MARINE ECOLOGY PROGRESS SERIESMar Ecol Prog Ser

Vol. 513: 211223, 2014doi: 10.3354/meps10945

Published October 22

INTRODUCTION

Under the influence of climate change, Arctic sea-ice has been retreating over recent decades, withbreak-up becoming earlier while freeze-up becomeslater (Gagnon & Gough 2005, Hochheim & Barber2010, Hochheim et al. 2010, Stammerjohn et al.2012). Arctic seabirds depend on the presence ofopen water to feed, and consequently spring icebreak-up and clearance of sea-ice sets the timetablefor subsequent breeding activities (Gaston et al.2009, Smith & Gaston 2012). Spring ice break-up

may also be influential in determining the timing ofother events within the food web (e.g. phytoplanktonblooms: AMAP 2011; ringed seal dispersal: Cham-bellant 2010), hence affecting seabird foraging suc-cess and potentially population dynamics (Gaston etal. 2009, Smith & Gaston 2012). The timing of bothbreak-up and freeze-up likely affect the ecology ofArctic seabirds, but, of the 2, the timing of ice break-up is the most likely to affect reproduction (Gaston etal. 2012). This is especially true for those species tiedto returning to large colonies which have sometimesexisted in the same location for centuries (e.g. Gaston

Environment Canada 2014 Publisher: Inter-Research www.int-res.com

*Corresponding author: tony.gaston@ec.gc.ca

Seabird diet changes in northern Hudson Bay,19812013, reflect the availability of schooling prey

Anthony J. Gaston1,*, Kyle H. Elliott2

1Environment Canada National Wildlife Research Centre, Ottawa, ON K1A 0H3, Canada2Department of Biological Sciences, University of Manitoba, Winnipeg, MB R3T 2N2, Canada

ABSTRACT: Ongoing climate change is altering Arctic marine ecosystems with major conse-quences for food-webs. Seabirds, by foraging over large marine areas but returning regularly totheir breeding colonies, provide a good medium for tracking such changes. We studied the preydelivered to nestling thick-billed murres Uria lomvia at a colony in northern Hudson Bay, Canada,over the period 19812013. During that period, ice conditions in the region altered substantially,with earlier break-up and clearance. This change was not gradual but mainly occurred in the mid-1990s. Breeding chronology also advanced, but not as fast as early summer ice clearanceadvanced. Ice conditions strongly affected diet, presumably reflecting the change in the timing ofbreeding relative to ice clearance. Of the dominant species in the diet, the proportion of Arctic codBoreogadus saida decreased and the proportion of capelin Mallotus villosus increased over the33 yr period, even after ice effects were taken into account, suggesting a progressive and cumu-lative effect of environmental change on the marine community. Those cumulative changes weremanifested primarily in the diet representation of the most frequent taxa (cod and capelin), withvariation in secondary diet components being driven by the availability of the dominant taxa. Dietdiversity, representing an increase in the proportion of benthic fish (stichaeids, zoarcids, pholidsand sculpins) and invertebrates (squid, amphipods and crustaceans), increased when ice coverwas low, and hatching was late relative to ice break-up. Chick growth rates were low when theproportion of benthic fish was high. Hence, although it appears that growth rates are influencedby diet composition, it is more likely that they reflect the availability of the dominant schoolingprey species, rather than any nutritional deficiency in benthic species.

KEY WORDS: Ice cover Climate change Forage fish Thick-billed murre Uria lomvia

Resale or republication not permitted without written consent of the publisher

FREEREE ACCESSCCESS

Mar Ecol Prog Ser 513: 211223, 2014

& Donaldson 1996) and for which emigration be -tween colony sites is likely to take generations,rather than happening from year to year (Gaston2004). Birds adopting this conservative strategy withrespect to nesting location cannot easily adjust tochanging conditions by shifting their breeding site,necessitating other adjustments to changing environ-mental conditions.

To understand the likely consequences for Arcticseabirds and for their prey organisms of the climatechanges currently underway, we need to betterunderstand the role played by ice break-up andother climate variables in determining inter-yearvariations in prey availability. In this paper weexamine a 33 yr dataset relating to the breedingbiology and nestling diet of thick-billed murres Urialomvia in northern Hudson Bay, Canada, to explorethe role of environmental conditions in determiningthe timing of breeding and changes in diet thathave already been identified (Gaston et al. 2003,2012) and the effects of such changes in diet onnestling growth.

Seabird diets can be useful for interpreting chan -ges in prey populations (e.g. Barrett 2002, Deguchiet al. 2004, Hedd et al. 2006, Piatt et al. 2007,Springer et al. 2007, Renner et al. 2012). Most stud-ies are based on nestling diets as these can beobserved or sampled non-destructively (Barrett etal. 2007). The food delivered to nestlings in manycases diverges from the adult diet, especially in spe-cies where food is transported externally, held inthe bill, in which case larger prey items, providinggreater efficiency in transport, make up a higherproportion of food items in nestling diets (Thaxter etal. 2013). Although nestling and adult diets may notbe identical they normally encompass a similarrange of prey organisms, and individual specializa-tions tend to be reflected in both adult and offspringdiets (Woo et al. 2008, Provencher et al. 2013). Akey result from numerous seabird dietary studies isthat seabird productivity is associated with one or afew key species, and that when those key speciesbecome less available the diet becomes more diver-sified (Hedd et al. 2006, Schrimpf et al. 2012, Hatch2013).

The physical and biological oceanography ofHudson Bay is poorly characterized and monitoredrelative to many other parts of the Arctic (Hochheimet al. 2010, Stewart & Barber 2010). Therefore,information from seabirds may be particularly valu-able in detecting changes in Hudson Bay marineecosystems. Furthermore, the direct entry of airmasses from the Pacific and Atlantic Oceans into

the Hudson Bay is buffered by mountains and dis-tance, and there are few topographical featureswithin the region to modify local climates (Maxwell1986, Stewart & Barber 2010), leading to a homoge-nous Arctic ecosystem relatively unaffected byexternal influences. Ice freeze-up and break-updates in the Hudson Bay are primarily associatedwith local surface air temperature and ocean circu-latory processes (Hochheim et al. 2010, Stewart &Barber 2010), although global processes particu-larly, energy waves associated with Pacific andAtlantic oscillations do impact freeze-up dates inthe Hudson Bay via their influence on air tempera-ture (Hochheim & Barber 2010, Hochheim et al.2011, Gaston et al. 2012). Likewise, because theHudson Bay is only connected to the Arctic andAtlantic Oceans via small straits at its northernedge, marine communities are impoverished rela-tive to other marine environments at similar lati-tudes, and changes in marine communities throughthe immigration of more boreal elements are likelyto be slow in occurring.

