Short-term effects of culling on the ecology and population dynamics of the yellow-legged gull

Authors

  • M. Bosch,

    1. Departament d'Ecologia, Universitat de Barcelona, Avda. Diagonal 645, E-08028 Barcelona, Spain;
    Search for more papers by this author
  • D. Oro,

    1. Ornithology Unit, Department of Zoology, IBLS, Glasgow University, Glasgow G12 8QQ, UK; and
    Search for more papers by this author
    • §

      Present address: Instituto Mediterráneo de Estudios Avanzados IMEDEA (CSIC-UIB), Miquel Marques 21, 07190 Esporles, Mallorca, Spain.

  • F.J. Cantos,

    1. Programa de desarrollo de la Estrategia de Biodiversidad, Dir. Gral. de Conservación de la Naturaleza, Subd. Gral. de Conservación de la Biodiversidad, Gran Vía de San Francisco 4, 28005 Madrid, Spain
    Search for more papers by this author
  • M. Zabala

    1. Departament d'Ecologia, Universitat de Barcelona, Avda. Diagonal 645, E-08028 Barcelona, Spain;
    Search for more papers by this author

M. Bosch, Departament d'Ecologia, Universitat de Barcelona, Avda. Diagonal 645, E-08028 Barcelona, Spain (fax 34 93 4111438; e-mail mbosch@porthos.bio.ub.es).

Summary

1.  Culls are being performed in many yellow-legged gull Larus cachinnans colonies in the Mediterranean region, but little is known of their effects on this species. Between 1992 and 1996, 25 000 breeding yellow-legged gulls were culled in the colony of the Medes Islands, north-western Mediterranean, because of the possible role of gulls in the transmission of pathogens and as predators on other bird species. In the present study the effect of the culls on several ecological parameters and on the population dynamics of the colony were analysed.

2.  Some breeding parameters (nest density, clutch and egg sizes, chick growth and breeding success, intraspecific predation, size and body condition of adults), diet (importance of the different prey categories and width of the trophic niche) and population dynamics of the colony were analysed during the years of cull. The annual culls did not include the entire colony, so culled and unculled areas were distinguished.

3.  Nests density decreased significantly both in the culled and the unculled areas. No significant difference in clutch size was detected between culled and unculled areas in any year, but clutch size decreased significantly through time within the culled areas. In three-egg clutches the mean egg volume was significantly larger in unculled than in culled plots, whereas no significant year effect was found. No differences were detected in the mass or in the body condition of chicks throughout the study, whereas both fledging success and breeding success increased significantly over the period of the study. Intraspecific predation had decreased significantly by 40% two years after the beginning of the study. Size and body condition of adults varied between years but no trend was observed.

4.  Despite the large decrease in breeding gull numbers and an expected reduction in intraspecific food competition, no changes in diet were detected during the study. The dietary niche width was very similar between years, and gulls continued to exploit the same foraging resources.

5.  Annual censuses showed that from 1960 to 1992, colony size increased at a rate of 5% per year. During the culling period, between 21% and 29% of the breeding adults were killed, and colony size decreased progressively at an estimated rate of 19% per year. The estimated annual adult survival according to ring recoveries during the culling period was 74%. A demographic model was developed to assess the observed changes in numbers of yellow-legged gulls. This suggested that gulls born in the Medes Islands were emigrating during the years of culling, with a particularly high estimated emigration rate of 25% per year since 1994.

6.  The planning of culling programmes by wildlife agencies has not always taken into account the multiple factors responsible for the population dynamics of colonies and the effects of culls. Culling at the Medes Islands probably failed to reduce the breeding numbers of yellow-legged gulls at the metapopulation level, due to the emigration of birds to neighbouring colonies that had recently increased in numbers. Thus, the potential problems linked to large numbers of gulls in a colony may simply have been transferred to other sites.

Introduction

During recent decades, gull populations have increased dramatically in Europe, north America and Australia, probably as a result of an increase in food availability derived from human activities (Blokpoel & Spaans 1991; Furness, Ensor & Hudson 1992; Pons 1992; Smith & Carlile 1993). In some cases, large numbers of gulls can have adverse effects on other bird populations through predation or nesting competition (Bradley 1986; Hario 1994; Guillemette & Brousseau, in press), and on humans through contamination of domestic water supplies, bird strikes at airports and nuisance in towns or disruption of commercial operations (Benton et al. 1983; Monaghan 1983; Burger 1985; Furness & Monaghan 1987; Blokpoel & Tessier 1992). To minimize such effects, wildlife agencies have implemented several management measures.

Measures that reduce food availability have proved to be effective in reducing gull numbers in the long term (Pons 1992; Pons & Migot 1995). However, these measures have been little used because they are often expensive (Spaans & Blokpoel 1991). In breeding colonies, periodic scaring, alteration of nesting habitat, and sterilizing or removing the eggs have been used to reduce numbers (Thomas 1972; Christens & Blokpoel 1991; Blokpoel & Tessier 1992; Wanless et al. 1996). In addition, a rapid reduction in numbers of breeding gulls has frequently been achieved by killing adults (Thomas 1972; Duncan 1978; Coulson, Duncan & Thomas 1982; Wanless & Langslow 1983; Smith & Carlile 1993).

However, several factors associated with the population dynamics of gull colonies are density-dependent, such as fecundity and recruitment (i.e. the entry of new individuals into the breeding population). On the Isle of May (Scotland), increases in egg size, body size and body condition of adults, and a reduction in the age at recruitment, were observed following culls of breeding adults (Duncan 1978; Coulson, Duncan & Thomas 1982). In addition, culling a single gullery has unpredictable effects at the metapopulation level because it may influence immigration and emigration rates between colonies (Coulson 1991; Defous du Rau 1995). Culling may thus be less effective than planned and its effects may be neutralized once it is discontinued (Spaans et al. 1991).

Most studies dealing with the effects of gull control have analysed variations in both the number of pairs and nest density (Thomas 1972; Wanless & Langslow 1983; Skira & Wapstra 1990; Alvarez 1992), but few have examined factors associated with the population dynamics of the colony (Spaans, de Wit & Van Vlaardingen 1987; Coulson 1991; Wanless et al. 1996). This paper reports the effects of a 5-year culling programme on several ecological parameters in a large colony of yellow-legged gulls Larus cachinnans Pallas in the north-western Mediterranean. Culls are being performed in many parts of their range (e.g. Spain, France, Portugal; Blondel 1963; Aguilar 1991; Alvarez 1992; Morais, Santos & Vicente 1998; Vidal, Medail & Tatoni 1998) but little is known about the effects and the effectiveness of this control measure in the Mediterranean region. In our study, the effect of reducing numbers of breeding adults on nest density, breeding performance, interspecific predation, body condition of adults and diet was analysed throughout the years of the culls. Survival and recruitment were also estimated using capture–recapture techniques, to model the effects of culling on the population dynamics of this colony (Lebreton & Clobert 1991). Finally, the suitability of culling as a measure to reduce the presence of gulls is discussed.

Methods

Study area

The study was conducted in the Medes Islands (42°0′N, 3°13′E, north-east Spain), an archipelago comprising two small islands (Meda Gran, 15 ha, and Meda Petita, 3 ha) and several calcareous rocks just 0·9 km off the coast of L'Estartit, Spain. The archipelago holds one of the largest breeding colonies of yellow-legged gulls in the Mediterranean basin, with more than 14 000 pairs in 1991, before the start of the cull (Bosch, Oro & Ruiz 1994a). Three main habitats, differing in vegetation cover, are distinguished in the colony: shrubs, dominated by Atriplex halimus Linneo; grass, dominated by grassy, ruderal plants (e.g. Hordeum murinum L.); and bare soil, with dispersed rocks and sparse vegetation (Bosch & Sol 1998; Torre & Bosch 1999). Since 1992, the colony has been culled annually by the regional Nature Conservancy Agency (Departament d'Agricultura, Ramaderia i Pesca, DARP, Generalitat de Catalunya, Spain) because of the possible role of the gulls in transmitting human pathogens and as predators on other bird species (although see Oro & Martínez-Vilalta 1994; Bosch & Muniesa 1996; Bosch 1996a). The annual culls do not cover the entire colony, so culled and unculled areas can be distinguished. The method of culling involves placing baits of bread with butter mixed with α-chloralose and secobarbital in nests containing eggs. The staff of the Nature Conservancy Agency visit the culled area on three consecutive occasions: first, when placing the first bait in each nest; secondly, when removing the dead gulls (most of them in their own nests) and placing new baits; and finally, when removing both newly killed gulls and any baits that have not been eaten. Several hours are allowed to elapse between consecutive visits, allowing narcosis, and during this time the area is not disturbed, to prevent gull dispersal away from the nesting areas (for more details see Sargatal, Saavedra & Romero 1992).

Breeding parameters

The study was carried out between 1992 (the first year of culling) and 1996. During this time, data for most of the parameters were collected for at least 3 consecutive years. Only data on chick size and growth and chick predation were analysed during 2 non-consecutive years.

Density of nests

The density of nests was estimated each year from 1994 to 1996, and analyses also included data from 1993 published in Bosch et al. (1994b), following the same methods. Densities were determined by counting the number of nests within strips selected randomly each year (388 in total) in the different sections into which the two islands of the archipelago had been divided. Strips measured 30 × 5 m, with the exception of those sampled in cliffs, which were larger in all cases (from 600 to 3000 m2). The sections of the archipelago had been differentiated previously according to physical features such as the slope: flat, moderate slope; Gregal Valley; high slope; shore rocks; and cliffs (for more details see Fortià & Hontangas 1991) (Fig. 1). As nest density variation between sections was similar on each island (Bosch et al. 1994b), data from the two islands were pooled. The ratios of strips in the different vegetation substrates of each section were maintained yearly, avoiding a possible bias of results due to a habitat effect (Bosch & Sol 1998). At the same time, the ratios between culled and unculled areas were maintained each year in the strip sampling procedures. Possible differences in nest densities between years were tested in each section using a Kruskal–Wallis (K–W) test.

Figure 1.

Map of the Medes Islands showing habitat type, culled and unculled areas.

Data from 1993 (after the first cull) showed no significant differences between sections when the cliffs were excluded from the analysis (Bosch et al. 1994b). Thus we pooled data from all the remaining sections and then differentiated between strips sampled in the culled area and those sampled in the unculled area. Afterwards, differences between the two areas were tested by a normal approximation to the Mann–Whitney U-test (Z distribution).

Breeding performance

To analyse breeding performance, we only considered data from an area of grassy habitat, avoiding a possible bias of results by a habitat effect (Bosch & Sol 1998). In this area we differentiated two plots annually: one submitted to culls (called ‘culled plot’ hereafter) and one excluded from the culls (called ‘unculled plot’ hereafter). Each plot had an approximate area of 0·2 ha (Fig. 1). Breeding parameters included clutch size, egg size, chick size and growth, and the breeding success of the pairs nesting in the study plots.

Clutch and egg size were recorded by checking each plot every 2–3 days throughout the laying period (approximately 35 days) from 1992 (just before the first cull) to 1996 (467 clutches and 1146 eggs in total). The length and width of the eggs were measured with callipers to the nearest 0·1 mm. Egg volume (ml) was calculated using the equation of Harris (1964) with Kv = 0·476, and the mean egg volume was then calculated for every clutch. First, the effects of year and plot (culled and unculled) on the clutch size frequency distribution were tested simultaneously using a three-dimensional contingency table (Zar 1996). As overall independence among variables was rejected (χ213 = 23·82, P < 0·05), the plot effect was analysed independently for each year, while the year effect was analysed separately for each plot. Both effects were tested by the likelihood ratio test (G-test) and Fisher exact test when expected frequencies were lower than five. Differences in the mean egg volume were initially tested by a three-way anova, considering the effects of the plots, years and clutch size simultaneously. Prior to this, egg volume data were checked for fit to a normal distribution (Kolmogorov–Smirnov test, Dn = 0·037, P = 0·622). As a very significant interaction was detected between plot and clutch size (F1,400 = 7·26, P = 0·007), possible variations in the mean egg volume were re-analysed separately for two-egg clutches and three-egg clutches by two-way anovas.

