SEARCH

SEARCH BY CITATION

Keywords:

  • apparent survival;
  • Dail–Madsen model;
  • golden-winged warbler;
  • habitat quality;
  • non-breeding season;
  • population dynamics;
  • recruitment;
  • Vermivora chrysoptera;
  • winter ecology

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

1. Identifying high-quality habitat is integral to effective species conservation efforts and requires information about habitat-specific abundance and demographics. This information is lacking for many species of conservation concern because of the inherent difficulties associated with implementing mark–recapture studies at large spatial scales.

2. The golden-winged warbler Vermivora chrysoptera is a Neotropical–Nearctic migratory bird experiencing a long-term population decline; yet no information about non-breeding season habitat quality or habitat selection exists to inform conservation efforts. We used a novel hierarchical model that requires only spatially and temporally replicated count data to estimate habitat-specific abundance, apparent survival, recruitment and detection probability of golden-winged warblers wintering in Costa Rica. We quantified habitat selection at the home-range level using radiotelemetry.

3. Golden-winged warblers were absent from tropical dry forest and were most abundant in premontane evergreen forest. Within their home ranges, golden-winged warblers selected microhabitat features associated with intermediate disturbance that reflected their preference for foraging in hanging dead leaves.

4. Consistent with other evidence of a declining population, local population size decreased over the duration of the study. The rate of decrease was higher during the non-breeding season than among seasons. We found no differences in apparent survival or recruitment among habitat types; however, our estimates of these parameters were imprecise.

5.Synthesis and applications. Golden-winged warblers are forest-dependent species during the non-breeding season and have specialized microhabitat requirements that make them vulnerable to ongoing tropical deforestation. However, advanced secondary forests can provide the requisite microhabitat features, and because we found no evidence of reduced survival in this habitat type, regenerating forest on degraded lands may be an effective component of a conservation strategy for this species. Our study also demonstrates that information on population dynamics and habitat quality can be obtained using repeated counts instead of mark–recapture methods.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Identifying high-quality habitat is a top priority in many conservation efforts. Habitat quality is defined by its occupants’ average survival and reproductive rates (Newton 1998), parameters that are typically estimated using mark–recapture methods. However, the costs associated with implementing mark–recapture studies at large spatial scales are often prohibitive, and capturing and handling animals can reduce survival. For these reasons, habitat quality information is lacking for most species, even declining members of well-studied groups such as Neotropical–Nearctic migratory birds (Nolan 1978; King, Rappole & Buonaccorsi 2006; Sauer, Hines & Fallon 2008). Habitat quality data for Neotropical–Nearctic migratory birds are particularly scarce during the non-breeding season. This is concerning because non-breeding season events can limit population growth (Rappole, Ramos & Winker 1989; Sherry & Holmes 1996; Rappole, King & Diez 2003; Studds & Marra 2005; Calvert, Walde & Taylor 2009), and habitat loss is rampant in many parts of the Neotropics (Myers et al. 2000; Asner et al. 2009).

One of the most rapidly declining Neotropical–Nearctic migratory birds is the golden-winged warbler Vermivora chrysoptera (Sauer, Hines & Fallon 2008), a species that breeds in southern Manitoba and Ontario, the Great Lakes states, and in the central and southern Appalachian Mountains (Confer 1992). The stationary non-breeding range has not been adequately delimited, but it is believed to extend from southern Mexico to the northern Andes of Venezuela, Colombia, and Ecuador (DeGraaf & Rappole 1995). Although studies suggest that breeding season factors such as hybridization with blue-winged warblers V. pinus and habitat loss have contributed to this population decline (Gill 1997; Buehler et al. 2007), information is needed from the entire annual cycle to fully understand what factors contribute to the decline of golden-winged warblers (Rappole, King & Diez 2003).

No detailed studies of golden-winged warblers have been conducted during the non-breeding season; however, anecdotal observations and incidental reports from community-level studies indicate that golden-winged warblers may be specialized in their habitat use (Bent 1963; Tramer & Kemp 1980; Blake & Loiselle 2000), which would increase their potential susceptibility to destruction or alteration of non-breeding habitat. For example, this species appears to be restricted to lower- and middle-elevation tropical wet forests that have experienced high levels of deforestation over the past 50 years (Powell, Rappole & Sader 1992; Robbins, Fitzpatrick & Hamel 1992; Blake & Loiselle 2000). Evidence also exists that golden-winged warblers are specialized dead leaf foragers, which could further limit the extent of usable habitat as this habitat feature is patchily distributed (Morton 1980; Tramer & Kemp 1980; Gradwohl & Greenberg 1982). This anecdotal evidence is not sufficient to inform conservation efforts and leaves many questions unanswered regarding the specifics of habitat selection and habitat quality.

Estimating habitat-specific abundance and demographics of golden-winged warblers at large spatial scales does not appear to be feasible using standard mark–recapture methods such as those based on constant effort mist-netting because capture rates are very low (Saracco et al. 2008; Chandler 2010). An alternative method of estimating habitat-specific abundance and demographics was recently developed by Dail & Madsen (in press). They proposed a hierarchical model that requires only spatially and temporally replicated count data and thus does not require capturing and marking individuals. In this study, we demonstrate how this model can be applied to quantify non-breeding season habitat quality of golden-winged warblers. A second objective was to quantify habitat selection at multiple spatial scales to understand why certain habitats were utilized more than others.

