Introduction

Range expansions have the potential to change migration routes in newly established breeding populations (Sutherland 1998). While some populations seem to adjust to the new flyways and distances (e.g. Berthold et al. 1992), others use ancestral migration routes along the paths of expansions (e.g. Bairlein et al. 2012).

The Paddyfield Warbler Acrocephalus agricola, which has a core breeding area in Central Asia, has recently expanded into eastern Europe (Gavrilenko 1954) and currently has established breeding populations as far west as the coast of the Black Sea in Bulgaria (Kennerley and Pearson 2010). The non-breeding range of the species is restricted to the area extending from southern Iran to northern Myanmar (Kennerley and Pearson 2010). However, almost no information is available on the migration strategies of Paddyfield Warblers flying along the Indo-European flyway and how different breeding populations are distributed across the non-breeding area.

The major aim of the study presented here was to collect individual migration tracks using light-level geolocation datalogger to test the hypothesis that the westernmost Paddyfield Warbler breeding population follows the expansion path, as has been suggested by previous orientation experiments (Zehtindjiev et al. 2010). We also sampled feathers from individuals on their Bulgarian breeding sites that were molted during the non-breeding period and subjected these to stable hydrogen (δ2H) isotope analysis with the aim to obtain insights into the variation of non-breeding sites among individuals of this particular breeding population.

Methods

We captured and marked Paddyfield Warblers staying at the Black Sea coast in Bulgaria (Durankulak Lake: 43°41′N, 28°33′E; Shablenska Tuzla Lake: 43°33′N, 28°35′E; Fig. 1) and deployed light-level geolocators using backpack harness on 34 adults during 2014 and 2015. For the stable isotope analysis, we sampled 25 innermost primaries at the study sites in 2015 and another one upon geolocator retrieval in 2016. Details on marking, tagging and its effect on the birds as well as the handling of samples are provided in Electronic Supplementary Material (ESM) 1.

Fig. 1
figure 1

a Timing of the post-breeding migration of a geolocator-tracked Paddyfield Warbler (Acrocephalus agricola). Bars represent the stationary periods defined by the median time of arrival and median time of departure. Lines represent the interquartile range (IQR) of the arrival and departure times, respectively. b Map showing the location of stationary periods during the post-breeding migration (white) and the non-breeding grounds (dark gray). The triangle indicates the breeding ground of the tracked bird (light gray). The lines within each circle denoting the location of stationary periods coincide with the IQR of the latitude and longitude of the site. The line connecting the breeding ground with the stationary locations is the median position during the non-stationary periods together with the IQR (light-gray area). The hatched area shows the breeding range and the cross-hatched area shows the non-breeding range of the species (BirdLife International 2017) (Color figure online)

Full size image

We used the R package FLightR to estimate geographic locations based on the light recordings (Rakhimberdiev et al. 2017). The method uses a template fit to estimate a spatial likelihood surface for each twilight period and applies a particle filter to sample from these likelihoods and to derive a posterior distribution describing the likeliest path along with credible intervals. Detailed information on the twilight calculation and calibration is given in ESM 1.

For the stable isotope analysis, we prepared a subsample (0.272 ± 0.007 mg [mean ± standard deviation]) of equilibrated and cleaned feather samples, and after combustion we calculated the δ2H ratios between the samples and the keratin standard [for more details see Popa-Lisseanu et al. (2012)]. The internal laboratory keratin standards were scaled to those established in Saskatoon in order to use the calibration equation from Procházka et al. (2013). More details are provided in ESM 1.

