The ocean carbon sink – impacts, vulnerabilities and challenges

Abstract. Carbon dioxide (CO2) is, next to water vapour, considered to be the most important natural greenhouse gas on Earth. Rapidly rising atmospheric CO2 concentrations caused by human actions such as fossil fuel burning, land-use change or cement production over the past 250 years have given cause for concern that changes in Earth's climate system may progress at a much faster pace and larger extent than during the past 20 000 years. Investigating global carbon cycle pathways and finding suitable adaptation and mitigation strategies has, therefore, become of major concern in many research fields. The oceans have a key role in regulating atmospheric CO2 concentrations and currently take up about 25% of annual anthropogenic carbon emissions to the atmosphere. Questions that yet need to be answered are what the carbon uptake kinetics of the oceans will be in the future and how the increase in oceanic carbon inventory will affect its ecosystems and their services. This requires comprehensive investigations, including high-quality ocean carbon measurements on different spatial and temporal scales, the management of data in sophisticated databases, the application of Earth system models to provide future projections for given emission scenarios as well as a global synthesis and outreach to policy makers. In this paper, the current understanding of the ocean as an important carbon sink is reviewed with respect to these topics. Emphasis is placed on the complex interplay of different physical, chemical and biological processes that yield both positive and negative air–sea flux values for natural and anthropogenic CO2 as well as on increased CO2 (uptake) as the regulating force of the radiative warming of the atmosphere and the gradual acidification of the oceans. Major future ocean carbon challenges in the fields of ocean observations, modelling and process research as well as the relevance of other biogeochemical cycles and greenhouse gases are discussed.

system may progress at a much faster pace and larger extent than during the past 20 000 years. Investigating global carbon cycle pathways and finding suitable mitigation strategies has, therefore, become of major concern in many research fields. The oceans have a key role in regulating atmospheric CO 2 concentrations and currently take up about 25 % of annual anthropogenic carbon emissions to the atmosphere. 10 Questions that yet need to be answered are what the carbon uptake kinetics of the oceans will be in the future and how the increase in oceanic carbon load will affect its ecosystems and their services. This requires comprehensive investigations, including high-quality ocean carbon measurements on different spatial and temporal scales, the management of data in sophisticated data bases, the application of state-of-the-art 15 Earth system models to provide future projections for given emission scenarios as well as a global synthesis and outreach to policy makers. In this paper, the current understanding of the ocean as an important carbon sink is reviewed with respect to these topics. Emphasis is placed on the complex interplay of different physical, chemical, and biological processes that yield both positive and negative air-sea flux values for natural important one. Other radiatively active trace gases like methane (CH 4 ), halocarbons, and nitrous oxide (N 2 O) have a higher greenhouse potential per molecule than CO 2 , but are less abundant in the atmosphere than CO 2 , so that CO 2 is the most important anthropogenic driving agent of climate change (Myhre et al., 2013). The focus of this review is, thus, on CO 2 and the oceanic ("carbon") sink. Future CO 2 emission scenarios 15 to drive climate models have been produced on empirical evidence concerning human behaviour and economics. In view of the on-going high energy use in wealthy nations and the accelerating energy production in emerging economies (especially China and India; see Raupach et al., 2007), current and recent annual CO 2 emission rates are at the levels of the most pessimistic emission scenario as produced a few years ago for 20 the climate projections of the 5th assessment report of the IPCC (RCP scenarios; van Vuuren et al., 2011a, b;Peters et al., 2013). Considering the key role of the oceans in the global carbon budget it is therefore fundamental to broaden our knowledge on their past, present, and future quantitative impact in regulating atmospheric CO 2 concentrations. Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | (Fig. 2, formulas 1 and 2). The more CO 2 gets absorbed by the ocean the lower the amount of CO 2− 3 becomes. In parallel, the concentration of hydrogen ions increases, causing a decrease in open ocean pH that is referred to as ocean acidification. Projections of future ocean pH suggest a potential total reduction by 0.4-0.5 units by the end of the 21st century as compared to pre-industrial levels, resulting in a pH of 7.7-7.8 5 (Haugan and Drange, 1996;Brewer, 1997;Caldeira and Wickett, 2003;. Furthermore, a shifting ratio of HCO − 3 : CO 2− 3 : CO 2 results in a decrease in CO 2 buffering: the larger the concentration of DIC in the ocean becomes, conversely the smaller the fraction of increased carbon added to the atmosphere that can be taken up by the ocean will be. Or in other words, the higher the cumulative CO 2 emissions 10 to the atmosphere become, the less effective seawater will be in dissociating a part of this CO 2 into HCO − 3 and CO 2− 3 . DIC is distributed in the oceans as passive tracer (like dye) by currents and turbulent mixing. In a simplistic model, transportation of carbon in the oceans mainly follows the large scale ocean circulation: in the northern North Atlantic, surface waters are moved 15 to the deep sea in a process of deep-water formation. The solubility of CO 2 gas in seawater increases with decreasing temperature. As newly formed deep water is cold, the downward transport of the carbon fraction dissolved in seawater due to high CO 2 solubility is also called solubility pump. However, the dissociation of CO 2 into bicarbonate and carbonate ions is antagonistic to the solubility and decreases with decreasing tem-Introduction ter path" between South America and Antarctica; Rintoul, 1991). The water that has spent the longest time away from contact with the atmosphere is found in the northern Pacific Ocean below depths of about 2000 m and is approximately 1500 years old. Comparably, the human perturbation of the carbon cycle has occurred only over the last 250 years, and diluting high anthropogenic carbon loads from the upper ocean 5 with large deep-water reservoirs by mixing processes will take at least 6 times as long. Also, the slower oceanic circulation and mixing become with on-going climate change, the smaller the uptake rate of surface waters for human-produced carbon will be and the less efficient the ocean carbon sink will become for absorbing further CO 2 additions to the atmosphere as carbonic acid dissociates less well into bicarbonate and 10 carbonate in water of high pCO 2 .