Materials and methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Study Area

This study was conducted from 2006 to 2010 between 800 and 1600 m on both slopes of the continental divide in the Cordillera de Tilarán, Costa Rica N10°13′ W84°39′ (Fig. S1, Supporting Information). The study area lies within the watersheds of the Río Jamaical on the Caribbean slope and the Río Aranjuez on the Pacific slope and encompasses an area of approximately 100 km2. The Pacific slope is much drier than the Caribbean slope because the northeasterly trade winds lose most of their precipitation as they rise and cool over the Caribbean slope (Clark, Lawton & Butler 2000). This rain shadow has a profound effect on the vegetation (Haber 2000). The lower elevations of the Pacific slope are characterized by fragments of tropical dry forests, classified as premontane moist forests in the Holdridge life-zone scheme, and are embedded within a mosaic of small coffee farms and cattle pastures. This agricultural landscape lies adjacent to the >28 000 ha Monteverde Reserve Complex (MRC), which includes the Monteverde Cloud Forest Preserve, the Children’s Eternal Rainforest and the Alberto Manuel Brenes Biological Reserve. The forests above 1200 m on both sides of the continental divide are classified as upper montane wet forest, and those below 1000 m on the Caribbean slope are premontane rain forests. Mean annual temperature ranges from approximately 18 °C at high elevations (1500 m) to 24 °C at lower elevations (700 m).

Field Methods

Habitat-specific abundance and demographics

To assess habitat-specific abundance and demographics, we surveyed golden-winged warblers at 97 points visited three times each during the 2008–09 and 2009–10 non-breeding seasons using a 20-min point count methodology. To avoid the possibility of including transients, surveys were conducted between 1 January and 15 March each season. We could have begun these surveys earlier in the season because golden-winged warblers appear to establish territories by late October; however, a pilot study during October–December 2006 failed to obtain sufficient detections during those months (4 of 84 survey points) because of the extremely high rainfall typical of that period (Chandler 2010). The end date was determined based upon radiotelemetry and resight data, which indicated that migratory movements do not begin until late March. Each survey was divided into an initial 10-min passive period and a subsequent 10-min period during which golden-winged warbler songs and call notes were broadcast from handheld speakers. Vocalizations were acquired from the Cornell Lab of Ornithology’s Macauly Laboratory and were played at a volume of 100 dB at a distance of 1 m from the speakers. For each individual detected, we recorded the sex and the 10-min time intervals in which it was observed.

The 97 points were stratified among four habitat types: closed-canopy primary forest (= 25), naturally disturbed primary forest (= 25), secondary forest (= 23) and agroforestry systems (= 24) (Fig. 1a). Closed-canopy primary forest was defined as forest with no sign of human disturbance and >70% canopy closure. Naturally disturbed primary forest was characterized by ≤70% canopy closure and was located along rivers or in areas with large canopy gaps. Secondary forests were >15 years old, even-aged, naturally regenerating forests resulting from cattle pasture abandonment. Agroforestry systems were mostly shade coffee plantations but included some ‘silvopastures’, pastures with dense or scattered trees. These habitat types represent the major categories present in the study area, with the exception of cattle pastures that pilot data as well as published information indicated are not used by golden-winged warblers (Powell, Rappole & Sader 1992; Blake & Loiselle 2000). All survey points were located at least 500 m apart to ensure that no individuals were detected at more than one point.

image

Figure 1.  Maps of habitat type distribution (a, b) and golden-winged warbler (GWWA) locations (c, d) in the Cordillera de Tilarán, Costa Rica. Solid line is the continental divide.

Download figure to PowerPoint

To quantify habitat characteristics, we established 50-m-radius plots at each point and measured the following variables: elevation, canopy height, percentage canopy cover, slope, aspect and diameter at breast height (dbh) of trees selected using a 10-factor cruising prism. Each plot was partitioned into quarters and the following microhabitat variables, which behavioural observations suggested were used by golden-winged warblers (Chandler 2010), were measured within each: hanging dead leaf index (0, 1–100, 101–1000, >1000 leaves), vine tangle index (none, vines but no tangles >1 m diameter, vines and 1–2 tangles, vines and >2 tangles) and epiphyte index (no moss or bromeliads, moss <2 cm thick and few bromeliads, moss 2–5 cm thick with numerous bromeliads, moss >5 cm thick).

Home-range level habitat selection

The point count surveys provided information about where golden-winged warblers located their home ranges within the study area (second-order habitat selection; Johnson 1980). Organisms also select habitat features within their home ranges (third-order habitat selection), which we studied using radiotelemetry. We captured golden-winged warblers using broadcast vocalizations and a clay decoy placed between two nets. Each individual was fitted with a 0·43-g Holohil BD-2N transmitter using an elastic backpack harness (Rappole & Tipton 1991). The weight of these units was approximately 6% of total body weight, which was thus higher than the recommended level of 3%; however, the increased weight was necessary to provide sufficient battery life needed to obtain reliable behavioural observations used in another component of our research (Chandler 2010). The transmitters did not appear to affect behaviour, except for one individual, not included in the analysis, that continually pecked at its unit.

Tracking began 1 day after the transmitter was attached and continued until battery failure or mortality. We located each bird daily and followed it for 2 h, recording locations every 30 min using a handheld global positioning system (GPS) unit. Wherever possible, we located the bird visually to record key habitat variables. We recorded approximately five relocations per day, spaced evenly to minimize observer bias. At each relocation point, we recorded the habitat type as primary forest, secondary forest, naturally disturbed forest or agroforestry system. We also measured canopy height and diameter at breast height (dbh) of all trees selected by 10-factor cruising prism. For each bird location, we established a 2-m-radius plots and measured the dbh of all stems >2 cm, the number of dead leaves (0–10, 11–50, >50), the presence or absence of vine tangles >1 m in diameter, the number of bromeliads and the thickness of epiphytic moss. The 2-m plot variables were only recorded during the second and third seasons of the study, and of these, bromeliads and moss were only recorded in the final season. Upon battery failure of radiotransmitters or mortality, we used kernel density estimators to delimit the 95% home-range boundaries. Within these boundaries, we took habitat measurements on a 20-m grid. At each grid intersection, we measured the same habitat variables as were measured at points where marked birds had been located.

