Discovering the migration and non-breeding areas of sand martins and house martins breeding in the Pannonian basin (central-eastern Europe)

Transcription

1 Journal of Avian Biology 48: , 2017 doi: /jav The Authors. This is an Online Open article Guest Editor: Anders Hedenström. Editor-in-Chief: Jan-Åke Nilsson. Accepted 15 November 2016 Discovering the migration and non-breeding areas of sand martins and house martins breeding in the Pannonian basin (central-eastern Europe) Tibor Szép, Felix Liechti, Károly Nagy, Zsolt Nagy and Steffen Hahn T. Szép, Inst. of Environmental Science, Univ. of Nyíregyháza, Nyíregyháza, Hungary. F. Liechti and S. Hahn, Dept of Bird Migration, Swiss Ornithological Inst., Sempach, Switzerland. K. Nagy and Z. Nagy, MME/BirdLife, Budapest, Hungary. The central-eastern European populations of sand martin and house martin have declined in the last decades. The drivers for this decline cannot be identified as long as the whereabouts of these long distance migrants remain unknown outside the breeding season. Ringing recoveries of sand martins from central-eastern Europe are widely scattered in the Mediterranean basin and in Africa, suggesting various migration routes and a broad non-breeding range. The European populations of house martins are assumed to be longitudinally separated across their non-breeding range and thus narrow population-specific non-breeding areas are expected. By using geolocators, we identified for the first time, the migration routes and non-breeding areas of sand martins (n 4) and house martins (n 5) breeding in central-eastern Europe. In autumn, the Carpathian Bend and northern parts of the Balkan Peninsula serve as important pre-migration areas for both species. All individuals crossed the Mediterranean Sea from Greece to Libya. Sand martins spent the non-breeding season in northern Cameroon and the Lake Chad Basin, within less than a 700 km radius, while house martins were widely scattered in three distinct regions in central, eastern, and southern Africa. Thus, for both species, the expected strength of migratory connectivity could not be confirmed. House martins, but not sand martins, migrated about twice as fast in spring compared to autumn. The spring migration started with a net average speed of 400 km d 1 for sand martins, and 800 km d 1 for house martins. However, both species used several stopover sites for d and were stationary for nearly half of their spring migration. Arrival at breeding grounds was mainly related to departure from the last sub-saharan non-breeding site rather than distance, route, or stopovers. We assume a strong carry-over effect on timing in spring. Various populations of long distance migratory birds in the western Palearctic have declined in recent decades (Sanderson et al. 2006). Candidate factors are climate change (Both et al. 2006), changes of habitats in breeding (Donald et al. 2001), migration, and non-breeding areas (Zwarts et al. 2009, Maggini and Bairlein 2011). To identify carry-over effects and seasonal interactions on population development (Harrison et al. 2011) we need comprehensive knowledge about the whereabouts of individuals during the entire annual cycle. In contrast to detailed information available for breeding periods, we are still lacking information for migration and non-breeding periods (Vickery et al. 2014), especially for long distance passerine species. In the Pannonian basin (central-eastern Europe), 58% out of 26 common long distance migrant species have declined This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. significantly since 1999, whereas only 8% show positive trends (Szép et al. 2012). The information deficiency in relation to the migration and non-breeding areas of these populations are immense, especially as they often play key roles on the dynamics of the entire European populations (BirdLife 2004). The sand martin Riparia riparia and house martin Delichon urbicum are typical examples: their populations suffered from strong declines with mean annual population growth rates of 2.7% in sand martins (during , Szép unpubl.) and 4.7% in house martins ( , Szép et al. 2012). Information on their distribution during the non-breeding period is almost entirely lacking. The sand martins breeding in eastern Hungary were found among the first where adverse climate conditions in potential African non-breeding areas (Sahel) could be correlated with the decreasing annual survival rates (Szép 1995). Despite the huge effort on ringing with almost individuals in eastern Hungary during more than 30 yr, there is no recovery in Africa for this breeding population. For other Hungarian and the neighbouring Czech and Slovakian 114