To estimate the origin of the feathers (moult location), we used the R package IsoriX version 0.4–1 (Courtiol et al. 2016). We calculated a spatial mixed model predicting a δ2H isoscape based on the measurements of the rainfall δ2H values by the Global Network of Isotopes in Precipitation (GNIP) corrected for altitudinal changes using the function ‘Isofit’ and ‘Isoscape’, respectively. We used all available precipitation data from August to October, when Paddyfield Warblers undergo a complete molt (Kennerley and Pearson 2010), and applied the transfer equation (δ2Hfeather = 1.28 δ2Hisoscape − 10.29) between the feather sample δ2H values and rainfall isoscape δ2H values of the closely related Eurasian Reed Warbler Acrocephalus scirpaceus from Procházka et al. (2013) using the function ‘Calibfit’. Finally, we assigned the feather sample δ2H values using the function ‘Isorix’. This function computes a map of P values from an assignment test based on differences between δ2H values of interest and the predicted isotopic value at each location of the isoscape. As we found extensive variation in the δ2H values in our samples, we applied the clustering function ‘pamk’ from the R package fpc (Hennig 2015) to distinguish distinct groups of δ2H values which were then used for depicting the molting origins. For additional details see ESM 1. All analyses were conducted in R 3.3.1® Core Team 2016).

Results

We recaptured one geolocator-tagged male in 2016. The bird left the breeding ground in the north-eastern direction on 15 August (median; interquartile range [IQR]: 15–16 August). The median post-breeding migration lasted 36 (IQR: 34–37) days during which the bird made five stopovers, varying in length from 1 to 11 days. The first three stopover sites were northeast of the breeding ground, while the last two stopover sites and the non-breeding site (mean arrival 20 September; IQR: 19–21 September) were to the southeast (Fig. 1). The total great circle distance between the breeding ground, the median estimates of the stopover sites and the non-breeding site was 5166 km, whereas the direct great circle distance between the breeding and the estimated non-breeding site was only 4763 km; the difference indicates a detour of approximately 7.8%. We found a large δ2H variation among the 25 feather samples (mean: − 71.4 ± 18.7 ‰; n = 25) and identified seven clusters of feather δ2H values covering the entire non-breeding range of the species (Table 1; Fig. S1 in ESM 2). Stable isotope assignment of the feather molted during the light level recording was in agreement with the geolocator estimate of the non-breeding site (Fig. S2 in ESM 2).

Table 1 Mean δ2H values for the seven isotopic clusters and the number of samples belonging to each cluster
Full size table

Discussion

Our results support the previously suggested migration direction along the pathway of the recent range expansion in the Paddyfield Warbler (Zehtindjiev et al. 2010). Migration routes can reflect historical changes in the distribution of avian species (Bairlein et al. 2012). Our result is interesting as a number of other species show remarkable plasticity in migration directions or migration routes that have lead to the establishment of new non-breeding sites (Berthold et al. 1992). These inconsistent patterns suggest that avian species have different abilities to modify their migration routes in response to range expansions or environmental changes and that the propensity to alter migration behavior will largely depend on the ecology, genetic control of migration and evolutionary history of each species (Sutherland 1998). The time elapsed since the expansion may play an additional role. The degree of migratory connectivity can help to better understand ecology and demographic changes within a species. We have shown a low degree of connectivity in the Paddyfield Warbler, a pattern that seems to be frequent in long-distance migrants (Finch et al. 2017).

We are aware that our spatial assignment based on δ2H values may be affected by using multiple-year precipitation data for calculating the δ2H isoscape. However, the variation in δ2H values within our samples is expected to be independent of yearly variations in the underlying δ2H isoscape (see also Vander Zanden et al. 2014), and the agreement between the geolocator estimate and the stable isotope assignment provides additional support for our main conclusions (Fig. S2 in ESM 2). The question of whether our tracking data from a single individual represent the migration route and phenology of all Paddyfield Warblers from the Bulgarian population is clearly open to discussion. However, previous orientation experiments found similar results (Zehtindjiev et al. 2010; Ilieva et al. in review), suggesting that at least part of this population follows the same migration route.

Our findings on migration direction and notably the retracing of the recent range expansion as well as the weak migratory connectivity provides novel insights into the Paddyfield Warbler. However, many questions remain unanswered, and we hope that this study will stimulate future research on the migration and non-breeding distribution across populations and species that use the rarely studied Indo-European flyway.