1 an Isopycnal View of the Nordic Seas Hydrography with Focus On

An Isopycnal View of the Nordic Seas Hydrography with focus on Properties of the Lofoten Basin.

T. Rossbya*, Vladimir Ozhiginb, Victor Ivshinb, and Sheldon Baconc

aGraduate School of Oceanography, University of Rhode Island, Kingston, RI, 02881 bPolar Research Institute (PINRO), Knipovich St. 6, 183763, Murmansk, Russia cNational Oceanography Centre, European Way, Southhampton, SO14 3ZH, U.K.

Submitted to: Deep Sea Research: December 2, 2008

*Corresponding author. Tel: +1 401 874 6521; fax: +1 401 874 6728. E-mail address: [email protected]

1 Abstract

Few basins in the world exhibit such a wide range of water properties as do the waters of the Nordic Seas with cold fresh waters from the Arctic in the western basins and warm saline waters from the Atlantic in the eastern basins. In this study we present a 50-year hydrographic climatology of the Nordic Seas in terms of depth and temperature on four different specific volume anomaly surfaces. This approach allows us to better distinguish between change due to variations along such surfaces and change due to depth variations of the stratified water column. Depth variations indicate changes in the mass field while property variations along isopycnals give insight into isopycnal advection and mixing, as well as diapycnal processes. We find that the warmest waters on each surface are found in the north, close to where the isopycnal outcrops, a clear indication of downward mixing of the warmer, more saline waters on shallower isopycnals due to convective cooling at the surface.

Using a specific volume anomaly surface that exists all year except in the coldest regions we show i) the role of topography in isolating water masses to either side of the mid- ocean ridges, ii) the great depth of the pycnocline in the center of the Lofoten Basin, deeper than almost anywhere else in the Nordic Seas, and iii) the increase in temperature (and hence salinity) near where it outcrops in the northeast. Close inspection indicates that the deepening results from the expulsion of anticyclonic eddies from the continental escarpment just offshore of the Lofoten Islands and their pooling in the center of the basin. This pooling of warm waters, which leads to an upper ocean anticyclonic density structure, is a key factor to the large heat losses in the Lofoten Basin.

Time series analysis of isopycnal depth in the Lofoten Basin shows it to be rather stable over time with a small but distinct annual cycle superimposed. However, in 1968-1969 it shoaled over 400 m. Almost certainly this resulted from excessive heat loss to the atmosphere during those two very cold winters. This excess loss also shows up as the greatest temperature anomaly in the entire 50-year record of this analysis. Interannual variations in pycnocline depth correlate with the NAO index.

Keywords: thermohaline circulation; water properties; isopycnal; Nordic Seas; Lofoten Basin; heat loss; interannual variability

3 Introduction The Nordic Seas have perhaps been studied longer and more thoroughly than any other ocean. Already in 1887 Mohn published a chart of the circulation of the Norwegian Sea clearly indicating the inflow of warm North Atlantic waters on the eastern side and flow south of Arctic waters in the west. This study was followed a couple of decades later by the pioneering Helland-Hansen and Nansen (1909) monograph on the hydrography of these northern waters. Using water mass analysis (reversing thermometers and accurate salinity titrations) and the dynamic method, the circulation patterns they deduced have stood the test of time impressively well. Even today their figure of salinity in the southern Norwegian Sea and across the Iceland-Faroe Ridge provides a remarkably accurate synthesis of the regional circulation. They detailed the route by which warm North Atlantic waters flowed north through the Norwegian Sea and beyond towards the Barents Sea and Svalbard. A striking aspect about the Helland-Hansen and Nansen study was its emphasis on horizontal distributions. They could do this thanks to the systematic hydrographic surveys throughout the Norwegian and Greenland Seas. Helland-Hansen and Nansen set the standard for subsequent studies including the major atlases of water properties at selected standard depths by Dietrich (1969) and Koltermann and Lüthje (1989).

In recent decades, with increasing awareness that in stratified oceans fluid exchange takes place along isopycnal surfaces, and with increasing computer power to interpolate and plot observations, it has become increasingly common to examine properties as a function of density in order to better resolve where along a section or a surface changes in water properties take place. Various measures of ‘isopycnality’ can be used; these are (with increasing accuracy) surfaces of constant sigma-t, sigmat-θ, specific volume anomaly, and gamma (neutral surface). When plotted as a function of isopycnal one can see precisely where and how changes in water properties take place along a section (e.g. Arhan, 1990; McCartney and Mauritzen, 2001). Parallel sections can be used to construct maps of properties on isopycnal surfaces. Thus, Bower et al. (1985), using temperature and oxygen distributions on two different sigma-theta surfaces, could show how the Gulf Stream acts as a ‘barrier’ to cross-stream exchange on shallow isopycnals and as a

4 ‘blender’ on deeper surfaces. Although such studies require 2-dimensional coverage in the horizontal, charting distributions on isopycnal surfaces can give valuable insight into pathways of spreading and mixing processes.

Our interest in developing an isopycnal surface view of the Nordic Seas was stimulated by the availability of a huge hydrographic database. Well over 300,000 stations have been archived over the last half-century. This allows one examine in considerable detail isopycnal surfaces directly, their depths and their physical properties, and how these vary over time. A change in depth of an isopycnal (surface) implies a change in the density profile and hence pressure field, a change of dynamical consequence, whereas a change in temperature/salinity composition on an isopycnal implies a change in water type. It does not impact the pressure field, but contains valuable information on advection, mixing (or lack thereof), and diapycnal processes. Knowing that the Nordic Seas exhibit strong contrasts both spatially and seasonally, it seemed appropriate to explore whether and how an isopycnal approach might help us to better understand the nature of these patterns and their eventual change over long time.

