Spatial and temporal variability of sea ice in the southern Beaufort Sea and Amundsen Gulf: 1980–2004



[1] Changing extent, location, and motion of the Arctic perennial pack affect the annual evolution of seasonal ice zones. Canadian Ice Service digital ice charts covering the southern Beaufort Sea and Amundsen Gulf are used to illustrate summer and winter conditions and trends between 1980 and 2004 for several sea ice stages of development. Results illustrate average sea ice conditions within the region in summer and winter for predominant sea ice types and changes in the relative concentration of sea ice types in summer and winter. In summer, a trend toward increased old sea ice concentration occurred near the mouth of Amundsen Gulf, with a trend toward decreasing summer first-year sea ice farther west. In winter, increasing thick first-year sea ice extent appears to be replacing young sea ice within the flaw lead system in the region. The dynamically driven breakup of sea ice in spring in the Amundsen Gulf is a highly variable event taking anywhere between 2 and 22 weeks to completely remove ice from the gulf. The timing and duration of the open water season depends upon the extent and timing of old ice influx. Freezeup occurs very quickly, proceeding from west to east with little temporal variability. The results of this paper are used to set the context for the Canadian Arctic Shelf Exchange Study (CASES) in terms of sea ice dynamic and thermodynamic processes.

1. Introduction

[2] Recent variability in Arctic sea ice thickness and extent has been attributed to seasonal changes in dynamic and thermodynamic forcing [Zhang et al., 2000]. A shift in the sea level pressure pattern over the central Arctic Ocean during the 1990s led to a weakening of the Beaufort Gyre [Lukovich and Barber, 2006] and increased observations of southerly and easterly winds in the western Arctic [Rothrock et al., 1999] which advect ice north and west away from the continental coast. Rothrock et al. [1999] also note that southerly winds were accompanied by positive anomalies in surface air temperatures which increased melt, leading to thinner perennial ice and more open water. Increased open water and thinner perennial sea ice contributes to increases in heat content in the upper water column further reducing ice thickness and extent [Rigor and Wallace, 2004]. Polyakov and Johnson [2000] also note that a 60 to 80 year pressure oscillation came in phase with the decadal decrease in sea level pressure during the 1990s, accounting for the persistence and amplitude of the negative pressure anomaly, and enhanced warm air advection. These factors contributed to sea ice extent reduction in the 1990s, and the record negative sea ice extent anomalies in the Beaufort Sea during 2002 [Serreze et al., 2003], 2003, and 2004 [Stroeve et al., 2005].

[3] Regional and interannual sea ice motion, such as the circulation of ice in the Beaufort Gyre and convergence and divergence of sea ice along the Beaufort Sea coast, have occurred since the beginning of the satellite record [Thorndike, 1986; Serreze et al., 1989]. Subdecadal cycles in sea ice thickness have also been observed [Mysak and Manak, 1989]. Thinning of perennial sea ice in the Arctic has been related to transpolar sea ice drift exporting sea ice out of the Arctic through Fram Strait at approximately 7-year intervals [Mysak and Manak, 1989].

[4] The longer-term in situ record suggests a continuing trend toward polar-amplified warming of surface air temperatures [Polyakov et al., 2002], increased cyclone activity [Zhang et al., 2004], and earlier melt onset and later refreeze [Belchansky et al., 2004]. Models however, vary widely in their predictions of Arctic sea ice extent and stage of development. Forecasting scenarios range from a stabilization of current conditions to the possibility of a seasonally ice-free Arctic [Meehl et al., 2007]. Results from the Community Climate System Model also show wide variation of predicted future decreases in perennial sea ice volume [Holland et al., 2006], which has been shown by Rothrock et al. [1999] to have been in significant decline since the 1960s. Despite broad disagreement on the rate and extent of the sea ice decay, the spatial pattern of summer melt favors retreat toward the west coast of the Canadian Arctic Archipelago, exposing a larger seasonally ice-free area [Flato et al., 2000].

[5] Sea ice thickness and extent determines the exchange of heat between the ocean and atmosphere [Maykut, 1982], is an important variable in determining the strength of sea ice in a region [Andreas et al., 1993], and has an obvious influence on the ice mass flux [Aagaard and Carmack, 1989] and therefore the freshwater storage in a region [Wadhams, 2000]. Polynyas and flaw leads may be open months before the annual melt begins, allowing for enhanced ocean-atmosphere heat exchange year round and albedo feedbacks which contribute to the instability or mobility of the surrounding ice pack. These areas of reduced sea ice concentration and/or thickness are important to biological and physical oceanography. Sea ice provides a substrate for significant primary production within and below the ice, as well as affording mammals an opportunity to forage and rear young [Dunbar, 1981]. Consequently, Inuit people have hunted in and around these areas for at least the past three thousand years [Smith et al., 1990].

[6] Monitoring long-term variation of sea ice extent is very important in both the estimation of current climate trends and the prediction of future change. Satellite-borne passive microwave (PMW) sensors are typically relied upon to monitor sea ice temporal and spatial variation of sea ice at the hemispheric scale. PMW sea ice extent data have shown significant decreases in northern hemisphere, the amount of which is dependent upon the time period and area(s) studied [i.e., Stroeve et al., 2005; Serreze et al., 2003; Zhang et al., 2000; Gloersen et al., 1999; Parkinson et al., 1999].

[7] At the regional scale, Drobot and Maslanik [2003] explored the interannual variability of Beaufort Sea ice conditions, noting that summer ice conditions north of Alaska were related to dynamic mechanisms in the preceding winter and thermodynamic processes during summer. Sea ice retreat and formation play an integral role in creating a physically favorable environment for biological production in western Amundsen Gulf [Arrigo and van Dijken, 2004]. Barber and Hanesiak [2004] note the scale dependence of sea ice anomalies in the southern Beaufort Sea and Amundsen Gulf, associating local sea ice anomalies with local-scale advection and regional-scale sea level pressure and surface temperatures. Timing of sea ice formation and breakup in Amundsen Gulf for the years 1964–1974 were examined by Hammill [1987], using ice summaries and analysis reports for the Canadian Arctic prepared by the Canadian Department of Transport (1964–1972) and The Sea Ice Atlas of Arctic Canada (1973, 1974). For the purpose of classifying ice retreat and formation into early, medium and late years, Amundsen Gulf was considered ice free when 80% of the gulf was open, and freezeup was considered complete when 80% of the gulf was covered in sea ice [Hammill, 1987]. The eleven years studied by Hammill [1987] were almost evenly distributed into early breakup (1968, 1970, 1971, 1973), medium breakup (1965, 1969, 1972) and late breakup (1964, 1966, 1967, 1974). Early ice formation occurred in 1964, 1967, 1974 while medium ice formation years were 1965, 1966, 1970, 1972 while ice formation occurred late in 1968, 1969, 1971 and 1973 [Hammill, 1987]. Hammill [1987] also describes the location of a polynya in Amundsen Gulf, although in 1964, 1968, and 1973 it did not form at all, and for only two of the other 9 years did it form in the same position.

