The role of eddies and energetic ocean phenomena in the transport of sediment from shelf to basin in the Arctic



[1] Particle fluxes over the Beaufort continental slope (70.548°N/140.045°W; 710 m) were measured at two depths (290 and 490 m) during the fall 1992 to spring 1993 period. Sequential sediment traps recorded a peak in particle flux under heavy ice cover in midwinter. The fine-grained, highly terrigenous material was consistent in composition with bottom sediments from the shelf and upper slope. The peak occurred coincident with high current speed and isopycnal displacements characteristic of cyclonic eddies, and upwelling-downwelling (UW-DW) cycles driven by winds and drifting ice. Such phenomena may be the operative mechanisms for transporting sediments from shelf to basin in the Arctic. Available evidence favors the origin of the midwinter peak to be rainout of particles from the cyclonic eddies; we suggest that the particles were earlier entrained into the eddy when it impinged upon the upper slope. Estimates show that sediment-laden water in an eddy could sustain rainout at 250 mg m−2 d−1 for ∼120 days, so that sediment could have been transported 100–500 km with the eddy before reaching our site. Current speeds generated by wind-driven UW-DW events were insufficient to resuspend sediment on the outer shelf, and the timing of strong events was unconvincing as a link to the enhanced sediment flux over the slope 50 km to the north. Bottom current must exceed 19–36 cm s−1 depending on the sediment type to resuspend material from the seabed of the Mackenzie shelf.

1. Introduction

[2] Knowledge of Arctic shelf-basin exchange processes is critical to the construction of realistic material budgets. During spring and summer, Arctic rivers such as the Mackenzie deliver vast quantities of sediment to the shallow marginal seas of the Arctic; the material moves across the shelf via cycles of deposition, resuspension and transport in nepheloid and surface layers [O'Brien et al., 2006]. Although particulate matter is clearly supplied to Arctic Basins laterally from the surrounding margins [Hwang et al., 2008; Fahl and Nöthig, 2007; Honjo et al., 2010], the mechanisms and timing of this flux remain unclear. Processes that circumvent topographic constraints and generate cross-shelf currents to transport shelf water and associated properties offshore include Ekman drift, wind-induced upwelling and downwelling (UW, DW), meandering jets and eddies [Kadko and Muench, 2005].

[3] The main circulation features of the western Arctic Ocean play central roles in sediment transport. A narrow, mostly eastward flowing boundary current, the Beaufort Shelfbreak Jet [Pickart, 2004; Nikolopoulos et al., 2009] and the large-scale anticyclonic Beaufort Gyre in Canada Basin [Proshutinsky et al., 2002] are both highly sensitive to wind forcing. Winds and drifting sea ice drive UW and DW along the Beaufort shelf margin particularly in areas like Mackenzie Trough [Williams et al., 2006; Carmack and Kulikov, 1998]. Eddies are ubiquitous in Canada Basin and the Beaufort Sea. Most are anticyclonic [e.g., Krishfield et al., 2002; Manley and Hunkins, 1985], but rarer, cyclonic eddies have also been observed and described in this area [e.g., Padman et al., 1990]. Mesoscale eddies are shed from the shelf break jet into the Canada Basin [Spall et al., 2008; Nikolopoulos et al., 2009], carrying particles, organisms and water properties from the shelf to the basin. This function of eddies could be critical for otherwise unproductive food webs of the central Arctic, for carbon mineralization and sequestration and for oxygen consumption [Llinás et al., 2009; Mathis et al., 2007; Kadko et al., 2008]. The generation of eddies in the Arctic remains speculative and may be complicated by the presence of ice cover. For example, modeling work by Chao and Shaw [2003] predicts the generation of eddies due to dense outflow through a submarine canyon into a baroclinic boundary current and Chao and Shaw [1998] have proposed a theory for eddy generation via surface buoyancy forcing (brine or fresh water) under rigid drifting ice; we note that the examples provided involve intense brine and freshwater sources that are unlikely to occur naturally.

[4] We examine physical processes accompanying a midwinter peak in sinking particle flux below the halocline over the Beaufort slope and consider the roles of eddies, current, wind and ice drift on the source and fate of these sinking particles.

2. Background

[5] Our AM1-92 mooring (70.548°N/140.045°W; 710 m) was deployed to the NW of Mackenzie trough during the winter of 1992–93 (Figure 1). The instruments on the mooring included Baker-Milburn sequential sediment traps [Baker et al., 1988] (290 and 490 m), Aanderaa RCM4 current meters (59, 161, 512 m), SBI Seacats measuring temperature and salinity (60, 162 m) and an ice-measuring sonar (APL-UW ULS at 55 m). The drag of strong current occasionally pulled instruments deeper in the water column, in one instance as much as 24 m; we have adjusted the density time series to constant depth to compensate for pull-down. Knowledge of concurrent ice conditions in the Beaufort was derived from the weekly charts provided by the Canadian Ice Service, from the ULS and from a drifting ice buoy, IABP #11252 [Colony and Rigor, 1993, 1995]. Wind velocity was derived from the National Centers for Environmental Prediction (NCEP) atmospheric reanalysis project (10 m elevation; 71.426°N/140.625°W).

Figure 1.

Location of AM1-92 mooring (70.548°N/140.045°W; 710 m), National Centers for Environmental Prediction (NCEP) wind grid point (10 m elevation; 71.426°N/140.625°W), drift track of IABP buoy, and inset of ice drift track for trap intervals 4–5 (dots C–F indicate timing maximum change of sigma-theta (see Figure 3g, C–F)). Depth contours are 50, 100, 200, 500, 1000, 1500, 2000, 2500, 3000, and 3500 m. Mooring diagram with instrument depths is to the right of the map.

[6] Sediment trap samples were preserved with mercuric chloride and analyzed for total dry weight (TDW) flux. After acidification to remove inorganic carbon, particulate organic carbon (POC) was measured on a Leeman CE440 Elemental Analyzer standardized with acetanilide. Biogenic silica (BIOSI) was determined by the method described by Mortlock and Froelich [1989] and chlorophyll a (CHLA) was analyzed according to Parsons et al. [1984]. Aluminum (Al) was measured by inductively coupled plasma with optical emission spectrometry (ICP-OES) analysis of acid solutions, produced by the dissolution of glasses prepared by fusion with lithium metaborate (protocols of Earth and Ocean Sciences, University of British Columbia, Vancouver, Canada). Measured POCtotal was divided into biogenic and refractory POC where the term refractory is used instead of terrestrial (as in the work of O'Brien et al. [2006]) because this fraction may also contain marine components. A ratio %POCrefractory: %Al of 0.16 was used to calculate the POCrefractory content as described by O'Brien et al. [2006], and the biogenic fraction (%POCbiogenic) was determined as %POCtotal minus %POCrefractory. POC fractions were converted to particulate organic matter (POM) using appropriate factors: 1.87 for biogenic organic matter [Anderson, 1995] and 1.22 for refractory organic matter [O'Brien, 2009]. Percent BIOSI was converted to %OPAL using a factor of 2.4. The terrigenous fraction was obtained by subtracting %POMbiogenic, %POMrefractory, and %OPAL from 100%.

[7] Note that the terrigenous fraction consists of clays, other minerals and (Ca, Mg) carbonates. The latter are derived from the biogenic production of calcium carbonate shells (e.g., by foraminifera and/or coccolithophoridae), inorganic precipitation, and/or dolomite and limestone rock fragments. Ideally, the biogenic fraction of the total CaCO3 would be identified, but unfortunately suitable data are not available. The contribution by marine carbonate shells is likely to be insignificant as few foraminifera and no coccolithophorids were observed microscopically. Carbonates in Mackenzie Shelf surface sediments are present mainly as dolomite [Pelletier, 1984] and occur in higher concentrations inshore (∼16–27% by weight) versus ∼8–16% offshore.

3. Results

[8] TDW fluxes in the fall/winter of 1992–93 (101–479 mg m−2 d−1; >80% terrigenous) exhibited a midwinter peak (trap intervals 4 and 5) of highly terrigenous (93–94%), fine-grained material at both 290 and 490 m (Figure 2a). In this peak, the TDW flux at 490 m exceeded that at 290 m by 78 mg/m−2 d−1 in trap interval 4 and by 8 mg m−2 d−1 in interval 5. It is difficult to attach significance to these differences due to uncertainties with respect to sediment trap collection efficiencies, to variable settling speeds of particles, and to the variations—in both time and space—of the flow fields above the traps [Siegel and Deuser, 1997].

Figure 2.

Fluxes (left axis and plots with open and filled black dots) and compositions (right axis and plots with open and filled red triangles) of trapped particles for the seven 35 day sediment trap intervals: (a) TDW fluxes partitioned into terrigenous flux, refractory particulate organic matter (POMrefractory) flux, biogenic particulate organic matter (POMbiogenic) flux, and biogenic opal (OPAL) flux; (b) aluminum (Al) flux and %Al; (c) total particulate organic carbon (POC) flux and %POC; (d) biogenic silica (BIOSI) flux and %BIOSI; and (e) chlorophyll a (CHLA) flux. Flux units are mg m−2 d−1 except for CHLA flux, which is ug m−2 d−1.

