Assessing the algal population dynamics using multiple machine learning approaches: Application to Macao reservoirs.

Area of Science:

• Environmental Science
• Water Resource Management
• Computational Ecology

Background:

• Reservoir water quality is crucial for human and animal health.
• Eutrophication poses a significant threat to reservoir water safety.
• Machine learning (ML) offers powerful tools for environmental process analysis, including eutrophication.

Purpose of the Study:

• To compare the performance of various ML models in analyzing algal dynamics using time-series water quality data.
• To investigate the influence of specific water quality parameters on algal growth in two Macao reservoirs.
• To identify the most effective ML model for data reduction and interpretation of algal population dynamics.

Main Methods:

• Analysis of water quality data from two reservoirs using multiple ML approaches.
• Implementation of stepwise multiple linear regression (LR), Principal Component (PC)-LR, PC-Artificial Neuron Network (ANN), and GA-ANN-Connective Weight (CW) models.
• Systematic investigation of water quality parameter influences on algal proliferation.

Main Results:

• The GA-ANN-CW model demonstrated superior performance, indicated by higher R-squared and lower error metrics (MAPE, RMSE).
• This model effectively reduced data size and improved the interpretation of algal population dynamics.
• Key water quality parameters like silica, phosphorus, nitrogen, and suspended solids were identified as directly impacting algal metabolism.

Conclusions:

• The GA-ANN-CW model is highly effective for predicting algal population dynamics from time-series data.
• This study enhances the application of ML in managing reservoir water quality and understanding eutrophication.
• Identifying critical water quality parameters aids in targeted interventions to control algal blooms.