2 populations, there are only eight recoveries from the African continent, Lake Chad (2), Morocco, east of Tunisia (4), DR Congo and two nearby recoveries from Israel and Lebanon in spring (Heneberg 2008, Szép 2009). Thus, sand martins are assumed to migrate on a broad front (Turner and Rose 1989), and should be widely distributed in sub-saharan Africa, eastern and southern Africa (Walther et al. 2010). The house martin is one of the ten most common Palaearctic African migrants (Hahn et al 2009) but spatial information during the non-breeding season is very scant (Hill 1997). For birds breeding in Germany there are just six recoveries from the African non-breeding areas spanning the Central African Republic, Cameroon, DR Congo, and Zambia (Bairlein et al. 2014). Moreover for the large population in the UK, there is only a single recovery from Nigeria (Hill 2002). House martins breeding in northern Europe had been recovered in southern Africa (Hill 2002, Valkama 2014), whereas the little available information for birds from central-eastern Europe points to migration routes across the Balkan peninsula, southern Italy and north-western Libya during autumn, and north-western Algeria, Malta, and the Balkans during spring (Cepák 2008, Králl and Karcza 2009). Albeit small, this tantalising information raises the suggestion that European house martins might be longitudinally separated in Africa with eastern populations overwintering in east Africa, central populations in Zambia, Zimbabwe and South Africa, and western populations distributed in the region of the Bight of Benin (Hill 2002). In this paper we investigate the migration and non-breeding areas of sand martins and house martins breeding in the Pannonian basin using geolocators, and compare our results with the ringing recoveries of the two species during the non-breeding season. Finally, we also study the diurnal and nocturnal use of cavities during the non-breeding season to explain the very low numbers of African recoveries especially for house martins. Methods Studied populations The sand martin population we studied breeds along the river Tisza in eastern Hungary and has been intensively monitored since 1986 (Szép et al. 2003b). The house martin colonies we studied are situated in two villages along the upper section of the river Tisza, where birds have been ringed annually since Deployment of geolocators In 2012, we equipped adult breeders of both species with SOI GDL2 ver. 1.2 (Swiss Ornithological Inst., Sempach, Switzerland) geolocators using a modified leg-loop harness (Supplementary material Appendix 1). Including the harness, these geolocators weigh 0.6 g. Sand martins were captured at the end of the breeding season, between 9 25 July 2012, in two colonies along the river Tisza, at Szabolcs ( N, E, 35 males, 34 females, colony size 1700 pairs), and at Gávavencsellő ( N, E, five males, six females, colony size 100 pairs). Average body mass of the geolocator-harnessed birds at deployment was 13.4 g (SD 0.80, n 80); thus geolocators mass was 4.5% of body mass. We recaptured five sand martins in May June of 2013 (two females and one male at Szabolcs and two females at Gávavencsellő), and received geolocator data from four birds (one geolocator failed). Additionally, another female with a geolocator was identified at Szabolcs using a digital camera, but was not recaptured. All recaptured birds were active breeders. The return rate of geolocator-harnessed birds varied between Szabolcs (5.8%) and Gávavencsellő (18.2%), but not significantly (Fisher s exact test, p 0.19). Return rates of geolocator-harnessed birds and controls (caught at the same catching events in 2012) showed significant difference at Szabolcs (control: 17.8%, n 129, geolocators: 5.8%, n 69 birds, Fisher s exact test, p 0.028), but not at Gávavencsellő (control: 21.7%, n 46, geolocators: 18.2%, n 11, Fisher s exact test, p 1.0). The return rate of females was double that of males, but the difference was not significant (female: 4/40, male: 2/40, Fisher s exact test, p 0.67). House martins were equipped with geolocators at Nagyhalász-Homoktanya ( N, E 21 males, 18 females and 1 adult of unknown sex) and at Tiszabercel ( N, E three males and seven females between 19 July and 3 August 2012). The mean body mass of geolocator-harnessed birds was 17.1 g (SD 1.04, n 50), while the geolocator mass was 3.5% of adult body mass. The colony at Nagyhalász-Homoktanya comprised 317 nests (41% with clutches), whereas at Tiszabercel the colony consisted of 43 nests (51% with clutches). Five geolocator-harnessed birds were recaptured at the colony of Nagyhalász-Homoktanya (three males, two females) in July of The return rate (12.5%) was lower than 38.9% return rate of control birds (n 18) (Fisher s exact test, p 0.035). At Tiszabercel neither geolocator nor control birds were recaptured. Light-level data analysis We calculated positions using the R-package GeoLight (Lisovski and Hahn 2012). However, we could not use light data from breeding ranges to calibrate sun elevation angles because of the non-natural sunset and sunrise that the birds experienced inside cavities where they nest. We therefore used median sun elevation angle ( 2.7 ) for all individuals of both species derived by the Hill Ekstrom calibration method (Lisovski and Hahn 2012) from long non-breeding stationary periods ( 50 d). This sun elevation angle was very close to the one measured with the same geolocator type in another study ( 2.8) that looked at more than 100 barn swallows (Liechti et al. 2014). Data filtering Light level was recorded every five minutes and varied between 0 (total darkness) and 63 (maximum value; for sensor specific details see Adamík et al. 2016). Based on these values, we defined three time periods per day: 1) a sunrise period that lasted from the first light value 0 after at least four hours of zero values until the light value reached the 115

