The migration system of barn swallows hirundo rustica breeding in Southwestern Spain and wintering across West Africa

ThE MIgRATIOn SySTEM Of BARn SwALLOwS HIRUNDO RUSTICA BREEDIng In SOUThwESTERn SpAIn AnD wInTERIng ACROSS wEST AfRICA EL SISTEMA MIgRATORIO DE LA gOLOnDRInA COMún HIRUNDO RUSTICA REpRODUCTORA En ESpAñA E InvERnAnTE En ÁfRICA OCCIDEnTAL Cosme LópEZ-CALDERón1*, Sergio MAgALLAnES2, Alfonso MARZAL2 and Javier BALBOnTín1 SUMMARy.—Many populations of migratory bird species are rapidly declining. As a requisite for targeting effective conservation efforts it is essential to determine the whereabouts of migrants yearround. however, our knowledge of migratory routes and spatial-temporal occurrence across periods of the annual cycle is still limited for most species. we used light-level geolocators to describe in detail the migration system of Barn Swallows Hirundo rustica breeding in southwestern Spain and wintering across west Africa. we were able to successfully retrieve year-round data for 38 individuals and reconstructed their migratory routes using FLightR R package. Many of the studied individuals remained for some time in summer wandering through southern Spain and northern Morocco, a period that we defined as pre-migration. The studied swallows started their autumn migration on average on August 18th, stopping over to refuel in northwestern Morocco and southern Mali. On average the tagged individuals arrived on September 3rd at their wintering areas, which were located across Ivory Coast and surrounding countries, in localities dominated by savannahs, grasslands and crops. After wintering, swallows started the spring migration January 26th on average, stopping over in Senegal and Mauritania. They arrived back at the Spanish breeding colonies february 18th on average (from mid-January to mid- March). Surprisingly, during the autumn migration, one of the tagged individuals travelled to England before returning south and spending a short wintering period in northwestern Spain.—López-Calderón, C., Magallanes, S., Marzal, A. & Balbontín, J. (2021). The migration system of Barn Swallows Hirundo rustica breeding in southwestern Spain and wintering across west Africa. Ardeola, 68: 335-354. Key words:Hirundo rustica, light-level geolocators, migratory routes, phenology, solar geolocation, stopover areas, wintering areas. RESUMEn.—numerosas poblaciones de aves migratorias están disminuyendo rápidamente. Como requisito previo a implementar medidas de conservación eficaces, es esencial determinar los lugares Ardeola 68(2), 2021, 335-354 DOI: 10.13157/arla.68.2.2021.ra2 1Department of Zoology, University of Seville, Avda. Reina Mercedes, E-41012 Seville, Spain. 2Department of Anatomy, Cellular Biology and Zoology, University of Extremadura, Avda. de Elvas, E-06006 Badajoz, Spain. *Corresponding author: [email protected] Twitter: @MarzalLab / @Magallanes_S_86 / @AlfonsoMarzal InTRODUCTIOn Migratory behaviour has evolved in a variety of taxonomic groups, such as those constituting zooplankton, insects, fish, amphibians, reptiles and mammals. however, migratory behaviour has undoubtedly developed to the greatest extent in birds, whose aerial journeys incur relatively low energy requirements (newton, 2008). Thus, more than 50 billion birds perform regular annual movements between breeding and wintering areas, often involving thousands of kilometres (Berthold, 1993). Specifically, it is estimated that about two billion birds migrate each autumn from Europe to sub-Saharan Africa (hahn et al., 2009). This bulk of individuals constitutes a strong ecological link between different continents across the world, demanding resources along migratory routes and potentially transporting associated species (which may be parasitic or invasive), pollen, seeds, nutrients or contaminants (Kays et al., 2015). Beyond the intrinsic scientific knowledge that has always fascinated ornithologists, it is necessary to know where migratory birds stay during the non-breeding period as a prerequisite for their conservation (faaborg et al., 2010). Indeed, many migratory bird species are declining in numbers, in particular long-distance migrants (Sanderson et al., 2006), aerial insectivores (nebel et al., 2010) and species with low variability in migratory performance (gilroy et al., 2016). Mortality rates are usually much higher during migration than during the stationary life-cycle stages (newton, 2008), and conditions experienced along the journey and during overwintering may affect subsequent reproductive performance (i.e. “carry-over effects”; norris & Marra, 2007; harrison et al., 2011). Thus, in order to inform effective management decisions, it is first necessary to determine migratory schedules as well as the distribution of habitats visited during migration and winter (faaborg et al., 2010). Ardeola 68(2), 2021, 335-354 LópEZ-CALDERón, C., MAgALLAnES, S., MARZAL, A.and BALBOnTín, J. 336 utilizados por las especies migratorias durante todo el año. Sin embargo, nuestro conocimiento sobre las rutas migratorias y el uso espacio-temporal a lo largo de distintos periodos del ciclo anual es todavía limitado para la mayoría de especies. En este estudio, hemos usado geolocalizadores para describir en detalle el sistema migratorio de la golondrina común Hirundo rustica reproductora en el suroeste de España e invernante en África Occidental. fuimos capaces de recuperar con éxito datos de intensidad lumínica durante toda la migración para un total de 38 individuos, y a partir de estos datos hemos reconstruido las rutas migratorias con el paquete de R FLightR. Muchos de los individuos de estudio permanecieron un cierto periodo de tiempo en verano vagando por el sur de España y norte de Marruecos, periodo que hemos definido como premigración. Las golondrinas de estudio comenzaron su migración de otoño en promedio el 18 de agosto, repostando en el noroeste de Marruecos y sur de Mali. En promedio, el 3 de septiembre los individuos marcados llegaron a sus zonas de invernada, distribuidas entre Costa de Marfil y países circundantes (lugares dominados por sabanas, praderas y cultivos). Después de la invernada, las golondrinas comenzaron la migración de primavera en promedio el 26 de enero, repostando a lo largo de Senegal y Mauritania. finalmente, los individuos del estudio llegaron de vuelta a las colonias de reproducción en promedio el 18 de febrero (desde mediados de enero hasta mediados de marzo). Sorprendentemente, durante la migración de otoño, uno de los individuos marcados viajó hasta Inglaterra y volvió a España, donde realizó un corto periodo de invernada en el noroeste del país.