Document Type : Research Paper
Authors
1 Member of scientific boAssistant Professor, Soil Conservation and Watershed Management Research Institute, Agricultural Research, Education and Extension Organization (AREEO), Tehran, Iran
2 Associate Professor, Soil Conservation and Watershed Management Research Institute, Agricultural Research, Education and Extension Organization (AREEO), Tehran, Iran
Abstract
This study examines the accurate estimation of suspended sediment in the Atrak River, particularly at the Hootan station. Suspended sediment in rivers, especially in semi-arid regions, poses significant challenges for water resource management and sediment control in dam reservoirs. In this research, a combination of classical and intelligent methods was used to estimate suspended sediment, including sediment rating curves, neural networks, and deep learning models. Key variables influencing sediment were identified using a random forest algorithm, and the data was divided into homogeneous groups. The ensemble learning model, XGBoost, was selected as the best model, demonstrating high accuracy in predictions. Results indicate that XGBoost outperformed other models with the lowest error and highest performance index. This model effectively manages highly skewed data and identifies complex nonlinear relationships. Additionally, the combined approach used in this study improved predictions compared to traditional methods. However, data quality and hydrological changes significantly impact model performance. This research highlights the importance of advanced machine learning techniques in analyzing hydrological data and emphasizes the need for a link between data science and water resource management. The findings of this study can serve as a reference for policymakers and water resource managers in enhancing sediment management and water quality in rivers.
Keywords
)