U.S. Dept. of Commerce / NOAA / OAR / PMEL /Publications
Since 1985 various investigators have proposed that Norwegian Sea deep water (NSDW) is formed by mixing of warm and saline deep water from the Arctic Ocean with the much colder and fresher deep water formed by convection in the Greenland Sea (GSDW). We here report on new observations which suggest significant modification and expansion of this conceptual model. We find that saline outflows from the Arctic Ocean result in several distinct intermediate and deep salinity maxima within the Greenland Sea; the southward transport of the two most saline modes is probably near 2 Sv. Mixing of GSDW and the main outflow core found over the Greenland slope, derived from about 1700 m in the Arctic Ocean, cannot by itself account for the properties of NSDW. Instead, the formation of NSDW must at least in part involve a source which in the Arctic Ocean is found below 2000 m. The mixing of the various saline outflows is diapycnal. While significant NSDW production appears to occur in northern Fram Strait, large amounts of saline Arctic Ocean outflow also traverse the western Greenland Sea without mixing and enter the Iceland Sea. During the past decade, deep convection in the Greenland Sea has been greatly reduced, while deep outflow from the Arctic Ocean appears to have continued, resulting in a markedly warmer, slightly more saline, and less dense deep regime in the Greenland Sea.
The presence of a large body of saline deep water in the Arctic Ocean (Arctic Ocean deep water, or AODW), with a probable formative connection to the surrounding shelf seas, is by now well established [Aagaard, 1981; Swift et al., 1983; Aagaard et al., 1985], as is the outflow of this water to the Greenland Sea through western Fram Strait [Aagaard et al., 1985; Smethie et al., 1988; Swift and Koltermann, 1988]. This earlier work suggests that the importance of the saline Arctic Ocean outflow lies first and foremost in its contribution to the Norwegian Sea deep water (NSDW), which is now thought to be formed by mixing of the outflow with the Greenland Sea deep water (GSDW), thus providing a rather direct connection between the deep thermohaline circulation of the Arctic Ocean and that of the seas to the south.
Our purpose here is to shed further light on that connection, based principally on hydrographic measurements in the western Greenland Sea and Fram Strait during 1987. We find that several distinct saline Arctic Ocean sources contribute to the intermediate and deep hydrographic structure of the Greenland Sea, and that while much of the saline Arctic Ocean outflow recirculates in the Greenland Sea, a significant portion continues into the Iceland Sea. We also find clear evidence of mixing between GSDW and deep waters from the Arctic Ocean, including the production of NSDW; some of this production occurs in northern Fram Strait. Variability in the supply of source waters (either warm and saline AODW or cold and fresher GSDW) will lead to changes in the hydrographic structure of the Greenland and Norwegian seas. Such a situation appears to have occurred during the past decade, with reduced convection in the Greenland Sea (but persisting outflow from the Arctic Ocean) resulting in a warmer and less dense deep regime.
During June 1987 we obtained a series of hydrographic sections in the Greenland Sea from the R/V Polarstern. The emphasis was on the relatively undersampled southwestern portion of the sea, together with the northern source waters in Fram Strait. Figure 1 shows those stations extending below 500 m. Other, shallower stations primarily supported the biological programs on board, and these provide no information on the deep water issues of interest here.
Figure 1. Hydrographic stations extending below 500 m, June 1987. The reference station for Norwegian Sea deep water characteristics shown in Figure 3 is denoted by the solid circle marked I. Mooring locations during 1987-1988 used to define the eastward recirculation of water from the East Greenland Current are shown by the three triangles labeled GS, and locations of moorings during 1987-1989 used to estimate transport in the saline core over the Greenland slope are shown by the two unlabeled triangles.
At each of the deep stations, conductivity-temperature-depth (CTD) profiling was accompanied by discrete sampling for nutrients and dissolved oxygen, using a 24-bottle rosette. The CTD, a Neil Brown Instrument Systems (NBIS) Mark III, was calibrated for temperature and pressure at the Scripps facility before and after the cruise. In situ calibration was done at each station using a full set of discrete salinity samples spanning the depth of the cast and 3-4 racks of deep-sea reversing thermometers. The salinities were determined with a Guildline Autosal standardized against standard seawater batch P-105 before and after the analysis of the samples from each station. These bottle salinities have a precision and accuracy of about 0.001 and 0.002, respectively. The time- and pressure-corrected CTD profiles are accurate to about 0.003 in salinity and 0.003°C. In regions of strong gradients the time constant mismatch between the temperature and conductivity sensors reduces the accuracy significantly further; however this is not a problem in the deep water sphere we consider here.
Dissolved oxygen and nutrients were determined by standard methods using a
Manostat titration system and a Technicon autoanalyzer, with due attention being
paid to the analysis standards and to temperature control. We judge the precision
to be 0.02 mL L
for
oxygen and 0.1-0.2, 0.01-0.02, and 0.1-0.2 µmoL L,
respectively, for nitrate, phosphate, and silicate.
