2.1. East Greenland 1978

In summer 1978 both Kong Oscars Fjord and the offshore zone outside it carried a drift ice cover, the southern East Greenland ice limit lying between lat. 70° 30′ N. (Scoresbysund) and lat. 71° N. throughout the experimental period. The ice concentration was typically 6–8/10, composed of floes of diameters 30–120 m in an intimate mixture of first-year and multi-year ice. The median diameter (across longest axis) of 22 floes visited was 45 m, and most floes lay in the range 30–60 m. Multi-year floes were extensively hummocked, with a snow cover of 1–2 cm and large areas covered by refrozen melt pools. The mean thickness of 14 measurements through smooth areas was 3.1 m, which underestimates the true mean. First-year floes were smoother and averaged 1.5 m in thickness. The water surface between the floes was filled by brash and, later in September, newly formed pancake ice and sheets of nilas.

Figure 3 shows a Landsat image taken on 18 September 1978 and is typical of ice conditions during the experiments. A gyre is clearly visible within the fjord, showing that the surface circulation consists of an inflow along the north side of the fjord; a gyre within the fjord where incoming East Greenland Current water entrains ice-free fjord surface water; and an outflowing jet of fjord water along the south side of the fjord. This circulation pattern was first mentioned by Reference KochKoch (1945) who stated that Kong Oscars Fjord and Scoresbysund are the only East Greenland fjords wide enough to possess such an internal surface circulation; Koch illustrated the Scoresbysund pattern (reproduced in Reference WadhamsWadhams, 1981[a]) which closely resembles that of Figure 3. Within the ice-covered zone the sea surface temperature was close to freezing point, varying from −1.1 °C to −1.9 °C (−1.62 °C was the mean of 18 measurements), and water samples taken to 25 m depth by a Van Dorn bottle showed that the water column was isothermal to that depth. Air temperatures during the day on the ice (10.00 to 18.00 h local time) varied within the narrow range −3 to +1 °C (mean −1.2 °C) with winds usually very light (mean of 1.1 m s⁻¹, maximum of 7 m s⁻¹ and 60% calm). Surface melting had lowered the surface water salinity to a mean of 21.7‰ within the ice zone (13 measurements varying from 15.7‰ to 27.5‰) while at 25 m the salinity reached 31.6‰.

Fig. 3.

A Landsat-3 image of Kong Oscars Fjord taken on 18 September 1978. The image is a sum of spectral bands 4 (0.5 – 0.6 μm), 5 (0.6 – 0.7μm) and 7 (0.8 – 1.1 μm) with enhancement on the DK.IDIMS digital image computer at Electromagnetics Institute.

2.2. East Greenland 1979

Conditions in the same area in 1979 were quite different. At the time of freeze-up in autumn 1978 a large amount of ice from the main East Greenland pack remained in Kong Oscars Fjord and became embedded in the fast ice cover which formed in the fjord. Reference WadhamsWadhams (1981[a], fig. 14) shows Tiros-N imagery from 22 July 1979 when the fast-ice cover was still complete together with a fringe of fast ice along the outer coast. Fast-ice break-up began in early August but was incomplete in that a “plug” of fast ice remained to block the mouth of Davy Sund until about 20 August, by which time the main pack had retreated north of the fjord mouth. When we arrived on 24 August the plug had broken up but its fragments remained as a coherent mass of floes at the mouth of Davy Sund. There was no offshore pack, while within Kong Oscars Fjord the ice consisted mainly of multi-year floes which had been trapped in the fjord all winter. Figure 4 shows two temperature, salinity, and pH profiles taken by Van Dorn bottle in the uppermost 15 m of fjord water on 30 and 31 August. The surface water was exceptionally warm and of low salinity from ice melt and run-off from land. Under these conditions the old floes in the fjord had become rotted and were heavily undercut around the waterline with active melt-water pools and thaw holes. Figure 5(a) is a side-looking airborne radar (SLAR) image of the fjord taken on 6 September using a Twin Otter of the Greenland Ice Patrol and a 10 GHz SLAR developed at the Electromagnetisk Institut. The remains of the plug are still in place in Davy Sund (right of picture) while the remainder of the fjord contains a 3/10 ice cover with a few patches of greater density. By 10 September (Fig. 5(b)) the plug had disappeared, Davy Sund was ice-free, and a circulation similar to that of 1978 had begun to develop in the fjord. Figure 1 shows the tracks of two floes on which we worked for extended periods, showing outward drift along the south side of the fjord at some 0.5 m s⁻¹. Time-lapse photography from a mountain on the south shore revealed an up-fjord drift near the north shore at a similar speed. This circulation rapidly drained the fjord of ice, since no new ice was entering from offshore. By 17 September the fjord was almost empty of ice and only on 20 September, our day of departure, did the pack advance southward and begin to enter the fjord again.