Previous publications relating to older subsets ofthe same data have dealt with nestling diet changesthat occurred during a period of step change innorthern Hudson Bay ice conditions (Gaston et al.2003, 2012) and with changes in the overall dietexpressed as principal components within a multi-variate framework (Smith & Gaston 2012). In thispaper we addressed (1) how individual prey taxaand overall prey diversity are affected by ice condi-tions, the timing of ice conditions relative to breed-ing, sea-surface temperature (SST) and air tempera-ture; and (2) how nestling growth is impacted bydiet variation and by changes in timing of breeding.In addition, we examined whether the general prin-ciple that diet diversifies when preferred prey isscarce (Esterbrook & Dunham 1976, Pyke et al.1977, Pyke 1984) applies to these marine predators.In doing so, we attempted to infer causality within ahierarchy of interrelated variables, based on theassumption that physical changes precede biologicalchanges and that birds acquire prey by adoptingparticular feeding strategies that may not be optimalin a given year.

MATERIALS AND METHODS

Observations were made at the thick-billed murreUria lomvia colony near Cape Pembroke, CoatsIsland (62.947 N, 82.015 W), during 1981, 19842011 and 2013.

212

Gaston & Elliott: Ice and seabird diet in Hudson Bay

Diet

Diet data were derived from watches carried out toobserve prey delivered to nestlings. During 1981 and19841987, fish were observed opportunistically, asand when work and conditions allowed. From 1988onwards most observations were made during 3 ormore all-day watches carried out at evenly spacedintervals over the chick-rearing period (feedingwatches), although additional opportunistic recordswere also included. Up to 1992 many sightings weremade through binoculars at ranges of up to 40 m,and at such distances some taxa could not be accu-rately distinguished. From 1993 onwards all feedingwatches were carried out from a blind where themajority of deliveries observed took place within 5 mof the observer, allowing accurate identification of>80% of prey brought in. Observers had binoculars,but deliveries took place very rapidly, and the major-ity of prey items recorded from 1993 onwards wereidentified with the naked eye. Because of the dis-tance at which observations were made, up to 1992many deliveries were classified as simply sculpins(Cottidae; 4 genera known from the colony; Elliott &Gaston 2008) or blennies (Stichaeidae, Pholidae,Zoarcidae; 7 genera; Elliott & Gaston 2008). Theoccurrence of prey taxa was scored for each year asthe proportion (percent by number) of all prey de -liveries identified in that year (Table S1 in the Sup-plement at www.int-res.com/articles/suppl/m513p211_ supp. xls). More than 95% of deliveries con-sisted of a single prey item. Occasionally >1 item wasdelivered, but in such cases the items were usually ofthe same species. In the rare cases (1 taxon was delivered at once, the 2 taxa were bothcounted. For some analyses the proportions of the 3schooling fishes (Boreogadus, Mallotus and Ammo -dytes) were combined.

The diversity of prey delivered was expressed usingthe Shannon diversity index: H = pi ln(pi), wherepi is the proportion of all items delivered constitutedby species i (Shannon 1948, Hill 1973). The index wascalculated for years from 1993 on wards, when differ-ent genera of stichaeids, zoarcids and pholids wereidentified. For inclusion in principal componentanalyses (PCA; see Statistical analyses), we estima -ted H for the period 19881992 (based on thereduced diversity of fish identified in 19881992) byinterpolating the actual value of H from the relation-ship between the actual H and reduced H from the19932013 dataset (R2 = 0.97). Sculpins were lumpedin all analyses, as the identification of different generawas considered less reliable than other taxa.

Timing of breeding

Each year a sample of 60 to 100 (mean = 87) murreeggs was marked during incubation and visitedevery 2 d to ascertain the date of hatching. Themedian date of hatching for this sample of eggs wasused as an index of timing of breeding (medianhatch, 1986 onwards), as well as the differencebetween the median date of hatching and the date of50% ice clearance from the local area, NorthernHudson Bay Narrows (relative date of hatch; seeGaston et al. 2012 and next subsection). Nestlingswere weighed every 2 to 3 d and the mass on Day 14(hatching = Day 1), either measured directly or esti-mated by linear interpolation, was used as a measureof nestling growth rate. The earliest age at whichnestling murres begin to depart from the colony vol-untarily is 15 d, so mass at Day 14 is unbiased bydeparture age (Gaston & Hipfner 2000). Inter-yearcomparisons were based on the mean nestling massat Day 14, based only on nestlings hatching within7 d on either side of the median date of hatching.

Climate and ice conditions

Ice conditions were derived from the Canadian IceService ice archive using the Icegraph 2.0 analysistool (http://iceweb1.cis.ec.gc.ca/IceGraph20). Localice break-up was estimated from the Canadian IceService region Northern Hudson Bay Narrows(NHBN), while regional ice break-up was based onthe Hudson Bay region (Fig. 1). Ice cover in bothregions changes from ~100% in winter to ~0% in latesummer (AugustSeptember). Break-up commencesin May (Prinsenberg 1986). Indices of annual ice conditions were derived from the extent of ice ineach region on 2 July, the date closest to having an average of 50% ice cover during the period19712012 for the 2 areas used. Sea-surface temper-atures (SST) were obtained from the NOAA websitehttp:// nomad3.ncep.noaa.gov/png, using the areabounded by 6165 N and 8084 W and averagingmonthly values for June and July. Local air tempera-tures were obtained from the Environment Canadamonitoring station at Coral Harbour, the closest cli-mate station to the field site. Mean monthly tempera-ture for June was used in modeling median date ofhatching, as the murres arrive at the colony in lateMay and most have laid their eggs (and hence fixedthe timing of hatching, the incubation period beingvery constant; Gaston & Hipfner 2000) by the end ofJune (de Forest & Gaston 1996).

213

http://www.int-res.com/articles/suppl/m513p211_supp.xlshttp://www.int-res.com/articles/suppl/m513p211_supp.xls

Mar Ecol Prog Ser 513: 211223, 2014

Statistical analyses

Models were constructed using theBest Subsets module of Statistica7.1s GLZ routine (Statsoft 2005), withmodel suitability determined fromcomparison of AICc values (Akaikeinformation constants for small sam-ples), following the recommendationsof Burnham & Anderson (1998) andAnderson (2008). All models contain-ing more variables than the topranked model were discarded, as wellas all multiple-variable models withAICc > 3. AICc weights (wi) were cal-culated based on the remaining mod-els. Three types of model were con-sidered: (1) The median date ofhatching was modeled for the effectof year, SST, mean June air tempera-ture and both local and regional icecover, as well as a 2-category variableice phase, divided into years to 1995and years from 1996 onwards, to rep-resent the step change in ice condi-

tions in Hudson Bay in the mid-1990s, when ice coverin late June dropped from an average of ~70 to~50%, as noted by Gaston et al. (2012). (2) The pro-portions of different prey taxa in the diet were mod-eled using the same variables plus date of hatchingin relation to date of 50% ice (relative hatch date) butomitting June air temperature as unlikely to bedirectly linked to foraging conditions. The first roundof analyses showed that regional ice cover did notenter into any top model so, given the close correla-tion with local ice cover and SST, final analyses werecarried out using only year, ice phase, local ice cover,SST and relative date of hatching. Only prey taxamaking up >1% of diet items, averaged over allyears, were analysed (this resulted in the exclusion of7 taxa identified in >1 yr). (3) Lastly, 14 d nestlingmass was modeled in relation to ice phase, relativedate of hatching, June air temperature (which couldinfluence nestling energy balance) and the propor-tions of all prey taxa averaging >3% of prey annually(Boreogadus, Mallotus, Ammodytes, all blennies,all sculpins and amphipods The misto). All thesetaxa were identified in all years when 14 d mass wasmeasured. Full details of model comparisons aregiven in Table S1 in the Supplement.