Statistical Methods

Habitat-specific abundance and demographics

To analyse the population dynamics of golden-winged warblers, we used the recently developed Dail–Madsen model (Dail & Madsen in press), which is a generalized form of the Royle (2004) binomial mixture model developed for open populations (i.e. those in which abundance can change over time). This model is very well suited to non-breeding season data because it assumes that abundance patterns are determined by an initial territory establishment process followed by gains and losses resulting from mortality, recruitment and movements. It also accounts for imperfect detection probability. The model requires both spatial and temporal replication and can be described as follows:

  • image
  • image
  • image
  • image
  • image

where Nit is the number of individuals at site i on survey occasion t, Git is the number of gains (recruits) between seasons, Sit is the number of survivors that do not permanently emigrate, and yit is the observed count at site i on survey occasion t. M is the number of plots and T is the number of survey occasions. The four structural model parameters are initial abundance (λ), recruitment rate (γ), apparent survival (ω) and detection probability (p), i.e. the probability of detecting Nit on a single survey. Because this is a model of data from unmarked individuals, it is not possible to distinguish between losses because of mortality and those because of permanent emigration; therefore, we use the phrase ‘apparent survival’, which can be defined as one minus the probability of losing an individual at a given plot. Population size within the surveyed area can be estimated at any point in time as inline image, and thus, the trend in abundance can be estimated by inline image.

The four parameters (λ, γ, ω and p) can be modelled as functions of covariates. We used a manual stepwise selection process based upon Akaike’s Information Criterion (AIC) to find the best combination of covariates that we had a priori reason to believe were influential. These included all variables listed previously as well as a precipitation proxy (distance from continental divide). We used distance from the continental divide as a proxy for precipitation because detailed precipitation data do not exist for our study area, whereas the relationship between distance from the continental divide and precipitation has been clearly established (Young, DeRosier & Powell 1998; Clark, Lawton & Butler 2000). We also considered three subcategories of diameter at breast height (dbh) size classes (<20, 20–50 and >50 cm). We evaluated quadratic terms for precipitation, average canopy height, elevation and epiphytes because field observations led us to believe that golden-winged warbler abundance might peak at intermediate levels of these variables. We modelled recruitment and apparent survival using these same predictor variables, and we included season in each model to estimate within- and among-season rates separately. In the detection probability component of the model, we considered wind, observer skill, precipitation, time of day, date, canopy height, basal area and an index of ambient noise produced by rivers. Wind and precipitation were measured on a 1–5 scale. Observer skill was defined as follows: 1 = limited point count experience, 2 = extensive point count experience on breeding grounds, 3 = some point count experience with golden-winged warblers during the non-breeding season and 4 = extensive experience surveying golden-winged warblers during the non-breeding season. River noise was scored on a scale of 0–5, where 0 represented no ambient noise from rivers and 5 was extremely loud ambient noise. Summary statistics for all predictor variables are presented in Table S1 (Supporting Information). Covariates that were included in models with ΔAIC values <2 were considered to be important predictor variables.

We estimated parameters of the Dail–Madsen model by maximizing the integrated likelihood function in the freely available R software (R Development Core Team 2010). We also added model-fitting capabilities to a custom version of the R package ‘unmarked’ (Fiske, Chandler & Royle 2010), which is included as Appendix S1 in Supporting Information. This implementation differs slightly from the original paper in that differences in time intervals between survey occasions were modelled in continuous time rather than discrete time steps because this made it easier to estimate monthly rates of apparent survival and recruitment as opposed to daily rates.

Home-range level habitat selection

We used a mixed-effects logistic regression model to analyse home-range level (third-order) habitat selection. When using logistic regression to analyse use vs. availability data, it is important to recognize that an unknown fraction of the availability data include points that were used. Thus, the logistic regression model is not predicting the probability of use; rather, it is the probability of use relative to availability (Keating & Cherry 2004). The null hypothesis is that individuals use habitat in proportion to availability. Treating variation among individuals as a random process made it possible to make inference at both the individual and population (individual average) levels. All models included a random effect for the intercept, and we considered random effects for each of the covariates as well. To identify the most parsimonious model, we used the same AIC-based model selection process as described previously. A summary of the distributions of these predictor variables is shown in Table S2 (Supporting Information). Analyses were conducted using R 2·11·0 (R Development Core Team 2010) and the package lme4 (Bates & Maechler 2010).

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Habitat-Specific Abundance and Demographics

We detected 59 golden-winged warblers during 546 surveys over two seasons (Fig. 1c). Only four of these detections (6·8%) occurred during the first 10 min of the survey, highlighting the importance of using broadcast vocalizations for surveying golden-winged warblers during the stationary non-breeding season. In 2009, we detected golden-wings at 25 of 94 points (26·6%). In 2010, we did not resurvey four points that were very difficult to access and we added three new points. Golden-winged warblers were not detected at any of the four points dropped in 2009. In 2010, we detected golden-wings at 17 of 93 points (18·3%). Ninety-five percent of detections were of single individuals; however, at one point, a female and a male were detected on different occasions in 2009. Two males were detected simultaneously at one point during two consecutive survey occasions in 2010. Only three females were detected, and thus, we were not able to model the sexes separately.

In addition to these 59 observations on point counts, we observed 89 other individuals (69 males and 20 females) incidentally while carrying out other duties associated with the project (Fig. 1d, Appendix S2). The spatial locations of point count detections were closely aligned with incidental observations of golden-winged warblers (Fig. 1c,d), except that several golden-winged warblers were incidentally detected along large rivers within the Manuel Brenes biological reserve but were never detected in this habitat during point counts.