Fig. 4.

Temperature, salinity, and pH profiles from the uppermost 15 m of water in Kong Oscars Fjord off Mestersvig, taken on 30 August (•) and 31 August (+) 1979.

Fig. 5.

Imagery of Kong Oscars Fjord taken by the 10 GHz TUD side-looking airborne radar (SLAR) on (a) 6 September, (b) 10 September 1979, showing disappearance of dense mass of ice floes from Davy Sund. The flight altitude was 2 700 m and the coastline has been enhanced in (b).

Winds during the period were again light, with 56% of the time calm, a mean of 1.4 m s⁻¹ and maximum of 6.2 m s⁻¹. Figure 6 is a wind rose showing that for most of the time the wind blew straight up the fjord. Air temperatures varied from −6 °C to +7 °C during the period from 06.00 to 18.00 h (mean +0.5 °C for period).

Fig. 6.

Wind rose for Mestersvig, 27 August–17 September 1979.

Ice thicknesses could not be determined because the rotted nature of the ice caused the corer to become stuck beyond about 1.5 m. One multi-year floe, possibly a piece of broken-out sikussak (Reference Walker and WadhamsWalker and Wadhams, 1979) had a freeboard of 2.5 m and was honeycombed with vertical holes of small diameter (1–2 cm). Floe diameters were similar to 1978.

2.3. Barents Sea and north-east of Svalbard, 1980

In the western Barents Sea the ice extent in the summer of 1980 was close to the 1971–1980 average, but in the eastern part more ice than normal occurred (Norsk Meteorologisk Institutt, 1980; Reference VinjeVinje, 1981). During late June much ice had drifted into the Barents Sea from the Arctic Basin, leaving wakes behind islands (Fig. 7), and by the beginning of July the ice edge lay between lat. 77 °N. at Svalbard and about lat. 75 °N. at Novaya Zemlya. During July the eastern Barents Sea opened up as far as Zemlya Frantsa losifa.

Fig. 7.

NOAA-6 image of the Barents Sea and western Eurasian Basin, 26 June 1980, showing ice streaming into northern Barents Sea across Nordaustlandet–Zemlya Frantsa losifa axis.

Ice in the Barents Sea was about 90% first-year during Ymer-80. Ice concentration varied greatly, with large areas of diffuse ice cover where the ice was in free drift and was in an advanced stage of melting. North of the line Svalbard–Kvitøya–Zemlya Frantsa losifa the ice concentration increased beyond 8/10 and the percentage of multi-year ice also increased (Fig. 2(a)). Multi-year ice was dominant north of lat. 82 °N. off Svalbard and north of lat. 81 °N. off Zemlya Frantsa Iosifa.

Ice cores gave only a few observations of floe thickness, and additional data were obtained on passage by scaling overturned ice blocks against a pair of marks on Ymer’s weather deck (Fig. 8). In this way the mean floe thickness within the Barents Sea was estimated at 0.9–1.2 m (Fig. 9). Floe thickness increased in the Arctic Basin and exceeded 1.5–2 m north of the line where multi-year ice became dominant.

Fig. 8.

The weather deck of ymer, photographed from bridge, showing scale used to estimate thickness of overturned ice blocks (60 cm wide).

Fig. 9.

Mean thickness of undeformed ice floes during leg 1 of Ymer-80, in metres.

Ice surface topography was profiled with a Eumig laser mounted on Ymer’s bridge and aimed obliquely at the ice surface 30 m from the ship’s side (Reference Leppäranta and PalosuoLeppäranta and Palosuo, [1981]). The method was usable in first-year ice where ice thickness was less than 1.5 m. In heavier ice the ship’s motion was uneven, producing large uncertainties in the record. Further, in heavy ice Ymer’s route was chosen carefully to avoid large ridges and highly deformed floes. Owing to problems with the laser, measurements could not be made in the eastern Barents Sea nor during leg 2. The usable data cover the early part of leg 1, from Kong Karls Land to lat. 82 °N. Ridges appeared as narrow peaks in the plotted laser profiles and they were manually tabulated using a cutoff height of 0.9 m. The results did not indicate significant variations over the study area, so all the profiles were combined in one distribution. Total path length was 72 km and 183 ridges were counted, a mean density of 2.6 km⁻¹ (± 30% due to uncertainty in ridge identification). The mean sail height was 1.34 m and the height distribution (Fig. 10) did not differ significantly from a negative exponential. When converted to a 1 m cutoff these figures become 1.44 m mean and 2.1 sails per km, which are lower than comparable values from further north in the western Eurasian Basin (Reference WadhamsWadhams, 1981[b], fig. 20). Visual observations from the bridge gave a similar range of ridge heights but a larger ridge density in the northern areas, by a factor of 1.5. This may be due to visual bias or to navigation. In the area north of Zemlya Frantsa Iosifa, not profiled by laser, ridges were higher and extensive hummock fields were seen.

Fig. 10.

Distribution of ridge sail height measured by laser during leg 1 of Ymer-80, showing fit to a negative exponential distribution. Track ran from Kong Karls Land to lat. 82 °N. long. 25 °E.

The contribution to the mean ice thickness from ridged ice, h _e, can be estimated from

(Reference HiblerHibler and others, 1974), where μ is the number of ridges per kilometre and

is the mean square sail height (m²). The result is h _e = 0.17 m, i.e. about 15% of the level ice thickness. This depends on the choice of cutoff height; assuming that the negative exponential distribution for sail heights fits below 0.9 m, a reduction of cutoff to 0.5 m increases h _e to 0.25 m. This is similar to values estimated by Reference HiblerHibler and others (1974) for the Beaufort–Chukchi province of the Arctic Basin in summer, but lower than their values for the rest of the Arctic Basin. Reference HiblerHibler (1980) has also given mean thickness contours for the Arctic Ocean, estimated from submarine sonar surveys. North of Nordaustlandet his 2 m contour occurs at lat. 81.5 °N. in April; our July data, combining level ice and ridges, suggest a 1.5 m mean thickness at that latitude with a gradient of 0.2 to 0.3 m per degree of increasing latitude.

A number of icebergs were observed in the Barents Sea during Ymer-80 (Fig. 11). Reference ZubovZubov (1945) reported that two areas of the Barents Sea (shaded in Fig. 11) were prolific sources of grounded iceberg sightings during the period 1899–1928. He believed that the eastern group of bergs originated from Zemlya Frantsa Iosifa and the western group from Nordaustlandet. Storøya, Kvitøya, and Victoria Island.

Fig. 11.

Iceberg sightings during leg 1 of Ymer-80; the “7” indicates seven bergs together. Shaded areas are locations of grounded iceberg sightings during 1899–1928, according to Reference ZubovZubov (1945).

2.4. Fram Strait, 1980

Ymer crossed the ice edge in Fram Strait at lat. 79° 15′ N., long. 00° 38′ E., encountering conditions typical of the marginal ice zone (Reference WadhamsWadhams, 1980, Reference Wadhams1981[a]), i.e. a well-defined ice edge fringed by small floes (less than 30 m diameter), with the floe size gradually increasing with distance from the edge until heavy consolidated pack ice began some 40–60 km inside the ice edge. Although well-defined, the edge was subject to meanders and, at the point of entry, a large vortex some 60 km in diameter caused by an instability of the front between polar surface water and Greenland Sea water (Reference Wadhams and SquireWadhams and Squire, in press). Six small icebergs were also seen at the ice edge (Reference WadhamsWadhams, 1981[a], fig. 21).

The ice in Fram Strait was an intimate mixture of first- and multi-year ice at 10/10 concentration. Near the Greenland coast, however, a zone of open water was encountered, which extended from the point of entry (lat. 80° 40′ N.) up to Nordøstrundingen. This polynya is characteristic of the region, and has been discussed in Reference WadhamsWadhams (1981[a]). Large tabular icebergs, possibly calved from Flade Isblink, were observed along the coast (Reference WadhamsWadhams, 1981[a], fig. 23). North of Nordøstrundingen a narrower lead continued to run north-west separating the offshore pack from very old multi-year fast ice at the mouth of Danmarks and Independence Fjords. This lead closed at lat. 82° 25′ N., long. 16° W, and very heavy deformed pack ice was encountered, forcing Ymer to return to the south-east.