Biological carbon pumps
While purely inorganic carbon cycling leads to a slight increase of DIC with depth, biological carbon cycling is responsible for most of the gradients existing in the real ocean DIC distribution. These gradients are mainly fuelled by uptake of DIC by biota 15 in the surface ocean to produce particulate matter, the vertical flux of these particles, and degradation of these particles on their downward way through the water column. Biological carbon binding occurs mainly in the ocean surface layer, where phytoplankton through the process of photosynthesis produces biomass that can be utilized by other organisms on higher trophic levels (classical food chain). Next to dissolved CO 2 , 20 phytoplankton requires light and nutrients for their growth, the latter two being critical limiting factors. About 25 % of the particulate organic carbon (POC), which is produced in the ocean surface layer, eventually sinks through the water column (Schlitzer, 2000) with most of it being remineralised and returned to the dissolved phase already within the upper 1500 m. Normally, less than 1 % of POC reaches the open-ocean seafloor 25 by sedimentation (Lee et al., 2004). In addition to POC, marine biota also produce dissolved organic carbon (DOC), which is discriminated from POC based on particle size (Turnewitsch et al., 2007). As increasingly small particles do not sink anymore through 1614 Introduction Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | the water column but become suspended due to the increasing importance of friction for small particles, DOC is transported through the oceans like DIC as a passive tracer. While a large fraction of DOC may persist and accumulate in the water column before being remineralised to inorganic substances, biologically labile DOC is converted quickly (within minutes to days) in the upper ocean, predominantly by microbial activ- 5 ity (Carlson, 2002). By utilising DOC, bacteria can build up exploitable biomass and part of the dissolved organic carbon may re-enter the classical food chain through the "microbial loop". However, as the microbial loop itself includes several trophic levels, a large part of the recycled DOC is converted back to inorganically dissolved carbon along the process (Azam et al., 1983;Fenchel, 2008). In addition to microbial degrada-10 tion, sorption onto larger particles, and UV radiation may constitute further important processes in the removal of dissolved organic matter (Carlson, 2002). The oceanic DOC pool is overall about one order of magnitude smaller than the marine DIC inventory but larger than the POC pool. Nevertheless, the highly reactive POC dominates the effect on variations in the oceanic DIC distribution. Most of the DOC is quite re-15 fractory which is consistent with its high radiocarbon age (4000-6000 years, Druffel et al., 1992 compounds increases slightly at lower temperature and strongly with increasing depth (pressure) (Mucci, 1983;Zeebe and Wolf-Gladrow, 2001). Shell material sinking together with POC through the water column is usually degraded at larger depths than the organic material. The composition of the sinking material determines also its sinking velocity. Phytoplankton (plant plankton) and zooplankton (animal plankton) grazing 5 on plant plankton or eating other zooplankton can modify the vertical particle flux by producing a variety of carbonaceous or siliceous shell material. Shallow seas including the continental margins are marked with high accumulation rates of organic carbon (Jahnke, 1996). In contrast, deep-sea sediments are mainly composed of the hard parts of calcareous and siliceous shell material (Leinen et al., 10 1986;Archer, 1996). In regions of vivid upward motion of water, such as at the Equator, in front of west coasts, in the Southern Ocean, and during vertical mixing in the North Atlantic, the biological pump can be substantial as new nutrients are supplied from below. This happens especially during plankton blooms, when light availability and stable surface water stratification enables temporarily strong photosynthesis lead- 15 ing first to strong production of phytoplankton and subsequent increase in zooplankton which grazes on the phytoplankton. Particle transport via the biological carbon pump, remineralisation, and ocean circulation are superimposed and are responsible for most of the gradients of dissolved carbon and nutrients in the water column: (1) regarding the vertical gradient, low concentrations result at the surface due to biological uptake, 20 while values increase with depth due to remineralisation. (2) In deeper layers, concentrations increase horizontally with age of the water along the trajectory of water flow when the respective water volume receives more and more remineralised products from the particles under degradation. The loop for the cycling of biological carbon through the ocean is closed, when the deeper waters well up and eventually return reasons. Therefore, many BSi-producers are found in upwelling areas, while CaCO 3 producers are more abundant in other oceanic domains (Dymond and Lyle, 1985). The sedimentary climate record shows that modifications of biological carbon cycling have significantly contributed to the glacial drawdown of atmospheric CO 2 during the repeated ice age cycles over the past million years (Balsam, 1983;Farrell and Prell, 15 1989;Oliver et al., 2010).
The organically bound and living biomass carbon reservoirs in the ocean are significantly smaller than the inorganic reservoir (approximate ratio of 1 : 50; Druffel et al., 1992;Ciais et al., 2013). The biological carbon pump does not sequester anthropogenic carbon added to the ocean itself on decadal to centennial time scales (as

Variability, time evolution, and kinetics of the ocean carbon sink
The cycling of carbon in the oceans is a complex interplay of different physical, chemical and biological processes, yielding both positive and negative air-sea flux values for natural and anthropogenic CO 2 depending on the oceanic region and the seasonal cy- 15 cle. Due to the rapid increase of atmospheric CO 2 concentrations in the past 250 years and the resulting implications for the global heat budget, it is of great importance to understand the driving forces of carbon sequestration in the oceans as well as their variability, i.e. to understand the role of the oceans as a sink for anthropogenic CO 2 . 20 The classical view about the marine uptake of anthropogenic CO 2 from the atmosphere is that the ocean sink averaged over the entire globe is operating continuously and reliably and is less variable than the exchange between the atmosphere and the land biosphere including soil and plants (though the classical view also includes that the ocean atmosphere transport of CO 2 co-varies with short-term climate variability). This Introduction view was supported by the basic inorganic carbon buffering mechanism and by the fact that the equilibration timescale between the ocean surface layer and the atmosphere is approximately 6-12 months. The variability of air-sea CO 2 gas exchange is dampened, because not only the CO 2 molecules are taking part in the equilibration process, but the entire surface layer volume needs to achieve chemical equilibria for the com-5 pounds HCO − 3 , CO 2− 3 , and dissolved CO 2 . Therefore, seasonal variations in DIC due to biological production and remineralisation occur quicker than for respective air-sea gas exchange fluxes to compensate for them. Thus, also, the seasonal cycle in the instrumental atmospheric CO 2 record is dominated by the seasonal variation of the land biosphere, especially for the Northern Hemisphere (Keeling et al., 2001). However, with 10 significantly improved observing systems in the past two decades, it has become obvious that on a regional scale air-sea carbon fluxes may considerably differ between years (Le Quéré et al., 2007;Schuster and Watson, 2007). There are indications that these regional and temporal variations have been smoothed out on decadal time scales over the past 20 years (McKinley et al., 2011), but nevertheless observations and mod-15 els suggest that the ocean sink is vulnerable to a decrease in efficiency during further climate change and further rising ambient CO 2 levels (Friedlingstein et al., 2006;Le Quéré et al., 2007;Watson et al., 2009;Arora et al., 2013).

