Corresponding author: J. K. Hutchings, International Arctic Research Center, University of Alaska Fairbanks, 930 Koyukuk Dr., Fairbanks, AK 99777, USA. (email@example.com)
 A new record minimum in summer sea ice extent was set in 2007 and an unusual polynya formed in the Beaufort Sea ice cover during the summer of 2006. Using a combination of visual observations from cruises, ice drift, and satellite passive microwave sea ice concentration, we show that ice dynamics during preceding years included events that preconditioned the Beaufort ice pack for the unusual patterns of opening observed in both summers. Intrusions of first year ice from the Chukchi Sea to the Northern Beaufort, and increased pole-ward ice transport from the western Arctic during summer has led to reduced replenishment of multiyear ice, older than five years, in the western Beaufort, resulting in a younger, thinner ice pack in most of the Beaufort. We find ice younger than five years melts out completely by the end of summer, south of 76N. The 2006 unusual polynya was bounded to the south by an ice tongue composed of sea ice older than 5 years, and formed when first year and second year ice melted between 76N and the older ice to the south. In this paper we demonstrate that a recent shift in ice circulation patterns in the western Arctic preconditions the Beaufort ice pack for increased seasonal ice zone extent.
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.
 Arctic sea ice experienced a dramatic decrease in extent during summer 2007 [Stroeve et al., 2008], which far exceeded recent and projected trends for sea ice loss [Stroeve et al., 2007, 2008]. Low summer minimum ice extents, over the last decade, have in part been due to an increased area of seasonal ice over the peripheral seas of the Arctic Ocean. In 2007, ice melt across the Laptev, East Siberian and Chukchi Seas was more extensive than previously witnessed in the satellite record (1979 onwards), Figure 1. Analysis of Russian ice charts [Mahoney et al., 2008], dating from the mid-1930s onwards, indicates that extensive ice retreat in the early years of the record may be comparable to recent ice loss in the Siberian Arctic. It must be noted that the variability and uncertainty in this historical data is unknown. However, the persistent circulation patterns from 1928 to 1935 indicate low ice extents may have been confined to the Siberian arctic [Overland and Wang, 2005b]. The 2007 melt extended into the Beaufort Sea, and coincided with the first recorded melt out of ice in the Eastern Beaufort and throughout the Canadian Archipelago, opening the McClure Strait and all other routes in the North West Passage (NWP) (J. Falkingham et al., 2007 Arctic Ice Retreat Concerns National Ice Services, press release, 2007, http://nsidc.org/noaa/iicwg/docs/IICWG_2007/IICWG_NEWS_RELEASE.pdf). Many of the passages in the NWP had been open before, but this was the first time, since monitoring of the NWP began with ice charts, that all the passages were open at the same time.
 Before the mid-1990s inter-annual variability in minimum ice extent, from 1979 onwards, correlated with atmospheric indices such as the Arctic Oscillation (AO) and North Atlantic Oscillation (NAO) [Rigor et al., 2002; Deser and Teng, 2008]. This suggests the minimum ice extent each summer was related to changes in mean atmospheric circulation, and therefore the frequency of cyclones entering the Arctic. However, in the last decade the summer ice extent has continued to retreat, breaking correlation with the AO. This suggests that recent decline in ice extent may not be directly attributable to the AO overlaying atmospheric circulation [Deser and Teng, 2008], and as Rigor and Wallace  note, the recent decline in sea ice cover may be related to a decrease in the age and thickness of ice drifting toward the coast driven by a different atmospheric anomaly pattern. Overland and Wang [2005b]also postulate that a shift in the pattern of atmospheric circulation, in 2000, leading to increased pole-ward winds across the East Siberian, Chukchi and Beaufort Seas has led to warmer Arctic surface temperatures and accelerated transport of ice across the Arctic, resulting in a decrease in summer ice extent in the western Arctic. This hypothesis is further supported byOgi and Wallace  and Ogi et al. , who show meridional wind anomalies correlate with September ice extent.
Nghiem et al.  found that the low summer ice extents across the Arctic in 2006 and 2007 were preceded by large losses of perennial ice area through transpolar drift across the Fram Strait. The mechanism causing this was examined by Ogi et al. , who demonstrate that summer winds can cause strong poleward ice drift. The Dipole Anomaly (DA) Index describes these circulation events [Wu et al., 2006] and positive DA is related to increased west to east ice drift and reduced summer ice extent [Wang et al., 2009]. The Beaufort Gyre plays a role in the loss of sea ice, as it loads perennial ice into the Transpolar Drift.
Shimada et al.  found that the spatial pattern of ice loss in the Chukchi and Beaufort follows changes in the pathways, and an increase in temperature, of Pacific Summer Water flowing into the region, suggesting increased oceanic heat flux to the ice may be responsible for its demise. Wang et al.  suggest that this increased heat flux through the Bering Strait is related to increasing atmospheric meridional circulation. Steele et al.  demonstrate that the Arctic upper ocean has warmed since the 1970s. This may be through an increase in solar radiation input to the ocean, and Perovich et al.  hypothesize that a large increase in solar radiation input to the upper Beaufort Sea caused a dramatic melt of multiyear ice observed in the Beaufort Sea in 2007. Jackson et al.  find a near surface temperature maximum forms with solar heating of the Beaufort Sea, and this heat is stored under the summer halocline, becoming available to retard winter ice growth and may persist through the winter.
