Introduction

Migratory birds take advantage of seasonally occurring resources for their reproduction and survival. To do so, they often travel considerable distances during the annual cycle and, hence, their migration programs, physiology, and morphology have been evolutionary adapted to the spatiotemporal occurrence of resources. However, rapidly changing global environmental conditions have pushed birds to adapt their migration strategies in response. There is strong evidence for such adaptations during the last decades, including shifting migration phenology (Horton et al. 2020), decreasing migration distance (Visser et al. 2009; Rotics et al. 2017), and propensity (Plummer et al. 2015) as well as shifts in breeding and non-breeding distributions (Hill et al. 1998; Curley et al. 2020). For many species and populations, we still lack basic knowledge on their year-round whereabouts and movement patterns. Recording and documenting this information for different species is of high priority for conservation as well as for understanding and predicting the impacts and consequences of climate change on migratory animals in the Anthropocene (Callaghan et al. 2018). Thus, acquiring a proper yardstick—a measure of the current state of things—is a high priority for future records (Visser and Both 2005).

Traditionally, the direction, destination, and timing of bird migration have been studied using individual marking—mainly with unique metal leg rings (Baillie et al. 2007; du Feu et al. 2016). Recovery records of ringed birds have been systematically collected since the beginning of the twentieth century, and they have provided most of the fundamental understanding of where and when different species and populations migrate (e.g. Bairlein 2001; Bairlein et al. 2014). With advanced analytical tools, ring-recovery data can be used to reconstruct population-specific migration routes (Musitelli et al. 2019) and assess migratory connectivity (Cohen et al. 2018) or spatially explicit survival probabilities (Schirmer et al. 2023). Nevertheless, an unavoidable downside of ringing-based studies is the generally low recovery rate of marked individuals in regions such as sub-Saharan Africa, and hence geographic biases in recovery data (Perdeck 1977; Fiedler et al. 2007). For selected species, this has resulted in virtually no recovery records from their non-breeding areas despite more than 100 years of bird ringing. Modern tracking technologies like light-level geolocators (Bächler et al. 2010; Lisovski et al. 2012), despite their limitations, can help partially fill in these knowledge gaps. In particular, our understanding of individual non-breeding sites and migration timing (e.g. van Wijk et al. 2013; Van Bemmelen et al. 2016; Pedersen et al. 2018), as well as links between different breeding and non-breeding populations (i.e. migratory connectivity; Webster et al. 2002; Bauer et al. 2015; Finch et al. 2017), can be improved using tracking technology. Hence, integrating ringing and tracking data can help to elucidate the unknowns in the movement ecology of hard-to-study species.

From the conservation point of view, understanding migratory connectivity can aid in developing appropriate conservation strategies. For example, if individuals from a single breeding population spread across a large non-breeding area, i.e. low migratory connectivity, conservation efforts on a specific area of the non-breeding grounds could benefit a large proportion of individuals from across a large area of the species breeding range. Conversely, if individuals from a single breeding population have low spread and population-specific non-breeding grounds, i.e. high migratory connectivity, conservation efforts on a specific area of the non-breeding grounds would aid only a small proportion of the species’ global breeding population (Webster et al. 2002; Cresswell 2014).

Common Whitethroat (Curruca communis; hereafter, whitethroat) is a long-distance migrant overwintering in the arid and semi-arid zones of sub-Saharan Africa (BirdLife International and Handbook of the Birds of the World 2019). Despite being one of the most common migratory species within the Afro-Palearctic migration system (Hahn et al. 2009), little information is available on its migration routes and non-breeding sites (but see da Prato and da Prato 1983; Fransson 1995). A recent study by Tapia-Harris et al. (2022) provided the first available tracks for the species from its non-breeding sites in Nigeria to breeding sites in Europe and back, revealing the use of multiple non-breeding sites. Here, we combine ring-recovery data with novel migration route simulation technique (Musitelli et al. 2019) and geolocator tracking to describe the migration patterns of whitethroats between Europe and Africa. We focus on describing (1) broad-scale patterns in general migration directions of various European populations, (2) links between breeding and non-breeding populations, (3) individual migration routes in autumn and spring, and (4) individual timing of migration. We further compare results from our tracking study conducted at two localities on the species’ breeding grounds in Europe with a recent tracking study carried out on the species’ non-breeding grounds in Nigeria (Tapia-Harris et al. 2022).

