Arctic cod (Boreogadus saida) is the dominant pelagic fish in Arctic seas and a staple food of many arctic predators including several seabird species. Marginal ice zones are known as important feeding locations for seabirds. The hypothesis that thick-billed murre (Uria lomvia), northern fulmar (Fulmarus glacialis) and black-legged kittiwake (Rissa tridactyla) congregate in areas of high Arctic cod food resource and low ice concentration was tested at different spatial scales. Arctic cod biomass was estimated by hydroacoustics as a resource proxy, and seabirds were counted and sampled for stomach analysis along eight longitudinal transects across the marginal ice zone in southern Baffin Bay in June–July 2016. With increasing length, the epipelagic age-0 Arctic cod migrated from open waters to ice-covered areas. Subsequently, age-1 and age-2 Arctic cod tended to concentrate in a subsurface layer (40–100 m) within the epipelagic layer. Arctic cod 5.7–16.1 cm long (late age-0 to age-5) were the main fish prey of the three seabird species, which preferentially captured age-1 cod (6–11.5 cm). At large spatial scale (western versus eastern Baffin Bay), thick-billed murre, northern fulmar and their Arctic cod resource proxy were generally more abundant on the western ice-covered side of Baffin Bay. No clear spatial match was found, however, when comparing seabird abundances and their food-resource proxy in different ice concentrations across the marginal ice zone or at small scale (5 km). At medium scale (12.5 km), only murre density was influenced positively by its Arctic cod resource. A lack of schooling behavior and a successful strategy to avoid predation by hiding under the ice could explain the absence of any strong spatial match between Arctic cod and its seabird predators at these different scales.

Arctic cod (Boreogadus saida) is a small forage fish distributed ubiquitously in Arctic seas (Fortier et al., 2015). Its pivotal role in transferring energy from lower to upper trophic levels is well documented (e.g., Welch et al., 1992; Darnis et al., 2012; Kortsch et al., 2015). Arctic cod occur as dispersed individuals or in different types of aggregation. Dense schools are observed in shallow coastal environments (Welch et al., 1993; Hop et al., 1997). Adult Arctic cod trapped in deep embayments by circulation can form large and dense winter aggregations at depth (Benoit et al., 2008). Offshore, Arctic cod are segregated vertically by size and age throughout the year: age-0 larvae and juveniles form an epipelagic scattering layer (<100 m), while fish of increasingly large size with depth form a mesopelagic scattering layer in the ~100–500-m depth interval (Falk-Petersen et al., 1986; Geoffroy et al., 2016). Epipelagic Arctic cod are mainly age-0 individuals <5 cm (Geoffroy et al., 2016), but some age-1 and age-2 Arctic cod are associated with sea ice during the summer ice melt (Gradinger and Bluhm, 2004; David et al., 2016). As well, swarms dominated by age-2 Arctic cod have been observed at the ice-water interface in the central Arctic Ocean in winter (Melnikov and Chernova, 2013).

In Arctic seas, many predators including seabirds feed primarily on Arctic cod (Bradstreet and Cross, 1982; Welch et al., 1992). Surface feeders such as the black-legged kittiwake (Rissa tridactyla) and the northern fulmar (Fulmarus glacialis) catch their prey mainly at the very surface, but they may also dive within the first 1 or 3 meters of the surface, respectively (Hobson and Welch, 1992; Garthe and Furness, 2001; Matley et al., 2012a). Deep-diving species such as the thick-billed murre (Uria lomvia) regularly dive to depths >100 m to feed, and occasionally below 200 m (Gaston and Hipfner, 2000; Elliott and Gaston, 2014). The influence of fish availability on arctic seabird distribution has been documented mostly for coastal or ice-free regions (e.g., Logerwell et al., 1998; Mehlum et al., 1999; Matley et al., 2012a; Suryan et al., 2016). Offshore ice edges and marginal ice zones (where the limit between pack ice and open water is more a gradient than a sharp edge; Smith, 1987) are productive areas (e.g., Mundy et al., 2009) that are often favored feeding locations for seabirds where they feed in openings on invertebrates and fish (Bradstreet, 1982). On the other hand, extensive ice cover can impede foraging (Hunt, 1990).

The degree of association among species typically varies with the spatial scale of observation, and a multiscale approach may help capture the spatial scales at which species are most likely to interact (Schneider and Piatt, 1986; Logerwell and Hargreaves, 1996; Logerwell et al., 1998; Mehlum et al., 1999). In the present study, we tested the hypothesis that seabirds congregate in areas of the marginal ice zone where fish biomass is high and/or ice concentration is low. The stomach contents of thick-billed murre, northern fulmar and black-legged kittiwake were analyzed to identify the fish prey of each species. The co-distribution of seabirds and their potential fish resource was compared at three spatial scales (west vs east Baffin Bay, 12.5 km and 5 km) as well as in relation to percent ice cover across the marginal ice zone of Baffin Bay.

### Study area

Baffin Bay is a large (689,000 km2) seasonally ice-covered sea with relatively narrow continental shelves surrounding an abyssal plain as deep as 2300 m. It is connected to the Arctic Ocean by the narrow and shallow Nares Strait, and to the Labrador Sea and the Atlantic Ocean by Davis Strait. The generally cyclonic circulation carries Atlantic waters northward along Greenland and Arctic waters southward along Baffin Island (e.g., Tang et al., 2004). Except for the North Water polynya and the southeastern part, Baffin Bay is covered with ice for most of the year and completely ice-free in August and September. Typically, ice breakup in April–May progresses from south to north along Greenland, resulting by July in a S–N oriented marginal ice zone with open water to the east and pack ice to the west (e.g., Tang et al., 2004, their Figure 5). As part of the international Green Edge expedition, the research icebreaker CCGS Amundsen conducted an ecosystem survey from 9 June to 12 July 2016 along eight longitudinal transects across the marginal ice zone of southern Baffin Bay (Figure 1). Depth along the transects varied from 144 to 2100 m (average of 928 m). The midnight sun prevailed during the entire duration of the survey.

Figure 1

Sampling locations and transects surveyed. Deployment locations of epipelagic net (red circles), bottom trawl (pink circles), simultaneous epipelagic net and pelagic trawl (blue circles), simultaneous epipelagic net and bottom trawl (yellow circles), and locations of seabird sampling (green rectangles). The CCGS Amundsen route (black line), and simultaneous hydroacoustic and seabird sighting transects (teal line) are presented. The order of the completed transects is indicated by 1–8. The background gradient (white–gray scale) represents the mean sea-ice concentration from 9 June to 12 July 2016. Coastal areas with no available ice data are indicated in black. The dashed box shows the domain of Figure 6. DOI: https://doi.org/10.1525/elementa.339.f1

Figure 1

Sampling locations and transects surveyed. Deployment locations of epipelagic net (red circles), bottom trawl (pink circles), simultaneous epipelagic net and pelagic trawl (blue circles), simultaneous epipelagic net and bottom trawl (yellow circles), and locations of seabird sampling (green rectangles). The CCGS Amundsen route (black line), and simultaneous hydroacoustic and seabird sighting transects (teal line) are presented. The order of the completed transects is indicated by 1–8. The background gradient (white–gray scale) represents the mean sea-ice concentration from 9 June to 12 July 2016. Coastal areas with no available ice data are indicated in black. The dashed box shows the domain of Figure 6. DOI: https://doi.org/10.1525/elementa.339.f1

### Sea-ice cover

Satellite-based estimates of daily sea-ice concentration (% cover) were provided by IFREMER/CERSAT at a pixel resolution of 12.5 × 12.5 km (ftp.ifremer.fr/ifremer/cersat). Each 12.5-km segment along the survey transects was assigned the sea-ice concentration of the nearest most overlapping pixel on the sampling day. No sea-ice cover data were available at a resolution scale of 5 km; values were assigned from the 12.5-km resolution data following the same method.

### Fish sampling and acoustic detection

The abundance and taxonomy of fish along the survey transects were assessed by acoustics validated by deploying a combination of ichthyoplankton sampler and trawls during both day and night. To document the epipelagic fish assemblage, a Double Square Net bearing two square conical nets (1-m2 aperture, 500-μm and 750-μm meshes, respectively) was deployed from the surface to depths of 75 to 103 m at 17 locations (Figure 1). The standard length (SL) of up to 50 Arctic cod was measured fresh after each net retrieval. All epipelagic fish were preserved in 95% ethanol. Pelagic fish were sampled with an Isaacs-Kidd Midwater Trawl (IKMT, 9-m2 rectangular mouth aperture with mesh sizes of 11 mm in the first section and 5 mm in the last section), deployed from the surface to depths of 244 to 520 m on six occasions (Figure 1). Demersal and benthic fish were sampled with a bottom beam trawl (3-m2 rectangular aperture with a headline and a footrope of 4.27 m and a cod-end mesh of 9.5 mm) deployed at seafloor depths of 199 to 359 m at five locations (Figure 1). The pelagic and bottom trawls were towed at 2–3 knots for 10 to 20 min. All fish collected by the trawls were measured before freezing at –20°C.

Fish density was estimated with a Simrad EK60® split-beam echosounder following the recording and processing methods of Bouchard et al. (2017), except that the acoustic signal was processed with background and impulsive noise removal operators (De Robertis and Higginbottom, 2007; Ryan, et al. 2015) instead of a time-varied threshold. Due to the depth of the transducers on the vessel hull (~7 m), potential vessel avoidance by fish, and acoustic near-field effects (Simmonds and MacLennan, 2005), effective sampling by the echosounder was assumed to start at 12 m; the top 12 m of the water column were excluded from the analysis. The difference in mean volume backscattering strength ΔMVBS (dB re: 1 m–1) between 38 and 120 kHz was used to discriminate pelagic fish and dense aggregations of large zooplankton.

In the present study, Arctic cod and 5.0–7.6 cm long Benthosema glaciale (mean SL of 6.3 cm) were the only two pelagic species caught in nets with a swim bladder and likely detected at 38 kHz. Arctic cod represented 96% of the total number of the two species captured in the three nets (epipelagic net, pelagic trawl, and bottom trawl). We therefore assumed that echo-integration cells with ΔMVBS120–38 in the range –10 dB to 5 dB were dominated by Arctic cod (Benoit et al., 2014; Geoffroy et al., 2016), mixed occasionally with small numbers of B. glaciale.

The mean fish target strength (TS, dB re: 1 m2) of each of the two species was estimated from their respective mean SL in net/trawl collections, using TS = 14.33 log10(SL)–65.13 for Arctic cod (Geoffroy et al., 2016), and TS = 30 log10(SL)–84.5 for B. glaciale (based on Scoulding et al., 2015). Mean TSs were converted to mean backscattering cross-sections (σb = 10TS/10, m2). The total area backscattering coefficient (ABC, m2 m–2), integrated over the depth intervals of 12–20 m and 12–200 m, was partitioned to each species based on their relative frequencies in nets/trawls and mean σb (Simmonds and MacLennan, 2005). The overall density of Arctic cod (individual m–2) in each depth interval was estimated by dividing its partitioned ABC for that interval by σb (Simmonds and MacLennan, 2005; Parker-Stetter et al., 2009). The overall density of Arctic cod was used to calculate the food resource of the three seabird species (see below).

A proxy for availability of Arctic cod to surface-feeding northern fulmar and black-legged kittiwake was estimated in echo-integration cells corresponding to the depth interval of 12–20 m by 5-km transect segments (small scale analysis, n = 216 cells) and 12.5-km segments (medium scale analysis, n = 104 cells), respectively. The corresponding echo-integration cells for the deep-diving thick-billed murre spanned the depth interval of 12–200 m. A large-scale analysis was also conducted by contrasting western and eastern Baffin Bay, using longitude 60°W as limit between the two regions.

### Estimated in situ length-frequency distribution of Arctic cod

The length-frequency distribution of Arctic cod in situ was estimated from TS of targets detected at 38 kHz as determined by the Echoview® single-echo detection algorithm for split-beam echosounder (method 2: threshold applied on targets after compensation for their position in the beam). Targets detected in the depth intervals of 12–20 m and 12–200 m were assumed to be of Arctic cod, given the overall low frequency of B. glaciale (4%) in nets/trawls and its tendency to remain at depths ≥200 m in daylight (Kaartvedt et al., 2009). To eliminate multiple echoes detection, cells 20 pings long and 20 m deep with a Sawada Index >0.04 were removed (Gauthier and Rose, 2001). A lower threshold of –70 dB and an upper threshold of –43 dB were set to exclude zooplankton and targets bigger than the largest Arctic cod (Geoffroy et al., 2016). The frequency distribution of SL (based on TS of targets detected) was determined for each echo-integration cell defined for the surface-feeding northern fulmar, black-legged kittiwake (12–20 m by 5 or 12.5 km) and for the deep-diving thick-billed murre (12–200 m by 5 or 12.5 km). When less than 10 targets were detected in a given cell, the length-frequency distribution for that cell was estimated by pooling TS of targets for that cell and the two adjacent cells along the transect.

### Seabird observation, sampling and stomach content analysis

Seabirds were identified and counted from the bridge of the icebreaker (16 m above sea level) during transit between sampling stations (Figure 1) at speeds ranging from 13 to 18.5 km h–1. The same observer enumerated all birds continuously on one side within a 300-m band perpendicular to the track of the ship using binoculars and a computer-based logger, grouping observations over two-minute blocks following Johansen et al. (2015). The time, coordinates, distance from ship and behaviour of birds (flying or resting on water) were noted for each observation. The analysis was focused on foraging birds that could potentially be influenced by their prey. Therefore, all birds sighted were included (flying or on water) for northern fulmar and black-legged kittiwake as they may dive directly from the air to feed (Matley et al., 2012b), while only thick-billed murre on water were included. Birds suspected to be ship followers (following or circling around the ship) were excluded from the analysis as well. Seabird numbers were calculated for 5-km and 12.5-km visual transects corresponding to the acoustic echo-integration cells and were converted into abundances (individual [ind.] km–1) for fulmar and kittiwake. Density (ind. km–2) was calculated for murre using the distance sampling method (Buckland et al., 2001), as all birds included in the analysis were on water allowing for accurate distance from ship estimation. Overall, 2013 2-minute observation periods, nearly uniformly distributed over 104 12.5-km segments (216 5-km segments), were recorded.

Seabirds were sampled from a zodiac launch deployed from the Amundsen on ten occasions in Greenland waters (Figure 1). Adult individuals were shot in flight or on water by a skilled seabird hunter (with appropriate hunting permit delivered by the Ministry of Fisheries, Hunting and Agriculture of Greenland). In total, 35 thick-billed murres, 30 northern fulmars and 9 black-legged kittiwakes were collected (Table 1). Collected birds were frozen at –20 °C. In the laboratory, the whole stomach (proventriculi and gizzard) was removed. Stomach contents were examined under a dissecting microscope, and fish otoliths were retrieved. Otoliths retained in the stomach were assumed to be sagittae, were identified to the lowest taxonomic level possible (following Härkönen, 1986; Campana, 2004), and their number was divided by two to estimate the number of fish ingested. Total length (TL, mm) of Arctic cod prey was estimated from otolith length (OL, mm) based on TL = 20.6631 + 24.2791 OL, and then transformed into SL (cm) using SL = (–0.2736 + 0.9187 TL)/10 (David et al., 2016). The age of individual Arctic cod prey was estimated with the age-length relationships reported in Falk-Petersen et al. (1986, their Figure 9).

Table 1

Relative frequency (%) of fish species ingested by seabirds and sampled by the epipelagic net, pelagic trawl and bottom trawl. DOI: https://doi.org/10.1525/elementa.339.t1

Fish speciesNorthern fulmarThick-billed murreBlack-legged kittiwakeEpipelagic netPelagic trawlBottom trawl

Boreogadus saida (Arctic cod) 59.5 94.4 96.4 97.1 11.8 26.6
Benthosema glaciale 14.3 0.0 3.6 0.0 64.7 0.0
Liparis gibbus 8.3 0.0 0.0 0.0 0.0 0.0
Lampanyctus macdonaldi 6.0 0.0 0.0 0.0 0.0 0.0
Liparis atlanticus 4.8 3.3 0.0 0.0 0.0 0.0
Liparis fabricii 2.4 1.1 0.0 1.9 11.8 0.0
Gymnocanthus tricuspis 2.4 0.0 0.0 0.0 0.0 0.0
Ammodytes sp. 1.2 0.0 0.0 0.0 0.0 0.0
Careproctus reinhardti 1.2 0.0 0.0 0.0 0.0 0.0
Triglops sp. 0.0 1.1 0.0 0.1 0.0 59.0
Othersa 0.0 0.0 0.0 0.9 11.8 14.4
Related data

Shannon’s indexb 0.6 0.1 0.1 0.1 0.6 0.5
Number of birds, nets or trawls sampled 30 35 17
Birds with non-empty stomach 29 25 – – –
Birds with otoliths in stomach 18 21 – – –
Mean number of fishc 2.9 2.7 3.1 50.4 11.3 96.2
Fish speciesNorthern fulmarThick-billed murreBlack-legged kittiwakeEpipelagic netPelagic trawlBottom trawl

Boreogadus saida (Arctic cod) 59.5 94.4 96.4 97.1 11.8 26.6
Benthosema glaciale 14.3 0.0 3.6 0.0 64.7 0.0
Liparis gibbus 8.3 0.0 0.0 0.0 0.0 0.0
Lampanyctus macdonaldi 6.0 0.0 0.0 0.0 0.0 0.0
Liparis atlanticus 4.8 3.3 0.0 0.0 0.0 0.0
Liparis fabricii 2.4 1.1 0.0 1.9 11.8 0.0
Gymnocanthus tricuspis 2.4 0.0 0.0 0.0 0.0 0.0
Ammodytes sp. 1.2 0.0 0.0 0.0 0.0 0.0
Careproctus reinhardti 1.2 0.0 0.0 0.0 0.0 0.0
Triglops sp. 0.0 1.1 0.0 0.1 0.0 59.0
Othersa 0.0 0.0 0.0 0.9 11.8 14.4
Related data

Shannon’s indexb 0.6 0.1 0.1 0.1 0.6 0.5
Number of birds, nets or trawls sampled 30 35 17
Birds with non-empty stomach 29 25 – – –
Birds with otoliths in stomach 18 21 – – –
Mean number of fishc 2.9 2.7 3.1 50.4 11.3 96.2

a Category “Others” includes Anarhichas lupus, Anarhichas minor, Artdiellus atlanticus, Gymnelius viridis, Hygophum sp., Hypoglossoides platessoides, Leptoclinus maculatus, Leucoraja ocellate, Lumpenus fabricii, Lycenchelys paxillus, Lycodes terraenovae, Nezumia bairdii, Reinhardtius hippoglossoides and Stomias boa.

b Measure of diversity of fish ingested or sampled.

c Mean number of fish in stomach (for all birds of a given species, with or without stomach content) and in net samples.

### Estimating the Arctic cod resource available to seabirds

The Arctic cod food resource (mg m–2) available to each seabird species in each echo-integration cell was calculated based on the overall density and length-frequency distribution of Arctic cod in situ in the size range of Arctic cod ingested by seabirds (4–18 cm). For the thick-billed murre, the food resource was calculated in the depth interval of 12–200 m corresponding to the maximum diving range; for the black-legged kittiwake and the northern fulmar, the resource was calculated in the depth interval of 12–20 m as a proxy for food availability in their maximum diving range of 1 and 3 m, respectively. The average weight (W in mg) of Arctic cod in each 2-cm SL interval of the frequency distributions was calculated by integrating the relationship W = (0.0055 SL3.19)*1000 (Geoffroy et al., 2016) over the length interval. Based on Fortier and Harris (1989) and Ponton and Fortier (1992), the Arctic cod resource (R) available for a given seabird species was estimated as

${\rm{R}} = \sum\nolimits_{{\rm{i}} = 1}^{\rm{n}} {{{\rm{W}}_{\rm{i}}}{{\rm{D}}_{\rm{i}}}}$
⁠, where Wi is the average weight of Arctic cod in SL interval i, Di is the density of SL interval i in the echo-integration cell (product of overall density by frequency of SL interval i) and n is the number of 2-cm length intervals in the frequency distributions (from 4 to 18 cm). Following the same method, the biomass of Arctic cod <4 cm was also estimated in each cell, hereafter referred to as small Arctic cod biomass.

Seabird abundance (fulmar and kittiwake) and density (murre) were then compared to their available Arctic cod food-resource proxy (4–18 cm length) and to small Arctic cod biomass (<4 cm length) at different scales (West-East Baffin Bay, 12.5 km, 5 km) along the survey transects. Seabird abundance/density was non-normal due to frequent 0-values and was rounded for statistical analyses. The influence of Arctic cod resource, small Arctic cod biomass and sea-ice concentration on bird distribution at medium (12.5 km) and small (5 km) scales was tested with zero-inflated negative binomial regressions. To test spatial differences at large scale (West-East Baffin Bay and in different ice concentrations across the marginal ice zone), non-parametric tests (Mann-Whitney U and Kruskal-Wallis tests) were used for Arctic cod food resource and small Arctic cod biomass, whereas negative binomial regressions were used for seabird abundance/density. All statistical analyses were performed with R® version 3.2.3 (R Core Team, 2015).

### Seabird fish prey and in situ fish assemblage

Northern fulmar (n = 2242), thick-billed murre (n = 2028) and black-legged kittiwake (n = 278), the three most abundant piscivorous seabirds observed along the survey transects, were included in the analysis. Other relatively frequent bird species (n > 30) were the glaucous gull (Larus hyperboreus), the little auk (Alle alle), the black guillemot (Cepphus grille) and the common eider (Somateria mollissima).

A majority of the northern fulmar, thick-billed murre and black-legged kittiwake sampled at sea presented a stomach content and most contained fish otoliths (Table 1). The mean number of fish ingested per bird was similar for the three species. About two thirds of the fulmars (11/18), murres (13/21) and kittiwakes (5/8) with otoliths in their stomachs also had other discernable fish remnants (e.g., flesh, eyes, bones) indicating that they had been feeding on fish recently (in the last 6 to 24h; Hawkins et al., 1997; Hilton et al., 2000). The low-diversity fish diet of murre and kittiwake (Shannon index = 0.1) was dominated by Arctic cod (>94%), and similar to the fish assemblage in the epipelagic net collections. Northern fulmar also preyed primarily on Arctic cod (~60%) but presented a much more diversified fish diet (Shannon index = 0.6), feeding on no less than 9 fish species including four Liparidae (Liparis atlanticus, L. fabricii, L. gibbus and Careproctus reinhardti). Both surface-feeding fulmar and kittiwake preyed on B. glaciale, a bathypelagic species that dominated collections in the deep pelagic trawl (Table 1). The fulmar also captured Lampanyctus macdonaldi, another bathypelagic fish.

The estimated mean SL of Arctic cod prey differed significantly (one-way ANOVA, p < 0.001) among the three seabird species (Figure 2). Overall, Arctic cod prey ranged from 5.7 to 16.1 cm in SL (mean = 9.16 cm, n = 150) corresponding to age-0 to age-5 fish (mostly age-1). By comparison to seabirds, the epipelagic net captured essentially small 0.6–1.4 cm (with one 4.9 cm exception) age-0 Arctic cod (Figure 2). The pelagic trawl collected larger age-0 and age-1 Arctic cod 5.2 to 6.8 cm, while Arctic cod in the bottom trawl ranged in SL from 5.7 to 17.2 cm, and from age-0 to age-5.

Figure 2

Standard-length frequency distribution of Arctic cod ingested by seabirds and sampled by the nets/trawls. Standard length (SL) of Arctic cod ingested by seabirds was estimated from otolith length. The age frequency in each length interval is provided. DOI: https://doi.org/10.1525/elementa.339.f2

Figure 2

Standard-length frequency distribution of Arctic cod ingested by seabirds and sampled by the nets/trawls. Standard length (SL) of Arctic cod ingested by seabirds was estimated from otolith length. The age frequency in each length interval is provided. DOI: https://doi.org/10.1525/elementa.339.f2

### The distribution of Arctic cod across the marginal ice zone

Coriolis forcing maintained the southward-drifting ice cover on the western side of Baffin Bay, which resulted in a well-defined marginal ice zone (Figure 1). On average over the study period (early June to mid-July), 100% ice cover prevailed on the western Canadian side, thinning progressively to fully open waters on the eastern Greenland side. Despite this general pattern, ice concentration along the survey transects could be highly variable at smaller spatio-temporal scales.

Vertically, Arctic cod distributed in two distinct continuous scattering layers: an epipelagic layer down to about 50 m and a mesopelagic layer above the bottom or between 200 and 300 m in deeper areas (Figure 3). Both scattering layers varied in thickness and density among regions with different ice cover. However, in a given location, scattering density varied little horizontally, and no discrete schools were observed in the scattering layers. Some filament-shaped horizontal aggregations were observed between the two main scattering layers in echograms recorded at nighttime and early morning (Figure 3a, b, e).

Figure 3

Examples of fish distribution in different ice concentrations. Acoustic data (Sv) measured at 38 kHz in the depth interval of 12–450 m over a distance of 2 km. Two main scattering layers of fish are present: an epipelagic (red arrow) and a mesopelagic (pink arrow) layer. The hatched part of the echograms represents the top 12 m excluded from the analysis. Areas below the bottom or considered as noise are in black. For each panel (a–e), ice concentration (% ice cover) is followed by local time at the middle of the transect. DOI: https://doi.org/10.1525/elementa.339.f3

Figure 3

Examples of fish distribution in different ice concentrations. Acoustic data (Sv) measured at 38 kHz in the depth interval of 12–450 m over a distance of 2 km. Two main scattering layers of fish are present: an epipelagic (red arrow) and a mesopelagic (pink arrow) layer. The hatched part of the echograms represents the top 12 m excluded from the analysis. Areas below the bottom or considered as noise are in black. For each panel (a–e), ice concentration (% ice cover) is followed by local time at the middle of the transect. DOI: https://doi.org/10.1525/elementa.339.f3

The length of Arctic cod estimated from TS of targets detected at 38 kHz increased with sea-ice concentration (Figure 4). In the depth layer of 12–20 m, the length-frequency distribution of age-0 Arctic cod <5 cm generally shifted to larger sizes with increasing % ice cover. However, the relative frequency of age 1+ Arctic cod (fish > 5 cm), the main fish prey of seabirds, changed little with sea-ice concentration. By contrast, in the depth layer of 12–200 m where the deep-diving thick-billed murre forages, the relative frequency of age-1+ Arctic cod prey increased with sea-ice cover to a maximum at ice concentrations between 40 and 100% (Figure 4).

Figure 4

Length-frequency distribution of Arctic cod detected in different ice concentrations. Length distribution in the depth intervals of 12–20 m and 12–200 m on a log10-based scale measured as target strength (TS, top x-axis) of targets detected at 38 kHz. The bottom x-axis provides the estimated standard length (SL) of Arctic cod based on TS (TS = 14.33 log10 [SL]–65.13; Geoffroy et al., 2016). Values for n are the number of targets detected. The size range of the Arctic cod ingested by northern fulmar (NF), black-legged kittiwake (BLK) and thick-billed murre (TBM) is presented. DOI: https://doi.org/10.1525/elementa.339.f4

Figure 4

Length-frequency distribution of Arctic cod detected in different ice concentrations. Length distribution in the depth intervals of 12–20 m and 12–200 m on a log10-based scale measured as target strength (TS, top x-axis) of targets detected at 38 kHz. The bottom x-axis provides the estimated standard length (SL) of Arctic cod based on TS (TS = 14.33 log10 [SL]–65.13; Geoffroy et al., 2016). Values for n are the number of targets detected. The size range of the Arctic cod ingested by northern fulmar (NF), black-legged kittiwake (BLK) and thick-billed murre (TBM) is presented. DOI: https://doi.org/10.1525/elementa.339.f4

The length of Arctic cod estimated from TS of targets detected at 38 kHz generally increased with depth over the interval of 12–200 m (Figure 5). Small age-0 Arctic cod dominated the population based on detections. Some age-1 and age-2 Arctic cod, the preferred prey of seabirds, were detected at nearly all depths under each ice-cover concentration. Moreover, relatively high SL were concentrated in a subsurface layer (40–100 m) for all ice concentration classes, indicating an increased frequency of larger Arctic cod even though small age-0 were preponderant. This subsurface layer of larger fish tended to thicken with ice concentration, spanning the depth interval of 40–80 m at 0–40% ice cover to that of 20–120 m at 40–100% ice cover.

