A Study of the High and Mid Latitude Biogeochemistry in the Atlantic Ocean: the Influence of Surface Processes

A Study of the High and Mid Latitude Biogeochemistry in the Atlantic Ocean: the influence of surface processes

Gisle Nondal

Dissertation for the Degree of Philosophiae Doctor (PhD)

October 2010

Geophysical Institute University of Bergen Norway

“Man wants to know; and when man no longer wants to know, he will no longer be man” Fridtjof Nansen (1861-1930)

A Study of the High and Mid Latitude Biogeochemistry in the Atlantic Ocean: the influence of surface processes

Gisle Nondal

PhD thesis in Chemical Oceanography

October 2010

Geophysical Institute University of Bergen Norway

My thesis consists of an introductory section, and a collection of papers, presented in partial fulfillment of the requirements for the degree of Philosophiae Doctor in Chemical Oceanography at the Geophysical Institute, University of Bergen, Norway.

List of papers:

1. Nondal, G., R. G. J. Bellerby, A. Olsen, T. Johannessen, and J. Olafsson, 2009: Optimal

evaluation of the surface CO2 system in the northern North Atlantic using data from voluntary observing ships, Limnology and Oceanograhphy: Methods, 7, 109-118.

2. Nondal, G., A. Olsen, R. G. J. Bellerby, and P. M. Haugan, 2010; Fluxes of O2, CO2, and atmospheric potential oxygen from autonomous profiling floats. Manuscript.

3. Nondal, G., K. M. Assmann, and R. G. J. Bellerby: Signals of Climate Change in Antarctic Intermediate Water, Manuscript to be submitted to Nature.

4. Olsen, A., A. M. Omar, R. G. J. Bellerby, T. Johannessen, U. Ninnemann, K. R. Brown, K. A. Olsson, J. Olafsson, G. Nondal, C. Kivimäe, C. Neill, and S. Olafsdottir, 2006: 13 Magnitude and origin of the anthropogenic CO2 increase and C Suess effect in the Nordic Seas since 1981, Global Biogeochemical Cycles, 20, doi:10.1029/2005GB002669.

5. Riebesell, U., K. G. Shulz, R. G. J. Bellerby, M. Botros, P. Fritsche, M. Meyerhöfer, G. Nondal, A. Oschlies, J. Wohlers, and E. Zöllner, 2007: Enhanced biological carbon

consumption in a high CO2 ocean, Nature, 450, 545-549.

6. Bellerby, R. G. J., K. G. Shulz, U. Riebesell, C. Neill, G. Nondal, E. Heegaard, T. Johannessen, and K. R. Brown, 2008: Marine ecosystem community carbon and nutrient uptake stoichiometry under varying ocean acidification during the PeECE III experiment, Biogeosciences, 5, 1517-1527.

7. Nondal, G., P. M. Haugan, K. S. Seim, and R. G. J. Bellerby, 2010: Climate Science: The New Frontier, in review for Climatic Change.

Financial support for this work was provided by the Geophysical Institute at the University of Bergen, Norway, a personal endowment by Mr. Trond Mohn c/o Frank Mohn A/S, and from the 6 Research Council of Norway through the Southern Ocean Biogeochemistry Education and Research (SOBER) project. Computer time was provided by the Norwegian program for supercomputing (NOTUR).

8 Acknowledgements

The writing of this thesis would never have been possible without a lot of support from numerous people:

First of all I would like to thank the supervising committee: Peter M. Haugan, Richard Bellerby, Laurent Bertino, Karen Assmann, and Ola M. Johannessen for valuable scientific guidance, for supporting my ideas (when they were scientifically sound), and for letting me know when I was moving off track. Thank you for reading and answering all the emails I sent! I am especially grateful to Richard (and the penguins!) for boosting my motivation on Boulders Beach in 2009, and for his constant support and innovative ideas. My existing scientific network would not have been possible to create without him, and I consider him an invaluable Mentor and a very good friend.

I am highly grateful for the scientific, as well as non-scientific, discussions with my dear friend, and colleague, Knut S. Seim, and really appreciate the numerous solutions he provided to my programming and Matlab problems. “Det er jo lett, går fort å få det til!”, yeah right…

Are Olsen is acknowledged for always setting aside the time to discuss and provide help if required. His great Matlab skills are highly acknowledged as well as always inspiring co-authorship.

Thank you to all of my friends for the fun times when not doing research. I especially thank Torgeir “Best Man” Hultstrand and Roar “Balla” Frantzen for making fun of me when I was getting too serious in scientific debates outside of the office. “CO2?!? Havet kan jo ikke bli surere av det!!”. Also thank you for all the nice hunting trips where I always let you catch more than me… Morten Strøm; thank you for letting me use your brilliant office in the garage, and for many nice jogs around Askøy!

I am in infinite debt to my family for always standing behind me through all the ups and downs in my life. A special thank you goes to my grandmother Bodil Nondal, who still tells her friends that I am studying to become “a weather man”. Glad i deg Besta!!

Thanks to all the people at the Mohn-Sverdrup - and Nansen Center, the Geophysical Institute, and the Bjerknes Centre for a fun and healthy working environment!

9 Finally I would like to thank my amazing and beautiful wife Helga Gunnarsdóttir Nondal for believing in me ever since I started the University of Bergen in 1999. Couldn’t have done this without you!!

This thesis is dedicated to my kids Björk Gisladóttir Nondal and Ísak Gislason Nondal. Pabbi elska ykkur!!

Gisle Nondal Bergen, October 2010.

10 Contents

1. Introduction 12 1.1 Factors regulating oceanic carbon uptake and distribution 12 1.2 Inorganic carbon chemistry 15

1.2.1 Anthropogenic CO2 17 1.3 Physical drivers 18 1.4 Focus areas 20 1.4.1 Northern North Atlantic and the Nordic Seas 20 1.4.2 South Atlantic Ocean 21 2. Approach 22 2.1 Measurements of surface biogeochemical properties 22 2.2 Impact of ocean circulation on biogeochemistry 23 2.3 Biological sensitivity to surface modifications 24 2.4 Conclusions from the papers 25 2.5 Epilogue 25 3. Future work 26 3.1 Remineralisation in the South Atlantic twilight zone 26

3.2 Regional variability of O2, CO2, and APO fluxes in the South Atlantic 26

3.3 Evaluating the biological response to increasing CO2 with ecosystem models 26 4. Appendix I – Autonomous profiling floats 28 5. References 31

12 1. Introduction

“… [H]uman beings are now carrying out a large scale geophysical experiment … [which] … if adequately documented, may yield a far reaching insight into the processes determining weather and climate. It therefore becomes of prime importance to attempt to determine the way in which carbon dioxide is partitioned between the atmosphere, the oceans, the biosphere and the lithosphere”.

The above statement written by Revelle and Suess in 1957 about the changes in the Earth’s climate due to mankind’s emission of fossil fuels still holds true today and represents a major challenge with regards to understanding the oceanic carbon cycle, and ocean biogeochemistry. After the lithosphere the ocean is the largest global carbon reservoir, containing approximately 50 times the amount of carbon in dissolved form, as in the atmosphere (Sundquist 1985, cited by Takahashi et al. 1993). Subduction of surface waters into the deep ocean removes carbon from the atmosphere on a time scale of centuries to a few millennia, and the sinking of planktonic debris to the sediments allows the return of carbon to the lithosphere. Presently, atmospheric carbon dioxide (CO2) concentrations have exceeded levels not seen for at least 650 000 years (Siegenthaler et al. 2005), principally caused by the combination of fossil fuel combustion, cement production, and land use changes. Oceanic CO2 uptake has significantly dampened the atmospheric CO2 increase, the ocean having taken up nearly half of the CO2 emitted since the onset of the industrial revolution

(e.g. Sabine et al. 2004). The uptake of anthropogenic CO2 (Cant) has caused a lowering of the average surface ocean pH of 0.1 units, equivalent to a 30% increase in the concentration of hydrogen ions (H+). This may have adverse effects on marine life (e.g. Raven et al. 2005). In parallel, oceanic dissolved oxygen (O2) concentrations are declining (e.g. Keeling et al. 2010, and references therein); mostly due to CO2 induced global warming, and reduced ventilation. This may exert additional stress on marine life living in vulnerable areas (e.g. Stramma et al. 2008). The combined effect of decreasing pH and O2 will impact marine systems through modifications of the respiration index (Brewer and Peltzer 2009, Hofmann and Schnellenhuber 2009).

1.1 Factors regulating oceanic carbon uptake and distribution Air-sea carbon fluxes, and the distribution of carbon within the world oceans, are largely controlled by three conceptual carbon pumps (fig. 1); (i) the solubility pump, (ii) the organic carbon pump (or soft tissue pump), and (iii) the carbonate counter pump (Volk and Hoffert 1985, Heinze et al. 1991; fig. 1).

13 The solubility pump

The solubility of CO2 in seawater is a function of temperature and salinity (Weiss 1974), where cold water with relatively low salinity can hold more carbon compared to warmer and more saline water.

Thus changes in the surface heat and/or freshwater content impacts the air-sea exchange of CO2 through the solubility pump, as depicted in figure 1. The same argument is also valid for O2 (Weiss 1970).

The organic carbon pump Living organisms on our planet are chiefly made up of five primary chemical elements, hydrogen (H), carbon (C), nitrogen (N), oxygen (O), and phosphorus (P), in addition to several secondary elements and micro constituents (trace elements; see e.g. Williams 1997). At the time of Revelle and Suess, Redfield (1958) identified a tight connection between the oceanic carbon cycle and the marine biogeochemical cycling of nutrients through marine biology.

Fig. 1. Diagram showing the three oceanic carbon pumps; (i) the solubility pump, (ii) the organic carbon pump (or soft tissue pump), and (iii) carbonate counter pump. From Heinze et al. (1991).

Redfield et al. (1963) found that in the marine environment phytoplankton, through photosynthesis, take up the major elements in a relatively constant stoichiometric ratio. This can be illustrated through the following photosynthetic reaction:

- 106CO2 + 16NO3 + H3PO4 + 122H2O ⇔ (CH2O)106 (NH3)16 (H3PO4) + 138O2 (1)

The classical “Redfield ratio” can be thus identified as C : N : P : O2 = 106 : 16 : 1 : -138, implying that in order to produce organic soft tissue, and drive their metabolism, phytoplankton take up 106,

16, and 1 mole(s) of C, N, P, respectively, while at the same time 138 moles of O2 are released. This represents the surface part of the organic carbon pump (fig. 1), where phytoplankton growth drives a flux of carbon from the atmosphere to the ocean. Equation (1) has been refined since the time of Redfield (e. g. Anderson and Sarmiento 1994), and is now commonly written as:

- 106CO2 + 16NO3 + H3PO4 + 78H2O ⇔ C106H175O42N16P + 150O2, (2) expressing the “updated” Redfield ratio of C : N : P : O2 = 106 : 16 : 1 : -150.

While mostly recycled within the upper layers, the remains of planktonic organisms sink out of the surface, into the intermediate and deep waters (fig. 1) thus depleting the surface of nutrients and carbon. Only a small part of the sinking organic material is buried in the deep ocean sediments, while most of it is remineralised through respiration (Ducklow 1995, del Giorgio and Duarte 2002). Respiration is described by the following equation:

- C106H175O42N16P + 150O2 ⇔ 106CO2 + 16NO3 + H3PO4 + 78H2O (3)

It is evident from (3) that through respiration O2 is utilized and CO2 is released. This part of the biological carbon pump causes vertical carbon gradients, with higher values at depth, and lower values in the surface. In order for biological production to be maintained in the world oceans, there has to be a return path of nutrients from the intermediate and deep ocean. This has classically been explained as taking place through upwelling driven by vertical mixing in the thermocline (fig. 1). Estimates of vertical mixing (Ledwell et al. 1993), and combined modeling and observations of radiocarbon (Toggweiler and Samuels 1993), do not support this view. A more likely return path of nutrients to the surface ocean has been identified as through the upwelling of old deep waters in the Southern Ocean that then return to lower latitudes at a shallower depth as mode and intermediate waters (Sarmiento et al. 2004, Williams et al. 2006). When the old, carbon-rich water masses come in contact with the atmosphere, this will favor a flux of CO2 from the ocean to the atmosphere.

The carbonate counter pump Apart from the, previously mentioned, metabolism and formation of soft tissue (through eq. 2), different planctonic species also produce hard parts (i.e. shells or skeletons), using silicate or calcareous material. Production through silicate (SiO4), by diatoms, does not directly influence the cycling of carbon (Heinze et al. 1991). Calcifying organisms (mainly coccolithophorids, foraminifera, and pteropods), on the other hand, build their shells or skeletons from calcium 15 carbonate (CaCO3), a process releasing CO2 (e.g. Frankignoulle and Canon 1994). This can be described by the following equation:

2+ - Ca + 2HCO3 ⇔ CaCO3 + CO2 + H2O (4)

The described mechanism represents the upper right-hand side of the carbonate counter pump, shown in figure 1 (Heinze et al. 1991), also called the alkalinity pump (Riebesell et al. 2009). As for soft tissue, a portion of the planktonic hard parts sink to depth, representing a CaCO3 flux from the surface to the deep ocean (fig. 1). In the deep layers, CaCO3 re-dissolves, causing an alkalinity increase in the respective water mass. When this water comes in contact with the atmosphere it has a greater potential to take up CO2. However this process operates over longer time scales (Boyle 1988, Heinze et al. 1991).

1.2 Inorganic carbon chemistry

When CO2, in gas phase (CO2 (g)), dissolves in seawater it first hydrates to form aqueous CO2 (CO2

(aq)) before reacting with water (H2O) to form carbonic acid (H2CO3):

CO2 (g) ⇒ CO2 (aq) (5)

CO2 (aq) + H2O (l) ⇔ H2CO3 (aq) (6)

The concentrations of CO2 (aq) and H2CO3 (aq) are commonly joined together into the hypothetical * species H2CO3 , as they are very difficult to discern analytically (DOE 1994). The dissolution of

CO2 in seawater can then be represented by the following chemical reactions:

* CO2 (g) + H2O (l) ⇔ H2CO3 (aq) (7) * + - H2CO3 (aq) ⇔ H(aq) + HCO3 (aq) (8) - + 2- HCO3 (aq) ⇔ H (aq) + CO3 (aq) (9)

* - The relative proportions of the main chemical compounds (H2CO3 ), bicarbonate (HCO3 ), and 2- carbonate ion (CO3 ) in seawater can be described by a so-called Bjerrum plot (fig. 2). Under current ocean conditions bicarbonate is the most abundant form of CO2 in seawater (~91%), * followed by carbonate (~8%) and H2CO3 (~1%).

16 The seawater carbonate system is fully described by the four parameters dissolved inorganic carbon

(Ct), Total alkalinity (At), fugacity of CO2 (fCO2), and pH.

Ct represents the sum of the three major carbon compounds in seawater: * - 2- Ct = [H2CO3 ] + [HCO3 ] + [CO3 ] (10)

* - 2- Fig. 2. Plot showing concentrations of H2CO3 , here labeled CO2 (blue), HCO3 (red), and CO3 (purple), as functions of pH. The green arrow indicates the range of pH likely to be found in the oceans now and in the future. From Raven et al. (2005).

At is a measure of the excess of proton acceptors (bases) over proton donors (acids) (Dickson 1981), and can be described by the following equation (DOE 1994):

- 2- - - 2- - - At = [HCO3 ] + 2[CO3 ] + [B(OH)3 ] + [OH ] +2 [PO4 ] + [SiO(OH)3 ] + [NH3] + [HS ] + … + - – [H ] – [HSO4 ] – [HF] – [H3PO4] - … (11) where acid or base species having small influence on the alkalinity have been neglected.

* fCO2 is determined by the concentration of H2CO3 through the following relationship:

[]H CO* = 2 3 fCO2 (12) K0

17 where K0 represents the solubility of CO2 in seawater (Weiss 1974). If the non-ideality of CO2 is neglected the term partial pressure of CO2 (pCO2) is used. pH is a dimensionless measure of the acidity of a solution (e.g. Raven et al. 2005), defined as the negative logarithm of the concentration of hydrogen ions (H+):

+ pH = -log10[H ] (13)

Depending on the definition of the total H+ concentration ([H+]), oceanic pH is reported on different scales (Dickson 1984).

The knowledge of any combination of two out of the four carbon system variables enables the calculation of the remaining two, by means of thermodynamic equations (e. g. Lewis and Wallace 1998).

+ It is evident from (7) – (9) that when dissolving CO2 in seawater, the H concentration is increasing, thus through (13) the pH is decreasing, a process termed “ocean acidification”. This term does not mean that the ocean will become acidic, however it represents a shift towards lower pH. As seen from (7) – (9) combined with (10), perhaps the term dissolved inorganic carbon enrichment is more descriptive (Hutchins et al. 2009).

Seawater is able to act as a buffer for CO2, a process described by the following equation:

2- - CO2 (aq) + H2O + CO3 ⇔ 2HCO3 (14)

It is thus evident that dissolving CO2 represents a shift in the seawater carbonate chemistry, towards less carbonate. This influences the ability of calcifying organism to build their shells or skeletons, through lowering the saturation state for CaCO3 (Gattuso et al. 1998, Langdon et al. 2000, Riebesell et al. 2000, Feely et al. 2004).

1.2.1 Anthropogenic CO2

The concept of anthropogenic CO2 (Cant) is used extensively in this thesis and thus warrants a brief introduction.

In preindustrial times, the air-sea flux of CO2 was driven by biology and meridional water mass transport. Since the onset of the industrial revolution, an additional CO2 load has been added to the 18 atmosphere by human activities. This is often referred to as anthropogenic CO2, or Cant. The anthropogenic CO2 flux is driven by this atmospheric CO2 increase which has perturbed the natural exchange of CO2 between the ocean and the atmosphere. While the ocean is believed to have been a source of carbon of 0.4±0.2 Pg C yr-1 preindustrially, it takes up 2.0±1.0 Pg C yr-1 of anthropogenic

CO2 in a reference year 2000 (Takahashi et al. 2009). Cant is not a directly measurable quantity, however the oceanic Cant uptake can be diagnosed by two different approaches (Wallace 1995):

(i) repeat biogeochemical surveys in order to investigate Ct changes over time, and (ii) using back- calculation techniques to separate Cant from the large natural background (e.g. Gruber et al. 1996).

1.3 Physical drivers As pointed out by Heinze et al. (1991), the above-described carbon cycle pumps are tightly intertwined with the redistribution of carbon, nutrients, oxygen, temperature, and salinity within the global overturning circulation (fig. 3). The Atlantic Ocean plays a key role in this circulation system, and thus for global climate (Broecker 1997, Ganaschaud and Wunch 2003, Knorr and Lohmann 2003, Toggweiler and Russel 2008). North Atlantic Deep Water (NADW) is formed in the high latitude North Atlantic (fig.3; e.g. Killworth 1983), preconditioned by watermasses entering the South Atlantic (Gordon et al. 1992, Gordon 1996, Iudicone et al. 2008) through a combination of the “cold-water path” through Drake Passage (Rintoul 1992), and the “warm water path” through the Agulhas Current (Gordon 1986, Knorr and Lohmann 2003, Biastoch et al. 2008).

This deep water formation process plays an important role in the uptake, and sequestration of Cant, together with mode and intermediate water formation in the south Atlantic and Southern Ocean (Sabine et al. 2004, Khatiwala et al. 2009, Ito et al. 2010; fig. 4). Subantarctic Mode Water (SAMW) is a layer of nearly uniform density, thought to originate in a relatively uniform manner from the thick wintertime mixed layers encircling the Southern Ocean (McCartney 1977). However, recent work (Sallée et al. 2010) suggests a more localized formation in areas of intense subduction.

SAMW has classically been identified as the precursor for Antarctic Intermediate Water (AAIW), a water mass found in the layer immediately below SAMW (McCartney 1977). In situ float observations in the Southeast Pacific (Schneider and Bravo 2006) question this hypothesis, and AAIW was recently proposed to form from winter water originating from the Bellingshausen Sea (Garabato et al. 2009).

Fig. 3 Schematic illustration of the global overturning circulation, centered on the Atlantic Ocean. From Kuhlbrodt et al. 2007.

-2 Fig. 4 Column inventory of anthropogenic CO2 in the ocean (mol m ). High inventories are associated with deep water formation in the North Atlantic and mode and intermediate water formation in the Southern Ocean. From Sabine et al. (2004).

20 1.4 Focus areas The topic of this thesis is to investigate the impact of surface ocean processes on biogeochemistry in the high and mid latitude Atlantic, thus focusing on the northern North Atlantic, and the Nordic Seas, and the Southern Ocean sector of the South Atlantic.

1.4.1 Northern North Atlantic and the Nordic Seas The Nordic Seas (Drange et al. 2005; marked by N in fig. 3), located in the northernmost North Atlantic, consists of the Greenland, Iceland, and Norwegian Seas. This area contains a significant part of the northern limb of the overturning circulation; with its near surface circulation dominated by northward flow of warm and saline Atlantic Water, and a southward flow of relatively fresh and cold polar water from the Arctic Ocean (fig. 5). On its way north, the Atlantic surface water experiences cooling, thus gains density, permitting mixing to greater depths. Additionally, through the solubility pump, this enables carbon uptake from the atmosphere (Skjelvan et al. 2005). Model studies have identified a large Cant sink in the northern North Atlantic and Nordic Seas (Orr et al. 2001, Wetzel et al. 2005), however other studies suggest that the largest uptake occurs further south, so that the waters entering the respective area have already equilibrated with the atmospheric

CO2 concentration en route, thus having little potential for further CO2 uptake (Álvarez et al. 2003, Macdonald et al. 2003, Roson et al. 2003).

Fig. 5 Surface circulation in the Nordic Seas. The warm and salty Atlantic Water is shown in red while the cold and fresh Arctic water is shown in blue. M represents the location of ocean weather ship station M. Courtesy Svein Østerhus, Geophysical Institute, University of Bergen. 21

The surface decrease in pH caused by the invasion of atmospheric CO2 has been estimated to about 0.1 units since 1750 (Wallace 2001, Skjelvan et al. 2005), with projected levels at the end of this century up to 0.35 units lower than present (Wallace 2001, Bellerby et al. 2005).

1.4.2 South Atlantic Ocean The South Atlantic Ocean (e.g. Wefer et al. 1996) serves as a medium through which NADW is transported southwards, and introduced to the other ocean basins (e.g. Rintoul 1991). This circulation suggests that water must return northward in order to balance the flow. The general upper ocean circulation features of the South Atlantic, together with its connections with the Pacific, Indian, and Southern Oceans, are shown in fig. 6. It is evident that the return flow stems from water of Pacific -, Indian -, as well as Southern Ocean origin, the main water masses of the return flow being mode and intermediate waters, and Antarctic Bottom Water (Iudicone et al. 2008).

Fig. 6 The arrows indicate the mean surface and intermediate circulation in the Atlantic and Indian Oceans. The blue X represents the “cold water path” through Drake Passage (Rintoul 1991) while the red X represents the “warm water path” (Gordon 1986, Gordon et al. 1992) via Agulhas Leakage. The cross-hatched areas are areas of strong mixing and eddy activity, the Brazil-Malvinas Confluence Zone shown to the left and the Agulhas Retroflection to the right. The circulation scheme was inspired by Gordon et al. (1992), Gordon (1986,2003), Stramma and Peterson (1990), Williams et al. (2006), Hanawa and Talley (2001) and Lutjeharms (2007).

In addition to serving a important conduits of Cant from the atmosphere to the interior ocean (Ito et al. 2010), the Southern Ocean intermediate waters are sensitive indicators of climate change and 22 variability (Banks et al. 2000). They distribute nutrients to the World Ocean and influence global ocean productivity (Sarmiento et al. 2004, Williams et al. 2006), and oxygenate the intermediate ocean countering the oxygen utilization of marine microbial processes. These water masses have experienced a significant warming (Gille 2002), increasing Cant (Sabine et al. 2004, Khatiwala et al. 2009), and regionally diminishing dissolved oxygen concentrations (Matear et al. 2000) presumably in response to climate change.

2. Approach In order to improve the knowledge regarding the impact of surface ocean processes on biogeochemistry, this thesis focuses on three approaches. (i) developing, and testing, methods for obtaining high quality in situ measurements, (ii) understanding the influence of ocean circulation on the dispersion of surface biogeochemical signals, and (iii) the potential biological feedbacks to changing biogeochemical settings. These three points will be discussed in detail below.

2.1 Measurements of surface biogeochemical properties In order to improve the current knowledge of surface ocean biogeochemistry, it is of prime importance to increase the number of measurements, either for stand-alone use, or to be used to constrain biogeochemical models. In Paper 1 we investigate the potential of using measurements from Voluntary Observing Ships (VOS; fCO2, sea surface temperature SST, sea surface salinity SSS) to obtain an accurate description of the full surface carbon system. VOS lines are well established in the North Atlantic, and have provided a wealth of insight regarding air-sea exchange of CO2 in the region (e.g. Watson et al. 2009). We employ a high-density carbon system database for the surface northern North Atlantic to develop new relationships between At and SSS, and Ct,

SSS, SST and nitrate. The optimal method for obtaining a full description of the CO2 system was determined to be the prediction of surface At using SSS, and combine this with in situ fCO2 to calculate the surface Ct distribution. Evaluation using independent datasets shows that surface Ct may be calculated with a mean bias of -1.0 μmol kg-1 and a standard error of calculation of 7.4μmol kg-1. The described method, combined with remotely sensed SSS and SST, has the potential to provide estimates of northern North Atlantic acidification, and Cant uptake, from space.

Paper 2 utilizes 2.5 years of surface temperature, salinity, and O2 measurements obtained by two Navigating European Marine Observer floats (NEMO) profiling floats (appendix I). Combined with atmospheric data (sea level pressure and 10m wind speed) from Kanamitsu et al. (2002), the O2 measurements enable direct calculations of the air-sea exchange of O2 along the float drift tracks.

By employing the LDEO surface pCO2 database (Takahashi et al. 2010) we identify seasonal

(summer and winter) relationships between surface pCO2 normalised to 5°C and the nominal year 2008, and SST, SSS, lon, and lat. These relationships were validated against an independent dataset 23

(Boutin 2008), and enable the estimation of surface pCO2 using float SST, SSS, and position.

Together with atmospheric data this enables evaluation of the air-sea flux of CO2 for 2008.

Combining the O2 and CO2 fluxes, we are able to calculate fluxes of atmospheric potential oxygen (APO; Stephens et al. 1998) across the float drift tracks, an important variable for improving atmospheric inversion studies (e.g Rödenbeck et al. 2008). If increasing the number of APO measurements, this parameter my also offer a valuable means for evaluating ocean biogeochemical models (Naegler et al. 2007). Although this paper demonstrates the versatility of profiling floats equipped with O2 sensors, it promotes the parallel use of CARIOCA buoys with O2 sensors. The optimal approach would be floats equipped with both CO2 and O2 sensors, as this would provide knowledge about the mixed layer dynamics as well, compared to the purely surface measurements of CARIOCA.

2.2 Impact of ocean circulation on biogeochemistry Ocean circulation has been identified as vital in the redistribution of carbon, nutrients, oxygen, temperature, and salinity in the global ocean. Paper 3 investigates potential biogeochemical changes to the properties of Antarctic Intermediate Water in the South Atlantic Ocean. We utilize high quality measurements of oceanic dissolved oxygen (O2) from 2008, obtained by means of two

NEMO floats (appendix I), which are compared to O2 measurements obtained in 1988 during the South Atlantic Ventilation Experiment (SAVE). The comparison revealed the fastest decrease in the oxygen content of Southern Hemisphere intermediate waters observed to date (6.5μmol kg-1 per decade over the past two decades). A model experiment using a state-of-the-art ocean circulation – carbon cycle model over the same time period shows that this oxygen reduction extends throughout the Southeastern Pacific and South Atlantic. The main circulation feature of the Southern Ocean is the Antarctic Circumpolar Current (ACC) flowing in an eastward manner around Antarctica, primarily driven by the westerly winds (e.g. Toggweiler and Russel 2008). The westerly winds have strengthened and shifted southward over the past 50 years, potentially due the combined effect of

CO2 induced warming and stratospheric ozone loss (Toggweiler 2009, and references therein). The westerlies are now located more squarely over the ACC, thus able to do more work in driving the current (Saenko et al. 2005, Toggweiler and Russel 2008). As a response to the shifting westerlies, the model displays increased northward water mass transport in the near surface layers, leading to enhanced upwelling south of 50°S. This is enhancing the supply of Circumpolar Deep Water from the ocean interior. Combined with a reduction in wintertime convection, the model simulates formation of AAIW with more pristine CDW characteristics, hence lower O2, higher Ct, and more nutrients (here represented by silicate). This signal will potentially be transported to the global oceans through the AAIW circulation.

24 In Paper 4, carbon system measurements from an extensive survey of the northern North Atlantic in 2002/2003 are compared to measurements from data obtained during the Transient Tracers in the Ocean, North Atlantic Study (TTO-NAS) in 1981, by means of an extended multilinear regression approach. The paper provides new knowledge regarding pathways of anthropogenic carbon into the Nordic Seas. Polar Water entering the Nordic Seas from the Arctic Ocean is undersaturated with respect to anthropogenic CO2, thus promoting uptake of CO2 from the atmosphere, while Atlantic Water entering from the south appears to be equilibrated.

2.3 Biological sensitivity to surface modifications

Compared to the total biological CO2 uptake, marine phytoplankton is responsible for nearly 50% of the uptake (Field et al. 1998). Understanding the potential effects of inorganic carbon enrichment, and concomitant ocean acidification, on the biological carbon pump is therefore of significant importance. Papers 5 and 6 are based on measurements obtained from a mesocosm carbon enrichment study conducted in Raunefjorden (60.3°N, 5.2°E) outside of Bergen, Norway in 2005. The carbon system in nine plastic enclosures (~27m3), called mesocosms, was manipulated, by CO2 aeration, to obtain triplicates of three CO2 concentrations, 350 μatm (1xCO2), 700 μatm

(2xCO2), and 1050 μatm (3xCO2). After adding nutrients to the system, the development, and subsequent decline, of a phytoplankton bloom was monitored over a 24-day period. The papers present results showing that the community consumed 27% and 39% more dissolved inorganic carbon at increased CO2, compared to present levels, whereas nutrient uptake remained unchanged.

The elemental ratio of C and N increased from 6.0 at low CO2 to 8.0 at high CO2, thus exceeding the Redfield ratio of 6.6 (106:16) in today’s ocean. Paper 6 also include the study of potential acidification effects on diatom silification, an effect very rarely studied (Hutchins et al. 2009). Although the formation of opaline material by diatoms does not directly influence the carbon cycle (Heinze et al. 1991), such organisms are responsible for about one-fifth of the photosynthesis on earth (Nelson et al. 1995), and generate most of the organic matter serving as food for marine life

(e. g. Armbrust 2009). Silicate drawdown was virtually identical across CO2 treatments, suggesting that the silicate cycle was not affected by the carbon enrichment (and subsequent acidification).

25 2.4 Conclusions Through this thesis, various aspects of the high and mid latitude Atlantic Ocean biogeochemistry have been investigated. The main conclusions from the papers can be summarized as:

• We have shown that measurements of fCO2 and sea surface salinity on Voluntary

Observing Ships can provide an accurate description of the full CO2 system.

• The versatility of using autonomous profiling floats for in situ ocean observing has been highlighted. We have shown that such platforms can provide flux estimations of

O2 (through direct measurements), CO2 (through empirical relationships), and that this combination provides flux estimates of APO.

• The fastest decrease in the oxygen content of Southern Hemisphere intermediate waters, to date, has been identified (6.5μmol kg-1 per decade over the past two decades). The cause of the reduction has, by means of a biogeochemical model, been traced to changing ocean circulation as a response to changing westerly winds.

• Polar Water, entering the Nordic Seas from the north was found to be undersaturated

with respect to atmospheric Cant, promoting local uptake of Cant within the Nordic Seas. On the other hand, Atlantic Water entering from the south appeared to be

equilibrated. This suggested that there is no room for further uptake of Cant in the parts of the Nordic Seas, which is dominated by Atlantic Water.

• Increased carbon uptake relative to nitrogen, as a response to Ct enrichment (ocean acidification), was documented in a natural plankton community for the first time. The silification process in diatoms did not seem to be affected by the changing carbon chemistry.

2.5 Epilogue The results, and conclusions, from the papers presented in this thesis are relevant to our understanding of climate. Fuelled by the “climategate” and “glaciergate” incidents we discuss, in Paper 7, two conceptual models describing the relationship, and interface, between science and policy with special focus on climate science. In this paper it is suggested that in order for climate science to recover from the present legitimacy crisis, climate science should go through a democratization, guided by the philosophy of Post Normal Science (Funtowicz and Ravetz 1990). We also point at examples of how the scientific community is already developing this methodology. 26 3. Future work Several aspects of the datasets, and results, presented in this thesis deserves to be further investigated.

3.1 Remineralisation in the South Atlantic twilight zone A significant part of the remineralisation occurs in the “twilight zone”: the layer underlying the euphotic zone and extending down to about 1000m (e.g. Buesseler and Boyd 2009). Traditionally, most of this was thought to take place in the upper mixed layer, with modest contribution of intermediate waters (e. g. Suess 1980, del Giorgio and Duarte 2002). However the intermediate waters are now thought to play a much more pronounced role in the remineralisation of organic matter (Moriarty and O’Donohue 1995, Vidal et al. 1999, del Giorgio and Duarte 2002). Recent work demonstrate the versatility of using autonomous floats for quantifying oceanic metabolism (Martz et al. 2008), and furthermore the floats used in Papers 2 and 3 were shown to capture the evolution of AAIW across the South Atlantic, a water mass in which a large fraction of the remineralisation in the Australian sector of the Southern Ocean is suggested to occur (Moriarty and O’Donohue 1995).

3.2 Regional variability of O2, CO2, and APO fluxes in the South Atlantic The manuscript outlined in Paper 2 utilizes the climatological frontal locations of Orsi et al. (1995) for the major South Atlantic Ocean fronts. Compared to Boutin et al. (2008) this is likely not representative for the situation as described by the NEMO floats (their fig. 1). Detailed frontal locations should thus be identified, following the methods of Boutin et al. (2008), in order to accurately separate the NEMO measurements obtained in the subantarctic zone (SAZ) and the polar zone (PZ). In addition, the use of satellite derived wind products like QuickScat (http://cersat.ifremer.fr/data/discovery/by_parameter/ocean_wind/mwf_quikscat), might improve the surface wind field compared to reanalysis products. This also deserves future investigation.

3.3 Evaluating the biological response to increasing CO2 in ecosystem models A recent paper by Oschlies et al. (2008) investigated the effect of enhanced carbon-to-nitrogen drawdown as identified in Papers 5 and 6, on global biogeochemical cycles. They employed a relatively simple marine ecosystem model having two major nutrients (nitrate and phosphate) and two phytoplankton functional groups. Increased C : N drawdown has recently been included in the European Regional Seas Ecosystem Model (ERSEM; e. g. Blackford et al. 2004) through the EU 7th framework programme Marine Ecosystem Evolution in a Changing Environment (MEECE; Bellerby et al. 2010). ERSEM is a mature plankton functional type model related to the NPZD type models. It represents the key processes of temperate shelf ecosystems; plankton community 27 complexity, the microbial loop, variable nutrient stoichiometry, variable carbon : chlorophyll ratios, and a comprehensive description of benthic biochemical and ecological processes (http://web.pml.ac.uk/meece/library/ersem.html). The units of currency of ERSEM are Carbon,

Nitrogen, Phosphorus, Silicon and O2. ERSEM has been coupled to the 1D General Ocean Turbulence Model (GOTM; http://www.gotm.net/), providing information on Temperature (T), Salinity (S), and mixing. The effect of enhanced C : N could therefore be tested, through the describe model pairing, on the extensive hydrographical and biogeochemical timeseries from ocean weather station M (66°N, 2°E; marked by M in fig. 5).

28 4. Appendix I Autonomous profiling floats In Papers 2 and 3 we utilize measurements obtained from autonomous profiling floats. As this technology is only mentioned briefly in the papers, this appendix provides a more thorough description.

4.1 Float technology The use of autonomous profiling floats for oceanic research started in the 1950’s, with the introduction of the “Swallow float” (Swallow 1955, Davis et al. 2001). Having varied in complexity from simple floats passively drifting with the ocean currents enabling measurements of current speed, the state-of-the-art floats are able to profile the water column (typically down to 1000 or 2000m) for up to 4 years. The Navigating European Marine Observer (NEMO) profiling float, manufactured by Optimare Sensorsysteme AG (http://www.optimare.de/cms/en/divisions/mms/mms-products/nemo.html) was employed in this work. The floats were programmed to drift at 650 m depth for ten days before descending to 1000m and subsequently rise to the surface while measuring T, S, and O2. At the surface they determine their position by onboard GPS, before transmitting data through Iridium satellite communication (fig. 7).

Fig. 7 Schematic showing a typical ten-day cycle for a NEMO float. Inspired by http://www.argo.ucsd.edu/pnp.html.

The GPS – Iridium pairing enables more data to be transmitted at a shorter period, limiting float surface exposure. Additionally, it provides a more accurate determination of the float location, compared to the majority of the floats, using the ARGOS system (http://www.argos-system.org/) 29 for both communication and localization. The NEMO floats move up and down in the water column by changing their volume, utilizing the principle of neutral buoyancy. This is achieved by pumping mineral oil from an internal reservoir into an external bladder. NEMO is highly suitable for polar applications by the inclusion of an ice detection algorithm. On the ascent, the float computes the median temperature of the seven near surface measurements (≥ 20 dbar). If the median is less than, or equal to, -1.79°C, the surface is likely to be ice covered and the float aborts the ascent, storing the data internally until next possible transmission. This prevents the float hitting the underside of the potential ice, damaging the antenna, and sensors, mounted on its top cap.

4.2 Sensor description Temperature, salinity, as well as pressure measurements were performed by means of a Seabird Electronics Inc. (SBE) 41 Microcat Conductivity Temperature Depth (CTD) sensor, specifically designed for the use on profiling floats (http://www.seabird.com/alace.htm). This sensor enables measurements with the following accuracies: T – 0.002°C; S – 0.002; and pressure – 2 dbar. O2 was measured by means of the Aanderaa Data Instruments (AADI) Oxygen Optode model 3830 (Tengberg et al. 2006). It has been proven suitable for aquatic use (Tengberg et al. 2006), as well as for use on autonomous floats (Körtzinger et al. 2004, 2005). Increasing pressure causes the O2 measurements to drop by 3.2 % per 1000 dbar, however this effect is linear, and fully reversible (Uchida et al. 2008). In addition, a salinity correction is required as the optode reports values as if submerged in fresh water (S=0). The correction to in situ salinity is performed through an equation including both in situ temperature and salinity.

4.3 Extended optode calibration The oxygen optodes are delivered having a nominal accuracy of 8 μmol kg-1 or 5% whichever is greater, as obtained through a batch calibration by the manufacturer. Field studies however indicate that this accuracy can be improved (i.e. Körtzinger et al. 2005, Tengberg et al. 2006). We thus performed an extended optode O2 calibration prior to float integration. This was done using water from Byfjorden outside of Bergen, Norway, manipulated to temperatures, and O2 concentrations, which the floats were expected to encounter in the specific study area. As the extended calibration was performed using temperature measurements from the integrated temperature sensors on the optodes, this promotes the use of optode temperaturte when adjusting to in situ salinity. NEMO60 -1 -1 initially measured O2 with a mean offset of -2.2 μmol kg , and an rms deviation of 4.0 μmol kg , while NEMO47 measured with an initial mean offset of -1.2 μmol kg-1, and an rms deviation of 4.1 μmol kg-1. To correct for the offset linear equations were fitted to the data, with the best equations of the form:

30 NEMO60 in situ 2 O2 = 0.969*O2 + 11.432, r = 0.9992 (A1)

NEMO47 in situ 2 O2 = 0.957*O2 + 13.919, r = 0.9994 (A2)

-1 After the individual corrections, NEMO60 measured O2 with a mean offset of -0.007 μmol kg and an rms error of 2.2 μmol kg-1, while NEMO47 with 0.002 μmol kg-1, and 1.8 μmol kg-1 respectively.

