The influence of Southern Ocean winds on the North Atlantic carbon sink

Abstract

Observed and predicted increases in Southern Ocean winds are thought to upwell deep ocean carbon and increase atmospheric CO₂. However, Southern Ocean dynamics affect biogeochemistry and circulation pathways on a global scale. Using idealized Massachusetts Institute of Technology General Circulation Model (MITgcm) simulations, we demonstrate that an increase in Southern Ocean winds reduces the carbon sink in the North Atlantic subpolar gyre. The increase in atmospheric CO₂ due to the reduction of the North Atlantic carbon sink is shown to be of the same magnitude as the increase in atmospheric CO₂ due to Southern Ocean outgassing. The mechanism can be described as follows: The increase in Southern Ocean winds leads to an increase in upper ocean northward nutrient transport. Biological productivity is therefore enhanced in the tropics, which alters the chemistry of the subthermocline waters that are ultimately upwelled in the subpolar gyre. The results demonstrate the influence of Southern Ocean winds on the North Atlantic carbon sink and show that the effect of Southern Ocean winds on atmospheric CO₂ is likely twice as large as previously thought in past, present, and future climates.

Key Points

• Increased Southern Ocean winds reduce the North Atlantic carbon sink
• The effect of Southern Ocean winds on atmospheric pCO₂ is doubled by the nonlocal feedback
• Atlantic tropical biology affects the North Atlantic subpolar gyre Revelle buffer factor

1 Introduction

Paleoproxy and historical data show large changes in atmospheric pCO₂, indicating variations in the atmosphere-ocean exchange of carbon. Changes in atmospheric forcing, such as Southern Hemisphere westerly wind stress, can have a strong influence on the carbon cycle. The impact of observed and projected Southern Ocean wind increase [Swart et al., 2014; Bracegirdle et al., 2013] on the Southern Ocean carbon flux has been extensively studied [e.g., Le Quere et al., 2007; Lovenduski et al., 2008; Lovenduski and Ito, 2009; Ito et al., 2010; Lauderdale et al., 2013; Munday et al., 2014]. Increased Southern Ocean winds are thought to increase upwelling of deep ocean natural carbon, elevating local ocean pCO₂. The elevated ocean pCO₂ leads to a reduction in local sink of anthropogenic CO₂ and a subsequent increase in atmospheric CO₂ levels.

However, little attention has gone into estimating the climate impacts of dynamical and biogeochemical changes induced by the Southern Ocean wind stress outside of the Southern Ocean. The Southern Ocean circulation forms part of the global meridional overturning circulation [Gnanadesikan, 1999; Marshall and Speer, 2012] and has a strong influence on transport pathways [Ito et al., 2010; Palter et al., 2010], including in the North Atlantic. Southern Ocean winds act to ventilate the deep ocean, which is richer in nutrients than the surface ocean. The wind-driven ventilation in the Southern Ocean mixes nutrients into the surface ocean, giving it a large effect on ocean biogeochemistry and biological productivity on a global scale [Marinov et al., 2006]. Anderson et al. [2002] and Sarmiento et al. [2004] show that increased Southern Ocean upwelling increases nutrient supply north of the Southern Ocean and enhances high- and low-latitude biological productivity. Changes induced in the Southern Ocean can therefore have far-reaching impact on ocean dynamical and biogeochemical properties.

Studies such as Lauderdale et al. [2013] examine the basin-scale response to Southern Ocean wind stress perturbations on the dynamics and distribution of carbon but focus on the local contribution of wind stress on atmospheric pCO₂. The nonlocal effect of the Southern Ocean winds on atmospheric CO₂, via change in global-scale transport pathways and in nutrient export from the Southern Ocean, remains to be estimated. We focus primarily on understanding the response of other major carbon sinks to changes induced in the Southern Ocean.

In the Atlantic, the major oceanic sink of atmospheric carbon (both natural and anthropogenic) is the subpolar gyre [Group, 1998; Dickson et al., 2008; Khatiwala et al., 2009; Kieke and Yashayaev, 2014]. Any changes in the North Atlantic ocean carbon chemistry will therefore greatly impact the magnitude of the local carbon sink and subsequently alter atmospheric CO₂ levels. Halloran et al. [2015], for example, show that the biogeochemical properties of the subpolar gyre are affected by the water advected from the tropics and Southern Ocean. Southern Ocean winds have the potential to influence the North Atlantic carbon sink by affecting the biogeochemical properties of the waters advected from the Southern Ocean to the North Atlantic.

The aim of this paper is to assess the nonlocal effect of Southern Ocean winds on atmospheric CO₂ and to demonstrate that increased Southern Ocean winds affect the North Atlantic pCO₂ through a novel mechanism. The mechanism can be summarized as follows: increased Southern Ocean wind stress enhances nutrient transport toward the tropics, thereby increasing tropical biological productivity. This in turn affects the chemistry of underlying subthermocline waters. The tropical subthermocline waters are subsequently advected to the North Atlantic subpolar gyre by equatorial currents and the meridional overturning circulation (MOC), reducing the capacity of the subpolar gyre to absorb atmospheric CO₂.

The paper is structured as follows. We provide a description of the model used in this paper in section 2, with details regarding the model design and mean state. Section 3 describes and quantifies the effect of Southern Ocean wind perturbations on the North Atlantic carbon sink. The subsequent section 4 will detail the key steps in the mechanism, and section 5 summarizes the main conclusion, caveats, and implications of the presented study.

2 Model Design and Control Experiment

To study the effect of Southern Ocean winds on the North Atlantic carbon sink, we use the Massachusetts Institute of Technology General Circulation Model (MITgcm) [Marshall et al., 1997] in an idealized configuration designed to mimic the Atlantic and Southern Oceans. The setup and parameters are similar to Munday et al. [2014] at 2° × 2° horizontal grid spacing with eddies parametrized using the Gent-McWilliams [Gent and McWilliams, 1990] scheme with a constant coefficient. The domain, shown in Figure 1a, is a sector 20° wide in longitude, extending to 60°N/S with a reentrant channel between 60° and 40°S to allow for a circumpolar current. The basin is 5000m deep, except at the channel boundaries where a 2500m deep sill represents Drake Passage. Coupled to the ocean, there is a well-mixed atmospheric box which only solves for atmospheric pCO₂, and the air-sea flux of carbon is parametrized following Wanninkhof [1992]. The model resolution is chosen such that a large set of perturbation experiments can be run for centuries while still capturing the main features of a large-scale circulation and the appropriate response of the ocean carbon cycle. Such a setup is widely used to conduct large-scale ocean studies [e.g., Wolfe and Cessi, 2010; Ito and Follows, 2003; Munday et al., 2013].

Figure 1

(a) Schematic of MITgcm sector. Latitudinal profiles of surface boundary conditions for the control experiment of (b) wind, with the dashed line showing a 2 × τ_SO perturbation, (c) temperature, and (d) salinity. Control experiment steady state (e) residual overturning stream function and (f) zonal mean velocity.

The model is forced with a zonal wind stress profile shown in Figure 1b with N/m² (where the subscipt “SO” indicates the Southern Ocean between 60° and 40°S and the overline indicates a spatial average). Heat and freshwater fluxes are prescribed through temperature and salinity restoring profiles (Figures 1c and 1d), which broadly match the observed distributions in the Atlantic. The resulting residual (sum of Eulerian and eddy) overturning stream function is shown in Figure 1e [Marshall and Radko, 2003].

The model also supports a series of equatorial and midlatitude zonal currents, with a westward equatorial current, an eastward undercurrent [Knauss, 1960; Philander, 1973; Wyrtki and Kilonsky, 1984; Philander et al., 1987], and a series of stationary eastward jets on either side of the equator [Tsuchiya, 1972; Firing et al., 1998; Rowe et al., 2000; Izumo et al., 2002; Qiu et al., 2013] (Figure 1f). While the structures are similar to observations, the magnitude of the currents in the MITgcm are smaller, likely due to the model geometry and its coarse horizontal and vertical resolution. The equatorial jets, alongside vertical and horizontal mixing, are important factors for ventilation of the shadow zone [Brandt et al., 2015]: a region of the subsurface ocean that is unventilated by the wind-driven gyre circulation.

The ocean model contains a biogeochemical component with seven tracers: dissolved inorganic carbon (DIC), alkalinity, preformed alkalinity, phosphate, preformed phosphate, dissolved organic phosphorous, and dissolved oxygen [Dutkiewicz et al., 2008; Follows et al., 2006]. Biological productivity is governed by phosphate and light limitation. Figures 2a–2c show the zonally averaged distribution of DIC, regenerated DIC, and dissolved oxygen. The maximum in DIC concentration is located in the abyssal cell at high latitudes where water is oldest. Regenerated DIC concentrations are generally high where dissolved oxygen concentrations are low since oxygen is used up in the remineralization of soft organic matter. The upper ocean distributions of DIC and alkalinity are in agreement with Global Data Analysis Project (GLODAP) climatology [Key et al., 2004].

Figure 2

Zonally averaged (a) total dissolved inorganic carbon, (b) regenerated dissolved inorganic carbon, (c) dissolved oxygen, and (d) dissolved phosphate. Dashed black lines in Figures 2a–2d indicate the extent of the channel, and the solid black lines are residual overturning stream function contours spaced 1Sv (sverdrup, 10⁶ m³/s) apart. (e) Zonally averaged meridional nutrient stream as defined by equation 5, with dashed lines showing surfaces of constant density and the thick black line showing the σ_θ=26 kg/m³ density surface.

The model shows the presence of an oxygen-minimum zone (OMZ) in the shadow zone between 100 and 1000 m depth, with corresponding high regenerated carbon concentrations. The lack of wind-driven ventilation in the shadow zone leads to a buildup of organic carbon and the associated removal of dissolved oxygen. The meridional and vertical extent of the OMZ compares well with other studies such as Ito and Follows [2005], although the magnitude is slightly higher in the presented model compared to other studies. For a more detailed description of the distribution of biogeochemical tracers in the model, see Munday et al. [2014].

Mixed layer biological activity is shown in Figure 3a, with higher activity at high latitudes than at the equator and little biological activity in the oligotrophic subtropical regions. Such patterns are supported by satellite chlorophyll observations and other models [Williams and Follows, 2011]. The efficiency of the biological pump in the model (given by the ratio of mixed layer nutrient concentration over the nutrient concentration at 100–200 m depth), at high latitudes is 0.1–0.2, while in the tropics it is between 0.6 and 0.8. These numbers are consistent with observations [Sarmiento et al., 2004]. The spatial pattern, including the large-scale meridional gradients of pCO₂, shown in Figure 3b, is comparable to observations [Takahashi et al., 2009; Valsala and Maksyutov, 2010]. As a result, there is a positive air-sea flux of carbon north of the southern channel and in the northern subpolar region (40°–60°N) and a negative air-sea flux of carbon in the southern channel and tropics, again consistent with observations [Takahashi et al., 2009] (Figure 3c). Overall, the large-scale distribution and magnitude of the biogeochemical tracers agrees well with other studies [e.g., Ito and Follows, 2005] and observations [Feely et al., 2001; Gruber, 1998].

Figure 3

Steady state biogeochemical quantities in the control experiment: (a) Mixed layer average biological activity (mmol/m³/year of phosphate), (b) mixed layer ocean pCO₂, and (c) air-sea CO₂ flux. Dashed black lines indicate the extent of the channel.

3 The Response of the North Atlantic Carbon Sink to Increased Southern Ocean Wind Stress

3.1 The North Atlantic Carbon Flux

To examine the response of the carbon system to Southern Ocean wind stress perturbations, the wind stress between 60° and 20°S, τ_SO, is multiplied by a constant in a set of numerical experiments (see Figure 1b). Each perturbed experiment is run for 400 years, after which time the dynamics of the upper 1000m has fully equilibrated. We will examine the carbon cycle response to a range of wind stress perturbations, while specific examples will be drawn from the 2 × τ_SO scenario. The 2 × τ_SO experiment was run for 3000 years to verify that the results are not dependent on the length of integration. The response of the circulation to changes in Southern Ocean wind is linear. The Ekman transport and strength of the upper overturning cell increase linearly with wind stress. Our results therefore scale with forcing scenarios without nonlinear behavior.

Figure 4 illustrates the impact of ocean pCO₂, pCO , in the North Atlantic subpolar gyre on atmospheric pCO₂, pCO . Figure 4 shows time series of atmospheric pCO₂ (Figure 4a), the area-averaged ocean pCO₂ for the North Atlantic subpolar gyre (40°–60°N, Figure 4b), and the Southern Ocean (40°–60°S, Figure 4c) in the 2 × τ_SO wind stress perturbation experiment. The initial adjustment of the Southern Ocean pCO occurs within the first few decades. The atmospheric pCO₂ shows an initial increase over a few decades which coincides with the time scale of the local response of the Southern Ocean pCO . However, the short-term response is followed by a long-term centennial adjustment in pCO due to the perturbed ocean carbon reservoir in the North Atlantic. We refer to the atmospheric pCO anomalies induced by the North Atlantic subpolar gyre as the nonlocal effect, with a time scale set by the adjustment of the North Atlantic subpolar gyre pCO .

Figure 4

Time evolution of (a) Southern Ocean mean surface pCO , (b) North Atlantic subpolar gyre mean surface pCO , (c) atmospheric pCO , (d) North Atlantic mean surface Revelle buffer factor, and (e) North Atlantic and (f) Southern Ocean net air-sea flux of carbon, in the 2 × τ_SO experiment. The dashed lines in Figures 4e and 4f show the control mean air-sea flux. The overline indicates a spatial mean between 40°–60°N and 40°–60°S for the North Atlantic and Southern Oceans, respectively.

The increase in North Atlantic pCO shown in Figure 4b depends on the adjustment of the mean surface Revelle buffer factor R (Figure 4d) [Revelle and Suess, 1957]. R measures the fractional ocean pCO₂ change (denoted by δ) given a fractional change in DIC and vice versa such that

(1)

where Alk is the alkalinity. The buffer factor R quantifies the amount of carbon that can be stored in the form of DIC for a given change in atmospheric CO₂ concentration. High buffer factor systems are able to absorb changes in ocean pCO₂ by storing carbon in the form of DIC. For example, in a high buffer factor region which absorbs atmospheric carbon, local DIC does not change significantly. Therefore, more carbon remains in the form of dissolved CO₂, leading to higher pCO values and less uptake of atmospheric CO₂. The value of R can be approximated in terms of the ratio of alkalinity over DIC, as shown in equation 1. As alkalinity levels increase, the buffer factor decreases and the ocean becomes more capable of absorbing atmospheric CO₂ anomalies. Conversely, increasing DIC concentrations reduce the ocean's capacity to buffer atmospheric CO₂ changes.

Figure 4d shows that the temporal evolution of North Atlantic R corresponds to the change in North Atlantic pCO₂ shown in Figure 4b. Changes in the buffer factor account for 94% of the change in North Atlantic pCO₂, while other changes such as temperature and salinity account for the remainder (using the approach of Lovenduski et al. [2007]). As a result of the increase in North Atlantic pCO₂, the yearly air-sea flux of carbon in the region, , is reduced compared to that of the control run (Figure 4e). After roughly 425 years of the wind stress perturbation, the cumulative flux perturbation in the North Atlantic subpolar region becomes zero such that the North Atlantic subpolar region becomes a net emitter of carbon relative to the control state. The increase in Southern Ocean wind therefore induces a reduction in the North Atlantic carbon sink. The air-sea carbon flux in the Southern Ocean (Figure 4f) also shows a long-term decrease due to slow buildup of atmospheric CO₂ driven by the North Atlantic, since the Southern Ocean pCO₂ remains constant after 100 years of wind stress perturbation. After 400 years, the net change in the North Atlantic carbon flux is comparable in magnitude to the change in the Southern Ocean carbon flux.

3.2 Relative Feedback of the North Atlantic on Atmospheric CO₂

To quantify the strength of the nonlocal climate feedback on atmospheric pCO₂ due to Southern Ocean wind stress, we compare it with the change in atmospheric CO₂ due to Southern Ocean outgassing, i.e., the local feedback. We define the nonlocal feedback parameter ε which relates a change in Southern Ocean pCO , , to the subsequent change in North Atlantic pCO₂, . If the pCO in the Southern Ocean is altered through changes in the large-scale circulation and biogeochemistry, the perturbed ocean dynamical and biogeochemical state will also cause changes in the pCO₂ elsewhere, such as in the North Atlantic (as shown in Figure 4). The relative changes in pCO₂, , and will therefore be related to one another:

(2)

We can use ε to express the atmospheric pCO₂ anomaly, ΔpCO , due to a change in Southern Ocean pCO , by considering the air-sea flux of carbon F_air-sea over the global ocean. The air-sea flux equation takes the following form [Wanninkhof, 1992]:

(3)

where k is the gas transfer velocity. We integrate equation 3 over the global ocean. At equilibrium, the net air-sea flux is zero such that . However, we split the integral into three different regions: the Southern Ocean, the North Atlantic subpolar gyre, and the rest of the ocean basin, denoted by the subscripts SO, NA, and E, respectively. By considering a perturbation in atmospheric pCO₂, ΔpCO , we can then write

(4)

where P_SO, P_NA, and P_E are the area fractions of the respective regions, weighted by the local wind stress (see supporting information). The first term in equation 4 is the change in atmospheric pCO₂ due to changes in the ocean pCO₂. In the case of ε = 0, there would be no nonlocal effect and the change in atmospheric pCO₂ would be due to change in Southern Ocean pCO₂ alone. If ε is positive and takes a value of 1, the net feedback is positive and doubles the local Southern Ocean feedback on atmospheric pCO₂, since P_SO and P_NA are roughly of equal magnitude. Conversely, if ε is negative with a value of −1, there would be an opposing nonlocal feedback which cancels out the local effect of the Southern Ocean wind stress. The second, third, and fourth terms in equation 4 quantify the change in atmospheric pCO₂ due to the enhanced turbulent air-sea transfer of CO₂. However, these terms mostly cancel each other out and only account for 5–10% of the total ΔpCO .

The value of ε can be estimated using both equations 2 and 4, since all terms apart from ε can be calculated and diagnosed from the MITgcm simulations. The results of the calculations are shown in Figure 5. While there are differences in the results between using equations 2 and 4, the order of magnitude and trend of the calculated values of ε are similar. Values of ε are positive and range between 0.5 and 0.9 for weak wind stress perturbations. The magnitude of ε increases as the wind stress increases; ε is roughly 1.2 for a doubling of Southern Ocean wind stress. For all experiments, ε increases with time (not shown): from less than 0.4 in the first decade after the wind stress perturbation is initially applied to the quasi-equilibrium values of roughly 1 (as shown in Figure 5). The climate feedback of Southern Ocean wind stress on atmospheric pCO₂ is therefore almost doubled by the nonlocal effect. To understand the reason behind the change in the North Atlantic carbon sink, the following sections will describe the mechanism by which Southern Ocean winds influence the North Atlantic buffer factor, , and set the long-term centennial adjustment time scale of atmospheric pCO₂ anomalies.

Figure 5

The feedback parameter ε as a function of wind stress in MITgcm estimated using equations 2 (blue) and 4 (red). The black line shows the linear trend.

4 Southern Ocean Winds, Equatorial Biology, and the North Atlantic Buffer Factor

4.1 The Effect of Southern Ocean Winds on Nutrient Streams

In the Southern Ocean, surface biology is inefficient due to strong light and iron limitation [Boyd et al., 2000; Coale et al., 2004; DeVries et al., 2012] such that the upwelled nutrients remain mostly unused. Instead, the nutrients are transported northward and into the ocean interior following the subsurface currents in what are commonly referred to as nutrient streams [Pelegri and Csanady, 1991; Pelegri et al., 1996, 2006; Williams et al., 2006, 2011].

A nutrient stream is an advective pathway through which nutrients are transported from regions of upwelling and nutrient excess, such as the Southern Ocean, to nutrient-poor regions such as the equator and the northern high latitudes (see Figure 2d). The advective pathways mainly follow streamlines of the overturning circulation. Ekman transport drives nutrients northward and out the Southern Ocean [Williams and Follows, 1998], where it is subducted below the mixed layer during the formation of Antarctic Intermediate Water. Transport is then predominantly isopycnal, with nutrients gradually reinducted into the mixed layer as isopycnals shoal toward the surface in the North Atlantic subpolar region [Williams et al., 2006]. North of the Southern Ocean where biology is far more efficient, the nutrient stream causes an increase in macronutrient-limited biological productivity [Sarmiento et al., 2004]. The nutrient stream S_P(x,y,z,t) is defined here as the meridional transport of phosphorus, in the form of phosphate PO , by the meridional residual velocity field v and is given by

(5)

where the square brackets indicate concentration. The zonally averaged nutrient stream S_P is shown in Figure 2e for the MITgcm control run.

Physical transport in the ocean interior is mostly adiabatic along isopycnals, so the location of the nutrient stream lies along surfaces of constant density and its strength is modulated by the biological flux of nutrients [Pelegri et al., 1996]. In the current numerical setup, the maximum transport occurs along the potential density surface σ_θ=26 kg/m³ (σ₂₆), shown in Figure 2e for the control run. Therefore, we interpolate the meridional transport of phosphate onto σ₂₆. Figure 6 shows the meridional nutrient transport, S_P (Figure 6a), the steady state concentration of PO (Figure 6b) along σ₂₆, and the mean mixed layer biological productivity, F_bio (Figure 6c). The spatial pattern of S_P shows the (positive) nutrient stream originating just north of the southern channel where intermediate mode water is formed in region 1. The nutrient stream follows the northward MOC transport along the western boundary in region 2, strengthening until it reaches the northern subpolar gyre, where the nutrients are inducted back into seasonal thermocline in region 3.

Figure 6

The nutrient stream (a, d) S_p, (b, e) phosphorous on the σ_θ=26 kg/m³ surface, and (c, f) the mean mixed layer biological productivity. Figures 6a–6c are for the control run, and Figures 6d–6f are after 400 years of a 2 × τ_SO perturbation. Only fields north of the southern channel are shown for clarity.

The effect of increased Southern Ocean winds on the nutrient transport is shown in Figures 6d–6f after 400 years of the 2 × τ_SO perturbation experiment. Due to the increase in Southern Ocean wind-induced vertical mixing, there is a global increase in concentrations of PO along the perturbed σ_θ=26 kg/m³ surface ( ) in the Southern Hemisphere (Figure 6e). The wind stress perturbation also causes an increase in the meridional velocity v along . As a result, there is an increase in the northward nutrient stream (Figure 6d). The increase in surface nutrients leads to additional nutrients being introduced into the nutrient-limited equatorial shadow zone in region 4 via eastward equatorial jets and mixing across the southern shadow zone boundary. Subsequently, mixed layer biological productivity is strongly enhanced north of the southern channel and in the eastern tropics where nutrients are upwelled due to wind stress divergence. For the 2 × τ_SO experiment, the increase in biological productivity is less than 10% in the southern channel, between 20% and 40% in the subpolar gyre, but more than 200% between 20°S and 20°N. In the northern subpolar gyre, the response in biological productivity is not as strong as that in the tropics. This relatively weak response in the subpolar gyre is partly due to high-latitude light limitation [Nielsdottir et al., 2009] and partly due to the depletion of the nutrient stream by the time it reaches the northern end of the basin. Most of the increase in biology is therefore centered around the eastern tropics.

4.2 Surface Biology and the Revelle Buffer Factor

An increase in equatorial biology can reduce atmospheric pCO₂ by drawing more carbon from the mixed layer to form biological tissue, therefore reducing ocean surface pCO and increasing the air-sea flux of carbon into the ocean. Conversely, it can increase atmospheric pCO₂ by affecting the chemistry of subthermocline waters in the shadow zone. Surface biology utilizes DIC and alkalinity at different rates. To understand the effect of biology on the buffer factor, let us consider the influence of mixed layer biology on the mixed layer chemistry. The key points discussed below are also summarized in the diagram in Figure 7.

Figure 7

Diagram of the influence of surface biology on the Revelle buffer factor R. As mixed layer biological activity increases, mixed layer DIC is reduced and alkalinity is increased, lowering the buffer factor. Underneath the mixed layer to roughly 1000 m depth, more soft matter is remineralized compared to hard tissue, leading to an increase in DIC and a reduction of alkalinity, resulting in a strong increase in the buffer factor. Deep in the water column, remineralization of soft and hard tissue is comparable giving an increase in both DIC and alkalinity, resulting in a more moderate increase in the buffer factor.

Production of soft matter reduces DIC by 117mol for every mole of phosphate and increases alkalinity by 16mol, following Redfield stoichiometry [Redfield, 1934; Anderson and Sarmiento, 1994; Weber and Deutsch, 2012]. The formation or dissolution of 1mol of calcite (hard tissue), on the other hand, reduces DIC by 1mol and alkalinity by 2mol. Surface biology therefore changes the local concentrations of DIC and Alk depending on the relative export and remineralization of soft matter and calcite. The relative export of hard to soft matter, r, is known as the “rain ratio.” Studies of climatological basin-wide and local seasonal data give a rain ratio between 0.05 and 0.1 in the open ocean; the former is used in the MITgcm experiments conducted here [Sarmiento et al., 2002; Jin et al., 2006; Emerson et al., 2011]. For all values of r, an increase in surface biology results in a decrease of the relative ratio of DIC/Alk within the mixed layer.

Surface biology also affects the vertical transport of DIC and alkalinity in the ocean. The soft matter that is produced in the ocean mixed layer falls through the water column until it is remineralized back into DIC, in the form of C_soft. The remineralization of soft matter therefore increases DIC and reduces alkalinity. Similarly, calcite shells produced in the mixed layer fall through the water column until they are dissolved into DIC in the form of C_carb. Dissolution of calcite increases DIC and alkalinity. The remineralization depth of soft matter is shallower than hard tissue such that just below the mixed layer, remineralization of organic matter is mostly due to soft tissue which increases the relative ratio of DIC to Alk.

The relative concentration of DIC and alkalinity of water parcels, influenced by surface biology, is pivotal for the climate system since they determine the value of the Revelle buffer factor R and therefore the capacity of the ocean to absorb atmospheric CO₂ (equation 1). In regions of high biological activity, surface waters are reduced in R and the underlying waters are therefore enhanced in R. If these water parcels are then advected away and upwelled at the surface elsewhere, they will increase R in the region of upwelling. Water parcels with an elevated buffer factor R will be less capable of absorbing atmospheric CO₂ anomalies, leading to a local reduction in the ocean carbon sink.

As Southern Ocean winds increase and surface biological productivity is enhanced across the global ocean, the strongest increase in the buffer factor R occurs below the thermocline in the equatorial shadow zone along (Figure 8), where the largest relative changes in DIC and Alk occur. After 50 years (Figure 8a), there is an increase in R in the southern channel associated with increased upwelling of DIC-rich waters (region 1). A positive R anomaly occurs in the equatorial shadow zone due to the increased biological activity (region 2). After another 100 years (Figure 8b), the anomaly in R in the southern channel has decayed, while the tropical anomaly signal intensified. Because ventilation of the shadow zone is sluggish compared to the rest of the ocean, the increase in the downward flux of organic matter from the surface causes a slow buildup of DIC and a slow reduction of alkalinity in the shadow zone leading to a rise of R anomaly for centuries.

Figure 8

Revelle buffer factor anomaly (2 × τ_SO, control) along the perturbed σ_θ=26 kg/m³ surface after (a) 50, (b) 150, (c) 250, and (d) 400 years. The blue arrows in Figure 8d show the perturbed flow field along the surface. The black arrows show the primary advection pathway out of the equatorial shadow zone, while the dashed arrows show a secondary advection pathway due to recirculation within the shadow zone.

The increase in R generated in the eastern tropics is transported predominantly westward within the shadow zone through the equatorial current, where it is then entrained by the nutrient stream in the western boundary of the shadow zone. As shown in Figure 8d, an additional recirculation of the R anomaly within the shadow zone itself is present, leading to further advection toward the southern edge of the shadow zone. After the R anomaly is entrained into the nutrient stream, it follows the MOC transport northward (region 3) into the subtropical gyre where it is upwelled into the surface. The positive R anomaly in the intermediate waters just north of the Southern Ocean, denoted by region 4 in Figure 8c, is only 10% to 30% of the equatorial signal since the intermediate waters are ventilated at the surface prior to subduction.

4.3 Verification of the Mechanism

The effects of Southern Ocean mixing and nutrient export on low- and high-latitude biological productivity have been extensively studied [Anderson et al., 2002; Sarmiento et al., 2004; Marinov et al., 2006]. However, the advective pathways of biogeochemical tracers from the equatorial shadow zone to the North Atlantic and their impact on the subpolar gyre buffer factor are fairly novel. Therefore, we run an additional experiment in which we release a passive dye tracer into the shadow zone along where concentrations of oxygen are less than 0.17mol/m³ in the control simulation (the oxygen-minimum zone). Figure 9a shows the initial distribution of the tracer along , while Figures 9b and 9c show the dye concentration after 10 and 20 years, respectively. The tracer is being entrained near the western boundary and advected northward toward the subpolar gyre, where it is then mixed into the surface ocean. The advective time scale from the shadow zone to the North Atlantic subpolar gyre is roughly 15 years.

Figure 9

Passive dye tracer along the unperturbed σ_θ=26 kg/m³surface after (a) 0, (b) 10, and (c) 20 years in an unperturbed control simulation.

Figures 8 and 9 show that the positive R anomaly from the shadow zone is advected to the North Atlantic subpolar gyre. As discussed earlier, increases in surface biological activity in the subpolar gyre can reduce the local mixed layer buffer factor. However, by the time the enhanced nutrient stream reaches the subpolar gyre it is nearly stripped of the nutrient surplus so that the increase in biological activity in the North Atlantic subpolar gyre is not sufficient to reduce the positive R anomaly of the upwelled waters. The result is therefore an increase in the North Atlantic subpolar gyre buffer factor due to the increased Southern Ocean wind stress.

Following this argument, the net change in the North Atlantic buffer factor should depend on the relative change in the ratio of equatorial and North Atlantic biological productivity, such that

(6)

where the overbar refers to the spatial averages between 10°S and 10°N for the subscript “eq” and between 40° and 60°N for the subscript “na.” Since an increase in the buffer factor reduces the capacity of the North Atlantic to absorb atmospheric CO₂, we also expect that the atmospheric pCO₂ will scale with such that

(7)

Using the MITgcm numerical simulations in which the wind stress is perturbed, we will explore the validity of equations 6 and 7. We run further simulations to explore the sensitivity of the results using wind stress perturbation experiments with a warmed spin-up state, with increased ocean stratification. The model is warmed by increasing the temperature restoring profile shown in Figure 1c by 30% across the domain. The warmed model reaches a new equilibrium after 3000 years, at which time we introduce the same Southern Ocean wind stress perturbations described in section 2.2, up to a 3 × τ_SO perturbation.

Furthermore, experiments were also run whereby only the biogeochemical component in MITgcm is perturbed while keeping the wind stress profile constant. The parameters altered are the nutrient limitation half-saturation constant K_N [Monod, 1949] and the light limitation half-saturation constant I_k [Jassby and Platt, 1976]. The effect of perturbing the biogeochemical component is to increase or decrease biological activity by an amount which depends on the local light or nutrient limitation, without affecting the circulation so that the changes seen in and are purely biogeochemical. Increasing K_N causes areas that were not limited in macronutrients (e.g., high latitudes) to become limited, effectively increasing . On the other hand, an increasing I_k reduces light limitation more at high latitudes compared to the tropics, giving a reduction in . The nutrient and light half-saturation perturbation experiments are designed to show that an increase in atmospheric pCO₂ can also be caused by perturbations in the biology due to factors other than wind stress. The proposed feedback mechanism is therefore not solely dependent on Southern Ocean wind stress but can be caused by other climate change-induced factors that affect surface biological activity, such as freshwater or micronutrient input.

The results for all numerical experiments to test equations 6 and 7 are shown in Figure 10. Figure 10a shows that changes in induce a linear change in relative to the control state, , following the prediction of equation 6. The relationship is linear for perturbations less than 60% of the mean state. Furthermore, Figure 10b shows that ΔpCO varies linearly with as predicted, where Δ indicates again the anomaly relative to the control state. The sensitivity of the atmospheric pCO anomaly to changes in the North Atlantic buffer factor in Figure 10b depends on the control value of pCO and DIC concentration values, explaining the different slopes. The MITgcm experiments therefore show that an increase in tropical biological activity relative to northern subpolar gyre biological activity results in an increase in atmospheric pCO₂, under a variety of wind stress and biogeochemical forcing scenarios.

Figure 10

(a) Change in as a function of calculated from experiments with wind stress perturbations alone (black squares), wind stress alone in the 40° longitude model (magenta), wind stress and temperature perturbations (cyan squares), light half-saturation constant perturbations (green squares), and nutrient half-saturation constant perturbations (red squares). (b) Change in pCO as a function of (legend as in Figure 10a).

5 Conclusion and Discussion

5.1 Summary of Findings

The goal of this paper is to quantify the nonlocal equilibrium climate feedback of Southern Ocean wind stress on atmospheric pCO₂, pCO , using idealized coarse-resolution MITgcm experiments. In addition to triggering local changes in the Southern Ocean, an increase in the westerlies over the Southern Ocean can lead to changes in the biogeochemical and dynamical quantities in the Northern Hemisphere. Primarily, the North Atlantic carbon sink is found to be reduced several decades after an increase in Southern Ocean winds. The resulting feedback on atmospheric pCO₂, measured by the nonlocal feedback parameter ε, is of roughly the same order as the local Southern Ocean feedback. Overall, the nonlocal effect of Southern Ocean wind stress perturbations on atmospheric pCO₂ magnifies the local effect by 80% to 110%. The reduction in the North Atlantic carbon sink therefore constitutes a significant climate feedback that cannot be measured by considering changes in the Southern Ocean carbon sink alone.

Driving the reduction in the North Atlantic carbon sink is an increase in the Revelle buffer factor. It is found that wind stress-induced upwelling strengthens the export of nutrients north of the Southern Ocean by increasing the nutrient stream and increases equatorial biological productivity. The increased equatorial biological productivity increases the Revelle buffer factor, R, in the eastern tropical shadow zone/oxygen-minimum zone located between 100 and 700m depth. Via ocean interior advective pathways, the waters enriched in R are entrained into the nutrient stream and transported to the North Atlantic where they are inducted into the seasonal thermocline, increasing the subpolar gyre buffer factor and ocean pCO₂, pCO . Due to the reduction in the North Atlantic carbon sink, atmospheric pCO₂ is increased. The mechanism is also verified by only perturbing the biogeochemical component of the model.

5.2 Caveats and Implications

The use of a coarse-resolution model introduces several caveats. Coarse-resolution models fail to accurately represent the extent of the oxygen-minimum zones. Due to the unresolved complexity in equatorial undercurrents, a phenomenon coined “nutrient trapping” [Najjar, 1990; Aumont et al., 1999] results in overestimation of the strength of the oxygen minimum in the tropics. Nonetheless, the proposed mechanism should still be valid since oxygen-minimum zones are observed in the ocean [Stramma et al., 2009]. However, nutrient trapping could amplify the strength of the nonlocal feedback due to an exaggerated biogeochemical response in the tropics. Unresolved eddies in the Southern Ocean also overestimate the wind stress-induced increase in nutrient upwelling, since a constant Gent-McWilliams coefficient is used in the model. However, both Southern Ocean pCO and the relative response of the North Atlantic and equatorial biology depend equally on upwelling in the Southern Ocean so we do not expect this to affect the relative magnitude of the nonlocal feedback, although it may affect the equilibration time scale. For a full appreciation of the effect of eddies, a high-resolution eddy-permitting or eddy-resolving simulation is required.

Lastly, the geometry of the sector model introduces several caveats: the size of the basin plays an important role in setting the equilibration time scale of the nonlocal Southern Ocean winds feedback on pCO . The feedback equilibration time scale depends on basin-wide advection: the buffer factor anomaly in the shadow zone has to be advected zonally from east to west across a larger oxygen-minimum zone before it is advected northward by the nutrient steam. To illustrate this dependence of the time scale on basin width, an identical MITgcm setup 40° wide in longitude is used where the equilibration takes twice as long in the 40° longitude model (see supporting information). However, the strength of the nonlocal feedback ε is similar. The width of the basin therefore determines the time scale of the nonlocal feedback but not the magnitude. Further study is required to determine the exact equilibration time scale with a realistic geometry. The model also lacks a Pacific Ocean and Indian Ocean. However, the lack of MOC-like large-scale overturning transport and lack of strong ocean carbon uptake in the Pacific and Indian basins means that the presence of these basins will not affect the magnitude of the feedback of Southern Ocean winds on atmospheric CO₂ via the North Atlantic subpolar gyre. The increase in biological carbon drawdown in the Pacific and Indian basins as a result of increased nutrient transport is likely to be small due to the simultaneous input of surface carbon.

The relationship between the North Atlantic buffer factor and equatorial biology has further implications for the climate as well, since the ratio of North Atlantic to equatorial biology can be altered by many processes such as stratification, cloud cover, other sources of macronutrients, and iron supply, to name a few. Our results show the influence of oxygen-minimum zones on North Atlantic surface chemistry. Given recent results from studies that show expansion of tropical oxygen-minimum zones [Stramma et al., 2012], a thorough understanding of the biogeochemical drivers of the North Atlantic buffer factor is therefore needed. The North Atlantic subpolar region constitutes one of the main sinks for anthropogenic CO₂, and our results show that changes in the buffer factor of as small as 0.5 could result in a positive atmospheric carbon feedback of 10–30ppm.

Furthermore, Southern Ocean winds and vertical mixing have been used to partially explain glacial-interglacial cycles of atmospheric CO₂. Francois et al. [1997] and Anderson et al. [2009] suggest that there was an increase in vertical mixing in the Southern Ocean at the end of the Last Glacial Maximum (LGM), 20,000–12,000 years B.P., possibly related to an increase in wind stress. So far, however, current theories of physical mechanisms have not been able to fully explain the glacial-interglacial cycles of atmospheric CO₂ [Sigman and Boyle, 2000; Solomon, 2007]. Including the effects of the nonlocal atmospheric feedback of the North Atlantic due to Southern Ocean vertical mixing could aid our understanding of glacial-interglacial atmospheric CO₂ cycles.

Kohfeld et al. [2005] show that there was higher biological carbon export in the Atlantic equatorial shadow zone in the LGM but states that “these areas are relatively small and do not affect the overall picture,” while our research shows that these areas should not be overlooked. Although global changes in ocean alkalinity have been acknowledged as being potentially important, there has been no consideration of spatial patterns. Kohfeld et al. [2005] does not give exact figures of the relative change in biological export in the tropics and North Atlantic. However, Francois et al. [1997] suggest that Southern Ocean vertical mixing was increased by 20% to 50% at the end of the LGM, which, given our model study, would correspond to an increase in atmospheric pCO₂ of 15–35ppm in addition to the local Southern Ocean outgassing.

Acknowledgments

B.B. was supported by a NERC CASE studentship with the Met Office. Further support from NERC was provided to L.Z. and D.R.M. This work made use of the facilities of HECToR and Archer. We acknowledge the MITgcm team for making their code publicly available. We would also like to thank the reviewers for their helpful suggestions and comments. Details and results from the numerical simulations are available from B.B.