Urban Forest Patches and Stopover Duration of Migratory Swainson's Thrushes

Abstract.

Long-distance migration between nonbreeding and breeding grounds involves use of multiple stopover sites where birds must refuel to meet the energetic demands of migration. Migrant forest birds that encounter urban landscapes experience high levels of habitat fragmentation and often use small, isolated forest patches. We investigated factors influencing stopover duration (length of stay) and migratory departure of Swainson's Thrushes (Catharus ustulatus) using forest patches within an urban landscape in Columbus, Ohio. During May, 2004–2007, we fitted 69 Swainson's Thrushes with 0.66-g radio transmitters, experimentally relocated the birds to seven mature forest sites that varied in area and degree of urbanization and monitored each individual daily until it departed. Mean minimum stopover duration was 3.7 days (± 3.4 SD), ranging from 1 to 12 days. Stopover duration was negatively related to advancement of migration (Julian date) and energetic condition at capture, and departure from stopover sites was associated with low wind speed and declining barometric pressure. We found no difference in stopover duration across the seven urban sites despite large variation in forest area and degree of urbanization. Swainson's Thrushes had strong site tenacity at the five largest (11.9–38.4 ha) sites, where 100% of birds remained until migratory departure; in contrast, 28% of individuals left the smallest (0.7 and 4.5 ha) sites during stopover. Results suggest that patches of remnant forest within cities may provide Swainson's Thrushes with suitable stopover opportunities without strongly influencing their migratory schedule, but minimum area requirements may limit the thrushes' use of some patches.

Resumen.

La migración de larga distancia entre territorios no reproductivos y reproductivos involucra el uso de múltiples sitios de parada donde las aves deben reabastecerse para alcanzar las demandas energéticas de la migración. Las aves de bosque que se encuentran con paisajes urbanos durante la migración experimentan altos niveles de fragmentación de hábitat y usualmente utilizan pequeños fragmentos de bosque aislados. Investigamos los factores que influencian la duración de las paradas (largo de la permanencia) y la partida migratoria de individuos de Catharus ustulatus que usan fragmentos de bosque en un paisaje urbano en Columbus, Ohio. Durante el mes de mayo de los años 2004 al 2007, colocamos radio transmisores de 0.66 g a 69 individuos de C. ustulatus, relocalizamos de modo experimental a las aves en siete sitios de bosque maduro que variaron en superficie y grado de urbanización y monitoreamos diariamente cada individuo hasta su partida. La duración promedio minima de la parada fue de 3.7 dias (± 3.4 DE) y varió entre 1 y 12 días. La duración de la parada se relacionó negativamente con el avance de la migración (fecha juliana) y la condición energética al momento de la captura. La partida desde los sitios de parada se asoció con una baja velocidad del viento y disminución en la presión barométrica. No encontramos diferencia en la duración de las paradas entre los siete sitios urbanos, a pesar de una gran variación en el tamaño de los fragmentos y el grado de urbanización. Catharus ustulatus presentó) una gran tenacidad al sitio en los cinco sitios de mayor tamaño (11.9–38.4 ha), donde el 100%) de las aves permanecieron hasta la partida de la migración; en contraste, el 28%) de los individuos dejaron los sitios de menor tamaño (0.7 y 4.5 ha) durante la parada. Estos resultados sugieren que los fragmentos de bosque remanente en las ciudades pueden brindar oportunidades de parada adecuadas para C. ustulatus sin afectar fuertemente sus esquemas migratorios, pero los requerimientos mínimos de área pueden limitar el uso de C. ustulatus de algunos parches.

Introduction

Birds have developed many strategies to accomplish their annual migrations between breeding and nonbreeding ranges (Alerstam and Hedenström 1998). With few exceptions, migratory birds cannot complete these journeys in a single flight and must use a series of stopover locations along their migratory route (Moore et al. 1995). The habitats used by migratory birds are vital for resting and restoring energy, making stopover a key period within migration. This is significant given that events during spring migration may influence migratory songbirds' subsequent reproduction (Smith and Moore 2003, Newton 2006), and may ultimately limit populations of some migrants (Sillett and Holmes 2002).