We also examined the data within a multivariateframework because we anticipated that many of the

214

Fig. 1. Areas used in the calculation of local (NorthernHudson Bay Narrows , NHBN double-hatched) and re -gional (Hud son Bay single-hatched) ice cover, as definedby the Canadian Ice Service (http:// iceweb1. cis.ec.gc.ca).

A: Coats Island colony; B: Coral Harbour

Variable All years Years after 1995n (yr) r p n (yr) r p

PhysicalLocal ice cover 30 0.46 0.01 17 0.27 0.29Regional ice cover 30 0.44 0.02 17 0.11 0.66JuneJuly SST 29 0.13 0.49 17 0.13 0.61June air temperature 30 0.27 0.17 17 0.15 0.58

DietMallotus 30 0.81

Gaston & Elliott: Ice and seabird diet in Hudson Bay

prey items and climatic variables would covary dueto similar underlying processes. We calculated a PCAfor arcsine-transformed diet components. We usedthe PCA primarily as a visualization tool, especiallywithin the context of examining the components anddimensionality of diet diversity. We only includedthose axes with eigenvalues >1.0. To examine for cor-relations between dietary components and climaticvariables within their respective multivariate spaces,we then conducted a redundancy analysis (RDA). Forthe multivariate analyses, we only used data from1988 to 2013, as that was the period when fish wereidentified to detailed taxonomic groupings. We onlyincluded continuous variables within the RDA, andtherefore excluded ice phase as a climatic variable.We used R 2.14.2 for all multivariate analyses.

RESULTS

Inter-correlations and time trends

Areas of local (NHBN) and regional (Hudson Bay)ice cover were strongly correlated over the period19812013 (r31 = 0.66, p < 0.001), but there was muchvariation in local ice cover not explained by thelarger regional situation. Both decreased signifi-cantly over the period of the study (Table 1, Figs. 2 &3) and both were correlated with June to July SST(local r31 = 0.63, regional r31 = 0.54, both p < 0.001).Taken over the entire period, Mallotus and Ammo -dytes increased, while Boreogadus and sculpinsdecreased (Table 1, Table S1 in the Supplement).After 1995 there were no significant time trends inthe environmental variables (ice, SST; Table 1,Fig. 2), timing of breeding (median hatch, relativehatch) or any diet component other than Mallotusand Boreogadus.

Timing of breeding

Median date of hatching of thick-billed murre Urialomvia nestlings was correlated with June to July SST(r25 = 0.74, p < 0.001) and ice cover on 2 July (regionaland local, both r25 = 0.66, p < 0.001). The top best sub-sets model for median hatching included ice phase, re-gional ice cover and SST (wi = 0.48, adjusted R2 = 0.69,p = 0.001; Table 1). A model including only ice phaseand SST was the only other model within AICc < 2(wi = 0.36), and SST was included in models comprisinga combined AICc weight of >0.99. The top-rankedmodel had an AICc weight 40 times greater than the

highest single-variable model (SST). There was notrend in date of hatching after 1995 (Fig. 3).

Despite the correlation between median date ofhatching and ice cover on 2 July, estimated change in

215

Fig. 2. Trends in primary physical variables and principleprey species, expressed as percent deviations from themean of annual indices: (a) local ice cover on 2 July, Julysea- surface temperature (SST) and mean JuneJuly air tem -peratures at Coral Harbour; (b) schooling fish; and (c) ben-

thic fish and Themisto

Mar Ecol Prog Ser 513: 211223, 2014

the former over the study period (~5 d, range: 1931 July) was much less than for the change in thedate of 50% ice cover (~11 d, range: 9 June15 July).Relative date of hatching was negatively correlatedwith local and regional ice cover (r25 = 0.55 and0.76, respectively, both p < 0.01), but was only mar-ginally affected by year (r25 = 0.37, p = 0.06).

Diet

In years after 1992 (when different benthic generawere identified separately), proportions of Leptocli-nus, Gymnelus, Eumesogrammus and all sculpinswere all positively correlated with one another (all r18> 0.52, all p < 0.02; Fig. 4) and Leptoclinus, Eumeso-grammus and all blennies were positively corre-lated with Themisto (all r18 > 0.54, all p < 0.02; Fig. 4).Among the prey taxa for which best subsets modelswere run, 3 yielded no worthwhile model fit (R2 < 0.1,model p > 0.1; Table 2): the stichaeid (prickleback)Leptoclinus, the squid Gonatus and the amphipodThemisto. Among the other 6 taxa modeled, none ofthe top models, or those within AICc < 2, includedregional ice cover, whereas 5 included local ice coverand 5 in cluded the step variable ice phase (Table 2).Despite the inclusion of the ice phase variable, yearentered into the top models for Boreogadus and Mal-lotus, the 2 most common taxa in the diet, and rela-tive date of hatching was included in the top modelsfor Mallotus and all blennies. June SST was in -cluded in the top model for Boreogadus, and wasthe only variable included for Eumesogrammus(Table 2).

Among the 6 taxa with models yielding R2 > 0.1,Mallotus increased over time, while Boreogadusdecreased. Boreogadus, Mallotus and Ammodytesincreased with local ice cover, while all blennies

216

1985 1990 1995 2000 2005 2010 2015198

200

202

204

206

208

210

212

214

Med

ian

hatc

h (d

ays

from

1 J

anua

ry)

0.0

0.2

0.4

0.6

0.8

1.0

1.2

Loca

l ice

ext

ent,

2 J

uly

Hatch date

Local ice

Fig. 3. Trends in median date of hatch for thick-billed mur-res Uria lomvia at Coats Island, Nunavut, 19842013 andlocal ice cover on 2 July. Lines are fitted by distance-

weighted least-squares procedure

Table 2. Variables included in top models (dark colours) predicting diet components for thick-billed murres Uria lomvia atCoats Island, Nunavut, 19812013 (All years) and other models within AICc < 2 (light colours)brown: positive; blue: neg-ative. AICc weights, adjusted R2 values, F statistics and full model probabilities for the top model are also given. Grey cells

were not included in the respective models. SST: sea-surface temperature; H: Shannons diversity index

Gaston & Elliott: Ice and seabird diet in Hudson Bay

and sculpins decreased (Table 2). Those taxa whichincreased with ice cover decreased with relativehatch date and vice versa, presumably because of thestrong correlation between local ice cover and rela-tive hatch, so Mallotus decreased in the diet as thetime between hatching and 50% local ice coverincreased, whereas all blennies increased.