Figure 5

Box and whisker plots of the length of Arctic cod by 20-m deep intervals in different ice concentrations. Box and whisker plots of the length on a log10-based scale measured as target strength (TS, top x-axis) of targets detected at 38 kHz. The black vertical line in the box is the median, the left and right boundaries of the box are the 25 and 75 percentiles, and the whiskers are the minimum and maximum values. The bottom x-axis provides the estimated standard length (SL) of Arctic cod based on TS (TS = 14.33 log10 [SL]–65.13; Geoffroy et al., 2016). Values for n are the number of targets detected. Size range corresponding to age-0, age-1 and age-2 are presented. DOI: https://doi.org/10.1525/elementa.339.f5

Figure 5

Box and whisker plots of the length of Arctic cod by 20-m deep intervals in different ice concentrations. Box and whisker plots of the length on a log10-based scale measured as target strength (TS, top x-axis) of targets detected at 38 kHz. The black vertical line in the box is the median, the left and right boundaries of the box are the 25 and 75 percentiles, and the whiskers are the minimum and maximum values. The bottom x-axis provides the estimated standard length (SL) of Arctic cod based on TS (TS = 14.33 log10 [SL]–65.13; Geoffroy et al., 2016). Values for n are the number of targets detected. Size range corresponding to age-0, age-1 and age-2 are presented. DOI: https://doi.org/10.1525/elementa.339.f5

### The co-distribution of Arctic cod and its seabird predators across the marginal ice zone

The three seabirds, their Arctic cod resource proxy and small Arctic cod were widely distributed across the sea-ice gradient of the marginal ice zone (Figure 6). The Arctic cod resource of the deep-diving thick-billed murre and small Arctic cod in the depth layer of 12–200 m were more abundant (p < 0.001, Mann–Whitney U test) under the dense ice cover on the western side of Baffin Bay than in the open waters to the east (Figure 6b, c; Table 2). The density of murre was concentrated in southern transects 4 and 8 and roughly followed this longitudinal gradient in food availability along transect 4 (Figure 6a–c). But both low and high concentrations of murre were observed in the resource-rich ice-covered western waters and in the resource-poor open eastern waters. Overall, murre were slightly more abundant in the resource-rich waters on the western side of the Bay (Table 2), although the difference was not statistically significant (p > 0.05, Negative binomial regression).

Figure 6

Spatial distribution of seabirds and Arctic cod across the marginal ice zone. Distribution of thick-billed murre (individuals [ind.] km–2), northern fulmar (ind. km–1) and black-legged kittiwake (ind. km–1), their Arctic cod food-resource proxy (mg m–2) and small Arctic cod (AC) biomass (mg m–2) in southern Baffin Bay from 9 June to 12 July 2016 at a resolution of 5 km. Black dots are values of 0. The background gradient (white–gray scale) represents the mean sea-ice concentration (% coverage) during the study period. DOI: https://doi.org/10.1525/elementa.339.f6

Figure 6

Spatial distribution of seabirds and Arctic cod across the marginal ice zone. Distribution of thick-billed murre (individuals [ind.] km–2), northern fulmar (ind. km–1) and black-legged kittiwake (ind. km–1), their Arctic cod food-resource proxy (mg m–2) and small Arctic cod (AC) biomass (mg m–2) in southern Baffin Bay from 9 June to 12 July 2016 at a resolution of 5 km. Black dots are values of 0. The background gradient (white–gray scale) represents the mean sea-ice concentration (% coverage) during the study period. DOI: https://doi.org/10.1525/elementa.339.f6

Table 2

Mean thick-billed murre density, northern fulmar and black-legged kittiwake abundances, their Arctic cod food-resource proxy and small Arctic cod (AC) biomass in two regions: western and eastern Baffin Bay.a DOI: https://doi.org/10.1525/elementa.339.t2

RegionbThick-billed murre (ind. km–2)Depth layer 12–200 mNorthern fulmar (ind. km–1)Black-legged kittiwake (ind. km–1)Depth layer 12–20 m

Food resource (mg m–2)Small AC (mg m–2)Food resource (mg m–2)Small AC (mg m–2)

Western 2.62 113.013 c 10.176 c 0.74 0.07 8.687 c 1.778 c
Eastern 1.26 21.694  2.418  0.33 0.14 3.274  1.038
RegionbThick-billed murre (ind. km–2)Depth layer 12–200 mNorthern fulmar (ind. km–1)Black-legged kittiwake (ind. km–1)Depth layer 12–20 m

Food resource (mg m–2)Small AC (mg m–2)Food resource (mg m–2)Small AC (mg m–2)

Western 2.62 113.013 c 10.176 c 0.74 0.07 8.687 c 1.778 c
Eastern 1.26 21.694  2.418  0.33 0.14 3.274  1.038

a Differences were not significant for seabird abundance/density (p > 0.05, Negative binomial regression).

b Ice concentration was significantly higher on western side than eastern side (p < 0.001, Negative binomial regression).

c Significantly higher (p < 0.05, Mann–Whitney U test).

As expected, less Arctic cod food resource seemed to be available to the surface-feeding northern fulmar and black-legged kittiwake, using the depth layer of 12–20 m as proxy, than to the deep-diving thick-billed murre in the depth layer of 12–200 m (Figure 6; Table 2). The resource proxy of both fulmar and kittiwake and small Arctic cod in the depth layer of 12–20 m were patchily distributed, with high values occurring both in the eastern open waters (e.g., transects 1 and 3) and under the western dense ice cover (transects 6, 7, and 8), with generally lower values in the intermediate ice gradient (Figure 6e, f). At the scale of the survey area, no spatial match between the distribution of fulmar or kittiwake and that of their potential food was obvious (Figure 6d–g). However, northern fulmar was marginally (p > 0.05, Negative binomial regression) more abundant on the western side of the Bay where its Arctic cod food-resource proxy and small Arctic cod were more abundant (p < 0.05, Mann–Whitney U test; Table 2). In contrast, black-legged kittiwake was more abundant on the eastern side (Table 2), although the difference was not statistically significant (p > 0.05, Negative binomial regression).

When grouped according to ice coverage, no clear pattern was found between the abundance/density of seabirds and their Arctic cod food across the marginal ice zone (Figure 7). On average, and despite a highest value in ice concentration 20–40% (p = 0.005, Kruskal-Wallis test), the food resource of thick-billed murre was distributed relatively uniformly across the marginal ice zone (Figure 7a). Small Arctic cod biomass (<4 cm length) in the depth interval of 12–200 m was much lower than the food resource in all ice concentrations and was significantly higher in 80–100% ice cover (p < 0.001, Kruskal-Wallis test; Figure 7a). Total Arctic cod biomass (≤18 cm length) in the depth interval of 12–200 m was mainly dominated by large age-0 and age-1 Arctic cod 4–8 cm across the different ice concentrations (Figure 8). Murre density increased from low–medium (0–60%) to high ice concentration (60–100%), and highest bird density was measured in 80–100% ice cover (p < 0.001, Negative binomial regression; Figure 7a).

Figure 7

Seabirds density/abundance and Arctic cod biomass in different ice concentrations. Mean thick-billed murre density (a; ind. km–2), northern fulmar (b; ind. km–1) and black-legged kittiwake (c; ind. km–1) abundances, their Arctic cod food-resource proxy (mg m–2) and small Arctic cod (AC) biomass (mg m–2) in different ice concentrations. Circles represent abundance/density of seabirds and bars represent Arctic cod food resource and small Arctic cod biomass. Asterisks (a) indicate ice concentration class with significant highest value (p < 0.05) of Arctic cod food resource or small Arctic cod biomass (Kruskal-Wallis test) and of murre density (Negative binomial regression). DOI: https://doi.org/10.1525/elementa.339.f7

Figure 7

Seabirds density/abundance and Arctic cod biomass in different ice concentrations. Mean thick-billed murre density (a; ind. km–2), northern fulmar (b; ind. km–1) and black-legged kittiwake (c; ind. km–1) abundances, their Arctic cod food-resource proxy (mg m–2) and small Arctic cod (AC) biomass (mg m–2) in different ice concentrations. Circles represent abundance/density of seabirds and bars represent Arctic cod food resource and small Arctic cod biomass. Asterisks (a) indicate ice concentration class with significant highest value (p < 0.05) of Arctic cod food resource or small Arctic cod biomass (Kruskal-Wallis test) and of murre density (Negative binomial regression). DOI: https://doi.org/10.1525/elementa.339.f7

Figure 8

Frequency of total Arctic cod biomass for each length interval in different ice concentrations. Biomass frequency (%) of each length interval (≤18 cm length) of Arctic cod was calculated for the depth intervals of 12–20 m and 12–200 m under ice conditions of 0–20, 20–40, 40–60, 60–80, and 80–100% ice cover. DOI: https://doi.org/10.1525/elementa.339.f8

Figure 8

Frequency of total Arctic cod biomass for each length interval in different ice concentrations. Biomass frequency (%) of each length interval (≤18 cm length) of Arctic cod was calculated for the depth intervals of 12–20 m and 12–200 m under ice conditions of 0–20, 20–40, 40–60, 60–80, and 80–100% ice cover. DOI: https://doi.org/10.1525/elementa.339.f8

The Arctic cod food-resource proxy of surface feeders was higher in low ice concentration (0–20%), although not statistically significant (p > 0.05, Kruskal-Wallis test; Figure 7b, c). Small Arctic cod biomass level in the depth layer of 12–20 m was similar to the food-resource proxy of seabirds (except in the 0–20% ice cover) and generally increased with ice concentration (Figure 7b, c). No length interval clearly dominated the total Arctic cod biomass (≤18 cm length) in the depth interval of 12–20 m in low–medium ice concentration (0–60% ice cover) whereas 2–4 cm Arctic cod (age-0) were predominant in high (60–100%) ice concentration (Figure 8). Both northern fulmar (20–40%) and black-legged kittiwake (0–40%) preferred low to medium ice cover (Figure 7b, c), but the differences were not statistically significant (p > 0.05, Negative binomial regression).

### Medium-scale spatial association

The Arctic cod food resource in the depth interval of 12–200 m had a significant positive effect (p = 0.03, Zero-inflated negative binomial regression) on thick-billed murre density at a medium scale of 12.5 km, but this food-resource proxy in the depth interval of 12–20 m had no significant effect (p > 0.05) on northern fulmar and black-legged kittiwake abundances (Negative binomial part of the regression models, Table 3). Small Arctic cod biomass in the depth interval of 12–20 m had a significant negative effect (p = 0.02) on kittiwake abundance (Negative binomial part of the regression models, Table 3). There was no significant effect of ice concentration on murre density or fulmar and kittiwake abundances (p > 0.05).

