Long‐term continental changes in wing length, but not bill length, of a long‐distance migratory shorebird

Abstract We compiled a >50‐year record of morphometrics for semipalmated sandpipers (Calidris pusilla), a shorebird species with a Nearctic breeding distribution and intercontinental migration to South America. Our data included >57,000 individuals captured 1972–2015 at five breeding locations and three major stopover sites, plus 139 museum specimens collected in earlier decades. Wing length increased by ca. 1.5 mm (>1%) prior to 1980, followed by a decrease of 3.85 mm (nearly 4%) over the subsequent 35 years. This can account for previously reported changes in metrics at a migratory stopover site from 1985 to 2006. Wing length decreased at a rate of 1,098 darwins, or 0.176 haldanes, within the ranges of other field studies of phenotypic change. Bill length, in contrast, showed no consistent change over the full period of our study. Decreased body size as a universal response of animal populations to climate warming, and several other potential mechanisms, are unable to account for the increasing and decreasing wing length pattern observed. We propose that the post‐WWII near‐extirpation of falcon populations and their post‐1973 recovery driven by the widespread use and subsequent limitation on DDT in North America selected initially for greater flight efficiency and latterly for greater agility. This predation danger hypothesis accounts for many features of the morphometric data and deserves further investigation in this and other species.


| INTRODUCTION
The morphometry of populations shifts constantly in response to environmental changes in biological and physical factors. The colors and markings of male guppies Poecilia reticulata quickly evolved lower intensity when predators were experimentally introduced (Endler, 1980).
Beak sizes of Galápagos finch species (Geospiza spp.) have undergone rapid and fluctuating changes in response to severe natural selection on their performance under varying abundance, size, and hardness of seeds produced in response to variation in rainfall (Grant, 1986). The wings of forest birds in eastern North America have become more pointed in boreal regions and less pointed in temperate regions over the past century, which Desrochers (2010) interpreted as resulting from selection pressure arising from landscape changes, specifically deforestation in boreal regions (favoring more flight) and afforestation in temperate regions (less flight). A 20-year decline in wing size observed in cliff swallows (Petrochelidon pyrrhonota) in Nebraska was attributed to mortality from automobile collisions favoring increased agility (Brown & Brown, 2013). Shorter, more convex, and less pointed wings increase lift and hence take-off speed, and improve flight agility, both important traits for evading predators (Swaddle & Lockwood, 1998;Burns & Ydenberg, 2002;Burns, 2003). However, they are less energetically efficient for sustained flight due to their greater drag (Savile, 1956;Rayner, 1988;Norberg, 1990;Hedenström & Møller, 1992;Pennycuick, Fuller, Oar, & Kirkpatrick, 1994;Vágási et al., 2016).
Morphometric change may result from natural or sexual selection, as in all of the above examples, but can also be due to phenotypic plasticity (Teplitsky, Mills, Alho, Yarrall, & Merilä, 2008). Declines in overall adult body size in several European passerine bird populations appear to be environmental effects on phenotype and were attributed to climate warming causing an increasing temporal mismatch between the peak in food abundance and the chick-rearing period (Husby, Hille, & Visser, 2011). Plasticity was also invoked to account for much of the changes in body proportions of arctic-breeding red knots (Calidris canutus, van Gils et al., 2016).
Whether by microevolution, phenotypic plasticity, or some combination of both processes, several authors have proposed that a decline in body size is a widespread general response to climate warming (Rode, Amstrup, & Regehr, 2010;Gardner, Peters, Kearney, Joseph, & Heinsohn, 2011;Sheridan & Bickford, 2011;Baudron, Needle, Rijnsdorp, & Marshall, 2014). Empirical papers have documented heterogeneity in the magnitude and direction of size responses of many taxa (Yom-Tov, Yom-Tov, Wright, Thorne, & Du Feu, 2006) and call for both empirical and theoretical studies to better understand the underlying mechanisms and physiological consequences of body size shifts (McNamara, Higginson, & Verhulst, 2016). Such studies contribute to our understanding of the extent to which, and how rapidly, species may respond to macroenvironmental changes (Botero, Dor, McCain, & Safran, 2014).
This study examines long-term variation in two measures of body size in a long-distance migratory shorebird, the semipalmated sandpiper (Calidris pusilla). The species breeds across arctic North America and winters along coastlines of Central and South America (Hicklin & Gratto-Trevor, 2010). Our investigation was stimulated by reports that semipalmated sandpipers captured at Johnson's Mills in the Bay of Fundy in the early 1980s had substantially longer wings and slightly longer bills than those caught during the late 1990s and early 2000s (Hicklin & Chardine, 2012). The species has a marked geographic cline in size across the breeding range, with bills and wings shorter in the west (Harrington & Morrison, 1979;Gratto-Trevor et al., 2012).
Western and some central Arctic breeders migrate southward through central North America . The remaining central Arctic breeders, and all eastern Arctic birds, migrate south via the Atlantic coast, particularly the Bay of Fundy. Hicklin and Chardine (2012) interpreted the shorter metrics they reported around 2000 as support for the hypothesis that eastern, long-billed populations had recently undergone large and disproportionate population reductions, resulting in relatively lower usage of this migratory site than populations from central breeding areas. Surveys in North America (Morrison et al., 2001(Morrison et al., , 2006 and South America (Ottema & Spaans, 2008;Morrison et al., 2012) have reported strong declines, thought to represent primarily eastern populations (Brown et al., 2017; but see Andres et al., 2012).
Interpreting changes in morphometrics of birds captured at migratory stopover sites is complicated because changes in the metrics can occur in several ways related to sample biases, even in the absence initially for greater flight efficiency and latterly for greater agility. This predation danger hypothesis accounts for many features of the morphometric data and deserves further investigation in this and other species.

K E Y W O R D S
allometry, Calidris pusilla, environmental change, phenotypical change, predation risk, semipalmated sandpiper of any real phenotypic changes. If body size differs among breeding areas, changing representation bias from different areas of origin would alter metrics (Hicklin & Chardine, 2012). If the sexes differ in size or passage time, a change in sex ratio or the timing of sampling relative to this progression could also alter population metrics. Last, changes in size distribution at a stopover site could result from mass or morphology-dependent shifts in habitat use (e.g., mass-dependence: Ydenberg et al., 2002;Ydenberg, Butler, Lank, Smith, & Ireland, 2004).
Any or all of these processes could apply to semipalmated sandpipers.
Identifying phenotypic change in the morphometrics of breeding populations of semipalmated sandpipers is more straightforward because none of these complications is a factor. We therefore undertook a range-wide investigation of historical changes in wing and bill morphology in this species at breeding and stopover sites, to determine the extent to which the large changes reported represented true morphological changes versus changes in migratory demographics and timing. Specifically, we tested the possibility that phenotypic change on the breeding range, rather than or in addition to changes in migratory population structure, could account for changes reported at stopover sites (Hicklin & Chardine, 2012). We compiled morphometric data measured on breeding populations of semipalmated sandpipers to help interpret the long-term changes recorded at migratory stopover sites, including data from two additional migratory sites to help assess whether the morphometric changes reported from the Bay of Fundy also occurred elsewhere.

| METHODS
We assembled field measurements of bill and wing length from >57,000 semipalmated sandpipers captured as live adults between 1972 and 2015 at five breeding sites across the Arctic and at three major southbound migratory stopover sites. Morphometric data originate from both published and unpublished reports. The locations, years, sample sizes, annual means and standard errors, references to methods and to persons supervising data collection are summarized in Table 1. Data from these "live birds" were compared with measures from 139 museum specimens of adults collected earlier, mostly during the 1950s and 1960s, with a few specimens dating back to the early twentieth century (Harrington & Morrison, 1979). To compare metrics from museum metrics with those from live birds, we adjusted for shrinkage by increasing the measured culmen length of each museum specimen by 1% (Engelmoer & Roselaar, 1998) and wing length by 2% (Prater, Marchant, & Vuorinen, 1977;Harrington & Morrison, 1979).
Museum specimens were sexed by gonadal inspection.
Live birds from breeding grounds were trapped on nests as routine parts of breeding biology studies (Table 1). To account for the known geographic cline in body size of semipalmated sandpipers across their Arctic breeding range (Harrington & Morrison, 1979), each record was assigned to "western" (Alaska); "central," including western Nunavut, Banks Island, Kitikmeot (formerly Mackenzie) and Kivallig (formerly Keewatin) Districts of the Northwest Territories; or "eastern" regions (Baffin Island, Belcher Island/eastern Hudson Bay), using the divisions portrayed in Gratto-Trevor et al. (2012). Adult semipalmated sandpipers at migration stopover sites were captured in mist nets or, after 1986 at the Bay of Fundy, primarily with Fundy pull traps at roosting sites (Hicklin, Hounsell, & Finney, 1989).
For each bird, bill length was taken as the exposed culmen, measured with calipers to the nearest 0.1 mm. Wing measurements were taken using wing rulers; precision was 1.0 mm at most sites, but 0.5 mm at some. At most sites in most years, "flattened" wing length, from a bent "elbow" (radius/ulna to carpus-metacarpus joint) to primary tip, was measured, which has become the worldwide standard for shorebirds (Prater et al., 1977). At Manomet (1972Manomet ( -1995 Hicklin, J. Paquet, P. Donahue, N. Garrity, pers. commun.; Table 1). We therefore treat the wing lengths from Manomet, and the Bay of Fundy 1997-2006, separately in our analyses and interpretations.

| Data analysis
Our analytical approach compares estimated metrics of samples or populations with balanced sex ratios. Sex-specific comparisons of live birds are difficult because semipalmated sandpipers have femalebiased sexual size dimorphism (Harrington & Taylor, 1982), but cannot be reliably sexed in the field; ~20%-30% of individuals remain ambiguous even with information from 11 skeletal variables (Cartar, 1984).
To estimate size distributions of "historical" wing and bill lengths for each region, we first generated normal size distributions from sexspecific means and standard deviations of the pooled museum specimens within each region (Harrington & Morrison, 1979;their Table 2).
We then randomly drew 1,000 males and 1,000 females from the simulated regional distributions and pooled the sexes to estimate a regional population mean and standard deviation (SD). These values are displayed in the left portion of each panel in Figure 1.
We compared baseline historical distributions with the earliest morphometrics from live breeding birds that were available from each region (eastern: 8 years, 1980-1987; central: 3 years between 1991 and 1994; western: 6 years, 1993-1998; Table 1 and Figure 1).
In breeding studies, both members of this socially monogamous species with biparental incubation were usually trapped at the nest, and these field samples were therefore well balanced by sex. We tested measurements from all sites pooled within each region against random samples drawn from the simulated historical distributions. To make comparisons with appropriate statistical power, we drew random samples from the historical distributions, with replacement, each with N equal to the number of museum specimens originally measured for that region and metric (Harrington & Morrison, 1979;their Table 2).
We then compared the live versus historical distributions with twosample t tests. We report the mean t-values and probabilities from tests against 30 random samples drawn from each region's historical distribution (Table 2).
T A B L E 1 Summary table of the morphometric data (wing length, exposed culmen length) of adult semipalmated sandpipers

Location and References
Year N wings or (wing, culmen)  : 1986, 1991, 1995). An alternative approach using least-squared annual means corrected for date of capture to adjust for sex ratio effects at migration sites, assuming common seasonal slopes over years (Hicklin & Chardine, 2012), produced results similar to those presented here.

Western Breeding Region
For comparison with other studies, we estimate rates of phenotypic change in darwins and haldanes (Hendry & Kinnison, 1999).
Almost all wing and bill metrics were taken on the same individuals, allowing us to calculate the phenotypic covariances of samples from sites to estimate what fraction of change might be attributed to a change in one metric driving change on the other and how these relationships might change over time. We present these as slopes of bill versus wing length. We used SAS 9.4 for data management and statistical calculations.

| RESULTS
The well-known longitudinal clines in bill and wing sizes of semipalmated sandpipers are evident in the regional historical distributions Individuals from different breeding regions mingle at migratory stopover sites and mean sizes were thus expected to be intermediate Wing lengths are flattened and straightened lengths, except as footnoted. The annual mean and standard error, and the identity of the banding supervisor are given for each of five breeding and three stopover locations. Sample sizes (n) are given for wing measurements; sample sizes for culmen are within 1% of the matching wing tally, except as noted with double entries: (wing, culmen). "Historical" wing and culmen lengths were simulated as described in the text, n in this case refers to the number of museum specimens originally measured, and standard deviations rather than = 6.08, p < .0001), probably due to a more male-biased sex ratio, as females depart breeding grounds earlier than males (Ashkenazie & Safriel, 1979;Gratto-Trevor, 1991). c Data gathered 2010-2014 in coordination with the Arctic Shorebird Demographics Network (Brown, Gates, & Liebezeit, 2014). d All wing data natural wing chord. Culmen length from 1985 to 1995 were adjusted from a "narina" bill measurement taken in those years, using the regres- Positive differences indicate that the historical mean is shorter. The differences in t and p values reported are the means of 30 two-sample t tests, in which random samples with size equivalent to the number of museum specimens originally measured were drawn from the simulated historical regional distributions, and compared against the corresponding live distributions.
to sizes of their breeding regions of origin. This prediction was met for culmen lengths (  Table 2).
In contrast, bill lengths from these years showed no significant differences from corresponding historical means in any region, and are if anything shorter by 1.09, 0.13, and 1.24 mm in western, central, and eastern regions, respectively (p = .12, .68 and .56, respectively; Table 2).
F I G U R E 1 Five decades of annual mean wing (left panels: a, c) and bill lengths (right panels: b, d) of adult semipalmated sandpipers, measured at breeding (upper panels: a, b) and migratory stopover sites (lower panels: c, d). The measures from live birds are plotted as annual means in mm with 95% CIs. The left portion of each panel displays historical regional wing and culmen distributions (mean ± SD), estimated based on pre-1970 museum specimens, as described in the text. Breeding sites are aggregated into three breeding regions (west, central, east; see text). Stopover measures were made during southward migration at three major stopover sites (James Bay, ON; Bay of Fundy, NB; Manomet, MA). Lines indicate statistically significant linear (solid) or quadratic (dashed) trends in annual mean values for individual breeding regions or stopover sites. A few points from James Bay and Manomet stopover sites were excluded from trend calculations because sampling did not occur throughout the season (see text, Table 1). Wing measurements are flattened chords, except for points within the dashed boxes, which are natural wing chords recorded at Manomet, or of annually variable methodology at the Bay of Fundy 1997-2006 (see text, Table 1). Bill lengths were measured as exposed culmen (Table 1) Within each region, wing lengths measured on live birds at breeding sites were longest in the earlier years and decline thereafter. Thus, the period of wing length increase terminated before or as these studies began. At the James Bay stopover site, a quadratic model fits the 1975-1982 data, with a 1980 peak length (annual mean wing = 30 9,924 + 313.33*year − 0.07917 year 2 , F [1,5] = 9.52, p = .027). As several thousand birds were captured at this site annually, the annual estimates themselves have narrow confidence limits, and this curve provides solid support for a 1980 peak in wing length at this location.
After 1980, mean wing length decreased on breeding sites by nearly 4% over the 35-year period ending in 2015 (Table 3, Figure 1).
Analysis of covariance shows no significant regional differences in rate Wing length data from the Bay of Fundy were analyzed separately during three time periods (see section 2, Table 3, Figure 1). Flattened wing lengths, all measured by PH (Table 1) T A B L E 3 Regression coefficients of linear models of change in mean annual morphometrics of adult semipalmated sandpipers on breeding grounds and at major southward migration stopover sites  1 and 3, Figure 1c,d).
There was substantial phenotypic covariance between the two body size metrics (  were themselves pooling of measurements made by multiple individuals whose individual biases would tend to offset one another.

| DISCUSSION
Finally, substantial change occurred within sites despite having consistent investigator training (e.g., Nome 1993-2014, and  Due to phenotypic covariance (Table 4), contemporaneous changes in wing and bill metrics could be driven to some extent, in principle, by overall changes in body size and/or correlated selection (Lande & Arnold, 1983). While some parallel patterns do occur in the short term, most strikingly visible in the consistently measured and large James Bay

| Interpreting changes at migratory sites
Processes other than true phenotypic change can complicate interpretation of the data from migratory stopover sites (see section 1).
Hicklin and Chardine (2012)  tionate changes in regional representation over the entire period, but offers no alternative explanation for the short bill sizes in the early 2000s (Bliss 2015).
Changes in wing lengths of breeding semipalmated sandpipers reported in the present study, unknown to Hicklin and Chardine (2012), combined with changes in the phenotypic covariance between wing and bills (Table 4 (Table 4), and phenotypic covariances can be reasonable surrogates for genetic covariances (Agrawal & Stinchcombe, 2009).
The extensive data from Manomet, while noisy, probably due to shifting measurement biases, nonetheless present a contrasting situation from this perspective. In contrast to all other post-1980 data, wing lengths increased, but bill lengths did not (Figure 1). This suggests true disproportionate use of the site by longer-winged migrant sandpipers.
A potential explanation for this is that since the late 1980s, longer-

| Ecological causes of morphometric patterns
Any single hypothesis to explain the phenotypic changes documented here must account for (1)  (see also Husby et al., 2011). A similar process could operate here, if shrinking body size is a universal response to global temperature increase (Gardner, Heinsohn, & Joseph, 2009;Gardner et al., 2011Gardner et al., , 2014). In the system described here, the lack of concordant changes in bill length suggests an effect specific to wing length per se rather than a generalized change in body size. More importantly, while the timing of the decline in wing length since 1980 broadly matches a global temperature increase, an increasing wing length prior to 1980 is opposite to what would be expected under this mechanism.
A second possible mechanism is that wing wear during migration and the nonbreeding season has increased, e.g., due to habitat or range changes (Fahrig, 2003 versal has yet been identified. A third possibility is that the changes in wing size reflect changes in selection on performance attributes of wing size and shape that affect the balance between the fitness consequences of energetic flight efficiency versus agility (Swaddle & Lockwood, 1998). Flattened wing length of small sandpipers correlates strongly with principle component indices of wing shape derived from measurements of all 10 primaries; longer wings of semipalmated sandpipers are more pointed and shorter wings more rounded (Fernández & Lank, 2007;Ortiz et al., unpublished). Several factors could alter this balance.
Longer wings improve long-distance flight efficiency, and even small changes can be important. Greater agility in shorebirds can also be favored by sexual selection for aerial displays, (Figuerola, 1999;Székely, Reynolds, & Figuerola, 2000), which would operate primarily on males. Here we are considering mixed sex populations, and there is no reason to believe that display modality is shifting toward greater aerial display or that there were changes in the intensity of competition among males. If anything, recently reported population declines of semipalmated sandpipers (Andres et al., 2012) might reduce population densities and therefore intrasexual competition on breeding grounds, which would not favor shorter wing lengths.
Wing shapes of small shorebirds may be influenced by human hunting, which has become a conservation concern for several species in the Western Hemisphere. Along the coasts of Guyana, French Guiana, and Suriname, long the core historical wintering range for semipalmated sandpipers, trembling or "choking wires" stretched from posts across beaches have long been used to knock low-flying birds out of the air, but mist nets are now also widely used (Ottema & Spaans, 2008;Morrison et al., 2012). Both techniques would select against less agile birds and thus favor shortened wing lengths. We lack information on historical shifts in hunting techniques or intensity (Watts & Turrin, 2016), thus we cannot easily try to match the timing of the effect of human harvest on the wing length changes we have documented.
Under this hypothesis, the peak wing length value observed ca. 1980 represents the end of a period during which low predation danger allowed the benefits of flight efficiency for migration to select for longer wings. When falcon numbers rebounded, the balance of selection reversed, increasingly favoring more defensive (shorter and rounder) wing morphology.
We estimate the post-1980 rate of phenotypic decline in wing length at 1,098 darwins and 0.176 haldanes, which fall well within the range estimated by other studies of microevolutionary change of a decade or longer in duration (see Table 1 in Hendry & Kinnison, 1999;Kopp & Matuszewski, 2013). Hence, it is plausible that these changes are primarily a direct genetically based response to selection. In addition to or along with selection, the change in size could be an induced defense, as in crucian carp (Carassius carassius), in which a deeper body form is induced by the presence of predatory pike (Esox lucius; Domenici, Turesson, Brodersen, & Bronmark, 2008).
The altered body form of carp lowers vulnerability and enables higher speed, acceleration, and turning rate during anti-predator responses.
It is in theory possible that individual semipalmated sandpipers alter feather morphology during the annual wing molt based on experience gained during the previous southward migration or a more generalized assessment of their danger landscape. Such a mechanism not been demonstrated in birds, to our knowledge, although wing morphology commonly changes toward higher performance (longer) wings between juvenile and adult phases of avian species (Alatalo, Gustafsson, & Lundbürg, 1984;Fernández & Lank, 2007). A further possibility is that adjustments to wing morphology result from a maternal effect in which the mother's assessment of predation danger leads to an adjustment of egg contents that alter her offspring's phenotype. As a current avian example, experimentally increasing levels of perceived predation risk increased yolk testosterone in clutches of great tits, and young hatching from treated nests grew wings at faster rates and had longer wings at maturity than those in control nests (Coslovsky & Richner, 2011;Coslovsky, Groothuis, de Vries, & Richner, 2012).

| CONCLUSION
The wing lengths of semipalmated sandpiper populations increased during the decades prior to 1980 and subsequently decreased with no accompanying systematic changes in bill length. The changes in wing length documented here coincide temporally with continentwide changes in predation danger attributable to the steep decline and subsequent ongoing recovery of falcon populations resulting from the postwar introduction of DDT and 1973 ban on its widespread agricultural use. We hypothesize that the ongoing changes in wing size reflect changes in the balance of selective pressures arising from efficient long-distance migration (longer, more tapered wings) versus defensive morphology (shorter, rounder wings). Several alternative hypotheses, including that a general body size reduction is a universal response to climate warming, do not account for the initial increase in wing lengths and/or the lack of systematic changes in bill size. If the predation danger hypothesis is true, it should apply more generally to small shorebirds and other avian taxa (Yom-Tov et al., 2006). We encourage continued exploration of historical changes during the preand post-DDT time frame to challenge this hypothesis.

ACKNOWLEDGMENTS
This study was conceived by RCY and developed together with DBL and CX. Authors 3-18 provided data and are listed in chronological order based on the date of collection of the first live measures contributed, which correlates well with the (log) number of live individuals measured (Pearson r 2 = −0.86). DBL and CX compiled the data. RCY, CX, and DBL developed methodology and analyzed the data. The MS was written by RCY, CX, and DBL with many other coauthors providing editorial input. SAF produced Figure 1. All authors acknowledge the contributions of the many biologists, students, and volunteers who helped collect the data as well as the landowners who facilitated access to field sites. We wish to recognize the Arctic Shorebird Demographics Network for contributing standardized measurement data from breeding grounds in recent years (2010)(2011)(2012)(2013)(2014), and in par-  Figure A1 plots the modeled deviation of each observer from the global mean. Seven of the 12 observers fall within 0.5 mm of the global mean. The mean difference between the 66 possible pairs of observers was 0.66 ± 0.46 SD mm, with a maximum difference of ~1.7 mm between individuals.
Ideally, estimates of interobserver variability should be based on measurements of the same bird. The James Bay data included 35 recaptures each measured independently within a season by somewhat random pairs of two of eight observers. We estimated the repeatability (intraclass correlation coefficient Lessells & Boag, 1987) of these measurements as 0.34. Based on an observer variance component of 0.72, 67% of observations between observers should fall within 0.85 mm (±1 SD) of the mean value, and 95% of values within 1.70 mm (±2 SD). Ten of the 35 measurements were identical; but the actual mean difference between measurements was more substantial than in the larger population comparisons, being 2.0 ± 1.76 (SD) mm, or that determined for western sandpipers in a third comparison. Sixteen western sandpipers, a simlarly sized close relative of We conclude that individual wing length measurement baises among observers involved in this study are on the order of ± <1.0 mm from a global mean.
F I G U R E A 1 Estimated measurement biases of 12 individuals on flattened wing lengths, in mm, expressed as deviations from a mean of 101.2 mm. All individuals measured >350 wings during one year at James Bay 1974-1982; some did so in multiple years. Each bar is the estimate of individual observer effect from a maximum likelihood model (SAS, Proc mixed) predicting wing length as a function of individual (random effect, all p > .0001), year (p ranged from >.0001 to 0.54 among years), date measured (p = .112), and a year*date interaction term (p ranged from .003 to .763; without the interaction term p for date <.0001)