3 maximum (63) level; 2) a sunset period that lasted from the last maximum light level until the last value above zero, followed by at least four hours of zero values; 3) a lightness period that lasted from the end of the sunrise period until the start of the sunset period. Both species breeds in dark burrows or half-cup nests and, occasionally, they use similar sites (e.g. holes, caves) outside the breeding season. This behaviour can be used to divide recorded sun events (sunrise and sunset) into two classes: natural and non-natural (Liechti et al. 2014, Gow et al. 2015). To filter such potentially biased light data, we defined non-natural sunrises and sunsets on the basis of whether or not the first/last light value was below or above an empirically derived threshold. When the level of first/last light value was less than 5 (84.5% of 5313 cases), the minimum length of sunrise/sunset varied between min; in other cases (first/last light value 5, 15.5%) the minimum length of these periods varied between 0 20 min. In the first case, we considered the sunrise or sunset as natural, while in the second as non-natural. Zero light values during sunrise/ sunset could also indicate the use of dark sites during the entire sunrise/sunset period. Such events were significantly more frequent when the first/last light value was 5 or higher (30.2% of 775 cases) than for lower values (6.6% of 4537 cases; c , p 0.001). Consequently, sunset and sunrise periods with a zero value were also defined as unnatural. Finally, differences in shading between consecutive sun events (sunrise and sunset) can cause an asymmetry in day and night length and lead to error in the calculation of noon and midnight, thus longitude. We therefore calculated the difference in the length of consecutive sun event periods and for further analyses, days and nights with differences larger than 55 min (upper 5% quartile) were excluded, as well as all non-natural sun events (example in Supplementary material Appendix 1, Fig. A1). Defining stationary periods Geolocation divides the temporal pattern of observations into two time steps per 24 h (daytime and night-time), and we assumed that considerable shifts in consecutive sunrises or sunsets indicate a movement. Generally, stationary periods were defined based on the time shifts between two consecutive, equal sun-events (sunrises or sunsets, respectively), which include the two time steps (daytime and night-time). We decided not to use the changepoint algorithm implemented in the GeoLight-software, instead applied some simple rules easy to follow. Our light data had a low impact of shading (except cavity-use), thus, from visual inspection alone it was obvious that some of the very short stopover periods (1 3 d) were missed by the standard algorithm. By using the same data as the standard algorithm we therefore developed a few arbitrary but objective rules to define stationary periods. There were almost no differences in stationary periods longer than 7 d, but we could identify in addition some short stopovers during migration. We believe that this pragmatic approach facilitated the opportunity to derive more detailed results than with a standard approach in our dataset. We determined for each time step whether it was a stationary or movement period by applying the following rules when all sun events were natural: 1) if the time shift between two equal and consecutive sun-events was less, or equal to five minutes, the second time period (daytime or night-time) was classified as a stationary period; 2) if a time shift was more than five minutes between two equal sun-events but smaller than between the two sun-events in the step before, we assume that the bird was stationary during the second time period (daytime or night-time); 3) if a time shift was more than five minutes between two equal sun-events, and larger than between the two sun-events in the step before, we supposed that the bird was moving during the second time period (daytime or night-time); 4) if condition 2 was true, but the difference in the duration of the actual day (or night, respectively) and the day before (or night, respectively) was larger than 10 min, then the period was classified as an uncertain stationary period. These rules are arbitrary and specifically adapted to the data we collected with this type of geolocator. Nevertheless, our classification is based on objective rules and, most importantly, is independent of calculated positions. Thus, all geographical positions within a stationary period (not interrupted by a movement period) were pooled to a single site, and if median geographic position between consecutive sites was within a 200 km radius, then sites were pooled. Sites were defined by their median and 90% quartile of all geographical positions within this time period. The seasonal pattern was defined by four specific stationary periods: 1) the end of the breeding period was defined by departure from the breeding grounds (leaving the first stationary site within 200 km from the breeding grounds); 2) arrival at the first, and departure from the last site north of the Mediterranean Sea defined the pre-migratory period; 3) arrival at the first and departure from the last site south of the Sahara ( 23.5 N) defined the non-breeding residence period; 4) arrival at breeding grounds was defined by a stationary period within 200 km or by the occurrence of frequent unnatural sun events due to nest cavity visits. The initiation and end of autumn migration was defined by the end of the pre-migratory period and the start of the nonbreeding residency. Spring migration was defined by the end of the non-breeding residency and the start of the breeding period. The area associated with the longest stationary period south of the Sahara ( 23.5 N) was defined as the main residence area