Researchers at The University of Manchester have created a new artificial intelligence (AI) method based on physics principles that allows for precise global predictions of how dissolved organic carbon moves between seawater and marine sediments. This work, led by Dr. Peyman Babakhani from the Department of Civil Engineering and Management and conducted in collaboration with Dr. Majid Sedighi, shows that even simple AI algorithms can effectively mimic complex environmental models that are usually too computationally intensive to use on a planetary scale. The research has been published in The Innovation journal. The study found that 11% of the particulate organic carbon reaching the seafloor is returned to seawater as dissolved organic carbon, while 24% attaches to minerals. Remarkably, about half of the solid-phase organic carbon in the top meter of marine sediments appears to come from dissolved carbon that has attached to minerals. These findings provide the first global measurement of how dissolved organic carbon cycles through sediments and emphasize its role in Earth's long-term carbon budget. To develop the modeling framework, the researchers tested various AI techniques, including deep learning, random forest models, and simpler artificial neural networks. Surprisingly, the simplest algorithms produced the most accurate predictions. The team confirmed their results by comparing the AI model's output with low-resolution global maps, where the original environmental model could still be solved numerically, and with algebraic solutions for variables that have known mathematical expressions. They also observed that increasing the complexity of the neural networks consistently reduced prediction accuracy, offering rare real-world support for the principle of parsimony—also known as Occam's razor—in AI model design. These findings have significant implications for climate science. Understanding carbon movement across the sediment-water interface is crucial for studying global climate dynamics, but this has been limited by computational challenges. The new AI-based framework offers a fast, scalable, and accurate way to model sediment carbon processes. It can be integrated into global climate models to explore potential ocean-based strategies for mitigating climate change. The research opens new possibilities for simulating how marine carbon reservoirs may respond to environmental changes in the coming decades. Dr. Babakhani, a lecturer in geoenvironmental engineering, noted, "The modeling framework developed in this study can play a substantial role in testing potential ocean-based climate change mitigation scenarios in silico. With this approach, we can finally explore global-scale carbon cycling processes that were previously impossible to quantify."