We shall also refer briefly to CTD data from Polarstern cruises in 1988 and 1989. The calibration procedures were similar to those in 1987, and the overall data quality is fully as good.
One or more intermediate salinity maxima connected with the outflow of deep water from the Arctic Ocean were found throughout most of the southwestern Greenland Sea during the 1987 cruise. Figure 2 shows the section extending NW-SE across the Greenland slope near 75°N. The saline core, which contains a salinity maximum of 34.919, lies over the slope just inshore of the 2000-m isobath, but its influence can be seen extending out over the abyssal plain. The salinity inversion structures with vertical scales as much as several hundred meters in the seaward extension of the core are reminiscent of those portrayed in Figure 10 of Aagaard et al. [1985] and would therefore seem to be a common feature in the spreading in the Greenland Sea of the salinity maxima which represent deep water from the Arctic Ocean. We did in fact also observe such salinity structures during subsequent cruises in both 1988 and 1989, although the details differ somewhat from year to year. We shall return to this latter issue.
Figure 2. Salinity distribution below about 500 m in the section between 75°-76°N (see inset for location). The most saline portion of the Arctic Ocean outflow is centered near 1200 m.
Figure 3 shows the potential temperature-salinity
characteristics below about 800 m at three stations: station 231, located over
the Greenland slope in the core of the saline outflow (cf. Figure
2); station 233, near the base of the slope where multiple cores with salinity
somewhat lower than at station 231 are apparent (cf. Figure
2); and station 264, near the supposed center of the deep convective region
of the Greenland Sea. Particularly notable is the presence of a salinity maximum
even in the latter region; a similar feature has been described by Clarke
et al. [1990]. Above 1900 m depth the potential temperature-salinity
values are plotted in Figure 3 at 50-m intervals
except for station 264, where the intervals are 100 m, as is also the case below
1900 m at station 233. Note that in moving away from the saline core over the
slope (station 231), the salinity maximum layers appear to be both freshened
and cooled, in accord with the proposed scheme of mixing between the Arctic
Ocean outflow and the GSDW. Examination of the entire data set for 1987 suggests
that the various salinity maxima can be represented as being of three types.
Type 1 is by far the most common salinity maximum below 1000 m. It almost always
represents the most saline water at that location (except for the upper layer
of warm water of recent Atlantic origin); its potential density is 28.075 ±
0.003; it generally lies at a depth of 1500 ± 300 m, except near the center
of the gyre where it is found down to about 2500 m; and at some stations it
is represented by several distinct maxima in the vertical. The absolute salinity
maxima at all three stations in Figure 3 are
of type 1. Type 2 is warmer and/or less saline, with a significantly lower potential
density, typically 28.067, and it is usually found above 1000 m. The upper salinity
maximum near 950 m at station 233 (Figure 3)
is an example of type 2. Because of its lower density, we will not consider
type 2 further here, but note that in like types 1 and 3, it also appears to
mix with relatively fresh and cold Greenland Sea waters. Type 3 is a colder
water with potential density 28.080 or more; it occurs at a relatively small
number of stations, where it is typically found near 2000 m. The deepest salinity
maximum at station 233, near 1800 m (Figure 3),
is an example of type 3. Note that the three thickest fingerlike salinity structures
shown in Figure 2 correspond to the three types
of salinity maximum we have described, with type 2 lying highest in the water
column and type 3 lowest. Note also that a typical density contrast between
waters of types 1 and 3 of 0.005 in 
increases to nearly 0.02 when the waters are compared at 2000 db.
Figure 3. Potential temperature-salinity characteristics below 800 m at stations 231, 233, and 264. These stations respectively represent the Greenland slope where the Arctic Ocean outflow of highest salinity is found, the region near the base of the slope containing multiple reduced salinity maxima, and the central Greenland Sea in the region of deep winter convection where the deep water is the freshest. Plotting intervals for stations 321 and 233 are 50 m above 1900 m and 100 below 1900 m depths, plotting intervals for station 264 are 100 m over the entire depth range. The numbers 1, 2, and 3 denote the salinity maximum type (see text), and the solid lines represent the associated isopycnals at surface pressure.
