Suspended Sediment Load Forecasting Using Machine Learning-based Time Series Model: The Case of Gumara River, Ethiopia
Domaine:
environment and energyclimateType de record:
paper
Créateur:
ZabFit
Éditeur:
Hôte:
This study develops and evaluates machine learning (ML) models for forecasting suspended sediment load (SSL) in the Gumara River, Ethiopia, using long-term meteorological and hydrological time series data. Neural network architectures including ANN-MLP, WANN, LSTM, GRU, CNN, and a hybrid ConvLSTM were trained and validated, with feature importance assessed through SHAP and hyperparameters optimized via grid search. The hybrid ConvLSTM achieved the highest predictive accuracy (R2=0.96 training; R2=0.956 testing), outperforming other models. Extreme Value Analysis (EVA) was applied to high- and low-flow events, with hysteresis loops and index values revealing both clockwise and counter-clockwise sediment–discharge dynamics. Overall, the integration of ML and EVA provided a comprehensive framework for capturing complex sediment behavior under varying hydrological conditions. Hybrid deep learning models provide a robust framework for enhancing sediment forecasting accuracy, thereby supporting sustainable river basin management, and enhance long-term reservoir resilience.
Similaires
Forecasting suspended sediment load using regularized neural network: Case study of the Isser River (Algeria)Forecasting the algerian load peak profile using time series model based on backpropagation neural networksPredicting Daily Suspended Sediment Load Using Machine Learning and NARX Hydro-Climatic Inputs in Semi-Arid EnvironmentFarhanilig/Machine-Learning-Based-Forecasting-of-Fuel-Prices-in-Somalia-Using-Time-Series-ModelsSuspended Sediment Load Simulation during Flood Events Using Intelligent Systems: A Case Study on Semiarid Regions of Mediterranean BasinA Drought Forecasting Model Using the Prophet Time Series Analysis Technique