Oecologia DOI 10.1007/s00442-015-3298-2

HIGHLIGHTED STUDENT RESEARCH

Survival and local recruitment are driven by environmental carry‑over effects from the wintering area in a migratory seabird K. Lesley Szostek1 · Peter H. Becker1 

Received: 2 June 2014 / Accepted: 15 March 2015 © Springer-Verlag Berlin Heidelberg 2015

Abstract  We estimated annual apparent survival rates, as well as local recruitment rates in different age groups and for different breeding status in the common tern Sterna hirundo using mark–recapture analysis on a long-term individualbased dataset from a breeding colony in Germany. Strong inter-annual variability in survival rates became apparent, especially in prospectors. Local recruitment also varied strongly between years and age groups. To explain these fluctuations, we linked survival and recruitment estimates to several environmental covariates expected to be limiting during the wintering period and migration, including the global climate indices of North Atlantic Oscillation and Southern Oscillation, fish abundance indices, and marine primary productivity in the West African wintering area. Contrary to expectations, global indices did not seem to be linked strongly to vital rates. Results showed that primary productivity had the strongest effect on annual survival, especially in young and inexperienced individuals. Primary productivity in the wintering area was also strongly associated with the probability of recruitment in the following breeding season, indicating that conditions during winter can have carryover effects on the life cycle of individuals.

Communicated by Scott McWilliams. Electronic supplementary material  The online version of this article (doi:10.1007/s00442-015-3298-2) contains supplementary material, which is available to authorized users. * K. Lesley Szostek lesley.szostek@ifv‑vogelwarte.de 1



Institute of Avian Research, Vogelwarte Helgoland, An der Vogelwarte 21, 26386 Wilhelmshaven, Germany

Keywords  Common tern · Marine primary productivity · Migration · NAOI · SOI · Sterna hirundo

Introduction The ecology of free-living animal populations is affected strongly by environmental conditions. Geographic, annual and seasonal variation in weather, predation, intra- and inter-specific competition, food availability and accessibility can affect their life cycle through survival, breeding propensity, reproductive success and timing of events such as age of first reproduction, seasonal onset of breeding or migration, among others (e.g. Stenseth et al. 2002; Crick 2004). In a rapidly changing global environment, it is a topic of major importance for ecologists to identify the forces driving population growth through impacts on demographic rates. A special case in this context is migratory species, which spend parts of their annual cycle in several distinct habitats, experiencing widely different environmental conditions. As a consequence, these species can experience various constraints not only during the breeding season but also throughout the wintering period and on migration (e.g. in birds: Jenouvrier et al. 2005, 2009; Sandvik et al. 2005; Schaub et al. 2005; Rolland et al. 2009; Emmerson and Southwell 2011; Sheehy et al. 2011). Demographic rates are affected from within individuals through age, sex, experience or individual quality (intrinsic factors), as well as from the environment through habitat quality, food availability, climate, density and competition (extrinsic factors). These affect demographic parameters both directly (e.g. through mortality) and indirectly (e.g. through body condition and breeding propensity). Some effects might appear with a delay: so-called carry-over effects manifest themselves in demographic rates long after the ecological