3.1. Technique

As each ice-core section was drawn, it was encased quickly in a mould consisting of a vertical Teflon-lined cylinder cut lengthways and enclosed in an insulated wooden frame. The mould was perforated with holes at 10 cm intervals, into which a digital thermometer probe was pushed. The core was then placed in a second frame and sawed into 10 cm sections, which were measured with calipers, weighed and placed in bottles. After melting in the laboratory (ship or hut) and warming to ambient temperature, the conductivity was measured with a bench salinometer (1978/9), a conductivity probe which was frequently recalibrated against standard salinity samples (1980, leg 2) or a vibration density meter, with the salinity calculated from density (1980, leg 1).

When salinity was calculated from conductivity, the standard relation of UNESCO (1971) was used:

where S is salinity in ‰ and R is the ratio of sample conductivity to the conductivity of standard sea-water (chlorinity 19.376‰) at 15 °C. The uncertainty in R was estimated to yield an uncertainty in S of ±0.1‰. This relation assumes that the ionic composition of the salts in sea ice is identical with that of standard sea-water. There is abundant recent evidence that this is not the case (see section 7 of this paper), so that the “salinities”, in common with many others quoted for sea ice from published sources, are subject to a small and indeterminate error. Alarger error arises from the fact that brine drainage occurs during the drawing and sectioning of the core, so that the salinities quoted are inevitably underestimates.

3.2. Results from East Greenland 1978/9

Figure 12 shows salinity profiles from the 1978 and 1979 cores, and in Figure 13 the overall data have been averaged in groups of ten points, in order of increasing depth. The error bars are the standard deviations of depth and salinity for each group of data points. Points with salinity greater than 7‰ were not included in Figure 13; samples with such high salinities always represented isolated peaks in otherwise smoothly varying salinity profiles, and consisted of thin layers of slushy ice within the fabric. It is likely that these slush layers contained sea-water which had penetrated into the floe either from the side or from below through open brine drainage channels, although they may have had frazil ice fabric during the winter, as has been seen in ice from the Weddell Sea (Reference GowGow and others, 1982) and Labrador Sea (Reference Winsor and LeDrewWinsor and LeDrew, 1978).

Fig. 12.

Salinity profiles from multi-year cores in (a) 1978 and (b) 1979.

Fig. 13.

Composite salinity profiles for 1978 and 1979 multi-year ice cores, averaged in groups of ten data points. Bars show standard deviation of depth and salinity for each group of points. Shaded area shows desalination during a year for floes trapped in a fjord.

We showed in section 2.2 that the multi-year ice sampled in Kong Oscars Fjord in 1979 came essentially from the same ensemble of floes as that sampled in 1978, because of the entrapment due to the fjord plug and the lack of replenishment from the offshore pack. The salinity profiles show a major difference between 1978 and 1979, although the 1979 profile reached only 1.7 m because of the difficulty of using the corer in rotten ice. Assuming that the number of cores is large enough to be an unbiased random sample, then Figure 13 shows that multi-year ice from the East Greenland Current—typified by the 1978 profile—suffers further brine drainage when confined in a fjord through the winter and part of the succeeding summer until it reached the salinity shown for the 1979 profile. The difference between the two (shaded area) is thus the brine lost from multi-year ice in a year of drainage, although under somewhat unusual conditions. Each profile in Figure 13 also displays a small increase in salinity at the upper surface—probably due to spray—and a peak of salinity at about 90 cm depth. This peak is also seen on the mean profile of multi-year hummocks from the Beaufort Sea (Reference Cox and WeeksCox and Weeks. 1974), and may represent a depth limit for flushing, which one would expect to be most effective above the waterline. In other respects the 1978 profile resembles the average profiles found by Reference SchwarzacherSchwarzacher (1959), in that the salinity rises to about 3‰ at 2 m depth and remains fairly constant at increasing depth, with some further increase at extreme depths near the ice–water interface.

3.3. Results from Ymer-80

As Figure 2(a) shows, most of the 19 cores taken during leg 1 of Ymer-80 were obtained in two probes northward into the Arctic Ocean, the first north of Nordaustlandet (cores 1–8, 10–20 July 1980) and the second north of Zemlya Frantsa Iosifa (cores 9–17, 22–28 July 1980). Core 18 was obtained west of Kvitøya on 30 July and core 19 in the Barents Sea north of Kong Karls Land on 31 July. Thus only one core comes from the Barents Sea itself; however cores 9–18 come from the zone of the Arctic Ocean which is believed to be a source area for ice which subsequently drifts down the east side of Svalbard in the weak, variable, wind-driven East Spitsbergen Current (Reference LundeLunde, [1965]; Reference Loeng and VinjeLoeng and Vinje, 1979; Reference WadhamsWadhams, 1981[a]). 13 of the cores were classed as first-year (mean thickness 1.31 m, standard deviation 26 cm) and 6 as multi-year ice (mean 2.65 m, standard deviation 39 cm).