The potential temperature-salinity characteristics of the type 1 salinity maximum at the various stations are remarkably well correlated. Figure 4 shows these characteristics for all the 1987 Greenland Sea stations south of 77°N, together with their linear regression (marked LR). Also shown in Figure 4 are the potential temperature-salinity relations below 1300 m at stations 203 and 204, the former located in northern Fram Strait over the lower Greenland slope, where the Arctic Ocean outflow first enters the Greenland Sea, and the latter in deeper water farther east, where its Arctic Ocean characteristics have been significantly altered. The type 1 points are clustered tightly about the regression line, with a correlation coefficient of 0.95 (n = 30). Furthermore, the regression line passes through the potential temperature-salinity characteristics of the GSDW as represented by the water below 3300 m at station 264, and it intersects the curve for station 203 just above 1700 m. Note also in Figure 4 that there is relatively little variation in core properties along the Greenland slope (compare station 231 over the central Greenland slope with stations 257 and 248 located 300-400 km to the southwest). These features suggest that the prominent intermediate salinity maximum in the Greenland Sea, which we have denoted type 1, represents water which has been withdrawn from the Arctic Ocean through western Fram Strait at about the 1700-m level (compare Aagaard et al. [1985]). This water flows southeastward along the slope, with the western part of the core being mixed relatively little. However, saline water on the eastern side of the core is drawn in toward the interior of the Greenland Sea, where it is mixed with cold, low-salinity Greenland Sea waters. Particularly notable is the presence of a slight salinity maximum even in the center of the deep Greenland Sea, where it lies above the colder and fresher GSDW (e.g., at station 263); Figure 4 suggest an Arctic Ocean contribution there of about 10%.
Figure 4. Potential temperature/salinity characteristics at the depth of the intermediate salinity maximum (solid circles with station numbers); the line LR shows their linear regression. At some stations there was more than one salinity maximum of type 1. The two lighter lines represent isopycnals at 1500 and 2500 db. The lighter solid circles labeled with their observation depths show the potential temperature/salinity characteristics below 1300 m at stations 203 and 204 (see Figure 1 for locations). The open circles denote type 3 salinity maxima. The deep water characteristics of the Greenland Sea (GSDW), Norwegian Sea (NSDW) and the adjacent Arctic Ocean (EBDW) are respectively derived from observations below 3300 m at station 264, from observations below 2500 m at the Norwegian Sea intercalibration station (denoted I in Figure 1) ,and from Swift and Koltermann [1988].
An important aspect of Figure 4 is that the
intermediate salinity maxima lie along a line (LR) strongly suggestive of two-component
mixing which is diapycnal at all but surface pressure, as is clear from the
orientation of the two subsurface isopycnals shown in Figure
4: 
= 35.149 and 
= 39.733. This contrasts with earlier suggestions that such mixing should be
isopycnal at ambient pressure [Aagaard
et al., 1985; Swift
and Koltermann, 1988]. We return to this issue in connection with the
formation of NSDW.
We have indicated in Figure 4 the characteristics of NSDW as exemplified by the property range below 2500 m determined by bottle samples at an intercalibration station in the northern Norwegian Sea during 1989 (denoted I in Figure 1). The NSDW characteristics adopted by Swift and Koltermann [1988] lie just outside this range on the cold and saline side. Note that these NSDW characteristics lie to the right of the type 1 regression line. Ignoring for the moment the possibility of temporal variability, this suggests that NSDW contains an Arctic Ocean contribution more dense and of deeper origin than what we have denoted as type 1 water. We also show in Figure 4 the potential temperature-salinity characteristics for the type 3 salinity maximum which was found at six stations. The control of the CTD salinity quality with independent bottle samples was very good at these stations, but the extreme maximum at station 258 (34.913) was seen only on the downcast. Since the CTD appears to have been well-behaved on both the preceding and following stations, and since agreement between the CTD and bottle samples was also acceptable on the upcast (with a potential temperature-salinity value at the depth of the deep salinity maximum comparable to those at stations 233 and 240), we do not feel justified in rejecting the salinity maximum on the downcast at station 258. Rather, the observations at this station may point to small-scale structure or fronts within the deep water, which were crossed during the time between the down- and upcasts. Regardless of uncertainty with respect to this station, it is clear that most of the type 3 water, which in potential temperature-salinity space lies extremely close to the properties of NSDW, could be formed by mixing between GSDW and water from the Arctic Ocean similar to that found at 2100 m at station 203. The latter is in fact close to the characteristics for such water adopted by Swift and Koltermann [1988] and labeled in Figure 4 as EBDW (Eurasian Basin deep water, a subset of AODW). Note also that such a mixing line lies closer to an in situ isopycnal than does the regression line for the type 1 salinity maxima, although the mixing would still be significantly diapycnal.
The potential temperature-salinity characteristics of station 204 (Figure 4) further strengthen the argument for mixing between GSDW and an Arctic Ocean source deeper and denser than that of the saline core over the Greenland slope (type 1). In particular, note that the potential temperature-salinity curve for station 204 is drawn down toward GSDW properties at a number of depths, including a thick layer between 2000-2500 dbar which approaches NSDW properties. Note also that this station lies in northern Fram Strait, suggesting that water with NSDW properties may be formed there by mixing between GSDW and EBDW. It is therefore possible that the presence of NSDW along the northern periphery of the Greenland gyre, first noted by Metcalf [1960], results at least in part from local formation of this water mass in Fram Strait, rather than simply representing a deep westward recirculation of water with a southern origin which has come northward through the eastern Greenland Sea as part of the large-scale cyclonic circulation.
