Sensitivities of marine carbon fluxes to ocean changeUlf Riebesell1, Arne Kortzinger, and Andreas OschliesMarine Biogeochemistry, Leibniz Institute of Marine Sciences, IFM-GEOMAR, Dusternbrooker Weg 20,24105 Kiel, Germany

Edited by Hans Joachim Schellnhuber, Potsdam Institute for Climate Impact Research, Potsdam, Germany, and approved August 31, 2009(received for review December 29, 2008)

Throughout Earth’s history, the oceans have played a dominant role in the climate system through the storage and transport of heat andthe exchange of water and climate-relevant gases with the atmosphere. The ocean’s heat capacity is �1,000 times larger than that of theatmosphere, its content of reactive carbon more than 60 times larger. Through a variety of physical, chemical, and biological processes, theocean acts as a driver of climate variability on time scales ranging from seasonal to interannual to decadal to glacial–interglacial. The sameprocesses will also be involved in future responses of the ocean to global change. Here we assess the responses of the seawater carbon-ate system and of the ocean’s physical and biological carbon pumps to (i) ocean warming and the associated changes in vertical mixingand overturning circulation, and (ii) ocean acidification and carbonation. Our analysis underscores that many of these responses have thepotential for significant feedback to the climate system. Because several of the underlying processes are interlinked and nonlinear, thesign and magnitude of the ocean’s carbon cycle feedback to climate change is yet unknown. Understanding these processes and their sen-sitivities to global change will be crucial to our ability to project future climate change.

climate change � marine carbon cycle � ocean acidification � ocean warming

The ocean is presently undergoingmajor changes. Over the past 50years, the ocean has stored 20times more heat than the atmo-

sphere (1). The Arctic Ocean surface lay-ers have recently experienced a pro-nounced freshening (2), and althoughthere is still considerable uncertainty incurrent observational estimates (3), cli-mate models run under increasing atmo-spheric CO2 levels essentially all predict aslowdown of the Atlantic meridional over-turning circulation (MOC), which is partof the global thermohaline circulation (4).Since preindustrial times, the ocean hasalso taken up �50% of fossil fuel CO2,which has already led to substantialchanges in its chemical properties (5).

Carbon fluxes from the sea surface intothe ocean interior are often described interms of a solubility pump and a biologi-cal pump (6). The abiotic solubility pumpis caused by the solubility of CO2 increas-ing with decreasing temperature. Inpresent climate conditions, deep waterforms at high latitudes. As a result,volume-averaged ocean temperatures arelower than average sea-surface tempera-tures. The solubility pump then ensuresthat, associated with the mean verticaltemperature gradient, there is a verticalgradient of dissolved inorganic carbon(DIC). This solubility-driven gradient ex-plains �30–40% of today’s ocean surface-to-depth DIC gradient (7).

A key process responsible for the re-maining two thirds of the surface-to-depthDIC gradient is the biological carbonpump. It transports photosyntheticallyfixed organic carbon from the sunlit sur-face layer to the deep ocean. Integratedover the global ocean, the biotically medi-ated oceanic surface-to-depth DIC gradi-ent corresponds to a carbon pool 3.5times larger than the total amount of at-

mospheric carbon dioxide (8) and has amean residence of a few hundred years.Hence, small changes in this pool, caused,for example, by biological responses toocean change, would have a strong effecton atmospheric CO2. Counteracting theorganic carbon pump in terms of its effecton air–sea CO2 exchange is a processtermed the carbonate counter pump (9),also known as the alkalinity pump. Theformation of CaCO3 shell material by cal-cifying plankton and its sinking to depthlowers the DIC and alkalinity in the sur-face ocean, causing an increase in CO2partial pressure. It is worth noting that theorganic and inorganic carbon pumps rein-force each other in terms of maintaining avertical DIC gradient, whereas they arecounteractive with respect to their impacton air–sea CO2 exchange.

Although the range of potentialchanges in the solubility pump and chemi-cal responses of the marine CO2 system isknown reasonably well, our understandingof biological responses to ocean change isstill in its infancy. Such responses relateboth to possible direct effects of risingatmospheric CO2 through ocean acidifica-tion (decreasing seawater pH) and oceancarbonation (increasing CO2 concentra-tion), and indirect effects through oceanwarming and changes in circulation andmixing regimes. These changes are ex-pected to impact marine ecosystem struc-ture and functioning and have the poten-tial to alter the cycling of carbon andnutrients in the surface ocean with likelyfeedbacks on the climate system.

Changes in the Solubility Pump. Alterationsin the physical state of the ocean will af-fect both the solubility pump and the bio-logical pump. For finite-amplitude pertur-bations, changes in both pumps interactnonlinearly and therefore cannot be con-

sidered separately. For conceptual simpli-fication, however, we will here focus onphysical impacts on the solubility pumponly before discussing impacts on the bio-logical pump in a later section.

Rising atmospheric CO2 leads to highersea-surface temperature (SST) and a con-current reduction in CO2 solubility. It isestimated that this positive SST feedbackwill reduce the oceanic uptake of anthro-pogenic carbon by 9–15% [�45–70 giga-tons carbon (Gt C)] by the end of the 21stcentury (10–13). Global warming will alsointensify the hydrological cycle. The directdilution effect on CO2 solubility is smallcompared with the temperature effect andis expected to largely cancel out at theglobal scale as enhanced precipitation inone area tends to be balanced by en-hanced evaporation in another area. Over-all, the expected effect of direct dilutionon anthropogenic CO2 uptake is of uncer-tain sign but estimated to be much smallerthan 1% (13). A potentially more signifi-cant impact of changes in the hydrologicalcycle on the oceanic CO2 uptake can ariseat high latitudes in the North Atlantic:Here, reduced surface salinities, togetherwith higher SSTs, would lower the densityof surface waters and thereby may inhibitthe formation of deep waters. This lower-ing in turn would reduce meridional pres-sure gradients and tend to slow down thethermohaline-driven part of the meridi-onal, overturning circulation. Climatemodel simulations indeed predict a gen-

Author contributions: U.R., A.K., and A.O. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

1To whom correspondence should be addressed. E-mail:uriebesell@ifm-geomar.de.

This article contains supporting information online at www.pnas.org/cgi/content/full/0813291106/DCSupplemental.

20602–20609 � PNAS � December 8, 2009 � vol. 106 � no. 49 www.pnas.org�cgi�doi�10.1073�pnas.0813291106

Dow

nloa

ded

by g

uest

on

July

16,

202

0

eral weakening of the North AtlanticMOC for the 21st century when forcedwith increasing greenhouse gas concentra-tions (4, 14). Complementary to the simu-lated weakening of the Atlantic MOC,climate models also predict an intensifica-tion of the ocean circulation in the South-ern Ocean, resulting from a larger warm-ing in mid and low latitudes comparedwith the waters around Antarctica. Inconsequence, the meridional pressure gra-dients are predicted to increase bothacross the Antarctic Circumpolar Current(15) and in the atmosphere above theSouthern Ocean (16). Estimates of thetotal impact of the circulation changes onthe solubility pump differ considerablyamong different models, predicting a re-duction in global oceanic carbon uptakeby some 3–20% (17). A major part of thelarge uncertainty can be attributed to ourincomplete understanding of the SouthernOcean’s role as a carbon sink.

