Over the past decade, the multiyear oceanographic time series from ocean weather station Mike at 66°N, 2°E indicate a warming by about 0.01°C yr−1 in the deep water of the Norwegian Sea. The time of onset of this warming is depth dependent, starting at 2000-m depth in 1987 but not at the 1200-m level until 1990. The warming abruptly halts around 1993 for a couple of years before it culminates in the absolute maximum temperatures in the end of the 50-yr-long record. This warming is attributed to variations in the amount and direction of interchanges between the three deep basins of the Nordic seas: the Arctic Ocean, the Greenland Sea, and the Norwegian Sea. The reduction of deep convection in the Greenland Sea from the early 1980s and the increased horizontal exchange with the relatively warm deep waters of the Arctic Ocean are the proximate cause of the warming, leading to a constant rise in the temperature of the “parent” Greenland Sea deep water (GSDW) from the early 1980s. These changing GSDW characteristics were passed on to the deep Norwegian Sea via the Jan Mayen Channel through Mohn Ridge, entering at a depth determined by the sill depth of this passage (2200 m) and propagating to shallower depths thereafter. The cessation of deep warming in the Norwegian Sea from 1993 to 1995—not shown by the GSDW—is attributed to the reversal of flow in the Jan Mayen Channel as GSDW production was greatly reduced, as confirmed by direct current measurements. The importance of the Arctic Ocean–Nordic seas system to global climate emphasizes the importance of identifying and understanding the mechanisms that control the interbasin dynamics of the Nordic seas and simulating them realistically in models.
Although they remain both sparse and sporadic, our longest hydrographic time series from the North Atlantic confirm the existence of natural variability at all space scales extending to those of the ocean basin and on timescales up to decades and centuries (Wunsch 1992). Even in the upper ocean these changes are not simple reflections of the atmospheric forcing but display at least two modes of ocean–atmosphere interaction (Kushnir 1994; Bjerknes 1962). While sea surface temperature (SST) variations on interannual timescales may display a coherent local relationship with the surface wind and atmospheric temperature, interdecadal changes in SST appear to be governed by basinwide dynamical interactions that take effect through changes in the large-scale ocean circulation.
In the deep-convection centers of the North Atlantic, where atmospheric forcing can lead to hydrographic change at subsurface depths, the ocean’s response is likely to be even more protracted. It has been demonstrated that the intensity of winter convective activity in the Labrador and Greenland Seas changes on decadal timescales due to slow changes in the forcing (Dickson et al. 1996; Lazier 1980; Meincke et al. 1997). There is evidence that these changes were correlated (in phase but of opposite sign) at these two main sites. As the products of this variable convection spread out slowly in the deep circulation, they carry the signal of decadal change throughout their respective basins, complicated by interactions with topography and with resident water bodies of different climatic history. Evidence of decadal change can therefore be expected to be present in many parts of the North Atlantic and over a considerable depth range.
The complex interactions between the decadal changes of the Arctic Ocean, Greenland Sea, and Norwegian Sea might be more important than most, however (Wunsch 1992). The Arctic Ocean and Greenland Sea are the most important sources of intermediate and deep water in the Northern Hemisphere, contributing significantly to North Atlantic deep water production and hence to the ventilation of the world ocean via the global thermohaline circulation (Kushnir 1994). The deep water of the Norwegian Sea (NSDW) is formed from a mixture of Greenland Sea deep water (GSDW) and Arctic Ocean deep water (AODW), carrying information on decadal change from both sources (Bjerknes 1962).
The main source of data for this paper are the hydrographic time series from Ocean Weather Station Mike (OWS M; 66°N, 2°E) in the Norwegian Sea (Gammelsrød et al. 1992). The OWS M operating above the eastern margin of the Norwegian deep basin (see Fig. 1) has been occupied since 1948. The hydrographic record from OWS M is the longest running homogeneous time series from the deep ocean. The hydrographic program at OWS M includes observations of temperature, salinity, and (since 1953) oxygen weekly at standard depth down to 2200 m. Water samples, taken with Nansen bottles equipped with protected reversing thermometers, are analyzed for dissolved oxygen content on board and salinity concentration on shore after each cruise, which usually lasts one month. The observation depths are corrected using unprotected thermometers. Two deep-sea mercury in glass thermometers on each Nansen bottle are used to measure the temperature. This method yields the water temperature to an accuracy better than ±0.02°C. As this method of measuring temperature has not changed significantly through the last 50 yr, the time series are indeed homogeneous.