Individual salinity profiles are given in Reference PalosuoPalosuo (1981[a], Reference Palosuo[b]), while Figure 14 shows the overall salinities for first-year ice cores averaged in groups of ten values in the same manner as Figure 13. The first-year cores show a rapid rise in salinity with increasing depth to a peak at 70 cm, followed by a trend towards a steady value of 3–3.5‰. Comparison with Figure 13 (1978 data) shows

• (a) similar surface salinity;

• (b) a similar feature of a kink in the salinity curve at approximately the waterline (70 cm for first-year, 90 cm for multi-year ice):

• (c) a similar salinity (3–3.7‰) in the lower part of the ice sheet.

Fig. 14.

Composite salinity profiles for first-year ice cores from Ymer-80, showing desalination between legs 1 and 2.

The main difference is that the kink occurs at a higher salinity value in first-year ice, i.e. the rise in salinity with depth is more rapid. The general form of the salinity profile is clearly that of summer, differing significantly from the salinity profile of winter first-year ice (Reference Cox and WeeksCox and Weeks, 1974; Reference WeeksWeeks, 1976) where the salinity at the surface is at least as high as the salinity at depth. Figure 15, in which all of the measured salinity values are shown for the first-year profiles, indicates no significant difference between cores on the two northward probes, except for a tendency to greater desalination in the uppermost 30 cm during the second probe, which can be accounted for by the extra two weeks of summer experienced by the ice. Similarly, the two last cores from Kvitøya and the Barents Sea resemble the other cores.

Fig. 15.

Envelopes of salinity–depth for first-year cores from leg 1 of Ymer-80.

The six multi-year cores on leg 1 comprised 29 measured salinity values, which are shown averaged in groups of 7 on Figure 16.

Fig. 16.

Composite multi-year salinity profiles from Ymer-80, averaging over seven data points for leg 1, ten for leg 2. Note desalination between legs 1 and 2.

On leg 2 the 25 cores comprised 7 multi-year, 16 first-year, and two of indeterminate ice type. The first-year salinity profiles are shown in Figure 17, plotted on a semilogarithmic scale to emphasize variations in the uppermost low-salinity layers. The following features are apparent:

• (a) The cores, with few exceptions, follow remarkably similar salinity–depth curves, suggesting that a single relationship can be used to make a good average fit to all of the first-year ice from the whole experimental region for this time of year. First-year ice from north of Fram Strait does not differ significantly from first-year ice from north of Svalbard or Zemlya Frantsa Iosifa. There are two major exceptions. A few cores (e.g. 082702, 083102) appear to have suffered a great deal more brine drainage from their upper layers than the norm. The results from East Greenland in 1978/79 suggest that these floes may have spent the early summer in a confined waterway such as a fjord; their position north of Fram Strait would allow an origin near a European Arctic land mass. Secondly, the last two cores, taken in mid-September (091201 and 091401) are unusually saline; the reason for this is unknown.

• (b) Repeated cores from the same floe show very similar results, i.e. 090202/3, 083101/2, 082802/3, showing that first-year floes in the melt season possess considerable lateral homogeneity. This is not normally the case for multi-year ice, where systematic differences exist between cores taken through hummocks and depressions (Reference Cox and WeeksCox and Weeks, 1974).

• (c) Normally the salinity monotonically increases with depth, except in the uppermost 30 cm where it sometimes increases to the surface. This can be ascribed to the action of spray which is deposited on the upper surface and which percolates down through the upper layers. The large variability of salinity at the very top of the core appears to be independent of ice thickness and occurs also with multi-year ice. This raises serious difficulties in the interpretation of ice type from the results of remote-sensing techniques such as radar scatterometry (Reference OnstottOnstott and others, 1982) which depend on the properties of the uppermost few centimetres of the ice. This may be a problem, especially in summer, throughout the marginal ice zones of the European Arctic.

Fig. 17.

Salinity profiles for first-year ice cores from leg 2 of Ymer-80, plotted on semilogarithmic scale. (a) is period 15–28 August, (b) is 31 August to 14 September.

