Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Band recoveries reveal alternative migration strategies in American Robins

Band recoveries reveal alternative migration strategies in American Robins Migration strategies may change in response to climate change with consequences for conservation efforts. We used 80 years (1934-2014) of band recovery data (N = 1,057) to describe spatial and temporal patterns in the migration behavior of American Robins. The distribution of recoveries suggests strong continental scale connectivity with distinct separation between eastern and western North America, with a more moderate degree of connectivity within these regions. We also found little evidence of differential migration between males and females. Despite previous studies that suggest the winter distribution of robins has shifted northward, our analysis shows no obvious change in migration distance over time. Surprisingly, we found that a significant proportion of across season band recoveries occurred locally (20%), in close proximity to the original banding locations. It's well known that large numbers of robins linger in northern breeding grounds well into the winter of some years, but the proximity of these birds to breeding areas was previously unknown. We found little evidence that the winter latitude of migrants or local recoveries shifted over time. However, there was a trend for increased frequency of local recoveries in recent decades, providing an alternative hypothesis for the northward shift in winter distribution. Keywords: Turdus migratorius, bird banding, climate change, partial migration, winter residency, distribution shift 1 Introduction Global change is causing taxonomically and geographically widespread changes to bird distributions and migration patterns [1]. In Nearctic and Palearctic migration systems, *Corresponding author David Brown, Eastern Kentucky University, United States, E-mail: david.brown@eku.edu Gail Miller, Eastern Kentucky University numerous species have advanced spring arrival dates and shortened migration distances in response to global change [2­4]. The winter distributions of many species have also shifted north [5­8], in part because of increased winter temperatures, and it is possible that such shifts are driven by shortened migration distances [9]. An alternative but not exclusive hypothesis argues that changes in distribution and timing of arrival can be influenced by the frequency of non-migration strategies [9]. Partial migration, in which some individuals within a population migrate and others are year-round residents, occurs in a great number of species worldwide, perhaps even a majority [10­12], and it is critical that we develop a solid understanding of how such behaviors will allow species to adapt to climate change. Connectivity between breeding and winter populations is another aspect of migration that may have important implications for conservation planning, but it is generally poorly understood for most species [13]. Populations with strong connectivity, in which most of a breeding population migrates to a common wintering area, may be more vulnerable to population declines caused by spatially focused environmental changes. Whereas, the mixing across seasons that occurs in populations with weak connectivity can spread the effects of spatially focused environmental perturbations that occur in one part of the annual cycle [13]. Research has also revealed sex-biased differential migration in many bird species, most commonly in which females winter further south than males [14]. One well supported theory for this pattern is that males remain relatively close to their breeding grounds to arrive earlier for the breeding season and acquire better territories [14]. If such behavior occurs on a continuum, then one end point would be for birds, especially males, to reside yearround on or near their breeding territory. For most species, but in particular for non-game species, there is little data available to describe migration patterns, let alone to test how these patterns might be changing over space and time. Recent work using lightlevel geolocators, stable isotopes, satellite tracking, miniaturized GPS, and other technologies are revealing © 2016 David Brown, Gail Miller, published by De Gruyter Open. This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivs 3.0 License. previously unknown migration behaviors [15­19]. Prior to such advances, bird band recoveries were a vital source of information on bird migration, and they continue to be utilized [20], often in conjunction with these new technologies [21,22]. Studying migration with banding data is an extrinsic technique that requires capture, marking, and a subsequent encounter. Each year, an average of approximately 1 million birds are banded in the United States and Canada (https://www.pwrc.usgs. gov/BBL/homepage/howmany.cfm). Birds are banded primarily as part of research programs, and the capture techniques and sampling designs underlying the bird banding differ widely among species and researchers. We refer to the initial observation and marking of a bird as "banding" and any subsequent recovery or capture as an "encounter." We use the term "recovery" more generally to refer to individual birds that were banded and subsequently encountered. Each year, a fraction of previously banded birds are encountered. Encounters occur through research programs that involve active or passive techniques to capture birds, harvest, and accidental encounters, such as dead birds found after collisions. These encounters typically occur at higher rates for game species (mean = 19%/year for waterfowl) in large part because of harvest. The high encounter rates for some waterfowl have allowed managers to develop population and migration models [23,24]. Non-game species, such as most passerines, have low rates of encounter (mean = 1.2%/year), and thus banding encounters have been less useful for researchers. Encounters can occur within or between migratory periods and years. Encounters within the same season and year may be useful for describing dispersal. To be useful for understanding migration patterns, encounters must span migratory periods, such as between winter and breeding seasons, which may occur within or across a single annual cycle. There are both advantages and disadvantages with this type of data [15]. Advantages include that it is relatively inexpensive, at least on a per bird basis, and the locations are relatively accurate for both the banding and encounter locations. Disadvantages are the low encounter rate for some species, and spatial autocorrelation, particularly for banding locations [25­27]. Perhaps no other bird species in North America is more strongly linked to popular conceptions of migration as the American Robin (Turdus migratorious)[28]. They are among the most abundant and geographically widespread species in North America [29]. Robins are ecologically flexible, inhabiting a wide variety of habitats including forest clearings, riparian zones, agricultural fields, and urban parks throughout the annual cycle [29]. The longevity record for American Robins is 13 years and 11 months [30], and average adult apparent survival is 0.51, by one range-wide estimate [31]. Across much of the United States, there is overlap between the breeding and winter distribution of American Robins, but it is not known to what extent this pattern is influenced by non-migrant, resident behavior. Audubon mid-winter bird counts show that the winter distribution of American Robins is highly variable among years [29], and in some years, large numbers of robins appear to stay well north of their typical winter distribution [32]. These patterns are likely linked to weather and food on northern breeding areas. Moreover, there appears to be a temporal trend in the winter latitude and longitude of American Robins, with the mean of the distribution shifting north and east over time [7,33]. Little is known about migratory connectivity in this species [31]. The social behavior of robins shifts from solitary or paired during the breeding season to gregarious and flocking in the winter, although some wintering birds appear to be solitary for parts of the winter [29]. American Robins are known to be facultative seasonal migrants, or even "vagrants" or "wanderers" in the winter with individuals and flocks moving in response to changes in weather and food availability [29,34]. Based on the frequent observations of large non-breeding season flocks, many birds are likely moving widely away from their breeding territories. Between year return rates to winter sites appears to be low [35], whereas the degree of within-winter site fidelity is largely unknown. There are anecdotal reports of robins that attempted to breed during winter months in states such as Michigan [36], but it is not known if these are birds that spent the spring and summer months in the same vicinity. In fact, we were unable to find peer-reviewed literature reporting robins overwintering on their breeding territory so as to spend the entire year in a single home range, although there are such accounts in popular media. The objective of this study was to provide a descriptive account of the patterns of migration by American Robins using bird band recovery data. The robin recoveries used in this study were collected across the range of robins using data that dates to the inception of federally managed bird banding operations. Specifically, we explored the possibility of migratory connectivity between populations of American Robins. We also estimated distances of migration and the frequency of different migration strategies, and tested for sex-based differences in such patterns. Finally, we tested for changes over time in wintering latitude, migration distance, and frequency of migration strategies, both overall and between sexes. 