Chick size and growth were measured in 1992 and 1994 (57 and 68 chicks, respectively) in the unculled plot. In the culled plot, data were not recorded because nesting gulls were killed and their eggs pricked before hatching. Bill length, wing length, tarsus length and weight of the chicks were measured every 3 days from hatching to fledging (chicks older than 30 days; Oro, Bosch & Ruiz 1995). To check for a difference between years in the slope relating chick body mass and age, an ancova analysis was carried out only with fledging chicks. To ensure independence of observations, only one measure of mass per chick was used, selected randomly from the measurements available in the linear segment of the growth curve (chicks aged between 15 and 30 days). Possible differences in chick growth rates between years were tested using the Mann–Whitney U-test. Chick body condition (an index that relates body mass and body size) at hatching was calculated using regression to control for differences in body size. Hatching mass was regressed on tarsus length, and the residuals were used as a measure of chick body condition (Oro, Jover & Ruiz 1996). Differences in chick mass and body condition were tested by t-test.

Breeding output was studied from 1993 to 1996 in the unculled area by surrounding a group of nests (between 42 and 57, depending on the year) with a 50-cm high net just before hatching and then recording the number of chicks that hatched and those that fledged (Oro, Bosch & Ruiz 1995). Although enclosures might have had an effect on breeding success (e.g. allowing or preventing intraspecific predation), this is the most suitable method to assess productivity for ground-nesting species (Bolton 1991; Griffiths 1992; Oro, Jover & Ruiz 1996). We assumed that these potential effects were similar each year and inter-year comparisons were not biased. Fledging success was the number of chicks fledged, as a percentage of eggs hatched, while the breeding success was number of chicks fledged, as a percentage of eggs laid. In the culled area, breeding output was not recorded as most of the nesting gulls were killed. The total number of fledglings reared in the colony each year was estimated using the breeding output (as the mean number of fledglings per pair) and the number of nests in the colony, corrected by the number of adults culled (see below). Although no information was available on the effects of culling at the individual nest level, culling procedures suggested that in most cases baits killed both members of the pair (estimated at 80%). We then subtracted the corrected number of culled nests from the total number of nests counted before each culling to estimate the total number of fledglings reared.

Intraspecific predation

In colonies of large gulls, some authors have shown that chick predation by conspecifics may decrease when nest density is reduced (Spaans, de Wit & Van Vlaardingen 1987; Kilpi 1989). If culling reduced nest density, we would expect a decrease in chick predation rates. The number of chicks that disappeared from the enclosures built to study the breeding performance in 1992 and 1994 was thus compared, and statistical differences were tested using the log-likelihood ratio G-test. Because the corpses of missing chicks were not found subsequently, we assumed that they had been preyed upon by adult conspecifics, rather than having died by disease or starvation (Hario 1994; Hario & Rudbäck 1996).

Adult size and body condition

Body measurements of 331 culled adult gulls were recorded from 1992 to 1996. We only considered data from gulls nesting in an area of grassy habitat to avoid a possible bias in results due to a habitat effect (Bosch et al. 1997). Gulls were taken just after having been killed in their own nests, and their body mass, head length (distance from the tip of the bill to the posterior ridge formed by the parietal–supraoccipital junction) and tarsus length (tarsometatarsus length) were measured. Body mass was measured with a hand-held 1500-g Pesola balance (± 10 g), while head and tarsus length were measured with callipers to the nearest 0·1 mm. Gulls were sexed by dissection. Body condition was calculated as the residual from a linear regression of body mass on tarsus length (Oro, Jover & Ruiz 1996). Because of the great sexual dimorphism of the species (Bosch 1996b), possible variations in body size and body condition were analysed separately for males and females.

Diet composition

Diet was studied by examining the regurgitates of fledging chicks, which provide a good indication of the food items collected by the adults (Mudge & Ferns 1982; González-Solís et al. 1997). Diet composition was studied from 1994 to 1996, but analyses also included data from 1992 published in Bosch, Oro & Ruiz (1994a) following the same methods. Regurgitates were collected and preserved for identification in the laboratory using reference collections of fish, invertebrates and mammals from the same areas. The quantification procedures always followed the rule of minimum numbers (i.e. by pairing bilateral elements such as otoliths or elytra, matching for different sizes, and scoring a simple item per food sample when remains of fur, feathers, garbage or plants were found, unless different types could be distinguished; González-Solís et al. 1997; Brown & Ewins 1996). The dry weight of each undigested item was calculated using an oven, keeping the samples at 60 °C until they reached a constant weight. For semi-digested prey, dry weight was estimated from the reference collection (for fish) and predictive functions (for invertebrates) quoted in Rogers, Buschbom & Watson (1977) and Díaz & Díaz (1990). We used two separate criteria to establish prey categories: taxonomy and foraging habitat. The relevance of the different prey categories was expressed by the percentage of occurrence (% P), numeric percentage (% N) and biomass percentage (% B). The width of the trophic niche was measured in relation to foraging habitats by Brillouin's diversity index (Pielou 1975), and a jack-knife procedure was used to estimate diversity at the population level, together with the associated variance (Zahl 1977).

Prey biomass percentages consumed in each regurgitate were compared between years using the K–W test and Mann–Whitney U-test with Bonferroni correction (Bosch, Oro & Ruiz 1994a; Buckley & McCarthy 1994). Diversities were compared using a modified Student's t-statistic (Hutcheson 1970).

Demographic data

Colony size and culling adults

From 1993 to 1996, the number of breeding pairs in the colony was estimated prior to the onset of the culls by multiplying nest density in each section by its area (Fortià & Hontangas 1991; Bosch et al. 1994b). To analyse the growth of the colony, we included data collected prior to the onset of the culls (Bosch, Oro & Ruiz 1994a).

