Study area and bird captures

In the first half of the 20th century the Spanish purple swamphen population was restricted to a few localities in southern Spain. In the Ebro Delta (NE Spain), it became extinct in the 1950s but recolonised this area in 1993, with breeding first confirmed in 1995 (Martí-Aledo, 2004). The origin of birds that started the recolonisation process is unknown. Nowadays the species is widespread and very abundant in this area, where groups of up to 150 individuals can be observed in some lagoons and rice fields.

This study was conducted during the “Delta Lagoon” Life+ project of the European Commission. Between January 2011 and May 2014, the project included intensive monitoring of purple swamphens in the marshes of the Alfacada lagoon in the Ebro delta (40° 40′ 44″ N, 0° 50′ 03″ E). The monitoring consisted of monthly capture sessions during two to three consecutive days, using seven large funnel traps located throughout the lagoon. The traps were baited with sorghum and inspected daily prior to and during capture sessions. All captured birds were marked with a metal ring and a plastic ring with an engraved two-digit individual code. Capture sessions were conducted monthly (except in August 2012 and 2013) but most morphological data were collected from winter to mid-spring (December to April) when most captures (90%) occurred.

Measurements

For each bird, we measured four bill dimensions: culmen plus shield length (culmen), nalospi measured from nostril to tip of bill (nalospi), depth at nostril (bill depth), maximum shield width (shield), head plus bill length (head), tarsus length (tarsus), tarsus plus mid-toe length without nail (tarsustoe; after Baker 1993), maximum flattened wing chord (wing), length of the second outermost primary (P2) and body mass (mass) (fig. 1). Bill measurements were taken following Craig et al. (1980), except for bill depth. All measurements were made by the same person (AB) with a digital caliper (precision: ± 0.02 mm), except for wing, tarsus-toe and P2, which were measured with a ruler (precision: ± 1 mm) and mass, measured with a spring balance (precision: ± 10 g). In order to assess the repeatability of measurements (all biometrical traits except mass), a sample of 109 individuals was measured twice during the same capture event, removing the caliper or ruler between consecutive measurements. We estimated an intraclass coefficient of correlation [ICC] following Sokal and Rohlf (1995) and Lessels and Boag (1987). Repeatability values were high for all assessed variables and ranged between 0.905 (bill depth) to 0.997 (nalospi). All repeatability values were significant at P < 0.0001.

Fig. 1.

Biometncal measurements of study purple swamphens (morphometric analysis) and the area analysed for scoring lesser, median and greater underwing-coverts.

[Medidas biométricas del estudio del calamón común (análisis morfométrico) y del área analizada para mareajes de las coberteras inferiores pequeñas, medianas y mayores de las alas.]

Age determination

During captures, we took a series of digital photographs in order to describe plumage characteristics (head, breast, underwing and upperwing, tail; figs. 2, S1). Juveniles have a partial post-juvenile moult, normally completed four months after hatching, and adults have a simultaneous complete moult in summer (Cramp and Simmons, 1980). The age at first breeding is unknown, but adult plumage is achieved after the first summer complete moult (see results). In southern Spain, reproduction was reported from February to September (Sánchez-Lafuente, 2004). However, at the Ebro Delta we only observed or captured chicks (< 2 months old) from July to September and juveniles from September onwards.

Fig. 2.

Examples of purple swamphen photographs used to score immature vs. adult plumages. A) and B) Immature plumage of breast, thighs and underwing-coverts (UWC), coded as 1-1-1 on 25 January 2011; C) immature plumage of breast and thighs (1-1-0 on 27 January 2011); D) immature plumage of breast and UWC (1-0-1 on 21 February 2012); E) immature plumage of breast only (1-0-0 on 27 March 2012); F) immature plumage of thighs and UWC (0-1-1 on 18 May 2011); G) immature plumage for UWC (0-0-1 on 26 January 2011 ); and H) adult plumage (0-0-0 on 26 January 2011 ). See methods for a more detailed explanation of plumage codes.

[Ejemplos de fotografías de calamón común usadas para puntuar el plumaje de los inmaduros y los adultos. A) y B) Plumaje de inmaduro en el pecho, los muslos y las infracoberteras (IC), codificado como 1-1-1 el 25 de enero de 2011; C) plumaje de inmaduro en el pecho y los muslos (1-1-0, 27 enero de 2011); D) plumaje de inmaduro en el pecho y las IC (1-0-1, 21 de febrero de 2012); E) plumaje de inmaduro solo en el pecho (1-0-0, 27 de marzo de 2012); F) plumaje de inmaduro en los muslos y las IC (0-1-1, 18 de mayo de 2011 ); G) plumaje de inmaduro únicamente en las IC (0-0-1, 26 de enero de 2011); y H) Plumaje adulto (0-0-0, 26 de enero de 2011). Para una mayor explicación de los códigos de plumaje ver el apartado de métodos.]

