The current paradigm of modern climatology and oceanography is that the thermohaline ocean circulation comprises a so-called global ocean “salinity conveyor belt” – a system of currents connecting the northern North Atlantic and northern North Pacific. A hypothesis is put forth here that a slight disparity in freshwater redistribution between the oceans in the Northern Hemisphere is sufficient to build up and maintain a global conveyor-type ocean thermohaline circulation.
 Although it was preceded by a number of studies of world ocean circulation as a global entity [e.g., Broecker and Denton, 1989; Cox, 1989; Gordon, 1986; Manabe and Stouffer, 1988, 1995; Rahmstorf, 1995; Stommel, 1958; Stommel and Arons, 1960], there are very few scientific metaphors that took over the research in earth sciences as strikingly as the idea of a global ocean “salinity conveyor belt”. The original drawing and many redrawings of the original art by Broecker  have found their way in numerous publications [see, e.g., Bigg, 1996], including popular books and magazines. The idea of the global “conveyor” is illustrated in the drawings as a loop of currents connecting the two most distant parts of the world ocean – the northern North Atlantic (NA) and northern North Pacific (NP). By means of simple computer experiments, we show that the NA-NP sea surface salinity (SSS) contrast that manifests disparity between the freshwater supplies to the two oceans is the key for the conveyor as a global entity.
 However, there are several yet unanswered questions. For example, what should be regarded as a “rather small” and “substantial alteration”? If the conveyor metaphor is essentially to bring into focus the identity of the THC running in a conveyor-type mode, then the question is how much, or how little additional rainfall or evaporation do we need to wipe out this identity? In other words, what is the fundamental mechanism governing the generation of the global conveyor belt?
 In many nonlinear physical systems, and the THC is definitely one of them, there is a threshold in critical parameters beyond which a complete change of a structured pattern occurs. What is the minimal local modification of this exchange that can cause global structural changes? In a way, we seek here a “minimalist's conveyor,” that is, the global conveyor emerging in response to small structural changes in very generalized surface salinity distributions.
 Our goal here is to determine the most important structural elements of the evaporation-minus-precipitation (E-P) disparity that cause the conveyor between different parts of the world ocean. (We simplify the entire hydrological cycle to E-P because only these two components are responsible for the large-scale basin-wide SSS distributions.)
2. Numerical Experiments
 To address this problem, we use the Modular Ocean Model (MOM) version 2 [Pacanowski, 1996] with relatively coarse resolution of 6° × 4° with 12 layers. The physics of the model and model parameters are all the same as in our recent work [e.g., Seidov et al., 2001; Seidov and Haupt, 2003]. There are five experiments that are summarized in Table 1 and Figure 1 (the setup here is consistent with the experimental design by Seidov and Haupt ). All runs start from an initially homogeneous ocean state with a temperature of 4°C and a salinity of 34.25 psu (practical salinity unit or pss - practical salinity scale) everywhere and continue for 10,000 years of model time. In all experiments, the temperature of the upper ocean layer relaxes to the annual mean observed SST [Levitus et al., 1994]. The annual mean wind stress is specified from Hellerman and Rosenstein . Thus, the only difference in the idealized experimental setups of Exps. 2–5 in Table 1 is in the freshwater fluxes across the sea surface.
Table 1. Freshwater Fluxes in Different Ocean Basins
Removed/added water is redistributed/removed over/from the entire sea surface to conserve water balance in the world ocean. Minus means removal of freshwater (more saltier surface water), and plus means added freshwater (fresher surface water). The rates are in Sv. NA- the North Atlantic Ocean; NNA- the northern NA; NP – the North Pacific Ocean; NNP – the northern NP; GIN – the Greenland-Iceland-Norwegian Seas.
control experiment (see text)
homogeneous salinity everywhere (34.25; see text)
 Because of the space limits, we only show two sets of figures. The SSS at the end of each run is shown in Figure 2 (except for Exp. 2, which unsurprisingly yielded SSS = 34.25 psu everywhere), and the meridional overturning in the Atlantic and Pacific Oceans in Figure 3.
 In Exp. 2 the freshwater fluxes across the sea surface were set to zero everywhere (Exp. 2 in Table 1). In this experiment, the salinity cannot change and remains constant (34.25 psu) everywhere. The THC in Exp. 2 is entirely different from Exp. 1 and is in complete contradiction to observations. The NA and NP Oceans are not connected by a deep ocean THC, and deepwater production occurs in both oceans – a complete contradiction to the known fact that no deepwater is produced in the NP Ocean. This result taken out of context is trivial – it is well known that salinity is responsible for the global THC as observed in the ocean. However, here we use this extreme case as the starting point to find out at what point the THC converts to a more familiar pattern of the global conveyor.
 Experiments 1 and 3 through 5 are illustrated in Figure 2. The SSS distribution built by the combined forces of freshwater fluxes and advection by ocean currents is shown for the three experiments, while the control run is restored to the annual mean observed SSS. Figure 3 shows the Atlantic (left) and Pacific (right) overturning in all 5 experiments (including Exp. 2, which is the run with “zero” hydrological cycle).
 Exp. 3 in Table 1 is the first in the series and the simplest possible inter-ocean freshwater redistribution that is, removing of freshwater in the central NA (in the area shown in Figure 1) and re-depositing it evenly over the entire world ocean. In this setup, the removal of freshwater creates a positive SSS anomaly in the central NA (in Table 1 we denote freshwater removal as negative and freshwater addition as positive values). The fluxes are constant within the areas shown in Figure 1, and very small residual fluxes in Exps. 3–5 are also constant. These residual fluxes are applied to the rest of the ocean surface for zeroing global freshwater fluxes over the world ocean.
