Abstract
The Arctic Ocean (AO) remains an important carbon dioxide (CO 2 ) sink, accounting for approximately 2%–9% of the global oceanic carbon uptake. Over recent decades, rapid sea‐ice loss has altered oceanic carbon uptake dynamics, introducing substantial interannual variability. Although previous work has emphasized snapshot case studies or long‐term trends, our study provides an investigation of the interannual variability of oceanic carbon uptake and its key environmental drivers in the western AO. Using a newly developed machine learning model, we reconstructed high‐resolution (8‐day, 6‐km) sea surface partial pressure of CO 2 ( p CO 2 ) maps for the May to August period from 2003 to 2022. The model was trained with >235,000 underway p CO 2 measurements and paired remote sensing products. Model‐derived oceanic carbon uptake exhibits >20% variability relative to its 20‐year mean, showing distinct dynamical regimes in response to sea‐ice cover variations. In the Canada Basin, interannual variability of oceanic carbon uptake is predominantly controlled by open‐water duration, whereas in the Bering Sea and Pacific subarctic regions, air‐sea CO 2 flux intensity is the dominant driver. The relative importance of open‐water duration versus CO 2 flux intensity in modulating oceanic carbon uptake is quantified through correlation‐based metrics. Principal component analysis of environmental variables further reveals that sea‐ice cover fluctuations induce a compensation between wind mixing and biological utilization on oceanic carbon uptake. Unlike traditional atmospheric inversion or biogeochemical models, this study presents an alternative data‐driven framework that incorporates seasonal dynamics into assessing interannual and long‐term changes in AO carbon uptake.

This publication has 54 references indexed in Scilit: