AI Framework Maps Global Ocean Sediment Carbon Cycling

Researchers at the University of Manchester have developed a physics-based AI framework capable of predicting the movement of dissolved organic carbon (DOC) between seawater and marine sediments on a global scale. This breakthrough allows scientists to quantify carbon cycling processes that were previously impossible to map due to extreme computational demands.

Using AI "emulators" to mimic complex mechanistic models, the team discovered that 11% of particulate organic carbon falling on the seafloor is returned to seawater as dissolved organic carbon. Additionally, 24% of particulate organic carbon is assimilated into minerals. The study further reveals that almost half of all solid-phase organic carbon found in the upper metre of marine sediments originates from dissolved carbon that was either sorbed onto or taken into minerals.

The Computational Barrier

Quantifying the carbon budget across the sediment-water interface is essential for understanding global climate dynamics. Until now, traditional mechanistic models were often too unstable, time-consuming, or computationally expensive to run at a global resolution. This created a significant gap in the scientific community's understanding of the planet's total carbon cycle.

To bridge this gap, the Manchester team tested various machine learning approaches. They found that simpler feedforward artificial neural networks yielded more accurate predictions than more complex random forest models or deep learning architectures, providing a reliable path toward global-scale simulation.

Implications for Climate Science

This framework is significant because it enables the integration of sediment carbon cycling into global circulation models for the first time. By providing a scalable tool to simulate how marine carbon reservoirs respond to environmental shifts, the research offers a new way to analyze the ocean's role in carbon sequestration.

Dr. Peyman Babakhani, a Lecturer in Geoenvironmental Engineering, noted that the framework can play a substantial role in testing potential ocean-based climate change mitigation scenarios in silico. This digital approach allows for rapid iteration and testing of hypotheses that would otherwise be impossible to observe in real-time.

Future Applications

With the ability to explore these global-scale processes, researchers can now better predict how the seafloor interacts with the water column under varying climate conditions. The next step involves using these emulators to refine global carbon budgets and test the efficacy of theoretical mitigation strategies without the need for prohibitively expensive physical experiments or unstable traditional simulations. This shift toward AI-driven emulation marks a turning point in how oceanographers approach the complexities of the deep-sea carbon sink.