[8] In this study, 25 years (1980–2004) of Canadian Ice Service (CIS) digital ice chart data over Amundsen Gulf and the southeastern Beaufort Sea are used to determine the extent and change of interactions between a seasonal sea ice zone and the perennial Arctic pack. This data set is unique in that it contains sea ice stage of development information, including classifications of total, old (a superset of second-year and multiyear sea ice), first-year (a superset of thin, medium and thick first-year), young and new sea ice. The CIS data set delineates stages of development by thickness, including new ice (hi < 10 cm), young ice (hi = 10–30 cm) and first-year ice (hi > 30 cm). Subsets of first-year ice include thick first-year (hi > 120 cm), medium first-year (70 cm < hi < 120 cm) and thin first-year (30 cm < hi < 70 cm) sea ice [Fequet, 2002]. Old sea ice is defined as sea ice which has survived a summer's melt, including second-year and multiyear sea ice and is not defined by thickness. Total sea ice concentration is the sum of old, first-year, young and new sea ice concentrations (expressed in tenths). A focus is placed on the conditions of sea ice development in Amundsen Gulf and southeastern Beaufort Sea where the seasonal sea ice zone and perennial Arctic pack interact. This thermodynamically and dynamically active area has been the focus of recent attention, including the Canadian Arctic Shelf Exchange Study (CASES), ArcticNet, and the International Polar Year (IPY) Circumpolar Flaw Lead (CFL) Study. This area of the continental shelf is the only North American analog to the vast Russian continental shelves. The interaction of annual and old sea ice in these regions gives rise to flaw lead formation, which has implications on ecology, shipping, and oil exploration.

[9] Amount, position, and seasonal to interannual cycles of these ice types allow for: (1) characterization of mean winter and mean summer sea ice concentration as a function of sea ice type in the region; (2) calculating the trend in concentration of each sea ice type during winter and summer over the time series, (3) quantifying patterns in the timing and duration of breakup and freezeup in the region and (4) discussion of the variability of breakup and freezeup in Amundsen Gulf in terms of regional dynamics and the evolving seasonal ice cover in this area. We consider this regional sea ice climatology an important component of the ocean-sea ice-atmosphere system in the southern Beaufort Sea and in particular a key element in how changes in sea ice conditions may affect physical-biological coupling in the region.

2. Study Area, Data, and Methods

2.1. Amundsen Gulf and the Southern Beaufort Sea

[10] Amundsen Gulf is connected to the Beaufort Sea through an outlet bounded to the north by Banks Island and to the south by Cape Bathurst (Figure 1). Landfast ice along the perimeter of the basin extends generally to the continental shelf break, where the water depth is 50 to 100 m. West of Banks Island lies the perennial Beaufort Sea ice pack. Its position is correlated with large-scale pressure patterns, and motion within the pack is largely determined by the local geostrophic wind [Thorndike and Colony, 1982]. Annually recurrent leads make up the extensive flaw lead system in the southeastern Beaufort Sea which persists from the completion of freezeup to break up in spring [Carmack and MacDonald, 2002; Barber and Hanesiak, 2004]. The regional flaw lead system forms in three main areas: (1) parallel to the continental coastline and north of the landfast ice edge along the coast of the Tuktoyaktuk Peninsula, Cape Bathurst, Franklin Bay and Cape Parry, (2) west of the landfast ice edge along the west and southwest coast of Banks Island and, (3) along the west coast of Victoria Island in eastern Amundsen Gulf (Figure 1).

Figure 1.

The southern Beaufort Sea and Amundsen Gulf, western Canadian Arctic. Black lines delineate the study area used in section 5. The area normally occupied by the flaw lead system is shown in gray.

[11] In some years between late April and mid-May, a lead also forms at the interface between the mobile ice pack and high-concentration sea ice in eastern Amundsen Gulf, running north-south between the south tip of Banks Island and Cape Bathurst. This dynamic area is further characterized by a constricted airflow which lends to accelerations of air velocities downwind [Torneby, 2006], relatively strong tidal mixing [Kliem and Greenberg, 2003], and upwelling along shallow shelves [Carmack and MacDonald, 2002]. Using this area of open water, ice sometimes enters Amundsen Gulf blocking ice advection out of the gulf and stalling the open water season.

2.2. Sea Ice Concentration Data

[12] CIS digital ice charts are available from an online archive at This sea ice data set integrates a number of data sources including aerial and marine surveys, satellite remote sensing data including Radarsat-1, NOAA AVHRR and Envisat ASAR as well as in situ observations. CIS digital ice charts are created manually by expert ice analysts at the CIS who derive total and partial concentrations for each sea ice stage of development. Agnew and Howell [2003] used CIS ice chart data as a baseline to evaluate PMW ice concentration data. Their results showed a considerable discrepancy between CIS and PMW data; tie points used in calibrating the NASA Team sea ice extent algorithm perform better at the hemispheric scale rather than for specific areas, performing poorly in seasonal ice zone areas throughout the annual cycle, and during melt and freezeup conditions, where they consistently underestimate total sea ice extent. They also showed the usefulness of the CIS data set in long-term regional monitoring of sea ice, particularly because the CIS data has comparatively high spatial resolution and are composed using multiple data sets to categorize sea ice cover. Conversely, ice concentration and extent data produced using PMW data contain only total ice concentration and no thickness information. The stability of the CIS digital data set through time may be affected by advances in remote sensing platforms and changing shipping routes which subject the data set to shifts in bias and variance, but the Canadian Ice Service Archive Documentation Series [2006] quality index indicates statistical analysis is appropriate in the region. We used 802 sea ice charts available for the western Arctic between 1 January 1980 and 31 December 2004. Prior to 1980, CIS data were only available for year weeks 24–48, which coincided with the majority of ice retreat and advance in the Beaufort Sea region but did not contain any winter sea ice data. Starting in 1980, winter chart data were produced monthly, while between year-weeks 23 (∼10 June) and year-week 48 (∼1 December), weekly charts of total and partial concentrations of sea ice and type are available.