With the expansion of cities worldwide, migrating birds encounter urbanization increasingly, yet the details of how migrants are influenced by urban habitats during stopover remain largely unstudied. In contrast to extensively forested landscapes, urban forests are more fragmented, smaller in area, have more non-native invasive plants and nest predators, and have a lower diversity and abundance of arthropods (McKinney 2002). For many breeding birds, these attributes are indicators of lower-quality habitat and may explain declines in avian diversity and species richness associated with increasing urbanization (Friesen et al. 1995, Blair 1996, Rodewald and Bakermans 2006). However, urban forest patches may provide important energetic resources for migrants during stopover (Seewagen and Slayton 2008) but could also function as an energetic sink for migrants by not allowing efficient accumulation of energetic stores.

Birds that use urban landscapes during migration likely are constrained by habitat availability and have few options when selecting areas for feeding and resting as they arrive. Recent studies indicate that during stopover migrants may settle into urban landscapes without regard to degree of urbanization. For example, in Columbus, Ohio, Rodewald and Matthews (2005) found species richness and abundance of migrants to be higher in upland forests than in riparian forests but not associated with urbanization within 1 km. Similarly, migrants' use of habitat in Cincinnati, Ohio, was positively associated with canopy cover of native trees and patch width but was not associated with the extent of urbanization (Pennington et al. 2008).

To evaluate the conservation importance of urban areas as stopover sites, it is necessary to determine not only the numbers of migrant birds that use urban environments but also to quantify behavioral responses of birds to these habitats. In spring migration, birds are under pressure to arrive at the breeding grounds to establish territories, and timing of arrival can influence reproductive success (Alerstam and Hedenström 1998, Drent et al. 2006). Among the most important decisions birds make during migration are the amount of time to stop over (Bachler and Schaub 2007) and when to depart on migration (Tsvey et al. 2007). During migration, an individual's decisions regarding the duration of stopover are influenced by its energetic state and time pressures (Hedenström and Alerstam 1997). Weather can influence stopover decisions strongly by promoting (e.g., in spring, low-pressure systems, warm temperatures, and favorable winds) or delaying departure (e.g., cooler temperatures, opposing or strong winds) (Richardson 1990, Schaub et al. 2004). Ultimately, stopover decisions are likely triggered by some minimum fuel load and further conditioned by local weather and geographical factors (Hedenström 2008). Migration timing, energetic state, and the ability to take advantage of favorable weather may be influenced by urbanization (or other land-use changes) if birds are unable to forage and refuel efficiently. This has the potential to lengthen stopovers, lower energetic state, and ultimately delay arrival in breeding areas.

The goal of our study was to examine if the stopover behavior of Swainson's Thrush (Catharus ustulatus) differs among different urban forest patches and whether urban areas represent suitable stopover habitat for the species. Swainson's Thrush is a common nearctic-neotropical migrant and has been used as a model species in migration research (e.g., Cochran et al. 1967, Wikelski et al. 2003, Bowlin et al. 2005, Tietz and Johnson 2007). Subspecies C. u. swainsoni spends the nonbreeding season in South American forests and migrates at night to breeding areas in boreal coniferous forests of North America (Mack and Yong 2000). Using radio telemetry, we quantified the stopover duration and departure decisions of Swainson's Thrushes within an urban landscape in the mid-western United States to identify how attributes of urban forests influence stopover behavior. We adopted an experimental approach by relocating Swainson's Thrushes to seven forests of different sizes to explicitly test for individual, site-level, and temporal influences on stopover duration. In addition, we considered the effects of weather conditions and Julian date on decisions associated with departure from these sites. Our primary working hypotheses were that stopover duration should be more strongly associated with differences among urban forest patches than with individual (energetic condition) or temporal factors, and that the departure of Swainson's Thrushes from these stopover sites should be more strongly associated with weather.

