Notice: Wiley Online Library will be unavailable on Saturday 30th July 2016 from 08:00-11:00 BST / 03:00-06:00 EST / 15:00-18:00 SGT for essential maintenance. Apologies for the inconvenience.
 In the last three decades, deep convection has come to a halt in the Greenland Sea. Hydrographic data reveal that during this period, temperature and salinity in the deep Greenland Sea have increased at mean rates without precedent in the last 100 years, and these trends are among the highest in the global deep ocean. The origin of these changes is identified as the advection of Arctic Ocean deep waters and the necessary transports to explain them are calculated (0.44±0.09 Sv). Despite the fact that the deep Greenland Sea hardly covers 0.05% of the global surface, the resulting trends constitute 0.3% of the World Ocean heat content increase per unit area of earth's surface and 0.1% of the global sea level rise. These results suggest that changes of the deep Arctic Mediterranean and their contribution to the global budgets need to be addressed.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Deep waters (> 800 m) of the Arctic Mediterranean do not immediately contribute to the Atlantic thermohaline circulation because they are separated from the North Atlantic by the Greenland-Scotland Ridge (Figure 1b). Yet, their ventilation implies modification of the shallower overflow waters that cross the ridge at rates of 6 Sv [Hansen and Osterhus, 2000; Eldevik et al., 2009]. The ventilation is different in the Arctic Ocean where relatively warm and saline deep waters are produced and the Nordic seas where, particularly in the Greenland Sea, open ocean convection produces relatively cold and fresh deep waters (Figure 1a). However, during the last three decades, deep convection has been strongly reduced in the Greenland Sea [Schlosser et al., 1991; Rhein, 1991; Meincke et al., 1992]. As a consequence, two major hydrographic changes have been observed: (1) the appearance and deepening of an intermediate temperature maximum and (2) the continuous warming and salinity increase of the deep Greenland Sea [Bönisch et al., 1997; Meincke et al., 1997; Ronski and Budéus, 2005; Karstensen et al., 2005; Budéus and Ronski, 2009].
 Besides the possible effects of these changes on the renewal of North Atlantic deep waters and thermohaline circulation, changes in thermohaline properties that are not density compensated alter the local sea level which might induce circulation changes in [Osterhus and Gammelsrod, 1999] and beyond the Greenland Sea.
 Furthermore, the warming of the deep Greenland Sea contributes to heat storage in the deep ocean. Recent discussions about heat that is “missing” in the atmosphere and in the upper ocean [Trenberth and Fasullo, 2010; Meehl et al., 2011; Levitus et al., 2012], but possibly sequestered in the deep ocean, require that also Arctic Mediterranean deep waters are considered, which have been sparsely represented in global ocean heat content change estimations so far [Levitus et al., 2009].
 In this paper, we update time series of temperature and salinity from the deep Greenland Sea after 2004 [Budéus and Ronski, 2009], present an explanation for their changes, and assess their contribution to deep ocean heat content changes and steric sea level rise. In section 4, we present evidence for the continuous absence of deep convection and in section 5, use the change of the thermohaline properties of the deep Greenland Sea to estimate the necessary transports from the deep Arctic to explain the observed changes. Estimates of the deep Greenland Sea contribution to the increase in World Ocean heat content and global sea level rise are presented in section 6. In section 2, the study area and the data used are described.
2 The Greenland Sea in the Arctic Mediterranean
 The Greenland Sea is the northernmost deep basin (maximum depth: 3700 m) of the Nordic seas and the only one with a direct deep connection to the Arctic Ocean through Fram Strait (Figure 1b). Open ocean deep convection provides the coldest and freshest deep waters of the Arctic Mediterranean to the deep Greenland Sea, and this freshening and cooling is counteracted by the advection of warmer and saltier waters from the deep Arctic Ocean [Rudels, 1986; Rudels et al., 2012, and references therein] (Figure 1a). Two different warmer and saltier deep waters with origin in the Amerasian and Eurasian basins of the Arctic Ocean enter the Nordic seas through Fram Strait and are part of the boundary current circumnavigating the Greenland Sea (Figure 1b). Amerasian Basin Deep Water (ABDW) that crosses the Lomonosov Ridge and passes the Eurasian Basin is found in Fram Strait and at the continental slope of Greenland at a mean depth of 1500 m. Eurasian Basin Deep Water (EBDW) flows southward through Fram Strait at deeper levels and is found there at depths of about 2300 m [Aagaard et al., 1991; Langehaug and Falck, 2012] (Figure 1a).
