Patterns of change in diets of two piscivorous seabird species during 35 years in the Pribilof...

Deep-Sea Research II 65-70 (2012) 273–291

Contents lists available at SciVerse ScienceDirect

Deep-Sea Research II

0967-06

doi:10.1

n Corr

E-m

journal homepage: www.elsevier.com/locate/dsr2

Patterns of change in diets of two piscivorous seabird speciesduring 35 years in the Pribilof Islands

Heather M. Renner a,n, Franz Mueter b, Brie A. Drummond a, John A. Warzybok a, Elizabeth H. Sinclair c

a Alaska Maritime National Wildlife Refuge, US Fish and Wildlife Service, 95 Sterling Hwy, Suite 1, Homer, AK 99603, USAb School of Fisheries and Ocean Sciences, University of Alaska, 315 Lena Point, 17101 Pt. Lena Loop Rd., Juneau, AK 99801, USAc NOAA/National Marine Fisheries Service, National Marine Mammal Laboratory, 7600 Sand Point Way NE, Seattle, WA 98115, USA

a r t i c l e i n f o

Available online 21 February 2012

Keywords:

Marine birds

Diets

Black-legged kittiwake

Thick-billed murre

Walleye Pollock

Arctic oscillation Index

Pribilof Islands

Bering Sea

Alaska

USA

45/$ - see front matter Published by Elsevier

016/j.dsr2.2012.02.014

esponding author. Tel.: þ1 907 226 4623; fax

ail address: [email protected] (H.M. Re

a b s t r a c t

As upper level predators in the marine ecosystem, seabirds reflect fluctuations in the marine

environment that influence their prey supply. Studies of seabird diets thus provide insight into the

physical and biological mechanisms that potentially drive population changes in both predators and

their prey. The eastern Bering Sea shelf, among the most productive marine ecosystems in the world,

has undergone significant restructuring in recent decades that is likely to continue in light of

anticipated climatic change. Using a dataset spanning 35 years at two of the Pribilof Islands in the

Bering Sea, we examined temporal patterns in diet and their relationships with oceanographic variables

for black-legged kittiwakes (Rissa tridactyla) and thick-billed murres (Uria lomvia), two piscivorous

seabirds with differing foraging strategies. Diets varied significantly among years and between islands

and species. Our substantially expanded dataset supported conclusions found in previous studies of this

system, including the importance of pollock, particularly age-0 class, in kittiwake diets and the absence

of capelin in diets of either species since the late 1970s. Diets of both species contained more gadids at

St. Paul Island and more squid and euphausiids at St. George Island, likely reflecting differences in

foraging location between islands. We found some relationships between kittiwake diet and broad-

scale oceanographic variables (Arctic Oscillation Index and regional summer sea surface temperature)

but not with local physical variables. Almost no time series data exist on availability and abundance of

zooplankton or forage fish species such as age-0 pollock, myctophids, and sandlance in the eastern

Bering Sea. Our measure of diet (number of individuals within each prey type) appears too coarse for

detecting complex relationships between local oceanographic variables and seabird responses, but may

provide invaluable information about changes in forage fish stocks, which are frequently expensive or

difficult to otherwise measure. Future diet analyses should increase emphasis on evaluating caloric

input (i.e., size and nutrient composition of each prey type) as well as attempts to measure the diet of

murre chicks.

Published by Elsevier Ltd.

1. Introduction

Studies of the foraging habits of marine predators can signalphysical and biological oscillations in the marine environment(Cairns, 1987; Montevecchi, 1993, 2007) that are otherwisedifficult or expensive to measure (Byrd et al., 2008a). Foragingstudies also provide insight into the factors driving demographicchange of the predators under study.

Seabirds have high energy requirements making them especiallysensitive to changes in the environment that affect prey availabilityand abundance (Piatt et al., 2007). Unlike the majority of marinespecies, seabirds are conspicuous and readily accessible to observers

Ltd.

: þ1 907 235 7783.

nner).

at large breeding colonies, making them ideal sentinels of change inthe marine ecosystem. The Pribilof Islands are home to a largemixed-species colony of seabirds, providing an excellent opportu-nity to investigate variation in this region of the North Pacific.

The eastern Bering Sea shelf, including the Pribilof Islands, isamong the most productive marine ecosystems in the world, withresources integral to the cultural and dietary components of AlaskaNatives, and fisheries that comprise �50% of the US commercialharvest by weight (NMFS, 2010). The Bering Sea shelf has changedsignificantly in recent decades, both physically and biologically(Grebmeier et al., 2006; Hunt et al., 2011), and is anticipated tochange even more in light of predictions of warming waters and areduction in the shelf’s seasonal ice cover (IPCC, 2007). The mostrecent decade was characterized by a multi-year period of unu-sually warm conditions (2001–2005) followed by a rapid transitionto unusually cold conditions (2007–2010), which had far-reaching

www.elsevier.com/locate/dsr2

www.elsevier.com/locate/dsr2

dx.doi.org/10.1016/j.dsr2.2012.02.014

mailto:[email protected]

dx.doi.org/10.1016/j.dsr2.2012.02.014

H.M. Renner et al. / Deep-Sea Research II 65-70 (2012) 273–291274

effects on zooplankton and fish populations (Coyle et al., 2011;Hunt et al., 2011; Stabeno et al., 2012). A number of focusedmultidisciplinary research programs have emphasized the study ofzooplankton and forage fish availability over the past two decades(i.e., FOCI [1984–present], SMMOCCI [1995–present], SEBSCC[1996–2002], GLOBEC [1990–present], BEST-BSIERP Bering SeaProject [2007–present]). These have resulted in some short-termdirected studies of forage fishes such as juvenile pollock andmyctophids (i.e., Sinclair and Stabeno, 2002; Wyllie-Echeverria,1996) as well as retrospective efforts based on forage fishesincidentally retained in bottom trawl surveys (Brodeur et al.,1997; Lauth, 2010). However, intensive long-term monitoringefforts to assess prey abundance and availability, such as thoseconducted by Russian researchers in the western Bering Sea(Sinclair et al., 1999), are lacking for both commercial and non-commercial fauna in the eastern Bering Sea ecosystem. In contrast,seabird diets in the Pribilof Islands have been monitored since1975, offering an opportunity to assess long-term changes inavailability of forage species important to upper trophic levelspecies, including commercial fishes. This long-term perspectivecan help elucidate the mechanisms that drive variability in preyavailability to seabirds and, potentially, to overall ecosystemproductivity.

This study focuses on diets of two apex predator seabirdspecies with different foraging strategies and constraints: black-legged kittiwakes (Rissa tridactyla) and thick-billed murres (Uria

lomvia). Both species are central place foragers during the nestingseason, raising chicks on narrow ledges on vertical sea cliffs andmaking daily foraging trips to feed young. The foraging capabil-ities of the two species, however, are quite different. Kittiwakes(small gulls) can only feed at or near the surface, but they areefficient flyers and able to store prey in their crop (Hatch et al.,2009). They make long foraging trips and usually deliver foodloads with multiple prey items to their chicks. In contrast, murresobtain their prey on extremely deep dives (regularly 4100 m;Gaston and Hipfner, 2000) but they are poor flyers with high wingloading and can only carry prey in their bill. Thus, they makeshorter foraging trips and are more limited in the amount of foodthey can deliver to their young with each trip. Murres generallyreturn from foraging trips with a single, large prey item toprovision their chick while sustaining themselves on differentand often considerably smaller prey items (Mehlum, 2001).

Kittiwakes and murres have been studied at their nestingcolonies in the Pribilof Islands since 1975 by a variety ofresearchers. Both species are now included in an ongoing annualseabird monitoring program operated by the Alaska MaritimeNational Wildlife Refuge (AMNWR; Dragoo et al., 2010), whichcollects data on seabird diets and a number of demographicparameters, including reproductive success, population trends,and survival. In 2008–2009, detailed new data became availableon both species, including foraging tracks for successful breeders,at-sea surveys of adult foraging distribution, and focused preysurveys.

In this paper we present a diet dataset spanning 35 years forpiscivorous black-legged kittiwakes and thick-billed murres ontwo islands in the Bering Sea. We describe diet patterns from1975 to 2009, comparing and contrasting St. Paul and St. Georgeislands, and the two species. We also evaluate relationshipsbetween adult diet composition and variables describing envir-onmental conditions and availability of prey.

In contrast to most seabird diet studies in this system, whichused diet data from the entire summer, we limited our analyses todiets from the chick-rearing portion of the breeding season. Dietsamples collected at or near the breeding colony during otherstages of the breeding season may not be comparable due totemporal variation in prey availability resulting from differences

in foraging behavior, as seabird energetic constraints can varysubstantially during the breeding season (Shaffer et al., 2003).During the incubation period, adults need only to feed themselvesand return to the colony to relieve their mate at less frequentintervals than after the chicks hatch (Hamer et al., 1993). Duringthe chick-rearing period, energy demands on adult seabirdsincrease dramatically (Golet et al., 2000a) as birds must not onlyfeed themselves but also bring energy-rich prey items back totheir young on a frequent schedule. Consequently, some seabirdsmay utilize different foraging strategies or collect different preyitems when provisioning chicks compared to when birds are justfeeding themselves (Ito et al., 2010). Therefore, we used only dietsamples collected during the period when birds were provision-ing chicks.

2. Study area

The Pribilof Archipelago is located in the southeastern BeringSea (�571N, 1691W) near the edge of the continental shelf.The two main islands, St. George and St. Paul islands, are situatedwithin the Middle Shelf Domain (characterized by water depthsbetween 50 and 100 m) but unique oceanographic features placethe island group in its own Pribilof Domain (Sullivan et al., 2008).St. George Island is approximately 65 km south of St. Paul Islandand is closer to the deeper Outer Shelf Domain (depths between100 and 200 m) and to the edge of the continental shelf (shelfbreak) and the pelagic zone of the Aleutian Basin (depthsexceeding 1000 m; Kinder and Schumacher, 1981; Stabenoet al., 1999).

The area along the shelf break has been referred to as the‘‘Green Belt’’ (Mizobata et al., 2008; Springer et al., 1996) becauseof high productivity resulting from dynamic physical oceano-graphic processes that enhance nutrient availability along theouter shelf and slope. This high productivity sustains largepopulations of forage fish species important to both commercialinterests (National Research Council, 1996) and to the marinemammals and seabirds that breed on the Pribilof Islands. Forexample, walleye pollock (Theragra chalcogramma), whichaccount for about 60% of the total landings of groundfish inAlaskan waters, are typically concentrated along the shelf breaksouth and northwest of the Pribilof Islands (Ianelli et al., 2010).

The Pribilof Islands support one of the largest breeding con-centrations of marine birds in the North Pacific region, estimatedto exceed 2 million individuals (Hickey and Craighead, 1977).St. George Island has approximately an order of magnitude moreledge-nesting seabirds than St. Paul Island (Hickey and Craighead,1977) due to more extensive cliff nesting habitat on St. GeorgeIsland, and possibly due to its closer proximity to the productiveshelf break (Byrd et al., 2008b). Black-legged kittiwakes and thick-billed murres are the most abundant species of piscivorousseabirds on the islands, where they aggregate with each otherand with their congeners, red-legged kittiwakes (R. brevirostris)and common murres (U. aalge).

3. Materials and methods

3.1. Data collection

Diet samples were collected from captured or shot birdsattending the cliff breeding colonies on St. Paul and St. Georgeislands, 1975–2009 (Table 1). Samples were collected from atleast one species on at least one island during 25 years, while 10scattered years had no data. Most sampling occurred during Julyand August, but some samples were collected as early as 15 April

Table 1Sample sizes and data sources used for analyses of seabird diets from the Pribilof Islands, 1975–2009. Data represent total samples collected each year (‘‘All’’) and those

samples collected during the chick-rearing period (‘‘Chick’’) only. Black-legged kittiwake (BLKI) data come from both adult and chick diet samples; thick-billed murre

(TBMU) data are adult samples only.

