Site-Specific Variation in Food Resources, Sex Ratios, and Body Condition of an Overwintering Migrant Songbird

Abstract

Territorial social behavior of wintering Nearctic—Neotropic migrant songbirds places males and females in direct conflict over access to winter space and resources. Outcomes of this intersexual competition can vary by species and habitat, but information has been collected for only a small subset of migrant species. We investigated the available food resources, sex ratios, and body condition of territorial Bicknell's Thrushes (Catharus bicknelli) wintering in the Dominican Republic between 1999 and 2008 at two ecologically distinct wet-forest sites, one in high-elevation cloud forest and the other in mid-elevation rainforest. Arthropod abundance was greater in cloud forest habitat, which was occupied by proportionally more males, the larger-bodied sex (74% male). By contrast, both sexes occurred at parity in rainforest habitat (53% male), where soft-bodied fruit was the predominant dietary resource. Body condition of cloud forest males was comparable to that of rainforest males, but cloud forest females were in poorer body condition than rainforest females. Females at the cloud forest site may face a greater likelihood of agonistic interactions with larger-bodied males and the thermoregulatory demands of roosting in colder night temperatures (0–12°C). We suggest that there are sex-specific advantages of wintering in these two habitats and that both are critical to supporting the full demographic structure of Bicknell's Thrush populations. Rainforest habitats, which are highly vulnerable to agricultural development in the Dominican Republic, may be particularly important to female survival during the winter period.

Resumen

El comportamiento social territorial durante el invierno de las aves canoras que migran entre el Neártico y el Neotrópico pone a los machos y a las hembras en conflicto directo por el acceso al espacio de invernada y a los recursos. Los resultados de esta competencia intersexual pueden variar entre especies y hábitats, pero sólo existe información al respecto para un pequeño subconjunto de especies migratorias. Entre 1999 y 2008 investigamos la disponibilidad de recursos, la proporción de sexos y la condición corporal de individuos territoriales de la especie Catharus bicknelli que pasaban el invierno en República Dominicana en dos lugares de bosque húmedo ecológicamente diferentes, uno en un bosque de niebla de gran elevación y otro en un bosque lluvioso de elevación media. La abundancia de artrópodos fue mayor en el bosque de niebla, área que fue ocupada proporcionalmente por mayor cantidad de machos (74%), que son el sexo de mayor tamaño. En contraste, ambos sexos se encontraban en igual proporción en el bosque lluvioso (54% machos), donde el recurso alimenticio predominante son las frutas blandas. La condición corporal de los machos del bosque de niebla era comparable a la de los machos del bosque lluvioso, pero las hembras del bosque de niebla tuvieron una condición corporal más pobre que la de las hembras del bosque lluvioso. Las hembras del bosque de niebla podrían enfrentar una mayor probabilidad de interacciones antagónicas con machos grandes y las demandas termorregulatorias de dormir a temperaturas frías en la noche (0–12⁰C). Sugerimos que existen ventajas específicas para cada sexo de invernar en estos dos hábitats, y que ambos son críticos para sostener la estructura demográfica de las poblaciones de C. bicknelli. Los hábitats de bosque lluvioso, que son áltamente vulnerables al desarrollo agrícola en la República Dominicana, pueden ser particularmente importantes para la supervivencia de las hembras durante el periodo de invierno.

THE UNEVEN DISTRIBUTION of food resources in wintering areas of Nearctic—Neotropic migrant songbirds is a critical proximate mechanism that drives habitat occupancy patterns and social systems during the winter period (Johnson and Sherry 2001, Sherry et al. 2005). Many migrant species are known to have winter territorial social systems (for review, see Greenberg and Salewski 2005) wherein males and females compete directly for exclusive access to food and shelter (Fretwell 1972). Understanding the influence of this intersexual competition on the distribution and fitness of individuals can help elucidate population-limiting factors in overwintering songbirds (Sherry and Holmes 1996, Holmes 2007) and identify how ongoing winter habitat loss may affect population dynamics (Stutchbury 1994, Wunderle 1995, Stutchbury et al. 2005).

A potential outcome of intersexual competition is segregation of the sexes into different habitats (Lynch et al. 1985, Marra and Holmes 2001, Latta and Faaborg 2002, Smith et al. 2010). An extensive series of studies, focused primarily on American Redstarts (Setophaga ruticilla), showed that individuals can segregate into sex-specific habitats as a result of body-size-mediated behavioral dominance (Marra 2000, Studds and Marra 2005, Holmes 2007). Larger-bodied birds (mostly males) secure sites with the highest-quality and most consistent food abundance, leaving subdominants (mostly females and juveniles) to occupy sites with reduced and seasonally declining food resources (Marra and Holmes 2001). Alternatively, sexual habitat segregation can result from competition avoidance, wherein the sexes evolve adaptations to distinct, minimally overlapping habitats (Morton et al. 1987). Female Hooded Warblers (S. citrina), for instance, preferentially occupy secondary forest, even when territorial vacancies are available in male-dominated primary forest (Morton et al. 1987, 1993; Stutchbury 1994). In a laboratory study, female Hooded Warblers chose artificial habitat that was structurally similar to secondary forest (Morton 1990), which suggests innate sex-specific habitat preferences that may have evolved to avoid energetically costly competition between the sexes. Finally, sexual habitat segregation may also be influenced by body-size-based differences in thermoregulatory ability, whereby the sex with the smaller ratio of surface area to volume is better able to withstand colder habitats and fasting periods at temperatures below metabolic thresholds (Ketterson and Nolan 1976).

We assessed the available food resources, sex ratios, and body condition of wintering Bicknell's Thrushes from 1999 to 2008 at two ecologically distinct wet broadleaf forest sites in the Dominican Republic, one in high-elevation cloud forest and the other in mid-elevation rainforest. The cloud forest site was characterized by low winter temperatures occasionally reaching 0°C, precipitation that arrived predominantly as cloud and fog water, and a dense forest understory. By contrast, the rainforest site experienced mild temperatures typical of the low-altitude tropics, had high levels of rainfall, and was composed of a relatively open forest understory (Townsend et al. 2011). Previous work at these sites showed a highly male-skewed sex ratio at the cloud forest site, where birds consumed an arthropod-heavy diet, and a female-skewed sex ratio at the rainforest site, where birds consumed mostly fruit (Townsend et al. 2010, 2011). At both sites, birds maintained exclusive territories, and no differences in territory sizes were detected among sites or between the sexes or age classes (Townsend et al. 2010).

Bicknell's Thrush is a wet-broadleaf-forest specialist throughout its winter range, and ≥90% of the total global population is estimated to overwinter on Hispaniola (Rimmer et al. 2001). Broadleaf forests in the Dominican Republic are highly vulnerable to development for agriculture, with ongoing deforestation both within and outside protected areas (Brothers 1997, Geisler et al. 1997, Latta 2005, Kerchner et al. 2010). Bicknell's Thrush is a species of high conservation concern (BirdLife International 2000, Lebbin et al. 2010), and although the extent to which winter habitat loss contributes to population declines is currently unquantified, it is believed to be significant (Rimmer et al. 2001).

In the present study, our objective was to determine whether sex-specific occupancy patterns and body condition differ between rainforest and cloud forest in order to inform more effective local conservation planning. Specifically, our goals were to (1) more fully document the variation in sex ratio between these habitats, (2) quantify differences in the dietary resources available in each habitat, and (3) compare body condition of birds in each habitat. Taken in concert, these metrics can be used to evaluate the evidence for sexual habitat segregation among wintering Bicknell's Thrush and to distinguish habitat-specific differences in factors that may limit populations of this globally vulnerable species.

Methods

Study sites.—Pueblo Viejo, a cloud forest site (hereafter “cloud forest”), and Loma la Canela, a rainforest site (hereafter “rainforest”), served as long-term focal study sites (see Townsend et al. 2010). The two sites are located in the southwestern and north-central areas of the country, respectively, and are ∼185 km apart.

