1 Anthropogenic Decline in High-Latitude Ocean Carbonate by 2100 James C. Orr1, Victoria J. Fabry2, Olivier Aumont1*, Laurent Bopp1, Scott C. Doney3, Richard M. Feely4, Anand Gnanadesikan5, Nicolas Gruber6, Akio Ishida7, Fortunat Joos8, Robert M. Key9, Keith Lindsay10, Ernst Maier-Reimer11, Richard Matear12, Patrick Monfray1† , Anne Mouchet13, Raymond G. Najjar14, Gian-Kasper Plattner8, Keith B. Rodgers1*, Christopher L. Sabine4, Jorge L. Sarmiento9, Reiner Schlitzer15, Richard D. Slater9, Ian J. Totterdell16, Marie-France Weirig15, Yasuhiro Yamanaka7, & Andrew Yool16 1LSCE, UMR CEA-CNRS, CEA Saclay, F-91191 Gif-sur-Yvette, France 2Dept. of Biol. Sciences, Cal. State University San Marcos, San Marcos, CA 92096– 0001, USA 3WHOI, Woods Hole, MA 02543-1543, USA. 4PMEL/NOAA Seattle, WA 98115-6349, USA. 5NOAA/GFDL, Princeton, NJ 08542, USA. 6IGPP, UCLA, Los Angeles, CA 90095-4996, USA. 7Frontier Research Center for Global Change, Yokohama 236-0001, Japan. 8Climate and Environmental Physics, University of Bern, CH-3012 Bern, Switzerland. 9AOS Program, Princeton University, Princeton, NJ 08544-0710, USA 10NCAR, Boulder, CO 80307-3000, USA. 11Max Planck Institut für Meteorologie, D-20146 Hamburg, Germany. 2 12CSIRO Division of Marine Research, Hobart, TAS 7001, Australia. 13Astrophysics & Geophysics Institute, University of Liege, B-4000 Liege, Belgium. 14Dept. of Meteorology, Penn. State University, University Park, PA 16802-5013, USA. 15Alfred Wegener Institute for Polar and Marine Research, D-27515 Bremerhaven, Germany. 16Southampton Oceanography Centre, Southampton SO14 3ZH, UK. *Present address: LOCEAN, 4 Place Jussieu, F-75252 Paris cedex 05, France †Present address: LEGOS, UMR 5566 CNES-CNRS-IRD-UPS, F-31401 Toulouse, France. The surface ocean is everywhere saturated with respect to calcium carbonate (CaCO3). Yet increasing atmospheric CO2 reduces ocean pH and carbonate ion concentrations [CO32−] and thus the level of saturation. Reduced saturation states are expected to affect marine calcifiers even though it has been estimated that all surface waters will remain saturated for centuries. Here we show, however, that some surface waters will become undersaturated within decades. When atmospheric CO2 reaches 550 ppmv, in year 2050 under the IS92a business-as- usual scenario, Southern Ocean surface waters begin to become undersaturated with respect to aragonite, a metastable form of CaCO3. By 2100 as atmospheric CO2 reaches 788 ppmv, undersaturation extends throughout the entire Southern Ocean (< 60°S) and into the subarctic Pacific. These changes will threaten high- latitude aragonite secreting organisms including cold-water corals, which provide essential fish habitat, and shelled pteropods, an abundant food source for marine predators. 3 Ocean uptake of CO2 will help moderate future climate change but the associated chemistry, namely hydrolysis of CO2 in seawater, increases the hydrogen ion concentration [H+]. Surface ocean pH is already 0.1 unit lower than preindustrial values. By the end of the century it will probably become another 0.3 to 0.4 units lower1,2, meaning [H+] will increase by 100 to 150%. Simultaneously, the aqueous CO2 concentration [CO2(aq)] will increase and [CO32−] will decrease, making it harder for marine calcifying organisms to form biogenic CaCO3. Substantial experimental evidence indicates that calcification rates will decrease in low latitude corals3,-5, which form reefs out of aragonite, and in phytoplankton that form their tests (shells) out of calcite6,7, the stable form of CaCO3. Calcification rates will decline along with [CO32−] due to its reaction with increasing concentrations of anthropogenic CO2 CO2 + CO32− + H2O → 2HCO3− (1) even though surface waters remain supersaturated with respect to CaCO3, a condition that previous studies have predicted will persist for hundreds of years8,4,9. Recent predictions of future changes in surface ocean pH and carbonate chemistry have largely focused on global average conditions1,2,10 or on the low latitudes4, where reef building corals are abundant. Here we focus on future surface and subsurface changes in high latitude regions where planktonic shelled pteropods are prominent components of the upper ocean biota in the Southern Ocean, Arctic Ocean, and subarctic Pacific11-15. Recently, it has been suggested that the cold surface waters in such regions will begin to become undersaturated with respect to aragonite only when atmospheric CO2 reaches 1200 ppmv, more than 4 times the preindustrial level (4×CO2) of 280 ppmv9. In contrast, our results suggest that some polar and subpolar surface waters will become undersaturated at ∼ 2×CO2, i.e., probably within the next 50 years. Changes in Carbonate 4 We have computed modern ocean carbonate chemistry from observed alkalinity and dissolved inorganic carbon (DIC), relying on data collected during the CO2 Survey of the World Ocean Circulation Experiment (WOCE) and the Joint Global Ocean Flux Study (JGOFS). These observations are centred around 1994, and have recently been provided as a global-scale, gridded data product GLODAP16 (see Supplementary Information). Modern surface [CO32−] varies meridionally by more than a factor of two, from average Southern Ocean concentrations of 105 µmol kg−1 to average tropical concentrations of 240 µmol kg−1 (Fig. 1). Low Southern Ocean [CO32−] is due to (1) low surface temperatures and carbonate thermodynamics (high solubility) as well as (2) large local amounts of upwelled deep water, which contain high [CO2(aq)] from organic matter remineralisation. These two effects reinforce one another, yielding a high positive correlation of present-day [CO32−] with temperature (e.g., R2 = 0.92 for annual mean surface maps). Changes in [CO32−] and [CO2(aq)] are also inextricably linked to changes in other carbonate chemistry variables (Fig. S1). We also estimated preindustrial [CO32−] from the same data, after subtracting data-based estimates of anthropogenic DIC17 from the modern DIC observations and assuming that preindustrial and modern alkalinity fields were identical (see Supplementary Information). Relative to preindustrial conditions, invasion of anthropogenic CO2 has already reduced modern surface [CO32−] by more than 10%, i.e., by 29 µmol kg−1 in the tropics and 18 µmol kg−1 in the Southern Ocean. Nearly identical results were found when instead of the data-based anthropogenic CO2 estimates, we used simulated anthropogenic CO2, i.e., the median from 13 models that participated in the second phase of the Ocean Carbon-Cycle Model Intercomparison Project, OCMIP-2 (Fig. 1c). To quantify future changes in carbonate chemistry we used simulated DIC from ocean models that were forced by two atmospheric CO2 scenarios: the continually 5 increasing IPCC IS92a scenario (788 ppmv in 2100) and the IPCC S650 stabilisation scenario (563 ppmv in 2100) (Fig. 1). Simulated perturbations in DIC relative to 1994 (the GLODAP reference year), were added to the modern DIC data; again, alkalinity was assumed constant. To provide a measure of uncertainty we report model results as the OCMIP median ±2σ. The median generally outperformed individual models in OCMIP model-data comparison (Fig. S2). By year 2100 as atmospheric CO2 reaches 788 ppmv under the IS92a scenario, average tropical surface [CO32−] declines to 149 ±14 µmol kg−1. This is a 45% reduction relative to the preindustrial ocean, in agreement with previous predictions8,4. In the Southern Ocean, surface concentrations dip to 55 ±5 µmol kg−1, i.e., 18% below the threshold where aragonite becomes undersaturated (66 µmol kg−1). These changes extend well below the sea surface. Throughout the Southern Ocean (all waters south of 60°S), the entire water column becomes undersaturated with respect to aragonite. During the 21st century under IS92a, the Southern Ocean's aragonite saturation horizon (the limit between undersaturation and supersaturation) shoals from its present average depth of 730 m (Fig. S3) all the way to the surface (Fig. 2). Simultaneously, in a portion of the subarctic Pacific, the aragonite saturation horizon shoals from depths of about 120 m to the surface. In the North Atlantic, surface waters remain saturated with respect to aragonite, but the aragonite saturation horizon shoals dramatically, e.g., from 2600 m to 115 m north of 50°N. The greater erosion in the North Atlantic is due to deeper penetration and higher concentrations of anthropogenic CO2, a tendency that is already evident in present-day data-based estimates18,17 and in models19,20 (Figs. S4 and S5). Less spectacular changes were found for the calcite saturation horizon. For example, in 2100 the average calcite saturation horizon in the Southern Ocean stays below 2200 m. Nonetheless, in 2100 Weddell Sea surface waters become slightly undersaturated with respect to calcite. 6 In the more conservative S650 scenario, the atmosphere reaches 2×CO2 in 2100, 50 years later than with the IS92a scenario. In 2100, average Southern Ocean surface waters remain slightly supersaturated with respect to aragonite. But the models also simulate that the Southern Ocean's average aragonite saturation horizon will have shoaled from 730 m to 60 m and that the entire water column in the Weddell Sea will have become undersaturated (Fig. 2). In the north, all surface waters remain saturated under the S650 scenario. North of 50°N, the annual average, aragonite saturation horizon shoals from 140 m to 70 m in the Pacific, whereas it shoals by 2000 m to 610 m in the North Atlantic. Therefore under either scenario, the OCMIP models simulated large changes in surface and subsurface [CO32−]. Yet these models account for only the direct geochemical effect of increasing atmospheric CO2 because they were all forced with prescribed modern climate conditions. In addition to this direct geochemical effect, ocean [CO32−] is also altered by climate variability and climate change. To quantify the added effect of future climate change, we analysed results from three atmosphere-ocean climate models that each included an ocean carbon cycle component (see Supplementary Information). These three models agree that 21st century climate change will cause a general increase in surface ocean [CO32−] (Fig. 3), mainly because most surface waters will be warmer. Moreover, the models agree that the magnitude of this increase in [CO32−] is small, typically counteracting less than 10% of the decrease due to the geochemical effect. High-latitude surface waters show the smallest increases in [CO32−], even small reductions in some cases. Therefore, our analysis suggests that physical climate change alone will not substantially alter high-latitude surface [CO32−] during the 21st century. Climate also varies seasonally and interannually, whereas our previous focus has been on annual changes. To illustrate how climate variability affects surface [CO32−], we used results from an ocean carbon cycle model forced with the daily NCEP 7 reanalysis fields21 over 1948-2003 (see Supplementary Information). These fields are observationally based and vary on seasonal and interannual time scales. Simulated interannual variability in surface ocean [CO32−] is negligible when compared with the magnitude of the anthropogenic decline (Fig. 3b). Seasonal variability is also negligible except in the high latitudes, where surface [CO32−] varies by about ±15 µmol kg−1 when averaged over large regions. This is smaller than the 21st century's transient change (e.g., ∼ 50 µmol kg−1 in the Southern Ocean). However, high-latitude surface waters do become substantially less saturated during winter, because of cooling (higher [CO2(aq)]) and greater upwelling of DIC-enriched deep water, in agreement with previous observations in the North Pacific22. Thus high-latitude undersaturation will be first reached during winter. Our predicted changes may be compared to those found in earlier studies, which focused on surface waters in the tropics8 and in the subarctic Pacific23,22. These studies assumed thermodynamic equilibrium between CO2 in the atmosphere and surface waters at their in situ alkalinity, temperature, and salinity. If the pCO2 in the equilibrium approach is taken only to represent seawater pCO2, then results agree with our non- equilibrium approach, when the sets of carbonate chemistry constants are identical (Fig. 4). However, assuming equilibrium with the atmosphere leads to the prediction that future undersaturation will occur too soon (at lower atmospheric CO2), mainly because the anthropogenic transient in the ocean actually lags that in the atmosphere. For example, with the equilibrium approach we predict that average surface waters in the Southern Ocean become undersaturated when atmospheric pCO2 is 550 ppmv (year 2050 under IS92a), whereas our non-equilibrium approach that uses models and data indicates that such will occur at 635 ppmv (year 2070). Despite these differences, both approaches indicate that the Southern Ocean surface waters will probably become undersaturated with respect to aragonite during this century. Conversely, both these approaches disagree with a recent assessment9 that used a variant of the standard 8 thermodynamic equilibrium approach, where an incorrect input temperature was used inadvertently. Uncertainties The three coupled climate-carbon models show little effect of climate change on surface [CO32−] (Fig. 3a vs. Fig. 1) partly because air-sea CO2 exchange largely