Ecosystem community structure shapes carbon dynamics in the ocean

Rebecca Mary Wright , Erik Buitenhuis, Rui Ying, Joe Guest, Philip Townsend, Tereza Jarnikova, Anna Sommer, Marie-Fanny Racault, Corinne Le Quéré

Understanding how marine ecosystems influence air–sea CO2 flux under climate change remains a major uncertainty in ocean carbon-cycle projections. Progress is hampered by insufficient representation of ecosystem complexity in climate models and insufficient observations for developing and calibrating complex marine ecosystem models. We present a novel three-branch ensemble of the PlankTOM12 global ocean biogeochemistry model, designed to isolate the influence of ecosystem community structure on carbon dynamics. PlankTOM12 represents ecosystem dynamics and their influence on carbon and other biogeochemical cycles through twelve interacting Plankton Functional Types (PFTs) spanning four trophic levels: heterotrophic bacteria/archaea, six phytoplankton PFTs, and five zooplankton PFTs. The branches share this ecosystem framework but differ in ecological parameterisations, producing distinct emergent communities. Branch 1 (GCB) is the historical configuration used in the Global Carbon Budget 2025. Branch 2 (OBio) focuses on reproducing observation-based global estimates of PFT carbon biomass, while Branch 3 (SMor) focuses on reproducing PFT vertical profiles. We evaluate each branch against ecosystem and biogeochemical observations and quantify differences in PFT biomass partitioning. The branches differ in total biomass, allocation among PFTs, and relative total phytoplankton and zooplankton biomass. These ecosystem differences propagate to primary production; sinking fluxes of organic carbon, calcium carbonate, and silica; carbon export and export efficiency. Ecosystem-driven differences also affect the latitudinal distribution of surface-ocean CO2 flux in the tropics and Southern Ocean. We conclude that marine ecosystems have the potential to influence air–sea CO2 flux under climate change and suggest priorities for future model–data development and integration.