Biochemical Oxygen Demand Prediction Based on Three-Dimensional Fluorescence Spectroscopy and Machine Learning

Area of Science:

• Environmental Science
• Analytical Chemistry
• Water Quality Monitoring

Background:

• Biochemical oxygen demand (BOD) is a key indicator of organic water pollution.
• Traditional BOD5 determination methods are complex and rely on precise dissolved oxygen measurements.
• There is a need for facile and efficient methods for water quality assessment.

Purpose of the Study:

• To develop a simplified method for predicting biochemical oxygen demand (BOD5) using fluorescence spectroscopy.
• To combine three-dimensional fluorescence spectroscopy with parallel factor analysis and machine learning for BOD5 prediction.
• To establish a non-contact, efficient approach for water quality monitoring.

Main Methods:

• Water samples were analyzed using national standard methods for BOD5 determination over a five-day incubation period.
• Three-dimensional fluorescence spectroscopy data were collected at regular intervals.
• Parallel factor analysis was used to extract fluorescence components, and a random forest model was employed to correlate these components with BOD5 values.

Main Results:

• A strong correlation was found between extracted fluorescence components and BOD5 values.
• The random forest model successfully predicted BOD5 with a high goodness of fit (R² = 0.878) and low mean square error (MSE = 0.28).
• The method demonstrated effective BOD5 prediction using non-contact fluorescence measurements.

Conclusions:

• Fluorescence spectroscopy combined with machine learning provides a facile and efficient method for BOD5 prediction.
• This approach simplifies water quality analysis by avoiding complex dissolved oxygen measurements.
• The developed technique offers a convenient solution for monitoring large volumes of water samples and assessing water quality.