CHAPTER 9

Cold Climate and Deep Water Formation

In the preceding chapter, we investigated the transient response of climate to a gradual increase in the atmospheric concentration of CO₂. In this chapter, based upon the study conducted by Stouffer and Manabe (2003), we discuss the total equilibrium response of climate to large changes in the atmospheric CO₂ concentration, given a sufficiently long time for the climate to adjust. Although we discussed the equilibrium response to CO₂ doubling in the preceding chapter, that discussion was based on results from an atmosphere/mixed-layer-ocean model, in which heat exchange between the surface layer and the deep ocean is held fixed and does not change with time. Here, we explore the role of the deep ocean in the equilibrium response of climate using the coupled model, in which heat exchange between the surface layer and the deep ocean is incorporated explicitly. Four very long time integrations of the coupled model were performed as described below.

Starting from the realistic initial condition that was discussed in the previous chapter, the time integrations of the coupled model were performed over at least several thousand years, which is long enough for the temperature of deep water to stabilize. The control integration was run with the atmospheric concentration of CO₂ held fixed at the standard value of 300 ppmv. In the 2×C and 4×C integrations, the CO₂ concentration initially increased at a compounded rate of 1% year⁻¹ before being held fixed at twice and four times the standard value, respectively. The CO₂ concentration in the ½×C integration initially changed at a compounded rate of −1% year⁻¹, but then was held unchanged at one-half the standard value. The time-varying forcing for each of these four integrations is depicted in figure 9.1. The duration of the time integrations was more than 15,000 years for the control, 4000 years for the 2×C, and 5000 years each for the 4×C and ½×C.

FIGURE 9.1 Temporal variation of the atmospheric concentration of CO₂ (ppmv; logarithmic scale) as prescribed in the coupled atmosphere-ocean model simulations.

Toward the end of all four runs, the global mean ocean temperature at a depth of 3 km was barely changing, indicating that the deep ocean of the model was near a state of thermal equilibrium (figure 9.2). The temperature of deep water toward the end of the four runs was 6.5°C in 4×C, 4.5°C in 2×C, 1°C in 1×C, and −2°C (i.e., the freezing point of seawater at the oceanic surface) in ½×C. A noteworthy aspect of the ½×C run is that the temperature of deep water stabilized earlier than in the other runs, as dense, cold and saline water occupies the deep ocean. The analysis presented here is based on the mean states of the coupled model averaged over the last 100 years of each integration.

As discussed in chapter 1, the atmospheric greenhouse effect increases approximately in proportion to the logarithm of the CO₂ concentration of air. This implies that a doubling of CO₂ concentration from 150 to 300 ppmv exerts approximately the same thermal forcing as a doubling from 300 to 600 ppmv or from 600 to 1200 ppmv, even though the magnitudes of change in CO₂ concentration are quite different from one another. Table 9.1 indicates, however, that the difference in surface temperature between ½×C and 1×C is 7.8°C, and is much larger than the 4.4°C difference between 1×C and 2×C, which in turn is larger than the 3.5°C difference between 2×C and 4×C. In short, the equilibrium response of surface temperature to CO₂ doubling decreases with increasing surface temperature, mainly because the strength of the albedo feedback of snow and sea ice decreases as the climate warms, as discussed in chapter 5.

FIGURE 9.2 Temporal variation of global mean deep water temperature (°C) at a depth of 3 km.

TABLE 9.1  Global mean surface air temperatures at equilibrium in the coupled model
		Model time integration
	½×C	1×C	2×C	4×C
Global mean surface temperature (K)	276.4	284.2	288.6	292.1
Temperatures averaged over the last 100-year period of the four time integrations.

FIGURE 9.3 Latitudinal profiles of (a) zonal mean surface air temperature; (b) zonal mean surface air temperature deviation from the control (1×C).

The latitudinal profiles of the zonal mean surface air temperature obtained from the four time integrations are illustrated in figure 9.3a. For close inspection, the differences between the control and other integrations are shown at magnified scale in figure 9.3b. In general, the differences in temperature increase from low to high latitudes, where the albedo feedback of snow and sea ice predominates. Of particular interest is the large difference in surface air temperature between the control and ½×C in the Southern Ocean around 60° S. This exceptionally large cooling is attributable in no small part to the very extensive perennial sea ice in the Southern Ocean in ½×C. Next, we will describe how the ocean’s structure in the ½×C simulation is quite different from its structure in the simulations with higher CO₂ concentrations.

FIGURE 9.4 Latitude-depth distribution of zonal mean temperature (°C): (a) ½×C; (b) 1×C.

Figure 9.4a illustrates the latitude-depth profile of zonal mean temperature obtained from the ½×C simulation. Dense cold water fills the very thick deep layer of ocean, outcropping at the oceanic surface in the high latitudes of both hemispheres. The temperature of much of this layer is almost isothermal and is close to −2°C, the freezing point of seawater at the oceanic surface, and is substantially lower than the temperature of deep water below the depth of 3 km in the 1×C simulation, which is about +1.5°C. Comparing figure 9.4a with figure 9.4b, which illustrates the profile from the 1×C simulation, one notes that the layer of cold deep water is much thicker in the ½×C simulation. The temperature profile of the ½×C ocean is quite different from that of the 1×C ocean shown in figure 9.4b. For example, the thermocline is shallower and the cold bottom water is thicker in the ½×C ocean. Although it is not shown here, the salinity of deep water is high, particularly in high southern latitudes. The formation of the thick layer of cold and saline deep water described above is attributable mainly to the deep convection that predominates in the Southern Ocean, where sea ice forms rapidly in winter at the oceanic surface, yielding patches of saline and cold water through brine rejection. Although a similar process operates in the 1×C experiment, as described in chapter 8, the rate of deep water formation is substantially smaller.