 Our analyses are based on hydrographic data (salinity, temperature, and pressure) from 1950 to 2010 in the central Greenland Sea (between 74°N and 76°N, 0°W and 6°W) (see Figure 3d). All data used are freely available from the ICES Dataset on Ocean Hydrography and the Pangaea database (details in Table 1). Before 1993, the area was sampled only occasionally, but between 1993 and 2010, an oceanographic section at 75°N was sampled annually during Polarstern summer expeditions [Budéus and Ronski, 2009] (Figure 2c). Since 1994, an SBE 911+ system was used providing accuracies for temperature and salinity of 0.001°K and 0.003, respectively [see Budéus and Ronski, 2009].
Table 1. Mean Potential Temperature, θ, and Salinity, S, of Deep Waters in the Greenland Seaa
Based on observations in the Eurasian and Amerasian Basins of the Arctic Ocean between the density ranges indicated and in the shown areas and time periods (see data in Figure 1).
 For estimating the import of Arctic Ocean deep waters into the Greenland Sea, hydrographic data of the Arctic Ocean were also used (Table 1).
3 Warming and Salinity Increase of the Deep Greenland Sea
 We argue that the persistence of the temperature maximum and the warming and salinity increase at depth provide evidence for the strong reduction of deep convective mixing so that inflow from the deep Arctic Ocean will fill the deep layers successively with EBDW.
3.1 Evolution of the Intermediate Temperature Maximum (InTeMax)
 According to previous studies, a prominent change in the Greenland Sea was the appearance of an intermediate temperature maximum (hereinafter InTeMax) at ∼ 800 meters depth at the beginning of the 1990s that was constantly deepening afterwards [Ronski and Budéus, 2005; Karstensen et al., 2005; Budéus and Ronski, 2009]. The analysis of the hydrographic data in the central Greenland Sea reveals its occurrence intermittently also during the 1980s. In 1982, the InTeMax appears at 400 m, and in 1988 and 1989 at ∼ 600 m [Meincke et al., 1997]. After 1993, the InTeMax is observed as a permanent and ubiquitous feature in the central Greenland Sea (74°N and 76°N, 0°W and 6°W).
 Due to the scarcity of data before 1993, the temporal evolution of the InTeMax in the central Greenland Sea is shown from 1993 onwards in Figure 2. For clarity in the representation, only profiles between 3°W and 4.5°W of the section along 75°N are included. The resulting time series of temperature profiles highlights the persistence of the InTeMax and its progressive deepening from 1000 m in 1993 to 1800–2000 m between 2005 and 2009 (Figure 2a). Although after 2007, the signal as an intermediate temperature maximum is not present anymore, the InTeMax conserves its characteristic θ, S, and density (Figure 2). The dots delimiting the upper and lower limits of ABDW that enters the Greenland Sea (ABDWGS) (Figure 2a) show that the InTeMax has the same potential density than this water mass. This confirms the suggestion by Meincke et al.  and Rudels et al.  that the InTeMax has its origin in ABDW that enters the Greenland Sea and penetrates from the rim towards the center of the gyre.
 However, the immediate question arising from this assertion is as follows: Why can the signature of ABDWGS as the InTeMax be recognized since the beginning of the 1980s only, and not before? To investigate this, property changes at the density range of ABDWGS(equal to that of the InTeMax) are examined from 1950 onwards.
 Until the early 1980s, the density range of ABDWGS was found near the surface in the central Greenland Sea, and both the potential temperature and salinity in this density range showed high variability (Figure 3). Since the beginning of the 1980s, a deepening of this density range was observed in the Greenland Sea (Figure 3c), coincident with the first appearance (1982) and occurrences at deeper levels (1988 and 1989) of the InTeMax. Simultaneously, the variability of the hydrographic properties reduced, and we conclude that this reflects a reduced impact of atmospheric forcing on this layer and thus less mixing with waters from above. Following this line of argument, the waters in the ABDWGS density range seem to have been isolated from the surface since the beginning of the 1990s, and they are affected by vertical mixing only occasionally in small eddies [Karstensen et al., 2005; Budéus and Ronski, 2009]. Since this time, the temperature and salinity in the density range of ABDWGS have been fairly constant (Figures 3a and 3b) and can be clearly identified through the InTeMax (Figure 2a) throughout the central Greenland Sea.
 After 2003, part of the InTeMax was below 2000 m the depth of the Jan Mayen Ridge (Figure 2a). During this time, temperature and salinity slightly increased between the isopycnals bounding ABDWGS (InTeMax) (Figures 3a and 3b), but the increase predominantly affected the lower part of the InTeMax (Figures 2a and 2b). As long as the density range of ABDWGS(InTeMax) has been entirely above the depth of the Jan Mayen Ridge, it could have freely crossed it. Thus, temperature and salinity remained constant above that depth, as in the time before when part of the warm water left the Greenland Sea towards south. However, below the sill depth, temperature and salinity increased, so that the vertical structure of the InTeMax became progressively blurred (Figures 2a and 2b). A concomitant effect could also be a decrease in the contribution of ABDWGS density range to the Denmark Strait overflow after 2003. It could partially explain the different contributions of Arctic Ocean deep waters found on the Denmark Strait overflow by Tanhua et al. —based on data from 1997—and by Jeansson et al. —based on data from 2002.
3.2 Warming and Salinification of Greenland Sea Deep Water (GSDW)
 Below 2000 m, the deep water in the Greenland Sea is confined towards the south by the Jan Mayen Ridge (see Figure 1a). At levels below this depth, temperature and salinity have risen progressively (Figures 2 and 4). The increase is so large that water properties do not resemble those used to define Greenland Sea Deep Water (GSDW) any longer (Table 1). In order to use a property-independent definition, we will consider all waters in the central Greenland Sea between 2000 m and the bottom as GSDW in a similar way as done by Bönisch et al. .
 The 60 year record from the central Greenland Sea shows the onset of an unprecedented warming and salinity increase in the early 1980s (Figures 4a and 4b). The coincidence of the beginning of these trends with the appearance of the InTeMax confirms the absence of deep convection, i.e., convection reaching the GSDW levels, since the beginning of the 1980s. A linear regression of potential temperature and salinity against time for the period between 1982 and 2010 results in mean trends of 0.131°C decade−1 and 0.01 decade−1, respectively. These trends are among the highest trends in the global deep ocean up to date [Kouketsu et al., 2011]. The resulting temperature increase of 0.3°C is the largest in the last 100 years in the deep Greenland Sea according to data published by Aagaard .
 The warming during the late 1950s and the cooling afterwards have been related to phases of more intense or reduced local deep convection and the advection of Arctic Ocean deep waters [Bönisch et al., 1997, and references therein], although the oscillation can hardly be distinguished in salinity (Figures 4a and 4b). The Arctic Ocean deep water dense enough to mix on isopycnals with the deep waters in the central Greenland Sea is EBDW that leaves the Arctic below 2000 m and is not able to cross the Jan Mayen Ridge. Therefore, in the absence of deep convection providing fresher and colder waters below 2000 m, EBDW accumulates in the deep Greenland Sea making it warmer and saltier as long as the original EBDW properties are not reached. Assuming that the present trends will remain constant in the future, GSDW will acquire the properties of EBDW that enters the Greenland Sea in ∼ 10 years in agreement with predictions by Karstensen et al. ; Langehaug and Falck .
4 Import Rate of EBDW into the Greenland Sea
 In the absence of deep convection, the relatively cold and fresh GSDW is only renovated by the advection of warm and salty EBDW. The persistence of this situation during the last three decades provides the ideal in situ experimental setup to estimate the import rate of EBDW from the Arctic Ocean that explains the temperature and salinity increase of the deep Greenland Sea with a simplified heat conservation equation:
As water masses spread on isopycnals, the import of EBDW into the Greenland Sea, TEBDW and the remaining terms of equation (1) are limited by the isopycnals σ2.5 = 39.723 and σ2.5 = 39.743 bounding the EBDW density range that enters the Greenland Sea (EBDWGS). Thus, in equation (1), dθ/dt|ρ is the potential temperature trend observed in the last three decades in the Greenland Sea between the isopycnals bounding EBDWGS (Table 2.II). θEBDW is the mean potential temperature of EBDW at the density range that enters the Greenland Sea (Table 1). θGSDW is the mean value of potential temperature in the central Greenland Sea in the same density range, and VGS is the volume of the central Greenland Sea occupied by this density range. A similar expression to equation (1) can be applied to salinity changes (dS/dt|ρ, Table 2.II).
Table 2. Significant Potential Temperature (dθ/dt) and Salinity (dS/dt) Trends at 0.01 Significance Level in the Central Greenland Sea (74°N–76°N and 0°W–6°W) From 1982 to 2010
Between the different pressure ranges, the corresponding heat content change, Q, and sea level rise, F, scaled to the area of Earth's surface associated to the shown trends are indicated (see section 6).
The necessary volume transport (T) of EBDW that enters the Greenland Sea (EBDWGS, σ2.5=39.723 to σ2.5=39.743) to explain the corresponding changes in θ and S From 1993 to 2010 are indicated (see section 5).
 Changes on isopycnals must be introduced in equation (1) because only these changes can be related to isopycnal advection. They do not coincide with changes on isobars shown in Figure 4 because changes on isobars (dθ/dt|p) are the sum of two components: changes on isopycnals (dθ/dt|ρ) and changes due to vertical displacements of isopycnals (heave, dp/dt|ρ∂θ/∂p) [Levitus et al., 1989; Bindoff and McDougall, 1994]. Changes on isopycnals at the density range of EBDWGS represent between 70% and 80% of those observed on isobars from 2000 m to the bottom (0.09 versus 0.131°C decade−1 and 0.008 versus 0.01 decade−1 for potential temperature and salinity, respectively). The remaining changes on isobars are due to the deepening of isopycnals (heave).
 In order to apply equation (1), it has to be noted that while the warming and salinification trends have remained constant during the last three decades, θGSDW and SGSDW have increased from year to year. In addition, due to the deepening of isopycnals, the specified density range bounding EBDWGS has deepened with time hereby reducing the volume, VGS, occupied by this density range from 3.43×1014 m3 in 1993 to 2.3×1014 m3 in 2009 as the cross-sectional area of the central Greenland Sea decreases towards the bottom. Thus, applying equation (1)—and its equivalent for salinity—to the trends dθ/dt|ρ and dS/dt|ρ in Table 2.II, θEBDW and SEBDW in Table 1, and the varying annual mean values of θGSDW, SGSDW, and VGS from 1993 to 2009 results in the mean import rates and error estimates (2 standard deviation) shown in Table 2.II. We consider the mean value of 0.44±0.09 Sv (average between dθ/dt|ρ and dS/dt|ρ estimates) for the import rate of EBDW into the central Greenland Sea. This estimate gives a residence time of GSDW of 30 years in agreement with estimates based on tracer observations [Schlosser et al., 1991; Rhein, 1991], suggesting that an increase of EBDW transport (0.12 Sv before the 1980s) [Bönisch and Schlosser, 1995] would have compensated the strong decrease in deep water formation rate after the 1980s (0.47–0.5 Sv before the 1980s versus 0.1 Sv after the 1980s) [Schlosser et al., 1991; Rhein, 1991; Bönisch and Schlosser, 1995].
5 Thermohaline Changes Contributions to Global Heat Content Change and Sea Level Rise
 Here, the contributions of the pronounced changes in potential temperature (dθ/dt|p) and salinity (dS/dt|p) in the central Greenland Sea to the local heat content change and sea level rise are calculated (all data concerning this section are detailed in Table 2.I). The results are scaled by the total Earth's surface area following the procedure by Purkey and Johnson  to be comparable with global energy budget studies, concerning not only the hydrosphere but also the remaining components of the Earth's climate system.
 Consequently, the estimates of the local heat content changes per unit area of Earth's surface (Q, equation (2)) shown on the right part of Table 2.I is obtained by multiplying the trend in potential temperature within a given pressure range (dθ/dt|p in Table 2.I) by the potential density (ρ), heat capacity (Cp), cross-sectional area (aGS), and thickness of the corresponding layer, and dividing the result of this product by the Earth's surface area (aES: 510072000 km2).