Year Species St. Paul Island St. George Island Total Source

Chick All Chick All Chick All

1975 BLKI 118 118 0 0 118 118 G.L. Hunt

TBMU 11 19 0 0 11 19

1976 BLKI 55 56 31 31 86 87 G.L. Hunt

TBMU 5 8 0 0 5 8

1977 BLKI 129 148 60 86 189 234 G.L. Hunt

TBMU 18 24 18 26 36 50

1978 BLKI 70 90 52 64 122 154 G.L. Hunt

TBMU 7 14 31 47 38 61

1979 BLKI 23 23 0 0 23 23 G.L. Hunt

TBMU 4 5 0 0 4 5

1981 BLKI 0 0 10 10 10 10 D.S. Lloyd

TBMU 0 0 8 8 8 8

1984 BLKI 29 29 67 67 96 96 M. Bradstreet

TBMU 10 10 42 42 52 52

1985 BLKI 4 5 12 35 16 40 STG: D.E. Dragoo

TBMU 0 0 15 40 15 40 STP: USFWS

1986 BLKI 0 0 38 56 38 56 D.E. Dragoo

TBMU 0 0 15 30 15 30

1987 BLKI 23 23 0 0 23 23 G.L. Hunt

TBMU 0 0 0 0 0 0

1988 BLKI 48 48 28 28 76 76 G.L. Hunt

TBMU 5 5 11 11 16 16

1992 BLKI 7 28 65 67 72 95 A.M. Springer

TBMU 0 25 24 52 24 77

1993 BLKI 8 19 11 24 19 43 STG: Univ of AK

TBMU 0 7 0 1 0 8 STP: A.M. Springer

1994 BLKI 0 27 5 11 5 38 STG: Univ of AK

TBMU 0 22 0 0 0 22 STP: A.M. Springer

1995 BLKI 0 22 0 0 0 22 A.M. Springer

TBMU 0 11 0 0 0 11

1997 BLKI 30 57 9 9 39 66 A.M. Springer

TBMU 0 8 5 5 5 13

1998 BLKI 30 43 39 39 69 82 A.M. Springer

TBMU 0 9 0 16 0 25

1999 BLKI 0 18 13 23 13 41 A.M. Springer

TBMU 0 8 16 24 16 32

2000 BLKI 5 14 9 18 14 32 A.M. Springer

TBMU 0 11 10 19 10 30

2001 BLKI 0 0 7 7 7 7 USFWS

TBMU 0 0 0 6 0 6

2003 BLKI 22 22 0 0 22 22 USFWS

TBMU 0 0 0 0 0 0

2006 BLKI 9 9 0 0 9 9 BLKI: USFWS

TBMU 0 0 15 21 15 21 TBMU: M. Ito

2007 BLKI 0 0 0 0 0 0 M. Ito

TBMU 0 0 14 14 14 14

2008 BLKI 33 33 36 36 69 69 USFWS

TBMU 29 29 27 27 56 56

2009 BLKI 22 22 24 24 46 46 USFWS

TBMU 48 48 30 30 78 78

Totals BLKI 665 854 516 635 1181 1489

TBMU 152 263 281 419 433 682

H.M. Renner et al. / Deep-Sea Research II 65-70 (2012) 273–291 275

and as late as 22 September; samples within the chick-rearingperiod ranged from 21 June to 22 September. Samples from black-legged kittiwakes consisted of stomachs from shot adult birds

(1975–1978, 1981, 1984–1988, 1992–2001), regurgitations fromcaptured adults (2008–2009) or chicks (1975–1978, 1984, 2003,2006, 2008–2009), and stomach contents from water-offloading


or ‘‘stomach pumping’’ adults (2008–2009; Barrett et al., 2007;Wilson, 1984). Samples from thick-billed murres consisted ofstomach contents from adult birds collected from 1975 to 2003,while water-offloading was used exclusively from 2006 forward.Collection time of samples across all years ranged throughout theday from early morning until dark.

Detailed prey processing protocols were not available for alldatasets, but generally each sample was thawed or drained ofpreservative, and prey were identified to the lowest taxonomiclevel possible using standard taxonomic keys and preservedreference material. Length was measured for intact prey itemsand estimated for partial prey items through measurement ofhard parts. Sagittal otoliths were removed from heads of fish toverify identity and (for juvenile pollock) to estimate fork length(and age class) using species-specific regression relationships ofotolith length vs. fork length (e.g., Frost and Lowry, 1981;Zeppelin et al., 2004). All loose otoliths and squid beaks wereextracted from diet samples to determine prey species andestimate the total number of that prey type in the sample (e.g.,a left and right matched pair of otoliths, or upper and lower squidbeaks of similar size, were assumed to represent one fish orsquid). In samples containing large numbers of loose otoliths ofsimilar size of a given species, the number of individual fish wasestimated to be half the number of otoliths. Numbers of squidwere estimated to be the greater of upper or lower beaks when allloose beaks could not be matched. Usually, otoliths were mea-sured (70.01 mm) under a microscope using an ocular micro-meter. In some cases, the total number of prey taken wasestimated by counting the number of identifiable parts (e.g., eyes,jaws, tails) and dividing by the number of those parts in a wholeorganism. To determine pollock age classes, we used a forklength cutoff of 80 mm to separate age-1 (Z80 mm) and age-0(o80 mm) pollock. This size was suggested to be more specificfor pollock in the vicinity of the Pribilof Islands in June, July, andearly August than other published mean sizes (T. Buckley and P.Ressler, AFSC, NOAA, pers. comm.). When pollock were partiallydigested, we estimated fish lengths using regression equationsbased on otolith measurements (Zeppelin et al., 2004). Some ofthe early data on otolith length and prey number used in thisstudy were not available to previous studies; the acquisition ofraw lab data sheets enabled these additional analyses.

3.2. Data analysis

Our diet data (Table 1) differ somewhat from those presentedpreviously by Decker et al. (1995), Hunt et al. (1996a), Sinclairet al. (2008), and Springer et al. (1986). In addition to accumulat-ing a number of additional years of diet data (2001–2009, andseven years within the period 1975–2000 covered by Sinclairet al. (2008)), we also restricted our use of diet samples, based ondate and sample type, slightly differently than other earlieranalyses (e.g., Sinclair et al., 2008; Vliestra et al., 2003). Allanalyses were limited to diet samples collected during thechick-rearing period (from hatching until chick fledging) becausewe found highly significant differences in diet between theincubation period and the chick-rearing season, similar to resultsfound for other species (e.g., Quillfeldt et al., 2010). We definedthe chick-rearing period for each species on each island for eachyear as beginning at the date of first hatch based on observedproductivity plots; when no observations were available in agiven year, we used the earliest recorded hatch date. Diet sampleswithin the chick-rearing period ranged from 21 June to 22September (differing from the range of 21 July–28 October usedby Sinclair et al. [2008]). We also restricted our dataset based onsample type, using only adult diet samples for murres becauseprey items delivered to murre chicks are frequently different than

those chosen by adults for self-feeding (Decker et al., 1995; Huntet al., 1996a; Mehlum, 2001). Although we recognize that chickdiet may be an important link to environmental correlates,unfortunately murre chick diet samples are difficult to collectand few data were available for an independent comparison (only181 samples across seven years, all prior to 1988). For kittiwakes,on the other hand, adult and chick diets were combined as inprevious studies (e.g., Decker et al., 1995; Hunt et al., 1996a;Sinclair et al., 2008) because it is not possible to distinguishbetween the two, given adults store prey in their crops for severalhours before regurgitating it to the young. Because the diet dataused in this paper were not collected and analyzed consistently inall years (e.g., biomass data collected in some but not all years,some prey items identified to species while others only to genusor family, etc.), we were limited in our diet data metric to eithernumerical abundance (how many individuals of a prey type werein each sample) or frequency of occurrence (how many samplescontained a given prey type, regardless of quantities). Bothmeasures have potential biases: abundance data underemphasizethe importance of fish and other larger diet items (a bird may eatonly a few larger fish but many smaller invertebrates), whereasfrequency of occurrence does not account for differences in thequantities of prey items across samples (a sample containing asingle individual is treated the same as a sample containinghundreds of individuals) and thus may underemphasize theimportance of more numerous diet items.

To compare adult diets between species, between locations,and over time using multivariate analyses, we used numericalabundances of prey types in diet samples (see Appendices A1 andB1). We repeated all multivariate analyses using frequency ofoccurrence, but conclusions were consistent with those usingabundances; thus we show results based on numerical abun-dances only. Potential biases in numerical data were reduced byfirst applying a fourth-root transformation to abundance data, tolessen the influence of relatively few samples with very high preycounts. Following Mueter and Norcross (2000), we then standar-dized abundance data to ensure that different prey types, whichmay vary greatly in quantity or size, receive equal weight in theanalysis. Data were standardized by dividing the abundance foreach sample by the maximum abundance for the correspondingprey type across all samples, thus scaling abundances for eachprey type from 0 to 1. Our rationale for the standardization wasthat the largest observed numbers of a given prey type likelyreflect the maximum ‘‘capacity’’ of a bird for this prey type. Bysetting the highest observed abundances for each prey type to 1,diets corresponding to full capacity receive the same weight,regardless of prey type. Thus standardized abundances should becomparable across prey types despite large differences in the sizeand individual quantities of different prey types in a sample. Inaddition to analyses of numerical abundances, we also examinedtrends in the frequency of occurrence of individual prey typesusing a logistic regression based on binomial presence/absencedata.

Prior to the analyses, prey items were grouped into ninetaxonomic groups or prey types (gadids, sandlance, myctophids,capelin, other fish, squid, amphipods, euphausiids, and otherinvertebrates). Unidentified gadids were assumed to be walleyepollock, as the only other identified gadid (Pacific cod Gadus

macrocephalus) occurred in just two years and in o1% of samples.Fish that were classified as ‘‘unknown‘‘ were excluded rather thangrouped with ‘‘other fish’’ to avoid spurious patterns that wouldlikely result from a large number of gadids classified as unknownin some years. This could potentially affect our results as itunderestimates the abundance of some prey types in sampleswith a large proportion of unknown fish (however, less than 2% ofsamples had more than five unknown fish). In addition, we


examined variability in the prevalence of age-0 and age-1 walleyepollock in seabird diets over time for those years when they wereidentified. It would have been preferable to further separate allinvertebrate prey to the species level, as euphausiids, amphipods,and squid are known to have species-specific distributionaldifferences (Sinclair et al., 1999; K. Benoit-Bird, pers. comm.).However, separating them did not affect conclusions from themultivariate analyses and caused large uncertainties becausethere were many more samples with unidentified euphausiids,amphipods, and squid than samples in which they were identifiedto species. Thus we present results based on aggregated groupsonly (euphausiids, amphipods, squid); however, we investigatedthe proportions of continental shelf vs. slope taxa over timeby grouping all species into shelf, slope, or ‘unknown’ groups(K. Benoit-Bird and S. Heppell, pers. comm.).

Samples that contained unknown counts for any of the preytypes were treated in two different ways to examine the sensitivityof results to uncertainty in the actual counts: (1) All unknowncounts (represented in the data set as ‘‘present’’) were changed to‘‘1’’ prior to analysis, which is probably reasonable for fish, but likelyunderestimates the true abundances of euphausiids and otherinvertebrates. (2) All samples containing unknown counts wereeliminated from the analysis of abundance data (but not fromanalyses of presence/absence data); hence sample sizes may differbetween these analyses. The conclusions were consistent betweenthe two approaches and we show results for the first approach only.

To statistically test for overall differences in diet compositionbetween species, between locations, and among years, we usedpermutation-based analyses of variance as implemented inPERMANOVAþ for PRIMER (Anderson et al., 2008). Dissimilarityamong diet samples was assessed by computing pairwise Bray–Curtis dissimilarities (Bray and Curtis, 1959) of the transformedand standardized data. The Bray–Curtis measure is commonlyused for species abundance data as it has the desirable featurethat it is not influenced by joint zeros (i.e., prey types that do notoccur in either sample have no influence on the dissimilaritymeasure). Samples with no diet items were eliminated from theanalysis because they provide no basis for comparing diets andbecause the Bray–Curtis measure of dissimilarity is not defined ifall counts are zero. Overall PERMANOVA analyses were followedby pairwise comparisons among levels of a variable for thosevariables that had more than two levels (e.g., years). PERMANOVAanalyses were based on only those years where both locationswere sampled, to avoid biasing the results.

We graphically assessed differences in diets between islandsand among years using ordination plots based on non-metric

Table 2Prey variables used in analyses with diet composition of black-legged kittiwakes and t

Variable Description

Local scaleAge-1 pollock abundance

bottom

Index of age-1 pollock abundance on the bottom with

surveys (1982–present; data provided by R. Lauth, AF

Age-1 pollock abundance

mid-water

Index of age-1 pollock abundance in midwater based

2006–present only; data provided by P. Walline and P

Capelin abundance Index of capelin abundance on the bottom within 100

(1982–present; data provided by R. Lauth, AFSC, NOA

Total forage fish abundance Index of total forage fish abundance (shrimp, herring, s

100 km radius of the Pribilof Islands, based on NMFS

NOAA, Seattle, pers. comm.).