compensates for changes in surface DIC caused by changes in marine productivity and circulation. In subsurface waters where such compensation is lacking, these models could under- or over-predict the extent to which [CO32−] will change as a result of changes in overlying marine productivity (see Supplementary Information). However, we are confident in the model-predicted trend, which only worsens the decline in subsurface [CO32−]. That is, all coupled climate models predict a more vigorous hydrological cycle with increased evaporation in the tropics and increased precipitation in the high latitudes24. Increased high-latitude precipitation drives freshening of surface waters and thus greater vertical stratification. This leads to a decrease in high-latitude nutrients, but an increase in light availability. The latter wins out in the Southern Ocean of the IPSL-Paris model so that at 2×CO2 there is a 10% increase in local surface carbon export of particulate organic carbon (POC)25. Subsequent remineralisation of this exported POC within the thermocline leads to increased DIC, which only exacerbates the decrease in high- latitude subsurface [CO32−]. The largest uncertainty by far, and the only means to limit the future decline in ocean [CO32−], is the atmospheric CO2 trajectory. To better characterise uncertainty due to CO2 emissions, we compared the six illustrative IPCC SRES scenarios in the reduced complexity, PIUB-Bern model. Under the moderate SRES B2 scenario, average Southern Ocean surface waters in that model become undersaturated with respect to aragonite when atmospheric CO2 reaches 600 ppmv in 2100 (Fig. 5). For the three higher emission SRES scenarios (A1FI, A2, A1B), these waters become undersaturated 9 sooner (between 2058 and 2073); for the two lower emission scenarios (A1T, B1), these waters remain slightly supersaturated in 2100. Thus if atmospheric CO2 rises above 600 ppmv, most Southern Ocean surface waters will become undersaturated with respect to aragonite. Yet even below this level, the Southern Ocean's aragonite saturation horizon will shoal substantially (Fig. 2). For a given atmospheric CO2 scenario, predicted changes in surface ocean [CO32−] are much more certain than are related changes in climate. The latter depend not only on the model response to CO2 forcing but also on poorly constrained physical processes, such as those associated with clouds. Ocean CO2 Uptake With higher levels of anthropogenic CO2 and lower surface [CO32−], the change in surface ocean DIC per unit change in atmospheric CO2 (µmol kg−1 per ppmv) will be about 60% lower in 2100 (under IS92a) than it is today. Simultaneously, the CO32− /CO2(aq) ratio will decrease from 4:1 to 1:1 in the Southern Ocean (Fig. 4). These decreases are due to the well-understood anthropogenic reduction in buffer capacity26, already accounted for in ocean carbon cycle models. On the other hand, reduced export of CaCO3 from the high latitudes would increase surface [CO32−], thereby increasing ocean CO2 uptake and decreasing atmospheric CO2. Due to this effect, ocean CO2 uptake could increase by 6 to 13 Pg C over the 21st century, based on one recent model study27 that incorporated an empirical, CO2-dependant relationship for calcification7. Rates of calcification could decline even further, to zero, if waters actually became undersaturated with respect to both aragonite and calcite. We estimate that the total shutdown of high-latitude aragonite production would lead to at most a 0.25 Pg C yr−1 increase in ocean CO2 uptake, assuming that 1 Pg C yr−1 of CaCO3 is exported globally28, that up to half of that is aragonite29,9, and that perhaps half of all aragonite is exported from the high latitudes. The actual increase in ocean CO2 uptake could be much lower because the aragonite fraction of the CaCO3 10 may be only 0.1 based on low-latitude sediment traps30, and the latitudinal distribution of aragonite export is uncertain. Thus increased CO2 uptake from reduced export of aragonite will provide little compensation for decreases in ocean CO2 uptake due to reductions in buffer capacity. Of greater concern are potential biological impacts due to future undersaturation. Biological Impacts The changes in seawater chemistry that we project to occur during this century could have severe consequences for calcifying organisms, particularly shelled pteropods, the major planktonic producers of aragonite. Pteropod population densities are high in polar and subpolar waters. Yet