A ten-point average salinity profile from the 14 first-year cores of leg 2 (excluding the 12 and 14 September cores) is shown in Figure 14. The mean salinity tends towards the same limiting value (3–3.5‰) as for leg 1 below 1 m, but in the uppermost metre there is more desalination. The weighted mean date for the leg 1 cores was 21 July and for leg 2 was 26 August, so the shaded area in Figure 14 represents desalination in first-year ice occurring during the 36 d between these dates. Again the kink is present at the waterline—a universal feature of summer salinity profiles.

Figure 18 shows the multi-year salinity profiles from leg 2, plotted on a semilogarithmic scale. There is a much greater spread of properties in the uppermost 1 m than in the case of first-year ice. The only repeated cores on the same floe (083001/2/3) do show very similar properties in the 20–100 cm range, though not in the 0–20 cm range, despite being sampled in well-separated parts of a highly weathered floe. Core 083001 is of particular interest since it was taken through a floe in excess of 6.1 m thick (the depth limit of the corer) off the mouth of Danmarks and Independence Fjords. The floe (Fig. 19) was exceptionally weathered at the surface, with a highly developed drainage system of streams, canyons, and mesas indicating great age, and a variation in surface topography of at least 2 m. The floe was very large, and in the absence of aerial survey it was unclear whether the floe was actually part of the fast ice cover in the fjord mouth or whether it had recently broken out from this cover. In approaching the floe, Ymer had coasted along the fast ice edge, which had consisted of ice of a similar type. Such extremely old, very thick fast ice was called “sikussak” by Reference KochKoch (1945), who described its occurrence in north-east Greenland. Reference Walker and WadhamsWalker and Wadhams (1979) collected reports of very thick floes in the Arctic Ocean and surmised that they had originated as old fast ice. Application of Reference Maykut and UntersteinerMaykut and Untersteiner’s (1971) thermodynamic model for ice growth with altered input parameters typical of a high Arctic fjord (ocean heat flux zero, snowfall 1 m per year) showed that continued growth to about 20 m thickness could occur over tens of years, as opposed to the 3 m equilibrium thickness in the open ocean. The subject is further reviewed in Reference WadhamsWadhams (1981[a]).

Fig. 18.

Multi-year salinity profiles from leg 2 of Ymer-80, plotted on semilogarithmic scale.

Fig. 19.

Surface appearance of very old multi-year fast ice sampled at mouth of Danmarks and Independence Fjords, 30 August 1980.

Figure 16 shows a composite ten-point averaged salinity profile from the leg 2 multi-year cores, compared with the averaged profile for leg 1 multi-year cores. The difference between them is considerable, and indicates the additional desalination which occurred during the period between the weighted mean dates for leg 1 (20 July) and for leg 2 (27 August). A comparison of Figures 16 and 14 shows the desalination of multi-year relative to first-year ice at the same time of year. We see a salinity which increases linearly with depth and which reaches 3‰ at 2 m instead of 1 m for the first-year profile. A comparison of Figures 16 and 13 shows a good correspondence between the 1980 leg 2 multi-year curve and the 1978 multi-year curve, except that the kink is smoothed out in the 1980 profile. This indicates that the 1978 profile from East Greenland was indeed typical of multi-year ice from the offshore pack and that therefore summer multi-year ice has a similar salinity structure in the low latitude (72°) East Greenland Current as it does in the Trans-Polar Drift Stream north of Fram Strait and Svalbard.

3.4. Average salinity

Reference Cox and WeeksCox and Weeks (1974) found a negative correlation between the average salinity S through the depth of the ice sheet and the ice thickness h in winter and a weak positive correlation in summer. Figure 20 shows S plotted against h for all of the cores which were sampled to the bottom. There is again a weak positive correlation with a linear regression

compared with Cox and Weeks’ result of

Fig. 20.

Average salinity plotted against ice thickness for all 1978–80 cores sampled to bottom. Data reported by Reference Cox and WeeksCox and Weeks (1974) are also shown. Upper regression line fits our data: lower is from Cox and Weeks.

Our composite salinity profiles lead us to expect a positive correlation between S and h for a given ice type (first-year or multi-year). However, this effect is masked by the fact that multi-year ice is more desalinated in its upper layers than first-year ice, so that Figure 20 is really a sum of two independent lines of positive gradient, one for first-year ice and the other (with a lower intercept) for multi-year ice. In fact, if separate linear regressions for the first-year and multi-year points in Figure 20 are carried out, the multi-year data show a stronger positive correlation with depth of

while the first-year data show a very weak correlation: