Towards a mechanistic understanding of marine oxygen dynamics

Abstract

Our current understanding of the ocean's oxygen cycle, its dynamics and current global decline (termed deoxygenation), has required a combination of observational methods and forms of numerical model to build. Yet while global and regional trends are well established, a process-led or mechanistic understanding of ocean oxygen dynamics will be necessary to anticipate the wider marine ecosystem impacts of deoxygenation. This thesis presents several distinct methodological approaches for investigating the governing processes behind oceanic dissolved oxygen concentrations. First an analysis of oceanic oxygen subduction, the transport of oxygenated seawater in to the ocean interior, is performed using an ensemble of Earth system models from the pre-industrial experiment of CMIP6 (Climate model intercomparison project, Phase 6). While the models global subduction is supported by the only available observational constraints, the contributing terms to the subduction budget differ greatly between models, and can cluster in their respective model development groups. This effect was caused by a spurious atmospheric bias present in half of the ensemble evaluated here, a result which motivated some investigation in to observational methods. Our subsequent analysis makes the novel application of a causal discovery algorithm to one of the most complete time series of ocean oxygen, in the Oyashio region off of the East coast of Hokkaido, Japan. Here we demonstrate that tidal mixing, likely at the remote source region of the Bussol Straits, can be identified in the oxygen time series, and that this effect can be quantified using a causal discovery analysis. The tidal influence is most pronounced below the density range typically within the surface mixed layer, implying a separation between the governing processes of the mixed layer oxygen and that of the ocean interior within the Oyashio region. Finally, we address the assumption of linearity in time series of ocean oxygen, using the Hawaii ocean time series (HOTS). An assessment of the evidence for nonlinearity is performed in physical and biogeochemical variables over the upper ~1000m of HOTS, using an iterative SMap analysis. A causal analysis, under the assumption nonlinear dynamics may be present, was then performed using convergent cross-mapping. Reasonable evidence for nonlinear dynamics is present in oxygen, AOU and the other biogeochemical time series, with a clear depth dependence on the strength of this evidence. No evidence was found for nonlinear causal interactions between oxygen and these biogeochemical time series.

Date of Award	13 May 2025
Original language	English
Awarding Institution	University of Bristol
Supervisor	Rory J Bingham (Supervisor) & Oliver Andrews (Supervisor)