Bird handbooks state that juveniles and immatures (EURING age codes 3 and 5; EURING 2010) have duller plumage and bare parts than adults and can retain some grey body feathers, typical of the juvenile plumage, after their partial post-juvenile moult (e.g. Cramp and Simmons, 1980; Taylor, 1996). Throughout our fieldwork, we used these characters to age birds and we also recorded new plumage characteristics that could help ageing. For the sake of simplicity, we used the term immature to describe birds that have lost their chick plumage (dull black down) but do not yet display a typical adult plumage before first summer complete moult.

To assess the reliability of ageing criteria applied in the field, we examined the digital pictures of captured birds at the end of fieldwork (2014) and described plumage characteristics of three body areas: 1) breast and belly, frontal view (breast); 2) frontal view of both thighs (thighs); and, 3) under - wing-coverts (UWC) (fig. 2). We only used data from birds for which we could clearly characterise these three body areas from good-quality pictures (425 captures in total). The same observer (SR) viewed all photographs and scored the presence (scored as 1) or absence (scored as 0) of grey or whitish feathers on the breast and thighs. In the same way, the presence (1) or absence (0) of underwing-coverts with whitish tips was recorded. Birds were classified as immature, independently of the age assigned during the fieldwork, if at least one of these three body parts (breast, thigh or UWC) had a score of 1. Although most birds were photographed on several occasions, we only used one picture for each individual, normally taken at first capture.

As field work progressed, we noticed that the underwing-covert characteristics could also be a reliable character for ageing (see results) so we took pictures of these systematically beginning in March 2013 (232 birds). The same observer (SR) separately scored the lesser, median and greater under - wing-coverts as follows; (0) no whitish feather tips, (1) fewer than five feathers with a whitish tip, (2) up to 50% of feathers with whitish tips, and (3) over 50% of feathers with whitish tips (fig. 1).

To investigate the extent and chronology of immature plumage, we used a total score (hereafter body plumage score) that consisted of the sum of scores for each body area (breast, thigh, UWC; maximum score = 3) and a total score for underwing-coverts (hereafter UWC score) that consisted of the sum of scores for lesser, median and greater underwing-coverts (maximum score = 9).

Blood collection and DNA-based sexing

Blood samples (approx. 0.2 ml) were collected from the brachial vein or the intertarsal vein, kept in ethanol and stored at -20 °C until analysis in the laboratory. After centrifugation and removing the ethanol supernatant from these samples, the blood cells were collected and DNA was extracted following the protocol of a commercial kit (QIAGEN Blood and Tissue kit). The quality of the DNA was assessed by spectrophotometry.

DNA samples were amplified using the P2 and P8 primers described for sex determination (Griffiths et al., 1998). The PCR program included 35 cycles of three steps; 30s at 94 °C for denaturing the DNA; 45s at 48 °C for annealing primers; and 45s at 72 °C for extension of the primers. This was preceded by five minutes at 94 °C for complete separation of duplex template DNA and followed by ten minutes at 72 °C for complete extension of amplicons. The amplicons were cut using the restriction enzyme Hae III. The digested products were run on 3% agarose gel and stained with ethidium bromide for visualisation. The CHD-Z and CHD-W can both be amplified using these primers. Digestion with Hae III generates distinctive banding patterns that differentiate the sexes.

Statistical analysis

We calculated multiple analyses of variance (MANOVA) to obtain multivariate descriptive statistics. We used Pillai's criterion to assess statistical differences between groups because sample size varied between groups (Hair et al., 2010). We tested for collinearity among study biometric traits using correlation analyses. We found pairs of redundant variables in the correlation matrix (wing - P2: r = 0.88; culmen - nalospi: r = 0.86; culmen - shield: r = 0.87). Hence, we dropped one variable from each pair of correlated variables (specifically, we dropped ‘wing’, ‘nalospi’ and ‘shield’) before performing the MANOVA analyses. We also excluded ‘tarsus-toe’ because this variable includes ‘tarsus’. Multivariate homoscedasticity was tested using the Kullback test (Kullback test X² ₆₃ = 36.697, P = 0.997) (Legendre and Legendre, 1998). We carried out univariate comparisons (two-way ANOVA) to test for sexual and age differences in each biometric trait, including the variables ‘sex’, ‘age’ and the ‘sex x age’ interaction as explanatory variables. Body mass was cube-root transformed to achieve a normal distribution. We tested for normality and heteroscedasticity by examining the quartile-quartile plots and the plots of residual versus fitted values, respectively (Crawley, 2007).