Finally, we point out that our classification of the water exhibiting the salinity maximum into three types, while useful, is somewhat arbitrary even within this single data set, and that we can expect observations which do not fit the scheme rigorously. For example, the second-deepest salinity maximum at station 233, near 1600 m (Figure 3), is actually intermediate between types 1 and 3. However, the properties of this water are extremely close to those of NSDW defined in Figure 4, suggesting that type classifications, mixing schemes, and water mass conceptualizations need to retain a certain flexibility. This becomes particularly important when comparing data from different cruises. For example, while we found the salinity structures over the Greenland slope during both 1988 and 1989 to be generally similar to those in 1987, the details differ somewhat. Had we been working solely from either of those later data sets, we might therefore have chosen slightly different criteria in classifying or characterizing the structures, e.g., with respect to their density ranges. Thus, while the salinity maxima for 1988 and their associated potential temperatures cluster very closely around the type 1 regression line for 1987 (Figure 4), the 1989 data suggest the salinity maxima during that year to be slightly lower in density (by about 0.001 in the mean). However, none of these differences in detail alter our fundamental conclusion that a spectrum of saline water types from the deep water regime of the Arctic Ocean interleaf and mix with fresher and colder waters which have been conditioned by surface processes in the Greenland Sea.
We have plotted in Figure 5 the areal distribution
of the salinity maximum for water with type 1 characteristics (shown by three-digit
salinity values), together with the salinity distribution on the surface
= 28.075 (heavy lines denoted by five-digit salinity values). This surface corresponds
to the mean potential density of the type 1 salinity maximum. We note that the
selection of a plotting surface is somewhat arbitrary in this instance, since
the mixing appears to be significantly diapycnal, but mapping the salinity distribution
on other surfaces yields similar results.
Figure 5. Salinity at the depth of the intermediate salinity maximum; only the last three digits are shown. Heavy lines denoted by five digits show the distribution of the salinity on an isopycnal surface corresponding approximately to the midpoint of the regression line in Figure 4.
Two features in Figure 5 stand out. One is the
suggestion that the saline outflow from the Arctic Ocean follows the Greenland
continental slope in a coherent fashion only to about 74°N. South of that latitude
the isopleths diverge, and a significant portion of the saline core appears
to recirculate eastward, closely following the deeper isobaths. This recirculation
is clearly seen in the records from current meters moored approximately along
the meridian 8°W during 1987-1988 (positions shown in Figure
1). Table 1 shows that at the northernmost mooring (GS-3 at 73°N), the long-term
mean flow, which is barotropic, was directed eastward at 5-6 cm s.
In contrast, the two sites farther south did not show statistically significant
mean flow at any level. The concentration of the recirculation north of 72°30
N
suggested by the current records is consistent with the stronger horizontal
gradient in the salinity field in that region shown in Figure
5. Instead of representing a broad eastward sweep throughout the southern
Greenland Sea, the recirculation therefore appears to be concentrated in a rather
narrow band, perhaps 50 km in meridional extent (contrast the schematic representation
in Figure 12 of Smethie
et al. [1988]).
| Standard | |||||
|---|---|---|---|---|---|
| Maximum Speed | Deviation | ||||
| Depth, | Record Length, | Mean Velocity, | (6-Hour Mean), | Principal | (6-Hour Mean), |
| m | Days | cm s |
cm s |
Axis, °T | cm s |
| Mooring GS-1 (71°55.7 N,
8°21.7W; 2424
m) |
|||||
| 74 | 80 | 0.3, 173°T | 3.0 | 330 | 0.8 |
| 214 | 355 | 1.6, 118°T | 12.3 | 275 | 2.6 |
| 2374 | 355 | 1.6, 87°T | 8.6 | 260 | 2.1 |
| Mooring GS-2 (72°27.6N,
8°11.0W; 2460
m) |
|||||
| 60 | 356 | 1.5, 299°T | 14.2 | 295 | 3.8 |
| 200 | 356 | 1.1, 301°T | 12.8 | 290 | 3.5 |
| 2410 | 356 | 1.3, 288°T | 14.8 | 315 | 2.2 |
| Mooring GS-3 (73°03.8N,
8°01.7W; 2620
m) |
|||||
| 80 | 357 | 6.0, 94°T | 20.7 | 290 | 3.8 |
| 220 | 357 | 5.5, 93°T | 20.1 | 290 | 3.6 |
| 2570 | 357 | 5.4, 90°T | 20.9 | 280 | 4.3 |
The second important feature in Figure 5 is the apparent continuation of the most saline water southward into the Iceland Sea, as indicated by the orientation of the isohaline 34.915. The sill depth between Greenland and Jan Mayen exceeds 1600 m, with a trough cutting southward across the western part of that passage near the location of station 248, where Figure 5 shows a maximum salinity of 34.917 near 1300 m. (Other stations in this southernmost section were in water insufficiently deep to define a salinity maximum.) Except for type 3 water, the saline cores are therefore not topographically constrained from entering the Iceland Sea. They cannot, however, continue out of the Iceland Sea along the Greenland slope, since the sill depth in Denmark Strait, between Greenland and Iceland, is only about 600 m. It therefore appears that saline outflow from the Arctic Ocean contributes directly and in significant quantities to the deep water of the Iceland Sea, where there is no low-salinity GSDW to dilute it, only the slightly less dense intermediate waters produced there. We would, for example, expect a high-resolution comparison of the deep waters of the Norwegian and Iceland seas to show the water in the latter basin to be slightly more saline. Clarke et al. [1990] have in fact shown that the deep Norwegian Sea contains a slight salinity maximum in the vertical, and we suggest that this feature originates with the intermediate salinity maximum carried southward from the Arctic Ocean into the Iceland Sea, whence its mixing products feed the Norwegian Sea. Very recently, Malmberg et al. [1990] have also presented evidence of Arctic Ocean outflow in the Iceland Sea; during the 3 years 1987-1989 the most pronounced occurrence appears to have been in 1987, the year of the observations reported here.
We had earlier suggested that the saline core of Arctic Ocean outflow might
also be distinguishable by a local silicate maximum [Aagaard
et al., 1985]. Our present detailed sampling does not show any nutrient
extrema associated with the salinity maximum, however (Figure
6). Instead, there is a small minimum in dissolved oxygen, with values near
7 mL L (Figure
6). Such a minimum is also present at middepth (6.9 mL L
near 1600 m) at station 203 in northernmost Fram Strait, where it is probably
a representative feature of upper Arctic Ocean deep water. Within the Greenland
Sea, therefore, the oxygen minimum appears to be advected from the north with
the saline core, rather than being formed locally by nonconservative processes.
Its origin within the Arctic Ocean is uncertain and will probably remain so
until the data base there is expanded considerably.
Figure 6. Distribution below 800 m of salinity, dissolved oxygen, and nutrients at two stations in the core of the saline Arctic Ocean outflow (see Figure 1 for locations).
However, the distribution in the Greenland Sea of the oxygen minimum provides independent evidence of two-part mixing similar to that suggested by the temperature/salinity correlations. To check the applicability of the two-component mixing concept, we performed a regression analysis on the bottle-based oxygen, temperature, and salinity data corresponding to the type 1 salinity maximum. We found that not only did the bottle-based temperature-salinity data provide a fit virtually identical to that of Figure 4, based on the CTD data, but the correlation between these two parameters and dissolved oxygen also corresponded closely to two-part mixing (Figure 7), with a correlation coefficient of -0.94 (n = 23). Over the spatial and temporal scales represented by these observations, dissolved oxygen in the Greenland Sea therefore appears to be a nearly conservative tracer, the distribution of which is consonant with the two-part diapycnal mixing suggested by the temperature-salinity correlation portrayed in Figure 4.
Figure 7. Dissolved oxygen-salinity characteristics at the depth of the intermediate salinity maximum; the line LR shows their linear regression. The characteristics of the deep Greenland Sea (GSDW) and of the outflow from the Arctic Ocean of the type 1 salinity maximum (203/1579 m) are respectively derived from samples at 3337 m at station 264 and at 1579 m at station 203.
We have argued that the outflow of relatively dense water from the Arctic Ocean to the Greenland Sea occurs over a range of depths and that its subsequent interactions with Greenland Sea waters result in several distinct intermediate and deep salinity maxima within the Greenland Sea. The principal saline core overlying the Greenland slope represents withdrawal of water from about 1700 m in the Arctic Ocean, but although it appears through mixing with GSDW to account for much of the salinity structure in the Greenland Sea, such mixing cannot by itself form NSDW, since the latter lies slightly off the mixing line on the cold and saline side. Instead, NSDW must at least in part involve a deeper source within the Arctic Ocean, apparently lying below 2000 m. A noteworthy aspect of the mixing, regardless of the Arctic Ocean source level, is that it appears to have a diapycnal component. We do not know the energy source for such mixing. For example, we find no evidence in the smaller-scale CTD profiles to suggest that double diffusive processes are important. An alternative energy source might be dynamic instabilities in the boundary current, although observations farther north suggest that there the flow over the slope is baroclinically stable [Foldvik et al., 1988]. Whether this is also true over less steep topography elsewhere remains to be explored.
Regardless of the details of the various sources and ensuing mixing, there is now clear evidence that the temperature-salinity properties of the various deep basins south of Fram Strait are controlled by the interplay between two different hydrographic regimes. One is the relatively warm and saline deep water sphere of the Arctic Ocean, the properties of which in part are accounted for by shelf processes, and the other is the much colder and fresher regime of the Greenland Sea, which is driven by open-ocean convection. A curious aspect of this situation is that the freshness of the Greenland Sea also is the result of water export from the Arctic Ocean, apparently primarily in the form of low-salinity sea ice [Aagaard and Carmack, 1989], so that in a sense both regimes are of Arctic Ocean origin, with the deep Greenland Sea attempting to mix again descendants of the salt and fresh water which were separated by freezing over the shelves of the Arctic Ocean.
Given this interplay within the Greenland Sea between source waters of very different characteristics, we might expect variability in the deep hydrographic structure if one or both of the sources also varied. Figure 8 shows two stations at essentially identical locations within the convective region of the central Greenland Sea; they represent conditions near the end of the convective season, but are seven years apart. Note that below 1500 m the deep water warmed markedly during the intervening years, generally, by 0.05°-0.08°C, depending on depth. The salinity increase during the same period appears to have been confined to below 2000 m and did not exceed about 0.002, which is close to the accuracy of these measurements. (In preparing Figure 8, we have increased the published [Clarke et al., 1984] 1982 Hudson salinities by 0.001, based on a careful comparison of the various standards used in the 2 years.)
Figure 8. Potential temperature-salinity characteristics at a site in the central Greenland Sea during 1982 and 1989, together with the depth of each observation. The inset shows the change in properties from 1982 to 1989, with positive values representing warming and increasing salinity.
Tritium/
He and CFC (chlorofluorocarbon)
measurements in the Greenland Sea over the same period suggest little deep convective
renewal during these 7 years [Schlosser
et al., 1991]. The observed thermohaline changes in the deep water therefore
likely resulted from a shutdown of the deep surface-driven convective regime
in the central Greenland Sea, which normally supplies cold low-salinity water,
while the outpouring of relatively warm and saline water from the Arctic Ocean
continued. In this connection, note in Figure 8
that the warming and slight increase in salinity reach a maximum in the depth
interval 2000-2500 m. This is the interval in which the type 3 salinity intrusion
appears to be concentrated (compare station 204, Figure
4). The absence of apparent warming during 1989 above 1500 m is consistent
with the observation that during the late winter of 1989 convection reached
to at least 1600 m depth after a period of several years of very limited convection
[Meincke
et al., 1990]. That is, the recent strengthening of the convection partially
reversed the warming trend of the 1980's, cooling (and presumably freshening;
see Figure 8) the water column above about 1500
m.
From this perspective, the importance of the warm Arctic Ocean outflow lies in its contribution of sensible heat to the Greenland Sea, which more than compensates for the effect on density of the small addition of salt by the same outflow. The net effect of this outflow is therefore to destabilize the middle and lower water column in the convective region of the central Greenland Sea. This preconditions the deep ocean to overturning after significant surface-driven convection develops in the upper water column. In this manner, the warm Arctic Ocean outflow serves as a regulatory mechanism which tends to maintain the long-term continuity of the deep ventilation in the Greenland Sea.
A crude transport estimate for the warm and saline outflow from the Arctic
Ocean can be made as follows. In both of the sections across the Greenland slope
between 75°-76°N portrayed in our Figure 2 and
in Figure 10 of Aagaard
et al. [1985], the cross-sectional area of water more saline than 34.910
is about 25 km
. The mean
salinity of this water is close to 34.913, and it principally represents the
type 1 salinity maximum. Direct current measurements in this section during
1987-1988 from two instruments moored 5 m above the bottom (locations given
in Figure 1) show a year-long mean flow of 13.2
cm s toward SSW along
the slope in water 1250-1300 m deep. Measurements of shorter duration (106-224
days) the following year, made 5 m and 600 m above the 1800 m isobath, i.e.,
slightly farther seaward, gave record-length means from 8.0-9.7 cm s
. A time-weighted mean value suggests that the flow through this section at
depths of 1200-1300 m is about 12 cm s.
For a cross section of 25 km,
this yields a transport of 3 Sv with a mean salinity of 34.913. From the regression
line of Figure 4, water of this salinity represents
a mixture containing two-thirds Arctic Ocean water, suggesting that the outflow
rate of the intermediate salinity maximum from the Arctic Ocean (type 1) is
about 2 Sv. A small amount of the type 2 water has also been included in this
calculation, while the type 3 water, which is important to the formation of
NSDW, has largely been omitted. A similar calculation can be made for a section
farther north, namely, that of Smethie et al. [1988; their Figure 5a]
near 78°N. The cross-sectional area of water in that section more saline than
34.910 is about 70 km,
and the mean salinity of this water is close to 34.917. If we assign the year-long
mean speed of 2.3 cm s
at 1378 m measured about 150 km farther north by Foldvik
et al. [1988; their Table 1] during 1984-1985, the transport of the
saline water through this section is 1.6 Sv. From the regression line of Figure
4, water of salinity 34.917 has a type 1 content of about 80%, suggesting
an outflow rate from the Arctic Ocean of 1.3 Sv. Our range of 1.3-2.0 Sv contrasts
with the smaller estimates of Smethie
et al. [1988] (0.8-0.9 Sv) and Heinze
et al. [1990] (0.7-1.0 Sv), both based on box models. Apart from issues
of long-term variability, our larger outflow values are reasonable considering
the various model assumptions, particularly (1) the capping in the models of
the deep water sphere at 1500-1700 m (thereby missing the large amount of saline
Arctic Ocean outflow above those depths) and (2) the assumption that all the
outflow mixes with GSDW (which is clearly not the case, since a significant
fraction continues southward into the Iceland Sea without mixing).