Regional scaleAge-0 pollock abundance Estimated abundance of age-0 pollock in the Eastern B

et al., 2010) lagged one year to reflect age-0 abundan

Age-1 pollock abundance Estimated abundance of age-0 pollock in the Eastern B

et al., 2010).

multidimensional scaling (NMDS). Abundances of each prey typewere averaged across samples within a given year and location.The resulting matrices of average diet composition were fourth-root transformed and standardized to species maximum asdescribed above for the sample-by-species matrix. Pairwise dis-similarities among diets aggregated by year and island werequantified using the Bray–Curtis measure of dissimilarity, andNMDS was used to approximate the Bray–Curtis dissimilarities intwo dimensions to display similarities among diet samples in anordination diagram. To interpret the ordination axes in terms ofthe underlying prey abundances, we computed Spearman rankcorrelations between the axis scores and the fourth-root trans-formed abundances of individual prey types. To further test forand illustrate differences in prey composition, we modeledindices of prey composition (i.e., axis scores for axis 1 and axis2 of the NMDS ordination) as a function of year and location usingan analysis of variance weighted by the number of samples ineach year/location combination.

To investigate temporal trends in the importance of individualprey types in seabird diets, we estimated the proportion of dietsamples containing a given prey type in a given year using alogistic regression based on presence/absence data by species andisland. We estimated the probability, pt, that a sample in year t

contains a given prey type (or, equivalently, the fraction ofsamples containing the prey type) and the associated uncertainty.Coefficients and their standard errors were estimated on the logitscale {log{pt/(1�pt)} and were back-transformed to probabilities(0–1) for plotting the estimated proportions with associatedconfidence intervals. No confidence intervals can be computed ifa given prey type does not occur in any sample (pt¼0) or occurs inall samples (pt¼1). Because diets and sampling years differedbetween St. George and St. Paul islands, we quantified time trendsin diet composition for each island separately.

Finally, we related diet composition to measures of preyavailability and physical environmental variables using the BESTprocedure in PRIMER v6 (Clarke and Gorley, 2006). Prey avail-ability variables were selected to describe abundances of foragefish species at local and regional scales (Table 2). Environmentalvariables were chosen to reflect spring and summer conditions atlocal, regional, and large basin-wide scales (Table 3). All local andregional temperature and ice indices were moderately to stronglycorrelated with each other (rZ0.5), but weakly correlated withwind indices and large-scale indices (rr0.5). The BEST procedureuses a Mantel-type test statistic to identify the subset of environ-mental and prey variables that results in the highest correlationbetween pairwise similarities in average diet composition among

hick-billed murres at the Pribilof Islands.

in 100 km radius of the Pribilof Islands, based on NMFS summer bottom trawl

SC, NOAA, Seattle, pers. comm.).

on NMFS echo-integration trawl survey (1996, 1997, 1999, 2000, 2002, 2004,

. Ressler, AFSC, NOAA, Seattle, pers. comm.).

km radius of the Pribilof Islands, based on NMFS summer bottom trawl surveys

A, Seattle, pers. comm.).

andfish, eulachon, capelin, sandlance, and juvenile pollock) on the bottom within

summer bottom trawl surveys (1982–present; data provided by R. Lauth, AFSC,

ering Sea, based on age-1 recruitment data from stock assessments (from Ianelli

ce.

ering Sea, based on age-1 recruitment data from stock assessments (from Ianelli

Table 3Physical environmental variables used in analyses with diet composition of black-legged kittiwakes and thick-billed murres at the Pribilof Islands.

Variable Description

Local scaleSummer SST Sea surface temperature (� July/August) measured within a 100 km radius around the Pribilof Islands measured during NMFS bottom

trawl surveys (1982–present; temperature data provided by R. Lauth, AFSC, NOAA, Seattle, pers. comm.).

Spring transition index Based on air temperatures measured at St. Paul Island airport, defined as the first day when smoothed daily air temperatures (using

LOWESS smoother, Cleveland et al., 1992) first exceeded 3 1C (data from the NOAA National Climatic Data Center;

www.ncdc.noaa.gov/oa/ncdc.html).

Summer wind mixing index Defined as mean daily wind speeds cubed at 10 m height at 571N, 1691W (just east of the Pribilof Islands), averaged over the period

1 June–31 August (data from the NOAA Bering Climate website, www.beringclimate.noaa.gov).

Strong wind event index Defined as the number of days with wind speeds in excess of 9 m/s (same location/time period as wind mixing; data from the NOAA

Bering Climate website, www.beringclimate.noaa.gov).

Regional scaleRegional summer SST Regional index of average sea surface temperature over the Eastern Bering Sea shelf during the summer breeding season (June–

August), based on measured temperatures during the NMFS summer bottom trawl survey (1982–present; temperature data provided

by R. Lauth, AFSC, NOAA, Seattle, pers. comm.). Index is adjusted for differences in the mean date of sampling as described in Mueter

et al., 2006).

Regional summer SST 2 Longer time series for regional index of average sea surface temperature over the Eastern Bering Sea shelf during the summer breeding

season (June through August), based on NOAA’s extended reconstructed SST data over the approximate trawl survey region (inshore of

approximately 200 m depth contour between 551N and 611N).

Spring ice retreat index Defined as the number of days with ice cover after 15 March in the vicinity of Mooring 2 site (56.91N, 164.11W) as a proxy for the

timing of ice retreat from the southeast Bering Sea shelf (data from the NOAA Bering Climate website, www.beringclimate.noaa.gov).

Cold pool index The spatial extent of the cold pool over the Eastern Bering Sea shelf, estimated as the fraction of the NMFS trawl survey region with

water temperatures below 2 1C during the survey period (1982–present; temperature data provided by R. Lauth, AFSC, NOAA, Seattle,

pers. comm.).

Cold pool index 2 Longer time series for cold pool index, based on NOAA’s extended reconstructed SST data. Index is based on a regression of bottom

temperatures on the average March–May SSTs (i.e., when the water column is vertically mixed, hence SSTs and BTs are identical).

Basin-wide scalePacific Decadal Oscillation

Index (PDO)

November–March large-scale index related to climate variability in the eastern Bering Sea. Monthly values obtained from the Joint

Institute of the Atmosphere and Oceans, University of Washington (http://jisao.washington.edu/pdo/PDO.latest).

Arctic Oscillation Index (AO) January–March large-scale index related to climate variability in the eastern Bering Sea. Monthly values obtained from NOAA,

National Weather Service, Climate Prediction Center (http://www.cpc.ncep.noaa.gov/).


years and the corresponding pairwise similarities in environmen-tal/prey conditions among years. The correlations were testedfor significance using an appropriate randomization procedure(Clarke and Gorley, 2006). Because complete data are required forthese tests (no missing values), we tested different sets ofvariables over different time periods to maximize the number ofyears used in the analysis (using fewer variables) or the numberof variables (using fewer years). Finally, we used multivariateregression to quantify the amount of variation in diet explainedby the variables identified by the BEST procedure.

4. Results

4.1. Overall comparison of diets between species, between

locations, and over time

There were obvious and highly significant differences in dietcomposition during the chick-rearing period between the twospecies based on a three-way PERMANOVA, with species, location,and year as explanatory variables and using only those six yearswith complete data (i.e., both species were sampled at bothlocations). All main effects and interactions were highly signifi-cant (po0.001), indicating significant differences in diet compo-sition between the two species, which accounted for the largestcomponent of variation (F¼44.7, po0.001), between locations(F¼15.7, po0.001), and among years (F¼8.57, po0.001). Sig-nificant interactions further suggested that the differencesbetween species were not consistent at both locations or indifferent years and that differences between locations were notthe same each year. Because of the large and significant differ-ences between species, all further analyses were done by speciesto examine differences between locations and among years.

4.2. General diet comparisons between locations and among years

by species

4.2.1. Black-legged kittiwakes

The diet composition of black-legged kittiwakes differed sig-nificantly between St. George and St. Paul islands and showedstrong and significant differences over time (n¼13 years whenboth islands were sampled; Fig. 1). Interannual variability in preycomposition was considerably larger than variability between thetwo islands. A significant interaction (Pseudo-F¼3.40, po0.001)suggests that the differences between the two islands were notconsistent over time. This is also evident in the NMDS ordination,which showed that diets were very similar between St. Georgeand St. Paul islands in most years (e.g., 1977, 1998), but differedgreatly between islands in some years (e.g., 2009).

Axis 1 (x-axis) in the NMDS plot primarily described a gradientin prey composition of sandlance, myctophids, and squid, rangingfrom many sandlance but few myctophids and squid on the leftside of the plot (small values of first axis scores) to the oppositepattern on the right side of the plot (large values of axis 1 scores;Table 4, Fig. 1). This axis clearly separates St. Paul Island fromSt. George Island, indicating a higher prevalence of myctophids andsquid in St. George Island diets and a higher prevalence ofsandlance at St. Paul Island (Fig. 1). Axis 2 (y-axis) in the NMDSplot described a gradient in abundance of amphipods, euphausiids,capelin, gadids, and myctophids, ranging from many amphipods(both shelf and slope), euphausiids (both shelf and slope), andcapelin but few gadids and myctophids at the top of the plot (largevalues of second axis scores), to the opposite pattern at the bottomof the plot (small values of second axis scores; Table 4, Fig. 1).

The weighted ANOVA model for axis 1 (R2¼84%) showed

significant differences among years (F¼2.53, p¼0.049) andbetween locations (F¼8.56, p¼0.012; Fig. 2). Axis scores weresignificantly higher, reflecting high sandlance and low myctophid

http://jisao.washington.edu/pdo/PDO

http://www.cpc.ncep.noaa.gov/

Table 4Spearman rank correlations between NMDS axes 1 and 2 and abundance of individual taxonomic groups in black-legged kittiwake and thick-billed murre diets.

Black-legged kittiwake Thick-billed murre

Prey item Axis 1 Axis 2 Axis 1 Axis 2

rho pa rho p rho p rho p

Total gadid 0.078 0.650 �0.407 0.014nn�0.236 0.268 0.318 0.130

Sandlance 0.846 0.000nn 0.157 0.360 �0.655 0.001nn�0.191 0.371

Myctophid �0.700 0.000nn�0.357 0.033nn

�0.654 0.001nn 0.291 0.167

Capelin �0.137 0.426 0.354 0.034nn�0.537 0.007nn 0.023 0.916

Other fish 0.115 0.503 �0.109 0.527 �0.126 0.558 0.331 0.114

Squid �0.454 0.005nn 0.261 0.124 �0.294 0.163 �0.813 0.000nn

Amphipod �0.214 0.211 0.722 0.000nn�0.514 0.010nn 0.090 0.677

Euphausiid �0.089 0.605 0.679 0.000nn�0.150 0.485 �0.678 0.000nn

Other invertebrate �0.205 0.230 0.236 0.166 �0.755 0.000nn 0.427 0.037nn

an¼significant at the 0.1 level; nn

¼significant at the 0.05 level.

Fig. 1. NMDS ordination based on average diet compositions of black-legged kittiwakes by colony and year for all year/colony combinations based on samples collected

during the chick-rearing period. Distances between points are approximately proportional to Bray–Curtis measure of dissimilarity in diet composition. Vectors indicate

correlations between ordination axes and fourth-root transformed prey numbers; lengths of the vectors indicate magnitudes of correlations. Vector origins were arbitrarily

placed on the graph and correspond to x¼0, y¼0.


abundance in 2001, 2003, and 2006, and higher abundances atSt. Paul Island than at St. George Island. This suggests kittiwakesate more sandlance in those warm years and agrees with theinter-island patterns in the NMDS plot (Fig. 1), with kittiwakes atSt. Paul Island consuming more sandlance and those at St. GeorgeIsland consuming more myctophids. The weighted ANOVA modelfor axis 2 (R2

¼95%) showed highly significant differences amongyears (F¼11.2, po0.001) but not between locations (F¼0.073,p¼0.792; Fig. 2). Axis scores were higher, indicating greaternumbers of amphipods, euphausiids, and capelin through 1981,decreased throughout the 1990s, and increased again in 2008 and2009 (Fig. 2). This suggests amphipods and euphausiids wereprevalent in the 1970s and early 1980s, as well as in 2008 and2009; capelin, however, were only observed in significant num-bers during the 1970s.

4.2.2. Thick-billed murres

Murres were sampled on both islands in only six years. Thediet composition of thick-billed murres differed significantly

between St. George and St. Paul islands (F¼9.57, po0.001;Fig. 3) and also showed strong and significant differences amongyears (F¼5.69, po0.001). The component of variation attributedto inter-island differences (2 3 7) is similar in magnitude to thecomponent attributed to interannual differences (2 9 9). Thiscontrasts with black-legged kittiwakes, where differencesbetween the islands accounted for a much smaller componentof variation than differences among years (174 vs. 489), implyingthat inter-island differences are more pronounced for murres(Figs. 1 and 3).