—López-Calderón, C., Magallanes, S., Marzal, A. y Balbontín, J. (2021). El sistema migratorio de la golondrina común Hirundo rustica reproductora en España e invernante en África occidental. Ardeola, 68: 335-354. Palabras clave: fenología, geolocalización solar, geolocalizadores, Hirundo rustica, rutas migratorias, zonas de invernada, zonas de repostaje. passerine species make up the majority of long-distance migratory birds (e.g. more than 73% of European breeding individuals; hahn et al., 2009). however, most detailed information on bird migration is only available for raptors and waterfowls (Kays et al., 2015). This gap of knowledge is obviously led by body size, which has traditionally hampered the attachment of satellite transmitters on to small birds such as migratory passerines (Kays et al., 2015). for this reason, gpS tags have been continuously miniaturised until for the first time a ~1g archival gpS tag was attached to a small migratory bird (hallworth & Marra, 2015). however, the temporal resolution of these gpS tags is still very limited (only 28 days; hallworth & Marra, 2015). Consequently, continued miniaturisation of this technology together with an increase in temporal resolution should be expected during the coming years (Kays et al., 2015). In the meantime, archival light-level geolocators represent the most developed tracking technique available to obtain location estimates during the passerine non-breeding period (McKinnon et al., 2013; finch et al., 2017; Lisovski et al., 2020). These devices record light levels at fixed time intervals during the whole migration cycle. Once the marked individuals are recaptured, the stored data can be downloaded for analysis. Then, following astronomical equations (e.g. Ekstrom, 2004, 2007), latitude is calculated according to the duration of daylight or to the rate of change in light levels; whereas longitude is calculated according to the time of solar noon or midnight (Lisovski et al., 2020). Light data recorded by geolocators is noisy due to different physical factors (e.g. proximity to equator, equinoxes, cloud cover) and biological factors (shading caused by dense vegetation or by the use of cavities). Consequently, efforts in this research area are currently focussed on statistical methods for quantifying uncertainty of position estimates taking into account all sources of error (Rakhimberdiev et al., 2015, 2017; Lisovski et al., 2020). Although light-level geolocators present many limitations and offer relatively low precision in position estimates, their simplicity makes them able to collect data during full migration cycles at reasonable economic cost. That is why, since the first published migratory track of a passerine species (Stutchbury et al., 2009), more than 7,800 small birds have been tagged across the world (Brlík et al., 2020). nonetheless, the attachment of geolocators onto small birds (body mass < 100g) slightly reduces survival rates and thus the further miniaturisation of these devices is still mandatory on ethical grounds (Brlík et al., 2020). After all, geologgers have so far provided invaluable scientific information on the whereabouts of migratory passerines, such as the distribution of their non-breeding habitats, migration phenology, migratory connectivity, factors influencing migration and carry-over effects throughout different stages of the life cycle (McKinnon et al., 2013; patchett et al., 2018). The Barn Swallow Hirundo rustica is a small passerine (15-20g) that breeds semicolonially and feeds on insects in flight. The available dataset on ringing recoveries for this species is one of the largest (Ambrosini et al., 2009) but it is still strongly biased towards African countries with high ringing effort (e.g. South Africa). Thanks to limited ringing recoveries, we know that populations breeding in southwestern Europe overwinter from west to Central Africa, whereas northern breeding populations winter from Central to South Africa (Ambrosini et al., 2009). Several breeding populations across Europe have been tracked recently with lightlevel geolocators, confirming these earlier results (Liechti et al., 2014; Arizaga et al., 2015; Klvaňa et al., 2018; Briedis et al., 2018). Taking into account the abundant scientific literature available for the Barn Swallow (e.g. a search in the web of Science Ardeola 68(2), 2021, 335-354 SpAnISh BARn SwALLOw MIgRATIOn 337 for Hirundo rustica produces a total of 1,669 studies), filling gaps of knowledge in its migratory behaviour should benefit other research areas that use this species as a study model. for instance, different research areas such as ecological links between continents (e.g. patchett et al., 2018), avian parasites and diseases (e.g. Kays et al., 2015), sexual selection (e.g. Møller, 1994) or the response of organisms to global Change (e.g. gilroy et al., 2016), among others, would benefit from precise information on the migratory behaviour of Barn Swallows. In this study, we used light-level geolocators to elaborate a detailed description for the migration system of Barn Swallows breeding in the southwestern Iberian peninsula. first, we reconstructed the migratory tracks of our study individuals. Second, we determined their migration schedules. Third, we identified the stopover and wintering areas used by our study population. MATERIALS AnD METhODS Field procedures During 2016-2019, we monitored five breeding colonies of Barn Swallows in southwestern Spain (provinces of Badajoz and Seville): “Asesera” (38º39’n, 7º13’w), “La Alegría” (37º29’n, 6º11’w), “Las Coladas” (37º36’n, 6º14’w), “La Calera” (37º34’n, 6º13’w) and “Chaparro” (37º37’n, 6º14’w). In our study sites Barn Swallows breed in traditional farms surrounded by Mediterranean Dehesa, an extensive agrosilvopastoral management system derived from the Mediterranean forest ecosystem. The number of breeding pairs differed across the study sites and years, ranging from eight to 30 per colony. from february to June, we captured adult Barn Swallows with mist nets and marked each individual with both metallic and colour rings. we trapped birds each year until 90- 100% of the population was captured at each breeding site. Individuals were sexed from the presence or absence of brood patch, by the length of outermost tail feathers and also by observation of breeding behaviour (Møller, 1994). from May to June 2016, 2017 and 2018, we attached respectively 64, 59 and 55 geolocator devices to adult Barn Swallows using leg-loop harness (Rappole & Tipton, 1991; Bowlin et al., 2010). The tag model was ML6540 (Biotrack, wareham, UK). In 2016, every tag incorporated a 5mm light stalk and all harnesses were made of elastic material (Street Magic cord, pepperell Braiding Company). In this year, the weight of both tag and harness averaged 0.59g (3.34% of body weight) and the harness loop diameter was 26-31mm depending on body size. geologger stalks place the light sensor outside the body feathers, thus reducing shading and improving the quality of light records. however, as a drawback, the stalk increases air drag with associated negative effects on survival (Bowlin et al., 2010; Scandolara et al., 2014; Morganti et al., 2018). In order to reduce this negative effect (Costantini & Møller, 2013), in subsequent years we only deployed stalk-less tags (so called “flat geolocators”). In 2017, 33 harnesses were made of elastic material and 26 of non-elastic material (cotton thread obtained from mist net material). In that year, the weight of both tag and harness averaged 0.56g (3.24% of body weight), loop diameter of elastic harnesses was 30-33mm and that of non-elastic harnesses was 34-40mm. finally in 2018, all harnesses were made of elastic material (as in 2016) with a loop diameter of 30mm. In this year, the weight of both tag and harness averaged 0.57g (3.08% of body weight). To test whether return rates of tagged birds were lower than for untagged, 48, 40 and 42 colour-ringed swallows were considered as controls in 2016, 2017 and 2018 respectively. Control individuals were: (1) individuals that Ardeola 68(2), 2021, 335-354 LópEZ-CALDERón, C., MAgALLAnES, S., MARZAL, A.and BALBOnTín, J. 338 were captured under the same field procedure as the tagged birds (i.e. captured on the same day that tags were deployed within our breeding colonies) and (2) adult individuals that were captured across the breeding season (e.g. weeks before we started tagging) but that were observed staying in the breeding colonies until we finished tagging. Sample sizes were 64 and 66 individuals for control types 1 and 2 respectively. Light data analysis The geolocators measured light levels each minute and recorded the maximum light value every two minutes. Raw light data were corrected for clock drift with Decompressor (British Antarctic Survey). Twilight transitions were individually inspected with package BAStag (wotherspoon et al., 2016) of R (R Core Team, 2017). Transitions with light peaks or non-linear transitions that substantially disrupted twilight time were manually excluded from further analysis. On average ± SD, this process resulted in the rejection of 17.82 ± 11.67% of twilight events in each tag record (range: 3.30-51.37%; n = 38). we never edited non-natural twilights; we deleted them but did not fill these gaps with an expected twilight time. This latter restriction should lead to higher uncertainty in position estimates during periods of low quality light data but we were confident that this is the most conservative approach. A minimum dark period of five hours was considered, and the arbitrary threshold value defining twilight was set at 16 (the geolocators measured light levels from 0 to 64). Additional information regarding the quality of our raw light records is available in the Supplementary Material, Appendix 1. we followed the general guidelines for solar geolocation analyses (Lisovski et al., 2020) and we reconstructed migratory routes from light data with package FLightR (Rakhimberdiev et al., 2017) of R(R Core Team, 2017). This method provides refined accuracy and precision of calculated positions by using a hidden Markov chain model and a template fit observation model (Rakhimberdiev et al., 2015). Because geolocator units measure light levels with varying precision, FLightR finds the linear relationship between the observed light values (measured by the tag) and the theoretical light values (calculated from current sun angle). This calibration (i.e. “template fit”) can be estimated by recording light measurements at a known position. Thus, we adjusted the selection of calibration periods for each migratory track in order to maximise the number of days with high quality light records at a known position. we used one, two or three different calibration periods whose duration (e.g. two, three weeks) varied among tags depending on data quality. for migratory tracks with more than one calibration period, we accounted for the loss of transparency in the surface of the light sensor (implemented in FLightR). The use of buildings for nesting had a strong impact on the light profiles (Liechti et al., 2014; hobson et al., 2015), and thus the main calibration period was taken usually after nesting but before departure from the breeding colony. Depending on the light data available at a known position, the calibration of a given tag included periods when it was already attached to the bird and, additionally, when it was exposed to the sun on a rooftop. In case we included any rooftop calibration period, we checked that the variability of slopes between observed and expected light values was not constrained so much for subsequent simulations to perform adequately. If that was the case, we did not include the rooftop calibration period but instead used a calibration period at the wintering areas for an estimated stationary position. we used a spatialmask, constraining our model to consider states of residence only over land, with a buffer distance of 30km out to sea. following Ardeola 68(2), 2021, 335-354 SpAnISh BARn SwALLOw MIgRATIOn 339 Rakhimberdiev et al.(2015), we incorporated as priors (Bayesian marginal probabilities) 0.1 for probability of migratory behaviour and 300 ± 150km for distance covered between consecutive twilights. we defined the start date of a given migratory track as the first day of the first on-bird calibration period, or alternatively as the first day of clean light records after breeding. The end date of a given migratory track was defined as the first day in which twilight times dramatically changed after the wintering period. This could be done by visual inspection of light profiles with function preprocessLight of the BAStag package (the same function as in the TwGeos package). In this way, the start of breeding behaviour after migration could be identified as the appearance of frequent shading events (i.e. use of buildings for nesting). Because twilight times were dramatically disrupted after the start of breeding behaviour (see “BOX 1” in Lisovski et al., 2020), we excluded this period from the analysis and we assumed that the tagged bird had arrived at its breeding colony on the end date of the track (Liechti et al., 2014; hobson et al., 2015). After having calibrated light measurements for each tag, and having once defined the spatial mask as well as the Bayesian priors, FLightR discretizes the space and uses the particle filter to approximate the posterior distribution (Rakhimberdiev et al., 2015). At the start date of each track (“twilight i”), the particle filter generates a sample of 106 particles (i.e. points with positions). for each particle, a new position is calculated at “twilight i + 1”, according to the observation model as well as the movement model that discerns periods of residence and true migration. All new particles are then resampled according to their weight and continue to the next stage (“twilight i + 2”). The process repeats for each twilight until generating 106 possible migratory pathways that start and finish at the breeding colony. from these 106 Bayesian simulations, the median position at a given twilight synthetises the most probable location where the tagged individual could be found at that particular twilight. This position estimate matches locations given by high-precision gpS tags adequately (Rakhimberdiev et al., 2016). In addition, the credible intervals (in latitude and longitude) including 95% of particle positions at the given twilight represent the uncertainty associated with the position estimate of that particular twilight. we ran a total number of 218 trial simulations in which different settings of the analysis were changed to find the best performance for each migratory route (biologically and technically). Such settings include minor differences in prior data filtering, calibration periods and start/end dates. Some trial simulations were even repeated keeping the same settings, and median position estimates did not change substantially. Thus, on average, we repeated the FLigthR simulation six times for each track. In general, those tracks with lower light data quality required more trials to be run. Once a given simulation performed adequately, we accepted that result as the migratory route for that individual. After we reconstructed the migratory route for each individual, we defined five key dates across the migration cycle. we defined (1) “departure date from breeding colony” as the first day on which the tagged bird moved from the breeding colony to a different position and did not return. we defined (2) “onset of autumn migration” as the first day on which the tagged bird started to move southwards steadily. we defined (3) “arrival date at wintering area” as the first day on which the tagged bird stopped moving southwards steadily. Some birds arrived at their wintering areas by moving eastwards with constant latitude; in those cases we defined arrival at the wintering area as when the tagged bird stopped moving in that direction. we de- Ardeola 68(2), 2021, 335-354 LópEZ-CALDERón, C., MAgALLAnES, S., MARZAL, A.and BALBOnTín, J. 340 fined (4) “onset of spring migration” as the first day on which the tagged bird started to move northwards steadily (i.e. departure date from wintering area). Some birds left their wintering areas by moving westwards with constant latitude; in those cases we defined the onset of spring migration as when the bird started to move in that direction. The arrival date at the wintering area and the onset of spring migration that we estimated based on our definitions were virtually the same as these dates inferred with the function find.times.distribution of FLightR. finally, we defined (5) “arrival date at breeding colony” as the first day on which the tagged bird arrived at its breeding colony (based on the Markov chain simulation). The function plot_slopes_by_location of the FLightR package provides a complementary way to determine arrival date at the breeding colony by retrieving the day in which calibration slopes show little variation in time after the wintering period. we also used this function to further confirm arrival date, which was especially useful when the particular swallow started to use buildings some days after its actual arrival at the breeding area. In these cases, we could select these periods for an additional on-bird calibration period at a known location. for the purposes of other ongoing studies, we monitored exhaustively the reproductive performance of our swallows; for example, we inspected each nest weekly from february to July. we therefore knew the breeding date of each tagged individual. This information was very useful to check that the arrival date at the breeding colony estimated by solar geolocation must be earlier than the breeding date estimated by our field monitoring. with these five key dates, we could divide the migration cycle into four different time periods: i) “pre-migration period”, characterised by erratic short-distance movements around the breeding colonies (i.e. the period between key dates 1 and 2); ii) “autumn migration period”, characterised by long movements southwards (i.e. between key dates 2 and 3); iii) “wintering period”, characterised by relatively short-distance movements within Africa (i.e. between key dates 3 and 4); and iv) “spring migration period”, characterised by long movements northwards (i.e. between key dates 4 and 5). Limiñana et al. (2008) developed a standardised methodology to define the pre-migration period. nonetheless, because their approach was designed for gpS tracks that allow a precision of < 1km in position estimates, we established our own criteria to define this stage. we determined the wintering area for each tracked swallow as polygons generated by the function plot_util_distr from the FLightR package (Rakhimberdiev et al., 2017). This function calculates the accumulated probabilities of occurrence during a specified period, i.e. between the arrival and departure dates at wintering areas. for example, the polygons obtained with this function setting the parameter percentiles equal to 0.5, indicate the grid points that accumulated more than 50% of probability of occurrence during winter. These polygons were circles of ~85km diameter because FLightR discretizes the space to reduce computation time (Rakhimberdiev et al., 2015). In order to identify the stopover areas used during pre-migration and migration, we generated our own function in R(R Core Team, 2017). we first found the median position estimates in which the tagged bird stayed for at least three consecutive twilights. If the median position estimate of twilight “i” matched that of twilight “i + n” but there were missing twilights among them due to poor quality light data, we filtered this location only if the bird remained there for at least three consecutive twilights (we obtained the stopover duration from the first and last twilight at this location). Then, because any median position estimate may change across repeated Markov chain simu- Ardeola 68(2), 2021, 335-354 SpAnISh BARn SwALLOw MIgRATIOn 341 lations, we defined the stopover area as the error square generated by 95% credible intervals in latitude and longitude for that median position estimate. A given stopover site is indicated by several twilights with the same median position but with specific associated uncertainty estimates that can be different. Therefore, for a given stopover site, we defined the credible interval in latitude and longitude as the shortest among the twilights involved. To depict the wintering and stopover areas used by our whole population under study we rasterized, respectively, polygons with accumulated probabilities of occurrence during winter, and error squares of stopover locations. pixels within these polygons were coded as one, and pixels outside these polygons as zero, resulting in one binary map per individual. finally, for each stage of the migration cycle, we summed over each individual raster layer. All gIS operations were conducted in Rversion 3.3.1 (R Core Team, 2017) with packages raster (hijmans, 2016) and sp (pebesma & Bivand, 2005). RESULTS In 2017, we recaptured 21 out of 64 tagged birds (32.8%) and 27 out of 48 control birds (56.2%). In 2018, we recaptured eight out of 59 tagged birds (13.5%) and eight out of 40 control birds (20%). finally, in 2019, we recaptured 16 out of 55 tagged birds (29.0%) and 20 out of 42 control birds (47.6%). Thus, survival rates were lower for tagged birds than controls across all our study years, this difference being statistically significant for 2017 ( χ2= 6.15; df = 1; p= 0.01 ) and 2019 ( χ2= 4.01; df = 1; p= 0.045 ) , but not for 2018 ( χ2= 1.15; df = 1; p= 0.28 ) . five of the recovered tags did not provide any useful light data and two tagged individuals (excluded from the Chi-squared tests above) had lost their geolocator upon recapture. Thus, this study was based on a total sample size of 38 year-round migratory tracks. Eleven individuals were tagged in more than one year but only two of them survived the second migration episode. Therefore, we only obtained longitudinal data on migratory routes for two individuals. Supplementary Material, Appendix 2 presents an interactive map with the 38 migratory pathways. figure 1 shows a subsample of three migratory tracks: two of them indicating the usual pathways for this migration system (figure 1A-1C), and the other depicting a single individual that migrated north instead of south (figure 1B). Similar routes for both autumn and spring migrations were followed by 14 individuals, whereas 24 individuals used different routes across migrations. Among birds that changed migratory pathways between seasons, ten followed an anti-clockwise loop, five a clockwise loop and nine of them a double loop (i.e. a figure- of-eight). The pre-migration period Barn Swallows tagged with geolocators left their breeding colonies on average ± SD on July 28th ± 17.23 days (range: June 23rd- August 25th; n = 38) (figure 2). Up to 32 study individuals then spent some time wandering across southern Spain and northern Morocco (figure 3A), whereas only six individuals started their migration directly from the breeding colonies. The pre-migration period averaged 21.61 ± 17.09 days (range: 0-55 days; n = 38). Before starting migration, the studied individuals stopped over on average 2.84 ± 2.40 times (range: 0-8 stopovers; n = 38). Mean stopover duration per individual averaged 8.98 ± 10.68 days (range: 