The cloud forest site, where studies were conducted in 1999, 2000, and 2002–2008, is composed of primary, undisturbed montane broadleaf forest at 1,600–1,800 m elevation in the Sierra de Bahoruco National Park (18°12′N, –71°32′W). The site is characterized by a dense understory exceeding 11,000 stems ha^-1, composed largely of vine tangles from storm-related blow-downs (Townsend et al. 2011), complete canopy cover with trees reaching heights between 15 and 20 m, and an abundance of lianas and epiphytes (Veloz 2007). Annual rainfall is ∼1,700 mm year^-1, and night temperatures in winter can fall below 0°C, with mean winter daytime temperatures of 15–18°C (Bolay 1997).

The rainforest site, where studies were conducted in 2000, 2004, 2005, 2007, and 2008, is situated in the Cordillera Septentrional at elevations of 350–600 m and is part of the Loma Quita Espuela Scientific Reserve (19°25′N, –70°8′W). This site has undergone periodic selective logging and small-scale shifting agriculture during the past 100 years and is a mosaic of early- and late-successional broadleaf forest with small patches (about 10– 20 ha) of undisturbed primary forest at the highest elevations (Sanchez and Hager 1997). The rainforest site has a relatively more open understory than the cloud forest site (<2,000 elements ha^-1) and fewer dense vine tangles (Townsend et al. 2011). Annual rainfall is ∼2,400 mm year^-1, and the mean annual temperature is 25°C, with little seasonal variation (Sanchez and Hager 1997).

Field methods.—We captured Bicknell's Thrushes in both habitats, using passive arrays of 6- and 12-m mist nets with 36-mm mesh, and by target netting using conspecific playback of vocalizations. Birds were aged as second-year (SY) or after-second-year (ASY) by the shape of their rectrices (Collier and Wallace 1989), and blood samples were analyzed by polymerase chain reaction (PCR) to determine sex (Griffiths et al. 1998, Townsend et al. 2010). Sex and age ratios of birds captured by passive netting did not vary from those captured by target netting. All captured birds were weighed to the nearest 0.1 g and scored for fat (0 = no visible fat in furcular cavity, 0.5 = trace on sides of furcular cavity, 1 = sheet lining furcular cavity, 2 = filling cavity, 3 = mounding above top of cavity). We also measured unflattened wing chord and tail to the nearest 0.5 mm, and tarsus, exposed culmen length, length of bill from tip to nares, and bill width and depth to the nearest 0.01 mm.

Blood was collected from each bird in heparinized capillary tubes via brachial venipuncture using sterile 27-gauge hypodermic needles. Blood was preserved for sex determination in Queen's lysis buffer (Seutin et al. 1991) and used for DNA extraction to determine sex.

Resource availability.—We used pitfall traps, leaf-litter collectors, and sweep nets to quantify arthropod abundance, and straight-line transects to quantify fruit availability.

Pitfall traps consisted of 24 plastic cups (∼0.7 L with a 12-cm-diameter opening) buried at 50-m intervals along mist-netting transects with a small volume of water in the bottom of each trap. Traps were buried so that the lip of the cup was below ground level and detritus was arranged around the lip so that arthropods traveling along or below the leaf litter fell into the cups and were captured (Thomas and Sleeper 1977). During each trapping session, we deployed traps for 48 h, then combined the contents of the 24 individual traps and preserved them in 95% ethanol. We used a dissecting microscope to identify individual arthropods to taxonomic order and measure length. We pitfall trapped in both habitats in January and March of 2007 and 2008, providing an index of both mid- and late-winter arthropod abundance. Leaf-litter collectors consisted of a 1-L coffee can (12-cm-diameter opening, 18-cm depth) pushed through the leaf litter parallel to the ground until the can was full. We collected a swath of litter (all leaf litter from surface to soil) approximately 10 × 50 cm at each sampling site (weight of litter collected: mean ± SD = 130 ± 20 g). In cloud forest, we collected litter within the known home ranges of 13 individuals established by radiotelemetry in 2005 (Townsend et al. 2010). On each home range, we randomly chose 5 known telemetry locations as sampling sites. We also collected cloud forest litter and rainforest litter in late winter 2007 along mist-net transects. We sorted arthropods to taxonomic order on a white sheet and estimated the lengths of all arthropods to the nearest 1 mm. One worker (J.M.T.) performed all leaf-litter sampling. Sweep netting consisted of a standard muslin net 38 cm in diameter with a 61-cm handle passed over vegetation at heights of 0.5 to 1.5 m (Janzen 1973). Sweep netting took place at the same sites as leaf-litter collection in both habitats. Each sample consisted of 50 continuous sweeps. We immediately sorted the contents of sweep nets to taxonomic order and estimated the lengths of each arthropod. One worker (J.M.T.) performed all sweeps.