In order to ensure high quality O2 measurements following float deployment, comparison to in situ measurements from independent datasets is required. Profile 36 of NEMO60 warranted comparison with profile 91_1 from the R/V Polarstern ANT24 cruise in late 2007/early 2008, as these stations overlapped. Ideally such comparison should be performed on the deep-water masses (>2000m), however the floats only enveloped the top 1000m. The comparison was thus performed on data from 700-1000m, identified as Antarctic Intermediate Water (Nondal et al. manuscript to be submitted). In order to account for internal-wave movements, the density referenced to 3000m was calculated, and a second order polynomial was fitted to the data (e.g. Lamb et al. 2002; not shown) before performing the comparison. This analysis showed that NEMO60 was measuring O2 with amean bias of -1.9 μmol kg-1, and an rms error of 1.9 μmol kg-1, suggesting negligible sensor drift over, at least, one year. As no independent dataset for evaluating the performance of NEMO47 was identified, we compared its measurements to the corresponding (spatial and temporal) measurements from NEMO60. Three locations suitable for such crossover analysis were identified, corresponding to profiles number 1 (15.01.08), 2 (25.01.08), and 6 (05.03.08) respectively. The analysis was performed as described above, with the results shown in table A1.

Table A1. Results from inter-float comparison.

Location Mean bias σrms 1 4.6 4.9 2 2.2 2.4 6 1.1 1.9 Average 2.6 3.1

Assuming negligible water mass differences at the respective locations, NEMO47 was able to -1 measure O2 with an average mean bias of 2.6 μmol kg , and an average rms deviation (σrms) of -1 3.1μmol kg , when compared to NEMO60 Although higher mean bias, and σrms compared to

NEMO60, no correction was applied to the NEMO47 O2. This will possibly be done when a suitable shipborne comparison has been performed. 31 5. References Alvarez, M., A. F. Rios, F. F. Perez, H. L. Bryden and G. Roson (2003). "Transports and budgets of total inorganic carbon in the subpolar and temperate North Atlantic." Global Biogeochemical Cycles 17(1). Anderson, L. A. and J. L. Sarmiento (1994). "Redfield Ratios of Remineralization Determined by Nutrient Data-Analysis." Global Biogeochemical Cycles 8(1): 65-80. Armbrust, E. V. (2009). "The life of diatoms in the world's oceans." Nature 459(7244): 185-192. Banks, H. T., R. A. Wood, J. M. Gregory, T. C. Johns and G. S. Jones (2000). "Are observed decadal changes in intermediate water masses a signature of anthropogenic climate change?" Geophysical Research Letters 27(18): 2961-2964. Bellerby, R., G. Nondal, M. Butenscön, Y. Artioli and J. Blackford (2010). "D2.5 Sub-model for

carbon enhancement in marine diatoms relative to pCO2." http://www.meece.eu/documents/deliverables/WP2/D2.5.pdf.

Bellerby, R. G. J., A. Olsen, T. Furevik and L. G. Anderson (2005). Response of the CO2 system in the Nordic Seas and northern North Atlantic to climate change. AGU GEophysical Monograph. 158. Biastoch, A., C. W. Boning and J. R. E. Lutjeharms (2008). "Agulhas leakage dynamics affects decadal variability in Atlantic overturning circulation." Nature 456(7221): 489-492. Blackford, J. C., J. I. Allen and F. J. Gilbert (2004). "Ecosystem dynamics at six contrasting sites: a generic modelling study." Journal of Marine Systems 52(1-4): 191-215.

Boutin, J., L. Merlivat, C. Henocq, N. Martin and J. B. Sallee (2008). "Air-sea CO2 flux variability in frontal regions of the Southern Ocean from CARbon Interface OCean Atmosphere drifters." Limnology and Oceanography 53(5): 2062-2079. Boutin, J. (2008): "Underway carbon dioxide measurement results in the Southern Ocean of Carioca buoy CARIOCA_20060127. " Laboratoire d´Oceanographie Dynamic et du Climat, Paris, doi:10.1594/PANGAEA.728794 Boyle, E. A. (1988). "The role of vertical chemical fractionation in controlling late quaternary atmospheric carbon dioxide." Journal of Geophysical Research 93(C12): 15701-15714. Brewer, P. G. and E. T. Peltzer (2009). "Limits to Marine Life." Science 324(5925): 347-348. Broecker, W. S. (1997). "Thermohaline circulation, the Achilles heel of our climate system: Will

man-made CO2 upset the current balance?" Science 278(5343): 1582-1588. Buesseler, K. O. and P. W. Boyd (2009). "Shedding light on processes that control particle export and flux attenuation in the twilight zone of the open ocean." Limnology and Oceanography 54(4): 1210-1232. 32 Davis, R. E., J. T. Sherman and J. Dufour (2001). "Profiling ALACEs and other advances in autonomous subsurface floats." Journal of Atmospheric and Oceanic Technology 18(6): 982-993. del Giorgio, P. A. and C. M. Duarte (2002). "Respiration in the open ocean." Nature 420(6914): 379-384. Dickson, A. G. (1981). "An Exact Definition of Total Alkalinity and a Procedure for the Estimation of Alkalinity and Total Inorganic Carbon from Titration Data." Deep-Sea Research Part a- Oceanographic Research Papers 28(6): 609-623. Dickson, A. G. (1984). "pH scales and proton-transfer reactions in saline media such as sea water." Geochimica et Cosmochimica Acta 48(11): 2299-2308. DOE (1994). Handbook of methods for the analysis of the various parameters of the carbon dioxide system in seawater. ORNL/CDIAC-74. A. G. Dickson and C. Goyet. Version 2. Drange, H., T. Dokken, T. Furevik, R. Gerdes and W. H. Berger (2005). The Nordic Seas: An integrated perspective. Washington DC, AGU Monograph 158, American Geophysical Union. Ducklow, H. W. (1995). "Ocean Biogeochemical Fluxes - New Production and Export of Organic- Matter from the Upper Ocean." Reviews of Geophysics 33: 1271-1276. Feely, R. A., C. L. Sabine, K. Lee, W. Berelson, J. Kleypas, V. J. Fabry, and F. J. Millero (2004).

"Impact of anthropogenic CO2 on the CaCO3 system in the oceans." Science 305: 362-366. Field, C. B., M. J. Behrenfeld, J. T. Randerson and P. Falkowski (1998). "Primary production of the biosphere: Integrating terrestrial and oceanic components." Science 281(5374): 237-240. Frankignoulle, M. and C. Canon (1994). "Marine Calcification as a Source of Carbon-Dioxide -

Positive Feedback of Increasing Atmospheric CO2." Limnology and Oceanography 39(2): 458-462. Funtowicz, S. and J. Ravetz (1990). Uncertainty and quality in Science for policy., Kluwer Academic Publishers. Ganachaud, A. and C. Wunsch (2003). "Large-scale ocean heat and freshwater transports during the World Ocean Circulation Experiment." Journal of Climate 16(4): 696-705. Garabato, A. C. N., L. Jullion, D. P. Stevens, K. J. Heywood and B. A. King (2009). "Variability of Subantarctic Mode Water and Antarctic Intermediate Water in the Drake Passage during the Late-Twentieth and Early-Twenty-First Centuries." Journal of Climate 22(13): 3661-3688. Gattuso, J. –P., M. Frankignoulle, I. Bourge, S. Romaine and R. W. Buddemeier (1998). "Effect of calcium carbonate saturation of seawater on coral calcification." Glob. Planet. Change 18: 37-46. Gille, S. T. (2002). "Warming of the Southern Ocean since the 1950s." Science 295(5558): 1275- 1277. 33 Gordon, A. L. (1986). "Inter-Ocean Exchange of Thermocline Water." Journal of Geophysical Research-Oceans 91(C4): 5037-5046. Gordon, A. L. (1996). "Oceanography - Communication between oceans." Nature 382(6590): 399- 400. Gordon, A. L. (2003). "Oceanography - The brawniest retroflection." Nature 421(6926): 904-905. Gordon, L. A., R. F. Weiss, W. M. Smethie, Jr. and M. J. Warner (1992). "Thermocline and intermediate water communication between the South Atlantic and Indian Oceans." Journal of Geophysical Research 97(C5): 7223-7240. Gruber, N., J. Sarmiento, and T. F. Stocker, (1996). "An improved method for detecting

anthropogenic CO2 in the oceans." Global Biogeochemical Cycles 10(4): 809-837. Hanawa, K. and L. D. Talley (2001). Mode Waters. Ocean Circulation and Climate: Observing and Modelling the Global Ocean. G. Siedler, J. Church and J. Gould. San Diego, California, Academic: 373-386.

Heinze, C., E. Maier-Reimer and K. Winn (1991). "Glacial pCO2 reduction by the World Ocean: experiments with the Hamburg Carbon Cycle Model." Paleoceanography|Paleoceanography 6(4): 395-430. Hofmann, M. and H.-J. Schnellnhuber (2009). "Ocean acidification affects marine carbon pump and triggers extended marine oxygen holes." PNAS 106(9): 3017-3022. Hutchins, D. A., M. R. Mulholland and F. X. Fu (2009). "Nutrient Cycles and Marine Microbes in a CO2-Enriched Ocean." Oceanography 22(4): 128-145. Ito, T., M. Woloszyn and M. Mazloff (2010). "Anthropogenic carbon dioxide transport in the Southrn Ocean driven by Ekman flow." Nature 463: 80-84. Iudicone, D., S. Speich, G. Madec and B. Blanke (2008). "The global conveyor belt from a Southern Ocean perspective." Journal of Physical Oceanography 38(7): 1401-1425. Kanamitsu, M., W. Ebisuzaki, J. Woollen, S.-K. Yang, J. J. Hnilo, M. Fioorino and G. L. Potter (2002). "NCEP-DEO AMIP-II Reanalysis (R-2)." Bul. of the Atmos. Met. Soc.: 1631-1643. Keeling, R. F., A. Körtzinger and N. Gruber (2010). "Ocean Deoxygenation in a Warming World." Annu. Rev. Mar. Sci. 2: 199-229.

Khatiwala, S., F. Primeau and T. Hall (2009). "Reconstruction of the history of anthropogenic CO2 concentrations in the ocean." Nature 462: 346-349. Killworth, P. D. (1983). "Deep Convection in the World Ocean." Reviews of Geophysics 21(1): 1- 26. Knorr, G. and G. Lohmann (2003). "Southern Ocean origin for the resumption of Atlantic thermohaline circulation during deglaciation." Nature 424(6948): 532-536. 34 Kortzinger, A., J. Schimanski and U. Send (2005). "High quality oxygen measurements from profiling floats: A promising new technique." Journal of Atmospheric and Oceanic Technology 22(3): 302-308. Kortzinger, A., J. Schimanski, U. Send and D. Wallace (2004). "The ocean takes a deep breath." Science 306(5700): 1337-1337. Kuhlbrodt, T., A. Griesel, M. Montoya, A. Levermann, M. Hofmann and S.Rahmstorf (2007). "On the driving processes of the Atlantic meridional overturning circulation." Reviews of Geophysics: 1-32. Lamb, M. F., C. L. Sabine, R. A. Feely, R. Wanninkhof, R. M. Key, G. C. Johnson, F. J. Millero, K. Lee, T. H. Peng, A. Kozyr, J. L. Bullister, D. Greeley, R. H. Byrne, D. W. Chipman, A. G. Dickson, C. Goyet, P. R. Guenther, M. Ishii, K. M. Johnson, C. D. Keeling, T. Ono, K. Shitashima, B. Tilbrook, T. Takahashi, D. W. R. Wallace, Y. W. Watanabe, C. Winn and C.

S. Wong (2002). "Consistency and synthesis of Pacific Ocean CO2 survey data." Deep-Sea Research Part Ii-Topical Studies in Oceanography 49(1-3): 21-58. Langdon, C, T. Takahashi, C. Sweeney, D. Chipman, D. Goddard, F. Marubini, H. Aceves, H. Barnett, and M. J. Atkinson (2000). "Effect of calcium carbonate saturation state on the calcification rate of an experimental coral reef." Global Biogeochemical Cycles 14: 639- 654. Ledwell, J. R., A. J. Watson and C. S. Law (1993). "Evidence for Slow Mixing across the Pycnocline from an Open-Ocean Tracer-Release Experiment." Nature 364(6439): 701-703.

Lewis, E. and D. W. R. Wallace (1998). "Program Developed for CO2 System Calculations." ORNL/CDIAC-105. Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory, U.S. Department of Energy, Oak Ridge, Tennessee. Lutjeharms, J. R. E. (2007). "Three decades of research on the greater Agulhas Current." Ocean Science 3: 129-147. Macdonald, A. M., M. O. Baringer, R. Wanninkhof, K. Lee and D. W. R. Wallace (2003). "A 1998- 1992 comparison of inorganic carbon and its transport across 24.5 degrees N in the Atlantic." Deep-Sea Research Part Ii-Topical Studies in Oceanography 50(22-26): 3041- 3064. Martz, T. R., K. S. Johnson and S. C. Riser (2008). "Ocean metabolism observed with oxygen sensors on profiling floats in the South Pacific." Limnology and Oceanography 53(5): 2094- 2111. Matear, R. J., A. C. Hirst and B. I. McNeil (2000). "Changes in dissolved oxygen in the Southern Ocean with climate change." Geochemistry Geophysics Geosystems 1(2000GC000086). McCartney, M. S. (1977). Subantarctic Mode Water. A Voyage of Discovery. M. Angel. New York, Pergamon: 103-119. 35 Moriarity, D. J. W. and O'Donahue (1995). "Organic carbon transport from the Southern Ocean and bacterial growth in the Antarctic Intermediate Water masses of the Tasman Sea." Marine Ecology Progress Series 119: 291-297. Naegler, T., P. Ciais, J. C. Orr, O. Aumont and C. Rodenbeck (2007). "On evaluating ocean models with atmospheric potential oxygen." Tellus Series B-Chemical and Physical Meteorology 59(1): 138-156. Nelson, D. M., P. Treguer, M. A. Brzezinski, A. Leynaert and B. Queguiner (1995). "Production and Dissolution of Biogenic Silica in the Ocean - Revised Global Estimates, Comparison with Regional Data and Relationship to Biogenic Sedimentation." Global Biogeochemical Cycles 9(3): 359-372. Nondal, G., K. M. Assmann, and R. G. J. Bellerby: Signals of Climate Change in Antarctic Intermediate Water, Manuscript to be submitted to Nature. Orr, J. C., E. Maier-Reimer, U. Mikolajewicz, P. Monfray, J. L. Sarmiento, J. R. Toggweiler, N. K. Taylor, J. Palmer, N. Gruber, C. L. Sabine, C. Le Quere, R. M. Key and J. Boutin (2001). "Estimates of anthropogenic carbon uptake from four three-dimensional global ocean models." Global Biogeochemical Cycles 15(1): 43-60. Orsi, A. H., T. Whitworth and W. D. Nowlin (1995). "On the Meridional Extent and Fronts of the Antarctic Circumpolar Current." Deep-Sea Research Part I-Oceanographic Research Papers 42(5): 641-673. Oschlies, A., K. G. Schulz, U. Riebesell and A. Schmittner (2008). "Simulated 21st century's

increase in oceanic suboxia by CO2-enhanced biotic carbon export." Global Biogeochemical Cycles 22(4). Raven, J., K. Caldeira, H. Elderfield, O. Hoegh-Guldberg, P. Liss, U. Riebesell, J. Shepherd, D. C. Turley and A. Watson (2005). "Ocean acidification due to increasing atmospheric carbon dioxide." The Royal Society, London Policy document 12/05. Redfield, A. C. (1958). "The Biological Control of Chemical Factors in the Environment." American Scientist 46(3): 205-221. Redfield, A. C., B. H. Ketchum and F. A. Richards (1963). The influence of organisms on the composition of seawater. The Sea. M. N. Hill. New York, Wiley. 2nd. edn. Revelle, R. and H. E. Suess (1957). "Carbon Dioxide Exchange between Atmosphere and Ocean

and the Question of an Increase of Atmospheric CO2 During the Past Decades." Tellus 9(1): 18-27. Riebesell, U., A. Kortzinger and A. Oschlies (2009). "Sensitivities of marine carbon fluxes to ocean change." Proceedings of the National Academy of Sciences of the United States of America 106(49): 20602-20609. 36 Riebesell, U., I. Zondervan, B. Rost, P. D. Tortell, R. E. Zeebe and F. M. M. Morel (2000).

"Reduced calcification of marine plankton in response to increased atmospheric CO2." Nature 407(6802): 364-367. Rintoul, S. R. (1991). "South Atlantic interbasin exchange." Journal of Geophysical Research 96: 2675-2692. Rodenbeck, C., C. Le Quere, M. Heimann and R. F. Keeling (2008). "Interannual variability in

oceanic biogeochemical processes inferred by inversion of atmospheric O2/N2 and CO2 data." Tellus Series B-Chemical and Physical Meteorology 60(5): 685-705. Roson, G., A. F. Rios, F. F. Perez, A. Lavin and H. L. Bryden (2003). "Carbon distribution, fluxes, and budgets in the subtropical North Atlantic Ocean (24.5 degrees N)." Journal of Geophysical Research-Oceans 108(C5). Sabine, C. L., R. A. Feely, N. Gruber, R. M. Key, K. Lee, J. L. Bullister, R. Wanninkhof, C. S. Wong, D. W. R. Wallace, B. Tilbrook, F. J. Millero, T. H. Peng, A. Kozyr, T. Ono and A. F.

Rios (2004). "The oceanic sink for anthropogenic CO2." Science 305(5682): 367-371. Saenko, O. A., J. C. Fyfe and M. H. England (2005). "On the response of the oceanic wind-driven

circulation to atmospheric CO2 increase." Climate Dynamics 25(4): 415-426. Sallee, J. B., K. Speer, S. Rintoul and S. Wijffels (2010). "Southern Ocean Thermocline Ventilation." Journal of Physical Oceanography DOI: 10.1175/2009JPO4291.1: 509-529. Sarmiento, J. L., N. Gruber, M. A. Brzezinski and J. P. Dunne (2004). "High-latitude controls of thermocline nutrients and low latitude biological productivity." Nature 427(6969): 56-60. Schneider, W. and L. Bravo (2006). "Argo profiling floats document Subantarctic Mode Water formation west of Drake Passage." Geophysical Research Letters 33(16): 4 pp. Siegenthaler, U., T. F. Stocker, E. Monnin, D. Luthi, J. Schwander, B. Stauffer, D. Raynaud, J. M. Barnola, H. Fischer, V. Masson-Delmotte and J. Jouzel (2005). "Stable carbon cycle-climate relationship during the late Pleistocene." Science 310(5752): 1313-1317. Skjelvan, I., A. Olsen, L. G. Anderson, R. G. J. Bellerby, E. Falck, Y. Kasajima, C. Kivimäe, A. Omar, F. Rey, K. A. Olsson, T. Johannessen and C. Heinze (2005). A review of the inorganic carbon cycle of the Nordic Seas and Barents Sea. AGU GEophysical Monograph. 158: 157-175. Stephens, B. B., R. F. Keeling, M. Heimann, K. D. Six, R. Murnane and K. Caldeira (1998).

"Testing global ocean carbon cycle models using measurements of atmospheric O2 and CO2 concentration." Global Biogeochemical Cycles 12(2): 213-230. Stramma, L., G. C. Johnson, J. Sprintall and V. Mohrholz (2008). "Expanding oxygen-minimum zones in the tropical oceans." Science 320(5876): 655-658. Stramma, L. and R. G. Peterson (1990). "The South-Atlantic Current." Journal of Physical Oceanography 20(6): 846-859. 37 Suess, E. (1980). "Particulate Organic-Carbon Flux in the Oceans - Surface Productivity and Oxygen Utilization." Nature 288(5788): 260-263. Sundquist, E. T. ”Geological perspectives on carbon dioxide and the carbon cycle.” The Carbon

Cycle and Atmospheric CO2: Natural Variations Archean to Present. E. T. Sunquist and W. S. Broecker. Washington, DC, Geophysical Monor. Ser. 32: 5-60. Swallow, J. C. (1955). "A Neutral-Buoyancy Float for Measuring Deep Currents." Deep-Sea Research 3(1): 74-81. Takahashi, T., J. Olafsson, J. G. Goddard, D. W. Chipman and S. C. Sutherland (1993). "Seasonal-

Variation of CO2 and Nutrients in the High-Latitude Surface Oceans - a Comparative- Study." Global Biogeochemical Cycles 7(4): 843-878. Takahashi, T., S. C. Sutherland and A. Kozyr (2010). "Global oceam surface water partial pressure

of CO2 database: Measurements performed during 1957-2009 (Version 2009)." ORNL/CDIAC-152, NDP-088(V2009). Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory, U.S. Department of energy, Oak Ridge, Tennessee. doi:10.3334/CDIAC/otg.ndp088(V2009). Takahashi, T., S. C. Sutherland, R. Wanninkhof, C. Sweeney, R. A. Feely, D. W. Chipman, B. Hales, G. Friederich, F. Chavez, C. Sabine, A. Watson, D. C. E. Bakker, U. Schuster, N. Metzl, H. Yoshikawa-Inoue, M. Ishii, T. Midorikawa, Y. Nojiri, A. Kortzinger, T. Steinhoff, M. Hoppema, J. Olafsson, T. S. Arnarson, B. Tilbrook, T. Johannessen, A. Olsen, R. Bellerby, C. S. Wong, B. Delille, N. R. Bates and H. J. W. de Baar (2009). "Climatological

mean and decadal change in surface ocean pCO2, and net sea-air CO2 flux over the global oceans (vol 56, pg 554, 2009)." Deep-Sea Research Part I-Oceanographic Research Papers 56(11): 2075-2076. Tengberg, A., J. Hovdenes, H. J. Andersson, O. Brocandel, R. Diaz, D. Hebert, T. Arnerich, C. Huber, A. Kortzinger, A. Khripounoff, F. Rey, C. Ronning, J. Schimanski, S. Sommer and A. Stangelmayer (2006). "Evaluation of a lifetime-based optode to measure oxygen in aquatic systems." Limnology and Oceanography-Methods 4: 7-17. Toggweiler, J. R. (2009). "Shifting Westerlies." Science 323(5920): 1434-1435. Toggweiler, J. R. and J. Russell (2008). "Ocean circulation in a warming climate." Nature 451(7176): 286-288. Toggweiler, J. R. and B. Samuels (1993). New radiocarbon constraint on the upwelling of abyssal water to the ocean's surface. The Global Carbon Cycle. M. Heimann. Berlin, Springer: 333- 366. Uchida, H., T. Kawano, I. Kaneko and M. Fukasawa (2008). "In Situ Calibration of Optode-Based Oxygen Sensors." Journal of Atmospheric and Oceanic Technology 25(12): 2271-2281. 38 Vidal, M., C. M. Duarte and S. Agusti (1999). "Dissolved organic nitrogen and phosphorus pools and fluxes in the central Atlantic Ocean." Limnology and Oceanography 44(1): 106-115. Volk, T. and M. I. Hoffert (1985). Ocean carbon pumps: analysis of relative strengths and

efficiencies in ocean-driven atmospheric CO2 changes. The Carbon Cycle and Atmospheric CO2: Natural Variations Archean to Present. E. T. Sunquist and W. S. Broecker. Washington, DC, Geophysical Monor. Ser. 32: 99-111.

Wallace, D. W. R. (2001). Storage and transport of excess CO2 in the oceans: The JGOFS/WOCE global CO2 survey. Ocean circulation and climate: observing and modelling the global ocean. G. Siedler, J. Church and J. Gould, Academic Press: 489-521. Wallace, D. W. R. (1995). "Monitoring global carbon inventories", OOSDP Background Rep. 5: 54pp. Ocean Obs. Syst. Dev. Panel, Texas, A&M Univ. College Station, Tex. Watson, A. J., U. Schuster, D. C. E. Bakker, N. R. Bates, A. Corbiere, M. Gonzalez-Davila, T. Friedrich, J. Hauck, C. Heinze, T. Johannessen, A. Kortzinger, N. Metzl, J. Olafsson, A. Olsen, A. Oschlies, X. A. Padin, B. Pfeil, J. M. Santana-Casiano, T. Steinhoff, M. Telszewski, A. F. Rios, D. W. R. Wallace and R. Wanninkhof (2009). "Tracking the

Variable North Atlantic Sink for Atmospheric CO2." Science 326(5958): 1391-1393. Wefer, G., W. H. Berger, G. Siedler and D. J. Webb (1996). The South Atlantic: present and past circulation. Berlin Heidelberg, Springer-Verlag. Weiss, R. F. (1970). "The solubility of nitrogen, oxygen and argon in water ans seawater." Deep- Sea Research 17: 721-735. Weiss, R. F. (1974). "Carbon dioxide in water and sea water: The solubility of a non ideal gas." Marine Chemistry 2: 203-215.

Wetzel, P., A. Winguth and E. Maier-Reimer (2005). "Sea-to-air CO2 flux from 1948 to 2003: A model study." Global Biogeochemical Cycles 19(2). Williams, R. G., V. Roussenov and M. J. Follows (2006). "Nutrient streams and their induction into the mixed layer." Global Biogeochemical Cycles 20(1). Williams, R. J. P. (1997). "The natural selection of the chemical elements." Cellular and Molecular Life Sciences 53(10): 816-829.

Paper 1

Optimal evaluation of the surface CO2 system in the northern North Atlantic using data from voluntary observing ships. Nondal, G., R. G. J. Bellerby, A. Olsen, T. Johannessen, and J. Olafsson Limnology and Oceanograhphy: Methods, 7, 109-118, 2009.

Optimal evaluation of the surface ocean CO2 system in the northern North Atlantic using data from voluntary observing ships Gisle Nondal1,2*, Richard G. J. Bellerby3,1, Are Olsen3,1,6, Truls Johannessen1,3, and Jón Olafsson4,5. 1Geophysical Institute, University of Bergen, Allégaten 70, N-5007 Bergen, Norway 2Mohn-Sverdrup Center for Global Ocean Studies and Operational Oceanography at the Nansen Environmental and Remote Sensing Center, Thormøhlensgt. 47, N-5006 Bergen, Norway 3Bjerknes Centre for Climate Research, University of Bergen, N-5007 Bergen, Norway 4Marine Research Institute, Skúlagötu 4, 121 Reykjavik, Iceland 5Faculty of Earth Sciences, University of Iceland, Askja, 107 Reykjavik, Iceland 6Departement of Chemistry, Göteborg University, SE 412 96 Göteborg, Sweden

Abstract

This work evaluates whether an accurate calculation of the entire CO2 system in the northern North Atlantic

can be carried out using a combination of in situ fugacity of CO2 (fCO2) and ancillary data often measured on Voluntary Observing Ships (VOS), i.e., sea surface temperature (SST) and sea surface salinity (SSS), as well as – nitrate (NO3 ). Two approaches are tested: (I) determination of At from SSS and then calculating Ct from mea- – sured fCO2 and estimated At; and (II) determination of Ct from SSS, SST, and NO3 and then calculating At from

measured fCO2 and estimated Ct. The optimal approach was found to be determination of At from SSS and then

calculating Ct from measured fCO2 and estimated At. This allowed At to be determined with a mean bias of –1.8 μ –1 μ –1 mol kg and root mean square (rms) deviation 6.2 mol kg and then Ct to be calculated with a mean bias of –1.0 μmol kg–1 and standard error of calculation of 7.4 μmol kg–1, as validated using independent data sets.

Introduction number of carbon system measurements available in the region, and as pointed out by Olsen et al. (2008), until The northern North Atlantic Ocean is considered to be an recently, there have not been enough measurements from the important uptake area for atmospheric carbon dioxide and North Atlantic Ocean to describe a full annual cycle. This sit- observations indicate substantial variations on interannual uation has been alleviated by autonomous systems that mea- timescales (Lefevre et al. 2004; Olsen et al. 2006; Omar and sure sea surface CO2 fugacity (fCO2) that have been installed Olsen 2006; Corbiére et al. 2007; Schuster and Watson 2007). on board Voluntary Observing Ships (VOS) by several institu- Unfortunately our understanding is presently limited by the tions, as summarized in the report from the Surface Ocean

CO2 Variability and Vulnerabilities (SOCOVV) meeting (available through http://www.ioccp.org/). Knowledge of two of the

*Corresponding author: E-mail: [email protected] four carbon system variables: inorganic carbon (Ct), total alkalinity (A ), pH, and fCO is required for the calculation of the Acknowledgments t 2 The authors would like to thank Abdirahman Omar at the Bjerknes remaining CO2 system variables through thermodynamic Centre for Climate Research (BCCR), Ingunn Skjelvan at UoB and BCCR, equations (Dickson et al. 2007). One of the key challenges in and Eva Falck at the Geophysical Institute (UoB) for their contributions using autonomous systems is to be able to make the systems to this paper. Comments from three anonymous reviewers considerably as reliable and cost efficient as possible, while at the same time improved the manuscript. This work acknowledges financial support from the EU through the CARBOOCEAN (511176 –2) and TRACTOR producing a set of high quality data that can be used to obtain (EVK2-2000-00080) projects, from the Norwegian Research Council a complete description of the surface ocean carbon chemistry. through the CABANERA (155936/700), A-CARB (178167/S30), Marine In practice, this requires development of reliable autonomous Ecosystem Response to Climate Change (MERCLIM, contract no: instruments for determination of At, Ct, or pH, which are not 184860/S30) and Bipolar Atlantic Circulation (BIAC, project currently available on the market. no:176082/S30) projects, and from Swedish National Space Board through RESCUE-II (d.nr 62/07:1). GN acknowledges an endowment by As a first step to overcome this, one can employ the rela- Mr. Trond Mohn c/o Frank Mohn AS. This is publication A212 of the tionships that exist between various CO2 system variables and Bjerknes Centre for Climate Research. nutrient and hydrographic variables (Brewer et al. 1995;

109 Nondal et al. Surface ocean CO2 system in North Atlantic

Millero et al. 1998; Tait et al. 2000; Lee et al. 2000; Olsen et al. Standard error in Ct calculated from At and fCO2—Assuming σ σ 2003; Lefevre et al. 2004; Lee et al. 2006). This study deter- that the measurement errors in At ( At) and Ct ( Ct) are inde- mines relationships that allow for the estimation of surface At pendent and random (Taylor 1982), the following formula and Ct, using a larger dataset than previous studies in the area. gives the standard error in fCO2 calculated from At and Ct

Two approaches were tested: (I) determination of At from SSS (Lueker et al. 2000; Omar 2003). 1 and then calculating C from measured fCO and estimated A ; ⎛ 2 2 ⎞ 2 t 2 t ⎛∂ ⎞ ⎛∂ ⎞ – ⎜ fCO2 fCO2 ⎟ and (II) determination of Ct from SST, SSS, and NO3 and then σσσ= ⎜ ∗ ⎟ +⎜ ∗ ⎟ (A1) fCO() A ,C A C 2 t t ⎜ ∂ t ∂ t ⎟ ⎝ At ⎠ ⎝ Ct ⎠ calculating At from measured fCO2 and estimated Ct. The cal- ⎝ ⎠ culations are then compared to measured data.

Using Eq. A1 for Ct calculated from At and fCO2 gives Materials and procedures 1 ⎛ 2 2 ⎞ 2 ⎛∂C ⎞ ⎛ ∂C ⎞ The focus area for this study is the northern North Atlantic, σσσ= ⎜⎜ t ∗ ⎟ +⎜ t ∗ ⎟ ⎟ (A2) C ()A , fCO A fCO t t 2 ⎜⎝ ∂ t ⎠ ⎝∂ 2 ⎠ ⎟ including the Nordic Seas (Greenland, Norwegian, and Ice- ⎝ At fCO2 ⎠ land Seas). The southern boundary of the study area was set to 45°N. The eastern boundary was set to 10°W up to 50°N; from Starting with the following expressions (Takahashi et al. there it follows the continental shelf edge northwards to 1993): about 83°N. The western boundary was set at the east Green- ∂fCO fCO ∂fCO fCO land coast or the ice edge when it had a more eastward exten- 22=∗RF (A3)22=∗TAF (A4) ∂C C ∂A A sion. South of Cape Farewell (the southern cape of Green- tt tt land), the western boundary was set to 45°W. ∂fCO C Field measurements—The measurements used in this study were where RF (Revelle Factor) = 2 ∗ t (A*) ∂C fCO obtained from the European Sub Polar Ocean Program (ESOP-2), t 2 the tracer and circulation in the Nordic Seas region (TRACTOR) ∂fCO A project (http://www.ices.dk/ocean/project/tractor/), the Marine and TAF (Total Alkalinity Factor) = 2 ∗ t (A**): ∂A fCO Research Institute in Reykjavik (MRI), the University of Bergen t 2 (UoB), the Bjerknes Centre for Climate Research (BCCR), and ∂C C the Global Ocean Data Analysis Project (GLODAP, Key et al. (A3) ⇒ tt= (A5) ∂fCO fCO∗ RF 2004). The locations of the measurements are shown in Fig. 1. 22 Data analysis—The empirical relationships presented in this ∂fCO C paper were determined using the software Sigmaplot 8.0. The (A3) ⇒=fCO 2 ∗ t (A6) 2 ∂ Ct RF CO2 system calculations were performed using an inorganic carbon speciation model (Lewis and Wallace 1998), with the ∂fCO A (A4) ⇒=fCO 2 ∗ t (A7) following settings: CO dissociation constants from Mehrbach 2 ∂ 2 At TAF et al. (1973) as refitted by Dickson and Millero (1987); total pH – scale; and dissociation constants for KSO4 from Dickson Combining Eqs. A6 and A 7: (1990). The refitted Mehrbach et al. (1973) constants were ∂fCO C ∂fCO A used because they have been shown to enable the most accu- ⇒ 22∗=t ∗⇒∂∗∗∂∗t fCO C A TTAF=∂ fCO ∗ A ∗∂ C ∗ RF ∂ ∂ 2 tt 2 tt Ct RF At TAF rate CO2 system calculations (Wanninkhof et al.1999; Nondal 2004), possibly because they were determined using real sea- ∂C ∂C CTAF∗ ⇒∗∂∗C A TAF =∗∂∗ A C RF ⇒∗ C TAF =∗ A RF ∗ t ⇒ t = t water (Mojica-Prieto and Millero 2002). As underway mea- tt t t t t ∂ ∂ ∗ At At ARFt surements of silicate and phosphate were not available, the (A8) effect of these nutrient concentrations in the calculations has been neglected. This will not change the results significantly Using Eq. A2 with the constants from Eqs. A5 and A8 gives 1 as the error from this is smaller than the precision of the in ⎛ 2 2 ⎞ 2 ⎛CTAF∗ ⎞ ⎛ C ⎞ situ measurements (not shown). ⇒=σσ⎜⎜ t ∗ ⎟ +⎜ t ∗σ ⎟ ⎟ (A9) C ()A , fCO A fCO t t 2 ⎜⎝ ∗ t ⎠ ⎝ ∗ 2 ⎠ ⎟ Error analysis for over-determination—Based on studies hav- ⎝ ARFt fCO2 RF ⎠ ing employed parts of the in situ measurements also used in this work (Chierici et al. 1999; Skjelvan et al. 1999), the fol- Standard error in At calculated from Ct and fCO2—Using Eq. A1 μ –1 lowing accuracies have been assumed: At ± 5.0 mol kg , Ct ± for At calculated from Ct and f CO2 μ –1 μ 5.0 mol kg , and fCO2 ± 3.0 atm. The procedure outlined in 1 ⎛ 2 2 ⎞ 2 the next section is applied for determining the standard errors ⎛ ∂A ⎞ ⎛ ∂A ⎞ σ σ ⇒=σσσ⎜⎜ t ∗ ⎟ +⎜ t ∗ ⎟ ⎟ (A10) A ()C , fCO C fCCO of calculation (C ()A , fCO , and A ()C , fCO ) when carbon system t t 2 ⎜ ∂ t ⎝∂ 2 ⎠ ⎟ t t 2 t t 2 ⎝⎝ Ct ⎠ fCO2 ⎠ variables are calculated from a combination of other carbon system variables.

110 Nondal et al. Surface ocean CO2 system in North Atlantic

Fig. 1. Maps showing the locations of the measurements used in this study. (a) Measurements used to derive the regression relationship between At and SSS for Atlantic-influenced water, as well as ice melt water (Eq. 6), are shown in black. The I/B Oden measurements used to obtain the At and SSS relationship for Arctic-influenced water (Eq. 7) with S < 34.5 are shown in red. (b) Measurements used to derive the regression relationship between Ct and SSS, SST, and nitrate. Black dots show measurements used to obtain Eq. 8, and red dots the measurements used to obtain Eq. 9. (c) R/V G. O. Sars

2003 cruise (58GS20030922), (d) dataset used for overdetermination when testing the At (SSS) relationship in Eq. 6. The red dots indicate stations with – At measurements, (e) GLODAP data from 1995 used to validate the Ct(SSS,SST,NO3 ) relationship (Eq. 8), and (f) dataset used for overdetermination when testing Eq. 8.

111 Nondal et al. Surface ocean CO2 system in North Atlantic

(R 2) is the percentage of the variability in the in situ mea- ∂A A surements that can be accounted for by the calculations (Allen (A4) ⇒ tt= (A11) ∂ ∗ fCO22fCO TAF et al. 2007).

∂A ARF∗ Assessment (A8) ⇒ t = t (A12) ∂ ∗ Ct CTAFt Approach I, determination of the CO2 system from measured

fCO2 and total alkalinity derived from salinity—Total alkalinity in Using Eq. A10 with the constants from Eqs. A11 and A12 gives the ocean is affected by several processes: fresh water addition

1 (e.g., river runoff, ice melting, precipitation); fresh water ⎛ 2 2 ⎞ 2 ⎛ ARF∗ ⎞ ⎛ A ⎞ removal (sea-ice formation, evaporation); nutrient cycling; σσ= ⎜⎜ t ∗ ⎟ +⎜ t ∗σ ⎟ ⎟ (A13) A ()C , fCO C fCO t t 2 ⎜ ∗ t ⎝ ∗ 2 ⎠ ⎟ and by the oceanic carbonate cycling (formation and dissolu- ⎝⎝CTAFt ⎠ fCO2 TTAF ⎠ tion of carbonate minerals). In the open ocean, the distribu- The RF value of 11.4, representative for the high latitude tion of surface alkalinity is mainly controlled by the same fac- North Atlantic, and a TAF value of –9.4 (Takahashi et al. 1993) tors that govern salinity (Broecker and Peng 1982), and were used for estimating the uncertainties in calculated vari- therefore surface alkalinity is frequently estimated from sur- ables via the formulas given above. The results of estimated face salinity data. For instance, Millero et al. (1998) identified and calculated variables are given as the mean of the residuals a linear relationship between At and sea surface salinity (SSS) between measured and estimated/calculated variables ± the for the Atlantic Ocean: root mean square (rms) error or standard error of calculation. A = 51.24 × SSS + 520.1 (σ = 9 μmol kg–1, n = 427) (4) Assessing model performance—To objectively evaluate the t performance of the various approaches in this study, three cri- To better account for the effect of nutrient cycling on At, teria will be employed, as motivated by Allen et al. (2007): (i) Lee et al. (2006) extended the work of Millero et al. (1998) and

The Nash Sutcliffe Model Efficiency (ME, Nash and Sutcliffe, identified an optimal polynomial relationship between At, SSS 1970), given by the following formula: and sea surface temperature (SST) in the North Atlantic of the

N form: ∑()IE− 2 nn A = 2305 + 53.97(SSS-35) + 2.74(SSS-35)2 –1.16(SST-20) ME =−1 n=1 (1) t N – 0.040(SST-20)2 (σ = 6.3 μmol kg–1, n = 326) (5) − 2 ∑()IIn n=1 The Arctic Ocean and western Nordic Seas are influenced where In is the in situ measurements, En is the calculated vari- by run-off water from Arctic rivers, which are known to con- μ ables, the overbar indicates the mean of the in situ variable, N tain high concentrations of At, typically above 1000 mol is the total number of measurements, and n is the nth value. kg–1(Olsson and Anderson 1997). This leads to a high intercept

The ME is a measure of the ratio of the model error to the vari- between At and SSS for water influenced by Arctic river runoff, ability of the data, and the performance is rated as follows: < negating the use of a single relationship between At and SSS as 0.2 poor, 0.2-0.5 good, 0.5-0.65 very good, and > 0.65 excel- suggested by Millero et al. (1998). In this study, following lent, according to Maréchal (2004) (Allen et al. 2007). Bellerby et al. (2005), two different relationships were identi- (ii) A Cost Function χ (Holt et al. 2005), given as: fied using data from the top 150 m: one relationship for Atlantic-influenced water and water influenced by ice melt 1 N χ 2 =−∑()EI2 (2) (Eq. 6); and one for Polar-influenced water (Eq. 7). The nσ 2 nn I n=1 Atlantic- and Polar-influenced water masses were defined to be σ where E are calculated variables, I are in situ variables, I is the separated by a salinity of 34.5 (Fig. 2), also following Bellerby standard deviation of the in situ data, and N is the number of et al. (2005). measurements. The Cost Function (CF) enables a comparison × of the accuracy of the estimated variables and lower values of At = 49.35 SSS + 582.00 CF indicate better performance. (r 2 = 0.86, σ = 9.7 μmol kg–1, n = 2478) (S > 34.5) (6) (iii) The correlation coefficient (R) defined as:

A = 15.29 × SSS + 1751.73 N t ∑ ()()IIEE−− (r 2 = 0.63, σ = 8.8 μmol kg–1, n = 455) (S < 34.5) (7) R = n=1 nnn n N (3) −−22 ∑()()IInn EE n n n=1 The locations of the measurements used for obtaining these relationships are shown in Fig. 1a. where E is calculated variables and I is in situ variables as It is evident from Fig. 2 that a portion of the measurements before, and the overbar denotes mean values. The square of R having salinity lower than 34.5 follow the mixing line of the

112 Nondal et al. Surface ocean CO2 system in North Atlantic

Table 1. Model performance for the relationships identified by Millero et al. (1998), Lee et al. (2006), and Corbiére et al. (2007) compared with the relationships identified in this study

Relationship ME* CF* R 2* Millero et al. (1998), Eq. 4 0.67 0.57 0.81 Lee et al. (2006), Eq. 5 0.72 0.53 0.82 Corbiére et al. (2007) 0.73 0.52 0.81 Eqs. 6 and 7 0.90 0.32 0.90 Relationship identified in this study 0.85 0.39 0.85 of the form of Lee et al. (2006). *High values for ME, and R 2, and low values for CF indicate the best correspondence between measured and estimated variables.

