A soft sensor for simulating algal cell density based on dynamic response to environmental changes in a eutrophic...

Area of Science:

• Environmental Science
• Water Quality Monitoring
• Machine Learning Applications

Background:

• Eutrophic lakes frequently experience harmful algal blooms, necessitating efficient monitoring.
• Traditional methods for measuring algal cell density (ACD) are labor-intensive and slow, hindering rapid bloom assessment.

Purpose of the Study:

• To develop a rapid, soft sensor-based approach for simulating ACD using machine learning.
• To improve water quality assessment and algal bloom management in eutrophic lakes.

Main Methods:

• Utilized a soft sensor approach with surrogate indicators and machine learning models.
• Applied ensemble learning, specifically extreme gradient boosting (XGBoost), to simulate ACD.
• Implemented multi-stage variable selection, identifying seven key predictors from 43 candidates.

Main Results:

• XGBoost models demonstrated superior performance over traditional algorithms.
• Key variables identified include dissolved oxygen, chlorophyll-a, Secchi disk depth, pH, COD_Mn, week of the year, and wind velocity.
• The final soft sensor model achieved a high R² of 0.761, indicating reliable spatiotemporal generalization.

Conclusions:

• The soft sensor approach offers a novel method for simulating ACD in eutrophic lakes.
• This tool facilitates rapid bloom condition assessment, aiding local administrations in emergency prevention and control efforts.