[9] Aluminum contents (7.6–8.8%; Figure 2b) were highest in the deep trap indicating sorting of terrigenous material during transport. POC contents (2.5–6.3%; Figure 2c) were highest in fall/early winter and lowest at the midwinter peak. Al, POC, and BIOSI fluxes all exhibited midwinter peaks (Figures 2b2d). CHLA fluxes decreased during fall and increased slightly at the midwinter peak (Figure 2e) suggesting remobilization of settled biological material from surface sediments.

[10] AM1-92 was initially in open water but by 22 September ice cover was 40–80%, with freezeup complete by early October. Ice cover modulates wind stress in a nonlinear fashion. In winter, DW favorable wind in the Beaufort pushes ice toward the coast where it meets resistance and slows, thereby reducing DW stress—this is evident in the ice drift and wind records (see asterisk in Figures 3b and 3d) where ice drift ceased despite strong winds to the E-NE. In contrast, UW favorable winds move ice to the WNW with no nearby obstruction and the ice moves with the wind exerting UW favorable stress on the ocean [Williams et al., 2006]. Consequently, it is important to consider UW and DW in the Arctic in terms of both ice movement and wind.

Figure 3.

(a) Ice draft (running average). (b) Daily average ice drift velocity—the asterisk indicates period where ice drift has ceased despite strong winds to the E-NE. (c) Daily average current velocity (59, 161, 512 m)—note different scale for the 512 m data. (d) Speed of ice drift (black) and wind speed (light blue). (e) Pressure record indicating depression of mooring due to strong currents. (f) Current speed (59, 161 m). (g) Sigma-theta (60, 162 m; hourly) and eddies labeled A–F. Cyclonic rotation of the eddy is depicted; the beginning and end of the 35 day trap intervals are indicated on the x axis.

[11] Ice drift occasionally reached high speeds in winter (Figures 3b and 3d) indicating considerable ice-ocean stress. Ice displacement repeatedly cycled between westerly and easterly during fall and winter. Sediment trap interval 4 exhibited an initial DW favorable period followed by 12 days of minimal ice drift despite strong DW favorable winds. The fifth interval exhibited alternating UW and DW favorable periods and the highest ice drift speed (48 cm s−1) was DW favorable. A DW impetus from the ice will be linked to an eastward acceleration of the water column over the continental slope, and vice versa. UW conditions predominated during intervals 6 and 7.

[12] During trap intervals 1–5, the pinching of isopycnals toward a central density surface and increased current speed (Figures 3f and 3g) suggest six encounters with cyclonic eddies (features A–F; Figure 3g) between 60 and 162 m. It is not possible to assign origin, size or age to these eddies, but we note that the Beaufort Shelf (Canadian or Alaskan) during freezeup is a plausible source for the colder fresher core water at 60 m. Although most eddies observed in Canada Basin and the Beaufort Sea have been anticyclonic [e.g., Krishfield et al., 2002; Manley and Hunkins, 1985], we observed only cyclonic eddies, which are rarer but have been reported in the Arctic [e.g., Padman et al., 1990].

[13] At 59 and 161 m at AM1-92, maximum current speeds (44 and 62 cm s−1, respectively) occurred during interval 4 (Figures 3c and 3f). At 512 m, well below the eddy and near the 490 m trap, current speeds were much lower (<10 m s−1; Figure 3c) and direction varied approximately parallel to bathymetry with a net vector to the SE particularly during intervals 4–7. Because strong orbital speeds of eddies dominated the currents at 59 and 161 m, determination of background flow was not possible.

[14] Events A and C (Figure 3g) persisted at AM1-92 for ∼4 days, and associated patterns of speed and direction with respect to the density anomalies are consistent with meandering over the mooring site (depicted in Figure 4). The maximum density excursions generally occurred with lower current speeds as would be expected in an eddy's core. Events A and B could have been the result of a single eddy that impinged on the mooring site twice—note that these events were not accompanied by increased particle fluxes. Likewise, features C, D, E and F may also have been repeated encounters with a single eddy (Figure 3g). The repeated backtracking of the ice cover during this period (inset, Figure 1) could have provided the impetus to the water column so as to make the eddy also move back and forth. Ice with deep keels covered AM1-92 during the winter (Figure 3a). Because the drag coefficient of rough ice is high, even slow ice movement can apply a large stress on the upper ocean. Figure 1 (inset) shows the timing of the passage of eddy features in relation to the ice drift. The appearances of eddies C, D, E, and F (Figure 3g) over AM1-92 coincide with abrupt changes of direction in ice drift from SE to NW, suggesting that an eddy may indeed have been pushed back and forth over the mooring with changing background flow along the continental slope. Unfortunately this intriguing idea is difficult to explore further with the data at hand.

