Suspended sediment load prediction using sparrow search algorithm-based support vector machine model.

Area of Science:

• Environmental Engineering
• Hydrology
• Artificial Intelligence

Background:

• Suspended sediment load (SSL) prediction is crucial for hydrological modeling and water resources engineering.
• Sediment transport is complex and non-linear, influenced by rainfall, flow strength, and sediment supply.
• Artificial intelligence (AI) offers advanced solutions for multifaceted problems in water resource engineering.

Purpose of the Study:

• To develop a robust Support Vector Machine with Sparrow Search Algorithm (SVM-SSA) model for suspended sediment load (SSL) prediction.
• To evaluate the performance of the SVM-SSA model against other hybrid models and a benchmark SVM model.
• To assess the model's accuracy using metrics such as MAE, RMSE, R², and E_NS.

Main Methods:

• Proposed a novel SVM-SSA model for SSL computation in the Brahmani river basin.
• Considered five different scenarios for model development, incorporating lagged sediment and discharge data.
• Compared SVM-SSA with SVM-BOA, SVM-GOA, SVM-BA, and a conventional SVM model.

Main Results:

• The SVM-SSA model demonstrated high accuracy in predicting SSL, particularly for scenario V (3-month lag for sediment and discharge).
• Achieved superior performance with RMSE = 15.5287, MAE = 15.3926, and E_NS = 0.96481.
• The conventional SVM model yielded the poorest results, highlighting the effectiveness of the proposed AI approach.

Conclusions:

• The SVM-SSA model is a precise and reliable approach for modeling suspended sediment load in rivers.
• The developed model meets the accuracy demands of practical engineering applications.
• The approach significantly reduces computational time while ensuring high prediction precision.