Converting our estimated outflow rate into a production rate for NSDW is problematic,
principally because of uncertainties about the production process and the fraction
of saline water which actually participates in NSDW production. Nonetheless,
a set of fairly reasonable assumptions suggest a production rate quite consistent
with calculations using the growing body of transient tracer data. For example,
if we assume for the moment that (1) the total outflow of saline waters of types
1 and 3 from the Arctic Ocean is 2 Sv, (2) only one half of this outflow mixes
with GSDW (the remainder entering the Iceland Sea; see Figure
5), and (3) the portion which is mixed does so in a 2:1 ratio (see Figure
4), then a production rate of new NSDW of 1.5 Sv is implied. Since the volume
of the Norwegian Sea below 1700 m is 8.6 × 10
km [Smethie
et al., 1988], a deep water formation rate of 1.5 Sv would give a replacement
time of 18 years. This is comfortably in the range of time scales estimated
from transient tracer box models [e.g.,
Schlosser, 1985; Smethie
et al., 1988; Heinze
et al., 1990], although certain details of these models may, as we've
suggested, not be realistic. In the present case, these model assumptions apparently
do not lead to unrealistic replacement times, possibly because in the real ocean
only a portion of the initially large outflow from the Arctic Ocean mixes with
GSDW to produce NSDW.
On the other hand, it is now clear that the steady state assumptions implicit in these kinds of calculations have limited applicability to the Greenland Sea, and in particular that the production of GSDW is highly variable, so that both the properties and the effective renewal rates of the NSDW can be expected to vary on time scales of a few years. It is also clear that the outflow of deep waters from the Arctic Ocean represents a complex pattern in potential temperature-salinity space. In particular, what we have here classified as type 1 water cannot by itself account for the various salinity maxima south of Fram Strait, but instead additional sources from deeper levels in the Arctic Ocean must also be invoked. Furthermore, the mixing regime and the circulation pattern are more complex than has been generally appreciated. For example, diapycnal mixing is important, there is significant NSDW production outside the western Greenland Sea (e.g., in northernmost Fram Strait), deep waters from the Arctic Ocean also feed the Iceland Sea, and large amounts of NSDW probably never enter the Norwegian Sea but instead recirculate to the north and/or are discharged into the Arctic Ocean [cf. Smethie et al., 1988; Swift and Koltermann, 1988]. We suggest that as attempts continue to construct plumbing diagrams of the thermohaline circulation of the various arctic seas and their interconnections, these diagrams will grow increasingly more complex and more interconnected and will increasingly indicate significant variability. This undoubtedly has important implications for climate modeling.
Acknowledgments. We are grateful to Gerd Rohardt, who took care of the CTD instrumentation and processing and assured the quality of this data set. The three reviewers of the manuscript provided exceptionally helpful suggestions. Partial support came from the Arctic Program, Office of Naval Research. NOAA Pacific Marine Environmental Laboratory contribution 1235; Alfred Wegener Institute for Polar and Marine Research contribution 290.
Aagaard, K., On the deep circulation in the Arctic Ocean, Deep Sea Res., 28, 251-268, 1981.
Aagaard, K., and E.C. Carmack, The role of sea ice and other fresh water in the arctic circulation, J. Geophys. Res., 94, 14,485-14,498, 1989.
Aagaard, K., J.H. Swift, and E.C. Carmack, Thermohaline circulation in the arctic mediterranean seas, J. Geophys. Res., 90, 4833-4846, 1985.
Clarke, A.J., J.L. Reid, and J.H. Swift, CSS Hudson Cruise 82-001, 14 February-6 April 1982, vol. 1, SIO Ref. 84-14, Scripps Inst. of Oceanogr., La Jolla, CA, 1984.
Clarke, A.J., J.H. Swift, J.L. Reid, and K.P. Koltermann, The formation of Greenland Sea deep water: Double diffusion or deep convection? Deep Sea Res., 37, 1385-1424, 1990.
Foldvik, A., K. Aagaard, and T. Tørresen, On the velocity field of the East Greenland Current, Deep Sea Res., 35, 1335-1354, 1988.
Heinze, C., P. Schlosser, K.P. Koltermann, and J. Meincke, A tracer study of the deep water renewal in the European polar seas, Deep Sea Res., 37, 1425-1453, 1990.
Malmberg, S.-A., S.S. Kristmannsson, and E. Buch, Greenland Sea project in the western part of the Iceland Sea from Jan Mayen to the Denmark Strait, CM 1990/C:27, Int. Counc. for the Explor. of the Sea, Copenhagen, 1990.
