Wetland-dependent migratory songbirds represent one of the most vulnerable groups of birds on the planet, with >67% of wetland-obligate species threatened with extinction. One of the major hurdles for conservation efforts is determining the migration routes, stopover sites, and wintering sites of these species. We describe an annual migration cycle revealed by geolocator tracking of Great Reed-Warblers (Acrocephalus arundinaceus) breeding in the Aras River wetlands of eastern Turkey. Because of its relatively large size and breeding ground fidelity, the Great Reed-Warbler is an excellent candidate for geolocator studies and can serve as an indicator species for other wetland songbirds, many of which are particularly threatened in the Middle East. All birds made use of at least 2 wintering grounds in South Sudan, on the Indian Ocean coast and on the western shores of Lake Malawi, as well as several important stopover sites. We also identified a counterclockwise migration path into and out of Africa. Throughout the year, these birds encountered 277 Important Bird Areas, >40% of which had little or no protection. Many species of wetland songbird, particularly threatened species, may be too rare or too small to be the focus of similar studies. Our results not only allow for comparisons with other Great Reed-Warbler populations, but also reveal previously unknown stopover and wintering locations to target conservation efforts that will help wetland-dependent bird species in the Middle East and East Africa.
Every year, between 2 and 5 billion birds migrate between Africa and the Palearctic (Moreau 1972, Hahn et al. 2009). However, populations of many Afro-Palearctic migrants have declined in the last few decades (Vickery et al. 2014). This is particularly true of wetland-dependent species (IUCN 2015). One major conservation challenge is determining critical habitats for these species, not just on their breeding grounds, but also at their wintering and migration stopover sites (Runge et al. 2014). In the past, tracking the migrations of most bird species in detail was impossible, but advances in light-based geolocator technology have helped to reveal migration and wintering sites for ever-smaller songbirds (Bridge et al. 2011, Peterson et al. 2015, Streby et al. 2015). Geolocators have been deployed primarily in North America and Europe (Stutchbury et al. 2009, Delmore et al. 2012, Hahn et al. 2013, Lemke et al. 2013, Salewski et al. 2013, Finch et al. 2015, Briedis et al. 2016); only rarely has the technology been used to track passerines elsewhere (Jahn et al. 2013, Koleček et al. 2016, Yamaura et al. 2016).
Our study species, the Great Reed-Warbler (Acrocephalus arundinaceus), is an Afro-Palearctic migratory insectivorous passerine dependent on wetlands for breeding (del Hoyo et al. 2006). The species breeds throughout the Palearctic, from the Iberian Peninsula to the Himalayas, and recent studies have identified wintering regions throughout sub-Saharan Africa (Lemke et al. 2013, Koleček et al. 2016). The Great Reed-Warbler is a good candidate species for geolocator tracking because (1) adults weigh 27 g on average (del Hoyo et al. 2006), thus a 1-g geolocator represents <4% of the bird's overall mass, with expected minimal negative consequences for the bird (Barron et al. 2010), and (2) Great Reed-Warblers exhibit high breeding site fidelity (Bensch et al. 1998, Hansson et al. 2002, Koleček et al. 2015), facilitating recapture and geolocator recovery. As a wetland specialist, the Great Reed-Warbler is of particular interest. Worldwide, wetland specialist species are among the most vulnerable to extinction, with 67% of wetland-obligate species under threat (analysis of IUCN (2015) data). Studying wetland species is essential for identifying the causes of declines in these species and for undertaking the necessary conservation actions. Because of its size, distribution, and relative abundance, the Great Reed-Warbler is an excellent indicator species for threatened wetland passerines across much of Europe, the Middle East, and Africa.
Previous studies using stable isotope analyses or mark–recapture data have suggested that this species exhibits some degree of wintering site fidelity (Nisbet and Medway 1972, Yohannes et al. 2008). Banding studies have suggested that the species typically moves in mid-winter, making use of multiple wintering sites (Hedenström et al. 1993). A previous study tracking the movements of Great Reed-Warblers breeding in Sweden confirmed the use of multiple wintering sites and found pronounced migration speed differences between spring and fall (Lemke et al. 2013). A more recent study tracking the migration of Great Reed-Warblers breeding across Europe and the Middle East reported a moderate degree of migratory connectivity and identified a counterclockwise migration for Great Reed-Warblers breeding in Eastern Europe (Koleček et al. 2016).