[13] Each CIS data file was converted from its native .e00 GIS format to a grid in a Lambert conformal conic projection where each grid cell was four km2 (n = 99977 grid cells). Four square kilometers was chosen because it is small enough to adequately capture coastlines and polygon edges without losing valid nearshore ice data but still allowed for manageable data volumes. Sea ice concentration (in tenths) for each stage of development was extracted.

[14] Mean summer sea ice concentration for each grid cell was calculated by averaging all digital ice chart data for July, August and September (JAS) in each year. Mean winter sea ice concentration for each grid cell was calculated by averaging all CIS chart data for each grid cell in January, February and March (JFM) in each year. It is possible for the monthly winter CIS data set to have misrepresented the presence of dynamic features within the study area such as leads. However, independent analysis using daily PMW data indicates that lead areas in the region are persistent both within the winter season and interannually [Barber and Hanesiak, 2004], making us confident that the monthly charts represent average winter conditions in the region. The mean statistic was used because the CIS data set contains sea ice concentration in decimal tenths (i.e., 0.3 tenths, 0.6 tenths, which convey trace concentrations and 9.7 tenths, which describes very high concentration sea ice which is typically not landfast).

[15] To determine the trend in sea ice concentration for each ice type in summer and winter, a least squares linear regression was calculated for the concentration at each grid cell over the 25-year period [Parkinson et al., 1999], where the slope of the regression indicates the trend (percent per year) for each ice type. Although we explored various nonlinear models in the regression analysis, an examination of the residual plots from the linear model appeared to be uniformly distributed through the range of the models. Therefore the linear model was judged most statistically appropriate for the trend analysis performed. For each trend at each grid cell, an F test (df = 23) was used to test the null hypothesis of slope = 0. We rejected the null hypothesis for all those grid cells containing trends with statistical significance greater than 90%, mapping only those trends which are significant at the 90, 95 and 99% level.

[16] Temporal variability in the onset of breakup and freezeup was examined by identifying the year-week (YW) for each grid cell when total ice concentration reached thresholds used by Hammill [1987]. The start of breakup was defined as the YW when sea ice concentration decreased to ≤80% and continued until the cell contained ≤20% sea ice. Freezeup began when the grid cell contained ≥20% sea ice and was completed when the cell contained ≥80% sea ice. Advection of large floes from the Beaufort Sea into the Amundsen Gulf during summer can introduce short-term increases in total ice concentration above 20% during summer, so freezeup onset was further constrained temporally, occurring no earlier than YW 33 (third week in August).

[17] Finally, Amundsen Gulf was defined by a north-south line between Cape Bathurst and Cape Kellett [after Kwok, 2006] and all grid cells were averaged to produce one value of sea ice concentration for the gulf (Figure 1). This was done so direct comparisons of interannual variability and trends in the timing and duration of breakup and freezeup could be made with data from Hammill [1987].

3. Sea Ice Climatology and Trends

3.1. Sea Ice in the Southern Beaufort Sea and Amundsen Gulf

3.1.1. Winter Sea Ice Climatology: 1980–2004

[18] During the winter months (JFM), the mean total sea ice coverage over the study region is 3.9 × 105 km2 (∼98%). The average winter area not covered by sea ice is 8900 km2 (∼2%). The region contains near-complete winter sea ice cover in each of the 25 years studied here. On average, three sea ice regimes are observed (Figure 2a): (1) a landfast (10/10ths) sea ice fringe in dark red, (2) a high-concentration mobile ice area in eastern Amundsen Gulf shown in yellow, and (3) a lower concentration mobile sea ice zone in the southern Beaufort Sea and western Amundsen Gulf shown in light blue. The average winter old ice coverage in the study region is 1.4 × 105 km2 (∼36%) within its average winter position (Figure 2b). Old sea ice is located offshore west of Banks Island and north of Cape Kellett. Old sea ice concentration increases from 2/10ths moving northwest as the central Arctic perennial pack is approached and the old ice concentration becomes 8–9/10ths (Figure 2b). Interannually, old ice concentration in the region is variable. The presence of old ice in Amundsen Gulf during winter has occurred 5 times (1986, 1987, 1992, 1993, 1997) (Figure 3).

Figure 2.

Average winter sea ice concentration by type between 1980 and 2004 (a) total sea ice, (b) old sea ice, (c) first-year, (d) thick first-year, and (e) young and new sea ice.

Figure 3.

Average winter old sea ice concentration by year from 1980 to 2004.

[19] Four of these occurrences (1986, 1987, 1993, 1997) are associated with a reduction of winter old ice concentration within the Arctic pack, while in 1992 an area of 1/10 old ice occurred in Thesiger Bay while the old ice concentration in the Beaufort Sea remained high (9/10ths).

[20] The study area is ∼62% (2.5 × 105 km2) covered by seasonal sea ice (first year, young and new ice) in winter on average (Figures 2c, 2d, and 2e). first-year sea ice occurs within the nearshore regions of the southern Beaufort Sea, and within Amundsen Gulf (Figures 2c and 4) where old ice concentrations are very low.

Figure 4.

Average winter first-year sea ice concentration by year from 1980 to 2004.

[21] The average annual sea ice exchange between Amundsen Gulf and the Beaufort Sea was 85,000 km2 to the Beaufort Sea, predominantly in fall and winter [Kwok, 2006]. The sea ice motion responsible for this export in fall and winter creates variability in sea ice stage of development within the seasonal sea ice zone as leads continually open and refreeze. In winter, substantial amounts of thin first-year, young and new sea ice are present on average during winter (Figures 2e, 5, and 6) , between the landfast ice edge and areas of thick first-year ice within Amundsen Gulf (Figure 2d).

Figure 5.

Average winter thin first-year sea ice concentration by year from 1980 to 2004.

Figure 6.

Average winter young and new sea ice concentration by year from 1980 to 2004.

