Study Sites

We conducted our study from 2012 to 2015 on 3 sites within a transitional zone between central hardwood forest and grassland in Boone, Randolph, and Howard counties in Central Missouri. Our study sites were the Thomas S. Baskett Wildlife Research and Education Center (38.73°N, 92.2°W; 890 ha) in 2012−2015; Rudolf Bennitt State Conservation Area (39.13°N, 92.25°W; 1146 ha) in 2013−2015; and Three Creeks Conservation Area (38.21°N, 92.28°W; 575 ha) in 2014−2015; hereafter Baskett, Bennitt, and Three Creeks, respectively. Study sites consisted of mixed-hardwood forest interspersed with successional red cedar (Juniperus virginiana) stands. Sawtimber density and snag density are similar across the 3 study sites (P > 0.1). All sites are situated within a matrix of forest patches, old-fields, pasture, and cropland. Acadian Flycatchers nested at all 3 sites and Ovenbirds nested at Bennitt and Baskett; in 2012, however, we only monitored Ovenbirds at Baskett.

Study Species

Breeding Bird Survey (BBS) data from 1966 to 2013 indicate that Ovenbird and Acadian Flycatcher populations have remained relatively stable nationwide (Sauer et al. 2014). Ovenbirds arrive the final week of April in our region, initiate breeding in early May, and generally stop nesting in late July. Ovenbirds will renest after nest failure but do not attempt a second clutch after successfully fledging young. Acadian Flycatchers typically initiate and continue nesting in central Missouri later than Ovenbirds, nesting from mid-May through mid-August (J. M. A. Jenkins personal observation). Acadian Flycatchers will renest after nest failure and sometimes initiate a second clutch after successfully breeding; 28% of females in our study region fledged a second brood when monitored into September in 2007−2009 (Hirsch-Jacobson 2011). Ovenbird clutches range from 2 to 5 eggs, and have a 12−13 day incubation period, with juveniles fledging ∼8 days after hatching (Porneluzi et al. 2011). Fledged Ovenbird young are unable to fly for the first 3−4 days out of the nest and are kept apart from siblings, camouflaged among the leaves and groundcover, and fed by adults. After 4 days the young become increasingly mobile and vocal, remaining mainly on the ground or low vegetation in semi-dependent family groups for 20−30 days (J. M. A. Jenkins personal observation). Acadian Flycatcher clutches typically contain 2 or 3 eggs, have a 13−15 day incubation period, and fledge 13−14 days after hatching (Whitehead and Taylor 2002). Unlike Ovenbird juveniles, Acadian Flycatcher young remain in close association with siblings throughout the dependent period, often perching “snuggled” together in the canopy (Mumford 1964). Juveniles remain partially dependent upon parents for 18–24 days postfledging (J. M. A. Jenkins personal observation).

Nest Monitoring and Radio Telemetry

We found nests from mid-May to mid-August each year by systematically searching appropriate habitat in study areas and observing adult behaviors. We monitored nests every 3−5 days, more frequently near hatching and fledging periods, recording the nest stage, number of eggs, nestlings, and occurrence of Brown-headed Cowbird (Molothrus ater; hereafter cowbird) parasitism (Martin and Geupel 1993). On the day before projected fledging, we captured all available nestlings and recorded nestling mass (± 0.1g). We supplemented Ovenbird nest captures with opportunistically hand-caught non-volant fledged juveniles (1−2 days out of the nest). We attached unique combinations of 3 colored leg bands and a standard aluminum U.S. Geological Survey leg band to all captured Ovenbirds and attached radio transmitters to 1 or 2 (rarely 3) individuals per brood. All Acadian Flycatcher nestlings received a standard aluminum U.S. Geological Survey band and one juvenile per nest received a single colored leg band and a radio transmitter. Transmitters were attached using a leg-loop harness made with flexible cording (Rappole and Tipton 1991). In 2012, transmitters weighed 0.55 grams, were 3.5–5% of Ovenbird juvenile mass at time of attachment, and had an expected battery life of 22 days (model A1015, Advanced Telemetry Systems (ATS), Itasca, Minnesota, USA). In 2013–2015, transmitters weighed 0.3 g, were 1.8–2.8% of Ovenbird mass and 2.3–3.3% of Acadian Flycatcher mass at time of attachment, and had an expected battery life of 44, 29, and 44 days depending on the model (2013 and 2015: model A2414, ATS; 2014: model PicoPip Ag337, Biotrack, Wareham, Dorset, UK).