L'Heureux et al. demonstrate that anomalous atmospheric circulation in 2007 (with high value for the Pacific North American index and persistent anti-cyclone over the Beaufort and East Siberian Seas) led to increased pole-ward ice and warm air transport across the Chukchi and west Beaufort Seas, and increased downwelling solar radiation [Kay et al., 2008]. Schweiger et al. , however found, in a model study that the solar radiation gain was balanced by a longwave radiation loss over the Eurasian Arctic, suggesting an increase in solar radiation input does not directly explain the dramatic retreat of ice extent in the Eurasian Arctic. However, they did find that in the Beaufort Sea the cloud anomaly may have contributed to the enhanced melt observed in summer 2007.
Shimada et al.  hypothesized that decreases in ice strength (thickness) due to export of ice from the Arctic Basin in the winter of 1997/1998 [Rigor and Wallace, 2004] led to a reduction in internal ice stress, enabling increased transport of ice along the Canadian and Alaskan coastline. This allows more efficient coupling between the atmosphere and ocean, strengthening upper ocean circulation, and increasing transport of warm Pacific Summer Water into the Chukchi and Beaufort Seas, retarding ice growth in winter and perpetuating the weak ice conditions. Ice drift and deformation rate has increased from 1979 to 2007 [Rampal et al., 2009], and ice drift increased in the Beaufort Sea between 1992 and 2009 [Spreen et al., 2011]. Hakkinen et al.  found that wind speed, and therefore wind stress to the ice and ocean, has increased gradually over the last 50 years. This increase does not extend into the Southern Beaufort, though the increase in seasonal ice area here in the last decade is likely to have changed the strength of the ice pack and hence enhanced wind stress transfer to the ocean.
 It is important to note that the last two of these mechanisms involve positive feedback for ice loss whereby a thinner, less compact ice pack is required to trigger an enhanced feedback. The trigger may be driven by the atmosphere either through changes to incoming shortwave or longwave radiation (though evidence is inconclusive on this mechanism), or through changing ice drift. Anomalous wind driven ice drift events, or changing ice and atmospheric circulation patterns, have acted [Rigor et al., 2002; Rigor and Wallace, 2004; Nghiem et al., 2006, 2007; Ogi et al., 2008] to create a younger ice pack in much of the Arctic [Comiso, 2002; Rigor et al., 2002; Rigor and Wallace, 2004; Maslanik et al., 2007; Kwok, 2007; Nghiem et al., 2007], preconditioning the ice pack to be more responsive to changes in surface energy balance and changes in wind stress. The role of changes in ice drift and deformation in the last decade, due to a thinning ice pack and increase in cyclonic weather systems in the Arctic, has not been examined in detail. In this paper we address whether ice drift has acted to accelerate ice loss in recent years in the Beaufort Sea. We will address whether deformation acts as a positive or negative feedback to this ice loss in a future paper. We further demonstrate, in this paper, that ice drift in the Beaufort Sea, since 2000, has changed with the atmospheric circulation changes explained by Overland et al. . This has resulted in further thinning of the Beaufort ice pack. Hence we expect the ice pack throughout the Beaufort Sea has been preconditioned to be more vulnerable than 20 years ago to interannual variability in thermodynamic forcing.
 Model projections suggest that the Arctic could pass a “tipping point” where all perennial ice would be lost [Lindsay and Zhang, 2005]. They suggest that ice thinning since 1998 is due to preconditioning, where perennial ice was flushed from the Arctic, and positive feedback occurred in response to increased surface warming and the preconditioned thinner ice pack. To understand if the ice loss mechanisms described above are a step toward a seasonally ice covered Arctic, we need an understanding of the variability of ice circulation within the Beaufort Gyre, and how this impacts the perennial ice pack. The condition of sea ice in the Beaufort Sea is controlled by the size and strength of the Beaufort Gyre, so a detailed investigation of the sea ice in this region provides much information about the ice circulation in the Beaufort Gyre.
 In this paper we examine the state of sea ice in the Beaufort Sea and how this has responded to changes in Arctic-wide sea ice circulation patterns. We investigate the role ice dynamics plays in the increase of Beaufort Sea seasonal ice pack over the last decade, especially in terms of understanding the spatial variability. We take a close look at the summers of 2006 and 2007, which had remarkably different ice covers. The data and models we use in our analysis are outlined insection 2.
 In section 3 we describe the Beaufort ice pack state at the end of summer in 2006 and 2007, in the context of the satellite record of ice extent. We use ship observations as they provide information about ice age, stage of melt and thickness that is not readily available from satellite data. Section 4 examines ice circulation patterns in the Arctic and Beaufort Sea, and particular anomalies in this circulation are highlighted as they help explain the unusual conditions observed in summers 2006 and 2007. We investigate variability in ice circulation over the last 30 years, and how this preconditions the ice pack for the melt observed in 2006 and 2007 in section 5. We summarize and discuss our findings in section 6.
2. Data Description
 The Canadian Coast Guard Ship Louis S. St. Laurent cruised the Beaufort Sea August 4th to September 14th 2006 and July 28th to August 28th 2007. During these two cruises an hourly ice watch was kept, where an observer recorded ice conditions in a 3 km radius area about the ship (visibility permitting) during a 10 minute interval on the bridge. The ice watch recorded sea ice concentration, ice type, floe size and surface topography following ASPECT conventions [Worby and Allison, 1999]. Stage of melt of each ice type was noted, following Russian convention (V. Smolyanitsky, Arctic and Antarctic Research Institute, personal communication, 2002). We used aerial observations to validate that bridge observations of ice type, concentration and melt-pond cover were regionally representative. The summer Arctic ice pack has distinct regions defined by often sharp boundaries between ice types. Helicopter flights occurred at least once in each ice region, and provide qualitative assessments of the ice conditions. A 2 m stick painted with 10 cm bands, mounted over the ships side from a lower deck, gave a visual aid to estimate the thickness of ice over-turned as the ship broke ice. The most prevalent thickness was noted for each ice type during each hourly watch. We did not always have a thickness estimate for each ice category observed, as the ship would preferentially travel through the open water or thinnest ice. This bias in the thickness observations means the data should not be used to calculate regional ice volume, but the data does provide an estimate (with 0.2 m accuracy) of the mode of the thickness distribution of the thinner ice types observed during a ten-minute observation period. This method is appropriate only for observing ice thickness below 2.5 m, as thicker ice is not overturned fully by the ship.
 It should be noted that our definition of multiyear (MY) ice differs from the standard World Meteorological Organization (WMO) practice of identifying ice type according to its thickness. In 2007 large areas of the Southern Beaufort were covered with MY ice less than 1.5 m thick, and would be classified as FY ice if the standard practice was followed. We determined whether ice was first year (FY) or MY from consideration of the ice surface topography. Any ice that showed hummocking due to survival of a previous melt season, which would be visible in the surface topography and spatial distribution of melt ponds, was designated as MY ice.
 We use a NASA team algorithm passive microwave ice concentration product [Gloersen and Cavalieri, 1986], provided by the National Snow and Ice Data Center [Cavalier et al., 1996] to map ice area in the Beaufort Sea. The minimum Beaufort Sea ice extent at the end of each summer since 1979 was calculated on the day it occurred, assuming the ice edge was at 15% concentration.
 The age of sea ice is estimated using a simple drift model (DM) [Rigor and Wallace, 2004], which uses the gridded fields of sea ice motion analyzed by the International Arctic Buoy Programme [Rigor et al., 2002], and tracks parcels of sea ice from September of one year to September of the following year. To determine which areas of ice survive the summer melt and age 1 year, we use the 90% sea ice concentration line in September of each year from the NASA Bootstrap 2 analysis from 1979 to 2007 [Comiso and Parkinson, 2008] and prior to 1979 [Walsh, 1978].
3. Summer 2006 and 2007 Ice Conditions
 In the last decade the Beaufort saw four years with lowest on-record Beaufort Sea ice extents (Figure 1), and reductions in summer ice minimum have been linked to increases in the extent of the Beaufort Sea seasonal ice zone [Drobot and Maslanik, 2003]. In 1998, the Beaufort Sea ice pack experienced an all-time low summer ice minimum (Figure 1), and not until 2008 did the Beaufort ice pack experience a subsequent record minima. Since 1998 the end of summer ice has been of average or below average extent. A casual look at the time series of Beaufort Sea ice minima suggests a change in the ice pack occurred around 1998. The western Beaufort experienced more ice retreat in the last decade than the eastern Beaufort. The latitude of the ice edge at summer minimum describes variability in summer ice melt [Eisenman, 2010]. Since 1998 there have been 8 years when the ice extent in the western Beaufort receded past 76N (Figure 2). Prior to 1998, in the satellite record, the minimum ice extent in the western Beaufort did not recede past 74.5N. The amplitude of the variability in ice edge position increased, after 1998, as the western Beaufort ice edge has retreated northward. The ice edge at summer minimum in the eastern Beaufort has not experienced dramatic changes over the last 30 years, although the 2008 retreat is a notable outlier. In the satellite passive microwave record of ice extent, the last decade has been characterized by an increase in the extent of the seasonal ice zone in the south-western Beaufort, and less change in the eastern Beaufort ice cover. Combining the ice watch information with satellite imagery of the Beaufort Sea provides a detailed description of ice conditions at the end of summer in 2006 and 2007.Figure 3 shows SSM/I sea ice concentration [Cavalier et al., 1996] calculated by the NASA Team Algorithm [Gloersen and Cavalieri, 1986], and the ship based observations are summarized in Figure 4, where most prevalent ice type and corresponding thickness of that type are plotted.
 We observed in 2007 that the MY ice, which we estimated to be 2 or 3 years old, in the western Beaufort, south of 76N, was thin (often 1/2 m) and rotten (as described by Barber et al. ). We were astonished to observe such an expanse of thin MY in the southern Beaufort (Figure 4), as were Barber et al. , and also surprised to find no ridges greater than 4m thick in our drill hole surveys taken at ice stations during the 2006 and 2007 cruises.
