On the role of inter-basin surface salinity contrasts in global ocean circulation



[1] The role of sea surface salinity (SSS) contrasts in maintaining vigorous global ocean thermohaline circulation (TOC) is revisited. Relative importance of different generalizations of sea surface conditions in climate studies is explored. Ocean-wide inter-basin SSS contrasts serve as the major controlling element in global TOC. These contrasts are shown to be at least as important as high-latitudinal freshwater impacts. It is also shown that intra-basin longitudinal distribution of sea surface salinity, as well as intra- and inter-basin longitudinal distribution of sea surface temperature, is not crucial to conveyor functionality if only inter-basin contrasts in sea surface salinity are retained. This is especially important for paleoclimate and future climate simulations.

1. Introduction

[2] Sea surface climatology is characterized by zonality, with the equator-to-poles sea surface temperature (SST), sea surface salinity (SSS), and sea surface density gradients the most prominent features of sea surface climatology (e.g., Levitus and Boyer [1994]; Levitus et al. [1994]). Together with the curl of wind stress, these latitudinal density contrasts are the major cause of global thermohaline circulation (TOC). Another, equally fascinating feature of sea surface climatology are the basin-wide contrasts in sea surface conditions in the Atlantic, Pacific, and Indian Oceans. The difference in sea surface conditions in these ocean basins appears to be fundamental to present-day TOC functionality [e.g., Gordon, 1986, 2001]. These contrasts can develop for a number of reasons [e.g., Warren, 1983], with evaporation minus precipitation (E-P) considered to be the major element. However, Manabe and Stouffer [1988, 1999] argue that the THC can also be the instrument for maintaining inter-basin SSS contrasts, enhancing the evaporation E-P impact. Since extensive review of the issue is given by Weaver et al. [1999], we do not discuss the reasons for these inter-basin contrasts. Our focus is on the role of these contrasts in TOC, rather than on their genesis.

[3] At present, three major water masses – the North Atlantic Deep Water (NADW), the Antarctic Deep Water (AABW), and the Circumpolar Deep Water (CDW) – govern the entire global deep-ocean current system [e.g., Gordon, 1986, 2001]. (Several intermediate water masses, of which the most important is the Antarctic Intermediate Water, may also contribute to structuring THC.) Importantly, deepwater is formed only in the North Atlantic (NA) and the Southern Ocean (SO) – none is formed in the North Pacific Ocean (NP). The deep-ocean current system, or thermohaline circulation, is also known as a “global ocean conveyor’ or “salinity conveyor belt” [Broecker, 1991; Broecker and Denton, 1989]. As the name suggests, salinity is the key in running the conveyor, with the NADW serving as an engine for the whole system. There are many papers on the subject (e.g., Bryan [1986]; Maier-Reimer et al. [1993]; Manabe and Stouffer [1988, 1997, 1999]; Marotzke and Willebrand [1991]; Rahmstorf [1995b]; Stocker et al. [1992, 2001]; Wang and Mysak [2000, 2001], just to name a few; for more references see, for example, Seidov et al. [2001]).

[4] Pushed by freshwater fluxes in high latitudes of the two hemispheres, the conveyor can behave as an ocean seesaw, or pendulum: Reduced or increased NADW formation is in counter phase with increased or reduced AABW [Broecker, 1998, 2000; Seidov and Maslin, 2001; Stocker, 1998]. Bond et al. [1997] found oscillations in surface ocean conditions within glacial cycles of the Pleistocene with a period of 1,500 years that also could be linked to freshwater pulses in the high-latitudinal NA. Alley et al. [2001] pointed out a noise-induced resonance in NA with a 1,500-year periodicity. The “noise”, ranging from tens to hundreds of years, could be envisioned primarily as meltwater functions.

[5] The role of asymmetry in the Atlantic-Pacific E-P in global TOC has been explored less thoroughly than high-latitudinal freshwater impacts. However, many studies have already addressed this issue (e.g., Cox [1989]; Maier-Reimer et al. [1993]; Marotzke and Willebrand [1991]; Mikolajewicz and Crowley [1997]; Mikolajewicz et al. [1993, 1997]; Murdock et al. [1997]; Rahmstorf [1995b]; Schiller et al. [1997]; Stocker and Wright [1991]; Weaver et al. [1999]; Weaver and Hughes [1994]; see also Weaver et al. [1999] for more references). The new feature of our work is its extension of understanding of the TOC's sensitivity to basin-scale variations in SSS, or to the freshwater balance needed to sustain these variations. Moreover, our work extends previously more idealized studies (e.g., Hughes and Weaver [1994]; Rahmstorf [1995a]; Stocker and Wright [1991]; Weaver and Hughes [1994]) by adding realistic geometry and a new experimental design.

