Behavioral Ecology and Sociobiology (2021) 75: 152 https://doi.org/10.1007/s00265-021-03082-5 ORIGINAL ARTICLE Year‑round spatial distribution and migration phenology of a rapidly declining trans‑Saharan migrant—evidence of winter movements and breeding site fidelity in European turtle doves Yvonne R. Schumm1 · Benjamin Metzger2 · Eric Neuling3 · Martin Austad4 · Nicholas Galea4 · Nicholas Barbara4 · Petra Quillfeldt1 Received: 24 June 2021 / Revised: 6 September 2021 / Accepted: 9 September 2021 / Published online: 13 October 2021 © The Author(s) 2021 Abstract Populations of migratory bird species have suffered a sustained and severe decline for several decades. Contrary to non- migratory species, understanding the causal mechanisms proves difficult (for migratory bird species) as underlying processes may operate across broad geographic ranges and stages of the annual cycle. Therefore, the identification of migration routes, wintering grounds, and stopover sites is crucial for the development of relevant conservation strategies for declining migrant bird species. We still lack fundamental data of the non-breeding movements for many migratory species, such as European turtle doves Streptopelia turtur, a trans-Saharan migrant. For this species, knowledge of non-breeding movements is mainly based on ringing data that are limited by a low recovery rate in Africa, and tracking studies with a strong bias towards indi- viduals breeding in France. We used Argos satellite transmitters to obtain detailed year-round tracks and provide new insights on migration strategies and winter quarters, of turtle doves breeding in Central and Eastern Europe. The tracking data along with analysis of land cover data confirm previously assumed use of multiple wintering sites and the use of a wide range of forest and agricultural landscapes at the breeding grounds. Tracking data in combination with environmental parameters demonstrated that most environmental parameters and niche breadth differed between breeding and wintering grounds. “Niche tracking” was only observed regarding night-time temperatures. Furthermore, we provide evidence for breeding site fidelity of adult individuals and for home range size to increase with an increasing proportion of agricultural used areas. Significance statement The European turtle dove, a Palearctic-African migrant species, is one of the fastest declining birds in Europe. The rapid decline is presumed to be caused mainly by habitat modification and agricultural changes. Here, we represent data on migration strategies, flyways, and behavior on European breeding and African non-breeding sites of turtle doves breeding in Central and Eastern Europe equipped with satellite transmitters. Our results confirm the use of different migration fly- ways and reveal an indication for “niche switching” behavior in terms of environmental factors during the different annual phases. The migratory behaviors revealed by the tracking approach, e.g., prolonged stopovers during autumn migration in Europe overlapping with time of hunting activities, stopovers in North Africa during spring migration, or evidence for loop migration, are important protection-relevant findings, particularly for the Central-Eastern flyway, for which no tracking data has been analyzed prior to our study. Keywords Argos satellite transmitter (PTT) · Migration routes · Satellite telemetry · Streptopelia turtur · Stopover sites · Winter quarters Introduction Communicated by W. Wiltschko Twice every year an estimated number of more than two * Yvonne R. Schumm billion birds, belonging to the Palearctic–African migration Yvonne.R.Schumm@bio.uni-giessen.de system, migrate between Europe and sub-Saharan Africa Extended author information available on the last page of the article Vol.:(012 3456789) 152 Page 2 of 16 Behavioral Ecology and Sociobiology (2021) 75: 152 (Hahn et al. 2009). In general, these migrants travel between As the seasonal movement patterns are not solely influ- their European breeding and sub-Saharan non-breeding enced by the internal state of organisms and biological fac- grounds crossing the Mediterranean Sea and the Sahara tors, but also by external factors, i.e., the environment and desert via several broad-scale migration corridors and fly- underlying context (Nathan et al. 2008), we compared the ways, formed by specific geographical structures and eco- environmental conditions at the European breeding and the logical barriers. Migration within this system strongly fun- sub-Saharan non-breeding region of turtle doves by selecting nels along two major flyways, namely the Western flyway, different environmental variables to describe and character- over the Iberian Peninsula crossing the strait of Gibraltar to ize the individual habitats of tracked birds. The environ- Northwest-Africa, and the Eastern flyway via the Balkan mental factors determining the distribution of migrants may Peninsula and the Middle East (Briedis et al. 2020). A third differ between breeding and non-breeding areas (Ponti et al. migration route via the Apennine peninsula and across the 2020), depending on if migrants move in geographical space strait of Sicily is the Central flyway (Marx et al. 2016). to track their favored environmental conditions to remain in One of the bird species migrating from Europe to the a specific subset of preferred niche space, so-called “niche African Sahel zone is the European turtle dove Streptope- tracking” (van der Graaf et al. 2006; Tingley et al. 2009; lia turtur (hereafter turtle dove). It is the smallest member Gómez et al. 2016). Alternatively, they may change their of the European Columbiformes and the only long-distance environmental niche (“niche switching”) between periods migrant among them. Formerly a widespread and common of the annual cycle. If different aspects of seasonal move- breeding bird species over a large part of the European ments reflect conservatism in ecological characteristics vs. Continent, Western Asia, and Northern Africa (Glutz von seasonal changes, the conserved patterns may greatly inform Blotzheim and Bauer 1987), the turtle dove has faced popu- related issues, such like habitat choice and timing of migra- lation declines over the past decades and is now listed as tion (Nakazawa et al. 2004). “Vulnerable” by the IUCN (BirdLife International 2019). We present findings of a satellite tracking study on tur- In Europe, numbers have decreased by around 80% between tle doves equipped with Argos transmitters during spring 1980 and 2017 (PECBMS 2020). The major reasons for the migration on Malta, located on the Central flyway, and dur- population decline are presumed to be habitat modification ing breeding season in two states of Germany, with hitherto and agricultural intensification at the breeding and wintering unknown assignment to the possible flyways. In addition areas as well as potentially also on stopover sites used during to the description of the different annual phases (breeding, migration (Browne and Aebischer 2004; Eraud et al. 2009; spring and autumn migration, stopover, and wintering) of Fisher et al. 2018). Unsustainable legal and illegal hunting each tracked bird, we analyzed the favored environmental activities along the migration routes are further contributing conditions at the breeding and wintering sites in order to test to the decline (Fisher et al. 2018; Lormée et al. 2019). for niche overlap in the seasonal niches. Analyses of ring recoveries confirm that turtle doves migrate along all three aforementioned flyways. Ringing studies further found evidence for a migratory divide in Material and methods Europe with Western European populations of turtle doves using the Western flyway and Central and Eastern European Bird capture and transmitter deployment populations migrating along the Central or the Eastern fly- way (Dimaki and Alivizatos 2014; Marx et al. 2016). How- From 2016 to 2020, turtle doves were caught during stopover ever, sampling turtle dove populations across Europe did not on their return migration in spring along the Central fly- reveal any genetic structure that would support discerned way on the Maltese islands (n = 8) using mist nets. In 2019 populations according to this migratory divide, but rather and 2020, turtle doves were caught during breeding time one large, panmictic population (Calderón et al. 2016). Fur- in two states of Germany (Central Germany: Hesse (n = 5) thermore, only 1.6% (14 out of 897) ring recoveries came and Eastern Germany: Brandenburg (n = 3)) using drop traps from Sub-Saharan Africa (Marx et al. 2016), indicating that baited with a mix of cereal seeds (Table 1). It was not pos- there is still a lack of knowledge regarding the wintering sible to record data blind, as our study involved individually grounds and the exact flyways of turtle doves. marked animals in the field. Advances in tracking techniques have started to shed The sex was determined by molecular analysis based on more light onto the migration routes of turtle doves (Eraud collected feather or blood samples (Griffiths et al. 1998) et al. 2013; Lormée et al. 2016). However, so far there has and by characteristics of plumage (Demongin 2016). Birds been a strong bias of studies towards turtle doves breeding were individually fitted with an Argos satellite tag (Solar 5 g in France, which use the Western flyway, whereas detailed PTT, Microwave Telemetry, USA), providing location fixes knowledge based on tracking of individuals migrating on