Shannons H was negatively correlated with all 3species of schooling fish (significant only for Mallo-tus, r18 = 0.58, p = 0.007) and positively correlatedwith all other taxa (p < 0.01 for all taxa except Gym-nelus, for which the relationship was not significant).When diet diversity was modeled in relation to phys-ical variables (year, ice, SST) and relative date ofhatching, only relative date of hatching appeared inthe top model (wi = 0.57), and the model includingboth relative date of hatching and year was withinAICc < 2 (wi = 0.20). However, if the proportion ofschooling fish was included, the top model includedyear, local ice cover, SST and sum of schooling fish,with wi = 0.97, and no other models were withinAICc < 2.

Chick mass at 14 d was correlated with the propor-tion of Leptoclinus, Gymnelus and all blennies inthe diet (all r = 0.45, p = 0.03), but not with year oreither ice cover variable. The top best subsets modelcontained only all blennies (Table 2). When the pro-portions of the 3 schooling genera were pooled, theresulting top model included both relative hatchingdate and all blennies, with a higher R2 (0.31 vs. 0.17)and the model including all blennies and Gonatuswas also well supported (AICc < 2). None of the non-prey variables (year, June air temperature, relativedate of hatching) gave wi > 0.05.

Within the multivariate analyses (Fig. 4), dietcould be decomposed into 3 principal axes, eachexplaining >10% of the variation in diet (Table 3).Variation in the proportion of secondary prey items(benthic prey and invertebrates) was associatedwith PC1 (Table 2, Fig. 4). Variation in the composi-tion of primary, schooling prey was associated withPC2 (Boreogadus relative to MallotusAmmodytes)and PC3 (Mallotus relative to Ammodytes; Table 3,Fig. 4). The Shannon diversity index was heavilyloaded on PC1, demonstrating that it was primarilylinked to variation in the proportion of secondaryprey items (benthic and invertebrate prey; Table 3,Fig. 4).

Climate variables significantly explained variationin diet (permutation test: F5,19 = 4.05, p = 0.005, 199permutations). Climate variables that explained dietcomponents could be decomposed into 2 main axes.The main correlate of diet was year, whichexplained primarily the variation in Boreogadus andMallotus (Fig. 5). The second axis was directly asso-ciated with climate variables, especially lo cal icecover, which in turn was correlated with SST andhatch date. That axis primarily re flected variation indiversity. Diversity was high when ice cover waslow.

217

0.4 0.2 0.0 0.2 0.4

0.4

0.2

0.0

0.2

0.4a

b

Principal Component 1

Prin

cip

al C

omp

onen

t 2

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

19981999

2000

2001

2002

2003

200420052006

20072008

2009

2010

20112013

4 2 0 2 4

4

2

0

2

4Boreogadus

MallotusAmmodytes

Leptoclinus

Gymnelus

Eumesogrammus

Cottidae

ThemistoGonatus

H

0.4 0.2 0.0 0.2 0.4

0.4

0.2

0.0

0.2

0.4

Principal Component 3

Prin

cip

al C

omp

onen

t 4

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

19981999

2000

2001

2002

2003

2004

2005

2006

2007

20082009

20102011

2013

3 2 1 0 1 2 3

3

2

1

0

1

2

3

Boreogadus

Mallotus

Ammodytes

Leptoclinus

Gymnelus

Eumesogrammus

Cottidae

Themisto

Gonatus

H

Fig. 4. Principal components analysis of thick-billed murreUria lomvia diet variation at Coats Island, Nunavut 19882013. Principal Component 4 had an eigenvalue

Mar Ecol Prog Ser 513: 211223, 2014

DISCUSSION

Timing of breeding by themurres Uria lomvia at CoatsIsland was associated with icephase, regional ice conditionsand SST, with SST apparentlythe strongest driver. Althoughregional ice cover was moreimportant than local ice cover inmodels predicting timing ofbreeding by the murres, theproportions of different preytaxa in the diet was more influ-enced by local ice cover, beingincluded in all top models,except those for the zoarchid

(eelpout) Gymnelus and the stichaeid (prickleback)Eumesogrammus. That result was also backed upby the strong support for ice cover in relation to dietdiversity within the redundancy analysis. Ice phasealso appeared in top models for 5 taxa, and giventhe small number of pre-step years included in thesample of years analysed for individual benthic gen-era, its exclusion from top models for Leptoclinusand Eumesogrammus is not surprising. The promi-nence of this variable among factors affecting murrediets confirms the importance of the step-change inmarine ecosystems that occurred in Hudson Bay inthe mid-1990s (Gaston et al. 2012), a change thatalso created a strong signal in freshwater ecosys-tems in the Hudson Bay lowlands (Rhland et al.2013). Similarly, in the Antarctic, ice conditionswere more important than year effects for the diet ofAdlie penguins (Pygoscelis adeliae) monitored over7 yr, with more fish consumed during years of lowerice cover (Ainley et al. 2003). Timing of hatchingrelative to the date of 50% ice cover appeared inthe top-ranked models for Mallotus (capelin) andall blennies, but date of hatching itself was notincluded in any, supporting the idea of Gaston et al.(2012) that timing of breeding per se is less impor-tant than timing relative to ice break-up, for thebiology of the murres.

To date, there is little evidence that birds areadjusting their timing of breeding to match changesin ice conditions. An initial advance in timing ofbreeding in the period up to 1996 (Gaston & Hipfner

218

Species Common name PC1 PC2 PC3

Variation explained (%) 37 26 13Boreogadus saida Arctic cod 0.138 0.568 0.188Mallotus villosus Capelin 0.106 0.512 0.373Ammodytes hexapterus Sandlance 0.174 0.426 0.417Leptoclinus maculatus Daubed shanny 0.426 0.234 0.139Gymnelus viridis Fish doctor 0.283 0.296 0.443Eumesogrammus praecisus Fourline snakeblenny 0.388 0.191Cottidae Sculpin 0.373 0.225Themisto libellula Amphipod 0.372 0.223Gonatus fabricii Squid 0.215 0.539Shannon diversity index (H ) 0.451 0.170 0.222

Table 3. Loadings from a principal components (PC) analysis of diet composition forthick-billed murres Uria lomvia at Coats Island, Nunavut, 19882013. Only axeswith eigenvalues >1.0 are shown; no other axis explained >10% of the variation in

diet. Loadings with absolute values

Gaston & Elliott: Ice and seabird diet in Hudson Bay

1998) was followed by relative stasis. This situationcontrasts with observations at Prince Leopold Island,in the high Arctic, where there was a strong correla-tion between date of 50% ice cover and date of lay-ing by thick-billed murres (Hipfner et al. 2005). Thelack of correlation between ice conditions and timingof breeding means that the difference between dateof 50% ice and date of hatching (relative hatch date)has increased, although the significance is marginal(p = 0.06).