Figure 4.

Bathymetric cross section at the AM1-92 mooring showing the mooring location and depths of traps and instruments. Transects of sigma-theta at 60 and 162 m depth across a measured cyclonic eddy (see Figure 3g) are plotted at appropriate spatial scale (10 km), seaward of the mooring at the upper right. The cartoon on the upper slope (left) depicts to scale the topographically trapped boundary current with associated resuspension and nepheloid layers. The stick plot (lower left) shows current velocity during sediment trap intervals 4 and 5, and the schematic to its right shows how the measured eddy current is linked to the position and drift (hatched arrow) of the cyclonic eddy relative to the mooring; the E and SE edges of the eddy are in contact with the AM1-92 mooring site at the times A and B, respectively. The hatched arrow indicates the direction of drift of the eddy.

[15] The co-occurrence of high particle fluxes and eddies, and the composition of these particles (Figure 2), raise the possibility that particles were eroded from the upper slope by the eddies at some earlier time and transported with the eddies to the site of collection. Active sediment resuspension via high eddy currents was not likely occurring at the time of our observations—the strong currents to the ENE during trap intervals 4 and 5 indicate that the E and SE boundary of the eddy was over the mooring, so that the eddy itself was to the northwest, in deeper water (depicted in Figure 4).

4. Discussion

[16] Earlier study of sediment erosion, transport and deposition in the southern Beaufort Sea has illustrated that an understanding of time-variable physical phenomena is vital to the interpretation of sediment dynamics in this area [e.g., O'Brien et al., 2006; Forest et al., 2008]. The increase in midwinter particle flux (at intervals 4–5) was 2.4–3.2 times the background for TDW flux and 1.4–1.7 times the background for POC flux. The background flux during the fall/winter period was about 150 mg m−2 d−1 for TDW and 6.8 mg m−2 d−1 for POC (see Figure 2); this indicates that mechanisms capable of resuspending intermediate amounts of sediment operate fairly continuously over the slope in this area. However, here we are concerned with understanding the transient event that contributed the midwinter peak in particle collection rate. The background TDW flux (∼150 mg m−2 d−1) is clearly not to be neglected in a quantitative assessment of regional sediment transport, but this is not the focus of this paper.

[17] The available physical data indicate that, despite heavy ice cover, there were processes capable of delivering highly terrigenous sediment to the sediment traps (290 and 490 m) at the 700 m isobath on the slope. The peak in particle flux co-occurred with two energetic phenomena—namely, eddies with high current and alternating UW and DW driven by ice-imposed surface stress. Both phenomena may be capable of eroding sediment from the upper slope and shelf to the south of AM1-92.

[18] Our mooring with sediment traps was 46 km seawards of the shelf break (at approximately the 80 m isobath) and the traps were at 210 m and 410 m above the seabed at the 700 m isobath (Figure 4). Keeping this in mind, there are three issues to consider: 1) Possible sources of sediment available for resuspension; 2) Source of energy to accomplish resuspension; and 3) Possible mechanisms to transport the mobilized sediment from the place of origin (likely the shelf/upper slope) to that of collection.

4.1. Source of the Resuspended Particles

[19] The composition of the fine-grained particles collected over the slope in the midwinter peak—high terrigenous content (93–94%), low %POC (2.5–2.8%), low %BIOSI (1.7–1.9%), and slightly elevated CHLA flux—strongly suggest that the collected material was resuspended bottom sediment from higher on the slope, from the shelf edge or from the shelf itself.

[20] Mid and bottom nepheloid layers, evidence of resuspension, have been repeatedly documented in this area; they are embedded in the bottom Ekman layer which provides a mechanism for cross-isobath transport of the suspended particles toward the shelf break [e.g., O'Brien et al., 2006; Forest et al., 2007]. The Mackenzie River delivers vast quantities of suspended sediments to the Mackenzie Shelf. It is a likely source of fluidized mud, which is increasingly regarded as prevalent on continental shelves that receive large river inflows [Wright and Friedrichs, 2006; McKee et al., 2004; Kineke et al., 1996].

[21] The source of the sediments trapped at AM1-92 could have been a readily mobilized fluidized mud or bottom sediment eroded by strong current either from the continental shelf, where waves, wind and ice drift can generate strong bottom currents [Hill et al., 1991], or from the continental slope, where energy to erode sediment could be provided by the strong current within an eddy or a boundary current [Darby et al., 2009]. The source could be hundreds of kilometers away, with the suspended sediment transported to the mooring within the boundary current or within eddies that detach from it.

