Existing information concerning the pack ice and relevant climate variables of the Canadian Arctic Archipelago north of Parry Channel is summarized. This knowledge is enhanced by newly available data on ice thickness derived from 123,703 drill holes completed during the 1970s. Pack ice in this area is a mix of multiyear, second-year, and first-year ice types, with the latter subordinate except in the southeast. Ice remains land fast for more than half the year, and summertime ice concentration is high (7–9 tenths). In a typical year, less than 20% of the old ice and 50% of the first-year ice melt. There are large interannual fluctuations in ice coverage and some suggestion of a decadal cycle. The average ice thickness in late winter is 3.4 m but subregional means reach 5.5 m. The pack is a mix of two populations, one consisting largely of multiyear ice imported from the zone of heavy ridging along the periphery of the Beaufort gyre and the other consisting of a mix of relatively undeformed first-year, second-year, and multiyear ice types that grow and age within the basin. The ice of the Sverdrup Basin is strongly influenced by a flux of heat (approximately 10 W m−2) that originates in the Atlantic-derived waters of the Arctic Ocean. The drift of ice through the basin is controlled in the present climate by the formation of stable ice bridges across connecting channels. The drift is episodic. Relaxation of these controls in a warmer climate may cause deterioration in ice conditions in Canadian Arctic waters.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 The Canadian Arctic Archipelago occupies 3.3 × 106 km2 of the North American continental shelf. Although 40% of the archipelago is land, the remaining marine area (1.9 × 106 km2) comprises one sixth of the total area of the Arctic Ocean and its peripheral seas (11.4 × 106 km2). The continental shelf has been deepened by glacial action to form a network of basins as deep as 600 m connected by wide straits of lesser depth (65–220 m). Islands lying north of 74.5°N in the Canadian Arctic Archipelago are known collectively as the Queen Elizabeth Islands. This paper discusses the ice cover of the waters separating these islands, for which there is no formal collective designation. We refer to the marine area as the Sverdrup Basin, following geological nomenclature for the region. The location of the Sverdrup Basin within the northern polar region is shown in Figure 1. Geographic features of the Queen Elizabeth Islands and the Sverdrup Basin are identified in Figure 2. Water depth exceeds 200 m over most of the Sverdrup Basin.
 Pack ice persists within the Canadian Arctic Archipelago throughout the year. Although the waters of the southern and eastern regions may clear wholly or in part by late summer, ice concentrations in the Sverdrup Basin are always high. The extreme difficulties of navigating the ice-plagued waters of this remote area and the harsh polar climate have inhibited study of its marine cryosphere. Scientific knowledge is superficial and incomplete.
 The Northwest Passage is a potential deepwater shortcut for shipping between the Atlantic and the Pacific. This route is minimally used at present because of the heavy multiyear ice that invades it from the Arctic Ocean in the west and the Sverdrup Basin in the north. Predicted changes in the climate of the Arctic may influence the development, evolution and fate of pack ice in these areas and thereby reduce the difficulty of navigating the Northwest Passage. From a similar perspective, an ameliorating ice climate could conceivably render viable the exploitation of large natural gas reservoirs within the Sverdrup Basin.
 The channels of the Canadian Arctic Archipelago provide pathways for the movement of seawater [Fissel et al., 1988] and pack ice between the Arctic and Atlantic Oceans. Freshwater fluxes through the Canadian Arctic as ice or as low salinity seawater have significance within both the Arctic and global climate systems [Stigebrandt, 2000]. The Arctic Ocean provides the return route to the Atlantic for excess precipitation over the Pacific created by atmospheric transport [Stigebrandt, 1984; Wijffels et al., 1992] and is therefore a critical link in the global hydrologic cycle. The freshwater in transit across the Arctic creates a stable upper ocean density structure that reduces the flux of oceanic heat to the surface [Melling et al., 1984; Aagaard and Carmack, 1989]. It thereby contributes to the continuance of perennial ice in the high Arctic [Maykut and Untersteiner, 1971]. The freshwater flux ultimately reaches the convective gyres of the Greenland, Irminger and Labrador Seas where it influences the intensity of the global thermohaline circulation [Hakkinen, 1999]. The magnitude of the freshwater flux through the Canadian Arctic and its partition between ice and seawater components are not known [Melling, 2000]. However, it may vary appreciably with change in climate. Paleoclimatic data suggest that ice (and marine mammals) moved more freely through the Canadian Arctic Archipelago at some times during the Holocene than at present [Dyke et al., 1996, 1997].
 The thickness and persistence of pack ice reflect the combined influences of surface temperature, cloud cover, snow accumulation, oceanic heat flux, ice ridging and transport. Whereas air temperature, snow cover and oceanic heat flux are principal influences on the thickness of land-fast ice [Maykut and Untersteiner, 1971; Brown and Cote, 1992; Flato and Brown, 1996], ridging is an important factor in determining thickness where the pack ice is mobile [Melling and Riedel, 1995; Wadhams, 1992]. The magnitude and relative importance of these influences in the Sverdrup Basin are likely different than in the Arctic Ocean.
 This manuscript reviews present knowledge concerning the ice of the Sverdrup Basin. A 10-year series of ice thickness measurements in the Sverdrup Basin [Wetzel, 1976, 1978, 1981] is central to the discussion. The data were acquired during the conduct of seismic survey from the ice surface during winter. With the possible exception of secret sonar data from submarines that may have transited the Sverdrup Basin, these provide the only information on noncoastal ice thickness in this area. Prolonged observations of coastal ice thickness and conventional ice reconnaissance data from the Canadian Ice Service are embraced by this review.
 The objectives of this study were to: 1) Document the thickness distributions associated with first-year ice, second-year ice, and multiyear ice in the Sverdrup Basin. 2) Investigate temporal and spatial variability in ice thickness. 3) Compare the characteristics of Sverdrup Basin pack ice with those of adjacent areas of the Canada Basin. 4) Develop a conceptual model to explain the differences.
2. Present Knowledge
2.1. Environmental Influences on Pack Ice
 The important environmental controls on the thickness and persistence of pack ice are surface temperature, cloud cover, snow accumulation, oceanic heat flux, ice ridging and transport. During the winter months, a broad band of low surface air temperature extends across the Arctic Ocean, joining temperature minima in northern Siberia and in the Canadian Arctic Archipelago [Rigor et al., 2000]. Winters are about 4°C colder in the Sverdrup Basin than in the central Arctic Ocean [Maxwell, 1980]. In summer surface air temperature over the Arctic Ocean remains within 1°C of freezing because of the ubiquitous presence of melting ice [Rigor et al., 2000]. The flux of sensible heat from the atmosphere to the ice is therefore small, and radiation is the dominant contributor to melting. Near the continental landmasses, where the average July August temperature rises to 5°C [Rigor et al., 2000], atmospheric heat may have a greater impact on melt. Within the Sverdrup Basin, where melting ice and snow-free land are closely juxtaposed, the surface air temperature in marine areas in summer is likely to be intermediate in value, although no data exist to demonstrate this [Maxwell, 1980].
 The climate of cloud cover for the Arctic Ocean and adjacent shelf seas is poorly defined. Accurate observation from the surface is hindered by polar winter darkness and is complicated by the common occurrence of ice crystals and multiple cloud layers. Satellite retrievals are hampered by cloud layering, low concentrations of atmospheric moisture and poor contrast between cloud and the snow covered land or sea surface in all bands of the electromagnetic spectrum [Curry et al., 1996]. The research that might identify differences between the cloud covers of the Sverdrup Basin and the Arctic Ocean has yet to be done.
 The average end-of-winter snow depth from Russian observations on pack ice northwest of the Sverdrup Basin exceeds 40 cm [Colony et al., 1998; Warren et al., 1999]. However, values from the Canadian climatology [Maxwell, 1980] for adjacent areas of the Queen Elizabeth Islands are lower (20–25 cm). It is not clear that this difference reflects regional geographic variation. It could be attributable to differences in substrate, in measurement strategy or in the choice of climatological epoch. Russian observations were made on multiyear floes whereas Canadian norms represent coastal sites on land. Russian data were acquired systematically via snow line surveys whereas Canadian observers were simply instructed to make several measurements, avoiding snowdrifts [Brown and Braaten, 1998]. Russian data are from 1954–1991, whereas Maxwell's uses the 1955–1972 period.
 The most probable explanation is a difference between the accumulations of snow on Arctic land and sea ice surfaces. The average end-of-winter snow depth on sea ice calculated from archived data for Isachsen in the central Sverdrup Basin is 40 cm [Canadian Ice Service, 1992]. This value is close to the Russian values for sea ice adjacent to the Canadian Arctic Archipelago, but almost twice that mapped at this site by Maxwell . The averaging period for the Isachsen data (1948–1978) encompasses Maxwell's epoch and is largely contained within the temporal domain of the Russian data. We conclude, in the absence of more exhaustive study, that the snow depth on Sverdrup Basin ice does not differ greatly from that on old ice in the Canada Basin.
 The flux of oceanic heat to the ice is smaller in the Arctic Ocean than in the Sverdrup Basin. In the Canada Basin, where a cold halocline intercepts heat diffusing from depths below 250 m, the average annual oceanic flux to the ice is only about 3 W m−2 [Maykut and Untersteiner, 1971]. The cold halocline is maintained by wintertime brine-driven convection within the flaw leads of the Beaufort, Chukchi and Laptev Seas [Melling, 1993]. The halocline of the Sverdrup Basin is isolated from this ventilation and therefore warms in response to sustained heat diffusion from below [Melling et al., 1984]. The rate of upward heat diffusion is enhanced in the Sverdrup Basin by turbulence generated by tidal stress on the sloping walls of the channels.
 Ridges are built in ice fields where stress accumulated from the action of wind and current over large areas of pack exceeds the local buckling strength of the ice. In general, ridge-building forces will be lower in the Sverdrup Basin than in the Arctic Ocean because wind stress can accumulate over only a few hundred kilometres of ice between islands. Moreover, ocean currents are too weak in most areas of the Sverdrup Basin to make a significant contribution to stress [Stronach et al., 1987]. For these reasons, deformation will have less impact on the average thickness of ice in the Sverdrup Basin than in adjacent areas of the Arctic Ocean, where up to 75% of the ice volume resides in ridges [Melling and Riedel, 1995].
 Because pack ice moves with the wind and current, its characteristics reflect not only the local climate but also that of the region of origin. Ice within the Canada Basin circulates within an anticyclonic gyre extending from 70° to 85°N that may trap ice for several decades [Colony and Thorndike, 1984] (see firstname.lastname@example.org.). This gyre maintains high ice pressure and shear along the northwest perimeter of the Queen Elizabeth Islands and thereby creates the thickest and most heavily ridged sea ice in the world [Bourke and Garrett, 1987]. Rugged floes that drift into the Sverdrup Basin from this zone may have characteristics significantly different from those formed locally.
2.2. Regional Ice Conditions
 Reconnaissance conducted by the Canadian Ice Service and its predecessors since the late 1950s is our primary source of information on the sea ice of the Sverdrup Basin. Canadian ice charts are prepared primarily for guidance to shipping. They are based on sources that have evolved over the years both in availability and utilization. Early visual reconnaissance from ships and aircraft has been supplemented progressively by visual and thermal band imagery from meteorological satellites (1–4 km resolution), by microwave band satellite imagery (25 km resolution) and by radar imagery from aircraft and satellites (25–200 m resolution). Some component of any trend in ice conditions charted by the Canadian Ice Service will certainly have its origin in changes in observational technology and coverage, rather than in changes in the environment [Carrieres, 2000]. The impact of technological change on this analysis has been minimized by simplifying the WMO ice classification to three distinct categories and by ignoring trends.
 The Sverdrup Basin is covered by land-fast ice at 10 tenths concentration for more than six months of the year [Canadian Ice Service, 2000]. The ice reaches its minimum extent early in September (Figure 3). In at least half of all years on this date, the ice remains land fast at 10 tenths concentration across all but one of the openings to the Arctic Ocean (Prince Gustaf Adolf Sea). The plug of multiyear fast ice in Sverdrup Channel is known to have cleared only in 1962, 1977, 1978 (probably), and 1998 and that in Nansen Sound only in 1962, 1971, and 1998 [Black, 1965; Jeffers et al., 2001; B. Alt, personal communication, 2001]. The plug in Peary Channel clears more frequently. Ice is typically present at 9 tenths concentration over most of the remaining area in late summer. Exceptions are the areas immediately to the south of the Ellef and Amund Ringnes Islands, in Belcher Channel and in western Norwegian Bay, where the median concentration ranges between 4 and 8 tenths and small ice-free areas can be found.