13

Oecologia

cause occurred (e.g. Saino et al. 2004; Norris and Taylor 2006; Sorensen et al. 2009; Sedinger et al. 2011; Szostek et al. 2014a). So far, there is little understanding about what effect the non-breeding environment has on pre-breeders and recruits (but see Nevoux et al. 2007; van Oudenhove et al. 2014) and how extrinsic factors influence first-time breeding, even though inexperienced individuals appear to be more vulnerable to outside influences (e.g. Ezard et al. 2006; Nevoux et al. 2007; Oro et al. 2010). For migratory seabirds, conditions in the marine environment are most limiting (Schreiber 2001; Cook et al. 2014), especially through the long- and short-term fluctuations of fish stocks (Montevecchi 1993; Gröger et al. 2010; Dänhardt and Becker 2011a). Hunting of fish can be impeded by adverse weather, such as storms, wind and rain (Finney et al. 1999; Daunt et al. 2006), while food availability is influenced by local weather conditions (Misund et al. 1997, 1998). These same weather conditions can obstruct migration through increased energetic demands, leading to increased migration time and effort, and can in extreme cases even kill substantial proportions of migrating individuals (Sillett and Holmes 2002; Newton 2006, 2007). Environmental circumstances during the wintering period determine the physical condition of individuals, which directly affects survival probability, breeding propensity and success of seabirds (Marra et al. 1998; Marra and Holmes 2001; Schreiber 2001; Sorensen et al. 2009; Sedinger et al. 2011). This has repercussions for individuals and the growth of populations (Sæther and Bakke 2000; Newton 2006). However, in complex systems, it is difficult to identify the factors that are most limiting to a population, as demographic parameters, such as survival and recruitment, are determined by a multitude of influences. In a system including different seasonal habitats and birds of differing age and status, many factors have to be considered when trying to make inferences at the population level. The common tern Sterna hirundo is a small long-distance migrant and plunge-diving seabird. For prey, they rely heavily on small, often juvenile fish in the upper water layers (Becker and Ludwigs 2004; Dänhardt and Becker 2011b). Their foraging abilities can be strongly impaired by conditions out at sea, such as wind or rain (Dunn 1973; Becker et al. 1985; Becker and Finck 1985). Although studies on demography in this species are extensive (e.g. Nisbet and Cam 2002; Ezard et al. 2006; Becker et al. 2008a, b; Szostek and Becker 2012; Zhang et al. 2015a), little is known as yet about how environmental conditions in their year-round habitat affect demographic rates. West African and North Atlantic fish stock represent the main food sources for European common terns in the wintering and breeding habitat and are likely to influence demographic parameters (Becker and Ludwigs 2004). These stocks, in turn, are dependent on primary

13

productivity in the region, which is the basis of all heterotrophic food webs (e.g. Ware and Thomson 2005). The studied colony of common terns mostly spends the winter in an upwelling zone off the northwest African coast, an attractive wintering area for many seabird species, where primary productivity tends to be higher than in other areas (McGregor et al. 2007; Arístegui et al. 2009). Variation in primary productivity can therefore function as proxy for food abundance (Ballance et al. 1997; Weichler 2004; Monticelli et al. 2007; Pinaud and Weimerskirch 2007; but see Grémillet et al. 2008). Local weather, primary productivity and fish stock are influenced over large geographical areas by global climate phenomena, such as the North Atlantic Oscillation Index (NAOI) (Hurrell et al. 2001, 2003) or the Southern Oscillation Index (SOI) (Trenberth and Caron 2000; Stenseth et al. 2003). These could therefore potentially be good integrated indicators of climate conditions over large geographical scales and might have more predictive power than local weather variables (Ottersen et al. 2001; Stenseth et al. 2002, 2003; Hallett et al. 2004; Hurrell and Deser 2009). Based on the long-term integrated population monitoring of a common tern colony we wanted to address: (1) how do survival rates of prospectors, local recruits and experienced breeders as well as local recruitment rates fluctuate annually and which groups show most variation? (2) Do environmental conditions during wintering or migration affect survival rate in different breeding states or local recruitment and if so, which covariates explain most of the variation?

Materials and methods Study species Common terns breeding in the German Wadden Sea area are a nationally endangered species (Südbeck et al. 2007). They migrate along the East Atlantic coast to winter in Western Africa (Becker and Ludwigs 2004). After their first migration, juvenile common terns stay in the wintering grounds for ca. 18 months, before returning to the breeding colony for the first time at 2 years of age (Becker and Ludwigs 2004). Most common terns spend at least one season prospecting the colony before breeding (Dittmann and Becker 2003). They usually make their first breeding attempt between the ages of 2 and 5 years (Ludwigs and Becker 2002; Szostek and Becker 2012). Recruitment age strongly affects breeding performance: older recruits have higher initial success, but younger recruits improve more between breeding attempts and have a higher overall fitness (Limmer and Becker 2010; Zhang et al. 2015a), even though early and successful reproduction can result in a