Table 3

Zero-inflated negative binomial regression coefficients of the effect of Arctic cod food resource, small Arctic cod (AC) biomass and ice concentration on thick-billed murre density, and northern fulmar and black-legged kittiwake abundances at medium scale of 12.5 km. DOI: https://doi.org/10.1525/elementa.339.t3

SpeciesParameterNegative binomial partZero-inflated part

EstimateZ-valueEstimateZ-value

Thick-billed murre Food resource (12–200 m) 0.019 a 2.171 0.264 0.188
Small AC (12–200 m) 0.053  0.786 0.061 0.017
Ice concentration 0.018  1.750 –0.052 –0.027
Northern fulmar Food resource (12–20 m) 0.031  0.584 –0.140 –0.902
Small AC (12–20 m) –0.351  –1.470 –0.502 –0.730
Ice concentration –0.001  –0.061 0.011 0.559
Black-legged kittiwake Food resource (12–20 m) 0.033  0.924 2.632 0.008
Small AC (12–20 m) –0.916 a –2.334 –35.619 –0.055
Ice concentration 0.003  0.219 1.178 0.011
SpeciesParameterNegative binomial partZero-inflated part

EstimateZ-valueEstimateZ-value

Thick-billed murre Food resource (12–200 m) 0.019 a 2.171 0.264 0.188
Small AC (12–200 m) 0.053  0.786 0.061 0.017
Ice concentration 0.018  1.750 –0.052 –0.027
Northern fulmar Food resource (12–20 m) 0.031  0.584 –0.140 –0.902
Small AC (12–20 m) –0.351  –1.470 –0.502 –0.730
Ice concentration –0.001  –0.061 0.011 0.559
Black-legged kittiwake Food resource (12–20 m) 0.033  0.924 2.632 0.008
Small AC (12–20 m) –0.916 a –2.334 –35.619 –0.055
Ice concentration 0.003  0.219 1.178 0.011

a Significant effect (p < 0.05).

Excess in absence of the three seabird species (0-values) at medium scale (12.5 km) was not influenced significantly (p > 0.05, Zero-inflated negative binomial regression) by Arctic cod food resource, small Arctic cod biomass or ice concentration (Zero-inflated part of the regression models, Table 3).

### Small-scale spatial association

Arctic cod food resource and small Arctic cod biomass measured in the depth intervals of 12–20 m and 12–200 m had no significant effect (p > 0.05, Zero-inflated negative binomial regression) on thick-billed murre density or on northern fulmar and black-legged kittiwake abundances at a small scale of 5 km (Negative binomial part of the regression models, Table 4). There was no significant effect (p > 0.05) of ice concentration on seabird abundance/density.

Excess in absence of the three seabird species (0-values) at small scale (5 km) was not significantly influenced (p > 0.05, Zero-inflated negative binomial regression) by Arctic cod food resource, small Arctic cod biomass or ice concentration (Zero-inflated part of the regression models, Table 4).

Table 4

Zero-inflated negative binomial regression coefficients of the effect of Arctic cod food resource, small Arctic cod (AC) biomass and ice concentration on thick-billed murre density, and northern fulmar and black-legged kittiwake abundances at small scale of 5 km.a DOI: https://doi.org/10.1525/elementa.339.t4

SpeciesParameterNegative binomial partZero-inflated part

EstimateZ-valueEstimateZ-value

Thick-billed murre Food resource (12–200 m) 0.008 0.999 –0.001 –0.197
Small AC (12–200 m) 0.004 0.084 –0.004 –0.062
Ice concentration 0.015 1.522 –0.033 –1.659
Northern fulmar Food resource (12–20 m) 0.020 0.651 0.016 0.440
Small AC (12–20 m) –0.217 –1.672 –2.200 –0.772
Ice concentration 0.012 0.846 0.033 0.916
Black-legged kittiwake Food resource (12–20 m) 0.041 0.899 0.017 0.507
Small AC (12–20 m) –0.106 –0.356 0.223 0.593
Ice concentration –0.023 –1.276 –0.006 –0.145
SpeciesParameterNegative binomial partZero-inflated part

EstimateZ-valueEstimateZ-value

Thick-billed murre Food resource (12–200 m) 0.008 0.999 –0.001 –0.197
Small AC (12–200 m) 0.004 0.084 –0.004 –0.062
Ice concentration 0.015 1.522 –0.033 –1.659
Northern fulmar Food resource (12–20 m) 0.020 0.651 0.016 0.440
Small AC (12–20 m) –0.217 –1.672 –2.200 –0.772
Ice concentration 0.012 0.846 0.033 0.916
Black-legged kittiwake Food resource (12–20 m) 0.041 0.899 0.017 0.507
Small AC (12–20 m) –0.106 –0.356 0.223 0.593
Ice concentration –0.023 –1.276 –0.006 –0.145

a There was no significant effect (p > 0.05).

### An ontogenetic migration of Arctic cod across the marginal ice zone?

Over the annual cycle in the offshore Beaufort Sea, Arctic cod were segregated vertically by size and age, with age-0 forming an epipelagic scattering layer (<100 m) and age-1+ distributing progressively deeper in the mesopelagic scattering layer (200–500 m) as their size increased (Geoffroy et al., 2016). The same general distribution into an epipelagic scattering layer dominated by age-0 Arctic cod and a mesopelagic layer dominated by larger fish was observed in southern Baffin Bay (Figure 3). Our TS analysis also confirmed a general increase in the length of Arctic cod with depth (Figures 4 and 5). It further revealed an ontogenetic horizontal migration of age-0 in the depth interval of 12–200 m from the open waters (0–20% ice cover), where small larvae <1.2 cm dominated, to the 40–100% ice pack, where metamorphosing and metamorphosed juveniles (2.3–6 cm) were increasingly frequent (Figure 4). Vertically, age-1 and age-2 Arctic cod were detected at all depths over the interval of 12–200 m, with increased frequency in the subsurface layer of 40–100 m (Figure 5). The emerging pattern is that of a progressive migration of age-0 Arctic cod from the productive open waters to the shelter of the ice cover, followed by some concentration of the developing age-1 and age-2 in the depth interval of 40–100 m. Thereafter, a further ontogenetic migration of age-2+ Arctic cod to depths >200 m occurs (Geoffroy et al., 2016). In this study, this pattern resulted in a general increase in the availability of Arctic cod resource to seabirds from the open eastern waters to the ice-covered western waters, in particular in the depth layer of 12–200 m accessible to the deep-diving thick-billed murre (Figure 6; Table 2). However, this general pattern is not fully supported by the distribution of Arctic cod resource as a function of ice concentration, which shows that, despite higher proportions in biomass of age-1+ Arctic cod (>6 cm) under the ice cover (Figure 8), the food resource for deep-diving seabirds in the layer of 12–200 m can be similar at the ice edge (0–20% cover) and under the ice cover (Figure 7a).

Vertically, we found no indication in the acoustic record that the age-1 and age-2 Arctic cod prey of seabirds were particularly abundant near the ice at the surface, their relative frequency and biomass being much higher in the overall layer of 12–200 m than in the narrow sublayer of 12–20 m (Figures 4 and 8). Yet, age-1 and age-2 Arctic cod with sympagic behavior are present at the ice-water interface and sporadically can swarm immediately under the ice (Gradinger and Bluhm, 2004; Melnikov and Chernova, 2013; David et al., 2016). Such near-surface swarms could not be detected acoustically given the 0–12 m acoustic gap above the effective sampling depths of our echosounder, although some sympagic individuals might have been detected below 12 m as, hypothetically, Arctic cod growing under the ice could migrate vertically to follow their zooplankton prey (see Melnikov and Chernova, 2013). As well, age-1+ fish may have shunned the ice-breaking noise from the ship because higher levels of Arctic cod resource were observed in the depth layer of 12–20 m at low ice concentrations, 0–20% (little or no icebreaking noise), than in medium–high concentrations of 20–100% (frequent noise) (Figure 7). In any case, the biomass of age-1+ Arctic cod detected in the depth interval of 12–20 m likely underestimated the actual food resource available to surface-feeding northern fulmar and black-legged kittiwake. In a given location, however, the epipelagic scattering layer (12 to ~50 m) proved remarkably uniform vertically (Figure 3). Accordingly, we assume that, although prey resource at the very surface might be influenced to some degree by top-down pressure from predators, spatial variations in Arctic cod resource in the depth layer of 12–20 m were a representative proxy of spatial variations in the food available to surface-feeding seabirds.

### A lack of coherence in the spatial co-distribution of seabirds and their Arctic cod food

As previously reported in the Barents Sea (Mehlum and Gabrielsen, 1993), the fish diet of black-legged kittiwake was comprised of the Arctic cod and the bathypelagic B. glaciale, suggesting that foraging was limited to deep offshore waters, presumably when these two pelagic fish move close to the surface. Northern fulmar preyed on a much wider fish spectrum, including the ice-associated Arctic cod and L. fabricii, the bathypelagic B. glaciale and L. macdonaldi, as well as several benthic or benthopelagic species not captured by the nets and trawls (Table 1). Its fish diet and surface-feeding behaviour suggest that, in addition to feeding in the offshore ice-covered waters, the fulmar likely fed in shallow waters on its way from the colonies to the marginal ice zone and possibly also on discard from trawlers or dead organisms (Hobson and Welch, 1992; Mehlum and Gabrielsen, 1993; Phillips et al., 1999). Although dominated by Arctic cod, the fish diet of thick-billed murre included the infrequent L. atlanticus and Triglops sp., suggesting some limited nearshore feeding as well (Elliott et al., 2009). As the timing of this study (9 June to 12 July) corresponded to the beginning of the breeding season of the three seabird species (McLaren, 1982), some fish ingested may have also been captured during migration from wintering grounds.