 The concentration of ice is lower to the south of the four straits that permit egress of ice from the Sverdrup Basin, namely Byam Martin Channel, Penny Strait, Cardigan Strait and Hell Gate. Strong currents in these straits cause divergence in the pack ice as it passes through them. Table 1 documents the average composition of pack ice during the melt season within five zones in the Sverdrup Basin outlined on Figure 7. The composition during the freezing season is essentially the same as that observed in July because the ice remains fast until midsummer. Second-year and multiyear types have been grouped into an old ice category because they are not reliably identified during ice reconnaissance (see below).
Table 1. Average Composition of Ice in the Sverdrup Basin on 1 September, Based on the 30 Year Climatology 1970–1999a
 The three zones (7–9) that comprise the northwest Sverdrup Basin have similar ice conditions. At the start of melt old ice covers about 86% of the area and thick first-year ice covers the remainder. During the summer, about one tenth of the sea surface become ice-free through loss of about 35% of the first-year ice and 5% of the old ice (zones 7 and 8). The unequal fractional losses indicate that ice melting, which eliminates 2 m of first-year ice more quickly than thicker old ice, contributes more to the loss than ice export. However, export is a more significant factor in zone 9 where the fractional declines (rightmost column) in first-year and old ice are closer in value. Zone 10 starts the summer with a much higher percentage of first-year ice (34%) than zones 7–9 and ends the melt season with more open water (30%), losing larger fractions of both the old and first-year ice types. Conditions in zone 11 continue this southeasterly trend to more first-year ice and more ice loss. About 6% of the ice in zone 11 have already disappeared by mid July through expansion of tidal polynyas in Hell Gate, Cardigan Strait and Belcher Channel. The total loss is almost 50%.
 Open water and new ice in September evolve to become first-year ice by the following July. Therefore the area of minimal ice cover for September in Table 1 should be approximately equal to the area of first-year ice in July. The small differences evident in the table can be attributed to movements of ice into and out of the Sverdrup Basin between October and December, before the pack becomes completely land fast. The fractional coverage of second-year ice can be estimated through similar logic. First-year ice that survives summertime melting becomes second-year ice on October 1. The fractional coverage of first-year ice in September can therefore be interpreted as the fractional coverage of second-year ice through the winter to the following melt season. By this argument Table 1 reveals that about 9% of zones 7–9, 17% of zone 10 and 23% of zone 11 have second-year ice cover in winter. The fractions of multiyear ice are then 80%, 54% and 29%, respectively. We conclude that second-year ice comprises between 11% and 45% of the old ice coverage of the Sverdrup Basin.
 Interannual variations in pack ice extent and composition in September are displayed in Figure 4, which is based on the 30 year record in Canadian weekly ice charts [Canadian Ice Service, 2000]. Data from zones 7–9, which have similar mean ice conditions, have been amalgamated in this presentation. Weak trends in these time series may have more to do with changes in reconnaissance and analysis techniques than with environmental change. The area of multiyear ice in zones 7–9 varies from year to year by more than ±20% from the average. The relative scarcity of multiyear ice in 1976, 1984, and 1995 is suggestive of a decadal variation in the fractional coverage by multiyear ice in the Sverdrup Basin. However, the abrupt decreases in the area of old ice in 1971 and in 1998–1999 run counter to this pattern. The same cycle forms an identifiable part of the variability in zone 10, again excepting 1998–1999. The amplitude of variations about the mean of zone 10 is ±60%, much larger than in zones 7–9. In zone 11 the area of old ice varies over ±100%, with this zone being essentially free of ice in 1981 and 1998. A decadal cycle in old-ice area is not obvious in zone 11.
 The variability in Figure 4 bears little obvious relationship to the decadal wind-forced variability of ice in the Arctic Ocean [Proshutinsky and Johnson, 1997]. The lack of correlation is not surprising since the ice of the Sverdrup Basin is land fast and therefore unresponsive to wind during most of the year. Jeffers et al.  argue that the widespread loss of multiyear ice in 1998 can be attributed in part to air temperature that was 2.5°C warmer than normal that summer. However, in other summers (1971, 1976, and 1984) with dramatically depleted multiyear ice the air temperature anomalies within the Canadian Arctic tundra zone were small, 0.8, 0.4 and 0.8°C (Environment Canada, Climate trends and variations bulletin for Canada. Meteorological Service of Canada Green Lane web site, http://www.msc-smc.ec.gc.ca/ccrm/bulletin/, 2001).
2.3. Ice Mobility
 Where an ice cover consists of loosely packed small floes, its drift through a channel is generally unimpeded. However, where large thick floes are closely packed, they may jam within the channel [Sodhi, 1977; Pritchard et al., 1979]. The continued drift of ice downstream of the blockage leaves an arch to mark the boundary between open water and fast ice. Ice arches form across channels as wide as 100 km within the Canadian Arctic Archipelago in winter. Much of the ice of the Sverdrup Basin is immobilized between October and late July by arches that link the islands along the northwest margin and bridge the routes of egress to the southeast [Marko, 1977; Melling, 2000]. As the ice weakens through the formation of shore leads and general melting in July, the arches collapse progressively toward the interior of the basin. However, not all arches disintegrate in summer and some areas remain land-locked for many years in succession. These areas are colored black in Figure 3. Note that the arch-shaped boundary of stable ice in any specific year will not necessarily be evident on a map of 30 year median concentration.
 With the collapse of arches in July, ice from the Sverdrup Basin begins to move southward through Hell Gate, Cardigan Strait, Penny Strait and Byam Martin Channel. This seasonal export of hazardous old ice into the northern shipping routes is well known to mariners [Bailey, 1957]. It feeds the early autumn outflow of old ice from Lancaster Sound and replenishes the stagnant ice fields in southern Viscount Melville Sound, M'Clintock Channel and Peel Sound. Unfortunately, there are no systematic observations of ice movement within the Sverdrup Basin.
 The persistent pressure of the Arctic polar pack on the outer coast drives heavily deformed floes into Sverdrup Basin from the northwest in summer and autumn if ice within the basin is mobile. However, none of the ice buoys tracked during the International Arctic Buoy Program (1979–2001) have moved from the Beaufort gyre into the Sverdrup Basin, perhaps because buoys cannot survive the passage through the dynamic interface between the polar pack and coastal ice [Hoar, 1980]. Ice islands that originate on the northern coast of Ellesmere Island are more robust markers of ice drift. T1 (30 km × 20 km) entered the Prince Gustaf Adolf Sea from the Arctic Ocean in the early 1960s, and drifted from MacLean Strait through Byam Martin and Austin Channels into Viscount Melville Sound (74°35′N 108°35′W) between 20 July and early October 1962 [Black, 1965]. Fragments of T1 turned eastward in Viscount Melville Sound and came to rest in M'Clintock Channel and Peel Sound. Other ice island fragments in 1962 drifted from the northern end of Peary Channel to Hassel Sound, to Wellington Channel and to Norwegian Bay. In 1988 ‘Hobson's Choice’ entered Peary Channel from a position 100 km northwest of Ellef Ringnes Island, following several years of drift along the Ellesmere coast. ‘Hobson's Choice’ drifted 150 km southeastward between August and October, stopped for the winter and drifted a further 21 km in September 1989. It stalled east of Ellef Ringnes Island until August 1991 and then moved through Hassel Sound and out through Penny Strait in less than four months [Jeffries and Shaw, 1993]. Several ice islands were locked in the ice of Wellingon Channel during the winter of 2000–2001.