Oecologia

survival cost (Zhang et al. 2015b). With increased breeding experience, parental provisioning skills improve (Limmer and Becker 2009). The oldest breeding individuals tend to have the highest breeding success (Nisbet et al. 2002), showing reproductive senescence only after a long period of increased success (Rebke et al. 2010; Zhang et al. 2015a). Skipped breeding happens most often after the first breeding attempt (ca. 20 %; Ludwigs and Becker 2007), after that it occurs at a lower, but constant rate (ca. 8 % annually; Szostek and Becker 2012). Mortality of individuals in the breeding area is very rarely recorded and the time of year at which mortality occurs most is often unknown and might be different for adult and subadult birds. However, environmental covariates might be able to tell us whether migration or the wintering period are more limiting to survival and local recruitment. Study site Since 1992, all common tern fledglings and 110 adults at Banter See colony (53°30′40′′N, 08°06′20′′E), located on the German North Sea coast, have been individually marked with transponders (TROVAN ID 100, 11 × 2 mm) as well as metal rings. In a long-term individual-based study of the entire colony, very detailed and complete lifehistories were collected through remote and automatic sensing using antennas with minimal disturbance to the colony, recording subadults, non-breeders and breeding individuals around the colony and on the nest site (Becker and Wendeln 1997; Becker et al. 2008b; Szostek and Becker 2012). For this study, data from 1992 to 2012 were used (n = 4800 marked individuals of which 1757 returned to the colony at least once). For each returned individual, information on breeding status (prospector, recruit, experienced breeder) and reproductive success was recorded every year. Thus, recruitment age was known for most marked individuals. Mark–re‑encounter analysis To estimate apparent survival and local recruitment rate, we examined a set of multi-state models (Brownie et al. 1993; Nichols and Kendall 1995; Lebreton and Pradel 2002) with three states: prospectors (P, pre-breeders), local recruits (R, local first-time breeders) and experienced breeders (B). ‘Prospectors’ were defined as birds attending the colony, but not breeding, without previous breeding experience. ‘Recruits’ were native first-time breeders; birds can only be in this state once in their lives. ‘Breeders’ were defined as individuals with at least 1 year of breeding experience. As skipped breeding occured rarely and survival of birds skipping a breeding season did not differ from other experienced breeders (Szostek and Becker 2012), birds were

considered ‘breeders’, even if they were attending the colony without breeding after having reached the state of experienced breeder. For each state apparent survival (Φ), detection probability (p) and transition between states (ψ) were estimated. We first defined a constant and a general model and then tested different scenarios for re-encounter rate, adding time or age-dependence to the constant model, while keeping the apparent survival parameter constant. Once a good fit was found, we did the same for the survival parameter and then the transition parameter until an overall best fit to the data was established. This model was then used for testing the effect of environmental covariates on apparent survival probability and transition probability. As 1-year-olds remain in the wintering habitat for their first summer after fledging (Becker and Ludwigs 2004), we constrained the survival parameter to 1 so that apparent survival for 2-yearolds equalled the apparent survival from fledging to age 2 years (=subadult survival). Goodness-of-fit tests were performed in U-CARE (Choquet et al. 2009) on the time- and state-dependent JMV model (a multi-state umbrella model; Brownie et al. 1993; Pradel et al. 2003). Since no capture in the traditional sense occurred, we excluded the trap-dependence parameter M.ITEC (Pradel 1993; Pradel et al. 2005; Szostek and Becker 2012). None of the remaining parameters was significant, suggesting a good model fit (details in Appendix Table B.1). Model ranking was based on Akaike’s information criterion (AIC) (Akaike 1973), with the lowest AIC value indicating the best fit of the model to the data (Anderson and Burnham 1999). Models were calculated in programme MARK 6.2 (White and Burnham 1999). Survival rates Survival rates had previously been found to vary between breeders and non-breeders younger than 5 years old (Szostek and Becker 2012). Consequently, we expected agedependence in prospectors and recruits, but not as much in experienced breeders (e.g. Ezard et al. 2006). Annual variation was considered likely in prospectors, but also possible in recruits and experienced breeders, as it has been shown that older and more experienced individuals are better at coping with environmental change (e.g. Ezard et al. 2006; Nevoux et al. 2007; Oro et al. 2010). For the same reason, we expected survival to be lowest in prospectors and highest in experienced breeders, with recruits somewhere inbetween. Since each breeding status naturally includes individuals of differing ages, survival was additionally estimated divided into age groups without the distinction of breeding status (for details, see Appendix Table A.1). Thus, the focus was mostly on recruits, as this status excludes individuals which were not able to overcome the “hurdle of