Time evolution and kinetics of the oceanic carbon sink
In general, one has to discriminate between the ultimate uptake capacity of the ocean 20 for anthropogenic CO 2 from the atmosphere and the marine uptake kinetics for this CO 2 . Both are societally relevant and need to be taken into account for emission reduction strategies and development of improved renewable energy systems.
The ultimate uptake capacity denotes the amount of anthropogenic carbon emitted to the atmosphere that in total eventually ends up in the ocean, long after the human-25 caused greenhouse gas emission perturbation has happened and when the ocean carbon cycle has achieved quasi-equilibrium. This time scale is of the order of several 10 000 years, because the ocean water column has to fully equilibrate with the 1619 Introduction CaCO 3 sediment on the seafloor, where a considerable portion of the CaCO 3 will become dissolved after repeated cycling of deep water (Broecker and Takahashi, 1977;Archer, 2005). In addition, high atmospheric CO 2 levels enhance the weathering rate of carbonates on land. This process also works effectively only on long time scales with potentially quicker changing hot spots (Archer, 2005;Beaulieu et al., 2012). The 5 ultimate storage capacity of the ocean critically depends on the total amount of carbon emitted. Burning of 5000 Gt C of potentially available fossil fuel reserves would lead to a higher long-term CO 2 level in the atmosphere and a reduced fractional ocean uptake capacity in comparison to, e.g. burning only 1000 Gt C (Archer, 2005). The impact on societies and life even after 100 000 years depends, thus, on our behaviour concerning 10 usage of fossil fuel reserves today. This fact as well has to be taken into account for greenhouse gas emission reduction strategies. The oceanic CO 2 uptake kinetics denote the speed with which human-produced CO 2 emissions to the atmosphere can be buffered by the oceans. Due to the limiting effect of gas exchange, CO 2 dissociation, turbulent mixing and ocean large-scale circulation, 15 only a certain percentage of the excess CO 2 in the atmosphere can be taken up at a given unit of time by the ocean (Maier-Reimer and Hasselmann, 1987;Joos et al., 2013). Regionally, this also depends on the seasonal variations in circulation, biological productivity, as well as light, temperature, sea-ice cover, wind speed, and precipitation. It is expected that climate change will lead to a more stable density stratification in the 20 ocean and a general slowing down of large-scale mixing and circulation (Meehl et al., 2007). The consequence will be a reduced uptake of anthropogenic carbon from the atmosphere at the ocean surface and also a lower downward mixing of anthropogenic CO 2 into deeper waters. In addition, high CO 2 in the atmosphere implies high CO 2 in surface waters and a reduction in the ocean's capability to dissociate the CO 2 into 25 the other compounds of DIC, i.e. a decreasing buffering ability with rising ambient CO 2 levels. We have, thus, a physical and a chemical driving force acting on the carbon balance simultaneously and slowing down the transfer of anthropogenic carbon from the atmosphere into the ocean. In a situation with reduced ocean ventilation, also the ESDD 5,2014 The ocean carbon sink -impacts, vulnerabilities, and challenges Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | biological pump will be affected and should be considered in the assessment on how the ocean carbon cycle is impacted. The oceanic CO 2 uptake kinetics depend on the rate of CO 2 emissions to the atmosphere: the faster the emissions are increasing, the stronger is the climatic effect on slowing down the uptake and the stronger the chemical effect on decreasing the CO 2 buffering. These effects are caused by water with 5 high anthropogenic carbon load that cannot be mixed into the interior of the ocean with the original efficiency and because the buffering ability of seawater decreases with increasing CO 2 partial pressure in the water. The oceanic bottleneck effect is obvious in several decade-long future scenarios with ocean models (Maier-Reimer and Hasselmann, 1987;Sarmiento and Le Quéré, 1996), and fully coupled Earth system models (Friedlingstein et al., 2006;Roy et al., 2011;Arora et al., 2013). The latter are complex computer programmes, which include dynamical representations of the various Earth system reservoirs (atmosphere, ocean, land surface, ice) and the simultaneous interaction between these reservoirs (Bretherton, 1985;Mitchell et al., 2012).
Earth system models are driven by solar insolation and greenhouse gas emissions 15 and deliver expected time-and space-dependent distributions of important climatic variables. These variables can be of physical nature, such as temperature, precipitation, salinity, wind fields, ocean currents, sea-ice cover, or of biogeochemical nature, such as CO 2 concentration in ocean and atmosphere, pH value in the ocean, nutrient and oxygen concentrations, soil organic carbon, or biological productivity. The tempo-20 rary build-up of high CO 2 concentrations in the atmosphere increases directly with the human-produced CO 2 emissions. At pessimistic scenarios with high annual emissions, the annual fraction of emissions buffered by the oceans is reduced, while pathways with reduced emissions enable a more efficient oceanic uptake rate. Inclusion of carbon dynamics in ocean and land models increases the sensitivity of climate models 25 with respect to radiative warming. This means that models with carbon cycle representations and respective carbon-cycle-climate-feedbacks lead to an overall stronger warming than with conventional climate models that do not include an interactive car-Introduction bon cycle. The range of this feedback is still large due to inherent model uncertainties and a partial lack of process understanding in all relevant disciplines.

Observations of ocean carbon variability
In the past two decades, the number of ocean carbon observations has considerably increased (Sabine et al., 2010). Data collection ranges from the surface to the deep 5 ocean, encompasses different oceanic regions and includes various time series to capture both spatial and temporal variations. Satellite measurements have been extremely useful to identify the geographical distribution of biological primary productivity at the sea surface over seasonal as well as interannual cycles and to derive wind fields of high value for quantification of gas transfer velocities across the air-water interface.