Chemical Response of the Marine CO2 Sys-tem. A significant portion of the carbonreleased by humankind in the Anthropo-cene era has been—and will continue tobe—redistributed to the world ocean. Theocean owes its huge CO2-uptake capacityto the presence of the carbonate ion(CO3

2�), which can react with excess CO2taken up from the atmosphere accordingto the following net buffering reaction(Note that the two analytically indistin-guishable species, CO2 and H2CO3, areusually combined as hypothetical speciesCO*2.):

CO*2 � CO32��H2O3 2HCO3

� [1]

The oceanic uptake capacity whichamounts to nearly 85% (not including thebuffering by sedimentary carbonates) ofthe entire anthropogenic carbon releasedto date (340–420 Gt C) (5), can only beaccommodated, however, on the turnovertime scale of the ocean, i.e., several centu-ries to millennia. The present oceanic up-take of �27% of the current anthropo-genic carbon emissions (5) therefore fallsmarkedly short of the thermodynamic up-take capacity. As the ocean continues totake up anthropogenic CO2, significantspeciation changes in the marine CO2 sys-tem take place that lead to a progressivereduction of the buffering capacity. Thesechanges will not only cause a decrease inthe ocean’s uptake ratio, i.e., the equilib-rium change in DIC per change in atmo-spheric CO2 concentration, but also leadto an amplification of the effects of ther-mal and biological forcing on the naturaloceanic carbon cycle. These changes inthe marine CO2 system, which we willexplore in the following, are inherentlynonlinear and will amplify with time.

To illustrate changes in the propertiesof the marine CO2 system during thecourse of the 21st century, we need toprescribe the development of atmosphericCO2 concentrations. With vastly differingemission scenarios and a host of openquestions, we opted for a rather simplisticapproach in which we extrapolated recentgrowth rates to yield an atmospheric CO2concentration of almost 700 parts per mil-lion by volume (ppmv) by the end of the21st century (see SI Text for details on thecalculations and all assumptions involved).The oceanic response exerted by this at-mospheric boundary condition will be re-gionally different depending mainly onseawater temperature and the generalcharacteristics of the marine CO2 system.To capture the entire sensitivity range, wedefine a cold- and warm-water caseroughly representing polar and tropicalsurface waters. In both cases, changes inCO2 system parameters pCO2, pH, andDIC as well as concurrent changes in CO2system speciation (HCO3

�, CO32�, CO*2)

were calculated (Fig. 1 A–C). These spe-ciation shifts also affect other CO2 systemproperties such as the CO2 buffering ca-pacity expressed by the uptake factor(�DIC/�xCO2

atm), the saturation state withrespect to the calcium carbonate mineralscalcite and aragonite as well as the sensi-tivities of pCO2 and pH to thermal andbiologically mediated chemical forcing(Fig. 1 D–E).

As expected, surface-ocean DIC in-creases along with pCO2 as a consequenceof uptake of ‘‘excess’’ atmospheric CO2(Fig. 1B), a process known as ‘‘ocean car-bonation.’’ The concurrent drop in sur-face-ocean pH is a mirror image of thepCO2 increase and amounts to 0.3–0.4 pH

units (Fig. 1A). This increase in the hy-drogen ion (H�) concentration of up to250%, also referred to as ‘‘ocean acidifica-tion,’’ is currently of growing concernamong marine biologists. Also note thatit is entirely unclear whether the corre-sponding decrease of the hydroxide ionconcentration by up to 60% could alsohave consequences for chemical reactionsor equilibria involving OH� (e.g., solubil-ity of iron hydroxides); this aspect war-rants a closer look. According to Eq. 1,the uptake of atmospheric CO2 in addi-tion to CO2

� increases HCO3� at the ex-

pense of CO32� (Fig. 1C). By converting

one molecule of CO32� to two molecules

of HCO3� this reaction is neutral, how-

ever, for total alkalinity (18).The marked drop in CO3

2� has distincteffects on two important properties of themarine CO2 system: First, the CO2 uptakecapacity, expressed as the equilibriumDIC increase per rise of atmospheric CO2(�DIC/�xCO2

atm), is attenuated to less thana third of its preindustrial value. Hence, atequilibrium one liter of seawater in 2100will sequester only one third of the atmo-spheric CO2 (per ppmv increase) it se-questered in 1750 (Fig. 1D). This is arather drastic change that must be (andusually is) taken into account in futureprojections. Secondly, the ubiquitous su-persaturation of surface waters with re-spect to the calcium carbonate mineralscalcite and aragonite falls to about a half(50–60%, Fig. 1D) of its preindustriallevel. Thus, cold surface waters in highlatitudes will be the first to no longer besupersaturated and may even turn corro-sive to carbonate minerals (19, 20).

Beyond these prominent and well-documented effects, there are other

Fig. 1. Trends and projections of changes in the properties of the marine CO2 system in cold (blue) and warm(red) surface waters between 1750 and 2100 under the prescribed atmospheric CO2 increase. (A) EquilibriumCO2 partial pressure (pCO2) and pH. (B) Equilibrium DIC and TA. (C) Concentrations of the three species CO2

�,HCO3

�, and CO32� (note that the concentration of CO2

� is multiplied by a factor of 10. (D) CO2-uptake ratio andsaturation states (�) of calcite and aragonite. (E) Temperature sensitivities of pCO2 and pH. (F) Chemical (i.e.,CO2) sensitivities of pCO2 and pH. See SI Text for details on the calculation and assumptions made therein.

Riebesell et al. PNAS � December 8, 2009 � vol. 106 � no. 49 � 20603

SP

EC

IAL

FE

AT

UR

E:

PE

RS

PE

CT

IVE

Dow

nloa

ded

by g

uest

on

July

16,

202

0

changes in the properties of the marineCO2 system that do not appear to havereceived as much attention. These changespertain to the inherent sensitivity of theCO2 system to thermal and chemical forc-ing, most importantly the impacts of theseasonal cycle of SST and net biogenicproduction of particulate organic and in-organic carbon on the ocean’s source/sinkpatterns for atmospheric CO2. Large re-gions of the world ocean show markedSST seasonality, which directly affects thetemperature-sensitive parameters pCO2and pH [but not the conservative proper-ties DIC and total alkalinity (TA), ifexpressed in gravimetric units]. As theanthropogenic perturbation of the marineCO2 system enhances the temperaturesensitivity of pCO2 (�pCO2/�T) by almosta factor of 2.5 in our extrapolation (Fig.1E), the effect of the natural seasonalSST forcing will be strongly amplified,leading to marked modulation of the sea-sonal air–sea CO2 flux pattern. In con-trast, seawater pH will only be marginallyaffected with even a slight reduction inthe temperature sensitivity (�pH/�T).

The second major driver of natural sea-sonal variability in the surface-ocean car-bon cycle are biologically driven chemicalchanges, i.e., withdrawal of carbonthrough net production of organic carbonas well as particulate carbonates (e.g., cal-cite and aragonite). The sensitivities ofpCO2 and pH with respect to organic car-bon production, which to a first-order ap-proximation affects DIC but not TA, areboth amplified by the anthropogenic per-turbation. The enhancement by a factor�3 by year 2100 for pCO2 (�pCO2/�DIC)is again much larger than for pH (�pH/�DIC), which will increase by a factor ofonly �1.4 (Fig. 1F). Overall, pCO2 is thusaffected most strongly by changes in phys-ical and (biologically mediated) chemicalforcing, which has implications for theseasonal CO2 sink/source pattern. Theseawater pH, in contrast, is much lessaffected.

The thermal and biochemical drivers ofthe natural seasonal CO2 sink/source pat-terns of the ocean have been explored indetail (e.g., 21–24). Unlike for oxygen, thetwo have counteracting effects on pCO2.In major parts of the world ocean, how-ever, one factor dominates, and a clearseasonal pCO2 cycle results. Especially inthese regions, changing thermal andchemical sensitivities will give rise to analteration of the seasonal cycle of pCO2with higher peak-to-peak amplitudes andpotentially also a change in the sign of thedisequilibrium. To make the superim-posed effects of these counteracting pro-cesses more palpable, we illustrate thesituation with a realistic example. Wehave chosen the Labrador Sea as a high-latitude example where the seasonal pCO2

cycle features a marked summer minimum(winter maximum) because of the domi-nance of biological over thermal forcing(Fig. 2A).