The number of culled gulls was estimated by the staff of the Nature Conservancy Agency, who collected culled birds in the colony and searched those that died at sea from a boat (Sargatal, Saavedra & Romero 1992; see also Duncan 1978 and Coulson, Duncan & Thomas 1982). No culled birds were recovered away from the Medes Islands (the closest neighbouring colonies are at least 50 km away).

Modelling demographic parameters and population dynamics of the colony

Gulls have been ringed in the Medes Islands colony since 1976 (see Appendix). During culls, Nature Conservancy Agency (DARP) staff checked the killed gulls for rings. One-hundred and fifty-seven gulls ringed as fledglings at the colony were recovered in culls from 1992 to 1996. Ring recovery data were analysed with permission from the Ringing Office (ICONA and GCA). The local survival rate was estimated using ring recovery models (dead animal recoveries) running the program mark (White & Burnham 1997), which computes estimates of survival (φ) and recapture (r) using numerical maximum likelihood techniques. The deviance and the number of estimated parameters were used to compute the Akaike information criterion (AIC) value for the model. The model with the lowest AIC value was accepted as the most parsimonious model (Lebreton et al. 1992). As recoveries were made during the period of culling, we estimated survival rate only during this period.

We also compared the survival estimates during the culling from recovery models (φc) with the estimated annual survival rate (sc) for the same period using the annual local population growth rate λ, calculated in the following way: as Nt is the local population size at time t, asymptotically Nt is proportional to N0λt, and estimated λ is:

λ     =    (Nt/N0)1/t(eqn 1)

We used the censuses of the colony from 1960 to 1996 to estimate λ for each time interval. From these λ-values, we calculated the mean of λ for two periods: one for the period from 1960 until the start of the culls in 1992 (i.e. prior to culling, λpc) and one for the period of culling, from 1992 to 1996 (λc). Since from 1960 to 1992 the colony was not censused every year, λpc was estimated as the average value of λ as:

λ pc     =     Σ λ / N (N = number of censuses carried out from 1960 to 1992)(eqn 2)

whereas λc was estimated as:

λ c     =     Σ ntt    +  n  -  1/ Σ Nt    +  1…t  + n(eqn 3)

We estimated the change in λ (Δλ) as:

Δ λ     =     λ pc    -  λc(eqn 4)

Similarly, the change in survival ss) after the onset of the culling was estimated as:

Δ s     =     spc    -  sc(eqn 5)

The terms Δλ and Δs are related by the parameter of elasticity e as:

(Δλ/λpc)/(Δs/spc) ≈ e, to give Δs/spc ≈ (Δλ/λpc)/e(eqn 6)

We could use the generation time (T) of the colony to estimate e, as e = 1 − 3/T (Ricklefs 1983); we took a value of 10 years as the generation time and 0·90 as the survival rate prior to culling (spc) in order to estimate Δs, and consequently the survival rate after the onset of the culling (sc) as:

sc     =     spc    -  Δs(eqn 7)

This value of sc is approximate because of the non-linearity of Δλ. We used a very simple local population change model:

Nt    +  1     =     s Nt    +  recruitment(eqn 8)

where Nt is the number of breeding pairs in year t, for comparing the terms sc and φc, such that: if sc < < φc, recruitment decreased and there was probably an emigration to other colonies; and if sc >> φc, recruitment increased and there was probably an immigration from other colonies.

To confirm these results on recruitment and emigration, a simple matrix model was used as a complementary approach. This model relies on iterating the relationship:

Nt    +  1     =     ANt(eqn 9)

where A is the matrix of the model and Nt is the population vector at time t. In our case, we used the Leslie matrix model, in which the matrix A is constant. We analysed the years between 1992 and 1996, i.e. the whole period affected by the culls. The model assumes that the local population is closed and its growth is infinite and that there is no density dependence. A further assumption of the model is that the sex ratio of fledged chicks is 50 : 50, as we only modelled the dynamics of females. Fecundity (f) was calculated as half the productivity of the colony. Based on the life history of large gulls (Wanless et al. 1996), we considered three age classes in our model: gulls aged 1 year (juveniles), gulls aged 2 and 3 years (immatures) and gulls older than 3 years (breeding adults). Although fecundity may increase with the age of breeding adults and age at first breeding is affected by culls (Coulson, Duncan & Thomas 1982; Coulson 1991), we assumed that fecundity was the same for all age classes and that the proportion of breeders for each age class was also the same. Because the matrix model was run for a short number of years (from 1992 to 1997) and the population growth rate λ of this species is very sensitive to adult survival rate and not to other demographic parameters (such as age of first breeding or productivity), we considered that these assumptions did not affect the outcome of the model. The Leslie matrix model for the Medes Islands colony is shown in Table 1. We used the adult survival estimate obtained from ring recoveries in the matrix, and the number of breeding pairs in 1992 at the start of the simulation. From these data, and using the same values of fecundity and immature survival, the initial numbers of juveniles and immature birds were calculated and entered into the model, which was run on the ULM software (Legendre & Clobert 1995). Deviations of the model estimates of breeding adults from the data obtained through the censuses of the colony were modelled using different emigration/immigration rates.

Table 1.  Simple Leslie matrix model with four age classes to predict the changes in population dynamics of the yellow-legged gull colony on the Medes Islands during the culling period 1992–97 (P = 0·5*f*p*sj, where f is fecundity (per female), p is the proportion of breeders among adult age classes, and sj is juvenile survival; s1, s2 and s3 are immature survival rates and s4 is adult survival; the vectors show the number of gulls of each age class (n1, n2, n3 and n4) in year t and t + 1. Biological parameter values used in the model were estimated from studies published on large gulls (herring gulls Larus argentatus, Chabrzyk & Coulson 1976; herring and lesser black-backed gulls L. fuscus, Wanless et al. 1996; herring gulls, Pons & Migot 1995; S.N. Freeman & B.J.T. Morgan, unpublished data on yellow-legged gulls): f = 1·5; sj = 0·65; s1 = s2 = s3 = 0·82; s4 = 0·90; P = 0·90
 
image

Results

Breeding parameters

Density of nests

Nest density tended to decrease throughout the years (from 1993 to 1996) in the different sections of the archipelago (Table 2). The reduction was significant in sections where more than one-third of the section was included in the culled area, with the highest reduction in sections in which more than two-thirds of the section was in the culled area; however, the reduction was not significant in sections excluded from or only just included in the culled area.