We located birds daily or nearly daily by homing using handheld receivers (model R410, ATS, and model R1000, Communication Specialists, Orange, California, USA) and handheld directional antennas (Yagi 3-element and H-Type, ATS). We observed radio-tagged birds for as long as possible without disturbing the individual, usually for 5−20 min. We recorded locations of nests and juveniles in Universal Transverse Mercator (UTM) coordinates with handheld GPS units (GPS error < 10 m). We recorded the coordinates of the location where we first sighted the individual, or if we flushed them, of the area where they resumed normal activity. We assumed our presence did not alter their habitat use. We relocated individuals until the transmitter signal was no longer detectable (transmitter battery failure or dispersal out of study area) or until we determined mortality. We assigned a cause to mortality events when possible. We assumed that an individual was dead if (1) the body or pieces of marked bird were found; (2) the transmitter was found with juvenile feathers or body parts; (3) the transmitter signal was lost, the bird was <14 days postfledging, and we resighted marked siblings or adults near the last location; (4) the damaged or digested radio was found near last location and bird was <4 days postfledge; or (5) we tracked signal to snake body or to same location in hollow tree for >3 days (assumed predation by snake or other). If a signal was lost after 15 days postfledging, we searched for the signal across the study area and surrounding forest patches on foot, by truck, or by helicopter. If the signal was still lost at the end of the potential battery life, we censored it after the last actual observation because we could not distinguish long-distance movement from mortality.

Vegetation Sampling

After family groups left the area, we sampled habitat structure at each nest and for at least half of fledgling locations. We used vegetation variables from the previous location for juvenile mortality events when a body was either not found or was found far from the family group, reasoning that the transmitter had been moved by a predator or scavenger. We estimated nest height to the nearest 0.5 m and estimated the percent of nest concealed by vegetation at nest height, 1 m from the nest in each cardinal direction, from above, and below (flycatchers only). We calculated canopy cover at each use point (nest or juvenile location) using the average of 4 spherical densiometer readings (1 in each cardinal direction). We averaged litter depth measurements taken at the central point and 2 m from the central point in each cardinal direction. We estimated percent green groundcover, live herbaceous or woody vegetation >0.3 m high, within an 11.3-m diameter circle. We estimated shrub density by counting woody stems <3 cm in diameter, at ∼1.3 m above ground, along two 22.6 m transects in cardinal directions, bisecting the use point and converted this count to density of stems per hectare (stems ha⁻¹). We measured the diameter at breast height (DBH) of all stems >3 cm DBH in a 10-factor basal area wedge plot and recorded trees as deciduous, coniferous (primarily cedars), or dead trees (snags). We calculated total basal area (BA) of all live trees and stem densities per hectare of saplings (3.0−12.5 cm DBH), pole timber (12.5−27.5 cm DBH), sawtimber (>27.5 cm DBH), and snags greater than 12.5 cm DBH (West 2009). We estimated understory foliage density using the average of 4 density board (0.3 m wide × 2.0 m tall) measurements taken from 11.3 m in each cardinal direction from the central point. Density board measurements were divided into 3 regions: low (0−0.3 m), mid (0.3−1.0 m), and high (1−2 m). We also created an overall groundcover metric by multiplying the low foliage density board measure with the percent green groundcover for each sample plot. We calculated distance to nearest non-forest edge for each nest and juvenile location remotely in ArcGIS (ESRI 2012) using the 30 m resolution Missouri 2005 Land Use Land Cover Database refined using aerial photos of field sites ( http://msdis.missouri.edu/; USDA-FSA Aerial Photography Field Office). Non-forest edge included all forest boundaries adjacent to ponds, roads, power line cuts, and other non-forest land uses that were visible from aerial photos; however, trails and non-improved roads with full canopy coverage were not considered edge.

