ALBATROSS POPULATIONS IN PERIL: A POPULATION …€¦ · ALBATROSS POPULATIONS IN PERIL: A POPULATION TRAJECTORY FOR BLACK-BROWED ALBATROSSES AT SOUTH GEORGIA JENNIFER M. ARNOLD,1,3

419

Ecological Applications, 16(1), 2006, pp. 419–432q 2006 by the Ecological Society of America

ALBATROSS POPULATIONS IN PERIL: A POPULATION TRAJECTORY FORBLACK-BROWED ALBATROSSES AT SOUTH GEORGIA

JENNIFER M. ARNOLD,1,3 SOLANGE BRAULT,1 AND JOHN P. CROXALL2

1Department of Biology, University of Massachusetts, 100 Morrissey Boulevard, Boston, Massachusetts 02125 USA2British Antarctic Survey, Natural Environment Research Council, High Cross, Madingley Road, Cambridge CB3 OET UK

Abstract. Simulation modeling was used to reconstruct Black-browed Albatross (Di-omedea melanophris) population trends. Close approximations to observed data were ac-complished by annually varying survival rates, reproductive success, and probabilities ofreturning to breed given success in previous years. The temporal shift in annual valuescoincided with the start of longline fishing at South Georgia and potential changes in krillabundance. We used 23 years of demographic data from long-term studies of a breedingcolony of this species at Bird Island, South Georgia, to validate our model. When we usedannual parameter estimates for survival, reproductive success, and probabilities of returningto breed given success in previous years, our model trajectory closely followed the observedchanges in breeding population size over time. Population growth rate was below replace-ment (lambda , 1) in most years and was most sensitive to changes in adult survival. Thissupports the recent IUCN uplisting of this species from ‘‘Vulnerable’’ to ‘‘Endangered.’’Comparison of pre-1988 and post-1988 demography (before and after the inception of alongline fishery in the breeding area) reveals a decrease in lambda from 0.963 to 0.910. Alife table response experiment (LTRE) showed that this decline in lambda was caused mostlyby declines in survival of adults. If 1988–1998 demographic rates are maintained, the modelpredicts a 98% chance of a population of fewer than 25 pairs within 78 years. For thispopulation to recover to a status under which it could be ‘‘delisted,’’ a 10% increase insurvival of all age classes would be needed.

Key words: albatross; Black-browed Albatross; bycatch; Diomedea melanophris; endangeredspecies; IUCN Red List; longline fishing; LTRE; Patagonian toothfish; population model; South Geor-gia; viability analysis.

INTRODUCTION

Over the past two decades, it has become increasingclear that numbers of albatrosses are declining and theirpopulations are at risk. Of the 24 species of albatrossesworldwide, 21 species are showing declines in greaterthan 50% of their populations and most species arenow classified as globally threatened (Croxall andGales 1998, Gales 1998, BirdLife International 2000,2004, IUCN 2002). Threats to albatross populationsinclude fisheries-related mortality, changes in preybase, plastic ingestions, human predation/disturbance,nonhuman predators, fire, floods, volcanic activity,habitat degradation, oil/chemical pollution, and disease(Croxall 1998, Gales 1998).

Of these threats, perhaps the greatest and one of thebest documented is mortality in longline fisheries(Brothers 1991, Cherel et al. 1996, Weimerskirch et al.1997, 2000, Prince et al. 1998, Tuck et al. 2001, In-chausti and Weimerskirch 2001). Albatrosses are par-ticularly vulnerable to mortality from longlines because

Manuscript received 23 October 2003; revised 13 April 2005;accepted 31 May 2005; final version received 7 July 2005. Cor-responding Editor: A. B. Hollowed.

3 Present address: USGS Patuxent Wildlife Research Cen-ter, 12100 Beech Forest Road, Laurel, Maryland 20708 USA.E-mail: [email protected]

most species are attracted to the bait and offal of fishingvessels, taking advantage of a supplementary foodsource (Thompson and Riddy 1995, Reid et al. 1996).Additionally, they are pelagic species that tend to feedover shelf-slope areas, the same areas used by longlinevessels (Gales 1998).

Although most threats to albatrosses are now wellrecognized (Croxall 1998, Croxall and Gales 1998,Gales 1998), the direct link between these threats andobserved population declines has been harder to es-tablish. Effective management actions rely on an un-derstanding of the magnitude of the population-leveleffect of specific threats, the sensitivity of the popu-lation to changes in vital rates of individuals in dif-ferent life history stages, and the effect of threats onspecific life history stages. In this paper, we introduceand describe a model based on the Black-browed Al-batross (Diomedea melanophris; see Plate 1), but withgeneral applicability to all albatross species.

With 680 000 pairs breeding worldwide (Croxall andGales 1998), the Black-browed Albatross is the mostpopulous albatross species. However, mainly becauseof recent declines in size of the breeding colonies atthe Falkland Islands, the species was moved from anIUCN listing of Lower Risk/Near Threatened to Vul-nerable (Croxall and Gales 1998, BirdLife International

420 JENNIFER M. ARNOLD ET AL. Ecological ApplicationsVol. 16, No. 1

PLATE 1. Offal from fishing boats provides an attractivefood source for Black-browed Albatrosses (Diomedea me-lanophris). Photo credit: Stephen Oswald.

2000, Woehler et al. 2002); in the 2004 revision of theIUCN Red List, this status was changed to Endangered(BirdLife International 2004). The breeding populationat the Falkland Islands, comprising ;80% of all Black-browed Albatrosses, has declined by 25% in the past20 years (Gales 1998, Huin 2001, 2002, IUCN 2002).The Bird Island, South Georgia, population (12% ofall Black-browed Albatrosses) has also declined overa similar period. These declines may be related to fish-ing activities around their breeding grounds and win-tering areas (Croxall and Gales 1998), as well as tochanges in the availability of Antarctic krill (Euphausiasuperba), the primary food source at South Georgia(Croxall et al. 1997).

Studies have shown that Black-browed Albatross isthe most frequent bycatch species (Brothers 1991,Gales et al. 1998, Weimerskirch et al. 2000). The highbycatch level of Black-browed Albatross in SouthernOcean fisheries may be explained by some combinationof their feeding habits, shelf-slope distribution, pro-pensity for following fishing vessels, and abundance.Black-browed Albatross colonies at the Falklands andelsewhere have benefited over the years from scaveng-ing opportunities brought about by trawl fisheries inthe vicinity of their breeding grounds (Thompson 1992,Thompson and Riddy 1995, Prince et al. 1998, Huin2002), and recent data (Sullivan and Reid 2004) basedon observations of current interactions around the Falk-lands suggest that substantial levels of albatross by-catch may have been associated with these fisheries asa result of collision with netsonde cables. In the Falk-

land Islands, finfish and squid fisheries were estimatedto have provided up to 15% of the forage base for theBlack-browed Albatross in the early 1990s (Thompsonand Riddy 1995). A potentially beneficial finfish/trawlfishery, such as that described for the Falklands, beganin South Georgia in the mid-1960s, but had collapsedby the late 1980s (Croxall et al. 1998, Prince et al.1998). Although the potential benefits no longer ex-isted, the bycatch-related mortality remained an in-creasingly important issue because of new and ex-panding longline fisheries in this area and elsewhere.