4.2. Energy for Bottom Sediment Resuspension

[22] Fluidized muds are linked to the short-term fate of the immense seasonal influx of suspended fluvial sediments via the Mackenzie River. A large portion of newly delivered fluvial and biogenic sediment may well reside in an easily remobilized pool within a few decimeters of the seabed. This material behaves as a dense fluid, not a solid. It likely experiences repeated renewal during storms and consolidates only slowly into cohesive bottom sediment. Mud like this may exist without consolidation for a long period of time, and be readily mobilized by bottom currents and moved with favorable forcing out to the slope.

[23] If there is no fluidized mud present, resuspension of bottom sediments requires sufficient energy in currents close to the seabed on the shelf and upper slope. In this section we investigate 1) the current speed required to resuspend bottom sediments and 2) the frequency that these speeds are likely to be encountered at the seabed at the shelf edge and over the upper slope.

[24] The key aspect of resuspension is the shear velocity at which particles are lifted from the seabed and here we estimate the range of bottom current velocities required to resuspend bottom sediments. Threshold values have been measured for a variety of sediment types. Walker et al. [2008] determined erosion thresholds of shear velocity on the Mackenzie Shelf as 1.05–1.30 cm s−1 for upper, less-consolidated sediments and 1.3–1.7 cm s−1 for the lower, more consolidated sediments. Shear velocity, u*, is defined by: τb = ρ (u*)2 where τb = bottom shear stress and ρ = water density. τb is customarily estimated from flow speed (UH) at a fixed distance (H) off the bottom, an appropriate drag coefficient (CH), and seawater density (ρ) according to the equation: τb = ρ CH (uH)2. Drag coefficients (C100) tabulated for H = 1 m are typically 0.0022 for mud and 0.0030 for sand/mud [Soulsby, 1997]. Based on these equations, applied to mud and mud/sand bottoms at 1 m elevation, speed of 19–28 cm s−1 would be required for resuspension of the upper layer and 24–36 cm s−1 for the lower layer. Such speeds have been observed midwater in this area in three main circumstances: 1) during storms and most commonly in water <10 m deep, where orbital circulation of surface waves can be appreciable at the seabed [Hill et al., 1991]; 2) during intensifications of the Beaufort shelf break jet [Forest et al., 2008]; and 3) in rotary currents associated with eddies [Krishfield et al., 2002]. Note that mobilization of fluidized mud occurs at significantly lower current speed.

[25] To assess the likelihood and frequency of bottom sediment resuspension, the data archive at the Institute of Ocean Sciences, Sidney, B.C. was searched for near-bottom current speed over the slope high enough to resuspend sediment at the seabed, namely, >19 cm s−1 and >24 cm s−1 for unconsolidated and consolidated sediments, respectively. Near-bottom current observations on the Beaufort Slope are scarce. The available data indicate that bottom current fast enough to resuspend sediments from the upper slope is very infrequent. Two examples from these archives are provided below.

[26] The first example comes from a site located at the shelf break halfway along the Mackenzie shelf edge (site ISC92-2; 70.952°N 133.727°W) in 83 m of water (ADCP data at 73 m depth). This record coincides with our sediment trap collection in the fall/winter of 1992–93, but the location is 225 km to the east. However, the ocean response at the shelf edge near AM1-92 will be very similar to that measured at ISC92-2, because impulses of wind force to the ocean in winter are mediated through continuous ice cover. The rigidity of the ice cover ensures that ocean forcing by ice will be very similar over hundreds of kilometers along the direction of drift. For the fall/winter period of 1992–93 (coinciding with AM1-92 trap intervals 3 to 6), the maximum current speed ∼10 m above the bottom at ISC92-2 was 27.7 cm s−1 (Figure 5). Current between 19 and 24 cm s−1—with sufficient energy to resuspend upper, less-consolidated sediments—occurred only for only 2.5% of the measurements, most of which occurred during interval 5. Current greater that 24 cm s−1—with enough energy to resuspend the lower more consolidated sediments—occurred for only 1% of the measurements, all of which occurred during interval 5. These contemporary data suggest that the stress exerted by drifting ice does not provide enough impetus at the right time for appreciable resuspension of bottom sediments at the shelf break. However, fluidized mud could have been mobilized, if present. This record provides only a glimpse of the energy available for resuspension but it does highlight the importance of collecting current data in a manner that would further our understanding of sediment mobilization and transport in this area.

Figure 5.