Survival Analysis

We used the logistic exposure method to estimate daily survival and relationships with covariates for both nests and postfledging juveniles (Shaffer 2004, Shaffer and Thompson 2007; GENMOD Procedure, SAS Institute, Cary, North Carolina, USA). Because we monitored multiple Ovenbird juveniles from the same brood, we adjusted standard errors for repeated measures using generalized estimating equations by identifying brood as the subject (SAS Institute 2008; Schreiber et al. 2016). We used an information-theoretic approach (Burnham and Anderson 2002) to determine support for effects of temporal, intrinsic, local vegetation (understory and tree level) covariates and distance to non-forest edge on survival. We first tested a priori nest and postfledging survival models within 4 subcategories to determine which were most supported and used these in the final additive model set. The temporal models for both juveniles and nests included a categorical effect of year (0−4 for Ovenbirds and 0−3 for Acadian Flycatchers), ordinal date, ordinal date², ordinal date³, a categorical variable of 2-seasons, a categorical variable of 3-seasons, and additive models of year and the other variables. Nest temporal model sets also included the categorical variable nest stage (lay, incubation, and nestling) and additive models of nest stage with other temporal variables. We created categorical variables representing seasonal intervals using the 50^th percentile of fledging dates to create 2 intervals and the 33^rd and 66^th percentile to create 3 intervals for each species. The intrinsic models for juveniles included mass at fledging, number fledged per brood, and their additive combination. The intrinsic models for nests included nest height (flycatchers only), cowbird nest parasitism status, nest concealment, and their combination. Vegetation models included all singular and additive combinations of groundcover, shrub density, litter depth, understory foliage density, sapling density, pole timber density, sawtimber density, total basal area ha⁻¹, and canopy cover. If distance to non-forest edge outperformed the null model, it was included in the final additive model set. We normalized all continuous variables before conducting analysis.

Each subcategory model-set was compared to a null model and evaluated using Akaike's Information Criterion adjusted for small sample sizes (AIC_c). The nest survival null model included intercept and a fixed effect of site. The postfledging survival null model included intercept, site, and age (days since fledging). We included site as a categorical fixed effect in all models to account for any variation due to differences in predator communities or landscape effects between our study locations. We include a fixed effect of age (days out of the nest) in all postfledging models because juveniles are most vulnerable immediately after leaving the nest (Anders et al. 1997, Brown and Roth 2004, Vitz and Rodewald 2011, Streby and Andersen 2013). We considered subcategory models competitive and moved them to the final model set if a model AIC_c was lower than the null model, was within 2 AIC_c units of the best model, and did not differ from more supported models only by the addition of uninformative parameters (Arnold 2010). If there were competitive models that described the same aspect of a hypothesis (such as 2-seasons vs. 3-seasons), only the top model was carried forward. Final model sets included all singular and additive combinations of the top subcategory models.

We estimated daily nest and daily postfledging survival for each species with the most-supported model or model-averaged coefficients and predictions when more than one model was competitive in the final model suite (Burnham and Anderson 2002). We present daily survival predictions for each covariate holding all other continuous factors at their means and categorical values at their observed frequencies. We calculated nest period survival for Ovenbirds based on a 25-day nesting cycle (4 lay days, 12 incubation days, and 8 nestling days). For Acadian Flycatchers we used a 30-day nesting cycle (2 lay days, 14 incubation days, 14 nestling days). Overall postfledging period survival rates were created using cumulative survival for the postfledging period: 1−23 days postfledging for Ovenbirds and 1−20 days postfledging for Acadian Flycatchers, based on the mean number of days juveniles were dependent upon parents (Jenkins 2016).