The Nordic Seas have a basic west-east organization with waters from the Arctic and the North Atlantic flowing south and north along the respective margins. But additional fluid pathways through the Nordic Seas exist thanks to the set of ridges that define the various sub-basins. Thus, the inflow of Atlantic water through the Shetland Channel comprises two branches, one (aka the inner branch of the Norwegian Atlantic Current) that flows along the continental margin north towards the Lofoten Basin, and the outer branch that flows north just west of the Vøring Plateau, and along the Mohn and Knipovich Ridges towards Fram St (Orvik and Niiler, 2002). See Figure 1 for the names and locations of the major ridges and basins. As we will see, the Jan Mayen Ridge, the Mohn Ridge and the Knipovich Ridge play an overarching role in the west-east organization of the Nordic Seas. To the west the Greenland and Iceland Seas are both very cold and fresh. To the east, the Norwegian Sea, which comprises the Norwegian Basin in the south and the Lofoten Basin to its northeast, exhibits quite counterintuitive properties with a deep thermocline in the Lofoten Basin and the warmest, saltiest waters even farther north between the Mohn/Knipovich ridge and the Barents Sea escarpment. In this isopycnal

5 study, we use specific volume anomaly (abbreviated as δ) to examine these properties in some detail.

We begin with a survey of several δ-surfaces in order to establish the mean state of the Nordic Seas from near the surface down across the main pycnocline. We then focus on one of these, the δ =2.1x10-7 m3kg-1 surface, which is the shallowest surface that on average does not outcrop in winter except along the margin of the Greenland Sea and thus remains well-defined throughout most of the year (almost all surfaces will outcrop in the center of the Greenland Sea). We use this surface as the stage to examine the annual cycle and variations on longer time scales. While this means foregoing a detailed discussion of corresponding developments other surfaces, both shallower and deeper, we intend that this isopycnal approach provide a complementary perspective to present-day hydrographic descriptions of the Nordic Seas as well as provide a starting point for more detailed studies in the future.

In the next section we describe the database, the procedures for quality control and preparation of the isopycnal fields used in this study. The following section describes the mean state of four δ-surfaces across the Nordic Seas for the 1951-2000 period. We then use one of these surfaces to document the annual cycle and RMS variability. The following section shows how - given the mean field description - one can determine the condition or state a region at any given time. The examples include Nordic Seas-wide variations, and an example of a very large anomaly or departure from the mean state that took place in the late 1960s. As a further illustration of the methodology we show that the state of the Nordic Seas at the time of the Helland-Hansen and Nansen (1909) study was quite typical of the last half-century. We then discuss some of the more striking findings such as the causes of the deep pycnocline in the Lofoten Basin, which has enormous implications for the heat balance in the Nordic Seas. A brief summary concludes the study.

Data preparation

The data used here combine data (1951-2000) in the ICES archives and Russian data at

6 PINRO in Murmansk, Russia. The database comprises more than 300,000 stations throughout the Nordic Seas from the Iceland-Faroe Ridge in the southwest to the Barents Sea in the northeast. Each station can be used to calculate depth, temperature and salinity of selected isopycnal surfaces. Given pressure, temperature and salinity (P, T and S), δ values were calculated for all measured depths for each hydrographic station. Based on adjacent δ values, density inversions were checked for each hydrographic station. Stations with a density decrease (inversion) of more than 0.005 kg/m3/m were eliminated (deleted). Pressure, temperature, salinity and δ were linearly interpolated between adjacent measurement depths to obtain water properties for desired δ-surfaces for each station in the data set. Four surfaces will be considered: δ = (4.1, 3.1, 2.1, 0.7)x10-7 m3kg-1

-3 (corresponding to ~σt = 27.7, 27.8, 27.9, 28.04 kgm ) and will be referred to as s41, s31, s21 and s07, respectively. For the s21 surface its properties will be labeled D21 (depth), T21 (temperature), and S21 (salinity). To save space salinity will not be plotted since it is effectively a mirror image of temperature on an isopycnal surface. It should be added here that the vertical spacing of measurements on many of the hydrocasts, which were taken primarily for fisheries surveys, is sometimes rather large. This means that linear interpolation between two bottles 200 m apart can lead to an additional estimation error that is beyond the scope of this study to examine in further detail. We rely on the very large volume of data to smooth out most of the scatter due to interpolation. Figure 2 shows the seasonal distribution of hydrographic stations used in this study. Clearly the summer months have the best coverage and the winter months the least. This is particularly true of the open waters of the Norwegian and Greenland Seas. The Barents Sea and all coastal waters have better coverage all-year round.

For each month between 1951-2000, D21, T21, and S21 were interpolated into grid nodes. Grid spacing was 30’ along parallels (30°W to 30°E) and 15’ along meridians (58°N to 80°N). The surface mapping system SURFER 8.01 (Golden Software, Inc) was used with the Triangulation with Linear Interpolation and Kriging methods applied for gridding D, T and S. Of the two methods the former is an exact interpolator (honors data points exactly when the point coincides with the grid node, meaning a coincident point carries a weight of 1.0). It creates a good representation of moderate-sized data sets (250-

7 1000 observations) and does not extrapolate values beyond the range of data. It is especially good with regularly spaced data. When stations are patchy as with our data set this method interpolates data between clusters of stations (over empty areas). This is the main weakness of this method. Kriging is much better for plotting maps from irregularly spaced data. It does not interpolate between data clusters but extrapolate values slightly outside the data area (cluster). This is a much smaller disadvantage with our data set in our opinion.

The gridded fields created at the previous stage are rather “patchy” reflecting station availability in a particular month of a particular year. To get climatic fields of D, T and S, pentadal averaging of the monthly gridded fields was implemented as follows. For the first pentad (1951-1955) January-March gridded fields (of D for example) are available from the previous stage (5 for each month, 15 in total). These 15 gridded fields (or layers) are averaged, i.e. interpolated values are averaged but only for corresponding grid nodes having 3 and more values. Doing so we look for the areas where data clusters overlap. We use 3 values per a grid node to serve as a criterion. This number is a compromise of some sort. Increasing this number from 1 to 5 cuts off areas with sparse data and reduces the coverage for that mean pentadal field. Similar computations are performed for the other 9 pentads (1956-1960,...,1996-2000). The result is one mean field for each pentad.