 In August 2006 an approximately 90,000 km2 area of open water developed in the Beaufort sea ice pack, which commanded much media attention. This feature was referred to as a polynya in 2006 (National Snow and Ice Data Center, Fall2006 sea ice update, press release, 2006, http://nsidc.org/news/press/2006_seaiceminimum/20060816_arcticseaicenews.html), but we demonstrate in section 4that the presence of this open water was not due to persistent wind opening or external heat input, as is usually the case for a polynya. As polynya is the only term available referring to an area of open water in the ice pack that persists over weeks to months, we continue to refer to the 2006 feature as a polynya, but as we argue below, we believe the salient feature to focus on is the ice tongue to the south. A tongue of MY ice was transported from the region north of the Canadian Archipelago into the south-western Beaufort during winter 2005–2006. This ice tongue consisted of ice greater than 2.5 m thick, and judging from surface conditions was older than any other ice observed during the two cruises.
 The northern edge of this polynya approximately followed the latitude of the 2002 summer minimum ice extent, which, prior to 2007, was the record minimum for the Arctic and had large ice retreat in the Beaufort Sea. The ice bounding the southern edge of the polynya was multiyear (MY) ice that had been transported into the Southern Beaufort from the Eastern Beaufort Canadian Archipelago region the previous winter. This ice was thicker, more ridged and more consolidated than the second year (SY) and first year (FY) ice that melted out to the north, and hindered navigation in the region just north of Barrow. We observed that the ice to the north of the polynya was predominantly FY ice that had been transported from the Chukchi in the previous winter. The ice on the southern boundary of the polynya and within the polynya was FY and SY ice, with isolated patches of MY ice in the west. The thinner, less ridged ice within and north of the polynya is more susceptible to melt than the MY ice to the south, suggesting that it was the presence of the MY ice tongue in the Southern Beaufort that allowed summer melt to create a polynya in the Beaufort pack ice.
 The year 2007 saw the lowest Arctic minimum sea ice extent on record. This year was characterized by an unusually early melt out of ice in the Chukchi and Beaufort, and an unprecedented loss of ice throughout most of the Laptev and East Siberian Seas. In the Beaufort, in-situ observations reveal vast expanses of thin ice in an advanced stage of melt. In the eastern Beaufort, along Banks Island, early ice melt opened the area to ship navigation to a greater extent than ever recorded before. We found 60% MY ice in the easterly legs of the 2007 cruise.
 Both 2006 and 2007 saw an opening of the Eastern Beaufort MY pack near Banks Island. Such a large ice retreat had previously been observed in this region only in 1988 and 1998. Conditions were so open (60% cover) in 2007 that we were able to travel into the data sparse region off the north-western shore of Banks Island.
 In both years we had no trouble traveling north of 76N to 79N in the western Beaufort. We found mostly FY or SY ice here, with a gradient in ice thickness increasing along the meridian. This is indicative of melt caused primarily by shortwave, solar radiation input. Although the 76N ice edge was distinct in concentration, it was more diffuse when measured in ice thickness.
 In summary, both 2006 and 2007 were unusual summers for the Beaufort Sea ice pack. In both years we estimated, from our visual observation and passive microwave ice extent, that all ice less than 3 years old, south of 76N, melted by the end of summer. In 2007 the MY ice observed visually in the south-western Beaufort melted by the end of summer. The years 1998 and 2002 also experienced a low ice extent minima in the Beaufort Sea, and in both years satellite observations indicate all the ice melted south of 76N in the Western Beaufort.
4. Ice Circulation Patterns in the Beaufort Sea
 Sea ice drift in the Arctic is characterized by the Beaufort Gyre (BG), which recirculates ice in the Western Arctic, and the Transpolar Drift Stream (TDS), which transports ice from the Eastern Arctic toward the North Atlantic (Figure 5). The extent of both of these features is variable and oscillates between times with a small (large) BG and increased (decreased) westward influence of the TDS. The size of the BG is related to the residence time of ice before export from the Arctic, estimated with annual mean ice velocity from buoy drift. From the mean velocity field, the time it takes ice in a particular location to exit the Arctic through Fram Strait (residence time) can be calculated, and is plotted as isochrones on Figure 5. The isochrones illustrate a region of recirculation in the Arctic, the BG, and a region of export, the TDS. For the following discussion we define the BG to encompass the annual mean region with residence time greater than five years, east of 120E. The TDS displays interannual variability in its speed, which is illustrated by the area with residence time less than one year. We find the mean of this annual average area over the 29 year buoy record is 1,000,000 km2, which matches estimates of Fram Strait export [Colony and Thorndike, 1985; Kwok and Rothrock, 1998; Kwok et al., 2004]. Figure 6 shows a time series of annual mean BG extent and TDS export, estimated with our isochrone method.
 From ice drift trajectories we can estimate the age of the ice throughout the Arctic. Results from a Drift age Model (DM) [Rigor and Wallace, 2004] are shown for the summers of 2005, 2006, and 2007 in Figure 7. Though there was ice greater than five years old (which we shall refer to as old ice from here on) along the southern bound of the Beaufort ice pack, in both 2006 and 2007, the majority of the ice in the Beaufort Sea was less than 5 years old. The DM model is further supported by the ship observations outlined in section 3, showing ice less than three years old in the north-western Beaufort in both years. The north-western corner of the Beaufort, north of 76N, was predominately FY ice. How did this come about?