Anecdotal reports have described how seabirds and marine mammals will congregate on kilometre-long schools of Arctic cod trapped in shallow waters, and deplete the available food resource over several days (Crawford and Jorgenson, 1993, 1996; Welch et al., 1993). In shallow Allen Bay (Nunavut), where Arctic cod form dense schools at depths between 7 and 20 m with small satellite aggregations close to the surface, the presence of schools was a significant predictor of the distribution of northern fulmar but not black-legged kittiwake (Matley et al., 2012a). The relatively weak association of seabirds with Arctic cod schools was attributed to a “local enhancement” mechanism (Thorpe, 1963) by which predators use memory and the presence of other predators (seabirds and marine mammals) to locate their food. In the offshore marginal ice zone, Arctic cod formed continuous scattering layers rather than dense and discrete schools. Little coherence between the distribution of seabirds and that of Arctic cod was found, despite the observed regional variations in the density of the scattering layers (Figure 6). Predators and prey resource were generally more abundant on the western ice-covered side of Baffin Bay (Table 2), suggesting some spatial match at large spatial scales. However, when comparing seabird abundance/density and their Arctic cod food-resource proxy in different ice concentrations across the marginal ice zone, no clear pattern was found (Figure 7). The only significant positive match was detected at the medium scale of 12.5 km, and only between murre density and Arctic cod resource (Table 3). These results suggest that seabirds respond to the “finer grain” in the spatial variability of their resource, and that local enhancement operates mainly at smaller scales than those resolved in the present study.

Another non-exclusive explanation for the observed lack of spatial match between seabirds and Arctic cod is that the fish prey successfully avoided its predators (Schneider and Piatt, 1986; Logerwell and Hargreaves, 1996). The observed migration of age-0 Arctic cod to ice-covered areas as their length attains the preferred prey length of seabirds suggests an adaptation to minimize avian predation. The subsequent concentration of age-1 and age-2 immature Arctic cod in the 40–100 m depth interval may be a response to limit predation by thick-billed murre and immature ringed seals Phoca hispida (Benoit et al., 2010). The further ontogenetic migration of age-2+ Arctic cod to depths >200 m could reduce predation by deep-diving mature seals and murre (Benoit et al., 2010; Geoffroy et al., 2016).

Seabird predators and their fish prey are more closely associated when prey density is high (e.g., Piatt, 1990; Mehlum et al., 1999; Fauchald and Erikstad, 2002; but see also Vlietstra, 2005). In the present study, as Arctic cod did not form dense schools, the food-resource proxy of seabirds was relatively low, with values ranging from 0 to 325 mg m–2 (mean value of 5 mg m–2) for the surface-feeding northern fulmar and black-legged kittiwake, and from 0 to 550 mg m–2 (mean value of 53 mg m–2) for the deep-diving thick-billed murre, and the biomass of small Arctic cod (<4 cm length) was even lower. Low food levels could be perceived as uniform by the predators, which would explain a lack of strong correlation between seabird and Arctic cod distributions at the different scales.

One could also argue that seabirds analysed in this study might have ingested smaller Arctic cod than the smallest Arctic cod found in bird stomachs (5.7 cm), and that bird distribution was thus more influenced by the distribution of these small Arctic cod than that of larger individuals considered as food resource. Our results at large, medium and small scale suggest, however, that such a greater influence was not the case. Except for the thick-billed murre in the 80–100% ice cover, at no scale was the distribution of seabirds better matched to that of small Arctic cod (<4 cm length) than to their potential food resource based on stomach analysis (4–18 cm length), suggesting that large Arctic cod as found in bird stomachs was their probable target.

Finally, as this study was conducted during the breeding season, the location of the colonies and breeding activity may have influenced seabird foraging behavior and, in addition to ice concentration and food resource, their offshore distribution across the marginal ice zone. Seabirds observed offshore were either non-breeders (immature birds or adults that skip breeding for a year) or breeders on foraging trips from their colonies. However, our observation data cannot distinguish breeding from non-breeding individuals. Foraging trips for breeding murre, kittiwake and fulmar may extend up to 300, 450 and 500 km from their colonies, respectively (Dunnet and Ollason, 1982; Weimerskirch et al., 2001; Falk et al., 2002; Ponchon et al., 2014; Yamamoto et al., 2016), although they have been documented to be mostly ≤100 km for murre and kittiwake in Baffin Bay (Boertmann and Mosbech, 2017). The location of their colonies may have therefore limited the area where observed seabirds could forage and the possibility of encountering high food resources. This situation was potentially the case for the thick-billed murre which was concentrated in the southern transects (Figure 6) relatively close to the Minarets colony on Baffin Island (~130 km to closest sampling transect), but it did not prevent its distribution from matching the highest Arctic cod food resource at medium scale.

### Sea ice and seabirds

Although ice concentration did not significantly influence the three seabird species at medium and small scales (Tables 3 and 4), our large-scale results suggest that surface-feeding kittiwake and fulmar might avoid zones of high ice concentration as too much ice can impede foraging, which is in agreement with previous reports (e.g., McLaren, 1982; Hunt, 1990; Mehlum, 1997; Mosbech and Johnson, 1999). Kittiwake and fulmar might prefer low to medium ice concentration (Figure 7), possibly because of the additional availability of sympagic prey compared to open waters (Bradstreet and Cross, 1982; Hobson and Welch, 1992; Lønne and Gabrielsen, 1992; Karnovsky et al., 2008). Thick-billed murre can be associated with higher ice concentration (McLaren, 1982; Laidre et al., 2008), as confirmed by our results (Figure 7). Like fulmar and kittiwake, murre benefit from an ice-covered area by feeding in openings on ice-associated invertebrates and fish (Bradstreet and Cross, 1982; Karnovsky et al., 2008).

Matley et al. (2012a) pondered if the association between surface-feeding fulmar and Arctic cod schools observed at scales of hundreds of metres in shallow Allen Bay would hold at larger scales. Our results indicate that it does not. This multiscale study points to a lack of horizontal variability in the offshore distribution of Arctic cod as a cause for the weak spatial correlation between seabird abundance and the biomass of their main food resource. In ice-free nearshore areas, the sporadic occurrence of dense schools of Arctic cod trapped near the surface by the shallow seafloor represents a trophic bonanza for seabird and marine mammal predators, one that may play a significant role in structuring the ecosystem (Welch et al., 1993; Davoren et al., 2003; Matley et al., 2012a). Offshore, the horizontal migration of age-0 Arctic cod under the ice cover at the precise moment in their ontogeny when they reach the length that makes them vulnerable to seabird predators, followed by the vertical migration of the age-1 to deeper layers, strongly suggests a strategy to minimize avian predation. It is tempting to speculate that this strategy and the lack of schooling behavior both contribute to reduce avian predation and underpin the general lack of spatial correlation between seabirds and their food across the marginal ice zone. We note, however, that the study of the co-distribution of fish and their seabird predators in offshore ice-covered waters is plagued by technical difficulties and potential observational biases, including the intrusive presence of a large research icebreaker. As both autonomous underwater vehicles and unmanned aerial vehicles become available for ecological studies, the coordinated deployment of an upward-looking fish echosounder under the ice and a seabird-observing drone above the ice could provide new insights into the predator-prey relationships between arctic seabirds and their fish prey across the marginal ice zone.

• Hydroacoustic data: Available at https://www.polardata.ca (CCIN Reference No.: 12841).

• Seabird observations data: Can be made available by contacting Anders Mosbech, Aarhus University, Denmark, at amo@bios.au.dk.

We thank the officers and crew of the CCGS Amundsen for their dedication and professionalism. We are grateful to the several technicians and colleagues for their contribution to seabirds observations, hydroacoustic data and net samples. Special thanks to Marine Cusa for seabird stomach analysis. The Green Edge project is conducted under the scientific coordination of the CERC on Remote sensing of Canada’s new Arctic frontier and the CNRS & Université Laval Takuvik Joint International Laboratory (UMI3376). Thanks to Québec-Océan and the Polar Continental Shelf Program for in-kind contributions of logistics and equipment. ML benefited from scholarships from the Natural Sciences and Engineering Research Council of Canada (NSERC). This is a contribution to ArcticNet, Québec-Océan at Université Laval, Green Edge, and the Canada Research Chair on the response of arctic marine ecosystems to climate warming.

The Green Edge project is funded by the following Canadian and French programs and agencies: the Network of Centres of Excellence ArcticNet, the Canada Foundation for Innovation (Amundsen Science), Canada’s Excellence in Research Chair (CERC) program, ANR (Contract #111112), CNES (project #131425), French Arctic Initiative, Fondation Total, CSA, LEFE and IPEV (project #1164).

The authors have no competing interests to declare.

