Linking the wintering and breeding grounds of warblers along the Pacific Flyway

Abstract Long‐distance migration is a behavior that is exhibited by many animal groups. The evolution of novel migration routes can play an important role in range expansions, ecological interactions, and speciation. New migration routes may evolve in response to selection in favor of reducing distance between breeding and wintering areas, or avoiding navigational barriers. Many migratory changes are likely to evolve gradually and are therefore difficult to study. Here, we attempt to connect breeding and wintering populations of myrtle warblers (Setophaga coronata coronata) to better understand the possible evolution of distinct migration routes within this species. Myrtle warblers, unlike most other warblers with breeding ranges primarily in eastern North America, have two disjunct overwintering concentrations—one in the southeastern USA and one along the Pacific Coast—and presumably distinct routes to‐and‐from these locations. We studied both myrtle and Audubon's warblers (S. c. auduboni) captured during their spring migration along the Pacific Coast, south of the narrow region where these two taxa hybridize. Using stable hydrogen isotopes and biometric data, we show that those myrtle warblers wintering along the southern Pacific Coast of North America are likely to breed at high latitudes in Alaska and the Yukon rather than in Alberta or further east. Our interpretation is that the evolution of this wintering range and migration route along the Pacific Coast may have facilitated the breeding expansion of myrtle warblers into northwestern North America. Moreover, these data suggest that there may be a migratory divide within genetically similar populations of myrtle warblers.

suggested two scenarios that might reduce the costs of migration and facilitate the exploitation of these distant, yet presumably suitable habitats. First, taxa might, over time, evolve adaptations that increase efficiency and enable longer migratory routes.
Blackpoll warblers (Setophaga striata), for example, execute an extraordinary, nonstop, transoceanic migration that connects their breeding range, which extends as far north as northern Alaska, to their wintering range in South America (DeLuca et al., 2015). Blackpoll warblers have likely evolved a suite of physiological, neurological, and morphological adaptations (e.g., Hussell & Lambert, 1980) that allowed them to extend their long migration as they expanded northwestward into Alaska following glacial retreat.
A second possibility to allow for breeding range expansion is for a species to expand or change its overwintering area or change its migratory route. These shifts might lessen the distance between wintering and breeding areas or avoid navigational barriers, like large lakes or mountain ranges. However, large shifts in migratory routes and/ or wintering locations appear to be evolutionarily difficult. For example, the northern wheatear (Oenanthe oenanthe) is the only Old World thrush in its genus to have colonized North America. Isotopic and geolocator data demonstrate that wheatears in the New World have retained their ancestral sub-Saharan wintering grounds, resulting in a remarkable cross-hemisphere migration (Bairlein et al., 2012). Shifts in migration routes are even more likely to be challenging if intermediate routes would take migrants across regions that are ecologically inferior or difficult to navigate across (Bensch, Andersson, & Akesson, 1999;Irwin, 2009;Irwin & Irwin, 2005); this may explain the rarity of expansions of long-distance migrants between Eurasia and North America.
Range and migration changes in high latitude species must have been substantial over the past several million years, particularly in response to the glaciations during the Pleistocene (Ruegg & Smith, 2002;Weir & Schluter, 2004). However, to study range and migratory changes in an evolutionary context, such shifts must either occur rapidly and over the period of study (e.g., the contemporary overwintering expansion of blackcap warblers, Sylvia atricapilla, into Britain and Ireland; Berthold & Terrill, 1988;Bearhop et al., 2005;Irwin, 2009) or biogeographic patterns must lead to a strong inference of evolutionary history. Here, we use the latter approach and take advantage of the distinctive breeding and overwintering patterns exhibited by the myrtle warbler (Setophaga coronata coronata; here we follow the taxonomic treatment of North American Nomenclature Committee, which treats the myrtle warbler as a subspecies of the yellow-rumped warbler species; it should be noted however that the IOC World Bird List treats myrtle warblers as a species distinct from other forms in the yellow-rumped warbler species complex).
Myrtle warblers exhibit several notable characteristics in terms of their breeding and wintering ranges. First, they are one of the few eastern warblers that have breeding ranges that are both mostly restricted to the boreal forest and also breed far into Alaska and the Yukon ( Figure 1a; biogeographic pattern similar to Figure 2c in Toews, 2017).
This contrasts with most other eastern boreal wood warblers, which have ranges that do not extend as far into the northwest. Second, myrtle warblers have two distinct overwintering concentrations in the F I G U R E 1 (a) Distribution and band recoveries of Audubon's and myrtle warblers. Banding data obtained from Brewer, Diamond, Woodsworth, Collins, and Dunn (2006) and the Canadian Bird Banding Office (2013). There is a hybrid zone between Audubon's and myrtle warblers where their ranges come into contact in western Canada. Overwintering records of myrtle (b) and Audubon's (c) warblers, up to 2015, retrieved from eBird (Sullivan et al., 2009) (Figure 1a; Hubbard, 1969;Barrowclough, 1980;Brelsford & Irwin, 2009).
Audubon's warblers winter along the Pacific Coast of California and Mexico ( Figure 1c). given the subtle differences in plumage and morphology, some authors have questioned the validity of such a "finely drawn subspecies" (Bent, 1953;Hubbard, 1970;Yarbrough & Johnston, 1965 (Hubbard, 1970). However, there have been no formal tests or direct evidence of connections between breeding and wintering populations.
There are many myrtle warblers that move through the lower mainland of coastal British Columbia during fall and spring migration, which is a region that is well within the breeding range of Audubon's rather than myrtle warblers. It is possible that these are those that winter along the Pacific Coast, providing us an opportunity to study the characteristics of this specific group. Here, we use morphological and isotopic analyses (Bowen, Liu, Vander Zanden, Zhao, & Takahashi, 2014;Hobson, 1999;Hobson, Van Wilgenburg, Wassenaar, & Larson, 2012;Kelly, Atudorei, Sharp, & Finch, 2002;Paxton, Yau, Moore, & Irwin, 2013;Toews, Brelsford, & Irwin, 2014) to infer the breeding region of these myrtle warblers on migration. We contrast this isotopic variation with Audubon's warblers migrating at the same time, but have a more restricted range of possible breeding locations within the southern half of British Columbia (e.g., Figure 1a) and are therefore used to indirectly validate our isotopic analysis.
For the breeding location of myrtle warblers, we distinguish between two possibilities. First, these birds could be primarily from the far northwestern breeding region of myrtle warblers, as is postulated for the range of hooveri (e.g., the Yukon and Alaska). Alternatively, these myrtle warblers could be from Alberta or other more eastern parts of the range of the coronata form, or they could be a broad mixture of coronata and hooveri. A finding that the western-migrating myrtle warblers are primarily from the Yukon/Alaska region would raise the possibility of the hooveri form having evolved a distinct migratory route compared to the coronata form, suggesting the possibility of a migratory divide between two forms of myrtle warblers (Helbig, 1991;Bensch et al., 1999;Irwin & Irwin 2005;Ruegg, 2008;Rohwer & Irwin, 2011;Delmore & Irwin, 2014 Hubbard (1970).
For a subset of Iona Island male warblers, we determined the stable hydrogen ratio (δ 2 Hf) in their covert feathers (n = 59 individuals, divided approximately equally between myrtle, n = 30, and Audubon's warblers, n = 29) sampled across the migratory period.
Stable hydrogen ratio here refers to the relative amounts of the two stable forms of hydrogen (deuterium over protium) divided by that ratio in a standard material. We call this ratio of ratios the "isotope value" of the feather. We took advantage of a distinctive pattern in the molt cycle of these warblers: In the fall, each bird molts all of its feathers during a prebasic molt, which takes place on the breeding grounds (Pyle, 1997). Prior to spring migration, these birds again molt three to four of their inner greater covert feathers on their wintering grounds during their prealternate molt (Gaddis, 2011 Ocean Water (i.e., δ 2 H VSMOW ). To compare isotope values between myrtle and Audubon's warblers, we used two-sample t tests as implemented in R 3.4.0 (R Core Team 2017). We also combined isotope and morphometric data to better assign individuals to specific breeding populations. For this, we used a linear regression between wing-plus-tail measures and the hydrogen value of the feathers, using the "lm" function in R.

| IsoMAP analysis
We estimated the geographic origin of the feathers using IsoMAP, which is a framework that allows for modeling, predicting, and analysis of "isoscapes" (Bowen et al., 2014;  There is not a 1:1 relationship between hydrogen in precipitation (δ 2 Hp) and hydrogen in feathers (δ 2 Hf). For organic samples, such as feather keratin (δ 2 Hf), it is therefore important to generate an empirically based transfer function between the two (Bowen et al., 2014). We used the two-part linear transfer function from Toews et al. (2014), which was modified from Hobson et al. (2012) and is based on hydrogen isotope values from passerine feathers grown at known locations. For feathers with δ 2 Hf below −53.6 ‰, we used δ 2 Hf = 0.5765*δ 2 Hp-61.34 as the transfer function; for higher values, we used δ 2 Hf = 1.345*δ 2 Hp-20.17. With IsoMAP, we then generated a geographic likelihood assignment surface for each feather using the "individual assignment" function, including the standard deviation of the residuals from the water/feather transformation function (9.96‰). The resulting likelihood surfaces were then averaged across individuals-separated by species and feather type-using the raster calculator in QGIS (QGIS Development Team 2017).

| RESULTS
Consistent with observations from previous years, a pulse of myrtle and Audubon's warblers moved through southern British Columbia during a short period between the middle and end of April (approximately 500 individuals of both species over 3 weeks were banded at the Iona Island Bird Observatory). The Audubon's warblers we sampled during this time had wing chords similar to other Audubon's warblers measured throughout the interior of British Columbia (Table 1).
The wings of migrant myrtle warblers were relatively long and were similar to the high latitude breeding populations of hooveri (Table 1), measuring approximately 2 mm longer than myrtle warblers sampled at lower latitudes from across their range (Table 1).
The combined measure of wing and tail lengths shows a strong latitudinal pattern: Above 60 o N latitude, the wing/tail composite measure was, on average, much higher than at lower latitudes ( Figure 2a). Our sample of myrtle warblers on migration has a distribution of wing plus tail lengths similar to those found at these high latitudes (Figure 2b) and is significantly different from those at lower latitudes ( Table 2).
Patterns of stable hydrogen content differed between the species and varied based on when the feather was molted (Figure 3). For both species, feathers grown on the wintering grounds had a much higher proportion of deuterium as compared to those grown on the breeding grounds, consistent with previous studies in this species complex comparing analogous feathers (Toews et al., 2014). There was no significant difference in δ 2 Hf between the species for feathers grown during the prealternate molt (i.e., grown on the wintering grounds; mean of myrtle δ 2 Hf: −69.75‰ ± 15.8 SD; mean of Audubon's δ 2 Hf: −63.01‰ ± 13.9 SD; t = −1.67, df = 53, p = .10). By contrast, in the feathers grown during the prebasic molt (i.e., grown during the previous breeding season), myrtle warblers were highly depleted in δ 2 Hf as compared to Audubon's warblers, and this difference was significant (mean of myrtle δ 2 Hf −160.75‰ ± 12.4 SD; mean of Audubon's δ 2 Hf T A B L E 1 Wing chord measures of myrtle and Audubon's warblers sampled previously by Hubbard (1970) and for the current study (captured on Iona Island, birds on migration in southern British Columbia). Audubon's warblers have similar wing lengths as those sampled throughout B.C.; Myrtle warblers from the current study have wing lengths similar to the hooveri subspecies, as measured by Hubbard (1970), and are much longer than birds sampled from Alberta eastward

| DISCUSSION
Here, we have used morphometrics and isotopes to link the wintering and breeding ranges of both western myrtle and Audubon's T A B L E 2 Morphometric analysis of myrtle warblers sampled previously by Brelsford and Irwin (2009)

| Wintering Pacific myrtle warblers breed at high latitudes
Myrtle warblers wintering along the Pacific Coast have been identified as a distinct subspecies-Setophaga coronata hooveri-primarily based on their long wings and tails (Bent, 1953;McGregor, 1899).
In their original description, hooveri had average wing (76.7 mm) and tail lengths (58.4 mm) that were the longest recorded in populations of myrtle warblers (McGregor, 1899). At the time, these birds were presumed to be "breeding probably in British Columbia and Alaska" (McGregor, 1899). This was indirectly confirmed by studies of breeding birds in the north by Hubbard (1970). Hubbard (1970) sampled myrtle warblers from across their range and found only those at high latitudes had similarly long wings and tails. Our morphometric data from myrtle warblers migrating through southern British Columbia indicate they also have relatively long wings and tails, which is consistent with the description of breeding and wintering samples of hooveri ( Figure 2 and Table 1).
A possible complication with this interpretation is that the large Pacific Coast myrtle warblers are not a wintering population drawn from a specific breeding location, but instead this region may repre- The likelihood surfaces based on these isotopic data are diffuse in many cases-particularly for the alternate feathers and the basic feathers from Audubon's warblers-yet three patterns are relevant to this discussion. First, there is a region of high likelihood from the alternate feathers (grown in the winter), from both myrtle and Audubon's warblers, along the Pacific Coast. While there is also a large region of high likelihood in the southwest and eastern USA, based on the known wintering locations of these two taxa (e.g., the colored outlines in Figure 1b and d), this band of high likelihood is more likely due to the distribution of δ 2 H on the landscape as opposed to being reflective of their true overwintering locations (Toews et al., 2014). It is important to note that similar isotope values do not necessarily mean that feathers from the two groups were grown in the same location.
However, at a broad scale, contrasting the two possible overwintering concentrations for myrtle warblers (e.g., Pacific Coast versus Gulf Coast), these results more strongly favor both Audubon's and myrtle warblers in our sample as wintering along the Pacific.
Second, for Audubon's warbler's basic feathers, the likelihood surface is very dispersed. However, it includes an area in central and eastern British Columbia where Audubon's are known to breed and is likely where these feathers were grown. Other regions to the north and east are beyond Audubon's warblers breeding range ( Figure 5c) and thus make the origin of feathers from these locations very unlikely.
Finally, the average likelihood surface derived from basic feathers from myrtle warbler feathers is highly concentrated and is confined to This suggests that a small number of individuals in our migrant sample may have bred in lower latitude populations but that the large majority of these myrtle warblers instead bred at very high latitudes during the previous spring and summer. This is consistent with isotopic studies of Wilson's warblers (Kelly et al., 2002;Paxton, van Riper, Theimer, & Paxton, 2007;Paxton et al., 2013), with northern breeding populations similarly showing lower hydrogen isotope values. Like Audubon's and the non-hooveri populations of myrtle warblers, Wilson's warblers consist of western and eastern genetically differentiated breeding populations that use distinct western and eastern migratory routes (Irwin, Irwin, & Smith, 2011;Paxton et al., 2013;Ruegg et al., 2014), presumably with a migratory divide between them.
Some caution is warranted in interpreting the likelihood surfaces of the geographic assignments from the myrtle basic feathers. This is because these feathers are highly depleted in deuterium. In fact, they are more depleted than most of the known-location feathers included in the δ 2 Hf:δ 2 Hp transformation function utilized here (Hobson et al., 2012;Toews et al., 2014). Therefore, by estimating δ 2 Hp values from these feathers, we are necessarily extrapolating outside of the range of input values used in the model. While the precise geographic regions where these feathers were grown should therefore be treated with caution, it is clear given the extremely low levels of deuterium that these feathers were grown at very high latitudes, regions that are also depleted of deuterium.

| The evolution of the migratory phenotype
These new data speak to three general conclusions that extend beyond migratory connectivity within the yellow-rumped warbler species complex. First, these data are consistent with previous research in other migratory taxa that have found a correlation between migratory distance and wing length (Nowakowski, Szulc, & Remisiewicz, 2014). This is suggested by the significantly negative correlation between wing-plus-tail and deuterium-depleted isotope values (Figure 4). This implies possible local adaptation of wing and tail morphology over a relatively small spatial scale, at least within the myrtle warblers. In great reed warblers (Acrocephalus arundinaceus), as in other avian taxa, wing length has been shown to vary positively with migration distance (Tarka et al., 2010). At a broader scale, across different groups within the yellow-rumped warbler species complex, there is evidence that wing shape differs predictably between migratory and sedentary populations, consistent with what would be expected given the selection pressures acting on the different life history strategies (Milá, F I G U R E 5 Likelihood surfaces of geographic origin of feathers based on their isotopic values, produced by IsoMAP. The likelihood surfaces are averaged by species and by where the feather was molted (i.e., on the breeding or wintering ground). We note that these maps are based on isotopes alone and include some regions that both warblers do not actually breed or winter. The extent of breeding ranges of myrtle and Audubon's warblers is shown in the blue and red outlines, respectively. The extent of wintering occurrence, outlined in purple, overlaps between the groups, although there are differences in local abundance in the different regions, as illustrated in Figure 1b and c Wayne, & Smith, 2008). The fact that myrtle warblers breeding at high latitudes have longer wings is consistent with these populations being morphologically adapted to a distinct migratory route.
A second conclusion from our study is that there may be a migra- Any divide within myrtle warblers likely occurs between genetically similar populations: Several myrtle warblers sampled from Alaska were included in genomic analyses of the species complex (Brelsford, Milá, & Irwin, 2011;Toews, Brelsford, Grossen, Mila, & Irwin, 2016).
These individuals from Alaska-within the presumed range of hooveriappear very similar genetically to other myrtle warblers sampled from across the range (Brelsford et al., 2011;Toews et al., 2016). However, these previous studies included only limited sampling of possible hooveri individuals. Moreover, important genes for migration may be restricted in the genome and therefore possibly not assayed by AFLP approach used by Brelsford et al. (2011) or the reduced genome sequencing approach employed by Toews et al. (2016). Additional genomic study of northern myrtle warbler populations will be useful in addressing any association between migration and more subtle patterns of genetic differentiation within the group.
Finally, these data are consistent with the evolution of a novel overwintering location within myrtle warblers along the Pacific Coast.
Myrtle warblers are one of the few eastern boreal warblers that have a breeding range that extends far into the northwest as well as a disjunct Pacific Coast wintering population. Our interpretation is that the evolution of this wintering range and migration route along the Pacific Coast may have facilitated the breeding expansion of myrtle warblers into the northwest.
We suggest two possible ways that an eastern warbler with an eastern migratory route could evolve a west-coast migratory route: through within-population evolution of the migratory program, or alternatively through introgression of alleles for the western route from other taxa. Hybridization with western-migrating Audubon's warblers may have introduced alleles that predispose individuals toward using the western route. Audubon's and myrtle warblers were likely separated for an extensive period during Pleistocene glaciations. If Audubon's warblers were already well adapted for wintering in California, the eastern-adapted myrtle warbler may have acquired those western-adapted alleles through hybridization. This remains a speculative idea at this time, but introgression is increasingly viewed as a common way that populations acquire advantageous alleles and may operate on a faster timescale than novel mutation.

| Other taxa and next steps
Other species have similarly distinct and separate migratory routes and/or overwintering populations and would be fascinating to examine. Palm warblers (Setophaga palmarum), for instance, winter mostly in Florida and the Caribbean, but also consistently along the coast of California, Oregon, and Washington. Distinct western and eastern subspecies have also been described within palm warblers (Wilson, 1996). Almost all palm warblers observed on the west coast are from the western subspecies (Wilson, 1996), but this subspecies also winters in large numbers in the southeast, suggesting a possible migratory divide within the western subspecies.
Understanding why several species have such clearly disjunct wintering populations will be useful in understanding the evolution of novel overwintering sites and migration routes. We suggest that the evolution of new wintering ranges and migration routes might generally facilitate breeding range expansions. Whether this is a general phenomenon will require additional data from this and other species groups.

CONFLICT OF INTEREST
None declared.

AUTHOR CONTRIBUTIONS
All authors worked together to conceive the idea for the project. JH collected the samples for the analysis. DPT analyzed the data and wrote the original draft of the manuscript. All authors contributed to the writing and editing of the manuscript.