 Our analysis of ice drift in the DM shows that from 1989 to present, ice has been entrained into the BG from the Chukchi Sea, and circulated to the south-western Beaufort in less than 5 years. Previous to 1989 Chukchi ice would have either been entrained into the Transpolar drift or recirculated in a much larger BG and therefore would have been at least 10 years old when entering the southern Beaufort. The Chukchi Sea has become seasonally ice covered after 1998. Also after 1998, FY ice observed in the north-western Beaufort was entrained from the seasonal ice zones in the Chukchi and East Siberian Seas the previous winter. In winter 2005/2006 entrainment of this FY ice occurred further to the south than previous years. In winter 2006–2007, we saw circulation of this, now second year ice in the BG, which explains why the MY ice covering much of the Beaufort Sea in 2007 was younger than 5 years.
 The DM and inspection of ice cover at the end of summer from satellite passive microwave ice area demonstrate that ice younger than 5 years old melted out completely south of 76N during the summers of 2006 and 2007. This observation, together with the latitudinally dependent end of summer FY ice thickness north of 76N, indicates that 76N is approximately the location of the thermodynamically controlled ice edge in the Beaufort Sea. South of 76N the end of summer ice extent is controlled by the presence of ice older than 5 years entrained from the north eastern Beaufort Sea and High Canadian Arctic. Hence it was the tongue of old ice in the southern Beaufort which formed the 2006 polynya in the ice pack. A lower extent of old ice in the Beaufort Sea in summer 2007 allowed increased area melt compared to 2006.
 Inspection of the DM model output indicates that large summer ice area in the eastern Beaufort does correspond to the presence of old ice in the south-east Beaufort. Years within the last decade, when the end of summer ice extends south in the western Beaufort (2006, 2005, 2001), correspond with an old ice tongue in the region. Prior to 1989 the old ice extended throughout the Beaufort Sea, and the minimum ice extent was consistently high. After 1989 an old ice tongue is always present throughout the southern Beaufort until the exceptional melt back to 77N in 1998. The reason the 2006 ice tongue formed an unusual polynya in summer 2006 is that during the prior winter ice was entrained from the Chukchi Sea to latitudes below 76N in the western Beaufort.
5. Seasonal Variability of Beaufort Ice Circulation
 It should be kept in mind that the drift patterns described in Figures 5 and 6 are the annual mean ice drift. This is particularly important when you consider that both the periods 1979–1987 and 1999–2007 have similar mean circulation patterns and yet remarkably different old ice distributions. In any given period the mean drift can deviate substantially from the patterns described by the BG and TDS.
 To understand how the ice drift preconditions summer melt in recent years, it is insightful to consider seasonal drift patterns in the south-east, south-west, north-west and north-east regions of the Beaufort Sea (Figure 8). Seasonal mean ice drift was calculated for each region from gridded-interpolated buoy and SSM/I ice velocity fields [Rigor et al., 2002]. Essentially the seasonal mean drift in the regions shown in Figure 8 illustrate BG circulation and imports/exports to the Beaufort Sea (Figure 9). It should be noted that on monthly timescales, large regional ice drift is often characterized by sustained wind forced events. For example in spring 2006 there was a southward push of ice throughout the Beaufort that resulted in a dramatic ice shove event observed at Barrow, after northerly winds were sustained for close to a month. This southward ice push, followed by normal westward drift in the southern Beaufort set up the MY ice tongue that characterized the 2006 summer.
 First we describe the seasonal mean circulation. Throughout the autumn, winter and spring, ice in the Eastern Beaufort drifts south, and the southward ice push in the winter preceding summer 2006 was not unusual here. This southward drift forms the tongue of old ice that extends into the southern Beaufort. In the autumn, winter and spring, ice drift is predominantly westward in the southern Beaufort, pulling this old ice tongue to the south-western Beaufort. This tongue either recirculates northward or drifts into the Chukchi Sea. Summer time ice drift in the entire Beaufort Sea is more variable. Since 1997 we see more frequent northward drift from the western Beaufort in summer (see label A onFigure 9), rather than westward drift toward the Chukchi. Essentially, by summer the old ice tongue may extend into the south-western Beaufort. Prior to 1997 it was most likely this ice was circulated into the Chukchi, whereas in the last decade this ice was more often re-circulated northward, in the summer, before it can reach the Chukchi. The northward drift occurs sporadically, varies in zonal extent, and depending on the following winter circulation, will push ice either into the TDS or BG to the north of 80N. Prior to 1989 this northward drift occurred further west in the East Siberian Sea. So recently there has been enhanced loading of Chukchi and Beaufort Sea ice into the TDS. In the northern Beaufort ice drift is variable throughout the year. However, in the last decade northern Beaufort ice drift has been predominantly eastward during spring and summer. This has allowed greater entrainment of Chukchi ice into the northern Beaufort. As there is reduced recirculation of old ice into the Chukchi and increasing area of total summer time pack melt in the Chukchi Sea, this entrained Chukchi ice, in the last decade, is FY ice. This shift in circulation has resulted in a younger ice pack in the central Beaufort, and narrowing of the old ice tongue, containing the oldest Arctic ice, in the southern and eastern reaches of the BG.