Here, we present information on the migration of a population of Great Reed-Warblers breeding in eastern Turkey, including wintering regions in Africa, and the birds' approximate migration routes. We also analyze the number of Important Bird Areas (IBAs) that these birds may have potentially encountered during the course of their migration and the degree of protection that these areas receive. IBAs are a network of ecologically critical sites deemed necessary for bird conservation by BirdLife International (2016). IBAs are designated because of their importance to endangered or range-restricted species, their support of large concentrations of species, or their function as migration bottlenecks (BirdLife International 2016). While IBAs can help to target conservation actions, the sites themselves often do not receive protection. Many IBAs and wetlands in Turkey are not formally protected (Şekercioğlu et al., 2011a,b), including the Aras River wetlands where our study took place, and which are threatened by the proposed Tuzluca Dam (Bilgin et al. 2016).
All tagging of Great Reed-Warblers with geolocators took place at the Aras River Bird Research and Education Center in northeastern Turkey (40.07°N, 43.35°E). The center is situated at the intersection of the Aras River and Iğdır Plains Globally Important Bird Areas (Bilgin et al. 2016), straddling Iğdır and Kars provinces. These IBAs lie along a major migration corridor and serve as critical breeding, wintering, and migration stopover sites for millions of birds of 284 species recorded to date.
In May of 2013, we attached 1-g geolocators designed by the British Antarctic Survey to 30 Great Reed-Warblers breeding in the Aras wetlands (Geolocator model MK6790 and MK 6740, Biotrack, Wareham, Dorset, UK; stalk length 13 mm at 45°). We caught birds during standard mist-netting sessions and also used playback to attract Great Reed-Warblers with breeding territories around the study site to increase the likelihood of tagging resident breeding birds likely to return the following year. Birds were outfitted with geolocators using rubber catheter tubing arranged in a “leg-loop” harness (Rappole and Tipton 1991), with the device resting on the bird's back and the light sensor extending caudally to minimize light interference from feathers. Geolocators comprised ∼3–4% of a bird's body mass, a percentage believed not to have negative impacts (Barron et al. 2010). Geolocators underwent a period of calibration lasting between 2 and 17 days. Unfiltered data from the calibration period showed geolocator recording latitude errors of 39 km (± 115 km SD) and longitude errors of 68 km (± 73 km SD). Unfiltered data from the time of deployment through July 15 showed geolocator recording latitude errors of 60 km (± 62 km SD) and longitude errors of 23 km (± 16 km SD).
In May of 2014, 4 males (hereafter, birds A, B, C, and D) with geolocators were recaptured. In May of 2015, 1 additional bird of unknown sex (hereafter, bird E) that was tagged in the 2013 season and had not been recaptured in 2014 was caught. The 4 birds recaptured in 2014 represented a year-to-year recapture rate of ∼13%, higher than the average recapture rate between spring seasons at our site (average = 4%, range = 1–7%). This suggests that geolocators did not reduce survival, although it must be noted that we specifically targeted 2 of the 5 birds with playback.
After geolocator recovery, data were downloaded and decompressed using BASTrak geolocator software (Fox 2009) and clock drift was accounted for (average drift = 7 min, range = 1–15 min; average and range are based on 4 out of 5 geolocators; see next paragraph). The light file was analyzed in R 3.1.1 (R Core Team 2014) using the GeoLight package (Lisovski and Hahn 2012). Twilight events were identified using a light threshold of 5. Sun elevation angle was calculated using the getElevation function (Lisovski and Hahn 2012). All points from time of capture through July 15 were used in calculating sun elevation, as it was unlikely that any birds would have left the area (all birds were caught in established breeding territories) and thus coordinates could be reliably assigned. The average sun elevation angle was −4.07 (range = −2.31 to −5.30), and values calculated for the breeding grounds were used throughout the year. All twilight transitions were converted to geographic points using the coord function (Lisovski and Hahn 2012), and loessFilter (Lisovski and Hahn 2012) was used to remove extreme outliers by comparing them with 2 interquartile ranges (average percentage of points removed = 20%, range = 9–29%).