during the non-breeding period. We defined the time period of autumn migration as the last date of the last stationary site ( 5 d) north of the Mediterranean Sea, and the first date of the first stationary site ( 5 d) south of the Sahara desert (latitude 23.5 N). The period of spring migration was defined vice versa. Use of cavities Sand martins and house martins are diurnal foragers of aerial plankton and use cavities for breeding as well as perhaps for resting (Cramp 1988). We quantified the use of cavities within the annual cycle and identified days with cavity visits during daylight when periods of light records of zero occurred for at least five minutes. Cavity use during nights was determined if either sunrise or sunset was defined as unnatural events. In the case of one bird (S4, sand martin) the geolocator had slipped to the side, thus data could not be 116

4 used for defining cavity usage (one third of positioning data was influenced). Data available from Movebank Data Repository: < doi: /001/ > (Szép et al. 2016). Results Autumnal pre-migration period Sand martins departed from their breeding sites before August (Szép unpubl.), although exact departure dates were not recorded because the logging season started 1 August. In August, all tracked birds were roaming in the Carpathian Bend or in Serbia, Romania, and Bulgaria until departing for their southbound migration (Fig. 1A, Supplementary material Appendix 1, Table A1). Tracked house martins departed between the end of July (K. Nagy unpubl.) and the first half of August from their breeding site to the Carpathian Bend (three individuals) as well as to western Ukraine and western parts of Romania. (Fig. 1C, Supplementary material Appendix 1, Table A2). Autumn migration The autumn migration coincided with the equinox period, and thus latitudinal estimates are not available during this period. However, longitudinal data indicated migratory tracks across the Balkan region with subsequent traverses of the Mediterranean Sea in both species (Fig. 1A, C). Sand martins showed a clear shift in longitude, with an easterly shift in the first half of the migration period and westerly in the second. The sand martins departed on average at 9 September (5 14 of September) and arrived after 17 d (16 20 d) at the non-breeding areas (mean: 26 September, range: September) (Fig. 2A). The overall migration speed averaged at 229 km d 1 (range: km d 1 ). Based on longitudinal positions house martins followed a more or less straight route to the non-breeding area in sub- Saharan Africa (Fig. 1C). They departed on average also at 8 September (5 12 September) and arrived after 24 d (21 30) at the non-breeding area (mean: 3 October, range: 2 5 October) south of the Sahara (Fig. 2C). The overall migration speed averaged at 193 km d 1 (range: km d 1 ). Non-breeding residence The main non-breeding areas of tracked sand martins were situated in the Lake Chad Basin (Fig. 1A), and thus, on average 4250 km ( km) separated from the breeding sites (great circle distance). Three of four birds used a single non-breeding site for on average 170 d (range: d), and one bird (S2) used three distant areas (400 km, 750 km, Fig. 1A). Before spring migration, between March and May, two individuals moved to pre-migratory sites near Lake Chad for about d (Fig. 1A). House martins were distributed widely across sub- Saharan Africa; we tracked two individuals in central Africa (4250 and 4500 km great circle distances from their breeding sites), two individuals in eastern Africa (Uganda, Ethiopia, distances of 4250 and 5200 km), and one individual in southern Africa (distance 8050 km) (Fig. 1C). Three of the five birds we tracked stayed in one area for an average of 159 d (range: d, Supplementary material Appendix 1, Table A1), while two birds (H3, H4) used two separated sites (400 km, 1200 km) (Fig. 1C). Between March and May, all the tracked birds moved to other sites, an average of 1350 km (range: km) from their main non-breeding residences (Supplementary material Appendix 1, Table A1, Fig. 1C). Surprisingly, only three of the five individuals we tracked moved northwards to sites closer to the southern Saharan border while two birds moved to sites in the west/ southwest and stayed there for at least three weeks before departing for their spring migration. Spring migration Three sand martins migrated straight to the north across the desert, while one bird followed a westerly loop (Fig. 1B). Sand martins departed between 11 April and 7 May, and used 5 6 stopover sites with an average stopover duration of 1.5 d (0.5 4 d) (Fig. 2B, Supplementary material Appendix 1, Table A1). Arrival at the breeding colonies was between 29 April and 24 May resulting in a migration duration of 14 d (range: 8 17 d) (Fig. 2B). Their overall migration speed was an average of 349 km d 1 (range: km d 1 ), while their net migration speed, excluding times spent at stopover sites (mean: 6 d, range: 6 11 d), was 621 km d 1 (range: km d 1, Supplementary material Appendix 1, Table A1). In contrast, house martins used two distinct migration routes related to difference in their main non-breeding sites. Birds from eastern Africa moved along an eastward loop across the Arabian Peninsula, Turkey, and the Balkan Peninsula and thus avoiding the crossing of the Mediterranean sea (Fig. 1D), while individuals overwintering in central Africa and South Africa moved straight across the desert and then crossed the central part of the Mediterranean sea through Malta/Sicily, southern Italy and Adriatic Sea, a very similar flyway to two of the tracked sand martins. Individuals overwintering in central and eastern Africa departed between 26 April and 8 May (Fig. 2D), and stopped over at 3 7 sites for an average of one day (0.5 4 d). They arrived at the breeding sites between 5 and 18 May, after 10 d (6 16 d) on migration (Fig. 2D, Supplementary material Appendix 1, Table A1). The bird overwintering in southern Africa left this site at the end of March and moved northwards to central Africa using three stationary sites (2 6.5 d). From these sites onwards, their migration was very similar to others (Fig. 2D); average overall migration speed was 594 km d 1 (range: km d 1 ), and net migration speed, excluding time spent at stopover sites (mean: 4 d, range: 2 9 d), was 1081 km d 1 (range: km d 1, Supplementary material Appendix 1, Table A1). In both species, individuals with the southernmost main non-breeding sites left sub-saharan Africa earlier and were first to arrive at the breeding grounds. Usage of cavities during the annual cycle The majority of birds of both species used cavities during the day when breeding but only occasionally during migration and the non-breeding residence period (Fig. 3A). Cavity use during the nights occured most frequently during the breeding period 117

5 Figure 1. Individual tracks of sand martins (A, B) and house martins (C, D) from breeding areas in the Pannonian basin to the nonbreeding areas in Africa (A, C) and back (B, D). Each colour represents an individual track. Subsequent stationary periods of an individual are connected by dashed lines, except for the autumn migration period during equinox times. The connecting lines are interpretations due to longitudinal information alone. Stationary periods are labelled by a digit (individual 1 5), a character (time period a autumn, w winter, s spring) and second digit (sequence of individual time period). Stationary sites with more than 20 positions were marked by the median and the 90% range of the positions, for shorter stopover periods (with circles) the median and an arbitrary standard error of 200 km for longitude, and 300 km for latitude are indicated. For sand martins all distant captures/recaptures from the studied population are indicated in panel (A) (autumn, 1 August to 11 September) and panel (B) (spring, 8 April to 24 May). For further details see Supplementary material Appendix 1, Table A1. and less often during non-breeding residence and migration periods (Fig. 3B). Results show that house martins used cavities more often on their African non-breeding (15.3 vs 0.7%, c , p 0.001) and breeding sites (80.3 vs 50.4%, c , p 0.001) than sand martins (Fig. 3B). Discussion This study has revealed, for the first time, the spatio-temporal migration patterns and non-breeding locations for individual sand martins and house martins. 118

6 (A) (B) w1 2w1 3w1 4w1 1w1 2w4 3w2 4w Distance from the breeding area (km) Distance to the breeding area (km) (C) (D) w1 2w1 3w1 4w1 5w1 1w2 2w3 3w3 4w7 5w Distance from the breeding area (km) Distance to the breeding area (km) Figure 2. Individual timing of migration and distances covered for the sand martins (A, B) and the house martins (C, D) in autumn (A, C) and spring (B, D). Distances (km) are given in relation to the breeding sites. Dots refer to arrival and departure at the specific site indicated in the legends (cf. Fig. 1). The Carpathian Bend is well known as an important area for the preparation of autumn migration in both species (Králl and Karcza 2009, Szép 2009), but our study now shows that some northern parts of the Balkan Peninsula are similar important. Recapture data from birds from our studied populations confirm this finding (Fig. 1A). When migrating, our track records show that sand martins move along the Balkan Peninsula, and cross the Mediterranean Sea at Greece in a narrow band, in contrast to recent recoveries from Italy and Malta that indicate a much wider autumn migration corridor (Cepák 2008, Heneberg 2008, Králl and Karcza 2009, Szép 2009). Mismatch between our records and others may be due to the specific weather conditions during the study year, as all individuals we studied departed for their autumn migration within a very narrow nine day time period. In addition, individuals from both species followed a route which included just a 500 km sea crossing. However, due to a lack of latitudinal information, we cannot distinguish whether, or not, this directional shift occurred in northern Africa (Fig. 1A). Contrary to our expectations, the studied sand martins spent the non-breeding season in an area with a radius of less than 700 km in northern Cameroon and the Lake Chad Basin. This result is consistent with two central-eastern European population ringed birds recovered from Lake Chad, but not fully with the 38 spring recapture/recoveries from our studied population, which are distributed along the wide west east range of the Mediterranean basin (Fig. 1B). While recoveries do point towards a more widespread nonbreeding range in Africa with low migratory connectivity, our results support the opposite view. Due to considerable differences in reporting rates among the different countries, ringing recoveries/recaptures have a considerable geographical bias. However, our small sample size does not allow for a final conclusion with respect to the strength of migratory connectivity. Obviously, the main non-breeding area of our studied population is situated more to the east compared to recoveries of the western and central European populations (Mead 2002, Walther et al. 2010, Bairlein et al. 2014). Trace element profiles of feathers grown by the British and Spanish martins while in Africa (overwintering in the western Sahel) differed from Hungarian birds (Szép et al. 2003a) in concordance with the geolocation result. In contrast, the house martins in our study are shown to occupy a much larger non-breeding range than sand martins. We identified three distant, non-breeding areas in central, eastern, and southern Africa that appear to have weak migratory connectivity. Very few central European birds have been recovered from central Africa (Bairlein et al. 2014), and records of non-breeding areas in eastern Africa represent 119

7 (A) Frequency of usage of cavities during day(%) (B) Frequency of usage of cavities during nights(%) Premigratory, autumn Premigratory, autumn House martin Migration, autumn House martin Migration, autumn Sand martin Wintering Season Sand martin Wintering Season Migration, spring Migration, spring Breeding Breeding Figure 3. Frequency of cavities usage (mean SD) of sand martins and house martins during the day (A) and the night (B) for the four non-breeding periods defined (see Methods) and the breeding season (significant differences are marked with ***: p 0.001). new results for central-eastern European breeding house martins. Non-breeding areas in southern Africa were already known for northern European and German populations (Hill 2002, Bairlein et al. 2014, Valkama 2014), but had not been recorded before for central-eastern European populations. All the non-breeding sites for Pannonian birds are situated east with respect to indirectly assigned non-breeding areas for a Dutch population (Hobson et al. 2012) and some north Italian individuals (Ambrosini et al. 2011). Nevertheless, there is considerable overlap in the non-breeding areas now known for Pannonian house martins with several other European populations; as a result, our findings do not support the hypothesis that there is a longitudinal separation in the non-breeding areas of European populations (Hill 2002, Ambrosini et al. 2011). Within the pre-migratory period, two of our five house martins moved considerably westward, towards an area where birds from the western and central European population have also been recovered during this time of the year (Bairlein et al. 2014). However, whether or not these birds spent their whole non-breeding period within this area remains unclear. One reason for the low recovery rate of house martins compared to other swallows might be their more frequent use of cavities (e.g. caves, trees, buildings) during the nonbreeding season in Africa. Such roosts are much more difficult to find, are occupied by a smaller number of individuals, and therefore are less attractive to local people than the huge roosts of the other two swallow species (Hill 2002). Spring migrations of our studied individuals took place after the equinox and, therefore, tracks and stopover *** *** behaviour could be estimated. The most striking difference in migratory routes was not between species, but rather between geographical positions of main non-breeding areas. Identified passage areas in the spring across the central part of the Mediterranean basin coincide very well with most recapture sites known for both Pannonian populations (Cepák 2008, Heneberg 2008, Králl and Karcza 2009, Szép 2009). Only two house martins overwintering in eastern Africa circumvented the Mediterranean Sea on their spring migration as is the case in other insectivorous birds (i.e. red-backed shrikes; Tottrup et al. 2012). Although autumn migration routes can only be reconstructed here using longitudinal information, there is good evidence that none of the individuals followed the same route in both autumn and spring. Migration duration in sand martins was very similar in autumn and spring, while house martins spent almost three times more days on their autumn compared to their spring migration. As a result, the overall travelling speed of house martins was about 16% lower than the sand martins in autumn, but 70% faster in the spring. Indeed, five out of the nine birds we tracked returned in spring in 10 d or less. Departure and arrival dates were also less synchronous in the spring compared to the autumn for all individuals of both species, but there was a high correlation in departure and arrival dates (r 0.88, Pearson) with the exception of the house martins in autumn. In both seasons, migratory distances were slightly longer for house martins than they were for sand martins ( 20%), taking into account the first (for autumn) and the last (for spring) sub-saharan stationary sites. There were also no obvious differences in overall migration speeds in the spring within species compared to dates, distances, or chosen route. Thus, in both species, arrival at breeding grounds was strongly determined by the date of departure from the last sub-saharan non-breeding site. During spring migration, when time spent at stopoversites could be estimated, individuals of both species where stationary for nearly half of their migration period. Net migration speeds were over 400 km d 1 for sand martins, rising to twice this value for house martins, over 800 km d 1 ; this strongly suggests that these birds use tail winds, as air speeds measured for migrating sand martins and