[22] First year sea ice concentration and extent is interannually variable along the west coast of Banks Island and in Amundsen Gulf (Figure 4). The position of the flaw lead system is denoted by new and young ice (Figures 2e and 6). The relative degree to which the flaw lead forms changes interannually, along with the relative concentrations of thin first-year, young and new sea ice that reside within it.

[23] Interannual variation of thin first-year (Figure 5) and young and new sea ice (Figure 6) show the extent to which the flaw lead system forms interannually. The flaw lead may form off the west coast of Banks Island and the north coast of the Tuktoyaktuk Peninsula to Cape Bathurst. This occurred in 1982, 1983, 1984, 1987, 1991, 1992, 2002, and 2003. The flaw lead system occurred within Amundsen Gulf along the southeast coast of Banks Island, the west coast of Victoria Island and/or across the mouths of Franklin and Darnley Bays. This occurred in 1980, 1981, 1986, 1988, 1989, 1994, 1995, 1997, 1998, 1999 and 2000 and 2004. The study region did not contain any thin first-year, young or new sea ice in 1993 or 1996 in winter (Figures 5 and 6). Ten of the twenty-five years contain evidence of reduced ice thickness or concentration between Cape Lambton and Cape Parry. Seven of those ten years occurred in the 1980s (1980, 1981, 1983, 1985, 1986, 1987, 1988) while only 1995, 1998 and 2004 contain this feature since 1988. This suggests that the nature of the flaw lead system in the region may have changed since the late 1980s.

[24] In winter, new, young and thin first-year sea ice types account for 2–9/10ths of the total sea ice concentration in the nearshore regions of Amundsen Gulf, along the west coast of Banks Island and the Tuktoyaktuk Peninsula. Western Amundsen Gulf between Thesiger Bay and Franklin Bay also contains substantial new and young sea ice on average (Figure 2e). The flaw lead areas may not necessarily be an area of reduced ice concentration (Figure 2a) but may be an area of reduced ice thickness, containing a substantial amount of thin first-year and young sea ice (Figure 6).

3.1.2. Summer Sea Ice Climatology: 1980–2004

[25] The mean summer sea ice area in the study region is 1.8 × 105 km2 (∼46%); the average summer open water area is 2.2 × 105 km2 (∼54%).

[26] In summer, total sea ice concentration increases from 1 to 2/10ths at the mouth of Amundsen Gulf moving northwest toward the perennial Arctic pack. Much of eastern Amundsen Gulf ice-free (Figures 7a and 8) , but a tongue of 1 to 2/10ths sea ice reaches from the diffuse edge of the summer pack into Amundsen Gulf and south into Franklin Bay on average (Figure 7a). The continental shelf north of the Tuktoyaktuk Peninsula and across the mouth of Liverpool Bay are also ice-free on average (Figure 7a).

Figure 7.

Average summer sea ice concentration by type between 1980 and 2004 (a) total sea ice, (b) old sea ice, and (c) first-year sea ice.

Figure 8.

Average summer total sea ice concentration by year from 1980 to 2004.

[27] Almost all of the summer sea ice in the region is composed of old (78%) and thick first-year (20%) sea ice on average (Figures 7b and 7c). Where the total sea ice concentration is ≤2/10ths (Figure 7a), about half is old ice (Figure 7b) and half is thick first-year (Figure 7c). Northwest of the 2/10ths total sea ice concentration contour (Figure 7a) the total concentration becomes heavily dependent upon old ice (Figure 7b). Medium first-year, thin first-year, young and new ice are present only in negligible amounts (<1/10th) spatially coincident with 8/10ths total sea ice. This may occur because the Arctic pack contains some open water areas allowing space for new ice production, while the ocean below is kept near freezing by the relatively high sea ice concentration and therefore high albedo. Any energy absorbed by the surface would first be used for bottom and lateral ablation of the sea ice present [Wadhams, 2000].

[28] Summer sea ice concentration in the region is variable between years. In 1991, the entire area remained almost completely ice covered in summer, save for a small ice-free area along the south coast of Banks Island and north of the Tuktoyaktuk Peninsula (Figure 8: 1991). Other heavy sea ice years include 1986, 1989, 1994, 1996, 2000, 2001, 2002, 2003, and 2004 (Figure 8). The last four heavy summer ice years (2001–2004), coincide with record hemispheric-scale reductions in the area occupied by the perennial Arctic pack [Serreze et al., 2003; Stroeve et al., 2005]. The northwest quadrant of the study area is normally occupied by a large area of 8–9/10ths sea ice. In 1998, the portion of the study area covered was very small and contained only ≤ 7/10ths sea ice concentration. Other years in which the regional ice concentration was comparatively light also occur, including 1982, 1988, 1993, 1995, 1998 and 1999 (Figure 8). These years are characterized by a retreat of the old ice pack to the north and northwest, leaving western Amundsen Gulf ice-free and even creating large ice-free areas along the west coast of Banks Island. Since 1998, summer sea ice in the southern Beaufort Sea and Amundsen Gulf has recovered to a degree (Figure 8); by 2000 the summer sea ice concentration and extent was similar to pre-1998 conditions. However, 2002, 2003 and 2004 suggest that the extent of high-concentration perennial sea ice in the southern Beaufort Sea has been reduced and has shifted east toward Banks Island, leaving decreased old sea ice concentration farther west (Figures 8 and 9) . The total sea ice concentration in the area depends almost completely on old sea ice in each year (Figures 8 and 9).

Figure 9.

Average summer old sea ice concentration by year from 1980 to 2004.

[29] Summer first-year sea ice concentration is low, and its extent is interannually variable (Figure 10). Prior to 1997 (except 1988 and 1995), first-year sea ice was generally included in the old ice pack (Figures 9 and 10). However, in 1997 almost no summer first-year sea ice occurred, and similar conditions prevailed during the summers of 1998 and 1999 (Figure 10).

Figure 10.

Average summer first-year sea ice concentration by year from 1980 to 2004.