We used linear discriminant function analyses (LDFA) to sex birds from biometrics. Discriminant analyses were conducted using all birds (irrespective of age in order to sex individuals of unknown age) or using each age class separately (adults and immatures). In a first LDFA, we used the same six variables as those used in the MANOVA analysis. In a second one, we included only two of these variables (‘mass’ and ‘culmen’), which showed the stronger sexual dimorphism (see results). Finally, in a third LDFA, we used the two variables previously recommended by Craig et al. (1980) to sex P.p. melanotus (‘nalospi’ and ‘bill depth’; note that our measure of bill depth was slightly different). Following Harrell (2001), we did not use a stepwise procedure to select the best model. Instead, we randomly divided the data into two sub-samples (Hair et al., 2010): the first to develop the discriminant function (70% of the sample of each sex); and the second to test the discriminant function (holdout sample, 30% of the sample of each sex). The use of the holdout method is considered to produce unbiased estimates of error rates (Chernick and LaBudde, 2011). For each model prior probabilities for each sex were set as equal.

In order to assess if ageing from plumage traits was reliable until the second complete summer moult, we analysed plumage score variation over time (‘month’, included as independent variable) using generalised linear models (GLMs) with a Poisson distribution and a log-link function.

We used the R 3.0.2 software (R Development Core Team, 2014) for all analyses (asbio, DiscriMiner, ICC, plyr and Rmisc packages).

Age and plumage characteristics
Ageing from body plumage scores

All of the 425 birds whose age was determined from their body pictures were assigned the same age as during visual inspection in the field (298 immatures and 127 adults overall). From September to November, all birds classified as immature showed breasts and thighs with grey or whitish feathers, and underwing-coverts with whitish tips (fig. 3). Thereafter, breast and thigh feathers were moulted gradually (partial winter moult) and total body plumage score decreased with date (score = 1.228 - 0.068 × month; GLM z = -3.552, P < 0.001; fig. 4a). In June, few immature birds were captured (n = 4); they showed adult like breast and thigh plumages, but had underwing-coverts with whitish tips (fig. 3). From September to June, we observed a high proportion of individuals with underwing-coverts with whitish tips (>90% at each month; fig. 3).

Fig. 3.

Monthly percentage of immatures showing typical immature-like plumage on breasts, thighs and underwing-coverts (UWC). Numbers above columns represent the total number of immatures sampled in each month.

[Porcentaje mensual de inmaduros mostrando el típico plumaje de inmaduro en su pecho, muslos y coberteras inferiores (IC). Los números bajo las columnas representan el número total de inmaduros muestreados en cada mes.]

Among the 298 immatures photgraphed, 21 were recaptured 13.4 ± 6.5 months later (range; 6–32). In all cases, they showed adult body plumage if at least one complete summer moult occurred between captures (table SI).

Ageing from underwing-covert scores

Of 232 birds with photographed under - wing coverts, 82 were classified as adults and 149 as immatures, and the age assignment from photos coincided with the assignment made during visual inspection in the field in all cases. Only one adult bird, captured as an immature and recaptured two years later, showed a few grey median underwing-coverts with whitish tips in the left wing only (fig. SI). In all other cases, adult birds captured in September-July showed brown to grey greater underwing-coverts and blue underwing feathers without whitish tips.

Fig. 4.

Temporal trends in (a) immature body plumage score (n = 294) and (b) underwing-covert score (n = 149). Lines show the fitted trend and 95% confidence interval. Dot sizes are proportional to the number of birds observed.

[Tendencia temporal en la puntuación (a) del plumaje del cuerpo (n = 294) y (b) de las infracoberteras (n = 149) en los inmaduros. La línea muestra la tendencia temporal obtenida en cada modelo y su intervalo de confianza al 95%. El tamaño de los puntos es proporcional al número de aves observadas.]

From September to June, all immature birds showed underwing-coverts with whitish tips (fig. 5). The median underwing-coverts showed a greater percentage of feathers with a whitish tip in all months (maximum score of 3; fig. 5). Hence, the median underwing-covert score seems to be the best for ageing purposes. The presence of underwing-coverts with whitish tips was a reliable character from September to June and the total underwing-covert score did not decrease throughout the year (GLM z = -1.391, P = 0.164; fig. 4b).