Methods

Ringing data

Ring-recovery data of whitethroats were provided by the EURING Data Bank on 13th October 2023 (du Feu et al. 2016) and by the SAFRING database (https://safring.birdmap.africa) on 28th March 2022. Additionally, 12 recoveries were extracted from Pearson et al. (2014). Upon our request, the EURING dataset was filtered for recoveries less than 20 km from the ringing site (local recoveries accounts for ca. 65% of all records) and finally contained 2803 recovery records of 2658 previously ringed individuals. The SAFRING database contained 95 recovery records of 87 individuals. Local recoveries accounted for 88% of all records, and only two recoveries had information of long-distance movements exceeding 200 km. To link potential breeding and non-breeding sites, we combined ring-recovery information from the above-mentioned sources and checked for records where at least one of the locations was within the species’ non-breeding range in sub-Saharan Africa.

Further, we used the EURING dataset to summarize and simulate migration routes taken by whitethroats within Europe and North Africa following the procedure outlined by Musitelli et al. (2019). This method allows simulations of partial and full migration pathways by connecting ring-recovery records that join at nearby geographic locations (on a given grid), but originate from different individuals. To this end, we used records with a movement distance of > 50 km between consecutive encounters (thereby excluding local movements) of the individual and analysed spring and autumn recoveries separately. Since we aimed at a general description of the species’ migration patterns in northern Africa and Europe, we considered all ring recoveries irrespective of the ringing and recovery year as long as both encounters were within the defined spring or autumn migration season. For the autumn migration simulation, we used ring-recovery records where both ringing and recovery occurred between June and December. For spring migration, we used records from a period between January and July. We intentionally included the breeding season in both specified periods as ring-recovery records where one of the nodes is located at the breeding grounds can indicate the migration pathways taken by the individual from and to its breeding site. In each season, we used only movements in the expected migration direction, i.e. northward in spring and southward in autumn, setting a minimum movement threshold between consecutive encounters to 0.2° across latitude. The resulting ring-recovery records used for route simulation amounted to 898 (538 of those were within the same year) for autumn migration and 421 (104 within the same year) for spring migration.

Analyses were done across one-degree grid cells and in cells with divergent migration directions (> 30-degree deviance in migration direction from a minimum of three ring recoveries). Eastward or westward movements were assigned randomly during the simulation procedure. Similarly, starting points for the route simulation were set at random. In spring, the simulation was run backward starting from the breeding grounds, as no information is available on the distribution and density of individuals at the non-breeding areas or on spring passage areas in North Africa (see Musitelli et al. 2019 for details).

Geolocator tracking

Whitethroats were tagged with light-level geolocators at two breeding locations in Czechia and Latvia between 2017 and 2020. In both study sites, birds were captured using mist nets. At the Central European site in Czechia (50.30° N, 16.44 °E), 23 and 31 individuals were tagged during the breeding seasons of 2018 and 2019, respectively, using the GDL-2.1 geolocators (Swiss Ornithological Institute). This included 23 males, 12 females, and 19 unsexed individuals.

At the northeastern European site in Latvia (56.24° N, 25.37° E), 20 individuals were tagged with Intigeo-P50Z11-7-DIP geolocators (Migrate Technology Ltd.) in late July–early August 2017, after the core breeding period of the species. The late deployment was unplanned and resulted from a delay in geolocator delivery. As a result, 4 tags were deployed on juvenile birds, 10 on females, 2 on males, and 4 on unsexed individuals. In each of the 2019 and 2020 breeding seasons, an additional 20 individuals—14 males and 6 females in both years—were tagged using GDL-2.1 geolocators (Swiss Ornithological Institute) between May and August. In all cases, geolocators including the harnesses weighed approximately 0.6 g, representing ca. 4% of the body mass of the tagged birds (mean: 14.4 ± 0.9 g).