Abundance adjusted for detection probability, but ignoring covariate effects, was <0·5 birds per points (Table 1). However, substantial variation in abundance existed among points as demonstrated by the inclusion of covariates in all supported models. The importance of four covariates was clear, although there was considerable uncertainty regarding the best combination of these variables. A quadratic effect of distance from the continental divide was present in all supported models (Table 2), indicating that abundance peaked at a distance of 1·45 km from the divide on the Pacific slope (Fig. 2a), which is an area within the premontane wet life-zone that receives approximately 2·5 m of annual rainfall. This amount corresponds to climatic conditions favouring intermediate levels of epiphytes (Fig. 1b). The second most supported effect was a quadratic relationship with canopy height, indicating that abundance peaked in forests with canopy heights of 21 m (Fig. 2b). Habitat type and hanging dead leaves were included in the list of supported models and indicated that golden-winged warbler abundance was highest in naturally disturbed primary forest and advanced secondary forest, and their abundance was positively associated with the number of hanging dead leaves.

Table 1.   Parameter estimates and 95% confidence intervals from the most general dynamic abundance model considered for golden-winged warblers surveyed at 97 point count locations in the Cordillera de Tilarán, Costa Rica, 2007–2009
ParameterEstimateSELowerUpper
  1. *Corresponds to 1 January 2009.

Initial abundance* (λ– individuals/plot)0·4980·1530·2720·911
Recruitment (γ– gains/month)0·0060·0080·0010·076
Within-season apparent survival (ωw– monthly rate)0·8700·1530·3220·990
Among-season apparent survival (ωa– monthly rate)0·9580·0350·8050·992
Detection probability (p– per survey)0·2740·0750·1520·442
Table 2.   Model selection results for golden-winged warbler abundance. λ is initial abundance, γ is recruitment, ω is apparent survival, and p is detection probability. Squared terms indicate quadratic effects. A dot signifies no covariate effect. Data are from 97 survey point locations in the Cordillera de Tilarán, Costa Rica, 2007–2009
λγωpΔAICR2
  1. AIC, Akaike’s Information Criterion.

Precip2 + CanHt2.SeasonWind + Obs0·000·30
Precip2 + CanHt2 + habitat.SeasonWind + Obs0·130·34
Precip2 + habitat.SeasonWind + Obs0·130·32
Precip2 + CanHt2.SeasonObs0·300·28
Precip2 + leaves.SeasonObs0·300·27
Precip2 + CanHt2.SeasonObs0·300·33
Precip2 + habitat.SeasonObs0·340·30
Precip2 + CanHt2 + habitat.Season.0·410·31
Precip2 + CanHt2.SeasonWind0·530·33
Precip2 + CanHt2.SeasonWind0·530·28
Precip2 + CanHt2.Season.0·590·27
Precip2 + leaves.Season.0·650·25
Precip2 + leaves.SeasonWind + Obs0·660·28
Precip2 + habitat.Season.0·690·28
Precip2.SeasonObs0·900·25
Precip2.SeasonWind1·290·24
Precip2.Season.1·430·23
Precip2 + CanHt2 + habitat + Leaves.SeasonWind1·650·33
Precip2 + CanHt2 + leaves.Season.1·650·27
Precip2 + CanHt2 leaves.SeasonWind + Obs1·650·30
Precip2.SeasonTime1·790·24
image

Figure 2.  Golden-winged warbler abundance in relation to distance from continental divide (top) and canopy height (bottom) in the Cordillera de Tilarán, Costa Rica, 2007–2009. Error band is ±1 SE.

Download figure to PowerPoint

Golden-winged warblers were detected at seven plots in 2010 where they had not been detected in 2009, and they were not detected at 15 plots in 2010 where they had occurred in 2009. After accounting for detection probability, the models indicated that the seven plots that appeared to be colonized were most likely used by golden-winged warblers in 2009, but those individuals were not detected. Thus, the recruitment rate was close to zero (Table 1). This near-zero recruitment rate made it unnecessary to account for seasonal differences as there was no variation to model. Apparent survival was not related to any of the habitat covariates we considered, but we included a season effect in all models to allow for differences in among- vs. within-season survival. The point estimate for within-season monthly apparent survival was 0·870, although the confidence interval was large (Table 1). This value contrasts with a relatively high among-season monthly apparent survival probability of 0·958. The low within-season apparent survival rate and low recruitment rate contributed to the overall population decline within our sample of point count plots (Fig. 3).

image

Figure 3.  Golden-winged warbler population size at the survey plots over time. Error bars are ±1 SE and were calculated using the parametric bootstrap. Data are from 97 point count plots in the Cordillera de Tilarán, Costa Rica, 2007–2009.

Download figure to PowerPoint

Detection probability was negatively related to wind and positively related to the observer skill index, indicating the importance of controlling them in the study design to the extent possible (e.g. training observers and rotating them among survey points), as well as accounting for them statistically.

Home-Range Level Habitat Selection

We radiotracked 24 golden-winged warblers during three non-breeding seasons, but only 11 individuals had sufficient relocations (>5) and home-range habitat data (>20 measurements) to model habitat selection. Only two of these individuals were females, and thus, we were not able to assess differences between the sexes. We were not able to include habitat type (primary forest, secondary forest or agroforestry system) in these models because very few home ranges included sufficient proportions of multiple habitat types to assess selection. Canopy height was strongly correlated with tree dbh (> 0·7), and thus, we only considered canopy height in these models.