A highly significant interaction (Pseudo-F¼6.43, po0.001)suggests that the differences between the two islands were notconsistent over time. This was also evident in the NMDS ordina-tion, which showed that diets were very similar betweenSt. George and St. Paul Islands in the early years (1977, 1978),but differed greatly between islands in other years (1984, 2008,2009). The main difference in diet composition between theislands was a relative preference for gadids at St. Paul Islandcompared to larger numbers of squid and euphausiids in dietsamples from St. George Island (Fig. 3).

1975 1985 1995 2005

-0.4

-0.2

0.0

0.2

0.4

0.6

0.8

Par

tial f

or Y

ear

STG STP

-0.4

-0.2

0.0

0.2

0.4

0.6

0.8

1975 1985 1995 2005

-0.4

-0.2

0.0

0.2

0.4

0.6

Par

tial f

or Y

ear

STG STP

-0.4

-0.2

0.0

0.2

0.4

0.6

Fig. 2. Partial fits and partial residuals from a weighted Analysis of Variance of indices of prey composition as a function of year and location. Estimates of the mean

response and 95% confidence interval of the mean response by year and location are shown for an ANOVA of axis 1 scores (top) and a separate ANOVA of axis 2 scores.

Fig. 3. NMDS ordination based on average diet compositions of thick-billed murres by colony and year for all year/colony combinations based on samples collected during

the chick-rearing period. For details see Fig. 1.


Axis 1 in the NMDS ordination of thick-billed murre dietsprimarily described a gradient in prey composition from manysandlance, myctophids, capelin, amphipods, and other inverte-brates on the left side of the plot (small values of first axis scores)

to the opposite pattern on the right side of the plot (small valuesof axis 1 scores; Table 4, Fig. 3). Axis 2 (y-axis) in the NMDS plotdescribed a gradient in abundance of squid, euphausiids, andother invertebrates in murre diets, ranging from many squid and


euphausiids but few other invertebrates at the bottom of the plot(small values of second axis scores), to the opposite pattern at thetop of the plot (large values of second axis scores; Table 4, Fig. 3).

The weighted ANOVA model for axis 1 (R2¼51%) showed no

significant differences among years (F¼2.49, p¼0.159) orbetween locations (F¼0.01, p¼0.940). The weighted ANOVAmodel for axis 2 (R2

¼81%) showed no significant differencesamong years (F¼0.54, p¼0.844) but a marginally significantdifference between locations (F¼5.39, p¼0.068). Axis scores werehigher, reflecting a greater proportion of squid and euphausiidsand fewer other invertebrates at St. George Island than at St. PaulIsland. These results differ from the PERMANOVA analysis, whichfound strong and significant differences among years and loca-tions. This difference is primarily due to inconsistency in theobserved diet differences among years, as evident by the interac-tion term (which accounts for the largest component of variation,excluding residual variation).

4.3. Temporal trends in diet composition


The proportion of gadids in kittiwake diets at St. Paul Islandgenerally decreased over time (but was relatively high in 2009),while that of sandlance increased (Fig. 4). Myctophids andsquid were most prevalent in the 1990s, amphipods were mostlyconsumed in the early years, and the occurrence of other taxawas variable without a clear trend. Diets at St. George Islandwere highly variable without obvious long-term trends (Fig. 5).One notable feature was the consistently high proportion of

1975 1985 1995 2005

0.0

0.4

0.8

0.0

0.4

0.8

0.0

0.4

0.8

0.0

0.4

0.8

0.0

0.4

0.8

0.0

0.4

0.8

gadid

1975 1985

sand

1975 1985 1995 2005

capelin

1975 1985

fis

1975 1985 1995 2005

amphipod

1975 1985

euph

Fig. 4. Probability of a given prey item occurring in the diet of black-legged kittiwake

collected during the time of chick rearing only. Confidence intervals for years with zer

myctophids in 2008 and 2009 samples, coupled with a lack ofany other fish species in the diet.


Trends in murre diet composition during chick rearing weredifficult to evaluate due to the limited number of years withsamples. Gadids were the most consistent diet component forthick-billed murres at St. Paul Island and were often identified inover 80% of samples (Fig. 6). The lowest proportion of gadids wasobserved in 2008 (38%), during which time thick-billed murresconsumed other fish and invertebrates. Sandlance and mycto-phids were taken sporadically, while capelin were consumed inthe 1970s and in 1988, and amphipods in the 1970s and in2008 only.

Gadids were also prevalent in thick-billed murre diets atSt. George Island, although found in smaller proportions than atSt. Paul Island (Fig. 7). In contrast, squid and euphausiids weregenerally much more prevalent at St. George Island than atSt. Paul Island. Other groups were taken sporadically or werefound in relatively few samples.

4.3.3. Pollock age classes

In addition to the aggregated species groups, we examined thefrequency of occurrence of age groups of walleye pollock (age-0and age-1) in the 18 years when the age classes could beseparated by otolith measurements. On average, both seabirdspecies consumed many more age-0 pollock (Fig. 8) than age-1pollock (Fig. 9), with no age-1 pollock occurring in diets in morethan half of the years. In those years when age-1 pollock wereconsumed, however, they were more prevalent in black-legged

0.0

0.4

0.8

0.0

0.4

0.8

0.0

0.4

0.8

1995 2005

lance

1975 1985 1995 2005

Myctophid

1995 2005

h

1975 1985 1995 2005

squid

1995 2005

ausiid

1975 1985 1995 2005

inverts

s sampled at St. Paul Island by year with 95% confidence intervals using samples

o occurrences cannot be estimated.

1975 1985 1995 2005

0.0

0.4

0.8

gadid

1975 1985 1995 2005

sandlance

1975 1985 1995 2005

Myctophid

1975 1985 1995 2005

0.0

0.4

0.8

capelin

1975 1985 1995 2005

fish

1975 1985 1995 2005

squid

1975 1985 1995 2005

0.0

0.4

0.8

0.0

0.4

0.8

0.0

0.4

0.8

0.0

0.4

0.8

0.0

0.4

0.8

0.0

0.4

0.8

0.0

0.4

0.8

amphipod

1975 1985 1995 2005

euphausiid

1975 1985 1995 2005

inverts

Fig. 5. Probability of a given prey item occurring in the diet of black-legged kittiwakes sampled at St. George Island by year with 95% confidence intervals using samples

collected during the time of chick rearing only. Confidence intervals for years with zero occurrences or 100% occurrences of a prey item cannot be estimated.


kittiwake diets than age-0 pollock, but generally occurred muchless frequently (o25%) in thick-milled murre diets than age-0pollock.

4.4. Comparison between shelf and slope species


At St. Paul Island, the frequency of occurrence of shelf prey wasgenerally much higher than that of slope prey across most years,particularly in the 1970s (Fig. 10). Annual proportions of shelf andslope taxa were negatively correlated, although the correlationwas not significant (r¼–0.233, p¼0.353). At St. George Island,the frequency of occurrence of slope taxa was approximately thesame as that of shelf taxa on average, but patterns varied acrossyears: shelf taxa dominated in the 1970s and 1980s, neither taxadominated in the 1990s, and slope taxa dominated in the recentcold years of 2008 and 2009 (Fig. 10). Annual proportions werestrongly inversely correlated over time (r¼–0.675, p¼0.002).At both islands, occurrence of slope taxa increased over time,although not significantly so (linear regression: St. Paul Island:t¼0.972, p¼0.345; St. George Island: t¼2.03, p¼0.060). It ispossible that these time trends are confounded with the propor-tion of ‘‘unknown taxa,’’ which was high in early years anddecreased over time (not shown).

Splitting euphausiids and amphipods into shelf and slopetaxa resulted in very similar NMDS ordinations (not shown) asthose generated when all euphausiids and amphipods weregrouped (Figs. 1 and 3), suggesting that shelf and slope zooplank-ton showed similar patterns of variability in diet samples.This implies that when shelf euphausiids or amphipods were

abundant in seabird diets, slope euphausiids or amphipods alsotended to be abundant. Indeed, fourth-root transformed abun-dances of shelf and slope euphausiids across individual diets werepositively correlated (r¼0.365, po0.001), although abundancesof the slope amphipod Themisto pacifica and the shelf amphipodThemisto libellula were not (r¼0.026, p¼0.418). This pattern inzooplankton contrasts with that of fish, in which the fourth-roottransformed abundances of myctophids (slope) in diet sampleswere negatively correlated with those of shelf fish such as gadids(r¼–0.305, po0.001) and sandlance (r¼–0.174, po0.001).


Diets of thick-billed murres at St. Paul Island were generallydominated by shelf species, although the occurrence of slopespecies was moderate to high in some years (Fig. 10). Slope taxawere more prevalent at St. George Island, where the frequency ofoccurrence of shelf and slope prey varied across years: shelf taxadominated in many earlier years but slope taxa dominated inrecent years (Fig. 10). Indeed, the frequency of occurrence of slopetaxa at St. George Island increased from about 25% in the late1970s to about 75% in recent years, although the increase was notsignificant (linear regression: t¼1.765, p¼0.101). There was astrong and significant inverse relationship between the frequencyof occurrence of shelf taxa vs. slope taxa across years at St. GeorgeIsland (r¼–0.592, p¼0.02) but not at St. Paul Island (r¼0.373,p¼0.323).

There was a strong positive rank correlation between trans-formed abundances of shelf and slope euphausiids in individualthick-billed murre diets (r¼0.493, po0.001), implying that

1975 1985 1995 2005

0.0

0.4

0.8

gadid

1975 1985 1995 2005

sandlance

1975 1985 1995 2005

Myctophid

1975 1985 1995 2005

0.0

0.4

0.8

capelin

1975 1985 1995 2005

fish

1975 1985 1995 2005

squid

1975 1985 1995 2005

0.0

0.4

0.8

0.0

0.4

0.8

0.0

0.4

0.8

0.0

0.4

0.8

0.0

0.4

0.8

0.0

0.4

0.8

0.0

0.4

0.8

amphipod

1975 1985 1995 2005

euphausiid

1975 1985 1995 2005

inverts

Fig. 6. Probability of a given prey item occurring in the diet of thick-billed murres sampled at St. Paul Island by year with 95% confidence intervals using samples collected

during the time of chick rearing only. Confidence intervals for years with zero occurrences cannot be estimated.


murres, like kittiwakes, do not choose one or the other but takeboth shelf and slope euphausiids when they are available. Therewas a negative rank correlation between gadids (shelf) and squid(slope; r¼–0.246, po0.001), but also between gadids (shelf) andeuphausiids (both; r¼–0.281, po0.001) in the diet, suggestingthat thick-billed murres tend to select either gadids or inverte-brates rather than being limited by a shelf vs. slope trade-off.For fish species, there were no correlations between transformedabundances of slope (myctophids) and shelf (gadids and san-dlance) prey (rr0.1, p40.1).

4.5. Relationships between diet composition and prey abundance


Local indices of forage fish abundance (1982–present) werenot significantly associated with black-legged kittiwake dietcomposition (Mantel’s rho¼0.119, p¼0.354). Measures of pollockavailability at the regional scale (using indices of age-0 and age-1abundance of the Eastern Bering Sea shelf stock, 1975–present)were only weakly associated with diet composition (Mantel’srho¼0.167, p¼0.083), and neither age-0 nor age-1 recruitmentindices were significantly correlated with the amount of totalgadids in kittiwake diets at either island (p40.2).


Thick-billed murre diets were not significantly associated withlocal (1982–present; Mantel’s rho¼0.173, p¼0.306) or regional(1975–present; Mantel’s rho¼0.048, p¼0.510) indices of prey

availability. Neither age-0 nor age-1 recruitment indices weresignificantly correlated with the amount of total gadids in murrediets at either island (p40.2).

4.6. Relationships between diet composition and environmental

variables


Among the environmental variables that were available for allyears (1975–2009; Pacific Decadal Oscillation Index, Arctic Oscil-lation Index, summer wind mixing index, strong wind eventindex, regional summer SST, and cold pool index), variability inthe Arctic Oscillation Index (AO) and regional summer SST, incombination, were most strongly and significantly associatedwith differences in black-legged kittiwake diet composition(Mantel’s rho¼0.24, p¼0.02). In a multivariate regression, thesetwo variables explained 16.6% of the variation in kittiwake dietcomposition (with 11.7% explained by the AO alone). The AO waspositively correlated with myctophid abundance and negativelycorrelated with sandlance and capelin abundance (Table 5).Regional summer SST was significantly correlated only withamphipod abundance in diets at St. George Island (Table 5).

When 2009 and years prior to 1984 were eliminated, allowingall environmental variables to be included in the analysis,variability in the AO, regional summer SST, and the springtransition index were most strongly associated with differencesin black-legged kittiwake diet composition (Mantel’s rho¼0.358,p¼0.022). The spring transition index was only weakly correlatedwith any individual diet items (Table 5).

