The multi-year development of long-lived convective chimneys in the Greenland Sea



[1] The first convective chimney in the Greenland Sea to have its temperature and salinity structure fully mapped was discovered in March 2001 near 75°N 0°W [Wadhams et al., 2002]. Later cruises have shown that this remarkable feature has survived for a further 26 months, being remapped in summer 2001, winter 2002, summer 2002 and most recently in April–May 2003, making it the longest-lived chimney yet seen in the world ocean. The chimney has an anticyclonically rotating core and experiences an annual cycle in which it is uniform in properties from the surface to 2500 m in winter, but is capped by lower-density water in summer. The latest cruise also discovered a second chimney, 70 km NW of the first, during a thorough survey of 15,000 km2 of the gyre centre which left the existence of further chimneys unlikely. We conclude that the 75°/0° chimney is not unique, but that Greenland Sea chimneys are rare and are probably rarer than in 1997, when several such features were discovered by a float survey. This has implications for deep water renewal and for the role of Greenland Sea convection in the North Atlantic circulation.

1. Introduction

[2] In March–April 2001 a deep convective chimney was discovered and mapped in the Greenland Sea [Wadhams et al., 2002]. It was of diameter 20 km, with a uniform temperature and salinity profile from the surface to 2500 m, much deeper than the 1200 m which has been considered the typical depth of winter convection in recent years. In summer 2001 the recently-discovered chimney developed a structure in which a cap of low-density water covered the uppermost 50 m, with further intrusion of ambient waters down to 500 m, leaving an unaltered anticyclonically rotating core at greater depth. The cap disappeared in winter 2002 allowing the chimney to be once again uniform in structure and open to the atmosphere, regaining the same shape, and remaining in an almost identical location, as in 2001 [Wadhams et al., 2004]. An APEX float implanted in the chimney in March 2001 remained in the core (shown by T,S structure) until December 2001, making only local movements, and demonstrated conclusively the persistence of the same chimney between the winter and summer 2001 cruises, and almost certainly its persistence up to the winter (February) 2002 cruise [Wadhams et al., 2004]. Capping again occurred in summer 2002 when the rotation rate was measured for the first time by ADCP [Budéus et al., 2004], showing a uniform surface-to-bottom anticyclonic rotation rate out to a radius of 9 km.

[3] In April–May 2003 two further cruises took place. In the first, by FS “Polarstern”, the 75°/0° chimney was rediscovered close to its original location and remapped. In the second, by RV “Lance”, the same chimney was found to have moved 28.4 km to the northward (bearing 6°) in 27 days while retaining an identical T-S structure. At the same time, a systematic grid survey of the entire central gyre region revealed that only one other chimney existed, some 70 km to the NW of the 75°/0° chimney (Figure 1). In this paper we describe the results from these datasets.

Figure 1.

Station grid from “Lance” May 2003 cruise, showing bottom topography, three successive locations of the 75°/0° chimney (yellow, blue and red squares), the location of chimney 2 (purple square), and the trajectory of intermediate water as shown by a float released at 1000 m in 2002 (red arrow). Thick black line is 3000 m depth contour.

[4] The chimney at 75°/0° is now the longest-lived and most extensively mapped convective chimney so far discovered in the world ocean; the only chimney to have been observed during three successive winters. Its long-term survival raises questions about difficulty of formation, reasons for longevity, role in water mass transformation, and final fate. The discovery of a second chimney demonstrates that the original chimney is not a unique site-specific phenomenon, but is a member of a class of feature found in the central Greenland Sea gyre and similar to structures seen in the few other regions in which open ocean convection occurs [Marshall and Schott, 1999]. The fact that only two chimneys currently exist in the survey region (and thus only perhaps 2–4 in the entire Greenland Sea gyre) shows that chimneys are not common and must be difficult to create. The only previous direct observation of a chimney in the area was a survey in 1997 [Gascard et al., 2002].