A longitudinal adjustment was required for locations calculated for bird D. When locations for this bird were determined following the same protocol that we used for birds A, B, C, and E, the data showed bird D's 2013 and 2014 breeding site to be ∼450 km east of its known capture location in eastern Turkey. Other points throughout the course of its migration appeared to be similarly affected. The error was unable to be addressed through clock drift adjustments and likely stemmed from improper time recording during data retrieval. To correct this error, median longitude was calculated for all points from the time of initial capture through July 15, 2013. The difference in longitude (6.28°) between this median and the longitude of the Aras River banding station (where the bird was known to be) was subtracted from all points for bird D.
Probable migration routes and geographic error propagation were inferred using the KFtrack package (Nielsen and Sibert 2004). Each bird's “most probable route” throughout the year (including geographic error; Figure 1) was exported and analyzed in ArcMap 10 (ESRI, Redlands, California, USA). We used BirdLife International data to examine how many Important Bird Areas (IBAs) were included along the migration path of each bird (BirdLife International 2016). IBAs were considered to have been potentially encountered if they fell within the geographic error propagation of a bird's “most probable route” according to KFtrack analysis. In addition to the “most probable route,” we analyzed the number of IBAs that fell within each bird's “most efficient route,” which was defined as the shortest great-circle route between long-term (>30 days) residency periods. “Most efficient route” points were calculated using the geosphere package for R (Hijmans et al. 2016). The same geographic errors calculated from the “most probable route” were used for the “most efficient route.” The number and the geographic area of IBAs that were potentially encountered along each route were compared using 2-sample t-tests. We also examined how many of the IBAs along the “most probable routes” had at least some degree of protection as determined by Birdlife International monitoring programs. Only the migration data for 2013–2014 were used (i.e. data from bird E for 2014–2015 were excluded) to analyze potential IBA encounters.
To determine periods of residency and movement, we used the ChangeLight function with a change point probability threshold of 0.03 and a minimum staging period of 3 days (Lisovski and Hahn 2012). Following the protocol of Koleček et al. (2016), we took average latitude and longitude for each identified stationary period. Subsequent stationary periods with average coordinates <250 km apart were lumped together. Resulting departure and arrival dates were used to calculate the lengths of stay at wintering grounds and stopover sites.
To infer wintering ground locations, we looked at each stationary period delineated by the ChangeLight analysis. Stationary periods of >30 days were considered to indicate probable wintering areas. Points from these periods were converted into kernel density plots with a single 90% contour layer using ArcMap (Figure 2). For stationary periods of <30 days, average latitudes and longitudes of locations were determined and plotted along with latitudinal error (Figure 2). Longitudinal error was too minimal to be evident. These locations were further split based on length of stay.
All 5 birds showed similar movement patterns and wintered in largely the same areas. Birds left eastern Turkey between July 31 and August 12 and spent an average of 29 days (range = 5–61 days) on migration before reaching their first wintering ground. All 5 birds traveled inland across the Arabian Peninsula and crossed over the Red Sea roughly midway between the Sinai Peninsula and the Bab-al-Mandeb Strait (Figure 1). Bird B migrated directly to its first wintering ground in South Sudan, while the other 4 birds made at least 1 stopover during fall migration. Stopover locations included central Sudan, western Ethiopia, central Kenya, and northern Tanzania (Figure 2); the locations of the latter 2 sites may have been influenced by equinox-based geolocator error. Average length of stopover during fall migration was 11 days (range = 5–21 days). All 5 birds spent the first part of the winter in South Sudan or northern Uganda (average length of stay = 91 days, range = 59–119 days). During this time, 2 birds appeared to make small relocations; Bird C moved ∼300 km south-southeast on November 17, and Bird E moved ∼700 km north on November 3 (Figure 2). These locations may represent additional wintering grounds or may be an artifact of geolocator error. The average distance between breeding grounds and first wintering sites was 3,646 km (range = 3,155–4,123 km).