Analyses of home-range level use vs. availability data indicated a consistently supported quadratic relationship between the probability of relative use by golden-winged warblers and canopy height, with a maximum probability at a canopy height of 12 m (Table 3, Fig. 4). Golden-winged warblers also preferred areas within their home ranges that had high basal area in 2-m plots and where vine tangles were present (Table 4). These microhabitat features were commonly found in large canopy gaps or advanced secondary forests.

Table 3.   Model selection results for use vs. availability logistic regression models for data from 11 golden-winged warblers radiotracked in the Cordillera de Tilarán, Costa Rica, 2007–2009. An intercept was included in all models as was a random effect term for variation among individuals
FixedRandomAIC
  1. AIC, Akaike’s Information Criterion.

CanopyHt2 + basal area + vinesCanopyHt939·70
CanopyHt2 + basal area + vinesIntercept only940·13
CanopyHt2 + basal areaCanopyHt940·19
CanopyHt2 + vinesCanopyHt940·90
CanopyHt2 + basal areaIntercept only941·42
CanopyHt2CanopyHt941·61
image

Figure 4.  Probability of use relative to availability for 11 radiotracked golden-winged warblers radiotracked in the Cordillera de Tilarán, Costa Rica 2007–2009. Thick black line is mean response among individuals.

Download figure to PowerPoint

Table 4.   Parameter estimates from the most supported logistic regression model of golden-winged warbler use vs. availability. Random effects are reported as standard deviations and can be interpreted as the among-individual variation in corresponding fixed effects. Data are from 11 golden-winged warblers radiotracked in the Cordillera de Tilarán, Costa Rica, 2007–2009
ParameterTypeEstimateSEZP
ß0Fixed−2·7410·307−8·914<0·001
Canopy heightFixed0·1780·0434·155<0·001
Canopy height2Fixed−0·0070·001−5·096<0·001
Basal area (2-m plot)Fixed0·0550·0301·8370·066
VinesFixed0·3410·2111·6150·106
ß0Random0·581   
Canopy heightRandom0·039   

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

One of the difficulties of conducting population-level studies of migratory birds during the non-breeding season is the logistical challenge of marking and recapturing sufficient individuals to make inferences at broad spatial scales. Thus, most research either has been restricted to small study areas or has ignored population dynamics and focused on indices of abundance or occurrence. We used a new class of hierarchical models (Dail & Madsen in press) developed for surveys of unmarked individuals to overcome these limitations, and our results represent the first application of this methodology for identifying high-quality habitat for a species of conservation concern.

Golden-winged warblers exhibited high degrees of specialization, in terms of both the habitats they selected to establish their home ranges (second-order habitat selection) and within home ranges (third-order habitat selection). Our analyses of point count data showed golden-winged warblers were most abundant close to the continental divide on the Pacific slope, which is dryer than the Caribbean slope and wetter than the lower Pacific (Young, DeRosier & Powell 1998). They were never detected in the dry forests far from the continental divide on the Pacific slope, which is consistent with previous qualitative findings (Bent 1963). Although we detected few golden-winged warblers far from the divide on the Caribbean slope during point count surveys, our incidental observations and records from other researchers demonstrate that they do occur in these lower-elevation wet forests (Powell, Rappole & Sader 1992). These findings along with the lack of support for elevation in the abundance models indicate that precipitation and not elevation drive second-order habitat selection. Elevation, however, may play a role outside of the range we studied. For example, golden-winged warblers are rarely reported above 2500 m. There also exist few records of this species in forests near sea level (Restall, Rodner & Lentino 2007; eBird 2010).

Within their range of preferred precipitation, golden-winged warblers were most abundant in forests characterized by intermediate disturbance, which included both natural disturbance in the form of wind-throws or river margins and anthropogenic disturbance such as the cutting of individual trees. Specifically, abundance was highest in forests with canopies 22 m tall and high levels of hanging dead leaves. The association with hanging dead leaves is probably due to their specialized foraging behaviour, which involves probing and prying open dead leaves to extract insects (Morton 1980; Tramer & Kemp 1980; Chandler 2010). Model selection results provided some evidence that golden-winged warblers were less abundant in undisturbed primary forest than in other habitat types, but this effect was relatively weak. These second-order habitat selection results suggest that golden-winged warblers are microhabitat specialists rather than habitat specialists, i.e. their preferred microhabitat conditions can be found in primary forest as well as in secondary forest and occasionally agroforestry systems, but the appropriate combination of these microhabitat variables is rare, which may be one reason why this species appears to be patchily distributed at low densities throughout its non-breeding range (Bent 1963).

Although we regularly encountered golden-winged warblers along the large rivers within the Manuel Brenes Biological Reserve, we never detected them at these locations during point count surveys. We believe this finding is because of a near-zero detection probability in that habitat type, which prevented us from statistically controlling this effect. Ambient noise caused by these rivers was extremely loud, and on several occasions, golden-winged warblers did not respond to broadcast vocalizations even when they were as close as 25 m. Typically, golden-winged warblers make loud call notes in response to the broadcast vocalizations and approach the observer aggressively (Chandler 2010). Many other species of Neotropical–Nearctic migratory birds were also seen in that habitat type, and future work should focus on assessing its conservation value.

We found a high degree of congruence between second- and third-order habitat selection patterns. At the home-range level, we found a similar quadratic relationship with canopy height as was evident in the second-order analyses, although the maximum probability of use peaked at a lower value for canopy height (12 m). We view this as additional evidence that birds select forest habitats affected by disturbance processes. Our analyses of second-order habitat selection indicated that golden-winged warblers selected home ranges with high numbers of hanging dead leaves and, within home ranges, golden-winged warblers preferentially used areas where vine tangles were present, probably because dead leaves were often prevalent in vine tangles. Thus, it appears as though they selected areas within their home ranges where foraging opportunities are greatest. This observation is consistent with previous research indicating that Neotropical–Nearctic species select habitat during the non-breeding season to optimize foraging opportunities (Rappole, King & Barrow 1999; Johnson & Sherry 2001). These microhabitat conditions were often found in large canopy gaps, along rivers, on steep slopes and in advanced secondary forests. Golden-winged warblers also occurred in agroforestry systems such as shade-grown coffee, but telemetry results indicate that individuals in shade coffee were in transit between adjacent patches of forest (R. B. Chandler unpublished data). It seems unlikely that shade coffee certification programmes could effectively mandate the retention of habitat features such as vine tangles and hanging dead leaves that would potentially make shade coffee suitable for golden-winged warblers. Therefore, forest protection and regeneration should be given higher priority than efforts to improve on-farm habitat conditions.

We did not have enough data on female golden-winged warblers to model sex-specific habitat selection, but we did resight 22 females incidentally while traversing the study area engaged in other activities. These observations, as well as the point count data and encounters with male and female golden-wing warblers at the same sites or even within the same flock on several instances, indicated that males and females occurred in similar habitats within the study area. Thus, we found no evidence of sexual habitat segregation, unlike what has been reported for other members of this family (Morton 1990; Marra & Holmes 2001). However, the apparent sex ratio bias in our study area suggests that the sexes may segregate at broader geographical scales.

Although golden-winged warblers appear to have specialized microhabitat requirements, their utilization of advanced secondary forests suggests that habitat restoration is possible in deforested areas. It does not, however, indicate that this species will persist without active conservation efforts or that it will not be strongly affected by deforestation (Hutto 1988). Although secondary forests are becoming more common in some tropical agricultural landscapes as people migrate from rural to urban locations (Grau & Aide 2008), net primary and secondary forest cover continues to decline from forest being cleared for permanent agriculture and human settlements (García-Barrios et al. 2009), and golden-winged warblers only used advanced stages of regeneration. Secondary forest is still rare in the Neotropics relative to more degraded land cover types (Asner et al. 2009), and in Costa Rica, financial incentives and conservation regulations are the principal reasons why most secondary forests exist (Pagiola 2008).

The habitat quality of secondary forests and agroecosystems depends upon the survival rates of the individuals in these habitat types. Evidence exists that human-modified habitats can serve as ecological traps, resulting in high densities of birds in habitats where survival rates are low (Rappole, Ramos & Winker 1989). We found no evidence of differences in apparent survival among habitats; though, low local abundance and few repeated visits per season limited our ability to separate detection probability from mortality or permanent emigration. Model selection results did indicate that apparent survival was lower within the stationary non-breeding season than among seasons. Although it is not possible to separate mortality from emigration using this model, we found that golden-winged warblers exhibit very strong within- and among-season site fidelity (Chandler 2010), suggesting the rate of decline was not attributed to movement out of the study area.

The Dail–Madsen model enabled us to estimate habitat-specific abundance, but because we used broadcast vocalizations to attract birds, the effective plot area was unknown, and thus, we could not directly estimate density. However, density can be calculated for various assumed plot sizes. Assuming that golden-winged warblers within a 100-m radius were available for detection, which is a conservative assumption given an average home range size of 8·7 ha (Chandler 2010), density would be 0·159 individuals per hectare. This is a low non-breeding season density estimate relative to most other Neotropical–Nearctic migrants (Rappole & Warner 1980; Bakermans et al. 2009), yet similar to estimates for non-breeding density of the golden-cheeked warbler Dendroica chrysoparia, an endangered species (Rappole, King & Diez 2003). This is despite the fact that reports from birdwatchers and ecotourists suggest our study area was located in a region where non-breeding density may be highest (eBird 2010).

Our results indicate that golden-winged warblers require forests within a narrow precipitation band that have microhabitat features such as hanging dead leaves and vine tangles used as foraging substrates. This habitat specialization combined with evidence of high territoriality, large home-range sizes (Chandler 2010) and low densities indicates that extensive areas of suitable habitat need to be protected to conserve this species. Unfortunately, less than half of the original forest cover still remains in their non-breeding ground range, and these forests continue to be cleared at an alarming rate (Sader & Joyce 1988; Myers et al. 2000). Much of the remaining forest has been degraded by human activities. This was true in our study area and may partially explain why within-season apparent survival was low and perhaps the low rate of among-season recruitment. We therefore suggest that conservation actions on the non-breeding grounds should be directed towards protecting what remains of lower- and middle-elevation tropical wet forests while encouraging efforts to regenerate forests on degraded lands.

Research directed at estimating vital rates such as survival and recruitment is critical to our understanding the non-breeding ecology of migrants and for developing conservation strategies for these species and their habitats. Historically estimating these population parameters was only achievable via mark–recapture studies that are often prohibitively expensive and logistically difficult to carry out except within a relatively localized area. Our study has shown that point count surveys analysed using hierarchical mixture models provide a feasible alternative for generating the estimates of demographic parameters at a landscape scale.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

We thank the members of the Cooperativa Montes de Oro as well as Victor Julio Arce Chavez, Arley Morales, Raul Raudales and Richard Trubey of the Mesoamerican Development Institute for their assistance with site selection, accommodation and travel. Many farmers in the study area graciously permitted us to work on their property. The Trejos, Gonzalez, Marin, Zunega, Fonseca and Salas families were especially generous. Carly Chandler was instrumental in the design and implementation of the field research. Seth Beaudreault, Carlos Orlando Delgado, Marcos Gonzalez, Johel Aguero, Jared Wolfe, Nicole Hazlet, Adam Anderson, Jeff Wells and Jeff Ritterson were excellent field assistants. Curtice Griffin, Peter Houlihan and John Rappole provided helpful guidance during the preparation of the manuscript. Javier Guevara of the Costa Rican Ministry of Environment and Energy provided us with necessary permits. Funding came from a Migratory Bird Conservation Act grant and from the US Forest Service’s International Programs. Two anonymous reviewers helped improve the manuscript.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information
  • Asner, G.P., Rudel, T.K., Aide, T.M., Defries, R. & Emerson, R. (2009) A contemporary assessment of change in humid tropical forests. Conservation Biology, 23, 13861395.
  • Bakermans, M.H., Vitz, A.C., Rodewald, A.D. & Rengifo, C.G. (2009) Migratory songbird use of shade coffee in the Venezuelan Andes with implications for conservation of cerulean warbler. Biological Conservation, 142, 24762483.
  • Bates, D. & Maechler, M. (2010) lme4: Linear Mixed-Effects Models Using S4 Classes. http://CRAN.R-project.org/package=lme4 .
  • Bent, A.C. (1963) Life Histories of North American Wood Warblers. Dover Publications, Inc., New York.
  • eBird (2010) eBird: An Online Database of Bird Distribution and Abundance. Version 2. Ithaca, New York. Available at: http://ebird.org/ebird/eBirdReports.
  • Blake, J.G. & Loiselle, B.A. (2000) Diversity of birds along an elevational gradient in the Cordillera Central, Costa Rica. The Auk, 117, 663686.
  • Buehler, D.A., Roth, A.M., Vallender, R., Will, T.C., Confer, J.L., Canterbury, R.A., Swarthout, S.B., Rosenberg, K.V. & Bulluck, L.P. (2007) Status and conservation priorities of Golden-winged Warbler (Vermivora chrysoptera) in North America. Auk, 124, 14391445.
  • Calvert, A.M., Walde, S.J. & Taylor, P.D. (2009) Nonbreeding-season drivers of population dynamics in seasonal migrants: conservation parallels across taxa. Avian Conservation and Ecology, 4, 5.
  • Chandler, R.B. (2010) Avian ecology and conservation in tropical agricultural landscapes with emphasis on Vermivora chrysoptera. PhD thesis, University of Massachusetts Amherst.
  • Clark, K.L., Lawton, R.O. & Butler, P.R. (2000) The physical environment. Monteverde: Ecology and Conservation of a Tropical Cloud Forest (eds N.M.Nadkarni & N.L.Wheelwright), pp. 1538. Oxford University Press, Inc, New York.
  • Confer, J.L. (1992) Golden-winged Warbler (Vermivora chrysoptera). The Birds of North America (ed A.Poole). Cornell Lab of Ornithology, Ithaca, NY.
  • Dail, D. & Madsen, L. (in press). Models for estimating abundance from repeated counts of an open metapopulation. Biometrics. DOI: 10.1111/j.1541-0420.2010.01465.x.
  • DeGraaf, R.M. & Rappole, J.H. (1995) Neotropical Migratory Birds. Cornell University Press, Ithaca, New York.
  • Fiske, I., Chandler, R. B. & Royle, J.A. (2010) unmarked: Models for Data from Unmarked Animals. http://cran.r-project.org/web/packages/unmarked/ .
  • García-Barrios, L., Galván-Miyoshi, Y.M., Valdivieso-Pérez, I.A., Masera, O.R., Bocco, G. & Vandermeer, J. (2009) Neotropical forest conservation, agricultural intensification, and rural out-migration: the Mexican experience. BioScience, 59, 863873.
  • Gill, F.B. (1997) Local cytonuclear extinction of the Golden-winged Warbler. Evolution, 51, 519525.
  • Gradwohl, J. & Greenberg, R. (1982) The effect of a single species of avian predator on the arthropods of aerial leaf litter. Ecology, 63, 581583.
  • Grau, H.R. & Aide, M. (2008) Globalization and land-use transitions in Latin America. Ecology and Society, 13, 16.
  • Haber, W.A. (2000) Plants and vegetation. Monteverde: Ecology and Conservation of a Tropical Cloud Forest (eds N.M.Nadkarni & N.L.Wheelwright), pp. 3994. Oxford University Press, Inc, New York.
  • Hutto, R.L. (1988) Is tropical deforestation responsible for the reported declines in Neotropical migrant populations? American Birds, 42, 375379.
  • Johnson, D.H. (1980) The comparison of usage and availability measurements for evaluating resource preference. Ecology, 61, 6571.
  • Johnson, M.D. & Sherry, T.W. (2001) Effects of food availability on the distribution of migratory warblers among habitats in Jamaica. Journal of Animal Ecology, 70, 546560.
  • Keating, K.A. & Cherry, S. (2004) Use and interpretation of logistic regression in habitat-selection studies. Journal of Willdlife Management, 68, 774789.
  • King, D.I., Rappole, J.H. & Buonaccorsi, J.P. (2006) Long-term population trends of forest-dwelling Nearctic-Neotropical migrant birds: a question of temporal scale. Bird Populations, 7, 19.
  • Marra, P.P. & Holmes, R.T. (2001) Consequences of dominance-mediated habitat segregation in american redstarts during the nonbreeding season. The Auk, 118, 92104.
  • Morton, E.S. (1980) Adaptations to seasonal change by migrant land birds in the Panama Canal zone. Migrant Birds in the Neotropics: Ecology, Behavior, Distribution, and Conservation (eds A.Keast & E.S.Morton), pp. 437453. Smithsonian Institution Press, Washington D.C.
  • Morton, E.S. (1990) Habitat segregation by sex in the hooded warbler: experiments on proximate causation and discussion of its evolution. The American Naturalist, 135, 319333.
  • Myers, N., Mittermeier, R.A., Mittermeier, C.G., da Fonseca, G.A. & Kent, J. (2000) Biodiversity hotspots for conservation priorities. Nature, 403, 853858.
  • Newton, I. (1998) Population Limitation in Birds. Academic Press, San Diego.
  • Nolan, V.J. (1978) Ecology and behavior of the Prairie Warbler Dendroica discolor. Ornithological Monographs, 26, 1595.
  • Pagiola, S. (2008) Payments for environmental services in Costa Rica. Ecological Economics, 65, 712724.
  • Powell, G.V.N., Rappole, J.H. & Sader, S.A. (1992) Neotropical migrant landbird use of lowland Atlantic habitats in Costa Rica: a test of remote sensing for identification of habitat. Ecology and Conservation of Neotropical Migrant Landbirds (eds J.M.Hagan & D.W.Johnson), pp. 287298. Smithsonian Institution Press, Washington D.C.
  • R Development Core Team (2011) R: A language and environment for statistical computing. R Foundation for Statistical Computing. Vienna, Austria. ISBN 3-900051-07-0, URL http://www.R-project.org/.
  • Rappole, J.H., King, D.I. & Barrow Jr, W.C. (1999) Winter ecology of the endangered Golden-cheeked Warbler. Condor, 101, 762770.
  • Rappole, J.H., King, D.I. & Diez, J. (2003) Winter- vs. breeding-habitat limitation for an endangered avian migrant. Ecological Applications, 13, 735742.
  • Rappole, J.H., Ramos, M.A. & Winker, K. (1989) Wintering wood thrush movements and mortality in southern Veracruz. The Auk, 106, 402410.
  • Rappole, J.H. & Tipton, A.R. (1991) New harness design for attachment of radio transmitters to small passerines. Journal of Field Ornithology, 62, 335337.
  • Rappole, J.H. & Warner, D.W. (1980) Ecological aspects of migrant bird behavior in Veracruz, Mexico. Migrant Birds in the Neotropics: Ecology, Behavior, Distribution, and Conservation (eds A.K.Keast & E.S.Morton), pp. 353393. Smithsonian Institution Press, Washington, DC.
  • Restall, R., Rodner, C. & Lentino, M. (2007) Birds of Northern South America: An Identification Guide, Volume 2: Plates and Maps. Yale University Press, New Haven, CT USA.
  • Robbins, C.S., Fitzpatrick, J.W. & Hamel, P.B. (1992) A warbler in trouble: Dendroica cerulea. Ecology and Conservation of Neotropical Migrant Landbirds (eds J. M.Hagan III & D. W.Johnston), pp. 549562. Smithsonian Institution Press, Washington, DC.
  • Royle, J.A. (2004) N-mixture models for estimating population size from spatially replicated counts. Biometrics, 60, 108115.
  • Sader, S.A. & Joyce, A.T. (1988) Deforestation rates and trends in Costa Rica, 1940 to 1983. Biotropica, 20, 1119.
  • Saracco, J.F., DeSante, D.F., Pyle, P., Morales, S., Alcántara, J.L. & de Vivar Alvarez, C.R. (2008) Assessing Winter Habitat Quality for Migratory Landbirds. A report to the Institute for Bird Populations. http://www.birdpop.org/DownloadDocuments/MoSIReport2007.pdf.
  • Sauer, J.R., Hines, J.E. & Fallon, J. (2008) The North American Breeding Bird Survey, Results and Analysis 1966–2007. Version 5.15.2008. USGS Patuxent Wildlife Research Center, Laurel, MD. Available at: http://www.mbr-pwrc.usgs.gov/bbs/bbs.html.
  • Sherry, T.W. & Holmes, R.T. (1996) Winter habitat quality, population limitation, and conservation of Neotropical-Nearctic migrant birds. Ecology, 77, 3648.
  • Studds, C.E. & Marra, P.P. (2005) Nonbreeding habitat occupancy and population processes: an upgrade experiment with a migratory bird. Ecology, 86, 23802385.
  • Tramer, E.J. & Kemp, T.R. (1980) Foraging ecology of migrant and resident warblers and vireos in the highlands of Costa Rica. Migrant Birds in the Neotropics: Ecology, Behavior, Distribution, and Conservation (eds A.Keast & E.S.Morton), pp. 285297. Smithsonian Institution Press, Washington D.C.
  • Young, B.E., DeRosier, D. & Powell, G.V.N. (1998) Diversity and conservation of understory birds in the Tilarán Mountains, Costa Rica. The Auk, 115, 9981016.

Supporting Information

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Appendix S1. Custom version of the R package ‘‘unmarked’’ used to fit the Dail-Madsen model. The package can be installed on Windows using the R command: install.packages(‘‘unmarked_0.0-0.zip’’, repo=NULL).

Appendix S2. List of golden-winged warblers, blue-winged warblers, and their hybrids encountered between 2006–2010 in the Cordillera de Tilarán, CostaRica.

Fig. S1. Location of study area (black square) with reference to elevation (m) in Costa Rica.

Table S1. Summary statistics of variables considered in abundance models for golden-winged warblers surveyed with point counts in the Cordillera de Tilarán, CostaRica, 2007–2009.

Table S2. Summary statistics of home range habitat availability variables for 17 radio-tracked golden-winged warblers surveyed at 97 point count station counts in the Cordillera de Tilarán, Costa Rica, 2007–2009.

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials may be re-organized for online delivery, but are not copy-edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

FilenameFormatSizeDescription
JPE_2001_sm_AppendixS1-S2.zip940KSupporting info item
JPE_2001_sm_FigS1.doc245KSupporting info item
JPE_2001_sm_TablesS1-S2.doc52KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.