Oecologia (2006) 147: 606–614 DOI 10.1007/s00442-005-0321-z
POPULATION ECOLOGY
Sue Lewis Æ David Gre´millet Æ Francis Daunt Peter G. Ryan Æ Robert J.M. Crawford Æ Sarah Wanless
Using behavioural and state variables to identify proximate causes of population change in a seabird
Received: 24 August 2005 / Accepted: 9 November 2005 / Published online: 2 December 2005 Springer-Verlag 2005
Abstract Changes in animal population size are driven by the interactions between intrinsic processes and extrinsic forces, and identifying the proximate mechanisms behind population change remains a fundamental question in ecology. Here we report on how measuring behavioural and state proxies of food availability among populations experiencing different growth rates can be used to rapidly identify proximate drivers of population trends. In recent decades, the Cape gannet Morus capensis has shown a major distributional shift with historically large colonies in Namibia decreasing rapidly, whilst numbers at South African colonies have increased, suggesting contrasting environmental conditions in the two regions. We compared per capita growth rates of five of the six
extant colonies with foraging range (using miniaturised Global Positioning System loggers), foraging work rate, food delivery rates and body condition of breeding adults. We found significant associations between the rate of population change, individual behaviour, energetic gain and body condition that indicate that recent population changes are associated with extrinsic effects. This study shows that behavioural and state data can be used to identify important drivers of population change, and their cost-effectiveness ensures that they are an appealing option for measuring the health of animal populations in numerous situations. Keywords Population regulation Æ Foraging behaviour Æ Distribution shift Æ Fishery interactions Æ Cape gannet
Communicated by Roland Brandl S. Lewis (&) Department of Zoology, University of Aberdeen, Tillydrone Avenue, AB9 2TN Aberdeen, UK E-mail:
[email protected] Tel.: +44-1330-826300 Fax: +44-1330-823303 F. Daunt Æ S. Wanless Æ S. Lewis Centre for Ecology and Hydrology Banchory, NERC, Hill of Brathens, AB31 4BW Banchory, Aberdeenshire, UK D. Gre´millet Centre d’Ecologie et Physiologie Energe´tiques, Centre National de la Recherche Scientifique, 23 rue Becquerel, 67087 Strasbourg Cedex 02, France P. G. Ryan DST/NRF Centre of Excellence at the Percy FitzPatrick Institute of African Ornithology, University of Cape Town, 7701 Rondebosch, South Africa R. J.M. Crawford Department of Environmental Affairs and Tourism, Marine and Coastal Management, Private Bag X2, 8012 Rogge Bay, South Africa
Introduction Population regulation lies at the core of the study of ecology and the mechanisms driving population change continue to intrigue biologists (Godfray and Rees 2002; Sibly and Hone 2002; Sibly et al. 2002, 2005). Fluctuations in animal abundance are caused by complex interactions between intrinsic factors, such as densitydependence, and extrinsic factors determined by the action of environmental processes (Newton 1998; Turchin 1999; Coulson et al. 2001; Godfray and Rees 2002; Berryman 2004). Two main approaches have been taken to understanding the determinants of population size (Sibly and Hone 2002). The ‘‘demographic paradigm’’ aims to establish the ultimate causes of population growth rate, by estimating the relative contributions of survival, fecundity, immigration and emigration. Using a variety of population modelling techniques, a great deal of progress has been made in understanding the contribution of these demographic rates to population
607
growth rates (Williams et al. 2001; Saether et al. 2002). In contrast, the ‘‘mechanistic paradigm’’ identifies the proximate causes in the environment (e.g. food supply, predation, parasites, competitors) that drive population growth rate (Krebs 2002). The relationship between population density and population growth rate is central to both paradigms. Under the assumption of constant environmental conditions, the relationship is generally found to be negative except when populations are very small (Sibly et al. 2005). However, the shape of the relationship is hard to estimate in the wild since the environment is rarely constant. Thus, to understand the drivers of population growth rate, the mechanistic paradigm, in which the effect of resource availability replaces the effect of population density, is preferable (Sibly and Hone 2002). Numerous studies have shown that growth rates of many animal populations are affected by food availability (reviews in Sinclair 1989). Indirect measures of food availability, such as weather conditions, have also been linked to population growth rate in the studies of Soay sheep Ovis aries and elk Cervus elaphus (Milner et al. 1999; Cook et al. 2004). Alternative indirect measures that have great appeal are the behaviour and state (e.g. body condition) of individuals within the population. This approach has the advantage over most demographic approaches of providing rapid assessments of causes of population change. Furthermore, recent advances in miniaturised electronic instruments provide opportunities to accurately quantify behavioural correlates of food availability in an ever-increasing range of species. Here, using a cross-sectional approach, we examine a case study in which the relationship between population growth rate and density suggests strongly that the environment is not constant across populations. The Cape gannet Morus capensis is a pelagic seabird that breeds at six colonies ranging in size from 400 to 70,000 breeding pairs in southern Africa. Five out of the six colonies are dependent on the Benguela upwelling system for food, and currently exhibit a positive relationship between population growth rate and colony size (Fig. 1). In this study, we take a mechanistic approach to understanding the variation in population growth rate across these five populations. We test the hypothesis that food availability around colonies during the breeding season provides the best explanation for differences in population growth rate. At each colony, we measured the following behavioural and state currencies of food availability: 1) foraging trip duration (and range); 2) foraging work rate; 3) density of conspecifics in foraging range; 4) quality of food delivered to the colony; and 5) adult body condition. Using these variables we can predict the counter-intuitive result of increasing growth rates with increasing colony size, illustrating the value of a mechanistic approach to understanding population growth rates.
Fig. 1 Relationship between the size of the breeding population of Cape gannets at the five Benguela colonies in 1990, and the recent rate of change in breeding pairs of Cape gannets from 1990 to 2000/ 2002
Methods Population growth rates The Cape gannet is a large, piscivorous seabird that feeds on a wide range of species by plunge-diving (Gre´millet et al. 2004; Ropert-Coudert et al. 2004). The entire population occurs in six colonies: five colonies are adjacent to the productive Benguela upwelling ecosystem off the west coast of southern Africa and one colony is off South Africa’s Eastern Cape Province. To assess long-term (50 year) changes in population size, we collated count data since the 1950s for the five Benguela colonies: Malgas (3303¢S, 1755¢E) and Lambert’s Bay (325¢S, 1818¢E) in South Africa’s Western Cape Province and Ichaboe (2617¢S, 1456¢E), Possession (2701¢S, 1512¢E) and Mercury (2543¢S, 1450¢E) in Namibia (Crawford 1999, Marine and Coastal Management unpublished data, P. Bartlett, unpublished data). Comparisons between current behavioural and state variables with population change were made using recent per capita population growth rates calculated from available counts between 1990 and 2002 (n=2–3 counts per colony). Behavioural and state variables We measured foraging parameters of breeding birds from Malgas, Lambert’s Bay, Ichaboe and Mercury using animal-borne instrumentation deployed during the 2003/2004 austral breeding season (13 November to 6 January). On average, birds initiate breeding earlier in South Africa and so we were able to sample colonies sequentially whilst keeping chick age constant (range
608
2–5 weeks) across colonies, thus minimising any differences in inter-colony behaviour associated with differences in chick age (Lewis et al. 2004). Adults were captured at the nest prior to leaving on a foraging trip and equipped with a miniaturised global positioning system (GPS) from Newbehavior, http://www.newbehavior.com and earth&Ocean Technologies, Kiel (see Gre´millet et al. 2004). Latitude, longitude, speed and time were recorded at 10 s intervals. Loggers were housed in waterproof and pressure tight, streamlined containers (95·48·24 mm; total mass of logger and housing: 65 g (ca. 2.5% adult body mass) and were attached to the central tail feathers (see Gre´millet et al. 2004). Nest sites of instrumented birds were monitored and when an adult returned it was recaptured and the logger removed. Wing cord (mm) and body mass (g) were measured on recapture using a stoppered wing rule and spring balance (±25 g). Adult gannets returning to the colony following a successful foraging trip often regurgitate food during handling. Regurgitated food samples were collected and weighed before storing for subsequent analysis. Previous trials on Malgas Island showed no significant difference in trip duration between birds equipped with similar sized loggers to those used here and controls (Gre´millet et al. 2004). We collected foraging data using GPS loggers from 94 birds (ca. 20 deployments at each colony), with only one trip recorded per bird to reduce problems associated with pseudo-replication. At Possession, birds were not equipped with loggers due to the need to minimise disturbance at this small colony. In this case we made observations of the number of changeovers at 25 nests with chicks (same age range as other colonies) in order to estimate trip duration (see Hamer et al. 1993). At all colonies, we recorded offspring neglect on 2–3 occasions by counting the number of chicks left unattended (n=50 nests on each occasion). The time a pair spent together during a changeover of breeding duties was recorded for a sample of 20–30 nests at each colony. From these data, we defined the following five behavioural, energetic and state currencies of food availability: 1) Foraging trip duration: The GPS tracks and changeover rates were analysed to determine foraging trip duration. For the four colonies where we deployed GPS loggers, we could also examine the relationship between trip duration (h) and maximum range (km) among colonies. 2) Foraging work rate: The following proxies of foraging work rate were determined: frequency of offspring neglect (an indicator of food stress, Lewis et al. 2004); time the pair were together during a changeover (an indicator of how much ‘spare time’ adults have); time spent on travelling, foraging and resting on the sea surface during a foraging trip. Gannet flight alternates between travelling between potential feeding sites and circling above these sites and plunge-diving for fish. We filtered GPS positions
associated with circling using a sinuosity index (full details in Gre´millet et al. 2004) and from this were able to distinguish foraging flight, travelling flight and resting on the sea surface. 3) Density of conspecifics in foraging range: Colony home ranges were calculated by selecting the foraging (i.e. circling) fix at the maximum distance from the colony for each bird (to avoid autocorrelation), and calculating the minimum concave polygon (60%) in RangesVI (Kenward et al. 2001), based on the assumption that the sample of birds carrying loggers was representative of each population. The degree of overlap between home ranges of adjacent colonies was estimated as the proportion of foraging fixes that occurred in the neighbouring colony’s home range. Mean density of breeding birds foraging in each home range was estimated using current colony sizes, adjusted by colony-specific rates of attendance (see results). 4) Energy delivered to colony: To determine the daily amount of energy brought back to the nest, we identified all prey items and calculated the percentage biomass of each species in each food sample. Using estimates of prey energy density given by Batchelor and Ross (1984) for sardine Sardinops sagax, anchovy Engraulis capensis, saury Scomberesox saurus, horse mackerel Trachurus capensis, snoek Thyrsites atun, Cape hake Merluccius capensis and squid Loligo reynaudi, we calculated the energy density (kJ/g) of each food sample. By multiplying this value by the mass of the food sample, we calculated the energy content of each food sample in kJ. The number of feeds per day was calculated from trip duration and the rate of unattendance. We multiplied feeds per day by the energy content of the food load to derive the energy brought back to the nest per day (kJ/day). This value does not correspond to the total mass of food caught at sea, but reflects the amount the bird brought back to the nest, either for chick provisioning, or for self-feeding. It can therefore be used as an index of foraging success. 5) Body condition: We divided adult body mass by wing cord to obtain an index of body condition.
Statistical analyses We obtained data from 15–30 adults at each colony. However, at Possession where detailed measurements on individuals could not be made, we estimated mean values from observations at 25–50 nests, depending on the variable. In order to control for unbalanced data sets among colonies and for potential pseudo-replication within each colony, analyses were performed by fitting weighted linear mixed models (using the method of residual maximum likelihood, Patterson and Thompson 1971) to currencies of food availability with the rate of recent population change as a covariate. Models were
609
weighted by the estimated colony-specific variances for each measurement, where the variance was set to equal 1 at Possession when a single mean value was used. The exception was the percentage of chicks unattended at the colony, which was compared using a Pearson’s correlation. All analyses were performed using the statistical package GenStat version 6.2 (VSN International Ltd 2003). Analyses are presented by referencing the Wald statistics to chi-squared distributions with appropriate degrees of freedom; F tests adjusting for uncertainty in estimated variance components gave qualitatively the same results. Since the five colonies are distributed in two regional clusters with related recent population trends (the three declining colonies are on the Namibian coast, and the two increasing colonies on the South African coast), we also tested whether any effect of population growth rate on behaviour, energy gain and condition was better explained by regional location. Therefore, we replaced population growth rate with region (Namibia vs South Africa) in the models, since it was not possible to fit both simultaneously because of the high correlation between them. For percentage of chicks left unattended, we carried out a t test to compare between the two regions.
Results Colony-specific rates of population change The among colony comparison of per capita growth rate and colony size did not provide evidence of densitydependent growth in Cape gannet colonies. Rather, it indicated a significant and positive relationship (Fig. 1, r=0.89, P