the based on Doppler calculations, fixed as a wing-loop back- Central and Eastern flyway remains very limited. pack using a 2-mm-wide Teflon ribbon (Ecotone, Poland) 1 3 Behavioral Ecology and Sociobiology (2021) 75: 152 Page 3 of 16 152 Table 1 Individual information of European turtle doves Streptopelia turtur equipped with Solar 5  g PTT Argos satellite tags during spring migration on Malta and during breeding period in Germany Bird ID Duty cycle Deployment date Catching loca- Sex Body mass [g] Device weight [% of End data Data trans- [ON/OFF] [dd.mm.yyyy] tion [Lat, Long] the birds’ body mass] t ransmissiona [dd. mission mm.yyyy] [days] #161046 10 h/48 h 13.05.2016 Malta NA 137 3.6 05.08.2017 450 35.95, 14.38 #161047 10 h/48 h 21.04.2017 Comino NA 136 3.7 30.04.2017 9 36.01, 14.34 #161048 10 h/48 h 22.04.2017 Comino f 118 4.2 27.04.2018 370 36.01, 14.34 #161049 08 h/15 h 22.04.2017 Comino m 142 3.5 20.09.2017 151 36.01, 14.34 #161050 08 h/15 h 23.04.2017 Comino m 129 3.9 14.08.2020 1209 36.01, 14.34 #181091 10 h/48 h 13.06.2019 Hesse f 160 3.1 25.09.2020 470 50.49, 08.92 #181090 10 h/48 h 24.06.2019 Brandenburg m 160 3.1 26.09.2019 94 51.92, 14.33 #181092 10 h/48 h 25.06.2019 Brandenburg m 161 3.1 09.05.2020 319 51.92, 14.33 #181089 10 h/48 h 25.06.2019 Brandenburg f 158 3.2 20.10.2019 117 51.92, 14.34 #200345 08 h/15 h 01.05.2020 Comino f 125 4.0 02.05.2020 1 36.01, 14.34 #200348 08 h/15 h 04.05.2020 Comino f 132 3.8 22.06.2020 49 36.01, 14.34 #200349 08 h/15 h 05.05.2020 Comino m 180 2.8 01.09.2020 119 36.01, 14.34 #200351 08 h/15 h 05.06.2020 Hesse m 148 3.4 30.10.2020 147 50.44, 08.55 #200352 08 h/15 h 07.06.2020 Hesse f 155 3.2 07.10.2020 122 50.44,08.55 #200353 08 h/15 h 08.06.2020 Hesse f 149 3.4 16.09.2020 100 50.49, 08.92 #200350 08 h/15 h 13.06.2020 Hesse m 173 2.9 12.10.2020 121 50.44, 08.55 a Transmission of locations stopped without known reasons or transmission manually terminated due to stable positions from the same (unhospi- table) location for consecutive weeks harness, following the method described by Lormée et al. data transmission stopped over the Mediterranean Sea (Sup- (2016). The overall weight of the tracking device was below plementary Fig. 1n). These three turtle doves were excluded the 5% of the birds’ body masses threshold, recommended from all further analyses, resulting in a final data set of 13 in literature (Fair et al. 2010). Satellite tags deployed were tracked turtle doves. programmed with a standard duty cycle of 10 h ON/48 h OFF or with a modified duty cycle of 8 h ON/15 h OFF Handling of tracking data (Table 1). All birds were released immediately after tagging at the location of capture. All location data as received from Argos of all location The transmitters of the individuals #161047 and #200345 classes (LC: 3, 2, 1, 0, A, B) were automatically uploaded stopped recording 9 days and 1 day after tagging, respec- to Movebank (movebank.org) in their original projection tively, while both birds were still on stopover on the Mal- (WGS84). We applied the “Douglas Filter” (Filter Method: tese Islands. These two individuals were probably killed by Best Hybrid, Douglas et al. 2012) to remove erroneous poachers. Individual #200348 sent data for 49 days. How- data and afterwards checked for possible remaining outli- ever, within that time, she did not settle at a breeding site ers manually. These filtered location data were used when but crossed the sea between Sicily and Libya twice until plotting the data in QGIS 2.18 (QGIS.org 2016). Tracks 1 3 152 Page 4 of 16 Behavioral Ecology and Sociobiology (2021) 75: 152 were displayed by using the “Points2One” plugin in QGIS birds was estimated with the R package “sp Classes and (Kapusta 2015). Methods for Spatial Data” (Pebesma 2020). Locations obtained before deployment were used to esti- To characterize the land cover in the occupied habitats the mate accuracy of the locations. The deviation of these loca- 95% KUDs were clipped in QGIS 2.18 with raster land cover tions (n = 63) was on average 2 km, ranging from 0.1 to data and the percentages of different land covers classes in 10 km. the 95% KUDs were calculated. For European breeding To determine the different phases (i.e., breeding, migra- grounds, land cover data were based on Corine Land Cover tion, stopover, and wintering) in the annual cycle of the CLC 2018 v.2020_20u1 (Copernicus Land Monitoring Ser- individuals, we used an approach similar to that described vice 2021) and for African wintering grounds on ESA CCI in Lormée et al. (2016): Clear switches in the pattern of Land Cover S2 prototype LC 20 m map of Africa 2016 (ESA the location data defined the onset and end of the different CCI Land Cover project 2021). phases. Breeding phase was the period when an individual Only complete periods in the life cycle of the birds were spent at least 45 days between April and September in one used in the detailed analyses of durations. To determine distinct area (Glutz von Blotzheim and Bauer 1987; Marx whether migratory movements of the turtle doves occurred et al. 2016). We defined the pre-migratory movements as during the night or day, we calculated for each location fix movements to a distinct site, onwards called post-breeding of complete autumn (n = 7) and spring tracks (n = 5), when site, where the period (minimum of 10 days) after likely the morning and evening civil twilight had been by using the nesting was spent before the onset of migration (Pagen et al. function “crepuscule” (R package maptools, Bivand et al. 2000; Vitz and Rodewald 2007). The onset of molt migra- 2020). With the “crepuscule” function we estimated when tion or autumn migration was specified as soon as move- the geometric center of the sun was 6° below the horizon ments > 100 km in direction to the wintering grounds, e.g., in the morning (civil dawn) and in the evening (civil dusk). southwards, south westerly or south easterly, occurred. Molt Night-time was defined as the period of time between civil migration is the temporal overlap in the molt and migration dusk and the consecutive civil dawn (Zúñiga et al. 2016). life history stages (Tonra and Reudink 2018). Since there Fixes included in this analysis were at least 1 h, but not is no uniform pattern whether molt migration occurred, as more than 6 h, apart from each other and only considered some individuals presumably molt the first inner primaries when individuals were actively migrating (n = 189 pairs of on or near their breeding sites, while others stop during fixes). If the direct distance between two consecutive fixes autumn migration to molt en route (Demongin 2016; Pillar was more than 25 km (mean flight speed during migration et al. 2016), the term “autumn migration” used throughout is approx. 50 km/h; Lormée et al. 2016), active migratory the paper includes possible molt migration (Pillar et al. 2016; movement was presumed (n = 51 pairs of fixes) and clas- Table 2). A stopover site was defined as consecutive set of sified as night-time, daytime, or between (i.e. one of the locations overlapping spatially for at least three days during fixes during night- and the other one during daytime). From migration period and being at least 100 km away from the the fixes during active migratory movement (n = 51 pairs of breeding site. Molt of the first inner primaries at stopover fixes), the mean flight speed was calculated. sites was assumed if the individuals staged after leaving the To compare environmental habitat parameters at breed- breeding site for at least 21 days (Mallet-Rodrigues 2012) ing and wintering grounds, nine habitat parameters were in Europe before October (Demongin 2016). These sites are obtained through the Environmental Data Automated Track referred to as “stopover molt sites” (Pillar et al. 2016). If Annotation System (Env-DATA, Dodge et al. 2013; inter- an individual stayed for at least 14 consecutive days in one polation: bilinear) on Movebank for filtered positions of distinct area south of 20° N (Glutz von Blotzheim and Bauer those turtle doves from which we obtained locations at both 1987; Eraud et al. 2013), this was specified as the start of wintering and breeding grounds (n = 5; Table 2). Locations the wintering period. received during winter movements (i.e., movement between Epanechnikov kernels (95% and 50% kernel utilization different wintering sites), stopovers, and active migration distributions “KUD”; Epanechnikov 1969) of filtered locali- were excluded. The environmental data included parameters zations received during the breeding and wintering period from MODIS land: net photosynthesis (PsnNet), gross pri- were calculated in R with the function the “kernelUD” in the mary productivity (GPP), total evapotranspiration, enhanced package “adehabitatHR” (Calenge 2015). We used a generic vegetation index (EVI), daily land surface temperature day grid of 200 cells and the smoothing parameter was estimated and night, and vegetation index (NDVI) as well as the with a href parameter. The area covered by the individual parameters elevation (ETOPO1) and human population den- sity adjusted (SEDAC GPW V3 and GRUMP V1 GRUMP 1 3 Behavioral Ecology and Sociobiology (2021) 75: 152 Page 5 of 16 152 Table 2 Details of the migration schedule of 13 European turtle doves Streptopelia turtur equipped with satellite tags from 2016 to 2020 Bird ID Spring migration Breeding Autumn migration Wintering Year Duration in days No. stopover No. stopover Duration in days Duration in days No. stopo- No. Duration in days [dd.mm to dd.mm] Africa Europe [dd.mm to [dd.mm to ver Europe stopover Africa [dd.mm to dd.mm] [country:days] [country:days] dd.mm] dd.mm] [country:days] [country:days] #161046 2016 [STa–22.05] 116 [22.05–14.09] 64 [14.09–16.11] 0 4 [LY:12, 156 [16.11–20.04] ML:6, NE:13; 4] 2017 23 [20.04–12.05] 1 [LY:9] 0 [12.05–ETa] #161048 2017 [ST–18.05] 114 [18.05–08.09] 16 [08.09–23.09] 1 [GR:3] 0 213 [23.09–23.04] 2018 [23.04–ET] #161049 2017 [ST–20.06] 2 [IT:18; 16] 63 [20.06–21.08] [21.08–ET] 1 [HU:25c] #161050 2017 [ST–05.05] 117 [05.05– 18 [29.08–15.09] 1 [IT:4] 1 [NE:6] 207 [15.09–09.04] 29.08]b 2018 23 [09.04–01.05] 1 [TN:15] 0 134 [01.05– 7 [11.09–17.09] 0 0 209 [17.09–13.04] 11.09]b 2019 20 [13.04–02.05] 1 [TN:8] 0 131 [02.05– 14 [09.09–22.09] 0 1 [MA:6] 205 [22.09–13.04] 09.09]b 2020 21 [13.04–03.05] 1 [DZ:10] 0 [03.05–ET] #181091 2019 [ST–01.09] 35 [01.09–05.10] 2 [FR:12, ES:6] 0 206 [05.10–27.04] 2020 50 [27.04–15.06] 3 [MR:6, 0 76 [15.06–29.08] [29.08–ET] 2 [FR:5;14] MA:16; 4] #181090 2019 [ST–27.08] [27.08–ET] 1 [HU:13] 1 [LY:6] #181092 2019 [ST–13.09] 22 [13.09–04.10] 1 [HU:3] 2 [LY:6, NE:3] 206 [04.10–26.04] 2020 [26.04–ET] #181089 2019 [ST–05.08] [05.08–ET] 2 [SK:11, 2 [TN:5, DZ:6] HU:12] #200349 2020 [ST–09.05] [09.05–ET] #200351 2020 [ST–05.09] [05.09–ET] 1 [FR:15] #200352 2020 [ST–29.08] [29.08–ET] 2 [FR:7, IT:20] 1 [TN:5] #200353 2020 [ST–01.09] [01.09–ET] 1 [ES:9] #200350 2020 [ST–31.07] [31.07–ET] 2 [GE:24c, 1 [DZ:14] ES:18] a ST, start of data transmission; ET, end of data transmission b Including time spent at the post-breeding site (see Supplementary Fig. 2) c Stopovers at stopover molt sites (see Supplementary Fig. 1f, k) 2000). These parameters were selected as they likely influ- function calculated in R with the function “kde2d” in the ence the habitat selection of turtle doves and had a sufficient MASS package (Venables and Ripley 2002). data coverage for both breeding and wintering sites. Statistical analyses were conducted using R 3.4.1 (R Development Core Team 2018). Means ± SE for the environ- Results mental parameters for each individual are given in Table 3. To compare the aforementioned environmental habitat Transmitters operated for 291 days on average (n = 13 turtle parameters at the different stages (breeding and wintering), doves, Table 1), resulting in 19,482 filtered location fixes a principal component analysis (PCA) was performed for during 18 breeding (hereof seven complete) and seven every single individual as well as for all individuals together. complete wintering periods, 12 migration cycles in spring The PCA extracted two significant components PC1 and (hereof five complete), and 15 in autumn (seven complete) PC2. Habitat niche plots were created from the two dimen- as well as 39 stopovers of which 2 are likely stopover molt sions of the habitat (PC1 and PC2) using kernel densities sites (Fig. 1; Table 2). 1 3 152 Page 6 of 16 Behavioral Ecology and Sociobiology (2021) 75: 152 Breeding period and post‑breeding areas The turtle doves arrived at their breeding grounds between 1. May and 20. June (median 18. May, n = 10) and spent a total of 63–134 days (median 107.3 days, n = 7) at the breeding grounds (Table 2). Breeding grounds of individuals tagged during spring migration on Malta were located in Italy (#161046, #161050, and #200349), Bulgaria (#161048), and Slovakia (#161049; Supplementary Fig. 1). Individuals (n = 8) tagged at the beginning of the breed- ing season in Germany stayed in the area where they were caught during the ongoing breeding season (Supplementary Fig. 1). The home range (95% KUD) used by the turtle doves (n = 18 breeding sites from 13 individuals) was on average 496 ± 335 km2 (min 263 km2 to max 1554 km2) and the core area (50% KUD) was 39 ± 29 k m2 (min 14 km2 to max 121 km2). Land cover in the 95% Epanechnikov kernels varied per breeding area and individual. In total 35 of 44 Corine Land Cover classes occurred in the breeding areas (Sup- plementary Table 1). Land cover classes that were present in every single turtle dove breeding habitat were non-irri- gated arable land (29.4 ± 15.5% of the 95% KUD), broad- leaved forest (20.9 ± 19.5%), discontinuous urban fabric (4.6 ± 2.4%), and industrial or commercial units (1.0 ± 0.3%; Fig. 2). The size of home ranges (95% KUD) increased with a higher percentage of agricultural areas (Supplementary Table 1) as land cover in the 95% kernel, while there was no significant relation between home range size and the pro- portion of forest and seminatural areas (GLM: agricultural areas: F1,10 = 5.47, p = 0.041; forest and seminatural areas: F1,10 = 3.53, p = 0.090). The same applied for the size of core areas (GLM: agricultural areas: F1,10 = 5.02, p = 0.049; forest and seminatural areas: F1,10 = 3.87, p = 0.078). Turtle doves for which we had data from the breeding grounds for consecutive years (n = 3; #161046, #161050, and #181091) showed high site fidelity, i.e., returned to the exact same breeding area (Supplementary Fig. 1a, b, d). The male individual #161050 performed short pre-migra- tory movements (< 1 day) into a defined post-breeding site around 20–30 km north-easterly to its breeding site (Sup- plementary Fig. 2), where it stayed for 39, 18, and 38 days in 2017, 2018, and 2019, respectively, before starting the autumn migration. The other birds, stayed in their breeding areas until they started autumn migration, but individuals #161049 and #200350 moved to stopover molt sites for a prolonged period in more southern latitudes within Europe (Table 2, Supplementary Fig. 1f, k), indicating a stopover molt migration for these individuals. Autumn migration Turtle doves left their breeding areas between 31. July and 14. September (median 30. August, n = 15). The overall 1 3 Table 3 Habitat variables at breeding and wintering grounds of tracked European turtle doves. Habitat variables (mean ± SE) were obtained through the Environmental Data Automated Track Annotation System (Env-DATA) on Movebank linked to Argos location fixes (n = number of filtered Argos locations) #161046 #161048 #161050 #181091 #181092 Breeding Wintering Breeding Wintering Breeding Wintering Breeding Wintering Breeding Wintering (n = 651) (n = 432) (n = 926) (n = 720) (n = 2948) (n = 3759) (n = 523) (n = 611) (n = 297) (n = 549) PsnNet [kg C m -2] 0.02 ± 0.01 < 0.00 0.02 ± 0.01 < 0.00 0.03 ± 0.02 < 0.00 0.03 ± 0.01 < 0.00 0.02 ± 0.01 < 0.00 GPP [kg C m-2] 0.03 ± 0.01 < 0.00 0.03 ± 0.02 < 0.00 0.05 ± 0.02 < 0.00 0.04 ± 0.02 < 0.00 0.03 ± 0.01 < 0.00 Evapotranspiration [kg m -2] 13.2 ± 4.2 0.4 ± 0.4 15.6 ± 9.4 0.6 ± 1.0 28.7 ± 14.6 0.5 ± 0.5 27.9 ± 10.6 1.8 ± 2.4 14.6 ± 4.5 0.5 ± 0.5 EVI 0.31 ± 0.05 0.13 ± 0.02 0.33 ± 0.08 0.15 ± 0.03 0.49 ± 0.13 0.14 ± 0.02 0.44 ± 0.08 0.18 ± 0.03 0.33 ± 0.03 0.11 ± 0.01 Surface temperature day [°C]a 36.5 ± 5.2 34.5 ± 5.9 33.7 ± 4.1 35.7 ± 5.7 31.3 ± 6.1 39.0 ± 4.9 29.2 ± 3.3 35.6 ± 5.5 28.2 ± 5.2 35.0 ± 5.4 Surface temperature night [°C]a 19.7 ± 3.1 18.4 ± 2.5 17.8 ± 3.5 18.2 ± 3.1 17.8 ± 3.9 19.2 ± 4.7 14.8 ± 2.8 17.3 ± 3.7 15.7 ± 4.3 17.6 ± 4.2 NDVI 0.45 ± 0.06 0.18 ± 0.04 0.51 ± 0.10 0.25 ± 0.05 0.68 ± 0.15 0.22 ± 0.04 0.68 ± 0.08 0.29 ± 0.06 0.64 ± 0.05 0.15 ± 0.03 Elevation [m amsl] 136.2 ± 84.3 343.7 ± 10.3 45.2 ± 35.6 297.2 ± 4.3 401.1 ± 156.4 239.7 ± 72.7 189.9 ± 33.5 291.7 ± 118.4 73.4 ± 8.1 307.5 ± 41.7 Population density [person/km²] 237.4 ± 170.2 89.6 ± 34.9 408.1 ± 219.1 27.2 ± 26.1 175.2 ± 140.9 22.3 ± 16.6 229.5 ± 46.8 18.9 ± 14.0 58.0 ± 62.1 35.7 ± 15.0 a Unit changed from Kelvin to °C Behavioral Ecology and Sociobiology (2021) 75: 152 Page 7 of 16 152 Fig. 1 Satellite tracks of 13 European turtle doves Streptopelia tur- shown as circles in the background of the track lines. Winter move- tur during migration between European breeding (red circles) and ments are displayed in dashed, blue lines. Black circles correspond African wintering grounds (blue circles). Tracks are given in differ- to stopover sites during autumn migration and gray circles to stopo- ent colors corresponding to different countries individuals had their ver sites during spring migration. Circle size corresponds to stopover breeding sites in: orange = Germany (dark orange = Hesse, Cen- duration: Small circle ≤ 10 days and big circle > 10 days. Background tral Germany; light orange = Brandenburg, Eastern Germany); dark colors indicate the terrain and gray lines indicate national borders red = Italy; pink = Slovakia: purple = Bulgaria. Detailed tracks for sin- (background map: Stamen terrain (map tiles by Stamen Design: gle individuals can be found in Supplementary Fig. 1. Autumn migra- http:// maps.s tamen. com; data by OpenStreetMap: www. opens treet tion is shown as solid line and spring migration as dashed line. Loca- map.o rg)) tion fixes (based on Doppler locations) received during migration are duration of the autumn migration including stopovers was in Europe after leaving their breeding site (Table 2; Fig. 1). 