1980 1990 2000 2010

0.0

0.4

0.8

gadid

1980 1990 2000 2010

sandlance

1980 1990 2000 2010

Myctophid

1980 1990 2000 2010

0.0

0.4

0.8

capelin

1980 1990 2000 2010

fish

1980 1990 2000 2010

squid

1980 1990 2000 2010

0.0

0.4

0.8

0.0

0.4

0.8

0.0

0.4

0.8

0.0

0.4

0.8

0.0

0.4

0.8

0.0

0.4

0.8

0.0

0.4

0.8

amphipod

1980 1990 2000 2010

euphausiid

1980 1990 2000 2010

inverts

Fig. 7. Probability of a given prey item occurring in the diet of thick-billed murres sampled at St. George Island by year with 95% confidence intervals using samples

collected during the time of chick rearing only. Confidence intervals for years with zero occurrences cannot be estimated.



Among the environmental variables that were available for allyears, there was no combination of variables that resulted in asignificant association with thick-billed murre diets (Mantel’srho¼0.17, p¼0.246). Similarly, when all variables were includedby excluding some years in which data were missing, there wereno significant associations with diet composition.

5. Discussion

5.1. Diet over time

This study expanded substantially on previous analyses oflong-term diet data (e.g., Sinclair et al., 2008) through further fieldsampling (2001–2009) and the acquisition of additional historicdata (seven years between 1979 and 1995). With the expandedtime series, many of the conclusions reported by Sinclair et al.(2008) were still supported despite different statistical techni-ques. The diet composition of both kittiwakes and murres showedstrong and significant differences over time and differed signifi-cantly between St. George and St. Paul islands, although differ-ences between islands were more strongly pronounced formurres. Specifically, we confirmed a continued absence of capelinin seabird diets in recent years, increased use over time ofmyctophids by black-legged kittiwakes, and a continued depen-dence on age-0 pollock with a low occurrence of age-1 pollock inrecent years. However, capelin were present in black-leggedkittiwake diets at St. Paul throughout the latter part of the1990s (Fig. 4) and the reliance on pollock has decreased sincethe late 1970s (Figs. 4–7).

The age class distribution of pollock in the diets of both seabirdspecies suggests that age-0 pollock are more frequently availableto seabirds than age-1 pollock, but in years when age-1 pollockare available, they are taken instead. Although pollock have longbeen present in the diets of Pribilof Island predators (Decker et al.,1995; Hunt et al., 1996b, Sinclair et al., 1994), its currentdominance in seabird diets may reflect compromises made byforaging birds due to changes in prey availability (see Hunt et al.,2002). In the late 1970s capelin was a significant component inthe diet of both species. Capelin is a schooling fish that hasrelatively high lipid content while pollock is amongst the leanestof fishes in the Bering Sea (Perez, 1994). Forage fishes with higherfat content have been linked to increased seabird reproductivesuccess (e.g., Golet et al., 2000b), and thus, capelin are consideredhigher quality prey for seabirds than pollock (Barrett and Furness,1990). Since the late 1970s, however, capelin has largely dis-appeared from diets of both kittiwakes and murres in the PribilofIslands. That pollock, a presumably less favorable diet item, hasbecome the most important prey species for piscivorous seabirdson the Pribilof Islands suggests birds are choosing prey based onaccessibility due to location and behavior, and not necessarilybecause it is preferred prey. Pollock represent a very large portionof the biomass in the Bering Sea, and a disproportionally largefraction of the age-0 pollock population is known to inhabit thePribilof Island nursery (Lloyd and Davis, 1989; Springer, 1992;Swartzman et al., 2005). Unfortunately, we know little about theabundance of juvenile capelin in the water column except thattheir presence is highly variable interannually. While absent fromseabird diets in recent years, capelin have continued to persist indiets of arrowtooth flounder (Atheresthes stomias), Pacific cod,Pacific halibut (Hippoglossus stenolepis), and pollock, at least

1975 1980 1985 1990 1995 2000 2005 2010

0.0

0.2

0.4

0.6

0.8

1.0

BLKI, St. Paul

1975 1980 1985 1990 1995 2000 2005 2010

BLKI, St. George

1975 1980 1985 1990 1995 2000 2005 2010

0.0

0.2

0.4

0.6

0.8

1.0

0.0

0.2

0.4

0.6

0.8

1.0

0.0

0.2

0.4

0.6

0.8

1.0

TBMU, St. Paul

1980 1985 1990 1995 2000 2005 2010

TBMU, St. George

Fig. 8. Probability of age-0 walleye pollock occurring in the diet of seabirds at the Pribilof Islands by year with 95% confidence intervals using samples collected during the

time of chick rearing only.


through 2001 (Yang et al., 2005). It seems reasonable to concludethat capelin are still present but have shifted to locations nolonger accessible to seabirds. This is consistent with summerdistributions of capelin inferred from bottom trawl surveys,which found capelin predominantly in the waters of the innerdomain from 2004 to 2008, placing them well out of foragingrange for Pribilof Island birds, although capelin were found wellwithin the Pribilof domain in 2009 (Hollowed et al., 2012).

As with capelin, irregular occurrences of other fatty foragespecies, such as sandlance and myctophids, in seabird dietssuggest that availability of these prey types may vary by year,foraging constraints, and preferences of predator species. Forexample, myctophids do not inhabit the waters of the BeringSea shelf but live in deep water of the pelagic zone where theydiurnally migrate into the epipelagic (Sinclair and Stabeno, 2002).Myctophids are also brought closer to the surface at night at theshelf edge and other upwelling zones (Frost and McCrone, 1979),making them available for far-ranging kittiwakes but difficult toaccess for murres with shorter foraging trips. Indeed, myctophidsrarely occur in murre diets while black-legged kittiwakes haveincreased their consumption of these fish since the 1970s at thesame time that they have reduced the number of pollockconsumed (Sinclair et al., 2008; this study). Pollock, due to theirrelatively consistent availability in shelf waters, may represent an

efficient alternative to higher quality prey species that are eitherno longer present or are difficult to access (Sinclair et al., 1994).

5.2. Comparison between species

Differences in diet between species probably reflect basicdifferences in foraging location and strategy. Data from GPStransmitters placed on foraging birds during the chick-rearingperiod in 2008 and 2009 indicated murres and kittiwakes foragedin different areas, with kittiwakes traveling farther from thecolonies than murres (R. Paredes, unpubl. data). These differencesmay partially be explained by differences in oceanographicconditions across the region and physiological adaptationsbetween species. The transfer of energy from primary productiv-ity to higher trophic levels in the waters of the southeasternBering Sea varies by location and depth; the middle domain isdominated by a benthic food web, whereas the outer domain andthe basin are dominated by a pelagic food web (Iverson et al.,1979; Schneider and Hunt, 1982). Murres, capable of diving forprey but unable to make long foraging trips, are able to exploit therich benthic food resources in the closer middle domain region.Kittiwakes, better adapted to longer flight but restricted tofeeding at the surface, can make longer trips to the outer domain

1975 1980 1985 1990 1995 2000 2005 2010

0.0

0.2

0.4

0.6

0.8

1.0BLKI, St. Paul

1975 1980 1985 1990 1995 2000 2005 2010

BLKI, St. George

1975 1980 1985 1990 1995 2000 2005 2010

0.0

0.2

0.4

0.6

0.8

1.0

0.0

0.2

0.4

0.6

0.8

1.0

0.0

0.2

0.4

0.6

0.8

1.0TBMU, St. Paul

1980 1985 1990 1995 2000 2005 2010

TBMU, St. George

Fig. 9. Probability of age-1 walleye pollock occurring in the diet of seabirds at the Pribilof Islands by year with 95% confidence intervals using samples collected during the

time of chick rearing only.


and slope waters where prey may be more readily available nearthe surface, particularly at night.

In addition to traveling longer distances, kittiwakes alsoexhibited less predictability in foraging location, ranging over awider area (primarily over the Aleutian Basin) than murres, whichforaged in more specific regions (K. Kuletz and N. Jones, unpubl.data). This suggests that the availability of prey for kittiwakes insurface waters may be more patchy than the availability ofmidwater prey for murres.

5.3. Comparison between islands

Diets of both species differed in the relative abundances of fishand invertebrates consumed at each island. Kittiwakes ate morepollock and sandlance at St. Paul Island, compared to moremyctophids, euphausiids, squid, and other invertebrates atSt. George Island. A similar pattern existed for murres, whosediets contained more pollock at St. Paul Island and more squidand euphausiids at St. George Island. These differences areconsistent with recent tracking studies, which found little overlapin foraging locations between the two islands in 2008 and 2009for either murres or kittiwakes (R. Paredes, unpubl. data). Becausekittiwakes travel much longer distances than murres, foraginglocations and therefore diets are more likely to overlap, which is

supported by our finding that diets of black-legged kittiwakes aremore similar between islands than diets of thick-billed murres.

If differences in diet between islands result primarily fromdifferences in foraging locations, diet differences may be used toinfer similarities or differences in forage locations among years.Both diet data and foraging track data for 2008 and 2009 suggestlarge inter-island differences in seabird diets. While tracking dataare not available for other years, a high degree of similarity inseabird diets between the islands in some years, such as in 1977or 1997, suggests that foraging locations were similar betweenislands in those years; hence foraging location data from 2008and 2009 might not be representative of all years. It is alsopossible that the prey base was larger or more broadly spread inyears with more inter-island overlap.

Conditions experienced by breeding birds on St. Paul andSt. George islands appear to vary between islands. Despite similarreproductive success on the two islands (Byrd et al., 2008a),population trends over the past 30 years exhibit different pat-terns, with numbers of kittiwakes and murres declining consis-tently on St. Paul Island while remaining relatively stable orrebounding after declines on St. George Island (Byrd et al., 2008b).A leading hypothesis to explain these differences is that St. PaulIsland is farther from the productive shelf edge than St. GeorgeIsland, requiring individuals at St. Paul Island to either forage inpoor areas closer to the island or to travel farther to access prime

1975 1985 1995 2005

0.0

0.2

0.4

0.6

0.8

1.0

1975 1985 1995 2005

1975 1985 1995 2005

0.0

0.2

0.4

0.6

0.8

1.0

0.0

0.2

0.4

0.6

0.8

1.0

0.0

0.2

0.4

0.6

0.8

1.0

1975 1985 1995 2005

TBMU, St. Paul

BLKI, St. Paul

TBMU, St. George

BLKI, St. George

Fig. 10. Probability of shelf (solid circles) and slope (open circles) taxa occurring in the diet of seabirds at the Pribilof Islands by year with 95% confidence intervals using

samples collected during the time of chick rearing only.

Table 5Spearman rank correlations between Arctic Oscillation Index, regional summer SST, and spring transition index variables, and abundance of individual taxonomic groups in

black-legged kittiwake diets at St. Paul and St. George islands.

Arctic oscillation index Regional summer SST Spring transition index

Prey item St. Paul Island St. George Island St. Paul Island St. George Island St. Paul Island St. George Island

rho pa rho p rho p rho p rho p rho p

Total gadid 0.070 0.782 �0.273 0.273 �0.154 0.542 �0.111 0.661 �0.203 0.434 0.052 0.843

Sandlance �0.417 0.085n�0.555 0.017nn 0.009 0.971 0.136 0.592 0.044 0.866 �0.260 0.314

Myctophid 0.583 0.011nn 0.439 0.068n 0.109 0.665 �0.061 0.810 0.198 0.447 �0.046 0.862

Capelin �0.060 0.812 �0.478 0.045nn�0.309 0.213 �0.213 0.397 0.121 0.645 0.083 0.752

Other fish 0.144 0.567 �0.283 0.255 �0.050 0.845 0.275 0.269 0.112 0.670 �0.297 0.248

Squid �0.305 0.218 �0.058 0.819 0.305 0.218 0.193 0.442 0.045 0.865 �0.113 0.665

Amphipod �0.215 0.392 0.028 0.912 �0.387 0.113 �0.631 0.005nn 0.170 0.514 0.389 0.123

Euphausiid �0.285 0.252 �0.197 0.433 �0.333 0.177 �0.146 0.564 0.035 0.894 0.009 0.972

Other invertebrate �0.246 0.325 0.171 0.498 �0.029 0.909 �0.183 0.467 0.086 0.743 �0.141 0.589

an¼significant at the 0.1 level; nn

¼significant at the 0.05 level.


food resources at the shelf break (Byrd et al., 2008b). Indeed, thereis evidence that birds on St. Paul Island may be more food-limited,as evident in higher stress levels (Benowitz-Fredericks et al.,2008) and greater variability in colony attendance of murres(Kitaysky et al., 2000) there compared to murres on St. GeorgeIsland. Our diet data and recent foraging track data (R. Paredes,unpubl. data), which show birds on each island are, in fact,foraging on different prey, usually in different places, supportthe hypothesis that food resources may contribute to some of thedemographic disparity between islands (Byrd et al., 2008b).