house martins average around 40 km h 1 (Liechti and Bruderer 2002). In contrast, net migration speeds during the spring tended to be higher for individuals using more distant nonbreeding residence areas; these birds must have either been able to accumulate more fuel reserves before departure, profited from abundant food resources en route, or benefited more from tail winds. As none of the individuals (in both species) surpassed another during spring migration, and because spring migration was generally fast (especially in house martins), we assume that there is a strong carry-over effect between departure from sub-saharan Africa and arrival at breeding grounds although more track records will clearly be needed to statistically support this assumption. To define stationary periods during spring migration in such details, we had to overcome the standard method used so far (GeoLight). We must admit that our approach provided reliable results only because of the high quality light data, with very minor shading effects (Lisovski et al. 2012). Based on our data, the fastest spring migration of a house martin (H5) was reconstructed using a movement 120

8 period of 4.5 d, resulting in an average ground speed of 52 km h 1. This is a realistic speed, well in line with known ground speeds measured by radar for small migrating passerines (Liechti and Bruderer 2002). Finally, as all the speeds we estimated were within a realistic range, our filtering seems to provide feasible results. Conclusions In contrast to the existing ringing recoveries from our sand martin population, the migration and non-breeding residency of the studied sand martins was fairly concentrated. Instead, for our studied house martins non-breeding residency of the five individuals was distributed across a huge area in subsaharan Africa, not supporting proposed population specific non-breeding ranges. Due to the small sample size these preliminary conclusions need to be verified by further investigations. The geolocator study provided substantial information for the whereabouts of the Pannonian populations of sand martin and house martin. However, there are also drawbacks of the method. For instance the lower return rates of geolocator birds compared to controls calls for cautious use of geolocators on such small birds, especially aerial feeders (see also Gomez et al. 2014, Scandolara et al. 2014) and careful interpretation of data on timing (Arlt et al. 2013). However, geolocation remains the only method to track the martins to date. Acknowledgements We thank Ákos Pelenczei, Zoltán Görögh, János Danku, Annamária Danku, Beáta Bokor, Ivett Kakszi, Zsolt Hörcsik, the members of the local chapter of the MME/BirdLife Hungary for the important help in the field work, and Edit Molnár for processing our field data. The deployment of the geolocators and related field works were carried out with the permission of the Hungarian National Inspectorate for Environment, Nature and Water (14/2104/5/2012). The study was donated by Madárvédelem határok nélkül HUSK/1101/2.2.1/0336 project of the EU and by the local chapter of the MME/BirdLife Hungary. The Swiss federal office for environment contributed financial support for the development of the data loggers (UTF-Nr. 254, 332, 363, 400). References Adamík, P., Emmenegger, T., Briedis, M., Gustafsson, L., Henshaw, I., Krist, M., Laaksonen, T., Liechti, F., Procházka, P., Salewski, V. and Hahn, S Barrier crossing in small avian migrants: individual tracking reveals prolonged nocturnal flights into the day as a common migratory strategy. Sci. Rep. 6: Ambrosini, R., Orioli, V., Massimino, D. and Bani, L Identification of putative wintering areas and ecological determinants of population dynamics of common house-martin (Delichon urbicum) and common swift (Apus apus) breeding in northern Italy. Avian Conserv. Ecol. 6, doi: /ACE Arlt, D., Low, M. and Part, T Effect of geolocators on migration and subsequent breeding performance of a longdistance passerine migrant. PLoS One 8: e Bairlein, F., Dierschke, J., Dierschke, V., Salewski, V., Geiter, O., Hüppop, K., Köppen, U. and Fiedler, W Atlas des Vogelzugs: Ringfunde Deutscher Brut- und Gastvögel. Aula Verlag. BirdLife Birds in Europe: population estimates, trends and conservation status. BirdLife International, Cambridge. Both, C., Bouwhuis, S., Lessells, C. M. and Visser, M. E Climate change and population declines in a long-distance migratory bird. Nature 441: Cepák, J Jiricka obcena (house martin). In: Cepák, J., Klvana, P., Škopek, L., Schröpfer, L., Jelínek, M., Horák, D., Formánek, J. and Zárybnický, J. (eds), Czech and Slovak bird migration atlas. Aventinum, pp Cramp, S The birds of the Western Palearctic, vol. 5. Oxford Univ. Press. Donald, P. F., Green, R. E. and Heath, M. F Agricultural intensification and the collapse of Europe s farmland bird populations. Proc. R. Soc. B 268: Gomez, J., Michelson, C. I., Bradley, D. W., Norris, D. R., Berzins, L. L., Dawson, R. D. and Clark, R. G Effects of geolocators on reproductive performance and annual return rates of a migratory songbird. J. Ornithol. 155: Gow, E. A., Wiebe, K. L. and Fox, J. W Cavity use throughout the annual cycle of a migratory woodpecker revealed by geolocators. Ibis 157: Hahn, S., Bauer, S. and Liechti, F The natural link between Europe and Africa 2.1 billion birds on migration. Oikos 118: Harrison, X. A., Blount, J. D., Inger, R., Norris, D. R. and Bearhop, S Carry-over effects as drivers of fitness differences in animals. J. Anim. Ecol. 80: Heneberg, P Brehule ricni (sand martin). In: Cepák, J., Klvana, P., Škopek, L., Schröpfer, L., Jelínek, M., Horák, D., Formánek, J. and Zárybnický, J. (eds), Czech and Slovak bird migration atlas. Aventinum, pp Hill, L. A Trans-Saharan recoveries of house martins (Delichon urbica), with discussion on ringing, roosting and sightings in Africa. Safring News 26: Hill, L. A House martin. In: Wernham, C. V., Toms, M. P., Marchant, J. H., Clark, J. A., Siriwardena, G. M. and Baillie, S. R. (eds), Britain and Ireland: the migration atlas: movements of the birds of Britain and Ireland. Poyser, pp Hobson, K. A., Van Wilgenburg, S. L., Piersma, T. and Wassenaar, L. I Solving a migration riddle using isoscapes: house martins from a Dutch village winter over west Africa. PLoS One 7: e Králl, A. and Karcza, Z Molnár fecske (house martin). In: Csörgő, T., Karcza, Z., Halmos, G., Magyar, G., Gyurácz, J., Szép, T., Bankovics, A., Schmidt, A. and Schmidt, E. (eds), Magyar madárvonulási atlasz. (Hungarian bird migration atlas.) Kossuth, pp Liechti, F. and Bruderer, L Wingbeat frequency of barn swallows and house martins: a comparison between free flight and wind tunnel experiments. J. Exp. Biol. 205: Liechti, F., Scandolara, C., Rubolini, D., Ambrosini, R., Korner- Nievergelt, F., Hahn, S., Lardelli, R., Romano, M., Caprioli, M., Romano, A., Sicurella, B. and Saino, N Timing of migration and residence areas during the non-breeding period of barn swallows Hirundo rustica in relation to sex and population. J. Avian Biol. 45: Lisovski, S. and Hahn, S GeoLight processing and analysing light-based geolocator data in R. Methods Ecol. Evol. 3: Lisovski, S., Hewson, C. M., Klaassen, R. H. G., Korner-Nievergelt, F., Kristensen, M. W. and Hahn, S Geolocation by light: accuracy and precision affected by environmental factors. Methods Ecol. Evol. 3: Maggini, I. and Bairlein, F Body condition and stopover of trans-saharan spring migrant passerines caught at a site in southern Morocco. Ring. Migr. 26: Mead, C Sand martin. In: Wernham, C. V., Toms, M. P., Marchant, J. H., Clark, J. A., Siriwardena, G. M. and Baillie, 121

9 S. R. (eds), Britian and Ireland: the migration atlas: movements of the birds of Britain and Ireland. Poyser, pp Sanderson, F. J., Donald, P. F., Pain, D. J., Burfield, I. J. and van Bommel, F. P. J Long-term population declines in Afro-Palearctic migrant birds. Biol. Conserv. 131: Scandolara, C., Rubolini, D., Ambrosini, R., Caprioli, M., Hahn, S., Liechti, F., Romano, A., Romano, M., Sicurella, B. and Saino, N Impact of miniaturized geolocators on barn swallow Hirundo rustica fitness traits. J. Avian Biol. 45: Szép, T Relationship between west-african rainfall and the survival of central-european sand martins Riparia riparia. Ibis 137: Szép, T Partifecske (sand martin). In: Csörgő, T., Karcza, Z., Halmos, G., Magyar, G., Gyurácz, J., Szép, T., Bankovics, A., Schmidt, A. and Schmidt, E. (eds), Magyar madárvonulási atlasz. (Hungarian bird migration atlas.) Kossuth, pp Szép, T., Møller, A. P., Vallner, J., Kovács, A. and Norman, D. 2003a. Use of trace elements in feathers of sand martin Riparia riparia for identifying moulting areas. J. Avian Biol. 34: Szép, T., Szabó, D. Z. and Vallner, J. 2003b. Integrated population monitoring of sand martin Riparia riparia an opportunity to monitor the effects of environmental disasters along the River Tisza. Ornis Hungarica 12 13: Szép, T., Nagy, K., Nagy, Z. and Halmos, G Population trends of common breeding and wintering birds in Hungary, decline of long-distance migrant and farmland birds during Ornis Hungarica 20: Szép, T., Liechti, F., Nagy, K., Nagy, Z. and Hahn, S Data from: Discovering the migration and non-breeding areas of sand martins and house martins breeding in the Pannonian basin (central-eastern Europe). Movebank Data Repository, < doi: /001/ >. Tottrup, A. P., Klaassen, R. H. G., Strandberg, R., Thorup, K., Kristensen, M. W., Jorgensen, P. S., Fox, J., Afanasyev, V., Rahbek, C. and Alerstam, T The annual cycle of a transequatorial Eurasian-African passerine migrant: different spatio-temporal strategies for autumn and spring migration. Proc. R. Soc. B 279: Turner, A. and Rose, C Swallows and martins: an identification guide and handbook. Houghton Mifflin Company. Valkama, J House martin (Delichon urbicum). In: Valkama, J., Saurola, P., Lehikoinen, A., Lehikoinen, E., Piha, M., Sola, P. and Luomus, W. (eds), The Finnish bird ringing atlas volume II. Finnish Museum of Natural History, pp Vickery, J. A., Ewing, S. R., Smith, K. W., Pain, D. J., Bairlein, F., Skorpilova, J. and Gregory, R. D The decline of Afro-Palaearctic migrants and an assessment of potential causes. Ibis 156: Walther, B. A., van Niekerk, A., Thuiller, W., Baumann, S., Dean, W. R. J., de Bruijn, B., Gutteridge, K., Jones, P. J., Nikolaus, G., Pearson, D. J., Robinson, S. P., Salewski, V., Schäffer, N., Taylor, P. B., Tushabe, H., Williams, P. H. and Rahbek, C A database of Western Palearctic birds migrating within Africa to guide conservation decisions. In: Harebottle, D. M., Craig, A. J. F. K., Anderson, M. D., Rakotomanana, H. and Muchai, M. (eds), Proceedings of the 12th Pan-African Ornithological Congress, Animal Demography Unit, pp Zwarts, L., Bijlsma, R. G., van der Kamp, J. and Wymenga, E Living on the edge: wetlands and birds in a changing Sahel. KNNV Publishing, Zeist. Supplementary material (Appendix JAV at < www. avianbiology.org/appendix/jav >). Appendix