2. Numerical Experiments

[6] In our research we used the GFDL Modular Ocean Model (MOM) version 2.2, with Gent-McWilliams isopycnal mixing [Gent and McWilliams, 1990] used as implemented in MOM-2 [1996]. A detailed discussion of the legitimacy of coarse resolution in first-order modeling of TOC can be found elsewhere [e.g., Seidov and Haupt 1999]. Here we only point out that many authors (e.g., Goodman [2001]; Manabe and Stouffer [1997]; Rahmstorf [1995a]; Toggweiler et al. [1989]) successfully used a 4° × 4° or equivalent resolution grid in sensitivity studies. In additional supporting runs, we compared the control case on 4° × 4° and 6° × 4° grids and did not find substantial differences. Additionally, Seidov and Haupt [1999] demonstrated that water transports, convection depths, and inter-basin water exchanges are simulated reasonably in experiments on the 6° × 4° grid. Therefore, we chose the 6° × 4° resolution at 12 levels as most economic yet still acceptable in a sensitivity study.

[7] Present-day annual mean sea surface climatology was employed in all experiments. This consisted of SST and SSS from Levitus and Boyer [1994]; Levitus et al. [1994] and wind stress from Hellerman and Rosenstein [1983]; see Seidov and Haupt [1999] for details. The world ocean model domain is bounded by Antarctica in the south and 80°N in the north. The Barents Shelf area is included; the eastern boundary in the North Atlantic Ocean sector is set at 40°E. For zonal averaging, the Atlantic, Indian, and Pacific Oceans are defined by their natural geographical boundaries. The northernmost boundary point for averaging within the Southern Ocean is the southern tip of Africa. The Mediterranean Sea and the Arctic Ocean are excluded.

[8] The numerical experiments described below are designed to explore the role of SSS basin-wide contrasts in maintaining TOC, and to ascertain whether SSS can be simplified (e.g., zonally averaged) without degrading the quality of modeled currents. The numerical runs described here are: (Exp. 1) control run with annual mean SST and SSS unchanged from the World Ocean Atlas [Levitus and Boyer, 1994; Levitus et al., 1994]; (Exp. 2) SST unchanged, but SSS constant and equal to 34.25 psu everywhere; (Exp. 3) SST unchanged, but SSS zonally averaged around the globe; (Exp. 4) SST unchanged, but SSS averaged zonally in each basin (i.e., basin-wide zonal averages retain inter-basin SSS contrast); (Exp. 5) SST and SSS averaged in each basin; and, finally, (Exp 6) SST globally zonally averaged, but SSS averaged in each basin (Table 1).

Table 1. Sea Surface Conditions in the Various Experiments
  1. a

    Exp. 1 is the “control run” with annually mean ocean surface climatology; SST and SSS are from Levitus and Boyer [1994]; Levitus et al. [1994], and wind stress TAU is from Hellerman and Rosenstein [1983]. {SST} means globally zonally averaged SST; the same notation is used for SSS. For [SST] and [SSS] zonal averaging is taken separately in each of four basins: In the Atlantic, Indian and Pacific Oceans between 34°S and the northern boundaries of these oceans (see text); and in the Southern Ocean between the Antarctica coastline and 34°S. Thus, [SST] means that [SST] replaces the 2-D observed SST in each of the four oceans. SSS34.25 means SSS equal 34.25 psu everywhere. The small crosses show what fields are used in each experiment.

2x     x
3x  x   
4x    x 
5    xx 
6  x  x 

[9] Global zonal averaging removes all longitudinal differences in sea surface climatology among ocean basins. However, latitudinal profiles of zonally averaged parameters preserve the main character of large-scale equator-to-pole sea surface variability. Basin-wide zonal averaging does an even better job of preserving latitudinal distributions within each basin (i.e., within four oceans: Atlantic, Indian, Pacific and Southern Ocean). The sea surface in such basin-wide averaging retains the ocean-to-ocean difference of sea surface parameters, although ocean-wide averaging removes longitudinal variations within each basin (with the exception of SO, where global and ocean-wide averaging are identical).