Meincke, J., K. Aagaard, H.v. Aken, E. Buch, A. Clarke, B. Cornuelle, E. Fahrbach, A. Foldvik, J.-C. Gascard, P. Gudmandsen, H.-J. Hirche, H. Hoeber, R. Knox, S. Kristmannsson, S. Legutke, J. Lynch, S.-A. Malmberg, B. Machenhauer, W. Munk, D. Quadfasel, M. Rhein, B. Rudels, P. Schlosser, F. Schott, J. Swift, T. Vinje, P. Wadhams, A. Warpakowski, and P. Worcester, Greenland Sea Project: A venture toward improved understanding of the oceans' role in climate, Eos Trans. AGU, 71, 750-751, 754-755, 1990.
Metcalf, W.G., A note on water movement in the Greenland-Norwegian Sea, Deep Sea Res., 7, 190-200, 1960.
Schlosser, P., Ozeanographische Anwendungen von Spurenstoffmessungen im Mittelmeerausstrom und im Europaischen Nordmeer, Ph. D. thesis, Ruprecht-Karls-Univ., Heidelberg, 1985.
Schlosser, P., G. Bönisch, M. Rhein, and R. Bayer, Reduction of deepwater formation in the Greenland Sea during the 1980's: Evidence from tracer data, Science, 251, 1054-1056, 1991.
Smethie, W.M., Jr., D.W. Chipman, J.H. Swift, and K.P. Koltermann, Chlorofluoromethanes in the arctic mediterranean seas: Evidence for formation of bottom water in the Eurasian Basin and deep-water exchange through Fram Strait, Deep Sea Res., 35, 347-369, 1988.
Swift, J.H., and K.P. Koltermann, The origin of Norwegian Sea deep water, J. Geophys. Res., 93, 3563-3569, 1988.
Swift, J.H., T. Takahashi, and H.D. Livingston, The contribution of the Greenland and Barents seas to the deep water of the Arctic Ocean, J. Geophys. Res., 88, 5981-5986, 1983.
Figure 1. Hydrographic stations extending below 500 m, June 1987. The reference station for Norwegian Sea deep water characteristics shown in Figure 3 is denoted by the solid circle marked I. Mooring locations during 1987-1988 used to define the eastward recirculation of water from the East Greenland Current are shown by the three triangles labeled GS, and locations of moorings during 1987-1989 used to estimate transport in the saline core over the Greenland slope are shown by the two unlabeled triangles.
Figure 2. Salinity distribution below about 500 m in the section between 75°-76°N (see inset for location). The most saline portion of the Arctic Ocean outflow is centered near1200 m.
Figure 3. Potential temperature-salinity characteristics below800 m at stations 231, 233, and 264. These stations respectively represent the Greenland slope where the Arctic Ocean outflow of highest salinity is found, the region near the base of the slope containing multiple reduced salinity maxima, and the central Greenland Sea in the region of deep winter convection where the deep water is the freshest. Plotting intervals for stations 321 and 233 are 50 m above 1900 m and 100 m below 1900 m depths, plotting intervals for station 264 are 100 m over the entire depth range. The numbers 1, 2, and 3 denote the salinity maximum type (see text),and the solid lines represent the associated isopycnals at surface pressure.
Figure 4. Potential temperature-salinity characteristics at the depth of the intermediate salinity maximum (solid circles with station numbers); the line labeled LR shows their linear regression. At some stations there was more than one salinity maximum of type 1. The two lighter lines represent isopycnals at 1500and 2500 dbar. The lighter solid circles labeled with their observation depths show the potential temperature-salinity characteristics below 1300 m at stations 203 and 204 (sea Figure 1 for locations). The open circles denote type 3salinity maxima. The deep water characteristics of the Greenland Sea (GSDW), Norwegian Sea (NSDW) and the adjacent Arctic Ocean (EBDW) are respectively derived from observations below 3300 m at station 264, from observations below 2500 m at the Norwegian Sea intercalibration station (denoted I in Figure 1) ,and from Swift and Koltermann [1988].
Figure 5. Salinity at the depth of the intermediate salinity maximum; only the last three digits are shown. Heavy lines denoted by five digits show the distribution of the salinity on an isopycnal surface corresponding approximately to the midpoint of the regression line in Figure 4.
Figure 6. Distribution below 800 m of salinity, dissolved oxygen, and nutrients at two stations in the core of the saline Arctic Ocean outflow (see Figure 1 for locations).
Figure 7. Dissolved oxygen-salinity characteristics at the depth of the intermediate salinity maximum; the line LR shows their linear regression. The characteristics of the deep Greenland Sea (GSDW) and of the outflow from the Arctic Ocean of The type 1 salinity maximum (203/1579 m) are respectively derived from samples at 3337 m at station 264 and at 1579 m at station203.
Figure 8. Potential temperature-salinity characteristics at a site in the central Greenland Sea during 1982 and 1989, together with the depth of each observation. The inset shows the change in properties from 1982 to 1989, with positive values representing warming and increasing salinity.