Although direct estimates of bycatch levels fromthese fisheries have been difficult to obtain, some ev-idence suggests that no longline fishery existed in thesummer foraging area prior to 1987/1988. Longlinefishing for Patagonian toothfish (Dissostichus elegi-noides) in the South Georgia area started in 1988/1989.Effort increased rapidly over the next decade. However,despite indications of high levels of seabird (especiallyalbatross) bycatch, reliable annual estimates only be-came available from 1997 onward (CCAMLR 2004,Croxall and Nicol 2004). In 1997, the estimated seabirdbycatch in the South Georgia area (CCAMLR Sub-area48.3) was 5755 birds, 40.4% of which were Black-browed Albatross. This suggested a minimum bycatchestimate of 2325 Black-browed Albatross (CCAMLR1997).

The temporal pattern in population decline coincideswith increases in longline fishing effort in Sub-area48.3 and other areas where birds once benefited fromtrawl fishery discards (Croxall et al. 1998, Prince et al.1998, Huin 2001, 2002). The shelf and shelf-slope ar-eas, to which Black-browed Albatrosses appear largelyrestricted, are becoming more frequently used by long-line vessels throughout the year (Prince et al. 1998,Nel et al. 2002). With regard to the South Georgiacolonies, the Patagonian toothfish fishery operated insummer foraging areas from 1988 (in recent years, theregulated fishery has been restricted to the wintermonths, particularly to reduce the magnitude of seabirdbycatch [CCAMLR 2004]); longline tuna and hake fish-eries have operated since the mid-1990s in their win-tering grounds, mainly in South African shelf-slopewaters (Barnes et al. 1997, Prince et al. 1998); andtrawl fisheries have operated in the Benguela Currentsystem for decades. Although these observations verifythat fishing activity occurred in regions occupied byBlack-browed Albatross, we cannot assess the influ-ence of these fisheries on Black-browed Albatross pop-ulations because no data on bycatch numbers or ratesare available.

Additional factors also may affect the populationtrend of Black-browed Albatross. Recent literature hassuggested a decline in sea-ice cover that may relate todeclines in krill populations (see Loeb et al. [1997] andAtkinson et al. [2004] for temporal trend data). Krilland krill-eating fish constitute a significant part of thediet of the Black-browed Albatross (Reid et al. 1996,

February 2006 421ALBATROSS POPULATIONS IN PERIL

TABLE 1. Model parameters, values used in the mean matrix calculation, sensitivities of lambda to changes in those values,and mean values for each parameter during the two study periods.

Parameter Definition

Estimatefor mean

matrix

Sensitivity oflambda toparameter

Meanpre-1988

Meanpost-1988

sJ juvenile survival for years 1–3 (early juveniles) 0.497 0.021 0.599 0.385sLJ juvenile survival for years 4 to recruit (late juve-

niles)0.780 0.033 † †

sA adult female survival (individuals become adultsonce they enter breeding population for one ormore seasons)

0.933 0.965 0.957 0.909

fB reproductive success (expressed as production offledged young) of breeding females

0.136 0.026 0.168 0.100

mLJ probability of moving from a late juvenile pseudosta-ge to the next stage or pseudostage in time t 1 1

0.470 0.009 no change no change

mSNBa probability of moving from ‘‘successful nonbreedera’’ (SNBa) to breeder in time t 1 1

0.571 0.0001 0.611 0.528

mUNBa probability of moving from ‘‘unsuccessful non-breeder a’’ (UNBa) to breeder in time t 1 1

0.710 0.0005 0.739 0.679

mNBb probability of moving from nonbreeder b (NBb) tobreeder in time t 1 1

0.990 0.0002 no change no change

PrS probability of returning as breeder in year t 1 1 ifsuccessful in year t

0.770 0.0005 0.765 0.773

PrU probability of returning as breeder in year t 1 1 ifunsuccessful year t

0.809 0.003 0.805 0.812

Notes: ‘‘No change’’ indicates that parameters are estimated based on mean and standard distribution of age at first breedingand overall probability of remaining in a nonbreeder stage for more than one year. In our model these values are held constantover time. The four highest sensitivities are in boldface.

† For late juvenile survival, we only had estimates for some pre-1988 years; thus, we used the mean estimate for yearsthat we did not have data. For this reason, no interperiod comparison can be done.

Croxall et al. 1997). The direct and indirect dependenceof Black-browed Albatross on krill is illustrated by thepositive correlation between poor ice years and yearsof low reproductive success for Black-browed Alba-tross (Prince 1985, Croxall et al. 1988, 1998, 2002,Reid and Croxall 2001). Thus, changes in krill abun-dance may also impact albatross populations throughreproductive success. Untangling the relationship be-tween krill and albatross population declines is verycomplicated, and the direct link between sea-ice chang-es, krill, and seabird population effects is difficult tomake. The most recent literature, however, suggeststhat declines in krill abundance will mainly affect re-productive success rather than adult survival. Thus,although we will not directly address impacts of krilldeclines in this study, our model is designed to illus-trate potential contributions of changes in reproductivesuccess to population trends and to relate those to thepotential contributions of changes in survival.

Given the time trends in survival and reproduction,and the current threats to albatross persistence, the aimsof this paper are: (1) to develop an age/stage classifiedmatrix population model with a biological structuredesigned to be generally applicable to all albatross spe-cies; (2) to validate this model using 23 years of lifehistory data (although not all vital rate data were col-lected to the resolution of one year; see Methods) forthe Black-browed Albatross at South Georgia (1976–1998); (3) to explore whether time trends in demo-graphic rates coincided with the start of the longlinefisheries (for Patagonian toothfish) at South Georgia

1988/1989; (4) to place fisheries threats in the contextof other threats to species persistence (including chang-es in prey base); (5) to determine the sensitivity oflambda (l, population rate of increase) to changes invital rates, and to the life history parameters that havemade the greatest contribution to differences in lambdaover the past two decades; (6) to project time to ex-tinction under different scenarios of reduced reproduc-tive success or increased mortality; and (7) to examinethe status of the Black-browed Albatross in terms ofIUCN criteria for threatened species.

METHODS

Field data

We used 23 years of demographic data from ColonyH, one of 25 Black-browed Albatross colonies at BirdIsland, to validate this albatross model (see Table 1 forparameter estimates). Together, the Bird Island coloniescomprise 15% of the total South Georgia population,which includes some 15 other main breeding sites. Col-ony philopatry of adults, once they have bred, is veryhigh, with no movements between different sites atSouth Georgia and very rare movements even betweenadjacent colonies at Bird Island (Prince et al. 1994; J.P. Croxall, P. A. Prince, A. G. Wood, and T. Burg,unpublished data). Most chicks return to breed at theirnatal colony, but some (5–10% of a cohort) move toadjacent colonies and a few (,1% of a cohort) to othersites at South Georgia. None of .30 000 Black-browedAlbatrosses banded at Bird Island have ever been re-


trapped at a breeding location away from South Geor-gia. However, genetic similarities between birds fromsouthern Chile, South Georgia, and Indian Ocean sub-antarctic islands indicate that some interchange has oc-curred, at least over historical time (Burg and Croxall2001). For modeling purposes, however, South Georgiacan be regarded as a closed population and Bird Islandvery nearly so. For details of the breeding colonies atBird Island, see Hunter et al. (1982).