only 5 species typically occur in such cold water regions and, of these, only 1-2 species are common at the highest latitudes31. High-latitude pteropods have 1-2 generations per year12,15,32, form integral components of food webs, and are typically found in the upper 300 m where they may reach densities of 100's to 1000's of individuals per m3 (refs. 11,13-15). In the Ross Sea, for example, the prominent subpolar-polar pteropod Limacina helicina sometimes replaces krill as the dominant zooplankton and is considered an overall indicator of ecosystem health33. In the strongly seasonal high latitudes, sedimentation pulses of pteropods frequently occur just after summer15,34. In the Ross Sea, pteropods account for the majority of the annual export flux of both carbonate and organic carbon35,34. South of the Antarctic Polar Front pteropods also dominate the export flux of CaCO336. Pteropods may be unable to maintain shells in waters that are undersaturated with respect to aragonite. Data from sediment traps indicate that empty pteropod shells exhibit pitting and partial dissolution as soon as they fall below the aragonite saturation horizon37,22,36. In vitro measurements confirm such rapid pteropod shell dissolution rates38. New experimental evidence suggests that even shells of live pteropods dissolve rapidly once surface waters become undersaturated with respect to aragonite9. Here we 11 show that when the live subarctic pteropod Clio pyramidata is subjected to undersaturation, similar to what we predict for Southern Ocean surface waters in 2100 under IS92a, marked dissolution occurs within 48 hours at the growing edge of the shell aperture (Fig. 6). Etch pits formed on the shell surface at the apertural margin, which is typically ∼ 7 µm thick, as the < 1-µm exterior (prismatic layer) peeled back (Fig. 6c), exposing the underlying aragonitic rods to dissolution. Fourteen individuals were tested. All of them showed similar dissolution along their growing edge, even though they all remained alive. If C. pyramidata cannot grow its protective shell, we would not expect it to survive in waters that become undersaturated with respect to aragonite. If the response of other high latitude pteropod species to aragonite undersaturation is similar to that of C. pyramidata, we hypothesise that these pteropods will not be able to adapt quickly enough to live in the undersaturated conditions that will occur over much of the high-latitude surface ocean during the 21st century. Their distributional ranges would then be reduced, both within the water column, disrupting vertical migration patterns, and latitudinally, imposing a shift towards lower latitude surface waters that remain supersaturated with respect to aragonite. At present, we do not know if pteropod species endemic to polar regions could disappear altogether or if they can make the transition to live in warmer, carbonate-rich waters at lower latitudes under a different ecosystem. If pteropods are excluded from polar and subpolar regions, their predators will be affected immediately. For instance, gymnosomes are zooplankton that feed exclusively on shelled pteropods39,33. Pteropods also contribute to the diet of diverse carnivorous zooplankton, mytophid and notothenoid fishes40-42, North Pacific salmon43,44, mackerel, herring, cod, and baleen whales45. Surface dwelling calcitic plankton such as foraminifera and coccolithophorids, may fare better temporarily. However, the beginnings of high-latitude calcite undersaturation will only lag that for aragonite by 50 to 100 years. The diverse benthic 12 calcareous organisms in high-latitude regions may also be threatened, including cold- water corals which provide essential fish habitat46. Cold-water corals appear much less abundant in the North Pacific than in the North Atlantic46, where the aragonite saturation horizon is much deeper (Fig. 2). Moreover, some important taxa in Arctic and Antarctic benthic communities secrete magnesian calcite, which can be more soluble than aragonite. These include gorgonians46, coralline red algae, and echinoderms (sea urchins)47. At 2×CO2, juvenile echinoderms stopped growing and produced more brittle and fragile exoskeletons in a subtropical 6-month manipulative experiment48. For high latitude calcifiers though, responses to reduced [CO32−] have generally not been studied. Yet experimental evidence from lower latitude, shallow dwelling calcifiers reveals a reduced ability to calcify with decreasing carbonate