 The total aperture for ice import along the northwest margin of the Sverdrup Basin is 315 km, although perhaps one third of this is frequently blocked by fast ice in summer (Figure 3). The principal openings are in Prince Gustaf Adolf Sea (95 km) and Peary Channel (90 km). The aperture for ice export to the southeast is much less (76 km in total) although stronger currents in these channels can facilitate a rapid transit by ice. The relatively small aperture for export contributes to the high summertime ice concentration in the Sverdrup Basin.
Marko  used ice charts from the 1960s and cloud-free satellite images during 1971–1977 to track the movements of ice in the Sverdrup Basin and connecting straits. Blocking by fast ice plays a dominant role in ice circulation. Ice west of Lougheed Island is often immobile in summer. In fact between autumn 1963 and summer 1967 there was no ice movement in the Prince Gustaf Adolf Sea and Byam Martin Channel. In some years there is substantial movement. An extensive clearing from the Sverdrup Basin in the summer of 1962 was following by a penetration of old ice from the Arctic Ocean late in 1962 and in the summer and autumn of 1963 [Black, 1965]. Ice in the Prince Gustaf Adolf Sea corridor was active again in 1972 and 1975. In 1973 ice moved freely through Hassel Sound. The massive barrier of Arctic Ocean pack in the north permits only southward drift, so that wind forcing is rectified. There may be some variation in drift to the east and west. For example, MacLean Strait may clear via Byam Martin Channel or Penny Strait depending on wind direction. However, ice responds to northerly wind only when a route to ice-free waters is available. For drift through Byam Martin Channel there must be leeway in Viscount Melville Sound or far to the south in Larsen Sound and Queen Maud Gulf [Marko, 1977].
 Typical average speeds of ice drift have been determined from the displacements of distinctive floes over intervals of several days, using either repeated aircraft reconnaissance or satellite images. Drift has been observed at 5–10 cm s−1 in the Sverdrup Basin, 10–20 cm s−1 in Byam Martin Channel and Penny Strait and 15–30 cm s−1 in Cardigan Strait and Hell Gate [Black, 1965; Marko, 1977; Jeffries and Shaw, 1993]. At 5 cm s−1, the 450 km transit of the Sverdrup Basin from the Arctic Ocean to Parry Channel can be completed in 105 days.
2.4. Ice Thickness
 The thickness of undeformed first-year ice and its seasonal and interannual variations are well documented by more than thirty years of weekly measurements at Arctic coastal meteorological stations [Canadian Ice Service, 1992; Brown and Cote, 1992]. Data recorded at Isachsen near the centre of the Sverdrup Basin (Figure 7) are presented in Figure 5. Ice growth begins on September 1 and ceases on June 1. The average maximum ice thickness is 2.2 m with an interannual variation of ±0.5 m. Snow accumulates most rapidly in autumn and in spring, to an average depth of 0.45 m. The interannual variation in maximum snow depth is large (±0.3 m) and constitutes the dominant control on interannual variations in ice thickness [Brown and Cote, 1992]. Ice begins to melt at the beginning of July following the complete melting of the snow cover (end of June). The temporal variation in ice and snow cover thickness at Isachsen during the 1970s is depicted in Figure 6.
 Knowledge of the full thickness spectrum of sea ice in northern Canadian waters is poor. The few observations that are published have been acquired using sonar operated for navigation on military submarines. These reveal a zone of very thick ice (4–7 m in average draft) that extends 100–200 km from the coast to the northwest of the Queen Elizabeth Islands [Bourke and Garrett, 1987]. In M'Clure Strait, ice on two transects in 1960 averaged 4.5 m in draft [McLaren et al., 1984]. Level ice comprised slightly more than half of these transects, with the most common drafts near 1.5, 2.0 and 3.7 m in February being attributable to undeformed first-year, second-year, and multiyear ice, respectively. Ridge keels deeper than 9 m were observed at an average separation of 1000 m and the deepest keel was 29 m. In May 1976 twelve multiyear pressure ridges with sail heights of 2.3–6.4 m were surveyed by scanning sonar in the Prince Gustaf Adolf Sea (78°26′N 105°10′W). The keels of half of these ridges extended at least 25 m down, two keels exceeded 30 m and one reached 37 m [Dickins and Wetzel, 1981].
3. New Data on Ice Thickness
 During the conduct of seismic surveying among the Queen Elizabeth Islands in the 1970s more than 120,000 sea ice thickness values were measured in holes drilled to deploy hydrophones. The data were acquired during the winters of 1971–1975 and 1977–1980. Originally proprietary within the oil industry, the data and statistical summaries have now been released on microfiche [Wetzel, 1976, 1978, 1981]. Unfortunately, the original data in digital form have been lost.
 A Motorola range-positioning system was used to locate the end and intermediate points of the seismic lines, and the exact points for the drilling were set using a theodolite and surveyor's tape. Ground control was based on Geodetic Survey stations. The alignment and interval spacing of boreholes was precise because of the stringent requirements for the spacing of hydrophones in seismic surveys.
 Thickness values were acquired in a systematic manner with minimal bias to thinner ice. Holes were drilled using a caterpillar-mounted rig, designed for rock drilling and blasting in support of road construction, which can operate in very rough terrain. Because land fast conditions develop in October, all ridges in the Sverdrup Basin are old and heavily drifted with snow by March. Drifted ridge sails are relatively easy to traverse. Moreover, since ridge keels rarely lie directly beneath ridge sails [Melling et al., 1993], failure to drill at the crest of a ridge does not necessarily mean that the thickest ice has been missed.
 Where the ice was extremely rough, a point located 100 feet to the right or left of the target was used. Holes were abandoned in very thick ice only occasionally, and this occurrence and the depth drilled were noted in the log. For example, 11 holes were abandoned at the 40-foot depth (12.2 m) and 4 at the 52-foot depth (15.9 m) in 1977, representing only 0.7% of 2204 holes drilled. In 1974, 120 holes (0.9% of 13,628) were terminated at 28 feet (8.5 m), which was the full length of the drill stem available to one of the crews at that time.