10
Targeted research cruises as well as the use of commercial ships (voluntary observing ships, VOS) equipped with automated systems are the backbone of surface ocean CO 2 concentration measurements, the data being synthesised in the SOCAT project ( Fig. 3) (Pfeil et al., 2013;Sabine et al., 2013;Bakker et al., 2014). Selected buoys and floats are used to capture the spatio-temporal variability of ocean carbon. The most 15 prominent network of floats was established in the framework of ARGO (Array for Realtime Geostrophic Oceanography) that delivers valuable temperature, salinity, and current data for a better understanding of mixed layer and subsurface dynamics. However nowadays, ocean floats are also successfully exploited as platforms for measuring e.g.  et al., 2009). These time series stations have often been established in areas of fairly low short-term variability in order to allow a reliable establishment of long-term trends in the observations. Though the observational basis for assessing changes in the oceanic carbon cy-5 cle is limited, a number of major findings have been achieved. Sabine et al. (2004) compiled a global map of the ocean water column storage of anthropogenic carbon for the year 1994. In this map, the North Atlantic and the Southern Ocean with adjacent regions are recognized as hot spot areas for anthropogenic carbon storage. By combining observations with statistical and process-based model approaches, it could 10 be shown that in these regions the annual uptake of CO 2 from the atmosphere has temporarily decreased, though the total inventory of the anthropogenic water column burden has monotonously increased. Both the North Atlantic and the Southern Ocean are deep-water production areas that would be very vulnerable regions with respect to climate-change induced slowing of oceanic carbon uptake. For the North Atlantic, ing atmospheric CO 2 for some time. This result could be achieved using models driven with realistic atmospheric forcing in combination with observations primarily from the Indian Ocean sector of the Southern Ocean (Le Quéré et al., 2007;). Partly, this change can be attributed to climatic oscillations in the Southern Hemisphere and their modifications due to changes in wind forcing associated with the decrease in 5 stratospheric ozone (Lenton et al., , 2013. Finally, also the tropical Pacific Ocean with the strongest known short-term climate variation of Earth called ENSO (El Niño Southern Oscillation) induces large temporary variability in ocean carbon uptake. The increased sea-surface warming during ENSO events and reduced upwelling of carbonrich waters result in a temporarily reduced outgassing and an enhanced oceanic car-10 bon uptake, respectively (Feely et al., 1999;Ishii et al., 2009). ENSO variations also have implications for air-sea fluxes in the tropical Atlantic as documented by Lefèvre et al. (2013). Regarding future scenarios for the evolution of ocean carbon sinks, Earth system models driven by solar insolation and greenhouse gas concentrations indicate the 15 strongest areas for sequestration of anthropogenic carbon are in the Southern Ocean as well as the tropical ocean (Tjiputra et al., 2010;Roy et al., 2011). The Southern Ocean seems to be the ocean fly wheel for changes in atmospheric CO 2 , not only for anthropogenic carbon uptake, but also for natural variations in atmospheric CO 2 (Sigman and Boyle, 2000;Heinze, 2002;Watson and Naveira Garabato, 2006). Long-20 term observational capacity for the Southern Ocean is critical to monitor the ocean sink strength for anthropogenic carbon.

The impact of human-produced carbon on warming and marine ecosystems
The ocean carbon sink provides a major service to human societies in removing anthropogenic CO 2 from the atmosphere and, thus, reducing the additional radiative forcing of the Earth system. On the other hand, dissociation of anthropogenic CO 2 in seawater increases ocean acidification, whose potential impacts on the diversity and function-Introduction ing of marine ecosystems are not yet fully understood. Understanding the role of the oceanic carbon sink in controlling Earth's heat budget and influencing marine life is of great importance to project future effects of climate change. Scenarios with Earth system models (advanced climate models, for a more detailed explanation see Sect. 3.2) reveal that the ocean sink may become less efficient in the future as higher cumulative 5 CO 2 emissions counteract the general tendency for oceanic CO 2 uptake. It, thus, remains to be explored what the ocean's ultimate uptake capacity for atmospheric CO 2 is, when it may be reached, and how until then the ocean may regulate the environmental effects of anthropogenic CO 2 . 10 The net carbon uptake rates of land and ocean determine the future time evolution of radiative forcing of the atmosphere and, hence, climate change for a given emission scenario (for a detailed definition of radiative forcing see Myhre et al., 2013). Joos et al. (2013) used different Earth system models to compute an average integrated global warming potential for a pulse emission of 100 Gt-C (Gt = Gigatonnes) into the 15 atmosphere. In the study it is also stressed that quantifying the global warming effect for certain retentions of CO 2 emissions to the atmosphere depends critically on the time horizon considered. For the 100 Gt-C pulse to the atmosphere, e.g. 25 ± 9 % of the pulse emission would remain in the atmosphere after 1000 years, during which the ocean and land would have absorbed 59±12 % and 16±4 %, respectively. This empha-20 sizes the long time horizon for the anthropogenic perturbation, which has to be taken into account even for a world with strongly reduced CO 2 emissions (Plattner et al., 2008). For higher total emission pulses, the overall retention in the atmosphere would be higher and likewise the global warming potential per kg CO 2 brought into the atmosphere (Maier-Reimer and Hasselmann, 1987;Archer, 2005) due to the weakening 25 buffering capacity of the ocean at high ambient CO 2 partial pressure. In recent years, a limit to future global warming of 2 • C above the average preindustrial surface temperature has been set as a, less ideal, but potentially achievable 1625 ESDD 5,2014 The ocean carbon sink -impacts, vulnerabilities, and challenges C. Heinze et al. target for greenhouse gas emission strategies. Recent experiments with a coarse resolution Earth system model taking into account multiple climate targets, i.e. limits for maximum amplitudes of specific variables such as surface air temperature increase, sea-level rise, aragonite saturation, and biomass production on land, reveal that CO 2 emissions need to be substantially reduced for achieving several mitigation goals si-5 multaneously, rather than for meeting a temperature target alone (Steinacher et al., 2013). Accounting for the carbon cycle climate feedback as well as other physical and biogeochemical feedbacks in climate models is of great importance for estimating the allowable emissions for a certain time line of atmospheric CO 2 concentration and global warming. Complex Earth system models are needed for this. Simplified reservoir models, such as Integrated Assessment Models, as often used in economical modelling and for construction of typical future scenarios, are insufficient for this purpose as they do not account for internal feedbacks in the Earth system in a dynamical way . 15 The term "ocean acidification" refers to the decrease of oceanic pH by 0.1 units over the past 250 years and the predicted lowering of pH by another 0.3-0.4 units until the year 2100 (Caldeira and Wickett, 2003;Raven et al., 2005). Its main cause is the uptake and dissociation of excess CO 2 from the atmosphere that leads to an increase in the oceanic hydrogen ion concentration. Thorough monitoring of ocean acidification is 20 of great importance, and by collecting values in observational carbon data bases (e.g. like SOCAT and fixed time series stations) as well as by conducting long-term carbon time-series measurements (e.g. as reported in Vázquez-Rodríguez et al., 2012) our understanding of this process and its spreading throughout Earth's oceans can be significantly advanced (Figs. 3 and 4). In addition, investigating the potential ef- 25 fects of "high CO 2 -low pH" conditions on the diversity and functioning of marine biota and ecosystems is currently the focus of many scientific studies. The interpretation of the observed responses in a species-and ecosystem-relevant context thereby sug-1626 ESDD 5,2014 The ocean carbon sink -impacts, vulnerabilities, and challenges gests that the two ocean acidification stressors high CO 2 concentration and decreased pH are very often only one part of a complex equation. Other environmental stressors like temperature, light availability, oxygen concentration, nutrient concentration, CaCO 3 saturation state or trace metal speciation (to name only a few) as well as time and physiological characteristics of the investigated organisms themselves have to be taken into 5 account when elaborating on ocean acidification impacts (Raven et al., 2005;Pörtner, 2008;Ries et al., 2009;Dupont et al., 2010). The most immediate response to an increase in CO 2 concentration and a decrease in seawater pH is expected for marine calcifying organisms, including corals, molluscs, crustaceans, echinoderms, coccolithophores, foraminifera as well as coralline and cal-10 careous algae. Maintenance and production of shells and skeletons may cost more energy in an environment with reduced pH, and altered organism physiology may increase the vulnerability of certain species and compromise their ecosystem functions (Bibby et al., 2007;McClintock et al., 2009;Tunnicliffe et al., 2009). Calcification rates are likely to decline with a reduced saturation value for aragonite and calcite, the two 15 most common forms of CaCO 3 in seawater (Feely et al., 2004;Guinotte and Fabry, 2008), caused by a decrease in CO 2− 3 concentration when CO 2− 3 , excess atmospheric CO 2 , and H 2 O react to HCO − 3 and hydrogen ions. Future projections indicate the potential undersaturation for both aragonite and calcite within the current century for all polar regions (see Fig. 5) and parts of the subpolar Pacific Ocean as well as the deep 20 North Atlantic Ocean (Orr et al., 2005;Fabry et al., 2008;Steinacher et al., 2009;Orr, 2011). Because aragonite dissolves at higher CO 2− 3 concentrations than calcite, corals and other aragonite-producing organisms are expected to experience corrosion of their hard shell materials due to ocean acidification first. At natural CO 2 seeps in Papua New Guinea, a decline in coral diversity was documented in areas of reduced pH as struc- 25 turally complex corals were replaced by massive Porites corals (Fabricius et al., 2011). The consequences arising from this diversity shift could be similar to those anticipated for a general reduction in coral cover and include a loss in biodiversity, habitat availability and quality as well as reef resilience (Fabricius et al., 2011). The decrease in ESDD 5,2014 The ocean carbon sink -impacts, vulnerabilities, and challenges CaCO 3 saturation as a result of ocean acidification combined with other environmental impact factors such as an increase in temperature can be critical (Kleypas et al., 1999;Hoegh-Guldberg et al., 2007;Veron et al., 2009;Fabricius et al., 2011). Recent scenario computations with Earth system models document that a drastic reduction of CO 2 emissions is required to preserve major coral reefs during the Anthropocene 5 (Ricke et al., 2013). However, aspects such as potential adaptation processes and migration need yet to be included in regional studies (Yara et al., 2012). The effects of ocean acidification on different groups of marine biota can be rather diverse and complex. For example, specimens of the economically and ecologically important blue mussel Mytilus edulis recovered from the North Sea showed drasti-10 cally reduced calcification rates, while specimens recovered from a coastal area of the Baltic Sea did not show any sensitivity to increased pCO 2 values (Gazeau et al., 2007;Thomsen et al., 2010;Schiermeier, 2011). Mussels from the Baltic seemed to be adapted to thriving in waters that generally experience strong seasonal pCO 2 fluctuations, and food availability may have potentially outweighed the effects of ocean 15 acidification ( Thomsen et al., 2010Thomsen et al., , 2013. In a study comparing different types of benthic marine calcifiers it could be shown that certain species experienced dissolution, while others were able to exploit the higher pCO 2 content in seawater and increased their net calcification. Physiological characteristics like the organism's ability to regulate pH, shell-protection with organic layers, biomineral solubility, and photosynthesis 20 utilization seemed to play a role (Ries et al., 2009). Species-specific reactions as well as an organism's life cycle stage are further factors that may have to be taken into account as it has been shown e.g. for echinoderms (Dupont et al., 2010(Dupont et al., , 2013Dupont and Pörtner, 2013). Results obtained for phytoplankton communities additionally stress the importance of community composition and/or shifts when assessing ocean acidifi-ESDD 5,2014 The ocean carbon sink -impacts, vulnerabilities, and challenges Ocean acidification does not only affect calcifying biota. Sensitivity towards ocean acidification has been detected for fish and other invertebrates, with increased risks of acidification of body fluids and tissues as well as hindered respiratory gas exchange (Raven et al., 2005). Beneficial effects were observed e.g. for seagrass (Palacios and Zimmerman, 2007;Hall-Spencer et al., 2008;Fabricius et al., 2011) and various algal 5 species (Hall-Spencer et al., 2008;Connell et al., 2013).