As shown by ref. 23, the observed an-nual pCO2 cycle in the central LabradorSea can be deconvoluted into a thermal(i.e., isochemical) and a chemical (i.e.,isothermal) component, where most ofthe chemical effect is mediated by carbonuptake via net biological production. As-suming no change in the thermal andchemical forcing, hypothetical seasonalpCO2 and pH cycles can be constructedfor the year 2100 (Fig. 2). The increase inthe peak-to-peak amplitudes is significantin both but much stronger in pCO2. Notethat although in year 2100 pH is lowerand shows a larger seasonal amplitudethan in 2004, saturation states of calciteand aragonite show a reduced seasonalamplitude even though they are alsolower. At first sight, the pCO2 effects mayappear to be a mere intensification of theseasonal pCO2 and hence CO2 source/sinkcycle with little net affect. In regionswhere on an annual average a disequilib-rium is observed, however, a significantnet effect may result. The Labrador Sea,for example, is nearly neutral in winterand a strong sink in summer (Fig. 3) withan overall annual CO2 sink of 2.7 molm�2 yr�1 in 2004 (23). In our extrapola-tion, the summer sink is enhanced

whereas the region turns into a CO2source in winter. If we further take intoaccount the marked seasonality in windspeed, the physical driving force of air–seaexchange, it becomes obvious that strongnet effects on the annual source/sink func-tion can occur (Fig. 3). In our case, theoverall net CO2 sink in 2100 is projectedto drop to about half the sink observed2004.

The well-documented climatologicalCO2 source/sink patterns of the presentocean (21, 22) are largely a consequenceof the two big, natural, counteractingforces, both of which are enhanced assurface-ocean pCO2 levels rise in con-cert with the atmospheric pCO2. Asshown above, the balance between thetwo responses—and hence the netocean-source/-sink function for the at-mospheric CO2 budget—will be af-fected, even without the need to invokeany changes in the general physical andbiological forcing of the marine carboncycle. If we further include the possibil-ity of such changes in the forcing itself,which may partly compensate and partlyenhance the responses presented above,it becomes clear that the reduction inthe buffering capacity of the ocean’sCO2 system is a critical system responsewith significant feedback potential.

Biotic Responses to Ocean Warming. Oceanwarming impacts the pelagic ecosystem

-200

-150

-100

-50

0

50

100

150

200

Sea

sona

l pC

O2

anom

aly

[µat

m] Observed: Körtzinger et al. (2008)

Observed (smoothed)Thermal forcingChemical forcing

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

-0.12

-0.10

-0.08

-0.06

-0.04

-0.02

0.00

0.02

0.04

0.06

0.08

0.10

0.12

Sea

sona

l pH

ano

mal

y

ObservedThermal forcingChemical forcing

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

A

B

Fig. 2. Seasonal cycle of pCO2 (A) and pH (B) anomalies at present (year 2004, continuous lines) and inthe future (year 2100, dashed lines) at a location in the central Labrador Sea (56.5° N, 52.6° W). Also shownare pCO2 observations (from 20) on which this analysis is based. See SI Text for details on the calculationand assumptions made therein.

20604 � www.pnas.org�cgi�doi�10.1073�pnas.0813291106 Riebesell et al.

Dow

nloa

ded

by g

uest

on

July

16,

202

0

both directly and indirectly in three ways:(i) through decreased supply of plant nu-trients because of a slowdown in verticalmixing and convective overturning; (ii)through increased thermal stratification,causing increasing light availability forphotosynthetic organisms suspended inthe upper mixed layer; and (iii) throughthe effect of increasing temperatureson the rates of biological processes. Whichof these factors will dominate in a particu-lar region and at a given time depends onthe hydrographic conditions, the composi-tion of the pelagic community, and theactivities of its individual components.

Changes in Circulation and Mixing. A re-duction in the overturning circulation canbe expected to reduce the supply of nutri-ents to the surface waters. In areas inwhich surface nutrient concentrations arealready zero, this reduction will notchange surface nutrient concentrationsand, for constant stoichiometry, surfaceDIC levels. Despite a possible reductionin biomass and its turnover, the first-ordereffect on air–sea CO2 exchange will besmall in these areas. Regions with non-zero surface nutrient concentrations, how-ever, can experience declining surfacenutrient and DIC concentrations for aslower circulation, which will allow moreatmospheric CO2 to enter the ocean.Without exchange with the underlyingwaters, utilization of the nutrients in theupper 50 m of the global ocean wouldcorrespond to a one-time additional car-bon uptake of �8 Gt C. Subduction ofthe newly nutrient-depleted waters andreplacement by nutrient-rich subsurfacewaters can, however, allow for a repeateduptake of atmospheric carbon. This pro-cess eventually leads to a reduction of pre-formed nutrients and a corresponding in-crease of regenerated nutrients andcarbon below the surface layer at the ex-pense of a decline in the atmospheric car-bon pool (25). Particularly in the SouthernOcean, the dynamics of converting pre-

formed nutrients into regenerated nutri-ents will be modulated by the concomi-tant reduction in the supply of the oftenlimiting micronutrient iron (26) underchanging light conditions (27).

Increased thermal stratification due torising SST affects both nutrient supplyand mixed-layer light intensities. In thetropics and midlatitudes, where thermalstratification restricts vertical mixing, typi-cally low surface nutrient concentrationslimit phytoplankton growth. Ocean warm-ing will further reduce mixing, diminishingthe upward nutrient supply and loweringproductivity (Fig. 4 Upper). At higher lati-tudes, phytoplankton is often light-limitedbecause intense vertical mixing circulatesalgal cells over deep mixed layers, result-ing in lower mean light intensity along aphytoplankton cell’s trajectory and hencelower net productivity. In these regions,ocean warming and a greater influx offresh water, mostly from increased precip-itation and melting sea ice, will contributeto reduce vertical mixing which may in-crease productivity (Fig. 4 Lower).

Temperature Effects on Biological Activities.Direct biotic responses to increasing tem-peratures will differ greatly among thevarious components of the pelagic ecosys-tem, most notably between the autotro-phic and heterotrophic communities. Thetemperature dependence of biologicalprocesses is commonly expressed by theQ10 factor—the factorial increase in therate for a 10° C increase in temperature.Although bacterial heterotrophic activitiestypically have a Q10 factor between 2 and3 (28), phytoplankton growth and photo-synthesis show only a moderate tempera-ture sensitivity (1�Q10�2) (29) and areprimarily controlled by incident light in-tensity. Bacterial growth efficiency, on theother hand, is an inverse function of tem-perature, causing an increased fraction ofthe assimilated carbon to be respired withrising temperature (30). These divergingtemperature sensitivities will likely induce

a complex nonlinear response to oceanwarming at the community and ecosystemlevels. Although shoaling of the uppermixed layer primarily affects the autotro-phic community through increased lightintensities in this layer, increasing seawa-ter temperature mainly affects the activityof the heterotrophic community. The netoutcome depends on the balance of lightand temperature sensitivities and dynamicinteraction of the different groups ofauto- and heterotrophic organisms andcannot be easily deduced from physiology-based first principles.

