Global cycling of dissolved organic carbon between seawater and sediments quantified using physics-based artificial intelligence.

The origin of dissolved organic carbon (DOC) in the ocean, which represents the largest active reservoir of exchangeable carbon on Earth, remains unknown due to the complexity of determining DOC exchange between marine sediments and seawater. Here, we emulate underlying processes of DOC dynamics using physics-based artificial intelligence (AI) to quantify DOC cycling and preservation in sediments at the global scale. We conclude that up to 11% of all particulate organic carbon deposited on the global seafloor returns to seawater as DOC following hydrolysis within the sediment, which is around half of the riverine DOC flux to the ocean. Around 50% of the total solid-phase organic carbon in the upper meter of sediments is formed through the sorption of DOC to minerals. Further, we show that the contribution of the abyssal plain to global DOC efflux and preservation, which is typically entirely overlooked in carbon cycle studies, is equivalent to half the contribution of the continental margins. This study also highlights the power of simpler AI algorithms compared to more complex models and extends the principle of parsimony in mathematical modeling to the context of physics-based AI.