2 Methods We used 80 years (1934-2014) of North American bird banding data to describe migration patterns of American Robins. All banding data for American Robin recoveries comprising locations where a bird was banded and subsequently encountered (N = 14,709) were acquired from the USGS Bird Banding Laboratory (BBL). This full data set of recoveries included birds banded as early as 1916. We narrowed the dataset by including only birds that were banded and encountered in both a breeding (May­August) and a winter (November­February) season, including data regardless of which of the two seasons the banding event occurred. Thus, we attempted to conservatively exclude banding and encounter events occurring during the main migration period for robins as well as encounters that occurred in the same season as banding. The total number of recoveries that included both a breeding and winter observation comprised 1,346 records dating back to 1916 and including all years up to 2014. The majority of birds were first banded during the breeding season with subsequent winter encounters (89%); the remainder were first banded on the winter grounds and encountered during the breeding season. We included encounters that occurred in non-subsequent seasons (42% of all breeding­ winter recoveries occurred in a non-subsequent season, and the average number of years between banding and recovery was 1.5 ± 1.4 SD). By doing so we assumed that the breeding season locations of individual birds did not change between years. Although some birds undoubtedly disperse to new breeding season locations between years, there is evidence that breeding site fidelity is high, as with many other passerines [37, D. Brown, unpublished data]. Band encounters can occur when birds are harvested, found dead, or captured by a bird banding permit holder other than the original bander, and thus typically in a different location. Using encounters from bird banding stations in studies such as this one introduces known biases because the distribution of bird banding stations is non-random and banders sometimes use methods to increase the chances of recapturing local birds [38]. Thus, we excluded all encounters that were coded as "previously banded bird trapped and released during banding operations" (i.e., birds encountered during banding). The remaining data consisted of birds that were found dead, and a small number that were shot in the 1930's (2.5%), and together form the primary database for analysis (N = 1,057, Figure 1). We included all birds regardless of age at banding and did not attempt to include age as a factor in any analysis (After-hatch-year = 36%, Hatch-year = 35%, Figure 1. Locations of all American Robin recoveries used in this study (N = 1,057) during the breeding season (A), and the winter season (B). Juvenile = 18%, Unknown = 11%), although in some figures we show only after-hatch-year birds to reduce clutter. We were also interested in addressing questions about differences between sexes, for which we only included after-hatch-year birds of known sex (N = 285). We used ArcGIS version 10 [39] to calculate the geodesic distance between banding and encounter locations for each bird. The precision of locations as reported to the BBL are to the nearest degree or half-degree of latitude and longitude, so distance calculations of zero are from within the same coordinate block, but not necessarily the exact same location. We classified all encountered robins as migrants or local recoveries. Birds with a distance between banding and encounter that exceeded 100 km were classified as migrants. Birds were assigned as local recoveries if the distance between banding and encounter events was less than 100 km , approximately 0.9° latitude or 1.15° longitude, following Fielder and colleagues [9]. In selecting the 100 km cutoff we chose to be conservative in our classification of local recoveries since our cutoff for classifying birds as migrants may include birds that moved regionally but did not truly migrate. For example, a bird residing for the winter 150 km from its breeding location is arguably not a migrant. To further describe the general migration patterns we conducted an exponential regression analysis of winter latitude and migration distance, excluding all birds classified as local recoveries. We tested for migratory connectivity between winter and breeding locations using visual interpretations of maps, a descriptive analysis of the frequency of robins within respective winter and breeding bands of longitude, and a statistical comparison among regions based on distance matrices. For the descriptive analysis, longitude bands were assigned arbitrarily to create a small number of relatively even classes: < -110°, -110° ­ -95°, -94.99 ­ -80°, and > -80°. This binning arrangement creates a single block approximately west of the Rocky Mountains (< -110°), and creates more blocks across the eastern US, where most of the robins were banded and encountered. Each bird was assigned to a winter and breeding longitude group. Because crossing longitude classes alone does not inform connectivity, we illustrate the relative frequency among groups, and do not attempt to apply statistical analysis to the pattern. To describe the overall degree of connectivity we used a Mantel correlation coefficient calculated between a breeding season distance matrix and a winter season distance matrix using Program R, Package Ade4 [40]. These within-season distance matrices were calculated assuming spherical geometry using the program Geographic Distance Matrix Generator, version 1.2.3 [41]. We conducted this analysis for the entire dataset of band recoveries, spanning all of North America, and then separately for eastern and western regions of the continent, using the Rocky Mountain divide to separate the regions. We tested for non-random patterns in the spatial distribution of local recoveries using a nearest neighbor analyis in ArcMAP. To test if local recoveries tended to occur in areas of highest breeding relative abundance we used Breeding Bird Survey 1996­2012 analysis results [42] to calculate the relative abundance of robins at each nonmigrant breeding location and compared the frequency of those observations with the total distribution of relative abundance across all BBS grid cells using a two-sample Kolmogorov-Smirnov test. We also compared patterns of recovery, and of proportions of migrant and local recoveries between eastern and western North America using the Rocky Mountain Continental Divide to separate the regions. We compared the average latitude of local recoveries to the average latitude of breeding robins. The average and standard deviation of the latitude of breeding robins was calculated from BBS data by weighting the latitude of the center of each BBS grid cell by the relative abundance and grid cell area. Because of the high number of grid cells within the BBS region (N = 23,000), we used a permutation procedure to generate 100 normal random latitudes of breeding robins. We compared the random breeding latitudes to the latitudes of all non-migrant winter recoveries using a permutation-based Welch's t-test. To investigate the possibility of sex-based variation in migration patterns we conducted a separate analysis of migratory status based on sex. We used two-sample t-tests to compare the winter latitude of males and females separately for migrants and local recoveries. We used a chi-square contingency table to test for differences in the frequency of migrants compared to local recoveries in males and females. We also performed a set of analyses to test for changes in migration patterns over time. First we used simple linear regression to model the relationship between year and migration distance, excluding all birds classified as local recoveries. To test for changes in the proportion of local recoveries, we used a generalized linear model with binomial error distribution and log link. The dependent variable was the migratory status (local recovery or migrant). We were primarily interested in how the relative frequency of migratory status changed over time. Preliminary review of the data suggested that the proportion of local recoveries was higher in recent decades, so we binned years into pre- and post-1980, which is a widely cited cutoff for the appearance of strong effects of extreme global warming [43], and used this binary variable as our primary predictor. We graphically present these results by decade to illustrate the basis for our binning approach. Breeding latitude was included as a covariate. Models with and without the binary time period variable were compared using likelihood ratio tests. All statistical analyses were performed with Program R [40]. 3 Results 3.1 Migrants American Robin band recovery events between winter and breeding seasons peaked during the 1930's and again during the 1960's and have declined to near all-time lows during the last two decades (Figure 2A). Interestingly, the total number of banded birds has not followed the same pattern. In the period for which all banding records were available electronically (1960­2010), an average (±SD) of 20,324 (± 4,969) birds were banded each year (range: 11,784­ 33,478), with the greatest numbers from 1985­1992. The vast majority, 91%, of all recoveries used in this study (N = 1,057) include both breeding and winter season locations in the eastern and central regions of United States, with most of the remainder coming from the Pacific coast region (Figure 1). Few recoveries occurred in the Rocky Mountain region. Eighty percent of all recoveries were classified as migrants (encountered > 100 km from banding location; Figure 2B). Of the migrants, 96% of individuals travelled 500­2,100 km. The longest migration distance was by an after-hatch-year bird of unknown sex that migrated 5,153 km from Fairbanks, AK (August 1959) to central Mississippi (February 1960). Migration distance tended to be greater for robins that wintered farther south (F1,838 = 468.37, P < 0.001, R2 = 0.35, y = 13,596e-0.075x; Figure 3A). Figure 2. Histograms of American Robin band recoveries. (A) Recoveries have declined in recent decades, despite relatively consistent levels of banding. (B) The distance between breeding and winter season locations of band recoveries. The zero-inflated distribution occurs because a relatively large proportion of American Robin across-season recoveries occurred locally (< 100 km from banding site). Figure 3. Migration distance (km) of American Robin band recoveries classified as migrants (> 100 km between banding and encounter locations) in relation to winter latitude (A), and across years (B). 3.2 Connectivity At the continental scale, there appears to be a strong separation of eastern and western populations, with only two records of individuals crossing the Rocky Mountains within the conterminous United States (Mantel test of breeding vs. wintering distance matrix: rm = 0.78, P = 0.001, Figure 4A). One of these was originally banded in eastern Kansas in May 1955 and encountered in January 1956 near Portland, Oregon (this bird does not appear in Figure 4A because it was aged as hatch-year). The other bird that crossed the Rockies was banded in February 1969 near San Francisco, CA, and then encountered in Michigan in June 1969. The anomalous movement of these two birds may indicate reporting problems for these records, such as if the reported locations do not reflect the location that the bird was recovered. There were only eight records of birds crossing -110° longitude, which crosses the Rocky Mountains in Wyoming. Within eastern North America, there was a moderate level of connectivity (rm = 0.34, P = 0.001). For example, most robins wintering in Florida migrated between coastal breeding areas, but 30% migrated more inland to or from Midwest or Great Lakes regions (Figure 4B). Likewise, birds wintering in Texas tended to migrate between Great Plains breeding areas, with fewer migrating to or from Midwest or Northeast regions. Eighty-five percent of birds with breeding season locations within the -95 ­ -80° longitude band were located within the same band during the winter season (Figure 5). Likewise, 65% of individuals with breeding season locations in the -110 ­ -95° longitude band had winter locations in the same band. Within western North America, the connectivity correlation coefficient was considerably lower (rm = 0.12, P = 0.009). 3.3 Local Recoveries Of all recoveries that included both winter and breeding season occurrences (N = 1,057), 20% were classified as local recoveries (i.e. winter-breeding migration distance < 100 km; Figure 2B). Sixty-eight percent of local recoveries occurred within the same degree of latitude and longitude (the precision of band reporting to BBL) of the original banding location and were classified as 0 km distance between the banding and encounter locations. The mean (±SD) distance between banding and encounter locations for local recoveries was 20.8 ± 44.45 km. The winter location of local recoveries occurred, on average, at higher latitudes than the winter location of migrants (mean local recovery winter latitude Figure 4. Lines connecting breeding season (May­August) and winter (November­February) locations of after-hatch-year American Robins. Panel (A) includes males, females, and robins of unknown sex across North America. Panel (B) shows male (blue) and female (red) American Robins for the central and eastern regions of North America. Figure 5. Percent of American Robin breeding season locations relative to winter season locations as categorized into longitude bands. All locations are based on pairs of banding and encounter events. = 40.35, SD = 3.90, median = 39.91; mean migrant winter latitude = 32.12, SD = 3.32, median = 31.41; t1,055 = 31.35, P < 0.001). The distribution of local recoveries exhibited a clustered distribution (nearest neighbor analysis: Ratio: 0.50, Z = -13.81, P < 0.001; Figure 6). There were disproportionately more local recoveries in western North America; 29% of all local recoveries occurred west of -110° longitude, whereas only 11% of all recoveries (i.e., including migrants and local recoveries) occurred in that region (63% of all recoveries in western North America were local, whereas 15% of all recoveries east of -110° longitude were local). There were relatively high levels of local recoveries across all 4 of the winter months, but rates were markedly higher for November (November: 56%, December: 18%, January: 15%, February: 18%). 3.5 Sex differences Of all the local recoveries, 63% were of unknown sex, whereas 76% of migrants were of unknown sex. Among birds of known sex, there was no difference in the frequency of local recoveries compared to migrants in males and females (2 = 1.07, df = 1, P = 0.30; females: 36.7% local recoveries, males: 43.5% local recoveries). The latitude of local recoveries did not differ between males and females (t77 = 0.66, P = 0.50, mean ± SE female: 40.46 ± 0.76; male: 39.80 ± 0.61). Likewise for migrants, winter latitude did not differ between males and females (t191 = 0.14, P = 0.88, mean ± SE males: 32.99° ± 0.34°; females: 32.06° ± 0.36°). 3.6 Temporal trends We found no relationship between migration distance and winter year, excluding local recoveries (F1,838 = 6.94, P = 0.09, R2 = 0.008; Figure 3B). When considering just migrants, winter latitude has not shifted over time (F1,838 = 0.29, P = 0.58, R2 < 0.001; Figure 7A). There was a statistically significant northward shift in winter latitude of local recoveries over time; however, there was a large degree of scatter in the analysis and a low degree of total variation explained (F1,216 = 9.77, P = 0.002, R2 = 0.04; Figure 7B). The percent of recovered American Robins occurring locally appears to have increased since 1980 (2 = 77.13, df = 1, P < 0.001; Figure 8). 3.4 BBS analysis of local recoveries Two of our analyses compared the location of local recoveries to information derived from BBS analysis. First, we found that local recoveries tended to occur in areas of greater breeding abundance, as estimated from BBS data (D = 0.30, P < 0.001). Second, local recoveries tended to occur farther south than the average latitude of breeding American Robins based on the BBS (mean local recovery latitude = 40.35, SD = 3.90, median = 39.91; mean BBS latitude = 43.81, SD = 6.48, median = 43.2; t134.7 = 3.98, P < 0.001). Figure 6. Location of winter season after-hatch-year American Robin recoveries by migration status (migrant [open circle with dot] or local recoveries [solid circle]) and sex (male = blue, female = red) for eastern and central North America. Local recoveries occurred across seasons (winter-to-breeding, or breeding-to-winter) within 100 km of banding locations. Inset shows winter locations of all American Robin local recoveries across North America (solid circles), including those of unknown age and sex. Figure 7. Winter latitude of American Robin band recoveries across time for migrants (A), and local recoveries (B). Local recoveries were classified based on encounters occurring < 100 km from banding location. Figure 8. The percent of breeding-winter recoveries classified as local recoveries by decade. Decades were further binned to pre- and post-1980 for statistical analysis. 4 Discussion 4.1 Residency Our results provide several interesting insights into the seasonal migration patterns of American Robins. American Robins are well known to be vagrants during the winter, moving locally or regionally in response to food availability. It appears that many individuals also remain in close proximity to their breeding site, at least for part of the winter. It is possible that many of these birds are actually migrants or at least vagrants that just happened to be captured near their breeding location before or following movements over larger extents. However, it seems highly unlikely that there would be so many total winter encounter events, including a high proportion that occurred in the middle of winter (i.e., January­February), that occur close to the breeding location if many of these birds were not residing in the area for most if not all of the non-breeding season. Because of the nature of the data used here, there is no way to address such contingencies. The local recoveries made up a significant proportion of the total population and occur across the winter distribution and in similar frequencies in both males and females. Of the individuals observed to be local recoveries, some may have in fact migrated in other years. We found no records of repeat recoveries, even when including encounters from banding operations, so we have no basis to indicate individuals migrate in some years and remain resident in others, but it is possible, especially given the typically vagrant nature of robins during the non-breeding season. We have twice observed robins that remained on their breeding territory in Kentucky during January and February (D. Brown, unpublished data). The two robins were among 35 individuals color-banded during the breeding seasons of 2010­2012, and one each was observed during the winters of 2010 and 2011. Both birds were males that were paired with successfully nesting females during the preceding and subsequent summers in the same area. In both years, the birds were observed acting aggressively towards individuals associated with passing winter flocks. A substantial proportion of local recoveries occurred in the Pacific coastal region, with dense clusters around San Francisco and Seattle, disproportionate to the frequency of migrant recoveries there. These patterns may have resulted from heterogeneous sampling. Alternatively, eastern and western populations of robins may have different migration strategies. There may also be genetic population level differences between regions, which seems likely given that descriptions of populations include distinct western varieties [29,44]. Geographic variation in morphology among populations (or subspecies) suggests distinct lineages, which may also be tied to geographic variation in migration. Environmental and habitat conditions differ considerably across the distribution of robins and the Pacific coastal region has a moderate winter climate compared to most of the Midwestern and Atlantic regions; these difference may help explain the higher incidence of local recoveries in the western region. The spatial frequency of local recoveries corresponds with the relative abundance of breeding birds based on BBS surveys. In other words, local recoveries are more likely to occur in areas of higher BBS abundance. This suggests that local recoveries are not coming from the edge of the range or sub-par habitats that have low relative abundance. However, as mentioned previously, there appears to be a relatively large proportion of local recoveries in our dataset occurring in two clusters in the Pacific coastal region. Chandler Robbins attributed the occasional overwinter occupancy of relatively northern regions by robins to a combination of factors including density dependence, abundant food, and early northward migration in response to mild winter climate [32]. In recent decades, the conditions for all three of these factors have changed to favor increased northern occupancy by robins during winter: (1) The distribution and population size of robins has expanded for decades [29,42]; (2) Expansion of ornamental and invasive fruit bearing shrubs in northern areas have increased winter food supplies [45], with a direct response by robins [46,47]; And, (3) winters have been warmer in recent decades [48]. We can now argue that this northerly occupancy pattern is driven partially by a tendency to remain near breeding areas. This pattern is consistent with theories of partial migration which predict a continuum of migration behavior that may shift in response to environmental factors and in some situations result in increased frequency of residency [12,49]. 4.4 Connectivity Our findings suggest that robins have strong connectivity at the continental-scale. American Robins in the western region of North America migrated between winter and breeding locations within the region, and likewise birds banded east of the Rocky Mountains were almost exclusively encountered in that region. One seeming exception to this pattern is the occurrence of three recoveries in the southeast and central US of birds banded in Alaska; however, this migration pattern may be specific to birds breeding in the extreme northwest (i.e., Alaska), and follows, in part, the pattern of some long distance migrants [52], and may be similar to other short-distance migrants such as White-crowned sparrows (Zonotrichia albicollis) [53]. Other species, including the Yellow Warbler (Setophaga petachia), Wilson's Warbler (Cardellina pusilla), Common Yellowthroat (Geothlypis trichas), Yellow-breasted Chat (Icteria virens), and Nashville Warbler (Oreothlypis ruficapilla) also show separation of populations between eastern and western areas with parallel migration patterns [21,54,55]. These patterns of connectivity in American Robins are consistent with taxonomic descriptions that have divided American Robins into distinct populations [44]. There is only one relevant North American based study with which we can compare our quantitative estimates of connectivity based on Mantel correlation coefficients between breeding season and winter season distance matrices. In that study, 21 Swainson's Thrushes (Catharus ustulatus) breeding in western North America and traced to wintering grounds using light-level geolocators had a Mantel correlation coefficient of connectivity (rm = 0.72) higher than what we found for American Robins banded in western North American, and similar to our continentwide estimate of connectivity [56]. Within the central and eastern regions there appears to be modest migratory connectivity with many birds traversing east-west, while others remain in relatively narrow longitudinal bounds. The graphical patterns of robins crossing between eastern and central North America reinforce these quantitative estimates of moderate to diffuse connectivity. The band recovery data has relatively few recoveries from the Rocky Mountain region, where robins breed and winter in moderate levels of relative abundance [42]. Some studies based on mitochondrial DNA have failed to detect genetic structure or migratory connectivity at finer scales [55]. However, other methodological approaches, such as geolocators, satellite transmitters, morphology, molecular analysis of single nucleotide polymorphisms, 4.2 Migrants Because of the known vagrant movement patterns of robins during the non-breeding season and the acknowledged problems inherent with sampling heterogeneity in bird banding data, we exercise caution in making strong inferences about the patterns observed with the migrant birds. Nonetheless, we can take away several meaningful insights from our analysis: there is no evidence that average migration distances have changed over time; migration distance is related to wintering latitude; there are complex patterns of spatial connectivity; and, sexbased difference in migration patterns are not apparent. 