 In Exp. 3 and the following two runs, the freshwater fluxes are small (varying from 0.035 Sv (1 Sv = 106 m3s−1) in the central NA to even smaller numbers in other parts of the world ocean). Although they follow observed or modeled patterns [e.g., Gent et al., 1998; Tziperman and Bryan, 1993; Wijffels, 2001], these fluxes have very generalized spatial distributions (see Figure 1), without the fine details of observed evaporation and precipitation patterns. The THC structure we were looking for – deepwater formation in the northern NA (NNA) and upwelling in the northern NP (NNP) – did not emerge in this run and, thus, a global conveyor did not materialize. Therefore, we continued to tune the most basic E-P pattern of Exp. 3 by adding new elements in Exps. 4 and 5.
 Some additional evaporation was assigned to the central NP in Exp. 4, with part of the evaporated water precipitating in the NNA, while the rest of the freshwater precipitating evenly around the world ocean (i.e., the freshwater flux is globally balanced and designed just to mimic the large-scale SSS distribution in the NA (Exp. 4), with a secondary features of SSS pattern added in the NNP in Exp. 5). The basic freshwater flux of Exp. 4 still maintains the main Atlantic-Pacific SSS contrast achieved in Exp. 3. Yet, it introduces a new element of meridional contrast in the NA that mimics freshwater redistribution within the NA to keep the Nordic Seas and some portion of the NNA substantially fresher than the subtropical waters.
 The overturning cell in the northern Pacific is greatly reduced in Exp.4, with the northward incursion of Antarctic Bottom Water and Antarctic Intermediate Water noticeably increased. Still, the intensity and the structure of the THC do not match the observed conveyor close enough. Obviously, the overturning in the Atlantic Ocean is too strong and there is still some deepwater formed in the NP, although the whole picture is different from Exp. 3. The last in the series, Exp. 5 is the most successful in depicting the conveyor. Importantly, the improvement of Exp. 5 over Exp. 4 is due to a very small freshwater outflow added to the “zeroed” freshwater fluxes over the subtropical NP Ocean (Table 1), with part of this freshwater removal precipitating in the NNP and the residual distributed over the world ocean (as described above for Exp. 3). However, this additional small freshening in the NNP in Exp. 5 leads to substantial improvement in Pacific overturning and, eventually, to the emerging of the fully developed global conveyor very similar to the one observed today and seen in the control simulations.
 Thus, a small structural element (here, an additional meridional variability of the freshwater balance in the Pacific Ocean) appeared to be critical to the fate of the THC. It is therefore tempting to consider the combined effects of salinizing of the central NA and freshening of the NNP, rather than the formation of North Atlantic Deep Water (NADW) alone, as the crucial elements for the THC dynamics observed today.
 There are two possibilities for the NNP to connect to the abyss – either through upwelling of the NADW (modified in the Southern Ocean) or deepwater production, thus cutting off NADW northward incursion. One might hypothesize that the redistribution of freshwater within the NP could be sufficient for suppressing deepwater production and thus giving way for the fully functional global conveyor, which is also a “salt machine”. Indeed, the NA-NP SSS contrast in the case of freshwater disparity between the two oceans would grow unrestrictedly if the build-up of salt concentration in the central NA would not be mediated by the THC. Basically, the Atlantic-Pacific freshwater disparity speeds up the conveyor, whereas the subtropical-polar disparity puts the brakes on the overturn (specifically, the conveyor in the Atlantic Ocean is too fast in Exp. 3, and slows down in Exps. 4 and 5).
 The following important conclusion can be drawn: the very nature of the global salinity conveyor depends critically on the salinity difference between the Atlantic and Pacific Oceans. If there is no sufficient SSS contrast of this nature, the global conveyor cannot develop. Thus, freshening of the Atlantic Ocean found in new analyses of salinity observations can potentially be a factor in the overall global conveyor functionality [e.g., Dickson et al., 2002].
 A warning should be issued here regarding breaking apart temperature and salinity in our idealized experiments. Although T and S are interrelated, keeping the SST unchanged while the initial SSS remains constant, and varying the air-sea water exchange is a simplification with the intended goal of isolating the fundamental role of building up inter-basin salinity contrasts. Moreover, observations indicate that temperature and salinity are at least partially density compensating in the Atlantic Ocean [e.g., Antonov et al., 2004; Levitus, 1989]. However, the inter-basin SSS contrasts between NA and NP co-exist with less pronounced contrasts in SST, so we may hope that our conclusions are still valid. A more sophisticated simulation using a coupled ocean-atmosphere model may shed more light on this issue.
 Moreover, AABW, in contrast to NADW, is largely thermally driven, as its dominance over the deep ocean was secured without any additional freshwater redistribution over the Southern Ocean. Moreover, the freshwater fluxes in the Southern Hemisphere are not as critical to the fate of the global conveyor as they are in the Northern Hemisphere (see above).
 A hypothesis can thus be put forward that the whole complexity of ocean-atmosphere-ice interactions needs to yield an astonishingly simple main pattern of water cycles responsible for the global salinity conveyor belt. The simplicity of this pattern may contribute to the stability of the observed climate, as minor details seem unimportant for maintaining the global conveyor as the key element of the climate system in the long run. In our simple experimental ocean-only approach we cannot prove this hypothesis; however, we hope that a fully coupled climate models can prove its viability.
 The authors are grateful to Igor Yashayaev and an anonymous reviewer for their very helpful and instructive comments. This study was supported by NSF (NSF projects #9975107 and ATM 00-00545). Acknowledgment is also made to the donors of the American Chemical Society Petroleum Research Fund for partial support of this research (ACS Petroleum Research Fund PRF #36812-AC8).