September-October 2003 [58GS20030922, Olsen et al. (2006)] (see Fig. 1c). It was assumed that no significant amount of water affected by Arctic river runoff influenced the data as no Fig. 2. Relationship between At and SSS in the northern North Atlantic. Black circles indicate Atlantic water (S > 34.5) and Atlantic water diluted measurements were made in, or close to, the EGC during this by local ice melt (S < 34.5). Red circles indicate polar-influenced water (S cruise, and Eq. 6 was therefore used when estimating the sur-

< 34.5) (red circles Fig. 1a) in the east Greenland Current. The blue line face At. A plot of the residuals is shown in Fig. 3. shows the linear regression relationship for the Atlantic water, including Alkalinity was estimated with a mean bias of –1.8 μmol kg–1 low salinity ice melt water, given by the following equation: A = 49.35 × t and an rms deviation of 6.2 μmol kg–1. SSS + 582 (r 2 = 0.86). The red line shows the linear regression relation- Determining C from estimated A —In this approach, C was ship for the Polar water in the East Greenland Current given by the fol- t t t × 2 determined from the measured fCO data and A derived lowing equation: At(S < 34.5) = 15.29 SSS + 1751.73 (r = 0.63). 2 t from salinity using the equations derived above. From the in situ alkalinity data available from the overdetermination Atlantic water. As the Atlantic water encounters and melts sea dataset (stations having alkalinity are shown with red dots ice, At changes along the regression line describing the Atlantic- in Fig. 1d), there appeared to be no influence of high alka- influenced water due to dilution (Anderson et al. 2004). Based linity river-runoff water. Therefore only Eq. 6 was employed. on the findings of Anderson et al. (2004), only data from the I/B Following this approach, it was possible to calculate the sur- μ –1 Oden cruise in 2002 (Fig. 1a, red circles) having salinity less than face Ct with a mean bias of –1.0 mol kg and a mean stan- 34.5 were used to determine the polar relationship, because dard error of calculation of 7.4 μmol kg–1 (standard error of those measurements were made in the East Greenland Current calculation obtained from A9). A plot of the residuals is (EGC) and Arctic Ocean, and clearly contain the signal of high shown in Fig. 4. alkalinity water from Arctic river runoff. All other mea- Approach II, determination of the CO2 system from measured – surements, also including low salinity ice melt water (S < 34.5), fCO2 and total carbon derived from SSS, SST, and NO3 —Con- were used to construct the Atlantic relationship. tinuous measurements of surface ocean nitrate have been To evaluate how the approach using two alkalinity regimes made feasible (http://www.n-virotech.com/ecolab.html; in the northern North Atlantic (Eqs. 6 and 7) compares to the Colijn and Petersen 2002), and might be included on VOS. approaches of Millero et al. (1998), Lee et al. (2006) (Eqs. 4 and Therefore, following the work of Lee et al. (2000), a relation- – 5, respectively), as well as a recent work of Corbiére et al. ship between Ct and SSS, SST, and NO3 was sought. The rela- (2007), the three model performance criteria described above tionship of Lee et al. (2000) was obtained by normalizing the – were employed. A relationship of the form of Lee et al. (2006) Ct to constant salinity and relating this to SST and NO3 . The was also identified using the same data as for determining Eqs. normalization possibly introduced erroneous trends to the

6 and 7, and tested in the same manner. The results are sum- relationship (Friis et al. 2003), and therefore in this study, Ct – marized in Table 1. was directly related to SSS, SST, and NO3 . The dataset used

It is evident from Table 1 that the two local linear relation- spanned 11 y (1991-2002), and the Ct measurements were ships (Eq. 6 and 7) are more appropriate for the northern adjusted to a single reference year to remove the effect of the

North Atlantic and the Nordic Seas, and that the use of a rela- anthropogenic Ct increase. The reference year was set to 1995. tionship similar to the one of Lee et al. (2006) does not Total alkalinity was assumed to have remained unchanged improve the results. over the last decades (Millero et al. 1998), and following Olsen Validating the alkalinity versus salinity relationship—To vali- et al. (2003), and more recently Körtzinger et al. (2008), it was date the relationships between alkalinity and salinity, they assumed that the northern North Atlantic surface water fCO2 μ –1 were tested on measurements from R/V G. O. Sars obtained in has tracked the atmospheric fCO2 increase of 1.4 atm y over

113 Nondal et al. Surface ocean CO2 system in North Atlantic

Fig. 3. At was estimated using Eq. 6, and the plot shows the deviation Fig. 4. The plot shows the measured – calculated Ct when estimating the from measured At. The solid black line indicates the mean bias. At using Eq. 6 and combining this with in situ fCO2. The black solid line represents the mean bias in the calculations.

the study period (Keeling and Whorf 2003). To determine the Validating Ct relationship—To test the performance of Eq. 8 rate of the surface water Ct increase due to the increasing fCO2, on in situ measurements obtained in 1995, all available 1995 the surface water (top 150 m) fCO2 was calculated from measurements in the study area from the GLODAP database 95 measured salinity, At, Ct, silicate, and phosphate. The calcu- were used (Fig. 1e). Eq. 8 enables the prediction of these Ct μ –1 μ lated fCO2 was then adjusted to 1995 using the with a mean bias of 0.5 mol kg and an rms error of 5.9 mol μ –1 –1 increase/decrease of 1.4 atm y , and then the Ct representa- kg . A plot of the residuals between measured and predicted 95 tive of 1995 was calculated from At, salinity, silicate, phos- Ct is shown in Fig. 5. phate, and adjusted fCO2. Based on these calculations, an Secondly, Ct was predicted, using Eq. 8, for the mea- μ –1 –1 adjustment of 0.8 mol kg y was applied on the Ct data, surements from the G. O. Sars cruise in 2003 (Fig. 1c). The sur- μ –1 –1 which agrees well with time series measurements from the face Ct normalized to 1995 by 0.8 mol kg yr was predicted Irminger Sea (Olafsson et al. 2002). with a mean bias of 3.3 μmol kg–1 and an rms error of 7.6 μmol The year was divided into two seasons, summer (April-Sep- kg–1. The residual plot is shown in Fig. 6. tember) and winter (October-March), and one relationship Determining At from estimated Ct—In this approach, At was was identified for each season: determined from measured fCO2 and Ct derived from SSS, SST, –1 and NO3 using Eq. 8. Fig. 7 shows a plot of the residuals. At 95 × × μ –1 Summer: Ct = 1176.52 – 4.32 SST + 26.55 SSS + 4.25 was calculated with a mean bias of 4.0 mol kg and a mean × – 2 σ μ –1 μ –1 NO3 , (r = 0.83, = 16.6 mol kg , n = 3544) (8) standard error of calculation of 22.3 mol kg (standard error of calculation obtained from A13). The large error associated with the calculation stems mostly from the large error (16.6 Winter: C 95 = –293.63 – 7.17 × SST + 69.58 × SSS + 1.28 t μmol kg–1) of the surface C estimates from Eq. 8. × NO –, (r 2 = 0.94, σ = 8.1 μmol kg–1, n = 207) (9) t 3 Comparing approaches I and II—To identify the optimal variable combination for VOS, the three criteria as described It is evident from Eqs. 8 and 9 that the summertime rela- before (Eqs. 1-3) were employed, and the results are shown in – tionship is more dependent on the NO3 concentration than Table 2. the wintertime relationship. This is due to the higher biologi- It is evident from Table 2 that when having in situ mea- cal activity during summer. The separation into two seasons surements of fCO2, the optimal variable combination to be will possibly lead to less noise in the wintertime calculations, used for estimating the carbon system using VOS is fCO2 and enabling more accurate calculations compared with employ- SSS. This approach has the highest model efficiency, highest ing one relationship for the entire year. correlation between measured and calculated variables, as well As there were limited amounts of measurements from the as the smallest value of the cost function. winter season available for validation, only the relationship in Eq. 8 will be used further in this paper. The very strong correla- Discussion tion identified in Eq. 9 encourages future verification of this rela- Empirical relationship between At and SSS—Anderson et al. tionship as more wintertime measurements become available. (2004) show the different effects on At by sea-ice melt water

114 Nondal et al. Surface ocean CO2 system in North Atlantic

Fig. 5. Ct from GLODAP was determined using Eq. 8, and the plot shows Fig. 6. Surface Ct normalized to 1995 was determined using Eq. 8. The the deviation from measured Ct. The black line indicates the mean bias. plot shows the deviation from measured Ct also normalized to 1995. The black line indicates the mean bias. and high alkalinity Arctic river-runoff water. Therefore, to employ et al. 2008). However as Findlay et al. (2008) did not include

Eq. 6 and 7 on continuous measurements of SSS, one needs to the effects of nutrient uptake on At in their model, this sug- know which alkalinity regime is present. The distinction gests that their estimate is an upper bound. A mesocosm between river water and melt water can be made using 18O mea- study between 15 May and 9 June 2005 containing calcifiers surements together with salinity (Frew et al. 2000), but as the common to Norwegian fjords and open waters of the Nor- measurements made on VOS do not include 18O, initiation of wegian and Barents Sea (Bellerby et al. 2008) indicate a μ –1 sampling of this variable from VOS is encouraged. The above reduction in At of about 20 mol kg under present CO2 con- discussion points to a shortcoming of this method at the ditions. As both studies show drawdown in At higher than boundary between the EGC and the waters to the east of it. the individual precision of At measurements, it indicates that

Olsen et al. (2008) show in their Fig. 4c that surface fCO2 in the caution should be made in estimating At using SSS during EGC is significantly lower than in the surrounding waters. At blooming events of calcifying organisms. It suggests that the 60°N, this corresponded to the 2750-m depth contour at the uncertainty in the proposed method increases during periods

East Greenland shelf edge. Underway fCO2 measurements of strong biological activity. – together with bathymetry may therefore be used to identify the Empirical relationship between Ct and SSS, SST, and NO3 —The

EGC. Based on this, it is proposed that caution should be exer- relationship between Ct normalized to constant salinity, SST, – cised when estimating surface At from SSS when a drop in fCO2 and NO3 identified by Lee et al. (2000) possibly contain arti- is observed approaching the east Greenland shelf. To the east of facts introduced through the normalization procedure (Friis et this, Eq. 6 should be used, and Eq. 7 should be used to the west. al. 2003). The approach identified in this study (Eq. 8), does

The effect on surface At due to biology has not been explic- not include salinity normalization and is able to estimate the 95 itly accounted for in Eqs. 6 and 7. Calcifying organisms (e.g., surface Ct from GLODAP, and Ct normalized to 1995 from G. coccolithophorids, foraminifera, and pteropods) use calcium O. Sars 2003 well (0.5 ± 5.9 μmol kg–1 and 3.3 ± 7.6 μmol kg–1 carbonate (CaCO3) when building skeletons and shells, respectively). thereby reducing At. As underway measurements of oxygen (O2) are currently

The precipitation of one mole of CaCO3 will lead to a being implemented on VOS lines, the inclusion of this param- reduction in At of 2 moles. This will happen in tandem with eter in the future studies of the surface carbon system is war- a reduction in the surface nitrate and phosphate concentra- ranted. As shown by McNeil et al. (2007), the prediction of – tions, each having the potential to increase At with one mole surface Ct was improved by using SST, SSS, NO3 , and O2 com- – each for one mole used nitrate and phosphate respectively pared to SST, SSS, and NO3 . (Brewer and Goldman 1976). Therefore the net effect on alkalinity from biology comes from a combination of the Comments and recommendations two above mentioned processes. Results from a modeling The understanding of the northern North Atlantic and study at Ocean Weather Ship M (66°N, 02°E), in the Norwe- Nordic seas carbon system is limited due to lack of in situ mea- μ –1 gian Sea, indicate an At reduction of 38 mol kg , in sum- surements, and VOS are presently being used to increase the mer, due to the effects of calcifying phytoplankton (Findlay data coverage. This study has presented and discussed two

115 Nondal et al. Surface ocean CO2 system in North Atlantic

Table 2. Performance of the two approaches used in this study

Approach ME* CF* R 2*

(I) At(SSS) G. O. Sars 0.86 0.36 0.88

Ct(At(SSS),fCO2) 0.95 0.22 0.96 – (II) Ct(SSS,SST,NO3 ) GLODAP 0.97 0.18 0.97 – Ct(SSS,SST,NO3 ) G. O. Sars 0.92 0.29 0.93 – At[Ct(SSS,SST,NO3 ),fCO2] 0.83 0.41 0.91 *Higher values of ME and R 2 and low values of CF indicate the best performance.

assessment of surface ocean acidification and anthropogenic

CO2 uptake, highlighting the importance for sustaining and expanding the VOS lines. References

Fig. 7. Measured – calculated At when estimating the Ct using the rela- Allen, J. I., J. T. Holt, J. Blackford, and R. Proctor. 2007. Error tionship given by Eq. 8 and combining this with in situ fCO adjusted to 2 quantification of a high-resolution coupled hydrodynamic- 1995 values. The solid black line shows the mean bias. ecosystem coastal-ocean model: Part 2. Chlorophyll-a. approaches for determining the CO2 chemistry in the north- nutrients and SPM. J. Mar. Sys. 68:381-404. ern North Atlantic from measured fCO2 and ancillary data. Anderson, L. G., S. Jutterström, S. Kaltin, E. P. Jones, and G.

Approach I used the strong relationship between At and SSS Björk. 2004. Variability in river distribution in the Eurasian μ –1 allowing At to be estimated with a mean bias of –1.8 mol kg Basin of the Arctic Ocean. J. Geophys. Res. 109(C01016): and an rms deviation of 6.2 μmol kg–1. Along with underway doi:10:1029/2003JC001773. fCO2, this enabled the calculation of Ct with a mean bias of Bellerby, R. G. J., A. Olsen, T. Furevik, and L. A. Anderson. μ –1 μ –1.0 mol kg and a standard error of calculation of 7.4 mol 2005. Response of the surface ocean CO2 system in the –1 kg . Approach II used a relationship between Ct and SST, SSS, North Atlantic to climate change. In H. Drange, T. M. – and NO3 and enabled Ct to be determined with a mean bias Dokken, T. Furevik, R. Gerdes and W. Berger (Eds.) The of 0.5 μmol kg–1 and an rms deviation of 5.9 μmol kg–1. Nordic Seas: An integrated perspective. Geophysical Mono-

Together with underway fCO2, this enabled the calculation of graph Series, AGU. μ –1 At with a mean bias of 4.0 mol kg and a mean standard error ———, K. G. Schulz, U. Riebesell, C. Neill, G. Nondal, T. of calculation of 22.3 μmol kg–1. Johannessen, and K. R. Brown. 2008. Marine ecosystem The optimal variable combination for VOS was identified as community carbon and nutrient uptake stoichiometry

SSS and fCO2 (approach I), as discussed above. This approach under varying ocean acidification during the PeECE III enables calculations of the carbon system with accuracy com- experiment. Biogeosciences 5:1517-1527. parable with that obtained when calculating carbon system Brewer, P. G., D. M. Glover, C. Goyet, and D. K. Shafer. 1995. variables using field measurements. It should be pointed out The pH of the North Atlantic Ocean: Improvements to the that the proposed method has no simple and efficient means global model for sound absorption in seawater. J. Geophys. of discerning between high alkalinity river water and Atlantic Res. 100(C5):8761-8766. water affected by local ice melt. Therefore sampling of 18O in ——— and J. C. Goldman. 1976. Alkalinity changes generated regions susceptible to this high alkalinity input is strongly rec- by phytoplankton growth. Limnol. Oceanogr. 21(1):108-117. ommended. Broecker, W. S., and T. H. Peng. 1982. Tracers in the sea. Future remote sensing products such as the European Space Eldigeo Press. Agency (ESA)’s Soil Moisture and Ocean Salinity (SMOS, Chierici, M., H. Drange, L. G. Anderson, and T. Johannessen. http://www.esa.int/esaLP/ESAMBA2VMOC_LPsmos_0.html) 1999. Inorganic carbon fluxes through the boundaries of mission, scheduled for launch in 2009, has the potential to the Greenland Sea Basin based on in situ observations and provide the first ever surface ocean salinity measurements water transport estimates. J. Mar. Sys. 22:295-309. from space, on which the method proposed in this work may Colijn, F., and W. Petersen. 2002. Collecting oceanography be employed. As Olsen et al. (2004) have described the feasi- data from ferries. ICES Newsletter no. 39. bility of estimating surface fCO2 using remotely sensed SST, Corbiére, A., N. Metzl, G. Reverdin, C. Brunet, and T. Taka- these two methods may be combined providing a full descrip- hashi. 2007. Interannual and decadal variability of the tion of the surface ocean CO2 system from remotely sensed oceanic carbon sink in the North Atlantic subpolar gyre. data. Over sufficiently long timescales this will enable an Tellus 59B:168-178.

116 Nondal et al. Surface ocean CO2 system in North Atlantic

Dickson, A. G., C. Sabine, and J. Christian (Eds.) 2007. Guide Lueker, T. J., A. G. Dickson, and C. D. Keeling. 2000. Ocean

to best practices for ocean CO2 measurements. PICES Spe- pCO2 calculated from dissolved inorganic carbon, alkalin-

cial Publication. ity, and equations for K1 and K2: validation based on labo- 1 ———. 1990. Standard potential of the (AgCl(s) + /2H2 (g) = ratory measurements of CO2 in gas and seawater at equilib- Ag(s) + HCl(aq)) cell and the dissociation constant of bisul- rium. Mar. Chem. 70:105-119. fate ion in synthetic sea water from 273.15 to 318.15 K. J. Maréchal, D. 2004. A soil-based approach to rainfall-runoff Chem. Thermodynamics 22:113-127. modelling in ungauged catchments for England and Wales. ——— and F. J. Millero. 1987. A comparison of the equilib- PhD thesis, Cranfield University. rium constants for the dissociation of carbonic acid in sea- McNeil, B., N. Metzl, R. M. Key, R. J. Matear, and A. Crobiere. water media. Deep-Sea Res. 34:1733-1743. 2007. An empirical estimate of the southern ocean air-sea

Findlay, H. S., T. Tyrrell, R. G. J. Bellerby, A. Merico, and I. CO2 flux.” Global Biogeochem. Cycles 21, GB3011, Skjelvan. 2008. Carbon and nutrient mixed layer dynamics doi:10.1029/2007GB002991 in the Norwegian Sea. Biogeosciences 5:1395-1410. Mehrbach, C., C. Culberson, J. E. Hawley, and R. M. Pytkow- Frew, R. D., P. F. Dennis, K. J. Heywood, M. P. Meredith, and S. icz. 1973. Measurement of the apparent dissociation con- M. Bosweli. 2000. The oxygen isotope composition of water stants pf carbonic acid in seawater at atmospheric pressure. masses in the northern North Atlantic. Deep Sea Res. I, Limnol. Oceanogr. 18:897-907. doi:10.1016/S0967-0637(00)00023-6. Millero, F. J., K. Lee, and M. Roche. 1998. Distribution of alka- Friis, K., A. Körtzinger, and D. W. R. Wallace. 2003. The salin- linity in the surface waters of the major oceans. Mar. Chem. ity normalization of marine inorganic carbon chemistry. 60(1-2):111-130.

Geophys. Res. Lett. doi:10.1029/2002GL015898, 2003. Mojica-Prieto, F. H., and F. Millero. 2002. The values for pK1 +

Holt, J. T., J. I. Allen, R. Proctor, and F. Gilbert. 2005. Error pK2 for the dissociation of carbonic acid in seawater. quantification of a high-resolution coupled hydrodynamic- Geochim. Cosmochim. Acta 66(14):2529-2540. ecosystem coastal-ocean model: Part 1. model overview and Nash, J. E., and J. V. Sutcliffe. 1970. River flow forecasting assessment of the hydrodynamics. J. Mar. Sys. 57:167-188. through conceptual models. Part I-a discussion of princi-

Keeling, C. D., and T. P. Whorf. 2003. Atmospheric CO2 ples. J. Hydrol. 10(3):282-290. records from sites in the SIO air sampling network. A Nondal, G. 2004. The carbon dioxide system in the northern compendium of data on global change. Carbon Dioxide North Atlantic: An assessment of the optimal variable com- Analysis Center, Oak Ridge National Laboratory, U.S. bination for voluntary observing ships. Cand. Scient. The- Department of Energy. (Available online at sis, University of Bergen, Norway. http://cdiac.ornl.gov/trends/co2/sio-keel-flask/). Olafsson, J., S. R. Olafsdóttir, and T. Takahashi. 2002. Time Key, R. M., and others. 2004. A global ocean carbon climatol- series observations of inorganic carbon and nutrients in the ogy: Results from GLODAP. Global Biogeochem. Cycles Irminger and Iceland Seas. Poster presented at EGU confer- 18:GB4031 ence in Nice, France. Geophys. Res. Abs. Vol. 5, 14234, Körtzinger, A., U. Send, D. W. R. Wallace, J. Karstensen, and M. 2003.

DeGrandpre. 2008. Seasonal cycle of O2 and pCO2 in the Olsen, A., R. G. J. Bellerby, T. Johannessen, A. M. Omar, and I. central Labrador Sea: Atmospheric, biological, and physical Skjelvan. 2003. Interannual variability in the wintertime implications. Global Biogeochem. Cycles. 22: air-sea flux of carbon dioxide in the northern North doi:10.1029/2007GB003029. Atlantic, 1981-2001. Deep-Sea Res. I 50(10-11):1323-1338. Lee, K., and others. 2006. Global relationships of total alkalin- ———, J. A. Triñanes, and R. Wanninkhof. 2004. Sea-air flux

ity with salinity and temperature in surface waters of the of CO2 in the Caribbean Sea estimated using in situ and world’s oceans. Geophys. Res. Lett., 33, L19605, remote sensing data. Rem. Sens. Environ. 89:309-325. doi:10.1029/2005GL027207. ——— and others. 2006. Magnitude and origin of the anthro- 13 ———, R. Wanninkhof, R. A. Feely, F. J. Millero. and T. H. pogenic CO2 increase and C suess effect in the Nordic Seas Peng. 2000. Global relationships of total inorganic carbon since 1981. Global Biogeochem. Cycles with temperature and nitrate in surface seawater. Global 20:doi:10.1029/2005GB002669. biogeochem. Cycles 14(3): 979-994. ———, K. R. Brown, M. Chierici, T. Johannessen, and C. Neill.

Lefevre, N., A. J. Watson, A. Olsen, A. F. Ríos, F. F. Pérez, and T. 2008. Sea-surface CO2 in the subpolar North Atlantic. Bio- Johannessen. 2004. A decrease in the sink for atmospheric geosciences 5:535-547.

CO2 in the North Atlantic. Geophys. Res. Lett. 31(L07306): Olsson, K., and L. G. Anderson. 1997. Input and biogeochem- doi:10.1029/2003GL018957. ical transformation of dissolved carbon in the Siberian shelf Lewis, E., and D. W. R. Wallace. 1998. Program developed for seas. Continent. Shelf Res. 17:819-833.

CO2 system calculations. ORNL/CDIAC-105. Carbon Diox- Omar, A. M. 2003. Carbon dioxide in northern high latitude ide Information Analysis Center, Oak Ridge National Labo- oceans: anthropogenic increase and air-sea flux variability. ratory, U.S. Department of Energy. Dr.Scient. thesis, University of Bergen, Norway.

117 Nondal et al. Surface ocean CO2 system in North Atlantic

——— and A. Olsen. 2006. Reconstructing the time history of Takahashi, T., J. Ólafsson, J. G. Goddard, D. W. Chipman, and

the air-sea CO2 disequilibrium and its rate of change in the S. C. Sutherland. 1993. Seasonal variation of CO2 and nutri- eastern subpolar North Atlantic, 1972-1989. Geophys. Res. ents in the high-latitude surface oceans: a comparative Lett., 33, L04602, doi:10.1029/2005GL025425. study. Global Biogeochem. Cycles 7(4):843-878. Schuster, U., and A. J. Watson. 2007. A variable and decreasing Taylor, J. R. 1982. An introduction to error analyses. The study of

sink for atmospheric CO2 in the North Atlantic. J. Geophys. uncertainty in physical measurements. Oxford Univ. Press. Res. 112(C11006):doi:10.1029/2006JC003941. Wanninkhof, R., E. Elwis, R. A. Feely, and F. J. Millero. 1999. Skjelvan, I., T. Johannessen, and L. Miller. 1999. Interannual The optimal carbonate dissociation constants for determin-

variability of fCO2 in the Greenland and Norwegian Seas. ing surface water pCO2 from alkalinity and total inorganic Tellus 51B:477-489. carbon. Mar. Chem. 65:291-301. Tait, V. K., R. M. Gershey, and E. P. Jones. 2000. Inorganic car- Submitted 9 June 2008 bon in the Labrador Sea: Estimation of the anthropogenic Revised 27 November 2008 component. Deep-Sea Res. I 47(2):295-308. Accepted 10 December 2008

118 Paper 2

Fluxes of O2, CO2, and atmospheric potential oxygen from II autonomous profiling floats. Nondal, G., A. Olsen, R. G. J. Bellerby, and P. M. Haugan. Manuscript in preparation.

Manuscript in preparation.

Fluxes of, O2, CO2 and atmospheric potential oxygen in the South Atlantic Ocean from autonomous profiling floats

Gisle Nondal1,2,3,4, Are Olsen1,2, Richard G. J. Bellerby1,2,3, and Peter M. Haugan3,1,2

[1]{ Uni Bjerknes Centre, Allégaten 55, N-5007 Bergen, Norway, Norway}

[2]{ Bjerknes Centre for Climate Research, Allégaten 55, N-5007 Bergen, Norway} [3]{ Geophysical Institute, University of Bergen, Allégaten 55, 5006 Bergen, Norway} [4]{ Mohn-Sverdrup Center/Nansen Center, Thormøhlensgate 47, N-5006 Bergen, Norway}

Correspondence to: G. Nondal ([email protected])

Abstract

Autonomous profiling floats are increasingly used platforms for global ocean in situ observations following recent advances in both float and sensor technology. After a deployment of two Navigating European Marine Observer (NEMO) profiling floats at 41.01°S, 55.03°W on 8 January 2008, water column measurements of temperature, salinity, and dissolved oxygen were made between the surface and 1000m. The floats were transported across the South Atlantic Ocean at the northern rim of the Antarctic Circumpolar Current. Fluxes of oxygen were calculated from direct measurements, and fluxes of carbon dioxide (CO2) were estimated from interpolation of the Lamont

Doherty Earth Observatory (LDEO) surface ocean pCO2 dataset. This enabled the calculation of fluxes of Atmospheric Potential Oxygen (APO) across the air-sea interface. The fluxes over the -2 -1 NEMO drift tracks were estimated to: 11.2±1.1 mmol CO2 m d from the atmosphere to the -2 -1 ocean, 0.012±0.006 mol O2 m d from the ocean to the atmosphere, and a mean APO flux of 0.001±0.006 mol APO m-2 d-1 directed from the ocean to the atmosphere.

1 Manuscript in preparation.

1. Introduction

The South Atlantic Ocean is considered an important source of oxygen (O2) to the atmosphere

(Gruber et al. 2001), while at the same time a significant sink for atmospheric carbon dioxide (CO2)

(McNeil et al. 2007, Takahashi et al. 2009). Atmospheric O2 variability is tightly coupled to variations in atmospheric CO2 through photosynthesis by land plants, and remineralisation and respiration of terrestrial organic matter (Severinghaus 1995, Gruber et al. 2001). Based on the work of Keeling and Shertz (1992) and Keeling et al. (1998), Stephens et al. (1998) proposed a new tracer called atmospheric potential oxygen (APO ~ O2 + 1.1CO2), a tracer removing the terrestrial signal, and to a large extent, the fossil fuel signal from atmospheric O2. APO thus represents the amount of O2 that would be left in a parcel of air after all of its CO2 have been consumed through photosynthesis (Stephens et al. 1998), hence variations in APO mainly reflect the air-sea exchange of O2 and to a lesser extent also CO2. The present understanding of the partitioning of O2 and CO2 between the ocean and the atmosphere is based largely on shipborne measurements with limited spatial and temporal coverage, inverse models (e.g. Gruber et al. 2001, 2009a), and simulations performed by coupled ocean general circulation - biogeochemistry models (e.g. Ito et al. 2010, Tjiputra et al. 2010). Measurements of APO has the potential to advance the current knowledge by providing boundary conditions to atmospheric transport models (Naegler et al. 2007, Rödenbeck et al. 2008), and serving as a sensitive test metric for ocean carbon cycle models (Stephens et al. 1998, Naegler et al. 2007, Rödenbeck et al. 2008). In order for this to be accomplished, there is, as concluded by Naegler et al. (2007), a strong need for in situ atmospheric APO measurements resolving interannual variability, in addition to evaluations of variability in oceanic fluxes of O2 and

CO2. High quality measurements from autonomous profiling floats have the potential to significantly alleviate this situation (Gruber et al. 2009). State-of-the-art floats are able to profile the upper water column (typically down to 1000m or 2000m), while measuring temperature (T), and salinity (S) (Davies et al. 2001), and recently also dissolved oxygen (O2) (Körtzinger et al. 2004, 2005, Tengberg et al. 2006, Gruber et al. 2009). Combined with atmospheric data (sea level pressure and 10m wind speed) this allows direct estimation of the air-sea exchange of O2. The development of autonomous CO2 system sensors, for floats, is still in its infancy and until such sensors become readily applied, a first step to overcome this issue is to employ the strong relationships existing between the partial pressure of CO2 (pCO2) and ancillary variables (e.g. Tans et al. 1990, Lee et al. 1998, Olsen et al. 2003).

2 Manuscript in preparation.

2. Methods This study focuses on the region 40-50°S, 56°W-45°E as this was the area covered by two floats deployed 8 January 2008 at 41.01°S, 55.03°W (fig. 1), with the initial objective of studying intermediate water masses in the “cold water path” (Rintoul 1991) of the Atlantic Meridional Overturning circulation (Nondal et al., manuscript to be submitted).

2.1 Float deployment and data Two Navigating European Marine Observer (NEMO) profiling floats (http://www.optimare.de/cms/en/divisions/mms/mms-products/nemo.html), equipped with Aanderaa Data Instruments oxygen optodes, were used in this work. Optimal float deployment strategies rely on an intimate knowledge of the study region and yet floats are of particular benefit in more poorly understood, and infrequently sampled regions. This paradox may be reconciled through the use of ocean general circulation models to propose deployment locations. In this work, experiments in which synthetic floats were released and subsequently tracked, in the HYbrid Coordinate Ocean Model (HYCOM; Bleck 2002) were used to determine optimal float deployment locations, from where the floats would likely be able to cross the entire South Atlantic Ocean during their expected lifetime. Three experiments were performed, all starting on model date 1 January 2001, and integrated for 360 days: (i) 141 synthetic floats were released south of the Falkland/Malvinas

Islands, drifting on the model 26.8 isopycnal (σ0), representing Subantarctic Mode Water (SAMW) in the southeast Pacific (e. g. Hanawa and Talley 2001) (black lines in fig. 2); (ii) 966 synthetic floats were deployed in the Brazil-Malvinas Confluenze Zone (BMCZ), drifting on the 26.6 isopycnal, representing SAMW in the South Atlantic (Hanawa and Talley 2001) (blue lines fig. 2); and (iii) 20 synthetic floats were released along 25°W, 35°-45°S, with 0.5 ° spacing between them. The latter experiment was performed for studying the evolution of the South Atlantic Current (SAC; Stramma and Peterson 1990), and the red lines in fig. 2 show the model trajectories for this experiment. It is evident that the model suggests that the floats would cross the South Atlantic Ocean in about 2 years if deployed in the BMCZ region, or south of about 55°S to the south of the Falkland/Malvinas Islands. This region was chosen as the optimal float deployment location, and deployments were made from R/V G. O. Sars on 8 January 2008 (41.01°S, 55.03°W).

3 Manuscript in preparation.

2.2 Calibration of O2 optodes The oxygen optodes are delivered having a nominal accuracy of 8.0 μmol kg-1 or 5% whichever is greater, as obtained through a batch calibration by the manufacturer. Field studies however indicate that this accuracy can be improved (i.e. Körtzinger et al. 2005, Tengberg et al. 2006). We thus performed an extended optode O2 calibration prior to float integration. This was carried out using water from Byfjorden outside of Bergen, Norway, manipulated to temperatures, and O2 concentrations, which the floats were expected to encounter in the specific study area. After the -1 individual optode corrections, NEMO60 measured O2 with a mean offset of -0.007 μmol kg and an rms error of 2.2 μmol kg-1, while NEMO47 with 0.002 μmol kg-1, and 1.8 μmol kg-1 respectively

(not shown). Crossover-comparison with in situ O2, as determined by Winkler titration, one year later suggested negligible optode drift over the respective time (not shown).

2.3 Sea surface partial pressure of CO2

We employ the LDEO surface ocean pCO2 database (Takahashi et al. 2010) identifying algorithms allowing for the prediction of the surface ocean pCO2 using float measurements. The pCO2 measurements were adjusted to the nominal year 2008 by assuming a steady oceanic pCO2 increase of 1.5 μatm yr-1 (Takahashi et al. 2009), the temperature was set to 5°C, and the year was divided into summer (Oct.-Mar.), and winter (Apr.-Sept.). As processes altering surface ocean pCO2 (mixing, biological activity, solubility, gas exchange) are related to both SST and SSS, these parameters warrant inclusion into the algorithm. Additionally, position (lon, lat) was included in order to account for the strong regional variability in the South Atlantic (Orsi et al. 1995, Boutin et al. 2008). This significantly improved the regression coefficients, as compared to only using SST and SSS. A multiple least squares regression resulted in the following relationships between surface pCO2, SST, SSS, lon, and lat:

Summer:

2008,t=5 2 pCO2 = –15.6682*SST + 8.8508*SSS + 0.5581*lon – 2.9220*lat, n = 8167, r = 0.91 (1)

Winter:

2008,t=5 2 pCO2 = – 17.8734*SST + 14.5210*SSS + 0.3427*lon + 0.7195*lat, n = 6760, r = 0.95 (2)

For evaluating the performance, and validity, of the seasonal algorithms we utilize the independent in situ measurements from CARIOCA buoy 9 of Boutin (2008), normalized to 2008. A plot of the t=5 residuals between measured and estimated pCO2 is shown in figure 4. The summertime t=5 relationship (1) enables estimation of surface pCO2 with a mean bias of 3.2μatm and an rms error 4 Manuscript in preparation. of 6.6μatm. The corresponding statistics for the wintertime relationship (2) is -0.2μatm, and 6.4μatm, respectively

2.4 Air-Sea gas exchange The flux, F, of any gas across the air-sea interface is a function of the concentration gradient across the interface, ΔC, and the transfer velocity, k:

F = kΔC (3)

If the gas follows Henry’s law, the concentration gradient can be expressed as (Liss and Slater, 1974):

CaH – Cw (4)

Where Cw represents the concentration of the gas in the water phase, Ca in the air phase, and H is the Henry’s law constant. CaH therefore represents the concentration of dissolved gas that would be attained in a water parcel in equilibrium with the air phase.

2.4.1 Oxygen

For oxygen with a high atmospheric concentration, CaH is regarded as equal to the saturation concentration, hence:

F = kO()[]− []O (5) O2 2 2 s

-2 -1 The O2 flux (mol m d ), at a sea level pressure of one atmosphere (1 atm), is expressed as:

k F = ()[]O − []O (6) O2 1000 2 2 s

-1 -3 where k is the transfer velocity (m d ), [O2] the surface O2 concentration (mmol m ), and [O2]s -3 represents the saturation concentration of oxygen (mmol m ; Weiss, 1970). Both [O2] and [O2]s were expressed in units of μmol kg-1, and converted to mmol m-3 using in situ density (kg m-3).

Negative F values indicate fluxes from the ocean to the atmosphere (i.e. a loss of O2 for the O2 ocean). As the annual mean sea level pressure is approximately 2% lower at 60°S compared to 30°S

5 Manuscript in preparation.

(Oberhuber 1988, Najjar and Keeling 1997), the O2 fluxes in this work need to be corrected for the pressure effect on O2 solubility. This was achieved through the following expression (Najjar and Keeling 1997): k ⎛ ⎛ p ⎞ ⎞ F = ⎜ ()[]O − []O +⎜1 − ⎟⋅ []O ⎟ (7) O2 1000⎝ 2 2 s ⎝ p0 ⎠ 2 s⎠

where p represents the corresponding (temporal and spatial) sea level pressure from the NCEP-DEO AMIP-II reanalysis (Kanamitsu et al. 2002), and p0 is the mean sea level pressure of 1013.25mb (Najjar and Keeling 1997). A lot of effort has been put in describing the relationship between the gas transfer velocity (k), and the wind speed (u) (e.g. Liss and Merlivat 1986, Wanninkhof 1992, Wanninkhof and McGillis 1999, Nightingale et al. 2000, Sweeney et al. 2006). We utilize the relationship between k (in m/d), the 10-m wind speed (u) and the Schmidt number (Sc) of Wanninkhof (1992):

− 1 ⎛ Sc ⎞ 2 k = 0.0744u2⎜ ⎟ (8) ⎝ 660⎠ where the wind-speed also comes from Kanamitsu et al. (2002). We choose the Wanninkhof (1992) formulation first of all as it is the most widely used. It has furthermore been identified as highly suitable for large-scale applications (Najjar and Keeling 2000), and found to agree well with mixed layer dynamics of CO2 and O2 (Körtzinger et al. 2008).

2.4.2 Carbon dioxide

For CO2, the concentration difference can be expressed through pCO2 according to:

F = K k() pCOatm − pCOsw (9) CO2 0 2 2

where K0 is the solubility of CO2 in seawater, k is the transfer velocity according to (8) with the

atm sw Schmidt number for CO2 from Wanninkhof (1992), and pCO2 and pCO2 represents the partial pressures of CO2 in the atmosphere and the ocean, respectively. K0 was calculated, using SST, through the relationship of Weiss (1974). Measurements of the atmospheric mole fraction of CO2

(XCO2), for 2008, were obtained from the NOAA CMDL Carbon Cycle Group flask sampling in atm order to determine pCO2 . The stations used are shown in figure 3. As suggested by Olsen et al.

(2003), atmospheric XCO2 may be parameterized as a function of latitude, however based on the

6 Manuscript in preparation.

small spread (i.e. the small standard deviation) in the XCO2 data from the stations used in this work

atm (not shown), constant XCO2 values were utilized for each month of the year. pCO2 was then calculated from XCO2 according to: atm = ()− pCO2 XCO2 pb pH2O (10)

where pb is the barometric sea level pressure from Kanamitsu et al. (2002), and pH2O is the water vapor pressure calculated from SST according to Cooper et al. (1998).

The APO flux is calculated from the combined fluxes as (Körtzinger et al. 2008):

F =−F −1.1F (11) APO O2 CO2

A negative flux represents an APO transport into the ocean, meaning that the atmosphere is losing matter.