Biometry

Adults were significantly larger than immatures (Pillai's criterion = 0.635, MANOVA F_{6 394} = 113.99, P < 0.001) and males larger than females (Pillai's criterion = 0.494, MANOVA F_{6 394} = 64.08, P < 0.001). The interaction between age and sex was also significant (Pillai's criterion = 0.041, MANOVA F₆ 394 = 2.81, P = 0.011). Given the significance of the overall test, we further investigated differences trait by trait using univariate analyses (table 1). For all traits, males were bigger than females, and adults bigger than immatures. For most variables the size ranking was as follows: adult males > immature males > adult females > immature females. However, adult females showed a bigger culmen and shield than immature males. The degree of sexual dimorphism also differed between study traits: mass was the most dimorphic variable, while P2 and wing length were the least dimorphic (table 2).

For all morphometric characters, 95% confidence intervals did not overlap between immature females and adult males, but some overlap was found between adult females and immature males for some characters (table 1). However, since ranges overlapped and individuals could not be ascribed to each sex and age group with a high degree of confidence using a single measurement, we explored the usefulness of classification functions. The discriminant analyses (LDFA) using all biometric traits (except redundant ones) correctly assigned sex in 94.9% of adults and in 88.5% of immatures (table 3). When pooling adults and immature birds the percentage of correct sexing was slightly reduced to 87.6% (table 3). If the discriminant function did not include the variable ‘mass’, which can show important short-term variations, the percentage of correct sexing was slightly reduced in all age categories (table 3). We obtained similar results using only the two most sexually dimorphic variables (same subset as for MANOVA analysis; table 2), which allowed predicting the sex of individuals of known age with an accuracy of 80–92% (table 3). Finally, the LDFA using the two variables recommended by Craig et al. 1980) to sex P. p. melanotus also gave similar results, with a prediction accuracy of 83–90% (table 3).

Fig. 5.

Monthly percentage of birds displaying lesser, median and greater underwing-coverts scores of 0-3. (O) no feather with whitish tips, (1) fewer than five feathers with whitish tips, (2) up to 50% of feathers with whitish tips, and (3) over 50% of feathers with whitish tips. Numbers above columns represent the total number of immatures sampled in each month.

[Porcentaje mensual de aves inmaduras que presentan infracoberteras pequeñas, medianas y grandes con puntuaciones entre 0-3. (0) todas las plumas sin bordes blanquecinos, (1) menos de cinco plumas con bordes blanquecinos, (2) hasta un 50% de las plumas con bordes blanquecinos, y (3) más de un 50% de las plumas con bordes blanquecinos. Los números encima las columnas indican el número total de inmaduros muestreados cada mes.]

Age determination

Body plumage differs between immatures and adults before the first partial winter moult is completed. After this first partial winter moult, immature and adult plumages differ only if some immature feathers are retained (Cramp and Simmons, 1980). Here, we found that body plumage characteristics were reliable for ageing all immatures until December (i.e. first-year birds; fig. 3). We also found that the presence of whitish tips on underwing-coverts was a reliable character to age all immature purple swamphens before the first summer complete moult (i.e. until June for second-year birds). This new criterion allows the ageing period to be extended by another six months. After their summer complete moult, second-year birds could not be distinguished from adults by plumage characteristics.

Sexual dimorphism

Few studies have reported morphometric data for the subspecies P. p. porphyrio. Information is available from a morphometric study carried out on museum specimens collected in Doñana (southern Spain; Hiraldo et al., 1974), and some descriptions published in bird handbooks, based on a very small sample sizes (e.g. Glutz von Blotzheim et al., 1973; Cramp and Simmons, 1980). As most of our morphometric measures were taken following Craig et al. 1980), they cannot be directly compared to those reported by Hiraldo et al. (1974). For the only two directly comparable measurements, wing length and shield width, we found similar values (t-test, all P > 0.05; pooling adults and immature birds, since age was not indicated by Hiraldo et al., 1974). We thus have no evidence that birds from the Ebro delta differ in biometry from those from southern Spain. This is consistent with the idea that the recent purple swamphen range expansion in Spain was due to a natural recolonisation process, probably helped by conservation actions (legal protection of both habitat and species and reintroduction projects) during the last 20 years of the 20th century (Sánchez-Lafuente et al., 1992, 2001).