The ½×C ocean is characterized not only by the thick layer of cold and saline deep water, but also by very extensive and thick perennial sea ice that extends to ~50° S in the Southern Ocean (figure 9.5). This is in contrast to the control, in which the coverage of sea ice undergoes large seasonal variation with little sea ice in summer (see figure 8.9). Albedo feedback plays an important role in maintaining this extensive sea ice cover, but an additional factor of importance is the upwelling of cold deep water driven by the intense westerly winds, which are much stronger than in the 1×C run. As this cold deep water upwells and reaches the surface, it freezes rapidly under the influence of the frigid overlying air. Brine rejection produces cold saline water that induces deep convective mixing. The combination of the upwelling of cold deep water and deep convective mixing prevents the poleward retreat of sea ice in spring and is responsible for the development of thick and extensive sea ice cover. In short, the Southern Ocean of the ½×C simulation may be characterized as a “gigantic sea-ice-producing machine” that also yields cold and saline deep water with near-freezing temperatures in the world’s oceans.

FIGURE 9.5 Geographic distribution of seasonal-mean thickness of sea ice (m) obtained from ½×C, (a) for June–July–August (JJA); (b) for December–January–February (DJF).

In the ½×C simulation, the atmospheric CO₂ concentration is half that of the 1×C simulation. This reduction in CO₂ is substantially larger than the reduction of the CO₂-equivalent greenhouse gas concentration at the LGM relative to its preindustrial value, as indicated by the analysis of air bubbles trapped in the Antarctic ice sheet (e.g., Neftel et al., 1982). Despite the larger greenhouse gas forcing in the ½×C simulation, it is likely that cold and saline deep water with near-freezing temperatures also occupied the deep ocean at the LGM, as indicated by the isotopic and chemical analysis of pore water conducted by Schrag et al. (2002) and Adkins et al. (2002). This suggests that a mechanism of deep water formation similar to that in the ½×C simulation may have operated at the LGM. According to the analysis of deep sea sediments conducted by Cooke and Hays (1982), sea ice covered the Southern Ocean at the LGM during much of the year, with the possible exception of summer (Crosta et al., 1998). One can therefore speculate that very extensive thick sea ice capped the Southern Ocean at the LGM, severely limiting the sea-to-air CO₂ flux in the primary region of deep water ventilation, as suggested by Stephens and Keeling (2000). Owing to sea ice capping at the oceanic surface and enhanced CO₂ solubility in the thick layer of near-freezing temperature, it is likely that deep water dissolved and sequestered huge amounts of carbon, thereby reducing the amount stored in the atmosphere. Meanwhile, the supply of nutrients to the upper layer of the Southern Ocean increased because of the intensification of upwelling of deep water, enhancing the biological production and drawdown of atmospheric CO₂, as suggested, for example, by Shackleton et al. (1983, 1992).

The state of the Southern Ocean obtained from the ½×C simulation resembles the state simulated by Shin et al. (2003) for the LGM, using a coupled atmosphere-ocean model developed by the National Center for Atmospheric Research. Common features of the two simulations include intense westerlies, extensive sea ice, and the formation of cold and saline deep water. The resemblance between these two states of the Southern Ocean suggests that the reduced concentration of atmospheric CO₂ has the dominant impact on the state of the LGM coupled atmosphere-ocean system in the Southern Hemisphere. This is in contrast to the Northern Hemisphere, where continental ice sheets had the dominant impact on the climate at the LGM, as discussed by Broccoli and Manabe (1987).

As noted in chapter 7, Broccoli (2000) made an attempt to simulate the condition of the oceanic surface at the LGM, using an improved version of the atmosphere/mixed-ocean model, in which the heat exchange between the mixed layer and subsurface layer of ocean was prescribed. Although the model simulated well the glacial-interglacial difference in SST reconstructed by CLIMAP at most latitudes, it underestimated the difference in the Southern Ocean poleward of ~40° S, as shown in figure 7.6. It is likely that this discrepancy is at least partially attributable to the absence of the combination of upwelling of cold deep water with near-freezing temperatures and deep convective mixing in the Southern Ocean in his LGM simulation.

In this chapter, we have described a slow but powerful feedback that is likely to play a very important role in the development of glacial climate. Operating in the Southern Ocean, it involves not only the albedo feedback of very extensive sea ice, but also other processes such as the upwelling of cold deep water, rapid freezing of seawater and brine rejection at the oceanic surface, deep convection, and the formation of cold and saline water in the deep ocean. The isotopic and chemical analyses of deep sea sediments suggest that a very thick layer of cold and saline deep water and very extensive sea ice also existed at the LGM, as noted above, and played a critically important role in maintaining the low atmospheric CO₂ concentration and the cold glacial climate.