[30] In 2000, summer first-year sea ice recovered somewhat but was not spatially coincident with the old ice pack; rather it occurred to the east and south of the old ice (Figures 9 and 10). The same conditions prevailed in 2001, 2002, 2003, before the summer first-year sea ice was again collocated with the old ice pack in 2004. It appears that low summer sea ice concentration and extent in 1997, 1998 and 1999 created a situation where the old ice and first-year sea ice became spatially disconnected for several years. Reductions in the summer old sea ice extent (Figure 9: 1998, 1999, 2000) forced increased first-year sea ice extent in winter (Figure 4: 1998, 1999, 2000, 2002, 2003) in order to completely cover the region in ice. The remnants of high winter first-year ice concentrations can be seen in the summer data (Figure 10: 2000, 2001, 2002 2003). By 2004, summer old and first-year sea ice had become spatially coincident similar to pre-1997 conditions but first-year sea ice replaced old ice in the west (Figures 9 and 10) while lower concentrations of first-year sea ice were present in the edge of the perennial pack (Figure 10).

[31] This shift in the location of seasonal and perennial sea ice gives us insight into how the ice regime will potentially operate in times of reduced or eliminated summer old ice [i.e., Holland et al., 2006]. It appears that under current climatic conditions, seasonal sea ice extent and concentration will increase to make up for the perennial ice loss. This has already been shown to impact the formation of the annually recurrent flaw lead system. In years when the perennial ice pack retreats to the northwest, seasonal sea ice is allowed to grow thicker in a greater portion of the region. This is due to reduction in dynamic processes within the ice field, normally caused by the Arctic pack moving against the landfast ice fringe. The seasonal sea ice is larger, thicker and less affect by any connection to the Arctic ice pack, devolving the flaw lead system.

3.2 Sea Ice Concentration Trends

3.2.1. Winter Trends in Sea Ice Concentration: 1980–2004

[32] Winter total sea ice concentration shows no trend because the region is nearly completely ice covered each year. Statistically significant trends in winter first-year, thick first-year and young ice have occurred (Figure 11).

Figure 11.

Winter sea ice (left) concentration trend (% per year) and (right) statistical significance for (a, b) first-year, (c, d) thick first-year, and (e, f) young sea ice. Black areas in the right column indicate statistical significance at the 99% level, dark gray areas indicate statistical significance at the 95% level, and light gray areas indicate statistical significance at the 90% level.

[33] Winter first-year sea ice showed an increasing trend of 1 to 2.5%/year west of Banks Island, within western Amundsen Gulf, across the Tuktoyaktuk Peninsula and along the west coast of Victoria Island (Figure 11a). The statistical significance of these trends is shown at three levels (90%, 95%, and 99%) in Figure 11b. The most dramatic increasing trend (>2% per year) in first-year ice occurred immediately west of Banks Island is statistically significant at the 99% level (Figure 11b). This area is part of the flaw lead system on average, containing 5 to 6/10ths new and young sea ice in winter (Figure 2e). A similar positive trend also occurred north of the Tuktoyaktuk Peninsula that contains 4–5/10ths new and young sea ice on average (Figure 2e) significant at the 95% level (Figure 11b). In much of western Amundsen Gulf, increasing trends in winter first-year sea ice occur (Figure 11a) where on average 1 to 3/10ths new and young sea ice occur in winter (Figure 2e). Positive trends in first-year sea ice concentration in western Amundsen Gulf are statistically significant at the 95% level (Figure 11b). The west coast of Victoria Island also contains a positive trend in first-year sea ice concentration (Figure 11a) significant at the 95% level (Figure 11b). First-year sea ice concentration has been increasing in the flaw lead system between 1980 and 2004.

[34] It follows that trends in thick first-year sea ice concentration have also occurred (Figure 11c). The largest changes (2.5% per year) occurred in the flaw lead area west of Banks Island, north of the Tuktoyaktuk Peninsula and within the southern Beaufort Sea across the mouth of Amundsen Gulf (Figure 11c). There is a trend toward more first-year ice in the flaw lead system, with much of it growing quite thick over the winter. There is also a trend toward more thick first-year sea ice in the western part of the study area, making up for losses in perennial sea ice in the same area.

[35] In the flaw lead system, thicker sea ice showed increasing trends and young sea ice concentrations have decreased. The area immediately west of Banks Island which on average contains 5–6/10ths new and young sea ice (Figure 2e) experienced young ice loss at the rate of >1.5%/year (Figure 11e) which is significant at the 99% level (Figure 11f). Smaller negative trends in young sea ice (0.5–1.5%/year) occurred parallel to the Tuktoyaktuk Peninsula, within western Amundsen Gulf, and extending west from the mouth of Amundsen Gulf (Figure 11e). The trends in most of these areas are significant at the 95% level (Figure 11f).

3.2.2. Summer trend in sea ice concentrations: 1980–2004

[36] A positive trend in summer old sea ice concentration (1 to 2% per year) is observed west of Banks Island reaching into the mouth of Amundsen Gulf (Figure 12a). This area is composed of 2 to 5/10ths total summer sea ice on average, where the total summer sea ice comprises equal parts old ice and thick first-year sea ice (Figures 7a7c). The trends are significant at the 95% level (Figure 12b). A statistically significant (95%) positive old ice trend in the flaw lead area west of Banks Island also occurred.

Figure 12.

Summer sea ice (left) concentration trend (% per year) and (right) statistical significance for (a, b) old sea ice and (c, d) first-year sea ice. Black areas in the right column indicate statistical significance at the 99% level, dark gray areas indicate statistical significance at the 95% level, and light gray areas indicate statistical significance at the 90% level.

[37] Significant trends in first-year summer sea ice were exclusively negative (Figure 12c), occurring within an area occupied by 6 to 8/10ths total summer ice concentration (Figure 7). Small areas of significant first-year sea ice reduction also occurred in eastern Amundsen Gulf.

3.3. Discussion: Sea Ice Concentration Trends

3.3.1. Winter Trends in Sea Ice Concentration: 1980–2004

[38] Recently, the Beaufort Sea perennial pack has undergone a change in the types of ice which make it up; perennial ice has replaced by seasonal ice in winter (Figure 4: 2002, 2003, 2004). The extent of perennial sea ice cover in the Arctic has been decreasing annually while the sea ice that remains is getting younger and thinner [Maslanik et al., 2007]. This transition from thick old sea ice to thinner old ice or seasonal ice has been attributed to (1) atmospheric-forced sea ice divergence and/or (2) thermodynamic loss.