Table 2.   Nest density (mean ± SE) of yellow-legged gulls in the different sections in which the Medes Islands were divided, expressed as the number of nests in strips of 30 × 5 m (150 m2). For each section, the intensity of culls is shown: high (more than two-thirds of the section was included in the culled area); moderate (between two-thirds and one-third of the section was included in the culled area; low (less than one-third of the section was included in the culled area); and nil (the section was excluded from the culled area). The number of 150-m2 sampled strips is shown in parentheses
  Year 
ZoneIntensity of culls1993199419951996K–W test
  1. † Nest density estimated by areas larger than 150 m2.

  2. Significant differences between sections: *P < 0·05; **P < 0·01; ***P < 0·005; ****P < 0·001.

FlatHigh6·8 ± 0·4 (24)4·3 ± 0·5 (23)3·2 ± 0·4 (29)2·3 ± 0·4 (31)46·18****
Moderate slopeModerate8·0 ± 0·7 (11)5·9 ± 0·7 (11)5·4 ± 0·5 (22)4·0 ± 0·4 (25)13·77***
Gregal ValleyHigh8·1 ± 0·5 (19)4·2 ± 0·5 (21)3·1 ± 0·4 (38)3·1 ± 0·3 (53)30·42****
High slopeModerate6·6 ± 1·0 (5)5·3 ± 0·8 (8)4·9 ± 0·8 (8)2·6 ± 0·7 (10)14·48***
Shore rocksLow5·8 ± 1·0 (5)4·4 ± 1·0 (5)3·6 ± 0·8 (8)3·0 ± 0·8 (8)6·30
CliffsNil1·8 ± 0·9 (6)1·5 ± 0·9 (6)1·4 ± 0·9 (6)1·3 ± 0·9 (6)1·66

If data corresponding to nests placed on cliffs were excluded from the analysis, differences in density between strips sampled in the unculled area and those sampled in the culled area were not significant in 1993 (Z = 0·7, P = 0·425), whereas in the remaining years differences were significant (1994: Z = 3·4, P < 0·001; 1995: Z = 4·0, P < 0·001; 1996: Z = 6·6, P < 0·001), and they increased from 1994 to 1996 (Fig. 2). In both culled and unculled areas, nest density decreased significantly through the years of culling (unculled area: H = 11·43, P = 0·010; culled area: H = 87·77, P < 0·001).

Figure 2.

Number of yellow-legged gull nests (mean ± SE) in strips of 30 × 5 m sampled in culled and unculled areas of the Medes Islands from 1993 to 1996.

Breeding performance

No significant difference in the frequency distribution of clutch size was detected between culled and unculled plots in any year (1992, before first cull: Fisher exact test, P = 0·272; 1993: Fisher exact test, P = 0·301; 1994: G1 = 1·24, P < 0·265; 1995: G1 = 2·82, P < 0·093; 1996: G1 = 0·09, P < 0·760) (Table 3). If each plot was considered separately, clutch size distribution did not differ significantly in the unculled plots between years (G1 = 5·40, P < 0·249). However, in the culled plots differences in clutch size between years were significant (G1 = 9·73, P < 0·045) and residuals showed that differences were due to a higher frequency of three-egg clutches both in 1992 and 1993. In two-egg clutches, the mean egg volume did not vary significantly between plots nor between years (plot: F1,54 = 1·76, P = 0·191; year: F4,54 = 2·12, P = 0·091; interaction: F4,54 = 1·82, P = 0·138). In contrast, in three-egg clutches the mean egg volume varied significantly between unculled and culled plots, whereas no significant year effect was found (plot: F1,346 = 14·38, P < 0·001; year: F4,346 = 0·87, P = 0·481; interaction: F4,346 = 0·89, P = 0·473). Multiple range analysis showed that eggs from the unculled plot were significantly larger than those from the culled plot (Table 4). Such differences were not detected when only data from 1992 (just prior to the first cull) were considered (t82 = 0·87, P = 0·389).

Table 3.   Clutch size (mean ± standard error) of yellow-legged gulls nesting in the culled and unculled plots sampled in a grassy area of the Medes Islands colony from 1992 (just before first cull) to 1996. The number of clutches sampled (n) is shown
 Unculled plotCulled plot 
 Clutch sizeClutch size  
Year123Mean ± SEn123Mean ± SEn
19924382·90 ± 0·05425512·91 ± 0·0456
19932382·95 ± 0·04403332·92 ± 0·0536
199412672·85 ± 0·0479213522·74 ± 0·0667
19955352·88 ± 0·054010262·72 ± 0·0836
19969362·80 ± 0·06456202·77 ± 0·0826
Table 4.   Egg volume (mean ± standard error) in three-egg clutches of yellow-legged gulls nesting in the culled and unculled plots sampled in a grassy area of the Medes Islands colony, from 1992 (just before first cull) to 1996. The number of clutches sampled (n) is shown
 Mean egg volume (ml)
YearUnculled plotnCulled plotn
199280·27 ± 0·963679·74 ± 0·8348
199382·49 ± 0·953879·09 ± 1·0233
199481·17 ± 0·864778·33 ± 0·8252
199582·50 ± 1·023378·89 ± 1·2522
199682·33 ± 1·112880·50 ± 1·3519

No differences were detected in the mass or body condition of chicks hatched in the first year of culling (1992) and those hatched 2 years after (1994) (chick hatching mass: t148 = −0·51, P = 0·61; body condition: t150 = −0·15, P = 0·88). A significant linear relationship between the mass of yellow-legged gull chicks and age was found at the Medes Islands colony in 1992 and 1994, including all the chicks measured with ages between 15 and 30 days (1992: r = 0·840, n = 25, P < 0·05, y = 34·82x − 165·43; 1994: r = 0·814, n = 41, P < 0·05, y = 29·56x − 32·28). However, there was no significant difference between years in the slope of the relationship between mass and age (ancovaF1 = 0·68, P = 0·413). Chick body condition throughout the growing period did not vary between these years (Fig. 3), nor did the growth rates of the parameters recorded (weight: Z = −0·90, P = 0·370; wing: Z = −1·14, P = 0·254; tarsus: Z = −0·18, P = 0·861; bill: Z = −1·01, P = 0·313). Furthermore, fledging condition did not vary significantly between the 2 years (t54 = −1·94, P = 0·06).

