Global marine phytoplankton dynamics analysis with machine learning and reanalyzed remote sensing

Area of Science:

• Marine Biology
• Oceanography
• Artificial Intelligence

Background:

• Phytoplankton are vital marine organisms, producing most of the Earth's oxygen and forming the base of the marine food web.
• Environmental factors like salinity and pH significantly impact phytoplankton growth and distribution.
• Advancements in AI offer new tools for analyzing complex environmental data.

Purpose of the Study:

• To develop an AI-driven system for predicting global phytoplankton levels.
• To assess the effectiveness of supervised machine learning regression techniques for this task.
• To provide a tool that complements existing in-situ oceanographic measurements.

Main Methods:

• Utilized supervised regression algorithms: Random Forest, Extra Trees, Bagging, and Histogram-based Gradient Boosting Regressor.
• Trained models on the Copernicus Global Ocean Biogeochemistry Hindcast dataset.
• Applied techniques to reanalysis data for predicting phytoplankton concentrations.

Main Results:

• Achieved a high coefficient of determination (R²) up to 0.96 in predicting phytoplankton levels.
• Demonstrated the efficacy of the selected machine learning models.
• Identified potential for AI in large-scale phytoplankton monitoring.

Conclusions:

• Supervised machine learning regression models show strong predictive power for global phytoplankton levels.
• The developed AI model can potentially be deployed for operational monitoring.
• This approach offers a valuable supplement to traditional oceanographic data collection.