4.3 Temporal patterns Based on evidence that the winter distribution of many neotemperate migrants, including robins, have shifted north in recent decades [7,33,50], we expected to find a decreasing migration distance over time. However, there is no such pattern in the recoveries of robins. One plausible explanation is that the breeding range has expanded northward concomitantly with the shift north in wintering range, and there is evidence that the range of robins has indeed expanded [29]. Another explanation for the northward shift of the winter range in the absence of changes in average migration distance is the increasing frequency of non-migrant behavior, which we have documented in this study. Other species of partial migrants, such as the Blackcap (Sylvia atricapilla) have increased the proportion of individuals maintaining year-round residency [9], and some populations of this species that were exclusively migratory have adapted to be partial migrants with individuals becoming year-round residents [49]. However, such patterns are very difficult to detect. In North America, a region of the world where birds are relatively well studied, a number of species are thought to be partial migrants (e.g., Eastern Bluebird [Sialia sialis], Eastern Towhee [Pipilo erythrophthalmus], Blue Jay [Cyanocitta cristata], Song Sparrow [Melospiza melodia]) [51], but we know little about relative frequencies of these behaviors. and bird band recoveries, have revealed connectivity in other species at finer scales [22,57,58] and weaker levels [59]. For example, Gray Catbirds (Cumetella carolinensis) breeding in the mid-Atlantic and Midwestern regions migrate to separate wintering areas [22], and genetically distinct breeding populations of Wilson's Warbler migrate to different wintering regions [58]. 4.5 Sex-based differences Surprisingly, we found no evidence of sex-based differences in migration, or in the latitude of winter recoveries. Differential migration, in which males winter farther north and migrate shorter distances than females, appears to be a widespread phenomenon among neotemperate migrants [14,60,61]. The facultative migration strategies and gregarious behaviors of robins, and the likely variation in these patterns among years, may have obscured our ability to detect differences in migration between males and females. Additionally, we used latitude as the primary indicator of distance from breeding grounds, which is simplistic and ignores how climate, habitat, and connectivity affect migration of males and females. Arguably, more robust and less biased datasets (e.g., museum collections, all banding records of a species, geolocators) are better suited for such tests [61,62], and it may be that we failed to detect any difference between males and females (i.e., Type II error). 4.6 Assumptions and Limitations As acknowledged earlier, one important bias of this dataset is that bird banding does not occur at random locations or times. We have assumed spatially and temporally homogenous sampling, but this assumption is clearly problematic, and thus the patterns we describe must be cautiously interpreted. Bird banding data include variation in banding, recaptures, and encounters, as well as habitat and other environmental factors [9]. Intensive banding within a region or time period increases the number of subsequent encounters and does not systematically account for variation in habitat. Recently developed models create possibilities to address such issues [25­27], but these models are beyond the scope of the descriptive approach that we have taken. However, our descriptive results should provide direction and testable hypotheses for future modelling efforts. If banding efforts have shifted southward in recent decades then the observation of increased frequency of local recoveries in recent decades may be a sampling artifact, since local recoveries were more likely to occur in the southern region of the breeding distribution. A related issue is that the number of recoveries has decreased in recent decades, while the total number of local recoveries has remained similar or increased. The former pattern would be expected if in fact the total number of robins banded each year had declined, but that has not been the case. Our work also suggests that winter season banding can be an important source of information about nonmigratory strategies. Of all banded robins that were later encountered in a different season (winter-to-breeding or breeding-to-winter), 30% of local recoveries were originally banded during the winter, whereas only 9% of migrants were banded during the winter. The number of robins recovered has decreased significantly in recent decades, despite relatively constant numbers of robins banded. We are unaware of changes in reporting of band recoveries that would cause such a pattern. Thus, there should be no bias in our main finding that many American Robins stay near their breeding grounds through much if not all of the winter, and that the population-level frequency of local recoveries has increased in recent decades. Our selection of months for the breeding season may have resulted in inclusion of some birds that had undergone post-breeding dispersal, and so their breeding season capture or recovery locations may differ from their actual breeding location. Additionally, within the winter season we may have included birds that were near their breeding locations, but that eventually migrated (i.e., November) or had already returned from migration (i.e., February). However, such lingering on the winter grounds should be weather and food dependent, so our conclusions about changes in these patterns over time should reflect behavioral responses to changes in environmental conditions across years. Emerging technologies are revolutionizing our understanding of bird migration [16]; however there is still much that can be learned from traditional sources of information such as citizen science [52], and the bird banding data reported here. Currently, guidelines for bird banding permit holders do not require reporting of recaptures of birds they originally banded. Such data would likely provide more insight into patterns of non-migrant behavior. Our study makes no direct links between climate and migration, and clearly more complex modelling would be required to untangle such relationships, especially given the variability in robin migration patterns, and the assumptions of using bird [15] Hobson K.A., Norris D.R., Animal migration: a context for using new techniques and approaches. In: Hobson, KA, Wassenaar, LI (Eds.), Tracking Animal Migration Using Stable Isotopes, Academic Press, London, 2008, 107-128 [16] Bridge E.S., Thorup K., Bowlin M.S., Chilson P.B., Diehl R.H., Fléron R.W., et al., Technology on the move: recent and forthcoming innovations for tracking migratory birds. Bioscience, 2011, 61, 689-698 [17] Bridge E.S., Kelly J.F., Contina A., Gabrielson R.M., MacCurdy R.B., Winkler D.W., Advances in tracking small migratory birds: a technical review of light-level geolocation. J. Field Ornithol., 2013, 84, 121-137 [18] McKinnon E.A., Fraser K.C., Stutchbury B.J.M., New discoveries in landbird migration using geolocators, and a flight plan for the future. Auk, 2013, 130, 211-222 [19] Hallworth M.T., Marra P.P., Miniaturized GPS tags identify non-breeding territories of a small breeding migratory songbird. Sci. Rep., 2015, 5, 11069 [20] Morris S.R., Covino K.M., Jacobs J.D., Taylor P.D., Fall migratory patterns of the Blackpoll Warbler at a continental scale. Auk, 2016, 133, 41-51 [21] Boulet M., Gibbs H.L., Hobson K.A., Integrated analysis of genetic, stable isotope, and banding data reveal migratory connectivity and flyways in the Northern Yellow Warbler (Dendroica petechia; Aestiva group). Ornithol. Monogr., 2006, 29-78 [22] Ryder T.B., Fox J.W., Marra P.P., Estimating migratory connectivity of Gray Catbirds (Dumetella carolinensis) using geolocator and mark-recapture data. Auk, 2011, 128, 448-453 [23] Johnson D.H., Nichols J.D., Schwartz M.D., Population dynamics of breeding waterfowl, In: Batt B.D., Afton A.D., Anderson M.G., Ankney D.H., Kadlec J.A., Krapu G.L., (Eds.), Ecology and Management of Breeding Waterfowl Minneapolis, Univ. Minnesota Press, Minneapolis, MN, 1992, 446-485 [24] Calenge C., Guillemain M., Gauthier-Clerc M., Simon G., A new exploratory approach to the study of the spatio-temporal distribution of ring recoveries: the example of Teal (Anas crecca) ringed in Camargue, Southern France. J. Ornithol., 2010, 151, 945-950 [25] Korner-Nievergelt F., Liechti F., Thorup K., A bird distribution model for ring recovery data: where do the European Robins go? Ecology and Evolution, 2014, 4, 720-731 [26] Cohen E.B., Hostetler J.A., Royle J.A., Marra P.P., Estimating migratory connectivity of birds when re-encounter probabilities are heterogeneous. Ecology and Evolution, 2014, 4, 1659-1670 [27] Thorup K., Korner-Nievergelt F., Cohen E.B., Baillie S.R., Large-scale spatial analysis of ringing and re-encounter data to infer movement patterns: a review including methodological perspectives. Methods Ecol. Evol., 2014, 5, 1337-1350 [28] Wauer R.H., The American Robin. Univ. of Texas Press, Austin, TX, 1999 [29] Vanderhoff N., Sallabanks R., James F., American Robin (Turdus migratorius). In: Poole A., (Ed.) Birds of North America Online, Cornell Lab of Ornithology, Ithaca, NY 2014, http://bna.birds. , cornell.edu/bna/species/462 [30] Klimkiewicz M.K., Clapp R.B., Futcher A.G., Longevity records of North American birds: Remizidae through Parulinae. J. Field Ornithol., 1983, 54, 287-294 band recovery data. There is still much to understand about the patterns of migration and vagrancy in robins, yet this study reveals more detailed patterns of breedingsite residency through winter months than was previously known. Acknowledgements: This work was supported by the Eastern Kentucky University Graduate Research Council. We thank Shannon Tegge for assistance with GIS analysis. We are grateful to the staff of the USGS Bird Banding Lab for providing the bird band recovery records, and the hard work of hundreds of banders who initiated this data. The authors have no conflict of interest to report. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Animal Migration de Gruyter

Band recoveries reveal alternative migration strategies in American Robins

Animal Migration , Volume (1) – Jul 19, 2016

Loading next page...