Figure 3.

Weight and wing length relationships for yellow-legged gull chicks of the Medes Islands colony in 1992 and 1994, including all the chicks measured up to death or fledging age (n = 57 in 1992, n = 68 in 1994). The same pattern was found between weight and tarsus length relationships.

Both fledging success and breeding success differed significantly between years (fledging success: G3 = 9·55, P = 0·023; breeding success: G3 = 8·45, P = 0·038) (Table 5). There was evidence of an increase in breeding performance during the culling period. Residuals of the contingency table showed that, for both parameters, success in 1993 was lower than expected, while in 1996 it was higher than expected. The total estimated number of young reared each year tended to decrease throughout the study (Table 5), except for a slight increase in 1995 due to the increase in the percentage of adults culled that year.

Table 5.  Fledging success (number of chicks fledged, as a percentage of eggs hatched) and breeding success (number of chicks fledged, as a percentage of eggs laid) of yellow-legged gulls breeding in an unculled grassy area of the Medes Islands colony from 1993 to 1996. From the young reared each year in the enclosure we estimated the mean number of fledglings per pair and the total number of young reared at the colony each year corrected by the culled pairs. The number of clutches sampled (n) is shown. Hatching success (number of eggs hatched, as a percentage of eggs laid) was similar between years, ranging between 91·5% and 94·5%
YearnFledging successBreeding successMean number of chicks per pairTotal fledglings reared
19935744·041·31,212 073
19945356·553·41,59222
19955257·752·81,58628
19964261·557·31,66371

Intraspecific predation

The percentage of chicks assumed to have disappeared as a consequence of intraspecific predation was significantly higher in 1992 (46%, n = 58 chicks) than in 1994 (28%, n = 64 chicks) (G1 = 4·46, P < 0·05).

Adult size and body condition

Weight was the only measurement of the adult gulls that differed significantly between years, both in males and females (males: F3,155 = 6·53, P < 0·001; females: F3,133 = 6·72, P < 0·001). However, differences did not result from any trend throughout the years of study (Table 6). Tarsus length, head length and body condition did not vary significantly between years in either sex (males: tarsus length F4,174 = 2·13, P = 0·079; head length F4,174 = 0·90, P = 0·466; body condition F3,155 = 2·56, P < 0·057; females: tarsus length F4,147 = 1·95, P = 0·106; head length F4,147 = 1·06, P = 0·381; body condition F3,133 = 2·42, P = 0·069).

Table 6.   Body measurements (mean ± SE) of male and female yellow-legged gulls breeding in a grassy area of the Medes Islands colony from 1992 to 1996. Body condition was calculated as the residual from a linear regression of body mass on both tarsus length and head length. The measurements are in mm or g
 MalesFemales
YearnWeightHeadTarsiBody conditionnWeightHeadTarsiBody condition
199220 129·4 ± 0·670·1 ± 0·5 15 119·4 ± 0·764·3 ± 0·6 
1993581159·3 ± 10·1129·9 ± 0·471·1 ± 0·31179·5 ± 2·145988·9 ± 11·5119·1 ± 0·465·6 ± 0·4986·5 ± 2·5
1994461154·0 ± 11·3130·4 ± 0·470·9 ± 0·41177·6 ± 3·132934·7 ± 13·7118·8 ± 0·564·5 ± 0·4978·0 ± 2·8
1995241230·8 ± 15·8130·4 ± 0·669·8 ± 0·51167·6 ± 4·9281007·9 ± 14·6119·5 ± 0·564·4 ± 0·5977·2 ± 4·7
1996311202·0 ± 13·9130·8 ± 0·570·3 ± 0·41172·4 ± 3·632999·4 ± 13·7120·1 ± 0·565·5 ± 0·4986·2 ± 3·2

Diet composition

A total of 237 regurgitates of fledging chicks were analysed from 1992 to 1996 (excluding 1993), and 769 prey items were identified. The relative importance of the different prey identified and the foraging habitats assigned are shown in Table 7. No significant differences in the biomass percentages of prey from crops (K–W: = 3·44, P = 0·33), sea (K–W: = 4·05, P = 0·26) or refuse (K–W: = 7·51, P = 0·07) were detected between years (Table 7).

Table 7.  Items of importance in the diet of yellow-legged gulls in the Medes Islands during 1992, 1994, 1995 and 1996 based on taxonomic and typological (i.e. foraging habitat) categories (see text for details). Diversity indices are also given at both taxonomic and typological levels. (%N = numeric percentage; %P = percentage of occurrence; %B = percentage of biomass)
Year
No. regurgitates
No. items
1992*
34
200
1994
39
70
1995
92
284
1996
72
215
%N%P%B%N%P%B%N%P%B%N%P%B
  1. * Data from Bosch, Oro & Ruiz (1994a).

Taxonomic
Order Clupeiforms1·05·95·910·018·015·23·28·76·86·516·713·5
Order Perciforms   2·95·15·11·13·33·33·38·38·0
Order Anguiliforms         0·51·41·4
Indeterminate fish4·526·523·24·37·77·01·13·32·70·92·81·9
Order Rodentia   1·42·62·6      
Subclass Pulmonata5·58·81·4   25·42·21·23·71·40·9
Order Oligochaeta20·58·83·031·42·62·335·97·66·442·35·62·5
Order Orthoptera   1·42·60·10·41·10·0   
Order Hymenoptera      0·41·10·0   
Order Lepidoptera      0·41·10·2   
Order Coleoptera38·517·62·24·32·60·20·71·10·12·81·40·5
Order Diptera1·05·90·0   0·41·10·00·91·40·1
Order Dermaptera1·05·90·1         
Order Isopoda9·011·81·4   2·12·20·47·02·80·7
Order Diplopoda      0·71·10·1   
Order Amphipoda1·02·90·0         
Olea europaea fruits (olives)3·02·92·7      1·42·81·9
Waste food (e.g. meat and organic rubbish)15·061·860·144·371·867·527·580·478·930·773·668·6
Typological
Sea6·532·429·117·130·827·45·315·212·811·229·224·8
Crops78·526·510·838·65·15·167·310·98·358·19·76·6
Refuse tip15·061·860·144·371·867·527·580·478·930·773·668·6
Population diversity (SE)
Taxonomic2·6 (0·3)  2·7 (0·7)  2·6 (0·4) 2·5 (0·3)   
Typological0·9 (0·3)  2·0 (0·4)  1·1 (0·2) 1·3 (0·2)   