A number of different approaches con-sistently predict an overall decrease inprimary and export production in thetropics and midlatitudes and a pole-wardmigration of geographic boundaries sepa-rating biogeochemical provinces:

(i) coupled atmosphere ocean generalcirculation models (AOGCMs) combinedwith mechanistic models of pelagic ecosys-tems that directly predict biological re-sponses (e.g., 31–33);

(ii) climate response patterns in theAOGCMs that are similar to observedmodes of interannual variability such asthe El Nino/Southern Oscillation or Pa-cific Decadal Oscillation (34, 35); and

(iii) empirical models based on observa-tional estimates of physical controls oncurrent distributions of chlorophyll andprimary production (36).

Biotic Responses to Ocean Acidification(OA). Oceanic uptake of anthropogenicCO2 and the resulting changes in seawaterchemistry will, in the course of this cen-

-200

-150

-100

-50

0

50

100

Air-

sea

pCO

2 [µ

atm

]

6

7

8

9

10

11

12

13

Win

d sp

eed

[m s

-1]

Present (2004)Future (2100)Wind speed

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Fig. 3. Monthly mean air–sea disequilibrium at a location in the central Labrador Sea (56.5° N, 52.6° W)in years 2004 (continuous line) and 2100 (broken line). Also shown is monthly mean climatological windspeed at this location.

Fig. 4. Schematic model illustrating the effect ofsea-surface warming on upper-ocean processes inlow (Upper) and high (Lower) latitudes. Graph de-picts effects on stratification and mixed-layerdepths, nutrient supply (yellow arrows), planktonbiomass, and particle flux (green arrows).

Riebesell et al. PNAS � December 8, 2009 � vol. 106 � no. 49 � 20605

Dow

nloa

ded

by g

uest

on

July

16,

202

0

tury, expose marine organisms to condi-tions which they may not have experi-enced during their recent evolutionaryhistory (37). The impact these changeswill have on marine life is still uncertain,but there are likely to be both winnersand losers. In fact, with regard to the po-tential biological effects of oceanic CO2uptake, the term ‘‘ocean acidification’’only encompasses one side of the story.Although the decrease in seawater pHmay pose a threat to the fitness of pH-sensitive groups, in particular calcifyingorganisms, ‘‘ocean carbonation’’—the in-creasing oceanic CO2 concentration—willlikely be beneficial to some groups ofphotosynthetic organisms, particularlythose that operate a relatively inefficientCO2 acquisition pathway.

Adverse effects of ocean acidificationare expected on various groups of calcify-ing organisms, including corals, bivalves,gastropods, and sea urchins, and possiblyon the many other groups that use CaCO3as internal or external structural elements,such as crustaceans, cnidaria, sponges,bryozoa, annelids, brachiopods, tunicates,squid, and fish (38). The loss of competi-tive fitness in these groups may lead tothe loss of biodiversity and the restructur-ing of marine ecosystems. To what extentsuch changes would affect marine produc-tivity, transfer of energy through the foodweb, or biogeochemical cycling is pres-ently unknown. Of relevance to the car-bonate pump are the pelagic calcifyinggroups, in particular the coccolithophores,foraminifera, and pteropods. Most studiesto date indicate a decrease in calcificationof these groups with increasing seawateracidification, both in laboratory experi-ments (e.g., 39–42) and field studies (43,44). However, this point remains contro-versial (45, 46).

Stimulating effects of ocean carbon-ation have been primarily observed forprocesses related to photosynthetic activ-ity. These effects include CO2-enhancedrates of phytoplankton growth and carbonfixation (47, 48), organic matter produc-tion (49, 50), and extracellular organicmatter production (51). CO2 sensitivity ofautotrophic processes is also evident fromchanges in the cellular carbon:nitrogen:phosphorus (C:N:P) ratios of marine mi-croalgae with CO2 concentration (52). Ashift in the ratio of carbon to nutrientdrawdown toward higher C:N and C:P atelevated pCO2 was observed during a me-socosm study in a natural plankton com-munity (53). Recent studies with the dia-zotrophic cyanobacterium Trichodesmiumrevealed an increase in both carbon andnitrogen fixation with increasing pCO2(54–56). As primary producers in the ma-rine realm encompass very phylogeneti-cally diverse groups of organisms (57) thatdiffer widely in their photosynthetic appa-

ratus and carbon-enrichment systems (58),it is presently difficult to assess which ofthese responses are specific to individualtaxa or represent general phenomenaacross phytoplankton taxa or within func-tional groups.

Uncertainties. It is important to note thatour present knowledge of pH/CO2 sensi-tivities of marine organisms is based al-most entirely on short-term perturbationexperiments, neglecting the possibility ofevolutionary adaptation. Also, little ispresently known regarding the effectsfrom multiple and interacting stressors,such as sea-surface warming and eutrophi-cation. Moreover, there is a complete lackof information on the transfer of re-sponses from the organism to the commu-nity and ecosystem levels and the replace-ment of OA-sensitive by OA-tolerantspecies.

Impacts on the Biological Carbon Pumps.For a comparison of the different biologi-cal responses to ocean change and theirpotential feedback to the climate system,we will apply four criteria to characterizeeach of the processes:

(i) the sign of the feedback, denotingwhether a feedback amplifies (positivefeedback) or dampens (negative feedback)the initial forcing;

(ii) the sensitivity, indicating the changeneeded to trigger the response;

(iii) the capacity, representing thefeedback strength relative to the forc-ing and relative to other feedbackmechanisms; and

(iv) the longevity, referring to thelength of time a feedback may operate(i.e., transient versus long-lasting).

In categories ii, iii, and iv, we use � 0for negligible, � for low, �� for moder-ate, and ��� for high values (Table 1).We wish to stress that our present under-

standing of biologically driven feedbackmechanisms is still rudimentary, so theassigned values mostly represent bestguesses. In some cases, we feel that ourunderstanding is too immature to evenmake a guess.

Responses to Ocean Warming and Circula-tion. The effects of sea-surface warmingon the marine biota and the resulting im-pacts on marine carbon cycling will differdepending on the prevailing light and nu-trient conditions in the upper mixed layerand the composition of the pelagiccommunity.� Decreased nutrient supply to already-nutrient-limited areas of the ocean’s sur-face layer is certain to lead to dimin-ished primary production (sensitivity���) and to a decline in the strengthof the biological pump, i.e., the amountof biogenic carbon transported to depth.On the other hand, as nutrient supply tothe surface layer slows down, so doesthe supply of dissolved inorganic carbon,slowing down the rate at which deepocean CO2 is brought back into contactwith the atmosphere. Because thechange in preformed nutrients is negligi-ble in nutrient-limited areas, the overalleffect on carbon sequestration can beexpected to be negligible (capacity �0)unless there is a change in C:N ratios inexport production.� Decreased nutrient supply to nutri-ent-replete (e.g., high-nutrient, low-chlorophyll) areas, may lead to lowerlevels in surface nutrients and DIC andin consequence reduce the totalamount of preformed nutrients. Associ-ated with these changes is an increasein regenerated nutrients and carbon inthe ocean interior (sensitivity ��),resulting in a negative feedback ofmoderate capacity (��), which canlast for several hundred years as long

Table 1. Biotic responses to sea surface warming and ocean carbonation/acidificationand their feedback potential to the climate system

Feedback process Sign of feedback Sensitivity Capacity Longevity

Responses to ocean warmingNutrient supply to oligotrophic ocean ��� �0 ��

Nutrient supply to HNLC areas negative �� �� ���

Nutrient utilization efficiency negative �� �� ���

Nutrient inventory positive �� ���

Organic matter remineralisation positiveResponses to ocean acidification

Calcification negative �† � �‡

Ballast effect positive ��� ���

Stoichiometry negative ��† �� ��

Extracellular organic matter production negative ��† ��� ‡

Nitrogen fixation negative ��† � ‡

Responses are characterized with regard to feedback sign, sensitivity, capacity, and longevity by usingbest guesses; we use �0 for negligible, � for low, �� for moderate, and ��� for high. Empty boxesindicate missing information/ understanding.†Available information mainly based on short-term perturbation experiments.‡Potential for adaptation presently unknown. HNLC, high nutrient, low chlorophyll.

