Sexual size dimorphism in the extreme SW breeding population of the European Storm Petrel Hydrobates pelagicus ( Aves : Procellariformes )

We used molecular sexing and morphological analysis to characterise sexual size dimorphism (SSD) in the European Storm Petrel Hydro­ bates pelagicus (Linnaeus, 1758) in the southernmost and westernmost breeding colony (El Hierro, Canary Islands).We analysed SSD in 79 individuals (46 males, 33 females). We performed discriminant analysis on 9 morphometric variables and correctly classified 70.5% of the original cases. By binary logistic regression, the accuracy in sexing birds previously sexed by DNA analysis was 70.5%. Females were larger than males in wing and tarsus length but there were no differences for bill size and shape. There was a large biometric overlap between sexes in this Canarian breeding colony of H. pelagicus. However, a moderate female-biased dimorphism in flight-relevant variables such as tail and wing traits could be of interest in trophic segregation of sexes and optimization of resource use in these pelagic-feeding birds.


Introduction
Sexual size dimorphism (SSD) is a widespread feature in many bird species (anderson 1994, Lezana et al. 2000, Campos et al. 2005).In most bird taxa, males are of the same size, slightly larger, or significantly larger than females (priCe 1984) and it is usually more pronounced in aggressive and predatory species (ThioLLay 1994).In raptors, females are larger and in extreme cases may have twice the weight of the males (ThioLLay 1994).By contrast, in some terrestrial, cursorial birds such as bustards, females are much lighter and smaller (CoLLar 1996).The evolution of SSD has probably occurred through sexual selection together with promiscuous mating (LaCk 1971), reflecting differences in sex-specific reproductive roles and responses to environmental factors (BLondeL et al. 2002, GonzáLez-soLís 2004), feeding behaviour (de Juana et al. 2004), trophic niche segregation of sexes and sexual selection (GuTiérrez-CorChero et al. 2007, navarro et al. 2009).
Several methods have been used to sex bird species which are monomorphic in plumage.Molecular sexing consists of a non-intrusive technique and the most popular, requiring only a small amount of DNA from blood or a single feather (TaBerLeT & BouveT 1991, risTow & wink 2004, suTherLand et al. 2004, infanTe & peris 2004).For many species of birds including Procellariiformes, sexing has been achieved by discriminant analysis using morphological measurements (risTow & wink 1980, CouLTer 1986, Lo vaLvo 2001, BuTLer & GosLer 2004, aLarCos et al. 2007).
The European Storm Petrel Hydrobates pelagicus (Linnaeus, 1758) is a small pelagic seabird species distributed in the NE Atlantic from S Iceland to NW France, N Spain, Salvages and Canary Islands; it is also present in the Mediterranean from N Spain to Greece (deL hoyo et al. 1992).Hydrobates pelagicus has been considered a monotypic species although some authors suggest that the Mediterranean population must be re-established as a valid subspecies (H.p. melitensis sChemBri, 1843).This distinction has been supported not only by biometrics (he mery & d'eLBee 1985, massa & CaTaLisano 1986, Ca Ta Lisano et al. 1988, LaLanne et al. 2001) but also by genetics (CaGnon et al. 2004).The Canary Islands correspond to the southern limit of the European Storm Petrel distribution range and represent the only known nesting place among the Macaronesian archipelagos.On the Canary Islands, a minimum of 1,000 breeding pairs of H. pelagicus have been estimated occupying small islets and rocks such as Alegranza (N of Lanzarote) and Roque Grande de Salmor, a rocky islet off El Hierro (noGaLes et al. 1993).Like other pelagic seabird species, H. pelagicus is a colonial breeder nesting in caves or small crevices, under heaps of boulders or under collapses of coastal cliffs (deLGado et al. 1985, noGaLes et al. 1993).
Geographical variation in morphology and ethology has great importance in species-specific recognition (BreTaGnoLLe 1989).Variation underlying speciation processes is methodologically difficult to delimit in these marine birds (BoLTon & Thomas 2001, CaGnon et al. 2004 and references there in).Morphological differences have been found for H. pelagicus even within the Mediterranean Basin (LaLanne et al. 2001), where there is, however, low genetic variation.Genetic differentiation is higher among the Atlantic populations.In addition, there is strong evidence of a lack of genetic exchange between Mediterranean and Atlantic populations.
Among the Atlantic Ocean populations, movements of Canarian H. pelagicus are poorly known.There have been records of birds reaching South Africa, and the islands would receive European migrants (marTín & Lorenzo 2001).Despite the wide distribution of H. pelagicus, the large population size of this species (130,000 -290,000 pairs after Carboneras 1992; 130,000 -400,000 pairs after franCo et al. 2004), and the morphological vari-ation throughout the distribution range, biometric data characterizing live specimens are scarce.There is to our knowledge no available biometric work from the breeding colonies of the Canary Islands.Moreover, because of post mortem shrinkage of museum skins (kuCzynski et al. 2003) our data from live birds become especially relevant.Here we assess SSD through molecular sexing and biometry, and analysed external morphological variation between sexes of the European Storm Petrel from its southern-and westernmost known breeding colony.