Annual survival rates of adults and juveniles wereestimated using capture–recapture models (see Princeet al. 1994, Croxall et al. 1998). Annual values forbreeding frequency, population size, reproductive suc-cess, and adult and juvenile survival from 1976 to 1996are summarized in Rothery and Prince (1990) andCroxall et al. (1998). For the present analysis, all dataup to (and including) 1998 were used, based on the2000 update of the Bird Island data. Each annual updatecorrects errors in the historical data, adds data for thelatest year, and reruns the analyses described in Croxallet al. (1998), thereby incorporating the data from thelatest year of the capture–recapture fieldwork. Annualvalues were available and used for all vital rates, exceptin the case of late and early juvenile survival. For latejuvenile survival (sLJ), calculations for this vital ratewere limited by age at maturity. By year 2000, manymarked birds may have been alive, but had not yetbegun breeding; thus, only individuals from the 1984cohort and earlier (i.e., age 16 years and older) wereincluded in the analysis. For this vital rate, we used amean value of 0.780 (the mean of all estimates for latejuvenile survival until 1984). For early juvenile sur-vival (sJ), we were missing estimates for 1976, 1980,1991, and 1995–1998; for years without annual esti-mates, we used a mean value, 0.584, calculated fromsurvival estimates for the years 1984 and earlier. Forboth early and late juvenile stages, these values arelikely to be overestimates of survival rate as survivaldeclined after 1984. Thus, they are likely to lead to anunderestimate of the extent of total population declineand of the relative role of juvenile survival in this de-cline.

Model structure

The population model structure (Fig. 1) is a mix ofage and stage classes, and assumes equal survival prob-abilities for males and females. It has a pre-breedingcensus format (Caswell 2001), meaning that populationvalues are assessed, with an annual time step, prior tobreeding events. New individuals produced during abreeding event in year t are thus first counted in yeart 1 1 (J2; Fig. 1). Prior to reproductive maturity, in-dividuals move through two annual early juvenile clas-ses, J2 and J3, with identical survival probability, thenthrough a late juvenile stage (LJ) lasting at least fouryears before birds can become breeders at age 7 yearsor older. Some remain in the LJ stage over a longerperiod; Black-browed Albatrosses at Bird Island begin

breeding between the ages of 7 and 19 (J. P. Croxall,P. A. Prince, and A. G. Wood, unpublished data). Thisprocess is modeled by a late juvenile stage made up offour identical ‘‘pseudostages’’ (see Fig. 1) of a durationdescribed by a negative binomial distribution (withskew to the right; Caswell 2001). This formulationcauses the LJ stage to last a minimum of four yearsand to have a mean duration of 4/mLJ , where mLJ is theprobability of moving to the next pseudostage. As aresult, individuals in a cohort will enter the breedingstage gradually rather than at a fixed age (see Validationof model). Early and late juvenile birds have differentsurvival probabilities (the annual survival rates); alladult stages have the same survival probability.

Like other albatross species, these birds are long-lived and slow to reproduce. They lay a single egg perbreeding season, and occasionally take years off be-tween breeding attempts (Croxall et al. 1998). Themodeled breeding stage, B, has a one-year duration,and is the only reproductive stage. Whether albatrossesattempt to breed in a given year is influenced by thesuccess of their breeding attempt in the previous year.Thus, following a breeding attempt, surviving individ-uals have three possible fates in the model. They can(1) become successful nonbreeders, SNBa, after a suc-cessful breeding event, modeled with the function fB(12 PrS), where PrS is the probability of remaining in thebreeder stage if successful at breeding; (2) become un-successful nonbreeders, UNBa, after an unsuccessfulbreeding event modeled with the function (1 2 PrU)(12 fB), where PrU is the probability of remaining in thebreeder stage if unsuccessful; or (3) remain in thebreeding stage B for another year, modeled with thefunction (fBPrS) 1 ((1 2 fB)PrU). Although the differ-ence between successful and unsuccessful breeders inthe probability of remaining in the breeder stage israther small for the Black-browed Albatross, it isknown to be much greater for other albatross species(Croxall et al. 1998); thus, we have designed the modelto accommodate such differences.

Three nonbreeder stages are specified. If birds in thenonbreeder stages SNBa and UNBa in the year follow-ing a breeding attempt remain nonbreeders for one ormore additional years, they enter the nonbreeder stage(NBb). As long as they survive, they remain in thisstage until they become breeders again with probabilitymNBb. An adult bird can thus re-enter the breeder stagefrom all three nonbreeder stages.

Unlike the case of the Wandering Albatross Diome-dea exulans (Tuck et al. 2001), no density dependencein parameter values was detected (Croxall et al. 1998).Thus, the model is expressed as a density-independentmatrix A that provides the parameters for projecting avector n of all stages from time t to t 1 1 as

n 5 Ant11 t (1)

(Caswell 2001). Equations formulating each elementai,j of matrix A are given in the Appendix. These equa-


FIG. 1. Life cycle diagram for the Black-browed Albatross. Key to variables: m is the probability of transition to thenext stage; s is survival, by stage; fB is reproductive success (expressed as production of fledged young) of breeding females;and PrS and PrU are the probabilities of remaining in the breeder stage if successful or unsuccessful, respectively, at breeding(see Table 1).

tions can easily be modified to include any form ofdensity dependence.

Model analyses

We obtained lambda, the population rate of increase,from analytical eigenvalue analysis of the projectionmatrix A (Caswell 2001). For population projectionsthrough the census period (1976–1998), annual A ma-trices were constructed using parameter values fromcensus and mark–recapture analysis results. Sensitiv-ities and elasticities (proportional sensitivities) oflambda to A matrix elements were calculated as

]lS 5 (2)i, j ]ai,j

a ]li, jE 5 . (3)i, j l ]ai,j

To obtain sensitivities of lambda to lower-level param-eters x, i.e., the survival probabilities, maturation prob-

abilities, and reproductive success listed in Table 1, thechain rule was used:

]a]l ]l i, j5 . (4)O

]x ]a ]xi,j i, j

A Life Table Response Experiment (LTRE; Caswell2001) was performed to analyze the effects on thechange in lambda of observed changes in parametervalues between two periods, pre-1988 and post-1988,which represent two different fisheries regimes andpossibly different underlying productivity regimes (At-kinson et al. 2004). The LTRE results measure the con-tribution of each lower-level parameter value to dif-ferences in lambda between the two periods. The con-tribution Cx of each parameter x to the difference inlambda between the two periods was calculated as

]a]l i, jC 5 (x 2 x ) . (5)Ox pre-1988 post-1988 ]a ]xi,j i, j

The contribution of a parameter, such as juvenile sur-


FIG. 2. Comparison of annual breeding population censusvalues for Black-browed Albatross from colony H at BirdIsland (gray line) and model predictions (black line) usingannual values for parameter estimates (r 5 0.959).

vival, to a change in lambda, is therefore a product ofthe change in value of this parameter between the twoperiods, and of the sensitivity of lambda to this param-eter. This calculation goes one step further than a sen-sitivity analysis in asking, ‘‘What actual change in thelife cycle between the two periods has had the largestrole in the observed change in lambda?’’

We explored the effect of the population decline onpopulation age structure by comparing the stable stagedistributions resulting from the mean A matrices forthe two periods.

Pseudoextinction was defined as a population declineto a size of .25 breeding pairs. This was an arbitrarydesignation based on a minimum number greater thanone pair, due to the colonial nature of the species. Be-cause of potential Allee effects, a larger number mayactually be necessary for persistence of the species, butis unknown. We defined, for early and late periods, fourpotential scenarios for changes in vital rates: (1) nochange from annual values in each period; (2) annuallate juvenile survival rates reduced by 2% (equivalentto the modal annual change in adult survival between1976 and 1998); (3) reproductive success reduced by40.5% (approximately equal to the decline observed inthis parameter between pre- and post-1988 periods);(4) adult survival reduced by 2% (equivalent to themodal annual change in this rate between 1976 and1998). To measure pseudoextinction probability, i.e.,the risk of population decline to fewer than 25 pairs,1000 simulation runs of 250 years were performed withvalues from each period and each scenario, and theproportion of the runs resulting in pseudoextinctionafter a given time horizon was noted. The simulationswere run by picking at random (with equal probability),with replacement, one of the annual projection matricesat every time step of a simulation, to multiply with thevector nt; this was done in turn using the set of annualA matrices from the pre-1988 period and from the post-1988 period. The initial population size was 9539 in-dividuals (i.e., the size of the population at Bird Island,South Georgia in 1995) distributed among the stagesaccording to the stable stage distribution obtained bythe mean matrix for the period of concern.

Recovery scenarios, based on return to the thresholdlambda required to change the IUCN Red List statusof this species, were obtained by running 1000 10-yearsimulations of the post-1988 period; as in the pseudo-extinction calculations, yearly parameter matrices wereselected randomly. For these simulations, we asked,‘‘What change in survival for all stages is necessaryfor this species to be either within or outside thresholdlambda values required for designation as Endangeredand Vulnerable under IUCN rules?’’ (IUCN 2002). Al-though IUCN criteria are not formulated in terms ofthreshold or minimum lambda values, they could be.Thus, a 50% population decline over 10 years (an ‘‘En-dangered’’ criterion) is obtained with a lambda valueof 0.933 or lower for 10 years; a 20% decline (a ‘‘Vul-

nerable’’ criterion) is obtained with a lambda of 0.978or lower (Caswell 2001).

RESULTS

Validation of model

We validated the fit of the model in two ways. Weprojected the trajectory of colony H at Bird Island,starting with its initial size of 230 pairs and using aprojection matrix with annual parameter values from1976 to 1998 (Table 1) obtained from the literature andunpublished data. For this analysis, the actual annualvital rates calculated for survival, fecundity, and otherdemographic parameters were used, rather than meansestimates from each period. We compared this with theactual trajectory for this colony from breeding colonycensuses during the same period. The comparison (Fig.2) shows a very close fit of model projections to annualfield counts (Pearson correlation coefficient r 5 0.959).The mean rate of increase, (lambda 5 0.936) is equiv-alent to a 6.4% population decline per year.

We also ran the model with only one initial cohortof 1000 individuals entering the first juvenile stage(J2), to validate the model predictions of progressionthrough juvenile stages and age at maturity. The rangeof age at maturity, as projected by our model, closelymimics the natural population (Fig. 3). Long-term dataindicate that, in the natural population, individuals re-cruit between ages 7 and 19 years (age at recruitment10.3 6 0.25 years, mean 6 SE) (J. P. Croxall, P. A.Prince, and A. G. Wood, unpublished data). In ourmodel, the first individuals enter the breeding popu-lation at age 7, most individuals have recruited by age17, and the mean age at recruitment is between ages11 and 12 years. Although the model results in a latermean age at maturity than was measured on Bird Island,our sensitivity analysis shows that the population rateof increase is very insensitive to changes in mLJ, theparameter controlling how fast individuals move from


FIG. 3. Progression of a cohort of 1000 individual Black-browed Albatrosses over 20 years of model projection usingthe mean parameter estimates.

FIG. 4. Lambda values (annual rate of increase) for 1975–1998, as predicted by the model when annual parameter es-timates for Black-browed Albatross are used.

late juvenile to adult age classes. Both observed andmodeled distributions of age at maturity are skewed tothe right; in our model, 50% of the individuals firstbreed by age 9 years.

Sensitivity

Sensitivity analysis (Table 1) suggests that changesin adult survival will have the greatest effect on lamb-da. With a sensitivity of 0.965, a small change in adultsurvival will produce a large change in lambda. Rel-ative to changes in adult survival, lambda is consid-erably less sensitive to reproductive success (sensitiv-ity 5 0.026), late juvenile survival (sensitivity 50.033), and early juvenile survival (sensitivity 50.021). Sensitivities of all remaining matrix elementsdo not exceed 0.01.

Comparing pre-1988 to post-1988 conditions

For both periods, most annual lambda values arebelow 1.0, indicating population decline (Fig. 4). How-ever, pre-1988 values for lambda are significantly high-er than post-1988 values (t21 5 3.42, P 5 0.001), sug-gesting that conditions have deteriorated in recent de-cades. Using the mean parameter values to obtain anoverall lambda for each period (pre-1988, 0.963; post-1988, 0.910), we obtain a 0.053 decline in lambda be-tween the two periods. Our basic sensitivity analysisshows that population growth rate is most sensitive tochanges in adult survival. This parameter has changedfrom a pre-1988 annual rate of 0.957 to a post-1988rate of 0.909 (Table 1). However, to understand whichfactors have made the greatest contribution to the de-cline in lambda between the two periods, it is necessaryto use both the sensitivity of lambda to the parametersand the measured difference in the parameters betweenthe two periods. We used a life table response exper-iment (LTRE) to examine the combined effects of thesefactors (Fig. 5). Although the decline in adult survivalover the two periods is considerably less than that ofearly juvenile survival, adult survival has the greatestimpact on the decline in lambda because of lambda’shigh sensitivity to this parameter. Conversely, the largereduction in early juvenile survival measured over thesame period has had a lesser influence on lambda be-cause of the low sensitivity of lambda to this rate.Declines in reproductive success had the third strongesteffect on lambda due to both mid-level sensitivity andthe large decrease in this parameter’s value betweenthe two periods (Fig. 5).

The pre-1988 stable stage distribution is composedof 10.7% early juveniles, 10.7% late juveniles, and78.6% adults, with 79.5% of these being breeders. Post-


FIG. 5. Life Table Response Experiment (LTRE) for the Black-browed Albatross at colony H, South Georgia. Plotsrepresent lambda sensitivity values from the matrix analysis, differences in actual measured values for vital rates betweenthe two periods, and contributions of each vital rate to the observed decline in lambda, respectively. Except for SB (successfulbreeders) and UB (unsuccessful breeders), stage abbreviations are listed in Fig. 1. Declines in survival of breeding adultsmake the greatest contribution to the observed decline in lambda between the pre-1988 and post-1988 periods. ‘‘Return’’indicates return to breeding. Note that contributions from LJ survival, LJ maturation, and NBb return cannot be calculatedbecause data on differences in these values between the two periods are not available.

1988, the distribution has shifted to 4.5% early juve-niles, 2.7% late juveniles, and 92.8% adults, with79.6% of these being breeders. This suggests that thepopulation is moving toward a state in which a largerpercentage of individuals will be in the adult stages.As a result, lambda becomes even more sensitive tochanges in adult survival in the post-1988 period com-pared to the pre-1988 period, and less sensitive to othervital rates. This is demonstrated by comparing the dif-ferences in sensitivities between early juvenile andadult survival for the two periods (pre-1988, 0.0292vs. 0.9454; post-1988, 0.0124 vs. 0.9837). We ac-knowledge that the evidence for this shift is built onthe assumption that the population reaches an asymp-totic state, and that this is unlikely, given the time frameof our study and the variance in the vital rates. How-ever, it is clear that if conditions continue as they are,this directional shift to a greater percentage of adultsin the population will continue to be realized.

Pseudoextinction probabilities

If vital rates remain the same as current rates (i.e.,post-1988 conditions), there is a 98% probability ofpseudoextinction in 78 years (Fig. 6). Even if the sit-uation improves to pre-1988 conditions, the model pre-dicts a 98% risk of pseudoextinction in 193 years. Asexpected from our sensitivity analysis, the greatest ex-tinction threat to the Black-browed Albatross comesfrom an increase in adult mortality. If an additional 128mature females were killed annually, the equivalent ofa 2% decrease in adult survival, there would be a 98%chance of pseudoextinction in just 63 years (15 feweryears than status quo) in a post-1988 regime, or in 124

years (69 fewer years than status quo) under pre-1988conditions. Declines in juvenile survival did not con-tribute substantially to increased extinction risk. How-ever, adult and juvenile survival declines are treated asseparate scenarios in Fig. 6, but probably would besimultaneous in reality, with additive effects on ex-tinction risk. Additional declines in reproductive suc-cess, or increased incidence of years with no repro-duction, will increase the probability of pseudoextinc-tion. A further reduction in reproductive success,40.5%, equal to the decline observed between pre- andpost-1988 periods, would only reduce time to pseu-doextinction by two or three years compared to statusquo under current conditions. If pre-1988 conditionsprevailed, this reduction would decrease the time topseudoextinction by 16 years. Thus, under current con-ditions, the effect of a 2% decrease in adult survivalon pseudoextinction probability is considerably largerthan a 40.5% decrease in fecundity.

Recovery scenarios

Under status quo at post-1988 conditions, .90% ofsimulated 10-year runs had a rate of increase lowerthan 0.933, the threshold lambda value equivalent tothe IUCN criterion of a 50% decline over 10 years usedto list a species as ‘‘Endangered’’ (Fig. 7a). For thisspecies to be downlisted to the ‘‘Vulnerable’’ category,its lambda would have to rise above threshold valuefor an ‘‘Endangered’’ listing, but remain below thethreshold value for a ‘‘Vulnerable’’ listing, 0.978(equivalent to a 20% population decline over 10 years).This downlisting would require a 5% increase in sur-vival rates for all stages compared to their post-1988


FIG. 6. Time to pseudoextinction (n , 25 pairs) for the Black-browed Albatross predicted by the model under scenarios(lines 1–4) based on potential effects of bycatch-related mortality or changes in krill abundance: (A) pre-1988 parameterestimates; (B) post-1988 parameter estimates. Simulations were run using an initial breeding population of 9539 pairs (1995–1996 estimate of the South Georgia population). Line 1 shows measured annual values; line 2 shows annual survival of LateJuveniles decreased by 2% (equivalent to the modal annual change in adult survival between 1976 and 1998); line 3 showsreproductive success decreased by 40.5% (approximately equal to the decline in this parameter between the pre-1988 andpost-1988 time periods); line 4 shows annual survival of adults decreased by 2% (equivalent to the modal annual change inthis rate between 1976 and 1998).

FIG. 7. Results from three simulation scenarios (each of1000 runs of 10 years) testing population status of Bird IslandBlack-browed Albatross relative to IUCN criteria for threat-ened species. The two vertical lines across all three graphsrepresent the threshold lambda values that correspond to cri-teria for ‘‘Endangered’’ status (l , 0.933) and for ‘‘Vulner-able’’ status (l , 0.978). (A) Parameter values are the actualdata from the post-1988 period; 92.9% of runs have l ,0.933. (B) Parameter values are post-1988 data, with survivalrates of all stages increased by 5%; 6.3% of runs have l ,0.933 and 91% have l , 0.978. (C) Parameter values arepost-1988 data, with survival rates of all stages increased by10%; all runs have l . 0.933 and 6.8% have l , 0.978.

values (Fig. 7b). For a change in population status from‘‘Endangered’’ to complete delisting, the rate of in-crease must be above the ‘‘Vulnerable’’ threshold of0.978 most of the time. For this to happen, all survivalrates must be augmented by at least 10% of their post-1988 values (Fig. 7c); under these conditions, the riskof obtaining a rate of increase below 0.978 is 6.8%.

DISCUSSION

Population modeling in albatross conservation

Natural resource managers are increasingly facedwith making regulatory and conservation-based deci-sions for long-lived species that spend most of theirlives in remote places. Even if accurate information onvital rates is available, understanding the sensitivity ofthese populations to anthropogenic or environmentalinfluences is difficult. Their remote and wide-ranginglifestyles make such effects difficult to observe in thefield. For long-lived species, population-level effectsmay not be detectable for years or even decades. Thispaper illustrates the effectiveness of using a populationtrajectory model to elucidate population-level effectsof anthropogenic stressors before the threat of extinc-tion.

Our results demonstrate that even the most populousspecies of albatross is at risk of extinction within a fewdecades. Although there are many threats to albatrosspopulations, the most likely factors contributing to de-clines of Black-browed Albatross are mortality from


fishing operations and a decrease in krill abundance.These two factors, respectively, will mainly decreaseadult survival and reproductive success (although thedirect link between reproductive success and krill abun-dance has been difficult to make). Increased mortalityfrom fishing is also expected to have an important in-fluence on juvenile survival either directly (long-lin-ing) or indirectly (prey availability) (Croxall et al.1998). Using data from the 12 years prior to and 11years after the inception of longline fisheries in theirsummer feeding areas, we took advantage of the rareopportunity to evaluate population-level changes con-current with this major change in fishing practice. Thisuse of matrix models in a natural experiment has en-abled us to describe a quantitative link between pop-ulation declines and longline fisheries that has beenhistorically difficult to demonstrate because of the rel-atively frequent occurrence of illegal fisheries and de-ficient reporting of bycatch incidents. Our model showsthat, in this species, declines in adult survival havemade by far the greatest contribution to population de-clines in the last two decades. From our projections, adecrease in adult survival by 2% reduces the time topseudoextinction by 15 years, bringing time to pseu-doextinction to 63 years. Changes in reproductive suc-cess or juvenile survival have also caused a reductionin lambda, but this impact is relatively small comparedto that on adult survival (Table 1, Fig. 5).

Our model suggests that the Black-browed Albatrosscolony at South Georgia has been declining since atleast 1976 and that the rate of decrease has becomesteeper in the last decade, as illustrated by the 0.053decrease in lambda between the pre- and post-1988periods. In the last decade of our study, 1988–1998,longline fisheries have intensified (beginning in thesummer foraging range), while trawl fisheries, of po-tential net benefit, have diminished within the range ofthe Black-browed Albatross (Croxall et al. 1998, Princeet al. 1998, Nel et al. 2002, Woehler et al. 2002). Al-though it seems clear that the current longline effortsare a threat to albatrosses worldwide, recent data fromtrawl fisheries near the Falklands suggest that earlytrawl fisheries, assumed to provide a positive net ben-efit for albatrosses, actually may have been responsiblefor a larger mortality than previously assumed (Sulli-van and Reid 2004). As a result, colonies exposed totrawl-fishing activities may not have experienced a netbenefit in demographic terms. This may explain thenegative lambda values in the early years of our study.Additionally, although trawl fisheries around SouthGeorgia are no longer in operation, trawl fisheries inthe range of other albatross colonies (e.g., FalklandIslands), may be a source of both juvenile and adultmortalities (Sullivan and Reid 2004).

Black-browed Albatrosses are highly dependent onkrill and krill-eating forage fish. In 1986, the diet ofBlack-browed Albatrosses at South Georgia consistedof 39.4% crustaceans and 29.5% fish. The most nu-

merous fish species identified in their diet (77%) wasPatagonotothen guntheri, which feeds predominantlyon krill. The remaining 23% of the fish species in theirdiet were also predominantly krill-eating fish (Reid etal. 1996). The increased frequency of years with limitedsea-ice cover and low krill production throughout ourstudy (Loeb et al. 1997, Reid et al. 1999, Reid andCroxall 2001) may have affected adult reproductivesuccess through poorer provisioning of offspring byparents (Huin et al. 2000, Reid and Croxall 2001). Thereduced reproductive success that we measured be-tween the two decades in the LTRE analysis (Fig. 5,Differences) is more likely to have been caused by suchenvironmental changes than by human causes.

Between 2000 and 2002, the global conservationstatus of the Black-browed Albatross was changedfrom ‘‘Low risk/near Threatened’’ to ‘‘Vulnerable’’(BirdLife International 2000, IUCN 2002) and veryrecently it was changed to ‘‘Endangered’’ (BirdLifeInternational 2004). Under IUCN criteria, a populationdecline by 50% or more within 10 years or three gen-erations qualifies for listing a species as ‘‘Endan-gered.’’ This corresponds to a population growth rateof l 5 0.933, or an annual rate of decline of 6.7% overa 10-year period (i.e., l 5 0.51/10; Caswell 2001). Ourresults show a mean lambda of 0.910 from 1988 to1998 for the Bird Island colony modeled here, belowthe endangered threshold. Under the recovery scenarioof no change from post-1988 conditions (Fig. 7a),lambda can be as low as 0.860 and there is only a 7%chance that lambda will be higher than the ‘‘Endan-gered’’ threshold. Assuming that the dynamics of thiscolony are a good proxy for most other colonies world-wide, our results strongly support the listing of Black-browed Albatross as ‘‘Endangered.’’

Although our model is based on data from only onecolony at Bird Island, all other colonies there havedecreased since 1990 (Croxall et al. 1998); a recentsurvey reports decreases from most other sites at SouthGeorgia (S. Poncet and G. Robertson, unpublisheddata). Furthermore, the species is declining in otherparts of its range as well. The largest population ofBlack-browed Albatrosses, at the Falklands, has de-creased by 25% in the last 20 years, probably due tolongline (and possibly trawl) fishing (Huin 2001, Sul-livan and Reid 2004). Populations in the French sub-antarctic islands in the Indian Ocean have also shownrecent declines (Weimerskirch and Jouventin 1998).Heard Island, Indian Ocean (Australia), and CampbellIsland, New Zealand both have shown slight increasesin recent years, but are small populations that are likelyto show declines in the near future as a result of in-creased longline fishing effort in their year-round feed-ing areas (Woehler et al. 2002).

Applicability of model to other albatross speciesand populations

The model structure was devised to be applicable toalbatross species in general. Four main characteristics


of albatross life cycles served as our blueprint: the longjuvenile period, variability in age at maturity, intra-specific differences in breeding frequency between suc-cessful and unsuccessful breeders, and interspecific dif-ferences in the probabilities of skipping years betweenbreeding attempts. Among albatross species, the av-erage juvenile period varies between 8 and 12 yearsand no birds have been known to become mature beforethe 4th year. By having the early juvenile age classesbe followed by a late juvenile set of classes, we separatebirds that cannot yet become mature from those thatcan. The structure of the late juvenile block of stagesprovides an appropriate minimal delay for the first birdsto reach maturity (seven years for Black-browed Al-batrosses at South Georgia); each of the four pseudo-stages is used as an age class, with longer delays pro-vided by the probabilities of remaining in each pseudo-stage. The range of ages at which maturity is reachedcan be controlled with the parameters of the negativebinomial distribution for the pseudostages: the numberof pseudostages (the parameter k of the distribution),and the rate of passage from one pseudostage to thenext (the parameter mLJ). This rate is estimated fromthe mean and variance of age at maturity in the naturalpopulation.

Although the Black-browed Albatross does not showa large difference between successful and unsuccessfulbreeders in rate of return to breeding, bigger differencesexist in other species. For example, Wandering Alba-trosses are essentially biennial breeders: breeders thatare successful in year n will not breed again in year n1 1 (Croxall et al. 1998). This is not surprising for along-lived species for which the average period be-tween mean fledging date and initiation of laying forthe next season is only about 5 days (Tickell 1968).Unsuccessful breeders may breed in year n 1 1 if theirbreeding attempt fails less than halfway through theseason (year n). We allow for such differences to bemodeled through differences in rates of return to breed-ing for successful (1 2 mSNBa) and unsuccessful breed-ers (1 2 mUNBa); the structure also allows for differentsurvival rates between these two classes. A biennialbreeding schedule is modeled by setting the rate ofreturn to breeding for successful breeders to zero, withthe probabilities of failed breeding early in the seasonaffecting the rate of returning to breed in the followingyear. This creates a two-year loop of ‘‘breeder–non-breeder a’’ and back to breeder. However, for specieswith biennial breeding, not all individuals breed in yearn 1 2. Thus, the final stage of the model, nonbreedingadults (NBb) that were nonbreeders the previous year,provides a flexible number of years between breedingattempts.

Tuck et al. (2001) modeled the impact of fisherieson two populations of Wandering Albatrosses, andfound that their model was a good fit for the Bird Islandand Crozet Islands populations up to the late 1980s,but thereafter did not accurately describe the Bird Is-

land population. Although they designed it for a some-what different purpose (estimation of bycatch by fish-eries), Tuck et al. (2001) used a life history structuresimilar to the one that we suggest for our model. Theyused age, rather than stage structure (a different rep-resentation of the life cycle), and applied a single ageat first breeding rather than a distribution. Density de-pendence, applied to juvenile survival for the Wan-dering Albatross, similarly can be incorporated in ourmodel structure by setting lower-level parameters asfunctions of the numbers of individuals in one or morestages. Future work on effects of density dependenceand individual differences in reproductive success iswarranted.

Management implications

The results of our sensitivity analysis suggest thatmanagement actions should be aimed at increasingadult and juvenile survival rates. As discussed earlier,bycatch in longline fisheries may be an importantsource of mortality for Black-browed Albatross. Ob-servers should be deployed on fishing vessels to doc-ument encounter and mortality rates of Black-browedAlbatross. Likewise, reliable catch-reporting protocolsshould be established to enable an analysis of the sea-sonal and spatial overlap of commercial fishing oper-ations and Black-browed Albatross. However, evenwith optimum observer reporting from regulated fish-eries, the lack of bycatch data from illegal and unreg-ulated fisheries may still compromise attempts to ac-curately model albatross–fishery interactions and theirimplications for albatross demography. Adults are like-ly to come into contact with longliners both locallywhere they breed and more widely in the areas thatthey frequent during the nonbreeding season. From ourmodel projections, the population age structure hasshifted from 78.6% adults to 92.8% adults, suggestingan increase in the sensitivity of lambda to changes inadult mortality. Increased mortality, however, is alsoan issue for both juvenile groups. Young juveniles areprobably critically dependent on the conditions in theBenguela current, South Africa, which is their primarydestination after fledging (Prince et al. 1998). As aresult, they may be particularly attracted to fishery op-erations in this region (traditionally trawls, but morerecently including longliners) (Phillips et al. 2005).Older juveniles will have survived their first experi-ences in the Benguela, but a proportion travel farthereast into the Indian Ocean and have more chance ofinteracting with the greater range of fisheries, espe-cially longline, that are widespread in this region.

Most incidental mortality of albatrosses associatedwith fisheries can be drastically reduced by relativelysimple modifications to fishing practice: appropriatetreatment of waste and offal, simple deterrents to trawlnets, or changes in trawling technique. Mitigation ofalbatross bycatch in longline fisheries is more complex,but invariably involves some combination of streamer


lines to prevent birds from accessing the baited hooksbefore they sink, additional weighting on the line sothat hooks sink, and fishing at night or outside thealbatross breeding season. Where these measures havebeen used in concert, as in South Georgia waters, a100-fold reduction in the number of birds killed infishing operations has been achieved over less than fiveyears (CCAMLR 2002, Croxall and Nicol 2004).Streamer lines alone, when properly constructed anddeployed, can reduce mortality by ;70%. Unfortu-nately, at present, a minority of regulated fisheries (andprobably no illegal, unregulated, and unreported fish-eries) use these methods. In these circumstances, itseems appropriate to advocate the mandatory use ofappropriate suites of measures designed to eliminate(or minimize) incidental mortality of albatrosses inthose longline fisheries whose distributions overlapwith those of albatross.

Our simulations indicate that a decrease of at least10% in mortality rate for a 10-year period is neededto remove Black-browed Albatross from the ‘‘Endan-gered’’ category and allow them to be delisted, if allother parameters remain at their post-1988 levels. Thisrequires a near-total reduction in bycatch mortality inadults, because the fate of this long-lived species hing-es mostly on adult survival. Even with this goalachieved, the species’ rate of increase would still re-main close to a lambda of 1, a point of no growth. Thisis due to the particularly low natural lambda for thisspecies, and for albatrosses in general (a characteristicof long-lived species with low reproductive rates), thevery reason why albatross populations have been re-duced so quickly by fishery bycatch. This low naturalgrowth rate also means that recovery to historical levelswill take decades, even under strict regulations andenforcement.

Reductions in reproductive success may also play arole in the decline of South Georgia Black-browed Al-batross. These effects are both direct and indirect. First,climate-induced habitat erosion appears to be causingmore reproductive failure in recent years, particularlyfor south-facing colonies (Croxall et al. 1998). Second,adults are highly dependent on krill and krill-eatingfish for provisioning their young (Croxall et al. 1997).An increase in the number of years with low sea-icein the Scotia Sea has been linked to years of low krillavailability (Loeb et. al. 1997, Croxall et. al. 2002),itself positively correlated with poor reproductive suc-cess of the Black-browed Albatross (Croxall et al.1999). The magnitude of these effects and the demo-graphic consequences of declining krill are currentlyunknown, but our LTRE analysis shows that the con-tribution of reproductive success to the change in lamb-da between the pre- and post-1988 periods is 34%(probably attributable to changes in krill availability).Juvenile survival, the change in which can be attributedto both fisheries bycatch and changes in krill avail-ability, is responsible for 8.6% of the change in lambda

between the two periods; adult survival contributed88.0%. Although both threats (changes in krill avail-ability and increased bycatch mortality) impact thepopulating growth rate, there is no possible direct hu-man intervention to mitigate the changing frequencyof years of poor krill abundance. However, the inci-dence of bycatch mortality can be controlled becauseit is due to human intervention.

Conclusions

Our model forecasts that if the vital rates for Black-browed Albatross do not increase soon, at least theSouth Georgia population may face extinction in thiscentury. Although total numbers for this speciesthroughout its range are still higher than those for otheralbatross species, and no local population extinctionshave yet been reported, the consistent and rapid rateof decline of the Bird Island population, its short ex-tinction horizon, and the need for a large reduction inmortality for recovery reinforce the recent decision byIUCN to list it as ‘‘Endangered’’ (BirdLife Interna-tional 2004). In the space of just five years, Black-browed Albatrosses have moved from near-threatenedto endangered, rendering them possibly the most threat-ened of all albatross species in terms of rate of pop-ulation decline.

The high mortality that many, if not most, albatrossspecies suffer as a result of longline fishing is clearlyan obstacle to achieving population recovery and im-proving conservation status, and mitigation is feasible.Additional potential threats come from reduced repro-ductive success as a result of declines in krill abun-dance, and potential indirect effects of declining krillabundance on survival through a change in the foragebase and an increased tendency for the birds to seekfishing vessels for food. We urge managers, conser-vationists, and politicians to take immediate actions toimplement now the known measures that would dras-tically reduce fisheries-related mortality for this spe-cies. This needs to be accompanied by improving themanagement of bycatch and incidental mortality by allRegional Fishery Organizations, thereby taking a majorstep toward restoring the conservation status of theworld’s albatrosses.

We are confident that our model can be easily adaptedfor other albatross species and encourage researchersto use the model to evaluate extinction and recoverypossibilities for these species. Demographic data formany albatross species remain untested by modernanalysis; our model should provide an excellent basisfor evaluating population status and trends, includingestimating extinction and recovery probabilities.

ACKNOWLEDGMENTS

We thank the British Antarctic Survey for use of their long-term data set, especially Julie Leland, Andy Wood and DirkBriggs. We also thank two anonymous reviewers for com-ments on an earlier draft and Stephen Oswald for assistancewith the final draft. Finally, we thank Auburn University and


the University of Massachusetts–Boston for their financialsupport while we conducted this project.

LITERATURE CITED

Atkinson, A., V. Siegel, E. Pakhomov, and P. Rothery. 2004.Long-term decline in krill stock and increase in salps withinthe Southern Ocean. Nature 432:100–103.

Barnes, K., P. G. Ryan, and C. Boix-Hansen. 1997. The im-pact of hake Merluccius spp. longline fishery off SouthAfrica on Procellariiform seabirds. Biological Conserva-tion 82:227–234.

BirdLife International. 2000. Threatened birds of the world.BirdLife and Lynx Editions, Barcelona, Spain.

BirdLife International. 2004. Threatened birds of the world2004. CDROM. BirdLife International, Cambridge, UK.

Brothers, N. 1991. Albatross mortality and associated baitloss in the Japanese longline fishery in the Southern Ocean.Biological Conservation 55:255–268.

Burg, T. M., and J. P. Croxall. 2001. Global relationshipsamongst black-browed and grey-headed albatrosses: anal-ysis of population structure using mitochondrial DNA andmicrosatellites. Molecular Ecology 10:2647–2660.

Caswell, H. 2001. Matrix population models: construction,analysis, and interpretation. Second edition. Sinauer As-sociates, Sunderland, Massachusetts, USA.

CCAMLR [Commission for the Conservation of AntarcticMarine Living Resources]. 1997. Report of the 16th meet-ing of the Scientific Committee. CCAMLR, Hobart, Aus-tralia.

CCAMLR. 2002. Report of the 21st meeting of the ScientificCommittee. CCAMLR, Hobart, Australia.

CCAMLR. 2004. Report of the 23rd meeting of the ScientificCommittee. CCAMLR, Hobart, Australia.

Cherel, Y., H. Weimerskirch, and G. Duhamel. 1996. Inter-actions between longline vessels and seabirds in Kerguelenwaters and a method to reduce seabird mortality. BiologicalConservation 75:63–70.

Croxall, J. P. 1998. Research and conservation: a future foralbatrosses? Pages 267–288 in G. Robertson and R. Gales,editors. Albatross biology and conservation. Surrey Beattyand Sons, Chipping Norton, Australia.

Croxall, J. P., and R. Gales. 1998. An assessment of theconservation status of albatrosses. Pages 46–65 in G. Rob-ertson and R. Gales, editors. Albatross biology and con-servation. Surrey Beatty and Sons, Chipping Norton, Aus-tralia.

Croxall, J. P., T. S. McCann, P. A. Prince, and P. Rothery.1988. Reproductive performance of seabirds and seals atSouth Georgia and Signy Island, South Orkney Islands1976–1986: implications for Southern Ocean monitoringstudies. Pages 261–285 in D. Sahrhage, editor. AntarcticOcean and resources variability. Springer-Verlag, Berlin,Germany.

Croxall, J. P., and S. Nicol. 2004. Management of SouthernOcean fisheries: global forces and future sustainability.Antarctic Science 16:569–584.

Croxall, J. P., P. A. Prince, and K. Reid. 1997. Dietary seg-regation of krill-eating South Georgia seabirds. Journal ofZoology London 242:531–556.

Croxall, J., P. A. Prince, P. Rothery, and A. G. Wood. 1998.Population changes in albatross at South Georgia. Pages69–83 in G. Robertson and R. Gales, editors. Albatrossbiology and conservation. Surrey Beatty and Sons, Chip-ping Norton, Australia.

Croxall, J. P., K. Reid, and P. A. Prince. 1999. Diet, provi-sioning and productivity responses of predators to differ-ences in availability of Antarctic krill. Marine Ecology Pro-gress Series 177:115–131.

Croxall, J. P., P. N. Trathan, and E. J. Murphy. 2002. Envi-

ronmental change and Antarctic seabird populations. Sci-ence 297:1510–1514.

Gales, R. 1988. Albatross populations: status and threats.Pages 20–45 in G. Robertson and R. Gales, editors. Al-batross biology and conservation. Surrey Beatty and Sons,Chipping Norton, Australia.

Gales, R., N. Brothers, and T. Reid. 1998. Seabird mortalityin the Japanese tuna longline fishery around Australia,1988–1995. Biological Conservation 86:37–56.

Huin, N. 2001. Census of the black-browed albatross pop-ulation of the Falkland Islands 2000/2001. Falklands Con-servation, Stanley, Falkland Islands.

Huin, N. 2002. Foraging distribution of the black-browedalbatross, Thalassarche melanophris, breeding in the Falk-land Islands. Aquatic Conservation: Marine and FreshwaterEcosystems 12:89–99.

Huin, N., P. A. Prince, and D. R. Briggs. 2000. Chick pro-visioning rates and growth in black-browed albatross Di-omedea melanophris and grey-headed albatross D. chry-sostoma at Bird Island, South Georgia. Ibis 142:550–565.

Hunter, I., J. P. Croxall, and P. A. Prince. 1982. The distri-bution and abundance of burrowing seabirds (Procellari-iformes) at Bird Island, South Georgia. I. Introduction andmethods. British Antarctic Survey Bulletin 56:49–67.

Inchausti, P., and H. Weimerskirch. 2001. Dispersal and meta-population dynamics of an oceanic seabird, the wanderingalbatross, and its consequences for its response to longlinefisheries. Journal of Animal Ecology 71:765–770.

IUCN. 2002. 2002 red list of threatened species. IUCN,Gland, Switzerland. ^http://www. redlist.org/&

Loeb, V., V. Seigel, O. Holm-Hansen, R. Hewitt, W. Fraser,W. Trivelpiece, and S. Trivelpiece. 1997. Effects of sea-ice extent and krill or salp dominance on the Antarctic foodweb. Nature 387:897–900.

Nel, D. C., P. G. Ryan, R. J. M. Crawford, J. Cooper, and O.A. W. Huyser. 2002. Population trends of albatross andpetrels at sub-Antarctic Marion Island. Polar Biology 25:81–89.

Phillips, R. A., J. R. D. Silk, J. P. Croxall, V. Afanasyev, andV. J. Bennett. 2005. Summer distribution and migration ofnonbreeding albatrosses: individual consistencies and im-plications for conservation. Ecology 86:2386–2396.

Prince, P. A. 1985. Population and energetic aspects of therelationship between black-browed and grey-headed alba-trosses and the Southern Ocean marine environment. Pages473–477 in W. R. Siegfried, P. R. Condy, and R. M. Laws,editors. Antarctic nutrient cycles and food webs. Springer-Verlag, Berlin, Germany.

Prince, P., J. P. Croxall, P. N. Trathan, and A. G. Wood. 1998.The pelagic distribution of South Georgia albatrosses andtheir relationships with fisheries. Pages 137–167 in G. Rob-ertson and R. Gales, editors. Albatross biology and con-servation. Surrey Beatty and Sons, Chipping Norton, Aus-tralia.

Prince, P. A., P. Rothery, J. P. Croxall, and A. G. Wood. 1994.Population dynamics of black-browed and grey-headed al-batrosses Diomedea melanophris and D. chrysostoma atBird Island, South Georgia. Ibis 136:50–71.

Reid, K., K. E. Barlow, J. P. Croxall, and R. I. Taylor. 1999.Predicting changes in the Antarctic krill, Euphausia su-perba, population at South Georgia. Marine Biology 135:647–652.

Reid, K., and J. Croxall. 2001. Environmental response ofupper tropic-level predators reveals a system change in anAntarctic marine ecosystem. Proceedings of the Royal So-ciety of London B 268:377–384.

Reid, K., J. P. Croxall, and P. A. Prince. 1996. The fish dietof black-browed albatross Diomedea melanophris and grey-headed albatross D. chrysostoma at South Georgia. PolarBiology 16:496–477.


Rothery, P., and P. A. Prince. 1990. Survival and breedingfrequency in albatrosses. The Ring 13:61–74.

Sullivan, B. J., and T. A. Reid. 2004. Longliners, black-browed albatross mortality and bait scavenging: what is therelationship? Polar Biology 27:131–139.

Thompson, K. R. 1992. Quantitative analysis of the use ofdiscards from squid trawlers by black-browed albatrossesDiomedea melanophris in the vicinity of the Falkland Is-lands. Ibis 134:11–21.

Thompson, K. R., and M. D. Riddy. 1995. Utilization of offaland discards from ‘‘finfish’’ trawlers around the FalklandIslands by the black-browed albatross Diomedea melano-phris. Ibis 137:198–206.

Tickell, W. L. N. 1968. The biology of the great albatrossesDiomedea exulans and D. epomophora. Antarctic ResearchSeries 12:1–55.

Tuck, G. N., T. Polacheck, J. P. Croxall, and H. Weimerskirch.2001. Modelling the impact of fishery by-catches on al-batross populations. Journal of Applied Ecology 38:1182–1196.

Weimerskirch, H., N. Brothers, and P. Jouventin. 1997. Pop-ulation dynamics of wandering albatrosses Diomedea ex-ulans and Amsterdam albatross Diomedea amsterdamensisin the Indian Ocean and their relationships with longlinefisheries: conservation implications. Biological Conser-vation 79:257–270.

Weimerskirch, H., D. Capdeville, and G. Duhamel. 2000.Factors affecting the number and mortality of seabirds at-tending trawlers and longliners in the Kerguelen area. PolarBiology 23:236–249.

Weimerskirch, H., and P. Jouventin. 1998. Changes in pop-ulation sizes and demographic parameters of six albatrossspecies breeding on the French sub-Antarctic islands. Pages84–91 in G. Robertson and R. Gales, editors. Albatrossbiology and conservation. Surrey Beatty and Sons, Chip-ping Norton, Australia.

Woehler, E. J., H. J. Auman, and M. J. Riddle. 2002. Long-term population increase of black-browed albatrosses Thal-assarche melanophrys at Heard Island, 1947/1948–2000/2001. Polar Biology 25:921–927.

APPENDIX

A detailed description of Matrix A and the transition equations used in the model for Black-browed Albatrosses at SouthGeorgia (Ecological Archives A016-019-A1).

Documents

ALBATROSS POPULATIONS IN PERIL: A POPULATION …€¦ · ALBATROSS POPULATIONS IN PERIL: A POPULATION TRAJECTORY FOR BLACK-BROWED ALBATROSSES AT SOUTH GEORGIA JENNIFER M. ARNOLD,1,3