Dietary niche width measured through diversity indices was very similar among years (Table 7), and differences were not significant in the 12 tests performed, at either taxonomic or typological level (all Hutcheson-modified Student's t-statistics P > 0·1).

Demographic data

Colony size and culling

The number of breeding pairs increased from c. 3000 in 1960 to c. 14 000 prior to the onset of the culls in 1991 at a steady rate of 5% per year (see below and Fig. 4). The peak in numbers was recorded in 1991, with c. 14 100 pairs. Since 1992, between 21% and 29% of the breeding adults have been killed, and the colony size progressively decreased to c. 6650 pairs in 1996, at an estimated rate of decrease of 19% per year (see below and Fig. 4).

Figure 4.

Number of breeding pairs of yellow-legged gulls (▪) in the Medes Islands, 1960–96 (data also from Bosch, Oro & Ruiz 1994a). The number of culled adults from 1992 to 1996 is also shown (●) (data from the Nature Conservancy Agency).The values of population growth rate λ are shown for the different time intervals and the mean for the periods prior to (λpc) and after (λc) the onset of the culls in 1992.

Modelling demographic parameters and population dynamics of the colony

Based on ring recoveries through the program mark, we found that annual adult survival during culling (φc) was constant (Table 8), estimated at 0·742 (SE = 0·07).

Table 8.  Modelling survival and recapture probabilities of culled yellow-legged gulls at the Medes Islands colony. The model with the lowest AIC value is the model that was finally selected (boldfaced model); Dev = deviance of the model, np = number of estimated parameters; AIC = 2*np*Dev
Biological hypothesisModelDevnpAIC
Time dependence of survival and recapture [ φ t, pt]687·328703·32
Age dependence of survival and time dependence of recapture [ φ a, pt]689·108705·10
Constant survival and time dependence of recapture [ φ, pt]689·475699·47
Time dependence of survival and constant recapture [ φ t, p]704·385714·38
Constant survival and recapture [ φ, p]752·542756·54

Once the values of λ for the periods prior to and after the onset of the culls (λpc and λc, respectively) were estimated (Fig. 4), we calculated Δλ as 0·232 and Δs as 0·286, to finally obtain a survival value after the onset of the culls of sc = 0·614. This value of sc was lower than that estimated from recoveries φc, suggesting that there was an emigration of gulls born in the Medes Islands. Similarly, the predicted population trend using a simple Leslie matrix model was consistent with the actual population counts only when high emigration rates were considered, especially after 1993 (Fig. 5).

Figure 5.

A comparison of observed numbers of breeding pairs on the Medes Islands colony during the culling period (1992–97) (solid line) and those predicted by a simple Leslie matrix model with no emigration (- -) and with different annual emigration rates: 0·05 (–·–), 0·10 (– - - –) and 0·25 (––).

Discussion

Effects of culling on nest density and related ecological parameters

In the short term, culling in the yellow-legged gullery of the Medes Islands affected several parameters associated with the breeding ecology of the colony. Nest density decreased throughout the culling period, while the area occupied by the colony remained the same as before the onset of the culls (cf. Duncan 1978; Coulson, Duncan & Thomas 1982; Wanless & Langslow 1983). The reduction in density was lower in those sections that were less intensively culled (see also Wanless & Langslow 1983). However, a significant reduction in density was detected even in the unculled area. We suggest three non-exclusive explanations: first, disturbances caused by the culls may have also decreased the density in the unculled areas (Kress 1983; Skira & Wapstra 1990); secondly, some birds nesting in the unculled areas may have moved to the culled areas, where nest site availability increased as a result of the cull; and thirdly, new recruits may have selected the lower density areas and thus relatively few recruited into the unculled areas because of the more extensive low densities made available by the culling (J.C. Coulson, personal communication). However, this reduction cannot be attributed to a decrease in the recruitment rate resulting from the reduction in breeding numbers (Chabrzyk & Coulson 1976; Coulson 1991), as effects were detected immediately after 1992, before the recruitment of the first cohorts affected by the culling. Some areas that were not culled, such as cliffs, may act as a refuge in future years, and as a result nest density should increase in these areas (Alvarez 1992). However, nest density on the cliffs at the Medes Islands did not increase after the start of the cull in 1992, probably because nest sites on the cliffs were already occupied at that time.

In several gulleries, a decrease in nest density has led to changes in several breeding parameters, due to a reduction in competition at the feeding areas (Coulson 1991) and/or a reduction in interactions (especially intraspecific predation; Parsons 1976; Burger 1984; Kilpi 1989, 1995) within the colony. A reduction in food competition caused by the decrease in breeding numbers may bring about an improvement in breeding performance (increase in clutch size, egg size, chick size, chick growth, breeding success and adult body condition; Coulson, Duncan & Thomas 1982; Spaans, de Wit & Van Vlaardingen 1987; Coulson 1991). In the Medes Islands, the pattern of variation of most of the breeding parameters did not seem to be linked with a reduction in food competition. No increase was recorded in egg size, chick size, chick growth or adult body condition as nest density decreased, whereas clutch size decreased slightly over time (throughout the study) and in the culled areas. The reduction in clutch size only in the culled areas may be the immediate result of recruitment to these areas of young breeders (Coulson, Duncan & Thomas 1982), which normally show a lower breeding performance (Reid 1988; Pugesek & Diem 1990; Pyle et al. 1991; Sydeman et al. 1991). Mean egg size in the unculled areas was significantly larger than in the culled areas, supporting this idea. This study may have been too brief to reveal an increase in size of both eggs and adults recorded in a culled gullery by Coulson, Duncan & Thomas (1982).