To determine total arthropod biomass, we calculated length— mass regressions specific to each of the orders of arthropods present in pitfall, litter, and sweep samples (Table 1). These were derived from reference samples that we measured to the nearest 0.01 mm, lyophilized in a freeze-dryer for 24 h, and weighed to the nearest 0.01 mg. To determine mass, we used the following equation: mass = a(length)b, where a and b are regression constants that were log transformed and then expressed in linear form for regression (Rogers et al. 1977, Johnson and Strong 2000). The final equation for estimating mass from length was mass in mg = eA (length in mm)b, where A = ln(a) and corresponds to the yintercept of the regression equation and b = the slope. All regression coefficients were P < 0.05.

To assess the late-winter availability of soft-bodied fruit, we established 15-m circular plots around 12 net sites set 50 m apart along the same transects used for mist netting in each habitat. In March 2007 and 2008, we counted the number of fruiting trees from the understory to the midcanopy (∼5 to 10 m height) within each 15-m plot and recorded the height of fruit-bearing trees. Canopy trees were not included because our telemetry observations indicated that Bicknell's Thrushes did not forage in the canopy (Townsend et al. 2010).

Table 1.

Regression statistics for mass (mg) by length (mm) of leaf-litter arthropods from wet-forest sites in the Dominican Republic. Equation: mass in mg = eA(length in mm)b.

Laboratory analyses.—We used Perfect gDNA Blood Mini kits (Eppendorf, Hauppauge, New York), following the manufacturer's protocol, to extract DNA from whole blood stored in lysis buffer. Homologous sections of sex-chromosome-based chromohelicase-DNA-binding (CHD) genes were then amplified by PCR. When viewed on a gel, PCR product from a male shows a single CHD-Z band, whereas PCR product from a female shows a second CHD-W band (Griffiths et al. 1998).

Statistical analyses.—For individuals captured in cloud forest (n = 166) and rainforest (n = 77), we examined both the proportion male and the proportion adult as response variables in separate generalized linear models (GLM), weighted by sample size of each habitat—year, with habitat as the predictor, specifying binomial errors and logit-link function. Parameter estimates from these models (β ± SE) are presented in logit scale. We analyzed arthropod biomass in pitfall traps using a two-way analysis of variance (ANOVA) with total biomass as the dependent variable and habitat (rainforest—cloud forest), season (middle—late), and the interaction of season with habitat as the explanatory variables. We further compared the abundance of each order of arthropods at the two sites using one-way ANOVA with biomass of individual orders as the dependent variable and habitat as the explanatory variable. We also used one-way ANOVA to assess differences between the two habitats in arthropod biomass estimated by litter collectors and sweep nets, and fruit abundance. To assess body condition, we created a standardized index of body size for 225 individuals, using the first principal component (PC1, which explained 82.1% of variation; hereafter “body size”) from a principal component analysis (PCA) of wing, tail, and tarsus measurements. We assessed variation in body size between habitats using ANOVA with body size as the dependent variable and sex, age, site, and sex * habitat interaction as the explanatory variables, with simple effects assessed post hoc by Tukey's HSD. We employed analysis of covariance (ANCOVA) to assess diurnal and sex-specific changes in body mass adjusted for body size in order to compare the physical condition of individuals. Body mass was the dependent variable, and body size (PC1) and time of day (hour) were added to the model as continuous covariates, along with the nominal explanatory variables year, habitat (cloud forest—rainforest), sex (male—female), age (SY—ASY), and season (middle—late) and the interactions sex * age, habitat * sex, and habitat * season, with simple effects assessed post hoc by Tukey's HSD. Data were analyzed using JMP, version 9.0 (SAS Institute, Cary, North Carolina). Results are presented as least squared means ± SE.

Results

Sex and age ratio.—The sex ratio in cloud forest varied significantly from 50:50 (χ² = 40.8, df = 8, P < 0.001) and was consistently malebiased across years, with a mean proportion male of 73.9 ± 3.8% (Table 2). By contrast, the sex ratio in rainforest fluctuated interannually (Table 2) and was not significantly different from 50:50 (χ² = 8.85, df = 4, P = 0.07), with a mean proportion male of 53.4 ± 9.4%. In a GLM with proportion male as the dependent variable and site as the predictor, cloud forest was significantly male biased compared with rainforest (β = -0.36 ± 0.14, χ² = 6.4, df = 1, P = 0.01). In a separate model comparing age ratios at the two sites, with proportion adult birds as the dependent variable and site as the predictor, the two sites did not differ (β = 0.06 ± 0.14, χ² = 0.25, df = 1, P = 0.62; mean proportion adult birds was 62.3 ± 0.03% in rainforest and 66.3 ± 0.07% in cloud forest). However, the sex ratio of adults differed between sites; there were proportionally more males in the adult population in cloud forest than in rainforest (β = 0.48 ± 0.19, χ² = 5.3, df = 1, P = 0.01; Table 2). The proportion male among SY birds was greater in cloud forest than in rainforest (Table 2) but the difference was not significant (β = 0.26 ± 0.23, χ² = 1.4, df = 1, P = 0.24). Similarly, the proportion of adult males in the overall sample was greater in cloud forest than in rainforest (Table 2), but not significantly (β = 0.22 ± 0.14, χ² = 2.45, df = 1, P = 0.118).

Resource availability.—The total available arthropod biomass estimated by pitfall trapping was greater in cloud forest than in rainforest (F = 45.96, df = 1 and 4, P = 0.003; Fig. 1). Seasonal change in available arthropod biomass was not significant (F = 1.43, df = 1 and 4, P = 0.298), nor was the interaction of habitat with season (F = 0.82, df = 1 and 4, P = 0.418). However, arthropod biomass declined from mid-to late winter in rainforest (2007: 43.6% decline; 2008:18.2% decline), whereas biomass in cloud forest declined by 19% in 2007 and increased by 18.1% in 2008. The composition of available arthropod biomass varied by site, with significantly greater Aranae (F = 13.71, df = 1 and 7, P = 0.01) and Archaeognatha (F = 9.04, df = 1 and 7, P = 0.024) biomass in cloud forest, significantly greater Orthoptera biomass in rainforest (F = 12.36, df = 1 and 7, P = 0.013), and marginally greater Formicidae biomass in rainforest (F = 4.36, df = 1 and 7, P = 0.08).

Litter collectors and sweep-net samples were combined in March 2007 to compare late-winter cloud forest and rainforest arthropod abundance. As with estimates from pitfall traps, available arthropod biomass was significantly greater in cloud forest than in rainforest (F = 9.50, df = 1 and 21, P = 0.006; Fig. 2A). Conversely, there were significantly more fruiting trees in rainforest (mean 2.41 ± 0.66 trees plot^-1) than in cloud forest (mean 0.23 ± 0.15 trees plot^-1; F = 11.26, df = 1 and 43, P = 0.002; Fig. 2B).

We used leaf-litter and sweep-net samples collected at the cloud forest site in 2005 to compare midwinter arthropod abundance on the known territories of individual male (n = 7) and female (n = 6) Bicknell's Thrushes. Arthropod abundance did not differ between male and female territories for either litter (F = 0.08, df = 1 and 12, P = 0.929) or sweep (F = 0.23, df = 1 and 12, P = 0.634) or for combined samples (F = 0.006, df = 1 and 12, P = 0.940). Arthropod abundance on these known territories was unrelated to body size of the territory holder (t = -0.87, df = 1 and 12, P = 0.403) or size of the home range (t = -0.2, df = land 12, P = 0.848).

Table 2.

Proportion of male Bicknell's Thrushes captured in each age class (SY = second-year, ASY = after-second-year), proportion of males in the overall sample, and proportion of adult males in the overall sample for individuals wintering at two sites in the Dominican Republic (n = number of individuals captured; asterisk denotes significant difference between sites).

Fig. 1.

Available arthropod biomass as indexed by mean mass ± SE of arthropods (mg) captured in 25 pitfall traps arranged along transects at a high-elevation cloud-forest site and a mid-elevation rainforest site in the Dominican Republic. Midwinter = January; late winter = March.

Fig. 2.

Mean late-winter (March) biomass of arthropods per combined leaf-litter and sweep-net sampling (A) and mean late-winter number of fruiting trees per 15-m² sampling area (B) at a rainforest and a cloud forest site in the Dominican Republic. Bars display standard error.

Body size.—Body size of Bicknell's Thrushes (n = 225) varied by sex (F = 131.18, df = 1 and 219, P< 0.001) and age class (F = 65.14, df = 1 and 219, P < 0.001). Post hoc comparison using Tukey's HSD showed that ASY males were significantly larger-bodied than any other sex— age grouping, SY females were significantly smaller-bodied than all other groupings, and ASY females formed a group with SY males that was of intermediate size (Fig. 3). We compared body size of each of these sex—age classes by site and found no site-based differences.

Body condition.— Comparison of body mass (ANCOVA; n = 222) between sexes and the two focal sites revealed significant effects of body size and time of day, and significant effects of the habitat * sex interaction (Table 3). Tukey's HSD pairwise comparisons of the simple effects in the habitat * sex interaction (cloud forest males, cloud forest females, rainforest males, rainforest females) revealed that cloud forest females were lighter than all other categorical groupings. Adjusted body mass of males did not differ between the habitats, nor did body mass of rainforest males differ from that of rainforest females (Fig. 4).

To further clarify site-specific diurnal changes in body mass, we analyzed each site separately using simple linear regression of mass adjusted for body size (PC1) with time of day. The body mass of cloud forest birds increased with time of day (F = 19.76, df = 1 and 141, P < 0.0001, r = 0.12), whereas birds in rainforest showed no trend in diurnal mass patterns (F = 1.56, df = 1 and 64, P = 0.22). The diurnal rate of mass increase in cloud forest was 0.09 ± 0.02 gh^-1.

Discussion

We identified clear site-based differences in sex ratios of wintering Bicknell's Thrushes. At the cloud forest site, males predominated in an environment with relatively colder night temperatures, high available arthropod biomass, and dense forest understory structure. Cloud forest males showed significantly greater adjusted body mass than cloud forest females. By contrast, at the rainforest site, the sexes occurred at parity in an environment with warmer temperatures, relatively low arthropod abundance, but greater availability of soft-bodied fruit and a more open understory. Rainforest males and females did not differ in their adjusted body mass and were comparable to cloud forest males, and both sexes in rainforest had significantly higher body mass than cloud forest females.

Fig. 3.

Body size of Bicknell's Thrush (n = 225) sex and age classes (SY = second-year, ASY = after-second-year) as indexed by principal component analysis. Box plots show the upper and lower quartiles. Whiskers show the sample minimum and maximum. The median is represented by a horizontal line spanning the width of the box. The mean is represented by a partial horizontal line in bold.

Variation in winter body mass can be the result of multiple factors, including the probability of food shortages, extreme variation in weather, risk of predation, and position within social hierarchies (Lima 1986, Gosler 1996, Marra 2000). Of these factors, the risk of predation and food shortages are unlikely to have a strong influence on Bicknell's Thrush body mass at the two focal sites in the present study. Diurnal predators occur at low densities at both cloud forest and rainforest (Townsend et al. 2009), and the high moisture levels of both habitats create conditions less vulnerable to the food shortages documented for xeric habitats during the late-winter dry season (Brown and Sherry 2006, Smith et al. 2010). The two habitats are, however, strongly differentiated by sex ratio, which may be a component of winter social hierarchy (Marra 2000, Latta and Faaborg 2002), and by temperature. These factors may have habitat- and sex-specific effects on body mass, and we suggest several possible mechanisms to explain the exceptionally lighter body mass of cloud forest females.

Table 3.

Results from ANCOVA of factors predicting body-mass variation in Bicknell's Thrushes overwintering at two sites in the Dominican Republic.

It is possible that the level of intersexual competition differs between cloud forest and rainforest and that this difference drives the contrasting patterns of habitat- and sex-specific body condition. Following the predictions of size-mediated behavioral dominance, females that maintained territories in cloud forest habitat, where males made up 74% of the population, would be expected to face high levels of agonistic interactions with their large-bodied male neighbors (Fretwell 1972, Marra and Holmes 2001). Although our data do not quantify the aggressive outcomes of male—female encounters, we have previously demonstrated that territoriality is prevalent at this site, with boundaries maintained by daily agonistic encounters between neighbors (Townsend et al. 2010). Females in cloud forest are numerically outnumbered by males and may face a greater probability of engaging in agonistic encounters with males. Such encounters might have an energetic cost that reduces the ability of cloud forest females to maintain winter body condition at levels comparable to those of males.

Smaller-bodied females may also suffer from the greater energetic requirements of nocturnal thermoregulation in the colder cloud forest climate. We documented a clear increasing body-mass trajectory with time of day in cloud forest, which suggests that cloud forest individuals experienced nocturnal mass loss due to winter night temperatures in the range of 0–12°C, well below the lower critical temperature at which Bicknell's Thrushes respond metabolically to temperature stress (17.5°C; Holmes and Sawyer 1975). Larger-bodied males, with a smaller surface-to-volume ratio, might retain a physiological advantage over smaller females in their ability to thermoregulate during the nocturnal fasting period in cloud forest habitat, as has been shown for temperate migrant songbirds wintering in cold climates (Ketterson and Nolan 1976). By contrast, there was no relationship between time of day and body-size-adjusted mass at the rainforest site, which suggests that the mass of birds declines less rapidly during the roosting period at this site, where night temperatures range from 15 to 20°C, near or above the lower critical temperature.

Fig. 4.

Bicknell's Thrush mass corrected for body size (least squared mean ± SE) as an estimate of physical condition for birds wintering at a high-elevation cloud forest site and a mid-elevation rainforest site in the Dominican Republic.

We initially postulated that cloud forest females should be larger than rainforest females, imparting competitive and thermoregulatory advantages that would allow them to thrive in cloud forest habitat (Stutchbury 1994, Stutchbury et al. 2005). However, cloud forest females were not larger-bodied than rainforest females and, therefore, gained no body-size-related thermoregulatory or agonistic competitive advantage. For cloud-forest-dwelling females, the combination of thermoregulatory costs and a competitive size handicap could work together to limit their potential overwinter fitness. By contrast, the better body condition of females that wintered at the rainforest site suggests that mid-elevation rainforest may be a “preferred” habitat for females (sensu Morton et al. 1987)—an area where they can compete on equal footing with males and survive the winter period in good physical condition. Our data suggest that the bulk of overwintering females in the Dominican Republic reside in warmer habitats where they survive on a fruit-heavy diet, a pattern similar to that of Black-throated Blue Warblers (S. caerulescens) in Puerto Rico, where females predominantly occupy secondary forest and rely on fruit and nectar resources (Wunderle 1992, 1995). Female Bicknell's Thrushes either actively avoid male-dominated cloud forest habitat or compete less successfully for cloud forest territories, thereby restricting most overwintering females to midelevation rainforest sites.

The overall size of the winter female population may, therefore, be limited by the availability of rainforest sites. In the Dominican Republic, there are ∼110,500 ha of available cloud forest habitat and 315,200 ha of available rainforest habitat (Tolentino and Peña 1998). Using a conservative density estimate of 1 bird/5 ha based on average territory sizes (Townsend et al. 2010), with a proportion male of 70% in cloud forest and 50% in rainforest, we estimate that there are potentially 31,520 females wintering in rainforest habitat and 6,630 in cloud forest habitat, an approximately fivefold difference. We suggest that lower-elevation rainforests are critically important to female survival on Hispaniola.

We stress, however, that our work on Hispaniola has not identified a sharply defined habitat-quality gradient for Bicknell's Thrush, as have several studies of paruline warblers on their Neotropical winter grounds (Latta and Faaborg 2001, 2002; Marra and Holmes 2001; Smith et al. 2010). The four species of wood-warbler for which fitness consequences correlate closely with variation in habitat quality all occupy mostly xeric habitats arrayed along a clear moisture gradient. This moisture differential is a key ecological determinant of winter habitat quality for these species, ranging from high-food-abundance wet sites to lower-food-abundance dry sites (Smith et al. 2010). It is possible that, by contrast, wet-forest specialists such as Bicknell's Thrush are less likely to be structured by extreme dominance relationships and starkly contrasting gradients of habitat quality. Further, both arthropods and fruit appear to serve as high-quality food resources for wintering Bicknell's Thrushes. We recommend further studies of other wet-forest specialists with omnivorous flexible winter diets, such as the other migrant thrushes in the genus Catharus and the Wood Thrush (Hylocichla mustelina), to better understand the differences in limiting factors between species occupying predominantly mesic versus xeric tropical habitats.

In the present study, we conclude that cloud forest is a malepreferred habitat where outcomes for cloud forest females are measurably poorer. By contrast, mid-elevation rainforest supports greater numbers of female thrushes with body condition comparable to that of both cloud forest and rainforest males. We stress the importance of both forest types to the overall population dynamics of wintering Bicknell's Thrushes. Mid- and low-elevation rainforests are highly threatened habitats on Hispaniola (Kerchner et al. 2010), and they have a clear value to the maintenance of female populations of this species. We recommend strong conservation measures to prevent further loss of rainforest habitat in the Dominican Republic and renewed efforts to fortify the preservation of high-elevation cloud forest habitat. We especially recommend a thorough quantification of the spatial extent of remaining wet forest on Hispaniola to identify those tracts most vulnerable to development. A spatially explicit wet-forest inventory, combined with the understanding of Bicknell's Thrush spatial organization and demographic structure presented in the present study and elsewhere (Rimmer et al. 2001; Townsend et al. 2009, 2010, 2011), would provide critical information to Hispaniolan land managers charged with making conservation decisions.

Acknowledgments

We gratefully acknowledge funding support from the Association of Field Ornithologists, the Carolyn Foundation, the Eastern Bird Banding Association, the John D. and Catherine T. MacArthur Foundation, The Nature Conservancy, the Stewart Foundation, the Thomas Marshall Foundation, the U.S. Fish and Wildlife Service, the U.S. Forest Service International Program, the Wilson Ornithological Society, and friends of the Vermont Center for Ecostudies and the Vermont Institute of Natural Science. Permission to band and place transmitters on birds was provided by the U.S. Geological Survey. Permission to conduct research in the Dominican Republic was provided by the Subsecretaria de Areas Protegidas y Biodiversidad. Laboratory facilities and assistance were generously provided by I. Lovette. We are especially thankful to J. Almonthe, J. Brocca, E. Cuevas, P. Diaz, S. Frey, E. Garrido, J. Hart, P. Johnson, J. Klavins, V. Mejia, R. Ortiz, and A. Townsend for their outstanding field work under difficult conditions.

Author notes

Associate Editor: C. M. Rogers