Methods

Study Area

Our study sites were located within metropolitan Columbus, central Ohio (city center: 39° 96′ N, 83° 00′ W), where buildings, roads, and mowed grasses increase along a rural-to-urban gradient (Rodewald and Shustack 2008) and isolated mature forests (80–100 years old) are the dominant habitat available to forest birds during migratory stopover (Rodewald and Matthews 2005). We conducted this study at seven sites of mature deciduous forest selected to vary in area and degree of urbanization within 1 km. We used this radius because it corresponds to linear distances of daily movement during stopover of two other migratory songbirds, the European Robin (Erithacus rubecula) (Chernetsov and Mukhin 2006) and Summer Tanager (Piranga rubra) (Aborn and Moore 1997). The mean minimum distance between sites was 3.6 km, and there was no overlap between sites (closest sites >2 km apart). After digitizing aerial photographs of Delaware and Franklin counties taken in 2002 and 2004, we quantified the area (ha) of patches where birds were released, the patches' distance to the city center, and eight other landscape metrics within a 1-km radius of the patches, including total length of roads (m), number of buildings, and percent of land covered by forest, mowed grass, agriculture, shrubland, residential development, and paved surfaces (Fig. 1, Table 1).

Mist Netting

We captured all Swainson's Thrushes at the Waterman Farm (12.6 ha) forest site, where we operated nine to thirteen 12-m × 3-m mist nets from sunrise to 10:00 (EDT), daily from 1 May to 2 June, 2004–2007. We removed captured birds from mist nets in under 10 min and banded each with a U.S. Geological Survey aluminum leg band. We recorded date, time of capture, wing chord (±0.5 mm), tarsus length (±0.01 mm), and body mass (±0.01 g). We estimated visible subcutaneous body fat on a 6-point ordinal scale (0–5) (Helms and Drury 1960). We aged birds by plumage as after second year, second year, or after hatch year (undetermined) (Pyle 1997); we were unable to sex them. After recording biometric data we placed each bird in a cloth holding bag.

Relocation Experiment

We relocated migrant Swainson's Thrushes experimentally to examine how differences in forest area and degree of urbanization influence stopover duration. Because migrants may settle at a stopover site before the day of capture, and it was not possible to know the date of captured individuals' arrival at a given site, this approach also allowed us to control for prior experience with a site.

We assigned captured birds to one of seven predetermined forest sites (Fig. 1, Table 1) so that sample sizes at each study site were similar and birds with different fat scores (0–1, 2–3, and 4–5) would be released at each site. We relocated thrushes to assigned sites by automobile. One of the seven sites (Waterman Farm) served as a control to test for potential effects of the relocation experiment. Birds released at the control site received the same treatment with respect to capture, handling, time held in the cloth bag, and attachment of a transmitter but were not driven to their release site.

Radio Telemetry

Upon reaching the release site, we removed birds from cloth bags and attached a 0.66-g radio transmitter (model BD-2, Holohil Systems Ltd., Carp, Ontario, Canada) to back feathers with LashGrip eyelash cement (Kenward 2001). Transmitters weighed approximately 2% of the thrushes' mean body mass (30.56 g ± 2.7 SD) and had a life span of at least 14 days. We visually examined each individual for signs of stress (e.g., bill open, eyelids partly shut, less alert) and released three birds that appeared stressed before attaching a radio transmitter. We released 91 experimental thrushes within 50–70 min of initial capture at a randomly selected point within each site. We recorded the behavior of thrushes upon release, including the distance flown on initial flight, height of first perch, and a qualitative assessment of whether the transmitter appeared to interfere with flight behavior.

We radio-located thrushes daily with a hand-held 3-element Yagi antenna (150 MHz, Advanced Telemetry Systems, Isanti, MN) and a 12-channel receiver (Communications Specialists, Inc., Orange, CA) with an estimated detection range of 0.7–1.2 km. We attempted to sight each bird detected by telemetry after approaching to within ca. 50 m. Birds that remained on site (i.e., did not migrate overnight) for >1 day were radiolocated twice on each subsequent day, once during the morning (06:30–12:30) and once in the afternoon (12:30—sunset), with locations separated by at least 4 hr. We collected these data until the bird initiated a migratory flight or the transmitter fell off (21% of individuals). We defined stopover duration as the difference (in days) between the dates of initial release and final radio-location. When technicians were unable to get a transmitter signal, they traversed the study site, spending at least 45 min listening for the frequency and looking for the bird. If necessary, we used a vehicle with a car-mounted whip antenna and systematically sampled all areas within a 1.5-km radius of the study site. After taking these steps, we assumed birds not radio-located to have departed from the study area. However, we checked the transmitter's frequency at each site for the next three days to confirm that the bird did not return.