20606 � www.pnas.org�cgi�doi�10.1073�pnas.0813291106 Riebesell et al.

Dow

nloa

ded

by g

uest

on

July

16,

202

0

as the supply of old preformed nutri-ents to the surface ocean from belowexceeds the subduction of new pre-formed nutrients (longevity ���).� Increased nutrient-utilization efficiencyin the high latitudes, on the other hand,has a potentially strong effect on the effi-ciency of the biological pump (36) (capac-ity ��). With the shoaling of mixed-layerdepths, currently nonused inorganic nutri-ents in high-latitude surface waters can becombined with CO2 to form organic mat-ter and be transported to depth (negativefeedback). This change in efficiency af-fects the partitioning of the nutrient poolbetween preformed and regenerated nu-trients and the associated partitioning ofcarbon among the atmospheric and oce-anic carbon reservoirs (25). In the deeplymixed high-latitude areas of the northernNorth Atlantic and Southern Ocean, asignificant shoaling of the mixed-layerdepth is required to have a noticeable ef-fect on nutrient utilization efficiency (sen-sitivity ��). Enhanced surface layer strat-ification in midlatitudes may also speed upthe biomass accumulation during seasonsof high productivity, e.g., during spring-bloom development, possibly affectingparticle aggregation and sinking and thedepth of remineralization. Provided thatpreformed nutrients continue to be sup-plied to the surface layer, this feedbackcould last for several hundred years (lon-gevity ���).� As sea-surface warming reduces deep-ocean ventilation, this slowdown will lowerthe supply of oxygen to the ocean interior(59). This process, which has been termed‘‘ocean deoxygenation’’ (60), is expectedto cause an overall decrease in deep-ocean oxygen content, including an expan-sion of oxygen-minimum zones (61, 62).Suboxic and anoxic conditions favor pro-cesses such as denitrification and anaero-bic ammonium oxidation, leading to theloss of bioavailable nitrogen in the ocean,with possible implications for marine pri-mary production. Provided that theocean’s nitrogen inventory ultimately de-termines the amount of carbon biologi-cally sequestered in the ocean, reducingthe nitrogen inventory would provide apositive feedback to the climate system.Suboxic conditions at the seafloor, on theother hand, favor the release of phosphatefrom sedimentary iron phosphates, possi-bly increasing the ocean’s content of bio-available phosphorus. A shift in theocean’s nitrate-to-phosphate contentcould affect the composition and produc-tivity of marine primary producers. Mostnotably, the extra phosphate may be usedby diazotrophic cyanobacteria, possiblycounterbalancing the nitrogen loss fromdenitrification by nitrogen fixation. Thenet outcome of these complex interactionsis difficult to assess, as is the sensitivity to

ocean warming. (Capacity ��, longevity���.)� Although sea-surface warming directlyaffects all biological rates, the higher tem-perature sensitivity of heterotrophic (rela-tive to autotrophic) processes is expectedto shift the balance between primary pro-duction and respiration/remineralization infavor of the latter (63). Additionally,warming shifted the partitioning of or-ganic carbon between the particulate anddissolved phase toward an enhanced accu-mulation of dissolved organic carbon (63).This shift may cause a decrease in thevertical flux of particulate organic matterand a decrease in remineralization depth.The overall effect could be a decline inthe strength and efficiency of the carbonpump (positive feedback). Because of thecomplexity of plankton trophic interac-tions it is difficult, however, to assess thesensitivity and capacity of this feedback.Also unknown is the extent to whichchanges in community composition maycompensate for the shifting balance be-tween auto- and heterotrophic processes,making it impossible to make a judgementon the longevity of this feedback.

Ocean Acidification/Carbonation. Our un-derstanding of the biological responses toCO2-induced changes in seawater carbon-ate is still in its infancy. As research onocean acidification gains momentum, new,unforeseen pH/CO2 sensitivities emerge,with partly opposing impacts on carbonsequestration in the ocean.� Decreased pelagic calcification and theresulting decline in the strength of thecarbonate pump lower the drawdown ofalkalinity in the surface layer, thereby in-creasing the uptake capacity for atmo-spheric CO2 in the surface layer. Assum-ing an annual rate of CaCO3 export of �1Gt C (64), the capacity of this negativefeedback is relatively small (65–68) (ca-pacity �). The sensitivity of this responsein coccolithophores appears to be highlyspecies-dependent (69), permittingchanges in species composition to dampenor eliminate the response at the commu-nity level (sensitivity �). Shifting speciescomposition and the potential for adapta-tion (presently unknown) could make thisa transient feedback (longevity �). Inview of the scarcity of information on pHsensitivities of foraminifera and ptero-pods, however, it is premature to specu-late on the significance of this feedbackprocess.� CaCO3 may act as ballast in particleaggregates, accelerating the flux of par-ticulate material to depth (70, 71; butsee also ref. 72). Reduced CaCO3 pro-duction could therefore slow down thevertical f lux of biogenic matter to depth,shoaling the remineralization depth oforganic carbon and decreasing carbon

sequestration (65) (positive feedback).The capacity of this feedback dependson the pH sensitivity of pelagic calcifiersand the quantitative importance ofCaCO3 as ballast for particle export,both of which are poorly understood. IfCaCO3 turns out to be a prerequisite fordeep transport of particulate organicmatter, and if calcification of pelagiccalcifiers remains sensitive to high CO2,this feedback could have a potentiallyhigh capacity (���) and extended du-ration (���).� A negative feedback also arises with in-creasing C:N drawdown, as observed inresponse to CO2 enrichment in a meso-cosm experiment (53). The capacity forthis process depends on the relative in-crease in carbon consumption in excess ofthe Redfield ratio, which based on theobserved 27% increase in response to adoubling in present day CO2 can be con-sidered moderate to high (capacity ��).It is not presently known whether the re-sponse observed for the mesocosm phyto-plankton community, which was domi-nated by diatoms (mostly Skeletonema sp.and Nitzschia spp.) and the coccolith-ophore Emiliania huxleyi, also applies toother phytoplankton taxa and functionalgroups, or whether this response may bemodified or lost because of adaptation tohigh CO2. Hence, it appears too early tomake a judgement on the sensitivity andlongevity of this feedback. If there is asignificant enhancement in the C:N draw-down at the surface that translates into anenhanced carbon export, it will lead tohigher oxygen consumption and a possibleexpansion of oxygen-minimum zones (73),linking to the oxygen-related feedbackdescribed in Responses to Ocean Warmingand Acidification.� Increased production of extracellularorganic matter under high CO2 levels(51) may enhance the formation of par-ticle aggregates (74, 75) and therebyincrease the vertical f lux of organic mat-ter (negative feedback) (76). This re-sponse may in fact have been responsi-ble for the observed increased in C:N:Pstoichiometry observed at elevatedpCO2 (53, 77). Provided that the re-sponse is of general nature in bloom-forming phytoplankton (presently un-known sensitivity), the capacity of thisprocess may be high because, similar tothe ballast effect, it could lead to anincrease in remineralization depth (ca-pacity ���). Again, not knowingwhether evolutionary adaptation willselect against this response prevents ajudgement on the longevity of thisprocess.� Enhanced nitrogen fixation at elevatedpCO2, as recently reported for the dia-zotrophic cyanobacterium Trichodes-mium (54–56), has the potential to in-

Riebesell et al. PNAS � December 8, 2009 � vol. 106 � no. 49 � 20607

Dow

nloa

ded

by g

uest

on

July

16,

202

0

crease the reservoir of bioavailablenitrogen in the surface layer, which in apredominantly nitrogen-limited oceanwould increase primary production andcarbon fixation (negative feedback). Be-cause cyanobacteria-produced biomass isgenerally thought to not be exported togreat depth, uncertainty exists about theextent to which the additional bioavail-able nitrogen would lead to enhancedcarbon sequestration or whether itwould remain suspended in the surfacelayer, eventually eliminating the ecologi-cal niche for nitrogen-fixing cyanobacte-ria. As CO2 sensitivity of nitrogen fixa-tion has thus far been reported for onlyone species, it is too early to speculateabout the sensitivity and longevity ofthis response.