 These fractions are almost an order of magnitude smaller than the 4.8% ‘deficit’ of deep holes in the Sverdrup Basin relative to the Beaufort Sea: 15.4% of the Sverdrup Basin ice exceeds 5 m thickness, whereas 20.2% of Beaufort Sea ice is in this category (see comparative data later in this paper). Therefore, the incidence of thick-ice avoidance in drilling is probably too small to be significant in these observations. A significant avoidance of thin ice is also not likely since the static ice fields of the Sverdrup Basin spawn few leads that might hinder vehicular access. The thinnest ice reported was 0.3 m.
 The precise time period during which data were acquired is not always specified in the reports. In 1977, measurements were made between March 4 and June 9. In 1979, the period was shorter, March 18 until May 20. Considering the daylight and temperature constraints on this type of activity, it is reasonable to assume that all the data were acquired during the March–May period. First-year ice in this area thickens by about 25 cm in March, 12 cm in April and 9 cm in May (Figure 5).
 Ice thickness was recorded as an integer in feet. Unfortunately, it is not clear exactly how the measurement was made, although it is likely that the thickness was estimated from the length of the drill stem below the ice surface at the point of breakthrough (Vern Wetzel, personal communication). If the drill stem was labelled in feet, the precision of thickness is probably about the same as the resolution, 1 foot. However, there is a tendency on some of the survey lines for the recorded ice thickness to be a multiple of 2, 3, 4, 5 or 6 feet (multiples of 2 and 5 feet are most common), especially for the thick ridged ice. In such cases the resolution is obviously degraded. It is possible that drill stem was used in sections of standard length (e.g., 5 feet) and that the thickness was taken as the number of sections on the drill at breakthrough. In this case the recorded thickness may have been rounded up from the true value.
 The original reports [Wetzel, 1976, 1978, 1981] present the thickness data as line printer plots versus distance and as statistical summaries calculated for each survey line. Thickness was measured at various separations dictated by seismic requirements (110, 165, 220, 440, 660, 880, or 990 feet). Because these values are large relative to the width of ice ridges, adjacent data are uncorrelated. For this reason the thickness-versus-distance series were judged to be of little interest and only the statistical summaries were typed into computer files for further processing.
 The analysis of the thickness data was straightforward for all but about 10% of the survey lines, which had data for thick ice grouped at a degraded thickness resolution with bins every N feet. These histograms were resampled in order to permit use of the entire data set at 1-foot resolution in preparing aggregate statistics. The resampling was based on the assumption that ice thickness grouped in bins at N-foot increments had been sampled from an exponential distribution with a scale length of 9.5 feet. This scale value is typical for the ridged ‘tail’ of the Arctic ice thickness distribution [Melling and Riedel, 1995] and consistent with values found in the Sverdrup Basin (see below). The population of each N-foot bin was shared with the preceding N–1 1-foot bins such that each bin contained 10% fewer samples than the adjacent bin for thinner ice. The 10% reduction was determined by the value of exp(−1/9.5).
 Information concerning the ice thickness data set for the Sverdrup Basin is summarized in Table 2. The locations of all the lines surveyed during 1971–1980 are mapped in Figure 7. The five zones used in the interpretation of the Canadian weekly ice chart are overlaid on this figure. Although weekly charts were not prepared for the winter months in the 1970s, the last chart of autumn and the first of following summer are representative of the ice surveyed during the seismic season in this area of land-fast ice conditions in winter.
Table 2. Summary of Ice Thickness Data From Boreholes in the Sverdrup Basin During 1971–1980
Average Ice, m
Maximum Ice, m
4 March to 9 June
18 March to 28 May
 Survey lines for each year were divided into groups to reduce sampling uncertainty in statistical analysis. Groups were planned in relation to the zone boundaries of Figure 7. Three quarters of the groups contained more than 1000 points and almost half contained more than 2000 points. Since the standard deviation in ice thickness is typically less than 3.5 m, the statistical uncertainty in group mean values is better than 0.1 m.
 Group mean values of ice thickness for the nine years of the survey are presented in Figure 8. Values range between 1.6 m in Norwegian Bay (1977) to 5.5 m in the Prince Gustaf Adolf Sea (1979). The high value represents 1108 measurements acquired across 56 km of ice. The distribution of ice thickness is represented here in its volume-weighted form. That is, the bar for each 1-foot thickness range is scaled in proportion to the fraction of the observations in the thickness range multiplied by the thickness value. Thus each bar represents the fraction of the total ice volume that falls within a 1-foot thickness bin. This presentation permits emphasis of the low-probability tail of the distribution without need for a logarithmic axis.
Figure 9 depicts the ice volume distribution based on the entire 422 km thickness of ice that was drilled. The average of the distribution is 3.4 m. The quartiles are 2.44, 3.19, and 5.42 m and the 10th and 90th percentiles are 2.05 and 8.4 m. Ice thicker than 10.4 m comprises only 5% of the volume. The mode (2.6 m) lies within a broad peak that spreads between 1.8 and 4 m. Clearly, this peak integrates relatively minor contributions from level first-year ice (typically 1.8 m thick in March and 2.2 m in May: Figure 5) with dominant contributions from second-year and multiyear ice. This aspect is consistent with the climatological mix of ice types that is revealed by Table 1.
 Although the separate modes of first-year, second-year, and multiyear ice cannot be discerned in the composite data set, the peaks are seen clearly in the histograms derived for local and contemporary groups of survey data. The thickness indicative of first-year ice was determined annually with reference to the ice thickness record at Isachsen during the 1970s (Figure 6). Because the maximum thickness of level first-year ice varies by as much as a metre from year to year (Figure 5) and the net growth of ice after its second winter may be only 0.5 m, this record was also a guide to the identification of second-year ice. We assumed that the thickness of second-year ice could not exceed 3 m.
 A three-category classification of the histograms was adopted according to the following principles: 1) An histograms may have a single or multiple modes. 2) A mode may correspond to first-year, second-year, or multiyear ice, according to the position and width of the peak. 3) The histogram may be narrow and symmetric or broad and positively skewed, according to whether the ice field is undeformed or ridged. Representative examples from the 54 survey groups are shown for single-mode distributions in Figure 10 and for multiple modes in Figure 11. The narrow multiyear ice peaks that can appear in the Sverdrup Basin contrast with the broader peaks typical of the Arctic Ocean [Wadhams and Horne, 1980]. Their sharpness suggests that some of the multiyear ice in this area has developed locally from broad expanses of essentially undeformed first- and second-year fast ice.