2. The 2003 Data

[5] Figure 1 shows the station grid of “Lance” during May 19–30 2003, designed to cover the whole central gyre region that was susceptible to overturning. The grid spacing was 10 n ml (18.5 km), slightly less than the overall diameter of the 75°/0° chimney, making it unlikely that a chimney could exist within the grid and not be detected by some departure of a station from a “conventional” regional T-S-profile. The figure also shows the location of the 75°/0° chimney as discovered by “Polarstern” on April 27, when it was centred at 74°50.5′N, 00°03.5′W. “Lance” went to this location for her first station, but found no evidence of the chimney, nor in a wider search pattern of 21 stations. The chimney was rediscovered during the grid survey and its centre and structure defined by a series of more closely spaced stations. The new position of its centre, on May 24, was 75°11.0′N, 00°4.3′E. Since April 27 the feature had moved a net distance of 28 km on a bearing of 6° in 27 days, a speed of ∼1 km/day.

[6] Figure 2a shows the temperature and salinity structure from two transects across the chimney in NE-SW and NW-SE directions at 2.5 n ml. (4.6 km) station spacing, drawn on the same scale as Figure 2b which shows the structure of the “Polarstern” chimney. Clearly these are the same feature; the similarity in their T-S properties is evident at all depths, which is remarkable given the spatial displacement, the time delay, and the fact that the transects were not necessarily in the same directions relative to the chimney's axes. The central core has a potential temperature of −1.014°C and salinity of 34.882, which represents a slight but detectable warming from its core properties in previous seasons and years (see Figure 3 and Table 1), although the salinity remained essentially identical.

Figure 2.

Temperature-salinity (T,S) structure from two transects across 75°/0° chimney by (a) “Lance” May 2003, and (b) “Polarstern” April 2003. Left column: potential temperature; right column: Salinity.

Figure 3.

Core temperature and salinity of 75°/0° chimney, from surveys done from 2001 onwards. Sections show potential temperature structure at different times, a: winter 2001; b: summer 2001; c: winter 2002; d: summer 2002; and e: winter 2003.

Table 1. Core Properties of the 75°/0° Chimney at 1750 db Depth as Observed During Successive Cruises
CruisePot Temp (°C)SalinityPot. DensityPressure (db)
Lance 2001 winter Stn:47,10−1.04234.87628.0551750
Lance 2001 summer Stn:39−1.04134.88328.0601750
Lance 2002 winter Stn:31−1.03734.88028.0581750
Polarstern 2002 Summer (AR18)−1.01934.88128.0571755
Lance 2003 winter Stn:34−1.01434.88228.0581750

[7] In addition, both in April and May 2003 the chimney core was covered by a dome of warmer, less saline water. It is not clear whether this represents an early beginning of the “summer capping”, as observed in 2001 and 2002, or whether this indicates that the chimney did not reopen completely to the surface during the winter of 2002–3.

[8] On May 30 the chimney was found again. The centre position was now 75°13.4′N, 00°20.8′W, a displacement of 13 km in 6 days (∼2 km/day) on a bearing of 290°. If we compare the chimney's trajectory with that of a second APEX float deployed near the chimney in March 2002 at a depth of 1000 m we see (Figure 1) that during April–May 2003 the chimney followed the regional trajectory of intermediate water rather than staying in virtually a fixed location as it did from 2001 until this year. It is possible, therefore, that the chimney was in the process of advecting out of the central gyre region.

[9] The grid survey also revealed the presence of a second chimney (“chimney 2”) centred at 75°34.0′N, 01°47.9′W, a position shown on Figure 1. This chimney was also capped, and had a similar shape and cold core structure to the 75°/0° chimney (Figure 4). However, the core potential temperature was higher, at −0.96° C instead of −1.02°C, while the core salinity was similar at 34.885 and the potential density also similar at 28.058. This interesting result suggests that we can distinguish chimneys by their core characteristics and that these two chimneys were formed at different times or locations.

Figure 4.

Structure of second chimney, May 2003, compared with 75°/0° chimney. Top row, left to right: potential temperature, salinity and potential density of 75°/0° chimney. Bottom row, left to right: potential temperature, salinity and potential density of chimney 2.

[10] Away from the chimneys the water structure in the central gyre region in April 2003 was consistent over the whole grid area. A characteristic of the central Greenland Sea is a Tmax layer at about 1500–1800 m; this first appeared at a shallower depth in 1990 and has been moving gradually deeper at about 150 m a−1 [Budéus et al., 1998]. One effect of a chimney is to displace the Tmax layer beneath it to 2200–2700 m, and in so doing to displace all the lower water masses so that the bottom temperature under the chimney increases, creating a “warm shadow”. A sensitive bottom-mounted temperature sensor could be a useful long-term detector of chimneys.

[11] A second characteristic of a chimney is that the core temperature, at −1.02°C for the 75°/0° chimney and −0.96°C for chimney 2, is significantly colder than the minimum temperature reached at similar depths regionally, typically −0.85 to −0.90°C. This in itself is evidence that the chimney's origin involved cooling, through surface heat exchange. These two characteristics enable anomalies in a station due to the influence of a nearby chimney to be readily recognised, hence our confidence that, if there are more than two chimneys in the Greenland Sea, the additional features must lie outside the survey area.

3. Origin of the Chimneys

[12] The results indicate that chimney formation is difficult in the Greenland Sea but that it is not unique to one location. This raises the question of the origin of chimneys. Wadhams et al. [2002] speculated on flow through the Greenland Sea Fracture Zone (the ridge to the NE of the grid in Figure 1); however, T,S profiles from the GSFZ resembled the “conventional” profiles found inside the grid area and no candidate water masses were identified.

[13] An alternative, and at this stage tentative, hypothesis harks back to theories in which the role of sea ice formation is stressed. During most winters until 1997 the central gyre region became covered by a locally-formed regime of frazil-pancake ice known as the Odden [Toudal, 1999]. Salt flux models [e.g., Wilkinson and Wadhams, 2003] which consider both ice growth and advection predict that in an Odden year the sea surface near 75°/0° acquires from brine rejection a sufficiently increased density for overturning during winter. The Odden ice tongue has not reached the gyre centre since 1997, however, possibly due to the prevalence of a phase of the Arctic Oscillation which leads to warm easterly winds over the region in preference to cold NW wind outbreaks. Gascard et al. [2002] deployed several neutrally-buoyant floats in the central gyre region in 1997 and found that four of them became trapped in separate rotating current systems at 240–530 m water depth. If we identify each of these with a chimney – one float remained for an extended period near 75°N 0°W, while the others migrated cyclonically around the basin – then this suggests a population of several chimneys in the area in 1997.

[14] Models of chimneys [Killworth, 1979, 1983] suggest that a population of 6–12 chimneys forming and dissipating within a single year will produce enough ventilation to account for observed changes in the Greenland Sea Deep Water. If we adopt the hypothesis that ice formation is necessary for chimney formation, then in 1997 several chimneys could have formed which have since dissipated or moved away, so that by 2003 we have only two surviving chimneys. We already know that a chimney can survive 26 months, implying a long-term stability which may be maintained by the surface forcing during winter, so a survival time of 6 years is not impossible. In this respect the chimney resembles submesoscale coherent vortices (SCVs) found at mid-depths in the Mediterranean outflow [Armi et al., 1989], which can also be long-lived, although they migrate over longer distances and have a less stable structure.

[15] One must keep in mind that winter convection will mix the fresher surface waters, or cap, with the rest of the water within the chimney. This will result in an overall freshening of the chimney over time unless an additional source of salt through the advection of Atlantic based water or sea ice formation occurs.

4. Conclusions

[16] Models suggest that if the volume of convection in the Greenland declines, then there will be a cooling impact on NW European climate after a few decades [Rahmstorf and Ganopolski, 1999]. It is therefore important to know whether, and to what extent, chimneys enable active convection to reach great depths. The question of whether a chimney requires an ice cover for its creation, and of how its development is affected by annual fresh water capping originating from the melt of ice transported from the Arctic, also implies a link between sea ice retreat and a change in the nature and/or magnitude of Greenland Sea convection.


[17] We are grateful to the Commission of the European Communities for support of the 2001–2 cruises under the CONVECTION project (5th Framework, Environment and Climate Programme, contract EVK2-2000-00058); to Norsk Polarinstitutt, Tromsø, for supporting the 2003 cruise of “Lance” and to Alfred-Wegener-Institut for support for “Polarstern” in 2003.