In Czechia, we retrieved five geolocators from males (three in 2020 and two in 2021), one of which stopped recording before spring migration. In Latvia, we recovered two geolocators—one from a female in 2018 and one from a male in 2021, 2 years after the deployment. The latter geolocator contained information on 1.5 migration cycles, providing repeated autumn tracks and information on non-breeding sites in  two consecutive years.

To obtain location estimates of whitethroats across their annual cycle, light intensity data collected by the geolocators were processed following the guidelines set by Lisovski et al. (2020). We first defined twilights using the R-package TwGeos (Lisovski et al. 2016) with a light-level threshold of one (log-scale), deleting twilights within 14 days of the equinoxes and false twilights created by shading. We obtained reference solar zenith angles from either in-habitat calibration based on light readings at the breeding location or Hill–Ekstrom calibration using light readings at the non-breeding site, depending on the quality and noise of light intensity data at each site. Stark changes in consecutive times of sunrise, sunset, noon, and midnight were used to distinguish movement and stationary periods (defined as stops greater than 2 days).

Next, following the Group model in the R package SGAT, we “grouped” similar twilights into single locations (stationary sites). Based on the twilight error distribution defined during calibration, flight speed distribution (gamma distributed; shape = 2.2, rate = 0.06), and a land mask, which limits stopovers to locations on land, SGAT estimates the most probable locations of the bird using a Bayesian model. We initiated the model by first running a “modifiedGamma” model with relaxed assumptions for 1000 iterations before tuning the model with final assumptions/priors for five runs with 300 iterations. Finally, we ran the model with 2000 iterations, producing the most likely tracks (median location estimates) and their associated 95% probability distributions.

Further, we identified Sahara crossing flights and their timing by manually inspecting daily light patterns and looking for abnormally long periods of the full (or increased) light pattern (see Adamík et al. 2016). We defined migration speed as the rate of movement over a complete migration period, including stopovers. Data analyses were done in R v.4.1.3 (R Core Team 2022).

Results

Ringing data

There were 43 ring recoveries that linked non-breeding grounds in sub-Saharan Africa with passage and breeding areas further north (Fig. 1). These indicated broad-scale longitudinal segregation between ringing and recovery sites, i.e. birds overwintering in West Africa originated from the western part of the breeding range, Central African overwintering sites were occupied by birds from central and northern Europe, and birds found in the eastern part of the non-breeding range originated from the eastern part of the breeding range (correlation between the ringing and recovery longitudes: r(41) = 0.935, p < 0.001).

Fig. 1
figure 1

Ring recoveries of Common Whitethroats (Curruca communis) linking non-breeding areas (light grey background) with breeding grounds (dark grey background) and passage areas in between. Black great circle lines indicate individual connections between ringing and recovery events, and the colour of the dots designates the timing of these events—a monthly legend is given in the colour wheel

Full size image

In autumn, migration route simulation suggested a potential migratory divide between the Western and the Eastern flyways in Central Europe at around 10° E (Fig. 2). In contrast, during spring, simulated routes showed a broad-front migration across Europe and the Mediterranean, with many of the modelled routes crossing the Mediterranean Sea either directly or via the Balearic and Italian islands, and Corsica (i.e. Central Mediterranean flyway) rather than circumventing via the Iberian Peninsula or the Middle East. As a general pattern, the simulated migration routes in spring were found more to the west than in autumn (mean longitude of routes: autumn = 12.2° E ± 13.2 (SD), spring = 8.7° E ± 11.3, t-test: t = − 4.502, df = 998, p < 0.001), revealing a clockwise loop migration pattern in whitethroats.

Fig. 2
figure 2

Simulation of the northern parts of migration routes of Common Whitethroats (Curruca communis) in a autumn and b spring based on ring recoveries. Maps show 500 simulated migration routes starting at random locations and moving southwards in autumn and northwards in spring

Full size image

Geolocator tracking

All seven geolocator-tracked individuals from Czechia and Latvia migrated to non-breeding sites in Central Africa and resided in the broader surroundings of Lake Chad between 6° and 15° E (Fig. 3). While in sub-Saharan Africa, all but one bird with available data utilized two distinct non-breeding sites, moving between them in mid-November and December. On average, these sites were located 320 ± 120 km (SD) from one another. The final and main non-breeding sites of the two Latvian birds were located more southward (N Nigeria) than the five Czech birds (S Niger, N Cameroon, and W Chad; Fig. 3). The average great circle distance between the breeding and main non-breeding residency sites was 3890 ± 120 km for the Czech birds and 5020 ± 40 km for the Latvian birds. For both populations, autumn migration routes traversed the Balkan Peninsula and tracked whitethroats crossed the Mediterranean between 15° and 25° E before arriving at the non-breeding sites. In spring, migration routes were located west of the autumn routes traversing Sardinia, Corsica, Sicily, and the Apennine Peninsula between 5° and 15° E. Thus, the annual migration tracks of whitethroats conformed to a clockwise loop migration pattern.

Fig. 3
figure 3

Geolocator tracks of seven Common Whitethroats (Curruca communis) breeding in Czechia (CZ) and Latvia (LV). Median stationary location estimates (dots and squares) are shown alongside their interquartile ranges (grey bars). Panels LV-2.1 and LV-2.2 show the same individual that was tracked for 1.5 years, thus providing information on two consecutive autumn migrations and non-breeding sites. Lower panel shows zoomed-in area of all non-breeding sites of the tracked individuals (light blue—Czech birds, dark blue—Latvian birds)

Full size image

Autumn migration started in late August and early September (mean ± SD; CZ: 25-Aug ± 4 days; LV: 29-Aug ± 5 days), and birds arrived at their respective non-breeding sites in the second half of September and early October (CZ: 20-Sep ± 9 days; LV: 26-Sep ± 7 days; Fig. 4). Major stopovers were made before crossing ecological barriers—the Mediterranean Sea and the Sahara Desert.

Fig. 4
figure 4

Individual migration schedules of seven Common Whitethroats (Curruca communis) tracked with geolocators. Red—breeding site, orange—autumn migration, blue—non-breeding period with the timing of site change indicated in light gray, green—spring migration, dark grey—Sahara crossing; colour tone designates the different breeding origin of the individuals: darker—Latvia, lighter—Czechia. IDs LV-2 and LV-2.1 refer to the same individual which carried the logger for two consecutive years. Incomplete lines indicate logger stoppage

Full size image

In spring, two migration strategies emerged: (1) the two Latvian and one out of four Czech birds initially moved west and made a ca. 4–5 weeks long stopover in the Sahel before crossing the Sahara Desert, while (2) the other three Czech birds with available spring tracks did not make prolonged stopovers before Sahara crossing. Further, most tracked birds stopped over after the Sahara crossing before they arrived at their respective breeding sites.

The three birds which had lengthy spring stopovers in the Sahel region spent on average 184 ± 8 days (SD) at their non-breeding residency sites, while the three birds without the prolonged stopovers spent 212 ± 6 days (SD) at their non-breeding sites. This 4-week difference mainly resulted from the earlier spring departure of the former group (mean ± SD: 29-Mar ± 2 days vs 16-Apr ± 7 days), which also had considerably longer spring migration compared to birds that did not make a prolonged stopover in the Sahel (47 ± 2 days vs 18 ± 2 days). The only Czech bird making a stopover in the Sahel arrived at the breeding site at a similar time to the two Latvian birds, but ca. 2 weeks later than the mean arrival of the three other Czech birds (17-May vs 4-May ± 6 days).

Migration speed in autumn varied considerably between individuals, averaging at 164 ± 34 km/day (range 99–207 km/day). Migration speed in spring was more than twofold faster for individuals which did not make the prolonged stopover in Sahel (225 ± 15 km/day) as compared to the individuals that did (99 ± 19 km/day).