• Contributed to conception and design: ML, AM, LF

• Contributed to acquisition of data: ML, SEG

• Contributed to analysis and interpretation of data: ML, SG, SEG, AM, LF

• Drafted and/or revised the article: ML, SG, SEG, AM, LF

• Approved the submitted version for publication: ML, SG, SEG, AM, LF

1
Benoit
,
D
,
Simard
,
Y
and
Fortier
,
L
.
2008
.
Hydroacoustic detection of large winter aggregations of Arctic cod (Boreogadus saida) at depth in ice-covered Franklin Bay (Beaufort Sea)
.
J Geophys Res
113
(
C06
):
S90
. DOI:
2
Benoit
,
D
,
Simard
,
Y
and
Fortier
,
L
.
2014
.
Pre-winter distribution and habitat characteristics of polar cod (Boreogadus saida) in southeastern Beaufort Sea
.
Polar Biol
37
:
149
163
. DOI:
3
Benoit
,
D
,
Simard
,
Y
,
Gagné
,
J
,
Geoffroy
,
M
and
Fortier
,
L
.
2010
.
From polar night to midnight sun: Photoperiod, seal predation, and the diel vertical migrations of polar cod (Boreogadus saida) under landfast ice in the Arctic Ocean
.
Polar Biol
33
:
1505
1520
. DOI:
4
Boertmann
,
D
and
Mosbech
,
A
.
2017
. Baffin Bay. An updated strategic Environmental Impact Assessment of petroleum activities in the Greenland part of Baffin Bay.
Scientific Report
218
:
320
.
Aarhus University, Danish Centre for Environment and Energy
,
Roskilde
.
5
Bouchard
,
C
,
Geoffroy
,
M
,
LeBlanc
,
M
,
Majewski
,
A
,
Gauthier
,
S
,
Walkusz
,
W
,
Reist
,
JD
and
Fortier
,
L
.
2017
.
Climate warming enhances polar cod recruitment, at least transiently
.
Prog Oceanogr
156
:
121
129
. DOI:
6
,
MSW
.
1982
.
Occurence, habitat use, and behaviour of seabirds, marine mammals, and Arctic cod at the Pond Inlet ice edge
.
Arctic
35
:
28
40
. DOI:
7
,
MSW
and
Cross
,
WE
.
1982
.
Trophic relationships at high Arctic ice edges
.
Arctic
35
:
1
12
. DOI:
8
Buckland
,
ST
,
Anderson
,
D
,
Burnham
,
K
,
Laake
,
J
,
Thomas
,
L
and
Borchers
,
D
.
2001
.
Introduction to distance sampling: Estimating abundance of biological populations
,
432
.
Oxford University Press
,
Oxford
.
9
Campana
,
SE
.
2004
.
Photographic atlas of fish otoliths of the Northwest Atlantic Ocean
,
284
.
NRC Research Press
,
Ottawa
.
10
Crawford
,
RE
and
Jorgenson
,
JK
.
1993
.
Schooling behavior of arctic cod, Boreogadus saida, in relation to drifting pack ice
.
Environ Biol Fish
36
:
345
357
. DOI:
11
Crawford
,
RE
and
Jorgenson
,
JK
.
1996
.
Quantitative studies of Arctic cod (Boreogadus saida) schools: Important energy stores in the Arctic food web
.
Arctic
49
:
181
193
. DOI:
12
Darnis
,
G
,
Robert
,
D
,
Pomerleau
,
C
,
,
H
,
Archambault
,
P
,
Nelson
,
RJ
,
Geoffroy
,
M
,
Tremblay
,
JE
,
Lovejoy
,
C
,
Ferguson
,
SH
,
Hunt
,
BPV
and
Fortier
,
L
.
2012
.
Current state and trends in Canadian Arctic marine ecosystems: II. Heterotrophic food web, pelagic-benthic coupling, and biodiversity
.
Clim Change
115
:
179
205
. DOI:
13
David
,
C
,
Lange
,
B
,
Krumpen
,
T
,
Schaafsma
,
F
,
van Franeker
,
JA
and
Flores
,
H
.
2016
.
Under-ice distribution of polar cod Boreogadus saida in the central Arctic Ocean and their association with sea-ice habitat properties
.
Polar Biol
39
:
981
994
. DOI:
14
Davoren
,
GK
,
Montevecchi
,
WA
and
Anderson
,
JT
.
2003
.
Distributional patterns of a marine bird and its prey: Habitat selection based on prey and conspecific behaviour
.
Mar Ecol Prog Ser
256
:
229
242
. DOI:
15
De Robertis
,
A
and
Higginbottom
,
I
.
2007
.
A post-processing technique to estimate the signal-to-noise ratio and remove echosounder background noise
.
ICES J Mar Sci
64
:
1282
1291
. DOI:
16
Dunnet
,
GM
and
Ollason
,
JC
.
1982
.
The feeding dispersal of fulmars Fulmarus glacialis in the breeding season
.
Ibis
124
:
359
361
. DOI:
17
Elliott
,
KH
and
Gaston
,
AJ
.
2014
.
Dive behaviour and daily energy expenditure in thick-billed Murres Uria lomvia after leaving the breeding colony
.
Mar Ornithol
42
:
183
189
.
18
Elliott
,
KH
,
Woo
,
KJ
,
Gaston
,
AJ
,
Benvenuti
,
S
,
Dall’Antonia
,
L
and
Davoren
,
GK
.
2009
.
Central-place foraging in an Arctic seabird provides evidence for Storer-Ashmole’s halo
.
Auk
126
:
613
625
. DOI:
19
Falk
,
K
,
Benvenuti
,
S
,
Dall’Antonia
,
L
,
Gilchrist
,
G
and
Kampp
,
K
.
2002
.
Foraging behaviour of thick-billed murres breeding in different sectors of the North Water polynya: An inter-colony comparison
.
Mar Ecol Prog Ser
231
:
293
302
. DOI:
20
Falk-Petersen
,
IB
,
Frivoll
,
V
,
Gulliksen
,
B
and
Haug
,
T
.
1986
.
Occurrence and age/size relations of polar cod, Boreogadus saida (LEPECHIN), in Spitsbergen coastal waters
.
Sarsia
71
:
235
245
. DOI:
21
Fauchald
,
P
and
,
KE
.
2002
.
Scale-dependent predator-prey interactions: The aggregative response of seabirds to prey under variable prey abundance and patchiness
.
Mar Ecol Prog Ser
231
:
279
291
. DOI:
22
Fortier
,
L
and
Harris
,
RP
.
1989
.
Optimal foraging and density-dependent competition in marine fish larvae
.
Mar Ecol Prog Ser
51
:
19
33
. DOI:
23
Fortier
,
L
,
Reist
,
JD
,
Ferguson
,
SH
,
Archambault
,
P
,
Matley
,
J
,
Macdonald
,
RW
,
Robert
,
D
,
Darnis
,
G
,
Geoffroy
,
M
,
Suzuki
,
K
,
Falardeau
,
M
,
MacPhee
,
SA
,
Majewsi
,
AR
,
Marcoux
,
M
,
Sawatzky
,
CD
,
Atchison
,
S
,
Loseto
,
LL
,
Grant
,
C
,
,
H
,
Asselin
,
NC
,
Harwood
,
LA
,
Slavik
,
D
and
Letcher
,
RJ
.
2015
. Arctic change: Impacts on marine ecosystems and contaminants. In:
Stern
,
G
and
,
A
(eds.),
From science to policy in the western and central Canadian Arctic: An Integrated Regional Impact Study (IRIS) of Climate Change and Modernization
,
432
.
ArcticNet Inc.
,
Quebec City
.
24
Garthe
,
S
and
Furness
,
RW
.
2001
.
Frequent shallow diving by a Northern Fulmar feeding at Shetland
.
Waterbirds
24
:
287
289
. DOI:
25
Gaston
,
AJ
and
Hipfner
,
JM
.
2000
. Thick-billed Murre (Uria lomvia). In:
Poole
,
A
and
Gill
,
F
(eds.),
The Birds of North America
,
497
.
The Birds of North America Inc.
,
. DOI:
26
Gauthier
,
S
and
Rose
,
GA
.
2001
.
Diagnostic tools for unbiased in situ target strength estimation
.
Can J Fish Aquat Sci
58
:
2149
2155
. DOI:
27
Geoffroy
,
M
,
Majewski
,
A
,
LeBlanc
,
M
,
Gauthier
,
S
,
Walkusz
,
W
,
Reist
,
JD
and
Fortier
,
L
.
2016
.
Vertical segregation of age-0 and age-1+ polar cod (Boreogadus saida) over the annual cycle in the Canadian Beaufort Sea
.
Polar Biol
39
:
1023
1037
. DOI:
28
,
RR
and
Bluhm
,
BA
.
2004
.
In-situ observations on the distribution and behavior of amphipods and Arctic cod (Boreogadus saida) under the sea ice of the High Arctic Canada Basin
.
Polar Biol
27
:
595
603
. DOI:
29
Härkönen
,
T
.
1986
.
Guide to the otoliths of the bony fishes of the Northeast Atlantic
,
256
.
Danbiu ApS. Biological Consultants
,
Hellerup
.
30
Hawkins
,
P
,
Butler
,
P
,
Woakes
,
A
and
Gabrielsen
,
G
.
1997
.
Heat increment of feeding in Brunnich’s guillemot
.
J Exp Biol
200
:
1757
1763
.
31
Hilton
,
GM
,
Furness
,
RW
and
Houston
,
DC
.
2000
.
A comparative study of digestion in North Atlantic seabirds
.
J Avian Biol
31
:
36
46
. DOI:
32
Hobson
,
KA
and
Welch
,
HE
.
1992
.
Observations of foraging northern fulmars (Fulmarus glacialis) in the Canadian High Arctic
.
Arctic
45
:
150
153
. DOI:
33
Hop
,
H
,
Welch
,
HE
and
Crawford
,
RE
.
1997
.
Population structure and feeding ecology of Arctic cod schools in the Canadian High Arctic
.
Am Fish Soc Symp
19
:
68
80
.
34
Hunt
,
GL
, Jr.
1990
.
The pelagic distribution of marine birds in a heterogeneous environment
.
Polar Res
8
:
43
54
. DOI:
35