saturation state9. Given that at 2×CO2, calcification rates in some shallow dwelling calcareous organisms may decline by up to 50%9, some calcifiers could have difficulty surviving even long enough to experience undersaturation. Certainly, they have not experienced undersaturation for at least the last 400,000 years49 and probably much longer50. Changes in high-latitude seawater chemistry that will occur by the end of the century could well alter the structure and biodiversity of polar ecosystems, impacting multiple trophic levels. Assessing these impacts is impeded by the scarcity of relevant data. Figure 1: Increasing atmospheric CO2 and decreasing surface ocean pH and [CO32−]. a Atmospheric CO2 used to force 13 OCMIP models over the industrial period and for two future scenarios: IS92a (I) and S650 (S). Increases in atmospheric CO2 lead to reductions in b surface ocean pH and c surface ocean [CO32−] (µmol kg−1). Results are given as global zonal averages for the 1994 data and the preindustrial ocean. The latter were obtained by subtracting data- based anthropogenic DIC17 (solid) as well as by subtracting model-based 13 anthropogenic DIC (OCMIP median, dotted line; OCMIP range, grey shading). Future results come from the 1994 data plus the simulated DIC perturbations to 2100 for the two scenarios; results are also shown for 2300 with S650 (thick dashed line). The small effect of future climate change simulated by the IPSL climate-carbon model is added as a perturbation to IS92a in 2100 (thick dotted line); two other climate-carbon models (PIUB-Bern and CSIRO) show similar results (Fig. 3a). The thin dashed lines indicating the [CO32−] for seawater in equilibrium with aragonite and calcite are nearly flat, revealing weak temperature sensitivity. Figure 2: The aragonite saturation state in 2100 as indicated by ∆[CO32−]A. The ∆[CO32−]A is the in situ [CO32−] minus that for aragonite equilibrated seawater at the same salinity, temperature, and pressure. Shown are the OCMIP-2 median concentrations (µmol kg−1 ) in year 2100 under scenario IS92a: a surface map; b Atlantic and c Pacific zonal averages. Thick lines indicate the aragonite saturation horizon in 1765 (white dashes), 1994 (white solid), and 2100 (black solid for S650 [S]; black dashes for IS92a [I]). Positive ∆[CO32−]A indicates supersaturation; negative ∆[CO32−]A indicates undersaturation. Figure 3: Climate-induced changes in surface [CO32−]. a The 21st century shift in zonal mean surface ocean [CO32−] due only to climate change, from three atmosphere-ocean climate models (CSIRO-Hobart, IPSL-Paris, and PIUB-Bern) that each include an ocean carbon cycle component (see Supplementary Information). b The regional-scale seasonal and interannual variability as simulated by an ocean carbon cycle model forced with reanalysed climate forcing. Figure 4: Key surface carbonate chemistry variables as a function of pCO2 (µatm). Shown are both [CO32−] (solid) and [CO2(aq)] (dashed) for 14 average surface waters in the tropical ocean (thick), the Southern Ocean (thickest) and the global ocean (thin). Solid and dashed lines are calculated from the thermodynamic equilibrium approach. For comparison, open symbols are for [CO32−] from our non-equilibrium, model-data approach vs. seawater pCO2 (circles) and atmospheric pCO2 (squares); symbol thickness corresponds with line thickness, which indicates the regions for area-weighted averages. The nearly flat, thin dotted lines indicate the [CO32−] for seawater in equilibrium with aragonite and calcite. Figure 5: Time series of average surface [CO32−] in the Southern Ocean for the PIUB-Bern reduced complexity model (see Fig. 3 and Supplementary Information) under the six illustrative IPCC SRES scenarios. The results for the SRES scenarios A1T and A2 are similar to those for the non-SRES scenarios S650 and IS92a, respectively. Figure 6: Shell from a live pteropod, Clio pyramidata, collected from the subarctic Pacific and kept in water undersaturated with respect to aragonite for 48 hours. The a whole shell has superimposed white rectangles that indicate the three corresponding magnified areas: b shell surface, which reveals etch pits from dissolution and resulting exposure of aragonitic rods; c prismatic layer, which has begun to peel back, increasing the surface area over which dissolution occurs; and d aperture region, which reveals advanced shell dissolution when compared to e a typical C. pyramidata shell not exposed to undersaturated conditions. References 1. Haugan, P. M. & Drange, H. Effects of CO2 on the ocean environment. Energy Convers. Mgmt. 37, 1019-1022 (1996). 