Oxygen Utilization Rate (OUR) Underestimates Ocean Respiration – A Model Study

We use a simple 1D model and several 3D global ocean biogeochemical models to evaluate the concept of computing the subsurface oceanic oxygen utilization rate (OUR) from the changes of apparent oxygen utilization (AOU) and water age on isopycnal surfaces. We find that OUR underestimates globally averaged total oceanic oxygen consumption (respiration) substantially. Most of this difference is observed in the upper 1000m of the ocean, with the discrepancies increasing towards the surface.

Hitherto the major problem in OUR determination has been considered to be the determination of water age, that is the time passed since the last contact of a body of water with the atmosphere, when its oxygen concentration is assumed to have reached equilibrium with air (“saturation”).. We find, however, more fundamental and large errors in the concept of OUR itself. AOU from which OUR is calculated is not only the imprint of respiration in the ocean’s interior, but is also strongly affected by advection and diffusive mixing. Similarly, water age is the net result of ageing and transports of the age tracer. Only when both tracers (AOU and age tracer) are affected by tracer transports in the same way does the OUR calculated represent the correct rate of oxygen consumption. This is the case only with uniform respiration rates, or when advection is the principal mode of transport. Inhomogeneous distribution of respiration yields underestimates (when maximum rates occur near the outcrops of isopycnals) or overestimates (when maxima occur far from the outcrops). Given the distribution of respiration in the ocean (and in the models employed), that is high rates near their high latitude outcrops and low rates below the oligotrophic gyres as well as rates decreasing with water depth everywhere, underestimates are the rule. Integrating these effects globally in an ocean biogeochemical model we find that AOU-over-age based estimates underestimate true model respiration by a factor of three.

In order to relate our model study to state-of-the art OUR estimations from the field, we finally compute OUR also using tracer ages from transient tracers (e.g. CFCs) and age estimates based on the transit time distribution (TTD) of the models.