Predicting suspended sediment load in Peninsular Malaysia using support vector machine and deep learning algorithms.

Area of Science:

• Environmental Science
• Hydrology
• Data Science

Background:

• High suspended sediment loads (SSL) negatively impact water resources, water quality, agriculture, and infrastructure.
• Accurate SSL prediction is crucial for effective water resource management and mitigating environmental damage.

Purpose of the Study:

• To develop and propose a single, universally applicable machine learning model for predicting river suspended sediment load (SSL) in Peninsular Malaysia.
• To evaluate the performance of Support Vector Machine (SVM), Artificial Neural Network (ANN), and Long Short-Term Memory (LSTM) algorithms for SSL prediction.

Main Methods:

• Utilized 11 river datasets from the Malaysian Department of Irrigation and Drainage, containing streamflow (SF) and SSL data.
• Developed and tested machine learning models, including SVM, ANN, and LSTM algorithms.
• Identified the optimal input scenario (current-day SF, previous-day SF, previous-day SSL) for the ANN model (ANN3).

Main Results:

• The ANN3 model demonstrated superior predictive performance, achieving the best results for 5 out of 11 datasets.
• ANN3 exhibited the highest average RM score (2.64), indicating high reliability and accuracy in SSL predictions across diverse datasets.
• ANN3 outperformed other tested models in terms of predictive accuracy and reliability.

Conclusions:

• The ANN3 model, utilizing specific streamflow and prior SSL data, is proposed as a universal model for accurate SSL prediction in Peninsular Malaysia.
• This research provides a reliable tool for environmental monitoring and management of water resources affected by suspended sediments.