Multi-decadal oxygen loss in the North Atlantic driven by stratification and surface nutrient depletion

Andrew John Margolskee, University of Washington Seattle, School of Oceanography, Seattle, WA, United States, Takamitsu Ito, Georgia Institute of Technology, Atlanta, GA, United States, Matthew C Long, National Center for Atm Res, Boulder, CO, United States, Hartmut Frenzel, University of Washington Seattle Campus, School of Oceanography, Seattle, WA, United States and Curtis A. Deutsch, University of Washington Seattle Campus, School of Oceanography, Seattle, United States
Global upper-ocean oxygen has been declining since the 1980's as global ocean temperature rises. The relationship between ocean oxygen and heat content stems from a combination of reduced O2 solubility in warmer surface waters, and an additional loss of O2 in the subsurface as those waters circulate more slowly and/or undergo higher rates of biological respiration. We analyzed the O2/heat relationship in historical observations and retrospective Earth System Model simulations, including a CESM hindcast and CMIP5 models, and found that models underpredict the historical decline of ocean oxygen per degree of warming by as much as 50%. This global bias is best observed in the subpolar North Atlantic, a region important for ocean ventilation that has relatively abundant historical observations of oxygen, nitrate, and temperature over the past 50 years. Historical hydrographic observations reveal significant interdecadal changes in surface nitrate that mirror those in subsurface O2: periods of warmer sea-surface temperature and ocean heat content exhibit lower sea-surface nitrate along with reduced upper ocean oxygen content. Furthermore, the relationship between surface nitrate and sea-surface temperature is weaker in models than in observations by ~50%, similar to the bias in the deep O2/heat relationship. Taken together, these results imply that the drawdown of surface nutrients during warm periods account for a substantial fraction of deoxygenation on decadal timescales, and that the bias towards weak deoxygenation is caused by net surface ecosystem production that is too stable. We discuss the roles of winter-time mixing, gyre circulation, and ecosystem processes in controlling the similarity in observed and modeled trends, the differences in their magnitude, and the implications for projections of future deoxygenation.