5.4. Relationships with forage fish abundance

Weak correlations between seabird diet and forage fishresources in our study may reflect limitations in forage stockdata. Available time series of forage fish abundance are eitherderived from bottom trawl surveys (between 3 m and 0.5 m offbottom; Honkalehto et al., 2008), which are unlikely to reflectforage fish availability for seabirds, or have very short time series.Furthermore, most historic fish surveys did not target forage fishstocks and only measure their presence as incidental bycatch.


Acoustic estimates of age-1 pollock and euphausiid abundancefrom midwater surveys (between 16 m from the surface and 3 moff bottom; Honkalehto et al., 2008) have recently becomeavailable (P. Walline and P. Ressler, pers. comm.; Ressler et al.,2012) and may provide a reasonable proxy for the availability ofthese species to diving seabirds like murres (but not kittiwakes,which forage exclusively at the surface). However, while euphau-siids in these midwater surveys increased from 2005 to 2009,their prevalence in thick-billed murre diets at St. George Islanddecreased between 2006/07 and 2008/09 (Fig. 7), and theiraverage abundance in diets was much lower than in the 1980sand 1990s (Appendix B1). For other prey species such as mycto-phids, squid, or sandlance, seabird diets provide one of the fewsources of information. Seabirds clearly provide important infor-mation about changes in the prey base of commercially importantfish species in the eastern Bering Sea. Indices of forage availabilityand abundance based on seabird diets could be improved byaccounting for interannual differences in foraging locations asdetermined by oceanographic conditions and behavioral choices(Satterthwaite and Mangel, 2012).

5.5. Relationships with physical oceanography

Atmospheric variability is believed to drive primary produc-tion, secondary production, forage fish recruitment, and, ulti-mately, predator demographic parameters in the Bering Sea(Coyle et al., 2008; Gaston et al., 2005; Hunt et al., 2002). Severalbroad categories of physical oceanography variables have beenlinked to prey distribution, e.g., sea surface temperature (Baird,1990, for forage fish; Decker et al., 1995, for walleye pollock;Coyle and Pinchuk, 2002, for zooplankton), sea bottom tempera-ture (Wyllie-Echeverria and Wooster, 1998, for groundfishes),water-column stability (Coyle et al., 2008, for zooplankton), windmixing (Jin et al., 2006, for phytoplankton; Brodeur et al., 2008,for jellyfish), and ice extent (Gaston et al., 2005; Wyllie-Echeverria and Wooster, 1998). Other studies have found associa-tions between physical variables and seabird diets (e.g., Sinclairet al., 2008) or reproductive success (e.g., Byrd et al., 2008a).While proposed mechanistic links between climate and toppredators are typically mediated through prey, direct effects onseabird survival have been documented. For example, sea tem-perature can affect the likelihood of storms (Norris, 2000; Wentzet al., 2000), which commonly cause widespread breeding failuresat seabird colonies (Mallory et al., 2009; Threfall et al., 1974).Winds, in general, likely play a significant role on both thephysical and biological structure of the shelf, not just throughstorminess, but also through their impact on currents and mixing.

Even if we assume that climate effects are mediated throughthe food chain, increasing decoupling between, e.g., a physicalvariable, primary production, secondary production, and predatorabundance, will make it difficult to detect a consistent relation-ship between the first and last step. While the strongest relation-ship should be evident between local forage abundances and aseabird’s response, our lack of knowledge regarding forage fishlife history makes selection of appropriate local environmentalmeasures of prey abundance challenging to assess. More studiessuch as that of Sinclair et al. (2008), which found a relationshipbetween summer SST, ice retreat index, and seabird and sealdiets, and this study, which found a relationship between ArcticOscillation and regional sea surface temperature with kittiwakediets, are needed to elucidate these intermediate relationships.

As apex predators, seabirds are several steps removed fromphysical climate variables, and the marine system is complexenough that we had little a priori means to choose specificvariables to investigate. For example, although we have demon-strated that sea temperature is important, choices about the

spatial and temporal scale at which to measure temperaturescan be crucial in finding strong linkages, but are somewhatarbitrary without a mechanistic understanding. Indeed, Byrdet al. (2008a) found a negative correlation between spring SSTand overall reproductive success of red-legged kittiwakes, but notsummer SST, and only when the data were detrended. We notethat the various SST datasets we investigated (some remotelysensed, some directly measured) are frequently not stronglycorrelated with each other. Further complicating the issue in aretrospective study such as ours, various climate datasets coverdifferent ranges of years, and data decisions often are based onwhat matches the time period of interest. Therefore, we cast awide net of available climate time series as potential correlates todiet, recognizing that some subset of any significant relationshipsare likely to be spurious.

We anticipated finding stronger evidence for a definitive linkbetween environmental variability specific to the eastern BeringSea and seabird diet composition. The lack of such linkagessuggests that diet composition may depend more on behavioralchoices made by seabirds in a variable foraging environment(Satterthwaite and Mangel, 2012) than on oceanographic varia-bility. While seabirds may feed opportunistically on readilyavailable prey, they are likely to return to predictable sourcesand may fail to find better prey in other locations, even whenavailable. For example, relatively high abundances of capelin,presumably a preferred food item, appeared to be available east ofthe Pribilof Islands in 2009 (Hollowed et al., 2012) but capelinwere not found in any diet samples.

Although we found a relationship between two broad mea-sures of climate variability and kittiwake diets, a mechanistic linksupporting this relationship remains elusive. Seabird diet and/ordemographic parameters have been associated with broad-scaleclimatic indices in many previous studies in various marinesystems (e.g., Bond et al., 2011; Irons et al., 2008; Sandvicket al., 2005). The link between climate and other seabird para-meters is usually assumed to be prey availability, and indeed,ways in which temperature or ice, for example, affect prey havebeen demonstrated (Wyllie-Echeverria, 1996). The AO is an indexof sea-level pressure variations north of 201N latitude and caninfluence movement and thickness of sea ice in the Arctic (Rigoret al., 2002) and sea ice variability (extent and timing of retreat)in the Eastern Bering Sea (Mueter et al., 2007), which in turninfluences stratification and mixing, as well as the temperatureand extent of the cold pool. However, neither this nor otherstudies have documented more direct links between broadmeasures and localized effects, such as between ice retreat or acold pool index and seabird diet, suggesting that the relationshipis spurious or reflects an unknown mechanism that is only weaklyrelated to more local temperature or ice conditions.

5.6. Conclusions

During their nesting season, seabirds are among the easiesttaxonomic groups to study in the marine system. We expect thattheir diet responds annually to changes in their prey base andultimately to large-scale drivers, including climate and physicaloceanography. However, these relationships are complex, noisy,and maybe more importantly, are mediated by behavioral adap-tations and learning, and will therefore be difficult to elucidate.Moreover, current measures of diet composition do not accountfor differences in prey quality, and better measures of caloric andnutrient intake rather than number of prey are needed. While wefound few direct relationships between prey composition andphysical drivers, we did document substantial changes in dietcomposition. Even our coarse measure of prey abundances inseabird diets provides useful information about forage stocks


(such as the disappearance of capelin), which are frequently muchmore difficult to measure directly. Many studies have demon-strated correlations between diet choice by seabirds and avail-ability of prey in the environment (e.g., Barrett and Krasnov,

Table A1Diet composition (average number of prey items) of black-legged kittiwakes by island

Island Year Total gadid Sandlance Myctophid Capelin

St. Paul 1975 2.22 0.08 0.28 0.20

St. Paul 1976 0.59 0.07 0.17 0.65

St. Paul 1977 1.40 0.02 0.47 0.50

St. Paul 1978 4.92 0.77 0 0.25

St. Paul 1979 3.30 0.05 0 0

St. Paul 1984 1.62 0.42 1.08 0

St. Paul 1985 0 1.00 0 0

St. Paul 1987 0.35 1.53 0 0

St. Paul 1988 2.70 1.61 0.45 0

St. Paul 1992 6.20 0 0.80 0

St. Paul 1993 1.86 0 0.57 0

St. Paul 1997 0.67 0 2.11 0.11

St. Paul 1998 10.43 1.75 0.39 0.04

St. Paul 2000 10.80 0 0.60 0

St. Paul 2003 0.29 4.67 0 0

St. Paul 2006 0 3.89 0.56 0

St. Paul 2008 0.23 0.77 1.17 0

St. Paul 2009 1.09 1.14 0.09 0

St. George 1976 0.06 0 0.76 0.06

St. George 1977 2.93 0.04 0.84 0.25

St. George 1978 4.10 1.30 0.52 0.36

St. George 1981 0 0.14 0 0

St. George 1984 0.63 0.02 0.81 0

St. George 1985 4.67 0 0.08 0.08

St. George 1986 4.97 0.89 1.14 0

St. George 1988 2.71 2.29 0.13 0

St. George 1992 0.20 0.08 0.80 0

St. George 1993 2.90 0 0.30 0

St. George 1994 0 0 4.00 0

St. George 1997 0.43 0 4.14 0

St. George 1998 0.14 2.83 0.92 0

St. George 1999 22.70 0 1.30 0

St. George 2000 6.78 0 4.00 0

St. George 2001 2.43 3.00 0.00 0

St. George 2008 0 0 2.39 0

St. George 2009 0 0 1.83 0

Table B1Diet composition (average number of prey items) of thick-billed murres by island and

Island Year Total gadid Sandlance Myctophid Capelin

St. Paul 1975 0.57 0 0 0

St. Paul 1976 2.50 0 0 0

St. Paul 1977 17.94 0 0.35 0.06

St. Paul 1978 11.40 0.20 0 0

St. Paul 1979 17.67 0 0 0

St. Paul 1984 25.90 0 0 0

St. Paul 1988 21.20 0 1.60 0.60

St. Paul 2008 2.27 0 0.04 0

St. Paul 2009 3.81 0 0.02 0

St. George 1977 11.56 0 0.11 0

St. George 1978 7.19 0.11 0 0

St. George 1981 4.88 0 0 0

St. George 1984 1.31 0.14 0 0

St. George 1985 70.27 0 0 0

St. George 1986 18.47 0 0 0

St. George 1988 30.27 0.18 0.09 0.09

St. George 1992 15.61 0 0 0

St. George 1997 1.40 0 0 0

St. George 1999 50.75 0.38 1.69 0

St. George 2000 30.60 0 0 0

St. George 2006 2.18 0 0 0

St. George 2007 1.69 0 0 0

St. George 2008 3.42 0 0 0

St. George 2009 0.79 0 0 0

1996; Hatch and Sanger, 1992; Hislop and Harris, 1985;Montevecchi and Myers, 1995), suggesting some seabird dietsare good indicators of changes in prey availability (Cairns, 1987;Montevecchi, 1993). Unfortunately, there are few time series data

and year.

Other fish Squid Amphipod Euphausiid Other invertebrate

0.02 0 2.47 2.12 1.47

0.04 0.04 3.93 6.59 0.17

0.54 0.24 1.79 2.73 0.99

0 0 2.20 3.23 0.97

0.10 0.05 1.25 17.35 2.40

0 1.00 0.27 0.31 5.31

0 0 0 0 0

0 0.12 1.18 6.47 3.47

0.11 0.02 0 0.11 1.11

0 0 0 5.00 1.00

0 0 0 0 0

0.30 0.11 0 0 0.74

0 0.07 0 1.07 4.43

2.40 0 0 0 1.80

0 0 0 0 0

0 0 0 0 0

1.07 0.10 0.53 0.27 1.77

2.82 0 2.55 10.73 0.05

0 0.12 2.71 6.71 0.12

0 0.34 0.38 1.29 0.11

0 0.06 2.64 37.36 58.26

0.86 0.14 0 141.57 0.29

0.12 0.42 0.16 5.44 0.21

0.08 0.25 0.17 10.92 3.50

1.36 0.03 0.28 2.67 2.17

0 0.13 0.04 10.25 1.96

0 0.80 0.63 15.88 7.35

0 0.10 0 0 9.80

0 0.20 2.00 0 86.80

0.57 0.29 0 0.14 0

7.08 0.42 0.06 14.67 1.14

0 0 0.10 0.20 0

0.56 0 0 1.67 0.44

0.43 0 0 2.43 0

0 0.03 1.31 83.78 3.11

0 0.08 0.42 59.83 0.17

year.