Johansen
,
KL
,
Boertmann
,
D
,
Mosbech
,
A
and
Hansen
,
TB
.
2015
.
Manual for seabird and marine mammal survey on seismic vessels in Greenland
, 4th ed.,
74
.
Aarhus University, Danish Centre for Environment and Energy
,
Roskilde
.
36
Kaartvedt
,
S
,
,
A
,
Klevjer
,
TA
and
Staby
,
A
.
2009
.
Use of bottom-mounted echo sounders in exploring behavior of mesopelagic fishes
.
Mar Ecol Prog Ser
395
:
109
118
. DOI:
37
Karnovsky
,
NJ
,
Hobson
,
KA
,
Iverson
,
S
and
Hunt
,
GL
.
2008
.
Seasonal changes in diets of seabirds in the North Water Polynya: A multiple-indicator approach
.
Mar Ecol Prog Ser
357
:
291
299
. DOI:
38
Kortsch
,
S
,
Primicerio
,
R
,
Fossheim
,
M
,
Dolgov
,
AV
and
Aschan
,
M
.
2015
.
Climate change alters the structure of arctic marine food webs due to poleward shifts of boreal generalists
.
Proc R Soc B
282
:
31
39
. DOI:
39
Laidre
,
KL
,
Heide-Jørgensen
,
MP
,
Nyeland
,
J
,
Mosbech
,
A
and
Boertmann
,
D
.
2008
.
Latitudinal gradients in sea ice and primary production determine Arctic seabird colony size in Greenland
.
Proc R Soc B
275
:
2695
2702
. DOI:
40
Logerwell
,
EA
and
Hargreaves
,
NB
.
1996
.
The distribution of sea birds relative to their fish prey off Vancouver Island: Opposing results at large and small spatial scales
.
Fish Oceanogr
5
:
163
175
. DOI:
41
Logerwell
,
EA
,
Hewitt
,
RP
and
Demer
,
DA
.
1998
.
Scale-dependent spatial variance patterns and correlations of seabirds and prey in the southeastern Bering Sea as revealed by spectral analysis
.
Ecography
21
:
212
223
. DOI:
42
Lønne
,
OJ
and
Gabrielsen
,
GW
.
1992
.
Summer diet of seabirds feeding in sea-ice-covered waters near Svalbard
.
Polar Biol
12
:
685
692
. DOI:
43
Matley
,
JK
,
Crawford
,
RE
and
Dick
,
TA
.
2012a
.
Summer foraging behaviour of shallow-diving seabirds and distribution of their prey, Arctic cod (Boreogadus saida), in the Canadian Arctic
.
Polar Res
31
. DOI:
44
Matley
,
JK
,
Fisk
,
AT
and
Dick
,
TA
.
2012b
.
Seabird predation on Arctic cod during summer in the Canadian Arctic
.
Mar Ecol Prog Ser
450
:
219
228
. DOI:
45
McLaren
,
PL
.
1982
.
Spring migration and habitat use by seabirds in Eastern Lancaster Sound and Western Baffin Bay
.
Arctic
35
:
88
111
. DOI:
46
Mehlum
,
F
.
1997
.
Seabird species associations and affinities to areas covered with sea ice in the northern Greenland and Barents seas
.
Polar Biol
18
:
116
127
. DOI:
47
Mehlum
,
F
and
Gabrielsen
,
G
.
1993
.
The diet of high-arctic seabirds in coastal and ice-covered, pelagic areas near the Svalbard archipelago
.
Polar Res
12
:
1
20
. DOI:
48
Mehlum
,
F
,
Hunt
,
GL
,
Klusek
,
Z
and
Decker
,
MB
.
1999
.
Scale-dependent correlations between the abundance of Brunnich’s guillemots and their prey
.
J Anim Ecol
68
:
60
72
. DOI:
49
Melnikov
,
IA
and
Chernova
,
NV
.
2013
.
Characteristics of under-ice swarming of polar cod Boreogadus saida (Gadidae) in the central Arctic Ocean
.
J Ichthyol
53
:
7
15
. DOI:
50
Mosbech
,
A
and
Johnson
,
SR
.
1999
.
Late winter distribution and abundance of sea-associated birds in south-western Greenland, the Davis Strait and southern Baffin Bay
.
Polar Res
18
:
1
17
. DOI:
51
Mundy
,
CJ
,
Gosselin
,
M
,
Ehn
,
J
,
Gratton
,
Y
,
Rossnagel
,
A
,
Barber
,
DG
,
Martin
,
J
,
Tremblay
,
JE
,
Palmer
,
M
,
Arrigo
,
KR
,
Darnis
,
G
,
Fortier
,
L
,
Else
,
B
and
Papakyriakou
,
T
.
2009
.
Contribution of under-ice primary production to an ice-edge upwelling phytoplankton bloom in the Canadian Beaufort Sea
.
Geophys Res Lett
36
:
L17601
. DOI:
52
Parker-Stetter
,
SL
,
Rudstam
,
LG
,
Sullivan
,
PJ
and
Warner
,
DM
.
2009
.
Standard operating procedures for fisheries acoustic surveys in the Great Lakes
,
168
.
Great Lakes Fish Comm Spec Pub 09-01
,
Ann Arbor
.
53
Phillips
,
R
,
Petersen
,
M
,
Lilliendahl
,
K
,
Solmundsson
,
J
,
Hamer
,
K
,
Camphuysen
,
C
and
Zonfrillo
,
B
.
1999
.
Diet of the northern fulmar Fulmarus glacialis: Reliance on commercial fisheries?
Mar Biol
135
:
159
170
. DOI:
54
Piatt
,
JF
.
1990
.
The aggregative response of common murres and Atlantic puffins to schools of capelin
.
Stud Avian Biol
14
:
36
51
.
55
Ponchon
,
A
,
Gremillet
,
D
,
Christensen-Dalsgaard
,
S
,
,
KE
,
Barrett
,
RT
,
Reiertsen
,
TK
,
McCoy
,
KD
,
Tveraa
,
T
and
Boulinier
,
T
.
2014
.
When things go wrong: Intra-season dynamics of breeding failure in a seabird
.
Ecosphere
5
:
4
. DOI:
56
Ponton
,
D
and
Fortier
,
L
.
1992
.
Vertical distribution and foraging of marine fish larvae under the ice cover of southeastern Hudson Bay
.
Mar Ecol Prog Ser
81
:
215
227
. DOI:
57
R Core Team
.
2015
.
R: A language and environment for statistical computing
.
R Foundation for Statistical Computing
,
Vienna, Austria
.
58
Ryan
,
TE
,
Downie
,
RA
,
Kloser
,
RJ
and
Keith
,
G
.
2015
.
Reducing bias due to noise and attenuation in open-ocean echo integration data
.
ICES J Mar Sci
72
:
2482
2493
. DOI:
59
Schneider
,
DC
and
Piatt
,
JF
.
1986
.
Scale-dependent correlation of seabirds with schooling fish in a coastal ecosystem
.
Mar Ecol Prog Ser
32
:
237
246
. DOI:
60
Scoulding
,
B
,
Chu
,
D
,
Ona
,
E
and
Fernandes
,
PG
.
2015
.
Target strengths of two abundant mesopelagic fish species
.
J. Acoust. Soc. Am
137
:
989
1000
. DOI:
61
Simmonds
,
J
and
MacLennan
,
D
.
2005
.
Fisheries acoustics: Theory and practice
, 2nd edn.,
456
.
Blackwell Publishing
,
Oxford
. DOI:
62
Smith
,
WO
.
1987
.
Phytoplankton dynamics in marginal ice zones
.
Oceanogr Mar Biol
25
:
11
38
.
63
Suryan
,
RM
,
Kuletz
,
KJ
,
Parker-Stetter
,
SL
,
Ressler
,
PH
,
Renner
,
M
,
Horne
,
JK
,
Farley
,
EV
and
Labunski
,
EA
.
2016
.
Temporal shifts in seabird populations and spatial coherence with prey in the southeastern Bering Sea
.
Mar Ecol Prog Ser
549
:
199
215
. DOI:
64
Tang
,
CCL
,
Ross
,
CK
,
Yao
,
T
,
Petrie
,
B
,
DeTracey
,
BM
and
Dunlap
,
E
.
2004
.
The circulation, water masses and sea-ice of Baffin Bay
.
Prog Oceanogr
63
:
183
228
. DOI:
65
Thorpe
,
W
.
1963
.
Learning and Instinct in Animals
, 2nd edn.,
558
.
Harvard University Press
,
Cambridge
.
66
Vlietstra
,
LS
.
2005
.
Spatial associations between seabirds and prey: Effects of large-scale prey abundance on small-scale seabird distribution
.
Mar Ecol Prog Ser
291
:
275
287
. DOI:
67
Weimerskirch
,
H
,
Chastel
,
O
,
Cherel
,
Y
,
Henden
,
J-A
and
Tveraa
,
T
.
2001
.
Nest attendance and foraging movements of northern fulmars rearing chicks at Bjørnøya Barents Sea
.
Polar Biol
24
:
83
88
. DOI:
68
Welch
,
HE
,
Bergmann
,
MA
,
Siferd
,
TD
,
Martin
,
KA
,
Curtis
,
MF
,
Crawford
,
RE
,
Conover
,
RJ
and
Hop
,
H
.
1992
.
Energy Flow through the Marine Ecosystem of the Lancaster Sound Region, Arctic Canada
.
Arctic
45
:
343
357
. DOI:
69
Welch
,
HE
,
Crawford
,
RE
and
Hop
,
H
.
1993
.
Occurence of Arctic cod (Boreogadus saida) schools and their vulnerability to predation in the Canadian Arctic
.
Arctic
46
:
331
339
. DOI:
70
Yamamoto
,
T
,
Kokubun
,
N
,
Kikuchi
,
DM
,
Sato
,
N
,
Takahashi
,
A
,
Will
,
AP
,
Kitaysky
,
AS
and
Watanuki
,
Y
.
2016
.
Differential responses of seabirds to environmental variability over 2 years in the continental shelf and oceanic habitats of southeastern Bering Sea
.
Biogeosciences
13
:
2405
2414
. DOI:
This is an open-access article distributed under the terms of the Creative Commons Attribution 4.0 International License (CC-BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. See http://creativecommons.org/licenses/by/4.0/.