The current analysis did not confirm the previoussuggestion that a decrease in Boreogadus in the dietcaused decreased nestling growth (Gaston & Hipfner1998). It appears that during the period of rapidlydiminishing ice cover in the 1990s, murres may havehad difficulty adapting their foraging behaviour tothe changing spectrum of prey, causing a reductionin the amount of food delivered to nestlings, eitherthrough a reduced rate of delivery or through thedelivery of smaller or less nutritious meals. Subse-quently, either they have become more successful atforaging on capelin Mallotus, or capelin has be comemore abundant locally, because post-1995 there wasno relationship between nestling mass and the pro-portion of Boreogadus delivered. Chick 14 d mass inseveral years after 2000 was high relative to that inthe early 1990s.

Year entered into the top models for the 2 mainprey taxa, Mallotus and Boreogadus, suggesting thattrends in these populations over time were, to someextent, independent of year-to-year fluctuations inice conditions. Likewise, in the multivariate ap -proach, the main environmental variable influencingdiet composition was year, and that was caused pri-marily by changes in the proportions of Mallotus andBoreogadus. This is not surprising, because individu-als of both genera delivered by murres to nestlingswere >1 yr old and therefore availability probablywould have been partially related to conditions inprevious years. Smith & Gaston (2012) found thatdetrended diet, as represented by PC1 and PC2 in aprinciple components analysis, was affected prima-rily by local SST, with the PCs heavily weighted bythe main prey species, Mallotus and Boreogadus. Inthe current study, year was not included in any topmodels for benthic fish (sculpins, all blennies, Lep-toclinus, Gymnelus, Eumesogrammus), suggestingthat there has been no secular trend in the availabil-ity of these fish. The importance of all blennies inmodels (and overall diversity within the multivariateanalyses) relating to chick mass at 14 d suggests thatthese fishes, of diverse phylogeny but all benthic andoften associated with the kelp Laminaria spp. (Cairns

1987, Scott & Scott 1988), tend to be taken more oftenin years when schooling prey (Boreogadus, Mallotus,Ammodytes) is less available than usual. This hypo -thesis is supported by the positive correlationsamong all the non-schooling prey taxa (benthic fish,as well as the epipelagic amphipod Themisto).Assuming these resources are relatively stable fromyear to year, it appears that when schooling fish arenot adequately available, the Coats Island murrepopulation diversifies its diet to include a variety ofprey types, but especially those associated with bot-tom habitat. A similar pattern was observed in com-mon murre Uria aalge diets in California (Ainley &Boekelheide 1990, Ainley et al. 1996). These obser-vations support theoretical predictions that diet willdiversify as preferred species become harder to find(Estabrook & Dunham 1976, Pyke et al. 1977, Pyke1984). The fact that diversity has declined over timeapparently reflects the fact that Mallotus has contin-ued to increase in the diet.

The most important parameter associated withvariation in diet, as measured by loadings on the firstprincipal component (Table 3), was the proportion ofsecondary prey items (benthic fish and inverte-brates). A similar analysis examining average energycontent delivered to offspring per year (Table 3 inSmith & Gaston 2012) also showed a strong loadingof invertebrates and benthic prey on PC1, suggestingthat high levels of secondary prey items are associ-ated with low-energy delivery rates. The second-and third-most important parameters associatedwith variation in diet, as measured by loadings on thesecond and third principal components (Table 3),were the proportion of the schooling prey items Boreogadus, Mallotus and Ammodytes. Years thatwere low in Boreogadus tended to be high in bothMallotus and Ammodytes (PC2), but, within low- Boreogadus years (PC3), years with a high propor-tion of Ammodytes were characterized as also havinga high proportion of deep-water secondary species(Gonatus and Eumesogrammus), and years with ahigh proportion of Mallotus also had a high propor-tion of shallow-water secondary species (Gymnelus).In high-Mallotus years gravid Mallotus were ob -served being fed to nestlings (Elliott et al. 2009a).Perhaps foraging behaviour during high-Mallotusyears involved the capture of spawning fish and asso-ciated foraging in shallow, near-shore environments,with the consequence that other shallow-water preywere captured disproportionately. In contrast, forag-ing behaviour during high-Ammodytes years mayhave involved foraging in deeper water for open-water schools.

219

Mar Ecol Prog Ser 513: 211223, 2014

Several benthic fish species are preyed on by spe-cialist individuals which use particular feedingstrategies (Woo et al. 2008, Elliott et al. 2008, 2009b).These birds maintain their specialization across years(Woo et al. 2008), and there is no evidence that, interms of total calories received, nestling nourishmentdiffers among the various prey specialties, or be -tween specialists and generalists. Consequently, thelowering of chick growth rates in years when ahigher proportion of benthic fish is being deliveredcould be the result of non-specialists turning to benthic prey when schooling prey is comparativelyscarce in the chick-rearing period and being less suc-cessful at finding such prey. Such conditions appearto be most likely when hatching is delayed relative tothe timing of 50% ice cover and when local ice coveris low. All the schooling prey showed a positive rela-tionship with local ice cover, whereas benthic fishhad a negative relationship (Table 1).

We document a change in marine communitieswithin Hudson Bay associated with changing localconditions. Although a few long-distance migrants,such as orcas Orcinus orca and razorbills Alca tordamay be able to directly and quickly respond tochanging conditions by moving into the bay to breed(Gaston & Woo 2008, Higdon & Ferguson 2009),many of the changes within the relatively isolatedmarine community must occur through variation inthe abundance of pre-existing community members(e.g. an American eel Anguilla rostrata in theSt Lawrence estuary cannot immediately take advan-tage of ideal oceanographic conditions in JamesBay). As such, murre diets at Coats Island do notdemonstrate the arrival of new prey, but rather achange in the proportion of community membersalready present, in particular, a shift from Bore-ogadus to Mallotus and Ammodytes. Mallotus domi-nates beluga Delphinapterus leucas diet in southernHudson Bay (Kelley et al. 2010), and Ammodytes hasdominated ringed seal Phoca hispida diet in westernHudson Bay since at least 1998 (Chambellant 2010).The changes seen at Coats Island may therefore represent a northward spread of Mallotus andAmmodytes, or of their spawning grounds, withinHudson Bay.

