Oecologia (2002) 132:286-295 DOI 10. 1007/s00442-002-0945-1
Michael A. Litzow *John F. Piatt Alexander K. Prichard *Daniel D. Roby Responseof pigeonguillemots to variableabundance of high-lipidand low-lipidprey
Received: 17 July 2001 / Accepted: 5 April 2002 / Published online: 7 June 2002 ? Springer-Verlag2002
Abstract Populationsof the pigeon guillemot (Cepphus with reductions in beta chick growth rates, with no columba)and other piscivores have been in decline for decline in beta chick survival.In contrast,the proportion several decades in the Gulf of Alaska and Bering Sea, of nests experiencingbrood reductionin the Outer Bay and a decline in abundanceof lipid-richschooling fishes (demersaldiet) increased>300% duringyears of below- is hypothesized as the major cause. We tested this average demersal abundance, although demersal fish hypothesis by studying the breedingbiology of pigeon abundancevaried only 4-fold among years. Our results guillemotsduring 1995-1999 while simultaneouslymea- supportthe hypothesisthat recoveryof pigeon guillemot suring prey abundancewith beach seines and bottom populationsfrom the effects of the Exxon Valdezoil spill trawls.Our study area (Kachemak Bay, Alaska)compris- is limitedby availabilityof lipid-richprey. es two oceanographicallydistinct areas. Populations of a lipid-richschooling fish, Pacific sand lance (Ammodytes Keywords Climate Exxon Valdez Food limitation* hexapterus),were higher in the warmerInner Bay than Junkfood hypothesis Seabird in the colder OuterBay, and sand lance abundancewas higherduring warm years. Populationsof low-lipid con- tent demersalfishes were similar between areas. Chick Introduction survivalto age 15 days was 47% higherin the InnerBay (high-lipid diet) than in the OuterBay (low-lipid diet), The pigeon guillemot (Cepphus columba) is a semi- and estimatedreproductive success (chicksfledged nest-') colonial seabird of the auk family (Alcidae) that has was 62% higher in the InnerBay than in the OuterBay. been one of the species affectedby a widespreaddecline Chick provisioningrate (kJ chick-' h-') increasedwith in populationsof piscivorous seabirdsand pinnipedsin the proportionof sand lance in the diet (r2=0.21),as did the Bering Sea and Gulf of Alaska. The best census data growth rate (g day-') of younger (beta) chicks in two- for the species in the Gulf of Alaska are from Prince chick broods (r2=0.14). Pigeon guillemots in the Inner William Sound, where populations have declined ap- Bay switched to demersal prey during years of below- proximately72% between 1972 and 1989-1993 (Agler averagesand lance abundance,and these birdsreacted to et al. 1999). This decline was partly due to the Exxon 38-fold interannualchanges in sand lance abundance Valdezoil spill, which immediatelykilled approximately 500-1,500 pigeon guillemots in the Sound (5-14% of M.A. Litzow (Di) the estimatedoverall decline; Piatt et al. 1990). Popula- Instituteof Marine Science, University of California, tions in oiled areas had not recoveredas of 1998 (Irons Santa Cruz, CA 95064, USA et al. 2000; Golet et al. 2002). e-mail: [email protected] Fax: +1-907-4811703 However,pigeon guillemot populationsin the region may have also been affected by changes in fish abun- J.F. Piatt* M.A. Litzow dance due to decadal-scale climate variability. Water Alaska Science Center,U.S. Geological Survey, 1011 TudorRd., Anchorage,AK 99503, USA temperaturesin the Gulf of Alaska oscillate between cold and warmregimes that are associatedwith decadal- A.K. Prichard in of the Aleutian Instituteof Arctic Biology, University of Alaska Fairbanks, scale shifts the location atmospheric Fairbanks,AK 99775, USA low (the Pacific Decadal Oscillation [PDO]; Mantua et al. 1997; Hare and Mantua2000). A shift to a warm D.D. Roby Oregon CooperativeFish and Wildlife Research Unit, water regime in 1977 resulted in a >90% decline in U.S. Geological Survey and Departmentof Fisheries and Wildlife, populationsof capelin (Mallotus villosus), a lipid-rich Oregon State University, Corvallis, OR 97331, USA schooling fish, and a >250% increase in populationsof 287 lipid-poordemersal fishes (mostly Gadidaeand Pleuron- Bay, Alaska) is bisected into two oceanographicallydis- ectidae;Anderson and Piatt 1999). This dramaticecosys- tinct sections by the Homer Spit, and the two sections tem change apparentlycascaded throughhigher trophic supportdistinct fish communities(Abookire et al. 2000). levels, as the proportionof capelin and other lipid-rich As a result,guillemots in the InnerBay feed theirchicks fishes in diets of piscivorous birds and mammals de- mostly sand lance (59% of meals), while guillemots in clined, and populationsof these high trophiclevel con- the OuterBay feed their chicks mostly low-lipid demer- sumers declined as much as 95% (Piatt and Anderson sal fishes (94% of meals; Litzow et al. 2000). By work- 1996; Kuletzet al. 1997; Merricket al. 1997; Agler et al. ing in the two partsof the bay we could study the repro- 1999). A subsequentregime shift in the late 1980s did ductive response of pigeon guillemots to differences not returnthe Gulf of Alaska food web to a pre-1977 in food availability while controlling for confounding state (Springer1998; Hareand Mantua2000). temporaleffects and effects operatingover large spatial One hypothesis to explain these populationdeclines scales (hundredsof kilometers). Specifically, we tested holds thatthe energydensity (kJ g-') of low-lipiddemer- two hypotheses. We