3. Results

[10] To save space, we show only meridional overturning stream function and oceanic heat transport between 30°S and the northern boundary in the Atlantic Ocean. Total meridional water overturning in the Atlantic Ocean is the most commonly used indicator of the behavior of the global conveyor. The reason is that the most important feature of the present-day ocean conveyor is a large NADW outflow from the North Atlantic in the deep ocean, which calls for an equally large compensating northward cross-equatorial flow in the upper layers, facilitating cross-equatorial oceanic heat transport. We begin with Exp. 1 and proceed according to Table 1.

[11] Figure 1a shows meridional overturning in the control case (Exp. 1, Table 1). About 15 Sv (1 Sv = 106 m3/s) of NADW is formed in the northern NA and the Nordic Seas, and approximately 4 Sv of AABW flows in the Atlantic Ocean below 3 km. Approximately 10 Sv outflows from the Atlantic Ocean between 1.5 km and 4 km, of which 4 Sv is returning AABW, and 6 Sv of NADW is leaving the Atlantic Ocean. This outflow requires 6 Sv of compensating northward flow in the upper layers (above 1.5 km). A recirculation flow of additional 4 Sv also occurs to the north of 30°S at intermediate levels, between 500 and 1 km. (Total southward cross-equatorial flow is close to 10 Sv, 6 Sv of which reaches 30°S and beyond, and 4 Sv of which returns to the NA with the compensating flow; see above.) The northward cross-equatorial water transport facilitates northward cross-equatorial oceanic heat transport – the signature of the present warm climate in the Northern Hemisphere in the Atlantic Sector (Figure 2, solid line). Basically, the results of the control case agree well with other coarse resolution ocean circulation simulations.

Figure 1.

Meridional overturning in the Atlantic Ocean in the experiments shown in Table 1; (a) control run with present-day annual mean sea surface climatology (see text); (b) Exp. 2 with SSS constant (SSS = 34.25 psu); (c) Exp. 3 with SSS zonally averaged over the globe; (d) Exp. 4 with SSS zonally averaged in each basin. Streamfunction is shown in Sv (1 Sv = 106 m3/s).

Figure 2.

Northward heat transport (in PW; 1 PW = 1015 W) in the Atlantic Ocean: (a) solid line shows Exp. 1 (see Table 1 and caption to Figure 1); (b) Exp. 2 (dashed-line); (c) Exp. 3 (dotted line); (d) Exp. 4 (dash-dotted line).

[12] Figure 1b depicts overturning in the experiment with global constant salinity (Exp. 2, Table 1). The 30°S outflow weakened substantially (to 2 Sv only). Although a sufficient amount of NADW is still formed (about 10 Sv), it is not dense enough to compete with AABW and CDW in order to maintain the global conveyor. Although there is still small NADW outflow at 30°S, and cross-equatorial NADW flow amounts only to 2 Sv, the cross-equatorial heat transport is noticeably weakened (Figure 2, dashed line). When SSS is zonally averaged in the entire world ocean (Exp. 3, Table 1), the conveyor collapses (Figure 1c). The explanation is simple: As discussed later (see next paragraph), zonal averaged SSS is lower in NA as it mediates between the NA and NP, thereby reducing high-latitudinal surface density and curtailing NADW production to less than 9 Sv (about half the control case). Thus, preserving latitudinal gradients of SSS within a basin (these gradients are removed by averaging across individual basins) is not essential in conveyor simulations. Moreover, a more idealized sea surface climatology with SSS held constant in Exp. 2 provides better results than a seemingly more realistic SSS field in Exp. 3. This is clearly seen in the ocean heat transport in Figure 2, line c (dotted line). However, Exp. 2 does not match Exp. 1 in either overturning or cross-equatorial heat transport (compare Figure 1a and 1c, and solid and dashed lines in Figure 2). Thus, although the results of Exp. 3 make an impression that SST can alone be responsible for most of THC features, a more careful analysis indicates that the absence of spatial variations in SSS leads to a serious deterioration of the results.