The water samples are analysed for salinity ashore after the cruise. It has turned out that the salinity of the samples increases as the storage time increases. A data quality program is in progress where we try to correct for varying storage time. To date we have analyzed few cruises. For the data presented here the standard error is estimated to ±0.003 practical salinity units (psu).
Aanderaa recording current meters (RCM) models 5 and 8 are used to obtain the current records from the channel along the Jan Mayen fracture zone, here called the Jan Mayen Channel. This is the deepest passage (sill depth ∼2200 m) through the Mohn Ridge separating the deep Greenland Basin from the deep Norwegian Basin; see Fig. 1.
The Aanderaa RCM 5/8 equipped with a Savonius rotor yields the water speed to an accuracy of ±1 cm s−1, and direction accuracy of ±5°.
Smoothed monthly mean temperatures for the three deepest standard sampling depths at OWS M (1200, 1500, and 2000 m) are compared for the period 1948–97, in Fig. 2. We would like to point out three notable features. First, it is evident that a significant warming has occurred in NSDW over the last decade. Second, it is equally clear that its occurrence is depth dependent, starting at 2000 m in 1986, then gradually penetrating upward, to 1500 m in 1987, reaching the 1200-m level in 1990. The third significant feature of these records is the fact that the warming abruptly stopped for a couple of years from 1993. Overall, the increase in temperature is about 0.10 ± 0.02°C; it is nearly constant with depth;and it culminates in absolute maximum values for the entire 50-yr record at the two upper levels in 1996 and at 2000-m depth in 1997. For the rest of the period the time series are rather synchronous except for a warm period at the 1200- and 1500-m levels in the late 1970s, which does not show up in the 2000-m record.
a. Comparison with the Greenland Sea warming
Since the low temperature of the Norwegian Sea deep water is maintained by the deep horizontal incursion of cold GSDW, it is likely that the basic cause of this rapid warming of NSDW would lie in changes within the parent water body in the Greenland Sea. Figure 3 shows a recent update of the variation in potential temperature for waters deeper than 2000 m in the central Greenland Sea (Clarke et al. 1990), using all records of sufficient precision. As shown, a period of cooling from the late 1950s to the early 1970s is followed by a dramatic and sustained warming thereafter, continuing into the 1990s. The GSDW is itself renewed from two sources, either by horizontal exchange with the deep waters of the Arctic Ocean through Fram Strait (sill depth 2600 m) or by vertical exchange as a result of local deep-reaching open-ocean convection. Cooling of GSDW can only be carried out from above, by convection, while warming can only be effected from outside the basin of the Greenland Sea, by lateral exchange with AODW, the warmest and most saline deep water in the Arctic Ocean–Nordic seas system.
The cooling and warming cycle in Fig. 3 has therefore been interpreted (Meincke et al. 1997, 1992), as evidence of intensifying Greenland Sea convection with deepening vertical exchange through the 1960s to the early 1970s followed by a progressive capping of convection and increased horizontal exchange with the Arctic Ocean since then. This hypothesis seems amply confirmed by the changing tritium/3He ratio and oxygen concentrations of the GSDW [see Schlosser et al. (1991) and Bönisch and Schlosser (1995), who found that the rate of bottom water formation in the Greenland Sea reduced to about 10%–20% around 1980, compared to the situation in the 1970s]. The suppressed convection in recent years has been accompanied by the collapse of the classic “doomed” isopycnal structures across the deep Greenland Sea (Meincke et al. 1997), with a compensating influx of AODW as a result (see also Bönisch et al. 1997).