Population Growth Models

We used a simple population growth model to illustrate the potential utility of incorporating empirical postfledging survival estimates into annual juvenile survival: λ = P_A + P_Jβ, where lambda (λ) is the population growth rate, P_A is the annual survival of adult females, P_J is the annual survival of juveniles, and β is the number of fledged juvenile females produced annually (Pulliam 1988). A population is stable if λ = 1. We calculated β using ½ Y_N × (1 − (1 − P_N )ⁿ ), where P_N is nest success, n is number of nest attempts, and Y_N is the mean number of young produced per fledged brood (Anders and Marshall 2005). We assumed a 50:50 sex ratio of fledglings. We did not consider the potential for double brooding in Acadian Flycatchers, or movement between populations in our models. Historically, when empirical estimates of juvenile survival are unknown, first year survival has been arbitrarily designated based on adult survival, typically as half of adult survival (Ricklefs 1973, Greenburg 1980, Temple and Cary 1988). When postfledging survival is known, P_J can be written as P_PF × P_W, where P_PF is postfledging period survival and P_W is overwinter survival. Unfortunately, P_W is unknown for the majority of migratory species, including our study species. Nonbreeding season studies of other migratory birds, including Black-throated Blue Warblers (Setophaga caerulescens), American Redstarts, (S. ruticilla), and Barn Swallows (Hirundo rustica), have found equal survival rates for nonbreeding adult and juvenile migratory birds (Marra and Holmes 2001, Sillett and Holmes 2002, Grüebler et al. 2014). However, inexperience may reduce juvenile survival during migration (e.g., Oppel et al. 2015), and juvenile birds may be excluded from high-quality wintering habitat, increasing mortality (Sherry and Holmes 1996, Marra and Holmes 2001). We assumed that P_PF < P_W and modeled 2 conservative scenarios for P_W: where P_W = P_A and where P_W = ¾ P_A. We compared lambda from postfledge inclusive models and a model produced using the generic juvenile survival value of half of adult annual survival, keeping all other model components stable. We used our calculated fecundity values and an adult annual survival rate of 0.62 for both species—0.62 is the mean of all published adult survival rates for Ovenbirds and is also commonly used for other small migratory passerines (Temple and Cary 1988, Donovan et al. 1995). There are no good estimates of annual survival for Empidonax flycatchers. We present all summary variables as means ± standard errors (SE).

Ovenbird Survival

We monitored 94 Ovenbird nests every 2.8 ± 0.06 days for a total of 308 observations. Thirty-nine Ovenbird nests successfully fledged a mean of 2.86 ± 0.05 young weighing 14.52 ± 0.06 g at time of capture. Predation was the primary cause of nest failure, accounting for 90% of all nest failures. Fifty-three percent of Ovenbird nests were parasitized by cowbirds. Of these 50 parasitized nests, 19 successfully fledged at least 1 Ovenbird. Most nests were first observed after the lay stage (Ovenbirds: 8% lay, 51% incubation, and 41% nestling). Fledging date for Ovenbirds ranged from May 26 to July 15. Ovenbird nests were 69.24 ± 2.08% visually concealed. Two Ovenbird nests failed due to adult mortality at the nest.

We attached transmitters to 50 Ovenbird fledglings from 36 known nests (7 from 3 nests in 2012, 11 from 6 nests in 2013, 17 from 14 nests in 2014, and 15 from 13 nests in 2015). An additional 12 non-volant fledged Ovenbirds were opportunistically captured and radio-tagged postfledging (1–2 days postfledge). Four Ovenbird transmitters fell off before radio failure (harness strap failure) and were censored after the last color-band observation (day 1, day 3, day 8, day 19). We recorded 29 Ovenbird mortalities (6 in 2012, 6 in 2013, 9 in 2014, and 8 in 2015). All recorded mortalities occurred in the first 10 days of the postfledging period, with 13 individuals that died before the first relocation. Three individuals from one nest were killed in a weather event on day one and were censored from the analysis. All other Ovenbird mortalities were categorized as predation. Multiple siblings in a brood were depredated in 2 cases, but none of the sibling mortalities took place on the same day. We observed postfledging Ovenbirds for 1−49 days for a total of 669 observations. We were able to sample vegetation at 556 of those observed locations. Birds that survived the study period with radios intact were monitored for 30.13 ± 1.26 days (min = 23 days). Ovenbird nests and juvenile locations differed in mean understory foliage density, litter depth, shrub density, groundcover, overall tree density, and distance from non-forest edge (Table 1).

TABLE 1.