With 10 pentadal gridded fields (or layers) available we again look for areas with good data overlapping and calculate average values for grid nodes having not less than 4 values. This number is also a compromise. Result is the mean field for January-March. Then computations are made for the other quarters (April-June, July-September and October-December). Exactly the same procedure is applied to construct the 50-year mean fields for T and S. This approach, with double averaging, reduces sensitivity to data clustering and bad data, and produces smooth and solid mean fields. The limitation is the loss of coverage in the north and west in winter and fall (where and when data are sparse and rare).

The mean state and annual cycle

8 The mean state Warm, salty waters from the northeast Atlantic fill the eastern basins and cold fresh waters from the Arctic fill the Greenland and Iceland Seas. Thus the Nordic Seas have primarily a north-south connectivity, unlike the strong east-west orientation of subtropical and tropical waters. The Atlantic and Arctic origins of the Nordic Seas waters leads to a fundamental difference between the eastern and western basins: the North Atlantic waters are stably stratified in temperature, but not in salinity: salinity decreases with increasing depth. The upper layer waters of the Greenland and Iceland Seas, on the other hand, are stably stratified in salinity due to the low salinity of Arctic waters. These patterns and the corresponding T/S-diagrams are well established in the hydrographic literature (e.g. Blindheim and Østerhus, 2005).

In addition to the strong spatial contrasts between basins, the Nordic Seas experience a large annual cycle, especially the warm region that surrenders large amounts of heat in winter such that the surface waters cycle through a wide range of temperatures between summer and winter. This poses a challenge from an isopycnal point of view as low- density waters disappear during the winter months. Thus, an annual mean would exclude a significant portion of the upper ocean and not be either particularly representative or useful. Instead we use the April-June Spring period as the base period, for two reasons: the data coverage is good, and because it follows the winter season, it carries better memory of that active period. Figure 3 shows mean depth of the four δ surfaces for the Spring quarter. Only the deepest two of these exist the entire year (except in the central Greenland Sea); the two lighter surfaces exist all year only where it is deeper than O(100- 200) m depending upon depth of mixed layer nearby in winter. (Rossby et al., 2007a, show a figure of δ at the surface in winter.) Thus, the top surface, s41, exists all year round only along the Norwegian and east Greenland margins. The former reflecting the flow of warm salty water from the Atlantic and the latter the flow of cold fresh water from the Arctic; however, lack of adequate coverage from the Greenland margin precludes including that area in this study. We see two troughs or deepening of the s41 δ- surface, one near 0°E, 68°N reflecting the underlying Vøring Plateau and the other just west of the Barents Sea due to the northward flow of Atlantic waters towards Svalbard.

9 Since this latter trough will show up repeatedly we label it the Svalbard trough. This trough shows up as a tongue of warm water when plotted as temperature at constant depth (cf. Walczowski and Piechura, 2007). The very shallow depth of this surface across the Norwegian, Iceland and Greenland Seas means that this δ surface exists here only seasonally. The next surface (s31) reveals a similar pattern with the Vøring Plateau trough extending farther north and broadly across the Lofoten Basin towards a somewhat more sharply defined Svalbard trough. The broad shallow area that outcrops in winter has been pushed west towards the Iceland and Greenland Seas. The next surface (s21), although only another ~200 m deeper, reveals a structured deepening in the Lofoten Basin stretching from the Lofoten escarpment (the very steep continental slope just offshore of the Lofoten Islands) to the center of the basin. The deepening in the Lofoten Basin means a large pooling of heat in this area compared to anywhere else in the Nordic Seas. The deepest surface, s07, follows s21 fairly closely with its maximum depth in the Lofoten Basin. In the discussion section we will show that this deepening almost certainly results from the aggregation of anticyclonic eddies previously shed at the Lofoten escarpment. This deepest surface also exhibits a deepening farther north in the Svalbard Trough, the cause of which is unknown. The shoaling in the center of the Greenland Sea stands out quite distinctly, as well as another in the central Iceland Sea. As with s21, this surface has a narrow extension into the Barents Sea, deeper on the southern side.

To see how these surfaces look in the vertical Figure 4 shows δ along two lines, one from the southern tip of the Lofoten Islands into the Greenland Sea (the classic Gimsøy section) and the other from the Lofoten Islands north towards Fram St (see Figure 3 for their locations). The data are from the April-June period with inevitable weighting towards May and June due to increased data availability. Note how the s41 surface is limited to close to the continental margin with its greatest slope at ~100 km, the s31 surface has its greatest tilt at about 350-400 km and the s21 surface at 500 km, all of them at about the same depth, 200-300 m. It seems plausible that this horizontal organization reflects the surfacing of progressively deeper isopycnals across the Lofoten Basin. Surfaces deeper than s21 shoal sharply at 500-600 km distance near the axis of the Mohn

10 Ridge, a vertical or frontal alignment that appears to be associated with the outer branch of the Norwegian Atlantic Current flowing north along the ridges. Note that because the Gimsøy section is routed just to the northeast of the deep pycnocline extremum in Figure 3, it doesn’t capture the full deepening of the pycnocline in the center of the Lofoten Basin. The tilt of the isopycnals at the continental slope tilt upwards for shallower isopycnals (δ>25-30) and downwards for the deeper surfaces indicating a reversal of baroclinic shear. The tilt of the shallow surfaces accords with the strong flow along the continental margin while the reversal of slope hints at an intensified flow to the northeast with increasing depth. The Fram St panel indicates a far more gradual shoaling to the north consistent with the fact that it does not cut across any sharp topography. The reverse tilt of the isopycnals in the far north very likely reflects the cross-over of warm waters towards the northern Greenland Sea. The data density for this section is lower and broader than for the repeat Gimsøy section.

A very different view of the water properties obtains when we examine the temperature/salinity properties on these surfaces, Figure 5. We show only temperature since by definition the salinity patterns will be the same (see Rossby et al., 2007b for an example of the corresponding salinity field). The four surfaces exhibit the expected east- west warm-cold contrast reflecting the Atlantic-Arctic origin of the water masses (coverage near Greenland is insufficient for this 50-year mean state description). However, the organization of maximum temperature gradient differs considerably from the pattern of tilt or baroclinity of these surfaces. Thus, the maximum temperature gradients have a similar pattern for all four surfaces and align with the major topographic features of the Nordic Seas: the Jan-Mayen, Mohn and Knipovich Ridges indicating that it is this ridge system that somehow (since the ridge crests are much deeper than these surfaces) isolates the Atlantic and Arctic water masses. Viewing temperature on isopycnals allows one to clearly distinguish between coldness due to shoaling isopycnals and coldness due to water type. These panels point to considerable control by the underlying ridge system.

The four panels reveal another organization, namely that the warmest and saltiest waters

11 of each layer are not found at the Atlantic inflow through the Shetland Channel, but farther north in the Lofoten Basin. Indeed, the warmest waters appear at or near where the surface outcrops in wintertime (recall that these surfaces are biased by more data in the latter months of the Spring season and hence span a larger area than the ones from the bottom of winter). Thus the deeper surfaces exhibit their maximum temperatures farther north. On s41 the warmest (and hence saltiest) water lies in the center of the Lofoten Basin at ~70°N. On s31 the warm anomaly shifts farther north and east, and even more so on the s21 surface where the anomalies are greatest in the northeast in the region of the Svalbard trough and extends east into the Barents Sea trough. The deepest surface brings some features on s21 into sharper relief, but attenuates others. The Greenland minimum in temperature is now more sharply defined, but the pool of spicier (=warmer and saltier, Munk, 1981) waters in the northern Lofoten Basin is broader and weaker. On the other hand spicy water shows up very sharply in the Barents Sea trough where the temperatures are at least 1°C higher than anywhere else.

Since the classical description of water mass transformation as waters drift north is one of cooling and freshening, the former due to heat loss and the latter due to mixing with fresher waters (e.g. Blindheim and Østerhus, 2005), it might seem strange that the waters in Figure 5 appear warmer and saltier in the north, but the explanation is simply a matter of perspective. Surface waters are more saline than waters at depth, thus vertical mixing due to cooling will result in not only cooler, but also fresher surface waters as they are mixed into the underlying layers. At the same time the downward mixing of salty waters results in water type that is saltier than the waters of the same density that have not yet been exposed to the atmosphere, hence their greater spiciness. So while the waters do cool and freshen as they sink, on the new denser isopycnal they appear warmer and saltier than the inflowing North Atlantic waters. The deeper density surfaces outcrop farther north than the shallower ones, hence the corresponding northward shift of spiciness. Use of the label ‘spicy’ seems particularly appropriate here given that the properties of these waters depend upon the downward mixing of warm and salty waters.

The two deepest surfaces and especially the deeper one show a remarkable increase in spiciness in the Barents Sea trough. The s21 surface suggests a source of spicy waters in

12 the Svalbard Trough, a distribution which is also linked to the Barents Sea trough. The s07 surface shoals in the Barents Sea trough (see Figure 3 s07: the middle depth contour in the trough is 200 m) so we cannot exclude the possibility that it outcrops in winter. The increase in spiciness is striking.

Figure 6 shows temperature as a function of δ along the Gimsøy and Fram St sections. These are plotted for late summer (rather than Spring) such that δ=50 (the lightest water included here) exists everywhere. The North Atlantic waters are flowing into the figure at the far right and are cooler and fresher than those in the center of the Lofoten Basin. North of the Mohn Ridge the waters are much colder and fresher on all surfaces. As a result maximum temperatures (and salinities, not shown) at constant δ are found in the central Lofoten Basin. The Fram St section shows clearly the northward shift of maximum spiciness on the deeper surfaces.

The annual cycle The 50-year mean field for each of the four seasons shows only little variation in depth of the s21 surface, Figure 7. The lack of data for the two winter seasons reflects both a lack of data to construct the 50-year mean as well as its disappearance in winter in the Greenland and Iceland Seas (Rossby et al., 2007b). But its localized deepening in the western Lofoten Basin is present all year varying between ~720, 650, 630 and 670 m for the four seasons, deepest in winter shoaling almost 100 m by late summer. The corresponding temperature pattern for this surface, Figure 8, remains quite stable throughout the year with the largest temperatures in the northeastern part of the Lofoten Basin and southern Barents Sea, with perhaps a hint of weakening in the fourth quarter. The emergence of cold and fresh waters in the Greenland Sea in the second quarter reflects the reheating of relatively fresh water, and may also include some melt water from the East Greenland icepack as Spring gets underway.

RMS variability The RMS depth variability of the s21 surface is greatest in the Lofoten Basin with sharp limits towards the Mohn Ridge and Vøring Plateau to its north and south, Figure 9. One also sees an extension of greater variability extending south along the eastern margin of

13 the Jan Mayen Ridge and along the Knipovich Ridge to the north. Significantly, the variability does not appear to reach the Norwegian continental slope except for a weak ridge of larger variability towards the Lofoten escarpment. The variability increases slightly in the meandering Iceland-Faroe Front north of the Faroes (contour >100m), but not along the Norwegian Atlantic Current to its north suggesting that the path of this baroclinic current is rather stable (cf. Nilsen and Nilsen, 2007). These patterns and their amplitudes are quite robust and not significantly dominated by the extreme 1968-1969 winters (next section). Curiously, the standard deviation hints at a local minimum where the depth is a maximum in the center of the Lofoten Basin suggesting that this deep pycnocline is tightly constrained with the variability associated somehow with an expansion, wobbling, and/or aggregation of new anticyclonic eddies (discussed below) in the center of the basin. The corresponding temperature variability on this surface exhibits an altogether different character, Figure 10. It follows the thermal (or property) front rather closely, along the Mohn Ridge and north just west of the Knipovich Ridge. The variability within basins is significantly less, especially in the southeast along the continental escarpment. In the west where the surface is shallow and outcrops in some areas in winter, the variability is noticeably larger. The broad expanse of greater activity in the Lofoten Basin (>0.6°C in the northern half) coincides with the region of greater spiciness. The actual variability in the Norwegian Sea may be somewhat less than indicated because some scatter may result from the low bottle density in many of the fisheries-research stations and hence greater interpolation in the vertical. But the pattern of greater variability over and along topography, which tends to separate water masses, stands out clearly. Close examination shows the T/S variability curving west at the northern end of the Knipovich Ridge axis and a seamount in the Fram St at 78°N. This agrees with the advection of spicy waters to the west in the Fram St north of the Greenland Sea gyre (Figure 5).

Interannual variations

Large-scale departures from the mean The 50-year mean field provides a well-defined framework to explore patterns of variability across the Nordic Seas. This is a potentially huge topic for study that can range

14 from internal variability in the ocean to using the reconstructed fields to evaluate ocean response in numerical simulations using known winds (e.g. Jonsson, 1991; Furevik and Nilsen, 2005). It is beyond the scope of this study to undertake such a comparison here, but we can illustrate by means of two examples when the δ surfaces were much shallower and deeper than normal. Figure 11 shows the anomaly of depth for s21 in summer 1955 (top panel) and summer 1991 (bottom panel). The fields appear noisy in part due to the non-simultaneity of the data used, but primarily due to eddy activity. The state of the Nordic Seas is strikingly different between these two periods, but they are by no means either unique or extreme. Throughout most of the 1960s into the early 1970s the surface was shallower than the 50-year average, and from 1989 deeper than average across the Nordic Seas. A greater depth means a thicker pool of warm water.

The corresponding water property patterns depend not only on advection and mixing, but also convective mixing of saltier waters from above. Since air temperatures tend to be higher (lower) during high (low) NAO intervals, we find significant temperature variations as well. Thus the 1955 shoaling shows the s21 surface to be spicier than average in the Norwegian and Lofoten Basins, and even to the immediate north of Iceland, Figure 12 (top panel). Temperatures are lower close to Greenland, perhaps because the waters (from the Arctic) are fresh at the surface such that convective overturning will freshen the entire layer (recall that figure is from the following summer such that a fresh layer at this density must appear cooler). But the waters may also be fresher due to melt of previous winter ice cover. Conversely, during high NAO periods, Figure 12 (bottom panel), there is less convective cooling and subsurface waters remain more insulated than normal. The eastern basin waters are very close to normal (only slightly cooler than the 50-year average reflecting perhaps the freshening that has been observed in recent decades, Dickson et al., 2002) and only significantly cooler in the area where spicy water production would have taken place; the western basins are warmer than normal.

The 1968-1969 Event In the late 1960s the s21 surface shoaled 400 m in the Lofoten Basin (69-71°N, 0-10°E), Figure 13 (red line). Almost certainly this event was caused by a couple of very bitter

15 winters: The blue line in Figure 13 shows the January-March temperatures for the same region 70-75°N, 0-5°E; the correspondence with the top panel is striking and clearly worth further study. Since the anomaly lasted for several years with a sharp rise and more gradual decay we use data from the best-covered summer months to estimate its magnitude and extent. Figure 14 shows depth of s21 in June 1968. The deep pycnocline in the Lofoten Basin has been eroded away everywhere except near the Norwegian continental margin. All that remains appears to be the flow north of the Slope Current along the western slope of the Vøring Plateau with a hint of its continuation north, and a deep eddy off the Lofoten Islands. Even the Svalbard Trough appears significantly weakened.

The corresponding temperature pattern, Figure 15, shows an increase from 3 to over 4°C over the northern half of the Lofoten Basin. One can, in fact, see a complementary pattern, where the surface shoals, the temperature increases: the 4°C contour traces the ~200m depth contour on the southern side and the Mohn/Knipovich ridge system on the northern side. This presence and spread of >4°C water stands out as the largest increase in spiciness observed in the entire 50-year period. The time series of temperature on this surface in the Lofoten Basin shows a sharp increase followed by a cooling in the early 1970s, Figure 16. This cooling continues and reaches a minimum around 1980 in conjunction with the passage of the Great Salinity Anomaly (Dickson et al., 1988) before recovering to ~0.5°C cooler (and fresher, Dickson et al., 2002) level than during the first two decades.

It is beyond the scope and purpose of this paper to go into a detailed analysis of this event, but we note that by comparing the heat storage in June 1968 to the 50-year mean the Lofoten Basin had ~ 2e9 Jm-2 less thermal energy in the top 1000 m that summer. To achieve such a deficit would require an additional ocean-atmosphere heat flux of > 200 Wm-2 over a 100 day period. Such a large heat flux implies a significantly larger rate of cooling, convective activity and downward mixing of saline waters, hence the striking increase in spiciness (temperature) on this surface when it is reformed in spring. Figure 14 gives further evidence that the build-up of the deep pycnocline must come from the continental escarpment. We return to this is the discussion.

16 The Helland-Hansen period

As a last example of how this 50-year climatology can be used we assess the state of the Norwegian Sea at the time of the Helland-Hansen study, Figure 17. By comparing these panels with Figure 3 one can see that they are quite similar to the 2nd half-century average, and well within the range of variability as discussed in the previous paragraphs. The Iceland-Faroe Front and the outer branch of the NwAC sit in just the right place (top panel) while the temperature front separating the Iceland and Norwegian Seas is co- located with the Jan-Mayen Ridge at 7-8°W (bottom panel).

Discussion

The principal advantage of the x-y-δ instead of the x-y-z framework lies in the ability to probe property variations along isopycnals on the one hand, and how these move or heave in the vertical on the other. Assuming a uniformly stratified fluid, no gradient in water property will appear on an isopycnal, whereas at constant depth large variations will arise across fronts or due to the heaving of the stratified water column. The absence of these on an isopycnal makes it far easier to study patterns due to isopycnal advection and mixing, and diapycnal processes. Fronts and heaving appear in the depth of the isopycnal. We might say that working in the δ-framework leads to a natural orthogonalization of the mass and property fields.

That the pycnocline is deep in the Lofoten Basin has of course been known for a long time, but a good understanding of its dynamics awaits development. Orvik (2004) conjectured that the deepening might be connected to the outer, baroclinic branch of the Norwegian Atlantic Current. And it may well be that the current could not exist without the deep pycnocline in the Lofoten Basin. But Figure 3 shows clearly how the deep pycnocline extends back towards the Lofoten escarpment. Hydrographic sections, such as the one plotted in Figure 18, reveal large anticyclonic eddies in the area west of the Lofoten escarpment. We infer that these eddies break away from the Slope Current and drift west towards the center of the basin where they coalesce with and maintain the permanent deepening of the pycnocline.

17 In addition to the hydrographic evidence for eddies, a recently completed study used acoustically tracked floats at ~150 m depth to study the circulation in the Norwegian Sea. Many floats drifted rapidly north in the Slope Current towards the Lofoten Basin. About 6-7 of the 17 floats approaching the escarpment broke away into the Lofoten Basin revealing an intense eddy activity (Rossby et al., 2008). In contrast, surface drifters do not show this breakaway tendency and instead continue northeast in the Slope Current (Poulain et al., 1996). We cannot say for sure, but it is tempting to speculate that the instability that produces these anticyclonic vortices results in lenses rather than surface- intensified eddies?

One can show that these eddies contain the required heat to maintain an annually- averaged heat loss of ~60 Wm-2 (Isachsen et al., 2007) over a 400x400 km2 area; larger than that of the deep pycnocline and less than the entire Lofoten Basin. (The eddies don't all coalesce, but spread out over a larger area as indicated by the float trajectories.) Given that each eddy has a volume of 1.5x1012 m3 (300m x π x (40 km)2) and that the drop in temperature due to cooling is 2°C, one requires a supply of 24 eddies/year. Much better information on eddy size distribution and rate of production would be needed in order to estimate fluxes better. The corresponding mass flux would be 24 eddies x 1.5x1012 m3 = 1.2 Sv, where 1 Sv(erdrup) = 106 m3s-1. This is a bit more than 1/3 of the 4.4 Sv mass flow in the Slope Current measured at the Svinøy section (Orvik and Skagseth, 2003). These numbers support the conjecture that the deep pycnocline and concomitant heat loss in the Lofoten Basin owes it existence to the steep continental slope off the Lofoten Islands. The reader is referred to Köhl (2007) for an interesting numerical study of the Lofoten Basin eddy, and to Wolfe and Cenedese (2007) for a discussion of some rotating tank experiments to simulate the vortex-generation process.

It was suggested earlier that the increase in spiciness in the north on each of these isopycnals reflected the downward convective mixing of saline waters. The T/S diagram in Figure 19 illustrates the point. Consider the segment AB of a hydrographic cast indicated by the dashed line where it crosses the σt=27.8 (~s31) curve. Wintertime cooling has cooled any overlying layers and point B is just about to be exposed to the atmosphere. Assuming no change in salinity due to precipitation or evaporation, the

18 parcel at B will match particle A on the σt=27.9 curve (~s21) when cooled 0.85°C. But due to its greater salinity its temperature at C will be 1.25°C warmer than unexposed fluid at point A. Since this geometric argument ignores many isopycnal and diapycnal mixing processes, most of which will need study, we can’t say what the trajectory of a cooled fluid parcel will be through the T/S diagram, but the magnitudes seem plausible.

The temperature patterns in Figure 5 suggest that ridges play a significant role in suppressing lateral exchange: mixing takes place within but not between basins, and especially not between the Lofoten Basin and Greenland Sea. But high gradients by themselves tell us little about the actual rate of exchange across the ridge since these can be maintained in the presence of mixing through a rapid supply or advection of new waters. For example, Rossby et al. (2008) report the drift of two floats along the outer branch of the Norwegian Atlantic Current from the Vøring Plateau north and along the Mohn Ridge, a nearly1000 km track in 3-4 months (~9 cms-1), suggesting a significant replenishment of Atlantic waters along the ridges. But the fact that the floats follow the ridges points to a steering role of topography and probable suppression of cross-ridge exchange. A similar process of reduced cross-ridge exchange was reported by Ollitrault and Colin de Verdière (2002): of 26 SOFAR floats deployed at 700 m depth in tight clusters over the western and eastern flanks of the mid-Atlantic Ridge just south of the Azores, only one from each side crossed the ridge. That result was all the more remarkable since no hydrographic signature of suppressed exchange can be seen on corresponding isopycnal surfaces. This isolation of water masses in their respective basins may have relevance to the meridional overturning circulation. It has been noted by many that increased ice melt could threaten the production of dense water in the Nordic Seas. With better knowledge of how ridges ‘operate’ on the overlying water column one will be better able to assess such issues.

But the east-west separation in the Nordic Seas by the Jan Mayen – Mohn – Knipovich Ridge system is by no means complete: Figure 5 shows the penetration of cold (fresh) waters into the eastern basins in the southern Norwegian Basin (just north of the Iceland- Faroe Ridge). On the lighter surfaces these waters are actively stirred into the North Atlantic waters (Rossby et al., 2008), and on the deepest surface (actually slightly deeper

19 than the s07 surface) the waters escape out into the deep North Atlantic through the Faroe-Bank Channel (Søiland et al., 2008). At the same time we see evidence for a flow of Atlantic water in the Greenland Sea at the northern end of the Knipovich Ridge. These are highly structured flow patterns controlled to a major extent by the shape and bathymetry of the Nordic Sea basins.