Other fish Squid Amphipod Euphausiid Other invertebrate

0.14 0 0.29 0 0.29

0 0 0 0 0

0.29 0.35 6.71 0.06 0.76

0.20 1.60 1.40 2.20 0.20

1.00 0 0 0 0

0 0 0 0 0

0.60 0.60 0 0.20 0

0.62 0.54 0.15 0.12 0.62

0.26 0.06 0 0 1.83

0 0.61 6.39 0 1.28

7 3.04 3.56 14.07 0.22

0 1.13 0 58.88 0

0 0.75 0.03 89.36 0.22

0.13 0.80 0 3.40 0.07

1.40 0 1.40 1.47 0.13

0.09 3.36 0 0 0.18

0.26 0.96 0 83.96 0

37.60 3.20 0 0 0

0.13 0.75 0 0.19 0.25

0.10 0.10 0 0.10 0.10

0 0.45 0 0.18 0

0 0.85 0 0.23 0

0.04 2.00 0 0.38 0

0 1.38 2.38 3.45 0


on forage fish abundances around the Pribilofs to test ourproposal of the utility of seabird diets as indicators of regionalprey availability. Nevertheless, given the immense value of thefishery resources in the Eastern Bering Sea, seabird diet studiesremain an essential component of ecosystem monitoring.

Acknowledgments

We are grateful to the many researchers who contributed dietdata from the Pribilof Islands: M.S.W. Bradstreet, D.E. Dragoo, G.L.Hunt, M. Ito, D. Lloyd, A.M. Springer, and the University of AlaskaFairbanks. Countless field workers have been bitten and regur-gitated on through rain, wind, and fog to collect these samples.Financial support for data collection was provided to G. L. Huntand his students by the NOAA/BLM OCSEAP program (1975–1979) and the NSF Office of Polar Programs in (1987, 1988); andto A.M. Springer (1992, 1997–2000) by Alaska Sea Grant CollegeProgram and North Pacific Universities Marine Mammal ResearchConsortium. Our BSIERP collaborators, especially D.B. Irons, N.Jones, A.S. Kitaysky, K. Kuletz, R. Orben, R. Paredes, and D.D. Roby,have provided invaluable contributions, support, and advicethroughout this project. We thank the many specialists, especiallyK. Turco, who carefully identified thousands of minute, partiallydigested prey items. Our forage fish analyses benefitted from dataand input from T. Buckley, R. Lauth, P. Ressler, and P. Walline ofNOAA. K. Benoit-Bird and S. Heppell contributed advice on preyspecies distribution. G.V. Byrd developed and tirelessly guidedthis project at every stage as well as providing valuable review ofearlier versions of the manuscript. Field work in 2008–2009 andthis retrospective analysis were funded by the North PacificResearch Board as part of BSIERP project B65. This is NPRBpublication number 341 and BEST-BSIERP Bering Sea Projectpublication number 57.

Appendix

See Appendix Tables A1 and B1.

References

Anderson, M.J., Gorley, R.N., Clarke, K.R., 2008. PERMANOVAþ for PRIMER: Guideto Software and Statistical Methods. PRIMER-E, Plymouth, UK.

Baird, P.H., 1990. Influence of abiotic factors and prey distribution on diet andreproductive success of three seabird species in Alaska. Ornis Scan 21,224–235.

Barrett, R.T., Furness, R.W., 1990. The prey and diving depths of seabirds onHornøy, North Norway after a decrease in the Barents Sea capelin stocks. OrnisScand 21, 179–186.

Barrett, R.T., Krasnov, Y.V., 1996. Recent responses to changes in stocks of preyspecies by seabirds breeding in the southern Barents Sea. ICES J. Mar. Sci.53, 713–722.

Barrett, R.T., Camphuysen, C.J., Anker-Nilssen, T., Chardine, J.W., Furness, R.W.,Garthe, S., Huppop, O., Leopold, M.F., Montevecchi, W.A., Veit, R.R., 2007. Dietstudies of seabirds: a review and recommendations. ICES J. Mar. Sci. 64,1675–1691.

Benowitz-Fredericks, Z.M., Shultz, M.T., Kitaysky, A.S., 2008. Stress hormonessuggest opposite trends of food availability for planktivorous and piscivorousseabirds in 2 years. Deep-Sea Res. II 55, 1868–1876.

Bond, A.L., Jones, I.L., Williams, J.C., Byrd, G.V., 2011. Diet of Auklet Chicks in theAleutian Islands, Alaska: Similarity Among Islands, Interspecies Overlap, andRelationships to Ocean Climate, J. Ornithol. doi:10.1007/s10336-011-0704-3.

Bray, J.R., Curtis, J.T., 1959. An ordination of the upland forest communities ofSouthern Wisconsin. Ecol. Monogr. 27, 325–349.

Brodeur, R.D., Wilson, M.T., Napp, J.M., Stabeno, P.J., Salo, S., 1997. Distribution ofjuvenile pollock relative to frontal structure near the Pribilof Islands, BeringSea. In: Forage Fishes in Marine Ecosystems. Proceedings of the InternationalSymposium on the Role of Forage Fishes in Marine Ecosystems. University ofAlaska Sea Grant AK-SG-97-01, Fairbanks, pp. 573–589.

Brodeur, R.D., Decker, M.B., Ciannelli, L., Purcell, J.E., Bond, N.A., Stabeno, P.J.,Acuna, E., Hunt Jr., G.L., 2008. Rise and fall of jelly fish in the eastern Bering Seain relation to climate regime shifts. Prog. Oceanogr. 77, 103–111.

Byrd, G.V., Sydeman, W.J., Renner, H.M., Minobe, S., 2008a. Responses of piscivor-ous seabirds at the Pribilof Islands to ocean climate. Deep-Sea Res. II 55,1856–1867.

Byrd, G.V., Schmutz, J.A., Renner, H.M., 2008b. Contrasting population trends ofpiscivorous seabirds in the Pribilof Islands: A 30 year perspective. Deep-SeaRes. II 55, 1846–1855.

Cairns, D.K., 1987. Seabirds as indicators of marine food supplies. Biol. Oceanogr.5, 261–271.

Clarke, K.R., Gorley, R.N., 2006. PRIMER v6: User Manual/Tutorial. PRIMER-E,Plymouth, UK.

Coyle, K.O., Pinchuk, A.I., 2002. Climate-related differences in zooplankton densityand growth on the inner shelf of the southeastern Bering Sea. Prog. Oceanogr.55, 177–194.

Coyle, K.O., Pinchuk, A.I., Eisner, L.B., Napp, J.M., 2008. Zooplankton speciescomposition, abundance and biomass on the eastern Bering Sea shelf duringsummer: the potential role of water-column stability and nutrients instructuring the zooplankton community. Deep-Sea Res. II 55, 1775–1791.

Coyle, K.O., Eisner, L.B., Mueter, F.J., Pinchuk, A., Janout, M.A., Cieciel, K., Farley,E.V., Andrews, A.G., 2011. Climate change in the southeastern Bering Sea:impacts on pollock stocks and implications for the Oscillating ControlHypothesis. Fish. Oceanogr. 20, 139–156.

Decker, M.B., Hunt Jr., G.L., Byrd, G.V., 1995. The relationships among sea-surfacetemperature, the abundance of juvenile walleye pollock (Theragra chalco-gramma), and the reproductive performance and diets of seabirds at thePribilof Islands, southeastern Bering Sea. In: Beamish, R.J. (Ed.), ClimateChange and Northern Fish Populations. Can. Spec. Publ. Fish. Aquat. Sci.121, pp. 425–437.

Dragoo, D.E., Byrd, G.V., Irons, D.B., 2010. Breeding Status, Population Trends andDiets of Seabirds in Alaska, 2007. U.S. Fisheries and Wildlife Service Report.AMNWR 2010/08, Homer, Alaska.

Frost, K.J., Lowry, L.F., 1981. Trophic importance of some marine gadids in northernAlaska and their body-otolith size relationships. Fish. Bull. 79, 187–192.

Frost, B.W., McCrone, L.E., 1979. Vertical distribution, diel vertical migration, andabundance of some mesopelagic fishes in the eastern Sub-arctic Pacific Oceanin summer. Fish. Bull 76, 751–770.

Gaston, A.J., Hipfner, J.M., 2000. Thick-billed murre (Uria lomvia). In: Poole, A. (Ed.),The Birds of North America, no. 497. Cornell Lab of Ornithology, Ithaca.Available online at: /http://bna.birds.cornell.edu/bna/species/497S.

Gaston, A.J., Gilchrist, H.G., Hipfner, J.M., 2005. Climate change, ice conditionsand reproduction in an Arctic nesting marine bird: Brunnich’s guillemot(Uria lomvia L.). J. Anim. Ecol. 74, 832–841.

Golet, G.H., Irons, D.B., Costa, D.P., 2000a. Energy costs of chick rearing in black-legged kittiwakes (Rissa tridactyla). Can. J. Zool. 78, 982–991.

Golet, G.H., Kuletz, K.J., Roby, D.D., Irons, D.B., 2000b. Adult prey choice affectschick growth and reproductive success in pigeon guillemots. Auk 117, 82–91.

Grebmeier, J.M., Overland, J.E., Moore, S.E., Farley, E.V., Carmack, E.C., Cooper, L.W.,Frey, K.E., Helle, J.H., McLaughlin, F.A., McNutt, S.L., 2006. A major ecosystemshift in the northern Bering Sea. Science 311, 1461–1464.

Hamer, K.C., Monaghan, P., Uttley, J.D., Walton, P., Burns, M.D., 1993. The influenceof food supply on the breeding ecology of kittiwakes (Rissa tridactyla) inShetland. Ibis 135, 255–263.

Hatch, S.A., Sanger, G.A., 1992. Puffins as predators on juvenile pollock and otherforage fish in the Gulf of Alaska. Mar. Ecol. Prog. Ser. 80, 1–14.

Hatch, S.A., Robertson, G.J., Baird, P.H., 2009. Black-legged kittiwake (Rissatridactyla). In: Poole, A. (Ed.), The Birds of North America, No. 92. Cornell Labof Ornithology, Ithaca. Available online at: /http://bna.birds.cornell.edu/bna/species/92S.

Hickey, J.J., Craighead, F.L., 1977. A Census of Seabirds on the Pribilof Islands. In:Environmental Assessment of the Alaskan Continental Shelf. Annual Reports ofPrincipal Investigators, vol. 2. BLM/NOAA/OMPA, Boulder, pp. 96–195.

Hislop, J.R.G., Harris, M.P., 1985. Recent changes in the food of young puffinsFratercula arctica on the Isle of May in relation to fish stocks. Ibis 127,234–239.

Hollowed, A.B., Barbeaux, S., Cokelet, E.D., Kotwicki, S., Ressler, P.H., Spital, C.,Wilson, C., 2012. Effects of climate variations on pelagic ocean habitats andtheir role in structuring forage fish distributions in the Bering Sea. Deep-SeaRes. II 65–70, 230–250.

Honkalehto, T., Williamson, N., Jones, D., McCarthy, A., McKelvey, D., 2008. Resultsof the Echo Integration-Trawl Survey of Walleye Pollock (Theragra chalco-gramma) on the U.S. and Russian Bering Sea Shelf in June and July 2007. U.S.Department of Commerce NOAA Technical Memo. NMFS-AFSC-190.

Hunt Jr., G.L., Decker, M.B., Kitaysky, A., 1996a. Fluctuations in the Bering Seaecosystem as reflected in the reproductive ecology and diets of kittiwakes onthe Pribilof Islands, 1975–1990. In: Greenstreet, S.P.R., Tasker, M.L. (Eds.),Aquatic Predators and Their Prey. Fishing News Books, Oxford, pp. 142–153.

Hunt Jr., G.L., Kitaysky, A.S., Decker, M.B., Dragoo, D.E., Springer, A.M., 1996b.Changes in the distribution and size of juvenile walleye pollock, Theragrachalcogrammus, as indicated by seabird diets at the Pribilof Islands and bybottom trawl surveys in the Eastern Bering Sea, 1975 to 1993. U.S. Departmentof Commerce, NOAA Tech. Rep. NMFS-126, pp. 125–139.

Hunt Jr., G.L., Stabeno, P., Walters, G., Sinclair, E., Brodeur, R.D., Napp, J.M., Bond,N.A., 2002. The eastern Bering Sea: evidence for change and a newhypothesis linking ecosystem control and climate. Deep-Sea Res. II 49,5821–5853.

Hunt Jr., G.L., Coyle, K.O., Eisner, L., Farley, E.V., Heintz, R., Mueter, F., Napp, J.M.,Overland, J.E., Ressler, P.H., Salo, S., Stabeno, P.J., 2011. Climate impacts on

dx.doi.org/10.1007/s10336-011-0704-3

http://bna.birds.cornell.edu/bna/species/497




eastern Bering Sea food webs: a synthesis of new data and an assessment ofthe oscillating control hypothesis. ICES J. Mar. Sci. 68, 1230–1243.