Between November 25 and January 4, all 5 birds left South Sudan and moved south and east, arriving at a second wintering ground either near Lake Malawi (bird E) or on the shores of the Indian Ocean near the Mozambique–Tanzania border (birds A–D) between December 21 and January 19 (Figures 1, 2). Birds A, C, and D moved directly to their second wintering ground, arriving within 1 day of departure, while birds B and E took 41 and 15 days, respectively, to reach their second wintering ground. Bird B made 2 extended stopovers, 1 in South Sudan from November 26 to December 7, the other in Kenya from December 8 to January 3. Bird E stopped over in northern Mozambique from January 9 to January 18. Bird A relocated ∼805 km north on January 29, possibly indicating an additional wintering ground (Figure 2). During the second part of the winter, none of the identified stationary periods for bird E met our criterion for a wintering ground (>30 days) as opposed to an extended stopover (Figure 2). This could have been caused by increased mobility on the part of the bird or geolocator error splitting a stationary period. For birds with an identifiable second wintering period, the length of stay averaged 73 days (range = 33–115 days). The average distance between the first and second wintering grounds was 2,161 km (range = 1,780–2,577 km).
During spring, migration routes were similar for all 5 birds. All birds moved north through the Horn of Africa and crossed into the Arabian Peninsula at the Bab-al-Mandeb Strait (Figure 1). All moved directly north over inland Saudi Arabia rather than tracing the coast (Figure 1). Patterns of movement, however, differed between birds. Bird A and bird D left on April 20 and April 19, respectively, and spent 14 and 12 days, respectively, migrating north. Both stopped over in northern Saudi Arabia–southern Iraq for 9 and 5 days, respectively. Birds B, C, and E left substantially earlier, between February 11 and March 6. All 3 birds made at least 4 stopovers, generally in eastern Kenya–western Somalia and southern Iraq (average length of stopover = 11 days, range = 4–29 days; Figure 2). Birds B and C each showed a stopover between Africa and Madagascar. These locations fell between stationary periods in the Horn of Africa and were likely artifacts of geolocator error. All 3 birds arrived back in eastern Turkey between May 1 and May 28. The average distance between the second wintering ground and the Aras breeding site was 5,533 km (range = 5,351–5,839 km).
During the 2014–2015 season, bird E followed a similar migration pathway as described in the 2013–2014 season (Figure 1). Bird E also exhibited similar patterns of stopover and wintered in similar locations in both years (Figure 2).
Together, the 5 birds potentially encountered 277 IBAs during the course of their migrations (total area = 62,269,453 ha). The majority of these sites (179, 65%) were potentially encountered by only 1 bird. However, 98 (35%) sites were potentially encountered by multiple birds and may be of general importance to migrants in the region (Appendix Table 1). Of all 277 IBAs potentially encountered, BirdLife International (2016) reports 115 (42%) as receiving little or no protection (area = 15,661,258 ha, 25% of total area), 8 (3%) as receiving some protection (7,443,453 ha, 12% of total area), 19 (7%) as mostly protected (area = 8,067,555 ha, 13% of total area), 130 (47%) as receiving complete protection (area = 29,464,187 ha, 47% of total area), and 5 (2%) as receiving an unknown level of protection (area = 1,633,000 ha, 3% of total area; Figure 3). There was no significant difference in the number of IBAs or IBA area encountered between the “most probable route” and the “most efficient route” (2-sample t-test, IBA number: t6.925 = −0.63, P = 0.55; IBA area: t9.666 = −0.35, P = 0.73).
Great Reed-Warblers breeding in Turkey made use of at least 2 wintering grounds in sub-Saharan Africa. This result is consistent with findings from mist-netting studies within Africa (Hedenström et al. 1993) and previous geolocator tracking of Great Reed-Warblers (Lemke et al. 2013, Koleček et al. 2016). South Sudan–northern Uganda, Lake Malawi, and the Indian Ocean coast near the Mozambique–Tanzania border all appear to be well-used wintering sites. The wintering sites in South Sudan and the Mozambique–Tanzania border reaffirm recently identified regions of importance for the Great Reed-Warbler (Koleček et al. 2016), but Lake Malawi appears to be a novel site. Though all 5 birds in our study made large-scale (i.e. >1,000 km) movements in mid-winter, several birds appeared to make smaller relocations as well, possibly indicating the use of more than 2 wintering grounds. As may be expected, these wintering sites are farther east than those used by Great Reed-Warblers that breed in Europe (Lemke et al. 2013), but agree closely with other wintering regions identified for populations of Great Reed-Warbler that breed elsewhere in Turkey (Koleček et al. 2016).