[13] The response of THC to freshening in the northern NA fits the ocean seesaw theory (see above). With SO water unchanged and NA surface freshening (as in Exp. 2), the seesaw theory would imply less NADW, which is exactly what happens. However, we emphasize that it is the middle latitude “freshening” rather then freshening of high-latitudinal NA that is critical in curtailing the conveyor. The global zonal averaging gives an average SSS of 34.15 psu at 60°N (against 34.45 psu south of Iceland at 60°N in the control case). Since there is very little NP water at 60°N, basin-wide averaging (Exp. 4) resulting in 34.18 psu in NA (32.7 psu in NP at 60°N). However, average freshening between 40°N and 60°N reduces SSS in NA by 1.1 psu, and between 40°N and 20°N by more than 1.4 psu. Such freshening is more than 50% of freshening of about 2 psu in the northern NA found in Manabe and Stouffer [1997], who simulated the Younger Dryas cooling episodes. Moreover, such freshening is also comparable with a meltwater scenario in Seidov and Maslin [1999], which also shows that meltwater must reach the northern NA to effectively shut off the conveyor. The bottom line here is that inter-basin salinity contrasts are at least as important as high-latitudinal freshwater impacts.

[14] The situation changes radically as soon as the inter-basin contrasts are preserved (Exp. 4, Table 1), even in a substantially simplified form. (Basin-wide zonal averaged values replace real SSS, removing intra-basin longitudinal gradients within each basin.) As 2-D SSS is replaced by zonally averaged distributions, the 3-D ocean circulation changes significantly. However, although dramatic change in total meridional overturning might be expected in all runs except for the control case, it only happens if the inter-basin SSS contrasts vanish (as in Experiments 2 and 3). Indeed, overturning in Exp. 4 that retains inter-basin contrasts (Figure 1d) is practically the same as in the control case, with all major parameters of the global conveyor, including cross-equatorial heat transport in the Atlantic Ocean (Figure 2, line d), matching those in the control case. Results for Experiments 5 and 6 (not shown) did not differ much from those for Exp. 4, leading us to conclude that inter-basin salinity contrasts are indeed the crucial factor in conveyor dynamics. Hence, although high-latitudinal freshwater regimes are important elements in ocean dynamics, inter-basin water vapor redistribution is imperative for global TOC. (E-P is the most significant element in the ocean-atmosphere hydrological cycle if freezing/melting is not involved – i.e., it is dominant everywhere except in polar and subpolar regions; river runoff, although very important locally and in the high latitudes where deep convection occurs, is somewhat less important than E-P in the hydrologic cycle in the middle latitudes.)

[15] It should be noted that some recent estimates suggest an oceanic fraction in total, atmospheric and oceanic, poleward oceanic heat transport is lower than in earlier studies (Trenberth and Caron [2001] and references in this publication). However, here we focus mainly on differences in simulated oceanic heat transports between the control case and the scenarios with different structures of SSS. Therefore, our main conclusions about the role of intra- and inter-basin SSS contrast do not depend on how realistically oceanic heat transport is modeled in the control case.

4. Conclusions

[16] Our main conclusion is that ocean-wide inter-basin sea surface salinity contrasts serve as the major controlling element in global TOC. Without this element, only intra-basin features of thermally and wind-induced gyres are well defined, with no true global deep-ocean conveyor operating in its present-day mode. Inter-basin salinity contrasts are at least as important as high-latitudinal freshwater impacts. On the other hand, thermal inter-basin contrasts, as well as longitudinal variation in SSS, are less important than latitudinal thermal gradients and inter-basin salinity contrasts. Moreover, details of sea surface salinity also decrease in importance as soon as inter-basin contrasts are retained. This is especially important for paleoclimate and future climate simulations, as the details of sea surface conditions may be somewhat neglected if only the inter-basin contrasts in the hydrological cycle are specified with confidence.


[17] We are grateful to two anonymous reviewers and AGU editorial staff for their useful comments that were very helpful for improving the manuscript. Discussions with Ron Stouffer were very useful and are very much appreciated. This study was supported in part by NSF (NSF projects #9975107 and ATM 00-00545) and the American Chemical Society (the ACS Petroleum Research Fund PRF #36812-AC8). Assistance from Lee Carpenter is very much appreciated.