 
/lp/de-gruyter/band-recoveries-reveal-alternative-migration-strategies-in-american-296hveTVgi
Publisher
de Gruyter
Copyright
Copyright © 2016 by the
ISSN
2084-8838
eISSN
2084-8838
DOI
10.1515/ami-2016-0004
Publisher site
See Article on Publisher Site

Abstract

Migration strategies may change in response to climate change with consequences for conservation efforts. We used 80 years (1934-2014) of band recovery data (N = 1,057) to describe spatial and temporal patterns in the migration behavior of American Robins. The distribution of recoveries suggests strong continental scale connectivity with distinct separation between eastern and western North America, with a more moderate degree of connectivity within these regions. We also found little evidence of differential migration between males and females. Despite previous studies that suggest the winter distribution of robins has shifted northward, our analysis shows no obvious change in migration distance over time. Surprisingly, we found that a significant proportion of across season band recoveries occurred locally (20%), in close proximity to the original banding locations. It's well known that large numbers of robins linger in northern breeding grounds well into the winter of some years, but the proximity of these birds to breeding areas was previously unknown. We found little evidence that the winter latitude of migrants or local recoveries shifted over time. However, there was a trend for increased frequency of local recoveries in recent decades, providing an alternative hypothesis for the northward shift in winter distribution. Keywords: Turdus migratorius, bird banding, climate change, partial migration, winter residency, distribution shift 1 Introduction Global change is causing taxonomically and geographically widespread changes to bird distributions and migration patterns [1]. In Nearctic and Palearctic migration systems, *Corresponding author David Brown, Eastern Kentucky University, United States, E-mail: david.brown@eku.edu Gail Miller, Eastern Kentucky University numerous species have advanced spring arrival dates and shortened migration distances in response to global change [2­4]. The winter distributions of many species have also shifted north [5­8], in part because of increased winter temperatures, and it is possible that such shifts are driven by shortened migration distances [9]. An alternative but not exclusive hypothesis argues that changes in distribution and timing of arrival can be influenced by the frequency of non-migration strategies [9]. Partial migration, in which some individuals within a population migrate and others are year-round residents, occurs in a great number of species worldwide, perhaps even a majority [10­12], and it is critical that we develop a solid understanding of how such behaviors will allow species to adapt to climate change. Connectivity between breeding and winter populations is another aspect of migration that may have important implications for conservation planning, but it is generally poorly understood for most species [13]. Populations with strong connectivity, in which most of a breeding population migrates to a common wintering area, may be more vulnerable to population declines caused by spatially focused environmental changes. Whereas, the mixing across seasons that occurs in populations with weak connectivity can spread the effects of spatially focused environmental perturbations that occur in one part of the annual cycle [13]. Research has also revealed sex-biased differential migration in many bird species, most commonly in which females winter further south than males [14]. One well supported theory for this pattern is that males remain relatively close to their breeding grounds to arrive earlier for the breeding season and acquire better territories [14]. If such behavior occurs on a continuum, then one end point would be for birds, especially males, to reside yearround on or near their breeding territory. For most species, but in particular for non-game species, there is little data available to describe migration patterns, let alone to test how these patterns might be changing over space and time. Recent work using lightlevel geolocators, stable isotopes, satellite tracking, miniaturized GPS, and other technologies are revealing © 2016 David Brown, Gail Miller, published by De Gruyter Open. This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivs 3.0 License. previously unknown migration behaviors [15­19]. Prior to such advances, bird band recoveries were a vital source of information on bird migration, and they continue to be utilized [20], often in conjunction with these new technologies [21,22]. Studying migration with banding data is an extrinsic technique that requires capture, marking, and a subsequent encounter. Each year, an average of approximately 1 million birds are banded in the United States and Canada (https://www.pwrc.usgs. gov/BBL/homepage/howmany.cfm). Birds are banded primarily as part of research programs, and the capture techniques and sampling designs underlying the bird banding differ widely among species and researchers. We refer to the initial observation and marking of a bird as "banding" and any subsequent recovery or capture as an "encounter." We use the term "recovery" more generally to refer to individual birds that were banded and subsequently encountered. Each year, a fraction of previously banded birds are encountered. Encounters occur through research programs that involve active or passive techniques to capture birds, harvest, and accidental encounters, such as dead birds found after collisions. These encounters typically occur at higher rates for game species (mean = 19%/year for waterfowl) in large part because of harvest. The high encounter rates for some waterfowl have allowed managers to develop population and migration models [23,24]. Non-game species, such as most passerines, have low rates of encounter (mean = 1.2%/year), and thus banding encounters have been less useful for researchers. Encounters can occur within or between migratory periods and years. Encounters within the same season and year may be useful for describing dispersal. To be useful for understanding migration patterns, encounters must span migratory periods, such as between winter and breeding seasons, which may occur within or across a single annual cycle. There are both advantages and disadvantages with this type of data [15]. Advantages include that it is relatively inexpensive, at least on a per bird basis, and the locations are relatively accurate for both the banding and encounter locations. Disadvantages are the low encounter rate for some species, and spatial autocorrelation, particularly for banding locations [25­27]. Perhaps no other bird species in North America is more strongly linked to popular conceptions of migration as the American Robin (Turdus migratorious)[28]. They are among the most abundant and geographically widespread species in North America [29]. Robins are ecologically flexible, inhabiting a wide variety of habitats including forest clearings, riparian zones, agricultural fields, and urban parks throughout the annual cycle [29]. The longevity record for American Robins is 13 years and 11 months [30], and average adult apparent survival is 0.51, by one range-wide estimate [31]. Across much of the United States, there is overlap between the breeding and winter distribution of American Robins, but it is not known to what extent this pattern is influenced by non-migrant, resident behavior. Audubon mid-winter bird counts show that the winter distribution of American Robins is highly variable among years [29], and in some years, large numbers of robins appear to stay well north of their typical winter distribution [32]. These patterns are likely linked to weather and food on northern breeding areas. Moreover, there appears to be a temporal trend in the winter latitude and longitude of American Robins, with the mean of the distribution shifting north and east over time [7,33]. Little is known about migratory connectivity in this species [31]. The social behavior of robins shifts from solitary or paired during the breeding season to gregarious and flocking in the winter, although some wintering birds appear to be solitary for parts of the winter [29]. American Robins are known to be facultative seasonal migrants, or even "vagrants" or "wanderers" in the winter with individuals and flocks moving in response to changes in weather and food availability [29,34]. Based on the frequent observations of large non-breeding season flocks, many birds are likely moving widely away from their breeding territories. Between year return rates to winter sites appears to be low [35], whereas the degree of within-winter site fidelity is largely unknown. There are anecdotal reports of robins that attempted to breed during winter months in states such as Michigan [36], but it is not known if these are birds that spent the spring and summer months in the same vicinity. In fact, we were unable to find peer-reviewed literature reporting robins overwintering on their breeding territory so as to spend the entire year in a single home range, although there are such accounts in popular media. The objective of this study was to provide a descriptive account of the patterns of migration by American Robins using bird band recovery data. The robin recoveries used in this study were collected across the range of robins using data that dates to the inception of federally managed bird banding operations. Specifically, we explored the possibility of migratory connectivity between populations of American Robins. We also estimated distances of migration and the frequency of different migration strategies, and tested for sex-based differences in such patterns. Finally, we tested for changes over time in wintering latitude, migration distance, and frequency of migration strategies, both overall and between sexes. 2 Methods We used 80 years (1934-2014) of North American bird banding data to describe migration patterns of American Robins. All banding data for American Robin recoveries comprising locations where a bird was banded and subsequently encountered (N = 14,709) were acquired from the USGS Bird Banding Laboratory (BBL). This full data set of recoveries included birds banded as early as 1916. We narrowed the dataset by including only birds that were banded and encountered in both a breeding (May­August) and a winter (November­February) season, including data regardless of which of the two seasons the banding event occurred. Thus, we attempted to conservatively exclude banding and encounter events occurring during the main migration period for robins as well as encounters that occurred in the same season as banding. The total number of recoveries that included both a breeding and winter observation comprised 1,346 records dating back to 1916 and including all years up to 2014. The majority of birds were first banded during the breeding season with subsequent winter encounters (89%); the remainder were first banded on the winter grounds and encountered during the breeding season. We included encounters that occurred in non-subsequent seasons (42% of all breeding­ winter recoveries occurred in a non-subsequent season, and the average number of years between banding and recovery was 1.5 ± 1.4 SD). By doing so we assumed that the breeding season locations of individual birds did not change between years. Although some birds undoubtedly disperse to new breeding season locations between years, there is evidence that breeding site fidelity is high, as with many other passerines [37, D. Brown, unpublished data]. Band encounters can occur when birds are harvested, found dead, or captured by a bird banding permit holder other than the original bander, and thus typically in a different location. Using encounters from bird banding stations in studies such as this one introduces known biases because the distribution of bird banding stations is non-random and banders sometimes use methods to increase the chances of recapturing local birds [38]. Thus, we excluded all encounters that were coded as "previously banded bird trapped and released during banding operations" (i.e., birds encountered during banding). The remaining data consisted of birds that were found dead, and a small number that were shot in the 1930's (2.5%), and together form the primary database for analysis (N = 1,057, Figure 1). We included all birds regardless of age at banding and did not attempt to include age as a factor in any analysis (After-hatch-year = 36%, Hatch-year = 35%, Figure 1. Locations of all American Robin recoveries used in this study (N = 1,057) during the breeding season (A), and the winter season (B). Juvenile = 18%, Unknown = 11%), although in some figures we show only after-hatch-year birds to reduce clutter. We were also interested in addressing questions about differences between sexes, for which we only included after-hatch-year birds of known sex (N = 285). We used ArcGIS version 10 [39] to calculate the geodesic distance between banding and encounter locations for each bird. The precision of locations as reported to the BBL are to the nearest degree or half-degree of latitude and longitude, so distance calculations of zero are from within the same coordinate block, but not necessarily the exact same location. We classified all encountered robins as migrants or local recoveries. Birds with a distance between banding and encounter that exceeded 100 km were classified as migrants. Birds were assigned as local recoveries if the distance between banding and encounter events was less than 100 km , approximately 0.9° latitude or 1.15° longitude, following Fielder and colleagues [9]. In selecting the 100 km cutoff we chose to be conservative in our classification of local recoveries since our cutoff for classifying birds as migrants may include birds that moved regionally but did not truly migrate. For example, a bird residing for the winter 150 km from its breeding location is arguably not a migrant. To further describe the general migration patterns we conducted an exponential regression analysis of winter latitude and migration distance, excluding all birds classified as local recoveries. We tested for migratory connectivity between winter and breeding locations using visual interpretations of maps, a descriptive analysis of the frequency of robins within respective winter and breeding bands of longitude, and a statistical comparison among regions based on distance matrices. For the descriptive analysis, longitude bands were assigned arbitrarily to create a small number of relatively even classes: < -110°, -110° ­ -95°, -94.99 ­ -80°, and > -80°. This binning arrangement creates a single block approximately west of the Rocky Mountains (< -110°), and creates more blocks across the eastern US, where most of the robins were banded and encountered. Each bird was assigned to a winter and breeding longitude group. Because crossing longitude classes alone does not inform connectivity, we illustrate the relative frequency among groups, and do not attempt to apply statistical analysis to the pattern. To describe the overall degree of connectivity we used a Mantel correlation coefficient calculated between a breeding season distance matrix and a winter season distance matrix using Program R, Package Ade4 [40]. These within-season distance matrices were calculated assuming spherical geometry using the program Geographic Distance Matrix Generator, version 1.2.3 [41]. We conducted this analysis for the entire dataset of band recoveries, spanning all of North America, and then separately for eastern and western regions of the continent, using the Rocky Mountain divide to separate the regions. We tested for non-random patterns in the spatial distribution of local recoveries using a nearest neighbor analyis in ArcMAP. To test if local recoveries tended to occur in areas of highest breeding relative abundance we used Breeding Bird Survey 1996­2012 analysis results [42] to calculate the relative abundance of robins at each nonmigrant breeding location and compared the frequency of those observations with the total distribution of relative abundance across all BBS grid cells using a two-sample Kolmogorov-Smirnov test. We also compared patterns of recovery, and of proportions of migrant and local recoveries between eastern and western North America using the Rocky Mountain Continental Divide to separate the regions. We compared the average latitude of local recoveries to the average latitude of breeding robins. The average and standard deviation of the latitude of breeding robins was calculated from BBS data by weighting the latitude of the center of each BBS grid cell by the relative abundance and grid cell area. Because of the high number of grid cells within the BBS region (N = 23,000), we used a permutation procedure to generate 100 normal random latitudes of breeding robins. We compared the random breeding latitudes to the latitudes of all non-migrant winter recoveries using a permutation-based Welch's t-test. To investigate the possibility of sex-based variation in migration patterns we conducted a separate analysis of migratory status based on sex. We used two-sample t-tests to compare the winter latitude of males and females separately for migrants and local recoveries. We used a chi-square contingency table to test for differences in the frequency of migrants compared to local recoveries in males and females. We also performed a set of analyses to test for changes in migration patterns over time. First we used simple linear regression to model the relationship between year and migration distance, excluding all birds classified as local recoveries. To test for changes in the proportion of local recoveries, we used a generalized linear model with binomial error distribution and log link. The dependent variable was the migratory status (local recovery or migrant). We were primarily interested in how the relative frequency of migratory status changed over time. Preliminary review of the data suggested that the proportion of local recoveries was higher in recent decades, so we binned years into pre- and post-1980, which is a widely cited cutoff for the appearance of strong effects of extreme global warming [43], and used this binary variable as our primary predictor. We graphically present these results by decade to illustrate the basis for our binning approach. Breeding latitude was included as a covariate. Models with and without the binary time period variable were compared using likelihood ratio tests. All statistical analyses were performed with Program R [40]. 3 Results 3.1 Migrants American Robin band recovery events between winter and breeding seasons peaked during the 1930's and again during the 1960's and have declined to near all-time lows during the last two decades (Figure 2A). Interestingly, the total number of banded birds has not followed the same pattern. In the period for which all banding records were available electronically (1960­2010), an average (±SD) of 20,324 (± 4,969) birds were banded each year (range: 11,784­ 33,478), with the greatest numbers from 1985­1992. The vast majority, 91%, of all recoveries used in this study (N = 1,057) include both breeding and winter season locations in the eastern and central regions of United States, with most of the remainder coming from the Pacific coast region (Figure 1). Few recoveries occurred in the Rocky Mountain region. Eighty percent of all recoveries were classified as migrants (encountered > 100 km from banding location; Figure 2B). Of the migrants, 96% of individuals travelled 500­2,100 km. The longest migration distance was by an after-hatch-year bird of unknown sex that migrated 5,153 km from Fairbanks, AK (August 1959) to central Mississippi (February 1960). Migration distance tended to be greater for robins that wintered farther south (F1,838 = 468.37, P < 0.001, R2 = 0.35, y = 13,596e-0.075x; Figure 3A). Figure 2. Histograms of American Robin band recoveries. (A) Recoveries have declined in recent decades, despite relatively consistent levels of banding. (B) The distance between breeding and winter season locations of band recoveries. The zero-inflated distribution occurs because a relatively large proportion of American Robin across-season recoveries occurred locally (< 100 km from banding site). Figure 3. Migration distance (km) of American Robin band recoveries classified as migrants (> 100 km between banding and encounter locations) in relation to winter latitude (A), and across years (B). 3.2 Connectivity At the continental scale, there appears to be a strong separation of eastern and western populations, with only two records of individuals crossing the Rocky Mountains within the conterminous United States (Mantel test of breeding vs. wintering distance matrix: rm = 0.78, P = 0.001, Figure 4A). One of these was originally banded in eastern Kansas in May 1955 and encountered in January 1956 near Portland, Oregon (this bird does not appear in Figure 4A because it was aged as hatch-year). The other bird that crossed the Rockies was banded in February 1969 near San Francisco, CA, and then encountered in Michigan in June 1969. The anomalous movement of these two birds may indicate reporting problems for these records, such as if the reported locations do not reflect the location that the bird was recovered. There were only eight records of birds crossing -110° longitude, which crosses the Rocky Mountains in Wyoming. Within eastern North America, there was a moderate level of connectivity (rm = 0.34, P = 0.001). For example, most robins wintering in Florida migrated between coastal breeding areas, but 30% migrated more inland to or from Midwest or Great Lakes regions (Figure 4B). Likewise, birds wintering in Texas tended to migrate between Great Plains breeding areas, with fewer migrating to or from Midwest or Northeast regions. Eighty-five percent of birds with breeding season locations within the -95 ­ -80° longitude band were located within the same band during the winter season (Figure 5). Likewise, 65% of individuals with breeding season locations in the -110 ­ -95° longitude band had winter locations in the same band. Within western North America, the connectivity correlation coefficient was considerably lower (rm = 0.12, P = 0.009). 3.3 Local Recoveries Of all recoveries that included both winter and breeding season occurrences (N = 1,057), 20% were classified as local recoveries (i.e. winter-breeding migration distance < 100 km; Figure 2B). Sixty-eight percent of local recoveries occurred within the same degree of latitude and longitude (the precision of band reporting to BBL) of the original banding location and were classified as 0 km distance between the banding and encounter locations. The mean (±SD) distance between banding and encounter locations for local recoveries was 20.8 ± 44.45 km. The winter location of local recoveries occurred, on average, at higher latitudes than the winter location of migrants (mean local recovery winter latitude Figure 4. Lines connecting breeding season (May­August) and winter (November­February) locations of after-hatch-year American Robins. Panel (A) includes males, females, and robins of unknown sex across North America. Panel (B) shows male (blue) and female (red) American Robins for the central and eastern regions of North America. Figure 5. Percent of American Robin breeding season locations relative to winter season locations as categorized into longitude bands. All locations are based on pairs of banding and encounter events. = 40.35, SD = 3.90, median = 39.91; mean migrant winter latitude = 32.12, SD = 3.32, median = 31.41; t1,055 = 31.35, P < 0.001). The distribution of local recoveries exhibited a clustered distribution (nearest neighbor analysis: Ratio: 0.50, Z = -13.81, P < 0.001; Figure 6). There were disproportionately more local recoveries in western North America; 29% of all local recoveries occurred west of -110° longitude, whereas only 11% of all recoveries (i.e., including migrants and local recoveries) occurred in that region (63% of all recoveries in western North America were local, whereas 15% of all recoveries east of -110° longitude were local). There were relatively high levels of local recoveries across all 4 of the winter months, but rates were markedly higher for November (November: 56%, December: 18%, January: 15%, February: 18%). 3.5 Sex differences Of all the local recoveries, 63% were of unknown sex, whereas 76% of migrants were of unknown sex. Among birds of known sex, there was no difference in the frequency of local recoveries compared to migrants in males and females (2 = 1.07, df = 1, P = 0.30; females: 36.7% local recoveries, males: 43.5% local recoveries). The latitude of local recoveries did not differ between males and females (t77 = 0.66, P = 0.50, mean ± SE female: 40.46 ± 0.76; male: 39.80 ± 0.61). Likewise for migrants, winter latitude did not differ between males and females (t191 = 0.14, P = 0.88, mean ± SE males: 32.99° ± 0.34°; females: 32.06° ± 0.36°). 3.6 Temporal trends We found no relationship between migration distance and winter year, excluding local recoveries (F1,838 = 6.94, P = 0.09, R2 = 0.008; Figure 3B). When considering just migrants, winter latitude has not shifted over time (F1,838 = 0.29, P = 0.58, R2 < 0.001; Figure 7A). There was a statistically significant northward shift in winter latitude of local recoveries over time; however, there was a large degree of scatter in the analysis and a low degree of total variation explained (F1,216 = 9.77, P = 0.002, R2 = 0.04; Figure 7B). The percent of recovered American Robins occurring locally appears to have increased since 1980 (2 = 77.13, df = 1, P < 0.001; Figure 8). 3.4 BBS analysis of local recoveries Two of our analyses compared the location of local recoveries to information derived from BBS analysis. First, we found that local recoveries tended to occur in areas of greater breeding abundance, as estimated from BBS data (D = 0.30, P < 0.001). Second, local recoveries tended to occur farther south than the average latitude of breeding American Robins based on the BBS (mean local recovery latitude = 40.35, SD = 3.90, median = 39.91; mean BBS latitude = 43.81, SD = 6.48, median = 43.2; t134.7 = 3.98, P < 0.001). Figure 6. Location of winter season after-hatch-year American Robin recoveries by migration status (migrant [open circle with dot] or local recoveries [solid circle]) and sex (male = blue, female = red) for eastern and central North America. Local recoveries occurred across seasons (winter-to-breeding, or breeding-to-winter) within 100 km of banding locations. Inset shows winter locations of all American Robin local recoveries across North America (solid circles), including those of unknown age and sex. Figure 7. Winter latitude of American Robin band recoveries across time for migrants (A), and local recoveries (B). Local recoveries were classified based on encounters occurring < 100 km from banding location. Figure 8. The percent of breeding-winter recoveries classified as local recoveries by decade. Decades were further binned to pre- and post-1980 for statistical analysis. 4 Discussion 4.1 Residency Our results provide several interesting insights into the seasonal migration patterns of American Robins. American Robins are well known to be vagrants during the winter, moving locally or regionally in response to food availability. It appears that many individuals also remain in close proximity to their breeding site, at least for part of the winter. It is possible that many of these birds are actually migrants or at least vagrants that just happened to be captured near their breeding location before or following movements over larger extents. However, it seems highly unlikely that there would be so many total winter encounter events, including a high proportion that occurred in the middle of winter (i.e., January­February), that occur close to the breeding location if many of these birds were not residing in the area for most if not all of the non-breeding season. Because of the nature of the data used here, there is no way to address such contingencies. The local recoveries made up a significant proportion of the total population and occur across the winter distribution and in similar frequencies in both males and females. Of the individuals observed to be local recoveries, some may have in fact migrated in other years. We found no records of repeat recoveries, even when including encounters from banding operations, so we have no basis to indicate individuals migrate in some years and remain resident in others, but it is possible, especially given the typically vagrant nature of robins during the non-breeding season. We have twice observed robins that remained on their breeding territory in Kentucky during January and February (D. Brown, unpublished data). The two robins were among 35 individuals color-banded during the breeding seasons of 2010­2012, and one each was observed during the winters of 2010 and 2011. Both birds were males that were paired with successfully nesting females during the preceding and subsequent summers in the same area. In both years, the birds were observed acting aggressively towards individuals associated with passing winter flocks. A substantial proportion of local recoveries occurred in the Pacific coastal region, with dense clusters around San Francisco and Seattle, disproportionate to the frequency of migrant recoveries there. These patterns may have resulted from heterogeneous sampling. Alternatively, eastern and western populations of robins may have different migration strategies. There may also be genetic population level differences between regions, which seems likely given that descriptions of populations include distinct western varieties [29,44]. Geographic variation in morphology among populations (or subspecies) suggests distinct lineages, which may also be tied to geographic variation in migration. Environmental and habitat conditions differ considerably across the distribution of robins and the Pacific coastal region has a moderate winter climate compared to most of the Midwestern and Atlantic regions; these difference may help explain the higher incidence of local recoveries in the western region. The spatial frequency of local recoveries corresponds with the relative abundance of breeding birds based on BBS surveys. In other words, local recoveries are more likely to occur in areas of higher BBS abundance. This suggests that local recoveries are not coming from the edge of the range or sub-par habitats that have low relative abundance. However, as mentioned previously, there appears to be a relatively large proportion of local recoveries in our dataset occurring in two clusters in the Pacific coastal region. Chandler Robbins attributed the occasional overwinter occupancy of relatively northern regions by robins to a combination of factors including density dependence, abundant food, and early northward migration in response to mild winter climate [32]. In recent decades, the conditions for all three of these factors have changed to favor increased northern occupancy by robins during winter: (1) The distribution and population size of robins has expanded for decades [29,42]; (2) Expansion of ornamental and invasive fruit bearing shrubs in northern areas have increased winter food supplies [45], with a direct response by robins [46,47]; And, (3) winters have been warmer in recent decades [48]. We can now argue that this northerly occupancy pattern is driven partially by a tendency to remain near breeding areas. This pattern is consistent with theories of partial migration which predict a continuum of migration behavior that may shift in response to environmental factors and in some situations result in increased frequency of residency [12,49]. 4.4 Connectivity Our findings suggest that robins have strong connectivity at the continental-scale. American Robins in the western region of North America migrated between winter and breeding locations within the region, and likewise birds banded east of the Rocky Mountains were almost exclusively encountered in that region. One seeming exception to this pattern is the occurrence of three recoveries in the southeast and central US of birds banded in Alaska; however, this migration pattern may be specific to birds breeding in the extreme northwest (i.e., Alaska), and follows, in part, the pattern of some long distance migrants [52], and may be similar to other short-distance migrants such as White-crowned sparrows (Zonotrichia albicollis) [53]. Other species, including the Yellow Warbler (Setophaga petachia), Wilson's Warbler (Cardellina pusilla), Common Yellowthroat (Geothlypis trichas), Yellow-breasted Chat (Icteria virens), and Nashville Warbler (Oreothlypis ruficapilla) also show separation of populations between eastern and western areas with parallel migration patterns [21,54,55]. These patterns of connectivity in American Robins are consistent with taxonomic descriptions that have divided American Robins into distinct populations [44]. There is only one relevant North American based study with which we can compare our quantitative estimates of connectivity based on Mantel correlation coefficients between breeding season and winter season distance matrices. In that study, 21 Swainson's Thrushes (Catharus ustulatus) breeding in western North America and traced to wintering grounds using light-level geolocators had a Mantel correlation coefficient of connectivity (rm = 0.72) higher than what we found for American Robins banded in western North American, and similar to our continentwide estimate of connectivity [56]. Within the central and eastern regions there appears to be modest migratory connectivity with many birds traversing east-west, while others remain in relatively narrow longitudinal bounds. The graphical patterns of robins crossing between eastern and central North America reinforce these quantitative estimates of moderate to diffuse connectivity. The band recovery data has relatively few recoveries from the Rocky Mountain region, where robins breed and winter in moderate levels of relative abundance [42]. Some studies based on mitochondrial DNA have failed to detect genetic structure or migratory connectivity at finer scales [55]. However, other methodological approaches, such as geolocators, satellite transmitters, morphology, molecular analysis of single nucleotide polymorphisms, 4.2 Migrants Because of the known vagrant movement patterns of robins during the non-breeding season and the acknowledged problems inherent with sampling heterogeneity in bird banding data, we exercise caution in making strong inferences about the patterns observed with the migrant birds. Nonetheless, we can take away several meaningful insights from our analysis: there is no evidence that average migration distances have changed over time; migration distance is related to wintering latitude; there are complex patterns of spatial connectivity; and, sexbased difference in migration patterns are not apparent. 