In the Medes Islands, the breeding success of yellow-legged gulls increased throughout the culling period. Again, a reduction in food competition could not have induced this pattern as chick size, chick growth and chick body condition did not vary between years. As in other studies (Spaans, de Wit & Van Vlaardingen 1987; Kilpi 1989), the lower nest density induced by culls reduced intraspecific predation on chicks, which was an important source of chick mortality, especially when nest density was highest (Carrera & Vilagrasa 1984). In this colony, nest spacing might be a convenient measure of the effects of reduced density on population regulation, as it appears to affect breeding success directly through intraspecific predation (cf. Jehl 1994). Results also suggest that short-term changes in adult body condition of yellow-legged gulls at the Medes Islands colony are not linked to a reduction in food competition. It is likely that food available to the gulls was already above the limiting threshold before the first cull in 1992. When food availability changes, it has been shown that yellow-legged gulls and other opportunistic seabirds normally vary their diet (review by Oro 1999 and references therein). However, diet composition in our study did not change when colony size was decreased, and gulls continued to exploit the same foraging resources. Furthermore, colony size has increased markedly during the last few decades, from 6000 to 28 000 breeding gulls in 30 years, also suggesting a lack of food limitation around the colony (Bosch, Oro & Ruiz 1994a).

Population dynamics of the medes islands colony and reliability of culling as a management measure

Between 1992 and 1996, over 25 000 adults were killed in the Medes Islands colony, reducing its size by more than 40%. The cull was similar in intensity to that performed on the Isle of May, Scotland, during the first years of culling (Duncan 1978; Coulson 1991). In the present study, modelling suggested that many birds emigrated after the onset of the cull, probably because intensive culling drove recruits away to other colonies, as previously recorded for colonies of herring gulls L. argentatus and lesser black-backed gulls L. fuscus culled in Britain (Duncan 1978; Coulson, Duncan & Thomas 1982; Coulson 1991). According to the density-dependent model of recruitment in herring gulls by Chabrzyk & Coulson (1976), a decrease in breeding density, resulting from a cull such as that recorded in the Medes Islands, makes a colony less attractive to new recruits.

Culling has been used widely as an effective short-term measure to reduce nest density and numbers of individuals, with the assumption that these reductions will in turn reduce problems caused by gulls (Spaans & Blokpoel 1991). As expected, nest density and numbers of individuals decreased in the Medes Islands gullery following the annual culls. However, culling also modified parameters related to the population dynamics of this colony, such as breeding success, recruitment and, indirectly, the dynamics of neighbouring yellow-legged gulleries. In a herring gull colony submitted to culls, Coulson (1991) estimated that c. 70% of young birds bred in a colony other than that in which they hatched, and they may breed up to 250 km from their natal colony (Chabrzyk & Coulson 1976). The culling at the Medes Islands colony may also have failed to dramatically reduce the breeding numbers of yellow-legged gulls at the metapopulation level, due to the spatial heterogeneity of the environment (Levins 1969; Wootton & Bell 1992) and the ability of gulls to disperse between local populations (i.e. colonies) (Coulson 1991; Munilla 1997; Morais, Santos & Vicente 1998; Oro & Pradel 1999). In fact, a large increase in the number of breeding yellow-legged gulls at neighbouring colonies has been recorded recently (Paterson 1997), especially in the Ebro Delta, 270 km south of the Medes Islands, where annual population growth has increased by 14% compared with that recorded before 1992 (D. Oro, unpublished data). Therefore, the culling at the Medes Islands might be reducing the number of breeding pairs by shifting them to other sites, and thus transferring any problem linked to the presence of gulls to other sites. Moreover, it is known that non-breeding adults, which may represent an important fraction of the local population and are not affected by culls, may have an improved recruitment probability when colony size is reduced (Coulson, Duncan & Thomas 1982; Klomp & Furness 1992). The planning of culling programmes by wildlife agencies has not addressed the multiple factors responsible for the population dynamics of a colony together with the relevant metapopulation features. Feare (1991) concluded that in the long term culling fails to reduce metapopulation size because of immigration into areas where birds are not killed together with the compensatory changes in production, survival and recruitment. Consequently, as long as the factors that cause a superabundance of gulls continue to operate, their impact on some bird and human populations is likely to persist (Spaans & Blokpoel 1991).

Acknowledgements

We thank Anna Galdeano, Núria Pocino, Daniel Sol, Toni Orantes, Jordi Piró, Kim Bosch, Oriol Rius, Vittorio Pedrocchi, Marilen Baixas, Chema Serrano, members of GCA and other friends for their help in the field work; and Port Autònom de Barcelona and the Servei de Vigilància de les Illes Medes for their logistic support. Jean-Dominique Lebreton, Roger Pradel and Giacomo Tavecchia helped with the demographic analyses. Guillermo Blanco, John Coulson, Daniel Sol, Sarah Wanless and an anonymous referee provided very helpful comments on the manuscript. Robin Rycroft from SAL improved the English text. We are also indebted to Enric Carrera for his effort in ringing large numbers of gulls prior to the cull, which made this research possible. We wish to state that we did not participate in the culls performed in the Medes Islands, nor were we responsible for these culls, nor did we advise these measures.

Received 18 January 1999; revision received 6 January 2000

Appendix

Table 9. Number of yellow-legged gulls ringed as young and subsequently recovered in the culls of the Medes Islands, according to their cohort. Data provided by the Oficina de Anillamiento (Ministerio de Medio Ambiente, Spanish Government).
   Year of culling
CohortNumber of ringed chicks19921993199419951996
197646994100
19771177287131
1978943186100
197961793011
198054342020
19811813165200
198295874010
19832057151000
19847720000
198510341420221
1986501000
1987000000
1988200000
1989000000
1990000000
199113870083218
19924000001

Ancillary