 To summarize, during autumn, winter and spring the Beaufort Gyre exhibits southward drift in the northeast, and westward drift in the south. Northward drift in the west and north occurs during summer and entrainment from the east to the northern Beaufort occurs during spring and summer. The westward and northward drift does not occur every year in the Beaufort Sea, and whether ice is recirculated in the BG from the south-east Beaufort depends on the alignment of northward and westward drift events in time. We find that the frequency of these northward and westward drift events has increased in the last decade, in the Beaufort Sea. The timing is such that old ice drifts northward in summer, and does not recirculate in the central BG. Instead it may enter the TDS or be recirculated back toward the Canadian Archipelago north of 80N. The eastern and southern drift occurs consistently, though at lower rates than the sporadic western and northern arms of the Beaufort Gyre.Figure 8 shows a cartoon of the typical drift patterns in the last decade, showing the Beaufort Sea circulation consisting of entrainment from the east in spring and summer, continuous circulation of this ice clockwise, and the summer time northward ice push that prevents MY ice traveling east into the Chukchi Sea.
 The plot of anomalies of seasonal mean ice drift (Figure 9) show that the 2005–2006 ice drift was not unusual. From the DM we find that the Beaufort ice area at the end of summer is correlated to the age of ice south of 76N, with squared correlation coefficients of 0.77 for the south-west Beaufort and 0.73 for the south-east Beaufort [Rigor, 2005].
 Scatterplots of ice extent at the end of summer minimum verses mean Spring and Summer ice drift are shown in Figure 10. We perform a linear regression of summer minimum ice extent on meridional ice velocity in the south-west Beaufort and zonal velocity in the south-east Beaufort.
 Recent years with large losses of sea ice in the western Beaufort are 1998, 2002, 2003, 2004, and 2007. We calculate correlation coefficients between summer minimum ice extent and meridional ice drift with and without these years to be −0.4 and −0.7 respectively. This indicates that the extreme minimum events are not related to northward ice drift in spring and summer.
 This finding that the northward migration of the ice edge can not be explained by northward ice drift (Figure 10, left) is in agreement with Kwok . Hence the extreme minima are melt backs of the ice pack, that are facilitated by either, or both, a change in thickness of ice in the southern Beaufort during summer and increased heat flux to the ice. Our analysis of the drift of the Beaufort ice indicates that changes in wind forced ice drift has resulted in a younger and therefore thinner ice pack in the southern Beaufort during the last decade, which facilitates thermodynamic ice loss.
 In summer 2007 we saw a reduction of sea ice cover adjacent to Banks Island. During winter, spring and summer, westward anomalies in ice drift were experienced in the south-eastern Beaufort, which might have contributed to a reduction in sea ice in the area. Several similar ice openings have occurred along Banks Island in the passive microwave satellite record, and we investigate if these are related to westward ice drift. With the data shown inFigure 10 (right) we find that the correlation of minimum ice extent to westward spring and summer ice drift is reduced when years with summer time opening along Banks Island are removed. This indicates that opening adjacent to Banks Island, and hence reduced ice extent in the eastern Beaufort, has some relation to westward ice drift during spring and summer. Westward ice drift did play a role in reducing the ice extent in the Beaufort Sea in summer 2007.
6. Relation to Pan-Arctic Ice Drift
 There have been several shifts in the pattern of circulation of sea ice in the Arctic Sea over the last three decades (Figure 6). In the 1970s and 1980s the BG was large, encompassing the entire Pacific sector of the Arctic Ocean, including the East Siberian Sea, Chukchi Sea and Beaufort Sea. Between 1988 and 1993 the BG shifted to the Chukchi and Beaufort Seas. These changes have been linked to atmospheric circulation anomalies [Rigor et al., 2002]. In 1989 there was a transition to high-AO conditions in winter, reducing the intensity and number of high pressure systems in the central Arctic while increasing the occurrence of cyclones. The reduced anticylonic recirculation in the BG in the central Arctic allowed the advection of perennial ice out of the Arctic [Rigor et al., 2002]. Between 1994 and 1996 the BG began returning to its large pre-1987 extent, moving MY ice into the Chukchi Sea. This did not last long enough for the MY ice to be recirculated back to the Beaufort Sea. In 1997 and 1998 the BG shrunk to cover only the Beaufort Sea, and ice from the Chukchi Sea moved into the TDS. This preconditioned the record reduction of ice in the western Arctic in 1998 [Maslanik et al., 1999], a depletion of perennial ice from the Arctic occurred again [Rigor and Wallace, 2004], and the remaining ice in the Central Arctic became considerably younger than in the 1970s and 1980s. Since 1998 the Chukchi Sea has maintained a seasonal ice cover. In 1999 the BG returned to its large extent encompassing the Beaufort, Chukchi and East Siberian Seas. Since then there have only been two years of reduced BG extent, however the ice within the Beaufort Sea remains young compared to the 1980s. Up until 2000 the end of summer Arctic ice extent was inversely correlated to the winter time AO index [Rigor et al., 2002; Overland and Wang, 2005a], indicating that reductions in summer ice extent is partially controlled by high ice export and a reduction in central Arctic ice age, when the polar vortex is strong. One may think that persistent low and neutral-AO conditions, would be required for the perennial ice pack to regenerate to its 1980s condition through recirculation of MY ice in a large BG. However, in recent years the AO has shifted back to more neutral values, with reduced polar vortex and increasing anti-cyclonic atmosphere and ice circulation in the Arctic, and although the annual average extent of the BG has returned to 1980s values (Figure 5), the area of perennial ice continues to shrink in the Arctic.