Attaching radio transmitters will influence animals in some way (Millspaugh and Marzluff 2001), but data on stopover decisions of individual songbirds cannot currently be obtained with other methods (Bachler and Schaub 2007). Swainson's and other Catharus and Hylocichla thrushes have served as model species in migration research since the late 1960s (e.g., Cochran et al.1967, Wikelski et al. 2003, Cochran and Wikelski 2005, Bowlin et al. 2005). Each of these studies attached radio transmitters (0.6–2 g) to thrushes and used radio telemetry to track nocturnal migratory flight. Other research has used radio telemetry to examine movements and habitat use of Swainson's Thrushes during post-breeding dispersal (White and Faaborg 2008) and fall stopover (Tietz and Johnson 2007). Wood Thrushes (Hylocichla mustelina) carrying 1.6-g transmitters (4% of body mass) for two years survived at rates similar to control birds and did not lose mass relative to control birds (Powell et al. 1998). Given the history of use of radio telemetry on Swainson's Thrushes, our use of relatively light transmitters (0.66 g or ca. 2% of body mass) should not have influenced experimental birds strongly. Moreover, our interest was in testing for differences in stopover decisions among study sites. With the exception of control birds, all individuals received the same treatment, so there should have been no systematic bias in our use of radio telemetry and experimental relocation.

Weather

We included weather in our analyses because it has been shown to have an important effect on the duration of migratory songbirds' stopover and timing of their departure (Rappole and Warner 1976, Yosef et al. 2006, Tsvey et al. 2007). We obtained daily weather data from Don Scott Airport, which is located 1–10 km from study sites. We used a suite of weather variables to explore departure decisions, including maximum daily temperature, mean daily temperature, mean daily barometric pressure, change in daily barometric pressure, mean wind speed, day (Julian date), and year to account for annual variability. Whenever an experimental bird(s) was in the study area, we classified days as either “non-departure” (i.e., no experimental birds migrated the previous night) or “departure” (i.e., one or more birds initiated migration).

Data Analysis

In analyses, we used the program R 2.8.0 (R Development Core Team 2008). To adjust for individual differences in body size, we incorporated structural measurements and body mass into one metric of energetic condition (Green 2001). We conducted a principal component analysis on a correlation matrix of wing and tarsus length. Scores from the first principal component represented a standardized structural metric of the bird. We regressed mass at time of capture against this structural component and used residuals as an index of energetic condition (Green 2001). We considered individuals with positive condition scores (residuals) to be in energetic condition better than the average of all migrants sampled. To evaluate this index of energetic condition with an alternative measure of condition (Schulte-Hostedde et al. 2005), we compared it with visual estimates of body fat and found a strong association (Spearman's ρ = 0.7, P < 0.05) between the two indices.

To examine the determinants of stopover duration we used an information-theoretic approach and Akaike's information criterion with a correction for small sample size (AICc) to rank candidate models (Burnham and Anderson 2002). We generated an a priori set of 13 candidate models by using a null model, energetic condition of the bird at capture, site, Julian date of capture, and year to examine patterns of stopover duration. We included site as a factor to examine how site-level differences influenced stopover duration; the various cover variables (e.g., percent cover of residential area) were not included because these would not properly account for degrees of freedom because of site-level dependencies among individuals and would thereby introduce bias to the model-selection criteria. Because data on stopover duration were skewed to the left and had a variance greater than the mean, we used a negative binomial distribution for analyses (White and Bennetts 1996). We calculated the log-likelihood ratio and used a G-test to evaluate the influence of the experimental relocation on stopover duration. We did not include age in the analysis because eight individuals were not aged because of conflicting criteria. Instead, we examined differences between age and stopover duration separately with a Mann—Whitney U-test. Because migratory passage often occurs in waves, we also tested for age-related differences in stopover duration through time by separating spring migration into three equal time periods, 2–11 May, 12–21 May, and 22–31 May. We used a χ2 test for heterogeneity (Zar 1999).

To examine the influence of weather on the thrushes' stopover period, we tested how weather affected departure from a site instead of how it affected stopover duration. This provided a more direct assessment of the influence of weather on the thrushes' departure decisions. We used Random Forest, a robust tree-based method, to quantify departure decisions associated with maximum daily temperature, mean daily temperature, mean daily barometric pressure, change in daily barometric pressure, and mean wind speed; day and year were included as predictor variables to account for temporal variation. This method is an extension of classification and regression trees (Breiman et al. 1984) that incorporates a bootstrapping component to generate a suite of classification trees (Cutler et al. 2007). We evaluated the resulting predictive model to determine the percentage of days correctly classified as departure or nondeparture. When using Random Forest in modeling departure decisions, we determined the importance of each variable by examining the mean decrease in the Gini index (i.e., splitting criterion for the classification trees, Brennen et al. 1984) following iterations of the model with and without permuting the variable.

Results

During spring migration of 2004 to 2007, we attached radio transmitters to 91 Swainson's Thrushes that were relocated to one of seven locations, which each received 11–15 birds. The birds were captured from 2 to 31 May, with peak migration (>50% of all captures) from 12 to 21 May. Of 91 individuals monitored, we recorded three mortalities (3.3%), including one building strike, one predation, and one of undetermined cause. Twenty-one percent of transmitters fell off birds before they initiated a migratory flight, and we did not use data from these in the analyses we report here. We monitored 69 individuals (8–12 individuals per site) from release until departure: Chadwick Arboretum (9), Don Scott Airport (9), Highbanks Metropark (12), Lazelle Woods (9), Rush Run Park (11), Waterman Farm (8), Woodward Park (11)]. Mean distance moved was 313 m (±182.7 SD) during the first three days of stopover. At the five largest release sites (11.9–38.4 ha), 100% of birds remained within the site until migratory departure, whereas at the two smallest sites (0.7 and 4.5 ha), 27.8% of individuals moved from the site of their initial release.

The mean duration of the birds' stopover was 3.7 days (± 3.4 SD), and stopovers ranged from 1 to 12 days. We detected no difference in stopover duration between relocated birds and birds released where captured at Waterman Farm (Fig. 2, G6 = 6.35, P = 0.38), indicating that experimental relocation did not bias our results on stopover duration. When the determinants of stopover duration were modeled, the top two models contained the majority of the model weight (w1 to ₂ = 0.62, Table 2). In our set of candidate models, stopover duration was most strongly influenced by Julian date (day); the energetic condition of the bird at capture also provided important information (Table 2). Stopover duration was negatively related to Julian date (β = -0.052 ± 0.018 SE) (Fig. 3), and, after we controlled for date (Julian date of capture) (second-ranked model), duration was also negatively associated with the bird's energetic condition (β = -0.014 ± 0.006 SE). Despite low power in the modeling effort (r2 = 0.18), the top model carried far more support than the null model (ΔAICc = 4.1), suggesting that averaging across the season and not considering Julian date would produce a pattern similar to random. There was virtually no support for stopover duration being influenced by site-level differences because ΔAICc was greater than 12.3 for all models that included site as a variable. Finally, stopover duration did not differ by age class (U61 = 353.5, P = 0.11), and there were no differences in the number of after-second-year and second-year birds when the season was partitioned into three equal periods, 2–11 May, 12–21 May, and 22–31 May (χ22 = 1.52, P > 0.25).

From 2004 to 2007, there were 94 days when a radio-tagged bird(s) was present at our study sites and either remained (nondeparture) or departed the following night (was not recorded again in, around, or at other sites). Model-classification accuracy was 72%, with departures and nondepartures correctly classified 66% and 77% of the time, respectively. The variables that contributed most to modeling departure decisions (i.e., resulted in greatest change in the Gini index) were Julian date, change in barometric pressure, and wind speed (Table 3). Declining barometric pressure and low wind speed both resulted in a higher likelihood of departure, and neither was associated with Julian date (Spearman's ρ = 0.04 and -0.16, respectively, P > 0.10).

Discussion

The length of time that a migratory bird remains at a stopover site has an important influence on the timing of migration and can ultimately have reproductive consequences if arrival in breeding areas is significantly delayed (Drent et al. 2006). As a result, stopover duration should provide important information for evaluating the relative quality of stopover habitats. We found that the duration of spring stopover of Swainson's Thrushes within an urban landscape did not vary by site, despite high variation in both the area of forest at the release site and urbanization (i.e., buildings and roads) in the surrounding landscape. Although the extent of a forest patch may be important for some species, the lack of a strong site-level influence on stopover duration suggests that Swainson's Thrushes were somewhat flexible in habitat needs and were able to meet their stopover requirements within urban forest patches. In Hylocichla and Catharus (including Swainson's) thrushes, Yong and Moore (1993) have experimentally shown that as energetic condition improves, individuals become more active at night (i.e., migratory restlessness) and presumably are more likely to depart on nocturnal migration. Therefore, if our study sites differed strongly in habitat quality, we should have observed differences among sites in stopover duration.

Urbanization can negatively affect many species of birds breeding in temperate forests (e.g., Friesen et al. 1995, Palominoa and Carrascal 2007). Our results suggest that during migratory stopover, Swainson's Thrushes are less sensitive to urban development. Increased habitat plasticity during stopover could allow migrants to take advantage of available resources (Petit 2000), a result consistent with Pennington et al. (2008), who found that migrating songbirds occurred in more urbanized sites than did locally breeding species. However, Swainson's Thrushes (and potentially other migrant landbirds) may still require a minimum area for forest stopover sites. We found that 100% of thrushes relocated to larger sites (11.9–38.4 ha) remained within the release site until migratory departure, whereas in small patches (0.7 and 4.5 ha) 28% of experimental birds moved away from the release site. Therefore, although we found no difference among urban sites in stopover duration, sites apparently were not equal in value to migrant Swainson's Thrushes. Thrushes that left the patch where they were released moved from 100 to 1000 m from the patch and settled in small hedgerows where they remained for 1 to 2 days. None of these thrushes was observed to return the patch of original release during stopover, and we did not observe individuals moving between forest study sites.

The relationship between sites and stopover duration within an urban landscape highlights that the variation in stopover duration within a site was greater than variation between sites. Migration theory suggests that individuals should base their stopover decisions on energetic state, yet field studies have produced inconsistent results (Chernetsov and Mukhin 2006). After accounting for Julian date, we found that energetic condition at capture was negatively related to stopover duration. At our urban study sites, the duration of Swainson's Thrushes' stopover declined with the advancing season, and by late May no age class stayed longer than two days. Although it is difficult to explain, early migrants may need to stop longer to acquire adequate resources because cooler temperatures may slow the emergence of arthropods. Alternatively, stopover duration may shorten later in migration if birds are under increased pressure to arrive at the breeding grounds on an optimal date.

Weather is an extrinsic factor that can influence migratory departure (Weber and Hedenström 2000, Tsvey et al. 2007), and understanding weather's influence on migratory behavior is important because weather can vary independently of other factors. We found that departure was influenced by a temporal component (Julian date) as well as changes in barometric pressure and mean wind speed. Departure was more likely under light winds following a drop in barometric pressure, a finding consistent with that of Richardson (1990). Our combined information on stopover duration and departure suggests that Julian date can influence how long a bird remains at a given stopover location but that individuals adjust the precise date of departure on the basis of weather.

These findings suggest that remnant forests within urban landscapes have conservation value for Swainson's Thrushes, and, potentially, other migrant landbirds. Beyond simple occurrence within small patches of urban forest (i.e., Rodewald and Matthews 2005, Pennington et al. 2008), we found that such areas provide Swainson's Thrushes opportunities for extended stopovers. Maintaining or restoring mature forest within urban landscapes may benefit migrant landbirds, but small, isolated habitat patches without links to the forest mosaic (e.g., our 0.7- and 4.5-ha study sites), may fail to meet the requirements of some species. We encourage research on additional species that addresses patch-area requirements and habitat quality within urban areas by specifically examining birds' movement patterns and refueling during stopover.

Acknowledgments

We thank Columbus Recreation and Parks and E. Grody for permission for research at Lazelle Woods, Woodward, and Rush Run parks. Columbus Metro Parks and J. O'Meara kindly gave permission for research at Highbanks Metro Park. Finally, we acknowledge all who recorded the data for this project; M. Labbe, S. Beaudreault, J. Fear, E. Dorsay, J. Cummings, J. Sauter, C. Stanton, J. Sullivan, and K. Kaminski. Funding was provided by the Federal Aid in Wildlife Restoration Program (W-134-P, Wildlife Management in Ohio) and administered jointly by the U.S. Fish and Wildlife Service and the Ohio Department of Natural Resources—Division of Wildlife. Field methods were approved by the Ohio State University Institutional Animal Care and Use Committee (IACUC permit 2009A0034).

Literature Cited