Current vectors (from ADCP data) plotted with respect to the along-shelf direction (ENE to WSW) at site ISC91-2 (70.886°N/133.732°W) located at the shelf break (83 m) midway along the Mackenzie shelf edge. Data is from the data archive at the Institute of Ocean Sciences, Sidney, B.C., Canada. Current directions favorable to upwelling (UW) and downwelling (DW) at the shelf edge are indicated to the right of the plot.

[27] The second example of current data close to the bottom comes from sites over the upper slope off the Mackenzie Shelf during the period 1984–1986. The data examined were from three locations between 134° and 139°W (bottom depths of 225, 302, and 280 m) with measurements at 10, 40, and 6 m above the bottom, respectively. Currents exceeding 30 cm s−1 (here considered as the reasonable resuspension threshold) were rare (3–8 events per year), contributing only 0.5% of the total measurements for two of the records (1985–86) and 0.04% of the total measurements for the third record (1984–85). One of these events of elevated currents (>30 cm s−1) was clearly associated with an eddy; other events were associated with strong winds, and for some, the origin could not be determined. Certainly, along a featureless continental slope, resuspension of sediments due to elevated bottom currents does occur but may be rare at any specific location. Elevated currents have been reported in the shelf break jet at the edge of the Alaskan Shelf [e.g., Pickart, 2004] and these could also be an important source of sediment resuspension.

[28] It is plausible that energy for resuspending sediment from the upper slope could originate in topographically trapped internal waves [McPhee-Shaw, 2006; Puig et al., 2004]. There are no data from this area to assess this possibility.

4.3. Transport Mechanisms

[29] Processes initiating sediment movement, resuspension, and advection at the shelf edge and over the slope are generally not as well understood as those over the shelves [Davies and Xing, 2005; Huthnance, 1995]. In the Arctic, these processes are less well understood due to the presence of ice cover and extreme stratification. Numerous studies have demonstrated that energetic events capable of moving sediment occur at shelf edges and that the specific geography and hydrography of each region is important—e.g. shelf width and shelf break depth, stratification, tidal influence, special topography such as canyons [Davies and Xing, 2005, and references therein]. Here we confine the discussion to two processes identified in our study as potentially capable of transporting enhanced levels of suspended particles out over the slope to AM1-92—namely, eddies and UW-DW cycles at the shelf edge.

4.3.1. Eddies

[30] Eddies, which repeatedly appear at AM1-92, provide the mechanism most closely correlated to the enhanced winter delivery of sediment to our traps. Cyclonic eddies associated with high current speeds co-occurred with the midwinter peak in particle flux over the slope. Knowledge of where and how these eddies were generated would provide a better understanding of how they acquire their sediment load. If eddies are to be implicated in the erosion of sediment then they must form in or later impinge upon places (viz. shallower water) where the eddy circulation can resuspend and move sediment. The subsequent motion of the eddy must then be capable of carrying entrained suspended particles across isobaths out into deeper water.

[31] Assigning origin, size, or age to the observed cyclonic eddies is not possible with the available data. The eddies observed in midwinter at AM1-92 were too distant from the shelf break—46 km to the south of AM1-92—to be actively eroding sediment at the time of observation, suggesting that they and any associated sediment burden originated remotely from AM1-92. The colder fresher core water at 60 m in the eddy might indicate that the eddy originated on the Beaufort Shelf during freezeup, implying that the eddy was generated in shallower water and raising the possibility that the high currents within the eddies may have contacted bottom sediment on the shelf or upper slope. Not all the eddies observed were linked to enhanced sediment capture—e.g., the passage of eddies A and B (Figure 3g) during trap intervals 1–2. These may have been generated without ever coming into contact with the seabed.

[32] To demonstrate the plausibility of eddies as means of sediment transport, we have combined available data with estimates for suspended sediment concentration within the eddy and for rainout rate to estimate the distance an eddy could travel after entraining sediment before losing all its particle load. The eddies observed during the winter peak (intervals 4–5; Figure 2a) had a minimum vertical extent of 150 m. Assuming a suspended sediment concentration of ∼200 mg m−3 within the eddy [e.g., Washburn et al., 1993; Bornhold, 1975] implies a minimum suspended sediment load of ∼30,000 mg m−2. Assuming a background flux of ∼150 mg m−2 d−1 (Figure 2a), the excess flux of ∼250 mg m−2 d−1 over a 70 day period (intervals 4–5)—if ascribed to rainout from an eddy—equates to 17,500 mg m−2 of trapped sediment over the winter peak. This amounts to 58% of the 30,000 mg m−2 sediment load in the eddy. Theoretically, this eddy could sustain a rainout rate of 250 mg m−2 d−1 for perhaps 120 days, which at postulated drift speeds of 1–5 cm s−1 corresponds to 100–500 km of travel before losing the load. A settling rate ≥15 m d−1 would be required for particles to settle to 290 and 490 m within our 35 day trap intervals.