3. Results

3.1 Air-Sea gas exchange

3.1.1 Oxygen The 6-h wind speed from the NCEP-DOE AMIP II reanalysis (Kanamitsu et al. 2002) for the year

2008, merged to float locations in space and time, was combined with float O2 measurements from the top 5m of the ocean for calculating the fluxes of oxygen using (7) and (8). Figure 5 shows the corresponding air-sea fluxes (mol m-2 d-1) for NEMO60 and NEMO47, respectively. Both plots depict clear annual cycles (solid lines representing a 60 day running mean of the total O2 flux), with outgassing of O2 during summer and O2 uptake during winter. However there appears to be larger variability in the region covered by NEMO47. The summertime outgassing results from a combination of surface warming (through increased net heat flux) and biological activity.

3.1.2 Carbon dioxide The flux of carbon dioxide was calculated through equations (1), (2), (8), (9) and (10), using the same atmospheric data as above. The results are shown in figure 6. As for the O2 flux there appears to be larger variability in the CO2 flux in the region covered by NEMO47, however both regions appear to be year-round sinks for atmospheric CO2.

7 Manuscript in preparation.

3.1.3 Atmospheric Potential Oxygen The APO flux, as calculated through equation (11), is shown in figure 7. By comparing fig. 7 with figures 5 and 6, respectively, it is evident that the APO flux is mostly dominated by the O2 flux, and to a lesser extent by the CO2 flux, as expected. Atmospheric APO has been calculated over the last few decades, from measurements obtained through the Scripps O2/N2 sampling network (http://scrippso2.ucsd.edu/apo-data). Figure 8 shows a plot of the APO for 2008 from Palmer Station, Antarctica (PSA; see fig. 3 for location). The seasonal APO cycle depicted in fig. 8, agrees well with the calculated air-sea APO fluxes in fig.7; atmospheric APO decreases during fall and winter as APO is transported from the atmosphere to the ocean, while atmospheric APO starts to increase during spring (~Oct.) where the APO flux is directed from the ocean to the atmosphere.

4. Discussion The waters of the South Atlantic sector of the Southern Ocean are separated from the saltier and warmer waters of the subtropical South Atlantic by strong meridional gradients in surface properties (Orsi et al. 1995). This separation is commonly known as the Subtropical Front (STF). The Antarctic Circumpolar Current (ACC) flows in an eastward manner around the globe to the south of the STF, with two additional fronts located within. These fronts are called the Subantarctic Front (SAF) and the Polar Front (PF), respectively (Orsi et al. 1995). The two distinct zonal bands separated by the fronts are called, from north to south, the subantarctic zone (SAZ), and the Polar Zone (PZ). The climatological positions (Orsi et al. 1995) of these fronts together with the respective stations sampled by the NEMO floats are shown in figure 9, indicating that NEMO47 sampled mostly in the SAZ, while NEMO60 collected approximately the same number of profiles in the SAZ as in PZ. The mean air-sea CO2 flux for 2008 as calculated from the NEMO47 data, was -2 -1 -2 -1 15.2±1.7 mmol m d . The mean O2 flux was 0.005±0.008 mol m d , and the consequent APO flux was calculated as -0.022±1.7 mol m-2 d-1. The calculations of the respective fluxes across the -2 -1 -2 -1 NEMO60 drift track resulted in a CO2 flux of 6.8±1.3 mmol m d , an O2 flux of -0.03 mol m d , and a resultant APO flux of 0.021±1.3 mol m-2 d-1. When averaging the fluxes over the areas -2 -1 covered by both floats, we identify a mean flux of CO2 of 11.2±1.3 mmol m d , a mean flux of O2 -2 -1 -2 -1 of -0.012±0.006 mol m d , and a mean APO flux of -0.001±0.006 mol m d . The combined O2 flux of 4.4±2.2 mol m-2 yr-1 (0.012±0.006 mol m-2 d-1) from the ocean to the atmosphere, is consistent with the inversion study of Gruber et al. (2001) identifying an O2 flux of about 5.2 mol m-2 yr-1 to the atmosphere for the subpolar South Atlantic in the region 36°-58°S (their fig. 5). The overall agreement between measured and estimated pCO2 (fig. 4), together with the combined O2 fluxes, provides confidence in our calculated APO fluxes. 8 Manuscript in preparation.

5. Conclusions This study highlights the versatility of using autonomous profiling floats for in situ ocean observing. We have demonstrated the possibility of estimating fluxes of atmospheric potential oxygen from floats, a parameter providing important constraint needed to improve global biogeochemistry models. This supports the current effort in supplying profiling floats with oxygen sensors, as formulated through a community white paper for the OceanObs´09 conference (Gruber et al. 2009). The approach described in this work warrants comparison with simultaneous in situ pCO2 and O2 measurements, promoting the inclusion of oxygen sensors on CARIOCA surface drifters. Our results furthermore highlight the regional variability in the South Atlantic Southern Ocean.

Acknowledgements Craig Neill and Kelly R. Brown are acknowledged for performing the extended optode calibrations. GN acknowledges financial support through an endowment by Mr. Trond Mohn, c/o Frank Mohn A/S, as well as through the Nansen Environmental and Remote Sensing Center in supplying the oxygen optodes. The authors also acknowledge financial support through the EU SOBER project (grant XXXX), BIAC.

9 Manuscript in preparation.

References: Bleck, R. (2002). "An oceanic general circulation model framed in hybrid isopycnic-Cartesian coordinates." Ocean Modelling 4(1): 55-88.

Cooper, D. J., A. J. Watson and R. J. Ling (1998). "Variation of pCO2 along a North Atlantic shipping route (UK to the Caribbean): a year of automated observations." Marine Chemistry 60: 147-164. Davis, R. E., J. T. Sherman and J. Dufour (2001). "Profiling ALACEs and other advances in autonomous subsurface floats." Journal of Atmospheric and Oceanic Technology 18(6): 982-993. Gruber, N., S. C. Doney, S. R. Emerson, D. Gilbert, T. Kobayashi, A. Kortzinger, G. C. Johnson, K. S. Johnson, S. C. Riser and O. Ulloa (2009). "Adding oxygen to Argo: Developing a global in situ observatory for ocean deoxygenation and biogeochemistry." Ocean Obs'09. Community white paper Version 20.7.09. Gruber, N., M. Gloor, S. M. Fan and J. L. Sarmiento (2001). "Air-sea flux of oxygen estimated from bulk data: Implications for the marine and atmospheric oxygen cycles." Global Biogeochemical Cycles 15(4): 783-803. Gruber, N., M. Gloor, S. E. M. Fletcher, S. C. Doney, S. Dutkiewicz, M. J. Follows, M. Gerber, A. R. Jacobson, F. Joos, K. Lindsay, D. Menemenlis, A. Mouchet, S. A. Muller, J. L. Sarmiento and T. Takahashi (2009a). "Oceanic sources, sinks, and transport of atmospheric

CO2." Global Biogeochemical Cycles 23. Hanawa, K. and L. D. Talley (2001). Mode Waters. Ocean Circulation and Climate: Observing and Modelling the Global Ocean. G. Siedler, J. Church and J. Gould. San Diego, California, Academic: 373-386. Ito, T., M. Woloszyn and M. Mazloff (2010). "Anthropogenic carbon dioxide transport in the Southern Ocean driven by Ekman flow." Nature 463(7277): 80-U85. Kanamitsu, M., W. Ebisuzaki, J. Woollen, S.-K. Yang, J. J. Hnilo, M. Fioorino and G. L. Potter (2002). "NCEP-DEO AMIP-II Reanalysis (R-2)." Bul. of the Atmos. Met. Soc.: 1631-1643.

10 Manuscript in preparation.

Keeling, R. F., A. C. Manning, E. M. McEvoy and S. R. Shertz (1998). "Methods for measuring

changes in atmospheric O2 concentration and their application in southern hemisphere air." Journal of Geophysical Research-Atmospheres 103(D3): 3381-3397. Keeling, R. F. and S. R. Shertz (1992). "Seasonal and Interannual Variations in Atmospheric Oxygen and Implications for the Global Carbon-Cycle." Nature 358(6389): 723-727. Kortzinger, A., J. Schimanski and U. Send (2005). "High quality oxygen measurements from profiling floats: A promising new technique." Journal of Atmospheric and Oceanic Technology 22(3): 302-308. Kortzinger, A., J. Schimanski, U. Send and D. Wallace (2004). "The ocean takes a deep breath." Science 306(5700): 1337-1337. Kortzinger, A., U. Send, D. W. R. Wallace, J. Karstensen and M. DeGrandpre (2008). "Seasonal

cycle of O2 and pCO2 in the central Labrador Sea: Atmospheric, biological, and physical implications." Global Biogeochemical Cycles 22(1). Lee, K., R. Wanninkhof, T. Takahashi, S. C. Doney and R. A. Feely (1998). "Low interannual variability in recent oceanic uptake of atmospheric carbon dioxide." Nature 396(6707): 155- 159. Liss, P. S. and L. Merlivat (1986). Sea-air gas exchange rates: Introduction and synthesis. The Role of sea-air Exchange in Geochemical Cycling. P. Buat-Ménard. Norwell, Mass., D. Reidel: 113-128. Liss, P. S. and P. G. Slater (1974). "Flux of Gases across Air-Sea Interface." Nature 247(5438): 181-184. McNeil, B. I., N. Metzl, R. M. Key, R. J. Matear and A. Corbiere (2007). "An empirical estimate of

the Southern Ocean air-sea CO2 flux." Global Biogeochemical Cycles 21(3). Naegler, T., P. Ciais, J. C. Orr, O. Aumont and C. Rodenbeck (2007). "On evaluating ocean models with atmospheric potential oxygen." Tellus Series B-Chemical and Physical Meteorology 59(1): 138-156. Najjar, R. G. and R. F. Keeling (1997). "Analysis of the mean annual cycle of the dissolved oxygen anomaly in the World Ocean." Journal of Marine Research 55(1): 117-151. Najjar, R. G. and R. F. Keeling (2000). "Mean annual cycle of the air-sea oxygen flux: A global view." Global Biogeochemical Cycles 14(2): 573-584. Nightingale, P. D., G. Malin, C. S. Law, A. J. Watson, P. S. Liss, M. I. Liddicoat, J. Boutin and R. C. Upstill-Goddard (2000). "In situ evaluation of air-sea gas exchange parameterizations using novel conservative and volatile tracers." Global Biogeochemical Cycles 14(1): 373- 387.

11 Manuscript in preparation.

Nondal, G., K. M. Assmann, and R. G. J. Bellerby: Signals of Climate Change in Antarctic Intermediate Water, Manuscript to be submitted to Nature. Oberhuber, J. M. (1988). An atlas based on the COADS set. The budgets of heat, buoyancy and turbulent kinetic energy at the surface of the global ocean., Max-Planck-Inst. für Meteorol. Hamburg, Germany. Rep. 15: 1-20. Olsen, A., R. G. J. Bellerby, T. Johannessen, A. M. Omar and I. Skjelvan (2003). "Interannual variability in the wintertime air-sea flux of carbon dioxide in the northern North Atlantic, 1981-2001." Deep-Sea Research Part I-Oceanographic Research Papers 50(10-11): 1323- 1338. Orsi, A. H., T. Whitworth and W. D. Nowlin (1995). "On the Meridional Extent and Fronts of the Antarctic Circumpolar Current." Deep-Sea Research Part I-Oceanographic Research Papers 42(5): 641-673. Rintoul, S. R. (1991). "South Atlantic interbasin exchange." Journal of Geophysical Research 96: 2675-2692. Rodenbeck, C., C. Le Quere, M. Heimann and R. F. Keeling (2008). "Interannual variability in

oceanic biogeochemical processes inferred by inversion of atmospheric O2/N2 and CO2 data." Tellus Series B-Chemical and Physical Meteorology 60(5): 685-705. Schneider, W. and L. Bravo (2006). "Argo profiling floats document Subantarctic Mode Water formation west of Drake Passage." Geophysical Research Letters 33(16): 4 pp.

Severinghaus, J. P. (1995). Studies of the terrestrial O2 and carbon cycle in sand dune gases and in Biosphere 2. PhD thesis., Columbia University, New York. Stephens, B. B., R. F. Keeling, M. Heimann, K. D. Six, R. Murnane and K. Caldeira (1998).

"Testing global ocean carbon cycle models using measurements of atmospheric O2 and CO2 concentration." Global Biogeochemical Cycles 12(2): 213-230. Stramma, L. and R. G. Peterson (1990). "The South-Atlantic Current." Journal of Physical Oceanography 20(6): 846-859. Sweeney, C., E. Gloor, A. R. Jacobson, R. M. Key, G. McKinley, J. L. Sarmiento and R.

Wanninkhof (2007). "Constraining global air-sea gas exchange for CO2 with recent bomb C-14 measurements." Global Biogeochemical Cycles 21(2). Takahashi, T., S. C. Sutherland and A. Kozyr (2010). "Global oceam surface water partial pressure

12 Manuscript in preparation.

Takahashi, T., S. C. Sutherland, R. Wanninkhof, C. Sweeney, R. A. Feely, D. W. Chipman, B. Hales, G. Friederich, F. Chavez, C. Sabine, A. Watson, D. C. E. Bakker, U. Schuster, N. Metzl, H. Yoshikawa-Inoue, M. Ishii, T. Midorikawa, Y. Nojiri, A. Kortzinger, T. Steinhoff, M. Hoppema, J. Olafsson, T. S. Arnarson, B. Tilbrook, T. Johannessen, A. Olsen, R. Bellerby, C. S. Wong, B. Delille, N. R. Bates and H. J. W. de Baar (2009). "Climatological

mean and decadal change in surface ocean pCO2, and net sea-air CO2 flux over the global oceans." Deep-Sea Research Part Ii-Topical Studies in Oceanography 56(8-10): 554-577. Tans, P. P., I. Y. Fung and T. Takahashi (1990). "Observational Constraints on the Global

Atmospheric CO2 Budget." Science 247(4949): 1431-1438. Tengberg, A., J. Hovdenes, H. J. Andersson, O. Brocandel, R. Diaz, D. Hebert, T. Arnerich, C. Huber, A. Kortzinger, A. Khripounoff, F. Rey, C. Ronning, J. Schimanski, S. Sommer and A. Stangelmayer (2006). "Evaluation of a lifetime-based optode to measure oxygen in aquatic systems." Limnology and Oceanography-Methods 4: 7-17. Tjiputra, J. F., K. Assmann and C. Heinze (2010). "Anthropogenic carbon dynamics in the changing ocean." Ocean Science 6(3): 605-614. Wanninkhof, R. (1992). "Relationship between Wind-Speed and Gas-Exchange over the Ocean." Journal of Geophysical Research-Oceans 97(C5): 7373-7382.

Wanninkhof, R. and W. R. McGillis (1999). "A cubic relationship between air-sea CO2 exchange and wind speed." Geophysical Research Letters 26(13): 1889-1892. Weiss, R. F. (1970). "Solubility of Nitrogen, Oxygen and Argon in Water and Seawater." Deep-Sea Research 17(4): 721. Weiss, R. F. (1974). "Carbon dioxide in water and sea water: The solubility of a non ideal gas." Marine Chemistry 2: 203-215.

13 Manuscript in preparation.

Figures and figure captions:

Fig. 1 Float deployment location is shown by the black square while the blue diamonds denote the NEMO60 stations. The red stars denotes the stations from NEMO47.

14 Manuscript in preparation.

. Fig. 2 Map showing spaghetti plots of the synthetic float drift from experiment 1 (black lines), experiment 2 (blue lines), and experiment 3 (red lines).

Fig. 3 Map showing the locations of the LDEO surface pCO2 winter- (blue) and summer data (red) used for obtaining

the relationships given in equations (3) and (4).The locations of the NOAA XCO2 sampling stations used in this work are shown with black squares; EIC=Easter Island, CGO=Cape Grim Observatory, CRZ=Crozet Island, TDF=Terra del Fuego, and PSA=Palmer Station Antarctica.

15 Manuscript in preparation.

2008,t=5 Fig. 4 The top panel shows the difference between pCO2 and estimated pCO2 from SST, SSS, longitude and lattitude through (1). Equation (1) is able to estimate pCO2 with a mean bias of 3.2 μatm, and an rms error of 6.6 μatm.

The bottom plot displays the difference between measured and estimated (through eq. 2) pCO2 for winter. Equation (2)

is able to estimate pCO2 with a mean bias of -0.2μatm, and an rms error of 6.4 μatm.

16 Manuscript in preparation.

-2 -1 Fig. 5 Air-sea fluxes of O2 (mol m d ) for NEMO47 (top plot) and NEMO60 (bottom plot) shown versus month of the composite year 2008. The solid black lines show a 60day running mean of the total flux.

17 Manuscript in preparation.

-2 -1 Fig. 6 Air-sea fluxes of CO2 (mmol m d ) for NEMO47 (top plot) and NEMO60 (bottom plot) shown versus month of the composite year 2008. The solid black lines show a 60day running mean of the total flux. 18 Manuscript in preparation.

Fig. 7 Figure showing APO flux (mol m-2 d-1) versus month of the year for NEMO47 (top plot) and NEMO60 (bottom plot). The solid lines represent 60 days moving averages of the data.

Fig. 8 APO (per meg) from Palmer Station, Antarctica, for the year 2008. Data obtained from http://scrippso2.ucsd.edu/apo-data.

19 Manuscript in preparation.

(b)

Fig. 9 (a) Location of NEMO47 measurements and climatological positions of the major South Atlantic Southern Ocean fronts from Orsi et al. (1995) (b) shows the corresponding for NEMO60.

20 Paper 3

Signals of Climate Change in Antarctic Intermediate Water. Nondal, G., K. M. Assmann, and R. G. J. Bellerby. To be submitted to Nature. III

Signals of Climate Change in Antarctic Intermediate Water

Gisle Nondal (1,2,3,4*), Karen M. Assmann (3,4), and Richard G. J. Bellerby (3,4,2)

1Mohn-Sverdrup Center/Nansen Center, Thormøhlensgate 47, N-5006, Bergen, Norway 2Geophysical Institute, University of Bergen, , Allegaten 55, N-5007, Bergen, Norway 3Uni Bjerknes Centre, Allegaten 55, N-5007, Bergen, Norway 4Bjerknes Centre for Climate Research, , Allegaten 55, N-5007, Bergen, Norway *To whom correspondence should be addressed. E-mail: [email protected]

Author contact information Cand. scient. Gisle Nondal Phone: +47 55 20 58 25, +47 95 77 18 46 E-mail: [email protected]

Dr. Karen M. Assmann Phone: + 47 55 58 37 09 E-mail: [email protected]

Prof. Dr. Richard G. J. Bellerby Phone: + 47 55 58 25 65, +47 90 89 92 56 E-mail: [email protected]

1 Signals of climate change and variability are communicated to the ocean interior through mixing and the sinking of surface water masses. The most prominent pathways are the generation of Southern Hemisphere mode and intermediate waters1. Through this ventilation, the Southern Ocean has a major role in controlling atmospheric carbon dioxide2-4, oxygenating the ocean interior5, and supplying nutrients to the global ocean6.

We present observations in the Southwestern Atlantic that show the fastest decrease in the oxygen content of Southern Hemisphere intermediate waters observed to date

(6.5μmol kg-1 per decade over the past two decades). A model experiment using a state-of- the-art ocean circulation – carbon cycle model7 over the same time period shows that this oxygen reduction extends throughout the Southeastern Pacific and South Atlantic. The model displays an intensification of Circumpolar Deep Water upwelling, a reduction in wintertime convection, and a subsequent shift in meridional overturning circulation. We also find that increasing carbon and nutrient fluxes from the Southern Ocean to the global ocean are associated with the change in ocean circulation.

Intermediate water masses originating in the Southern Ocean are sensitive indicators of climate change and variability1. They distribute nutrients to the World Ocean and influence global ocean productivity6, oxygenate the intermediate ocean countering the oxygen utilization of marine

8 microbial processes , and are an important conduit of anthropogenic CO2 (Cant) from the atmosphere to the interior ocean2-4. These water masses have experienced a significant

9 warming , increasing Cant (refs. 2,3), and regionally diminishing dissolved oxygen concentrations

(O2; ref. 5) presumably in response to climate change.

Being identifiable in all ocean basins10, Antarctic Intermediate Water (AAIW) is the major intermediate water mass with regards to global ocean connectivity. It is primarily formed at the Antarctic polar front west of Drake Passage from winter surface water originating in the

Bellingshausen Sea11, and subsequently transported into the South Pacific12, and through Drake

Passage, into the South Atlantic Ocean13 (Fig. 1). 2

We investigated changes in the circulation and biogeochemical transport of AAIW, combining observations of O2, temperature (T) and salinity (S) from autonomous floats in the western South Atlantic Ocean (Methods), and a newly developed state-of-the-art coupled ocean general circulation – carbon cycle model7. The model is configured on a global grid (Methods), with the water column divided into layers of constant density (isopycnals). Such a setup is favorable, as the vertical coordinate mimics the real structure of the water column, and tracer transport in the interior ocean is know to take place along such isopycnal surfaces.

Sub-surface O2 is likely to be more sensitive to changes in circulation and ventilation compared with other physical and biogeochemical variables14, and it is furthermore a powerful water mass tracer. We compared the new float O2 dataset from 2008 with observations from

15 1988 (GLODAP ; data locations shown in Fig. 1), and identified a significant AAIW O2 reduction of 12.9±4.2μmol kg-1 (Fig. 2; see also Methods). This corresponds to 6.5μmol kg-1 per decade and is, to the best of our knowledge, the fastest observed O2 decline for Southern

Hemisphere waters16.

The model simulates a similar statistically significant AAIW O2 reduction as identified from the observations over the past two decades (Fig. 2 and top plot Fig. 3, see also Methods), with concomitant increasing trends in total inorganic carbon (Ct), and nutrients over the intervening years (Fig. 3). The modeled O2 decrease can be traced into the AAIW formation region west of Drake Passage, which shows the largest change (Fig.4). From there it spreads northward in the Pacific and through Drake Passage into the Atlantic where it propagates eastward through the region where our observations were taken.

AAIW in the Southeast Pacific is made up by wintertime Antarctic Surface Water mixing with upwelled Circumpolar Deep Water (CDW), with the end-product (AAIW) subducting at the

Antarctic polar front. Over recent decades, the Southern Annular Mode (SAM), the main mode of southern hemisphere atmospheric variability (e.g. ref. 17), has shifted to a positive state associated with lower sea-level pressures than normal over Antarctica and higher sea level pressures over mid-latitudes (40-65°S). This is associated with an intensification and southward 3 shift of the Southern Hemisphere westerly winds, attributed to a combination of greenhouse-gas induced warming18, and ozone loss over the Antarctic continent17,19. As a response to the shifting westerlies, the northward water mass transport in the near surface layers (i.e. the Ekman transport) increases, leading to enhanced upwelling south of 50°S (Supplementary Fig. 1). This is enhancing the supply of CDW from the ocean interior. As this water mass was last in contact with the atmosphere more than five centuries ago20, it contains insignificant amounts of carbon originating from fossil fuel combustion21. Increased upwelling of CDW thus brings water with more natural carbon and nutrients, but less O2 to the surface, hence introducing a distinct biogeochemical signature. Gas exchange and biological production processes both act to partially normalize the carbon and O2 disequilibria, the extent of which is strongly dictated by the residence time of water at the surface.

Over the past 50 years, negative trends in simulated wintertime surface densities

(Supplementary Fig. 2) suggest a decrease in wintertime convection. Together with a faster

CDW⎯to⎯AAIW transformation as a result of the spin-up of the near surface meridional

22 circulation (i.e. the Deacon Cell ), and thus a shorter time for air-sea gas exchange of O2 and

CO2 at the surface, the combined result is formation of AAIW that retains stronger CDW characteristics. This mechanism implies a weakening potential for CO2 uptake from the atmosphere in the region of AAIW formation, due to increasing natural carbon flux from the ocean interior. This is a likely contributor to the modeled23, and observed24, reduction in the efficiency of the total Southern Ocean atmospheric CO2 sink. Despite reduced CO2 uptake from the atmosphere, AAIW remains a net carbon sink as it retains more natural carbon due to the reduced air-sea gas exchange (less outgassing) at the surface. This enhances ocean acidification

(lowers pH; not shown) despite the reduced net carbon uptake from the atmosphere.

We show, by employing O2 as a tracer, that the excess carbon in the Southeast Pacific resulting from enhanced CDW upwelling is transported through AAIW into the ocean basins.

Our results significantly add to the present state of knowledge with regards to Southern Ocean carbon transport. We are able to extend the short-term (2005-2006) modeling study of ref. (4) 4 identifying that the carbon transport has been dominated by Ekman dynamics since, at least,

1960.

The future scale and persistence of the circulation and biogeochemical changes is dependent on the intensity and position of the wind fields19. These are sensitive to both climate change25,26 and stratospheric ozone concentrations17,19. Although the stratospheric ozone levels are projected to recover during the course of the 21st century, the ever increasing atmospheric load of greenhouse gases will likely lead to a continuation of the current trend in the SAM, and therefore to a strengthening and southward shift in the Southern Hemisphere westerlies18. This will enhance the nutrient transport to the global ocean, potentially affecting primary productivity6.

5 Methods

Profiling floats

In situ measurements were made by means of Navigating European Marine Observer (NEMO) floats. (http://www.optimare.de/cms/en/divisions/mms/mms-products/nemo.html). In “park and profile mode”, these floats drift at 650m, descending to 1000m every 10 days, subsequently rising to the surface measuring T, S, and O2. At the surface the floats transmit data by means of

Iridium satellite communication, before returning to the parking depth.

NEMO O2 measurements

Each float was equipped with an optical oxygen sensor (Aanderaa Data Instruments Oxygen

Optode model 3830). Comparing NEMO stations with available in situ measurements shows that

-1 the O2 measurements are consistent within 2.8µmol kg , which we thus state as the accuracy of the measurements.

Comparing NEMO and Southern Atlantic Ventilation Experiment (SAVE) O2 data

The NEMO profiles from 2008 were interpolated onto common depths, averaged over a full year, and compared with the SAVE O2 measurements from 1988. The interpolation was performed using a Piecewise Cubic Hermite Interpolating scheme (pchip) in Matlab®, as suggested by ref. (29).

Observations of S, T and Si* – the latter a tracer that represents the relative depletion of silicate and nitrate6 –, obtained during the World Ocean Circulation Experiment (WOCE), identify the 700-1000m depth range as Antarctic Intermediate Water (AAIW). Using the same criteria for model analysis, we identify the layer having a constant density of 36.53 (σ2;

Methods) as the best model equivalent in both water mass characteristics and depth.

Statistical analysis of float data

Random “noise” on the order of ± 2.8µmol kg-1 for NEMO (optode accuracy), and ± 1µmol kg-1 for GLODAP (typical accuracy of wet-chemical titration), was introduced to the measurements.

We then performed a Student’s t-test 10000 times in Matlab®, using the ttest2 algorithm. This showed that the O2 difference in the depth range of 700-1000m was always significantly different than zero at the 95% level (for n=10000).

MICOM – HAMOCC coupled model system

The model is a coupling between the Miami Isopycnic Coordinate Ocean Model (MICOM), and the Hamburg Model of Ocean Carbon Cycling (HAMOCC). It is configured on a global grid with a nominal resolution of 2.4° with grid spacing ranging from 60km in the Arctic and

Southern Oceans, to 180km in the Subtropical Gyres. The water column is divided into a stack of

34 layers of constant density (isopycnals) referenced to 2000m depth (σ2 – coordinates), with a non-isopycnic surface mixed layer. The model was spun up for 950 years in order to reach equilibrium. The results analyzed in this study are from a hindcast simulation forced by NCEP reanalyzes (1960 – 2007).

Supplementary Information is linked to the online version of the paper at www.nature.com

Acknowledgements

We acknowledge financial support from the Norwegian Research Council through the Southern

Ocean Biogeochemistry, Education, and Research project (SOBER; contract no. XXXX), and the BIpolar Atlantic thermohaline Circulation project (BIAC; contract no. XXXX). GN acknowledges financial support through an endowment by Mr. Trond Mohn, c/o Frank Mohn

A/S. The authors acknowledge Knut S. Seim for discussion through the NEMO data analysis process; Laurent Bertino for comments regarding statistical methodology; and the Captain and crew on the R/V G. O. Sars, and cruise leader Henrik Søiland for deploying the NEMO floats. 7 Comments from Martin Miles, Richard Matear, Tor Eldevik, Keith Nicholls, and Peter M.

Haugan significantly improved the quality of the paper. This is publication XXXX of the

Bjerknes Centre for Climate Research.

References

1 Banks, H. T., Wood, R. A., Gregory, J. M., Johns, T. C. & Jones, G. S. Are observed decadal changes in intermediate water masses a signature of anthropogenic climate change? Geophysical Research Letters 27, 2961-2964 (2000).

2 Sabine, C. L. et al. The oceanic sink for anthropogenic CO2. Science 305, 367-371 (2004).

3 Khatiwala, S., Primeau, F. & Hall, T. Reconstruction of the history of anthropogenic CO concentrations in the ocean. Nature 462, 346-349 (2009). 4 Ito, T., Woloszyn, M. & Mazloff, M. Anthropogenic carbon dioxide transport in the Southern Ocean driven by Ekman flow. Nature 463, 80-84 (2010). 5 Matear, R. J., Hirst, A. C. & McNeil, B. I. Changes in dissolved oxygen in the Southern Ocean with climate change. Geochemistry Geophysics Geosystems 1 (2000). 6 Sarmiento, J. L., Gruber, N., Brzezinski, M. A. & Dunne, J. P. High-latitude controls of thermocline nutrients and low latitude biological productivity. Nature 427, 56-60, doi:10.1038/nature02127 (2004). 7 Assmann, K. M., Bentsen, M., Segschneider, J. & Heinze, C. An isopycnic ocean carbon cycle model. Geosci. Model Dev. 3, 143-167 (2010). 8 del Giorgio, P. A. & Duarte, C. M. Respiration in the open ocean. Nature 420, 379-384, doi:10.1038/nature01165 (2002). 9 Gille, S. T. Warming of the Southern Ocean since the 1950s. Science 295, 1275-1277 (2002). 10 Talley, L. D. in Mechanisms of Global Climate Change at Millenial Time Scales Vol. Geophys. Mono. Ser. 112 eds Clark, Webb, & Keigwin) (American Geophysical Union, 1999). 11 Garabato, A. C. N., Jullion, L., Stevens, D. P., Heywood, K. J. & King, B. A. Variability of Subantarctic Mode Water and Antarctic Intermediate Water in the Drake Passage during the Late-Twentieth and Early-Twenty-First Centuries. Journal of Climate 22, 3661-3688, doi:10.1175/2009jcli2621.1 (2009).

12 Iudicone, D., Rodgers, K. B., Schopp, R. & Madec, G. An exchange window for the injection of Antarctic Intermediate Water into the South Pacific. Journal of Physical Oceanography 37, 31-49 (2007). 13 Sloyan, B. M. & Rintoul, S. R. Circulation, renewal, and modification of Antarctic mode and intermediate water. Journal of Physical Oceanography 31, 1005-1030 (2001). 14 Körtzinger, A., Riser, S. C. & Gruber, N. Oceanic oxygen - the oceanographer's canary bird of climate change. Argonautics, 2-3 (2006). 15 Key, R. M. et al. A global ocean carbon climatology: Results from Global Data Analysis Project (GLODAP). Global Biogeochemical Cycles 18, doi:10.1029/2004gb002247 (2004). 16 Keeling, R. F., Körtzinger, A. & Gruber, N. Ocean Deoxygenation in a Warming World. Annu. Rev. Mar. Sci. 2, 199-229 (2010). 17 Thompson, D. W. J. & Solomon, S. Interpretation of recent Southern Hemisphere climate change. Science 296, 895-899 (2002). 18 Arblaster, J. M. & Meehl, G. A. Contributions of external forcings to southern annular mode trends. Journal of Climate 19, 2896-2905 (2006). 19 Toggweiler, J. R. & Russell, J. Ocean circulation in a warming climate. Nature 451, 286- 288, doi:10.1038/nature06590 (2008). 20 Toggweiler, J. R. & Samuels, B. Effect of Drake Passage on the Global Thermohaline Circulation. Deep-Sea Research Part I-Oceanographic Research Papers 42, 477-500 (1995). 21 Vazquez-Rodriguez, M. et al. Anthropogenic carbon distributions in the Atlantic Ocean: data-based estimates from the Arctic to the Antarctic. Biogeosciences 6, 439-451 (2009). 22 Webb, D. J. & de Cuevas, B. A. On the fast response of the Southern Ocean to changes in the zonal wind. Ocean Science 3, 417-427 (2007).

23 Le Quere, C. et al. Saturation of the Southern Ocean CO2 sink due to recent climate change. Science 316, 1735-1738, doi:10.1126/science.1136188 (2007). 24 Metzl, N. Decadal increase of oceanic carbon dioxide in the Southern Indian Ocean surface waters (1991-2997). Deep-Sea Research II 56, 607-619 (2009). 25 Fyfe, J. C., Boer, G. J. & Flato, G. M. The Arctic and Antarctic oscillations and their projected changes under global warming. Geophysical Research Letters 26, 1601-1604 (1999). 26 Miller, R. L., Schmidt, G. A. & Shindell, D. T. Forced annular variations in the 20th century intergovernmental panel on climate change fourth assessment report models. Journal of Geophysical Research-Atmospheres 111, doi:10.1029/2005jd006323 (2006).

9 27 Brewer, P. G. & Peltzer, E. T. Limits to Marine Life. Science 324, 347-348, doi:10.1126/science.1170756 (2009). 28 Hofmann, M. & Schnellnhuber, H.-J. Ocean acidification affects marine carbon pump and triggers extended marine oxygen holes. PNAS 106, 3017-3022 (2009). 29 Tanhua, T. et al. Quality control procedures and methods of the CARINA database. Earth. Syst. Sci. Data. Discuss. 2, 205-240 (2009).

Figures and figure legends

Fig. 1 Map showing the location of the 55 profiles obtained in 2008 by means of two NEMO subsurface drifting floats (red dots). The blue dots show the locations of the 32 profiles from the 1988 South Atlantic Ventilation

Experiment (SAVE) campaign. The dark grey arrows indicate the main ocean currents in the South Atlantic Ocean.

The green ellipse indicates the key region of Antarctic Intermediate Water (AAIW) formation, and the green thick arrows show the injection pathways of AAIW into the South Pacific, and Atlantic Oceans.

Fig. 2. The solid line shows the mean in situ O2 difference between 2008 (55 profiles) and 1988 (32 profiles) in the depth range 400–1000m, while the red shading indicates the 95% confidence interval of the mean.

Fig. 3. Simulated time series showing mean oxygen (top panel), total inorganic carbon (mid panel), and silicate

(lower panel) for the model section across 47°S (all units in μmol l-1). The dashed line in the mid panel represents a model run without CO2 emissions, hence representing “natural” DIC. This demonstrates that the DIC increase is not caused solely by an invasion of anthropogenic CO2, but includes increased natural carbon flux from the ocean interior.

-1 Fig. 4 Map showing the trend in O2 concentration (μmol l ) on model isopycnal 36.53 (Methods), in the Southern

Ocean model domain, from 1960-2007. Trends shown in colour are significant at the 99% level. The black solid line indicates the location of the Polar Front.

Supplementary information

Supporting figures and figure legends

Supplementary Fig. 1 Map showing the trend in summer (DJF) Ekman pumping velocity diagnosed from

NCEP/NCAR wind stress data from 1960–2007 in m s-1 year-1. The trend is significant at the 99% level.

Supplementary Fig. 2 Simulated trend in JAS surface density between 1960–2007 (kg m-3 year-1). The trend is significant at the 99% level.

16 Paper 4

13 Magnitude and origin of the anthropogenic CO2 increase and C Suess effect in the Nordic Seas since 1981 Olsen, A., A. M. Omar, R. G. J. Bellerby, T. Johannessen, U. Ninnemann, K. R. Brown, K. A. Olsson, J. Olafsson, G. Nondal, C. Kivimäe, C. Neill, and S. Olafsdottir Global Biogeochemical Cycles, 20, doi:10.1029/2005GB002669, 2006.

GLOBAL BIOGEOCHEMICAL CYCLES, VOL. 20, GB3027, doi:10.1029/2005GB002669, 2006 Click Here for Full Article

13 Magnitude and origin of the anthropogenic CO2 increase and C Suess effect in the Nordic seas since 1981 Are Olsen,1,2 Abdirahman M. Omar,3,4 Richard G. J. Bellerby,3,4 Truls Johannessen,1,2 Ulysses Ninnemann,5,2 Kelly R. Brown,3 K. Anders Olsson,3,4 Jon Olafsson,6 Gisle Nondal,1 Caroline Kivima¨e,1 Solveig Kringstad,1 Craig Neill,3 and Solveig Olafsdottir6 Received 12 December 2005; revised 26 May 2006; accepted 12 June 2006; published 29 September 2006. 13 13 [1] This study evaluates the anthropogenic changes of CO2 (DCant) and C(D Cant) in the Nordic seas, the northern limb of the Atlantic Meridional Overturning Circulation, that took place between 1981 and 2002/2003. The changes have been determined by comparing data obtained during the Transient Tracers in the Ocean, North Atlantic Study (TTO-NAS) with data obtained during the Nordic seas surveys of R/V Knorr in 2002 and R/V G.O. Sars in 2003 using an extended multilinear regression approach. The estimated 13 D Cant and DCant and their relationship to each other and to water mass distribution suggest that the Polar Water entering the Nordic seas from the north is undersaturated with respect to the present atmospheric anthropogenic CO2 levels and promotes a local uptake of Cant within the Nordic seas. In contrast, the Atlantic Water entering from the south appears equilibrated. It carries with it anthropogenic carbon which will be sequestered at depth as the water overturns. This preequilibration leaves no room for further uptake of Cant in the parts of the Nordic seas dominated by Atlantic Water. The upper ocean pCO2 in these regions appears to have increased at a greater rate than the atmospheric pCO2 over the last 2 decades; this is reconcilable with a large lateral advective supply of Cant. 13 Citation: Olsen, A., et al. (2006), Magnitude and origin of the anthropogenic CO2 increase and C Suess effect in the Nordic seas since 1981, Global Biogeochem. Cycles, 20, GB3027, doi:10.1029/2005GB002669.