Arithmetic mean values (± SE) of vegetation from all locations used by Acadian Flycatcher nests (n = 264) and juvenile locations (n = 442), and Ovenbird nests (n = 94) and juvenile locations (n = 556), in Boone County, Missouri, from 2012 to 2015. F and P values are given for generalized linear model with fixed effect of site and stage (nest vs. postfledge). Degrees of freedom (df) for the difference between nesting and postfledging Ovenbird and Acadian Flycatcher models was 1 and 657 and 1 and 702, respectively.

We had an effective sample size of 754 days for Ovenbird nest survival models and 856 days for Ovenbird postfledging survival models. We model averaged 2 of 22 additive nest survival models, including the covariates site, 3-season, sawtimber density, and snag density (Table 2 and Figure 1). Of the model-averaged covariates, sawtimber density and 3-seasons did not include zero in their 95% confidence intervals (Appendix Table 5 and Figure 1). We model averaged 5 of 8 additive Ovenbird postfledging survival models, including the covariates site, mass, age, sapling density, basal area ha⁻¹, and distance to edge (Table 2 and Figure 2). Age was the only model-averaged effect with a 95% confidence interval that did not overlap zero (Appendix Table 6 and Figure 2). Uninformative parameters from supported models not included in model averaging included groundcover (β = −0.27) and foliage density (β = −0.35). Daily nest survival was lowest for nests that fledged early in the season (Figure 1A). Daily nest survival decreased as sawtimber density increased (Figure 1C). Overall nest period (25 days) and postfledging period (0−23 days) survival for Ovenbirds was 0.27 ± 0.06 and 0.50 ± 0.09, respectively.

TABLE 2.

Summary of model-selection results from the best-ranked additive candidate models of top a priori intrinsic, temporal, edge, and vegetation subcategory models explaining survival of Ovenbird nests and postfledging juveniles in central Missouri, 2012−2015. Null models are also included for comparison. Models are ranked according to Akaike's Information Criterion adjusted for small sample sizes (ΔAIC_c). Models with a lower AIC_c have more substantial support. Number of parameters (K) in each model includes the intercept, site and each additional explanatory variable. Deviance (Dev) and Akaike's model weights (w_i) are also shown.

FIGURE 1.

Predictions of the best-supported models showing the effects of season (A), site (B), sawtimber density (C), and snag density (D) on daily survival of Ovenbird nests in Missouri, 2012−2015. Estimates are provided for the range of effects sampled while holding other variables at their mean or observed frequency. Error bars (A and B) and shaded areas (C and D) represent 95% confidence intervals.

FIGURE 2.

Predictions of the best-supported models showing the effects of age (A), sapling density (B), mass (C), basal area (D), distance to non-forest edge (E), and site (F) on daily survival of postfledging Ovenbirds in Missouri, 2012−2015. Estimates are provided for the range of effects sampled while holding other variables at their mean or observed frequency. Shaded areas (A–E) and error bars (F) represent 95% confidence intervals.

Acadian Flycatcher Survival

We monitored 264 Acadian Flycatcher nests every 3.56 ± 0.04 days for a total of 1,258 observations. Eighty Acadian Flycatcher nests successfully fledged a mean of 2.39 ± 0.03 young weighing 11.36 ± 0.04 g at time of capture. Predation was the primary cause of nest failure, accounting for 90% of all nest failures. Thirteen percent of Acadian Flycatcher nests were parasitized by cowbirds. No parasitized Acadian Flycatcher nests successfully fledged young. The majority of nests were first observed after the lay stage (3% lay, 53% incubation, and 42% nestling). Acadian Flycatcher nests were between 1.25 and 15 (5.98 ± 0.17) meters high and were 44.14 ± 1.19% visually concealed. Acadian Flycatchers fledging dates ranged from June 12 to August 5.

We deployed radio transmitters on 45 Acadian Flycatcher fledglings (11 in 2013, 13 in 2014, and 21 in 2015). We observed Acadian Flycatchers for 1−46 days for a total of 541 observations. We sampled vegetation at 442 juvenile Acadian Flycatcher locations. Birds that survived the study period with radios intact were monitored for 27.11 ± 1.57 days (min = 15 days). The majority of juvenile observations came from Baskett (77%), with 7% from Rudolf and 16% from Three Creeks. Five transmitters fell off before radio failure (harness strap failure) and were censored after the last color-band observation (3 at 1-day and 2 at 2-days postfledging). We recorded 10 Acadian Flycatcher mortality events; all but one occurred before independence from parents (1 in 2013, 3 in 2014, and 6 in 2015). Seven birds died before the first relocation. One carcass was found 2 days postfledging, undamaged but wet after a severe storm. The other 9 mortalities were classified as depredation. Acadian Flycatcher nests and juvenile locations differed in mean canopy cover, total tree density, and total basal area (Table 1).

The effective sample sizes for Acadian Flycatchers were 4,002 days for nesting and 695 days for postfledging. There were 24 models in the final nest model set and 88 models in the final postfledging model set. Two models were model averaged to create final nest survival estimates, containing the covariates site, stage, 3-season, and understory foliage density (Table 3 and Figure 3). The nest covariates season and stage had 95% confidence intervals that did not overlap zero (Appendix Table 7 and Figure 3). There were 2 postfledging models that were supported in the final postfledging model selection set; however, only the top model was used for estimates because sawtimber density was included within estimates of total tree density (Table 3). There was a negative effect of tree density (β = −0.75) in the top model, which corresponded to a positive effect of sawtimber density (β = 0.78) in the collinear model. The model postfledging covariates for age and tree density had confidence intervals that did not overlap zero (Appendix Table 8 and Figure 4). Nest and postfledging survival was lowest early in the season (Figure 3A and Figure 4A). Daily nest survival decreased with increasing understory foliage density (Figure 3D) while daily postfledging survival was negatively related with tree stem density (Figure 4C). Overall nest period (30 days) and postfledging period (1–20 days) survival of Acadian Flycatchers was 0.30 ± 0.03 and 0.89 ± 0.11, respectively.

TABLE 3.

Summary of model-selection results from the best-ranked additive candidate models of top a priori intrinsic, temporal, and vegetation subcategory models explaining survival of Acadian Flycatcher nests and postfledging juveniles in central Missouri, 2013−2015. Null models are included for comparison. Models are ranked according to Akaike's Information Criterion adjusted for small sample sizes (ΔAIC_c). Models with a lower AIC_c have more substantial support. Number of parameters (K) in each model includes the intercept, site, and each additional explanatory variable. Deviance (Dev) and Akaike's model weights (w_i) are also shown.

FIGURE 3.

Predictions of the best-supported models showing the model averaged effects of season (A), site (B), nest stage (C), and understory foliage density (D) on daily survival of Acadian Flycatcher nests in Missouri, 2013−2015. Estimates are provided for the range of effects sampled while holding other variables at their mean or observed frequencies. Error bars (A–C) and shaded area (D) represent 95% confidence intervals.

FIGURE 4.

Predictions of the best supported models showing the effects of season (A), age (B), tree density (C), litter depth (D), mass (E), and site (F) on daily survival of postfledging Acadian Flycatchers in Missouri, 2013−2015. Shaded areas (A–E) and error bars (F) represent 95% confidence intervals.

Population Growth

Population growth models using the arbitrary juvenile survival estimate of ½ P_A projected populations in decline for both species (Table 4). Acadian Flycatcher models incorporating empirical postfledging period survival with either high or conservative winter survival predicted either a 26% or 13% increase in lambda, respectively, above the arbitrary estimate (Table 4). The effect of incorporating empirical postfledging survival estimates was not as great for Ovenbird population growth estimates. The model incorporating Ovenbird postfledge survival with high winter survival projected a 3% higher growth rate compared to the arbitrary estimate and the model incorporating Ovenbird postfledge survival with conservative winter survival projected a 5% lower growth rate compared to the arbitrary estimate (Table 4).

TABLE 4.

Population growth estimates (λ) and model parameter values for Missouri forest fragment populations from 2012 to 2015 including annual survival rates for adults (P_A) and juveniles (P_J), juvenile postfledging period survival (P_PF) and overwinter survival (P_w), nest success (P_N), mean number of nest attempts (n), and mean number of fledglings per fledged brood (Y_N).