Summary

The isopycnal analysis employed here enables us to distinguish between dynamical change as measured by changes in depth of an isopycnal surface, and property variations on that surface. In conventional x-y-z framework it can be difficult to distinguish between the two due to the basic stratification of the water column. In this study we take advantage of the large volume of stations in the Nordic Seas to explore the spatial structure of the density field and how it varies, not using the classical T/S diagram, but on selected isopycnal surfaces. A central result of the study shows that the mean fields of depth and property have distinctly different patterns. This applies to their patterns of variability as well. We find that the mid-Atlantic ridges serve as a dynamical (in the sense that the ridges, although deep, constrain fluid motion throughout the entire water column) barrier to interbasin exchange. Thus the principal property fronts coincide with topographic features, not necessarily with the path of currents. This is not so obvious when we examine fields as a function of depth because what looks like a large property change across a front reflects different stratification profiles to either side.

Once the basic or mean state has been determined, it can serve very effectively as a basis for exploring anomalies and their characteristics in the Nordic Seas. This is where the large database becomes very useful. From one year to the next one can observe very substantial variations in both isopycnal layer depth and temperature/salinity, and these can persist for several years. The 50-year database could be used to better understand the response of the Nordic Seas to variations in wind stress on the one hand, and severity of winters on the other. As an example of the latter, we show how the deep pycnocline in the Lofoten Basin shoals due to convective heat loss and in so doing redistributes its upper layer salt onto deeper layers, thereby increasing their spiciness. But variations in

20 windstress and rates of production of anticyclonic eddies can also contribute to the depth of the Lofoten pycnocline. In short, this isopycnal approach allows one to more clearly distinguish between the effects of mechanical and thermodynamic forcing. These products, or reanalyses to borrow a phrase from meteorology, may prove useful for testing and verifying numerical models and simulations of the Nordic Seas circulation.

Acknowledgements

The genesis and much of this work took place as a charge from the ICES Working Group on Ocean Hydrography during which many of the ideas presented here first took shape. It is a true honor for us to acknowledge the working group for its suggestions and encouragement. We particularly thank Drs. Kjell Arne Mork, Øysten Skagseth, Henrik Søiland and Kjell Arild Orvik for many helpful discussions. We thank Mr. Linus Magnusson of the Meteorological Institute, University of Stockholm for the air temperature data in figure 13. We also thank our institutions, the Polar Research Institute of Marine Fisheries and Oceanography, and the University of Rhode Island for their support. We thank ICES for its assistance in making the large ICES hydrographic database available for this study.

21 References

Arhan, M., 1990. The North Atlantic Current and Subarctic Intermediate Water. Journal of Marine Research 48, 109-144.

Blindheim, J., Østerhus, S., 2005. The Nordic Seas, Main Oceanographic Features, pp. 11-37. In: Drange, H., Dokken, T., Furevik, T., Gerdes, R., Berger, W. (Eds.), The Nordic Seas: An Integrated Perspective. Geophysical Monograph 158, American Geophysical Union, Washington, DC, 366 pp.

Bower, A., Rossby, H.T., Lillibridge, J., 1985. The Gulf Stream - barrier or blender? Journal of Physical Oceanography 15 (1), 24-32.

Dickson, R.R., Meincke, J., Malmberg, S.A., Lee, A.J., 1988. The ‘Great Salinity Anomaly’ in the northern North Atlantic 1968–1982. Progress in Oceanography 20, 103– 151.

Dickson, R.R., Yashayaev, I., Meincke, J., Turrell, W., Dye, S., Holfort, J., 2002. Rapid freshening of the deep North Atlantic over the past four decades. Nature 416, 832-837.

Dietrich, G., 1969. Atlas of the hydrography of the northern North Atlantic Ocean. Conseil International pour l’Exploration de la Mer. Charlottenlund Slot, Danemark. 144 pp.

Furevik, T., Nilsen, J.E., 2005. Large-scale atmospheric circulation variability and its impacts on the Nordic Seas ocean climate – A Review, pp. 105-136. In: Drange, H., Dokken, T., Furevik, T., Gerdes, R., Berger, W. (Eds.), The Nordic Seas: An Integrated Perspective. Geophysical Monograph 158, American Geophysical Union, Washington, DC, 366 pp.

22 Helland-Hansen, B., Nansen, F., 1909. The Norwegian Sea: Its Physical Oceanography based on Norwegian Researches 1900-1904. In: Report on Norwegian fishery and marine investigations, vol. 2. Bergen, Norway, 390 pp + 25 plates.

Isachsen, P. E., C. Mauritzen, and H. Svendsen, 2007. Dense water formation in the Nordic Seas diagnosed from sea surface buoyancy fluxes. Deep-Ses Research-I,54, 22- 41.

Jonsson, S., 1991. Seasonal and interannual variability of wind stress curl over the Nordic Seas. Journal of Geophysical Research 96 (C2), 2649-2659.

Koltermann, K.P., Lüthje, H., 1989. Hydrographischer Atlas der Grönland und Nördlichen Norwegischen See (1979-1987). Deutsches Hydrographisches Institut, Hamburg. 274 pp.

Köhl, A., 2007. Generation and stability of a quasi-permanent vortex in the Lofoten Basin. Journal of Physical Oceanography 37, 2637-2651.

McCartney, M.S., Mauritzen, C., 2001. On the origin of the warm water inflow to the Nordic Seas. Progress in Oceanography 51, 125-214.

Mohn, H., 1887. The Norwegian North Atlantic Expedition 1876-1878. Volume 17. The North Ocean, its depths, temperature and circulation. Christiania. 212 pp.

Munk, W., 1981. Internal waves and small-scale processes. In: Warren, B.A., Wunsch, C., (Eds.), Evolution of Physical Oceanography, MIT Press, Cambridge, Massachusetts, pp. 264–291.

Nilsen, J.E.Ø., Nilsen, F., 2007. The Atlantic water flow along the Vøring Plateau: Detecting frontal structures in oceanic station time series. Deep-Sea Research I, 54, 297- 319.

Ollitrault, M., Colin de Verdière, A., 2002. SOFAR Floats Reveal Midlatitude Intermediate North Atlantic General Circulation. Part I: A Lagrangian Descriptive View. Journal of Physical Oceanography 32, 2020-2033.

Orvik, K.A., Niiler, P., 2002. Major pathways of Atlantic water in the northern North Atlantic and Nordic Seas toward Arctic. Geophysical Research Letters, 29, 1896, doi: 10.1029/2002GL015002.

Orvik, K.A., Skagseth, Ø., 2003. Monitoring the Norwegian Atlantic slope current using a single moored current meter. Continental Shelf Research 23, 159-176.

Orvik, K.A., 2004. The deepening of the Atlantic water in the Lofoten Basin of the Norwegian Sea, demonstrated by using an active reduced gravity model. Geophysical Research Letters 31, L01306, doi:10.1029/2003GL018687, 2004.

Poulain, P.-M., A. Warn-Varnas, and P. P. Niiler, 1996. Near-surface circulation of the Nordic Seas as measured by Lagrangian drifters. Journal Geophysical Research, 101,18,237-18,258.

Rossby, T., Ozhighin, V., Ivshin, V., Bacon, S., 2007a. An Isopycnal Analysis of the Nordic Seas Hydrography. In: Report of the Working Group on Oceanic Hydrography. ICES CM 2007/OCC:05, 12-31.

Rossby, T., Ozhighin, V., Ivshin, V., Bacon, S., 2007b. Isopycnal analysis of near-surface waters in the Norwegian-Barents Sea region. Clivar Exchanges 13, 2-4.

Rossby, T., Prater, M., Søiland, H., 2008. Pathways of inflow and dispersion of warm waters in the Nordic Seas. Journal of Geophysical Research. Submitted.

Søiland, H., Prater, M., Rossby, T., 2008. Rigid topographic control of currents in the

24 Nordic Seas. Geophysical Research Letters 35, L18607, doi:10.1029/2008GL034846.

Walczowski, W., Piechura, J., 2007. Pathways of the Greenland Sea warming. Geophysical Research Lettters 34, L10608, doi:10.1029/2007GL029974.

Wolfe, C.L., Cenedese, C., 2006. Laboratory experiments on eddy generation by a buoyant coastal current flowing over variable topography. Jornal of Physical Oceanography 36, 395-411.

Figure 1. The bathymetry of the Nordic Seas. Depth contours at 200, 500 m (brown), 1000, 2000, 3000 m (blue). (KR = Knipovich Ridge, MR = Mohn Ridge, CR = ‘connecting ridge’, JMR = Jan Mayen Ridge, GS = Greenland Sea, BS = Barents Sea, LB = Lofoten Basin, IS = Iceland Sea, VP = Vøring Plateau, NS = Norwegian Sea)

Figure 2. Hydrographic data for the period 1950-2000 grouped according to season.

Figure 3. Mean depth of δ surfaces s41, s31, s21 and s07 in Spring (April – June) left to right, top to bottom. The two black lines denote the axes of the Mohn and Knipovich Ridges and the two blue ones indicate the Gimsøy and Fram sections shown in Figures 4 and 6.

Figure 4. Specific volume anomaly (δ) along two sections as indicated in Figure 3. Top panel: Gimsøy section from the Lofoten Islands (68.4°N, 14.0°E) into the Greenland Sea. Bottom panel: Lofoten-Fram St. section. Both 50 year averages for April-June period.

Figure 5. Mean temperature on s41, s31, s21, and s07 in Spring (April – June). The two black lines denote the axes of the Mohn and Knipovich Ridges.

Figure 6. Fifty-year average temperature as a function of δ in August-September for the Gimsøy and Fram St sections indicated in Figure 3.

Figure 7. Seasonal variation in depth of the s21 surface. Although the surface outcrops in the Greenland/Iceland Sea in winter, it exhibits otherwise very little seasonal variation the maximum depth in the Lofoten Basin shoaling only.

Figure 8. Temperature on the s21 surface.

Figure 9. Spatial pattern of s21 depth variability from the 50-year record. Contour interval = 25 m.

Figure 10. Temperature variability on the s21 surface. Contour interval = 0.2°C.

Figure 11. Top panel: Deviation of the s21 surface from its 50-year mean during summer half-year of 1955. Other years when the s21 was broadly shallower than normal include 1956, and most of the 1960s through 1972. These were generally low NAO index years. Bottom panel: Deviation of the s21 surface during the summer half-year of 1991. Other years when the s21 was broadly deeper than normal include 1975, and the 1990s beginning 1989. These were mostly high NAO index years.

Figure 12. Top: Temperature anomaly on the s21 surface in summer 1955. Bottom: Corresponding temperature anomaly on the s21 surface for the summer 1991.

-200 -300 -400 m -500 -600 Depth, -700 -800 -900

2.0 °C 1.0 0.0 -1.0 -2.0 -3.0

temperature, temperature, -4.0 -5.0 Air -6.0 -7.0 1950 1960 1970 1980 1990 2000 Year

Figure 13. Fifty-year time-series of s21 depth in the central part of the Lofoten Basin (69- 71°N, 0-10°E) in June-August (upper panel) and January-March air temperature for the corresponding region (65-75°N, 0-5°E) (bottom panel). Heavy lines are their 5-year running means.

Figure 14. Depth of s21 in June 1968. The deep pycnocline in the Lofoten Basin does not fill the basin as it normally does.

Figure 15. Temperature on the s21 surface in June 1968.

40 4.5

°C 4.0

3.5

3.0

2.5

Temperature, Temperature, 2.0

1.5 1950 1960 1970 1980 1990 2000 Year

Figure 16. Fifty-year time-series of temperature on the s21 surface in the Lofoten Basin (69-71°N, 0-10°E) in June-August (thin line) and its 5-year running means (heavy line).

Figure 17. Depth (top panel) and temperature (bottom panel) of the s21 surface in the Norwegian Sea at the time of the Helland-Hansen (1909) study.

Figure 18. Temperature along a standard section at 69°20’N in July 2000. Note the deep thermal structure at about 9°E. Many other examples of such structures exist in the hydrographic data base. Often the isopycnals exhibit a reverse tilt at shallow depths such that the maximum swirl velocity may not be at the surface.

Figure 19. TS plotting diagram illustrating ‘increase in temperature’ due to cooling.