The switch away from Boreogadus occurred rap-idly and so far irreversibly in the mid-1990s, despitesubstantial short-term, year-to-year variation in iceconditions (Gaston et al. 2012; Table 1, Figs. 3 & 5).The increased use of capelin from the mid-1990s isconsistent with the movement towards the Arcticand/or an increase in abundance of a subarctic fishpreviously at the margins of its historical distribution,

under conditions of increasing ocean temperatures(Carscadden et al. 2013).

The step change in the mid-1990s impactednestling condition at departure, as the fledging massof murre nestlings declined steeply after the mid-1990s (Gaston et al. 2012). Mass at departure is correlated with subsequent survival to recruitment(Hipfner 2001). In a similar manner, the diet of manyseabirds in the northern Atlantic and Pacific haveshown stepwise changes, particularly, though notexclusively, in the abundance of Ammo dytes andMallotus (Barrett et al. 1997, Anderson & Piatt 1999,Barrett 2002, Montevecchi 2007, Hatch 2013). Thosechanges often reflected apparent warm-to-cold orcold-to-warm regime shifts in physical oceanography(Anderson & Piatt 1999, Montevecchi 2007, Hatch2013). The response of the biological community isoften rapid and difficult to reverse, once environmen-tal change reaches a tipping point (Overland et al.2010, Spencer et al. 2012). In our case, the cold-to-warm shift was clearly associated with a reduction inice cover, as ice is an important habitat for Bore-ogadus (Crawford & Jorgenson 1993, Gradinger &Bluhm 2004). Our dataset differs from datasetsderived from temperate systems in that (1) the cold-to-warm shift can be clearly linked to a change inphysical habitat (ice) and (2) some regime shiftsappear to have occurred periodically for centuries,but the mid-1990s shift in Hudson Bay forms part of ageneral trend in Arctic sea ice cover, possibly associ-ated with anthropogenic climate change (AMAP2011) and therefore unlikely to be reversed. Our abil-ity to detect the trends that we have described isgreatly enhanced by the availability of a wide rangeof secondary prey at Coats Island, a situation thatalso exists at other low Arctic murre colonies in Can-ada (Gaston 1985).

CONCLUSIONS

Compared with changes in the timing of ice break-up in Hudson Bay, change in the timing of breedingof murres Uria lomvia has been small and timing hasnot advanced since the step change in about 1995.The subsequent mismatch between ice conditionsand murre egg-laying has resulted in ice phase andlocal ice conditions being the most widespread fac-tors affecting diet variables. Both appear in top mod-els for 5 out of 6 taxa examined. When ice cover waslow on 2 July and hatching was late relative to icebreak-up, diet included a higher than average pro-portion of benthic fish and a lower than average pro-

220

Gaston & Elliott: Ice and seabird diet in Hudson Bay

portion of schooling fish. Diet diversity was inverselyrelated to the proportion of schooling fish. Chickmass at 14 d of age was inversely related to the pro-portion of benthic fishes in the diet, suggesting thatfeeding nestlings on these fish was sub-optimal forthe population as a whole, although, based on previ-ous work (Elliott et al. 2008, 2009a), individual spe-cialists can provision their nestlings adequately onbenthic fishes. Based on 14 d nestling mass as anindicator of diet quality, diet diversity increasedwhen feeding conditions were sub-optimal, support-ing previous observations of seabird diet breadth inrelation to reproductive success and other measuresof breeding conditions (e.g. Takahashi et al. 2001,Hedd et al. 2006, Waluda et al. 2012).

When diet analysis is based on the proportions ofdifferent taxa in the prey, it is inevitable thatincreases in one taxon will be at the expense of oneor more other taxa. Consequently, it seems appropri-ate to base interpretations on convergences anddivergences among prey taxa, rather than on individ-ual trends. The 2 most important taxa in nestlingdiets (Mallotus, Boreogadus) showed secular trends,even after environmental variables were taken intoaccount, suggesting that cumulative environmentaleffects were causing progressive changes in localmarine communities. There was no evidence for sec-ular trends in other diet elements. Inter-year varia-tion in the dominant diet taxa probably relates to rel-ative availability, but variation in secondary diet taxais probably driven by the availability of the dominanttaxa.

Acknowledgements. Research at Coats Island has beenfunded by Environment Canada with logistic support fromthe Polar Continental Shelf Project of Natural ResourcesCanada. K.H.E. received financial support through aNSERC Vanier Canada Graduate Scholarship, ACUNSGarfield Weston Northern Studies scholarship and AINAJennifer Robinson Scholarship. We thank all those numer-ous individuals who contributed to observations of fooddeliveries at Coats Island, especially Mark Hipfner andKerry Woo. Thanks to Paul Smith for statistical advice.

LITERATURE CITED

Ainley DG, Boekelheide RJ (1990) Seabirds of the FarallonIslands. Stanford University Press, Stanford, CA

Ainley DG, Spear LB, Allen SG, Ribic CA (1996) Temporaland spatial patterns in the diet of the common murre inCalifornia waters. Condor 98: 691705

Ainley DG, Ballard G, Barton KJ, Karl BJ, Rau GH, Ribic CA,Wilson PR (2003) Spatial and temporal variation of dietwithin a presumed metapopulation of Adlie penguins.Condor 105: 95106

AMAP (Arctic Monitoring and Assessment Programme)(2011) Snow, water, ice and permafrost in the Arctic(SWIPA): climate change and the cryosphere. AMAP,Oslo

Anderson DR (2008) Model based inference in the life sci-ences: a primer on evidence. Springer, New York, NY

Anderson PJ, Piatt JF (1999) Community reorganization inthe Gulf of Alaska following ocean climate regime shift.Mar Ecol Prog Ser 189: 117123

Barrett RT (2002) Atlantic puffin Fratercula arctica and com-mon guillemot Uria aalge chick diet and growth as indi-cators of fish stocks in the Barents Sea. Mar Ecol Prog Ser230: 275287

Barrett RT, Bakken V, Krasnov JV (1997) The diets of com-mon and Brnnichs guillemots Uria aalge and U. lomviain the Barents Sea region. Polar Res 16: 7384

Barrett RT, Camphuysen CJ, Anker-Nilssen T, Chardine JWand others (2007) Diet studies of seabirds: a review andrecommendations. ICES J Mar Sci 64: 16751691

Burnham KP, Anderson DR (1998) Model selection andinference: a practical information-theoretic approach.Springer-Verlag, New York, NY

Cairns DK (1987) Diet and foraging ecology of black guille-mots in northeastern Hudson Bay. Can J Zool 65: 12571263

Carscadden JE, Gjster H, Vilhjlmsson H (2013) A com-parison of recent changes in distribution of capelin(Mallo tus villosus) in the Barents Sea, around Icelandand in the Northwest Atlantic. Prog Oceanogr 114:6483