15 2. Brewer, P. G. Ocean chemistry of the fossil fuel CO2 signal: the haline signal of "business as usual". Geophys. Res. Lett. 24, 1367-1369 (1997). 3. Gattuso, J.-P., Frankignoulle, M., Bourge, I., Romaine, S. & Buddemeier, R. W. Effect of calcium carbonate saturation of seawater on coral calcification. Global and Planetary Change 18, 37-46 (1998). 4. Kleypas, J. A. et al. Geochemical consequences of increased atmospheric carbon dioxide on coral reefs. Science 284, 118-120 (1999). 5. Langdon, C. et al. Effect of elevated CO2 on the community metabolism of an experimental coral reef. Global Biogeochem. Cycles 17, 1011, doi:10.1029/2002GB001941 (2003). 6. Riebesell, U. et al. Reduced calcification of marine plankton in response to increased atmospheric CO2. Nature 407, 364-367 (2000). 7. Zondervan, I., Zeebe, R., Rost, B. & Riebesell, U. Decreasing marine biogenic calcification: A negative feedback on rising atmospheric pCO2. Global Biogeochem. Cycles 15, 507-516 (2001). 8. Broecker, W. S. & Peng, T.-H. Fate of fossil fuel carbon dioxide and the global carbon budget. Science 206, 409-418 (1979). 9. Feely, R. A. et al. The impact of anthropogenic CO2 on the CaCO3 system in the oceans. Science 305, 362-366 (2004). 10. Caldeira, K. & Wickett, M. E. Anthropogenic carbon and ocean pH. Science 425, 365 (2003). 11. Urban-Rich, J., Dagg, M. & Peterson, J. Copepod grazing on phytoplankton in the Pacific sector of the Antarctic Polar Front. Deep-Sea Res. 48, 4223-4246 (2001). 16 12. Kobayashi, H. A. Growth cycle and related vertical distribution of the thecosomatous pteropod Spiratella "Limacina" helicina in the central Arctic Ocean. Marine Biology 26, 295-301 (1974). 13. E. A. Pakhomov, H. M. V., Atkinson, A., Laubscher, R. K. & Taunton-Clark, J. Structure and grazing impact of the mesozooplankton community during late summer 1994 near South Georgia, Antarctica. Polar Biology 18, 180-192 (1997). 14. Fabry, V. J. Aragonite production by pteropod molluscs in the subarctic Pacifc. Deep-Sea Res. I 36, 1735-1751 (1989). 15. Bathmann, U., Noji, T. T. & Bodungen, B. Sedimentation in the Norwegian Sea in autumn. Deep-Sea Res. 38, 1341-1360 (1991). 16. Key, R. M. et al. A global ocean carbon climatology: Results from Global Data Analysis Project (GLODAP. Global Biogeochem. Cycles 18, GB4031, doi:10.1029/2004GB002247 (2004). 17. Sabine, C. L. et al. The ocean sink for anthropogenic CO2. Science 305, 367-370 (2004). 18. Gruber, N. Anthropogenic CO2 in the Atlantic Ocean. Global Biogeochem. Cycles 12, 165-191 (1998). 19. Sarmiento, J. L., Orr, J. C. & Siegenthaler, U. A perturbation simulation of CO2 uptake in an ocean general circulation model. J. Geophys. Res. 97, 3621-3645 (1992). 20. Orr, J. C. et al. Estimates of anthropogenic carbon uptake from four three- dimensionsal global ocean models. Global Biogeochem. Cycles 15, 43-60 (2001). 21. Kalnay, E. et al. The NCEP/NCAR 40-year reanalysis project. Bull. Am. Meteor. Soc. 77, 437-471 (1996). 22. Feely, R. A. et al. Winter-summer variations of calcite and aragonite saturation in the northeast pacific. Mar. Chem. 25, 227-241 (1988). 17 23. Feely, R. A., Byrne, R. H., Betzer, P. R., Gendron, J. F. & Acker, J. G. Factors influencing the degree of saturation of the surface and intermediate waters of the North Pacific Ocean with respect to aragonite. J. Geophys. Res. 89, 10,631-10,640 (1984). 24. Sarmiento, J. L. et al. Response of ocean ecosystems to climate warming. Global Biogeochem. Cycles 18, GB3003, doi:10.1029/2003GB002134 (2004). 25. Bopp, L. et al. Potential impact of climate change of marine export production. Global Biogeochem. Cycles 15, 81-99 (2001). 26. Sarmiento, J. L., Le Quéré, C. & Pacala, S. Limiting future atmospheric carbon dioxide. Global Biogeochem. Cycles 9, 121-137 (1995). 27. Heinze, C. Simulating oceanic CaCO3 export production in the greenhouse. Geophys. Res. Lett. 31, L16308, doi:10.1029/2004GL020613 (2004). 28. Iglesias-Rodriguez, M. D. et al. Representing key phytoplankton functional groups in ocean carbon cycle models: Coccolithophorids. Global Biogeochem. Cycles 16, 1100, doi: 10.1029/2001GB00145 (2002). 29. Berner, R. A. Sedimentation and dissolution of pteropods in the ocean. In Andersen, N. R. & Malahoff, A. (eds.) The Fate of Fossil Fuel CO2 in the Oceans, 243-260 (Plenum, New York, 1977). 30. Fabry, V. J. Shell growth rates of pteropod and heteropod molluscs and aragonite production in the open ocean: Implications for the marine carbonate system. J. Mar. Res. 48, 209-222 (1990). 31. Bé, A. W. H. & Gilmer, R. W. A zoogeographic and taxonomic review of euthecosomatous Pteropoda. In Ramsey, A. (ed.) Oceanic Micropaleontology, vol. 1, 733-808 (Academic Press, London, 1977). 18 32. Dadon, J. R. & de Cidre, L. L. The reproductive cycle of the Thecomosomatous pteropod Limacina retroversa in the western South Atlantic. Marine Biology 114, 439- 442 (1992). 33. Seibel, B. A. & Dierssen, H. M. Cascading trophic impacts of reduced biomass in the Ross Sea, Antarctica: Just the tip of the iceberg? Biol. Bull. 205, 93-97 (2003). 34. Accornero, A., Manno, C., Esposito, F. & M.C.Gambi. The vertical flux of particulate matter in the polyna of Terra Nova Bay. Part II. Biological components. Antarctic Science 15, 175-188 (2003). 35. Collier, R., Dymond, J., Susumu Honjo, S. M., Francois, R. & Dunbar, R. The vertical flux of biogenic and lithogenic material in the Ross Sea: moored sediment trap observations 1996-1998. Deep-Sea Res. II 47, 3491-3520 (2000). 36. Honjo, S., Francois, R., Manganini, S., Dymond, J. & Collier, R. Particle fluxes to the interior of the Southern Ocean in the western Pacific sector along 170°W. Deep-Sea Res. II 47, 3521-3548 (2000). 37. Betzer, P. R., Byrne, R., Acker, J., Lewis, C. S. & Jolley, R. R. The oceanic carbonate system: a reassessment of biogenic controls. Science 226, 1074-1077 (1984). 38. Byrne, R. H., Acker, J. G., Betzer, P. R., Feely, R. A. & Cates, M. H. Water column dissolution of aragonite in the Pacific Ocean. Nature 312, 321-326 (1984). 39. Lalli, C. M. Structure and function of the buccal apparatus of Clione limacina (Phipps) with a review of feeding in gymnosomatous pteropods. J. Exp. Mar. Biol. Ecol. 4, 101-118 (1970). 40. Foster, B. A. & Montgomery, J. C. Planktivory in benthic nototheniid fish in McMurdo Sound, Antarctica. Environ. Biol. Fish. 36, 313-318 (1993). 19 41. Pakhovmov, E., Perissinotto, A. & McQuaid, C. D. Prey composition and daily rations of myctophid fishes in the Southern Ocean. Marine Ecology Progress Series 134, 1-14 (1996). 42. La Mesa, M., Vacchi, M. & Sertorio, T. Z. Feeding plascity of Trematomus newnesi (Pisces, Nototheniidae) in Terra Nova Bay, Ross Sea, in relation to environmental conditions. Polar Biology 23, 38-45 (2000). 43. Willette, T. M. et al. Ecological processes influencing mortality of juvenile pink salmon (Oncorhynchus gorbuscha) in Prince William Sound, Alaska. Fisheries Oceanography 10, 14-41 (2001). 44. Boldt, J. L. & Haldorson, L. J. Seasonal and geographical variation in juvenile pink salmon diets in the Northern Gulf of Alaska and Prince William Sound. Transactions of the American Fisheries Society 132, 1035-1052 (2003). 45. Lalli, C. M. & Gilmer, R. Pelagic Snails (Stanford Univ. Press, Stanford, 1989). 46. Freiwald, A., Fosså, J. H., Grehan, A., Koslow, T. & Roberts, J. M. Cold-water coral reefs: Out of sight no longer out of mind. No. 22 in Biodiversity Series (UNEP- WCMC, Cambridge, UK, 2004). 47. Dayton, P. K. Polar benthos. In Smith, W. O. (ed.) Polar Oceanography, Part B, Chemistry, Biology, and Geology, 631-685 (Academic Press, San Diego, 1990). 48. Shirayama, Y. & Thornton, H. Effect of increased atmospheric CO2 on shallow- water marine benthos. J. Geophys. Res. in press (2005). 49. Petit, J. R. et al. Climate and atmospheric history of the past 420,000 years from the Vostok ice core, Antarctica. Nature 399, 429-436 (1999). 50. Pearson, P. N. & Palmer, M. R. Middle Eocene seawater pH and atmospheric carbon dioxide concentrations. Science 284, 1824-1826 (1999). 20 Supplementary Information accompanies the paper on www.nature.com/nature Acknowledgements We thank M. Gehlen for discussions and J.-M. Epitalon, P. Brockmann, and the Ferret developers for help with analysis. All but the climate simulations were made as part of the OCMIP project, which was launched in 1995 by the Global Analysis, Interpretation, and Modeling (GAIM) Task Force of the International Geosphere-Biosphere Program (IGBP) with funding from NASA. OCMIP-2 was supported by the EU GOSAC project and the U.S. JGOFS SMP funded through NASA. The interannual simulation was supported by the EU NOCES project, which is part of OCMIP-3. Competing Interests The authors declare that they have no competing financial interests. Correspondence Correspondence and requests for materials should be addressed to JCO (email: orr@cea.fr).