Tipping Points. Although several of thecause-and-effect relationships describedabove have the potential to drastically al-ter the ocean’s ecosystems and biogeo-chemistry, it is yet unclear if any of thepredicted changes may turn out irrevers-ible or may develop into runaway biogeo-chemical feedbacks. In the following, wewill give four examples wherein sensitivi-ties of ecological and biogeochemical pro-cesses to ocean change may develop intotipping points, possibly driving the oceanto a new state.

(i) A decrease of nutrient supplyto the surface layer because of sea-surface warming in presently produc-tive areas of the ocean may stimulate adominance shift in the phytoplanktoncommunity from large diatoms tosmall-celled f lagellates and cyanobacte-ria. A shift to smaller-celled phyto-plankton causes an extension of themarine food web by one or more tro-phic levels (78). With 90% of the bio-mass and energy lost from one trophiclevel to the next, this shift would dras-tically reduce the efficiency of energytransfer from the level of primary pro-ducers to the top predators. Accelerat-ing heterotrophic over autotrophic pro-cesses in a warming ocean will furtherreduce the transfer efficiency of prima-ry-produced organic matter up thefood chain. Moreover, as the growthefficiency of bacteria is an inversefunction of temperature, a larger frac-tion of assimilated carbon is respiredat elevated temperatures (30, 63). Asthese responses all contribute to accel-erating the spinning of the wheel atthe lower trophic levels, a warmer andmore stratified ocean may shift into anew and lower state of energy-transferefficiency, involving reduced fish pro-duction.

(ii) Depending on the still incom-pletely understood role of CaCO3 asmineral ballast for organic carbon ex-

port, ocean acidification might, via areduction in CaCO3 formation, lead toa substantial decrease of the efficiencyof the biological pump. This decreasewould lead to a change in upper-oceannutrient and oxygen status (68) withlikely impacts on the pelagic commu-nity structure. The reduction in exportwould also present a strong positivefeedback on atmospheric CO2 (68)and, in turn, on ocean acidification(65). However, the postulated increasein the production of extracellular or-ganic carbon under high CO2 levelsmay partly compensate for the loss ofCaCO3 ballast by favoring aggregationof organic matter and non-CaCO3 min-erals (72). Given the current state ofour knowledge, the efficiency of thebiological pump could change signifi-cantly in either the positive or negativedirection.

(iii) With regard to ocean acidification,a distinct tipping point may arise whenseawater turns undersaturated with re-spect to calcium carbonate and becomescorrosive for the shells and skeletons ofcalcifying organisms. In the SouthernOcean, for example, wintertime under-saturation for aragonite is projected tooccur at CO2 levels of 450 ppmv, whichunder business-as-usual (IntergovernmenalPanel on Climate Change IS92a scenario)CO2 emissions will occur by 2030 and nolater than 2038 (79). Wintertime satura-tion states are of particular relevance toone of the key species of polar ecosys-tems, the aragonite-producing pteropodLimacina helicina. As this species under-goes its larval development primarily dur-ing winter months, it may experience cor-rosive seawater conditions at a particularlysensitive phase of its life cycle within thenext couple of decades. As Limacina is animportant component in polar ocean foodwebs, linking lower levels of the food webto the top predators, its disappearancecould mark the disruption of polar pelagicecosystems.

(iv) For coral reef ecosystems, a tippingpoint will be reached when reef erosionexceeds reef accretion. At CO2 levels of560 ppmv, calcification of tropical corals isexpected to decline by 30% (80, 81). Atthis stage, loss of coral structure in areasof high erosion may outpace coral calcifi-cation, in which case reefs will no longerbe sustainable. As the carbonate satura-tion horizon—the depth below which cal-cium carbonate dissolves—shallows be-cause of ocean acidification, cold-watercorals become exposed to corrosive wa-ters. These slow-growing corals, whichinhabit cold waters down to 3,000-m waterdepth, have built extensive reef systems onthe shelves and along the continental mar-gins extending from Northern Norway tothe west coast of Africa. With unabated

CO2 emissions, 70% of the presentlyknown reef locations will be in corrosivewaters by the end of this century (82).Coral reefs provide the habitat for theocean’s most diverse ecosystems, are thebreeding grounds for commercially impor-tant fish, protect shorelines in tropicalareas from erosion and flooding, and gen-erate billions of dollars annually intourism.

Summary. The global carbon cycle is bothdriven by and a driver of Earth’s climatesystem. In this system, climate changetherefore goes hand in hand with achange in carbon cycling and a redistribu-tion of reactive carbon among the carbonreservoirs in the atmosphere, terrestrialbiosphere, and ocean. The distribution ofcarbon between these reservoirs is theresult of a multitude of interconnectedphysical, chemical, and biological pro-cesses, many of which are sensitive to cli-mate change themselves. A major chal-lenge in Earth system science isdetermining which of these processes actas primary drivers in the natural climatecycle and how they interact in controllingcarbon fluxes between the reservoirs.

The same challenge applies when tryingto forecast the system’s response to majorperturbations of the carbon cycle, such asthe release of CO2 from fossil-fuel burn-ing and land-use changes. At first sight,the ocean’s role in this system appears tobe that of a giant buffer, sequesteringenormous amounts of CO2 and therebydampening CO2-induced climate change.A closer look reveals that both the chang-ing climate and the extra load of CO2 se-questered by the ocean alter the oceaniccarbon cycle to the extent of modifying itscapacity for further uptake of anthropo-genic CO2

. The net outcome of thesemodifications is still uncertain, largely dueto our limited understanding of the under-lying mechanisms. This uncertainty holdstrue particularly for those processes in-volving biologically mediated components,which because of the complexity and plas-ticity of biotic responses and interactionsare extremely difficult to untangle.Progress in our understanding of theseinteracting processes and their sensitivitiesto ocean change requires the concertedeffort of all relevant disciplines, from mo-lecular and ecosystem biology,marine and atmospheric chemistry, physi-cal oceanography and palaeoceanography,to atmosphere–ocean- and Earth-systemmodeling, and with close interactions be-tween observationalists, experimentalists,and modelers.

ACKNOWLEDGMENTS. This work was supportedby the European Union Integrated Project Carbo-Ocean “Marine carbon sources and sinks assess-ment” contract no. 511176 (GOCE).

20608 � www.pnas.org�cgi�doi�10.1073�pnas.0813291106 Riebesell et al.

Dow

nloa

ded

by g

uest

on

July

16,

202

0

1. Levitus S, Antonov J, Boyer T (2005) Warming of theworld ocean, 1995–2003. Geophys Res Lett 32:L02060410.1029/2004GL021592.

2. Curry R, Mauritzen C (2005) Dilution of the northernNorth Atlantic Ocean in recent decades. Science308:1772–1774.

3. Cunningham SA, et al. (2007) Temporal variability ofthe Atlantic meridional overturning circulation at26.5° N. Science 317:935–938.

4. Gregory JM, et al. (2005) A model intercomparison ofchanges in the Atlantic thermohaline circulation inresponse to increasing atmospheric CO2 concentration.Geophys Res Lett 32: L12703.