 Why is this? The record low Arctic ice minima of 2007 was preceded by two winters where perennial ice was loaded into the Transpolar drift and anomalous northward winds were experienced, resulting in export of this ice through Fram Strait, on the “Polar Express” [Nghiem et al., 2007], and compaction of the perennial ice pack against Canada [Nghiem et al., 2007]. Over the last 3 decades there have been three distinct periods when the Polar Express acted to remove perennial ice area from the Arctic: 1988–1989 [Rigor et al., 2002], 1997–1998 [Rigor and Wallace, 2004], 2006–2007 [Nghiem et al., 2007]. The first two events, when the TDS moved ice from the western to eastern Arctic, accelerating export of this ice to Fram Strait, partly explain the reduced extent of perennial ice within the Arctic. An accelerated TDS (Figure 6) in winter 2006–2007 resulted in exceptional transport of ice from the East Siberian and Laptev Seas toward Fram Strait and the Canadian Archipelago. This event set up the ice pack for exceptional melt back in the Eurasian Arctic the following summer [Nghiem et al., 2007]. The studies summarized above explain how ice has been transported out of the central and Eurasian Arctic. They do not, however, directly explain the changes observed in the Beaufort Sea.
 The ice circulation in the Beaufort is characterized by entrainment of ice from the western Arctic into the Canadian Arctic and north-western Beaufort, which may then get swept into the BG. At the eastern edge of the Beaufort, adjacent to Banks island southward drift brings old ice, the majority older than 10 years, into the Beaufort from regions north of the Canadian Archipelago. This old ice has been collected from drift across the Arctic ocean, and is often thought of as recirculation in the BG. We will refer to the area where ice collects in the Canadian Arctic as the “Holding Pen”.
 A stream of this old ice drifts south in the far eastern Beaufort, from the Holding Pen. This ice is swept into the Southern Beaufort and in some years drifts westward toward the Chukchi Sea. In 2006 this tongue of old ice did not melt out during the summer, and later drifted northward toward the Transpolar drift, in the Polar Express event described by Nghiem et al. .
 Two years (1997–1999) with cyclonic circulation during winter time, resulted in a confined BG that recirculated ice in under 5 years in the Beaufort Sea. Ice from the Chukchi Sea entered the Transpolar Drift (the second Polar Express event), and in summer 1998 the Chukchi Bite formed as ice retreated from the Chukchi Sea. This event was followed by several summers where ice drifted north into the central Arctic from the western Beaufort, preventing recirculation of old ice in the Chukchi and Beaufort Seas. Ice entrained into the Beaufort Gyre in the north western Beaufort Sea has been FY ice originating from the Chukchi Sea in recent years. Hence the pack in the north-western Beaufort and central Beaufort Gyre has been relatively young in the last decade.
 The high 2005 and 2006 end of summer ice extents can be explained by the presence of old ice in the southern Beaufort. This old ice became entrained into northward drift [Nghiem et al., 2007], and entered the Holding Pen to the north of Ellesmere Island and Greenland, where some exited from the Arctic through Fram Strait or Nares Straight [Nghiem et al., 2008].
 Ice volume reduction in the Arctic has been documented since 1987 [Rigor et al., 2002; Rigor and Wallace, 2004], however volume reduction in the Beaufort Sea was not observed until 1998, when dramatic ice reduction in the Chukchi and western Beaufort was linked to extreme cyclonic circulation. We demonstrate that this played a role in the increased extent of the seasonal ice zone in the western Beaufort since 1999, triggering the decrease of ice age in the Beaufort Sea. It is anomalous northward ice drift in summer that has resulted in the maintenance of a seasonal ice zone in the Chukchi Sea. The Chukchi ice becomes entrained in the Beaufort Gyre, which has resulted in ice within the central Beaufort Gyre that is less than five years old. This ice melts out south of 76N every summer, and as no older ice in the last decade has been transported from the eastern Beaufort into the Chukchi, the Beaufort Gyre has not been replenished with old ice in the Beaufort Sea, and the Chukchi and Beaufort have maintained large seasonal ice zones. The increased seasonal ice zone is especially apparent in several years since 1998 when end of summer ice extent was low, and similar to those we observed in 2007. Our observations support [Ogi et al., 2010], who find summer time meridional ice drift contributes to low summer ice extents in the Beaufort Sea.
 This paper presents a case study of the Beaufort Sea ice pack in 2006 and 2007. The Beaufort Sea ice pack in summer 2006 was unusual compared to other summers in the satellite passive microwave record (1979 to present), due to a large ice-free polynya in the mid-Beaufort. Summer 2007 had a record Arctic minimum ice extent, and a large ice loss in the Beaufort that has been typical since 1998.
 Ice drift sets up the spatial distribution of ice types in the Beaufort Sea. Since 1998, large summer ice extents are associated with the drift of an old ice tongue into the southern Beaufort. Intrusions of young ice, from the Chukchi Sea have become more common. These bring FY ice into the northern Beaufort Sea, often in late winter. Replenishment of MY ice in the south-western Beaufort has become reduced, as the MY ice, less than 5 years old, reaching this area melts out in summer. These changes result in a younger, thinner ice pack in most of the Beaufort.