 Together, the sum of all pressure intervals result in an increase of Q of 2.33±0.08×10−3 W m−2 (2.87±0.1×1018 J from 1982 to 2010) (Table 2). Recent World Ocean estimations of Q are 0.18 W m−2 from 1955 to 2010 [Levitus et al., 2012] and 0.64 W m−2 from 1993 to 2008 [Lyman et al., 2010] for waters from the surface to 700 m. We compare the mean of these two values with our results for the last three decades. Considering that this mean value would be representative of climate warming that is neither found in the atmosphere nor in the upper ocean and is suggested to be accumulating in the deep ocean [Trenberth and Fasullo, 2010; Meehl et al., 2011], the warming in the deep central Greenland Sea during the last three decades constitutes 0.28% of that expected heat storage in the deep ocean. This contribution is 5.8 times larger than the percentage contribution of the central Greenland Sea area deeper than 2500 m (2.4885e+05 km2) to the global Earth's surface (0.05%). More importantly, this figure is much larger than the contribution per unit area of Earth's surface of most ocean basins. An illustrative example is that, weighted by their respective areas, the contribution of the deep Greenland Sea to the changes of the World Ocean heat content is 33 times larger than that of the deep Pacific Ocean [Kawano et al., 2010; Kouketsu et al., 2011; Levitus et al., 2012].
 Interestingly, most of the pronounced warming of the deep Greenland Sea (∼70%) which is responsible for its high contribution to the increase in the World Ocean heat content is due to the advection of Arctic Ocean deep waters. Therefore, a crucial issue to investigate is to which extent the deep Arctic Ocean properties are changing and whether or not these changes compensate for those observed in the Greenland Sea.
 The contribution of the changes in potential temperature to the global sea level rise, Fθ (Table 2.I), is obtained by multiplying the trend of potential temperature within a given pressure range by the thermal expansion of sea water (α) of the layer of interest, its cross-sectional area (aGS) and thickness, and dividing by the Earth's surface area (aES). A similar expression is used to calculate the haline contribution to global sea level rise, FS, where dθ/dt|p must be substituted by dS/dt|p and α by β, the haline contraction of sea water.
Thus, the thermosteric contribution of these changes (Table 2.I) to sea level rise constitutes 0.12% of the global estimation of approximately 1.6–1.8 mm year−1, estimated on the basis of tide gauge data [Levitus et al., 2012]. However, the increasing salinity at depth compensates for part of this rise, and together, the haline and thermosteric contributions explain 0.09% of the global sea level rise. This contribution is smaller than that to the global heat content change because the thermal expansion and haline contraction of seawater are relatively small at cold temperatures and high pressures [Purkey and Johnson, 2010]. Nevertheless, it is still two times larger than the contribution of the central Greenland Sea to the area of the global Earth's surface (0.05%).
 (1) The appearance and persistence of an intermediate temperature maximum in the central Greenland Sea and (2) the warming and salinity increase below have been proposed in this work as evidence for the halt of deep convection since 1982. The origin of both changes in the Greenland Sea has been identified as the advection of Arctic Ocean deep waters: the former related to the advection of ABDW and the later to EBDW.
 ABDW has always penetrated towards the central Greenland Sea, but prior to the 1980s, its signal was completely eroded by deep convection, and it was hence not evident in the hydrographic record. Later, the density range of ABDW that entered the Greenland Sea deepened progressively, and it was identifiable by the intermediate temperature maximum. Since 2003, the density range of ABDW that enters the Greenland Sea has reached the depth level of the Jan Mayen Ridge, starting to show temperature and salinity increase.
 This behavior resembles that of EBDW in the Greenland Sea, which cannot cross the Jan Mayen Ridge and accumulates in the central Greenland Sea, hereby increasing the temperature and salinity of the deep waters in the absence of deep convection. The strong reduction of deep convection during the last three decades has enabled us to estimate the import of EBDW into the Greenland Sea (0.44±0.09 Sv) necessary to explain the observed changes in GSDW.
 In general, warming and salinification in the deep Greenland Sea are among the highest trends in the deep oceans. Despite the comparatively small area of the central Greenland Sea, the contributions of these trends to the World Ocean heat content increase and sea level rise are significant. Because trends in GSDW are mainly due to advection of Arctic Ocean deep waters, changes in the deep Arctic Ocean have to be addressed too, in order to reveal a complete estimate of the Arctic Mediterranean deep waters' role in the present climate change.
 The authors wish to thank Peter Schlosser and another anonymous reviewer for their valuable and helpful comments on the manuscript.
 The editor thanks Peter Schlosser and an anonymous reviewer for their assistance in evaluating this paper.