The concentration of dissolved oxygen in ocean waters is a major determinant of the distribution and abundance of marine species globally. Deoxygenation – a reduction of dissolved oxygen in water – is considered one of the four major human-induced stressors on ocean ecosystems alongside warming, changes in primary productivity and ocean acidification. The decline in oxygen (O₂) solubility with increased ocean temperature is responsible for approximately 15% of the reduction in O₂ concentration, with the remaining 85% associated with reduced O₂ supply due to increased ocean stratification and increased deep-sea microbial respiration.

The average dissolved oxygen concentration in the ocean is presently 162 µmol per kg. Concentrations range from over 500 µmol per kg in productive Antarctic waters to zero in coastal sediments and in deep layers of isolated water bodies, such as the Black Sea and the Cariaco Basin. Most organisms are not very sensitive to oxygen levels as long as the concentrations are high enough. But once O₂ drops below a certain threshold, the organism suffers from a variety of stresses, leading ultimately to death if the concentrations stay too low for too long. Such conditions are termed hypoxic.

Thresholds for hypoxia vary greatly between organisms. We use the term hypoxia for water masses with O₂ concentrations below 80 µmol per kg. Zones with lower O₂ are effectively “dead zones” for many higher animals. The most intense (O₂ less than 20 µmol per kg) and largest oxygen minimum zones (OMZs), or suboxic layers, are mainly localized in the subsurface of upwelling regions in the Eastern Pacific and Northern Indian open oceans.

Sub-surface oxygen concentrations averaged between 200m and 600m from World Ocean Atlas 2009 (µmol per kg)

Light and dark red stripes indicate waters with O₂ <100 µmol per kg and O₂ <20 µmol per kg respectively.

The biggest threat related to open ocean deoxygenation is that of declines in biodiversity, through unknown ecosystem shifts with impacts on organisms within the affected areas as well as in their vertical and horizontal proximities. In most marine systems hypoxia alters physiological and metabolic rate processes, organism abundance, lifestyles, composition, complexity, diversity, and size structure resulting in mortality of benthic (ocean bottom dwelling) fauna, fish kills, habitat loss, and overall physiological stress. In the future overall reduced biodiversity is expected, associated with avoidance, mortality, or lowered growth and reproductive rates of hypoxia-sensitive species.

Open ocean deoxygenation has been recorded in nearly all ocean basins during the second half of the 20th century. Though observational studies remain relatively sparse and localized, over recent decades a mostly negative trend can be seen in the oxygen content in several basins of the world’s ocean including Black and Baltic Seas, the Arabian Sea, and the California, Humboldt, and Benguela Current systems. A recent global-scale observational study supports the evidence of a widespread ocean O₂ decrease between the 1970s and the 1990s.

For this report we use simulations produced by the Coupled Model Intercomparison Project Phase 5 (CMIP 5). Ten models simulate future climate states and dynamics based on four possible climate scenarios, depending on greenhouse gas emissions in the years to come. Scenarios are referred to as RCPs (Representative Concentration Pathways) and describe simulations in which emissions peak at various periods. We concentrate on two scenarios: RCP 4.5 in which emissions peak around 2040, then decline, and RCP 8.5 in which emissions continue to rise throughout the 21st century.

CMIP5 projections show an overall decline in oceanic dissolved oxygen concentration of 2 to 4% in 2090s relative to the 1990s depending on the complexity of the model and global warming scenario chosen. All models support the prediction that further global warming will exacerbate hypoxia conditions, through reduced oxygen solubility in warmer water, enhanced upper ocean stratification and reduced winter ventilation of the water column in higher latitudes.

Model-mean time series of global ocean O₂ content (%) over 1870-2100 using historical simulations (black line) and four RCP scenarios

Colours represent RCP scenarios: RCP 2.6 – blue, RCP 4.5 – green, RCP 6.0 – lavender and RCP 8.5 – red. Values are plotted relative to 1990s mean.

By 2100, all models project an increase in the volume of hypoxic waters, from +1% to +9%. For suboxic waters, there is much less agreement: simulated volumes do not agree with observations, and changes due to climate change may be either negative (a 4% reduction) or positive (up to 30% growth). In the business as usual RCP 8.5 scenario the North Pacific, the North Atlantic, the Southern Ocean, the subtropical South Pacific and the South Indian oceans all undergo deoxygenation, up to −50 µmol per kg in the North Pacific. In contrast, the tropical Atlantic and the tropical Indian show increasing O₂ concentrations in both scenarios. The equatorial Pacific shows increasing O₂ in the east and decreasing O₂ in the west. Over the mid-latitudes, patterns of projected changes in subsurface O₂ are broadly consistent with observations collected over the past several decades. Yet there is no such model–data agreement over most of the tropical oceans.

Understanding the global trend remains a challenging task. Projected O₂ changes in the subsurface layer (200-600m) show a complex pattern with both increasing and decreasing trends, influenced by circulation, production, remineralization, and temperature changes. Projected changes in the total volume of hypoxic and suboxic waters remain relatively uncertain in the current models. Hypoxia results from complex interactions between physical and biogeochemical processes, which currently cannot be understood by measurements or models alone due to our limited observational capacity and several model limitations. Models are invaluable tools for studying system dynamics, generalizing discrete observations and predicting future states. Model analyses and predictions are therefore used in this assessment, however their limitations in simulating even today’s oxygen concentrations are not insignificant and therefore we present data-based assessment for matters un-resolvable by the current generation of earth system models.