The causes of such radical changes are discussed by Dickson et al. (1996), who put this into the context of the North Atlantic oscillation. Here, the key point is that these events seem consistent with our observations from the deep water of the Norwegian Sea, already described. The almost complete cessation of deep-reaching convection in the Greenland Sea, the increased flux of warm AODW through Fram Strait, and latterly the collapse of the Greenland Sea dome (Meincke et al. 1997) all contributed to rapid warming of the GSDW by some 0.17°C between the early 1980s and 1995. This warming GSDW, with its increasing admixture of AODW, would then have passed through the Mohn Ridge (Fig. 1) to renew the deep water of the Norwegian Sea (Smethie et al. 1988; Heinze et al. 1990). Exchange of water masses between the two deep basins takes place through a channel, which has a threshold depth of 2200 m and is situated just north of the Jan Mayen Island. Current meter moorings deployed in this Jan Mayen Channel on several occasions for periods up to 9 months in the early 1980s (Sælen 1990) all show a steady flow from the Greenland Sea to the Norwegian Sea, with an average speed of about 7–8 cm s−1 (see also Swift and Koltermann 1988).
These deep, weak currents are in geostrophic balance, which means that they are steered by the bottom topography. Thus, during the 1980s the water flowing through the Jan Mayen Channel would follow the 2000-m isobath, circuit the western and southern margins of the Norwegian Sea basin (see arrows Fig. 1) to reach our monitoring point at OWS M with a delay of a few years, and with the warming reduced to about half its original amplitude (i.e., to ≈0.1°C) by mixing along its path.
The maximum warming below 2000-m depth in the Greenland Sea between 1982 and 1989 was centered around the 2000–2500-m level (Aagaard et al. 1991). Thus the existence of a sill at 2200-m depth in the Jan Mayen Channel may explain why the warming signal at OWS M is seen first at 2000 m. Although the warming at 2000 m in the Greenland Sea has been accompanied by a salinity increase, the density has been decreasing. Thus the buoyancy of the deep water entering the Norwegian Sea has increased, explaining why so little warming is found in the Norwegian Sea below this depth. Temperature profiles obtained across the southern Norwegian Sea in 1982 (Clarke et al. 1984) and 1994 (Østerhus et al. 1996) show that in the central basin the warming was much less (0.02°C) below 2200 m. The warming of the adiabatic layer below 2200 m is driven by diffusion from above and by geothermal heating from below (Østerhus et al. 1996).
The warming in the Greenland Sea in the intermediate water (200–2000 m) between 1980 and 1993 has been even greater than the warming in the deep water (Bönisch et al. 1997). Thus the warming seemingly penetrating upward at OWS M may be explained by horizontal advection from the Greenland Sea, with increasing mixing and decreasing advection velocity upward in the water column. At shallower depths there is also an increasing number of exchange locations between the two basins. The warming seen at 1500 and 1200 m at OWS M, but not at 2000 m in the last half of the 1970s (Fig. 2), may stem from a warming observed down to 1500 m in the Greenland Sea between 1973 and 1977 (Bönisch et al. 1997). Also at OWS M the temperature increase is accompanied by an increase in buoyancy. This is illustrated in Fig. 4, where we compare the density (σ2) profiles for three different years. The buoyancy increase at 2000 m corresponds to a lifting of the isopycnals to the 1650-m level in 1990 and to 1450 m in 1995 compared to 1985.
b. Reversal of deep basin exchange
Since the warming of GSDW appears to have continued to date (Fig. 3), the cessation of warming observed in the NSDW during 1993–95 is certainly unexpected (Fig. 2). This suggests that as GSDW production has (virtually) ceased, the transport through the Jan Mayen Channel may have reduced or even reversed, cutting off the deep Norwegian Sea from the influence of the GSDW and its changes. To check this point, a current meter mooring was deployed there between November 1992 and July 1993. The results are compared with all preexisting measurements from April–November 1981 (Sælen 1990) and September 1983–July 1984 in Fig. 5. Though too short to be conclusive as of yet, Fig. 5 indicates that the deep water flow between the Greenland Sea and the Norwegian Sea has indeed reversed, now exhibiting an average current of 0.8 cm s−1 into the Greenland Sea.
The reversal of the flow through the Jan Mayen Channel is related to the density distribution in the two basins. In Fig. 6 we have compared the time development of density just above the threshold level of the channel. The data representing the central Greenland Sea at 2000-m level are from Bönisch et al. (1997), and the Norwegian Sea is represented by OWS M. Figure 6 shows that the density at the 2000-m level was highest in the Greenland Sea in the 1980s. In the early 1990s the density in the two basins did not differ much, while in recent years the tendency is that the Norwegian Sea exhibits higher densities than the Greenland Sea. Comparing data from CTD stations taken in 1989 and 1990, Bourke et al. (1993) conclude that the flow through the Jan Mayen Channel had ceased that winter.
Inserted in Fig. 6 we have shown a temperature–salinity diagram comparing the development of the densities in the central parts of the two basins. The 1982 data are from the R/V Hudson cruise (Clarke et al. 1990), and the 1994 data are from the cruises with R/V Johan Hiort and R/V Håkon Mosby. The tendency shown here is the same, the density at 2000-m depth was higher in the Greenland Sea than in the Norwegian Sea in 1982, while the situation has reversed in 1994.
c. Relevance for the Scotland–Iceland–Greenland overflow
About 5–6 Sverdrups (106 m3 s−1, Sv) of deep waters that formed and transformed in Arctic Ocean and Nordic seas overflow the Scotland–Iceland–Greenland sills and contribute to the roughly 15 Sv of North Atlantic deep water (Dickson et al. 1990). However, only about 15% of this overflow water circulates through the deep basins north of the sills (Bönisch and Schlosser 1995). Due to the reduced supply from the Greenland Sea, there are indications that the upper level of NSDW north of the sills has become deeper (Turrell et al. 1999; Østerhus et al. 1996). This has resulted in a change in the composition of the water flowing through the deepest of the sills, the Faroe Shetland Channel. Here it is observed that the amount of NSDW contributing to the bottom water has reduced from 60% during the period 1960–80 to 40% after 1990 (Turrell et al. 1998). Fogelqvist et al. (1998) found that in 1994 the percentage of NSDW was around 30%. This has resulted in a decrease of the density of the overflowing water. As the density of this overflow water decreases, its ability to ventilate the abyss of the Atlantic water will diminish, and if this tendency continues it may in turn lead to changes in the deep North Atlantic circulation. Mauritzen (1996) has suggested a circulation scheme for the 1980s where the classic picture of deep water production in the gyres of the Greenland and Iceland Seas is unimportant for the overflow water. This may be true for the last two decades, but deep water production in these two gyres has significantly contributed to the overflow water in the 1960s, (Turrell et al. 1999).
Although global temperatures have been abnormally high in a succession of recent years, culminating (in 1997) in the highest mean temperatures since measurements were first properly recorded (in 1860), the warming in the deep water of Nordic seas is not obviously or directly linked to global warming. Instead, the multidecadal changes in the convective intensity of the Greenland Sea, from which these changes stem, appear to reflect a long-period shift in the North Atlantic oscillation (Dickson et al. 1996), although the causes of this correlation remain unexplained. However, the key importance of northern seas to the climate system as the projected site of an enhanced global warming signal and as a significant contributor of deep water to the global thermohaline circulation is undisputed. It is therefore important that we identify, understand, and attempt to simulate the causal mechanisms that control the variability in the interbasin exchanges in the Nordic seas with the use of regional, dynamical models.
Our observations indicate a reversal of the deep-basin exchanges in the Nordic seas. Though certainly rare, reversals of influence between the three ocean basins in question may not be unprecedented, even in our limited observational record. Earlier this century, at a time of active convection, the Greenland Sea was described by Wüst (1942) as directing deep water to both the Norwegian Sea and the Arctic Ocean, although his dataset was sparse. Further measurements to confirm the change of the flow in the Jan Mayen Channel are in progress.
The warming observed in the deep waters in the Nordic seas represents a substantial climate signal. In terms of equivalents, the warming at OWS M equates to a deepening of isotherms by 400 m at a 2000-m depth or 200 m at a 1200-m depth, to a 6-mm rise in sea level due to thermal expansion, or to a heat flux of about 1 W m−2, which is close to the estimated radiative forcing due to anthropogenic greenhouse gases.
This project has been supported by Norsk Forskningsråd and Nordic WOCE. Sincere thanks are due to Robert Dickson for his generous spending of time and helpful assistance in completing this paper. We are also grateful to the crew on board the weather ship Polarfront and to Johan Blindheim for offering ship’s time for mooring deployment. Two anonymous reviewers contributed with several constructive comments.