Ianelli, J.N., Barbeaux, S., Honkalehto, T., Kotwicki, S., Aydin, K., Williamson, N.,2010. Assessment of the walleye pollock stock in the eastern Bering Sea.In: Stock Assessment and Fishery Evaluation Report for the GroundfishResources of the Bering Sea/Aleutian Islands Regions. North Pacific FisheriesManagement Council, Anchorage, pp. 53–156.

Intergovernmental Panel on Climate Change (IPCC), 2007. Climate Change 2007:The Physical Science Basis. Contribution of Working Group I to the FourthAssessment Report of the Intergovernmental Panel on Climate Change.Cambridge University Press, Cambridge and New York.

Irons, D.B., Anker-Nilssen, T., Gaston, A.J., Byrd, G.V., Falk, K., Gilchrist, G., Hario, M.,Hjernquist, M., Krasnov, Y.V., Mosbech, A., Olsen, B., Petersen, A., Reid, J.B.,Robertson, G.J., Strom, H., Wohl, K.D., 2008. Fluctuations in circumpolar seabirdpopulations linked to climate oscillations. Glob. Change Biol. 14, 1455–1463.

Ito, M., Takahashi, A., Kokubun, N., Kitaysky, A.S., Watanuki, Y., 2010. Foragingbehavior of incubating and chick-rearing thick-billed murres Uria lomvia.Aquat. Biol. 8, 279–287.

Iverson, R.L., Coachman, L.K., Cooney, R.T., English, T.S., Goering, J.J., Hunt, G.L.,McCauley, M.C., McRoy, C.P., Reeburgh, W.S., Whitledge, T.E., 1979. Ecologicalsignificance of fronts in the southeastern Bering Sea. In: Livingston, R.J. (Ed.),Ecological Processes in Coastal and Marine Systems. Plenum, New York,pp. 437–466.

Jin, M., Deal, C.J., Wang, J., Tanaka, N., Ikeda, M., 2006. Vertical mixing effects onthe phytoplankton bloom in the southeastern Bering Sea midshelf. J. Geophys.Res. 111 (C03002), 1–20.

Kinder, T.H., Schumacher, J.D., 1981. Hydrographic structure over the continentalshelf on the southeastern Bering Sea. In: Hood, D.W., Calder, J.A. (Eds.), TheEastern Bering Sea Shelf: Its Oceanography and Resources. University ofWashington Press, Seattle, pp. 31–51.

Kitaysky, A.S., Hunt Jr., G.L., Flint, E.N., Rubega, M.A., Decker, M.B., 2000. Resourceallocation in breeding seabirds: responses to fluctuations in their food supply.Mar. Ecol. Prog. Ser. 206, 283–296.

Lauth, R.R., 2010. Results of the 2009 Eastern Bering Sea Continental Shelf BottomTrawl Survey of Groundfish and Invertebrate Resources. U.S. Department ofCommerce, NOAA Technical Memo. NMFS-AFSC-204.

Lloyd, D.S., Davis, S.K., 1989. Biological information required for improvedmanagement of walleye pollock off Alaska. In: Proceedings of the InternationalSymposium on the Biology and Management of Walleye Pollock, Anchorage1988. University of Alaska Sea Grant AK-SG-89-1, Fairbanks, pp. 9–31.

Mallory, M.L., Gaston, A.J., Gilchrist, H.G., 2009. Sources of breeding seasonmortality in Canadian Arctic seabirds. Arctic 62, 333–341.

Mehlum, F., 2001. Crustaceans in the diet of adult common and Brunnich’sguillemots Uria aalge and Uria lomvia in the Barents Sea during the breedingperiod. Mar. Ornithol. 29, 19–22.

Mizobata, K., Saitoh, S., Wang, J., 2008. Interannual variability of summer biochem-ical enhancement in relation to mesoscale eddies at the shelf break in thevicinity of the Pribilof Islands. Bering Sea. Deep-Sea Res. II 55, 1717–1728.

Montevecchi, W.A., 1993. Birds as indicators of change in marine prey stocks. In:Furness, R.W., Greenwood, D.J. (Eds.), Birds as Monitors of EnvironmentalChange. Chapman and Hall, London, pp. 219–266.

Montevecchi, W.A., 2007. Binary responses of Northern gannets (Sula bassana) tochanging food web and oceanographic conditions. Mar. Ecol. Prog. Ser. 352,213–220.

Montevecchi, W.A., Myers, R.A., 1995. Prey harvests of seabirds reflect pelagic fishand squid abundance on multiple spatial and temporal scales. Mar. Ecol. Prog.Ser. 117, 1–9.

Mueter, F.J., Norcross, B.L., 2000. Changes in species composition of the demersalfish community in nearshore waters of Kodiak Island, Alaska. Can. J. Fish.Aquat. Sci. 57, 1169–1180.

Mueter, F.J., Ladd, C., Palmer, M.C., Norcross, B.L., 2006. Bottom-up and top-downcontrols of walleye pollock (Theragra chalcogramma) on the Eastern Bering Seashelf. Prog. Oceangr. 68, 152–183.

Mueter, F.J., Hunt Jr., G.L., Litzow, M.A., 2007. The Eastern Bering Sea shelf: AHighly Productive Seasonally Ice-Covered Sea. ICES CM2007/D:04.

National Marine Fisheries Service [NMFS], 2010. Fisheries of the United States,2009. National Marine Fisheries Service, Office of Science and Technology,Silver Spring.

National Research Council, 1996. The Bering Sea Ecosystem. National AcademyPress, Washington, DC.

Norris, J.R., 2000. Interannual and interdecadal variability in the storm track,cloudiness and sea surface temperature over the summertime North Pacific.J. Clim. 13, 422–430.

Perez, M.A., 1994. Calorimetry measurements of energy value of some Alaskanfishes and squids. U.S. Department Commerce, NOAA Technical Memo. NMFS-AFSC-32.

Piatt, J.F., Sydeman, W.J., Wiese, F., 2007. Introduction: a modern role for seabirdsas indicators. Mar. Ecol. Prog. Ser. 352, 199–204.

Quillfeldt, P., Michalik, A., Viet-Kohler, G., Strange, I.J., Masello, J.F., 2010. Inter-annual changes in diet and foraging trip lengths in a small pelagic seabird, thethin-billed prion Pachyptila belcheri. Mar. Biol. 157, 2043–2050.

Ressler, P.H., Robertis, A.D., Warren, J.D., Smith, J.N., Kotwicki, S., 2012. Developingan acoustic index of euphausiid abundance to understand trophic interactionsin the Bering Sea ecosystem. Deep-Sea Res. II 65–70, 184–195.

Rigor, I.G., Wallace, J.M., Colony, R.L., 2002. Response of sea ice to the ArcticOscillation. J. Clim. 15, 2648–2663.

Sandvick, H., Erikstad, K.E., Barrett, R.T., Yoccoz, N.G., 2005. The effect of climateon adult-survival in five species of North Atlantic seabirds. J. Anim. Ecol.74, 817–831.

Satterthwaite, W.H., Mangel, M., 2012. Using behavioral models to predict impactsof climate change on seabirds and fur seals. Deep-Sea Res. II 65–70, 304–315.

Schneider, D., Hunt, G.L., 1982. Carbon flux to seabirds in waters with differentmixing regimes in the southeastern Bering Sea. Mar. Biol. 67, 337–344.

Shaffer, S.A., Costa, D.P., Weimerskirch, H., 2003. Foraging effort in relationto the constraints of reproduction in free-ranging albatrosses. Funct. Ecol.17, 66–74.

Sinclair, E.H., Stabeno, P.J., 2002. Mesopelagic nekton and associated physics of thesoutheastern Bering Sea. Deep-Sea Res. II 49, 6127–6145.

Sinclair, E.H., Loughlin, T.R., Pearcy, W.G., 1994. Prey selection by northern furseals (Callorhinus ursinus) in the eastern Bering Sea. Fish. Bull. 92, 144–156.

Sinclair, E.H., Balanov, A.A., Kubodera, T., Radchenko, V.I., Fedorets, Y.A., 1999.Distribution and ecology of mesopelagic fishes and cephalopods. In: Loughlin,T.R., Ohtani, K. (Eds.), Dynamics of the Bering Sea: A Summary of PhysicalChemical and Biological characteristics, and a Synopsis of Research on theBering Sea. North Pacific Marine Science Organization (PICES), University ofAlaska Sea Grant AK-SG-99-03, Fairbanks, pp. 485–508.

Sinclair, E.H., Vlietstra, L.S., Johnson, D.S., Zeppelin, T.K., Byrd, G.V., Springer, A.M.,Ream, R.R., Hunt Jr., G.L., 2008. Patterns in prey use among fur seals andseabirds in the Pribilof Islands. Deep-Sea Res. II 55, 1897–1918.

Springer, A.M., 1992. A review: Walleye pollock in the North Pacific—how muchdifference do they really make? Fish. Oceanogr. 1, 80–96.

Springer, A.M., Roseneau, D.G., Lloyd, D.S., McRoy, C.P., Murphy, E.C., 1986. Seabirdresponses to fluctuating prey availability in the eastern Bering Sea. Mar. Ecol.Prog. Ser. 32, 1–12.

Springer, A.M., McRoy, C.P., Flint, M.V., 1996. The bering sea green belt: shelf-edgeprocesses and ecosystem production. Fish. Oceanogr. 5, 205–223.

Stabeno, P.J., Schumacher, J.D., Salo, S.A., Hunt Jr., G.L., Flint, M.V., 1999. Physicalenvironment around the Pribilof Islands. In: Loughlin, T.R., Ohtani, K. (Eds.),Dynamics of the Bering Sea. University of Alaska Sea Grant AK-SG-99-03,Fairbanks, pp. 193–215.

Stabeno, P.J., Kachel, N.B., Moore, S.E., Napp, J.M., Sigler, M., Zerbini, A.N., 2012.Comparison of warm and cold years on the southeastern Bering Sea shelf andsome implications for the ecosystem. Deep-Sea Res. II 65–70, 31–45.

Sullivan, M.E., Kachel, N.B., Mordy, C.W., Stabeno, P.J., 2008. The Pribilof Islands:temperature, salinity and nitrate during summer 2004. Deep-Sea Res. II 55,1729–1737.

Swartzman, G., Winter, A., Coyle, K., Brodeur, R., Buckley, T., Ciannelli, L., Hunt, G.,Ianelli, J., Macklin, A., 2005. Relationship of age-0 pollock abundance anddistribution around the Pribilof Islands, to other shelf regions of the easternBering Sea. Fish. Res. 74, 273–287.

Threfall, W., Eveleigh, E., Maunder, J.E., 1974. Seabird mortality in a storm. Auk91, 846–849.

Vliestra, L.S., Springer, A.M., Byrd, G.V., Hunt Jr., G.L., 2003. Final Report on thePopulation and Trophic Ecology of Seabirds in the Eastern Bering Sea, with anEmphasis on the Pribilof Islands colonies: Part I. NOAA Coastal Ocean Program,Southeastern Bering Sea Carrying Capacity Program (SEBSCC).

Wentz, F.J., Gentemann, C., Smith, D., Chelton, D., 2000. Satellite measurements ofsea surface temperature through clouds. Science 288, 847–850.

Wilson, R.P., 1984. An improved stomach pump for penguins and other seabirds.J. Field Ornithol. 55, 109–112.

Wyllie-Echeverria, T.W., 1996. A relationship between the distribution of one yearold pollock and sea ice characteristics, in: Brodeur, R.D., Livingston, P.A.,Loughlin, T.R., Hollowed, A.G. (Eds.), Ecology of Juvenile Walleye PollockTheragra chalcogramma. U.S. Department of Commerce, NOAA TechnicalReport. NMFS 126, pp. 47–56.

Wyllie-Echeverria, T., Wooster, W.S., 1998. Year-to-year variations in Bering Seaice cover and some consequences for fish distributions. Fish. Oceanogr 7,159–170.

Yang, M.-S., Aydin, K., Greig, A., Lang, G., Livingston, P., 2005. Historical review ofcapelin (Mallotus villosus) consumption in the Gulf of Alaska and easternBering Sea. U.S. Department of Commerce, NOAA Technical Memo. NMFS-AFSC-155.

Zeppelin, T.K., Dominic, T.J., Call, K.A., Orchard, T.J., Gudmundson, C.J., 2004. Sizesof walleye pollock (Theragra chalcogramma) and Atka mackerel (Pleurogram-mus monopterygius) consumed by the western stock of Steller sea lions(Eumetopias jubatus) in Alaska from 1998 to 2000. Fish. Bull. 102, 509–521.

Patterns of change in diets of two piscivorous seabird species during 35 years in the Pribilof...

Documents

Transcript of Patterns of change in diets of two piscivorous seabird species during 35 years in the Pribilof...