 The classification of ice volume histograms permits estimation of pack ice composition in the Sverdrup Basin during winter in the 1970s: 37% multiyear, 34% second year, and 29% first year. These fractions, from boreholes predominantly in zones 7–10, are appreciably different from those in Table 1 from ice reconnaissance of the same zones: 54–80% multiyear, 10–17% second year, and 10–29% first year. An ‘excess’ of first and second-year ice in the thickness data set could result from overestimation (in this analysis) of the thickness attained by these types of ice in waters outside the coastal zone. An ‘excess’ of multiyear ice in the reconnaissance data set could reflect the difficulty of distinguishing ice types by surface appearance where all are relatively undeformed.
 There is certainly some ambiguity in the Canadian weekly ice charts of the Sverdrup Basin. Significant differences in classification exist in the overlap region between roughly contemporary charts for the eastern and western Canadian Arctic. For example, the eastern chart for 20 June 1973 has zone 10 coded as 1/4/5 (MY/SY/FY) whereas the western chart for 22 June displays the code 0/10/0. Significant differences in classification exist between versions of the same chart for late autumn and the following June. The western chart for 29 October 1971 codes the entire Sverdrup Basin as 10/0/0 whereas the western chart for 19 June 1972 indicates 0/10/0 for the same area. The obvious difficulty in discriminating ice types via reconnaissance prompted the decision to amalgamate multiyear and second-year ice concentrations for much of the discussion in this paper.
 We also scored the classification of each thickness group against the charted classification. A group was flagged if its thickness-based classification (FY, SY, and MY) was not charted at 4 tenths or more in any one of the ice chart polygons that overlapped the line group, in any one of the two (late autumn and early summer), or sometimes four (east and west Arctic), charts for the group. One third of the borehole classifications were flagged even under this weak criterion.
4.1. High Thickness Associated With Multiyear Ice
 The ice volume histograms for ice in the Sverdrup Basin have modes centred at values between 3 and 5 m that indicate the presence of relatively undeformed multiyear ice. Such ice is common over large areas of the central Arctic Ocean [Wadhams and Horne, 1980]. However, since some extremely thick (6–9 m) modes exist in the Sverdrup Basin (Figure 11), this area harbours a second population of multiyear ice that is distinct from that of the Beaufort gyre. These thick modes infringe upon the thickness range associated primarily with ice ridges in the Arctic Ocean.
 The thick modes appear chiefly in the northern parts of Prince Gustaf Adolf Sea, MacLean Strait and Danish Strait. These channels provide the least obstructed routes for the southward drift of ice from the Arctic coast (Figure 3). The likely source for ice entering the Sverdrup Basin is a distinct zone of multiyear, seasonally land fast ice along the periphery of the Queen Elizabeth Islands. We suggest that the extreme modes of multiyear ice in the Sverdrup Basin originate within this zone [Serson, 1972; Walker and Wadhams, 1979; Hoar, 1980].
4.2. Level Ice is Relatively Abundant
 Precise estimates of the level ice fraction in the Sverdrup Basin cannot be derived from these observations because it is impossible to determine whether the ice thickness remains constant over the distance between widely spaced boreholes. However, an upper bound to the level ice fraction can be determined on the assumption that the peak in the ice volume histogram represents level ice (see Melling and Riedel  for discussion of the validity of this assumption). The peak in Figure 9, integrated to a maximum thickness of the 4 m incorporates about 80% of the ice drilled (62% of the ice volume). Such a high incidence of level ice is unusual in the North American Arctic except in very young ice fields [Melling and Riedel, 1996]. Areal level ice fractions of 50–60% are typical of the polar pack [Wadhams and Horne, 1980].
 The abundance of undeformed ice in the Sverdrup Basin is a reasonable consequence of the limitation on ice stress by fetch within the Queen Elizabeth Islands and by land-fast conditions that persist for half the year. Large tracts of undeformed first-year ice form every winter over 10–30% of the Sverdrup Basin. Because 10–20% of the ocean surface retains first-year ice at the end of summer, the second-year and multiyear ice categories are augmented annually by level ice in this amount. We suggest that the low-thickness multiyear ice modes are sustained locally within the Sverdrup Basin.
4.3. Very Thick Ice is Relatively Scarce
 The probability density function of ice thickness based on all the data from the Sverdrup Basin is plotted in log linear form in Figure 12. Ice thicker than 4 m contributes a tail of exponentially decreasing probability that extends to a maximum thickness of about 25 m. A probability density function representative of mixed-type pack ice of the same average thickness in the Beaufort Sea is plotted for comparison [Melling and Riedel, 1995]. The pack ice west of the Canadian Arctic Archipelago at 76°N has similar properties [McLaren et al., 1984]. The Beaufort data, measured as ice draft using sonar, have been scaled by 1.15 to convert to thickness. This value is a compromise between the value for level ice (about 1.1) and that for ridges (about 1.2).
 The mode at just below 2 m in the Beaufort Sea reflects a greater incidence of first-year ice relative to the Sverdrup Basin. The Beaufort curve also has a tail of exponentially decreasing probability for thick ice. However, the characteristic scale of this roll off is more than 50% larger than in the Sverdrup Basin. Thus ice thicker than 4 m is progressively less common in the Sverdrup Basin as the thickness increases. Ice 20 m thick is 10 times less common than in the Beaufort. For 25 m ice the ratio is 20. The thickest ice from more than 12,000 km of survey in the Sverdrup Basin was 26 m. The 350 km transect in the Beaufort Sea yielded ice that was 32 m thick, but ice estimated to be 44 m thick (37 m keel) has been observed in this area (Melling, unpublished data).
 Because of limited fetch in the Sverdrup Basin, ice-borne stress will not reach values sufficient to build large ridges, particularly in old ice. It is thus probable that the deep ridges in the Sverdrup Basin migrate from the zone of high ice deformation along the outer coast. The incidence of big ridges in this ice will be similar to, or perhaps even higher than, the incidence further south in the Beaufort Sea.
 If so, why is very thick ice rare in the Sverdrup Basin? Why does the deficiency increase with increasing thickness? The answer to both questions probably involves melting. There is evidence from three studies that ridges melt more rapidly than level ice [Rigby and Hanson, 1976; Wadhams, 1992; Schramm et al., 2000]. The heat source for melt is apparently the ocean, since ice is lost from the submerged portion of the ridge. It is not likely that the heat reached the ocean via solar radiation, because the ice concentration remains high in the Sverdrup Basin throughout the summer.