Across all tracked individuals, the mean duration of each of the four main annual cycle parts was as follows: autumn migration = 27 ± 6 days (7% of the annual cycle), non-breeding residency = 198 ± 16 days (54%), spring migration = 32 ± 16 days (9%), and breeding site residency = 108 ± 5 days (30%; Fig. 4).

The single Latvian bird with a repeat autumn track showed similar migration patterns in both years with a 2–3 weeks long stopover at the Balkan Peninsula and a non-breeding site in the same region in north-eastern Nigeria (Fig. 3). The distance between the mean estimates of the two non-breeding sites was 220 km, but 95% CIs of the site estimates were overlapping and, thus, it cannot be statistically excluded that the bird visited the same non-breeding site in two consecutive years. The start of autumn migration in the two years differed by only four days, while the timing of the Sahara crossing and arrival at the non-breeding residency site(s) differed by 12 and 15 days, respectively (Fig. 4).

Discussion

Our results suggest relatively strong longitudinal separation in the global population of whitethroats and provide evidence for the existence of clockwise loop migration, at least for Central, Northern, and Eastern European populations. Moreover, migration route simulations based on ring recoveries suggest a migratory divide in Central Europe in autumn—a pattern characteristic for many species in this region (Cepák et al. 2008). These results are further supported by our geolocator tracking, as autumn migration routes of individuals from both tracked populations converged over the Balkan Peninsula. Our results of geolocator tracking largely coincide with tracking data gathered on the species’ non-breeding grounds in Nigeria (Tapia-Harris et al. 2022), with both studies linking breeding sites in Central and Eastern Europe with non-breeding sites in Lake Chad area.

Ring recoveries and geolocator tracks alike suggest more westerly migration routes in spring compared to autumn. These findings are also supported by observational data in, for example, Morocco, where whitethroats are widespread in spring, but uncommon in autumn (Thévenot et al. 2003). Such loop migration has typically emerged as an adaptation to seasonally uneven distribution of resources, e.g. food availability at stopovers and atmospheric conditions during flight bouts (La Sorte and Fink 2017; Vansteelant et al. 2017). A contributing factor to the observed pattern might be the later onset of spring in Southeast Europe and the Balkan Peninsula than in southwest Europe, central Mediterranean islands, and the Apennine Peninsula (Briedis et al. 2024) where food availability peaks earlier in the season, enabling feeding opportunities for migrants. A modelling attempt by Kranstauber et al. (2015) also found that optimal routes in respect to wind regime were located more westward in spring compared to autumn for migratory birds travelling between Europe and sub-Saharan Africa. Our results on the migration pattern of whitethroats are in line with these findings, suggesting that birds following the clockwise loop migration pattern might benefit from favourable atmospheric conditions in both seasons. However, atmospheric conditions and the distribution of favourable winds are predicted to change considerably in the future due to climate change (Arias et al. 2021). This might pose risks to the long-established migratory programmes of whitethroats and other species following similar migration routes.

One of the main conclusions in an earlier study by Tapia-Harris et al. (2022) was that whitethroats exhibit low migratory connectivity. Birds were tracked from a single location on the non-breeding ground in Nigeria and migrated to breeding sites spread between Central Europe and western Russia, between 48°–58° N and 18°–33° E, suggesting population mixing in sub-Saharan Africa. However, the geographic scale, the number of sampled populations, and distances between them are fundamental aspects when evaluating migratory connectivity (Cohen et al. 2018). Due to the differences in data availability and its spatial coverage, ring recoveries on the non-breeding grounds (Fig. 1) and geolocator tracks (Fig. 3) in our study suggest contrasting patterns of migratory connectivity when scrutinized separately. The pronounced longitudinal segregation in ring-recovery records between breeding and non-breeding regions implies population-specific non-breeding areas and, thus, strong migratory connectivity (Webster et al. 2002). On the contrary, longitude of the non-breeding sites largely overlap between geolocator-tracked individuals from Latvia and Czechia, implying population mixing on the non-breeding grounds and, thus, low migratory connectivity (Finch et al. 2017) for populations breeding in Central and Northeastern Europe. Tracking studies within the Afro-Palearctic migratory network are often biased towards (western) Europe (see Briedis et al. 2020), while for many long-distance migrants, breeding distributions often stretch further east to the Ural Mountains and beyond, where a large proportion of the global population breeds (Keller et al. 2020). As a result, only a subpart of the global breeding range is often sampled (tracked) when evaluating migratory connectivity (e.g. see Adamík et al. 2023). Our example of comparing ring-recovery records with tracking data illustrates the risk for misinterpretation when assessing migratory connectivity if only a part of the global distribution is sampled. This may be particularly important when quantifying species-specific strength of migratory connectivity (e.g. in a comparative framework; Finch et al. 2017) as such quantitative estimates are highly sample sensitive.

