Breeding latitude predicts timing but not rate of spring migration in a widespread migratory bird in South America

Abstract Identifying the processes that determine avian migratory strategies in different environmental contexts is imperative to understanding the constraints to survival and reproduction faced by migratory birds across the planet. We compared the spring migration strategies of Fork‐tailed Flycatchers (Tyrannus s. savana) that breed at south‐temperate latitudes (i.e., austral migrants) vs. tropical latitudes (i.e., intratropical migrants) in South America. We hypothesized that austral migrant flycatchers are more time‐selected than intratropical migrants during spring migration. As such, we predicted that austral migrants, which migrate further than intratropical migrants, will migrate at a faster rate and that the rate of migration for austral migrants will be positively correlated with the onset of spring migration. We attached light‐level geolocators to Fork‐tailed Flycatchers at two tropical breeding sites in Brazil and at two south‐temperate breeding sites in Argentina and tracked their movements until the following breeding season. Of 286 geolocators that were deployed, 37 were recovered ~1 year later, of which 28 provided useable data. Rate of spring migration did not differ significantly between the two groups, and only at one site was there a significantly positive relationship between date of initiation of spring migration and arrival date. This represents the first comparison of individual migratory strategies among conspecific passerines breeding at tropical vs. temperate latitudes and suggests that austral migrant Fork‐tailed Flycatchers in South America are not more time‐selected on spring migration than intratropical migrant conspecifics. Low sample sizes could have diminished our power to detect differences (e.g., between sexes), such that further research into the mechanisms underpinning migratory strategies in this poorly understood system is necessary.

ation is vital for developing a basic understanding of the evolution and regulation of migration, as well as for evaluating the fitness consequences of employing a given migratory strategy, since processes that occur in one season can influence the survival and reproduction of an individual in subsequent seasons (i.e., through carryover effects, reviewed by Harrison, Blount, Inger, Norris, & Bearhop, 2011).
Migratory birds that initiate reproduction late due to delayed arrival at the breeding site risk lower reproductive success, due to a steep decline in number of young fledged as the breeding season progresses (e.g., Murphy, 1986;Verboven & Visser, 1998). For example, in Hoopoes (Upupa epops), the duration of spring migration is negatively related to breeding territory quality and number of fledglings produced (van Wijk, Schaub, & Bauer, 2017). In contrast, an earlier arrival at the breeding site can result in increased reproductive success by increasing the probability of successfully raising a brood of nestlings or by allowing an individual to attempt multiple broods (Smith & Moore, 2005). Because spring migration strategy and timing of arrival on breeding grounds can have profound consequences on a migrant's ability to successfully reproduce in a given year (Kokko, 1999;McKinnon, Stanley, & Stutchbury, 2015;Smith & Moore, 2005;Visser, Holleman, & Gienapp, 2006), arrival timing is presumably under strong selection (Smith & Moore, 2005), especially in males, which must often compete for breeding territories and mates (Morbey & Ydenberg, 2001;Tøttrup & Thorup, 2008).
Despite the importance of migratory timing on reproductive success, we lack a detailed understanding of the selective pressures molding bird migration strategies across taxa and, except for a handful of model species, between conspecific populations. Hundreds of migratory bird species spend their entire lives within tropical and south-temperate latitudes, and their migratory strategies are poorly known. Intratropical migratory birds breed, migrate, and overwinter entirely within the tropics (including Mesoamerica and the Caribbean, South America, Africa, and Asia). Austral migrants breed at southern temperate latitudes of Africa, South America, and Australia and overwinter closer to the Equator (reviewed by Chesser (1994), Dingle (2008), Faaborg et al. (2010)). Research on these poorly understood systems promises novel insights into how much variation in migratory strategies exists across regions and taxa, and ultimately the mechanisms that underpin the strategies employed in these different systems.
Two lines of evidence suggest that intratropical migrants should employ a different spring migration strategy than austral migrants that breed at south-temperate latitudes. First, birds that breed at tropical latitudes experience a different ecological context than those that breed at temperate latitudes. Compared to seasonality at temperate latitudes, seasons in the tropics are primarily defined by variation in rainfall (e.g., Gottsberger & Silberbauer-Gottsberger, 2006;Oliveira & Marquis, 2002;Wikelski, Hau, & Wingfield, 2000), which in turn drives timing of leafing, flowering, and fruiting of tropical plants (e.g., Araujo, Vieira-Filho, Barbosa, Diniz-Filho, & Silva, 2017;Myneni et al., 2007;Patrícia et al., 2000), and consequently the abundance of arthropods (Amorim, DeÁvila, Camargo, Vieira, & Oliveira, 2009;Cotton, 2007;Develey & Peres, 2000;Jahn et al., 2010;Pinheiro, Diniz, Coelho, & Bandeira, 2002). Such seasonality in precipitation is more unpredictable between years than between the two groups, and only at one site was there a significantly positive relationship between date of initiation of spring migration and arrival date.

5.
This represents the first comparison of individual migratory strategies among conspecific passerines breeding at tropical vs. temperate latitudes and suggests that austral migrant Fork-tailed Flycatchers in South America are not more time-selected on spring migration than intratropical migrant conspecifics. Low sample sizes could have diminished our power to detect differences (e.g., between sexes), such that further research into the mechanisms underpinning migratory strategies in this poorly understood system is necessary.

K E Y W O R D S
Argentina, Brazil, cerrado, life history, light-level geolocator, Pampas is temperature (Lisovski, Ramenofsky, & Wingfield, 2017). In the Southern Hemisphere, the degree of seasonality (i.e., predictability and amplitude) peaks at south-temperate latitudes (i.e., at ~35°S; Lisovski et al., 2017), such that austral migrants experience more predictable seasonality than intratropical migrants. Second, optimal migration theory postulates that birds that migrate longer distances should be more time-selected on migration (Alerstam & Lindström, 1990;Hedenström, 2008). Thus, because austral migrants that overwinter in the tropics and breed at south-temperate latitudes breed at seasonally more predictable sites and migrate longer distances than intratropical migrants, austral migrants should be more timeselected on spring migration than intratropical migrants.
Our objective was to compare migratory strategies of birds that migrate different distances in South America. To do so, we deployed light-level geolocators on migratory Fork-tailed Flycatchers (Tyrannus s. savana) in South America. This subspecies breeds from central Brazil to central Argentina (Mobley, 2004), with Brazilian populations breeding at tropical latitudes and overwintering in northern South America (Jahn, Giraldo, et al., 2016;Jahn, Seavy, et al., 2016); they are thus intratropical migrants. Populations that breed at south-temperate latitudes in Argentina also overwinter in northern South America (Jahn et al., 2013) and are thus Neotropical austral migrants (Cueto & Jahn, 2008). Fall migration patterns (Jahn, Giraldo, et al., 2016;Jahn, Seavy, et al., 2016;Jahn et al., 2013), tracking of environmental conditions (MacPherson et al., 2018), and, for one site in Brazil, timing of arrival of males versus females (Bejarano & Jahn, 2018) have been studied in Fork-tailed Flycatchers, but little is known about spring migration strategies.
Because Fork-tailed Flycatchers that breed at south-temperate latitudes (hereafter, "austral migrants") migrate a longer distance in spring than those that breed at tropical latitudes (hereafter, "intratropical migrants"), we hypothesize that austral migrants will employ a more time-selected spring migration strategy than intratropical migrants. We predict that, compared to intratropical migrants, austral migrants on spring migration should exhibit: (a) an overall greater migration rate and (b) a spring migration rate that is positively related to the date of departure on spring migration. We also evaluate stopover duration of intratropical versus austral migrants. Optimal migration theory predicts that time-selected migrants should minimize time spent on stopover when fuel deposition rates are high (Alerstam & Hedenström, 1998;Lindström & Alerstam, 1992), continuing to migrate upon reaching optimal fuel load (Alerstam & Lindström, 1990).
We make no predictions regarding stopover duration of Fork-tailed Flycatchers, since we have no information on their refueling rates during stopover.

| MATERIAL S AND ME THODS
We captured flycatchers during the breeding season (September to December in Brazil; Marini, Lobo, Lopes, Franca, & Paiva, 2009; October to January in Argentina; Jahn et al., 2014), at four sites:     Jahn, Seavy, et al., 2016;Pyle, 1997). Wing chord, tarsus length, and tail length were collected following methods in Ralph, Guepel, Pyle, Martin, and DeSante (1993), and body mass was measured to the nearest 0.1 g using a portable digital scale (Ohaus LS 200). Finally, flycatchers were outfitted with a light-level geolocator using a leg-loop harness (Rappole & Tipton, 1991)  We deployed 55 geolocators at DF and 58 geolocators at EEI (Brazil). We recovered 4 (7%) geolocators at DF, and recovered 8 (14%) at EEI, all of which had usable data (two geolocators had been deployed on the same individual in different seasons at EEI, see explanation below). We deployed 103 geolocators at RED and 70 geolocators at RPL (Argentina). We recovered 16 (16%) at RED, of which nine had usable data, and we recovered 9 (13%) at RPL, of which eight had usable data. This rate of geolocator recovery is within the range of that reported in studies of other migratory birds (Bridge et al., 2013). None of the recaptured flycatchers exhibited any sign of injury from the geolocator or harness. We were not able to recapture many flycatchers because they had become net shy and much less responsive to predator models and playbacks, especially in Brazil.
Overall, we analyzed data from one female and 10 male intratropical migrant flycatchers from Brazil, and six female and 11 male austral migrant flycatchers from Argentina (Table 1).

