Zemskova, V., Prasad, S., and Yin, H. (2026)
Presented at:
Ocean Sciences Meeting 2026Oceans play a critical role in moderating atmospheric carbon dioxide levels as they are one of the largest carbon sinks. Understanding where carbon is stored and how it circulates within the ocean interior is instrumental in estimating the ocean’s ability to mitigate climate change. However, it has been challenging to quantify the rate of carbon uptake by the ocean due to the paucity of observational measurements of dissolved carbon concentrations in the ocean interior and subsequently understanding what processes guide the spatio-temporal distribution of carbon. To address these challenges, we use the output from the ECCO-Darwin model, which is a global circulation model constrained by observational data and that couples physical and biogeochemical processes. Such dataset allows us to analyze the ocean carbon storage over three decades (1992-2022) at a regular spatio-temporal resolution. Using unsupervised machine learning methods, we study the three-dimensional distribution of ocean dissolved inorganic carbon to identify regions that (1) have similar vertical profiles and (2) similar physical and biogeochemical processes that govern the carbon budget. The separation of the ocean into such carbon-specific regions, in contrast with simpler geographical regions or clustering based on other physical and biogeochemical properties, can help improve ocean carbon modelling and facilitate analysis of spatio-temporal trends in ocean carbon storage.