We found significant sexual and age size dimorphism in the purple swamphen for all morphometric measurements taken in this study. Hiraldo et al. (1974) also found significant sexual size dimorphism for four of nine biometric traits studied. However, when we reanalysed their published data (table 2 in Hiraldo et al., 1974), we found evidence that the nine study characters differed significantly between sexes (t-test, all P < 0.01; table S2). In our study, we found that, for all characters, males were bigger than females and, for a given sex, adults were bigger than juveniles. However, this relationship was not met for culmen and shield width, which were bigger for adult females than immature males. These two characters are indicative of shield surface area, and the size of this ornament probably reflects reproductive and/or social status in P. p. melanotus (Craig et al., 1980; Dey et al., 2012, 2014). Shield size is testosterone-dependent in several species of the family Rallidae (Gullion, 1951; Eens et al., 2000) and may function as an indicator of individual quality, as found for many testosterone-dependent traits in birds (e.g. Mougeot et al., 2009). In the purple swamphen, shield size (determined by culmen length and shield width) is likely to be sexually determined and can be modified according to social status or season (Craig et al., 1980; Dey et al., 2014). The observed age differences in size suggest that the immatures can still grow slightly until they reach the adult size, as evidenced by some of the study characters in the sample of birds captured as immatures and recaptured as adults (see table S3).

TABLE 1

Morphometric measurements of the purple swamphen from the Ebro Delta (NE Spain) according to age and sex (mean ± standard deviation, 95% confidence interval and range). Sample sizes = number of individuals with all traits measured. The results of two-way ANOvA indicate significant differences between sexes and age groups (df = 1,399 for all tests; mass was cube-root transformed for analysis). ^* variables included in the MANOvA and the linear discriminant function analyses.

[Medidas morfométricas del calamón común en el delta del Ebro (noreste de España) de acuerdo con la edad y el sexo (media ± desviación estándar, 95% de intervalo de confianza y rango). Tamaños muestrales = número de individuos con todas las características medidas. Los resultados de dos tipos de ANoVA muestran diferencias significativas entre sexos y edades grupales (df = 1.399 para todos los tests; la masa se transformó araíz cúbica para los análisis). ^* Variables incluidas en el MANoVA y los an´lisis de función lineal discriminante.]

Continued

Table 2

Mean dimorphism index (MDI % = (mean female/mean male) × 100) of purple swamphen biometric traits. ^* Variables included in the MANOVA and the linear discriminant function analyses.

[Índice de Media de Dimorfismo (MDI % = (media hembra/media macho) × 100) de las características biométricas del calamón común. ^* Variables incluidas en el MANOVA y los análisis de función line cd discriminante.]

The mean dimorphism index of each measurement was moderate within each age class, and body mass and culmen or shield width were the most dimorphic characters (table 2). Craig et al. 1980) found that these three measurements did not allow reliable determination of sex in P.p. melanotus, because of seasonal variations and an influence of social status. In contrast, we found here that, despite seasonal and social status variations, these variables could be used to predict sex accurately using linear discriminant functions. Discriminant functions including two variables (body mass and culmen) correctly assigned the sex of 80–95% of purple swamphens. The linear discriminant function proposed by Craig et al. 1980), using bill depth and nalospi, also correctly classified 74%–90% of our birds.

According to Dechaume-Moncharmont et al. (2011) the method that we used to test the accuracy of discriminant functions, splitting the data into training and test sets, produces similar results to the jackknife cross-validation, but with a much larger variance. On the other hand, when sample size is small or moderate (< 200 birds), as in our dataset for adults, high discriminant success can be obtained by chance. However, our sample size for immatures and both ages combined was large enough to obtain reliable classification results (Dechaume-Moncharmont et al., 2011). The lack of apparent size differences, at least in wing length and shield width, between two of the most separate populations of the purple swamphen in the Iberian Peninsula (Ebro Delta and Doñana, about 750 km apart) suggests that it may be possible to apply these discriminant functions to sex birds elsewhere in southern Europe. However, the usefulness of these discriminant functions previously needs to be contrasted with a sample of known-sex birds if morphological differences between populations, or measurement differences between researchers, are detected.

TABLE 3

Results of the discriminant analyses applied for sexing purple swamphens of known age (adult and immature) and of unknown age (All birds = adults plus immatures). An individual is classified as male if constant + Σ(coefficient × variable) > 0 and female if the value is ≤ 0. ^* Discriminant functions with the morphological traits proposed by Craig et al. (1980).

[Resultados del análisis discriminante aplicado para el sexado de calamones de edad conocida (adulto e inmaduro) y de edad desconocida (todas lasaves = adultos más inmaduros). Un individuo se clasifica como macho si la constante + Σ (coefficient × variable) > 0 y hembra si el valor es ≤ 0. ^* Funciones discriminantes con las características morfológicas propuestas por Craig et al. (1980).]