7–64 days (median 25.1 days, n = 7), with a total stopover Prolonged stopovers were mainly made in Europe (58.8% duration between 0 and 35 days (median 11.9 days, n = 7). of stopovers; 10 of 17 stopovers) and less often in Africa Stopovers were taken in 57.1% of autumn migrations (23.1%; 3 of 13 stopovers). in Europe (median 4 days, n = 7) and in 57.1% in Africa Of the 12 turtle doves four individuals (#181091, (median 8 days, n = 7, Table 2, Supplementary Fig. 3). How- #200351, #200353, and #200350), all with breeding ever, considering the data of all tracked individuals, i.e., grounds in Hesse, started in south-westerly direction. The partial tracks included, several birds (3 of 4 individuals in two females #181091 and #200353 crossed the Mediterra- the Western flyway and 4 of 8 in the Central-Eastern fly- nean Sea at or close to the strait of Gibraltar, while the two way) made prolonged post-breeding stopovers (> 10 days) males #200351 and #200350 crossed the Mediterranean Sea 1 3 152 Page 8 of 16 Behavioral Ecology and Sociobiology (2021) 75: 152 Fig. 2 Proportional occurrence [%] of Corine Land Cover classes (Copernicus Land Monitoring Service 2021) in 95% Epanechnikov kernels of satellite-tracked European turtle doves Streptopelia turtur at different breeding grounds. a Bulgaria (#161048), b Slovakia (#161049), c Southern Italy (#161046 and #161050), d Central Italy (#200349), e Central Germany (#181091, #20035, #200352, #200353, and #200350), f Eastern Germany (#181090, #181092, and #181089). Only land cover classes accounting for a fraction of more than 1% are shown. All remaining classes < 1% have been summed up to “O: others.” Further details to the other classes can be found in Supplementary Table 1 further east and therefore had a longer sea crossing (Supple- a more south-westerly direction. Turtle doves breeding in mentary Fig. 1j, k). The remaining turtle dove with breed- Eastern Europe (#161048 and #161049) started their autumn ing ground in Hesse (#200352) migrated south over Corsica migration in a southern direction over mainland and islands and Sardinia. Individuals breeding in Italy (#161046 and of Greece (Fig. 1, Supplementary Fig. 3). #161050) migrated along the Central flyway. Turtle doves Overall, turtle doves in our study were largely noctur- with breeding grounds in Brandenburg (#181090, #181092, nal migrants. The majority of location fixes (81.8%) dur- and #181089) migrated first in a south-easterly direction to ing active migratory movements were recorded during the stopover sites in Eastern Europe and from there changed in night, compared to 9.1% during daytime and 9.1% between. 1 3 Behavioral Ecology and Sociobiology (2021) 75: 152 Page 9 of 16 152 Fig. 3 Proportional occurrence [%] of ESA CCI Land Cover classes Individuals #161046, #161048, #161050, and #181092 using the (ESA CCI Land Cover project 2021) in 95% Epanechnikov kernels Central-Eastern flyway and wintering in Cameroon, Nigeria, Niger, of satellite-tracked European turtle doves Streptopelia turtur at differ- Mali, Burkina Faso, and Ghana. Only land cover classes accounting ent wintering grounds. a Western part of Western Africa. Individual for a fraction of more than 1% are shown. All remaining classes < 1% #181091 using the Western flyway and wintering in Senegal and have been summed up to “O: others.” Further details to the other South-Western Mali. b Central and Eastern parts of Western Africa. classes can be found in Supplementary Table 2 The mean flight speed during the active migration was resting and roosting sites, mainly characterized land cover 45.7 ± 12.8 km/h. at the wintering sites. Open water was available at all win- We lost most of our tagged individuals during autumn tering areas (Fig. 3, Supplementary Table 2). Remarkably, migration (61.5%, 8 of 13 individuals) compared to the other we observed a much higher proportion of tree cover area annual stages (wintering 0%, spring migration: 15.4% and (41.2%) at the wintering sites of individual #181091 (migrat- breeding: 23.1%). ing along the Western flyway) compared to the other four turtle doves (2.3 ± 1.9% tree cover), which used the Central- Wintering period Eastern flyway and spend the wintertime in the Eastern parts of Western Africa. Five of our satellite-tracked turtle doves arrived at their wintering grounds. The birds spent 156–213 days (median Spring migration 200.3 days, n = 7) wintering after arriving between 15. Sep- tember and 16. November (median 2. October, n = 7). Turtle The turtle doves started their spring migration between doves overwintered in Western and Central Africa south of 9. April and 5. May (median 18. April, n = 7). The dura- the Sahara (Figs. 1 and 3). While one individual (#161046) tion of the spring migration including stopovers was spent the entire wintering period at one wintering site (Sup- 20–50 days (median 27.4 days, n = 5). Stopovers lasted plementary Fig. 1b), the other four individuals used multiple 8–26 days (median 13.6 days, n = 5). African stopover sites (2–6) distinct wintering sites with a southward shift during were located in Mauritania, Morocco (Western flyway) the wintering period. The 95% KUD of wintering sites used and Algeria, Libya, and Tunisia (Central-Eastern flyway). by the turtle doves (n = 20 wintering sites) was on average None of the five individuals of which we have a complete 65 ± 154 km2 (min 18 km2 to max 510 km2) and the 50% spring migration track staged in Europe (Table 2, Supple- KUD was 5 ± 21 km2 (min 1 km2 to max 67 k m2). From one mentary Fig. 3). However, #161049 from which we have individual (#161050), we have locations from the wintering a partial track of its spring migration made two stopovers period for consecutive years. While its wintering duration in Italy, which lasted 18 and 16 days, before reaching its is quite consistent (Table 2), the wintering localities varied Slovakian breeding ground (Supplementary Fig. 1f). Indi- between the wintering periods (Supplementary Fig. 1a). vidual #161046 showed overshooting behavior during spring A mix of crop- and grassland, likely used for forag- migration, i.e., first flying further north before returning to ing, and tree and shrub covered areas, presumably used as its breeding ground (Supplementary Fig. 1b). The three 1 3 152 Page 10 of 16 Behavioral Ecology and Sociobiology (2021) 75: 152 Fig. 4 Habitat niches of satellite-tracked European turtle doves Streptopelia turtur (n = 5) for wintering (blue) and breed- ing (red) grounds. Obtained from Argos-positions and kernel densities of principal compo- nent scores of environmental parameters (n = 9) obtained through the Environmental Data Automated Track Annotation System (Env-DATA) on Move- bank. PC1 (eigenvalue 5.34) was determined mainly by the five variables (PsnNet, NDVI, EVI, GPP, and Evapotranspira- tion) and PC2 (eigenvalue 1.37) mainly by the surface tempera- ture at night individuals we have complete spring migration tracks fol- temperature at night and for the individuals #161046, lowing the Western (#181091) and Central flyway (#161046 #161048, and #181092 additionally by the daytime surface and #161050) for autumn migration followed a very similar temperature and for #181091 by elevation (Supplementary route for spring migration (Fig. 1). The partial migration Table 3). tracks of individuals #161048 and #161049 following the When comparing the ecological niches based on the Central flyway for spring migration and a flyway further to tested parameters, turtle doves showed a change in environ- the east during autumn migration, suggesting a clockwise mental conditions represented by PC1 between the winter loop migration (Supplementary Fig. 1c, f). and breeding season, with the 95% kernels of the single indi- Similarly to autumn, turtle doves migrated mainly during viduals and combined data of the PCs hardly overlapping night-time (77.8%). The mean flight speed during the active and the 50% kernels not overlapping (Fig. 4, Supplementary spring migration (50.7 ± 12.6 km/h) was not significantly Fig. 4), indicating different occupied niches in breeding and faster than during autumn migration (independent t-test: wintering sites with respect to vegetation and biomass pro- t = − 0.53, df = 38, p = 0.601). duction. However, PC2 (temperature, especially at night- time) did not differ remarkably. Niche description approach Environmental parameters differed between breeding and Discussion wintering grounds (Table 3). The PCA for all individuals combined extracted two significant components: PC1 (eigen- Year-round data from our satellite transmitters allowed us value 5.34) was determined mainly by the five variables, to trace the timing and route followed by turtle doves from which are related to vegetation and biomass production (Psn- breeding grounds in Italy, Germany, Bulgaria, and Slovakia Net, NDVI, EVI, GPP, and Evapotranspiration) and PC2 to the sub-Saharan wintering regions and vice versa, and (eigenvalue 1.37) mainly by the surface temperature at night to compare parameters of the breeding and wintering sites. (Fig. 4, Supplementary Table 3). The significant components extracted by the PCA differed Breeding and wintering areas slightly when single individuals were analyzed separately. PC1 was characterized for all single individuals mainly by The turtle dove breeding season starts immediately after PsnNet, NDVI, EVI, GPP, and Evapotranspiration. For the arrival on the breeding grounds (Browne and Aebischer individuals #161046, #161048, and #181092 additionally 2001). Assuming a total brood duration of around 45 days by elevation and for individual #181091 also by the popula- (Glutz von Blotzheim and Bauer 1987) and considering the tion density and surface temperature during the day. PC2 average number of days (107) tracked individuals spent at of all individuals was mainly determined by the surface the breeding areas, not more than two broods are possible. 1 3 Behavioral Ecology and Sociobiology (2021) 75: 152 Page 11 of 16 152 For some individuals spending even fewer days (#161049: habitats can be very diverse, depending on their nature (agri- 63; #181091: 76) in the breeding area, only one brood was cultural or natural), location, and time (Hanane 2012; Dias possible. One of the main findings comparing the British et al. 2013; Mansouri et al. 2019). Unlike the aforementioned population during the 1960s and 1990s was that turtle doves studies, based on predetermined areas, e.g., grid squares, we curtailed their breeding season, which ties in with a reduc- checked the land cover in the actually used habitats accord- tion of nesting attempts and productivity per pair (Browne ing to the satellite tracking data. Our results support that and Aebischer 2001, 2003, 2004). Our results indicate that habitat composition varies between different locations, e.g., the time spent in breeding areas may have shortened for tur- preponderance of coniferous forest in Brandenburg, broad- tle doves all over Europe. leaved forest in Hesse or olive groves in Italy (Fig. 2, Sup- For the first time, mean size of home ranges and core plementary Table 1). Our land cover analysis showed that areas (496 and 39 k m2, respectively) could be calculated land cover types suitable for nesting activities (e.g., forests, based on satellite tracking data. Glutz von Blotzheim and olive groves, or shrubs) and areas most likely used for for- Bauer (1987) state that turtle doves often move 3–6 km or aging (e.g., non-irrigated arable land, pastures, crop culti- more from their nesting site for foraging. Even greater forag- vations, or heathland) were present in every home range. ing distances, sometimes > 10 km, were recorded (Browne This reinforces the assumption that the close proximity of and Aebischer 2001). Home ranges based on 100% mini- suitable nesting and feeding areas is a key requirement for mum convex polygons (MCPs) of radio-tagged turtle doves good quality habitats (Browne et al. 2004; Dias et al. 2013). in Britain were between < 1 and 11.30 k m2 (Browne and Our findings indicate that a higher proportion of agricul- Aebischer 2001) and based on 90% MCPs 0.86 ± 0.16 km2 tural areas within the home range leads to an increase in (Dunn et al. 2020). Our calculated home range areas seem home range size. This is in line with Dunn et al. (2020), far larger than these ones. Differences might be due to vary- associating small home ranges with a high proportion of ing calculation methods: On one hand, radio-transmitters are non-farmed habitats and Chiatante et al. (2020), reporting constrained by line-of-sight range between transmitter and that areas with a high proportion of crops were avoided. As receiver, easily leading to missed fixes during foraging trips. large areas of the intensively farmed arable landscape are On the other hand, satellite data have larger error ranges not suitable for feeding, those breeding turtle doves with due the Doppler method, possibly leading to the larger size a high proportion of intensively farmed arable land within of calculated home ranges. In addition, we calculated the their home range are forced to forage over large distances to KUD sizes based on fixes received during the entire time reach good quality food resources. It is likely that the long individuals were at their breeding grounds, while Dunn distances covered affect the adults’ body condition through- et al. (2020) calculated home ranges derived solely from out the breeding season, and hence may negatively influence fixes during incubation and chick stage. As habitat use of their overall breeding performance (Browne and Aebischer turtle doves differs during the breeding season (Browne and 2001). It must be noted that land cover categories used in Aebischer 2001; Mansouri et al. 2019), different foraging the aforementioned studies and in our study mainly describe areas used over the seasonal progress may have added up in landscape types, but do not consider management proce- our calculation. dures. Breeding numbers of turtle doves show an overall It is suggested that individual turtle doves are not site- decline particularly from the 1970s onwards (Fisher et al. faithful (Browne and Aebischer 2001; Dunn and Mor- 2018). While there was no major land cover type change in ris 2012). In contrast, all our turtle doves returning to the Europe between 1950 and 2000 (Gerard et al. 2010), many breeding grounds (n = 3) returned to the same breeding site agricultural and silvicultural management procedures have occupied in the previous year. For #161050, this was the been modified drastically (Baessler and Klotz 2006; Dal- case for four consecutive years. Tracking results therefore limer et al. 2009; Wesche et al. 2012; EEA 2020). Therefore, propose that adult turtle doves are highly faithful to their future studies should take into account information about breeding sites. agricultural and silvicultural management, such as the use In general terms, turtle doves nest in trees or bushes in a of herbicides, conventional or organic farming, timing of landscape characterized by a patchy habitat mosaic of open harvest, understorey, or forest margin management, to be land, nearby to wooded areas and an adjacent water source able to draw a more precise picture of turtle dove habitat (Lutz 2007; Fisher et al. 2018). Habitat selection patterns requirements. and habitat requirements were investigated by numerous At their winter quarters, turtle doves are also susceptible studies mainly based on observational absence and presence to agricultural changes, e.g., increased cultivation, overgraz- data (Supplementary Table 4). These studies show that tur- ing, and cutting of trees (Lutz 2007; Fisher et al. 2018). It tle doves occur over a wide range of forest and agricultural was shown that the overwinter survival of adult turtle doves landscapes, depending on the availability of certain habi- is strongly related to the cereal production at the winter quar- tat types at the regional level, and that nesting and feeding ters (Eraud et al. 2009). Suitable wintering habitats appear to 1 3 152 Page 12 of 16 Behavioral Ecology and Sociobiology (2021) 75: 152 be defined by an abundant food supply, an accessible water July (Fisher et al. 2018). Plotting the habitat niches (Fig. 4) source and large trees or patches of woodland for roosting. shows that the environmental parameters are more widely If one of these key factors is absent, the habitat will typi- dispersed for data from the breeding grounds compared to cally only be used temporarily (Zwarts et al 2009). Previ- more uniform parameters at the wintering grounds, indicat- ous tracking confirmed that turtle doves wintering in West ing differences in the niche breadth for both seasons and a Africa make movements of several hundreds of kilometers more pronounced intraspecific difference in individual habi- during the wintering season (Eraud et al 2013; Lormée et al. tat choice at the breeding compared to wintering sites. This 2016). Our results confirm the use of more than one win- matches the fact that turtle doves occur over a wide range tering site for the majority of tracked individuals (4 of 5) of forest and agricultural landscapes at European breeding with a predominantly southward shift during the wintering grounds, but winter along a relatively narrow and more uni- period (Fig. 1). It is likely that the winter movements are form, with regard to climate and vegetation, latitudinal band linked to the availability of food resources, i.e., tracking food along the Sahel and Sudan savannah. resources that become temporally available by the maturing The main conclusion to be drawn from the niche tracking and harvesting of cereal crops in different regions (Eraud approach is that habitat requirements and preferences deter- et al. 2013; Lormée et al. 2016). Average sizes of 95% and mined at breeding sites cannot be assumed for wintering 50% KUDs (65 and 5 km2, respectively) were very similar sites and vice versa but need to be investigated separately with the size of winter sites (95% MCPs: 60 and 87 km2, due to the apparent observed niche switching in turtle doves. 50% MCPs: 2 and 3 k m2) calculated by Lormée et al. (2016). A narrower niche breadth, during the wintering compared to The habitat mosaic used at wintering sites consisted pre- the breeding season, might suggest that turtle doves might dominantly of crop- and grassland as well as a more varying be more vulnerable to future changes, such as land cover proportion of areas covered by trees or shrubs. The propor- conversion or climate changes, in their winter than in their tion of tree and shrub covered areas appears to be higher for breeding ranges. individuals wintering in the Western part compared to the ones in Central and Eastern part of Western Africa (Fig. 3). Migration and stopovers However, this pattern should be verified with more individu- als in order to derive possible connections between, e.g., Like previous studies (Murton 1968; Lormée et al. 2016), survival, body condition, or migration performance and our data clearly show that turtle doves are mainly nocturnal overwintering region and differing land cover types there. migrants. The migration durations shown by our tracked birds are similar to other studies (autumn migration: 21.3 Niche tracking versus niche switching and 22 days; spring migration: 28.3 and 20–21 days, Eraud et al. 2013; Lormée et al. 2016, respectively). Even if the We still lack a general understanding whether seasonal mean duration did not vary remarkably between spring migration occurs in order to track