 The distribution of ice types and melt was unusual in 2006 due to the presence of an old MY ice tongue in the south-west Beaufort and younger ice to the north of this, entrained from the Chukchi Sea, that melted out during the summer, forming an unusual polynya in the ice cover. Over the last decade the ice pack has become younger in the Beaufort Sea due to: (1) cyclonic circulation in 1997 and 1998 that led to a shrinking of the BG and entrainment of Chukchi ice after the exceptional melt back of 1998 preconditioned the Beaufort Sea with a young ice pack and (2) increase in pole-ward ice drift in summer since 2000 which has reduced recirculation of ice in the Chukchi and Beaufort Seas. This preconditioned the Beaufort to experience record minimum ice extents as younger (less than 5 years) ice melts out south of 76N. The Beaufort ice drift that preconditioned summer 2006 was not unusual. However, this was the first time since 1998 with persistent southward winds that compressed the old MY ice tongue below 75N. As the Beaufort Sea north of the ice tongue was covered with ice less than five years old, this melted out in summer 2006 forming a polynya in the ice pack.
 We find that the recent, 1998, 2002, and 2007, increase in summer melt of Beaufort ice cover is preconditioned by ice drift. Drift of ice older than five years into the southern Beaufort due to sustained southward drift in winter, leads to increased ice extent. Minimum extent prior to 1998 was highly correlated to ice age, and maintains some correlation with ice age after 1998. Long term changes in ice drift are likely due to a combination of changes in wind and ocean surface forcing on the ice pack as well as changes in ice pack strength related to changing spatial patterns of the ice thickness distribution. Westward ice drift in spring and summer probably lead to opening and thin ice in the eastern Beaufort, which leads to reduced minimum ice extent. This will be explored further in a future paper (J. K. Hutchings and I. G. Rigor, manuscript in preparation, 2012).
 To summarize, prior to 1998 the Beaufort minimum ice extent was primarily controlled by ice dynamics, through the advection of old ice into the southern Beaufort. From 1998 to date, there have been several years when the ice in the western Beaufort was MY ice less than 5 years old. These years have seen minimum ice extents in the western Beaufort that are controlled by thermodynamic ice melt. Hence the Beaufort Sea ice minimum has gone from a regime controlled purely by ice dynamics to a regime where both thermodynamics and dynamics play a role. However, it is ice dynamics (the advection of old ice into the south-western Beaufort) that accounts for the large inter-annual variability in the Beaufort Sea summer ice edge.
 Recent modulations of atmospheric circulation patterns (especially a shift to increased summer time meridional transport since 2000) have been important in setting up a younger Beaufort Sea ice cover in the last decade. We determined that the Beaufort ice pack has been maintained in a state of young ice cover, even under a return to increased anti-cyclonic ice circulation, due to an increase in pole-ward ice drift during summer that has impeded recirculation of MY ice in the Beaufort Sea. To get off the current trajectory would require a reduction in this pole-ward transport in concert with an maintenance of anti-cyclonic ice drift, perhaps sustained over several years as entrainment of ice into the central Beaufort Gyre is sporadic.
Kwok and Cunningham demonstrate that the Southern Beaufort has become a sink, though melt, for multiyear ice area. We find ice that has survived transit of the southern Beaufort in summer, has been impeded from drifting into the Chukchi through increased pole-ward ice transport in summer. In the last decade the majority of the Beaufort Sea has maintained an ice cover under five years old through increased entrainment of ice from the Chukchi, and no recirculation of ice older than five years into the Chukchi. Since 1999 this ice circulation has acted to maintain a reduced ice cover in the Beaufort. It is interesting that in summer 2010 multiyear ice was transported into the Chukchi Sea, and lost area to melt during its summer transport to the East Siberian Sea. The area of this ice available for recirculation in the Beaufort Gyre is much reduced compared to ice that followed a similar trajectory two decades ago.
 For the ice age distribution within the Beaufort Sea to return to pre-1997 status would require survival of MY ice in the Southern Beaufort and westward circulation of this toward the Chukchi Sea. Then this MY ice must survive summer and become entrained back into the Beaufort Gyre. In summer 2010 MY ice did drift into the Chukchi Sea, however this drifted further west into the East Siberian Sea, experienced substantial melt and may become loaded into the Transpolar Drift.
 Further work is required to determine whether the ice dynamics in the Beaufort Sea in the last decade are within natural variability or due to long term changes in winds, and thinning of the Arctic ice pack, which may be on a trajectory toward a new climatic state with an Arctic of reduced ice cover.
 Much gratitude is extended to the volunteer ice observers on the Louis S. St. Laurent: Patrick McKeown, Helen Drost and Mike Dempsey. Many thanks to Alice Orlich, a full time observer on the 2007 cruise. The SSM/I passive microwave NASA Team data was provided by the National Snow and Ice Data Center. The International Arctic Buoy program provided drifting buoy data. Mark Ortmeyer provided advanced processing of IABP data for 2005–2007. This work was funded by the National Science Foundation (OPP 0520574, OPP 1023662) and the ship observation program was partially funded by JAMSTEC. The Institute of Ocean Science, B.C., Canada, Fiona Mclaughlin and Andrey Proshutinsky, Woods Hole Oceanographic Institute, supported berths and helicopter time on the Louis S. St. Laurent. Many thanks to the captain and crew of the Louis S. St. Laurent on JWACS-JOIS cruises in 2006 and 2007. This paper greatly benefited from the comments of four anonymous reviewers, Son Nghiem and Andrey Proshutinsky.