1. Introduction approaches [Gruber et al., 1996] with inherent methodological uncertainties. Our understanding of where and when [2] The increase of the atmospheric CO concentration 2 C is taken up by the oceans, how it is transported within caused by fossil fuel combustion, cement production, and ant them, and where it is stored is therefore uncertain. This land use change is significantly dampened by oceanic CO 2 situation is now being alleviated by continued observations uptake. Current annual net oceanic CO uptake corresponds 2 which enable us to observe and quantify carbon concentra- to roughly one quarter of the fossil fuel and cement tion changes in the interior ocean associated with C input. manufacturing emissions [Prentice et al., 2001], and the ant [3] Decadal changes have been quantified in the Pacific accumulated oceanic uptake since the start of the industrial Ocean [Peng et al., 2003], the Indian Ocean [Peng et al., revolution corresponds to approximately 50% of the emitted 1998; Sabine et al., 1999], and in the Southern Ocean fossil CO [Sabine et al., 2004]. The size of the oceanic 2 [McNeil et al., 2001a] by comparing recent carbon data inventory of anthropogenic carbon (C ) is, however, hard ant with the data collected during GEOSECS and other expe- to estimate owing to the absence of oceanic carbon obser- ditions that were carried out during the 1960s and 1970s. In vations from the preindustrial era and estimates such as the the present paper observations obtained during the Transient quoted Sabine et al. [2004] have to rely on indirect Tracers in the Ocean, North Atlantic Study (TTO-NAS) of 1981 are compared with data obtained during two recent 1Geophysical Institute, University of Bergen, Bergen, Norway. surveys of the Greenland, Iceland, and Norwegian seas. 2Also at Bjerknes Centre for Climate Research, University of Bergen, This allows for the quantification and increased understand- Bergen, Norway. ing of the C input to the northernmost parts of the Atlantic 3Bjerknes Centre for Climate Research, University of Bergen, Bergen, ant Norway. Ocean over the last 2 decades. 4Also at Geophysical Institute, University of Bergen, Bergen, Norway. [4] The Greenland, Iceland, and Norwegian seas, collec- 5Department of Earth Science, University of Bergen, Bergen, Norway. tively known as the Nordic seas, are considered important 6 University of Iceland and Marine Research Institute, Reykjavik, for the natural exchange of carbon between the ocean and Iceland. the atmosphere, representing the preindustrial carbon fluxes. 6 3 À1 Copyright 2006 by the American Geophysical Union. Annually, approximately 6 Sv (Sverdrup, 10 m s )of 0886-6236/06/2005GB002669$12.00 warm and salty Atlantic Water (AW) from farther south

GB3027 1of12 13 GB3027 OLSEN ET AL.: NORDIC SEAS ANTHROPOGENIC CO2 AND C GB3027 enter the region [Hansen and Østerhus, 2000]. This water is increased in the Barents Sea as well, as presented by Omar cooled and transformed to subsurface waters within the et al. [2003]. They applied an approach that accounted for Nordic seas themselves, and also in the Arctic Ocean and changes in biology and hydrography, so their estimated associated shelf areas. The subsurface waters exit the annual mean increase of 1.3 atm yrÀ1 between 1967 Nordic seas over the Greenland-Scotland ridge and contrib- and 2001 is solely due to anthropogenic carbon. This ute to a total northern North Atlantic overturning circulation increase is slightly lower than the atmospheric increase of 15 Sv [A´ lvarez et al., 2002]. The cooling of the of 1.4 atm yrÀ1 in the same time interval. Since no northward flowing AW drives a flux of CO2 from the gradients could be observed within the Barents Sea, the atmosphere into the ocean so that the waters that overturn authors suggested that the anthropogenic carbon had been in the Nordic seas and contribute to the global oceanic deep advected into the region from the Nordic seas. water pool are loaded with carbon. As the waters eventually [6] This paper focuses on the region between the areas upwell this carbon is released back to the atmosphere, along addressed by Lefe´vre et al. [2004] and Omar et al. [2003]: with carbon that was added to the deep water from decaying the Nordic seas. The area was surveyed during the summers organic material. of 2002 and 2003 using R/V Knorr and R/V G.O. Sars. [5] The Nordic seas are also considered an important Data describing the inorganic carbon system were acquired region for oceanic uptake of Cant. The air-sea flux of during both cruises as well as data for temperature, salinity, anthropogenic CO2 is not driven by biology and meridional and nutrient concentrations. These data are compared with water transport as the preindustrial air-sea CO2 flux, but by the data acquired in the same area during the TTO-NAS of the increase of the atmospheric CO2 concentration which the summer of 1981, aiming to determine the changes of the has perturbed the natural exchanges between the ocean and Cant concentrations over the 2 decades that separate these the atmosphere. Given approximately 1 year the surface surveys. To this end an extended version of the multilinear layer of the world oceans will equilibrate with the atmo- regression (MLR) approach of Wallace [1995] is employed, sphere [Broecker and Peng, 1974] and further uptake will the eMLR approach as first introduced by Sonnerup et al. rely on exposure of older water to the atmosphere. The [2000] and later formalized by Friis et al. [2005]. The 13 12 largest air-to-sea transfer of Cant is thus considered to take eMLR method is also employed on data of the C-to- C place in regions with extensive communication between ratio, 13C, acquired during the TTO-NAS and the Sars surface and deep waters such as the tropical and Southern cruises so that the magnitude of the oceanic 13C Suess effect Ocean upwelling regions and the North Atlantic convection in the region and its relationship with the anthropogenic 13 areas. A large air-to-sea flux of anthropogenic CO2 and an CO2 changes can be estimated. The C Suess effect is a efficient surface-to-deep ocean carbon transport is expected consequence of the fact that fossil fuel CO2 is relatively to occur in the Nordic seas. Indeed, with respect to the latter, enriched with the light carbon isotope 12C so that emissions 13 12 Cant estimates do reveal deep penetration in the Nordic seas lead to a decrease of the C-to- C ratio in the atmosphere. spreading into the deep Atlantic [Gruber, 1998; Sabine et Isotopic exchange across the air-sea interface transmits this al., 2004]. As regards the air-sea flux of anthropogenic perturbation to the ocean, giving rise to the oceanic 13C carbon on the other hand, the situation is somewhat unclear. Suess effect. Information on the magnitude of this effect can Three-dimensional ocean carbon cycle models do tend to be used to quantify oceanic uptake of Cant [Quay et al., show a large air-sea Cant flux in the northern North Atlantic 1992; Tans et al., 1993; Heimann and Maier-Reimer, 1996; 13 and Nordic seas [Orr et al., 2001; Wetzel et al., 2005], but Quay et al., 2003], and the ratio of oceanic Cant and C these results have, however, recently been challenged by changes provides information on the recent atmospheric estimates of oceanic Cant transport which seem to agree exposure histories of water masses [McNeil et al., 2001b; that the largest uptake of Cant takes place in the tropical- Ko¨rtzinger et al., 2003]. to-subtropical Atlantic, so that the waters that enter the subpolar regions are already equilibrated with the atmo- 2. Data and Methods sphere’s anthropogenic CO2 concentration, leaving little room for further uptake [A´ lvarez et al., 2003; Macdonald [7] The positions of the data used in the present work are et al., 2003; Roso´n et al., 2003]. In fact, the Roso´n et presented in Figure 1. Data selection has been limited to al. [2003] estimate indicates that the subtropical-to-subpolar offshore of the 250-m isobath to avoid coastal influence, Atlantic is a source of anthropogenic CO2 to the atmo- and at/or north of the Greenland-Scotland ridge. sphere; that is, the uptake of CO2 from the atmosphere has 2.1. TTO-NAS Data been reduced since preindustrial times. This result is in agreement with output from simple model calculations [8] The TTO-NAS was carried out from April to October [Wallace, 2001; Anderson and Olsen, 2002; Bellerby et of 1981. The TTO-NAS survey consisted of 7 legs and the al., 2005]. Observations from the North Atlantic do indeed Nordic seas were surveyed during leg 5, in late July and seem to verify that the surface ocean pCO in this area has early August. The data used here were acquired at stations 2 142 to 159. The data were obtained from the Carbon increased at a greater rate than the atmospheric pCO2 over the last 20 years [Lefe´vre et al., 2004], implying a reduced Dioxide Information Analysis Center (CDIAC), Oak Ridge, uptake of atmospheric CO . However, the extent to which Tennessee. The total concentration of inorganic carbon (Ct) 2 in water was determined at sea as well as on shore after the this increase is a consequence of anthropogenic CO2 or of biological and/or temperature effects was not resolved by cruise. The at-sea determinations were carried out using Lefe´vre et al. [2004]. Surface ocean pCO appears to have potentiometric titration while the onshore determinations 2 were carried out in the late C. D. Keeling’s laboratory at

2of12 13 GB3027 OLSEN ET AL.: NORDIC SEAS ANTHROPOGENIC CO2 AND C GB3027

temperature, salinity, and nutrients data acquired during this cruise will be described by R. G. J. Bellerby et al. (manuscript in preparation, 2005). The data for Ct that are employed here were determined by gas extraction of acidified water samples followed by coulometric titration [Johnson et al., 1985, 1987]. The accuracy was set by running CRM supplied by Andrew Dickson of Scripps Institution of Oceanography, and the precision has been determined to better than ±2 mol kgÀ1 by comparison of deep samples acquired in the western Greenland basin where the Ct is regarded as being homogenous (R. G. J. Bellerby et al., manuscript in preparation, 2005).

2.3. R/V G.O. Sars Data [11] The R/V G.O. Sars cruise was carried out from September to October of 2003. The cruise covered the Barents Sea Opening, the Greenland Sea, the Faeroe- Figure 1. Map of the sampling locations in the Nordic Shetland Gap, the Faeroe Bank Channel, and the Svinøy seas. The TTO sampling locations are shown as crosses, Section that extends NW off the coast of Norway from Knorr locations as solid circles, and Sars locations as open 62°N. Data were also acquired at a few stations in the diamonds. The thin lines mark the position of the 250-m Iceland Sea (Figure 1). Temperature and salinity data were isobath. obtained using a SBE 911plus CTD profiler. The initial accuracy of the conductivity and temperature sensors were 0.0003 S/m and 0.002°C, respectively. The CTD profiler Scripps Institution of Oceanography using the gas extrac- carried two temperature and two conductivity sensors and the standard deviation of each pair of sensors averaged over tion manometric technique described by Weiss et al. [1982]. ° Subsequent analyses revealed that the Ct values determined all stations was 0.0023 C for temperature and 0.0012 for at sea were higher than the values determined on shore. This salinity. The CTD salinity data were corrected to bottle motivated the release of a revised TTO Ct dataset which samples by a polynomial fit. The standard deviation of the difference between the CTD and the bottle values was was calculated from the TTO total alkalinity (TA) and pCO2 data using the carbon system constants of Merbach et al. 0.0047 with a correlation coefficient of 0.9998. [1973] [Brewer et al., 1986]. Recently Tanhua and Wallace [12] The nitrate and silicate data were analyzed on the [2005] have reanalyzed the TTO carbon chemistry data ship using a Chemlab 3-channel autoanalyzer following from legs 2, 3, 4, and 7. They find that the TTO TA data are methods of Grasshoff [1970]. Phosphate was analyzed biased high with respect to modern data tied to Certified using a modified version of the Murphy and Riley [1962] Reference Material (CRM) and recommend that they should method. The accuracy and precision of the nutrient analyses be adjusted down by 3.6 mol kgÀ1. They moreover have been assessed by running nutrient test samples recommend that the TTO Ct values should be recalculated supplied by the QUASIMEME program [Wells and Cofino, using the adjusted TA data and currently accepted constants 1997] and laboratory reference material [Aminot and À1 Ke´rouel, 1998]. The laboratory reference material and then adjusted upward by 2.4 mol kg on the basis of comparison with the manometric Ct values obtained during was run at each station and indicates precisions (1 )of À1 À1 TTO. These recommendations were followed in the present ±0.13 mol kg for nitrate, ±0.02 mol kg for À1 work. Leg 5 Ct values thus computed were on average phosphate, and ±0.27 mol kg for silicate over the 0.26 mol kgÀ1 higher than the manometric data; this duration of the cruise. Immediately after the cruise the compares favorably with the offset between the manometric laboratory reference material was analyzed together with and revised Ct data of 2.5 mol kgÀ1, the latter being lower. test samples from the QUASIMEME program. The test 13 12 sample results were in very good agreement with the [9] The TTO-NAS data of the C-to- C ratio of dissolved inorganic carbon, 13C, were acquired from CDIAC. QUASIMEME assigned values. The data were measured by the Carbon Dioxide Research [13] Ct was determined using the same method as during Group at Scripps Institution of Oceanography using proce- the Knorr 2002 cruise. Accuracy was set by running CRM dures described by Gruber et al. [1999]. All data that were as for the Knorr cruise. The combined sample handling flagged as suspicious in the original data set were discarded, and measurement error of these data were determined to be 0.4 mol kgÀ1 by analyzing samples drawn from as well as data points where the replicate analyses differed by more than three times the estimated combined error from 10 different Niskin bottles closed at the same depth (1 ). 13 sampling and analysis, 0.12% (3X0.04%), following [14] The C samples acquired on this cruise were Gruber et al. [1999]. collected in 100-mL bottles and poisoned with HgCl2 and sealed with rubber stoppers immediately after collection. 2.2. R/V Knorr Data The samples were stored cold and in the dark. Each sample was analyzed in triplicate using a Finnigan GasBench [10] The R/V Knorr cruise was carried out in June 2002 and covered most regions of the Nordic seas (Figure 1). The coupled to a Mat 252 mass spectrometer within 3 months of

3of12 13 GB3027 OLSEN ET AL.: NORDIC SEAS ANTHROPOGENIC CO2 AND C GB3027

Table 1. Regression Coefficients and Statistics for the Predictive Equations Which Were of the Form Ct = a Â S+bÂ +c Â NO3 +dÂ 13 Si + k and C=aÂ S+bÂ +c Â NO3 +dÂ Si + k Relationship a b c d k r2 rms n 1981 Ct 65.0535 À5.9032 3.9031 À0.5033 À170.8064 0.96 5.88 274 1981 13C 0.04554 0.07225 À0.02909 0.0138 À0.09162 0.98 0.062 18 2002/03 Ct 18.75019 À3.2875 5.83105 À2.18864 1442.3472 0.95 4.62 2058 2003 13C À0.574599 0.026418 À0.067470 0.057407 21.594664 0.81 0.070 199

the cruise. On the basis of 16 samples taken below 2000 m differences in deep waters, 0.2 mol kgÀ1, indicating in the Greenland Sea with very similar physical and issues with the phosphate analyses on one of the cruises. chemical water mass characteristics, the precision has been There were no significant deep water differences between determined to be at least ±0.07% (1 ) for these results, just these cruises with respect to the other variables used in this above the long-term precision of the machine based on study. A cutoff value of 0.5 mol kgÀ1 was applied to the standard replicates (±0.05%). nitrate data to avoid the potential impact of carbon overconsumption at very low nutrient levels [Falck and 2.4. The eMLR Calculations Anderson, 2005]. The regression analyses were carried out [15] The extended multilinear regression (eMLR) using the software package R [R Development Core Team, approach builds on the multilinear regression (MLR) 2004]. The coefficients and statistics of each relationship are approach for determining Cant changes introduced by listed in Table 1. The residuals are plotted as a function of Wallace [1995] and subsequently employed on several fitted values in Figures 2a and 2b, and versus depth in occasions [Sabine et al., 1999; McNeil et al., 2001a; Peng et Figures 3a and 3b. In general the relationships predict Ct al., 2003]. The MLR method has also been used to determine the oceanic 13C Suess effect in the Southern Ocean [McNeil et al., 2001b]. The underlying idea of this method is that whereas the carbon system in itself is perturbed as Cant enters the ocean, the natural variations can be accounted for using a number of predictive parameters such as salinity, temperature, apparent oxygen utilization (AOU), nutrients, TA, etc. The approach takes advantage of the fact that changes of the Cant concentration will affect the accuracy of empirically derived functions that estimate Ct, provided the predictive parameters are only influenced by natural processes. The MLR method applies an equation, derived for one region with data obtained at one time, to predict Ct concentrations using data obtained in the same region but at a different time. Any systematic bias revealed through comparing measured with predicted Ct values is ascribed to changes in the concentration of Cant. The eMLR technique was first introduced by Sonnerup et al. [2000] for 13C data. Later on Friis et al. [2005] further formalized and reviewed the eMLR technique for Ct data. Rather than evaluating the offset between measured and predicted Ct values, the eMLR technique employs two regression equations derived from two data sets collected at different times. The two equations are applied on one set of predictive variables and differences in the predicted Ct values reveal the Cant change. This method minimizes the measurement error associated with the independent variables since they enter the prediction twice, once when they are employed in the modern relationship and once when they are employed in the historical relationship [Friis et al., 2005]. 13 [16] The predictive equations for Ct and C were derived using salinity, potential temperature (), nitrate, and silicate as independent variables. These parameters were found to be sufficient and necessary. Adding the parameters Figure 2. Residuals (measured - predicted) of (a) Ct and AOU and TA did not improve the fit. Nitrate data were (b) 13C as a function of predicted values. Residuals of the preferred to phosphate as comparison of the phosphate data Knorr/Sars regressions are shown as solid circles, and collected at the Knorr and Sars cruises revealed significant residuals of the TTO regressions as open circles.

4of12 13 GB3027 OLSEN ET AL.: NORDIC SEAS ANTHROPOGENIC CO2 AND C GB3027

Figure 3. Mean and standard deviation of residuals of (a) Ct and (b) 13C binned into intervals of 0–99 m, 100–199 m, 200–299 m, 300–499 m, and into 500-m intervals below 500 m. The residuals of the Knorr/Sars regressions are in black whereas those of the TTO regressions are in grey.

13 À1 and C to within ±6 mol kg and ±0.07% as evaluated [19] A schematic of the Nordic seas upper ocean circula- from the root mean square (rms) values (Table 1). The tion is presented in Figure 4. It is dominated by the accuracy of the Ct relationships appears uniform over northward flow of warm and saline AW in the east and the whole range of Ct values as evaluated from Figure 2. the southward flow of fresher and colder Polar Water (PW) The predictions are biased slightly high deeper than 2000 m, from the Arctic Ocean in the west. These water types meet around 1 to 2 mol kgÀ1, with the TTO results being higher and mix in the area between and form Arctic Surface Water, (Figure 3). Precision of the Ct predictions appears stable and as this water is cooled during winter several types of with depth to ±5 mol kgÀ1; this is taken as a measure of Arctic Intermediate Water (ArIW) are formed. The deep the uncertainty of the Cant estimates, except in the upper waters are not only formed locally such as Greenland Sea 100 m where precision decreases to ±10 mol kgÀ1. The Deep Water (GSDW), but are also advected into the region precision of the 13C predictions appears to be within an from the Arctic Ocean. Canadian Basin and Eurasian Basin acceptable ±0.1% at all depths, but it was not possible to Deep Waters enter the Nordic seas through the Fram Strait, evaluate deep TTO precisions due to the limited number of meet and mix with GSDW and moves south into the deep samples. Lofoten and Norwegian basins as Norwegian Sea Deep 13 [17] The anthropogenic changes of Cant (DCant) and C Water (NSDW). 13 (D Cant) were estimated as the difference between Knorr/ [20] The distributions of temperature and salinity along a Sars equations and the TTO equations when both were south-to-north section based on the TTO data are displayed applied to the TTO data. The equations were employed on in Figure 5. The positions of the stations have been the TTO data because these data define the region where all specified in Figure 4. The upper 600 m of the southernmost of the regression equations are known to be valid. station (Sta. 142) in the Faeroe Bank Channel are dominated Moreover, DCant was only determined for samples within by the relatively warm and salty (S > 35) AW entering the the salinity, , nitrate, and silicate ranges for which both the region, and below this the cold overflow water exiting to Knorr/Sars and the TTO relationships were valid, i.e., the North Atlantic prevails. Station 143 in the southern the common validity range. The same was the case for the Norwegian Basin is to the west of the Atlantic inflow which 13 D Cant estimates. These ranges are provided in Table 2. turns sharply east after crossing over the Iceland-Faeroe

3. Hydrographic Setting

[18] A description of the hydrographic features of the Table 2. Common Application Ranges of the Predictive Equations TTO section is necessary before considering the results of 13 13 for Ct and for C the DCant and D Cant calculations. The one presented here is based on the more comprehensive descriptions of the Parameter Range Ct Range 13C hydrography of the Nordic Seas by Hopkins [1991], Hansen Salinity 34.172–35.249 34.784–35.084 and Østerhus [2000], Rudels et al. [2005], and Blindheim À1.288–10.49 À1.15–8.45 and Østerhus [2005]. Nitrate 0.5–15.8 2.74–15.01 Silicate 0.2–13.4 0.75–12.25

5of12 13 GB3027 OLSEN ET AL.: NORDIC SEAS ANTHROPOGENIC CO2 AND C GB3027

Figure 4. Schematic of Nordic seas surface circulation. The solid lines indicate the flow of warm Atlantic Water, and the dashed lines show the flow of the cold Polar Water. The gray dotted line shows the position of the section given in Figures 5–8, along with the location (crosses) and numbers of the TTO stations used for its construction.

Ridge, thus salinities do not exceed 35 at this station. The 146 and 147 is the Arctic Front, north of which the upper part of this station has salinities in excess of 34.95 increased influence of PW is manifest as a substantial and captures the Recirculated Faeroe Current. This current freshening above 1000 m. The Greenland Sea appears carries AW that has recirculated and been mixed with ArIW otherwise quite homogenous with respect to both tempera- in the Norwegian Basin. The lens of water with S < 34.9 ture and salinity, but then again we stress that the resolu- between 400 and 600 meter is Norwegian Sea Arctic tions applied for salinity and temperature data are Intermediate Water (NSArIW), and NSDW dominates be- inadequate to resolve the finescale structures. As the section low this. The AW is reencountered at the next stations, 144 moves northward through station 149 over the Boreas Basin and 145, and it reaches its deepest extent at the latter of and then turns east onto the West Spitsbergen shelf, the these, in the Lofoten Basin. NSArIW is present at both of increasing Atlantic influence at intermediate depths these stations as well, as a salinity minimum at 700 and becomes evident with salinities ranging from 34.9 up to 1000 m, respectively, and below this NSDW prevails. The 35. Surface waters, though, are quite fresh throughout. In sharp gradient in salinity over the Mohns Ridge at stations the Boreas Basin, this is due to PW and particularly when

Figure 5. TTO data-based sections of (a) temperature and (b) salinity from the Faeroe Bank Channel to the West Spitsbergen Shelf along the track shown in Figure 4. The numbers on the upper x axis refer to the station numbers in Figure 4.

6of12 13 GB3027 OLSEN ET AL.: NORDIC SEAS ANTHROPOGENIC CO2 AND C GB3027

water in the Faeroe Bank Channel has DCant values in the same range as the upper NSDW, between 8 and 10 mol kgÀ1. Similar to salinity, DCant decreases sharply across the Arctic Front over the Mohns Ridge and the surface to deep ocean DCant gradient is smallest in the Greenland Sea as a result of deep convective mixing in this region. Approaching the West Spitsbergen Shelf where the influence of AW increase, the À1 DCant liesbetween12and14mol kg .

4.2. Anthropogenic d13C Changes 13 [22] The estimated changes of C attributable to the 13 13 oceanic C Suess effect, D Cant, are displayed in Figure 7. The deep parts of the Norwegian Basin and most of the Faeroe Bank Channel are excluded as they were outside the common validity range of the 13C equations. There are less surface samples, so the contours are not drawn all the way to the surface. 13 [23] The 1981 to 2003 oceanic C Suess effect exhibits substantial regional variation in the Nordic seas, and Figure 6. Estimated anthropogenic CO2 increase ( mol appears to be even more closely tied to water mass distri- kgÀ1) in the Nordic seas from 1981 to 2002/2003, along the bution than anthropogenic CO2. In the core of the AW in the track shown in Figure 4. The numbers on the upper x axis northern Norwegian Basin and the Lofoten Basin the 13C refer to the station numbers in Figure 4. The gray shading of decrease ranges between À0.4% and À0.6%, and corre- the upper 100 m indicates the higher uncertainty of the sponds to annual decreases between À0.018% yrÀ1 and D À1 À Cant estimates in this depth range (10 mol kg À0.024% yr 1. The changes are smaller in the Greenland À1 À compared to 5 mol kg elsewhere). Sea, only À0.1% which corresponds to À0.004% yr 1.As the AW is reencountered west of Spitsbergen larger changes are evident: between À0.3% and À0.4%. close to the shelf it is due to the West Spitsbergen Coastal Current, a continuation of the East Spitsbergen Current that carries water from the northern Barents Sea. 5. Discussion [24] The results on the changes of anthropogenic CO2 13 4. Results concentration and the C Suess effect over the last 20 years, the magnitude of these changes, and their spatial variation 4.1. Anthropogenic CO2 Increase [21] The estimated change in anthropogenic CO2 concentration along the TTO section between 1981 and 2002/03 is displayed in Figure 6. The largest variations of DCant are evident in the upper 100 m. This approximately represents the seasonal mixed layer and is the zone with the largest methodological uncertainty (±10 mol kgÀ1). These estimates should be interpreted with care and have been shaded gray in the figure. Below this DCant values range from less than 4 mol kgÀ1 to more than 20 mol kgÀ1 with systematic spatial variations that are fairly coherent with the water mass distribution. The largest values are associated with the Atlantic domain of the Nordic seas. In the core of the Atlantic inflow, above 600 m in the FBC, DCant lies between 18 and 20 mol kgÀ1. This corresponds to annual À1 À1 changes of around 0.9 mol kg yr . The DCant decreases slightly moving into the Norwegian Sea but values are clearly higher here than in the rest of the section; in the AW at stations 144 and 145 as well as in the recirculated AW of station 143 values are between 14 and 18 mol kgÀ1, corresponding to annual changes of 0.7 to 0.8 mol kgÀ1 yrÀ1. The values decrease with depth and in the NSDW DCant is À1 Figure 7. Estimated 13C change (%) attributable to the less than 6 mol kg . The isolines in the Norwegian Sea 13 slope northward and are deepest in the Lofoten Basin oceanic C Suess effect in the Nordic seas from 1981 to where also the Atlantic layer is deepest. The overflow 2003, along the track shown in Figure 4. The numbers on the upper x axis refer to the station numbers in Figure 4.

7of12 13 GB3027 OLSEN ET AL.: NORDIC SEAS ANTHROPOGENIC CO2 AND C GB3027 and relation to water mass distribution provides a framework that enables us to specify how anthropogenic CO2 is brought into the Nordic seas. 13 [25] The atmospheric C has decreased at a fairly steady rate of À0.027% yrÀ1 since at least 1962 up to 2003 (calculated from the Southern Hemisphere record of Francey et al. [1999], extended with the record from the Cape Grim Observatory (B. H. Vaughn and J. W. C. White, personal communication, 2005). The 13C decreased in the AW by up to À0.024% yrÀ1, so this water mass is close to equilibrium with the atmospheric 13C Suess effect. How- ever, this equilibrium cannot have been established solely within the Nordic seas because it takes 10 to 12 years to establish isotopic equilibrium between ocean and atmosphere [Broecker and Peng, 1974; Lynch-Stieglitz et al., 1995] so the transit time of about 1 year from the Iceland- Scotland gap up to the Lofoten Basin [Furevik, 2000, 2001] is insufficient. Thus the changes must have been advected laterally from farther south. The AW has been given time to equilibrate with the atmosphere as it travels toward the Figure 8. Equivalent pCO2 growth rate in the Nordic seas Nordic seas via the Gulf Stream–North Atlantic Current– 1981–2002/2003 (atm yrÀ1), along the track shown in North Atlantic Drift system. Although a large CO2 uptake Figure 4. The numbers on the upper x axis refer to the within the Nordic seas would preferentially introduce station numbers in Figure 4. The growth rate was computed isotopically light carbon to the ocean due to kinetic as the mean annual increase in CO2 partial pressure required fractionation [Lynch-Stieglitz et al., 1995], the Ct change to increase the TTO Ct data by the estimated anthropogenic À1 TTO 0 of 15 to 20 mol kg means that this kinetic effect can increase if at surface, i.e., [pCO2(Ct + DCant,At ,P,,S, TTO 0 only account for a total decrease of around À0.06% Si, PO4) À pCO2(Ct ,At,P,, S, Si, PO4)]/21.5, where (assuming starting conditions of Ct of 2160 mol kgÀ1 at preformed alkalinity (At0) was determined according to 13CofÀ1%, and a one-way addition of 15 mol kgÀ1 Nondal [2004], pressure (P) was set to zero, and S are in 13 CO2 at À10%). The present estimates for the AW C Suess situ potential temperature and salinity, and Si and PO4 were effect are similar to what has been seen upstream of the set constant to 8 and 1, respectively (typical wintertime Nordic seas. Ko¨rtzinger et al. [2003] found a mean 13C conditions). The error of the growth rate estimates was À Suess effect of À0.026% yrÀ1 between 35°N and 60°N determined to ±0.64 atm yr 1 in the upper 100 m and À using isopycnal analysis, and annual changes at Hydrostation ±0.37 atm yr 1 elsewhere by propagating the random error À1 À1 S close to Bermuda have been estimated to À0.025% yr associated with DCant (±10 and ±5 mol kg , respectively, [Bacastow et al., 1996; Gruber et al., 1999]. section 2.4) according to Dickson and Riley [1978]. Errors [26] As for anthropogenic CO2 it appears as if the AW in the other variables used for pCO2 determination goes beyond maintaining equilibrium with the atmosphere including dissociation constants, were not considered because the pCO2 growth rate in this water actually exceeds because these have the same effect on two pCO2 estimates the atmospheric growth rate based on the equivalent pCO2 and thus cancel out as the difference is computed. increase displayed in Figure 8. Data from Mauna Loa shows that from 1981 to 2002 the atmospheric CO2 concentrations increased from 339.9 ppm to 373.1 ppm and further to of probable growth rates in the inflow region of between 1.6 375.6 ppm in 2003 [Keeling and Whorf, 2005]. This gives À1 À1 À1 and 2.4 atm yr , and from 1.2 to 2.2 atm yr in the AW an annual mean atmospheric CO2 increase of 1.6 ppm yr over the time span covered by the present study (1981 to farther north and so it is not crystal clear that pCO2 growth 2002/03, 21.5 years). A similar trend was observed in data rates in the AW are exceeding those of the atmosphere. obtained within this time interval at Ocean Weather Station However, an inspection of the residuals of the MLR Mike (66°N, 2°E) in the Norwegian Sea [Tans and Conway, equations revealed no systematic bias with region or water mass properties which could ultimately have resulted in a 2005]. The equivalent pCO2 increase in the regions systematic overestimation of the AW pCO2 growth rates. dominated by AW in the Nordic Seas appears larger than À1 this. In particular, in the Faeroe Bank Channel the estimated Moreover, the ±0.37 atm yr is a measure of the random growth rates are above 2 atm yrÀ1, and in the most saline error of each individual estimate and so the error in the layers of the northern Norwegian Basin and the Lofoten gridded data presented in Figure 8 is probably smaller since Basin they range between 1.6 and 1.8 atm yrÀ1. These independent random errors tends to cancel each other during excessive growth rates appear below the upper 100 m, the averaging. This, and the fact that excessive growth rates have also been observed in the source region for the AW seasonal mixed layer. At these depths the error of the DCant estimates was ±5 mol kgÀ1 which corresponds to an entering the Nordic Seas [Lefe´vre et al., 2004; Friis et al., uncertainty of ±0.37 atm yrÀ1 in the estimated growth 2005; Omar and Olsen, 2006] lead us to conclude that it is rates (see caption of Figure 8 for details). This gives a range highly likely that the pCO2 growth rates in the AW feeding the Nordic Seas are exceeding those of the atmosphere.

8of12 13 GB3027 OLSEN ET AL.: NORDIC SEAS ANTHROPOGENIC CO2 AND C GB3027

Local uptake of anthropogenic CO2 cannot sustain these growth rates. We suggest that the excessive increase is the manifestation of a negative feedback on the North Atlantic CO2 solubility pump brought about by the advection and cooling of water loaded with anthropogenic CO2. This phenomenon can be conceptualized in a number of ways [Wallace, 2001; Anderson and Olsen, 2002; Vo¨lker et al., 2002; Bellerby et al., 2005]. We have chosen the following approach. The AW that enters the Nordic seas originates farther south. As the water moves northward it is cooled and the increased solubility induces a transfer of CO2 from the atmosphere to the ocean. This cooling-induced uptake is part of the natural preindustrial exchange of CO2 between the ocean and the atmosphere; the CO2 solubility pump. As a result of the uptake of Cant, the Ct content of the source water from the Gulf Stream system has increased. Assume that the pCO2 of the warm source waters increases at the same rate as the atmosphere; this implies a given anthropogenic change in Ct. The water moves northward Figure 9. Ratio of 1981–2002/2003 Nordic seas anthro- 13 D D13 D and carbon concentrations increase as a consequence of pogenic C and Ct changes ( Rc = Cant/ Cant)asa cooling-induced uptake. When Ct increases, the pCO2 function of salinity. Solid circles are data obtained between response for a given increase in Ct becomes larger; that is, 100 and 1000 m, and open circles are data obtained deeper the buffer capacity decreases. Therefore, as the water moves than 1000 m. The line shows the linear regression in the 100 2 northward, the effect on pCO2 of the given anthropogenic to 1000 m depth range, r = 0.93. change in Ct increases. Thus the increase in the south, tracking the atmospheric increase, may bring about changes greater than the atmospheric increase farther north. 13 CO sink [Takahashi et al., 2002], and modeling results [27] The anthropogenic changes of both C and Ct that 2 [Orr et al., 2001; Wetzel et al., 2005]. However, the impact have been estimated for the AW in the Nordic seas thus of deeper waters appears to be limited. This becomes clear indicate that this water mass enters the region preequili- through examining the ratio of the anthropogenic isotopic brated with the atmosphere. Within the Nordic seas it meets and carbon changes, DRc = D13C /DC . The atmo- with PW and these two mix. The PW has recently exited ant ant spheric perturbation history of 13C and anthropogenic CO from under the Arctic ice and has not been in contact with 2 has been similar and so one might assume similar oceanic the contemporary atmosphere for long, so both DC and ant penetration as well. This would imply that the ratio of the D13C are lower for this water mass than for the AW. The ant changes, i.e., DRc, should be constant. This theory was mixing with PW dilutes the anthropogenic carbon system invoked by Heimann and Maier-Reimer [1996], who, when signature of the AW, yielding the ties with water mass 13 introducing the ‘‘dynamic constraint’’ method for estimating distribution seen for both DCant and D C in the Nordic 13 oceanic uptake of anthropogenic CO2, determined the seas. The relationship to water mass is stronger for D C À À ant global DRc to be À0.016% (mol kg 1) 1. However, than for DC (Figures 6 and 7). Regression of DC versus ant ant because of the different equilibration times of 13C and CO , salinity in the upper 1000 to 100 m yielded an r2 of 0.40, 2 significant spatial variations in DRc may arise. In particular, which is much less than the r2 of 0.80 for D13C versus ant for situations where the surface residence time of a water salinity (not shown). We believe the different strengths of mass has been inadequate for establishment of isotopic the relationships are the consequence of the uptake of equilibrium the ratio will be less negative than for situations anthropogenic CO from the atmosphere which has 2 were isotopic equilibrium has been established. In the increased the initially low DCant of the PW after it has 13 Southern Ocean, DRcwasshowntovarybetween entered the Nordic seas, while the low D C signature À À ant À0.007% (mol kg 1) 1 for water masses with has been preserved to a greater extent. This effect is due to short surface residence time and down to À0.015% the atmosphere-ocean CO equilibration rate being ten times À À 2 (mol kg 1) 1 for waters with longer surface residence greater than the 13C equilibration rate; only 1 year is time [McNeil et al., 2001b]. Nordic seas DRc estimates are required to equilibrate the surface ocean with respect to plotted as a function of salinity in Figure 9. They take on a atmospheric CO changes whereas 10 years are required to 2 wide range of values and, noticeably, vary as a strong equilibrate with respect to 13C changes [Broecker and Peng, function of salinity above 1000 m. This variation reflects 1974; Lynch-Stieglitz et al., 1995]. Thus it appears as if the the differences and mixing between the PW and AW as influx of PW gives rise to a local sink for anthropogenic discussed above. The AW, which is the most saline water CO in the Nordic seas. 2 mass, is characterized by the most negative DRc which [28] One might ask to what extent it is not mixing with À À seems to stabilize at À0.032% (mol kg 1) 1 at the deep, rather than Polar Waters, that dilutes the anthropo- highest salinity values. The fact that this is similar to the genic signature of the AW. This is in accordance with the values observed by Ko¨rtzinger et al. [2003] in the AW classical view of the northern high latitude anthropogenic upstream of the Nordic seas and close to the value expected

9of12 13 GB3027 OLSEN ET AL.: NORDIC SEAS ANTHROPOGENIC CO2 AND C GB3027

13 for a water mass whose CO2 and C changes matches within the Nordic seas. Finally, convection in the Greenland those of the atmosphere corroborates that the AW is Sea does seem to lead to surfacing of deep water which equilibrated with the atmosphere when entering the Nordic takes up anthropogenic CO2 from the atmosphere. seas. DRc approaches zero as salinity decrease, i.e., as the [31] Given this, we expect that the air-sea flux of anthro- influence of PW increase. This confirms that the PW has not pogenic CO2 in the Nordic seas exhibits a distinct spatial been exposed to the present atmosphere for very long and so pattern that is intimately linked to surface water mass is closer to atmospheric equilibrium with respect to CO2 distribution. In the eastern part, the Atlantic region, the 13 than with respect to C. uptake of Cant from the atmosphere is probably limited, if [29] The deeper waters, defined here simply as waters not even negative; that is, it may be a source of anthropo- below 1000 m, takes on a wide range of values independent genic CO2 to the atmosphere. This latter phenomenon of salinity. We identify three reasons why this is so. First, would be apparent as a decreased undersaturation compared Nordic seas deep waters are, despite of the small variation to preindustrial times. Our findings indicate that the under- of salinity, formed from a variety of water masses at a saturation has decreased at least over the last two decades. variety of places. The deep water originates in the Arctic The area toward the west, on the other hand, across the Ocean as well as in the Nordic seas so variations in the Arctic Front, is most likely a sink of anthropogenic CO2 magnitude of the DRc are to be expected. Second, limited from the atmosphere as the surface waters of this region are and variable atmospheric exposure times for deep waters appreciably influenced by PW and deep mixing. This during convective events will introduce DRc variability. feature, i.e., a quasimeridional gradient approximately fol- 13 Finally, the relative uncertainty of the DCant and D Cant lowing the position of the Arctic Front is not evident in estimates at these depths is greater. The impact of the deep estimates of the air-sea flux of anthropogenic CO2 from waters on the variability of DRc in the upper 1000 m seems coarse-resolution ocean carbon model runs. There are dis- to be limited, as most of the points here lie on the straight crepancies between different models, but most of them seem line connecting the PW and AW. Only at salinities around to agree that there is a basin-wide uptake of anthropogenic 34.9 does this slope seem perturbed by deep mixing with CO2 in the Nordic seas [Orr et al., 2001; Wetzel et al., many points higher than the expected value. This is 2005]. This flux is brought about by deep water formation apparently a result of mixing with deeper waters, as these in the models [Wetzel et al., 2005]. The difference between typically have higher DRc than what is expected from the our inferred spatial pattern and model output supports the PW-AW mixing line. As can be seen from Figure 5, it is need for instance for better spatial resolution and improved only in the homogenous Greenland Sea that these waters representation of convective processes in order for models reach the surface and support an anthropogenic carbon to properly reproduce oceanic uptake of anthropogenic uptake. CO2. [32] When collapsing pCO2 data collected over a number 6. Conclusions and Implications of years to a 1995 climatology Takahashi et al. [2002] assumed no temporal trend of surface ocean pCO2 north of [30] The increase of the anthropogenic carbon concen- 45°N in the Atlantic. They argued that the anthropogenic 13 trations and the C decrease in the Nordic seas have been signal in surface ocean pCO2 in the region should be quantified using an extended version of the multilinear completely diluted by deep mixing. On the other hand regression approach of Wallace [1995]. The distribution of Olsen et al. [2003] assumed that the whole region followed the changes and in particular their relationship to each other the atmospheric pCO2 increase. Our results show that showed a high dependency on water masses and circulation. neither of these assumptions are totally correct. Olsen et al. This dependency has illuminated the processes that bring [2003] appear closer to the truth than Takahashi et al. anthropogenic CO2 into the Nordic seas. First, anthropo- [2002] but the temporal evolution of surface ocean pCO2 genic CO2 is advected into the region with the AW. The differs from region to region. The limited spatial coverage isotopic changes were too large to have taken place solely of the TTO data and increased random uncertainty of the within the Nordic seas where the residence time is only a DCant estimates in the upper 100 m prohibit a comprehen- few years. Advection and cooling of water from farther sive analysis, but as a first-order estimate we suggest that south in the Atlantic provides a reasonable explanation for the annual pCO2 increase varies as function of salinity 2 the excessive pCO2 increases evident in the core of the AW according to: DpCO2 = 2.074 Â S À 71.13, r = 0.6, n = 76, [Wallace, 2001; Anderson and Olsen, 2002]. The large RMS = 0.2. This equation was derived through the linear D negative Rc values of the AW corroborate this view; they regression of the annual pCO2 increase as estimated for reflect that the water has been exposed to the present Figure 8 and salinity in the upper 500 to 100 m. 100 m was atmosphere for a considerable time and are similar to the used as the upper cut off because of the increased random D Rc values observed by Ko¨rtzinger et al. [2003] south of uncertainty of the DCant estimates shallower than this. the Scotland-Greenland ridge. Second, and in contrast to the [33] The ratio of the anthropogenic isotopic and CO2 AW, in the PW the anthropogenic CO2 changes are changes, DRc has been used as a conversion factor to get relatively large compared to the isotopic changes, perhaps from estimates of the oceanic 13C Suess effect to anthropo- D best reflected in the trend of increasing Rc with decreasing genic CO2 estimates. Several authors have multiplied their salinity. This observation and the fact that the PW exits from estimates of 13C changes by À0.016% (mol kg À1)À1, the under the Arctic ice imply that the influx of PW promotes DRc estimate of Heimann and Maier-Reimer [1996], to get an uptake of anthropogenic CO2 from the atmosphere to Cant estimates [Bauch et al., 2000; Ortiz et al., 2000].