tested the prediction of the junk sal fishes is inadequateto meet the energeticdemands of food hypothesisthat the high lipid contentof sand lance birds and mammals,even when these fish are abundant should increase reproductivesuccess for pigeon guille- (the "junkfood" hypothesis;Piatt and Anderson 1996; mots with access to these prey. A possible disadvantage Merricket al. 1997;Rosen andTrites 2000). This is an at- of schooling prey not accounted for in the junk food tractiveexplanation for the decline in pigeon guillemots, hypothesisis that midwaterfishes may be more variable since the proportionof Pacific sand lance (Ammodytes in abundance than demersal fishes (Bradstreet and hexapterus),a lipid-rich schooling fish, in chick diets Brown 1985; Cairns 1987b). We therefore tested the declinedconcurrent with the populationdecline, and sand hypothesis that switching between sand lance and de- lance were replacedin diets by a varietyof low-lipidde- mersal prey allows pigeon guillemots to moderatethe mersalfishes (Hayes and Kuletz 1997). Furthermore,nat- effects of below-averagesand lance abundance.On the ural or oil spill-causeddeclines in sand lance abundance basis of past work we expected the effects of diet quality have been hypothesizedto constrainrecovery of guille- to be more noticeable in younger (beta) chicks than in mot populationsfrom the impactof the Exxon Valdezby older (alpha) chicks (Ainley et al. 1990a; Shultz and limiting the ability of the populationto replaceoil-killed Sydeman 1997; Cook et al. 2000). We tested our hypoth- individuals(Golet et al. 2002). Captive feeding experi- eses with data on fish abundancefrom beach seines and ments with tufted puffins (Fratercula cirrhata), black- bottomtrawls, and data on the rate of energyflow to the legged kittiwakes(Rissa trydactyla)and Steller sea lions nest, chick growth and survival rates, durationof chick (Eumetopiasjubatus) supportthe junk food hypothesis rearing,and reproductivesuccess. (Kitayskyet al. 1999; Romano 2000; Rosen and Trites 2000). These studies have found that even when diets of high- and low-lipidfish containthe same amountof ener- Materialsand methods gy, piscivores fed the low-lipid diet experiencereduced growth,reduced lipid reservesand elevatedlevels of cor- Study area ticosteronestress hormones.However, these experiments Kachemak Bay (59?35'N, 151019'W) is located on the east shore used diets comprising solely a gadid (walleye pollock of lower Cook Inlet, Alaska. The bay is bisected into oceano- Theragra chalcogramma) for low-lipid treatments. Diets graphically distinct inner and outer sections by the Homer Spit of piscivores in the wild typically comprise a mixture (Fig. 1). The Outer Bay is dominated by input from the Gulf of Alaska and is well mixed and relatively cold and saline, while the of low- and high-lipid prey (Hatch and Sanger 1992; Inner Bay is influenced by river runoff and tends to be more strati- Merrick et al. 1997; Suryan et al. 2000), so captive- fied, warmer,and less saline (Abookire et al. 2000). During sum- rearingexperiments using single-speciesdiets may over- mer 1996-1998 median monthly surface temperaturesaveraged estimate the importanceof high-lipid prey. Piscivores 0.9?C higher, and median monthly surface salinity averaged also possess a variety of 3.1 PSU lower, in the Inner Bay (Abookire et al. 2000). Pigeon behavioraland physiological guillemots nest in approximately 30 small colonies of 2-15 nests traits (e.g., flexible foraging effort, variableduration of each and in numerous solitary sites on the south shore of Kache- offspring rearing)that serve as buffers against variable mak Bay. Radio-tagged pigeon guillemots in the area foraged only food supply in the wild (Cairns 1987a; Burgerand Piatt within the part of the bay they nested in (Litzow et al., unpub- 1990; Monaghanet al. 1994; Uttley et al. 1994). Field lished data), so we consider Inner Bay and Outer Bay colonies to representindependent studiesof the responseof free-rangingpiscivores to vari- foraging habitats. ability in abundanceof high-lipidand low-lipid prey are needed to further understandingof the links between Prey abundance ocean climate variabilityand the populationecology of We measured prey abundance with trawls and seines set in areas high trophiclevel predators.However, few studies have where guillemots from study colonies were observed foraging. simultaneouslymeasured prey abundanceand demogra- Detailed methods for our measurements of fish abundance are phy of piscivoresin the high-latitudeNorth Pacific. reported elsewhere (Robards et al. 1999a; Abookire et al. 2000; In this paperwe relatethe breeding of Litzow et al. 2000). We measured sand lance abundance with biology pigeon catch per unit effort (CPUE; fish set-') in 226 beach seines set guillemotsto spatialand temporalvariability in oceanog- every 2 weeks at ten sites around guillemot colonies during chick raphy and prey abundance.Our study area (Kachemak rearing. We made a single set at each site during a given tide stage 288 and sampled from 16 June to 25 July in 1995 and from 1 June to 8 for at least 14 h (n=15 nests). From 1996 to 1999 we conducted all August in 1996-1999 (n=18 sets in 1995, 46 in 1996, 60 in 1997, day watches (0600-2200 hours; n=10 nests in 1996, 18 in 1997, 57 in 1998, and 45 in 1999). We classified sand lance >80 mm 18 in 1998, 8 in 1999). Provisioning adults were observed from total length as "forage size", as these are