A proposal for the NERC ARCICE programme by
Sheldon Bacon (Southampton Oceanography Centre) and
Peter Wadhams (Scott Polar Research Institute, University of Cambridge)

The Nordic Seas have the longest history of continuous study of any part of the World Ocean. Ice conditions were first described by Scoresby (1815, 1820), and measurements describing the main circulation features were made in the late 19th and early 20th centuries in the expeditions described in publications by Helland—Hansen and Nansen (1909) and Kiilerich (1945); features such as the northwards flux of saline Atlantic water in the east, the southward flux of cold, fresh polar waters in the west, and at depth, the doming of the Greenland Sea gyre, were described. In spite of this long history and of recent technological advances (satellites, acoustic tomography, for example), there remain many gaps in our knowledge of the three-dimensional oceanic circulation in the Nordic Seas and of the role played by ice, freshwater, salt and heat fluxes in the maintenance of the local oceanic climate. Furthermore, we are only just beginning to learn of the potential of the Nordic Seas region to influence the global oceanic and atmospheric climate. In this proposal we address elements of the local, the broader regional (European) and global aspects of the area’s influence on climate.

The Nordic Seas are crucial in the global oceanic vertical circulation. The Greenland Sea is one of the few sites in the world where open—ocean deep convection creates new deep waters; and the Nordic Seas are the source of the intermediate waters which overflow the Greenland—Scotland sill to contribute to the formation of North Atlantic Deep Water, the lower limb of the ‘Conveyor Belt’ in the Atlantic. We intend to study the influence of sea ice on the magnitude and variability of the vertical ocean fluxes generated by these mechanisms. By ‘vertical fluxes’ we mean the local vertical exchange of water properties in the Greenland Sea by convection, both to greater and to lesser depths; the transfer from intermediate depths in the Nordic Seas of Arctic Intermediate Water (AIW) to great depth in the North Atlantic via overflows between Greenland and Scotland; and the Atlantic thermohaline circulation as a whole, where warm salty water travels north at the surface, to return southwards at depth after cooling and freshening in the Nordic Seas. By studying the role of sea ice we may infer the effect of climate change both on sea ice production and on oceanic circulation, and their effects in turn on the maintenance of the climate of Europe.

Aims: We will investigate three essential areas of the climatic impact of Arctic sea ice. These are:

These aims are strongly coupled: the freshwater flux through Fram Strait plays a role in determining the stability of the Greenland Sea surface water, so that deep convection events in the Greenland Sea Gyre, and the intensity of more widely spread shallow convection, are not purely dependent on local factors. We will address these questions via ship—based, remotely—sensed, modelling and historical data studies. This work will build upon the achievements of the Greenland Sea Project of AOSB and the European Subpolar Ocean Programme (ESOP) of the EU in the Greenland Sea (Wadhams, 1996), and of the World Ocean Circulation Experiment (WOCE) in the northern North Atlantic.

Objectives: These are the specific objectives of this program:

O1 maps to A1, O2 to A2, and both O1 and O2 to A3. O1 requires ship-based measurements during the winter, supplemented by buoy and satellite data. For this we propose the charter of an ice-capable ship such as the Jan Mayen for 12 days. O2 requires ship-based measurements during the summer supplemented by (different) satellite data. For this we propose the use of the Ross for 20 days. The winter and summer measurement programs and their detailed motivations are described separately below, followed by the requirement for the analysis of seasonality enabled by a two-season measurement program. Active links to other proposals within ARCICE, and to other external programs, are described within the text.

Motivation:

Understanding of the role of local sea ice production in Greenland Sea convection was the chief aim of ESOP-1, and progress in achieving it has been documented in the papers of the ESOP-1 report (Wadhams, Wilkinson and Wells, 1996, shortly to appear as a special issue of Deep-Sea Res.). Much remains to be done in the way of detailed in situ measurements, because during the ESOP-1 period the Odden ice tongue failed to develop in 1994 and 1995, so that 1993 was the main experimental period (Wadhams et al., 1996). Briefly, the interactions occur within the range of influence of the Jan Mayen Polar Current (JMPC), which diverts eastward from the East Greenland Current (EGC) around 72-73°N. Analysis of satellite imagery shows that during most winters the polar pack ice moving southward in the EGC continues southward, while the polar surface water of the JMPC acquires a local ice cover which forms in December and typically lasts until late April. The ice forms initially as slicks of frazil ice (a suspension of ice crystals in water) which freeze together into cakes and then acquire raised rims from the pumping action of wave-induced relative motion. This is pancake ice; the cakes can reach diameters of 2-3 m with thicknesses up to 70 cm, more typically 20-40 cm. Pancake icefields are characteristic of regions which experience intense wave action, which prevents the ice from maturing into a continuous sheet. The pancake icefield which covers much of the surface area of the JMPC forms a tongue-shaped feature called Odden. Daily changes in shape occur; the tongue may lose contact with the ice of the EGC forming an island, or it may merge into the EGC to form a bulge. When the tongue is well defined, the open water separating it from the EGC is called Nordbukta (fig. 1), although it is possible for Nordbukta to become completely ice-covered.

The water in the JMPC cools in early winter towards the freezing point, and when the cold atmosphere (which tends to be cyclically cold, with cold outbreaks from Greenland alternating with warmer air mass penetrations from the Greenland Sea) generates an ice cover, most of the salt from ice formation is rejected into the water column (frazil ice has a salinity of about 13 psu, while young pancakes are typically 6-8psu). The salt flux can occur rapidly, and continues with continued cooling, since a frazil-pancake icefield allows a high ocean-atmosphere heat flux so that the ice growth rate is little reduced as the ice grows thicker. Depending on the initial salinity of the surface water (which partly depends on the amount of fresh water present from ice melt during the summer) the added salt may be sufficient to trigger convective plume activity. Model studies (eg Backhaus and Kämpf, 1996) differ on whether a single such freezing event can produce penetrative convection, or whether a series of cooling — freezing — ice removal — further freezing sequences are needed. It is noteworthy that on the few occasions when apparent convective structures have actually been observed, it has been in Nordbukta rather than Odden. It is therefore possible that pancake ice needs to form preferentially on one side of Odden (i.e. the NW side, fringing Nordbukta) and be transported across to the SE side to melt, giving a net salt flux into Nordbukta and a net freshwater flux at the outer ice edge. This is what was suggested by the tracks of the three GPS/Argos buoys released during the 1997 Jan Mayen experiment.

The unsolved questions which can be resolved by a further field season of careful ice-ocean work supported by remote sensing studies are:

1. How are the net salt rejection and the salt flux distributed geographically and temporally over the Odden? Is there a region where the flux over the winter is definitely positive?

2. What are the maximum values reached by the salt flux? Does it occur in pulses separated by melt? Do these episodes add enough density to trigger convection to the observed depth?

3. Can we actually see convective plumes forming? What are their dimensions, how long do they take to develop, and where in the Odden/Nordbukta system are they found?

4. Do they leave physical evidence in remote sensing imagery such that their presence, once measured, can at other times be inferred?

5. How is the depth achieved by convection related to total ice production during the winter? Is it entirely due to such local factors, or does the fresh water budget of the EGC play a vital role?

6. Given predicted rates of regional warming, will Odden cease to exist, and if so will convection also cease?

7. Has the failure of convection to renew the Greenland Sea Deep Water in recent decades also corresponded to a reduction in the contribution made by convection to AIW?

Measurements and techniques:

During an Odden event, in early March 2000, we plan to carry out a cruise into Odden to map the ice properties, and to obtain a physical and quantitative understanding of the ice-ocean interaction process and the resulting salt flux and convective activity. The James Clark Ross will not be available so a ship will need to be chartered for this purpose. We have identified RV Jan Mayen of the University of Tromsø as a very suitable vessel for this work. She was used in March 1997 during the ESOP-2 EU MAST programme. She is ice-strengthened, with a stern slipway, handling gear for heavy equipment, and on-board cold rooms, and comes with a CTD and technician to operate it. Experience showed that she can work effectively in the worst of weather, sea and ice conditions, with 24-hour service from the deck crew. Since only one day of transit time is required to reach Odden from Tromsø, we calculate that a 12-day charter will be adequate for the work proposed.

The cruise plan will depend on the shape and extent of Odden at the time, but will comprise a grid of ice and ocean (CTD) stations covering Odden and the open water region of Nordbukta to the north-west of it (fig. 1). The stations will be arranged such that, as far as is possible, they lie under the daily ascending and descending SAR swaths of ERS-2 / Envisat transits. The stations will cover the following work:

1. Direct sampling of locally formed frazil-pancake ice, by lifting pancakes on board using a steel basket device modelled on that of the Alfred Wegener Institute, and recovering frazil by a sampling pipe with a plankton mesh end. We will make use of the cold room facilities for ice fabric (thin section) analysis and sectioning of pancake samples for salinity and subsequent Oxygen—18 isotope analyses;

2. Deployment of 6 SPRI-built Pancake GPS/Argos buoys to map ice motion within Odden and hence determine ice source and sink regions. In most parts of the Arctic sea ice motion can be mapped by overlaying sequential SAR, AVHRR or passive microwave images and using cross-correlation techniques to map velocity vectors. However, this is not possible in Odden because of the way in which frazil-pancake slicks dissolve and re-form in a diffuse and fast-changing icefield where the size of ice elements is much less than the resolving power of the satellite sensor. The buoys will map the field of ice motion, as well as wind speed and direction and air temperature and SST. They performed very successfully in 1997 (fig. 2), with the GPS navigation option giving much more frequent and accurate fixes than ARGOS alone (fig. 3). The buoys were developed by J. Wilkinson (SPRI) and D. Meldrum (DML) under NERC grant GR2 / 02692 to SPRI. The new set will incorporate design improvements based on experience with the set of 3 deployed as a pilot study in 1997.

3. Direct detection of convective plumes. Three methods will be used, besides closely spaced CTDs wherever evidence of convective activity is seen. These are: (i) Acoustic means, deploying a source from a float and running the ship along a line a few hundred metres away to act as a "screen" for acoustic Schlieren imaging, a technique pioneered in ESOP by B. J. Uscinski, DAMTP Cambridge. The work would be done by him using existing equipment; (ii) Towing a T/S chain in open water (Nordbukta), a means by which Valdivia mapped convective plume activity in a 1994 ESOP cruise. The chain is owned by DERA, was upgraded by ESOP funds, and is being made available by DERA to the UK academic community; (iii) Deployment of Autosub to run at about 100 m under Odden with a T/S along-track package, a suitable depth for detecting the most vigorous activity of incipient convection according to the model of Backhaus and Kämpf (1996). (Subject of a separate submission to the Autosub Science Mission thematic programme);

4. A CTD survey of the region. CTD casts will be carried out simultaneously with ice sampling activities, with a basic set to the bottom and others to 2000 m. Near-surface CTDs will be done from the ship’s Zodiac using a Seabird portable CTD, to assess effects of recent pancake ice melt at locations undisturbed by the ship.

In support of the shipborne programme, we will use ERS-2 / Envisat SAR and SSM/I passive microwave imagery from the Odden for the winters of 1998-9, 1999-2000 and 2000-1. PW is an ESA PI for SAR and acquires data free of charge. SAR data from the cruise period will be validated using observations from the ship (and data from possible concurrent C-130 overflights carried out by the Danish Met. Inst.) to yield a relationship between ice type and SAR brightness and texture, and to yield relationships between observed pancake concentrations and apparent concentration given by the SSM/I bootstrap algorithm. The satellite data can then be used to infer the geographical distribution of thickness, salinities and ages of young ice types over the Odden (as in Wadhams et al., 1996). Combining these data with the buoy tracks, and using the whole winter record, we will infer the spatial and temporal distribution of salt fluxes into the surface waters. Spectral analysis of SAR imagery from the Odden ice edge region will yield the relationship between storm events and the temporary disappearances of Odden; and also independent estimates of ice thickness using the change in wave dispersion, a technique evolved by Wadhams and Holt (1991) and refined by Wadhams, Parmiggiani and Tadross (1995). The interannual variability of the role of Odden will be assessed by using passive microwave data from past years (1973 onwards) to examine Odden growth and decay rates and to relate the development of Odden to wind and temperature fields, and to the fresh water input from Fram Strait estimated from satellite, surface and sub-surface measurements.

In a complementary bid to the Autosub thematic programme we are asking for 4 days of additional Jan Mayen ship time for deployment and recovery of Autosub: the ice thickness profiling and water structure profiling that can be carried out from this vehicle will greatly enhance the value of the winter programme.

Motivation:

In the Atlantic, the global thermohaline circulation is manifested as north-going surface waters which sink in the Nordic Seas and return southwards, after overflowing the Greenland—Scotland Ridge, as deep water. Known colloquially as the ‘Conveyor Belt’, it keeps Europe 5 to 8 ûC warmer (Broecker, 1991) than it would be were the Conveyor Belt to shut down. Previous studies (eg Dickson, Gmitrowicz and Watson, 1990) have suggested that there exists no long-term variability in the outflow of deep water from the Nordic Seas to the Atlantic. In fact, over the last four decades outflow has approximately doubled and then halved again from the northern source of deep water for the Conveyor Belt (Bacon, 1997a&b; Bacon, 1998; McCartney et al., 1998). Furthermore, this variability is forced by variability in polar atmospheric air temperature, which is driven by high-latitude atmospheric processes related to the North Atlantic Oscillation. We are now seeing that shallow processes can respond to atmospheric forcing in such a way that the Nordic Seas exert important control on the variability of the Conveyor Belt. Understanding the present modes of variability of the Conveyor Belt’s sources is crucial for all climate studies.

Bacon (1998), from which we take fig. 4, shows the 40-year history of the DWBC variability in the northern North Atlantic derived from 22 hydrographic sections measured between 1955 and 1997, in the vicinity of Cape Farewell, Greenland. The vertical current shear near the bottom here is high, so a large part of the flow can be measured with the shear derived from the hydrographic data. There is no significant probability that this signal could result from random processes, nor is it possible that processes south of the Greenland—Scotland sill could be responsible. We look to the Nordic Seas for the source of the variability. Most of the water which flows over the Greenland—Scotland sill is AIW. Several models describe its formation: by winter heat loss to the atmosphere in the centre of the Iceland and Greenland Sea Gyres (eg Swift and Aagaard, 1981); in the east of the Nordic Seas (Mauritzen, 1996) over the Norwegian Atlantic Current; or by isopycnal mixing (Strass et al., 1993) in the west, in the eastern boundary of the EGC (where flow from the Arctic via Fram Strait is implicated). Rudels (1995) suggests that brine rejection by freezing on the Arctic/Barents Sea shelves may be involved, by operating on the loop of Atlantic waters which enter the Arctic to return, modified, through Fram Strait. Thus virtually the entire area of the upper waters of the Nordic Seas, with the exception of the waters of polar origin in the coastal side of the EGC, has been implicated in the formation of AIW.

The winter mean air temperature over the Nordic Seas, represented by measurements from Jan Mayen Island, ideally placed at the junction of the ridges separating the Iceland, Greenland and Norwegian Basins, is a likely link in the causal chain. Fig. 5 shows air temperature history and (inset) the overflow flux plotted against air temperature. Most of the latter points divide into two groups: low air temperature associated with high outflow, and vice-versa. This strongly suggests that the air temperature sets AIW density and/or layer thickness, which in turn cause the observed changes in outflow flux, and in the sense predicted by hydraulic exchange theory. Atmospheric temperature and AIW temperature are connected via Bönisch et al. (1997) plate 1 (central Greenland Sea temperature, 1952—1995), where the pattern of fig. 4 is repeated: in the 1950s—60s, AIW is warm (low outflow), in the 1970s—80s cold (high outflow), and in the 1990s, strongly warming (low outflow again). However some points (fig. 5 inset) fall outside this relationship. They are of very low air temperatures and of moderate to low outflow flux, and originate in 1966/7. The unusually low air temperatures of the mid-1960s are associated with very high ice cover (Mysak et al., 1990) — this event is cited (Häkkinen, 1993) as the root cause of the Great Salinity Anomaly (Dickson et al., 1988) — so it appears that the ice may have insulated the AIW formation region from the extreme cold of the air, resulting in the application of a 'different set of rules': low air temperature / low outflow conditions.

We intend, therefore, to study the production of AIW and its variability as it responds to (a) surface haline forcing, meaning the production of ice and the rejection of salt into the upper layers of the Greenland and Iceland Seas, and the melting of that ice, which may be local or non-local, and the advection of ice from elsewhere, and (b) surface heat flux, and their interaction, on both interannual and decadal timescales. We believe ice cover to be implicated in the ‘normal’ variability scenario, whereby the overflow of AIW into the Atlantic responds quasi-linearly to Nordic Seas air temperature variability, and we believe ice cover to be a crucial control in the case where the overflow practically ‘stops’ due to insulation by ice of the AIW production regions from the atmosphere. We need to investigate the precise mechanisms involved, in particular the involvement of the Odden, which is formed over the low-salinity JMPC, the eastward extension of the EGC which forms the south side of the Greenland Sea Gyre. The ice formed rejects salt into the layer below, and the Odden has been much reduced in area during the last two decades. Weakening of the Odden reduces the salt rejection into the waters beneath, which alters their density contrast with adjecent water masses, which changes the nature of the mixing and/or water mass production regimes in the part of the Nordic Seas. Since the JMPC operates at the junction of the Iceland, Greenland and Norwegian Seas, the potential for change here to travel throughout the system is obvious.

We stress at this point that this work is quite novel: the discovery that the overflows have varied in the recent past (decades) is quite counter to the prevailing wisdom, whereby the overflows themselves do not vary on timescales longer than a fortnight because the production of AIW is so widely physically and geographically based that it can (supposedly) withstand such climatic variability as has been observed thus far. At the same time, there is a growing body of both measured evidence from Greenland ice cores and Atlantic sediment cores, and related modelling (eg, Bond et al., 1993; Rahmstorf, 1995) showing the changes on millenial (and down to decadal) timescales of air and sea temperatures and circulation patterns, and associated ice cover, at European latitudes, all reflected in very great changes in the thermohaline structure of the oceans.

In the light of the foregoing, we propose measurements which will (a) clarify the summer circulation of the Iceland and Greenland Seas, (b) monitor, using novel satellite measurements and fore-/hind-casting techniques, the barotropic circulation and ice fluxes there, and so (c) resolve the issue of the cause of the observed overflow variability. The measurements and techniques are detailed next.

Measurements and Techniques:

We wish to use half of a cruise (20 days) on the RRS James Clark Ross in the summer to make detailed hydrographic measurements (CTD, lowered acoustic Doppler profiler, chemical tracers) along sub-satellite tracks which form the boundary of a box enclosing parts of the Greenland and Iceland Seas (see fig. 6). The tracks will be chosen to lie on (a) the upstream Denmark Strait, (b) from Iceland through the Iceland Sea, across the Jan Mayen Current to the central Greenland Sea, and (c) from there to the Greenland coast across the East Greenland Current. ERS-2 ground tracks lie roughly parallel (ascending) to the Greenland coast, and perpendicular (descending) to the major current systems (note that track selection will depend on cruise timing, given the 35-day repeat orbit of ERS-2). By using the hydrographic measurements in conjunction with inverse methods, we will generate the absolute circulation through the box, by imposing net flux constraints as appropriate (Bacon, 1994, 1997), with surface heat fluxes and Ekman fluxes estimated independently via meteorological measurements (ARCICE: Mobbs et al. proposal). With an approximate track length of 2200 km, we estimate 6.5 days steaming time at 8 knots (making some allowance for reduced speed at night / in ice). In order adequately to resolve the Rossby radius, we require stations every 30 km: 75 stations at about 6 stations per day require 12.5 days station time. Total 19 days, plus one day contingency, equals 20 days.

Having obtained one estimate of the absolute circulation, we will use the sea surface height variability signal along the ship track obtained from the ERS2 / Envisat satellite altimeters to estimate the absolute variability in the circulation through the box (again informed by inverse-type constraints to minimise error) after the manner of Challenor et al. (1996). Sea surface height data under the winter ice in the EGC will be obtained through our collaboration with S. Laxon (UCL-MSSL) by the new method of Laxon (ARCICE: Laxon proposal) so the full annual circulation cycle will be determined for the region. We will estimate the seasonal variability of baroclinic velocity and properties (a) through the more limited winter measurement program proposed for the Odden / Greenland Sea gyre region, and (b) through study of historical data to represent the mean upper ocean seasonal signal in the region. Ice mass fluxes will be estimated using ice thickness from sonar (T. Vinje) and altimeter (S. Laxon) data and ice velocity (R. Kwok).

We will use the resulting circulation and flux measurements to test the relevant existing models of AIW formation for water mass type and quantity; we will compute the sensitivity of the AIW formation rate in the measurement region to changes in surface fluxes of heat, salt, ice and freshwater by the various mechanisms available; and we will apply the results to the study of historical ice and hydrographic data with a view to identifying more precisely the processes which link Jan Mayen air temperature variability to Denmark Strait Overflow variability (Bacon, 1998). Measurement results will be compared with results obtained from the OCCAM GCM (ARCICE: Webb proposal) and the NPS GCM through collaboration with R. Tokmakian (NPS, Monterey).

Swift and Aagaard (1981) established some basic terminology for the Nordic Seas, particularly the Arctic domain, the transitional region separated in the west by the Polar front from the Polar domain, and in the east by the Arctic front from the Atlantic domain. The vertical stability in the upper waters of the Arctic domain is sufficiently low that winter buoyancy removal at the surface (direct heat loss and/or salt injection by ice formation) can produce dense waters to descend to intermediate or (in the Greenland Sea) greater depths. In their study of the Iceland Sea, the largest over-winter change in water mass volume occurs in the upper AIW. How does it happen that there is no seasonal pulsation in the Denmark Strait Overflow (Dickson et al., 1990 etc)? Furthermore, how significant is the seasonal variation of AIW volume over the Arctic domain compared with the overflow flux proper? Posing this latter question differently, from where does the AIW which overflows come, in such quantity and with such (relative) seasonal invariance? In the spirit of Swift and Aagaard (1981), rather than rely on a 'static' seasonal water-mass conversion rate to estimate the overflow, we will use the summer closed box of sections, including the Greenland Sea as well as the Iceland Sea, extrapolated to winter, to estimate conversion and transport (export) rates. To what extent do the measurements of Strass et al. (1993) show overflow water generation in the eastern fringes of the western boundary current, as isopycnal mixing between boundary-current-side and gyre-side waters, or is entrainment of recirculated gyre-side waters really responsible? If the latter is at least partially the case, the winter modification by the Odden of Jan Mayen Current waters, which return to the east of the EGC around the Greenland Sea gyre, would be be crucial.

The extrapolation to winter relies on both getting the barotropic circulation right via the satellite altimetry, and on getting the baroclinic structure right, for which we require in-situ winter measurements. We will need to use historical data to assess winter mean properties all around the box; the in-situ winter data, as well as re-setting conditions for calculation in the Greenland Sea, will inform the selection of historical data appropriate to our measurements.

The magnitude of winter convection in the Greenland Sea is a function of the stratification established in the previous summer as well as the rate of buoyancy removal in the current winter by cooling and ice formation. Therefore the proper sequencing of satellite monitoring and winter/summer field programmes is extremely important, as can be seen in the timeline below. At the start of the project the development of the 1998-9 Odden will be mapped using passive and active microwave imagery in combination with meteorological data from Jan Mayen Island to assess the salt flux and possible magnitude of the 1999 convection process. The summer 1999 James Clark Ross cruise will then measure the result of the subsequent melt in the region, with the ice extent and flux continuing to be monitored. Then the subsequent 2000-1 Odden will be mapped, with the in situ measurements of ice conditions and water mass modification during March 2000 supplying direct data on the winter phase of the same year's cycle. Finally, developments up to spring 2001 will be followed via continued summer and winter satellite mapping.

The following diagram shows the main events associated with this programme. We anticipate a 3-year duration, employing 2 staff, with summer and winter ship time and continuous satellite monitoring. Typical Odden event timing is shown.

TOTAL: £291.2K (NB: excluding all Ross—related costs)

The scientific direction of the project will be overseen by Bacon (SOC), Laxon (MSSL-UCL) and Wadhams (SPRI), who will meet thrice-yearly, with supplementary attendance by other collaborators. Financial control will be split between SOC and SPRI as befits relevant activities. Data will be disseminated by timely submission to BODC, Bidston and to ICES, Copenhagen and will be listed at the NERC Arctic Metadata Centre at SPRI, Cambridge. Results will be published in the refereed scientific press and presented at the 2000 or 2001 Fall AGU in San Francisco. The NERC and SOC press offices will be kept informed of issues of public interest. A web site for up-to-date news of progress will be established (location to be decided — probably SOC).

Back