All federal permits for the data collection, and the use of feather material was issued by the Austrian Federal Ministry of Education, Science and Research and the Department for Nature Conservation of the provincial government of Burgenland. The capture and ringing of birds followed standard protocols and birds were released unharmed on site.
Study sites and mist netting
In the course of one year, from February 2017 until February 2018, Reed Buntings (n = 372) were caught at the eastern shore of Lake Neusiedl in Illmitz (mean annual precipitation = approximately 600 mm, mean annual temperature = 10.1 °C, elevation = 117 m a.s.l.). The main field site (reed) was located along a dam (west–east orientation) that leads through an approximately 1 km wide reed bed at the eastern shore of the lake (47.76938° N, 16.75636° E). A second site was located east of the lake in a more open habitat at the shore of an alkaline lake (salt pan), with less extensive reed stands, abundant halophytic Chenopodiaceae species and early successional vegetation with oleaster and other shrubs, close to some wine yards (47.80746° N, 16.79040° E). This second field site was only studied during winter in order to assess whether different populations or morphotypes that potentially mix in this wintering area are locally segregated between microhabitats.
Birds were captured with mist-nets, during winter they were additionally attracted using a playback mix of song records of all relevant subspecies to increase the otherwise low capture rates. Each bird was marked with a metal ring, and the sex and age were determined using published criteria (Jenni and Winkler 1994). Wing length (maximum chord) and the length of P8 were measured with a metal ruler to the nearest 0.5 mm. Bill depth (at nostrils), bill length (to skull) and tarsus were measured with plastic callipers to the nearest 0.1 mm.
Stable isotope analysis
For stable isotope analysis, we collected a sample of each bird’s fifth secondary flight feather. With regard to the species-specific moulting patterns (Jenni and Winkler 1994), the fifth secondary was determined to reflect the stable isotope signature of the local food web of the breeding or natal area of the birds. Adults undergo their post-breeding complete molt between mid of July and end of September. As migration of adults even in their northernmost breeding areas usually start as late as in the second third of September (Glutz von Blotzheim and Bauer 1997), the 5th secondary should be already molted in the vast majority of adult birds before leaving their breeding grounds. Collected feathers were stored separately in clean paper envelopes for later isotope analysis. The samples were then randomly subsampled to maintain enough data when subdividing the dataset into sex, age and seasonal classes. Stable isotope analyses were performed at the stable isotope laboratory of the Leibniz Institute for Zoo and Wildlife Research (IZW, Berlin, Germany). Prior to isotope-ratio mass spectrometry, the samples were cleaned for 24 h in a 2:1 chloroform:methanol solution to remove surface contaminants and oils. Then, the samples were dried in a drying oven for 24 h at 50 °C. For stable carbon isotope analysis, 0.3–0.4 mg of cleaned feather vanes was placed into tin capsules. A Delta V‐Advantage stable isotope ratio mass spectrometer (Thermo Fisher Scientific, Bremen, Germany), connected to a Flash EA 1112 Series (Thermo Fisher Scientific) elemental analyzer in continuous mode, was used to measure stable carbon isotopes. The stable isotope ratios were calculated as per mil (‰) deviations from the ratios of the international standard, Vienna Pee Dee Belemnite (V-PDB) and expressed in the delta (δ) notation. The laboratory standard of known 13C/12C ratio, tyrosine, was used in conjunction with other international standards (NBS 19, NBS 22, USGS 24 and L‐SVEC). Measurement precision was estimated to be ± 2.1‰ (one standard deviation).
For stable hydrogen isotope analysis, 0.27 ± 0.01 mg of cleaned feather vanes was placed into silver capsules and left open (IVA Analysetechnik e.K., Meerbusch, Germany). When loading the 96 well microtitre plate, we added capsules with reference keratin material (sheep, goat and human). Then, the filled microtitre plate was placed open at least for 24 h at 50 °C in a drying oven to speed up equilibration and remove extra moisture. After loading the samples into a Zero Blank autosampler (Costech Analytical Technologies Inc., Cernusco sul Naviglio, Italy) they were flushed with chemically pure helium for 1 h. Chemically pure helium was also used as carrier gas. Subsequently, samples and standards were pyrolysed at 1350 °C in a high‐temperature elemental analyzer (HEKAtech GmbH Analysentechnik, Wegberg, Germany) with a silicon carbide tube filled halfway with glassy carbon chops and a carbon/water trap. H2, N2 and CO were separated at 80 °C (HTO Element Analyzer, HEKAtech GmbH Analysentechnik) and hydrogen stable isotope ratios were measured with a Delta V Advantage (Thermo Fisher Scientific). The stable hydrogen isotope ratios were expressed in the delta notation as per mil (‰) deviations from the international standard V‐SMOW. Keratin standards were previously calibrated to the USGS42 standard (Voigt and Lehnert 2018): δ2H = −111.65‰ SWE–SHE (powdered wool from sheep, Sweden), -61.54‰ ESP–SHE (sheep, Spain), -26.44‰ AFR–GOAT (goat, Tanzania) and −72.9 ± 2.2‰ USGS42 (human hair, Tibet). The analytical precision of repeated measurements of keratin standards equalled ± 1.6‰ (one standard deviation). We used these keratin standards to determine the stable isotope ratios of non-exchangeable hydrogen following Voigt and Lehnert (2018).
For geometric morphometric analysis a photograph of the bill in profile (n = 109) was captured on a millimetre grid (Nikon D7000®, Nikon, Melville, New York, USA). Before analysis, the photographs were edited in Adobe Photoshop CS2. A tps file was built from each image using tpsUtil version 1.74 (Rohlf 2016a). TpsDig2 version 2.30 (Rohlf 2016b) was then used to place seven landmarks (discrete homologous points) and eight semi-landmarks (points on a curve, determined by extrinsic criteria) (Zelditch et al. 2004) on each beak image. The semi-landmarks were placed by reference to a standardized grid superimposed onto each image, as detailed in Foster et al. (2008).
To remove non-shape variation, the landmark coordinates were rotated, translated and scaled through Generalized orthogonal least-squares Procrustes Analysis (GPA) as a ‘standardization’ step implemented in the R package geomorph (Adams et al. 2018). The grand mean was calculated (i.e. the consensus of all specimens) and shape variables were then generated (Bookstein 1997). The centroid size (Csize) was computed as the square root of the sum of the squares of the distances from all landmarks to their centroid (Bookstein 1991).
For each bird, a digital picture (Nikon D7000 camera®, Nikon, Melville, New York, USA) was taken of its fully extended right wing. For each individual, all measurements were performed on the same picture. A millimetre grid was placed in the background to provide a scale. The length of the primaries was digitally measured as described in Vanhooydonck et al. (2009), using the image processing program ImageJ (Schneider et al. 2012). The measured length of the primary feathers 1–8 was then used to calculate wing pointedness (C2-index) and convexity (C3-index) according to Lockwood et al. (1998). Higher values of C2 relate to a relatively more rounded wing, whilst higher values of C3 correspond to a more convex wing (Arizaga et al. 2006).
Statistical analyses were conducted in R 3.5.2 (R Core Team 2018). Unless specified otherwise in the methods, required packages for this analysis included car (Fox and Weisberg 2011), ggpubr (Kassambara 2019), geomorph (Adams et al. 2018), stats (R core Team 2018), IsoriX (Courtiol et al. 2019) and densityClust (Pedersen et al. 2017). Statistical significance was accepted at p < 0.05.
First, the relationship between δ2H and δ13C values of feathers was evaluated with Spearman´s rank correlation. Feather δ2H and δ13C values were then used to perform density-based cluster analysis provided by the R-package densityClust (Pedersen et al. 2017). The assigned cluster membership of all observations (n = 110) was extracted and used as grouping factor in univariate and multivariate analysis.
To estimate the geographic provenance of the sampled feathers (moulting locality/natal origin), the R-package IsoriX (Courtiol et al. 2019) was used. A spatial mixed model was calculated by predicting a δ2H isoscape based on the rainfall δ2Hp values (corrected for altitudinal changes) from the Global Network of Isotopes in Precipitation (GNIP). All available precipitation data from June until October, when Reed Buntings undergo their complete or post-juvenile moult (Jenni and Winkler 2007), were used. To define the relationship between the feather δ2Hf values and rainfall isotope δ2Hp values, a transfer equation (δ2Hf = 1.28, δ2H ISOSCAPE-10.29) based on δ2Hf values of Eurasian Reed Warblers (Acrocephalus scirpaceus) from known location that is reported in Procházka et al. (2013) was applied. To fit the calibration function, IsoriX uses a linear mixed-effects model (LMM) that considers the prediction variance, which varies spatially, as well as the prediction covariances between predicted isotope values across the isoscape (for details see Courtiol et al. (2019). Based on this calibration model, the geographic assignment of Reed Bunting feathers of unknown origin was performed and a probability map of geographic origin was created.
Statistical analysis and visualization of beak shape variation was performed using the R package geomorph (Adams et al. 2018). These data were then used in multivariate analysis to assess beak shape variation. A covariance matrix from GPA-aligned Procrustes coordinates was used to perform a number of statistical tests to appraise beak-shape differences in predefined geographical groups (clusters 1–3). However, GPA does not take account of the relationship between size and shape. Therefore, an allometric regression was performed using Procrustes ANOVA with permutation procedure in order to assess the association between log centroid size and shape variables. Furthermore, Procrustes ANOVAs were calculated to assess the relative amount of shape and size variation attributable to the categorical predictor (cluster), with age and sex as covariates. In these analyses, significance was evaluated with a residual randomization permutation procedure of 1000 iterations. A principal component analysis (PCA) of shape variation was performed to visualize morphological differences among groups in a two-dimensional plot of tangent space for a set of Procrustes shape variables (PC1 and PC2). Deformation grids were additionally computed to display the shape of specimens at the end of the range of PC1 and PC2.
Linear measurements of beak shape were reduced to one variable, which represents an overall bill shape index (bill shapeIND), by dividing bill depth by bill length. Overall body size, estimated as the first principal component of an analysis including the variables tarsus and wing length (PC1BODY SIZE 71.3% explained variance), was included as covariate in linear regression models to test for relationships between body size and linear bill measurements. Season was used as additional grouping variable, with groups being defined according to the Reed Buntings’ phenology at these latitudes, as referred in Csörgő et al. (2009) and Glutz von Blotzheim and Bauer (1997), as spring (16 Feb–30 Apr), summer (1 May–11 Sep), autumn (12 Sep–18 Nov) and winter (19 Nov–16 Feb). Further, age classes were reduced to two groups (first year, adult). Undetermined individuals were not considered in the analysis.
A data subset (n = 66) containing all observations from winter (dataWINTER) was used to test for differences of morphological traits as well as for variation in feather stable isotope ratios of wintering Reed Buntings that were caught at different sampling locations (reed, salt pan).
Parametric (one-way ANOVA and Student’s t test) and non-parametric (Wilcoxon–Mann–Whitney test, Kruskal–Wallis test) univariate tests were applied in order to evaluate the effects of age, sex, habitat and cluster on each of the morphological traits and feather stable isotope values. Whenever there were significant differences between the two sex and age classes, tests were performed for males and females or for adults and first-year individuals separately. As post hoc analysis, Tukeyʼs tests or Dunnʼs tests with Benjamini–Hochberg adjustment (1995) were used.