the size classes fed to anchored boats (using binoculars) or from blinds (using tele- guillemot chicks (Litzow et al. 2000). scopes), and each chick meal was identified to the lowest possible We measured demersal fish abundance with bottom trawl taxonomic level. CPUE (fish 1,000 m-2). We classified Lumpenus pricklebacks We measured diet composition as the proportion(by number) 80-200 mm total length and other demersal fish 80-150 mm total of sand lance in the diet. length as forage size (Litzow et al. 2000). We excluded hexa- grammids and gadids from trawl data because these fishes account for <1% of chick diets in Kachemak Bay (Litzow et al. 2000). Chick provisioning rates Bottom trawls were conducted once each year during mid-August (n=11 sets in 1996, 13 in 1997, 14 in 1998, and 11 in 1999). Pigeon guillemots provision chicks at a lower rate when they are We related sand lance abundance during 1996-1998 to previ- newly hatched or nearly fledged (Drent 1965; Emms and Verbeek ously published average July-August sea surface temperature 1991), so we only included nests with chicks aged 8-30 days in (SST) from the Inner and Outer Bay (Abookire et al. 2000). calculations of delivery rate (meals chick-' h-') and energy provi- We also used 1999 SST data that were collected using identical sioning rate (kJ chick-' h-'). methods (A. Abookire, personal communication). We visually estimated fish length relative to the length of a pigeon guillemot bill, in multiples of half bill-lengths. In order to calibrate our visual estimates of meal length we collected chick Nestling diet composition meals during 1996-1998 and measuredthe bill length (gape to tip) of six adult pigeon guillemots from Prince William Sound Pigeon guillemots carry single fish in their bills when provisioning (mean=43.2 mm). The length of collected meals (mean=128 mm, chicks, and usually rest on the water in front of the colony before SD=39 mm, n=79) was significantly greater than visual length delivering to the nest, making prey identification relatively easy. estimates generated with this standard for bill length (mean= We collected diet data for at least 1 year at each of ten guillemot 115 mm, SD=32 mm, n=2167; t2244=3.54,P<0.001). We therefore colonies (Fig. 1). We observed chick provisioning at two to five corrected our visual estimates upwards by multiplying them by nests during feeding watches. Watches were conducted during 128/115. We estimated energy provisioning rates by converting 3.5 h shifts distributed evenly across different tide stages and length to mass with length-weight regressions from fish caught in times of day (0600-2000 hours) in 1995. Each nest was watched trawls and seines (Table 1). We translatedthe resulting mass esti- mates into meal energy content using published values for prey taxa energy density. Energy density was calculated by averaging values for every species in a particularprey group that was sam- ALASKA t pled in appropriatesize classes in the cited references (Table 1; ~~~Inner r~iti~~~~~ - V|93 Van Pelt et al. 1997; Robards et al. 1999b; Anthony et al. 2000). This method provides an unbiased comparison of the energy con- Bay Area tent of diets containing different prey taxa (Emms and Verbeek * of* 1991). We included date in our calculations in situations when in- K - detail 93N formationon seasonal change in energy density was available.
Outer * e Number Bay of Nests Chick growth rates and reproductivesuccess We visited nests every 5 days to determine the fate of eggs and * 10 - 14 chicks and to weigh chicks with spring-loaded scales. During 1997 and 1998 we visited nests every 2 days after chicks were , 0 2 4 6 N 59?24' >30 days old so that we could accurately estimate fledging age. Chicks were assigned a rank based on their age: alpha (the older Kilometers of a two-chick brood), beta (the younger in a brood) or singleton 151048'W 151?36' 151V24' 151'12' 15100' (when only one egg in a clutch hatched). Because of the cryptic nature of nests, we often discovered nests only after chicks had Fig. 1 Location of pigeon guillemot (Cepphus columba) study hatched. In these situations we estimated age for chicks ?10 days nests and colonies in Kachemak Bay, Alaska. Stars indicate colo- old based on flattened wing chord, using measurementsfrom a set nies where provisioning watches occurred of known-age chicks for comparison (G. Divoky, personal com-
Table 1 Regressions of mass (y) on length (x) and wet mass energy density values used in estimating the energy content of pigeon guillemot nestling diets. Sources: a Robardset al. (1999b); b Anthony et al. (2000); c Van Pelt et al. (1997)
Prey group Scientific name Regression kJ g-I Source
Pacific sand lance (July) Ammodyteshexapterus y=(2xl106x)3.1224 5.25 a Capelin Mallotus villosus y=(3x 107X)3.684 5.04 b Pacific sand lance (August) Ammodyteshexapterus y=(2xl0-x)3.1224 5.02 a Prickleback Lumpenusspp. y=(4x I0"x)2.8977 4.76 b Gunnel Pholidae y=(8x 10-7x)3.2825 4.69 b Ronquil Ronquilusjordani, Bathymastersignatus y=(3xl0x)3'177 4.11 b Sculpin Cottidae y=(lxlO-5x)3.0128 4.10 b Salmonid Salmonidae y=(6x I0"x)30781 4.04 b Gadid Gadidae y=(7x 10-x)3.0241 3.31 b Flatfish Pleuronectidae y=(9x 10 x)30354 3.28 b Greenling Hexagrammidae y=(2x 10x)3.3399 3.16 b,c 289 munication). 83% of in of Age explained variability wing length 45 - Demersal fishes known-age chicks in this age range. We used growth rate (slope of P =0.19 linear regression of mass on age) of chicks age 5-20 days for comparisonsof chick growth (Emms and Verbeek E 1991). 30- We used the Mayfield method (Johnson 1979) to estimate nest- 0 ing success. We used a mean value for incubationlength (31 days; Ewins 1993). Chicks may fledge any time post-30 days (Ewins 15 1993), and since it is difficult to determine whether chicks older *ir than 30 days have fledged or been depredated,we calculated sur- 5/18 9/31 vival to age 30 days. To account for age dependant mortality we 0- calculated separate survival estimates for chicks age 1-15 days and 16-30 days. We make statistical comparisons of reproductive 500 success in terms of five parameters:clutch size, proportionof eggs Sand lance of that surviving incubation, proportion surviving eggs hatched, 400 P = 0.006 proportionof chicks surviving to age 15 days, and proportionof also chicks surviving 16-30 days. We present estimates of repro- 7 ductive success (chicks fledged nest-') that are the product of e 300- these five parameters.Reproductive success during the study was also affected by nest predation. Since predatorstypically removed 200- the entire contents of nests, we used the incidence of brood reduc- tion as a measure of reproductive success independent of preda- 100 -J tion. We defined brood reduction as situations where a beta chick died or disappeared from a nest while the alpha survived. Every 4/115 16/111 case of brood reduction that we observed occurred when chicks 0 _ | were s20 days old, so we estimated the incidence of brood reduc- InnerBay Outer Bay tion as the Mayfield daily survival rate of beta chicks age 0-20 days in nests with surviving alpha chicks. Fig. 2 Food availability for breeding pigeon guillemots in two ar- eas of Kachemak Bay, Alaska, 1995-1999. Catch per unit effort (CPUE) is from beach seines for sand lance, bottom trawls for Statistical analysis demersal fishes. Sample sizes at base of columns are number of sites / numberof individual sets. Data are grand means of average Daily survival rates estimated from the Mayfield method were CPUE at different site-years in the two areas, error bars ?1 SE. compared with a Z-test (Johnson 1979). When provisioning data P-values are from nested ANOVA were collected more than once from a single nest in one year we averaged data from different watches before analysis. We did not detect an effect of chick age on provisioning rate for chicks age 8-30 days (linear regression; FI,38=0.22,r2=0.01, P=0.65), so we ignored the effect of age in subsequentanalyses. Results Multiple beach seine sets made at a single site violate assump- tions of independence, so we treated individual sets as subsamples Prey availability with nested ANOVA (Zar 1999). We tested for year differences using year, site, and set(site) as factors, and tested for area differ- ences using area, site(area) and set(site) as factors. Bottom trawls We caught56,037 forage-sizesand lance and 936 forage- were set only once at each site in each year, so we used only used size demersalfish. The most commondemersal fishes in a nested approachfor our test of area differences. Our 5-year study trawls were rock sole (Pleuronectes bilineatus), yellow- did not provide adequate power to use a regression approach to fin sole (P. asper), crescent gunnels (Pholis laeta), examine the annual-scale response of pigeon guillemot breeding northernronquils (Ronquilusjordani), arctic shannies biology to variability in prey abundance,so we used a categorical approach. When ANOVA indicated significant differences in (Stichaeus punctatus), and flathead sole (Hippoglosso- CPUE among years we compared breeding parameters during ides elassodon). Sand lance CPUE was 76% higher in years of above-average and below-average CPUE for the domi- the Inner Bay than in the Outer Bay (Fig. 2; nested nant prey type in each study area. Classification of years as ANOVA, F1224=7.63, P=0.006). We failed to detect "above average" and "below average" was judged superiorto sta- tistical pairwise comparisons for purposes of defining food abun- a significant area difference in CPUE of all demersal dance categories because the extremely high variability in CPUE fishes (Fig. 2; nested ANOVA, F1,47=1.79, P=0.19), data and the logistical difficulty of conducting seines and trawls althoughsampling effort (49 sets) limited the power of meant that we could only detect extremely large pairwise differ- this comparison. ences. By classifying years as above- and below-average for each Sand prey type we were able to use all of our breeding biology data in lance CPUE varied38-fold amongyears (Fig. 3; tests for response to variabilityin food supply. Each year was clas- nested ANOVA, F4,212=3.58, P=0.008). Log-transformed sified as above- or below-average through comparison with the sand lance CPUE was above averagein 1995, 1998, and interannualgrand mean of log (x+l) transformedCPUE for each 1999, and below averagein 1996 and 1997. CPUE for all prey type. CPUE data were log-transformedto correct for hetero- demersal fishes as a group varied 4-fold among scedasticity, and because the dietary functional response of pigeon years guillemots shows a logarithmic relationship with variability in (Fig. 3; ANOVA, F345=3.49, P=0.02). Log-transformed abundanceof preferredprey (Litzow et al. 2000). One-tailed tests demersalCPUE was above average in 1997 and below were used for the junk food hypothesis and for tests of the effects averagein 1996, 1998 and 1999. of declining food abundance,since in both cases hypotheses made Interannualvariability in sand lance abundancewas directionalpredictions. All other tests were two-tailed. All propor- tional data used in regression analysis were arcsine-transformedto apparentlydriven by variabilityin ocean climate;CPUE satisfy assumptionsof normality.We set a=0.05, and we reportall increasedwith increasingSST (Fig. 4; ANCOVAof area means ?1 SE. and SST, SST: F1,5=11.22, P=0.02). We did not detect a 290
