Predicting water quality through daily concentration of dissolved oxygen using improved artificial intelligence

Area of Science:

• Hydrology
• Environmental Science
• Computational Intelligence

Background:

• Dissolved oxygen (DO) is a critical hydrological parameter and a primary indicator of water quality.
• Accurate DO prediction is essential for effective water resource management and ecological health monitoring.

Purpose of the Study:

• To introduce and evaluate four novel integrative methods for predicting DO concentration.
• To compare the performance of these methods using real-world hydrological data.

Main Methods:

• Employed multi-layer perceptron neural network (MLPNN) as the predictive model.
• Utilized four optimization algorithms: Teaching-Learning-Based Optimization (TLBO), Sine Cosine Algorithm, Water Cycle Algorithm (WCA), and Electromagnetic Field Optimization (EFO).
• Trained and tested models using hydrological data from the Klamath River, Oregon (USGS station).

Main Results:

• All models demonstrated reliability in DO prediction, with WCA-MLPNN showing initial superiority in the training phase (MAE: 0.9624).
• EFO-MLPNN and TLBO-MLPNN slightly outperformed WCA-MLPNN in the testing phase, indicated by Pearson correlation coefficients (Rp) and Root Mean Square Errors (RMSE).
• EFO-MLPNN was identified as the most efficient tool considering complexity and optimization time.

Conclusions:

• The developed integrative models, particularly EFO-MLPNN, offer improved accuracy for machine learning-based DO modeling.
• These novel methods provide advanced tools for hydrological parameter prediction and water quality assessment.
• The findings suggest significant advancements in computational intelligence applications for environmental monitoring.