Chambellant M (2010) Hudson Bay ringed seal: ecology in awarming climate. In: Ferguson SH, Loseto LL, MalloryML (eds) A little less Arctic. Springer-Verlag, New York,NY, p 137158

Crawford RE, Jorgenson JK (1993) Schooling behaviour ofarctic cod, Boreogadus saida, in relation to drifting packice. Environ Biol Fishes 36: 345357

de Forest LN, Gaston AJ (1996) The effect of age and timingof breeding on reproductive success in the thick-billedmurre. Ecology 77: 15011511

Deguchi T, Watanuki Y, Niizuma Y, Nakata A (2004) Inter-annual variations of the occurrence of epipelagic fish inthe diets of the seabirds breeding on Teuri Island, north-ern Hokkaido, Japan. Prog Oceanogr 61: 267275

Elliott KH, Gaston AJ (2008) Masslength relationships forfishes delivered to nestling thick-billed murres Urialomvia at Coats Island, 19812006. Mar Ornithol 36: 2534

Elliott KH, Davoren GK, Gaston AJ (2008) Time allocationby a deep-diving bird reflects prey type and energy gain.Anim Behav 75: 13011310

Elliott KH, Bull R, Davoren GK, Gaston AJ (2009a) Under-water and above-water search patterns of an Arctic sea-bird: reduced searching at small spatiotemporal scales.Behav Ecol Sociobiol 63: 17731785

Elliott KH, Woo K, Gaston AJ (2009b) Specialization in murres: the story of eight specialists. Waterbirds 32: 491506

Estabrook GF, Dunham AE (1976) Optimal diet as a functionof absolute abundance, relative abundance, and relativevalue of available prey. Am Nat 110: 401413

Gagnon AS, Gough WA (2005) Trends in the dates of icefreeze-up and breakup over Hudson Bay, Canada. Arctic58: 370382

Gaston AJ (1985) The diet of thick-billed murre chicks in theeastern Canadian arctic. Auk 102: 727734

221

http://dx.doi.org/10.1006/jmsc.1997.0243http://dx.doi.org/10.1086/283076http://dx.doi.org/10.1675/063.032.0402http://dx.doi.org/10.1007/s00265-009-0801-yhttp://dx.doi.org/10.1016/j.anbehav.2007.09.024http://dx.doi.org/10.1016/j.pocean.2004.06.010http://dx.doi.org/10.2307/2265547http://dx.doi.org/10.3354/meps177115http://dx.doi.org/10.1007/BF00012412http://dx.doi.org/10.1139/z87-196http://dx.doi.org/10.1139/f93-213http://dx.doi.org/10.1093/icesjms/fsm152http://dx.doi.org/10.1111/j.1751-8369.1997.tb00249.xhttp://dx.doi.org/10.3354/meps230275http://dx.doi.org/10.3354/meps189117http://dx.doi.org/10.1650/0010-5422(2003)105[95%3ASATVOD]2.0.CO%3B2http://dx.doi.org/10.2307/1369852

Mar Ecol Prog Ser 513: 211223, 2014

Gaston AJ (2004) Seabirds: a natural history. Yale UniversityPress, New Haven, CT

Gaston AJ, Donaldson G (1996) Peat deposits and thick-billed murre colonies and Hudson Strait and northernHudson Bay: clues to post-glacial colonization of the areaby seabirds. Arctic 48: 354358

Gaston AJ, Hipfner JM (1998) The effect of ice conditions innorthern Hudson Bay on breeding thick-billed murres(Uria lomvia). Can J Zool 76: 480492

Gaston AJ, Hipfner JM (2000) Thick-billed murre (Urialomvia). In: Poole A, Gill F (eds) The birds of North Amer-ica, No. 497. The Birds of North America Inc., Philadel-phia, PA

Gaston AJ, Woo K (2008) Razorbills (Alca torda) follow sub-arctic prey into the Canadian Arctic: colonization resultsfrom climate change? Auk 125: 939942

Gaston AJ, Woo K, Hipfner JM (2003) Trends in forage fishpopulations in northern Hudson Bay since 1981, as deter-mined from the diet of nestling thick-billed murres Urialomvia. Arctic 56: 227233

Gaston AJ, Gilchrist HG, Mallory ML, Smith PA (2009)Changes in seasonal events, peak food availability andconsequent breeding adjustment in a marine bird: a caseof progressive mis-matching. Condor 111: 111119

Gaston AJ, Smith PA, Provencher J (2012) Discontinuouschange in ice cover in Hudson Bay in the 1990s and someconsequences for marine birds and their prey. ICES JMar Sci 69: 12181225

Gradinger RR, Bluhm BA (2004) In situ observations on thedistribution and behaviour of amphipods and Arctic cod(Boreogadus saida) under the sea ice of the High ArcticCanada Basin. Polar Biol 27: 595603

Hatch SA (2013) Kittiwake diets and chick production signala 2008 regime shift in the Northeast Pacific. Mar EcolProg Ser 477: 271284

Hedd A, Bertram DF, Ryder JL, Jones IL (2006) Effects ofinterdecadal climate variability on marine trophic inter-actions: rhinoceros auklets and their fish prey. Mar EcolProg Ser 309: 263278

Higdon JW, Ferguson SH (2009) Loss of Arctic sea ice causing punctuated change in sightings of killer whales(Orcinus orca) over the past century. Ecol Appl 19: 13651375

Hill MO (1973) Diversity and evenness: a unifying notationand its consequences. Ecology 54: 427432

Hipfner JM (2001) Fitness-related consequences of relayingin an Arctic seabird: survival of offspring to recruitmentage. Auk 118: 10761080

Hipfner JM, Gaston AJ, Gilchrist HG (2005) Variation inegg-size and laying date in thick-billed murre popula-tions breeding in the Low Arctic and High Arctic. Condor107: 657664

Hochheim KP, Barber DG (2010) Atmospheric forcing ofsea ice in Hudson Bay during the fall period, 19802005.J Geophys Res 115: C05009, doi: 10.1029/2009JC005334

Hochheim K, Barber DG, Lukovich JV (2010) Changing seaice conditions in Hudson Bay, 19802005. In: FergusonSH, Loseto LL, Mallory ML (eds) A little less Arctic.Springer-Verlag, New York, NY, p 3951

Hochheim KP, Lukovich JV, Barber DG (2011) Atmosphericforcing of sea ice in Hudson Bay during the spring period19802005. J Mar Syst 88: 476487

Kelley TC, Loseto LL, Stewart REA, Yurkowski M, FergusonSH (2010) Importance of eating capelin: unique dietaryhabits of Hudson Bay beluga. In: Ferguson SH, Loseto

LL, Mallory ML (eds) A little less Arctic. Springer- Verlag,New York, NY, p 5369

Maxwell JB (1986) A climate overview of the Canadianinland seas. In: Martini IP (ed) Canadian inland seas.Elsevier Oceanography Series 44. Elsevier Science, Ams-terdam, p 7999

