Real-time chlorophyll-a forecasting using machine learning framework with dimension reduction and hyperspectral data.

Area of Science:

• Environmental Science
• Remote Sensing
• Data Science

Background:

• Water quality monitoring is essential, with chlorophyll-a concentration being a key indicator.
• Hyperspectral imaging offers detailed spectral information beyond traditional RGB cameras.
• Existing methods for chlorophyll-a prediction have limitations in real-time application and precision.

Purpose of the Study:

• To develop a real-time forecasting mechanism for chlorophyll-a concentrations.
• To leverage hyperspectral data and machine learning algorithms for improved water quality assessment.
• To enable early detection of algal blooms through accurate chlorophyll-a prediction.

Main Methods:

• Utilized hyperspectral imaging to capture detailed spectral data of water bodies.
• Applied Partial Least Squares (PLS) for dimensionality reduction of hyperspectral data.
• Implemented and optimized various machine learning algorithms, including LightGBM, for prediction.

Main Results:

• Achieved high prediction accuracy with R² values of 0.9 or above and low RMSE.
• The LightGBM model demonstrated superior performance with a mean R² of 0.963 and mean RMSE of 2.679.
• Successfully developed a real-time chlorophyll-a forecasting system.

Conclusions:

• The developed system provides a robust and precise method for real-time chlorophyll-a forecasting.
• This approach has significant potential for early algal bloom detection and water quality management.
• Integration with IoT platforms facilitates widespread application in environmental monitoring systems.