a specific niche between and autumn (27 and 25 days, respectively), the duration summer and winter distribution ranges, i.e., migrants follow- of autumn migration was more variable inter-individually ing a fixed set of environmental conditions throughout the (7–64 vs. 20–50 days), but also intra-individually (#161050: annual cycle (Zurell et al. 2018). The PCA in our analysis spring migration: 20–23  days and autumn migration extracted ecological habitat parameters related to vegeta- 7–18 days, Table 2). As migration consists of flight and refu- tion and biomass production to mainly determine PC1. This eling periods, the total migration duration is determined by fits the suggestion that Afro-Palearctic migrating landbirds flight speed as well as variables reflecting fuel deposition track the vegetation green-up in spring and depart before performance. The latter, e.g., total stopover duration, are vegetation senescence in autumn (Briedis et al. 2020). The expected to have a much stronger impact than flight behav- extracted parameters for PC1 might also be interpreted as ior on the duration of migration (Houston 2000; Nilsson a proxy for food availability, what would correspond to the et al. 2013; Schmaljohann 2018). Our results show a similar observed winter movements, which are most likely con- mean speed flight during active spring and autumn migration nected to the tracking of different available food resources (approx. 46 vs. 51 km/h, respectively) as well as a similar over time (Eraud et al. 2013; Lormée et al. 2016). PC2, that stopover duration (12.4 vs. 11.9 days). was more constant for both periods, was mainly determined On spring migration, turtle doves were expected to stop by temperature (Supplementary Table 3). For breeding over in the southern border area of the Sahara to refuel prior grounds, it was shown already, that mainly climate variables, to crossing the desert enabling them to cross the Sahara, in particular “minimum temperature in January” and “pre- North Africa, the Mediterranean Seas as well as much of cipitation of the warmest quarter,” shape distribution models Southern Europe without additional stopovers (Zwarts et al. of turtle doves (Marx and Quillfeldt 2018) and that distribu- 2009). Only individual #181091 showed that behavior, stag- tion is linked to an isotherm of a minimum of 16–19 °C in ing in Mauritania, while the remaining individuals possibly 1 3 Behavioral Ecology and Sociobiology (2021) 75: 152 Page 13 of 16 152 may have fueled at their wintering sites already. Instead, we migration routes (Fig. 1). This was particularly notable in found that all complete spring migration tracks included a differing longitudes at which birds arrived at the African stopover in North Africa (Fig. 1). This is in line with the continent (ranging from 9.5, 13.8 to 18.0°O). Likewise, results of Eraud et al. (2013) and Lormée et al. (2016), three individuals from one capture site in Hesse (#200350, showing that birds staged before crossing the Mediterranean #200351, and #200352) showed a similar variability in time Sea. As adult turtle doves have completed the flight-feather for migration onset (31.07, 29.08, and 05.09.2020, respec- molt at that time, it is likely that these stopovers in North tively) and migration routes (Fig. 1). In other bird species, Africa are used to refuel before heading further north (Eraud individuals from one breeding area or colony also follow dif- et al. 2013), emphasizing the importance of these stopover ferent migration routes (Bächler et al. 2010; Schmaljohann sites for successful arrival at breeding grounds. et al. 2012; Trierweiler et al. 2014; Wellbrock et al. 2017). Contrary to the clear pattern during spring migration, the The fact that different individuals from the same breeding stopover pattern during autumn migration showed higher site performed diverse movement patterns during autumn inter- and intraspecific variation. Importantly, in relation to migration suggests that several suitable areas for overwinter- current efforts on adaptive hunting management, the autumn ing coexist, assuming turtle doves taking different migration migration of the majority of individuals (58.3% and exclud- routes also spend the winter period in different sub-Saharan ing the individuals with breeding sites in Italy even 70.0%) areas as suggested by our tracking results and ringing data included prolonged stopovers (> 10 days) in Europe, e.g., (Marx et al. 2016). This indicates a rather weak linkage in France and Spain (Western flyway) or for the Central- between breeding and non-breeding grounds, i.e., a rather Eastern flyway Slovakia and Hungary, representing the most weak migratory connectivity. A rather weak migratory con- important country for autumn stopovers (Supplementary nectivity is in line with the non-existent genetic structuring Fig. 3). These stopovers as well as autumn migration move- across flyways (Calderón et al. 2016). ments match timewise with the legal hunting activities in the In general, our results confirm the three main migration respective European countries (Fisher et al. 2018). routes previously suggested based on mark-recovery data It is likely that the molt of the first inner primaries took (Dimaki and Alivizatos 2014; Marx et al. 2016). However, place at the post-breeding and stopover molt sites (Demongin compared to ring recoveries, the tracking data provide a 2016) along with building up reserves for migration. A com- more detailed picture of the routes. Thus, we show that not mon suggestion for explaining this shift in habitat on the all individuals following the Western migration route fly breeding area is that post-breeding adults might seek out along the strait of Gibraltar but cross the Mediterranean Sea for more abundant food resources and denser vegetation for already further east by leaving from the Spanish mainland, cover, as they may be more vulnerable because of compro- indicating that turtle doves do not necessarily avoid larger mised flight capabilities while they molt (Vitz and Rode- sea crossings. Moreover, the expected course of the Central wald 2007; Tonra and Reudink 2018). Turtle dove #161050 flyway through Italy and Malta was not taken by #200352, was the only carrying out an autumn migration without any which instead crossed the sea further west through Corsica stopover but staged during the other years (Table 2). Also the and Sardinia (Fig. 1). migration routes of #161050 were not exactly the same over The hypothesis for a loop migration pattern, i.e., using the years (Supplementary Fig. 1a). A larger inter-individual a flyway lying west or east of the spring route for autumn variation in autumn than in spring migration was also found migration, in turtle doves hitherto assumed based on geolo- in other tracked bird species (Alerstam et al. 2006; Vardanis cator data (Eraud et al. 2013) is partly supported by our data. et al. 2011; Stanley et al. 2012). This individual variation in In particular, the partial tracks of #161048 and #161049 migration routes may indicate that the birds navigate mainly indicate a clockwise loop migration, i.e., an Eastern route for by other means, e.g., responding to variation in environmen- autumn migration and a Central route for spring migration. tal conditions, than a detailed route recapitulation based on However, #181091, following the Western route, #161046 the recognition of landmarks (Vardanis et al. 2011). Clearly and #161,050, following the Central route for both migra- more turtle dove tracks are needed to statistically confirm tory directions, provide no evidence for a consistent loop that the species might be quite flexible in space, i.e., flex- migration pattern. These findings together with the ringing ibility in migration route. studies, which demonstrated a regular mixing between the By tagging different individuals in the same year as Central and Eastern flyway (Marx et al. 2016), indicate that well as from the same breeding sites, we can show diverse loop migration might be more common for turtle doves fol- movement patterns for individuals sharing a common breed- lowing the Eastern flyway in autumn than for individuals ing site. Individuals #181089, #181090, and #181092 all following the Western or Central flyway. Two main likely tagged at the same forest in Brandenburg started the autumn factors may result in loop migration patterns in some turtle migration with a difference of up to over 1 month (05.08, doves: regional variations in habitat availability and forag- 27.08, and 13.09.2019, respectively) and followed different ing conditions during the two seasons (Tøttrup et al. 2008; 1 3 152 Page 14 of 16 Behavioral Ecology and Sociobiology (2021) 75: 152 Stach et al. 2016) and adaptation to prevailing wind patterns the state office for occupational safety, consumer protection and health, (Patchett and Cresswell 2020; Lisovski et al. 2021), such Brandenburg (license number 2347–11-2018). as different flight altitudes in relation to trade winds and Consent for publication All authors declare that they read the final antitrades during spring and autumn migration in the Sahara version and give consent for the article to be published in Behavioral (Bruderer et al. 2018). Ecology and Sociobiology. Turtle doves spend about two-thirds of the year away from their breeding grounds, at stopover sites, on active migration Open Access This article is licensed under a Creative Commons Attri- and in the wintering grounds, highlighting the importance of