[39] Atmospheric-forced divergence reduces sea ice concentration in the periphery of the old ice pack enhancing melt through the ice-albedo feedback mechanism. Yu et al. [2004] showed increased fractional open-water and first-year ice areas in the central Arctic Basin between the 1960s and the middle 1990s, balanced by an 11% reduction in level-multiyear and ridged ice thicker than 2.0 m. They attributed these changes in ice mass balance to increased export through Fram Strait in the late 1980s and early 1990s driven by variability in the North Atlantic Oscillation (NAO) and Arctic Oscillation (AO), and to changes in thermal forcing [Yu et al., 2004]. As the AO index becomes increasingly positive the resulting surface level pressure pattern leads to divergent ice motion within the southern Beaufort Sea as well as positive surface air temperature anomalies, resulting in thinner than normal spring sea ice conditions [Rigor et al., 2002]. Rothrock et al. [1999] note that the mean draft of perennial sea ice in the Arctic Ocean decreased by 1.3 m (∼40%) between the 1990s and data acquired between 1958 and 1976, with mean draft decreases of about 1 m in the Beaufort Sea. This resulted from thinning of perennial sea ice in the Arctic Ocean as a whole and not advection of thinner stages of development from one region to another [Rothrock et al., 1999].

[40] The transition from thinner (new, young and thin first-year) to thicker first-year ice was observed within the seasonal ice zone, especially in the flaw lead system. In the absence of a decreasing trend in winter air temperatures [Howell et al., 2008] the increasing trend of thick first-year ice west of Banks Island may be attributed to the reduction in old ice concentration allowing seasonal sea ice in these areas to grow thicker. first-sea ice has been replacing young ice in winter in the flaw lead system (Figures 11a, 11c, and 11e).

3.3.2. Summer Trends in Sea Ice Concentrations: 1980–2004

[41] The increasing trend in old sea ice (∼1 to 2.5% per year) west of Banks Island (Figure 12a) is likely due to increased cyclonic motion and therefore divergence. Very high old ice concentration (>9/10ths) areas in the northwest may be losing ice to the 7 and 8/10ths old ice areas, increasing their concentration. These trends in summer old ice concentration can be attributed to differential reduction in perennial sea ice between the eastern to western Arctic (divided by 0° and 180° longitude) [Nghiem et al., 2006]. Nghiem et al. [2006] showed the east Arctic Ocean lost 48% of its perennial sea ice between consecutive Decembers (2004 and 2005). while the west Arctic Ocean gained 0.95 × 106 km2 perennial sea ice over the same period, replacing seasonal ice. Further, recent research has shown that summer cyclonic motion in the pack creating divergence is occurring with greater frequency [Lukovich and Barber, 2006], resulting in more old ice being dynamically driven toward Amundsen Gulf in the summer, causing an increase in the old ice concentration near the mouth of Amundsen Gulf. Decreasing summer first-year ice trends in the region (Figure 12c) may be attributed to enhanced melt in summer due to divergence in the pack allowing the ice-albedo feedback to take affect.

[42] The yearly minimum extent of perennial sea ice in the Arctic has been declining since the start of the passive microwave record at a rate of between 3 and 7% per decade, depending on the years studied [Parkinson et al., 1999; Comiso, 2002; Serreze et al., 2003; Stroeve et al., 2005]. Our work demonstrates that statistically significant (>95%) trends in summer old ice in the southern Beaufort Sea have increased despite hemispheric losses. first-year sea ice occurred between 1980 and 2004, and further that summer losses of first-year ice have also occurred.

4. Breakup and Freezeup Within the Southern Beaufort Sea and Amundsen Gulf

4.1. Average Breakup

[43] As a result of the high old ice concentration in some areas, breakup does not always occur in all grid cells (Figures 13b and 13d). In this discussion, we consider only those grid cells which break up each year (100% occurrence) (Figures 13b and 13d). On average, the flaw lead area north of the Tuktoyaktuk Peninsula and west of Banks Island, joined across the mouth of Amundsen Gulf begin to break up first around YW 21 (Figure 13a, shown in dark blue). Western Amundsen Gulf begins to break up between YW 22 and YW 24 (Figure 13a). Eastern Amundsen Gulf begins breaking up after YW 25 (Figure 13a, shown in light blue). Finally, the landfast sea ice in the region begins to break up around YW 30.

Figure 13.

Mean breakup (a) start year-week, (b) start occurrence (%), (c) start mean standard deviation (weeks), (d) end year-week, (e) end occurrence (%), and (f) end mean standard deviation (weeks) for each grid cell between 1980 and 2004. Occurrence is defined as the amount of times each grid cell began to break up or finished breaking up in 25 years, expressed as a percent.

[44] Breakup within the flaw lead system and across the mouth of Amundsen Gulf takes 1–2 weeks (Figures 13a and 13d). Eastern Amundsen Gulf finishes breaking up around YW 28 on average and finally the landfast ice in the region finishes breaking up after YW 30. This substantial reduction in concentration in the region over 1–2 weeks is due to the relative thinness of sea ice in the flaw lead area in winter allowing thermodynamic weakening to occur more easily than in the surrounding thick sea ice. This thin, weak sea ice is then advected away from the area in as little as a few days. Our results are similar to those of Hammill [1987], who concluded that the ice retreat pattern in the region was related to the locations of atmospheric pressure patterns in the area.

[45] While a broad expanse of open water (the “Cape Bathurst Polynya”) occurred between Thesiger Bay and Franklin Bay occurred in 15 of 25 years, several occurrences of what Hammill [1987] termed an “eastern polynya” were found in our time series. Like the formation of the Cape Bathurst Polynya, the location of the eastern polynya was always determined by where the mobile pack edge met landfast ice. In 1981, 1994, 1996, and 1999, the mobile pack edge reached deep into the gulf, so that a large open water area formed near Victoria Island. In other years, the primary open water area was a shore lead extending eastward from Cape Parry. These instances were noted in 1988, 1989, 1993, and 1998. With the exception of 1998, when above-average surface air temperatures melted a thin ice cover, all other occurrences of a southern flaw lead in the gulf tended to be later in the season, forming in late May (YW20) and early to mid-June (YWs 22–24). The remaining two years in which the first open water area did not occur in Amundsen Gulf were 1992 and 2000, when the Mackenzie River shelf was the primary source of breakup and open water. In both cases, these were late melt-onset years, occurring in mid-June (YW 25).