Ocean acidification and its impact on marine ecosystems
Projecting the precise impact of ocean acidification on the diversity and functioning of marine organisms and ecosystems is challenging. A meta-analysis of 228 published studies by Kroeker et al. (2013) revealed a decrease in calcification, growth, survival, development, and abundance across a wide range of taxa, but also showed a certain 10 degree of variability among groups suggesting different scales of sensitivity. It is not well established to which degree organisms can adapt to quasi-permanent changes in ocean pH due to rapid anthropogenic carbon input. It is also not known, if and in what way consequences like the physiological impairment of vulnerable species and the reduction and/or shifts in biodiversity may be mastered provided that ecosystem 15 functionality shall be preserved. With regard to the sustainable development of marine resources, future research will need to focus on multiple stressor studies over various time scales to reveal the functional impact of ocean acidification (and climate change in general) on marine ecosystem services and provide both comprehensive monitoring and solution-oriented results.

Future impact research
For future modelling approaches, not only the effects of atmospheric and oceanic warming as well as ocean acidification have to be considered, but also the influence of multiple stressors. These include physical and chemical drivers as well as circulation and stratification changes, freshening, changes in ice cover, deoxygenation, anthro-25 pogenic nitrogen input, changes in dust supply, marine pollution by offshore activities (e.g. Deepwater Horizon disaster; Mearns et al., 2011), and plastic waste (also on the micro-scale; Gross, 2013) or overfishing and bottom trawling. Earth system models 1629 ESDD 5,2014 The ocean carbon sink -impacts, vulnerabilities, and challenges that represent the marine carbon cycle and related biogeochemical cycles have been successfully used to establish the regional combination of some major stressors and the future evolution of these combinations ). Yet, robustness in regional projection is strongly dependent on the considered stressors and regions, and identifying the onset of emission induced change is still a challenging task that is espe-5 cially sensitive to the considered emission-scenario (see Fig. 5). The combined action of stressors has to be accounted for in designing correct future scenarios for the next generation of Earth system model climate projections (Steinacher et al., 2013). A critical variable within this context is the sustained generation of exploitable biomass in the ocean for human food production, where overall biological carbon fixation rates will 10 presumably decrease with a more stagnant ocean circulation (Steinacher et al., 2010).