[33] We acknowledge that this estimate represents an upper limit for the time period that the eddy could retain sediment; it does not factor in lateral transport within nepheloid layers. Our estimates establish that the observed eddies had the capacity to carry and rain out sediment within the time period of our study (e.g., if the eddy entrained sediment in the late fall) and over the distances appropriate to our study (shelf edge 46 km south of AM1-92).

[34] Dramatic topographic features such as Barrow Canyon and Mackenzie Trough may be important. The mouth of Barrow Canyon to the west of AM1-92 has been linked to eddy generation. The water in the core of an eddy observed by Pickart et al. [2005] east of Barrow Canyon, was highly turbid and because this eddy was close to its proposed source, the turbid water suggested resuspension accompanied eddy generation. A strong eddy was also observed in Mackenzie Trough, southeast of AM1-92 in September 2002 [Williams et al., 2006], suggesting that topography favors eddy generation here as well.

[35] The mechanism of eddy generation is an active area of research due to the high prevalence of eddies in the Beaufort Sea and Canada Basin. Chao and Shaw [2003] theorize through models that dense outflow through a submarine canyon into a baroclinic boundary current generates vertically stacked counter-rotating eddies, separated at the pycnocline with anticyclonic rotation at depth—the top cyclonic eddy is theorized to be eroded by friction against a rigid ice cover. The rotation of the eddies observed at AM1-92 (with all measurements from below the pycnocline) is cyclonic, contrary to the model results of Chao and Shaw [2003]. This theory, therefore, provides an unsatisfactory mechanism of generation for the eddies observed at AM1-92. Although rarer, cyclonic eddies are not unprecedented in the Arctic [e.g., Padman et al., 1990] but their mode of generation is speculative. For example, a model by Chao and Shaw [1998] predicts that ice cover drifting over a freshening source will generate a pair of vertically stacked counter-rotating eddies with an anticyclonic eddy above the pycnocline (which is easily eroded away due to frictional forces supplied by the drifting ice) and a cyclonic eddy extending into the halocline below. Conversely, ice cover drifting over a brine source generates a cyclonic eddy above the pycnocline (eroded away by the ice cover) and an anticyclonic eddy extending below the halocline. It must be emphasized that these models use more intense brine and freshening sources that are likely to be found in nature—salinity changes of 3–5 and 7 psu/d over 50 m depth.

[36] The role of cyclonic eddies in this area is important—but difficult to assess—given the link established here between these and enhanced particle fluxes. This association has been demonstrated in other studies as well. For example, in the southern East China Sea the offshore transport of river-borne terrigenous particles was facilitated by a cyclonic eddy [Hsu et al., 1998], and in the lee of the Hawaiian Islands, empty diatom frustules from a bloom were exported implying that the associated cyclonic eddy was providing a silica pump [Maiti et al., 2008]. In the Bering Sea, Mizobata et al. [2002] documented upwelling of nutrients to the surface in the center of a cyclonic eddy. These studies lead to the implication that, in addition to enhancement of particle fluxes in the Arctic Ocean, cyclonic eddies may play an important role during the ice covered period in supplying nutrients to the surface, particularly noteworthy in these highly stratified waters.

[37] Although it is clearly not possible to identify the mode of generation and place of origin of the observed eddies and any associated sediment entrainment, it is important to note that many of the elements considered important for the generation of eddies close to the shelf edge are present—namely, drifting ice, elevated currents, pools of freshened shelf-origin waters at the shelf edge [Macdonald et al., 1999], anomalous water properties within the halocline, and submarine canyons. In addition, there is a precedent for suspended sediments and other shelf properties being carried out over the Beaufort Slope within eddies [e.g., Llinás et al., 2009; Mathis et al., 2007; Kadko et al., 2008].

4.3.2. UW and DW Driven by Ice-Imposed Surface Stress

[38] Enhanced fluxes of terrigenous particles over the slope could plausibly be a consequence of ice-driven UW-DW oscillations, with sediment eroded from the upper slope or supplied by reservoirs of fluidized mud on the shelf. These reservoirs could be mobilized and transported out over the slope in nepheloid layers during the DW phase of the cycle. Driven by winds and drifting ice, episodes of UW and DW favorable conditions were present during the 1992–93 fall/winter periods at ISC92-2, as demonstrated by the current patterns in the along-shelf direction at this site (Figure 5). With continuous ice cover, it is reasonable to assume similar forcing over appreciable distance along the shelf edge.