Melling et al.  argue that the upward flux of oceanic heat increases as Arctic waters flow into the Canadian Archipelago. The increase is likely an outcome of stronger mixing driven by tidal currents in the narrow shallow waterways between islands. Mixing weakens the stratification and increases the temperature of the pycnocline. These changes in temperature and salinity profiles are clear in Figure 13, which compares data from the Canada Basin and Norwegian Bay.
 The magnitude of the flux to the ice is not known. However, we can estimate the divergence of the upward flux from the difference between the temperature profiles in Figure 13, which represents the integral of flux divergence over transit time through the archipelago. The uppermost 250 metres of Norwegian Bay contain 340 MJ m−1 more sensible heat than the same layer in the Canada Basin. If currents move Arctic water across the Sverdrup Basin at 2 cm s−1, the upward flux of sensible heat must be at least 15 W m−2 merely to warm the halocline. Flow at this speed across the input section (315 km) between the surface and sill depth (approximately 100 m) transports enough seawater (0.6 Sv) to match present estimates of throughflow [Melling, 2000]. Initially, the upward diffusing heat will be trapped within the halocline where the vertical temperature gradient is weak. As the halocline warms, flux divergence within the halocline will decrease and the sensible heat flux to the ice will increase. The impact of supplying heat at 5–15 W m−2 to the bottom of thick sea has a dramatic effect on thickness, by reducing the growth in winter and increasing ablation in summer [Maykut and Untersteiner, 1971; Walker and Wadhams, 1979]. The effect is most pronounced on deep ridge keels, in which temperature gradients are weak [Schramm et al., 2000]. Because the conductive heat flux to an ice-snow interface at −20°C is less than 10 W m−2 if the ice is thicker than 4 m, an oceanic heat flux of 10 W m−2 will halt ice accretion on thicker ice except during the coldest part of the winter and will ablate ice at other times.
4.4. Evidence for Loss of Thick Ice Through Preferential Melting
 A decrease in deep keels and very thick ice with distance from the Arctic Ocean is a logical repercussion of the preceding section. The concept can be explored using the observations from 1977, which encompassed the greatest geographic range from northwest to southeast in the Sverdrup Basin. Mean, modal and maximum ice thickness are plotted against distance from the Arctic coast in Figure 14. Near the outer coast there are three modes, one associated with seasonal ice and two with very thick multiyear ice. The seasonal mode shows little change with distance until the transect reaches Norwegian Bay (last point). The thickness of the intermediate multiyear mode decreases by 25% and disappears in Norwegian Bay. The thickest mode disappears within 250 km of the coast. The maximum drops dramatically by a factor of 3 and the mean decreases from 5.5 m to less than 2 m. These data are consistent with preferential melting of thick ice as it drifts southeast across the Sverdrup Basin.
4.5. Temporal Change in 1970s in Core Zone
Figure 14 might also be interpreted as evidence for an increase during the 1970s in the thickness and ridging of ice imported from the Arctic Ocean. In order to explore this possibility, data were examined for a core zone north of 77°N between 100° and 110°W, which was most consistently observed throughout the decade. The area encompasses Prince Gustaf Adolf Sea, MacLean Strait and Danish Strait. Results are displayed in Figure 15. Modes of ice of local origin (thickness <3.6 m) show no trend in this area during the 1970s. Their variation resembles that of land-fast ice at Isachsen. The very thick multiyear ice modes are more common in the second half of the decade. The maximum ice thickness increases by a factor of two over the same period, and the mean increases by 20%, from 3.5 to 4.2 m. Clearly, changes in the supply of ice from the Arctic Ocean are an important factor in the ice climate of the Sverdrup Basin.
 The ice charts of the Canadian Ice Service provide some insight. Figure 4 reveals a low old ice fraction (about 2 tenths) in zones 7–10 in September 1976. Since thick first-year ice covers most of the remaining area at this time, the old ice must have been lost in the last quarter of 1975. Indeed, both Figure 4 and satellite imagery presented by Marko  suggest that some of the loss may already have occurred by September of that year. The increase in old ice fraction to about 7 tenths by September 1977 must reflect in part the October 1 birthday for second-year ice, but may also be evidence for an influx of heavy multiyear ice from Arctic Ocean during the autumn of 1976 or the summer of 1977. Incidental notes by Wetzel  refer to “appreciable movements of ice to the south and east” during the latter period. Zones 7–9 again lost old ice in the last quarter of 1977, only to acquire more, either by inflow or conversion, in the late autumn of 1978. The concentration of old ice in the Sverdrup Basin did not drop again to 1976 levels until the infamous summer of 1998 [Jeffers et al., 2001].
 In September 1971, the concentration of old ice was also very low in zones 7–9, although not in zone 10. Again the prevalence of thick first-year ice at this time means that older ice was flushed from these zones during the last quarter of 1970. By similar argument, the old ice must have been replenished in the last quarter of 1971. The survey of the following spring, 1972, was the only one during the first half of the decade to find a mode of very thick multiyear ice (Figure 15). Ice drift was again significant in the Prince Gustaf Adolf Sea corridor during the following summer, 1972 [Marko, 1977].
 The ice of the Sverdrup Basin is land fast (10 tenths) between October–November and late July. It becomes mobile in most areas in late summer, but its concentration remains high. At the time of minimum extent in early September, the average concentration is 9 tenths in the northwestern zones of the basin, 7 tenths in the south and 5 tenths in Norwegian Bay. Interannual fluctuations in late-summer ice coverage obscure any evidence of trend. A decadal cycle contributes variability to the time series of both the total and the multiyear ice concentrations. Because the reputedly extreme conditions of 1998 are similar to occurrences in 1962 and 1971, there is little basis on which to view them as evidence for anthropogenic change.
 The wintertime composition of the pack derived from Canadian ice charts as (my, sy, fy) ranges between (8, 1, 1) tenths near the outer coast to (5, 2, 3) tenths in the south and (3, 2, 5) in Norwegian Bay. Composition derived from ice thickness measurements during the 1970s, (4, 3, 3) tenths, incorporates a larger presence of second-year ice. This discrepancy may indicate that level multiyear ice in the Sverdrup Basin can be thinner than the 3 m minimum assumed in this analysis. Depending on zone, 5–20% of the old ice melts in summer and 20–50% of the first-year ice.