4.3 Temporal patterns Based on evidence that the winter distribution of many neotemperate migrants, including robins, have shifted north in recent decades [7,33,50], we expected to find a decreasing migration distance over time. However, there is no such pattern in the recoveries of robins. One plausible explanation is that the breeding range has expanded northward concomitantly with the shift north in wintering range, and there is evidence that the range of robins has indeed expanded [29]. Another explanation for the northward shift of the winter range in the absence of changes in average migration distance is the increasing frequency of non-migrant behavior, which we have documented in this study. Other species of partial migrants, such as the Blackcap (Sylvia atricapilla) have increased the proportion of individuals maintaining year-round residency [9], and some populations of this species that were exclusively migratory have adapted to be partial migrants with individuals becoming year-round residents [49]. However, such patterns are very difficult to detect. In North America, a region of the world where birds are relatively well studied, a number of species are thought to be partial migrants (e.g., Eastern Bluebird [Sialia sialis], Eastern Towhee [Pipilo erythrophthalmus], Blue Jay [Cyanocitta cristata], Song Sparrow [Melospiza melodia]) [51], but we know little about relative frequencies of these behaviors. and bird band recoveries, have revealed connectivity in other species at finer scales [22,57,58] and weaker levels [59]. For example, Gray Catbirds (Cumetella carolinensis) breeding in the mid-Atlantic and Midwestern regions migrate to separate wintering areas [22], and genetically distinct breeding populations of Wilson's Warbler migrate to different wintering regions [58]. 4.5 Sex-based differences Surprisingly, we found no evidence of sex-based differences in migration, or in the latitude of winter recoveries. Differential migration, in which males winter farther north and migrate shorter distances than females, appears to be a widespread phenomenon among neotemperate migrants [14,60,61]. The facultative migration strategies and gregarious behaviors of robins, and the likely variation in these patterns among years, may have obscured our ability to detect differences in migration between males and females. Additionally, we used latitude as the primary indicator of distance from breeding grounds, which is simplistic and ignores how climate, habitat, and connectivity affect migration of males and females. Arguably, more robust and less biased datasets (e.g., museum collections, all banding records of a species, geolocators) are better suited for such tests [61,62], and it may be that we failed to detect any difference between males and females (i.e., Type II error). 4.6 Assumptions and Limitations As acknowledged earlier, one important bias of this dataset is that bird banding does not occur at random locations or times. We have assumed spatially and temporally homogenous sampling, but this assumption is clearly problematic, and thus the patterns we describe must be cautiously interpreted. Bird banding data include variation in banding, recaptures, and encounters, as well as habitat and other environmental factors [9]. Intensive banding within a region or time period increases the number of subsequent encounters and does not systematically account for variation in habitat. Recently developed models create possibilities to address such issues [25­27], but these models are beyond the scope of the descriptive approach that we have taken. However, our descriptive results should provide direction and testable hypotheses for future modelling efforts. If banding efforts have shifted southward in recent decades then the observation of increased frequency of local recoveries in recent decades may be a sampling artifact, since local recoveries were more likely to occur in the southern region of the breeding distribution. A related issue is that the number of recoveries has decreased in recent decades, while the total number of local recoveries has remained similar or increased. The former pattern would be expected if in fact the total number of robins banded each year had declined, but that has not been the case. Our work also suggests that winter season banding can be an important source of information about nonmigratory strategies. Of all banded robins that were later encountered in a different season (winter-to-breeding or breeding-to-winter), 30% of local recoveries were originally banded during the winter, whereas only 9% of migrants were banded during the winter. The number of robins recovered has decreased significantly in recent decades, despite relatively constant numbers of robins banded. We are unaware of changes in reporting of band recoveries that would cause such a pattern. Thus, there should be no bias in our main finding that many American Robins stay near their breeding grounds through much if not all of the winter, and that the population-level frequency of local recoveries has increased in recent decades. Our selection of months for the breeding season may have resulted in inclusion of some birds that had undergone post-breeding dispersal, and so their breeding season capture or recovery locations may differ from their actual breeding location. Additionally, within the winter season we may have included birds that were near their breeding locations, but that eventually migrated (i.e., November) or had already returned from migration (i.e., February). However, such lingering on the winter grounds should be weather and food dependent, so our conclusions about changes in these patterns over time should reflect behavioral responses to changes in environmental conditions across years. Emerging technologies are revolutionizing our understanding of bird migration [16]; however there is still much that can be learned from traditional sources of information such as citizen science [52], and the bird banding data reported here. Currently, guidelines for bird banding permit holders do not require reporting of recaptures of birds they originally banded. Such data would likely provide more insight into patterns of non-migrant behavior. Our study makes no direct links between climate and migration, and clearly more complex modelling would be required to untangle such relationships, especially given the variability in robin migration patterns, and the assumptions of using bird [15] Hobson K.A., Norris D.R., Animal migration: a context for using new techniques and approaches. In: Hobson, KA, Wassenaar, LI (Eds.), Tracking Animal Migration Using Stable Isotopes, Academic Press, London, 2008, 107-128 [16] Bridge E.S., Thorup K., Bowlin M.S., Chilson P.B., Diehl R.H., Fléron R.W., et al., Technology on the move: recent and forthcoming innovations for tracking migratory birds. Bioscience, 2011, 61, 689-698 [17] Bridge E.S., Kelly J.F., Contina A., Gabrielson R.M., MacCurdy R.B., Winkler D.W., Advances in tracking small migratory birds: a technical review of light-level geolocation. J. Field Ornithol., 2013, 84, 121-137 [18] McKinnon E.A., Fraser K.C., Stutchbury B.J.M., New discoveries in landbird migration using geolocators, and a flight plan for the future. Auk, 2013, 130, 211-222 [19] Hallworth M.T., Marra P.P., Miniaturized GPS tags identify non-breeding territories of a small breeding migratory songbird. Sci. Rep., 2015, 5, 11069 [20] Morris S.R., Covino K.M., Jacobs J.D., Taylor P.D., Fall migratory patterns of the Blackpoll Warbler at a continental scale. Auk, 2016, 133, 41-51 [21] Boulet M., Gibbs H.L., Hobson K.A., Integrated analysis of genetic, stable isotope, and banding data reveal migratory connectivity and flyways in the Northern Yellow Warbler (Dendroica petechia; Aestiva group). Ornithol. Monogr., 2006, 29-78 [22] Ryder T.B., Fox J.W., Marra P.P., Estimating migratory connectivity of Gray Catbirds (Dumetella carolinensis) using geolocator and mark-recapture data. Auk, 2011, 128, 448-453 [23] Johnson D.H., Nichols J.D., Schwartz M.D., Population dynamics of breeding waterfowl, In: Batt B.D., Afton A.D., Anderson M.G., Ankney D.H., Kadlec J.A., Krapu G.L., (Eds.), Ecology and Management of Breeding Waterfowl Minneapolis, Univ. Minnesota Press, Minneapolis, MN, 1992, 446-485 [24] Calenge C., Guillemain M., Gauthier-Clerc M., Simon G., A new exploratory approach to the study of the spatio-temporal distribution of ring recoveries: the example of Teal (Anas crecca) ringed in Camargue, Southern France. J. Ornithol., 2010, 151, 945-950 [25] Korner-Nievergelt F., Liechti F., Thorup K., A bird distribution model for ring recovery data: where do the European Robins go? Ecology and Evolution, 2014, 4, 720-731 [26] Cohen E.B., Hostetler J.A., Royle J.A., Marra P.P., Estimating migratory connectivity of birds when re-encounter probabilities are heterogeneous. Ecology and Evolution, 2014, 4, 1659-1670 [27] Thorup K., Korner-Nievergelt F., Cohen E.B., Baillie S.R., Large-scale spatial analysis of ringing and re-encounter data to infer movement patterns: a review including methodological perspectives. Methods Ecol. Evol., 2014, 5, 1337-1350 [28] Wauer R.H., The American Robin. Univ. of Texas Press, Austin, TX, 1999 [29] Vanderhoff N., Sallabanks R., James F., American Robin (Turdus migratorius). In: Poole A., (Ed.) Birds of North America Online, Cornell Lab of Ornithology, Ithaca, NY 2014, http://bna.birds. , cornell.edu/bna/species/462 [30] Klimkiewicz M.K., Clapp R.B., Futcher A.G., Longevity records of North American birds: Remizidae through Parulinae. J. Field Ornithol., 1983, 54, 287-294 band recovery data. There is still much to understand about the patterns of migration and vagrancy in robins, yet this study reveals more detailed patterns of breedingsite residency through winter months than was previously known. Acknowledgements: This work was supported by the Eastern Kentucky University Graduate Research Council. We thank Shannon Tegge for assistance with GIS analysis. We are grateful to the staff of the USGS Bird Banding Lab for providing the bird band recovery records, and the hard work of hundreds of banders who initiated this data. The authors have no conflict of interest to report.

Journal

Animal Migrationde Gruyter

Published: Jul 19, 2016

References