全球变化知识资源中心

The Southern Ocean (south of 40° S) ecosystem plays a fundamental role in global biogeochemical cycling through its effect on nutrient distributions and the air–sea balance of CO₂  (refs 5,6). Despite its remoteness, this region also hosts valuable krill and toothfish fisheries⁷. Recent trends in the Southern Ocean foodweb⁸, which can be linked partly to regional ocean warming and sea-ice retreat⁹, prompt the concern that further progression of anthropogenic stressors on sensitive marine organisms can have ripple effects far beyond the Southern Ocean. One key threat to Southern Ocean biota is the rapid progression of ocean acidification^{2, 10, 11}, which is caused by the uptake of anthropogenic CO₂. Along with a decreasing pH, the uptake of anthropogenic CO₂ decreases the CO₃²⁻ concentration and thereby the saturation state (Ω) of the CaCO₃ minerals aragonite (arag), calcite (calc) and magnesian calcite. These minerals chemically dissolve once Ω decreases below the well-established thermodynamic threshold of Ω = 1. Many marine calcifiers are sensitive to a decreasing Ω in the ocean and develop species-dependent responses already well above this thermodynamical threshold¹². For example, aragonite-forming organisms such as soft clams and pteropods exhibit a negative net calcification rate at Ω_arag ~ 1.5 (ref. 12) and close to 1 (ref. 2), respectively. In the following, we will refer to these species-dependent thresholds as biological thresholds.

Here, we use monthly output from ten Earth system models from the Coupled Model Intercomparison Project, Phase 5 (CMIP5, see Methods) to study the history and future development of Southern Ocean low-Ω_arag and -Ω_calc events. These CMIP5 models are the most advanced, global, state-of-the-art climate–carbon-cycle models. By using a multi-model ensemble we expect to add robustness to our analysis as potential shortcomings of individual models are of less consequence for the overall results¹³. Our analysis focuses mostly on aragonite undersaturation events, as aragonite is the most soluble CaCO₃ mineral and because Ω_arag = 1 closely lines up with the biological threshold of pteropods. Aragonite undersaturation in sea water can occur sporadically and naturally as a result of background variability; superimposed on this natural variability is the long-term ocean acidification trend^{14, 15}. The Southern Ocean is at particular risk of becoming undersaturated with respect to aragonite in the near future^{1, 11}, as thermodynamics and upwelling of CO₂-rich deep waters cause a naturally low Ω_arag environment¹⁶. The duration of these aragonite undersaturation events is an important indicator for the survival chances of organisms sensitive to these conditions, such as pteropods, as it quantifies how long these organisms will be exposed to lower calcification and increased dissolution rates, higher energetic cost, and suppressed metabolism, eventually leading to reduced growth and reproduction^{17, 18}. Furthermore, the rates at which the intensity and duration of aragonite undersaturation events change are crucial, as they may be faster than the evolutionary processes that could eventually lead to adaptation to low-Ω_arag habitats⁴. Compared with previous studies that were based on annual mean values of Ω_arag simulated by ocean-only models¹¹ or the extrapolation of relatively sparse measurements of Southern Ocean near-surface carbon-cycle parameters¹, monthly output from the ten CMIP5 coupled Earth system models employed in our study enables us to explore the spatial characteristics and temporal evolution of the habitat of pteropods and other sensitive organisms with unprecedented detail and robustness, and in the presence of natural climate variability and greenhouse warming.

Under the high-emissions Representative Concentration Pathway 8.5 (RCP8.5; ref. 19, see Methods), the ensemble mean of ten CMIP5 models documents that the duration and spatial extent of aragonite undersaturation events in the Southern Ocean will change rapidly over the next 40 years (Figs 1 and 2). According to the modelling results, regions in the Bellingshausen and Ross seas already experience sporadic short surface aragonite undersaturation events under present-day conditions (Fig. 1a). Although such short events are masked out in multi-year data products such as the Global Ocean Data Analysis Project (Supplementary Fig. 5a and reference in figure caption) as well as in the comparable ten-year ensemble means (Supplementary Fig. 5b), observations and salinity-based estimates of surface Ω_arag from these regions sometimes attain values near 1, thus supporting the model results^{20, 21}. The absence of simulated surface aragonite undersaturation events before 1965 indicates that their present occurrence may already be a result of the uptake of anthropogenic CO₂ (Supplementary Fig. 2a).

a–f, Decadal average of ensemble-mean duration of aragonite undersaturation events at the surface at present day (a) and around 2055 (b) and 2095 (c), and at 100 m at present day (d) and around 2055 (e) and 2095 (f). Duration was rounded to the closest integer. See Supplementary Fig. 2 for pre-industrial and additional years.

CMIP5 models and simulations.

Our analysis was based on the output of historical and Representative Concentration Pathway 8.5 (RCP8.5; ref. 19) simulations, conducted with ten Earth system models from the Coupled Model Intercomparison Project, Phase 5 (CMIP5), including CESM1-BGC, CanESM2, GFDL-ESM2G, GFDL-ESM2M, HadGEM2-ES, HadGEM2-CC, IPSL-CM5A-LR, IPSL-CM5A-MR, IPSL-CM5B-LR and MPI-ESM-LR (see Supplementary Table 1  for model details and references). All models are fully coupled climate–biogeochemical models that simulate interannual variability and future climate change. All models include an annual cycle and were forced with atmospheric CO₂ concentrations following RCP8.5 (ref. 19). Monthly outputs for the surface aragonite saturation state (Ω_arag) from 1860 to 2100 for each model were downloaded from the official CMIP5 data portal (https://pcmdi9.llnl.gov/search/cmip5). Individual CMIP5 model output was regridded to a 1° × 1° horizontal resolution and subsequently averaged to derive an ensemble mean.

Model evaluation.

Comparison of the model ensemble mean of Ω_arag for the period 1991–2000 with Ω_arag from the Global Ocean Data Analysis Project shows that the observed Southern Hemisphere large-scale pattern of meridionally decreasing Ω_arag is well represented by the ensemble mean (Supplementary Fig. 5 and references in figure caption). The spatial correlation coefficients between observed and simulated Southern Ocean Ω_arag (south of 40° S) range between 0.81 and 0.91 for the different models (Supplementary Fig. 6). All models simulate the amplitude of variations well, shown by the normalized standard deviation between 0.78 and 0.9 in the Taylor diagram. Cruise data from the Pacific sector of the Southern Ocean³³ suggest that the model mean underestimates Ω_arag slightly (~0.1 units) in this region (Supplementary Fig. 7a). The modelled trend of Ω_arag was also compared with available time series from other regions, including the Bermuda Atlantic Timeseries Station (BATS); the European Station for Timeseries in the Ocean, close to the Canary Islands (ESTOC); and the Hawaii Ocean Timeseries (HOT), which indicates that the ensemble mean follows the observed decreasing trend of Ω_arag closely (Supplementary Fig. 7 and references therein).

Duration of aragonite undersaturation events.

The duration of aragonite undersaturation events is defined as a consecutive sequence of months with surface waters undersaturated with respect to Ω_arag (Ω_arag < 1). Similarly, the duration of aragonite undersaturation events at 100 m depth is defined as a consecutive sequence of months with depth of the aragonite saturation horizon shallower than 100 m. A ten-year running mean duration of surface aragonite undersaturation events per year and model was calculated to avoid the influence of internal model variability. The duration of aragonite undersaturation events was then averaged across each ensemble member and rounded to the closest integer. The duration for other Ω_arag thresholds and for calcite was obtained accordingly.

Model agreement on first occurrence of aragonite undersaturation events.

A ten-year running mean of the multi-model ensemble mean suggests that surface aragonite undersaturation events first occurred in the Bellingshausen and Ross seas around 1965 (Supplementary Fig. 1a). Even though there are no observations from the 1960s to evaluate these results, recent data show Ω_arag around 1 in these regions^{20, 21}. However, IPSL-CM5A-MR and IPSL-CM5A-LR already simulate sporadic aragonite undersaturation events in coastal areas off Antarctica and CanESM2 off the coast of central Chile at the beginning of the twentieth century (Supplementary Fig. 1b), suggesting a large uncertainty in the timing of the first anthropogenic aragonite undersaturation events within the multi-model ensemble. All models agree that coastal waters around Antarctica will experience surface undersaturation before 2060 (Supplementary Fig. 1c).

Spatial inhomogeneity.

To elucidate whether the temporal progression of the duration and spatial spread of aragonite undersaturation events is driven mainly by the natural latitudinal [CO₃²⁻] distribution and superimposed anthropogenic CO₂ trend, or whether small-scale regional ocean circulation features associated with zonal inhomogeneities also have an influence, a fictitious data set of Ω_arag (Ω^fic) for each model using the following equation: Ω_(x,y,t)^fic = Ω_(x,y)^prein + _zmδΩ_(t,y) was generated. Ω^prein represents a ten-year climatological mean of pre-industrial Ω_arag conditions (1861–1870) and _zmδΩ is the zonal mean temporal trend of Ω_arag for the period from 1900 to 2099 (Fig. 2d). The duration of aragonite undersaturation events per model and as an ensemble mean were then calculated on the basis of Ω^fic. Finally, the duration of aragonite undersaturation events due to spatial inhomogeniety was defined as Dⁱⁿ = D(Ω_(x,y,z)) − D(Ω_(x,y)^prein + _zmδΩ_(t,y)).

1. McNeil, B. I. & Matear, R. J. Southern Ocean acidification: A tipping point at 450-ppm atmospheric CO₂. Proc. Natl Acad. Sci. USA 105, 18860–18864 (2008).
   • PubMed
   • Article
2. Bednaršek, N. et al. Extensive dissolution of live pteropods in the Southern Ocean. Nature Geosci. 5, 881–885 (2012).
   • CAS
   • Article
3. Hunt, B. P. V. et al. Pteropods in Southern Ocean ecosystems. Prog. Oceanogr. 78, 193–221 (2008).
   • Article
4. Sunday, J. M. et al. Evolution in an acidifying ocean. Trends Ecol. Evol. 29, 117–125 (2014).
   • ISI
   • PubMed
   • Article
5. Sarmiento, J. L., Gruber, N., Brzezinski, M. A. & Dunne, J. P. High-latitude controls of thermocline nutrients and low latitude biological productivity. Nature 427, 56–60 (2004).
   • CAS
   • ISI
   • PubMed
   • Article
   URL: