4.1. Build up of inorganic carbon in deep waters
Isolating the deepest water in the world's ocean from the atmosphere may be expected to alter the natural concentration of atmospheric carbon dioxide for a number of linked reasons. In particular, the ‘biological pump’ tends to partition carbon towards the deep sea, the carbon being re-equilibrated with the atmosphere when the water is returned to the surface. Increasing the residence time of the water in the deep ocean, therefore, allows more time for biologically transported material to build up in the bottom waters, thereby removing it from the surface water and contact with the atmosphere. Furthermore, increasing the total carbon content of the deep water by remineralisation of organic material lowers its pH, causing the lysocline to become shallower, and some carbonate in deep sea sediments or in particles falling to the ocean floor to dissolve more readily. This increases the alkalinity of the ocean, which further reduces the pCO2 of surface waters. Given a step change in the deep ocean pH, due for example to a change in deep sea ventilation, this ‘carbonate compensation’ might be expected to continue until the rate of output of alkalinity to sediments matches the rate of input to the oceans as a whole—which is due to weathering products coming from the land via rivers, and to a first approximation might be considered constant. This implies that the lysocline will, on a time scale set by the sedimentary interaction, return to roughly the same depth that it was previously at, but now with a higher mean alkalinity. There is good evidence that the mean position of the lysocline did not change much between the LGM and the present day, though it shallowed somewhat in the Atlantic and deepened a little in the Pacific (Anderson and Archer, 2002). We know of no other source than the ocean for the CO2 that has increased in the atmosphere (and also in the terrestrial biosphere) since glacial time. If this carbon had been removed from the ocean without any change in alkalinity, it would have caused the lysocline to deepen. The fact that today the lysocline seems to be in the same position as it was at the LGM, therefore, suggests that carbonate compensation has gone essentially to completion since the LGM.
4.2. Modelling the effects of circulation change
Box models of the ocean–atmosphere carbon system were used by the Harvardton Bears papers, and continue to be used to explain glacial-to-interglacial atmospheric CO2 change (Toggweiler, 1999; Stephens and Keeling, 2000). Three-dimensional ocean carbon general circulation models (OCGCMs) have also been applied to this problem since the late 1990s (Broecker et al., 1999; Archer et al., 2000a; Bopp et al., 2003; Toggweiler et al., 2003a). It has been found that box models usually display substantially higher sensitivity to high-latitude processes than do the OCGCMs (Broecker et al., 1999). The high-latitude mechanisms put forward, for example by Toggweiler (1999), Watson et al. (2000) and Stephens and Keeling (2000), are relatively ineffective when simulated in such ocean carbon models. The reasons for this different behaviour have been investigated by Archer et al. (2000b) and Toggweiler et al. (2003a,b). There is not as yet a general agreement on the causes, however. For example, Archer et al. experimented with several different OCGCMs, finding that when a level model was modified to rotate the mixing tensor into alignment with isopycnal surfaces (Redi, 1982), thus removing one (but not the only) source of unphysical diapycnal mixing, this resulted in a more acute high-latitude sensitivity. However, an isopycnic co-ordinate model, which should have little unwanted cross-isopycnal diffusion, showed relatively weak high-latitude dependence, so these authors were unable to unambiguously ascribe the low sensitivity of OCGCMs to diapycnal diffusion alone.
It is clear that altering diapycnal mixing in a model may substantially affect the CO2 concentration of its atmosphere. This is important because, as discussed above, the distribution of diapycnal mixing in the real ocean is poorly known. Existing OCGCMs almost certainly do not reproduce it well and, furthermore, numerical diffusion effects may mean that the true diapycnal leakage in many ocean carbon models is not well understood. If diapycnal diffusion is zero, the potential density of a water parcel is fixed once it has left the surface (because conserved properties are invariant in a Lagrangian frame of reference when diffusion is zero, see for example, Gill (1982), while isopycnal diffusion cannot, by definition, transport density). In such an ocean, water parcels must then return to the surface with the same density as they left it. Since most deep ocean isopycnals outcrop only in the high-latitude areas, the interaction of the bulk of the ocean with the atmosphere will then be governed largely by conditions in those outcropping regions. However, in the presence of diapycnal mixing, water parcels return to the surface with a lighter density and a higher temperature than they left it, tending therefore to outgas CO2 on reaching the surface. Thus, increasing the diapycnal mixing in a model will tend to raise the predicted atmospheric CO2 concentration, and make it more difficult to partition the CO2 into the deep sea and away from the surface. These considerations suggest (in accord with Archer et al.'s results using level models) that a reduction of diapycnal mixing should cause a model to have a greater sensitivity to high-latitude processes. Since in box models the diapycnal mixing can be specified to be arbitrarily low, whereas in coarse-resolution OCGCMs the effective diffusivity may be raised by numerical effects and be poorly characterized, it is at least possible that the real ocean behaves more like a box model than such an OCGCM in this particular regard.
Toggweiler et al. ascribe the difference in behaviour of OCGCMs and box models to the smaller outcrop areas of polar waters in the former, and to unrealistic treatment of ice cover of these outcrops. In the current generation of OCGCMs being used to address this problem, production of bottom water is normally parameterized by deep convection, occurring in the open ocean at the grid scale (i.e. typically a degree or more in linear dimension). In the real Southern Ocean, however, bottom water formation is intimately associated with shelf and ice formation processes. Convection is localized to much smaller scales and reaches the bottom only over the continental shelf. Toggweiler et al.'s work suggests that, for the CO2 problem, the source of the densified water formed over the shelf and the degree of mixing with ambient waters during transfer to the deep sea are critical in determining the natural concentration of atmospheric CO2 at steady state with the ocean. Because such processes are not accurately parameterized in current OCGCMs, and are arbitrarily specified in box models, neither representation can be physically justified.
Thus, despite the obvious superiority of GCMs over box models when reproducing wind-driven horizontal circulations, it is by no means certain that this superiority carries over to simulations by OCGCMs of the effect of circulation or biology changes on atmospheric CO2. In the following we describe the use of a simple box model to demonstrate that, at least in that context, the mechanisms we have described have a substantial effect on atmospheric CO2. We use a box model because it is more transparent and easier to understand in comparison to readily available OCGCM-based carbon models, but we caution that existing OCGCMs would not show as much sensitivity to the Southern Ocean processes. Nevertheless the box model is not set up to be deliberately sensitive to high-latitude regions (the overall vertical exchange rate is not artificially low, for instance).
The model is similar in general type to that used by Toggweiler (1999) to show that changing ventilation rates in the deep sea does affect atmospheric CO2.Stephens and Keeling (2000) employed a similar model to investigate the effect of blocking gas exchange by dense ice cover in the Southern Ocean. The model is shown in Figs 5a–d, where the circulation and steady state modern and glacial solutions are displayed, together with other details in Table 2. The model has some variations from those of earlier authors to allow investigation of the particular processes we have been discussing. It has eight boxes to represent the global ocean, based on the major water mass types. The surface ocean is divided into Polar Antarctic, sub-Antarctic, northern North Atlantic and the rest (the ‘warm surface ocean’). The subsurface ocean is divided into AAIW, NADW and AABW. A separate reservoir is included to represent the coldest AABW in the deep Southern Ocean.
Figure 5. The reservoir-flux model of the ocean and atmosphere used to investigate the influence of changing vertical circulation on atmospheric CO2. Figures 5a and 5b label the boxes in the model for the ‘modern’ and ‘LGM’ circulations—see Table 2 for further details on the reservoir sizes. Figures 5c and 5d show the magnitudes in Sv (1 Sv = 106 m3 s−1) of the flows connecting each box for the two circulations, where a single-headed arrow indicates a flow in one direction, and a double-headed arrow indicates an exchange flux. Figures 5e and f show the values for the four variables in the modern and a LGM run (C = total carbon, P = phosphate, A = total alkalinity, all in μmol kg−1, and T = temperature in °C). The LGM configuration shown is that of column 5 in Table 3, before carbonate compensation is applied. Thus the inventories of phosphate, alkalinity and carbon are the same in Figs 5e and 5f.
Download figure to PowerPoint
Table 2. Parameters of the box model (see also Fig. 5)
|Volume of global ocean 1.35 × 1018 m3||Area of global ocean 3.5 × 1014 m2||Volume of the atmosphere 1.66 × 1020 moles|
| ||Southern Ocean||sub-Antarctic||Warm||North Atlantic|
| ||Fractional area||0.07||0.08||0.8||0.05|
| ||Fraction of export production from surface boxes remineralised in each sub-surface box|
| ||Fractional volume|| |
|Antarctic intermediate||0.2|| ||0.4||0.4||0.4|
|North Atlantic deep||0.29||0.1||0.4||0.4||0.4|
|Deep Polar Southern||0.1||0.9|| |
|Antarctic Bottom Water||0.27|| ||0.2||0.2||0.2|
Figures 5a and 5c show ‘modern’ and ‘LGM’ circulations as represented in the model. In the modern case, the Atlantic meridional overturning circulation is represented by a 20 Sv formation of NADW, of which 5 Sv upwells into the AAIW reservoir while the rest upwells in the Southern Ocean (Webb and Suginohara, 2001). A total of 20 Sv of southern component bottom water is formed (see e.g. Sloyan and Rintoul, 2001), of which half is quickly returned to the surface by the mixing processes discussed above, while the rest goes on to form the AABW found in the bottom of the major ocean basins. The deep Southern Ocean is mixed with NADW, so that the bottom water of the Southern Ocean is warmed and modified before spreading to the rest of the world ocean. A total of 13 Sv of AAIW is formed at the equatorward flank of the ACC (see e.g. Talley, 1999), of which 8 Sv is drawn from the Sub-Antarctic surface and 5 Sv from the Polar Southern Ocean. The net advective input to the AAIW reservoir is thus 18 Sv (including 5 Sv produced by upwelling of NADW) and this is assumed to upwell into the warm surface reservoir.
Vertical mixing between reservoirs is represented by exchange fluxes of water. In the main ocean basins, these approximately mimic what we know about vertical mixing there. For reservoirs 1000 m deep covering the temperate and tropical oceans (an area of ∼3 × 1014 m2 s−1) a vertical diffusivity of 10−4 m2 s−1 corresponds to an exchange flux of ∼10−4× 3 × 1014/1000 m3 s−1= 30 Sv. Rates of 40, 30, and 20 Sv were typically used for the exchange between, repectively, AABW–NADW, NADW–AAIW, and AAIW–surface, corresponding to vertical mixing that increases with depth. The relatively high value for the main thermocline is meant to reflect not only diapycnal exchange there (which alone, given diffusivities of ∼10−5 m2 s−1, would suggest a value <5 Sv) but also ventilation of the thermocline by wind-driven and eddy effects.
In the LGM circulation, NADW formation is turned off and the AABW and NADW reservoirs combined to give a single large AABW reservoir. Increased density contrast is assumed to slow the exchange between this water and the AAIW above it, and following our earlier discussion we set a value, which is one third of that at the same boundary in the modern circulation (10 Sv, corresponding to a diapycnal mixing rate of ∼3.3 × 10−5 m2 s−1). The return pathway for AABW to the surface of the Southern Ocean is turned off. The total rate of AABW formation is decreased slightly (to 16 Sv from 20 Sv), The temperature of this bottom-forming water is reduced to −1°C to reflect the more extreme conditions in the Polar Southern Ocean.
The model tracks temperature (T), phosphate (P), alkalinity (A) and total dissolved inorganic carbon (C), modelling biological productivity and particle remineralisation by means of fixed Redfield ratios C(organic):P:C(inorganic) of 106:1:25. Particulate fluxes generated in the surface reservoirs are remineralised in the underlying boxes in fixed proportions (see Table 2). Atmospheric CO2 is included as a reservoir, with pCO2 calculated using a constant gas transfer velocity over the entire ocean surface. When required, carbonate compensation can be crudely simulated by restoring the carbonate concentration of the AABW reservoir to the value in the modern ocean, by adding or subtracting to the total A and C inventories in the ratio 2: 1.
Table 3 collects the results of several experiments shifting between the modern day and glacial conditions. Column 1 gives the modern configuration. In column 2 we shift to the glacial circulation, keeping temperatures, inventories and particulate fluxes from the surface boxes constant. This results in a ∼35 ppm drawdown of atmospheric CO2. If carbonate compensation is included, this increases to 57 ppm. The assumption of a constant particle flux is of course entirely ad hoc. Column 3 shows the result of setting export production in both the sub-Antarctic and Polar oceans almost to zero. In this case atmospheric CO2 still decreases below the modern value, by some 27 ppm if carbonate compensation is included. Despite the cessation of biological activity, surface polar nutrient concentrations still decrease a little from modern values in this case. This is because the upwelling that brings nutrient-rich water to the surface is reduced, exposing the surface Polar ocean to eddy exchange with the near surface, relatively nutrient-poor water to the North.
Table 3. Steady-state box model results
| ||1. Modern conditions||2. Glacial circulation, temperatures and modern Antarctic export production||3. As 2, but no Antarctic production export||4. As 2, but proxy- Antarctic consistent export production||5. As 4, but glacial (compare Figs temperatures 5e and 5f)||6. As 5, but adjust for volume 3% ocean change and 500 Pg terrestrial biosphere|
|Atmospheric pCO2 (μatm)||280||245||269||232||208||249|
|Atmospheric pCO2 after carbonate compensation (μatm)||280||223||253||211||187||210|
|Polar Antarctic surface PO4 (μM)||1.9||1.07||1.53||1.15||1.15||1.15|
|Sub-Antarctic Surface PO4 (μM)||1.1||0.37||0.76||0.30||0.30||0.30|
|Polar Antarctic export production (mol C m−2 yr−1)||1.20||1.29||0.05||0.56||0.56||0.56|
|Sub-Antarctic export production (mol C m−2 yr−1)||1.87||1.86||0.03||2.68||2.68||2.68|
In column 4, we apply export productivities to the sub-Antarctic and Polar Southern Ocean that are consistent with proxy data—polar export production is substantially reduced from the modern value but sub-Antarctic productivity is increased. The increase in the sub-Antarctic region can more than compensate for the decrease in the polar region in terms of the effect on atmospheric pCO2. Column 5 reveals the effect of then imposing LGM surface temperatures (a further reduction of ∼25 ppm). Finally, in column 6 we account for two processes expected to increase atmospheric CO2 in glacial time. These are the decrease in volume of the oceans due to build up of ice on land, which raises the carbon and phosphate concentrations and alkalinity, and the transfer of around 500 Pg of carbon from the oceans to the land as the terrestrial biosphere grew in biomass between the LGM and the present day. Final atmospheric concentrations are 70 ppm below the modern value in the case where carbonate compensation is allowed.