The ocean carbon sink in relation to the land carbon sink
The atmospheric CO 2 concentration is determined by the CO 2 emissions and the CO 2 exchanges between the land biosphere and atmosphere as well as between the atmosphere and ocean. Quantification of the regional as well as global land carbon sink is 15 associated with high uncertainties due to the direct coupling of CO 2 consumption and release on the land surface with the atmosphere in combination with the heterogeneity of the land biosphere, its constant change and different forms of land use including forestry changes. Complex soil processes like the degradation of organic material and permafrost melting processes (Schuur et al., 2009) Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | system modelling frameworks employed for the projections as summarised in the 5th assessment report of IPCC (Collins et al., 2013) included N limitation on land, and related processes and feedbacks are under discussion. In comparison to the land carbon sink, the large-scale oceanic sink is considered to be less variable on an interannual time scale (though considerable perturbations of the 5 ocean carbon cycle are linked with, e.g. the ENSO cycles; Feely et al., 2006) and, even though a 3-dimensional approach is required due to water motion, somewhat easier to quantify. This traditional view is exploited to estimate the year-to-year land sink for anthropogenic carbon from the atmospheric observations and ocean models (evaluated through observations). The terrestrial carbon sink is then the residual of CO 2 emissions, atmospheric CO 2 concentrations, and ocean-atmosphere CO 2 fluxes (Canadell et al., 2007;. Until precise quantifications of the land carbon sink become available through direct observations and modelling, estimating it through the ocean carbon sink is a valid option. However, with increasing detail in oceanic carbon sink determinations, oceanographers are starting to run into similar heterogeneity 15 problems in the oceans as geo-ecologists on land, especially when the continental margins, the shelf seas, and coastal and estuarine systems are taken into account (Borges, 2005;Liu et al., 2010b;Regnier et al., 2013). These likewise heterogeneous systems are so far not (or at best partially) included in global Earth system model scenarios, because the resolution of these models does not allow for the resolution of the 20 respective topographic features and super-computers are currently insufficient to run respective high-resolution models as yet (Mitchell et al., 2012). Measurements of the O 2 /N 2 ratio in the atmosphere and marine oxygen budgets can help to further specify the land carbon sink (Keeling et al., 1996).
The interannual variability of land-atmosphere carbon fluxes appears to be higher 25 than the respective variations for ocean-atmosphere fluxes when computing the land carbon sink as the residual between oceanic uptake and atmospheric CO 2 retention (Canadell et al., 2007). On a multi-millennial time scale, peat formation and organic carbon burial in lakes contribute to slow long-term accumulation on land (Einsele et al., 5,2014 The ocean carbon sink -impacts, vulnerabilities, and challenges  Gorham et al., 2012). Due to the overall smaller carbon inventory of the land biosphere as compared to the inorganic ocean carbon pool (Fig. 6), it is expected that the ocean through inorganic buffering and CaCO 3 sediment dissolution would ultimately account for the major part of removal of the human-induced addition of CO 2 to the atmosphere (Archer, 2005).

6 Major ocean carbon challenges and key knowledge gaps
Some aspects of marine carbon cycling can be regarded as well-established research fields, such as the inorganic carbon buffering system. However, other elements are more difficult to approach, partly due to inherent principle difficulties and partly based on the lack of technological and infrastructural effort. Within this section, some major 10 ocean carbon challenges and key knowledge gaps in ocean carbon research will be addressed.

Observational data bases
Based on measurements, our knowledge of inorganic and organic carbon cycling has significantly improved over the past decade. This is especially due to measurements Measurements of dissolved oxygen are of key importance for carbon cycle research. Oxygen data are the basis for improving estimates of the land carbon sink (Keeling et al., 1996) and for identifying any emergent fingerprint (Andrews et al., 2013), an extensive O 2 measurement programme is needed. In addition, measurements of at least two carbon variables of the marine inorganic carbon system are necessary. Here, 5 pH and pCO 2 are likely the ones where the techniques first will be available on floats, though this combination is not optimal for deriving the other inorganic carbon variables. Another option would be to measure DIC and alkalinity as the latter easily can be measured in seawater and determines together with DIC the marine inorganic carbon system (see Wolf-Gladrow et al., 2007). In combination with O 2 measurements on 10 automated float systems, this altogether would provide a significant advance in ocean carbon observations. Pilot studies conducted in recent years yielded promising results for a world-wide application of such systems (Gruber et al., 2010;Fiedler et al., 2013).
For improved estimates of the biological carbon pump variations, reliable shallow flux estimates as well as state-of-the-art biogenic CaCO 3 (aragonite, calcite) and bio- 15 genic silica (BSi) production maps would be desirable. Respective maps for CaCO 3 export production are at present possibly associated with large errors and give partly incongruous results Balch et al., 2007). Highly accurate total alkalinity observations and a reliable CaCO 3 surface map could be used as reference points for future developments of biocalcification under high CO 2 (Ilyina et al., 20 2009). Satellite observations have greatly improved our understanding about primary production in the ocean (Henson et al., 2012), but remote sensing efforts have still to be better exploited and extended in order to fill the gaps of fragmental in-situ observations, especially also for production of hard part shell material.
Anthropogenically induced elevated carbon levels in the ocean (C ant ) cannot be ob- 25 served directly, which is why indirect methods have to be used (Gruber et al., 1996;Hall et al., 2002;Touratier and Goyet, 2004;Friis et al., 2005). Even though year-to-year changes in DIC are measurable in ocean surface waters, it is a challenge to determine them in deeper layers as the anthropogenic perturbation in seawater is relatively small ESDD 5,2014 The ocean carbon sink -impacts, vulnerabilities, and challenges when compared to the natural background. Over the past years, major international networks and projects (EU framework programmes, OCB, PICES, SOLAS, IMBER, IOCCP etc.) have helped to make much scientific progress in ocean carbon research worldwide. However, extensions and new projects are required to continue the work (GEO/GEOSS, GOOS, FOO, ICOS etc.).

5
In contrast to the atmosphere, oceanic meso-scale circulation features are marked by short spatial scales and large time scales. While an atmospheric pressure system has a typical length scale of 1000 km and a lifetime of days to weeks, comparable oceanic meso-scale eddies have scales of 2-100 km and several months. Therefore, selected oceanic observations can be aliased through meso-scale motion and may not 10 reflect the long-term mean state.
Time series stations in the ocean are still rare and mostly cover low to mid-latitudes (e.g. HOTS, BATS, ESTOC, PAP, PAPA, DYFAMED). These time series have provided a lot of insight into the long-term evolution of carbon cycle tracers, e.g. the local decline of mean sea surface pH has been documented as unequivocal proof of progressing 15 ocean acidification (Santana-Casiano et al., 2007;Bates et al., 2014). An expansion of time series stations at higher latitude areas would be desirable as, e.g. the change in sea surface pCO 2 and pH would be largest over time, although the mean signal there would be somewhat more blurred by interannual variability Bauerfeind et al., 2014). 20 Apart from the issues described above, general challenges for determination of oceanic carbon budgets within the Earth system exist, which possibly never can be met adequately: (1) the annual net uptake rate of anthropogenic carbon from the atmosphere is small as compared to the gross upward and downward fluxes occurring over one year in different oceanic regions. That means that we always will have to quan-25 tify small net exchange fluxes as difference of large gross fluxes into and out of the ocean.
(2) The pristine carbon fluxes between the atmosphere and the ocean as well as the pre-industrial 3-dimensional distributions of DIC have not been measured and need to be reconstructed (Khatiwala et al., 2009(Khatiwala et al., , 2013. It is unlikely that ocean carbon ESDD 5,2014 The ocean carbon sink -impacts, vulnerabilities, and challenges variables can be reconstructed with high accuracy for the pre-industrial from potential proxy record development.

Process and impact knowledge
A major obstacle for improvements in future projections of the Earth system for selected future scenarios of driving factors is the lack of sufficient process understanding, 5 process quantification, and process identification. Though some major biogeochemical principles are known, detailed dynamical formulations of processes are scarce and in their infancy. There is a considerable uncertainty about the gas transfer velocity of CO 2 and other gases across the air-water interface (Carpenter et al., 2012;Garbe et al., 2014). While the global ocean carbon sink estimates may not too strongly depend on 10 this choice (otherwise projections with simple two box models for the global ocean would not have worked at all; Oeschger et al., 1975), the projected local CO 2 concentration in ocean surface waters is highly influenced by the chosen gas transfer velocity values, also for appropriate regional validation of ocean models. The co-limitation of biological production by various factors is an established concept, however, crucial 15 details are not uniformly established, such as the potential variation of carbon to nitrogen ratios in biogenic matter under different environmental conditions (Riebesell et al., 2007;Jiang et al., 2013). Marine particle fluxes and their dynamics are still poorly understood and not yet adequately quantified in a dynamic way in response to external drivers (Klaas and Archer, 2002;Gehlen et al., 2006). The ongoing and future im-20 pacts of high CO 2 on marine organisms have yet to be clarified (Gattuso and Hansson, 2011). Formulations on how to quantify the production as well as degradation of phytoand zooplankton particulate matter (organic, inorganic) are not mature enough or not even existing for providing step-change improvements of complex ocean models as well as coupled Earth system models. This includes, in particular, potential adapta- 25 tion of organisms and ecosystems to conditions not experienced since the geologic past (Langer et al., 2006 (Kattge et al., 2011). Approaches for the simulation of ocean ecosystems with multiple plankton functional types have been 5 initiated (Le Quéré et al., 2005), but trait data bases for marine organisms are not yet available in a suitable format and information from mesocosm and laboratory experiments is scarce and may not be straightforwardly transferable to the real Earth system.

Integrative modelling and combination with measurements
For simulations of the ocean carbon sink and its impact, suitable models are needed 10 to explain past and present events as well as to predict potential future pathways. Biogeochemical ocean general circulation models are employed either through observed forcing or within coupled Earth system models (reviewed in Heinze and Gehlen, 2013). There is a trade-off between their resolution (space and time) and a technically feasible length of the simulation period. High-resolution models with eddy dynamics 15 (large-scale turbulent mixing) are often too computationally expensive for integrations exceeding a few decades. However, multiple future scenarios calculated over decades, centuries, and millennia are necessary to achieve reliable future projections. In addition, biogeochemical models whose water mass properties shall be fully predicted by the models need very long and costly spin-up periods in order to bring the tracer dis-20 tributions including the carbon cycle tracers into quasi-equilibrium. Integration periods need to be at least as long as one full oceanic circulation cycle of about 1500 years. Even for still fairly coarse resolutions this is currently not easily done and quite costly. Global model simulations of deep-sea carbon distributions as well as other deep-sea properties are therefore often limited to a lower resolution as compared to their distri- 25 butions in surface or shallow waters Séférian et al., 2013;Tjiputra et al., 2013). 5,2014 The ocean carbon sink -impacts, vulnerabilities, and challenges Models need systematic improvement by combining them with and comparing them to observational data. By applying data assimilation procedures (Brasseur et al., 2009), existing discrete observations of oceanic variables can be interpolated (gap filling) and free adjustable parameters in models (such as, e.g. the particle sinking velocity) can be calibrated. Data-driven diagnostic models (Usbeck et al., 2003) are important for 5 suggesting first order values of free parameters in dynamical process descriptions and can be implemented in complex forward models, which can be used for predictions as well. Systematic model assessment with observations and model optimisation with data assimilation have made progress in recent years, but for integrated biogeochemical cycle simulations these approaches need to be extended. Skill score metrics, which 10 can be used to rank models according to their ability to reproduce physical and biogeochemical variables simultaneously, may become a valuable tool for future simulations. A simplified short cut method in order to assess the quality of future projections of Earth system models is the emergent constraint approach (Cox et al., 2013;Hoffman et al., 2014;Wenzel et al., 2014). In this approach, an interrelation is sought between a spe-15 cific Earth system sensitivity as resulting across an ensemble of comparable models and a corresponding observational trend or variability (see also Flato et al., 2013). This method has just started to also be used for addressing ocean biogeochemical problems (Hoffman et al., 2014) and respective constraints have to be identified for this research field. Model scenarios can diverge depending on slight modifications of the 20 starting (initial) and boundary conditions during a model run as well as due to internal variability in the model. Therefore, for a given CO 2 emission scenario the expected evolution of the results can differ. Ensemble simulations are necessary for establishing a range of statistically valid, potential outcomes that are associated with different degrees of probability. Due to the immense costs for multiple integrations of complex Introduction