bution 4.0 International License, which permits use, sharing, adapta- these periods in the life cycle when considering conservation tion, distribution and reproduction in any medium or format, as long efforts. Less strictly defined migration flyways, some indi- as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes viduals carrying out loop migration and an apparent rather were made. The images or other third party material in this article are weak migratory connectivity result in an overall observed included in the article's Creative Commons licence, unless indicated pattern of migration occurring in a broad front instead of otherwise in a credit line to the material. If material is not included in funneling at specific sites. Consequently, turtle doves should the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will be considered as one (panmictic) population, spending time need to obtain permission directly from the copyright holder. To view a in many different countries during migration, demanding copy of this licence, visit http://c reati vecom mons.o rg/l icens es/b y/4.0 /. concerted conservation actions across all relevant countries to provide protection along all flyways in order to protect the entire population of turtle doves breeding in Europe. References Supplementary Information The online version contains supplemen- Alerstam T, Hake M, Kjellen N (2006) Temporal and spatial patterns of tary material available at https://d oi.o rg/1 0.1 007/s 00265-0 21-0 3082-5. repeated migratory journeys by ospreys. Anim Behav 71:555–566. https:// doi. org/ 10. 1016/j. anbeh av.2 005. 05. 016 Acknowledgements We thank Charles Coleiro, Paulo Lago Barreiro, Bächler E, Hahn S, Schaub M, Arlettaz R, Jenni L, Fox JW, Afanasyev Tim Micallef, Edward Jenkins, Jennifer Greiner, Leslie Koch, and V, Liechti F (2010) Year-round tracking of small trans-saharan Hagen Deutschmann for help with fieldwork, capture, and tagging as migrants using light-level geolocators. PLoS One 5:e9566. https:// well as Mark Gauci, the Maltese Ringing Scheme and BirdLife Malta doi.o rg/ 10.1 371/ journa l. pone. 000956 6 for logistic support on Malta. We gratefully acknowledge the Natur- Baessler C, Klotz S (2006) Effects of changes in agricultural land-use schutz Bund Deutschland e.V. (BirdLife Germany), BirdLife Malta and on landscape structure and arable weed vegetation over the last the Hessische Gesellschaft für Ornithologie und Naturschutz (HGON) 50 years. Agric Ecosyst Environ 115:43–50. https:// doi. org/ 10. for funding and logistic support. 1016/j. agee. 2005. 12.0 07 BirdLife International (2019) Streptopelia turtur. The IUCN Red List of Threatened Species 2019. https://w ww.i ucnre dlist.o rg/s pecie s/ Author contributions YRS, BM, and PQ conceived the study. All 226904 19/ 15437 3407 authors planned and performed fieldwork. YRS, BM, and PQ designed Bivand R, Lewin-Koh N, Pebesma E et al (2020) maptools: tools for han- methodology and conducted the analyses. All authors contributed criti- dling spatial objects. http://r- forge.r-p rojec t. org/p roje cts/ mapto ols/ cally to the manuscript draft. Briedis M, Bauer S, Adamík P et al (2020) Broad-scale patterns of the Afro-Palaearctic landbird migration. Glob Ecol Biogeogr Funding Open Access funding enabled and organized by Projekt 29:722–735. https:// doi. org/1 0.1 111/ geb. 13063 DEAL. Financial support was received from the Naturschutz Bund Browne SJ, Aebischer NJ (2001) The role of agricultural intensification Deutschland e.V. (BirdLife Germany) and BirdLife Malta for the Argos in the decline of the turtle dove Streptopelia turtur. English Nature transmitters and fieldwork and the Hessische Gesellschaft für Orni- Research Report 421, Peterborough thologie und Naturschutz (HGON) funded fieldwork in Hesse. Browne SJ, Aebischer NJ (2003) Temporal changes in the migration phenology of turtle doves Streptopelia turtur in Britain, based on Data availability Data generated and/or analyzed during the current sightings from coastal bird observatories. J Avian Biol 34:65–71 study are included in the published article and its supplementary infor- Browne SJ, Aebischer NJ (2004) Temporal changes in the breeding mation files. ecology of European Turtle Doves Streptopelia turtur in Britain, and implications for conservation. Ibis 146:125–137. https:// doi. Declarations org/ 10.1 111/j. 1474- 919X. 2004. 00235.xBrowne SJ, Aebischer NJ, Yfantis G, Marchant JH (2004) Habitat availability and use by Turtle Doves Streptopelia turtur between Conflict of interest The authors declare no competing interests. 1965 and 1995: an analysis of Common Birds Census data. Bird Study 51:1–11. https://d oi. org/ 10.1 080/ 00063 65040 94613 26 Ethics approval All applicable institutional and/or national guidelines Bruderer B, Peter D, Korner-Nievergelt F (2018) Vertical distribution for the care and use of animals were followed. Capture, handling and of bird migration between the Baltic Sea and the Sahara. J Orni- tagging was carried out under license number BR029/16 of the Wild thol 159:315–336. https://d oi.o rg/ 10.1 007/s 10336- 017-1 506-z Bird Regulation Unit (WBRU) and the Maltese Environment and Calderón L, Campagna L, Wilke T et al (2016) Genomic evidence of Resources Authority (ERA) and under licenses of the regional council demographic fluctuations and lack of genetic structure across fly- Hesse (license number TVA-51/2017), including an ethical approval ways in a long distance migrant, the European turtle dove. BMC according to the national animal protection law (TierSchG §7a), and Evol Biol 16:237. https://d oi. org/1 0. 1186/s 12862-0 16-0 817-7 1 3 Behavioral Ecology and Sociobiology (2021) 75: 152 Page 15 of 16 152 Calenge C (2015) Home range estimation in R: the adehabitatHR pack- Prog Phys Geogr 34:183–205. https:// doi. org/ 10. 1177/ 03091 age. https:// cran.r- proje ct. org/ web/p ackag es/a deha bitatH R 33309 360141 Chiatante G, Porro Z, Meriggi A (2020) The importance of riparian Glutz Von Blotzheim UN, Bauer KM (1987) Handbuch der Vögel Mit- forests and tree plantations for the occurrence of the European teleuropas. Band 9. Columbiformes – Piciformes. AULA-Verlag, Turtle Dove Streptopelia turtur in an intensively cultivated agro- Wiesbaden ecosystem. Bird Conserv Int:1–15.https://d oi.o rg/1 0.1 017/S 0959 Gómez C, Tenorio EA, Montoya P, Cadena CD (2016) Niche-track- 27092 00005 32 ing migrants and niche-switching residents: evolution of cli- Copernicus Land Monitoring Service (2021) Corine Land Cover (CLC) matic niches in New World warblers (Parulidae). Proc R Soc B 2018, Version 2020_20u1. https:// land. coper nicus. eu/ pan- europ 283:20152458. https:// doi.o rg/1 0.1 098/ rspb. 2015. 2458 ean/c orine-l and- cover/c lc201 8 Griffiths R, Double MC, Orr K, Dawson RJG (1998) A DNA test to Dallimer M, Tinch D, Acs S, Hanley N, Southall HR, Gaston KJ, sex most birds. Mol Ecol 7:1071–1075. https://d oi.o rg/1 0.1 046/j. Armsworth PR (2009) 100 years of change: examining agricul- 1365- 294x. 1998.0 0389.x tural trends, habitat change and stakeholder perceptions through Hahn S, Bauer S, Liechti F (2009) The natural link between Europe the 20th century. J Appl Ecol 46:334–343. https:// doi. org/ 10. and Africa - 2.1 billion birds on migration. Oikos 118:624–626. 1111/j. 1365- 2664.2 009.0 1619.x https://d oi.o rg/1 0. 1111/j.1 600- 0706.2 009. 17309.x Demongin L (2016) Identification guide to birds in the hand. Beaure- Hanane S (2012) Do age and type of plantings affect turtle dove Strep- gard-Vendon, pp 184–185  topelia turtur nest placement in olive agro-ecosystems? Ethol Dias S, Moreira F, Beja P, Carvalho M, Gordinho L, Reino L, Oliveira Ecol Evol 24:284–293. https:// doi. org/ 10. 1080/ 03949 370. 2011. V, Rego F (2013) Landscape effects on large-scale abundance pat- 634439 terns of turtle-doves Streptopelia turtur in Portugal. Eur J Wildl Houston AI (2000) The strength of selection in the context of migra- Res 59:531–541. https:// doi.o rg/ 10. 1007/ s10344- 013-0 702-2 tion speed. Proc R Soc Lond B 267:2393–2395. https:// doi. org/ Dimaki M, Alivizatos H (2014) Ringing studies of the turtle dove 10. 1098/ rspb. 2000.1 296 Streptopelia turtur (Aves: Columbidae) during passage through Kapusta P (2015) Points2One. QGIS Plugin. https:// plugi ns.q gis.o rg/ Antikythera Island, southwestern Greece. J Nat Hist 49:419–427. plugin s/ point s2one/ https://d oi.o rg/ 10. 1080/ 00222 933.2 013.8 30790 Lisovski S, Neumann R, Albrecht T et al (2021) The Indo-European Dodge S, Bohrer G, Weinzierl R, Davidson SC, Kays R, Douglas D, flyway: opportunities and constraints reflected by Common Rose- Cruz S, Han J, Brandes D, Wikelski M (2013) The Environmen- finches breeding across Europe. J Biogeogr. published online. tal-Data Automated Track Annotation (Env-DATA) System: link- https:// doi. org/ 10. 1111/j bi. 14085 ing animal tracks with environmental data. Mov Ecol 1:3. https:// Lormée H, Barbraud C, Peach W, Carboneras C, Lebreton JD, Moreno- doi. org/ 10.1 186/ 2051- 3933-1-3 Zarate L, Bacon L, Eraud C (2019) Assessing the sustainability Douglas DC, Weinzierl R, Davidson SC, Kays R, Wikelski M, Bohrer of harvest of the European Turtle-dove along the European west- G (2012) Moderating Argos location errors in animal tracking ern flyway. Bird Conserv Int:1–16.https://d oi.o rg/1 0.1 017/S 0959 data. Methods Ecol Evol 3:999–1007. https:// doi. org/ 10. 1111/j. 27091 90004 79 2041- 210X. 2012. 