13

Oecologia

recruitment” (Becker and Bradley 2007), blurring the influence of poor quality individuals on variation in survival during the prospecting and recruitment process. Re‑encounter rates From previous analyses done on this colony, re-encounter rates were estimated to be very high and constant for breeding birds (nearing 100 %) and lower for non-breeding birds (2- to 3-year-olds: 83 %, 4-year-olds: 89 %, older non-breeders: 72 %; see Szostek and Becker 2012). Consequently, we expected some age- and time-dependence in prospectors (in individuals up to 5 years old) but not in recruits and breeders. Transition rates Several transitions were deemed impossible and were consequently fixed to zero in all models. These were: transition from prospector to breeder (ψP-B), transition from recruit to prospector (ψR-P), transition from breeder to recruit (ψB-R) and transition from breeder to prospector (ψB-P). Transition from recruit to breeder (ψR-B) was the only way a bird could transition once recruited as long as it survived, so this transition was considered inevitable and therefore ψR-B was fixed to 1. Only transition rate from prospector to recruit (ψP-R) was estimated, which is a measure of local recruitment rate (Crespin et al. 2006; Jenouvrier et al. 2008). Model selection was started with the fully time- and state-dependent model (general model) and a model where all parameters were set as constant over time (constant model). In order to keep the number of models low, we first selected a good model for the re-encounter parameter, adding time- and age-dependence, while keeping the other parameters constant. Then, we did the same to the transition parameter and finally the survival parameter. Time- and age-dependence of the re-encounter and survival parameter was added first to the experienced breeder stage (B), then to the recruit stage (R) and finally to the prospector stage (P), since we expected most of the variation to be in the inexperienced stages. We considered 49 models without covariates (full model set in Appendix Table A.2). Defining the wintering area In order to narrow down the covariates of food abundance and primary productivity spatially, we needed to define the range where common terns from our colony spend their winter. We combined knowledge from ring-recoveries of individuals marked at Banter See (Helgoland Ringing Centre, unpublished data) and geolocator data (Becker et al., unpublished data) to narrow down the likeliest area. We used the area between 27°N, 27°W; 27°N, 16°W; 5°N,

13

Fig. 1  Study site. Banter See common tern colony on the German North Sea coast (star) and the wintering area of this colony of common terns Sterna hirundo off the north-western coast of Africa (rectangle)

27°W and 5°N, 16°W, which roughly encompasses the ocean area between the Canary Islands in the north to the sea off Sierra Leone in the south and including the Cape Verde Islands (see Fig. 1). This region is a very important wintering area for many species of seabird (e.g. Cory’s shearwater Calonectris diomedea: Ristow et al. 2000; Péron and Grémillet 2013; several shearwater species: González-Solís et al. 2009; northern gannets Morus bassanus: Kubetzki et al. 2009; common terns: Brenninkmeijer et al. 2002). It is also possible that juvenile common terns do not spend all their time in the same wintering area as the adults, as ring-recovery data suggest that some mostly juvenile individuals also migrate to another large upwelling area off the coast of South Africa (cf. Fransson et al. 2008; Bairlein et al. 2014). Spatial segregation of different age groups has also been found in the wandering albatross Diomedea exulans (Lecomte et al. 2010) and Cory’s shearwater (Péron and Grémillet 2013), while in black-legged kittiwakes Rissa tridactyla, previous breeding success influenced winter habitat use (Bogdanova et al. 2011). It