1-47 days; n = 30), whereas maximum stopover duration per individual was on average 14.65 ± 12.30 days (range: 1-47 days; n = 30). Ardeola 68(2), 2021, 335-354 LópEZ-CALDERón, C., MAgALLAnES, S., MARZAL, A.and BALBOnTín, J. 342 The post-nuptial or autumn migration period The tracked Barn Swallows started autumn migration on average ± SD on August 18th ± 4.34 days (range: August 10th-26th; n = 38), and the autumn migration period lasted on average 15.57 ± 5.11 days (range: 8-35 days; n = 37) (figure 2). The key dates were very different for the single individual that migrated to the north and is consequently excluded from the summary results hereafter. During the autumn migration, individuals usually stayed to refuel in northwestern Morocco and southern Mali (figure 3B). The studied individuals stopped over on average 1.38 ± 1.36 times (range: 0-5 stopovers; n = 37). Mean stopover duration per individual averaged 1.80 ± 0.89 days (range: 1-5 days; n = 25) and the maximum stopover duration per individual averaged 2.18 ± 1.24 days (range: 1-5 days; n = 25). The wintering period Barn Swallows arrived at their wintering areas on average ± SD on September 3rd ± 5.90 days (range: August 21st-September 18th; n = 37), and the wintering period averaged Ardeola 68(2), 2021, 335-354 SpAnISh BARn SwALLOw MIgRATIOn 343 fIg. 1.—Migratory tracks of three different Barn Swallows tagged in 2016, 2017 and 2018, respectively. Subfigures A, B and C, represent birds ringed 2L12365, 2L12590 and 1y83581, respectively. Dots indicate median position estimates given by the Markov chain simulation run with FLightR (Rakhimberdiev et al., 2017). Dot colours indicate the month of the year (see legend). Credible intervals of median positions are shown in Supplementary Material, Appendix 2. [Rutas migratorias de tres golondrinas marcadas en 2016, 2017 y 2018, respectivamente. Subfiguras A, B y C representan los individuos anillados 2L12365, 2L12590 y 1Y83581, respectivamente. Los puntos indican las posiciones medianas estimadas en la simulación desarrollada por FLightR (Rakhimberdiev et al., 2017). El color de los puntos indica el mes del año (leyenda). Los intervalos de confianza de las posiciones medianas se muestran en el Material Suplementario, Anexo 2.] we found that during the spring migration all but two of the 38 tracked individuals stopped over to refuel, and they did so up to seven times and for a maximum period of 15 days. They stopped-over to refuel 3.3 times on average during spring migration and stayed for 2.5 days at these stopover sites. In comparison, Sand Martins Riparia riparia stopped over 5-6 times for 1.5 days on average and house Martins Delichon urbicum stopped over 3-7 times for one day on average (Szép et al., 2017). The longer migration distances of these martins may require more frequent and briefer stop-overs than those of Barn Swallows from southern Spain, perhaps because longer flights demand more energy resources and permit less delay. The main stopover site, used by 11 of the 38 tagged birds during their spring migration, ranged from Senegal to southwestern Mauritania. Senegal offers a range of African biomes from savannah to the Sahel, and hence Barn Swallows can stopover here to gain sufficient fat reserves for crossing the Sahara desert (Rubolini et al., 2002; Briedis et al., 2018). The identified stopover areas in Mauritania correspond to the Sahel but also to certain oases such as Lake gabbou, Lake Aleg and riparian corridors. Barn Swallows may use these Mauritanian oases as temporary refuges to recover from fatigue while crossing the desert as well as for foraging areas to increase fat loads for the rest of the journey (Maggini & Bairlein, 2011; Arizaga et al., 2013). An exception to the rule: an individual with reverse migration A single individual did not migrate south to Africa (figure 1B). Instead, it travelled north in autumn until it reached England. To the best of our knowledge, this reverse migration has not been reported previously for any other palearctic Barn Swallow. Although this was unexpected, it is possible that a small fraction of a given population could perform reverse migration route and still survive the winter in northern latitudes. for instance, two Barn Swallows were sighted along the coast of Dorset (UK) during December 2017 (http://dorsetbirds.blogspot.com/2017/12/; accessed 21 January 2018), which indicates that migrating individuals can use this region in winter. The wintering quarters used by this individual in northwestern Spain were very close to a known winter roost of Barn Swallows in Aveiro, northern portugal (van nus & neto, 2017), where environmental conditions are clearly good enough for Barn Swallows to overwinter. One out of 38 constitutes 2.6% of the tagged population, which might explain the observations of Barn Swallows in England and northern portugal during winter. It is nonetheless possible that climate change could be favouring the persistence of enough insects during winter (Bale & hayward, 2010) to allow the survival of individuals with rare migration phenotypes. Such extraordinary migrations may be driven by mutations in specific genes or by epigenetic mechanisms (reviewed in Merlin & Liedvogel, 2019). Thus, future studies should try to identify such genes or epigenetic mechanisms that could explain the variability in migratory phenotypes of Barn Swallows. COnCLUSIOnS In summary, we have extensively deployed light-level geolocators to investigate the migration system of Barn Swallows breeding in southwestern Spain. Such information relates to the phenology of migration and the geographical distribution of the nonbreeding habitats. Solar geolocation is starting to elucidate what have been long-standing gaps in ornithological knowledge. Ardeola 68(2), 2021, 335-354 LópEZ-CALDERón, C., MAgALLAnES, S., MARZAL, A.and BALBOnTín, J. 350 ACKnOwLEDgEMEnTS.—This study was funded by research projects of the Junta de Andalucía (p12-RnM-2144), Ministerio de Economía y Competitividad (CgL2015-646509) and Junta de Extremadura (IB16121). The first author was supported by an operating grant from the Junta de Andalucía (p12-RnM-2144). we thank Juan Sangran and his family, Borja Lora and his family, plácido and the Marzal family for allowing us to work on their properties. Thanks also to Martín, Manuel, Antonio, fernando, Javier, gabriel, yosi, Sebastián, Juana and other staff that work on the farms. we also thank Marina Bollo, Sergio Reinoso, Luz garcía and María Isabel Reviriego for helping in fieldwork and data base management. finally, we especially thank Álvaro Merino Alejandre for programing the interactive map showed as Supplementary Material, Appendix 2. Methods were evaluated and approved by the Aranzadi Ringing Scheme, Sociedad Española de Ornitología (SEO/BirdLife), Junta de Andalucía and Junta de Extremadura. All efforts were made to avoid animal distress and to minimise handling time according to “guidelines to the Use of wild Birds in Research” (J. fair, E. paul & J. Jones (Eds.), 2010. washington D.C.: Ornithological Council). All the experiments complied with the current laws of Spain, where the field sampling was performed. AUThOR COnTRIBUTIOnS.—Study conception: CLC and JBA; Methodology: CLC and JBA; Computation: CLC; formal analysis: CLC; Data collection: CLC, SMA, AMR and JBA; Resources: AMR and JBA; Data curation: CLC; writing the initial draft: CLC; Critical review, commentary or revision: CLC, SMA, AMR and JBA; Data presentation: CLC; Supervision: AMR and JBA; project administration: JBA; funding acquisition: AMR and JBA. REfEREnCES Åkesson, S., Klaassen, R., holmgren, J., fox, J.w. & hedenström, A. (2012). Migration Routes and Strategies in a highly Aerial Migrant, the Common Swift Apus apus, Revealed by Light- Level geolocators. PLoS ONE, 7: e41195. Allen, J. (2016). Land cover classification (1 year). nASA Earth Observations http://neo.sci.gsfc. nasa.gov/view.php?datasetId=MCD12C1_T1. Ambrosini, R., Møller, A.p. & Saino, n. (2009). A quantitative measure of migratory connectivity. Journal of Theoretical Biology, 257: 203-211. Arizaga, J., Maggini, I., hama, f., Crespo, A. & gargallo, g. (2013). Site - and species-specific fuel load of European-Afrotropical passerines on arrival at three oases of southeast Morocco during spring migration. Bird Study, 60: 11-21. Arizaga, J., willemoes, M., Unamuno, E., Unamuno, J.M. & Thorup, K. (2015). following year-round movements in Barn Swallows using geolocators: could breeding pairs remain together during the winter? Bird Study, 62: 141- 145. Bale, J.S. & hayward, S.A.L. (2010). Insect overwintering in a changing climate. Journal of Experimental Biology, 213: 980-994. Berthold, p. (1993). Bird migration. A general survey. Oxford University press, Oxford. Bowlin, M.S., henningsson, p., Muijres, f.T., vleugels, R.h., Liechti, f. & hedenström, A. (2010). The effects of geolocator drag and weight on the flight ranges of small migrants. Methods in Ecology and Evolution, 1: 398-402. Briedis, M., Kurlavičius, p., Mackevičienė, R., vaišvilienė, R. & hahn, S. (2018). Loop migration, induced by seasonally different flyway use, in northern European Barn Swallows. Journal of Ornithology, 159: 885-891. Brlík, v. et al. (2020). weak effects of geolocators on small birds: A meta-analysis controlled for phylogeny and publication bias. Journal of Animal Ecology, 89: 207-220. Costantini, D. & Møller, A.p. (2013). A metaanalysis of the effects of geolocator application on birds. Current Zoology, 59: 697-706. Ekstrom, p.A. (2004). An advance in geolocation by light. Memoirs of National Institute of Polar Research, Special Issue, 58: 210-226. Ekstrom, p. (2007). Error measures for templatefit geolocation based on light. Deep Sea Research Part II: Topical Studies in Oceanography, 54: 392-403. Ardeola 68(2), 2021, 335-354 SpAnISh BARn SwALLOw MIgRATIOn 351 faaborg, J. et al. (2010). Conserving migratory land birds in the new world: Do we know enough? Ecological Applications, 20: 398-418. finch, T., Butler, S.J., franco, A.M.A. & Cresswell, w. (2017). Low migratory connectivity is common in long-distance migrant birds. Journal of Animal Ecology, 86: 662-673. gilroy, J.J., gill, J.A., Butchart, S.h.M., Jones, v.R. & franco, A.M.A. (2016). Migratory diversity predicts population declines in birds. Ecology Letters, 19: 308-317. hahn, S., Bauer, S. & Liechti, f. (2009). The natural link between Europe and Africa-2.1 billion birds on migration. Oikos, 118: 624- 626. hallworth, M.T. & Marra, p.p. (2015). Miniaturized gpS Tags Identify non-breeding Territories of a Small Breeding Migratory Songbird. Scientific Reports, 5: 11069. hama, f., gargallo, g., Benhoussa, A., Zerdouk, S. & Idrissi, h.R. (2013). Autumn body condition of palaearctic trans-Saharan migrant passerines at an oasis in southeast Morocco. Ringing and Migration, 28: 77-84. harrison, X.A., Blount, J.D., Inger, R., norris, D.R. & Bearhop, S. (2011). Carry-over effects as drivers of fitness differences in animals. Journal of Animal Ecology, 80: 4-18. hijmans, R.J. (2016). Raster: Geographic Data Analysis and Modeling. R package version 2.5-8. https://CRAn.R-project.org/package= raster. hobson, K.A., Kardynal, K.J., van wilgenburg, S.L., Albrecht, g., Salvadori, A., Cadman, M.D., Liechti, f. & fox, J.w. (2015). A Continentwide Migratory Divide in north American Breeding Barn Swallows (Hirundo rustica). PLoS ONE, 10: e0129340. Kays, R., Crofoot, M.C., Jetz, w. & wikelski, M. (2015). Terrestrial animal tracking as an eye on life and planet. Science, 348: aaa2478. Klvaňa, p., Cepák, J., Munclinger, p., Michálková, R., Tomášek, O. & Albrecht, T. (2018). Around the Mediterranean: An extreme example of loop migration in a long-distance migratory passerine. Journal of Avian Biology, 49: 01595. Liechti, f., Scandolara, C., Rubolini, D., Ambrosi- ni, R., Korner-nievergelt, f., hahn, S., Lardelli, R., Romano, M., Caprioli, M., Romano, A., Sicurella, B. & Saino, n. (2014). Timing of migration and residence areas during the nonbreeding period of barn swallows Hirundo rustica in relation to sex and population. Journal of Avian Biology, 46: 254-265. Limiñana, R., Soutullo, A., López-López, p. & Urios, v. (2008). pre-migratory movements of adult Montagu’s harriers Circus pygargus. Ardea, 96: 81-90. Lisovski, S., Bauer, S., Briedis, M., Davidson, S.C., Adams, K.L.D., hallworth, M.T., Karagicheva, J., Meier, C.M., Merkel, B., Ouwehand, J., pedersen, L., Rakhimberdiev, E., Charron, A.R., Seavy, n.E., Sumner, M.D., Taylor, C.M., wotherspoon, S.J. & Bridge, E.S. (2020). Light-level geolocator analyses: A user’s guide. Journal of Animal Ecology, 89: 221-236. Lope, f. de (1980). La invernada de las golondrinas Comunes ibéricas en África. II Reunión Iberoamericana para la Conservación y Zoología de Vertebrados, Cáceres. Maggini, I. & Bairlein, f. (2011). Body condition and stopover of trans-Saharan spring migrant passerines caught at a site in southern Morocco. Ringing and Migration, 26: 31-37. McKinnon, E.A., fraser, K.C. & Stutchbury, B.J.M. (2013). new discoveries in landbird migration using geolocators, and a flight plan for the future. The Auk, 130: 211-222. Merlin, C. & Liedvogel, M. (2019). The genetics and epigenetics of animal migration and orientation: birds, butterflies and beyond. Journal of Experimental Biology, 222: 191890. Møller, A.p. (1994). Sexual selection and the barn swallow. Oxford University press. Morganti, M., Rubolini, D., Åkesson, S., Bermejo, A., de la puente, J., Lardelli, R., Liechti, f., Boano, g., Tomassetto, E., ferri, M., Caffi, M., Saino, n. & Ambrosini, R. (2018). Effect of light-level geolocators on apparent survival of two highly aerial swift species. Journal of Avian Biology: e01521. nebel, S., Mills, A., McCracken, J.D. & Taylor, p.D. (2010). Declines of aerial insectivores in north America follow a geographic gradient. Avian Conservation and Ecology, 5: 1. newton, I. (2008). The migration ecology of birds. Academic press, London, UK. Ardeola 68(2), 2021, 335-354 LópEZ-CALDERón, C., MAgALLAnES, S., MARZAL, A.and BALBOnTín, J. 352 norris, D.R. & Marra, p.p. (2007). Seasonal interactions, habitat quality, and population dynamics in migratory birds. The Condor, 109: 535. patchett, R., finch, T. & Cresswell, w. (2018). population consequences of migratory variability differ between flyways. Current Biology, 28: R340-R341. pebesma, E.J. & Bivand, R.S. (2005). Classes and methods for spatial data in R. R news 5 (2). https://cran.r-project.org/doc/Rnews/. R Core Team (2017). The R Project for Statistical Computing. R foundation for Statistical Computing. http://www.R-project.org. Rakhimberdiev, E., winkler, D.w., Bridge, E., Seavy, n.E., Sheldon, D., piersma, T. & Saveliev, A. (2015). A hidden Markov model for reconstructing animal paths from solar geolocation loggers using templates for light intensity. Movement Ecology, 3: 25. Rakhimberdiev, E., Senner, n.R., verhoeven, M.A., winkler, D.w., Bouten, w. & piersma, T. (2016). Comparing inferences of solar geolocation data against high-precision gpS data: annual movements of a double-tagged blacktailed godwit. Journal of Avian Biology, 47: 589-596. Rakhimberdiev, E., Saveliev, A., piersma, T. & Karagicheva, J. (2017). fLightR: an R package for reconstructing animal paths from solar geolocation loggers. Methods in Ecology and Evolution, 8: 1482-1487. Rappole, J.h. & Tipton, A.R. (1991). new harness design for attachment of radio transmitters to small passerines. Journal of Field Ornithology, 62: 335-337. Rubolini, D., pastor, A.g., pilastro, A. & Spina, f. (2002). Ecological barriers shaping fuel stores in barn swallows Hirundo rustica following the central and western Mediterranean flyways. Journal of Avian Biology, 33: 15-22. Sanderson, f.J., Donald, p.f., pain, D.J., Burfield, I.J. & van Bommel, f.p. (2006). Long-term population declines in Afro-palearctic migrant birds. Biological Conservation, 131: 93-105. Scandolara, C., Rubolini, D., Ambrosini, R., Caprioli, M., hahn, S., Liechti, f., Romano, A., Romano, M., Sicurella, B. & Saino, n. (2014). Impact of miniaturized geolocators on barn swallow Hirundo rustica fitness traits. Journal of Avian Biology, 45: 417-423. Schaub, M. & Jenni, L. (2000). Body mass of six long distance migrant passerine species along the autumn migration route. Journal of Ornithology, 141: 441-460. Stutchbury, B.J.M., Tarof, S.A., Done, T., gow, E., Kramer, p.M., Tautin, J., fox, J.w. & Afanasyev, v. (2009). Tracking Long-Distance Songbird Migration by Using geolocators. Science, 323: 896-896. Szép, T., Liechti, f., nagy, K., nagy, Z. & hahn, S. (2017). Discovering the migration and nonbreeding areas of sand martins and house martins breeding in the pannonian basin (centraleastern Europe). Journal of Avian Biology, 48: 114-122. van nus, T. & neto, J.M. (2017). Urban Roost of wintering Barn Swallows Hirundo rustica in Aveiro, portugal. Ardea, 105: 1-6. walther, B. & pirsig, L. (2017). Determining conservation priority areas for palearctic passerine migrant birds in sub-Saharan Africa. Avian Conservation and Ecology, 12: 1. Ardeola 68(2), 2021, 335-354 SpAnISh BARn SwALLOw MIgRATIOn 353 ©Cosme López-Calderón wotherspoon, S., Sumner, M. & Lisovski, S. (2016). BAStag: Basic data processing for light based geolocation archival tags. R package version 0.1.3. SUppLEMEnTARy ELECTROnIC MATERIAL Additional supporting information may be found in the online version of this article. See the volume 68(2) on www.ardeola.org Appendix 1. Additional information on light data quality. [Información adicional sobre la calidad de los registros de luz.] Appendix 2. Interactive map with 38 migratory routes (it must be opened with google Chrome). [Mapa interactivo con 38 rutas migratorias (debe ser abierto con Google Chrome).] Submitted: May 05, 2020 Minor revision: July 27, 2020 Second version arrived: August 08, 2020 Minor revision: September 23, 2020 Third version arrived: October 05, 2020 Accepted: October 06, 2020 Editor: pascual López Ardeola 68(2), 2021, 335-354 LópEZ-CALDERón, C., MAgALLAnES, S., MARZAL, A.and BALBOnTín, J. 354