[39] During trap interval 4, currents at ISC92-2 (Figure 5) were predominantly DW favorable, the maximum current was 20.4 cm s−1, and the duration of elevated currents was short—a scenario unlikely to have eroded and transported bottom sediment sufficiently to account for the midwinter peak in sediment fluxes over the slope. Only 0.7% of the data records during interval 4 showed currents capable of eroding an upper unconsolidated layer of sediment and there were no currents during this interval capable of eroding a lower consolidated layer. Note that during interval 4, ice drift virtually ceased for a period of ∼12 days despite strong E-NE winds and thus reduced the DW impetus during this period (Figure 3b).

[40] During trap intervals 3 and 5, the strongest currents (maximum of 27.7 cm s−1) were UW favorable and therefore unfavorable to the transport of resuspended bottom sediments from the shelf edge out over the slope. Overall, the magnitude, timing, and directions of currents at ISC92-2 indicate insufficient forcing at the shelf edge favorable to the transport of suspended sediments to the traps at our mooring site over the slope during the fall/winter of 1992–93.

[41] To the east of AM1-92, Forest et al. [2007, 2008] reported episodic events of high particle fluxes over the upper slope (at 200 m over the 300 m isobath) in the winters of 2004, 2005, and 2006—these events were attributed to resuspension of bottom sediments over the upper slope due to abrupt pulses (>30 cm s−1) of Pacific-origin water within the Beaufort Shelfbreak Jet and to strong rotary currents of submerged eddies formed by baroclinic instability of the shelf break current. In the study by Forest et al. [2008], the marked increase in particle fluxes (TDW flux > 10 g m−2 d−1) in January 2005 coincided with an intense storm with strong and prolonged NW winds (>20 m s−1;, which produced extreme DW and a strong current pulse along the continental slope (data archive at the Institute of Ocean Sciences). Note that the mooring sites in the studies by Forest et al. [2007, 2008] were closer to the shelf break than our mooring at AM1-92.

[42] The data in our study do not strongly support wind- and ice-driven UW-DW oscillations as the mechanism responsible for the midwinter enhancement of particle fluxes over the slope at site AM1-92 during 1992–93, although the work by Forest et al. [2007, 2008] demonstrates that it can be a contributing process at times in this area.

5. Conclusions

[43] Candidate mechanisms—namely, transport by cyclonic eddies and resuspension/mobilization of bottom sediment by UW/DW oscillations and by intensification of the Beaufort Shelfbreak Jet—support our contention that cross-shelf transport of sediment in this area is not in steady state, but rather occurs in response to energetic and relatively infrequent oceanographic events. Our study clearly demonstrates that sedimentation on the continental slope is very dynamic even in the dead of winter under heavy ice cover.

[44] The midwinter peak in particle fluxes over the slope during 1992–93 was most likely due to transport and rainout by cyclonic eddies. To our knowledge, this is the first report linking a cyclonic eddy with enhanced particle fluxes under heavy ice cover in the Arctic.

[45] Worthy of further investigation in the vicinity of Mackenzie Trough are 1) the relationship between ice movement and eddy drift; 2) the capacity of eddies generated in this area to entrain and transport suspended material; 3) the frequency and mechanics of eddy generation including the role of topography; and 4) the circumstances under which sediments are resuspended and moved in nepheloid layers on and beyond the shelf.

[46] Projected change in ice and atmospheric climates will likely have some of its greatest effects on processes occurring at the shelf edge [e.g., Carmack and Chapman, 2003]. With respect to having an effect on export of shelf particles into the interior Arctic Ocean, changes in the generation and migration of eddies, the frequency and intensity of occurrence of UW and DW, the frequency of exceeding conditions required to trigger dense bottom water outflows, and the energy content of internal waves all appear important.

[47] The loss of ice over the shelf edge will likely lead to altered patterns of material transport over the slope and into the basins. To understand these aspects of change, therefore, we need to initiate and maintain long-term physical and biogeochemical time series over the shelf, slope and basin paying particular attention to areas of complex topography like Mackenzie Trough.


[48] We acknowledge our collaborators Knut Aagaard (then at Pacific Marine Environmental Lab, Seattle) and Richard E. Moritz (Applied Physics Lab, University of Washington). The project was funded in Canada by the Northern Oil and Gas Action Plan (NOGAP). Special thanks to the Canadian Coast Guard and the Polar Continental Shelf Project for logistical support, and our many colleagues involved in the collection and analysis of these data.