10 of 12 13 GB3027 OLSEN ET AL.: NORDIC SEAS ANTHROPOGENIC CO2 AND C GB3027

This procedure cannot be recommended since different References surface residence times of water masses produce substantial A´ lvarez, M., H. L. Bryden, F. F. Pe´rez, A. F. Rios, and G. Rosoń (2002), variations in DRc at regional scales, as was also shown by Physical and biogeochemical fluxes and net budgets in the subpolar and McNeil et al. [2001b] for the Southern Ocean. The DRc temperate North Atlantic, J. Mar. Res., 60, 191–226. A´ lvarez, M., A. F. Rios, F. F. Pe´rez, H. L. Bryden, and G. Rosoń (2003), values of the AW are exceptional, and much more negative Transports and budgets of total inorganic carbon in the subpolar and than the values determined by Tanaka et al. [2003] at 44°N temperate North Atlantic, Global Biogeochem. Cycles, 17(1), 1002, in the Pacific. This reflects the special circulation of the doi:10.1029/2002GB001881. Aminot, A., and R. Ke´rouel (1998), Pasteurization as an alternative method Atlantic. Tanaka et al. [2003] suggested that their value was for preservation of nitrate and nitrite in seawater samples, Mar. Chem., applicable in general in subpolar and polar regions; this is 61, 203–208. not correct. Anderson, L. G., and A. Olsen (2002), Air-sea flux of anthropogenic carbon dioxide in the North Atlantic, Geophys. Res. Lett., 29(17), 1835, [34] The present findings support the results of anthropo- doi:10.1029/2002GL014820. genic carbon transport calculations in that the AW brings Bacastow, R. B., C. D. Keeling, T. J. Lueker, M. Wahlen, and W. G. Mook 13 with it a large amount of anthropogenic CO2 to the northern (1996), The C Suess effect in the world surface oceans and its North Atlantic [A´ lvarez et al., 2003; Macdonald et al., implications for oceanic uptake of CO2: Analysis of observations at ´ Bermuda, Global Biogeochem. Cycles, 10, 335–346. 2003; Rosoń et al., 2003]. In particular Alvarez et al. [2003] Bauch, D., J. Carstens, G. Wefer, and J. Thiede (2000), The imprint of 13 estimated that 56% of the anthropogenic carbon sequestered anthropogenic CO2 in the Arctic Ocean: Evidence from planktic C ° data from watercolumn and sediment surfaces, Deep Sea Res., Part II, 47, north of 24.5 N was brought in by oceanic transport 1791–1808. whereas 44% was taken up from the atmosphere within the Bellerby, R. G. J., A. Olsen, T. Furevik, and L. G. Anderson (2005), region. Just how large the ratio is in the Nordic seas has yet Response of the surface ocean CO2 system in the Nordic seas and north- to be determined, but the advective flux must be significant. ern North Atlantic to climate change, in The Nordic Seas: An Integrated Perspective, Geophys. Monogr. Ser., vol. 158, edited by H. Drange et al., Given our findings it is reasonable to assume that the AW pp. 189–197, AGU, Washington D. C. is saturated with anthropogenic CO2. Temperature and Blindheim, J., and S. Østerhus (2005), The Nordic seas: Main oceano- TA of the AW entering the Nordic seas is around 8°C and graphic features, in The Nordic Seas: An Integrated Perspective, 2325 mol kgÀ1, respectively; this gives a total anthro- Geophys. Monogr. Ser., vol. 158, edited by H. Drange et al., pp. 11– À1 37, AGU, Washington D. C. pogenic CO2 concentration of 53 mol kg (atmospheric Brewer, P. G., T. Takahashi, and R. T. Williams (1986), Transient tracers in pCO2 280 atm vs. 380 atm) which is reasonable the ocean—Hydrography data and carbon dioxide systems with revised compared to Ko¨rtzinger et al. [1998]. About 6 Sv overturn carbon chemistry data, NDP-004/R1, Carbon Dioxide Inf. Anal. Cent., Oak Ridge Natl. Lab., Oak Ridge, Tenn. in the Nordic seas [Hansen and Østerhus, 2000]. Combin- Broecker, W. S., and T.-H. Peng (1974), Gas exchange rates between air ing these numbers we estimate that overturning leads to an and sea, Tellus, 26, 21–35. annual sequestration of roughly 0.12 Gt Cant in the Dickson, A. G., and J. P. Riley (1978), The effect of analytical error on the evaluation of the components of the aquatic carbon-dioxide system, Mar. Nordic seas. This is 6% of the annual oceanic accumulation Chem., 6, 77–85. of 1.8 Gt Cant [Prentice et al., 2001]. Future studies should Falck, E., and L. G. Anderson (2005), The dynamics of the carbon cycle in elaborate our initial, conceptual appraisal of how mixing the surface water of the Norwegian Sea, Mar. Chem., 94, 43–53. with PW dilutes the anthropogenic carbon system signature Francey, R. J., C. E. Allison, D. M. Etheridge, C. M. Trudinger, I. G. Enting, M. Leuenberger, R. L. Langefelds, E. Michel, and L. P. Steele of the AW within the Nordic seas and quantify the relative 13 (1999), A 1000-year high precision record of C in atmospheric CO2, importance of air-sea fluxes and oceanic transport for Tellus, Ser. B, 51, 170–193. importing anthropogenic carbon into the region. Friis, K., A. Ko¨rtzinger, J. Pa¨tsch, and D. W. R. Wallace (2005), On the temporal increase of anthropogenic CO2 in the subpolar North Atlantic, [35] What we would finally like to emphasize here is the Deep Sea Res., Part I, 52, 681–698. fact that the Cant transport can bring around the excessive Furevik, T. (2000), On anomalous sea surface temperatures in the Nordic , 1044–1053. pCO2 increase seen in the core of the AW [Anderson and Seas, J. Clim., 13 Furevik, T. (2001), Annual and interannual variability of the Atlantic Water Olsen, 2002; Wallace, 2001]. Similar large increases have temperatures in the Norwegian and Barents Seas: 1980–1996, Deep Sea been seen farther south by Lefe´vre et al. [2004], Friis et al. Res., Part I, 48, 383–404. [2005], and Omar and Olsen [2006]. This gives us reason to Grasshoff, K. (1970), A simultaneous multiple channel system for nutrient analysis in seawater with analog and digital data record, Tecnicon Q., 3, believe that the AW feature is the northernmost manifesta- 7–17. tion of a trend that occurs in all regions fed with water from Gruber, N. (1998), Anthropogenic CO2 in the Atlantic Ocean, Global Bio- the Gulf Stream; that is, surface pCO2 levels in the whole geochem. Cycles, 12, 165–191. North Atlantic Drift–Norwegian Atlantic Current system Gruber, N., J. L. Sarmiento, and T. F. Stocker (1996), An improved method for detecting anthropogenic CO2 in the oceans, Global Biogeochem. are currently approaching equilibrium with the atmosphere. Cycles, 10, 809–837. Gruber, N., C. D. Keeling, R. B. Bacastow, P. R. Guenther, T. J. Lueker, [36] Acknowledgments. Financial support from the EU through M. Wahlen, H. A. J. Meijer, W. G. Mook, and T. F. Stocker (1999), IP CARBOOCEAN (511176-2) and TRACTOR (EVK2-2000-00080) is Spatiotemporal patterns of carbon-13 in the global surface oceans and greatly appreciated. The Cape Grim Observatory 13C record from 1991 to the oceanic Suess effect, Global Biogeochem. Cycles, 13, 307–335. 2003 was kindly provided by Bruce H. Vaughn and James W.C. White at Hansen, B., and S. Østerhus (2000), North Atlantic–Nordic seas ex- the Institute of Arctic and Alpine Research, University of Colorado. The changes, Prog. Oceanogr., 45, 109–208. authors would like to thank the officers and crews of R/V Knorr and R/V Heimann, M., and E. Maier-Reimer (1996), On the relations between the G.O. Sars and the chief scientists of the Knorr cruise, Jim Swift and Bill oceanic uptake of CO2 and its carbon isotopes, Global Biogeochem. Smethie, for their cooperation and organization. Yoshie Kasajima kindly Cycles, 10, 89–110. provided the Sars temperature and salinity data, and Jim Swift kindly Hopkins, T. S. (1991), The GIN Sea: A synthesis of its physical oceano- provided the Knorr temperature, salinity, and nutrients data. James Orr and graphy and literature review 1972–1985, Earth Sci. Rev., 30, 175–318. Rolf Sonnerup provided thorough, intelligent and helpful comments on the Johnson, K. M., A. E. King, and J. M. Sieburth (1985), Coulometric TCO2 manuscript. This is contribution A 131 from the Bjerknes Centre for analyses for marine studies: An introduction, Mar. Chem., 16, 61–82. Climate Research. Johnson, K. M., J. M. Sieburth, P. J. Williams, and L. Brandstrom (1987), Coulometric total carbon dioxide analysis for marine studies: Automation and calibration, Mar. Chem., 21, 117–133.

11 of 12 13 GB3027 OLSEN ET AL.: NORDIC SEAS ANTHROPOGENIC CO2 AND C GB3027

Keeling, C. D., and T. P. Whorf (2005), Atmospheric CO2 records from Quay, P., R. Sonnerup, T. Westby, J. Strutsman, and A. McNichol (2003), sites in the SIO air sampling network, in Trends: A Compendium of Data Changes in the 13C/12C of dissolved inorganic carbon as a tracer for on Global Change, http://cdiac.esd.ornl.gov/trends/co2/sio-keel.htm, anthropogenic CO2 uptake, Global Biogeochem. Cycles, 17(1), 1004, Carbon Dioxide Inf. Anal. Cent., Oak Ridge Natl. Lab., Oak Ridge, Tenn. doi:10.1029/2001GB001817. Ko¨rtzinger, A., L. Mintrop, and J. C. Duinker (1998), On the penetration of R Development Core Team (2004), R: A Language and Environment for anthropogenic CO2 into the North Atlantic Ocean, J. Geophys. Res., 103, Statistical Computing, R Found. for Stat. Comput., Vienna. (Available at 18,681–18,689. http://www.R-project.org) Ko¨rtzinger, A., P. D. Quay, and R. E. Sonnerup (2003), Relationship be- Rosoń, G., A. F. Rios, F. F. Pe´rez, A. Lavin, and H. L. Bryden (2003), 13 tween anthropogenic CO2 and the C Suess effect in the North Atlantic Carbon distribution, fluxes, and budgets in the subtropical North Atlantic Ocean, Global Biogeochem. Cycles, 17(1), 1005, doi:10.1029/ Ocean, J. Geophys. Res., 108(C5), 3144, doi:10.1029/1999JC000047. 2001GB001427. Rudels, B., G. Bjo¨rk, J. Nilson, P. Winsor, I. Lake, and C. Nohr (2005), The Lefe´vre, N., A. J. Watson, A. Olsen, A. F. Rıós, F. F. Pe´rez, and interaction between waters from the Arctic Ocean and the Nordic Seas T. Johannessen (2004), A decrease in the sink for atmospheric CO2 in north of the Fram Strait and along the East Greenland Current: Results the North Atlantic, Geophys. Res. Lett., 31, L07306, doi:10.1029/ from the Arctic Ocean-02 Oden expedition, J. Mar. Syst., 55, 1–30. 2003GL018957. Sabine, C., R. M. Key, K. M. Johnson, F. J. Millero, A. Poisson, J. L. Lynch-Stieglitz, J., T. F. Stocker, W. S. Broecker, and R. G. Fairbanks Sarmiento, D. W. R. Wallace, and C. D. Winn (1999), Anthropogenic (1995), The influence of air-sea exchange on the isotopic composition CO2 inventory in the Indian Ocean, Global Biogeochem, Cycles, 13, of oceanic carbon: Observations and modeling, Global Biogeochem. 179–198. Cycles, 9, 653–665. Sabine, C. L., et al. (2004), The oceanic sink for anthropogenic CO2, Macdonald, A. M., M. O. Baringer, R. Wanninkhof, K. L. Lee, and D. W. Science, 305, 367–371. R. Wallace (2003), A 1998–1992 comparison of inorganic carbon and its Sonnerup, R. E., P. D. Quay, and A. P. McNichol (2000), The Indian Ocean transport across 24.5°N in the Atlantic, Deep Sea Res., Part II, 50, 3041– 13C Suess effect, Global Biogeochem. Cycles, 14, 903–916. 3064. Takahashi, T., et al. (2002), Global sea-air CO2 flux based on climatological McNeil, B. I., B. Tilbrook, and R. J. Matear (2001a), Accumulation and surface ocean pCO2, and seasonal biological and temperature effects, uptake of anthropogenic CO2 in the Southern Ocean, south of Australia Deep Sea Res., Part II, 49, 1601–1622. between 1968 and 1996, J. Geophys. Res., 106, 31,431–31,445. Tanaka, T., Y. W. Watanabe, S. Watanabe, S. Noriki, N. Tsurushima, and McNeil, B. I., R. J. Matear, and B. Tilbrook (2001b), Does carbon 13 track Y. Nojiri (2003), Oceanic Suess effect of 13C in subpolar region: The anthropogenic CO2 in the Southern Ocean?, Global Biogeochem. Cycles., North Pacific, Geophys. Res. Lett., 30(22), 2159, doi:10.1029/ 15, 597–613. 2003GL018503. Merbach, C., C. H. Culberson, J. E. Hawley, and R. M. Pytkowicz (1973), Tanhua, T., and D. W. R. Wallace (2005), Consistency of TTO-NAS inor- Measurement of apparent dissociation constants of carbonic acid in sea- ganic carbon data with modern measurements, Geophys. Res. Lett., 32, water at atmospheric pressure, Limnol. Oceanogr., 18, 533–541. L14618, doi:10.1029/2005GL023248. Murphy, J., and J. P. Riley (1962), A modified single solution method for Tans, P. P., and T. J. Conway (2005), Monthly atmospheric CO2 mixing the determination of phosphate in natural waters, Anal. Chim. Acta, 27, rations from the NOAA CMDL Carbon Cycle Cooperative Global Air 31–36. Sampling Network, 1968–2002, in Trends: A Compendium of Data on Nondal, G. (2004), The carbon dioxide system in the northern North Atlan- Global Change, http://cdiac.esd.ornl.gov/trends/co2/cmdl-flask/cmdl- tic: An assessment of the optimal variable combination for voluntary flask.html, Carbon Dioxide Inf. Anal. Cent., Oak Ridge Natl. Lab., observing ships, M. S. thesis, 71 pp., Univ. of Bergen, Bergen, Norway. Oak Ridge, Tenn. Olsen, A., R. G. J. Bellerby, T. Johannessen, A. M. Omar, and I. Skjelvan Tans, P. P., J. A. Berry, and R. F. Keeling (1993), Oceanic 13C/12C uptake (2003), Interannual variability in the wintertime air-sea flux of carbon observations: A new window on ocean CO2 uptake, Global Biogeochem. dioxide in the northern North Atlantic, 1981–2001, Deep Sea Res., Part Cycles, 14, 353–368. I, 50, 1323–1338. Vo¨lker, C., D. W. R. Wallace, and D. A. Wolf-Gladrow (2002), On the role Omar, A. M., and A. Olsen (2006), Reconstructing the time history of the of heat fluxes in the uptake of anthropogenic carbon in the North Atlan- air-sea CO2 disequilibrium and its rate of change in the eastern subpolar tic, Global Biogeochem. Cycles, 16(4), 1138, doi:10.1029/ North Atlantic, 1972–1989, Geophys. Res. Lett., 33, L04602, 2002GB001897. doi:10.1029/2005GL025425. Wallace, D. W. R. (1995), Monitoring global ocean carbon inventories, Omar, A., T. Johannessen, S. Kaltin, and A. Olsen (2003), Anthropogenic OOSDP Background Rep. 5, 54 pp., Ocean Obs. Syst. Dev. Panel, Texas increase of oceanic pCO2 in the Barents Sea surface water, J. Geophys. A&M Univ., College Station, Tex. Res., 108(C12), 3388, doi:10.1029/2002JC001628. Wallace, D. W. R. (2001), Storage and transport of excess CO2 in the Orr, J. C., et al. (2001), Estimates of anthropogenic carbon uptake from four oceans: The JGOFS/WOCE global CO2 survey, in Ocean Circulation three-dimensional global ocean models, Global Biogeochem. Cycles, 15, and Climate: Observing and Modelling the Global Ocean, edited by 43–60. G. Siedler, J. Church, and J. Gould, pp. 489–521, Elsevier, New York. Ortiz, J. D., A. C. Mix, P. A. Wheeler, and R. M. Key (2000), Anthropo- Weiss, R. F., R. A. Janke, and C. D. Keeling (1982), Seasonal effects of genic CO2 invasion into the northeast Pacific based on concurrent temperature and salinity on the partial pressure of CO2 in sea water, 13 CDIC and nutrient profiles from the California Current, Global Nature, 300, 511–513. Biogeochem. Cycles, 14, 917–929. Wells, D. E., and W. P. Cofino (1997), A review of the achievements of the Peng, T.-H., R. Wanninkhof, J. L. Bullister, R. A. Feely, and T. Takahashi EU project ‘QUASIMEME’ 1993–1996, Mar. Pollut. Bull., 35, 3–17. (1998), Quantification of decadal anthropogenic CO2 uptake in the ocean Wetzel, P., A. Winguth, and E. Maier-Reimer (2005), Sea-to-air CO2 flux based on dissolved inorganic carbon measurements, Nature, 396, 560– from 1948 to 2003: A model study, Global Biogeochem. Cycles, 19, 563. GB2005, doi:10.1029/2004GB002339. Peng, T.-H., R. Wanninkhof, and R. A. Feely (2003), Increase of anthropogenic CO2 in the Pacific Ocean over the last two decades, Deep Sea Res., Part II, 50, 3065–3082. ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀ Prentice, I. C., C. D. Farquahar, M. J. R. Fasham, M. L. Goulden, R. G. J. Bellerby, K. R. Brown, C. Neill, K. A. Olsson, and A. M. Omar, M. Heimann, V.J. Jaramillo, H. S. Kheshgi, C. Le Que´re´, R. J. Scholes, and Bjerknes Centre for Climate Research, University of Bergen, Allegaten 55, D. W. R. Wallace (2001), The carbon cycle and atmospheric carbon diox- Bergen, N-5007, Norway. ide, in Climate Change 2001: The Scientific Basis—Contribution of Work- T. Johannessen, C. Kivimaë, S. Kringstad, G. Nondal, and A. Olsen, ing Group I to the Third Assessment Report of the Intergovernmental Panel Geophysical Institute, University of Bergen, Allegaten 70, Bergen N-5007, on Climate Change, edited by J. T. Houghton et al., pp. 183–237, Norway. ([email protected]) Cambridge Univ. Press, New York. U. Ninnemann, Department of Earth Science, University of Bergen, Quay, P. D., B. Tilbrook, and C. S. Wong (1992), Oceanic uptake of fossil Allegaten 41, Bergen N-5007, Norway. fuel CO2: Carbon-13 evidence, Science, 256, 74–79. J. Olafsson and S. Olafsdottir, University of Iceland and Marine Research Institute, Sudargata 1, 101 Reykjavik, Iceland.

12 of 12 Paper 5

Enhanced biological carbon consumption in a high CO2 ocean Riebesell, U., K. G. Shulz, R. G. J. Bellerby, M. Botros, P. Fritsche, M. Meyerhöfer, G. Nondal, A. Oschlies, J. Wohlers, and E. Zöllner Nature, 450, 545-549, 2007.

Vol 450 | 22 November 2007 | doi:10.1038/nature06267 LETTERS

Enhanced biological carbon consumption in a high CO2 ocean U. Riebesell1, K. G. Schulz1, R. G. J. Bellerby2,3, M. Botros1, P. Fritsche1, M. Meyerho¨fer1, C. Neill2, G. Nondal2,3, A. Oschlies1, J. Wohlers1 &E.Zo¨llner1

The oceans have absorbed nearly half of the fossil-fuel carbon diox- by about a factor of two relative to the present value (380 matm), and 1 10 ide (CO2) emitted into the atmosphere since pre-industrial times , could increase by a factor of three by the middle of the next century . causing a measurable reduction in seawater pH and carbonate This will cause seawater pH to further drop by 0.3 and 0.6 pH units, 2 saturation .IfCO2 emissions continue to rise at current rates, respectively, in addition to the 0.12 pH-unit decrease that has occurred upper-ocean pH will decrease to levels lower than have existed since pre-industrial times3. Changes in seawater chemistry of this mag- for tens of millions of years and, critically, at a rate of change 100 nitude are expected to have adverse effects, not only on individual times greater than at any time over this period3. Recent studies have species but also at the community and ecosystem level11. shown effects of ocean acidification on a variety of marine life To investigate the effect of rising CO2 on a natural plankton com- 4–6 forms, in particular calcifying organisms . Consequences at the munity, we conducted a mesocosm CO2 perturbation study in the community to ecosystem level, in contrast, are largely unknown. Raune Fjord in southern Norway. Nine enclosures, each containing 3 Here we show that dissolved inorganic carbon consumption of a 27 m of ambient water, were aerated with CO2-enriched air to achieve natural plankton community maintained in mesocosm enclosures concentrations of 350 matm (1 3 CO2), 700 matm (2 3 CO2) and m at initial CO2 partial pressures of 350, 700 and 1,050 atm increases 1,050 matm (3 3 CO2). After nutrient addition, the development with rising CO2. The community consumed up to 39% more dis- and decline of a plankton bloom was monitored over 24 days. solved inorganic carbon at increased CO2 partial pressures com- A rapid decline in CO2 partial pressures (pCO2 ) owing to photosyn- pared to present levels, whereas nutrient uptake remained the thetic carbon fixation led to a minimum pCO2 at day 12, at which same. The stoichiometry of carbon to nitrogen drawdown point nitrate concentrations were close to exhaustion (Fig. 1). The increased from 6.0 at low CO2 to 8.0 at high CO2, thus exceeding greater drop in pCO2 at increased CO2 levels was caused by the lower the Redfield carbon:nitrogen ratio of 6.6 in today’s ocean7. This buffer capacity of sea water under these conditions. No difference excess carbon consumption was associated with higher loss of between CO2 treatments was observed in the drawdown of nitrate organic carbon from the upper layer of the stratified mesocosms. (Fig. 1c), phosphate and silicate (see Methods, Supplementary Fig. 1). If applicable to the natural environment, the observed responses Phytoplankton biomass, depicted by chlorophyll a concentrations have implications for a variety of marine biological and biogeo- (Fig. 1b), peaked on day 10. This corresponded to the maximum in chemical processes, and underscore the importance of biologically particulate organic carbon (Methods, Supplementary Fig. 2) and driven feedbacks in the ocean to global change. coincided with phosphate exhaustion. The phytoplankton bloom Throughout Earth’s history, the ocean has had a crucial role in was initially dominated by diatoms, which reached a maximum modulating atmospheric carbon dioxide through a variety of phys- 1–2 days before the peak of the bloom owing to silicate limitation ical, chemical and biological processes. The same processes are on day 9 (Fig. 2). This was followed by a dominance of the prymne- involved in the ocean’s response to anthropogenic perturbations of siophytes, primarily the coccolithophore Emiliania huxleyi, which 2 the global carbon cycle. A key process responsible for about three- had maximum cell abundances of (3.5–6.2) 3 106 cells l 1 on quarters of the surface to deep-ocean gradient in dissolved inorganic days 10–11. Whereas diatoms and prymnesiophytes accounted for carbon (DIC) is the biological carbon pump8. This transports carbon 85–90% of phytoplankton biomass during the bloom, prasinophytes, bound by photosynthesis from the sunlit surface layer to the deep dinoflagellates and cyanobacteria dominated after the decline of the ocean. Integrated over the global ocean, the biotically driven surface bloom. Neither phytoplankton composition nor succession differed , m 21 to deep-ocean DIC gradient of 220 mol kg (ref. 9) amounts to significantly between CO2 treatments. , 5 15 2,500 petagrams of carbon (Pg C) (1 Pg 10 g), that is, 3.5 times A distinct CO2 treatment effect is evident for cumulative DIC the atmospheric carbon pool. Small changes in this pool, for example, drawdown owing to production of organic matter (DDICorg, 2 caused by biological responses to ocean change, would have a strong Fig. 3a), averaging 79, 100 and 110 mmol kg 1 at the peak of the affect on atmospheric CO2. bloom (day 12) in water at 1 3 CO2,23 CO2 and 3 3 CO2 concen- At present, one of the most far-reaching global perturbations of the tration, respectively. Thus, for the same uptake of inorganic nutri- marine environment iscaused by themassiveinvasion of fossil fuelCO2 ents, net community carbon consumption under increased CO2 into the ocean, making it the second largest sink for anthropogenic exceeded present rates by 27% (2 3 CO2) and 39% (3 3 CO2). carbon dioxide after the atmosphere itself. CO2 entering the ocean Continuous oxygen measurements in mesocosms 2 (3 3 CO2), 5 21 alters the seawater carbonate equilibrium, decreasing pH and shifting (2 3 CO2) and 8 (1 3 CO2), which revealed up to 20 mmol kg 22 dissolved inorganic carbon away from carbonate (CO3 )towards higher O2 concentrations at increased CO2 (data not shown), indi- 2 more bicarbonate (HCO3 )andCO2. For a ‘business-as-usual’ emis- cate enhanced net photosynthesis to be the source of the observed sion scenario in the year 2100 (IS92a), the CO2 concentration will rise CO2 effect. Possible causes for this include increased photosynthetic

1Leibniz Institute of Marine Sciences, IFM-GEOMAR, 24105 Kiel, Germany. 2Bjerknes Centre for Climate Research, 3Geophysical Institute, University of Bergen, Alle´gaten 70, 5007 Bergen, Norway. 545 ©2007 Nature Publishing Group LETTERS NATURE | Vol 450 | 22 November 2007

carbon fixation owing to lower energetic cost of carbon acquisition, the lowest and highest CO2 treatment. Enhanced particle sinking at reduced photorespiration owing to a higher CO2:O2 ratio, and increased CO2 may explain why POC and TEP concentrations did decreased organic matter degradation, for example, owing to changes not show a significant difference between treatments. In fact, positive in food quality. correlations between TEP production and CO2 concentration were Whereas the cumulative carbon to nitrate (C:N) drawdown obtained in natural plankton assemblages dominated by diatoms and remained slightly below the Redfield C:N ratio of 6.6 under cyanobacteria16 and in batch culture incubations with E. huxleyi17. 1 3 CO2 (Fig. 3c), this ratio increased to 7.1 and 8.0 under 2 3 CO2 The extent to which the observed CO2 sensitivity in phytoplankton and 3 3 CO2, respectively. The ratio of particulate organic carbon to carbon consumption and stoichiometry in carbon-to-nutrient draw- particulate organic nitrogen (POC/PON) accumulating in the surface down can be extrapolated to other marine ecosystems remains to be layer, in contrast, closely adhered to the Redfield C:N ratio independ- seen. In accordance with our results, enhanced carbon fixation of up ent of CO2 treatment (see Methods, Supplementary Fig. 2). Deviation to 15% in response to three-times increased CO2 was observed during 2 of DIC/NO3 drawdown from POC/PON build-up, previously incubations of natural phytoplankton assemblages from the nutrient- reported from both mesocosm and field studies12,13, was attributed poor central Atlantic18. Higher carbon-to-nitrate drawdown at to dissolved organic carbon (DOC) release14. DOC concentrations increased CO2 concentrations also occurred in a previous mesocosm 19 also increased during bloom development in our study, but surface- CO2 perturbation experiment during a bloom of E. huxleyi . The 2 layer DOC accumulations of ,25 mmol kg 1 remained well below concordant results indicate the existence of a widespread mechanism levels expected from the difference between DIC drawdown and under high CO2 conditions. The phytoplankton groups dominating POC build-up (see below). A possible fate of DOC is indicated by a in the mesocosm studies—diatoms and coccolithophores—are also fourfold increase in the concentration of transparent exopolymer the main primary producers in high productivity areas and are the particles (TEP), which are known to originate from dissolved pre- principal drivers of biologically induced carbon export to the deep sea. cursors and to accelerate particle aggregation and sinking15. Apart from seawater acidification, the ocean in a high CO2 world The loss of organic carbon from the upper mixed layer (UML) will experience other changes, including higher surface temperatures, through sinking is given by the difference between DIC drawdown enhanced stratification and decreased mixed-layer depths. Although and build-up of POC and DOC, DC 5 DDIC 2 (DC 1 the direct effect of rising temperatures on net community carbon-to- loss org POC nutrient ratios is unclear, at the organism level most experimental DCDOC). Between 15% and 30% of organically fixed carbon was lost from the UML daily, with consistently higher loss rates at increased CO2 (Fig. 3b). The 12-day cumulative carbon loss at 3 3 CO2 a 15 21 exceeded that under 1 3 CO2 by about 30 mmol kg , which closely ) resembles the difference in net inorganic carbon drawdown between –1 10 g kg μ a 1,200 (

diatom 5

1,000 α

800 Chl atm)

μ 0

( 600 2 b 15 CO

p 400

200 ) –1 10 0 g kg

b μ ( 15 prym 5 α ) –1 10 Chl g kg μ

( 0 α 5 c 15 Chl ) –1 0 10 c g kg

15 μ ( ) –1

remain 5 10 α Chl mol kg μ 0 5 0 5 10 15 20 Nitrate ( Days

0 Figure 2 | Development of dominant phytoplankton groups. Chlorophyll a 10 0 5 15 20 a a b Days equivalents in the UML of: ,diatoms(Chl diatom); , prymnesiophytes (Chl aprym), predominantly the coccolithophore E. huxleyi;andc,remaining Figure 1 | Growth conditions in experimental mesocosms. a,CO2 partial chlorophyll a (Chl aremain), primarily containing prasinophytes, a a a pressures (pCO2 ); b, chlorophyll concentrations (Chl ); and c, nitrate dinoflagellates and cyanobacteria. Group-specific chlorophyll equivalents concentrations in the upper 5.5 m mixed layer (UML) during the were derived from high-performance liquid chromatography pigment analysis experiment. Values are means of triplicate CO2 treatments with initial pCO2 using CHEMTAX algorithms (for details see Methods). The light-green curve of 350 matm (green), 700 matm (grey) and 1,050 matm (red). Error bars represents the range of total chlorophyll a as depicted in Fig. 1b. Colour code is denote 61 standard deviation. The vertical line indicates time of deep water as in Fig. 1. The vertical line indicates time of deep water injection into UML injection into UML owing to vertical mixing (day 12). owing to vertical mixing (day 12). Error bars denote 61 standard deviation. 546 ©2007 Nature Publishing Group NATURE | Vol 450 | 22 November 2007 LETTERS

studies indicate an increase in the C:N ratio with increasing tem- CO2, reaching its maximum strength long after atmospheric CO2 perature20. Decreased mixed-layer depth tends to increase phyto- has transgressed to levels tolerable with regard to climate change plankton carbon-to-nutrient ratios21, which would augment the and ocean acidification. direct CO2 effect on C:N:P stoichiometry observed in this study. Aside from its effect on oceanic carbon sequestration, a more Increased stratification, in contrast, would decrease the supply of efficient biological carbon pump would lead to an expansion of deep nutrients to the surface layer, reducing overall primary and export ocean oxygen minimum zones with possible consequences for mar- production. Thus, concurrent with a decrease in the strength of the ine biogeochemical cycling. Increasing C:N ratios would also lower biological pump caused by a lower nutrient supply, the pump’s effi- the nutritional value of primary-produced organic matter, which ciency is likely to increase at increased CO2. may affect the efficiency of bacterial degradation and zooplankton Shifting of the C:N:P stoichiometry of marine primary production reproduction25, thus having further implications for marine ecosys- has long been recognized as one of the most powerful mechanisms tem dynamics. Changing carbon-to-nutrient ratios in oceanic prim- determining ocean–atmosphere carbon partitioning22. Carbon rela- ary and export production may thereby act as a forceful driver for tive to nitrogen drawdown in excess of the Redfield C:N ratio, often ecosystem processes and biogeochemical cycles in the future ocean. referred to as carbon over-consumption23, is reported for various oceanographic regimes12,13,23,24, but was generally considered to be a METHODS SUMMARY transient phenomenon. Whether this also applies to the observed The study was conducted between 15 May and 9 June 2005 at the Espegrend CO sensitivity of plankton carbon consumption and to what extent Marine Biological Station (at Raunefjorden, 60.3u N, 5.2u E) of the University of 2 3 the mechanism(s) underlying this trend are applicable to other Bergen, Norway. Nine polyethylene mesocosm enclosures (,27 m , 9.5 m water diatom- and coccolithophore-dominated new production systems depth) were moored to a raft, and filled with unfiltered, nutrient-poor, post- bloom fjord water pumped from 13.5 m depth adjacent to the raft. The enclo- is presently unknown. Assuming the observed response can be extra- sures were covered by gas-tight tents made of ethylene tetrafluoroethylene foil, polated to new production systems in the ocean, we calculate an which allowed for 95% light transmission of the complete spectrum of sunlight. excess CO sequestration potential by the biological carbon pump 2 The carbonate system in the mesocosms was manipulated by CO2 aeration to of 116 Pg C until 2100 (for details see Methods). The efficiency of this obtain triplicates of three concentrations, 350 matm (1 3 CO2), 700 matm negative feedback mechanism develops only gradually with rising (2 3 CO2) and 1,050 matm (3 3 CO2) (for details see ref. 19). CO2 aeration of the water column was ended after 3 days, when target CO2 levels were reached. Continuous flushing of the tents with air adjusted at target CO concentrations a 0 2 ensured that starting values were maintained in the overlying air throughout the 20 experiment. )

–1 The water column was stratified by addition and subsequent mixing of 800 litres 40 of freshwater into the upper 5.5 m of the mesocosms, resulting in a salinity (S) 5 mol kg 60 gradient of 1.5 p.s.u. between the surface mixed layer (S 30.6 p.s.u.) and the μ

( underlying water column. Continuous mixing of this upper layer by peristaltic 2 org 80 pumps (flow rate 450 l h 1) maintained a homogenous distribution of dissolved

DIC compounds. To promote the development of a phytoplankton bloom, nitrate and

Δ 100 21 2 phosphate were added to yield initial concentrations of 14 mmol l NO3 and 120 21 3– 0.7 mmol l PO4 .Owing to left-over silicate in post-bloom waters, the experi- m 21 b 80 ment started at a Si(OH)4 concentration of 3.2 mol l . After nutrient addition, the development and decline of a phytoplankton bloom was closely monitored over a 24-day period (for details on sampling and measurements see Methods).

) 60 –1 Full Methods and any associated references are available in the online version of 40 the paper at www.nature.com/nature. mol kg μ ( 20 Received 16 June; accepted 17 September 2007. loss Published online 11 November 2007. C Δ 0 1. Sabine, C. L. et al. The oceanic sink for anthropogenic CO2. Science 305, 367–371 (2004). 0 5101520 2. Feely, R. A. et al. Impact of anthropogenic CO2 on the CaCO3 system in the Days oceans. Science 305, 362–366 (2004). 3. Caldeira, K. & Wickett, M. E. Anthropogenic carbon and ocean pH. Nature 425, c 15 365 (2003).

4. Langdon, C. et al. Effect of elevated CO2 on the community metabolism of an )

–1 experimental coral reef. Glob. Biogeochem. Cycles 17, 1011 (2003). 10 5. Gattuso, J.-P., Frankignoulle, M., Bourge, I., Romaine, S. & Buddemeier, R. W. Effect of calcium carbonate saturation of seawater on coral calcification. Glob. mol kg

μ Planet. Change 18, 37–46 (1998). 6. Riebesell, U. & Zondervan, I. Rost, B. Tortell, P. D., Zeebe, R. E. & Morel, F. M. M. 5 Reduced calcification in marine plankton in response to increased atmospheric

Nitrate ( CO2. Nature 407, 634–637 (2000). Δ 7. Redfield, A. C., Ketchum, B. H. & Richards, F. A. in The Sea 2nd edn (ed. Hill, M. N.) 26–77 (Wiley, New York, 1963). 0 0 50 100 8. Volk, T. & Hoffert, M. I. in The Carbon Cycle and Atmospheric CO2: Natural Variations Δ μ –1 Archean to Present (eds Sunquist, E. T. & Broecker, W. S.) Monograph Vol. 32 DICorg ( mol kg ) 99–110 (Am. Geophys. Union, Washington DC, 1985).

Figure 3 | CO2 sensitivity of carbon consumption and carbon loss. 9. Gruber, N. & Sarmiento, J. L. in The Sea: Biological–Physical Interactions in the a, Drawdown of dissolved inorganic carbon owing to organic matter Oceans Vol. 12 (eds Robinson, A. R., McCarthy, J. J. & Rothschild, B. J.) 337–399 production (DDIC ) in the upper 5.5 m mixed layer (UML) with time. (Wiley, New York, 2002). org 10. Houghton, J. T., et al. Climate Change 2001: The Scientific Basis (Cambridge Univ. b, Loss of organically bound carbon (DC ) from the UML with time. The loss Press, Cambridge, UK, 2001). vertical line in a and b indicates time of deep water injection into UML owing D 11. Raven, J. et al. Ocean acidification due to increasing atmospheric carbon dioxide. to vertical mixing (day 12). c, DICorg versus nitrate drawdown in the UML. Policy Document 12/05. Roy. Soc. Rep. 12 (2005). The black line represents Redfield C:N ratio of 6.6. Values are means of 12. Antia, N. J., McAllister, C. D., Parsons, T. R., Stephens, K. & Strickland, J. D. H. triplicate CO2 treatments. Error bars denote 61 standard deviation; colour Further measurements of primary production using a large-volume plastic sphere. code is as in Fig. 1. Limnol. Oceanogr. 8, 166–183 (1963). 547 ©2007 Nature Publishing Group LETTERS NATURE | Vol 450 | 22 November 2007

13. Sambrotto, R. N. et al. Elevated consumption of carbon relative to nitrogen in the 22. Broecker, W. S. Ocean geochemistry during glacial time. Geochim. Cosmochim. surface ocean. Nature 363, 248–250 (1993). Acta 46, 1689–1705 (1982). 14. Banse, K. Uptake of inorganic carbon and nitrate by marine plankton and the 23. Toggweiler, J. R. Carbon overconsumption. Nature 363, 210–211 (1993). Redfield ratio. Glob. Biogeochem. Cycles 8, 81–84 (1994). 24. Ko¨rtzinger, A., Koeve, W., Ka¨hler, P. & Mintrop, L. C:N ratios in the mixed layer 15. Engel, A., Thoms, S., Riebesell, U., Rochelle-Newall, E. & Zondervan, I. during the productive season in the Northeast Atlantic Ocean. Deep-sea Res. I 48, Polysaccharide aggregation as a potenial sink of marine dissolved organic carbon. 661–688 (2001). Nature 428, 929–932 (2004). 25. Sterner, R. W. & Elser, J. J. (eds) Ecological Stoichiometry (Princeton Univ. Press, 16. Engel, A. Direct relationship between CO2 uptake and transparent exoploymer Princeton, 2002). particles production in natural phytoplankton. J. Plankton Res. 24, 49–53 (2002). 17. Heemann, C. Phytoplanktonexsudation in Abha¨ngigkeit von der Supplementary Information is linked to the online version of the paper at Meerwasserkarbonatchemie. Thesis, Univ. Bremen (2002). www.nature.com/nature. 18. Hein, M. & Sand-Jensen, K. CO2 increases oceanic primary production. Nature Acknowledgements We thank the participants of the Pelagic Ecosystem CO 388, 2 526–527 (1997). Enrichment study (PeECE III, http://peece.ifm-geomar.de). We acknowledge 19. Engel, A. et al. Testing the direct effect of CO2 concentration on a bloom of the J. Egge, J. Nejstgaard and the staff of the Espegrend Marine Biological Station for coccolithophorid Emiliania huxleyi in mesocosm experiments. Limnol. Oceanogr. helping to organize and set up the mesocosms. This work was supported by the EU 50, 493–504 (2005). projects CARBOOCEAN ‘Marine carbon sources and sinks assessment’ (GOCE), 20. Woods, H. A. et al. Temperature and the chemical composition of poikilothermic CABANERA and University of Bergen (LOCUS) funding. organisms. Funct. Ecol. 17, 237–245 (2003). 21. Diehl, S., Berger, S. & Wo¨rhl, R. Flexible algal nutrient stoichiometry mediates Author Information Reprints and permissions information is available at environmental influences on phytoplankton and its abiotic resources. Ecology 6, www.nature.com/reprints. Correspondence and requests for materials should be 2931–2945 (2005). addressed to U.R. ([email protected]).