Study site
During the summer of 2007, field work was performed in Roque Grande de Salmor, a rocky islet of 0.03 km 2 (NW El Hierro island; 27º 49′ N 17º 59′ W, Canary Islands; Fig. 1).This is one of the largest colonies in the Canaries and the largest of El Hierro, with up to 200 -300 pairs (noGaLes et al. 1993, marTín & Lorenzo 2001).

Biometrics
A total of 79 individuals of H. pelagicus (46 males and 33 females) were caught by mist nets.Birds were ringed with conventional metal rings and no tape lures were used.All the biometric measurements were taken by G. deLGado CasTro.
Wing length (maximum chord) was taken using a stopped rule (svensson 1992) to 1 mm.Tail length was measured using an unstopped rule to the nearest millimetre.Following hemery & d' eLBee (1985), five morphological variables of bill were measured to 0.1 mm using digital callipers: bill length (along the dorsal mid-line from the edge of the feathers at the base of the culmen to the most distant part of the hook), bill depth 1 (BD1, mandibles + narines), bill depth 2 (BD2, mandibles at level of the opening of narines), narine width (BW1), bill width (BW2, mandibles at level of the opening of narines) and "maximum" tarsus length (strictly, tarsometatarsus length; SuTherLand et al. 2004).Body mass was estimated using a 50 g Pesola balance to 0.1 g.

Molecular sexing methods
Blood samples were collected by venipuncture of the brachial vein.Total DNA was isolated from 100 μl of blood using standard Proteinase K (Merck, Darmstadt) and phenol/chloroform procedures (samBrook et al. 1989).Sex identification was conducted following kahn et al. (1998).PCR was performed with 30 -60 ng of template DNA in 25 μl reaction volume containing 8 pmol of the primer H1272 and 9 pmol of the primer L1237, 0.1 mM of dGTP, dCTP, and dTTP, 0.045 mM dATP, 1 μCi [α-33 P]-dATP (AmershamBiosciences), 0.6 units of Taq-Polymerase (Pharmacia Biotech, Freiburg) and 2.5 μl of 10 × amplification buffer [10 mM Tris-Hcl (pH 8.5), 50 mM KCl and 1.5 mM MgCl 2 ].Each reaction was overlaid with two drops of mineral oil.Thermo cycling was performed with a Trio Thermo block TB1 (Biometra, Göttingen).Following the initial 5 min denaturation at 94°C, the program consisted of 31 cycles of 30 s at 94°C, 40 s at 56°C, 40 s at 72°C and 5 min at 72°C for final elongation.DNA fragments were separated by vertical PAGE (Polyacrylamide Gel Electrophoresis; length 40 cm) for 2 h at 65 W using a Base Acer Sequencer (Stratagene).After drying, the denaturing gels were exposed for 24 h to X-ray films (BioMax MR Film, Kodak).The bands were analyzed visually.The presence of two bands was scored as female and one band as male (kahn et al. 1998).

Data analysis
A Kolmogorov-Smirnoff test was used to check for normality of variables.We used a Student t-test to analyse biometric differences between the sexes.For nonnormal variables (wing length, tail length and mass) we used a non-parametric Mann-Whitney test.A Principal Component Analysis (PCA) was carried out in order to find a non-redundant pool of explanatory variables for biometrics of Hydrobates pelagicus.We applied the varimax rotation to achieve a better interpretation of the data pool.Discriminant analysis (sokaL & roLf 1995) was used to generate a function which is a linear combination of the morphometric variables measured from individuals of known sex, and to maximize the variance between groups (i.e.sexes).The method applied for calculating the discriminant function was Wilk's Lambda, and the probability for including a predictor variable in the model was set to 0.05 (0.1 for exclusion).Finally, a binary logistic regression function was calculated.All statistical analyses were performed using SPSS 12.0.

Sexing
The discriminant linear model based on the original combination of 9 morphometric variables was highly significant (Wilk's Lambda = 0.752; F = 12.361, P < 0.001) and included wing length and tail length as the best discriminators (Tables 4 and 5).The discriminant analysis correctly classified 70.5% of the original cases.
A binary logistic regression function was calculated to predict sex from a full set of variables (n = 79 individu- als).Backwards deletion of the variables was used so that only those elements that were significant (P < 0.05) were retained.This generated a significant binary logistic regression function (Chi-square = 23.27,df = 2, Nagelkerke R 2 = 0.347, P < 0.001), including only wing and tail length as significant parameters: X = 0.49 (wing length) -2.87 (tail length) + 18.07.

Discussion
We found that sexing H. pelagicus from the Canary Islands by using simple morphometric measurements renders a low degree of accuracy.There is a high degree of overlap between sexes in evolutionarily relevant biometrics such as bill size and shape.Also, as in the few studies published, we found a moderate female-biased sexual size dimorphism.Our results suggest the existence of large geographical variation in biometric parameters of H. pelagicus.Mean wing length of individuals caught at different North Atlantic breeding colonies have been reported to range between 116.7 and 123.7 mm (Table 2).Data regarding weight show similar trends (furness & BaiLLie 1981, and references therein).Nevertheless, furness & BaiLLie (1981) showed that storm petrel breeding on St. Kilda (Scotland) presented longer wings than wanderer birds, an observation which is also supported by fowLer et al. (1986) at Yell (Shetland), where statistical differences were found between breeding and non-breeding birds.Although a proportion of birds at Roque Grande de Salmor may be non-breeding birds (at least 70% of the birds caught regurgitated food on nets), these were clearly lighter (Table 1).The Mediterranean birds are the larger ones with a wing length range of 129.25 -123.58 mm.The tail length is quite similar between Atlantic and Mediterranean populations (Table 2).Accuracies in sex assignation vary widely in the literature (Appendix 1).In studies using discriminant analysis for larger species, similar degrees of accuracy have been either enough (i.e.Tundra Swan, 74.1%;miLLer et al. 1988) (1983), where a discriminant function based on the same variables accounted for 88.5% and 80.0% for males and females, respectively.We achieved the best separation between sexes by using the wing and tail discriminant functions.
Our results on SSD in the European Storm Petrel support findings of previous works with biometrics from live birds (James 1983).Tail and wing were larger in females than in males in this El Hierro breeding colony.Other biometric variables showed no statistically significant differences.For example, bill size and shape did not differ between sexes in this population of H. pelagicus.Bill size and shape is one of the most relevant sexually dimorphic variables with repercussions in foraging, sexual selection and speciation in seabirds (e.g. navarro et al. 2009).In this population, however, H. pelagicus showed a slight departure from monomorphy (the degree to which sexes are of equal size) in tail wing size, with probable effects in resource partition between sexes and intersexual competition (fiGueroLa 1999).In H. pelagicus, the variables involved in dimorphism are directly related to flight performance, which might be relevant for foraging optimization in these pelagic-feeding birds.Furthermore, the observed intersexual variation may have repercussions in the optimization of intraspecific resource use (navarro et al. 2009).The influence of low SSD in aspects of evolutionary biology and conservation of H. pelagicus such as assortative mating and mating success, phylopatry, foraging strategies, namely differential niche exploitation, especially when sea productivity fluctuates due to drastic environmental change, or sexual selection, deserves further research.

APPENDIX
Examples of sex determination in birds and degrees of accuracy by Determinant Function Analysis (DFA) based on various measurements.

Fig 1 .
Fig 1. Study area.Location of the breeding colony of Hydrobates pelagicus studied in El Hierro, Canary Islands.

Table 1 .
Population biometrics.Biometrics for the European Storm Petrel at Roque Grande de Salmor (NE El Hierro).Data are: means ± standard deviation (range), sample size: n.All data in mm except mass (g).

Table 3 .
PCA on biometric variables.Results of the Principal Component Analysis (PCA) on biometric measurements of Hydro bates pelagicus in Roque de Salmor, El Hierro (Canary Islands).Minor scores for each component and each variable have not been included.See Methods: Biometrics for explanation of codes.

Table 4 .
Discriminant-functions coefficients.Coefficients of the discriminant functions for both sexes and of the final canonical discriminant functions for the model (CDF).

Table 5 .
Correlations between variables and functions.Correlations between the discriminant variables and the canonical discriminant function.*Variablesnotincluded in the analysis; correlations of higher order are shown in bold.PCA biometrics diagram.Biplot of a PCA run on the measured set of biometrical variables, with samples (individuals) separated by sex.Our values are also smaller than those above 92% found in Larus argentatus(fox et al. 1981)and Puffinus ten uirostris(einoder et al. 2008).Lo vaLvo (2001) found that bill depth and body mass were the best discriminants (92 and 84% of correctness) for Calonectris diomedea in Sicily.The values presented by our Hydrobates population were lower than those of James