Along with winter site locality, we had occasion to investigate winter site fidelity in 1 individual from which 2 years of data were recovered. There were no obvious year-to-year differences in migration route or winter ground choice, suggesting that Great Reed-Warblers may exhibit winter site fidelity. However, these data come from only 1 individual.
We were able to identify many stopover locations, 2 of which were consistently used by several birds (Figure 2). These 2 locations, near the Kenya–Somalia border and in southern Iraq, may be critical stopover sites for birds traveling between Africa and the Middle East because they flank the geographic barrier formed by the deserts of the Arabian Peninsula and the Horn of Africa. Reliable stopover locations are limited in this region (Scott 1995), and Great Reed-Warblers appear to undertake a continuous migration across the Arabian Peninsula, which would explain the extended stopovers before and after crossing. Records of Great Reed-Warblers from either of these stopover sites range from occasional to nonexistent (Scott 1995, Stevenson and Fanshawe 2002, Sullivan et al. 2009). This highlights the importance of geolocator data for identifying regions of importance for bird species that otherwise go largely undetected.
Migration routes showed that all 5 birds moved in a similar counterclockwise pattern (Figure 1). First, from Turkey they flew southwest across Saudi Arabia and the Red Sea and into Sudan. Next they travelled south and east to either the Indian Ocean coast or the shores of Lake Malawi. Finally, they flew more or less straight north (with the exception of bird E, who first flew northeast to the Indian Ocean coast before continuing in a more northerly direction) across the Bab-al-Mandeb Strait, over inland Saudi Arabia and Iraq (Figure 1). While loop migrations are well documented, they are generally clockwise in both passerines and nonpasserines across the globe (Meyburg et al. 2003, Goodrich and Smith 2008, Klaassen et al. 2010, Schmaljohann et al. 2012, Willemoes et al. 2014). This is thought to be due to the dominant winds running east from ∼30°N to 40°N and west from 30°S to 30°N. Some Afro-Palearctic migrants are known to exhibit counterclockwise migration (Bairlein 2001, Berthold 2001), and our study provides more evidence of this phenomenon. Great Reed-Warblers in particular present an interesting model for studying loop migrations, as some populations migrate clockwise while other populations, principally those wintering in east Africa, migrate counterclockwise (Koleček et al. 2016).
We found that birds traveled on average 3,646 km (range = 3,155–4,123 km) from their breeding grounds to their first wintering site, 2,161 km (range = 1,780–2,577 km) from their first wintering site to their second wintering site, and 5,533 km (range = 5,351–5,839 km) from their second wintering site back to their breeding grounds. These lengths parallel recently identified migration distances for another population of Great Reed-Warbler breeding in Turkey, with reported distances of 3,510 km (range = 3,391–3,743 km), 1,813 km (range = 1,251–2,285 km), and 5,123 km (range = 4,479–5,605 km) for 3 similar migration legs (Koleček et al. 2016). These data suggest the potential for Turkish populations of Great Reed-Warbler to migrate more than 12,000 km each year. Except for Great Reed-Warbler populations breeding in Sweden, these distances are the longest yet recorded for the species (Lemke et al. 2013, Koleček et al. 2016).
IBAs and Protection
The birds in this study potentially made use of 277 IBAs throughout the nonbreeding season. Although many of these areas have a substantial degree of protection (on paper, at least), >40% do not. Runge et al. (2015) found that >90% of migratory bird species do not have adequate protection in at least one part of their yearly range (Runge et al. 2015). Loss of habitat at migration bottlenecks in particular can lead to population-level effects (Runge et al. 2014). This study identified the Bab-al-Mandeb Strait as a consistently used migration bottleneck. However, this critical crossing is greatly understudied and the 3 IBAs most immediately associated with this area (Kadda Guéïni, Les Sept Frères, and Bab al-Mandab – Mawza) all receive little to no protection (BirdLife International 2016, Ç. Şekercioğlu personal observation). If human impact in this critical region inhibits the movements of migratory birds or the stopover potential of migration sites, it could result in substantial negative consequences for all birds traveling through this corridor. We found no difference in the number or area of IBAs along the “most probable route” compared with the “most efficient route” of the birds in our study, suggesting that birds are not deviating from the most efficient migration route due to habitat quality.