5. Sabine CL, et al. (2004) The Oceanic sink for Anthropo-genic CO2. Science 305:367–371.

6. Volk T, Hoffert MI (1985) in The Carbon Cycle andAtmospheric CO2: Natural Variations Archean toPresent, Geophysical Monogr Ser, eds Sunquist ET,Broecker WS (Washington, DC), Vol 32, pp 99–111.

7. Toggweiler JR, Gnanadesikan A, Carson S, Murnane R,Sarmiento JL (2003) Representation of the carbon cyclein box models and GCMs: 1. Solubility pump. GlobalBiogeochem Cycles 17:1026, 10.1029/2001GB001401.

8. Gruber N, Sarmiento JL (2002) in The Sea: Ideas andObservations on Progress in the Study of the Seas, edsRobinson AR, McCarthy JJ, Rothschild BJ (Wiley, NewYork), vol. 12, pp 337–399.

9. Heinze C, Maier–Reimer E, Winn K (1991) Glacial pCO2reduction by the world ocean: Experiments with theHamburg carbon cycle model. Paleoceanography6:395–430.

10. Sarmiento JL, Le Quere C (1996) Oceanic carbon dioxideuptake in a model of century-scale global warming.Science 274:1346–1350.

11. Joos F, Plattner G-K, Stocker TF, Marchal O, Schmittner A(1999) Global warming and marine carbon cycle feed-backs on future atmospheric CO2. Science 284:464–467.

12. Matear RJ, Hirst AC (1999) Climate change feedback onthe future oceanic CO2 uptake. Tellus 51B:722–733.

13. Plattner GF, Joos F, Stocker TF, Marchal O (2001) Feed-back mechanisms and sensitivities of ocean carbon up-take under global warming. Tellus 53B:564–592.

14. Schmittner A, Latif M, Schneider B(2005) Model projec-tions of the North Atlantic thermohaline circulation forthe 21st century assessed by observations. Geophys ResLett 32:L23710, 10.1029/2005GL024368.

15. Schmittner A, Oschlies A, Matthews HD, Galbraith ED(2008) Future changes in climate, ocean circulation,ecosystems and biogeochemical cycling simulated for abusiness-as-usual CO2 emission scenario until 4000 AD.Global Biogeochem Cycles GB1013: doi:10.1029/2007GB002953.

16. Russell JL, Dixon KW, Gnanadesikan A, Stouffer RJ,Toggweiler JR (2006) The southern hemisphere west-erlies in a warming world: Propping open the door tothe deep ocean. J Clim 19:6382–6390.

17. Greenblatt JB, Sarmiento JL(2004) in The Global Car-bon Cycle, eds Field CB, Raupach MR (Island Press,Washington, D.C.), pp 257–275.

18. Wolf–Gladrow DA, Zeebe RE, Klaas C, Kortzinger A,Dickson AG (2007) Total alkalinity: The explicit conser-vative expression and its application to biogeochemicalprocesses. Mar Chem 106:287–300.

19. Orr JC, et al. (2005) Anthropogenic ocean acidificationover the twenty-first century and its impact on calcify-ing organisms. Nature 437:681–686.

20. Steinacher M, Joos F, Frolicher TL, Plattner GK, DoneySC (2009) Imminent ocean acidification in the Arcticprojected with the NCAR global coupled carbon cycle-climate model. Biogeosciences 6:515–533.

21. Takahashi T, et al. (2002) Global air-sea CO2 flux basedon climatological surface ocean pCO2, and seasonalbiological and temperature effects. Deep-Sea Res II49:1601–1622.

22. Takahashi T, et al. (2009) Climatological mean anddecadal change in surface ocean pCO2, and net sea-airCO2 flux over the global oceans. Deep-Sea Res II56:554–577.

23. Kortzinger A, Send U, Wallace DWR, Karstensen J, De-Grandpre M(2008). The seasonal cycle of O2 and pCO2in the central Labrador Sea: Atmospheric, biologicaland physical implications. Global Biogeochem Cycles22:GB1014, 10.1029/2007GB003029.

24. Kortzinger A, et al. (2008). The seasonal pCO2 cycle at49° N/16.5° W in the northeast Atlantic Ocean andwhat it tells us about biological productivity. J GeophysRes 113:C04020, 10.1029/2007JC004347.

25. Ito T, Follows M (2005) Preformed phosphate, soft tissuepump and atmospheric CO2. J Mar Res 63:813–839.

26. Brzezinski MA, et al. (2002), A switch from SI(OH)4 toNO3 depletion in the global Southern Ocean, GeophysRes Lett 29:1264, 10.1029/2001GL014349.

27. de Baar, HJ, et al. (2005) Synthesis of iron fertilizationexperiments: From the iron age in the age of Enlight-enment, J Geophys Res 110, C09S16, 10.1029/2004JC002601.

28. Pomeroy LR, Wiebe WJ (2001) Temperature and sub-strates as interactive limiting factors for marine het-erotrophic bacteria. Aquat Microb Ecol 23:187–204.

29. Eppley RW (1972) Temperature and phytoplanktongrowth in the sea. Fish Bull 70:1063–1085.

30. Rivkin RB, Legendre L (2001) Biogenic carbon cycling inthe Upper Ocean: Effects of microbial respiration. Sci-ence 291:2398–2400.

31. Cox PM, Betts RA, Jones CD, Spall SA, Totterdell IJ (2000)Acceleration of global warming due to carbon cyclefeedbacks in a 3D coupled model. Nature 408:184–187.

32. Bopp L, et al. (2001) Potential impact of climate changeon marine export production. Global Biogeochem Cy-cles 15:81–100.

33. Christian JR, Verschell MA, Murtugudde R, BusalacchiAJ, McClain CR (2002) Biogeochemical modelling of thetropical Pacific Ocean: I. Seasonal and interannual vari-ability. Deep-Sea Res II 49:509–543.

34. Boyd PW, Doney SC(2002) Modelling regional responsesby marine pelagic ecosystems to global climate change.Geophys Res Lett 29:1806, 10.1029/2001GL014130.

35. Behrenfeld MJ, et al. (2006) Biospheric primary produc-tion during an ENSO Transition. Nature 444:752–755.

36. Sarmiento JL, et al. (2004) Response of ocean ecosys-tems to climate warming. Global Biogeochem Cycles18:GB3003, 10.1029/2003GB002134.

37. Raven J, et al. (2005) Ocean acidification due to increas-ing atmospheric carbon dioxide. (Royal Society, Lon-don) Royal Society Report, Policy Document 12/05.

38. Fabry VJ, Seibel BA, Richard AF, Orr JC (2008) Impacts ofocean acidification on marine fauna and ecosystemprocesses. ICES J Mar Science, Journal du Conseil 2008,65:414–432

39. Bijma, JH, Spero J, Lea DW, Bemis BE(1999) in Use ofProxies in Paleoceanography: Examples from the SouthAtlantic, eds Fischer G, Wefer, G. (Springer, Berlin), pp489–512.

40. Riebesell U, et al. (2000) Reduced calcification in marineplankton in response to increased atmospheric CO2.Nature 407:634–637.

41. Sciandra A, et al. (2003) Response of coccolithophoridEmiliania huxleyi to elevated partial pressure of CO2 un-der nitrogen limitation. Mar Ecol Progr Ser 261:111–122.

42. Feng Y, et al. (2008) Interactive effects of increasedpCO2, temperature and irradiance on the marine coc-colithophore Emiliania huxleyi (Prymnesiophyceae).Eur J Phycol 43:87–98.

43. Delille B, et al. (2005) Response of primary productionand calcification to changes of pCO2 during experimen-tal blooms of the coccolithophorid Emiliania huxleyi.Global Biogeochem Cycles 19:GB2023, 10.1029/2004GB002318.

