the two clusters yielded strong predictive power:
- Cluster 1 (northern/cooler sub-basins): R² = 0.89, NSE = 0.85
- Cluster 2 (southern/semi-arid sub-basins): R² = 0.77, NSE = 0.72
Spatial SDR maps generated using the best-fitted empirical equation matched the ground-measured sediment data from stations with >81% accuracy.
Critical erosion hotspots (SDR > 0.4) are notably concentrated in the southern and southwestern regions of the basin, particularly in the Simineh Rood and Godarchai sub-basins. These areas are typified by steep slopes, prevalence of easily erodible geological units, and intensive agriculture—all factors that elevate erosion potential. In contrast, central and northern sub-basins displayed moderate or lower SDR values, often benefiting from greater natural vegetation cover and more stable geomorphological setting.
The study demonstrates that regional modeling, combining PCA and cluster-based regression, offers a high-accuracy, cost-effective solution for spatial erosion assessment in data-limited basins. The identification of key predictor variables and high mapping reliability provide valuable guidance for policy makers and practitioners. Future studies can further improve predictive capabilities by incorporating near real-time remote sensing and exploring machine learning approaches for capturing non-linear sediment transport dynamics.
Integrating advanced statistical approaches (PCA, Cluster Analysis, multivariate regression) with hydrological and spatial data proves effective for predicting and mapping water soil erosion in large and complex catchments. The resulting SDR regional model and spatial erosion maps offer actionable tools for prioritizing erosion control and watershed conservation efforts within the Urmia Lake basin.
Writing—original draft preparation, methodology, software, Conceptualization, funding acquisition, resources, M.T.
Data available on request from the authors.
The authors gratefully acknowledge the financial support of the The Vice-Presidency for Science, Technology and Knowledge (Project code: 24-29-29-034-980445) for conducting this research.
The authors avoided data fabrication, falsification, plagiarism, and misconduct. The study involved only soil sampling and laboratory analysis, without any interaction with humans or animals; hence, ethical approval was not applicable.
The author declares no conflict of interest
)