Investigating machine learning models in predicting lake water quality parameters as a 3-year moving average

Area of Science:

• Environmental Science
• Ecology
• Data Science

Background:

• Lake water quality is crucial for ecosystem health, influencing biotic and abiotic interactions.
• Accurate monitoring of water quality parameters is essential for ecosystem management.

Purpose of the Study:

• To employ various machine learning (ML) methods for predicting key water quality parameters.
• To evaluate the performance of different ML models in forecasting lake water quality.

Main Methods:

• Utilized ML algorithms including Classification and Regression Tree (CART), Chi-Squared Automatic Interaction Detector (CHAID), C5 tree, Quick, Unbiased, and Efficient Statistical Tree (QUEST), Multilayer Perceptron (MLP), and Radial Basis Function (RBF) neural networks.
• Applied C5 tree and QUEST for data classification and group prediction.
• Employed other models for predicting water quality parameter concentrations using a 3-year moving average.

Main Results:

• The CART decision tree demonstrated high accuracy in predicting dissolved oxygen (DO) concentration (NSE = 0.978, bias = 0.126).
• The C5 tree achieved superior performance in data classification, correctly identifying 33 out of 36 groups.

Conclusions:

• Machine learning models show significant potential for accurate lake water quality prediction.
• CART and C5 tree models offer robust solutions for predicting specific water quality parameters and classifying data.