METHODS bloom, this was not depicted in the alkalinity measurements owing to mixing in Sampling. Depth-integrated water samples were taken daily at 10:00 by means of of higher alkalinity deep water on day 12. Because mixing across the halocline D a 5-m long, 6-cm diameter tube that was lowered into the mesocosms, closed at was similar in all CO2 treatments, the divergence of estimated DICcalc, however, the top, pulled up onto the raft and emptied into sampling bottles. Vertical indicates a CO2 effect on calcification after the peak of the bloom. This phase of profiles of temperature and salinity inside the mesocosms were obtained using an E. huxleyi bloom, in which nutrient limitation impedes further cell division, is a hand-operated CTD (SAIV A/S, model SD204). often characterized by excess calcification, leading to shedding of calcite platelets (coccoliths) and their accumulation in the water column. Overall, E. huxleyi Measurements. CO2 partial pressure in seawater and the overlying air was measured using an infrared gas analyser (Li-Cor 6262) coupled to an equilibrator26. cell numbers remained much below those typically observed in blooms of this species, which explains the comparatively low amount of calcium carbonate The pCO system was mounted on a cart that was moved between the mesocosms. 2 2 Water was pumped from the mesocosm at 2 litre min 1 and returned to the precipitation. mesocosm after passing through the equilibrator. Temperature at the inlet of Excess carbon sequestration. Provided that the mechanisms underlying the the pump and in the equilibrator was measured simultaneously using platinum observed CO2 sensitivity of carbon consumption and carbon-to-nutrient stoi- RTD digital thermometers. The system was calibrated before and after water chiometry can be generalized to other diatom- and coccolithophore-dominated analyses with air standards with nominal mixing ratios of 345, 415 and new production systems, the excess CO2 sequestration potential by the ocean’s biological carbon pump was calculated for the periods from 1750 to present, and 1,100 p.p.m. Immediately after each measurement of pCO2 , samples for total 10 from 1750 to 2100 assuming: business-as-usual CO2 emissions (IS92a ) until the alkalinity and DIC were taken using the same water pump to fill bottles with 2 end of this century; present-day oceanic carbon export of 12 Pg C yr 1 (current ground glass stoppers. Total alkalinity and DIC samples were poisoned with 2 2 estimates range between 8 Pg C yr 1 and 16 Pg C yr 1)37; and an e-folding time of HgCl2 on collection and were filtered through glass fibre (GF/F) filters before analysis. Total alkalinity was measured using the classical Gran electrotitration 100 yr for CO2 sequestered by biological carbon export to get back into contact 2 38 method27. The reproducibility of measurements was usually within 4 mmol kg 1. with the atmosphere . 28 21 Supplementary Fig. 4a depicts DDIC as a function of atmospheric p , DIC was measured by colometric titration with a precision of 2 mmol kg . org CO2 D m For the analysis of phytoplankton pigments, 250–500 ml of the water samples with the present day value ( DICorg at 380 atm) set to 1. Extrapolating the observed trend to pre-industrial CO levels (p 5 280 matm) yields a value of were filtered through 25 mm Whatman GF/F filters. Filters were frozen at 2 CO2 , 220 uC until analysis. For pigment extraction, filters were homogenized in plas- 0.95; that is, for a given amount of inorganic nutrients, biological carbon tic vials (11 ml) together with 1 ml acetone (100%) and a mixture of glass beads consumption under pre-industrial conditions was about 95% of today’s level. m (2 and 4 mm) by shaking (5 min) in a cooled Vibrogen cell mill. The extracts were On the basis of this relationship, an increase in atmospheric CO2 from 280 atm m centrifuged (,850 g/5,000 r.p.m. for 10 min, cooled at 210 uC). The extraction to 380 atm (present day) corresponds to an excess carbon sequestration of 22 Pg process was performed under dimmed light to prevent photo-oxidation of the (range 14–29 Pg) for the past 150 yr (Supplementary Fig. 4b). Without this pigments. Concentrations of pigments (chlorophyll and carotenoids) were negative feedback mechanism, atmospheric CO2 would be approximately m determined by reverse-phase high-performance liquid chromatography29. 11 atm higher than its present value. By the end of this century, this process Identification of pigments was carried out by comparing their retention times would sequester an additional 94 Pg C to the deep ocean, bringing total excess and absorption spectra obtained with a diode array spectrophotometer carbon sequestration to 116 Pg C (range 76–154 Pg C). This reduces the increase m (WATERS) with those of pigment standards. Calibration was carried out with in atmospheric CO2 by a total of 58 atm at 2100. commercially available standards. Calculation of the composition of the phyto- 26. Wanninkhof, R. & Thoning, K. Measurement of fugacity of CO2 in surface water 30 plankton communities was executed using the CHEMTAX program , convert- using continuous and discrete sampling methods. Mar. Chem. 44, 189–205 ing the concentrations of marker pigments to equivalents of chlorophyll a with (1993). suitable pigment-to-chlorophyll a ratios. 27. Gran, G. Determination of the equivalence point in potentiometric titrations of Nitrate, nitrite, phosphate and silicate were determined from GF/F-filtered seawater with hydrochloric acid. Oceanol. Acta 5, 209–218 (1952). samples with an autoanalyser (AA II)31. Ammonium was measured according to 28. Johnson, K. M., Williams, P. J., Brandstrom, L. & Sieburth, J. McN. Colometric total 21, ref. 32. POC and PON were measured on 0.25 litre or 0.5 litre samples filtered carbon analysis for marine studies: automation and calibration. Mar. Chem. 117–133 (1987). gently (200 mbar) through precombusted (450 uC, 5 h) glass fibre filters (GF/F, 29. Barlow, R. G., Cummings, D. G. & Gibb, S. W. Improved resolution of mono- and Whatman). Filters were fumed overnight with saturated HCl to remove all divinyl chlorophylls a and b and zeaxanthin and lutein in phytoplankton extracts u particulate inorganic carbon, dried for 6 h at 60 C and measured on an ele- using reverse phase C-8 HPLC. Mar. Ecol. Prog. Ser. 161, 303–307 (1997). mental analyser (EuroEA 3000, EuroVector). DOC was measured using the high- 30. Mackey, M. D., Mackey, D. J., Higgins, H. W. & Wright, S. W. CHEMTAX — a temperature catalytic oxidation method33. program for estimating class abundances from chemical markers: application to Calculations. Inorganic carbon drawdown owing to organic matter production HPLC measurements of phytoplankton. Mar. Ecol. Prog. Ser. 144, 265–283 (1996). was calculated from measured changes in DIC and alkalinity according to: 31. Hansen, H. P. & Koroleff, F. in Methods of seawater analysis 3rd edn (eds Grasshoff, K., Kremling, K. & Ehrhardt, M.) 159–228 (Wiley VCH, Weinheim, 1999). DDICorg, 5 jDDICj 2 0.5jDTAj, where TA is total alkalinity. DDIC was corrected for CO air–sea gas exchange following ref. 34 with chemical enhancement 32. Holmes, R. M., Aminot, A., Ke´rouel, R., Hooker, B. A. & Peterson, B. J. A simple and 2 precise method for measuring ammonium in marine and freshwater ecosystems. factors given in ref. 35. DTA was calculated from changes in DIC and p using CO2 Can. J. Fish. Aquat. Sci. 56, 1801–1808 (1999). stoichiometric dissociation constants given in ref. 36. Calcium carbonate 33. Qian, J. & Mopper, K. Automated high-performance, high-temperature D 5 precipitation was estimated from alkalinity drawdown as DICcalc 0.5 combustion total organic carbon analyser. Anal. Chem. 68, 3090–3097 (1996). (DTA 1 Dnitrate). Owing to strong wind and wave action in the fjord on 34. Delille, B. et al. Response of primary production and calcification to changes of day 12, some water in the mesocosms from below the halocline was mixed pCO2 during experimental blooms of the coccolithophorid Emiliania huxleyi. Glob. Biogeochem. Cycles 19, GB2023 (2005). into the surface layer. Because surface water was lower in salinity, pCO2 , DIC and total alkalinity owing to the initial freshwater addition, injection of deep 35. Kuss, J. & Schneider, B. Chemical enhancement of the CO2 gas exchange at a smooth seawater surface. Mar. Chem. 91, 165–174 (2004). water led to a slight increase in pCO2 (Fig. 1a), DIC and total alkalinity on this day. This also affected the calculations of DDIC , DC (Fig. 3a, b) and DDIC 36. Zeebe, R. E. & Wolf-Gladrow, D. CO2 in seawater: equilibrium, kinetics, isotopes. org loss calc Elsevier Oceanogr. Ser., 65, (Elsevier, Amsterdam, 2001). (Supplementary Fig. 3). 37. Oschlies, A. Model-derived estimates of new production: new results point Biogenic calcification. Inorganic carbon drawdown owing to biogenic calcifica- towards lower values. Deep-Sea Res. 48, 2173–2197 (2001). D tion ( DICcalc) calculated from changes in surface layer alkalinity showed no 38. Maier-Reimer, E., Mikolajewicz, V. & Winguth, A. Future ocean uptake of CO2: distinct CO2 treatment difference during bloom development (Supplementary interaction between ocean circulation and biology. Clim. Dynam. 12, 711–721 Fig. 3). Although calcification continued for some days after the peak of the (1996).

Paper 6

Marine ecosystem community carbon and nutrient uptake stoichiometry under varying ocean acidification during the PeECE III experiment. Bellerby, R. G. J., K. G. Shulz, U. Riebesell, C. Neill, G. Nondal, E. Heegaard, T. Johannessen, and K. R. Brown. Biogeosciences, 5, 1517-1527, 2008.

Biogeosciences, 5, 1517–1527, 2008 www.biogeosciences.net/5/1517/2008/ Biogeosciences © Author(s) 2008. This work is distributed under the Creative Commons Attribution 3.0 License.

Marine ecosystem community carbon and nutrient uptake stoichiometry under varying ocean acidiﬁcation during the PeECE III experiment

R. G. J. Bellerby1,2, K. G. Schulz3, U. Riebesell3, C. Neill1, G. Nondal4,2, E. Heegaard1,5, T. Johannessen2,1, and K. R. Brown1 1Bjerknes Centre for Climate Research, University of Bergen, Allegaten´ 55, 5007 Bergen, Norway 2Geophysical Institute, University of Bergen, Allegaten´ 70, 5007 Bergen, Norway 3Leibniz Institute for Marine Sciences (IFM-GEOMAR), Dusternbrooker Weg 20, 24105 Kiel, Germany 4Mohn-Sverdrup Center and Nansen Environmental and Remote Sensing Center, Thormølensgate 47, 5006 Bergen, Norway 5EECRG, Department of Biology, University of Bergen, Allegaten´ 41, 5007 Bergen, Norway Received: 19 November 2007 – Published in Biogeosciences Discuss.: 13 December 2007 Revised: 3 September 2008 – Accepted: 3 September 2008 – Published: 10 November 2008

Abstract. Changes to seawater inorganic carbon and nutrient 1 Introduction concentrations in response to the deliberate CO2 perturbation of natural plankton assemblages were studied during the Consequent to the increase in the atmospheric load of car- 2005 Pelagic Ecosystem CO2 Enrichment (PeECE III) ex- bon dioxide (CO2), due to anthropogenic release, there has periment. Inverse analysis of the temporal inorganic carbon been an increase in the oceanic carbon reservoir (Sabine et dioxide system and nutrient variations was used to determine al., 2004). At the air-sea interface, the productive, euphotic the net community stoichiometric uptake characteristics of surface ocean is a transient buffer in the process of ocean- p a natural pelagic ecosystem perturbed over a range of CO2 atmosphere CO2 equilibrium, a process retarded by the slow scenarios (350, 700 and 1050 μatm). Nutrient uptake showed mixing of the surface waters with the intermediate and deep no sensitivity to CO2 treatment. There was enhanced car- ocean. As such, the greatest changes to the CO2 system are bon production relative to nutrient consumption in the higher occurring in the surface waters. This build up of CO2 is al- CO2 treatments which was positively correlated with the ini- ready altering the carbonate chemistry of the oceans and pro- tial CO2 concentration. There was no significant calcifica- jections on decadal to centennial timescales point to changes tion response to changing CO2 in Emiliania huxleyi by the in seawater pH (Caldeira and Wickett, 2003; Bellerby et al., peak of the bloom and all treatments exhibited low partic- 2005; Blackford and Gilbert, 2007) and carbonate species − ulate inorganic carbon production (∼15 μmol kg 1). With (Orr et al., 2005) that may have ramifications for the success insignificant air-sea CO2 exchange across the treatments, the of organisms or whole marine ecosystems (e.g. Raven et al., enhanced carbon uptake was due to increase organic carbon 2005; Riebesell, 2004; Kleypas et al., 2006). production. The inferred cumulative C:N:P stoichiometry Exposure of phytoplankton to pH and CO2 levels relevant of organic production increased with CO2 treatment from to those anticipated over the coming decades leads to modi- 1:6.3:121 to 1:7.1:144 to 1:8.25:168 at the height of the fications in physiological or morphological properties which bloom. This study discusses how ocean acidification may in- may have consequences for ecological structure and biogeo- cur modification to the stoichiometry of pelagic production chemical cycling. The general assertion is that increasing and have consequences for ocean biogeochemical cycling. CO2 has deleterious effects on the growth and productivity of marine calcifiers (e.g. Riebesell et al., 2000; Delille et al., 2005; Orr et al., 2005), although there are notable exceptions (Langer et al., 2006). Overconsumption of carbon, a common response to nutrient and environmental stress is mag- nified under high CO2 conditions (Zondervan et al., 2002; Correspondence to: R. G. J. Bellerby Engel et al., 2005). Changing CO2 concentrations in aquatic ([email protected]) systems has been shown to influence phytoplankton species

Published by Copernicus Publications on behalf of the European Geosciences Union. 1518 R. G. J. Bellerby et al.: Stoichiometry PeECE III succession (Tortell et al., 2002). Changes to the nutritional inlet tubes to the respective instruments immediately prior to quality (higher C:P), in response to increased CO2,ofphy- measurement. toplankton as a food source results in lower growth rate and fecundity in zooplankton (Urabe et al., 2003) 2.2 Calculations Efforts to understand potential consequences and feed- As the system had settled down by day 2 (16 May), fol- backs of increasing CO have employed laboratory and 2 lowing the initial CO and nutrient perturbations (Schulz mesocosm studies either at the individual species level or 2 et al., 2008), all changes to the biogeochemical fields were on natural and perturbed ecosystems (Riebesell et al., 2000; referenced to this date. Calculation of additional carbon Delille et al., 2005). In this study, a natural ecosystem was dioxide system variables, from measured CT and AT used perturbed with nutrients over a range of atmospheric CO 2 the CO2SYS program (Lewis and Wallace, 1998) adopting scenarios extending previous studies to include the effects of the dissociation constants for carbonic acid (Dickson and very high CO concentrations postulated for the 22nd cen- 2 Millero, 1987), boric acid (Dickson, 1990a) and sulphuric tury. acid (Dickson, 1990b) and the CO2 solubility coefficient from Weiss (1974). Seawater pH is reported on the total hydrogen scale. 2 Methods Particulate inorganic carbon (PIC) production was calculated from temporal changes (t) in total alkalinity with 2.1 Experimental appropriate correction for alkalinity contributions from net nitrate and phosphate consumption (Goldman and Brewer, A mesocosm experiment was performed between 15 May 1980): and 9 June 2005 at the University of Bergen Marine Bi- (AT − [NO ] − [PO ]) ological station in Raunefjorden, Norway. Nine polyethy- − . . 3 4 PIC= 0 5 t (1) lene enclosures (∼25 m3, 9.5 m water depth) were moored to a raft equipped with a floating laboratory. The enclosures In order to evaluate biological contributions to the inor- where filled with fjord water from 12 m depth, and manip- ganic carbon system it was necessary to first account for the p ulated in order to obtain triplicates of three different CO2 exchange of CO2 between the seawater and the overlying × μ × μ concentrations (1 CO2 (350 atm), 2 CO2 (700 atm) and mesocosm atmosphere (CO ( )). Gas exchange was calcu- × μ 2 ex 3 CO2 (1050 atm)). Addition of fresh water to the upper lated according to Delille et al. (2005) further employing the 5.5 m of the enclosures ensured the generation of a mixed chemical enhancement factors of Kuss and Schneider (2004). layer separated from the underlying water by a salinity gra- Net community production (NCP) was calculated from dient of 1.5. Nitrate and Phosphate were added to the up- temporal changes in CT allowing for modifications due to per mixed layer resulting in initial respective concentrations PIC production or dissolution and net CO2 gas exchange −1 of 16 and 0.8 μmol kg . A comprehensive description of thus: the mesocosm setup, CO and nutrient perturbation and sam- 2 ( − − ) pling strategy, as well as the nutrient measurement method- − CT + . . AT [NO3] [PO4] + CO2(ex) NCP= t 0 5 t t ology can be found in Schulz et al. (2008). (2) Samples for determining the carbon dioxide system were taken from seawater pumped from 1 m depth in each of the Removal of the contributions of net calcification and air-sea p ) enclosures. The partial pressure of carbon dioxide ( CO2 CO exchange gives the net perturbation of the inorganic car- p 2 was determined in air equilibrated with seawater CO2 us- bon system from community biological activity – the sum ing an infrared gas analyser (Li-Cor 6262) (Wanninkhof and of autotrophic and heterotrophic processes within the mixed Thoning, 1993). Gas calibration of the instrument against layer. Comparison of the inferred net organic production and high quality air standards containing mixing ratios of 345, nutrient uptake rates gives the pelagic community molar sto- 415 and 1100 ppm enveloped the daily seawater measure- ichiometry changes as a response to changing CO2. ment program. Following the pCO2 measurements, using the same sampling methodology, samples for total alkalinity 2.3 Statistical analysis (AT ) and total dissolved inorganic carbon (CT ) were drawn into 500 ml borate bottles and immediately poisoned with Statistical analyses were performed on data from all meso- HgCl2.AT was measured using Gran potentiometric titration cosms. The design of the statistical treatments followed a (Gran, 1952) on a VINDTA system (Mintrop et al., 2000) typical repeated measurement scheme with an outer covari- − with a precision of ≤4 μmol kg 1.CT was determined using ate as the treatment. The response (yij ) was observed for coulometric titration (Johnson et al., 1987) with a precision each particular treatment (treatij ) at time j=0, ..., ni within − of ≤2 μmol kg 1. For both AT and CT measurements, sam- the i’th mesocosm, i=1, ..., m. Appropriate to the form ples were filtered through GF/F filters placed in the sample of the relationship to be tested both linear and non-linear

Biogeosciences, 5, 1517–1527, 2008 www.biogeosciences.net/5/1517/2008/ R. G. J. Bellerby et al.: Stoichiometry PeECE III 1519

Fig. 1. Temporal development of the carbon dioxide variables and nutrient concentrations within the mesocosm upper mixed layers: (A) partial pressure of carbon dioxide; (B) total dissolved carbon dioxide; (C) Total alkalinity; (D) Nitrate; (E) Phosphate; and (F) Silicate. The mean treatment values (350, 700 and 1050 μatm), with standard deviation, are represented by the green, grey and red lines, respectively. mixed models were applied to the treatment data (Pinheiro The logistic relationship follows: and Bates, 2000). φ1i yij = + εij (3) 1 + exp −(day −φ i)/φ i 2.3.1 Logistic mixed effect model (L-NLME) ij 2 3 Where φ1i is the asymptote, φ2i is the xmid and φ3i is the The logistic relationship was estimated using the self-starting scale. The residual component followed a normal distribu- algorithm of SSlogis of the R-library nlme (Pinheiro and tion, εij ∼N(0,σ2). The mixed effect element is included in Bates, 2000). This relationship was applied for both NCP the parameter φi=β+bi where bi is the cluster-speciﬁc ran- and PIC as response variable (yij ) against corresponding day. dom effect, bi∼N(0,), and β is the ﬁxed effect. The treatment effects are included through φi=β+γxi+bi, where xi is the contrast matrix corresponding to the factorial treatment vector. www.biogeosciences.net/5/1517/2008/ Biogeosciences, 5, 1517–1527, 2008 1520 R. G. J. Bellerby et al.: Stoichiometry PeECE III

Where φ1 is the asymptote, φ2 is the intercept and φ3 is the logarithm of the rate constant.

2.3.3 Linear mixed model (LME)

For the relationship between dNO3 as the response (yij ) against NCP as predictor it was sufﬁcient with a linear mixed effect model (Pinheiro and Bates, 2000):

yij = β0 + b0i + βiNCPij + β2treat1,i,j + β3treat2,ij +β4NCPij treat1,ij + β5NCPij treat2,ij + εij (5)

3 Results

3.1 Temporal evolution of the carbon dioxide system

In accordance with the agreement for the PeECE special issue, only treatment means are represented in this study – except the statistical analyses which were performed on the data from individual mesocosm enclosures. Daily treatment means, with standard deviations, of seawater pCO2,CT and nutrient concentrations for the mixed layers in the three perturbation scenarios are shown in Fig. 1. Once the scenario CO2 concentrations had been reached in the seawater, only the atmospheric concentrations over the water were maintained at the prescribed treatment level and the seawater CO2 system was allowed to respond as in a naturally occurring plankton bloom. There is a clear treatment dependant response in pCO2 (Fig. 1a). Under the 1×CO2 scenario, pCO2 dropped by 102 μatm by the end of the bloom stage (Day 12), whilst pCO2 reductions were 335 and 570 μatm in the 2× and 3×CO2 scenarios, respectively. Similarly, the CT reductions also show the same order of response (Fig. 1b). Under the 1×CO2 treatment, CT showed a reduction of 94 μmol kg−1, increasing to 107 μmol kg−1 −1 for 2×CO2 and to 118 μmol kg for 3×CO2. Initial concentrations of AT (Fig. 1c) were higher in the 3×CO2 treatment corresponding with the higher initial salinity (Schulz et al., 2008). Changes in AT throughout the bloom stage Fig. 2. Calculated carbon dioxide system variables within the meso- indicated a modest treatment response before Day 10 with −1 cosm upper mixed layers: (A) Carbonate ion concentration; (B) sat- reductions in AT of 21, 23.5 and 24.8 μmol kg (1×CO2 ) uration state of calcite ( ; and (C) pHT . The colour assignment of through 3×CO2 treatments). Total alkalinity increased af- treatment is as in Fig. 1. ter the bloom peak in the 2× and 3× CO2 treatments but remained constant in the 1× CO2 treatment. 2− The absolute change in carbonate ion [CO3 ] concentra- 2.3.2 Asymptotic mixed effect model (A-NLME) tions, on the other hand, showed no treatment dependence and increased by 50, 52 and 49 μmol kg−1 in the low to high

An asymptotic relationship was assumed when a sigmoidal treatments, respectively (Fig. 2a). Correspondingly, as (the saturation state of calcium carbonate) is controlled mainly by relationship were inappropriate. The self-starting algorithm − [CO2 ], there were similarly increases in the calcite satura- SSasymp (Pinheiro and Bates, 2000) was used for the rela- 3 d y ) tion state; 1.2, 1.27 and 1.19 (Fig. 2b). Seawater pH (pHT ) tionship of PO4, as response ( ij , against NCP, following × the relationship: values, however, increased the most in the 3 CO2 treatment from 7.64 to 7.96 (pH=0.32); under the 2×CO2 scenario the pH increased from 7.81 to 8.07 (pH=0.26) and in the yij = φ1i + (φ2i − φ1i) exp − exp(φ3i)NCPij (4) 1×CO2 treatment from 8.11 to 8.27 (pH=0.16).

Biogeosciences, 5, 1517–1527, 2008 www.biogeosciences.net/5/1517/2008/ R. G. J. Bellerby et al.: Stoichiometry PeECE III 1521

Fig. 3. The main contributors to changes in the inorganic carbon concentration throughout the experiment (A) Cumulative particulate inorganic carbon (PIC) production and (B) net community production (NCP). The contribution of CO2 air-sea gas exchange to the net CT change was small (<1%) and not shown. The colour assignment of treatment is as in Fig. 1.

There was no significant difference in particulate inorganic on day 11. From the onset of the experiment, concentrations carbon (PIC) production between CO2 treatments (L-NLME, of nitrate diminished at a steady rate until Day 12 then de- p=0.77, n=170) (Fig. 3a). The dominant calcifier in the clined slowly until the end of the experiment. bloom was the prymesiophyte Emiliania Huxleyi (Paulino et The stoichiometry of net community inorganic nutrient to al., 2008) and PIC production increased to a maximum of calculated organic carbon uptake is shown in Fig. 4. As ex- −1 15 μmol kg on Day 10 corresponding to the start of the pected, with the measured nutrient concentrations exhibiting demise of the Emiliania Huxleyi population (Paulino et al., no scenario dependence, there is little deviation in nutrient 2008). stoichiometry between treatments. In the pre-bloom period, Net community production (NCP) (Fig. 3b), the net or- until day 5, there is a greater uptake of nitrate to phosphate ganic carbon production, contributed the most to the changes relative to the Redfield ratio (16:1) (Redfield et al., 1963). in inorganic carbon dioxide shown in Fig. 1b with contri- In the initial stage of the bloom, relative phosphate uptake butions to the CT changes from PIC and air-sea CO2 ex- increases rapidly on Day 6. Thereafter, the system behaves change (not shown) of only 15–20%. The cumulative NCP very close to Redfieldian until the end of the bloom when N:P has a maximum on Day 12 and shows a clear treatment rose slowly to a final cumulative value of 21 in all treatments. response with maximum production, relative to Day 2 (L- p< n × Nitrate uptake is high compared to silicate uptake until NLME, 0.001, =170). The 3 CO2 treatment produced Day 4, after which the diatom bloom started (Paulino et al., 110 μmol kg−1, falling to 96 μmol kg−1 and 80 μmol kg−1 × × 2008) and silicate was consumed at about 1:1 with nitrate un- in the 2 CO2 and 1 CO2 treatments. After Day 12 there til silicate reached a first minimum on Day 7. There followed is a fall in calculated NCP with further treatment divergence a small peak in silicate concentration on Day 8 (Fig. 1f) after in NCP until the end of the experiment - again favouring the which silicate became depleted on Day 9–10. 3×CO2 treatment. In contrast to the nutrient stoichiometric ratios, there are 3.2 Temporal evolution of nutrients significant treatment dependant differences between the carbon to nitrate (C:N) (LME, p<0.001, n=170) (Fig. 4c, g) The treatment dependant changes in the major nutrients ni- and carbon to phosphate (C:P) ratios (A-NLME, p<0.001, trate, phosphate and silicate are shown in Fig. 1c–e. Both n=170) (Fig. 4d, h). Overconsumption of carbon results nutrient uptake and the timing of nutrient minima showed in a higher C:N uptake ratio under higher CO2 exposures. no dependency on CO2 treatment although there were inter- Cumulative C:N uptake was higher in the 3×CO2 scenario treatment variations in the planktonic assemblages (Paulino throughout the entire bloom, quickly reaching a peak of 7.5– et al., 2008). Silicate was not added during the experi- 8.25 by Day 6 and remaining at this level until Day 12. mental set-up, however a residual concentration of about The pre-bloom responses in the 1× and 2×CO2 treatments −1 3 μmol kg was inherited from the fjord water used to fill were similar until day 5 after which the 2×CO2 treatment in- the mesocosm enclosures. Silicate concentrations dropped creased C:N uptake to 7.1. A similar order of response was sharply due to diatom uptake and reached a minimum on day seen in the C:P ratio with the 3× treatment showing a greater 9–10. Phosphate concentrations showed only modest reduc- carbon overconsumption throughout the experiment. By the tions until day 5 after which rapid uptake led to a minimum height of the bloom, the cumulative C:N:P stoichiometry of www.biogeosciences.net/5/1517/2008/ Biogeosciences, 5, 1517–1527, 2008 1522 R. G. J. Bellerby et al.: Stoichiometry PeECE III

Fig. 4. Net community nutrient stoichiometric uptake ratios referenced to Day 2. Illustrated are the respective daily changes and parameter- parameter relationships for nitrate to phosphate (A, E); silicate to nitrate (B, F); organic carbon to nitrate (C, G); and organic carbon to phosphate (D, H). The black line represents the conventional Redﬁeld ratio (Redﬁeld et al., 1963) (E, G, H)) and a silicate-nitrate relationship of 0.95:1 from the mean of diatom species Si/N reported in Table 3 of Brezinski (1985) (0.95:1±0.4) (F). The colour assignment of treatment is as in Fig. 1.

Biogeosciences, 5, 1517–1527, 2008 www.biogeosciences.net/5/1517/2008/ R. G. J. Bellerby et al.: Stoichiometry PeECE III 1523 net organic production increased with CO2 treatment from contemporary treatment. This study does not find any rela- 2− 1:6.3:121 to 1:7.1:144 to 1:8.25:168. In contrast to the C:N tionship between [CO3 ] and the timing or degree of cal- response, the C:P increased significantly in the post-bloom 2− cification. Further, although the [CO3 ] and calcite values phase. showed marked inter-treatment variations throughout the experiment there was no difference in cumulative net community calcification by the peak of the bloom as also identified 4 Discussion by Schulz et al. (2008) using direct measurements of PIC. This finding varies from the results of Delille et al. (2005) 4.1 Modification of the carbonate system who reported a 40% reduction in PIC production between 1×CO2 and 2×CO2 treatments. Conversely, this study does As changes to AT (including contributions from nutrient al- not show the promotion of higher calcification with increas- kalinity) were similar and low and gas exchange minimal in ing CO2 as found by Langer et al. (2006). all treatments, modifications to CT were mainly due to or- The peak concentrations of Emiliania Huxleyi in this study ganic carbon production. As such, the greatest changes were were 4–6×106 cells L−1 (Paulino et al., 2008) which is rep- related to the treatment dependant carbon overconsumption resentative of the concentrations found in the North Atlantic with increasing uptake with increasing initial pCO2. The and Norwegian Sea (Tyrrell and Merico, 2004). The Delille effect on pCO2 due to the greater carbon uptake is exacer- et al. (2005) experiment had much higher concentrations (4– − bated due to the low buffer capacity of the high CO2 sea- 6×107 cells L 1) more representative of the intense blooms water resulting in a rapid reduction of pCO2 in the 3×CO2 found in the Norwegian Fjords (Tyrrell and Merico, 2004). treatment. The opposite response is seen with pHT with the Extrapolating the Delille et al. (2005) calcification response largest increases found in the 3×CO2 scenario. Due to the to the PeECE III study cell numbers translates to a pre- −1 weakly buffered system in a future ocean, pelagic ecosys- dicted reduction in PIC production of 6 μmol kg (1×CO2– tems will undergo greater seasonal changes in their ambient 2×CO2 treatment). This corresponds to a change in carbon- − CO2 fields. Increasing CO2 will reach a point where changes ate alkalinity of 12 μmol kg 1 well within the measurement to the carbonate system will move outside the contemporary precision of the current methodology. A closer look at Emil- “carbonate system envelope” and accordingly, in this experi- iania Huxleyii concentrations reveals that there was an in- ment, at no point did any of the calculated carbonate system crease in cell numbers (Paulino et al., 2008), perhaps facili- variables in the future 2×CO2 and 3×CO2 scenarios overlap tated through lower virus (EhV) concentration (Larsen et al., × with the range of the contemporary 1 CO2 scenario. 2008), with increasing CO2. Therefore, the cells in the higher CO2 treatments may have been healthier and thus able to cal- 4.2 Calcification cify more efficiently. A further artefact of the experimental design was the lack of deep water total alkalinity measure- The growth and health of many calcifying marine organisms ments. The small salinity changes throughout the experiment have been shown to respond to the saturation state of CaCO3 (Schulz et al., 2008) suggest a potential input of AT from be- ) ( calcite in seawater (e.g. Gattuso et al., 1998; Riebesell et low the pycnocline which are accounted for in the inverse al., 2000; Langdon et al., 2003; Kleypas et al., 2006; Delille method. et al., 2005; Langer et al., 2006). The dominant calcifier during the experiment was the prymesiophyte Emiliania Huxleyi 4.3 Carbon consumption and nutrient stoichiometry which is common to the Norwegian fjords and open waters of the Norwegian and Barents Sea. Calcification in Emilia- Elemental stoichiometry of biological production in the sur- nia Huxleyi has been shown to be controlled by light, nutri- face ocean controls, to a great extent, nutrient balance and ent (especially phosphate) and carbonate ion concentrations cycling of the global ocean. Ocean nutrient stoichiometry is and calcite saturation state (see review in Zondervan, 2007). controlled on short timescales by the resource allocation in The photon flux density (PFD) concentrations throughout the marine organisms (e.g. Redfield et al., 1963; Klausmeier et experiment (Schulz et al., 2008) always exceeded the thresh- al., 2004) and on longer scales by continental nutrient supply old for saturation of 150–300 μmol photons m−2 s−1 found (Broecker, 1982). The degree of biogeochemical connectiv- by Nielsen (1997) and Zondervan et al. (2002) and there- ity between the surface and the interior ocean is also con- fore it is assumed that there was no light limitation on cal- trolled by the rate and stoichiometry of sedimenting mate- cification. In this study there was no treatment difference rial. In this work, molar stoichiometric nutrient uptake ratios in nutrient utilisation although the concentrations of Emilia- are inferred from the uptake of dissolved inorganic nutrients nia Huxleyi varied between treatments (Paulino et al., 2008). and thus reflect the community balance at a particular time. Merico et al. (2006) postulated that the seawater carbonate This method, therefore, cannot discriminate species specific ion concentration could be a control on the onset of calcifi- uptake as suggested by Klausmeier et al. (2004) but may be cation and Delille et al. (2005) showed that calcification was used to characterise distinct patterns in community nutrient delayed by 1 day during a 2×CO2 treatment compared to the balance under differing CO2 regimes. www.biogeosciences.net/5/1517/2008/ Biogeosciences, 5, 1517–1527, 2008 1524 R. G. J. Bellerby et al.: Stoichiometry PeECE III

Overconsumption of carbon relative to nutrient supply has ficient transfer of this carbon overconsumption to depth, the been reported in several studies of the planktonic response increase in atmospheric CO2 proposed by the end of the cen- to increased CO2 (Riebesell et al., 1993, 2000; Banse, 1994; tury may be reduced by 58 μatm (Riebesell et al., 2007). Delille at al., 2005; Engel et al., 2005). The increase in car- The efficiency of carbon-nutrient perturbations in the upper bon uptake is usually seen at the end of the bloom from TEP ocean on atmospheric CO2 depends on the rate and ultimate production in response to nutrient stress (Engel, 2002; En- depth of sedimentation which has been shown to be signif- gel et al., 2004). Hein and Sand-Jensen (1997) reported an icantly controlled by the ballast of the sedimenting material immediate (2 h) increase in primary production in relation to (Klaas and Archer, 2002). The ballast is mainly dependant increased CO2 concentrations. Changes in dissolved aque- on the proportion of CaCO3 to particulate organic carbon ous CO2 may determine the phytoplankton cell size distri- (CaCO3:POC) and has been used to test the sensitivity of bution (Engel et al., 2008). Egge et al. (2007) also reported changes to the carbonate pump on atmospheric CO2 (Heinze, increased primary production, based on in situ 14C incuba- 2004; Ridgwell et al., 2007). Previous studies have identi- tions, in this study in the higher CO2 treatments towards the fied a reduction in the calcification rate of pelagic calcifiers peak of the bloom. However, the overconsumption derived (summarised in Ridgwell et al., 2007). Calcification was in- from chemical uptake estimations is seen from the onset of sensitive to CO2 level in this study and Schulz et al. (2008) the experiment and is proportional to the initial treatment show that there was increased export of organic carbon from CO2 concentration (see also Riebesell et al., 2007). Schulz the mixed layer under the 3×CO2 treatment, suggesting a et al. (2008) report no changes to the stoichiometry of dis- lowering of the ballast effect. This would suggest that the at- solved and particulate organic matter. This apparent con- mospheric control proposed by Riebesell et al. (2007) would tradiction may be due to an increase in the carbon exported be diminished if there were no compensatory processes to from the mixed layer relative to nutrient concentration in the offset the reductions in the ballast effect. However, Engel high CO2 treatments. Allgaier et al. (2008) state that the ap- et al. (2008) show that there was an increase in the cell size parent increase in DOC production did not stimulate bacte- during the PeECE II experiment. This cell size change with rial secondary production due to N and P limitation reported increases in TEP production and, thus potentially higher ag- by Tanaka et al. (2008) and that TEP production, DOC exu- gregation of particulates shown to be a response of similar dation and ensuing enhanced sedimentation must have been pelagic plankton communities (Engel, 2002; Engel et al., high. The products and processes of export production were 2004), may be a mechanism to increase sedimentation rates poorly sampled in the waters below the pycnocline and thus in a high CO2 environments. no concrete explanation is possible of the fate of the miss- Carbon overconsumption by osmotrophs changes the nu- ing organic carbon. Schulz et al. (2008) suggest from am- tritional quality of food for zooplankton. Lower growth, pro- monium measurements that there must have been significant ductivity and fecundity has been documented in zooplank- remineralisation in the lower 1.5 m of the mesocosms, espe- ton fed on phytoplankton grown under high CO2 (Sterner cially in the high CO2 treatments fuelled by increased carbon and Elser, 2002; Urabe et al., 2003,; Anderson et al., 2005). export. The role of benthic organisms living on the walls of Carotenuto et al. (2007) hypothesise that the high C:N uptake the mesocosm enclosures, which has been shown to impact during the PeECE III study reduced recruitment of zooplank- biogeochemical cycling during mesocosm studies (Chen et ton nauplii. Changes to the ecological structure of pelagic al., 1997; Berg et al., 1999; Petersen et al., 1999), was not community will also control the export of stoichiometric sig- studied. nals through changes to the size structure and ballast of sedimenting material. 4.4 Consequences for ecosystem functioning, ocean bio- There exists now a large array of experimental data, from geochemical cycling and atmospheric CO2 control laboratory and mesocosm studies of individual species and ecosystem studies, on ecological and biogeochemical re- The elemental stoichiometry of biological export has an im- sponses to ocean acidification. The requirement now is for portant role in controlling atmospheric CO2 concentrations detailed statistical metaanalysis to breakdown the complex- (Broecker, 1982; Volk and Hoffart, 1985; Omta et al., 2006). ity and often contradiction of experimental results, to iden- From first principles, the stoichiometry of osmotroph up- tify the dominant controls and responses to ocean acidifica- take is transferred to the standing stock of pelagic production. With this information, the observationalists and mod- tion. Non-Redfield signatures of biological production have ellers, presently identifying large contemporary changes in been documented in the surface ocean (Sambrotto et al., the oceanic carbon sinks (e.g. Schuster and Watson, 2007; 1993; Anderson and Sarmiento, 1994; Falck and Anderson, Canadell et al., 2007), will be able to focus efforts on the 2005; Koeve, 2006) and at depth (Kortzinger¨ et al., 2001; oceanic processes sensitive to changes in the CO2 system. Pahlow and Riebesell, 2000). This study has indicated an increase in the carbon:nutrient stoichiometry of pelagic os- Acknowledgements. The staff at the Marine Biological Station, motroph productivity in a high CO2 world. If the PeECE University of Bergen, in particular Tomas Sørlie and Agnes results are representative of global productivity, with an ef- Aadnesen, and the Bergen Marine Research infrastructure (RI)