44. Engel A, et al. (2005) Testing the direct effect of CO2concentration on a bloom of the coccolithophoridEmiliania huxleyi in mesocosm experiments. LimnolOceanogr 50:493–504.

45. Iglesias-Rodriguez MD, et al. (2008) Phytoplankton cal-cification in a high-CO2 world. Science 320:336–340.

46. Riebesell U, et al. (2008) Phytoplankton calcification ina high CO2 world. (technical comment) Science 322,10.1126/science.1161096.

47. Riebesell U, Wolf-Gladrow DA, Smetacek V (1993) Car-bon dioxide limitation of marine phytoplanktongrowth rates. Nature 361:249–251.

48. Hein M, Sand-Jensen K (1997) CO2 increases oceanicprimary production. Nature 388:526–527.

49. Zondervan I, Rost B, Riebesell U (2002) Effect of CO2concentration on the PIC/POC ratio in the coccolith-ophore Emiliania huxleyi grown under light-limitingconditions and different daylengths. J Exp Mar BiolEcol 272:55–70.

50. Leornardos N, Geider RJ (2005) Elevated atmosphericCO2 increases organic carbon fixation by Emiliania hux-leyi (Haptophyta) under nutrient-limited, high-lightconditions. J Phycol 41:1196–1203.

51. Engel A (2002) Direct relationship between CO2 uptakeand transparent exopolymer particles production in nat-ural phytoplankton. J Plankton Res 24:49–53.

52. Burkhardt S, Zondervan I, Riebesell U (1999) Effect ofCO2 concentration on the C:N:P ratio in marine phyto-plankton: A species comparison. Limnol Oceanogr44:683–690.

53. Riebesell U, et al. (2007) Enhanced biological carbon con-sumption in a high CO2 ocean. Nature 450:545–549.

54. Hutchins DA, et al. (2007) CO2 control of Trichodes-mium N2 fixation, photosynthesis, growth rates, andelemental ratios: Implications for past, present andfuture ocean biogeochemistry. Limnol Oceanogr52:1293–1304.

55. Levitan O, et al. (2007) Elevated CO2 enhances nitrogenfixation and growth in the marine cyanobacteriumTrichodesmium. Global Change Biol 13:531–538.

56. Ramos JBe, Biswas H, Schulz KG, LaRoche J, RiebesellU(2007) Effect of rising atmospheric carbon dioxide onthe marine nitrogen fixer Trichodesmium. Global Bio-geochem Cycles 21:GB2028, 10.1029/2006GB002898.

57. Falkowski PG, et al. (2004) The evolution of moderneukaryotic phytoplankton. Science 305:354–360.

58. Giordano M, Beardall J, Raven JA (2005) CO2 concen-trating mechanisms in algae: Mechanisms, environ-mental modulation, and evolution. Ann Rev Plant Biol56:99–131.

59. Bopp L, Le Quere C, Heimann M, Manning AC, MonfrayP(2002) Climate-induced oceanic oxygen fluxes: Impli-cations for the contemporary carbon budget. GlobalBiogeochem Cycles 16:1022, 10.1029/2001GB001445.

60. Keeling RF, Kortzinger A, Gruber N (2009) Ocean de-oxygenation in a warming world. Ann Rev Mar Sci, inpress.

61. Matear RJ, Hirst AC (2003) Long term changes in dis-solved oxygen concentrations in the ocean caused byprotracted global warming. Global Biogeochem Cycles17:1125, 10.1029/2002GB001997.

62. Stramma L, Johnson GC, Sprintall J, Mohrholz V (2008)Expanding oxygen-minimum zones in the tropicaloceans. Science 320:655–658.

63. Wohlers J, et al. (2009) Changes in biogenic carbon flowin response to sea surface warming. Proc Nat Acad Sci106:7067–7072.

64. Milliman JD (1993) Production and accumulation ofcalcium carbonate in the ocean: budget of a nonsteadystate. Global Biogeochem Cycles 7:927–957.

65. Heinze C (2004) Simulating oceanic CaCO3 export pro-duction in the greenhouse. Geophys Res Lett31:L16308, 10.1029/2004GL020613.

66. Gehlen M, et al. (2007) The fate of pelagic CaCO3production in a high CO2 ocean: a model study. Bio-geosciences 4:505–519.

67. Ridgwell A, Zondervan I, Hargreaves J, Bijma J, LentonT (2007) Assessing the potential long-term increase ofoceanic fossil fuel CO2 uptake due to ‘‘CO2-calcificationfeedback’’. Biogeosciences 4:481–492.

68. Hofmann M, Schellnhuber HJ (2009) Oceanic acidifica-tion affects marine carbon pump and triggers ex-tended marine oxygen holes. Proc Nat Acad Sci USA106:3017–3022.

69. Langer G, et al. (2006) Species-specific responses ofcalcifying algae to changing seawater carbonate chem-istry. Geochem Geophys Geosyst 7:Q09006, 10.1029/2005GC001227.

70. Armstrong RA, Lee C, Hedges JI, Honjo S, Wakeham SG(2002) A new, mechanistic model for organic carbonfluxes in the ocean based on the quantitative associa-tion of POC with ballast minerals. Deep-Sea Res49:219–236.

71. Klaas C, Archer DA (2002) Association of sinkingorganic matter with various types of mineral ballastin the deep sea: Implications for the rain ratio. GlobalBiogeochem Cycles 16:1116, 10.1029/2001GB001765.

72. Passow U (2004) Switching perspectives: Do mineralfluxes determine particulate organic carbon fluxes orvice versa? Geochem Geophys Geosyst 5:1–5.

73. Oschlies A, Schulz KG, Riebesell U, SchmittnerA(2008) Simulated 21st century’s increase in oceanicsuboxia by CO2-enhanced biotic carbon export.Global Biogeochem Cycles, 22:GB4008, 10.1029/2007GB003147.

74. Engel A, et al. (2004) Transparent exopolymer parti-cles and dissolved organic carbon production byEmiliania huxleyi exposed to different CO2 concen-trations: A mesocosm experiment. Aquatic MicrobEcol 34:93–104.

75. Engel A, Thoms S, Riebesell U, Rochelle-Newall E,Zondervan I (2004) Polysaccharide aggregation as apotential sink of marine dissolved organic carbon. Na-ture 428:929–932.

76. Schartau M, et al. (2007) Modelling carbon overcon-sumption and the formation of extracellular particu-late organic carbon. Biogeosciences 4:433–454.

77. Arrigo KR (2007) Marine manipulations. Nature450:491–492.

78. Legendre L, Rivkin RB (2002) Pelagic food webs: Re-sponses to environmental processes and effects on theenvironment. Ecol Res 17:143–149.

79. McNeil BI, Matear RJ (2008) Southern Ocean acidifica-tion: A tipping point at 450-ppm atmospheric CO2. ProcNatl Acad Sci USA 105:18860–18864.

80. Langdon C, et al. (2000) Effect of calcium carbonate sat-uration state on the calcification rate of an experimentalcoral reef. Global Biogeochem Cycles 14:639–654.

81. Reynaud S, et al. (2003) Interacting effects of CO2 par-tial pressure and temperature on photosynthesis andcalcification in a scleractinian coral. Global Change Biol9:1660–1668.

82. Guinotte JM, et al. (2006) Will human-induced changesin seawater chemistry alter the distribution of deep-seascleractinian corals? Frontiers Ecol Environm 4:141–146.

Riebesell et al. PNAS � December 8, 2009 � vol. 106 � no. 49 � 20609

Dow

nloa

ded

by g

uest

on

July

16,

202

0