Interestingly, most tracked birds visited two non-breeding sites, moving between them in November–December. These results match the findings of Tapia-Harris et al. (2022), who tracked six whitethroats from their non-breeding sites in central Nigeria and found that birds utilized two non-breeding sites. First, a more northernly located site in the Sahel region and a second, more southern site in central Nigeria, moving between them in November–early December. Thus, non-breeding site itinerancy seems to be a common overwintering strategy, for whitethroats migrating to central Africa, but also for another congeneric species, the Barred Warbler migrating to Eastern Africa (Wong et al. 2024). Overall, there is an accumulating number of tracking studies that have described similar itinerancy patterns in a suite of species that migrate to the Sahel region (Briedis et al. 2016a; Thorup et al. 2017; Koleček et al. 2018; Wong et al. 2022). This phenomenon was already described by Moreau (1972), and the decrease in resource availability in the Sahel region following the rainy season is thought to be the main driver (Zwarts et al. 2009).

While ring-recovery data provide only fragmentary information on individual migration timing, geolocator tracking can deliver full annual schedules of the tracked birds. This includes timing of critical events—departure and arrival at the breeding and main non-breeding sites—and derived information like migration duration and speed (Briedis et al. 2019, 2020). As expected from geographic patterns in long-distance migration timing (Conklin et al. 2010; Briedis et al. 2016b, 2020), more northernly breeding Latvian birds lagged behind the Czech whitethroats at all key migration events. The peculiar differences in spring migration tactics, with some birds making a prolonged stopover before Sahara crossing and others not, may be related to deteriorating conditions at the non-breeding sites as migration approaches—the so-called Moreau’s paradox (Schlaich et al. 2016). As the dry season in the Sahel progresses in March and April, long-distance migrants breeding in Europe need to prepare and fuel for the cross-desert migration to their breeding sites (Moreau 1972). It may be that the difference in the two strategies found in whitethroats showcases a high degree of phenotypic plasticity in the species regarding spring movement strategies. Individuals residing at non-breeding sites where conditions become increasingly unfavourable may flexibly leave these sites and fuel for migration elsewhere. At the same time, individuals residing at non-breeding sites with plentiful resources may fuel for the first leg of migration—the cross-Sahara flight—directly at their non-breeding sites. As a result, they spend approximately 60% of the annual cycle at a single site in the Sahel, highlighting the importance of the Sahel region for whitethroats (Zwarts et al. 2009; Tapia-Harris et al. 2022).

Every autumn, roughly 80 million whitethroats migrate from the breeding areas in Europe to their non-breeding regions in sub-Saharan Africa (Hahn et al. 2009). Yet, century-long bird ringing efforts have resulted in a mere 43 ring-recovery records from the species’ non-breeding areas. While tracking studies are typically of small sample size and temporal coverage (often one or only a few years), limiting the strength of conclusions that can be drawn from them, supplementing tracking data with other data types, e.g. ring recoveries and stable isotopes (Boulèt et al. 2006; Lisovski et al. 2019), can provide detailed insights into the ecology of little-known and hard-to-study species. Gathering and publishing such basic information on species' natural histories are becoming increasingly rare in modern ornithological science (Callaghan et al. 2018). Yet, such knowledge remains a cornerstone for conservation, management, and future record in the light of global change.