4.2. Average Freezeup

[46] In the seasonal ice zone, the mean freezeup start year-week varies from YW 37 offshore of the Tuktoyaktuk Peninsula and in Franklin Bay to YW 41 nearby Victoria Island (Figure 14a). Our analysis shows that the first areas to freeze are those with some sea ice (about half old ice and half thick first-year) in summer (Figure 7). Areas covered by less sea ice in summer develop an ice cover later. Eastern Amundsen Gulf contains only trace concentrations of sea ice in summer, and are the last to begin to freezeup. Sea ice present at the end of summer can cause the ocean below to be cooler because it increases the albedo of the surface, decreasing the amount of energy the ocean can absorb. If sea ice is already present even in very low concentrations, it acts as a congelation agent for sea ice growth. Both factors lessen the oceanic heat available to reduce sea ice growth, causing the near-surface ocean column to approach its freezing point more quickly than if sea ice was not present [Wadhams, 2000].

Figure 14.

Mean freezeup (a) start year-week, (b) start occurrence (%), (c) start mean standard deviation (weeks), (d) end year-week, (e) end occurrence (%), and (f) end mean standard deviation (weeks) for each grid cell between 1980 and 2004. Occurrence is defined as the amount of times each grid cell began to freezeup or finished freezing up in 25 years, expressed as a percent.

[47] The region freezes up very quickly (Figures 14a and 14d). The average duration of freezeup within the seasonal ice zone is 1–2 weeks. The atmosphere, ocean and sea ice regime seem to meet a threshold time in the year, between YW 38 and 41 (∼23 September to ∼14 October), where the atmospheric temperature is sufficiently low, the near-surface ocean column reaches its freezing point and there is some small concentration of sea ice in the region to kick-start ice growth.

5. Breakup and Freezeup Timing and Duration in Amundsen Gulf

5.1. Breakup Variability

[48] Duration of breakup in the seasonal ice zone showed a much greater variability than that of freezeup (Figure 15). Hammill [1987] indicates that between 1964 and 1974, sea ice breakup was almost evenly divided between early, medium, and late years; if open water occurred in July, it was an early breakup year; if in August, a medium, and in September, a late breakup year. The sea ice concentration curves for 1980–2004 (Figure 15) show that there has been a substantial shift toward earlier breakup since 1964–1974. The average breakup start week between 1980 and 2004 was YW 22, seven weeks earlier than the earliest breakup start found by Hammill [1987]. The range of breakup start between 1980 and 2004 was broad, with the earliest occurring at week 18 in 1991, 1993, 1994, and 1998. The latest breakup start was recorded in 2000, at week 29, which corresponds to the week of early breakup for 1964–1974 [Hammill, 1987].

Figure 15.

Sea ice concentration in Amundsen Gulf averaged over all grid cells for each year studied by year-week. Total (black), old (blue), first-year (green), and new and young (red).

[49] Each breakup over the 25-year time series is summarized in Table 1. When the whole of Amundsen Gulf is examined, the average breakup lasted more than 8 weeks, which is almost 2 weeks longer than between 1964 and 1974 [Hammill, 1987]. The longest breakup duration (22 weeks) occurred in 1991. 1991 was particularly anomalous as the open water duration was only one week as a result of old ice influx into Amundsen Gulf in summer. The year in which breakup occurred most rapidly was 2000, taking only 3 weeks. However, this was also the latest year for breakup start, allowing thermodynamic melt to thin and weaken the sea ice in Amundsen Gulf. It was also a low sea ice extent in the Beaufort Gyre [Serreze et al., 2003], which created a larger than average seasonal ice zone (Figure 9).

Table 1. Year-Week in Each Year Where the Average Total Sea Ice Concentration in Amundsen Gulf Reaches Spring Breakup, the Open Water Season, and Fall Freezeupa
Year<80%Start OWLength OWEnd OW>80%Breakup SlopeFreezeup Slope
  • a

    OW denotes open water. The slope of the total concentration curves (Figure 15) is shown for breakup and freezeup.


[50] The slope of the total sea ice concentration curve through time in spring is indicative of breakup processes (Table 1). Smaller negative slopes indicate gradual, melt and/or restriction of clearing the Amundsen Gulf due to exchanges of old ice with the Beaufort Gyre. Floes from the gyre may enter the gulf, effectively blocking the exit of sea ice through the mouth of Amundsen Gulf. Larger negative slopes characterize years where ice is advected out of Amundsen Gulf quickly. The year with the largest negative slope was 2000, when ice was removed from the gulf within 3 weeks. Other years with large negative slopes include the El Niño years 1982, 1987–1988, 1993, and 1996, as well as 1998, when anomalous warming characterized the entire Canadian Archipelago. The smallest negative slope was in 1991, when open water duration was brief, and in 1985 and 1994 (Table 1).

[51] Breakup timing is highly variable between years, controlled in part by the position of the Beaufort Sea ice pack in winter and spring. In order for breakup to occur in Amundsen Gulf, the sea ice has to be thermodynamically weak enough to be moved, and it must have an area to move to. Breakup is sometimes retarded by increased concentrations of old ice and/or first-year sea ice occurring at year-weeks within the breakup period (Figure 15).

5.2. Open Water

[52] Open water was defined as the amount of time where the sea ice concentration fell below 20%. The average week in which open water occurred in Amundsen Gulf was around YW 29.6 ± 3.3 weeks between 1980 and 2004. This is 2 to 5 weeks earlier than medium ice retreat reported by Hammill [1987]. The earliest occurrences of open water occurred in 1988 (YW 25), 1993 (YW 25), and 1998 (YW 23) (mid-June). In two of the three cases (1988, 1998), the preceding refreeze occurred relatively late (YW45), and warming associated with El Niño (in 1988) or a hemispheric anomaly (in 1998) may have contributed to the rapid ice retreat. The latest appearances of open water were observed in 1991 and 2002, at week 39 (∼01 October) and 36 (∼09 September), respectively (Figure 15). The delay in open water formation during both 1991 and 2002 was due to anomalously large influxes of old ice to Amundsen Gulf beginning on around YW28 and ending around YW38 delaying the open water season.

[53] The duration of open water averaged 10.4 ± 4.4 weeks during 1980–2004. Using the Hammill [1987] medium ice breakup and medium freezeup dates as indicators of the open water duration during 1964–1974, the average length of open water for the Hammill [1987] data set was about 8 weeks. Those years with the earliest breakup and latest freezeup, corresponding to the longest open water seasons, were 1988 (16 weeks), 1998 (20 weeks), and 1999 (15 weeks). Those with the shortest ice-free duration were 1991 (1 week), 1996 (3 weeks), and 2002 (3 weeks) when old ice influx to Amundsen Gulf occurred (Figure 15).