00245.x Lormée H, Boutin JM, Pinaud D, Bidault H, Eraud C (2016) Turtle Dunn JC, Morris AJ (2012) Which features of UK farmland are impor- Dove Streptopelia turtur migration routes and wintering areas tant in retaining territories of the rapidly declining Turtle Dove revealed using satellite telemetry. Bird Study 63:425–429. https:// Streptopelia turtur? Bird Study 59:394–402. https:// doi. org/ 10. doi.o rg/1 0. 1080/ 00063 657.2 016. 118508 6 1080/ 00063 657. 2012. 725710 Lutz M (2007) Management plan for turtle dove (Streptopelia turtur) Dunn JC, Morris AJ, Grice PV, Peach WJ (2020) Effects of seed-rich 2007–2009. European Commission, Technical Report 007. https:// habitat provision on territory density, home range and breeding ec.e uropa.e u/e nviro nment/n ature/c onser vatio n/w ildbi rds/h untin g/ performance of European Turtle Doves Streptopelia turtur. Bird docs/t urtle_ dove. pdf Conserv Int:1–20. https:// doi. org/ 10. 1017/ S0959 27092 00006 35 Mallet-Rodrigues F (2012) Replacement and growth of primary feath- EEA (2020) European Environment Agency: Forest dynamics in ers in captive rock pigeons, Columba livia (Aves: Columbidae). Europe and their ecological consequences. https:// www. eea. Zoologia 29:121–125. https://d oi.o rg/1 0.1 590/S 1984-4 67020 1200 europa. eu/ themes/ biodi versi ty/ fores ts 02000 04 Epanechnikov VA (1969) Non-parametric estimation of a multivariate Mansouri I, Al-Sadoon MK, Rochdi M, Paray BA, Dakki M, Elghad- probability density. Theor Probab Appl 14:153–158. https:// doi. raoui L (2019) Diversity of feeding habitats and diet composition org/1 0. 1137/ 111401 9 in the turtle doves Streptopelia turtur to buffer loss and modifica- Eraud C, Boutin JM, Riviere M, Brun J, Barbraud C, Lormée H (2009) tion of natural habitats during breeding season. Saudi J Biol Sci Survival of Turtle Doves Streptopelia turtur in relation to western 26:957–962. https:// doi.o rg/ 10.1 016/j.s jbs.2 018.1 1. 006 Africa environmental conditions. Ibis 151:186–190. https:// doi. Marx M, Korner-Nievergelt F, Quillfeldt P (2016) Analysis of ring org/ 10.1 111/j. 1474- 919X.2 008.0 0876.x recoveries of European Turtle Doves Streptopelia turtur - flyways, Eraud C, Rivière M, Lormée H, Fox JW, Ducamp JJ, Boutin JM (2013) migration timing and origin areas of hunted birds. Acta Ornithol Migration routes and staging areas of trans-Saharan turtle doves 51:55–70. https:// doi. org/1 0.3 161/0 0016 454AO2 016.5 1.1. 005 appraised from light-level geolocators. PLoS One 8:e59396. Marx M, Quillfeldt P (2018) Species distribution models of European https:// doi.o rg/1 0.1 371/j ourn al.p one. 00593 96 Turtle Doves in Germany are more reliable with presence only ESA CCI Land Cover project (2021) S2 prototype LC 20m map of rather than presence absence data. Sci Rep 8:16898. https:// doi. Africa 2016. https:// 2016af rica landc over2 0m. esrin. esa. int org/ 10. 1038/ s41598- 018-3 5318-2 Fair J, Paul E, Jones J (2010) Guidelines to the use of wild birds in Murton RK (1968) Breeding, migration and survival of Turtle Doves. research. Ornithological Council, Washington DC Brit Birds 61:193–212 Fisher I, Ashpole J, Scallan D et al (2018) International single spe- Nakazawa Y, Peterson A, Martínez-Meyer E, Navarro-Sigüenza A cies action plan for the conservation of the European Turtle-dove (2004) Seasonal niches of nearctic-neotropical migratory birds: Streptopelia turtur (2018 to 2028). European Commission Techni- implications for the evolution of migration. Auk 121:610–618. cal Report xxx-2018 https://d oi. org/1 0. 2307/ 409042 5 Gerard F, Petit S, Smith G et al (2010) Land cover change in Europe Nathan R, Getz WM, Revilla E, Holyoak M, Kadmon R, Saltz between 1950 and 2000 determined employing aerial photography. D, Smouse PE (2008) A movement ecology paradigm for 1 3 152 Page 16 of 16 Behavioral Ecology and Sociobiology (2021) 75: 152 unifying organismal movement research. Proc Natl Acad Sci U S Proc Natl Acad Sci U S A 106:19637–19643. https:// doi. org/ 10. A 105:19052–19059. https:// doi. org/1 0. 1073/p nas.0 8003 75105 1073/ pnas.0 90156 2106 Nilsson C, Klaassen RHG, Alerstam T (2013) Differences in speed and Tonra CM, Reudink MW (2018) Expanding the traditional definition duration of bird migration between spring and autumn. Am Nat of molt-migration. Auk 135:1123–1132. https:// doi.o rg/ 10.1 642/ 181:837–845. https://d oi.o rg/1 0. 1086/ 670335 AUK- 17- 187.1 Pagen RW, Thompson FR, Burhans DE (2000) Breeding and post- Tøttrup AP, Thorup K, Rainio K, Yosef R, Lehikoinen E, Rahbek C breeding habitat use by forest migrant songbirds in the Missouri (2008) Avian migrants adjust migration in response to environ- Ozarks. Condor 102:738–747. https:// doi. org/ 10. 1093/ condor/ mental conditions en route. Biol Lett 4:685–688. https:// doi. org/ 102.4. 738 10. 1098/ rsbl.2 008. 0290 Patchett R, Cresswell W (2020) Regional wind patterns likely shape Trierweiler C, Klaassen RHG, Drent RH, Exo KM, Komdeur J, Bair- a seasonal migration detour. J Avian Biol 51:e02466. https:// doi. lein F, Koks BJ (2014) Migratory connectivity and population- org/ 10.1 111/ jav. 02466 specific migration routes in a long-distance migratory bird. Proc Pebesma E (2020) sp Classes and Methods for Spatial Data v1.4–1. R Soc B 281:20132897. https:// doi.o rg/ 10. 1098/ rspb. 2013.2 897 https://w ww. rdocum enta tion. org/ packa ges/ sp van der Graaf AJ, Stahl J, Klimkowska A, Bakker JP, Drent RH (2006) PECBMS (2020) Population Trends of Common European Breeding Surfing on a green wave – how plant growth drives spring migra- Birds. https:// pecbms. info/ trends- and- indic ators/ speci es- trends/ tion in the Barnacle Goose Branta leucopsis. Ardea 94:567–577 speci es/ strep topel ia-t urtur/ Vardanis Y, Raymond KHG, Strandberg R, Alerstam T (2011) Individ- Pillar AG, Marra PP, Flood NJ, Reudink MW (2016) Moult migration uality in bird migration: routes and timing. Biol Lett 7:502–505. in Bullock’s orioles (Icterus bullockii) confirmed by geolocators https:// doi. org/1 0. 1098/ rsbl. 2010. 1180 and stable isotope analysis. J Ornithol 157:265–275. https:// doi. Venables WN, Ripley BD (2002) Modern Applied Statistics with S, org/ 10. 1007/ s10336- 015- 1275-5 4th edn. Springer, New York Ponti R, Arcones A, Ferrer X, Vieites DR (2020) Seasonal climatic Vitz AC, Rodewald AD (2007) Vegetative and fruit resources as deter- niches diverge in migratory birds. Ibis 162:318–330. https:// doi. minants of habitat use by mature-forest birds during the post- org/ 10.1 111/i bi. 12784 breeding period. Auk 124:494–507. https://d oi.o rg/ 10. 1093/a uk/ QGIS.org (2016) QGIS Geographic Information System. Open Source 124.2. 494 Geospatial Foundation Project. https:// www. qgis. org/ en/ site/ Wellbrock AHJ, Bauch C, Rozman J, Witte K (2017) Same proce- R Development Core Team (2018) R: a language and environment dure as last year? Repeatedly tracked swifts show individual con- for statistical computing. R Foundation for Statistical Computing, sistency in migration pattern in successive years. J Avian Biol Vienna, Austria. http:// www.R-p roje ct. org 48:897–903. https:// doi. org/1 0. 1111/ jav. 01251 Schmaljohann H (2018) Proximate mechanisms affecting seasonal dif- Wesche K, Krause B, Culmsee H, Leuschner C (2012) Fifty years of ferences in migration speed of avian species. Sci Rep 8:4106. change in Central European grassland vegetation: Large losses https:// doi.o rg/ 10.1 038/ s41598- 018- 22421-7 in species richness and animal-pollinated plants. Biol Conserv Schmaljohann H, Buchmann M, Fox JW, Bairlein F (2012) Tracking 150:76–85. https:// doi. org/1 0.1 016/j.b iocon. 2012. 02.0 15 migration routes and the annual cycle of a trans-Sahara songbird Zúñiga D, Falconer J, Fudickar A, Jensen W, Schmidt A, Wikelski migrant. Behav Ecol Sociobiol 66:915–922. https:// doi. org/ 10. M, Partecke J (2016) Abrupt switch to migratory night flight in 1007/s 00265-0 12-1 340-5 a wild migratory songbird. Sci Rep 6:34207. https:// doi. org/ 10. Stach R, Kullberg C, Jakobsson S, Ström K, Fransson T (2016) Migra- 1038/ srep34 207 tion routes and timing in a bird wintering in south Asia, the com- Zurell D, Gallien L, Graham CH, Zimmermann NE (2018) Do long- mon rosefinch Carpodacus erythrinus. J Ornithol 157:671–679. distance migratory birds track their niche through seasons? J Bio- https:// doi. org/1 0.1 007/ s10336- 016- 1329-3 geogr 45:1459–1468. https://d oi. org/ 10. 1111/ jbi. 13351 Stanley CQ, MacPherson M, Fraser KC, McKinnon EA, Stutchbury Zwarts L, Bijlsma R, van der Kamp B, Wimenga E (2009) Living on BJM (2012) Repeat tracking of individual songbirds reveals the edge: wetlands and birds in a changing Sahel. KNNV Publish- consistent migration timing but flexibility in route. PLoS One ing, Zeist 7:e40688. https:// doi.o rg/1 0. 1371/j ourn al.p one. 00406 88 Tingley MW, Monahan WB, Beissinger SR, Moritz C (2009) Birds Publisher's note Springer Nature remains neutral with regard to track their Grinnellian niche through a century of climate change. jurisdictional claims in published maps and institutional affiliations. Authors and Affiliations Yvonne R. Schumm1 · Benjamin Metzger2 · Eric Neuling3 · Martin Austad4 · Nicholas Galea4 · Nicholas Barbara4 · Petra Quillfeldt1 1 Department of Animal Ecology and Systematics, 3 Naturschutzbund Deutschland E. V. (NABU), Charitéstraße Justus Liebig University, Heinrich-Buff-Ring 26-32, 3, 10117 Berlin, Germany 35392 Giessen, Germany 4 Birdlife Malta, 57/28 Marina Court, Triq l-Abate Rigord, 2 26/1 Immaculate Conception Street, Gzira GZR 1141, Malta Ta’ Xbiex, Malta 1 3