Oecologia

seems likely that juvenile terns search for alternative foraging grounds in years when conditions are adverse in their accustomed wintering area. Environmental covariates In order to match fluctuations in demographic parameters to specific environmental covariates, the year was subdivided into biologically relevant seasons. Since different age classes follow different spring migration times (shown by arrival dates; Dittmann and Becker 2003; Ezard et al. 2007; Becker et al. 2008a; and geolocator data, Becker et al., unpublished data), these were accounted for in the timing of environmental variables. Autumn migration most likely occurred at the same time for all stages, as recently fledged individuals tend to migrate together with their parents (Becker and Ludwigs 2004). We defined our relevant time-periods thus: winter period: October–March; spring migration: April (experienced breeders), May (recruits), June (prospectors); breeding season: May–August; autumn migration: September. The timeframe for each environmental covariate was deliberately limited, in order to avoid spurious correlations (cf. Frederiksen et al. 2014). Since 1-year-old individuals remain in the wintering grounds for ca. 18 months before returning, we related the best performing environmental covariate not only from the previous season but also from the preceding year with their survival probability, to determine which time of the year juveniles are subject to the highest mortality. Therefore, we divided the 1st and 2nd calendar years into equal periods (winter: November–January; spring: February–April; summer: May–July; autumn: August–October), as they do not correspond to specific seasonal stages for juveniles as they do in older individuals that migrate every year. We chose both large-scale (global climate indices) and smaller-scale (fishery data and regional productivity) environmental covariates (as listed below), because we expected them to have differing effects on vital rates (e.g. Milner et al. 1999; Catchpole and Freeman 2000; Coulson et al. 2001). Details for all environmental covariates and data sources are listed in Table 1. The North Atlantic Oscillation Index NAOI (Climate Data Guide at the National Centre for Atmospheric Research; http://climatedataguide.ucar.edu/guidance/ hurrell-north-atlantic-oscillation-nao-index-station-based) describes weather patterns in the North Atlantic and Western and Central Europe (Hurrell 1995; Hurrell et al. 2001, 2003). In addition, higher abundances of marine fish, such as herring Clupea harengus have been observed after years of low NAOI (Attrill and Power 2002). Common terns are most likely to be affected by it during migration (Favero and Becker 2006). In several bird species, associations of the NAOI with vital rates have already been found (e.g.

Forchhammer et al. 1998; Grosbois and Thompson 2005; Crespin et al. 2006; Descamps et al. 2010; Zipkin et al. 2010). Although the Southern Oscillation Index SOI (International Research Institute for Climate and Society (IRI) Data Library; http://iridl.ldeo.columbia.edu/docfind/databrief/ cat-index.html) describes air pressure differences in the Pacific Ocean, it has effects on the Northern Hemisphere through teleconnections (Trenberth and Caron 2000; Stenseth et al. 2002), such as stronger upwelling in Western Africa during La Niña events (Roy and Reason 2001), which would influence food abundance in the wintering area of common terns. Southern Oscillation is therefore most likely to affect common terns in their wintering habitat, as it has been shown to affect demographic rates in other bird species (e.g. Brichetti et al. 2000; Sillett et al. 2000; Nevoux et al. 2007; Jenouvrier et al. 2009). Food abundance is likely to affect survival in all seasons, ages and stages. We used herring and sprat Sprattus sprattus stock data (ICES 2013), the most common food species (along with smelt Osperus eperlanus) for common terns in the Wadden Sea (Becker and Ludwigs 2004; Dänhardt and Becker 2011a). For the wintering area, we took a total abundance index of small pelagic fish for the region (FAO 2011), including the most common prey of common terns: sardines Sardina pilchardus; Sardinella sp. and anchovies Engraulis encrasicolus (Dunn and Mead 1982; Becker and Ludwigs 2004), and also horse mackerel Trachurus trachurus, T. trecae, Caranx rhonchus, and chub mackerel Scomber japonicus. The measures of abundance for west African fish stock were total catch data, not corrected for fishing effort (FAO 2011, pp. 118–120). Another proxy for food abundance is primary productivity (College of Science, Oregon State University, USA; http://www.science.oregonstate.edu/ocean.productivity/ standard.product.php), because it fuels the food web of all heterotrophic organisms. We used data on net marine primary production from chlorophyll in the wintering area of our study species. Covariate models Environmental covariates were included after choosing a well-supported but parsimonious model. We assumed no difference in survival between the sexes (Nisbet and Cam 2002; Ezard et al. 2006; Szostek et al. 2014b; Zhang et al. 2015a). The covariates were tested for inter-correlation (Appendix Table B.2). Since the variable ‘autumn NAOI’ correlated with ‘winter SOI’, it was excluded from the analysis as redundant. Models were fitted with each covariate alone, linearly and as interaction, to see if it would improve the model fit. Covariates were added both to the apparent survival parameter and to the transition parameter