+ 75 Demersal fish P = 0.02 2 E 1,5 - I r =0.21 0 0-.------P= 0.001 0 -~50 ^) r-=030 0 0 0 0 0 Yea 0 :E ~ ~ 0 00,5 >02500 00 0 z 11 13 14 EL 00 0 ___ ~ 0 0
2 2 SnCEand lance 0.008 <30 r =0.14 0 P =0.047 7+
CD, ~~~~20 0000o 00 ------1------0 0 0 m 0 0~~~~~ 1 0 Om0,5- T 8/18 9/46 9/60 10/57 10/45 growtrate.2 20wt rae r o eachcsaeS2 as 1995 1996 1997 1998 1999 Year Proportionof sand lance in diet
Fig. 3 Interannualvariability in CPUE of sand lance and demersal Fig. 5 The effect of high-lipid prey on nestling provisioning and fishes in Kachemak Bay, Alaska, 1995-1999. Note log scales. growth rates. Growth rates are for beta chicks age 5-20 days. Dashed lines indicate interannualmeans used to classify years for Fourteen data points lie on the y-axis in the top panel, 12 in the tests of response of pigeon guillemots to variabilityin food supply. bottompanel Sample sizes given at base of demersal fish columns are the num- ber of bottom trawl sites; one set was made at each. Sample sizes given at base of sand lance columns are number of beach seine sites I total numberof sets. Error bars ?1 SE sal fish = 10.6?0.3 g, sand lance = 9.9+0.4 g, t819=1.38, P=0. 16). 0 InnerBay 0 Outer Bay Growthrates of beta chicks also increased with the presenceof sand lance in the diet (Fig. 5; linear regres- + 3 sion, n=28, r2=0.14, P=0.02). There was no area effect on growthrates of alphaand singletonchicks as a group; mean and standarderror were identical for the two areas (18.3?0.6 g day-'). However, beta chick growth was 37% higher in the sand lance diet area (Inner o73r 0.73 Bay = 18.9?0.6 g day-'; OuterBay = 13.8?1.4 g day-'; 0 0 one-tailedt5g=3.66, P<0.001). Averagegrowth rate of all 10 10,5 11 11,5 12 12,5 13 chicks in this study (17.7?0.4 g day-') was 9% greater Mean SST (? C) than the average growth rate (16.2?0.8 g day-') from five studies reviewed by Golet et al. (2000; one-sample Fig. 4 The effect of variable sea surface temperature(SST) on P=0.0001). sand lance abundance (log-transformed CPUE in beach seines). t160=3.93, Each point represents annual data from one study area. Tempera- Over all years of the study, chick survival to age ture data are July-August averages during 1996-1999 (Abookire 15 days was 47% higher in the Inner Bay (sand lance et al. 2000; A. Abookire, personal communication). ANCOVA diet) than in the OuterBay (demersaldiet; Z=3.75, one- showed a significant SST effect on CPUE (P=0.02) tailed P<0.001), and estimatedreproductive success was 62% higherin the InnerBay (0.47 chicks fledged nest-1, n=115) thanin the OuterBay (0.29 chicks fledged nest-', similar relationshipbetween SST and demersal CPUE n=136). Daily survivalrates of beta chicks were higher (F1 5=0.73, P=0.43). in the InnerBay (0.998?0.002, n=36) than in the Outer Bay (0.971?0.002, n=45; Z=3.72, one-tailed P<0.001). This difference in survival rate producedstrong spatial Breedingbiology in relationto diet patterns in brood reduction, which was observed in 3% of Inner Bay two-chick nests (n=36) and 36% of Estimatedchick provisioningrates (kJ chick-' h-1) in- Outer Bay nests (n=45). Chicks in the Inner Bay also creased with the proportionof sand lance in the diet fledged an averageof 3 days youngerthan chicks in the (Fig. 5; linearregression, n=46 nests, r2=0.21,P=0.001). Outer Bay (Inner Bay = 35+0.5 days; Outer Bay = 38+ This increase in provisioningrate was due to the high 0.6 days; t54=3.24,one-tailed P=0.001). We detected no caloric value of sand lance, as we failed to detect in- area differencein clutch size, egg survival, the propor- creases in prey delivery rate (linear regression, n=46 tion of survivingeggs that hatchedor survivalof chicks nests, r2=0.02,P=0.38), or estimatedmeal mass (demer- age 16-30 days (one-tailedP>0.25). 291
* Sand lance 0 Demersalfish * Sand lance 0 Demersal fish
0 514 362 -~~ 259 ~ 584 0,8 E ) 0,8 - COj 0,6- a 0,6 c E EC 0 o 0 0,2 0 0 1 Oa2 XE0. 2 0, 0
7 30- 0~~~~~ 20 - 1 2096 X
2 10 0- 301 C c,, 17 c 0 CD7 ~~~7 o
0 20~ 20 ~ ~~ -~~0) ~ 10 C\ p -C(D , 00,92
0)
0).X0 LA 10 10 - --- - 10 .C ~ ~~sn lanc CPU 0,98- 1--.7 Rp e pe10~~~~~~~~0 g l tf i o a a 0,960,96~~~~ - ~ ~ deera CPU ~0,98 0,94 -0.. A ct0,96 c 0,92- >~~~~~ 12 17 ~0,94CY 0,9 _ ~~~~~~~~~~o Above Below 'i 0,92 i C average average 17 19 0,9 - 1_- - _ _ - Log-transformed Below Above demersal CPUE average average
Fig. 6 Response of pigeon guillemots feeding on demersal prey to Log-transformed interannualvariability in prey abundance: Outer Kachemak Bay sand lance CPUE breeding parametersin years of above-average (1997) and below- average (1996, 1998, 1999) demersal fish abundance.P-values are Filg. 7 Response of pigeon guillemots feeding on sand lance to one-tailed, errorbars ? 1 SE interannual variability in prey abundance: Inner Kachemak Bay breeding parametersin years of above-average (1995, 1998, 1999) and below-average (1996, 1997) sand lance abundance. P-values are one-tailed, error bars ?1 SE Responsesto temporalprey variability
Pigeon guillemots in the two areas reactedto low food rates decreased 3% during below-average demersal years availability with adjustmentsto different breeding pa- (Fig. 6; Z=1.65, one-tailed P=O.05). This decline in daily rameters.In the OuterBay sand lance were rarelyfed to survival was large enough that the incidence of brood re- chicks, and parents did not increase the proportionof duction increased from 17% during the above-average sand lance in chick diets when demersalabundance de- demersal year (n= 12 nests) to 53% during below-average clined; demersalfish actuallymade a largercontribution years (n=17 nests). to diets during years of below-averagedemersal CPUE Pigeon guillemots in the Inner Bay more than doubled (Fig. 6; Fischerexact test, one-tailedP=0.98). Estimated the proportion of demersal fish in chick diets during provisioningrates fell 37% in the OuterBay duringbe- years of below-average sand lance abundance (Fig. 7; low-averagedemersal years (Fig. 6; t16=2.30,one-tailed Fischer exact test, one-tailed P
In spite of the high interannualvariability in sand lance Pigeon guillemotpopulations and ocean climate abundance,pigeon guillemots nesting at the Inner Bay (sand lance diet, switching to demersal fish) were less Spatial differences in physical oceanographybetween susceptible to the effects of decreased food abundance Inner and Outer KachemakBay created spatial differ- thanpigeon guillemotsat the OuterBay (entirelydemer- ences in fish abundance and community composition sal diet). Sand lance and demersalfish abundancefluctu- (Abookireet al. 2000). These differencesin fish distribu- ated asynchronouslyduring the study (Fig. 3), which tion led to differences in pigeon guillemot diet (Litzow would make switchingto demersalfishes duringyears of et al. 2000), which in turn producedgeographic differ- low sand lance abundancea particularlyeffective strate- ences in reproductivesuccess (this study). Thus, we gy. Guillemotsmay also have bufferedagainst temporal demonstratea link between ocean climate and pigeon variabilityin food supply with behaviornot measuredin guillemot demography,supporting the hypothesis that this study, such as flexible time-activity allocation changes in fish abundancedue to temporalvariability in (Burgerand Piatt 1990). oceanographyhave adversely affected pigeon guillemot Evidence that prey switching can act as an efficient populationsin the Gulf of Alaska since the 1977 PDO buffer for seabirdsagainst declines in the abundanceof regime shift. The link between pigeon guillemots and preferredprey is equivocal, and maintenanceof some oceanographyis also demonstratedby the effect of tem- thresholdamount of lipid-richfish in the diet may be a poral variabilityin SST on the abundanceof high-lipid key factor.Thus black-leggedkittiwakes in Norway and prey (Fig. 4). Alaska were able to maintainhigh reproductivesuccess Sand lance populationsare apparentlyhigher in Inner by switching among alternatelipid-rich prey (Barrettet KachemakBay because the combinationof nutrientin- al. 1987; Suryanet al. 2000). Common and thick-billed put from rivers and increasedstratification results in in- murres (Uria aalge and U. lomvia) in Canada were able creased primaryproduction (Abookire et al. 2000). Our to maintainconstant reproductive success when the pro- findings that guillemot reproductivesuccess was higher portionof lipid-richcapelin in chick diets declined from in a warm area and that sand lance were more abundant -75% to 12-45% (for common) and 20-30% to -3% during warm years complement studies of interannual (for thick-billed;Bryant et al. 1999). Conversely,other climate variabilityin southeastAlaska (Paul et al. 1991; species in California and Norway (pelagic cormorant Paul and Coyle 1993) and the Sea of Okhotsk(Kitaysky Phalacrocorax pelagicus, pigeon guillemot, Atlantic and Golubova2000). These studies have shown that the puffin Fraterculaarctica) suffered large declines in re- meso-zooplanktonprey of lipid-richschooling fishes are productivesuccess when lipid-richfish were not avail- more abundant during warm years than during cold able to replacescarce preferredprey (Ainley et al. 1995; years, and that piscivore reproductivesuccess increases Barrett1996). as a result. We found that pigeon guillemots were well described Althoughseveral hypotheseshave been advanced,the by the Cairns(1987a) model of responseto variableprey mechanismslinking fish populationswith decadal-scale abundance,which hypothesizes that breeding seabirds climate variabilityare not currentlyunderstood (Mantua should respond to increasingly severe food shortages et al. 1997; Andersonand Piatt 1999; Hare and Mantua with modification of increasingly critical reproductive 2000; Mueter and Norcross 2000). The complexity of parameters.Pigeon guillemots feeding mostly on lipid- this problem is highlighted by an apparentparadox: rich sand lance sufferedreduced beta chick growthrates while both high-lipid prey abundanceand piscivore re- duringyears of below-averagesand lance abundance,but productivesuccess were higher in warm spatial regimes did not experience a decline in beta chick survival (this study) and during warm annualregimes (Kitaysky (Fig. 7). Although there was some evidence that provi- and Golubova 2000; this study), warm decadal-scale sioning rates for this group declined during years of PDO regimes are associatedwith populationdeclines of below-average sand lance abundance(Fig. 7), the de- high-lipid fishes and piscivores (Anderson and Piatt cline was not as severe as that experiencedby the low- 1999). 294 Acknowledgements For their work in the field we are grateful to Cairns DK (1981) Breeding, feeding, and chick growth of the Yumi Arimitsu, Dave Black, Bryan Duggan, Matt Kopec, Jennifer black guillemot (Cepphus grylle) in southern Quebec. Can Litzow, Jeff Moy, April Nielsen, Cynthia Restrepo, Pam Seiser, Field Nat 95:312-318 Becka Seymour, John Shook, Brian Smith and Sadie Wright. Cairns DK (1987a) Seabirds as indicatorsof marinefood supplies. Thanks to Alisa Abookire, Jared Figurski and Martin Robards Biol Ocean 5:261-271 for sharing their work on fish populations. Bradford Keitt, Greg Cairns DK (1987b) Diet and foraging ecology of black guillemots Snedgen and Tom Van Pelt helped with logistics. George Divoky on northeasternHudson Bay. Can J Zool 65:1257-1263 provided wing length data, Alisa Abookire shared SST data and Cairns DK (1992) Populationregulation of seabird colonies. Curr Claire Armisteadmade the study site figure. Funding and logistical Ornithol9:37-61 supportwere provided by the Alaska Science Center (U.S. Geolog- Cook MI, Monaghan P, Burns P (2000) Effects of short-term ical Survey), the Exxon Valdez Oil Spill Trustee Council, the hungerand competitive asymmetryon facultativeaggression in Alaska MaritimeNational Wildlife Refuge (U.S. Fish and Wildlife nestling black guillemots Cepphusgrylle. Behav Ecol 11:282- Service), the U.S. Minerals ManagementService, the University of 287 Alaska Fairbanks,Oregon State University,and the Alaska Depart- Crawford RJM, Dyer BM (1995) Responses by four seabird spe- ment of Fish and Game. Special thanks to Sera Baxter and Mike cies to a fluctuating availability of Cape Anchovy Engraulis and Connie Geagel for their hospitality in Kachemak Bay. This capensis off South Africa. Ibis 137:329-339 paper was greatly improved by comments from M.A.L.'s thesis Drent RH (1965) Breeding biology of the pigeon guillemot, committee: MarkCarr, Don Croll, Jim Estes and Bernie Tershy;by Cepphuscolumba. Ardea 53:99-160 discussion with Alisa Abookire, Josh Donlan, Greg Golet, Sasha Drent RH, Daan S (1980) The prudent parent: energetic adjust- Kitaysky and Mike Shultz; and by comments by Mark Chappell ments in avian breeding. Ardea 68:225-252 and an anonymousreviewer. Emms SK, Verbeek NAM (1991) Brood size, food provisioning and chick growth in the Pigeon Guillemot Cepphus columba. Condor 93:943-95 1 Ewins PJ (1990) The diet of black guillemots Cepphus grylle in References Shetland. Holarct Ecol 13:90-97 Ewins PJ (1993) Pigeon Guillemot (Cepphus columba). In: Abookire AA, Piatt JF, Robards MD (2000) Nearshore fish distri- Poole A, Gill F (eds) The Birds of North America, No 49. The butions in an Alaskan estuary in relation to stratification,tem- Birds of North America, Philadelphia perature,and salinity. EstuarCoast Shelf Sci 50:45-59 Golet GH, Kuletz KJ, Roby DD, Irons DB (2000) Adult prey Agler BA, Kendall SJ, Irons DB, Klosiewski SP (1999) Declines in choice affects chick growth and reproductive success of marine bird populationsin Prince William Sound, Alaska coin- Pigeon Guillemots. Auk 117:82-91 cident with a climactic regime shift. Waterbirds22:98-103 Golet GH, Seiser PE, McGuire AD, Roby DD, Fischer JB, Ainley DG, Boekelheide RJ, Morrell SH, Strong CS (1990a) Kuletz KJ, Irons DB, Dean TA, Jewett SC, Newman SH Pigeon guillemot. In: Ainley DG, Boekelheide RJ (eds) Sea- (2002) Long-term direct and indirect effects of the Exxon birds of the Farallon Islands. Stanford University Press, Stan- Valdez oil spill on Pigeon Guillemots in Prince William ford, Calif. pp 276-305 Sound, Alaska. Mar Ecol Prog Ser (in press) Ainley DG, Strong CS, Penniman TM, Boekelheide RJ (1990b) Hare SR, MantuaNJ (2000) Empiricalevidence for North Pacific Feeding ecology. In: Ainley DG, Boekelheide RJ (eds) Sea- regime shifts in 1977 and 1989. Prog Oceanogr47:103-145 birds of the Farallon Islands. Stanford University Press, Stan- Hatch SA, Sanger GA (1992) Puffins as samplers of juvenile ford, Calif. pp 51-127 pollock and other forage fish in the Gulf of Alaska. Mar Ecol Ainley DG, Sydeman WJ, Norton J (1995) Upper trophic level Prog Ser 80:1-14 predatorsindicate interannualnegative and positive anomalies Hayes DL, Kuletz KJ (1997) Decline of pigeon guillemot popula- in the California Current food web. Mar Ecol Prog Ser tions in Prince William Sound, Alaska, and apparentchanges 118:69-79 in distributionand abundanceof their prey. In: Forage Fishes Anderson PJ, Piatt JF (1999) Trophicreorganization in the Gulf of in Marine Ecosystems, Proc Int Symp Role Forage Fish Mar Alaska following ocean climate regime shift. Mar Ecol Prog Ecosystems. Alaska Sea GrantCollege Program,University of Ser 189:117-123 Alaska, Fairbanks,pp 699-702 Anthony JA, Roby DD, Turco KR (2000) Lipid content and ener- Irons DB, Kendall SJ, Erickson WP, McDonald LL, Lance BK gy density of forage fishes from the northernGulf of Alaska. (2000) Nine years after the Exxon Valdezoil spill: effects on J Exp Mar Biol Ecol 248:53-78 marine bird populations in Prince William Sound, Alaska. Barrett RT (1996) Prey harvest, chick growth, and production of Condor 102:723-737 three seabird species on Bleiksoy, North Norway, during years Johnson DH (1979) Estimatingnest success: the Mayfield method of vafiable food availability.Can Wild Serv Occ Pap 91:20-26 and an alternative.Auk 96:651-661 BarrettRT, Anker-Nilssen T (1997) Egg-laying, chick growth and Kitaysky AS, Golubova EG (2000) Climate change causes con- food of Black Guillemots Cepphus grylle in North Norway. trasting trends in reproductiveperformance of planktivorous Faun Norv Ser C Cinc 20:69-79 and piscivorous alcids. J Anim Ecol 69:248-262 Barrett RT, Anker-Nilssen T, Rikardsen F, Valde K, Rov N, Kitaysky AS, Piatt JF, Wingfield JC, Romano M (1999) The adre- Vader W (1987) The food, growth and fledging success of nocortical stress-responseof Black-legged Kittiwakechicks in Norwegian puffin chicks Fratercula arctica in 1980-1983. relation to dietaryrestrictions. J Comp Physiol B 169:303-3 10 Ornis Scand 18:73-83 Kuletz KJ (1983) Mechanisms and consequences of foraging be- Blake BF (1986) Diet and fish stock availability as possible fac- havior in a population of breeding Pigeon Guillemots. MSc tors in the mass death of auks in the North Sea. J Exp Mar thesis, University of California,Irvine Biol Ecol 76:89-103 Kuletz KJ, Irons DB, Agler BA, Piatt JF, Duffy DC (1997) Long- Bradstreet MSW, Brown RGB (1985) Feeding ecology of the term changes in diets and populationsof piscivorous birds and Atlantic Alcidae. In: Nettleship DN, Birkhead TR (eds) The mammals in Prince William Sound, Alaska. In: Forage fishes Atlantic Alcidae. Academic Press, London, pp 263-318 in marine ecosystems, Proc Int Symp Role Forage Fish Mar Bryant R, Jones IL, Hipfner JM (1999) Response to changes in Ecosyst. Alaska Sea Grant College Program, University of prey availability by Common Murres and Thick-billed Murres Alaska, Fairbanks,pp 703-706 at the Gannet Islands, Labrador.Can J Zool 77:1278-1287 Litzow MA, Piatt JE, Abookire AA, Prichard AK, Robards MD Burger AE, Piatt JF (1990) Flexible time budgets in breeding (2000) Monitoring temporal and spatial variability in sandeel common murres: buffers against variable prey abundance. (Ammodyteshexapterus) abundance with pigeon guillemot Stud Avian Biol 14:71-83 (Cepphuscolumba) diets. ICES J Mar Sci 57:576-586 295 Mantua NJ, Hare SR, Zhang Y, Wallace J, Francis R (1997) Robards MD, Anthony JA, Rose GA, Piatt JF (1999b) Changes in A Pacific interdecadal climate oscillation with impacts on proximate composition and somatic energy content for Pacific salmon production.Bull Am Meteorol Soc 78:1069-1079 sand lance (Ammodytes hexapterus) from Kachemak Bay, Merrick RL, Chumbley MK, Byrd GV (1997) Diet diversity of Alaska relative to maturity and season. J Exp Mar Biol Ecol Steller sea lions (Eumetopias jubatus) and their population 242:245-258 decline in Alaska: a potential relationship. Can J Fish Aquat Roby DD (1991) Diet and postnatal energetics in convergent taxa Sci 54:1342-1348 of plankton-feedingseabirds. Auk 108:131-146 Monaghan P, Walton P, Wanless S, Uttley JD, Burns MD (1994) Romano MD (2000) Effects of diet on growth and development of Effects of prey abundance on the foraging behaviour, diving nestling seabirds. MSc thesis, Oregon State University, Cor- efficiency and time allocation of breeding Guillemots Uria vallis aalge. Ibis 136:214-222 Rosen DAS, Trites AW (2000) Pollock and the decline of Steller Mueter FJ, Norcross BL (2000) Changes in species composition of sea lions: testing the junk-food hypothesis. Can J Zool the demersal fish community in nearshore waters of Kodiak 78:1243-1250 Island, Alaska. Can J Fish Aquat Sci 57:1169-1180 Shultz MT, Sydeman WJ (1997) Pre-fledging weight recession in Orians GH, Pearson NE (1979) On the theory of central place Pigeon Guillemots on Southeast Farallon Island, California. foraging. In: Horn DJ, Mitchell RD, Stairs GR (eds) Analysis Colon Waterbirds20:436-448 of ecological systems. Ohio State University Press, Columbus, Springer AM (1998) Is it all climate change? Why marine bird pp 155-177 and mammal populations fluctuate in the North Pacific. In: Paul A, Coyle K (1993) Taxa composition and biomass of the sur- Holloway G, Muller P, Henderson D (eds) Biotic impacts of face-dwelling crustaceans during spring pycnocline formation extratropical climate variability in the Pacific. University of in Auke Bay, southeastern Alaska. J Crustac Biol 13:594- Hawaii, Manoa, pp 109-119 600 Suryan RM, Irons DB, Benson J (2000) Prey switching and vari- Paul A, Coyle K, Haldorson L (1991) Interannualvariations in able foraging strategies of Black-legged Kittiwakes and the copepod nauplii prey of larval fish in an Alaskan bay. ICES effect on reproductivesuccess. Condor 102:374-384 J Mar Sci 48:157-165 Sydeman WJ, Emslie SD (1992) Effects of parentalage on hatch- Piatt JF, Anderson P (1996) Response of common murres to the ing asynchrony, egg size, and third-chick disadvantage in Exxon Valdez Oil Spill and long-term changes in the Gulf WesternGulls. Auk 109:242-248 of Alaska marine ecosystem. Am Fish Soc Symp 18:720- Uttley JD, Walton P, Monaghan P, Austin G (1994) The effects 737 of food abundance on breeding performance and adult time Piatt JF, Lensink CJ, Butler W, Kendziorek M, Nysewander DR budgets of Guillemots Uria aalge. Ibis 136:205-213 (1990) Immediate impact of the Exxon Valdez oil spill on Van Pelt TI, Piatt JF, Lance BK, Roby DD (1997) Proximate marinebirds. Auk 107:387-397 composition and energy density of some North Pacific forage RobardsMD, Piatt JF, Kettle AB, Abookire AA (1999a) Temporal fishes. Comp Biochem Physiol 18A: 1393-1398 and geographic variation in fish communities of lower Cook Zar JH (1999) Biostatistical analysis, 4th edn. Prentice Hall, Inlet, Alaska. Fish Bull 97:962-977 Upper Saddle River, N.J.