Predicting global distributions of eukaryotic plankton communities from satellite data

Area of Science:

• Marine ecology
• Oceanography
• Remote sensing

Background:

• Previous models focused on phytoplankton size or groups.
• Global plankton dynamics require integrated approaches for phytoplankton and heterotrophic protists.

Purpose of the Study:

• To identify marine plankton community types using a global metabarcoding dataset.
• To predict the biogeography of these community types using satellite remote sensing data.
• To analyze spatiotemporal distribution and long-term trends of plankton communities.

Main Methods:

• Inferred a co-occurrence network from a planetary-scale eukaryotic 18S V4 rDNA metabarcoding dataset.
• Identified six distinct plankton community types.
• Applied machine learning to predict community types using 17 satellite-derived parameters.
• Validated model accuracy at 67%.

Main Results:

• Successfully identified six plankton community types.
• Developed a predictive model for community type distribution using satellite data.
• Demonstrated improved prediction accuracy with 17 parameters compared to chlorophyll a and temperature alone.
• Mapped global spatiotemporal distributions over 19 years, showing seasonal changes and long-term trends.

Conclusions:

• Satellite remote sensing can effectively predict marine plankton community types and their global distribution.
• The model highlights seasonal shifts in subarctic-subtropical boundary regions.
• Observed long-term distribution trends suggest plankton communities are responding to ocean warming.