We estimate that each Great Reed-Warbler in our study visited a minimum of 11 different countries on 2 continents, for a total of 17 countries for all individuals. Many of the wetlands and IBAs where these birds spent an extended period of time spanned 2 or more political boundaries. This makes conservation efforts especially difficult in a region where conservation laws and practices vary considerably across borders (Dallimer and Strange 2015), and demonstrates the importance of international collaboration for the conservation of migratory species.
Also of note was the consistent use, during spring migration, of the Bab-al-Mandeb Strait, a substantially understudied part of the world. When migrating animals are funneled into narrow passages (e.g., due to topographic constraints), there is great potential for large population effects to occur if these passages deteriorate in their quality as migration corridors (Runge et al. 2014). Egyptian Vultures (Neophron percnopterus) breeding in eastern Turkey (E. Buechley and Ç. Şekercioğlu personal observations) and other Great Reed-Warbler populations breeding in Turkey (Koleček et al. 2016) also appear to favor crossing at the Bab-al-Mandeb Strait, suggesting that this area is critical for many migrants.
Our findings have wide-ranging conservation implications. Because Great Reed-Warblers are larger and more abundant than most wetland passerines, they are an excellent study species and can act as an indicator for other wetland songbirds that are not suitable for a study such as this one. Of the 608 species worldwide that primarily inhabit wetlands, more than a quarter (165 species, 27%) are listed as Near Threatened, Vulnerable, Endangered, Critically Endangered, or Extinct by the International Union for the Conservation of Nature (IUCN 2015). Of the 80 wetland-obligate bird species (i.e. species that reside only in wetlands), 54 species (68%) are threatened with extinction, including the Aquatic Warbler (Acrocephalus paludicola) and Basra Reed-Warbler (Acrocephalus griseldis). Wetland habitats are rare and increasingly threatened by a number of factors, including draining, reed and peat removal, burning, dam building, pollution, and climate change (Junk et al. 2013). It is important to note that birds may exhibit different habitat preferences during the breeding and nonbreeding seasons (Petit 2000), and thus the use of certain regions by Great Reed-Warblers does not necessarily imply suitability for other wetland breeders. Nevertheless, Great Reed-Warblers are good indicators for migratory wetland birds in the arid part of the world studied herein, and additional similar studies to identify important wetlands are a critical step for ensuring year-round protection for migratory species.
We thank Turkey's Department of Nature Conservation and National Parks and Iğdır Directorate of Forestry and Water Affairs for granting us the permits for this study. We are grateful to KuzeyDoğa Society, Kafkas University, the hospitable people of Iğdır's Yukarı Çıyrıklı village, and dozens of volunteers at Aras River Bird Research and Education Center for their support. We thank Ethan Ethington for his analyses of IBAs and Staffan Bensch for his feedback during the fieldwork.
Funding statement: Funding for this study was provided by the Christensen Fund, the National Geographic Society, the Whitley Fund, and the University of Utah. We also thank KuzeyDoğa donors, and in particular Bilge Bahar, Devrim Celal, Seha İşmen, Lin Lougheed, Burak Över and Batubay Özkan for their support. None of the funders had any input into the content of the manuscript, and none of the funders required their approval of the manuscript before submission or publication.
Ethics statement: All tagging was permitted by Turkey's General Directorate of Nature Conservation and National Parks and Forestry, and took place on government-owned land. No protected species were sampled. While birds were being processed, they were housed and handled in such a way as to pose no risk to welfare. During periods of inclement weather, all banding was stopped for the safety of the animals. No sacrificing was required.
Author contributions: E.B., E.Ç., and Ç.H.Ş. conceived the idea, design, or experiment (supervised research, formulated question or hypothesis); J.J.H., E.B., L.A., E.Ç., J.M.H., Y.Ş., and Ç.H.Ş. performed the experiments (collected data, conducted the research); J.J.H. and Ç.H.Ş. wrote the paper; J.J.H. developed or designed methods; J.J.H. and M.C. analyzed the data; and M.A.K. and Ç.H.Ş. contributed substantial materials, resources, or funding.
APPENDIX TABLE 1.
A list of all 277 BirdLife International Important Bird Areas (IBAs) in the Middle East and sub-Saharan East Africa potentially encountered by Great Reed-Warblers over the course of the 2013–2014 nonbreeding season. All IBAs include the number of Great Reed-Warblers (GRW) that potentially visited, as well as location, area, and categorical degree of protection according to BirdLife International (2016).