Work

Enhancing the prediction of suspended sediment load in the Blue Nile River using machine learning under data-limited conditions

Year: 2025

Type: article

Abstract: The Blue Nile River This study examines sediment dynamics in the Blue Nile River and evaluates the performance of several machine learning models in predicting sediment transport. Sediment transport in this region presents a major challenge for reservoir operation, agricultural productivity, and sustainable water management. To address this, a data-driven modelling approach incorporating Artificial Neural Network (ANN), Bayesian Neural Network (BNN), Multiple Linear Regression (MLR), Decision Tree (M5Tree), Multivariate Adaptive Regression Splines (MARS), and Random Forest (RF) models are developed to predict sediment discharge and concentration using long-term observational data. Results reveal that predicting sediment concentration is inherently more complex than predicting sediment load due to its nonlinear behaviour. Among the applied techniques, BNN and MARS models achieve the highest predictive accuracy for sediment load, producing consistent and transferable results across different sediment conditions within the Blue Nile Basin. Model sensitivity analysis highlights that prediction accuracy improves significantly when both sediment concentration and rainfall are included as inputs. Incorporating spatial data and applying logarithmic transformations further enhance the model robustness. The findings demonstrate that the proposed data-driven methodology provides reliable sediment load estimates, supporting improved reservoir management, irrigation planning, and mitigation of sedimentation impacts. Future extensions should integrate physical process representations and broader environmental parameters to enhance prediction accuracy and model generalization. • Data-driven modelling for prediction of suspended sediment load. • Evaluation of machine learning techniques for suspended sediment load under data-limited conditions. • Application to discharge and sediment load in the Blue Nile River.

Source: Journal of Hydrology Regional Studies

DOI: 10.1016/j.ejrh.2025.102979

Authors: Ishraga S. Osman, Mohammed Seaid, Mohamed A. Osman

Institutions: University of Khartoum, Durham University, Université Mohammed VI Polytechnique

Language: English

Topic: Hydrological Forecasting Using AI

Subfield: Environmental Engineering

Field: Environmental Science

Domain: Physical Sciences

SDG: Clean water and sanitation

Open Access status: gold

Estimated APC paid: $2,600