| Data analysis
Raw light-level data were transformed into geographic loca- per run. We treated the first two runs as a burn-in period while summarizing location estimates between each run. We used the resulting median daily location to initialize each subsequent run. We kept every second iteration from the posterior distribution, from which we drew our geographic inference.
We We define the beginning of the winter period as the initiation of first stationary period in fall that was at least 30 days duration. Flycatchers exhibited multiple stationary periods throughout the nonbreeding season (Table 1), often moving across longitudes ( Figure 2), such that changes in longitude were not useful to delimit wintering versus migratory periods. Therefore, to estimate the initiation and termination of spring migration, we used large changes in the duration of stationary periods. We define termination of the winter period and initiation of spring migration as the end of the first stationary period that was at least 30 days long and that was not followed by a stationary period of 30 days or more. We define termination of spring migration as the first date the bird was within the 95% credible interval of the breeding site longitude.

| RE SULTS
We found no significant difference between males versus females from Argentinian sites in the rate, distance and duration of spring migration, or in the duration of time spent on stopover, date of departure, or date of arrival at the breeding site (p > 0.05 in all cases; Table 1). Thus, we combined data from both sexes for further analyses, except for analysis of variation in arrival date at the breeding site. although most intratropical migrants used only one (Table 1).

| General patterns
Initially, the spring migration route for both intratropical and austral migrants crossed central Amazonia (Figure 2). Intratropical migrants generally arrived soon thereafter at the breeding site, whereas most austral migrants passed through Bolivia or western Brazil and eventually into northern Argentina and Paraguay, before arriving back at the breeding site ( Figure 2).

| Distance and duration of spring migration
Spring migration distance was significantly different between sites (Table 2), with flycatchers from both sites in Brazil migrating a significantly shorter distance than those from sites in Argentina (Table 2 and Figure 3). However, there was no significant difference in the spring migration distance between flycatchers from sites in Brazil or between those from sites in Argentina (Table 2 and Figure 3).
Overall, austral migrants travelled on average ~4,438 km (±578.6) on spring migration at a mean rate of ~134 km/day (±39.6), whereas intratropical migrants migrated on average 3,355 km (±388.9) in spring at a mean rate of 144 km/day (±25.7; Table 1, Figure 3).  TA B L E 2 Results of linear model output of migration duration, distance, and arrival and initiation dates of spring migration of Fork-tailed Flycatchers in South America migrants spent a mean of ~24 days on migration (±5.0; Table 1 and Figure 3). There was no significant effect of wing chord (LM:

| Timing and rate of spring migration
Date of departure on spring migration was significantly different between sites, with flycatchers migrating to sites in Argentina departing significantly later than flycatchers migrating to sites in Brazil, and no significant difference in departure date between flycatchers migrating to Brazilian sites or between flycatchers migrating to Argentinian sites (Tables 1 and 2). Likewise, arrival date at the breeding site was significantly different between sites (Tables   1 and 2), with flycatchers from sites in Argentina arriving significantly later than flycatchers migrating to sites in Brazil, and flycatchers arriving at EEI, Brazil, significantly later than at DF, Brazil (Tables 1 and 2). However, there was no significant difference in arrival date between flycatchers migrating to Argentinian sites (Tables 1 and 2). Overall, intratropical migrants initiated spring migration in early August and arrived back at their breeding sites in late August. In contrast, austral migrants initiated spring migration in early September and arrived back at their breeding sites in early October (Table 1). Thus, intratropical migrants initiated spring migration on average 35 days earlier than austral migrants and arrived earlier at the breeding site than austral migrants (on average 45 days earlier; Table 1).
The daily spring migration rate was not significantly different among flycatchers migrating to the four sites (LM: F = 1.19, df = 3, p = 0.338).

| Seasonal effects on rate and timing of spring migration
There was a positive but not significant relationship between date of initiation and arrival date of spring migration for flycatchers at all sites, except EEI, Brazil (r = 0.81, t = 3.12, df = 5, p = 0.026; Figure 4), and there was a positive but not significant relationship between rate of spring migration and the date of initiation for flycatchers at all sites ( Figure 5).

| D ISCUSS I ON
Overall, we found little support for the hypothesis that austral migrants are more time-selected on spring migration than intratropical migrant conspecifics. Austral migrant flycatchers did not migrate at a significantly higher rate than conspecifics breeding at tropical latitudes, and we did not find a significantly positive relationship between date of initiation of spring migration and the rate of spring migration, a relationship that has been found in other  (Bravo et al., 2017).
Flycatchers from across all sites spent a similar amount of time on stopovers. In other migratory systems, spring migrants with larger fuel reserves have been found to depart earlier from stopover sites compared to lean birds (e.g., Goymann, Spina, Ferri, & Fusani, 2010), although a variety of factors ultimately affect an individual's timing of departure from a given stopover site (Schmaljohann & Eikenaar, 2017). Given that we have little information on the stopover ecology of either intratropical or austral migrants in South America, future research on flycatcher migration would benefit from stopover ecology research (e.g., refueling rates en route).
In summary, we found little evidence that austral migrant flycatchers are more time-selected than intratropical migrant conspecifics, which could be affected by potentially high error rates associated with detecting arrival times around the spring equinox, as well as the effect of selection pressures that we did not measure, such as sex-specific selection pressures and availability of suitable stopover habitat. Primary productivity at tropical and south-temperate latitudes can be highly variable between years (e.g., Goetz, Prince, Small, & Gleason, 2000;Nobre et al., 2006) (MacPherson et al., 2018), arriving at the wintering grounds in northern South America at the beginning of the wet season, which peaks in July and August (Poveda, Waylen, & Pulwarty, 2006) and where Fork-tailed Flycatchers undergo an annual flight feather molt (Jahn, Giraldo, et al., 2016;Jahn, Seavy, et al., 2016). Winter represents a critical period during which they must properly time flight feather molt prior to spring migration, since Fork-tailed Flycatchers generally avoid molting and migrating simultaneously (Jahn et al., 2017). Given that northern South America is susceptible to notably lower rainfall levels in some years (i.e., during the "El Niño" phase of the El Niño/La Niña climatic cycle, Poveda et al., 2006), understanding the relationship between interannual variation in food resource availability, which is key to completing feather molt (Jahn, Giraldo, et al., 2016;Jahn, Seavy, et al., 2016), and the timing of events in the flycatcher's annual cycle, will provide novel insights into the constraints molding the annual cycle and population dynamics of this and other species migrating within South America.
Further research on intratropical bird migration promises novel insights into how bird migration is molded by different environmental conditions. For example, early arrival at the breeding site by intratropical migrant Fork-tailed Flycatchers has been shown to incur reproductive benefits (Bejarano & Jahn, 2018).
Research on the extrinsic (e.g., food resource availability) and To the best of our knowledge, this is the first comparison of individual migratory strategies between conspecific passerines breeding at tropical versus temperate latitudes. We still lack a conceptual framework on the full spectrum of ecological and evolutionary processes that shape avian migratory strategies, such as how a bird's endogenous migration program is affected by envi-  Hockey, 2005). Thus, gaining a foothold on the proximate and ultimate drivers of bird migration in the tropics and south-temperate latitudes will require a multidisciplinary, long-term, and taxonomically broad approach. The recent advent of novel analytical techniques and miniaturized tracking technologies, such as loggers that provide combined activity and location data (Bäckman et al., 2017;Liechti et al., 2018), provides the tools necessary to employ such an approach, and ultimately shed new light on how songbirds are able to overcome the multiple challenges facing their annual spring journeys. In turn, such information will be valuable for developing effective conservation plans for migratory birds on a planet undergoing rapid habitat and climatic changes.

ACK N OWLED G M ENTS
We dedicate this paper to the memory of the late Kimberly G.
Smith, who inspired much of our early work. We thank M. Briedis and an anonymous reviewer for numerous comments that greatly improved the manuscript. We also thank numerous research assistants without whom this project would not have been possible, and to Bertrando Campos for the picture of the Fork-tailed Flycatcher.
We are grateful to the Fundación Elsa Shaw de Pearson for logistical support and to private landowners for access to their lands.
This research was funded by the National Geographic Society (Nos.

CO N FLI C T O F I NTE R E S T
None declared.

AUTH O R ' S CO NTR I B UTI O N S
A

DATA ACCE SS I B I LIT Y
The data used in this study (raw light-level data, geographic positions derived from light levels, and morphological data) are available on Movebank (movebank.org, study name: "Migratory patterns of