Specific regional foci for ocean carbon cycle studies
There are at least 6 major regional domains, which warrant more attention in the coming years of ocean carbon cycle research: 1. The Southern Ocean is quantitatively the most important region for worldwide carbon dynamics (today: Mikaloff Fletcher et al., 2006;glacial/interglacial: Watson 5 and Naveira Garabato, 2006; future: Tjiputra et al., 2010;Roy et al., 2011), but it is also one of the least well year-round observed regions (Takahashi et al., 2009;Swart et al., 2012;Pfeil et al., 2013;Sabine et al., 2013) due to its remoteness and high seasonality. Research priorities include the improvement of data coverage for carbon variables, dissolved oxygen, and related tracers. The water mass formation, mixing and deep convection processes, in particular in the Southern Ocean, are the "Achilles heel" of global ocean models, and a step-change improvement is needed in order to achieve more physically based deep-water production representations in ocean models as well as Earth system models (Lenton et al., 2013). This includes also the representation of Antarctic shelf regions and respective 15 water-mass formation mechanisms relevant for large-scale simulations.
2. Highly dynamic systems such as shelf areas, coastal zones, estuaries and continental margins will need to be accounted for in global carbon cycle quantifications. This is of key importance for impact studies as shallow seas are major spawning and living grounds for commercially exploited fish and food production. In addition, 20 anthropogenic stressors such as mega cities, pollution from riverine loads and deposition of reactive nitrogen (Duce et al., 2008) have to be considered. Progress has recently been made in providing advanced combined river runoff and river load data for use in biogeochemical models (Mayorga et al., 2010). Ocean biogeochemical models should include both pelagic ocean sediment models ) and shallow sediment representations to involve high fluxes and regeneration rates of organic sediments as well as respective low oxygen and ESDD 5,2014 The ocean carbon sink -impacts, vulnerabilities, and challenges C. Heinze et al. anoxic reactions and matter transformations like methanogenesis or denitrification (Naqvi et al., 2010;Mogollón et al., 2012). Land-ocean coupling of natural and anthropogenically perturbed systems (Regnier et al., 2013) needs inclusion in global Earth system models, especially with regard to quantifying nation-wide closed carbon budgets. 3. The Arctic Ocean is a hot spot of climatic and environmental changes, and represents the area in which ocean acidification accelerates most rapidly (Steinacher et al., 2009). Like the Southern Ocean, the Arctic is highly undersampled, making it difficult to determine reliable CO 2 sink estimates (Schuster et al., 2013). New process understanding (Wåhlström et al., 2012 has to be integrated into 10 large-scale ocean models. Shifts in water mass formation processes, including the cold halocline structure at the Arctic Ocean surface domain (Aagaard et al., 1981;Anderson et al., 2013), need to be identified. A strongly reduced Arctic sea-ice cover and changes in annual sea-ice formation will have fundamental consequences for both organic and inorganic carbon cycling as well as ocean 15 circulation and mixing (Loeng et al., 2005). The net effect on ocean carbon sink behaviour for a summer ice-free Arctic Ocean is not yet firmly assessed. Future studies need to include both sea-ice physics and sea-ice biogeochemistry. In addition, the potential climatically and tectonically induced degassing of CH 4 from Arctic Ocean sources needs to be further monitored as a potentially significant 20 greenhouse gas source (Biastoch et al., 2011;Shakhova et al., 2014).
4. The tropical ocean is another key sink area for anthropogenic carbon (Mikaloff Fletcher et al., 2006;Roy et al., 2011). Future research needs to focus on ENSOrelated variability in its carbon sink potential as well as on it being a region of high phytoplankton production rates in respective upwelling areas, where large- 25 scale impacts of ocean acidification may be measured already during an early stage (Ilyina et al., 2009) 6. Coastal upwelling areas have proven to be useful study areas for ocean acidifica-10 tion, deoxygenation, and biological carbon pump studies and will remain a major focus of future monitoring (Feely et al., 2008;Paulmier et al., 2008;Gruber et al., 2011). It will therefore be crucial to appropriately resolve the physically and biogeochemically highly dynamic regimes along continental margins both in observational campaigns and modelling efforts.

Using the ocean natural laboratory for case studies on complex couplings
The ocean and Earth system need to be better used as laboratories to understand processes and the resulting effects on a global scale. This can, for example, be achieved by using a biogeographic approach, where ecosystems are analysed along natural gradients in both space and time. Natural, environmental variability needs to be better 20 exploited to obtain results for impact research. Transient large-scale variations of the Earth system and the ocean carbon cycle's role in these patterns need to be explained.

Combination with other biogeochemical cycles and greenhouse gases
The ocean carbon cycle needs to be studied and assessed in combination with other biogeochemical cycles in a more focussed way than in the past. The oceanic ESDD 5,2014 The ocean carbon sink -impacts, vulnerabilities, and challenges C. Heinze et al. sources/sinks of CH 4 , N 2 O, and CO 2 , all three being natural and anthropogenic greenhouse gases, are controlled by coupled elemental cycles involving among others carbon compounds, nutrients, and gases. Only integrative approaches can ensure a full understanding of the coupled cycles and a full exploitation of respective observational evidence. The simultaneous quantifications of the oxygen and carbon cycles are vital 5 for closing the global carbon budget including the terrestrial biosphere. Nutrient cycles and their anthropogenic perturbations directly control the biological carbon cycling on land and in the oceans. Their more detailed dynamical implementation in land and ocean models is needed, including a better understanding of nutrient limitations (including effects of micronutrients such as iron) under changing environmental conditions. 10

Conclusion
The ocean carbon sink has two parallel effects: (1) parts of the anthropogenic CO 2 emissions are absorbed by the ocean and, thus, the radiative forcing associated with the human-caused excess CO 2 is reduced.

ESDD
Trends are based on a comprehensive suit of Earth system models and IPCC emission scenarios. The choice of models and scenarios is based on the IPCC AR5 report and references denoted within (Plattner et al., 2001;Orr et al., 2005;McNeil and Matear, 2008;Feely et al., 2009;Steinacher et al., 2009Steinacher et al., , 2010Keeling et al., 2010;Cocco et al., 2013). Note that trends in oxygen and net primary production are only analysed at the final year of the IPCC scenarios (year 2100), and the projected trends are most likely starting already at lower atmospheric CO 2 concentrations. 5,2014 The ocean carbon sink -impacts, vulnerabilities, and challenges C. Heinze et al.  Reservoir mass numbers and annual exchange fluxes are given in Pg C (10 15 g C) and

ESDD
Pg C yr −1 , respectively. Black numbers refer to pre-industrial values (before 1750). Red flux numbers represent annual anthropogenic fluxes averaged over the years 2000-2009 and red reservoir numbers depict cumulative changes of anthropogenic carbon between 1750-2011 (90 % confidence interval). A positive cumulative change denotes an increase in (gain of) carbon since the onset of the Industrial Era. Land-atmosphere carbon fluxes caused by rock weathering, volcanism, and freshwater outgassing amount in total to a flux of 0.8 Pg C yr −1 and are represented by the green number. Purely land-based processes like further rock weathering, burial, and export from soils to rivers are not depicted in the scheme above. The star (*) indicates that the given accumulation number refers to a combined value for Surface Ocean and Intermediate and Deep Ocean.