5.3. Freezeup Variability

[54] Freezeup in Amundsen Gulf between 1980 and 2004 was fairly consistent with the results of Hammill [1987], occurring over a period of about 3 weeks (Table 1) while medium ice advance took 2 weeks to complete between 1964 and 1974 [Hammill, 1987]. This rapid freezeup started on average, around YW40 and was completed by YW43 for the years 1980 – 2004. This agrees with the timing of medium ice growth conditions in Hammill [1987] (∼YW 40 to 42). Compared to breakup there was little variability in the start of freezeup in either the 1964–1974 [Hammill, 1987] or our 25-year data set. The earliest instance of freezeup onset occurred in 1996 on week 33, which also recorded the longest freezeup duration (8 weeks). In all 25 years, freezeup was complete between weeks 41–45. During 1998, 2002, and 2004, complete freezeup occurred within 1 week. These were anomalous years in terms of ice conditions in the archipelago (in 1998 due to anomalously warm surface air temperatures) or the southern Beaufort Sea (in 2002, 2004 [Stroeve et al., 2005]).

[55] Focusing on the average conditions, in which freezeup occurred at YW 40 and was completed by YW 43, a survey of the available radiation was made. Using a midpoint within the Amundsen Gulf (70°N, 118°W) as a geographical reference, the geometry and envelope of daily radiation incident at the top of the atmosphere was calculated. During YW 38, the maximum amount of shortwave radiation at the top of the atmosphere decreases to half its summertime maximum, from 910 W/m2 to 455 W/m2. It is also at this time that the number of daylight hours where the sun is above the horizon, corresponding to a solar zenith angle > 90°, decreases below 12 h. While further radiative transfer modeling is necessary to determine the actual amount of radiation reaching the surface and its role in ice refreeze onset, these basic characteristics suggest that a radiative threshold is reached by YW39 at this latitude and longitude.

6. Conclusions

[56] In this study, we examined the spatial and temporal variability of various sea ice types in the Southern Beaufort Sea and Amundsen Gulf region over 1980–2004, using CIS digital ice chart data. In winter, high concentrations of mobile sea ice dominated the southern Beaufort Sea and Amundsen Gulf on average. Winter sea ice in Amundsen Gulf is almost entirely composed of seasonal ice types, which interact with the perennial sea ice in the southern Beaufort Sea to create and maintain a flaw lead system, composed of predominantly new, young, and thin first-year sea ice, surrounded by mostly thick first-year sea ice. The results showed significant shifts in sea ice type and therefore mass balance within the flaw lead system between perennial and annual sea ice in winter over the 25-year time series. Linear regression analysis indicated an increasing trend in thick first-year sea ice in the flaw lead system. In the flaw lead system in winter, thick first-year sea ice also showed a significant increasing trend, balanced by decreasing young sea ice concentrations. These shifts in sea ice type have serious implications for the presence of the flaw lead network in the region. Reduction in the persistence or extent of the flaw lead system could lead to substantial changes in the way Amundsen Gulf breaks up in spring, reducing the biological productivity of the region and affecting the lives of the area's inhabitants.

[57] In summer the sea ice present in the region was mainly old ice, with some thick first-year sea ice; much of the southern Beaufort Sea was ice covered each year. Heavy sea ice years occurred in the 1990s and 2000s when summer sea ice extended from the southern Beaufort Sea perennial pack into Amundsen Gulf. Sea ice was present in eastern Amundsen Gulf in summer through the 1980s, and then again in 2002 and 2004. Following a massive reduction in summer sea ice extent in 1998, the region rebounded, regaining much of the summer sea ice extent which occurred prior to 1998. We found a eastward shift in the high-concentration perennial pack ice within the southern Beaufort Sea toward the west coast of Banks Island. We found a trend toward increasing old ice concentration west of Banks Island, and a trend toward decreasing first-year ice concentrations in the 5–8/10ths total sea ice contours. Old sea ice has reduced the amount of open water in the southern Beaufort Sea and Amundsen Gulf in summer over the 25 years studied, and the summer transition zone between open water and high-concentration sea ice in the southern Beaufort Sea gained first-year sea ice. This has important implications particularly for navigation of the Northwest Passage, as old sea ice may continue to block the western waterway to ship navigation.

[58] Rapid sea ice breakup in the study area occurs dynamically; presumably after the sea ice reaches a thermodynamic threshold which weakens the ice, making it susceptible to motion. Breakup begins within nearshore areas of reduced ice thickness along the coast of the Tuktoyaktuk Peninsula and the west coast of Banks Island respectively, followed by sections of Amundsen Gulf from west to east. The area covered by thin first-year, young and new ice is in decline, which may reduce the ability of sea ice in the region to break up normally. Thicker sea ice is more difficult to remove dynamically and thermodynamically. Freezeup in the seasonal ice zone is complete within 1 to 2 weeks occurring first within the seasonal to perennial ice transition zone offshore, then in western and finally in eastern Amundsen Gulf Additional sea ice in the region in summer may enable freezeup to happen sooner each year by keeping the ocean cooler and seeding new ice growth with ice already in the region.

[59] The timing and duration of breakup within Amundsen Gulf is widely variable in comparison with freezeup. The duration of breakup, open water and freezeup depend to some degree on the presence of old sea ice within the region. Old ice influx can delay the end of breakup (and therefore the beginning of the open water season), and can also influence the timing of freezeup in Amundsen Gulf. Breakup occurred much earlier on average during 1980–2004 than during 1964–1974.

[60] Generally, melt began earlier, leading to a slightly longer open water season after which a rapid refreeze covered the basin at or about the same time as in the Hammill [1987] data set. Refreeze was generally consistent between the two data sets, occurring the week following a 50% reduction in available shortwave energy at the top of the atmosphere and a reduction in daylight hours to 12 or less.


[61] Thanks to funding support from the ArcticNet Network of Centres of Excellence program, the CASES NSERC research network, and the Canada Research Chairs program with grants to D.G.B. R.G. thanks NSTP for their support. Thanks go to P. Minnett for his insight and comments. The Canadian Ice Service provided the digital chart database. Erica Key was supported through a grant from the NSF Office of Polar Programs (OPP-0327187).