 The maximum thickness of first-year ice in the Sverdrup Basin varies interannually between 1.8 and 2.6 m. A large interannual variation in snow cover (15–75 cm) is the dominant influence on ice thickness. The average thickness of the Sverdrup Basin ice pack is 3.4 m in late winter, a value close to the Arctic Ocean average. It is much less than the 5–7 m value considered typical of adjacent areas of the outer coast [Bourke and Garrett, 1987]. On a 50 km scale, the average thickness ranges between 1.6 and 5.5 m.
 Level ice is 1 times more abundant in the Sverdrup Basin than in the Canada Basin. Narrow level ice modes occur at thickness values typical of first-year, second-year, and multiyear ice. The prevalence of these narrow modes indicates that an appreciable fraction of the pack develops within the Sverdrup Basin where ridge-building forces are fetch-limited and weak.
 Very thick, quasi-uniform multiyear ice occurs in the northwestern part of the Sverdrup Basin. This ice is almost certainly forced between the islands in summer and autumn by the persistent pressure of the Arctic polar pack on the outer coast. There are thus two populations of pack ice in the Sverdrup Basin. One is a lightly ridged mix of first-year, second-year, and multiyear ice that is equilibrated to local conditions of growth, melt and deformation. The other is predominantly multiyear ice comprised of thick uniform floes and large ridges that originates in the very dynamic ice environment on the outer coast.
 Deep ridge keels are much less common in the Sverdrup Basin than in adjacent areas of the Arctic Ocean. The difference cannot be a consequence of diluting imported ice with local less deformed ice since the deficit in the Sverdrup Basin increases with keel depth. The difference exists because ice of deep draft melts preferentially in response to an enhanced flux of oceanic heat within the Sverdrup Basin. This effect contributes to the progressive loss of deep ridge keels with the southeastward movement of ice. The response of the thick ice in the Sverdrup Basin to the absence of a cold halocline layer is evidence for the strong sensitivity of ice thickness to oceanic heat flux. It emphasizes the importance of the cold halocline to the preservation of the present cover of perennial ice over the Arctic Ocean.
 The flux of ice through the Sverdrup Basin is intermittent. It is interrupted annually by the onset of land-fast conditions and interannually in late summer and autumn by factors that are not understood. There is evidence of a three-phase cycle of ice movement in the western half of the basin: 1) A stable phase when there is a modest export of ice through southern channels during the second half of the year; this phase may continue for several years. 2) An occasional rapid flushing of ice into Parry Channel from zones 8 (south), 10 and 11 during the late summer and autumn of a single year. 3) In the following summer/autumn, a rapid flushing of ice from one or more of zones 7–9, with a compensating influx of heavy old ice from the Arctic coastal region. The selective melting of thick ice is most influential during stage 1.
 Flushing events (stages 2–3) in the western half of the basin occurred between September 1970 and December 1971, between August 1975 and December 1978, and between July 1998 and September 2000. There were lesser events in the middle 1980s and middle 1990s. It is noted that the transit of ‘Hobson's Choice’ ice island, which was completed within one melt season (1991), is evidence for a more rapid flushing of the eastern half of the basin at times when the ice plugs break out of Peary and Sverdrup Channels.
 The current that carries ice to the southeast across the Sverdrup Basin is generally attributed to a drop in sea level across the Canadian Archipelago to Baffin Bay, although there is only weak direct evidence for its existence. The effect of current on ice drift is enhanced by the regional pattern of atmospheric pressure that favours northwesterly winds. The ice moves intermittently because its resistance to compression and buckling impedes free movement through the narrow channels that connect the Sverdrup Basin to Baffin Bay. Heavy ice leaving the Sverdrup Basin may drift west into Viscount Melville Sound, south into M'Clintock Channel and Peel Sound and east into Jones and Lancaster Sounds.
 It is commonly assumed that a warming climate will bring lighter ice conditions to Arctic waters. This study raises the possibility that the effect of a warming climate may well be contrary in northern Canadian waters. The supply of very heavy multiyear ice to the Northwest Passage is presently controlled by the tendency of pack ice to block the southern exits from the Sverdrup Basin in winter and the northwestern entry points with varying efficiency for years at a time. The thick, heavily ridged multiyear ice from the Arctic Ocean thus makes slow progress through the Canadian Arctic Archipelago. During the years spent in transit, deep ridge keels are severely ablated under the influence of oceanic heat and the ice thins appreciably. Ice strength depends on ice temperature. Because the melting season will be longer in a warmer climate, ice strength sufficient to maintain bridges will exist for a shorter time than at present. Thus the ice bridges that ring the Sverdrup Basin in winter will form later in the year and break up earlier. Flushing events will become more frequent. On average the heavy ice from the northwest will move more rapidly through the Sverdrup Basin. Not only will the flux of ice to northern shipping routes increase, but this ice will be thicker and more heavily ridged, having been subject to oceanic heat flux for a shorter time in transit. The impact of increased multiyear ice hazard on navigational use of the Northwest Passage will undoubtedly be negative.
 The impact of this climate-warming scenario on the freshwater flux to the Labrador Seas may be less significant. This impact is mediated by the influence of freshwater on the intensity of deep convection in the Labrador Sea in late winter, which in turn forces one component of the global thermohaline circulation. An increased freshwater flux may reduce the surface density of the Labrador Sea to the point where convection is suppressed. In the present scenario, the postulated increase in ice area flux would increase the supply of freshwater to the Labrador Sea, but the postulated decrease in ice ablation within the Canadian Arctic Archipelago would not. The latter change simply alters the partition of the freshwater flux between ice and seawater components within the Canadian Arctic Archipelago. The ultimate impact of freshwater in the Labrador Sea is felt only when all ice has melted.
 This study was stimulated by a chance discovery of the incredible set of drill hole observations collected and compiled by Vern Wetzel (retired), formerly of Sun Oil Ltd, Calgary. I am indebted to Mr. Wetzel for his insight, hard work and assistance in clarifying the technical aspects of his impressive 10 year observational programme. I thank Nick Melling and Rob Bowen for their effort in computer entry, mapping and quality control for transcription errors. Richard Chagnon of the Canadian Ice Service endured my repeated requests for an early release of the GIS version of the Canadian weekly ice charts, without which the chart analysis would have been tedious and imprecise. Ze'ev Gedalof lent his GIS skills to several iterations of the ice chart analysis. I acknowledge funding support from the federal Programme of Energy Research and Development (Project 23119 Arctic Canada Watch and Project 23112 Ice Thickness in a Changing Climate), from the Climate Change Action Fund (Extreme Summer of 1998 Project) and from Fisheries and Oceans Canada.