Oceanic oxygen minimum zones are located near nutrient-rich coastal upwelling regions that support high biological production near the surface. When dead organic material sinks down, it is chemically decomposed into basic nutrients, like nitrates and phosphates, but also into greenhouse gases, such as carbon dioxide and nitrous oxide. The decomposition consumes oxygen, decreasing the oxygen content of the surrounding water. In turn, lower oxygen concentrations lead to higher production of greenhouse gases.

The regions in which oxygen minimum zones can form are not only determined by high biological productivity, but also by water that has been isolated from the atmosphere for a relatively long time. Areas in which the water is almost stagnant because of the absence of currents going in or out exist in the eastern tropical oceans; they coincide well with the locations of oxygen minimum zones. Upper-ocean densities decrease with warmer temperatures, increasing the stability of the water column. This can lead to less atmospheric oxygen supply to the ocean, and so an increase in the extent of oxygen minimum zones and/or their strength.

In our study, we employ a climate model similar to those typically used in the last IPCC report. We release particles into the modelled ocean circulation and investigate how they are transported over time. By comparing results from pre-industrial model runs with high-carbon dioxide model runs, we can estimate the effect that a changed oceanic circulation will have on the development of oxygen minimum zones. We find that under increased atmospheric carbon dioxide, oxygen minimum zones become less connected to the surface ocean and so exchange with the atmosphere is hindered. As a result, the feedback between biogeochemical processes in oxygen minimum zones and in the surface ocean is likely to be weakened in the future.

However, we have to keep in mind that this prediction is based on a model that does not resolve all small-scale processes that might be needed to represent the system adequately. For now, we are isolating and describing just one of many feedback loops in a typical IPCC-type model to understand which processes and feedbacks are acting in IPCC predictions. While this in itself is important, future results will have to be tested in models that explicitly resolve more important processes to assess how well our model actually represents the real world.