13

13

Definition

Abundance of sprat (all size classes) in the North Sea (annual average)

Winter food: abundance of small pelagic fish off the northwest African coast (annual average)

Sprat

wFood

Weight in tonnes

Abundance index: landings per unit effort (lpue)

Abundance of herring (recruitment, size Estimated number of individuals class 0) in the North Sea (annual average)

mg C/m2/day

Unit

Herring

Winter SOI: Southern Oscillation Index during the winter period (October– March): standardised mean monthly values sNAOI (P/R/B) Spring NAOI: North Atlantic Oscillation during the spring migration for Prospectors (June), Recruits (May) and Experienced Breeders (April): monthly principal component values Prim Primary productivity in the wintering area (monthly values, averaged from October to March)

wSOI

Variable

1,138,524–1,450,632

Food and Agriculture Organization of the United Nations (FAO), (2011), Fisheries and Aquaculture Report No. 975, page 120

College of Science, Oregon State University, USA (http://www.science.oregonstate.edu/ocean.productivity/standard. product.php) 21,423,521 –86,616,423 International Council for the Exploration of the Sea (ICES), (2013), Report of the Herring Assessment Working Group for the Area South of 62 N (HAWG), page 111, Table 2.6.3.12 736–4921 International Council for the Exploration of the Sea (ICES), (2013), Report of the Herring Assessment Working Group for the Area South of 62 N (HAWG)

1088–1293

−1.52 to 2.33

1992–2009

1992–2012

1992–2012

2002–2012

1992–2012 International Research Institute for Climate and Society (IRI) Data Library (http://iridl.ldeo.columbia.edu/docfind/ databrief/cat-index.html) 1992–2012 Climate Data Guide at the National Centre for Atmospheric Research (http:// climatedataguide.ucar.edu/guidance/ hurrell-north-atlantic-oscillation-naoindex-station-based)

−1.80 to 2.23

Available data

Source

Range (min–max)

Table 1  Covariates used in mark–re-encounter analysis: abbreviation, definition, source and available timeframe

Oecologia

Oecologia

Multi‑state model without covariates The highest ranking multi-state model without covariates (full model set in Appendix Table A.2) was: ФP(t*sub,.ad) ФR(t) ФB(t), pP(t*sub-5y,.ad) pR(6 age) pB(.),ψPR(t*sub-5y,. ad). This model suggested that prospectors’ apparent survival was time-dependent for subadults, but constant for older individuals without further age distinction. Recruits’ and experienced breeders’ survival was time-dependent, irrespective of age (Fig. 2). Re-encounter rate was timeand age-dependent in prospectors up to the age of 5 years and constant thereafter; for recruits, it was age- but not time-dependent up to the age of 6 years and constant thereafter; in experienced breeders, re-encounter rate was constant. Transition rate from prospector to recruit (recruitment rate) was time-dependent and varied between the ages of 2 and 5 years and was constant among birds older than 5 years (when the period of recruitment is generally considered complete). This model was the basis for all multi-state covariate models. Multi‑state model with covariates The highest ranking multi-state model included a positive linear relationship between apparent survival of prospectors (only subadults) and primary productivity in the wintering area (Fig. 3), as well as an interaction of transition rate (ψPR) with primary productivity (Table 2; Appendix Table A.3). This model explained 78.2 % of variation (ANODEV: F = 54.137; df = 1, 177; P 

Survival and local recruitment are driven by environmental carry-over effects from the wintering area in a migratory seabird.

We estimated annual apparent survival rates, as well as local recruitment rates in different age groups and for different breeding status in the commo...
701KB Sizes 0 Downloads 9 Views