Biogeosciences, 5, 1517–1527, 2008 www.biogeosciences.net/5/1517/2008/ R. G. J. Bellerby et al.: Stoichiometry PeECE III 1525 are gratefully acknowledged for support in mesocosm logis- Discuss., 4, 3913–3936, 2007, tics. This work was partially funded through the EU project http://www.biogeosciences-discuss.net/4/3913/2007/. CARBOOCEAN “Marine carbon sources and sinks assessment” Chen, C.-C., Petersen, J. E., and Kemp, W. M.: Spatial and temporal (contract no. 511176), the Norwegian Research Council projects scaling of periphyton growth on walls of estuarine mesocosms, CABANERA “Carbon flux and ecosystem feedback in the northern Mar. Ecol.-Prog. Ser., 155, 1–15, 1997. Barents Sea in an era of climate change” (contract no. 155936/700) Delille, B., Harlay, J., Zondervan, I., Jacquet, S., Chou, L.,Wollast, and MERCLIM “Marine Ecosystem Response to a Changing R., Bellerby, R. G. J., Frankignoulle, M., Borges, A. V., Riebe- Climate” (contract no. 184860/S30) and the University of Bergen sell, U., and Gattuso, J.-P.: Response of primary production and LOCUS project. This is publication number A181 of the Bjerknes calcification to changes of pCO2 during experimental blooms Centre for Climate Research. of the coccolithophorid Emiliania huxleyi, Global Biogeochem. Cy., 19, GB2023, doi:10.1029/2004GB002318, 2005. Edited by: M. Dai Dickson, A. G.: Standard potential of the reaction: AgCl(s) + 1/2 H (g) = Ag(s) + HCl(aq), and the standard acidity constant of 2 − the ion HSO4 in synthetic seawater from 273.15 to 318.15 K, J. Chem. Thermodyn., 22, 113–127, 1990a. References Dickson, A. G.: Thermodynamics of the dissociation of boric acid in synthetic sea water from 273.15 to 318.15 K, Deep-Sea Res., Allgaier, M., Riebesell, U., Vogt, M., Thyrhaug, R., and Grossart, 37, 755–766, 1990b. H.-P.: Coupling of heterotrophic bacteria to phytoplankton Dickson, A. G. and Millero, F. J.: A comparison of the equilibrium bloom development at different pCO2 levels: a mesocosm study, constants for the dissociation of carbonic acid in seawater media, Biogeosciences, 5, 1007–1022, 2008, Deep-Sea Res., 34, 1733–1743, 1987. http://www.biogeosciences.net/5/1007/2008/. Egge, J. K., Thingstad, T. F., Engel, A., Bellerby, R. G. J., Anderson, L. and Sarmiento, J.: Redfield ratios of remineralization and Riebesell, U.: Primary production during nutrient-induced determined by nutrient data analysis, Global Biogeochem. Cy., blooms at elevated CO2 concentrations, Biogeosciences Dis- 8, 65–80, 1994. cuss., 4, 4385–4410, 2007, Anderson, T. R., Hessen, D. O., Elser, J. J., and Urabe, J.: Metabolic http://www.biogeosciences-discuss.net/4/4385/2007/. stoichiometry and the fate of excess carbon and nutrients in con- Engel, A.: Direct relationship between CO2 uptake and transpar- sumers, Am. Nat., 165(1), 1–11, 2005. ent exopolymer particles production in natural phytoplankton, J. Banse, K.: Uptake of inorganic carbon and nitrate by marine plank- Plankton Res., 24, 49–53, 2002. ton and the Redfield ratio, Global Biogeochem. Cy., 8, 81–84, Engel, A., Thoms, S., Riebesell, U., Rochelle-Newall, E., and Zon- 1994. dervan, I.: Polysaccharide aggregation as a potenial sink of ma- Bellerby, R. G. J., Olsen, A., Furevik, T., and Anderson, L. A.: rine dissolved organic carbon, Nature, 428, 929–932, 2004. Response of the surface ocean CO2 system in the Nordic Seas Engel, A., Zondervan, I., Aerts, K., Beaufort, L., Benthien, A., and North Atlantic to climate change, in: Climate Variability in Chou, L., Delille, B., Gattuso, J.-P., Harley, J., Heemann, the Nordic Seas, edited by: Drange, H., Dokken, T. M., Furevik, C., Hoffmann, L., Jacquet, S., Nejstgaard, J., Pizay, M.-D., T., Gerdes, R., and Berger, W., Geophysical Monograph Series, Rochelle-Newall, E., Schneider, U., Terbruggen,¨ A., and Riebe- AGU, 189–198, 2005. sell, U.: Testing the direct effect of CO2 concentration on a Berg, G. M., Glibert, P. M., and Chen, C,-C.: Dimension effects of bloom of the coccolithophorid Emiliania huxleyi in mesocosm enclosures on ecological processes in pelagic systems, Limnol. experiments, Limnol. Oceanogr., 50, 493–504, 2005. Oceanogr., 44(5), 1331–1340, 1999. Engel, A., Schulz, K. G., Riebesell, U., Bellerby, R., Delille, B., and Blackford, J. C. and Gilbert, F. J.: pH variability and CO2 induced Schartau, M.: Effects of CO2 on particle size distribution and acidification in the North Sea, J. Marine Syst., 1–4, 229–241, phytoplankton abundance during a mesocosm bloom experiment 2007. (PeECE II), Biogeosciences, 5, 509–521, 2008, Brezinski, M.: The Si:C:N ratio of marine diatoms: interspecific http://www.biogeosciences.net/5/509/2008/. variability and the effect of some environmental variables, J. Falck, E. and Anderson, L.G.: The dynamics of the carbon cycle in Phycol., 21, 347–357, 1985. the surface water of the Norwegian Sea, Mar. Chem., 94, 43–53, Broecker, W. S.: Ocean geochemistry during glacial time, 2005. Geochimica Cosmochimica Acta, 46, 1689–1705, 1982. Gattuso, J.-P., Frankignoulle, M., Bourge, I., Romaine, S., and Bud- Caldeira, K. and Wickett, M. E.: Anthropogenic carbon and ocean demeier, R. W.: Effect of calcium carbonate saturation of sea- pH, Nature, 425, 365–365, 2003. water on coral calcification, Global Planet. Change, 18, 37–46, Canadell, J. G., Le Quer´ e,´ C., Raupach, M. R., Field, C. B., Buiten- 1998. huis, E. T., Ciais, P., Conway, T. J., Gillett, N. P., Houghton, R. Goldman, J. C. and Brewer, P. G.: Effect of nitrogen source and A., and Marland, G.: Contributions to accelerating atmospheric growth rate on phytoplankton-mediated changes in alkalinity, CO2 growth from economic activity, carbon intensity, and effi- Limnol. Oceanogr., 25(2), 352–357, 1980. ciency of natural sinks, Proceedings of the National Academy of Gran, G.: Determination of the equivalence point in potentiometric Science, USA, doi:10.1073/pnas.0702737104, 2007. titrations of seawater with hydrochloric acid, Oceanologica Acta, Carotenuto, Y., Putzeys, S., Simonelli, P., Paulino, A., Meyerhofer,¨ 5, 209–218, 1952. M., Suffrian, K., Antia, A., and Nejstgaard, J. C.: Copepod feed- Hein, M. and Sand-Jensen, K.: CO2 Increases Oceanic Primary ing and reproduction in relation to phytoplankton development Production, Nature, 388(6642), 526–527, 1997. during the PeECE III mesocosm experiment, Biogeosciences www.biogeosciences.net/5/1517/2008/ Biogeosciences, 5, 1517–1527, 2008 1526 R. G. J. Bellerby et al.: Stoichiometry PeECE III

Heinze, C.: Simulating oceanic Ca CO3 export produc- L14613, doi:10.1029/2006GL026213, 2006. tion in the greenhouse, Geophys. Res. Lett., 31, L16308, Orr, J. C., Fabry, V. J., Aumont, O., Bopp, L., Doney, S. C., Feely, doi:10.1029/2004GL020613, 2004. R. A., Gnanadesikan, A., Gruber, N., Ishida, A., Joos, F., Key, Johnson, K. M., Sieburth, J. M., Williams, P. J., and Brandstr¨ om,¨ L.: R. M., Lindsay, K., Maier-Reimer, E., Matear, R., Monfray, P., Coulometric total carbon analysis for marine studies: automation Mouchet, A., Najjar, R. G., Plattner, G. K., Rodgers, K. B., and calibration, Mar. Chem., 21, 117–133, 1987. Sabine, C. L., Sarmiento, J. L., Schlitzer, R., Slater, R. D., Tot- Klaas, C. and Archer, D. E.: Association of sinking organic matter terdell, I. J., Weirig, M. F., Yamanaka, Y., and Yool, A.: An- with various types of mineral ballast in the deep sea: Implica- thropogenic ocean acidification over the twenty-first century and tions for the rain ratio, Global Biogeochem. Cy., 16(4), 1116, its impact on calcifying organisms, Nature, 437, 7059, 681–686, doi:10.1029/2001GB001765, 2002. doi:10.1038/nature04095, 2005. Klausmeier, C. A., Litchman, E., Daufresne, T., and Levin, S. A.: Pahlow, M. and Riebesell, U.: Temporal trends in deep ocean Red- Optimal nitrogen-to-phosphorus stoichiometry of phytoplank- field ratios, Science, 287, 831–833, 2000. ton, Nature, 429, 171–174, 2004. Paulino, A. I., Egge, J. K., and Larsen, A.: Effects of increased Kleypas, J. A., Feely, R. A, Fabry, V. J., Langdon, C., Sabine, C. atmospheric CO2 on small and intermediate sized osmotrophs L., and Robbins, L. L.: Impacts of Ocean Acidification on Coral during a nutrient induced phytoplankton bloom, Biogeosciences, Reefs and Other Marine Calcifiers. A Guide for Future Research, 5, 739–748, 2008, Report of a workshop sponsored by NSF, NOAA and the USGS, http://www.biogeosciences.net/5/739/2008/. 88 pp., 2006. Petersen, J. E., Cornwell, J. C., and Kemp, W. M.: Implicit scaling Koeve, W.: C:N stoichiometry of the biological pump in the North in the design of experimental aquatic ecosystems, Oikos, 85, 3– Atlantic: Constraints from climatological data, Global Bio- 18, 1999. geochem. Cy., 20, GB3018, doi:10.1029/2004GB002407, 2006. Pinheiro, J. C. and Bates, D. M.: Mixed-Effects Models in S and Kortzinger,¨ A., Hedges, J. I., and Quay, P. D.: Redfield ratios revis- S-plus, Springer, New York, 527 pp., 2000. ited: removing the biasing effect of anthropogenic CO2, Limnol. Raven, J., Caldeira, K., Elderfield, H., Hoegh-Guldberg, O., Liss, Oceanogr., 46, 964–970, 2001. P., Riebesell, U., Shepherd, J., Turley, C., and Watson, A.: Kuss, J. and Schneider, B.: Chemical enhancement of the CO2 gas Ocean acidification due to increasing atmospheric carbon diox- exchange at a smooth seawater surface, Mar. Chem., 91, 165– ide, Royal Society Report, 2005. 174, 2004. Redfield, A. C., Ketchum, B. M., and Richards, F. A.: The influence Langdon, C., Broecker, W. S., Hammond, D. E., Glenn, E., Fitzsim- of organism on the composition of seawater, in: The Sea, 2. Ed., mons, K., Nelson, S. G., Peng, T.-S., Hajdas, I., and Bonani, edited by: Hill, M. N., Wiley, 26–77, 1963. G.: Effect of elevated CO2 on the community metabolism of Riebesell, U., Wolf-Gladrow, D. A., and Smetacek, V.: Carbon an experimental coral reef, Global Biogeochem. Cy., 17, 1011, dioxide limitation of marine phytoplankton growth rates, Nature, doi:10.1029/2002GB001941, 2003. 361, 249–251,1993. Langer, G., Geisen, M., Klas,¨ J., Riebesell, U., Thoms, S., Young, Riebesell, U., Schulz, K. G., Bellerby, R. G. J., Fritsche, P., Mey- J. R.: Species-specific responses of calcifying algae to changing erhofer,¨ M., Neill, C., Nondal, G., Oschlies, A., Wohlers, J., and seawater carbonate chemistry, Geochem. Geophy. Geosy., 7 (9), Zollner,¨ E.: Enhanced biological carbon consumption in a high Q09006, doi:10.1029/2005GC001227, 2006. CO2 ocean, Nature, 545–548, doi:10.1038/nature06267, 2007. Larsen, J. B., Larsen, A., Thyrhaug, R., Bratbak, G., and Sandaa, Riebesell, U., Zondervan, I., Rost, B., Tortell, P. D., Zeebe, R. E., R.-A.: Response of marine viral populations to a nutrient induced and Morel, F. M. M.: Reduced calcification of marine plankton in phytoplankton bloom at different pCO2 levels, Biogeosciences, response to increased atmospheric CO2, Nature, 407, 364–367, 5, 523–533, 2008, 2000. http://www.biogeosciences.net/5/523/2008/. Riebesell, U.: Effects of CO2 enrichment on marine phytoplankton, Lewis, E. and Wallace, D. W. R.: Program Developed for CO2 Sys- J. Oceanogr., 60, 719–729, 2004. tem Calculations. ORNL/CDIAC-105. Carbon Dioxide Informa- Ridgwell, A., Zondervan, I., Hargreaves, J. C., Bijma, J., and tion Analysis Center, Oak Ridge National Laboratory, U.S. De- Lenton, T. M.: Assessing the potential long-term increase of partment of Energy, Oak Ridge, Tennessee, 1998. oceanic fossil fuel CO2 uptake due to CO2-calcification feed- Merico, A., Tyrrell, T., and Cocakar, T.: Is there any relationship be- back, Biogeosciences, 4, 481–492, 2007, tween phytoplankton seasonal dynamics and the carbonate sys- http://www.biogeosciences.net/4/481/2007/. tem?, J. Marine Syst., 59, 120–142, 2006. Sabine, C. S., Feely, R. A., Gruber, N., Key, R. M., Lee, K., Bullis- Mintrop, L., Fernandez-P´ erez,´ F., Gonzalez´ Davila,´ M., Kortzinger,¨ ter, J. L., Wanninkhof, R., Wong, C. S., Wallace, D. W. R., Till- A., and Santana Casiano, J. M.: Alkalinity determination by brook, B., Millero, F. J., Peng, T.-H., Kozyr, A., Ono, T., and potentiometry- intercalibration using three different methods, Rios, A. F.: The oceanic sink for anthropogenic CO2, Science, Ciencias Marinas, 26, 23–37, 2000. 305, 367–371, 2004. Nielsen, M. V.: Photosynthetic characteristics of the coccol- Sambrotto, R. N., Savidge, G., Robinson, C., Boyd, P., Takahashi, ithophorid Emiliania huxleyi (Prymnesiophyceae) exposed to el- T., Karl, D. M., Langdon, C., Chipman, D., Marra, J., and Codis- evated concentrations of dissolved inorganic carbon, J. Phycol., poti, L.: Elevated consumption of carbon relative to nitrogen in 31, 715–719, 1995. the surface ocean, Nature, 363, 248–250, 1993. Omta, A. W., Bruggeman, J., Kooijman, S. A. L. M., and Dijkstra, Schulz, K. G., Riebesell, U., Bellerby, R. G. J., Biswas, H., Mey- H. A.: Biological carbon pump revisited: Feedback mechanisms erhofer,¨ M., Muller,¨ M. N., Egge, J. K., Nejstgaard, J. C., Neill, between climate and the Redfield ratio, Geophys. Res. Lett., 33, C., Wohlers, J., and Zollner,¨ E.: Build-up and decline of organic

Biogeosciences, 5, 1517–1527, 2008 www.biogeosciences.net/5/1517/2008/ R. G. J. Bellerby et al.: Stoichiometry PeECE III 1527

matter during PeECE III, Biogeosciences, 5, 707–718, 2008, Urabe, J., Togari, J., and Elser, J. J.: Stoichiometric impacts of in- http://www.biogeosciences.net/5/707/2008/. creased carbon dioxide on a planktonic herbivore, Glob. Change Schuster, U. and Watson, A. J.: A variable and decreasing sink for Biol., 9, 818–825, 2003. atmospheric CO2 in the North Atlantic, J. Geophys. Res., 112, Volk, T. and Hoffert, M. I.: Ocean carbon pumps, analysis of rela- C11006, doi:10.1029/2006JC003941, 2007. tive strengths and efﬁciencies in ocean-driven atmospheric CO2 Sterner, R. W. and Elser, J. J.: Ecological Stoichiometry, Princeton changes, in: The Carbon Cycle and Atmospheric CO2: Natu- University Press, Oxford, UK, 2002. ral Variations Archean to Present, edited by: Sunquist, E. T. and Tanaka, T., Thingstad, T. F., Løvdal, T., Grossart, H.-P., Larsen, A., Broecker, W. S., Geophysical Monograph Series, Vol. 32, Wash- Allgaier, M., Meyerhofer,¨ M., Schulz, K. G., Wohlers, J., Zollner,¨ ington, DC, 99–110, 1985. E., and Riebesell, U.: Availability of phosphate for phytoplank- Wanninkhof, R. and Thoning, K.: Measurement of fugacity of CO2 ton and bacteria and of glucose for bacteria at different pCO2 in surface water using continuous and discrete sampling meth- levels in a mesocosm study, Biogeosciences, 5, 669–678, 2008, ods, Mar. Chem., 44, 189–205, 1993. http://www.biogeosciences.net/5/669/2008/. Weiss, R. F.: Carbon dioxide in water and seawater: the solubility Tortell, P. D., DiTullino, G. R., Sigman, D. M., and Morel, F. M. M.: of a non-ideal gas, Mar. Chem., 2, 203–215, 1974. CO2 effects on taxonomic composition and nutrient utilisation Zondervan, I., Rost, B., and Riebesell, U.: Effect of CO2 con- in an Equatorial Paciﬁc phytoplankton assemblage, Mar. Ecol.- centration on the PIC/POC ratio in the coccolithophore Emilia- Prog. Ser., 236, 37–43, 2002. nia huxleyi grown under light-limited conditions and different Tyrrell, T. and Merico, A.: Emiliania huxleyi: bloom observations daylengths, J. Exp. Mar. Biol. Ecol., 272, 55–70, 2002. and the conditions that induce them, in: Coccolithophores: from Zondervan, I.: The effects of light, macronutrients, trace metals and molecular processes to global impact, edited by: Thierstein, H. CO2 on the production of calcium carbonate and organic carbon R. and Young, J. R., Berlin, Germany, Springer, 75–97, 2004. in coccolithophores – A review, Deep-Sea Res. II, 54, 521–537, doi:10.1016/j.dsr2.2006.12.004, 2007.

www.biogeosciences.net/5/1517/2008/ Biogeosciences, 5, 1517–1527, 2008

Paper 7

Climate Science: The New Frontier Nondal, G., P. M. Haugan, K. S. Seim, and R. G. J. Bellerby. Under review for Climatic Change.

VII

Climatic Change manuscript No. (will be inserted by the editor)

Climate Science: The New Frontier

Gisle Nondal · Peter M. Haugan · Knut S. Seim · Richard G. J. Bellerby

Received: date / Accepted: date

Abstract Climate change has become one of the most important themes of discussion in the world today, greatly amplified by the Intergovernmental Panel on Climate Change (IPCC), and Vice President Al Gore, being awarded the Nobel Peace Prize in 2007. Decades of research have led to a well-established hypothesis that anthropogenic greenhouse gas emissions contribute significantly to contemporary global warming. Nevertheless, scientists are not able to provide future climate projections with sufficient methodological certainty. This represents a new situation in which the focus has shifted towards the epistemological uncertainties. The public debate regarding the legitimacy of the IPCC, and its climate projections, has been ongoing since the start, however it has peaked after the so-called ”Cli- mategate” incident (November 2009), and the IPCC having to modify a statement regarding the rate of melting of Himalayan glaciers (January 2010). Although the major scientific basis of the IPCC study still hold after these episodes, climate change research is currently facing a concerted challenge to its legitimacy. We discuss two conceptual models describing the relationship-, and interface-, between

G. Nondal Geophysical Institute, University of Bergen, Allégaten70, N-5007 Bergen, Norway Mohn-Sverdrup Center/Nansen Center, Thormøhlensgate 47, N-5006 Bergen, Norway Tel.: +47 55 20 58 25 Fax: +47 55 20 58 01 E-mail: [email protected] P. M. Haugan Geophysical Institute, University of Bergen, Allégaten70, N-5007 Bergen, Norway Bjerknes Centre for Climate Research, University of Bergen, Allégaten55, N-5007 Bergen, Norway

K. S. Seim Norwegian University of Science and Technology (NTNU), Trondheim, Norway

R. G. J. Bellerby Bjerknes Centre for Climate Research, University of Bergen, Allégaten55, N-5007 Bergen Norway Geophysical Institute, University of Bergen, Allégaten70, N-5007 Bergen, Norway 2 Gisle Nondal et al. science and policy with regards to climate change. We suggest remedying the current crisis by a democratization of climate science, following the philosophy of Post Normal Science, and point at current examples of how the scientific community is already developing this methodology. Keywords Post Normal Science · Climate change · IPCC

1 Introduction

”Man wants to know; and when man no longer wants to know, he will no longer be man” Fridtjof Nansen (1861-1930)

This statement, made by the Norwegian scientist and humanist Fridtjof Nansen, illustrates mankind’s continuous quest for enhanced knowledge and certainty. This has been a cornerstone in science and philosophy at least since the time of the ancient Greeks (Fjelland, 2002). The basic principle of science is the Scientific Method. Through this method, scientists promote hypotheses, that are open to falsification, using available observations and experimental results. The Scientific Method does not produce absolute truths or knowledge; however, when a hypothesis endures continuous scrutiny, it can approach the status of a theory, and thus an accepted view in the research community. In the scientific sphere, openness with regards to scientific data, documentation, substantiation, scepticism, and criticism, are well-established key features. The concept of scientific uncertainty has been recognized for several hun- dred years, and statistical methods have been invented to quantify the uncertainties. Fjelland (2002) argues that we are now facing a new situation in which the uncertainty has moved into the limelight. The global climate change issue is a pronounced example of this new situation. At present, it is clear that our planet has experienced a significant temperature increase over the past 50 years (IPCC, 2007c,a,b), and the atmospheric carbon dioxide (CO2) concentration has exceeded levels not seen in at least 650 000 years (Siegenthaler et al., 2005). Al- though the potential effect of CO2 on climate was first described by Arrhenius (1896), it was only twenty-two years ago, that James Hansen at NASA God- dard Institute for Space Studies, testified in front of a US Senate committee, explaining that he could say with ”99 percent confidence” that a recent persis- tent rise in global temperature was occurring. He further stated that ”the greenhouse effect has been detected and is changing our climate now”. This testimony (based on the findings of Hansen et al. (1988)) is considered the dawn of the contemporary global warming discussion (http://discovermagazine.com/1988/jun/ 23-special-report-endless-summer-living-with-the-greenhouse-effect), although already the Charney report (Charney, 1979) contained much of the scientific basis. Also in 1988, the World Meteorological Organization (WMO), and the United Nations Environment Programme (UNEP) established the Intergovern- mental Panel on Climate Change (IPCC) as a scientific body whose goal was to evaluate ”any change in climate over time whether due to natural variability or as a result of human activity”. IPCC stated in its fourth assessment report that it is very likely (more than 90 %) that the observed warming over the last 50 years is Climate Science: The New Frontier 3 attributed to the increasing greenhouse gas concentrations in the atmosphere, with CO2 identified as the principal anthropogenic greenhouse gas (IPCC, 2007c,a,b). Although IPCC states with more than 90 % certainty that contemporary global warming is linked to anthropogenic CO2 emissions, it is still a matter of contro- versy how important these emissions are, and what effects continuous emissions will have on the climate, and ultimately for life on Earth. Climate models project ◦ ◦ a temperature increase between 1.5 C and 6 C by year 2100, having potential adverse effects on life as we know it (IPCC, 2007c,a,b). The computer climate models are known to contain shortcomings causing substantial uncertainties in the estimates. The reliability of these projections is therefore a matter of prominent debate. This represents a new frontier, in which science faces a problem where it can no longer provide answers with sufficient certainty (May, 2001; Pielke, 2001). How can society and policy makers relate to such a shift towards more focus on epistemological uncertainties? (e.g. ”what can we know about this phenomenon?” or ”how do we know what we know?”, (Funtowicz and Strand, 2007)).

2 Models of science and policy

Funtowicz and Strand (2007) outline diﬀerent conceptual models describing the relationship, and interface, between science and policy with regards to environmental issues. We will in the following discuss two such models and their role in the global climate change issue.

2.1 The modern model

The modern model assumes that science provides objective, valid, and reliable information with uncertainties that can be controlled (or eliminated). This information is then passed on to the policy makers, telling them everything they need to know in order to make decisions. In this model, ”... there is only one correct description of the system, and it is to be provided by science” (Funtowicz and Strand, 2007). The scientiﬁc approach is what Ravetz (2004) refers to as ”main- stream science”. This employs reductionistic thinking in which complex objects are reduced to their parts, which may further be reduced to simple elements. Af- ter reducing the whole to its parts, the parts can be re-assembled again, providing enhanced knowledge.

2.2 The model of extended participation

An alternative to (various versions of) the modern model is called the model of extended participation (Funtowicz and Strand, 2007). It is based on the philosophy of post-normal science (PNS), described by Funtowicz and Ravetz (1990). PNS should, by deﬁnition, be employed when dealing with problems having large uncertainties, where the values are in dispute, the decision stakes are high, and the problem at hand requires urgent decisions (Funtowicz and Ravetz, 1990; Ravetz, 2004). Climate science is a prime candidate for application of PNS (Bray and von Storch, 1999; Saloranta, 2001; Hulme, 2007). The methodology should not replace 4 Gisle Nondal et al. what Kuhn calls ”normal” science, but aid in making the correct decisions based on the current knowledge at hand. One important aspect of this method is the inclusion of non-experts, creating an ”extended peer community” (Ravetz, 2004). In other words, non-scientists should be included in e.g. discussions about scien- tiﬁc results, and in science strategy debates, and there should be much interaction between scientists and non-scientists to lead up to decision-making.

3 Discussion

3.1 The modern model and climate science

It is well established that climate scientists gain insight into mechanisms driving the climate system by reducing it to processes on smaller and smaller scales. The problem is that the climate system is complex, and governed by non-linear dynamics. The concept of non-linearity can, in its simplest form, be described by the Logistic Map (May, 1976) represented by the following equation:

Xnext = aXinitial(1 − Xinitial) (1)

This equation can for example be used for studying long-term changes in animal populations May (2003). The behavior of the equation is thoroughly described by May (1976, 2003), but the main point to notice is that although being relatively simple, it exhibits great sensitivity to the starting conditions. Small differences in the starting conditions can cause large variations in the long-term behavior of the system. Such a behavior was noted by Lorenz (1963), when studying a simple model of the atmosphere. He famously denoted this as ”the butterfly effect”. Hysteresis and the possibility of abrupt jumps between distinctly different stable modes are illustrated by the simple non-linear system of equations used by Stommel (1961) to investigate sensitivity of global ocean circulation to surface fresh water forcing. Thus for non-linear systems, ”adding back does not work ...” (Fjelland, 2002).

3.2 Climate science and society

The IPCC, together with vice president Al Gore, were awarded the Nobel Peace Prize in 2007 ”... for their efforts to build up and disseminate greater knowledge about man-made climate change, and to lay the foundations for the measures that are needed to counteract such change.” (Nobel Committee, Press release 2007; http://nobelprize.org/nobel_prizes/peace/laureates/2007/press.html ). The Nobel Committee further stated that, ”... the IPCC has created an ever-broader informed consensus about the connection between human activities and global warming”. This latter statement has been interpreted as controversial. The word consensus has a very powerful meaning that could be exploited to create a false sense of agreement (Pielke, 2001). In the US, for example, over 31000 persons with Bachelor or higher degrees in science and technology (9000 PhDs), have signed ”the Global Warming Petition” (http://www.petitionproject.org/), where they ”... urge the United States government to reject the global warming agreement that Climate Science: The New Frontier 5 was written in Kyoto, Japan in December, 1997, and any other similar propos- als...”. The number of these who are scientists directly involved in climate research remains unknown. Skepticism and criticism are keystones in science. Un- fortunately, several commentators and self-proclaimed ”climate watchdogs” have either naively, or deliberately, mis-represented climate researchers with misleading interpretations of the present state of understanding of the processes controlling contemporary climate change. In addition it has been suggested that IPCC selec- tively pick out the scientists that contribute to their assessment reports. In 1709, the English writer Alexander Pope wrote in the poem An Essay on Criticism that ”A little learning is a dangerous thing” (http://poetry.eserver. org/essay-on-criticism.html). Here he is referring to the ease of disparagement, and to the often dogmatic views that hide behind much of the criticism. If the stakes in climate change were not so high, then the comments and criticisms aimed at scientists would be simply unreasonable, but when empty rhetoric and incorrect interpretation of scientific results is used as a basis for influencing public opinion, and ultimately future international policy, it is important that the public and politicians are made aware of the present state of knowledge. Non-governmental organizations (NGOs) are prominent in the debate on climate as well as other environmental issues. There is a wide variety of NGOs and it is hard to make generalizations about their roles or working methods. However, the primary mission of NGOs is to play a significant role in decision-making. Many of them do this by providing pointed statements and conclusions and try to back these up with science that supports their chosen policy option. They are experts in framing the issues and creating media attention sometimes by focusing on sym- bol cases. Even if their primary role is not to conduct research, some NGOs have academic staff. Some also invite scientists on science advisory boards. This offers another arena for scientists to engage in policy-making. In November 2009, a data server at the Climate Research Unit (CRU) at the University of East Anglia was hacked, and thousands of emails, including documents, from prominent climate researchers were stolen and posted on the Inter- net (e.g. Nature Editorial (2009)). This incident has gained tremendous attention worldwide, and has been popularly termed ”Climategate”, clearly analogous to the Watergate scandal in the US in the 1970s. Although the global temperature increase seen in the historical temperature reconstruction dataset at CRU (and used by IPCC) remains robust in the wake of Climategate (Nature News, 2010; Nature News Feature, 2010), the leaked emails have been interpreted as serious violations of basic scientific principles. It has been claimed that scientists have systematically attempted to withhold data, which should be publicly available under the US and UK Freedom of Information Acts, and also violated good practice concerning the peer-review process. Furthermore, IPCC recently released a statement saying that a paragraph in their latest assessment report ”... refers to poorly substantiated estimates of rate of recession and date for the disappearance of Himalayan glaciers” (IPCC, 2010). The claim that Himalayan glaciers could disappear by 2035 was not based on peer- reviewed scientific literature, but on a media interview with a scientist performed in 1999. This error not only showed that the internal communication between different parts of the IPCC was not working according to the expected standard, but also has nurtured skepticism in the public at large towards the IPCC leadership 6 Gisle Nondal et al. which is seen to create unfounded attention by deliberate overestimation of climate change impacts. Returning to Pope (1709); ”most critics, fond of some subservient art, still make the whole depend upon a part”. This statement illustrates that although neither Climategate nor the Himalayan glacier case has impacted the main IPCC conclusions, they have contributed to a present legitimacy crisis with regards to climate science. A recent survey shows that 57 % of Americans polled late 2009/early 2010 believe that climate change is happening, compared to 71 % in October 2008 (Leis- erowitz et al., 2010). A question of vital importance is thus how climate science can regain the trust of the society?

3.3 Climate sciences new social contract with society

To our mind, the answer to the above question lies in the model of extended participation. By employing this model (and thus PNS), a scientific quality assurance system split into an internal and an external component is favorable (Funtowicz and Strand, 2007). The internal component should correspond to the traditional scientific ”peer-review” process. This component ensures that results published in scientific journals have been objectively scrutinized by leading experts in the field, thereby minimizing the probability of fundamental errors in the research. The statement made by the Roman poet Juvenal ”Quis custodiet ipsos custodes?” (∼ who will guard the guardians?) shows that this system is not fool-proof (Fun- towicz, 2001). However it is worth pointing out, again, that skepticism is at the very heart of the Scientific Method. Thus any climate scientist (or more generally every scientist) should, by nature, be a skeptic. This point should be made very clear to the public. The external component would be the extended peer community. Gibbons (1999) referred to ”Sciences new social contract with society”, when discussing the ongoing shift in modern science from ”reliable” towards ”socially robust” knowledge. One can also easily draw a parallel to civic science, thoroughly described by (Bäckstrand, 2009). Bearing in mind the poem by Alexander Pope, is there an optimal way of connecting science with society? We believe that a statement from President Franklin D. Roosevelt is very relevant here:

”Democracy cannot succeed unless those who express their choice are prepared to choose wisely. The real safeguard of democracy, therefore, is education.” - Franklin D. Roosevelt, September 27, 1938

The democratization of science is at the heart of the model of extended participation. Saloranta (2001) suggested that the climate science involving IPCCs second assessment report (IPCC, 1996b,a) could to a large extent be characterized as PNS, as IPCC did employ an extended-peer community, as well as an enhanced management of uncertainties. Saloranta (2001) also stated that these efforts associated with the early stages of the IPCC seemed successful in bridging the gap between the scientific and non-scientific communities and making tangible progress towards binding agreements for limiting emissions. While the IPCC mechanism presently does not seem so successful, we see clear opportunities and also signs in the community towards testing methods for a more societally integrated science: Climate Science: The New Frontier 7

1. Use of modern communication technology. The Internet is an enormous resource in modern society. The use of Google (www.google.com) and Wikipedia (www.wikipedia.org) as means of obtaining knowledge has skyrocketed in recent years, and the use of blogs (short for web logs) as a forum for public debate has increased significantly (Farrell and Drezner, 2008). A lot of discussions about contemporary climate change can be found in the ”blogosphere”, but as suggested by Ashlin and Ladle (2006), ”blogs on environmental topics show that they vary greatly in accuracy, which indicates a need for participation by informed scientists”. This represents a new challenge for scientists in which they are urged to participate more actively in online forums. As stated to Nature by the climatologist Robert Pielke Sr. when talking about his blog, ”Now Im making my arguments to a broader community to see how well they stand up. I also use it as a professional diary and it has increased my network. The feedback has been wonderful” (Nature 2006, 440, p597-599). Other examples of such activity include the blogs RealClimate (www.realclimate.org), ClimatePol- icy (www.climatepolicy.org), ClimateEthics (http://climateethics.org/), and ClimateCentral (www.climatecentral.org) where scientists present their views and results directly to the public. 2. Open-access publishing. The primary means of obtaining scientific results is through publications in scientific journals. However the majority of these are only available through relatively expensive subscriptions, making them difficult to obtain outside scientific institutions. In addition, the review process is confidential, and only available to the author(s) and the editor(s). There is currently a trend towards more open-access, freely available, journals. Coperni- cus Publishing (http://publications.copernicus.org/) is, to the best of our knowledge, the first publisher fully endorsing the model of extended participation. Manuscripts submitted to its various journals (currently ten Geosciences and Earth System Science journals) have to go through a public online review process, where input from both scientists and the public is welcome. This will contribute to bridging the gap between scientists and non-scientists. 3. Formal involvement of non-scientists in scientific activities. Research projects are initiating the use of so-called Reference User Groups (RUGs). An example is the European Project on OCean Acidification (EPOCA). ”The results of this multi-disciplinary collaboration will be presented to the non-scientific community (business leaders, organisations and the general public) to inform about the risks and what we can do to avoid the tipping points related to ocean acidification... . The RUG, ... is working with EPOCA to examine in detail the user related issues (the types of data, analyses and products that are most use- ful to managers, policy advisors, decision makers and politicians, the format and nature of key messages arising from the EPOCA research, and the dis- semination procedures) RUG members also feedback key science developments into their own sector/parent organisation during the life time of the project.” (http://www.epoca-project.eu/index.php/Outreach/RUG/).

4 Concluding remarks and comments

Climate change has, without a doubt, become one of the most important themes of discussion in the world today. There is still pronounced debate regarding mankinds 8 Gisle Nondal et al. contribution to the observed climate changes. Since the stakes are high, the uncertainties are large and the challenge of future climate changes requires urgent action, this is a prime candidate for applying the PNS approach, through the model of extended participation. After a promising start (Saloranta, 2001), it presently seems that the IPCC process has failed. To our knowledge, there are no in-depth analyses of the actual penetration and impact of extended participation on climate science in the past decade. Perhaps the reductionistic modern model has been dominating in practice despite some formally post normal mechanisms and attributes. We have shown that contemporary climate research has new tools available for adapt- ing and developing methodology for extended participation. We are conﬁdent that both science and the public have, and will, beneﬁt from this approach, and that it is a way forward for climate science to regain the trust of the society.

”New frontiers of the mind are before us, and if they are pioneered with the same vision, boldness, and drive with which we have waged this war we can create a fuller and more fruitful employment and a fuller and more fruitful life.” - President Franklin D. Roosevelt, November 17, 1944.

Acknowledgements Comments by R. Strand to a previous version of this manuscript, sig- niﬁcantly improved the quality of the paper. GN acknowledges ﬁnancial support through an endowment by Mr. Trond Mohn, c/o Frank Mohn A/S.

References

Arrhenius, S., 1896. On the influence of carbonic acid in the air upon the temperature of the ground. Philosophical Magazine and Journal of Science 41 (Series 5), 237–276. Ashlin, A., Ladle, R. J., 2006. Science communication - environmental science adrift in the blogosphere. Science 312 (5771), 201–201, times Cited: 5. Bäckstrand, K., 2009. Civic science for sustainability: Reframing the role of experts, policymakers and citizens in environmental governance. Global Environ- mental Politics 3:4, 24–41. Bray, D., von Storch, H., 1999. Climate science: An empirical example of postnor- mal science. Bull. Amer. Meteorol. Soc. 80, 439–455. Charney, J. G. e. a., 1979. Carbon dioxide and climate: A scientific assessment. Tech. rep., National Academy of Sciences, 22pp. Farrell, H., Drezner, D. W., 2008. The power and politics of blogs. Public Choice 134 (1-2), 15–30, times Cited: 7. Fjelland, R., 2002. Facing the problem of uncertainty. Journal of Agricultural & Environmental Ethics 15 (2), 155–169, times Cited: 1 11th International Congress of Logic, Methodology and Philosophy of Science AUG 20-26, 1999 KRAKOW, POLAND. Funtowicz, S., 2001. Peer review and quality control. In: International Encyclope- dia of the Social and Behavioral Sciences. pp. 11179–11183, iSBN: 0-08-043076-7. Funtowicz, S., Ravetz, J., 1990. Uncertainty and quality in Science for policy. Kluwer Academic Publishers. Climate Science: The New Frontier 9

Funtowicz, S., Strand, R., 2007. Biosafety first: Holistic Approaches to Risk and Uncertainty in Genetic Engineering and Genetically Modified Organisms. Tapir, Trondheim, Ch. Models of Science and Policy, pp. 263–278. Gibbons, M., 1999. Science’s new social contract with society. Nature 102, Supp2., C81–C84. Hansen, J., Fung, I., Lacis, A., Rind, D., Lebedeff, S., Ruedy, R., Russell, G., Stone, P., 1988. Global climate changes as forecast by goddard institute for space studies 3-dimensional model. Journal of Geophysical Research-Atmospheres 93 (D8), 9341–9364, times Cited: 532. Hulme, M., 2007. The appliance of science. The Guardian (http://www.guardian. co.uk/society/2007/mar/14/scienceofclimatechange.climatechange). IPCC, 1996a. Climate change 1995: impacts, adaptions and mitigation of climate change: Scientific-technical analyses. In: Watson, R. T., Zinyowera, M. C., Moss, R. H. (Eds.), Contribution of Working Group II to the Second Assessment of the Intergovernmental Panel on Climate Change. Cambridge University Press, UK, p. 18. IPCC, 1996b. Climate change 1995: The science of climate change. In: Houghton, J. T., Filho, L. G. M., Callender, B. A., Harris, N., Kattenberg, A., Maskell, K. (Eds.), Contribution of Working Group I to the Second Assessment of the Intergovernmental Panel on Climate Change. Cambridge University Press, UK, p. 878. IPCC, 2007a. Climate change 2007: Climate change impacts, adaption and vul- nerability. IPCC, 2007b. Climate change 2007: Mitigation of climate change. IPCC, 2007c. Climate change 2007: The physical science basis. IPCC, 2010. Ipcc statement on the melting of himalayan glaciersGeneva, 20. Jan- uary 2010. Leiserowitz, A., Maibach, E., Roser-Renouf, C., 2010. Climate change in the american mind: Americans’ global warming beliefs and attitudes in january 2010Yale University and George Mason University. New Haven, CT: Yale Project on Cli- mate Change. Lorenz, E. N., 1963. Deterministic nonperiodic flow. Journal of the Atmospheric Sciences 20 (2), 130–141, times Cited: 5139. May, R. M., 1976. Simple mathematical-models with very complicated dynamics. Nature 261 (5560), 459–467, times Cited: 2175. May, R. M., 2001. Risk and uncertainty. Nature 411, 891. May, R. M., 2003. It must be beautiful. Great equations of modern science. London: Granta Books, Ch. The best possible time to be allive: The logistic map, pp. 28–45. Nature Editorial, 2009 462, 545. Nature News, 2010 463, 860. Nature News Feature, 2010 463, 285–287. Pielke, R. A., 2001. Room for doubt. Nature 410 (6825), 151–151, times Cited: 12. Ravetz, J., 2004. The post-normal science of precaution. Futures 36 (3), 347–357, times Cited: 11. Saloranta, T. M., 2001. Post-normal science and the global climate change issue. Climatic Change 50 (4), 395–404, times Cited: 11. Siegenthaler, U., Monnin, E., Kawamura, K., Spahni, R., Schwander, J., Stauffer, B., Stocker, T. F., Barnola, J. M., Fischer, H., 2005. Supporting evidence from 10 Gisle Nondal et al.

the epica dronning maud land ice core for atmospheric co2 changes during the past millennium. Tellus Series B-Chemical and Physical Meteorology 57 (1), 51–57, times Cited: 26. Stommel, H., 1961. Thermohaline convection with 2 stable regimes of ﬂow. Tellus 13 (2), 224–230, times Cited: 450.