Montevecchi WA (2007) Binary dietary responses of north-ern gannets Sula bassana indicate changing food weband oceanographic conditions. Mar Ecol Prog Ser 352: 213220

Overland JE, Alheit J, Bakun A, Hurrell JW, Mackas DL,Miller AJ (2010) Climate controls on marine ecosystemsand fish populations. J Mar Syst 79: 305315

Piatt JF, Harding AMA, Shultz M, Speckman SG, van PeltTI, Drew GS, Kettle AB (2007) Seabirds as indicators ofmarine food supplies: Cairns revisited. Mar Ecol Prog Ser352: 221234

Prinsenberg SJ (1986) On the physical oceanography ofFoxe Basin. In: Martini IP (ed) Canadian inland seas.Elsevier, Amsterdam, p 217236

Provencher JF, Elliott KH, Gaston AJ, Braune BM (2013)Networks of prey specialization in an Arctic mono -morphic seabird. J Avian Biol 44:551560

Pyke GH (1984) Optimal foraging theory: a critical review.Annu Rev Ecol Syst 15: 523575

Pyke GH, Pulliam HR, Charnov EL (1977) Optimal foraging: a selective review of theory and tests. Q Rev Biol 52: 137154

Renner HM, Mueter F, Drummond BA, Warzybok JA, Sinclair EH (2012) Patterns of change in diets of two piscivorous seabird species during 35 years in the Pri-bilof Islands. Deep-Sea Res II 6570(Spec Issue): 273291

Rhland KM, Paterson AM, Keller W, Michelutti N, Smol JP(2013) Global warming triggers the loss of a key Arcticrefugium. Proc R Soc Lond B Biol Sci 280: 20131887

Schrimpf MB, Parrish JK, Pearson SF (2012) Trade-offs inprey quality and quantity revealed through the behav-ioural compensation of breeding seabirds. Mar Ecol ProgSer 460: 247259

Scott WB, Scott MG (1988) Atlantic fishes of Canada. Uni-versity of Toronto Press

Shannon CE (1948) A mathematical theory of communica-tion. Bell System Tech J 27: 379423 and 623656

Smith PA, Gaston AJ (2012) Environmental variation and thedemography and diet of thick-billed murres. Mar EcolProg Ser 454: 237249

Spencer M, Mieszkowska N, Robinson LA, Simpson SD andothers (2012) Region-wide changes in marine ecosystemdynamics: state-space models to distinguish trends fromstep changes. Glob Change Biol 18: 12701281

Springer AM, Byrd GV, Iverson SJ (2007) Hot oceanogra-phy: planktivorous seabirds reveal ecosystem responsesto warming of the Bering Sea. Mar Ecol Prog Ser 352: 289297

Stammerjohn S, Massom R, Rind D, Martinson D (2012)Regions of rapid sea ice change: an inter-hemisphericseasonal comparison. Geophys Res Lett 39: L06501, doi: 10.1029/2012GL050874

Statsoft (2005) Statistica 7.1. Statsoft Inc., Tulsa, OKStewart DB, Barber DG (2010) The oceansea iceatmos-

phere system of the Hudson Bay complex. In: FergusonSH, Loseto LL, Mallory ML (eds) A little less Arctic.Springer-Verlag, New York, NY, p 137

Takahashi A, Kuroki M, Niizuma Y, Kato A, Saitoh S (2001)

222

http://dx.doi.org/10.1007/s002270100594http://dx.doi.org/10.1029/2012GL050874http://dx.doi.org/10.3354/meps07080http://dx.doi.org/10.1111/j.1365-2486.2011.02620.xhttp://dx.doi.org/10.3354/meps09589http://dx.doi.org/10.3354/meps09750http://dx.doi.org/10.1098/rspb.2013.1887http://dx.doi.org/10.1086/409852http://dx.doi.org/10.1146/annurev.es.15.110184.002515http://dx.doi.org/10.3354/meps07078http://dx.doi.org/10.1016/j.jmarsys.2008.12.009http://dx.doi.org/10.1006/jmsc.1997.0265http://dx.doi.org/10.3354/meps07075http://dx.doi.org/10.1016/j.jmarsys.2011.05.003http://dx.doi.org/10.1029/2009JC005334http://dx.doi.org/10.1650/0010-5422(2005)107[0657%3AVIESAL]2.0.CO%3B2http://dx.doi.org/10.1642/0004-8038(2001)118[1076%3AFRCORI]2.0.CO%3B2http://dx.doi.org/10.2307/1934352http://dx.doi.org/10.1890/07-1941.1http://dx.doi.org/10.3354/meps309263http://dx.doi.org/10.3354/meps10161http://dx.doi.org/10.1007/s00300-004-0630-4http://dx.doi.org/10.1093/icesjms/fss040http://dx.doi.org/10.1525/cond.2009.080077http://dx.doi.org/10.14430/arctic618http://dx.doi.org/10.1525/auk.2008.07195http://dx.doi.org/10.1139/z97-222

Gaston & Elliott: Ice and seabird diet in Hudson Bay

Importance of the Japanese anchovy (Engraulis japoni-cus) to breeding rhinoceros auklets (Cerorhinca mono-cerata) on Teuri Island, Sea of Japan. Mar Biol 139: 361371

Thaxter CB, Daunt F, Grmillet D, Harris MP and others(2013) Modelling the effects of prey size and distributionon prey capture rates of two sympatric marine predators.PLoS ONE 8: e79915

Waluda CM, Hill SL, Peat HJ, Trathan PN (2012) Diet vari-ability and reproductive performance of macaroni pen-guins Eudyptes chrysolophus at Bird Island, South Geor-gia. Mar Ecol Prog Ser 466: 261274

Woo KJ, Elliott KH, Davidson M, Gaston AJ, Davoren G(2008) Individual specialization in diet by a generalistmarine predator reflects specialization in foraging.J Anim Ecol 77: 10821091

223

Editorial responsibility: Yves Cherel, Villiers-en-Bois, France

Submitted: December 6, 2013; Accepted: July 11, 2014Proofs received from author(s): September 26, 2014

http://dx.doi.org/10.1111/j.1365-2656.2008.01429.xhttp://dx.doi.org/10.3354/meps09930http://dx.doi.org/10.1371/journal.pone.0079915

cite43: cite28: cite5: cite14: cite42: cite3: cite27: cite13: cite1: cite41: cite12: cite40: cite25: cite38: cite11: cite24: cite37: cite8: cite23: cite51: cite36: cite6: cite49: cite22: cite35: cite4: cite48: cite21: cite2: cite19: cite20: cite33: cite18: cite46: cite32: cite45: cite31: cite16: cite9: cite29: cite44: cite30: cite15: