Natal forest fragment size does not predict fledgling, premigration or apparent annual survival in Wood Thrushes

ABSTRACT Determining the drivers and mechanisms for first year survival of migratory songbirds has been an understudied area in population dynamics due to the difficulty in tracking juveniles once they have dispersed from the natal site. With the advancement in miniaturization of radio-tags (battery life ∼400 days) and the development of the Motus Wildlife Tracking System, we tracked 189 Wood Thrush (Hylocichla mustelina) nestlings through independence and to fall migration departure, and their return the following spring. Natal forest fragment size and landscape forest cover (at different spatial scales) were not strong predictors of fledgling, pre-migration, or apparent annual survival; and onset of fall migration was best predicted by fledge date but not natal fragment size. Survival probability was lowest the first 16 days post-fledging (70%, or 0.86 weekly survival probability), very high for juveniles as they explored the landscape prior to fall migration (89%, or 0.99 weekly survival probability) and low during their first migration and wintering season (26%, or 0.95 weekly survival probability). To our knowledge, this is the first study to directly estimate annual apparent juvenile survival in a migratory songbird using year-round radio-tracking. Our study suggests that small forest fragments are important for the conservation for forest songbirds because they can support high survival of juveniles. LAY SUMMARY Understanding the main drivers of survival during the full annual cycle of juvenile songbirds is urgently needed to better predict population growth. A key question for forest songbirds is whether nesting in small forest fragments negatively affects juvenile survival. The advancement in radio-tags (∼400-day battery life for small tags) alongside the Motus Wildlife Tracking System allowed us to estimate apparent juvenile survival at three life stages (fledgling, pre-migration, and migration/wintering) for a declining migratory forest songbird, the Wood Thrush. We show that natal fragment size does not predict survival probability for fledgling or juvenile Wood Thrushes. Weekly survival was lowest for fledglings on their natal territory, high for juveniles as they explore the landscape prior to fall migration, and lower during their first migration and wintering season. Our study suggests that even small forest fragments on the breeding grounds are important as they can support high juvenile survival for a forest songbird and that survival is driven primarily by factors outside of the breeding grounds. RESUMEN Determinar los factores y mecanismos que afectan la supervivencia del primer año de las aves canoras migratorias ha sido un área poco estudiada en la dinámica de poblaciones debido a la dificultad de rastrear a los juveniles una vez que se han dispersado del sitio natal. Con el avance en la miniaturización de los marcadores de radio (vida útil de la batería ∼400 días) y el desarrollo del Sistema de Seguimiento de Vida Silvestre Motus, seguimos a 189 polluelos de Hylocichla mustelina durante el proceso de independencia hasta la partida de la migración de otoño, y su regreso la primavera siguiente. El tamaño del fragmento de bosque natal y la cobertura forestal del paisaje (a diferentes escalas espaciales) no fueron buenos predictores de la supervivencia de los volantones, la supervivencia previa a la migración o la supervivencia aparente anual; y el inicio de la migración de otoño se predijo mejor por la fecha de emancipación, pero no por el tamaño del fragmento natal. La probabilidad de supervivencia fue más baja en los primeros 16 días después de la emancipación (70%, o una probabilidad de supervivencia semanal de 0.86), muy alta para los juveniles mientras exploraban el paisaje antes de la migración de otoño (89%, o una probabilidad de supervivencia semanal de 0.99) y baja durante su primera migración y temporada de invierno (26%, o una probabilidad de supervivencia semanal de 0.95). Hasta donde sabemos, este es el primer estudio que estima directamente la supervivencia aparente anual de juveniles en un ave canora migratoria utilizando seguimiento por radio durante todo el año. Nuestro estudio sugiere que los pequeños fragmentos de bosque son importantes para la conservación de las aves canoras de bosque porque pueden mantener una alta supervivencia de los juveniles.


LAY SUMMARY
• Understanding the main drivers of survival during the full annual cycle of juvenile songbirds is urgently needed to better predict population growth.• A key question for forest songbirds is whether nesting in small forest fragments negatively affects juvenile survival.
• The advancement in radio-tags (~400-day battery life for small tags) alongside the Motus Wildlife Tracking System allowed us to estimate apparent juvenile survival at three life stages (fledgling, pre-migration, and migration/wintering) for a declining migratory forest songbird, the Wood Thrush.• We show that natal fragment size does not predict survival probability for fledgling or juvenile Wood Thrushes.
• Weekly survival was lowest for fledglings on their natal territory, high for juveniles as they explore the landscape prior to fall migration, and lower during their first migration and wintering season.• Our study suggests that even small forest fragments on the breeding grounds are important as they can support high juvenile survival for a forest songbird and that survival is driven primarily by factors outside of the breeding grounds.

INTRODUCTION
Understanding which breeding, migration, and overwintering factors influence survival is important in modelling the population dynamics of migratory birds (Hannon andMartin 2006, Naef-Daenzer andGrüebler 2016) and, subsequently, developing effective conservation policies (Cox et al. 2014, Streby et al. 2014, Vernasco et al. 2018).Survival during the first year of life for a long-distance migratory songbird is fraught with challenges and risks as they leave their natal territory for the first time, migrate thousands of kilometers and overwinter in the tropics (Naef-Daenzer and Grüebler 2016).The drivers and mechanisms behind first-year survival have been an understudied area in population dynamics of songbirds (Cox et al. 2014, Drummond et al. 2019) because of the difficulty in tracking their movements once they disperse from the natal territory (Mitchell et al. 2011, Tarof et al. 2011).Because almost all juvenile songbirds disperse away from their natal area to unknown breeding locations (e.g., Rushing et al. 2021), mark-resight methods and archival tracking devices that require recapture after migration are typically of little use for studying juvenile survival or fall migration (McKinnon et al. 2014).Stable isotopes in feathers can be used to estimate juvenile dispersal; for instance, in Ovenbirds (Seiurus aurocapilla), most first-year breeders originated from < 200 km from the study area (Hachè et al. 2014) whereas in Cerulean Warblers (Setophaga cerulea) most first-year breeders were longer-distance immigrants (Jones and Islam 2023).Detailed studies of the full annual cycle of juvenile songbirds are urgently needed (Cox et al. 2014, Streby et al. 2014, Raybuck et al. 2019, Carle-Pruneau et al. 2021) to better understand what conservation actions can improve juvenile recruitment and population stability.Annual survival probability of juvenile migratory songbirds can be partitioned into 3 non-overlapping stages: fledgling survival to independence, pre-migration survival, and migration/wintering survival (Tarof et al. 2011).Factors affecting survival no doubt differ between the fledgling period (first 16 days post-fledging when still-developing young have left the nest but receive food and protection from parents) and the pre-migration period (>16 days post-fledging to on-set of fall migration) when full-grown juveniles care for themselves and often begin local dispersal movements through the landscape prior to migration.Numerous radio-tracking studies have shown that the highest risk of mortality for young songbirds is the first days after they fledge and cannot fly well (Naef-Daenzer and Grüebler 2016, Drummond et al. 2019, Raybuck et al. 2019).Fledglings typically have an overall ~50% chance of surviving to 3 weeks after fledging, when most songbirds reach independence from the parents (Eng et al. 2009, Jones et al. 2017, Raybuck et al. 2019).Although many studies have estimated fledgling survival in migratory songbirds, to our knowledge there have only been 2 studies that tracked independent juveniles until they departed on fall migration weeks or months after fledging, and both are for swallows.Tarof et al. (2011) radio-tracked nestling Purple Martins (Progne subis) after fledging and resighted birds at their pre-migration roost and at breeding colonies the next spring; pre-migration survival of independent juveniles was very high (87%).Evans et al. (2019) used the Motus automated telemetry array at the landscape scale and estimated 62% of juvenile Barn Swallows (Hirundo rustica) survive the first month after gaining independence.
Our study on Wood Thrush (Hylocichla mustelina) is the first we know of to estimate annual apparent juvenile survival in a migratory songbird using year-round radio-tracking, and to estimate pre-migration juvenile survival for a forest songbird.Tracking juvenile Wood Thrushes through their first year of life has been made possible with the automated long-distance radio telemetry collaborative Motus Wildlife Tracking System (Taylor et al. 2017) combined with the miniaturization of radio-tags with > 365 days battery life.This tracking system consists of an array of receiver towers that allow researchers to track tagged wildlife movements over regional and continental spatial scales.With this system many aspects of migratory bird behavior can be quantified such as patterns of connectivity (McKellar et al. 2015, Marchand et al. 2020), identifying stopover locations and duration (Smetzer andKing 2018, Beauchamp et al. 2020), and departure and arrival times (Mitchell et al. 2015, Müller et al. 2018, Dossman et al. 2023).Although much of the research to date using this system has been on adult bird movement, some studies have addressed the knowledge gap in relation to juvenile dispersal, survival, and migration strategies (Brown and Taylor 2015, Crysler et al. 2016, Cormier and Taylor 2019).
We studied Wood Thrushes because this species is a Neotropical migrant that has declined across most of its breeding range (Rushing et al. 2016, Sauer et al. 2019), so is a model species for understanding forest bird declines.In Canada, Wood Thrushes are currently listed as a Species at Risk as populations have been declining; in Ontario a decline of 3.2% yr -1 was detected from 2005-2015 (Sauer et al. 2017) representing a cumulative 27.4% decline of the population.The deciduous forests of southern Ontario are highly fragmented, but it is unknown to what extent this contributes to low juvenile survival and hence population declines in this region.
For forest birds, a key predictor of nesting success is regional forest fragmentation and forest fragment size, with reduced productivity in small fragments due to a higher incidence of nest predation and brood parasitism by Brownheaded Cowbirds (Molothrus ater) (Lampila et al. 2005, Falk et al. 2011, Richmond et al. 2012).Nest predation may decrease with increased nest distance from the edge if predators are more abundant in the agricultural matrix (Etterson et al. 2014).The presence of Brown-headed Cowbird nestlings may also increase detectability and nest predation risk (Hannon et al. 2009).Predation risk within a forest fragment can also depend on the amount regional forest cover in the surrounding landscape (Cox et al. 2012, Etterson et al. 2014).Nest predation and cowbird parasitism have been shown to increase with decreasing regional forest cover (Lloyd et al. 2005) but this relationship can depend on predator type (Cox et al. 2012).Chiavacci et al. (2018) found that higher regional forest cover increased the risk of songbird nest predation by black rat snakes (Pantherophis obsoletus) and Cooper's Hawks (Accipiter cooperii) but not by raccoons (Procyon lotor).The extent to which predation rate increased with forest cover also increased with the spatial scale at which the landscape was measured (e.g., 200 m versus 2,500 m).However, studies examining fledgling survival have not found a strong relationship to natal forest fragment size.Rush and Stutchbury (2008) radio-tracked fledgling Hooded Warblers (Setophaga citrina) and found no difference between survival rates in large versus small fragments.Eng et al. (2009) radiotracked fledgling Hooded Warblers in Ontario and found that forest fragment size was a far weaker predictor of fledgling survival than variation in forest structure resulting from prior logging practices.A parallel study on Rose-breasted Grosbeaks (Pheucticus ludovicianus) found that fledgling survival was not predicted by either fragment size or prior logging treatment (Moore et al. 2010).For Wood Thrushes in a fragmented landscape in southern Indiana, fledgling survival was significantly higher for sites with greater local (2km radius) forest cover in drought years, but the relationship was reversed in non-drought years (Vernasco et al. 2018).Other possible drivers of fledgling survival in songbirds include nestling body condition, habitat quality, and timing of fledging (Brown and Roth 2004, Mitchell et al. 2011, Vitz and Rodewald 2011, Blomberg et al. 2014).
Timing of migration is an important factor in full-life cycle demographic models because it can impact future survival and reproduction through negative carry-over effects.For instance, poor-quality winter territory (e.g., lower arthropod abundance) can delay spring migration timing and reduce subsequent reproductive success (Norris et al. 2004).For species that are territorial on their wintering grounds, poor body condition or late fall migration may make it more difficult to secure a high-quality winter territory.For instance, adult Louisiana Waterthrush (Parkesia motacilla) that were in poor condition prior to fall migration obtained lower quality winter territories and ended the nonbreeding season in poor condition (Latta et al. 2016).Although fall migration timing does not appear to affect arrival time on the wintering grounds for adult Wood Thrushes (Stutchbury et al. 2011), it is not known how fall migration timing of juveniles affects winter territory quality or over-winter survival.A radio-tracking study in Mexico found that first-year Wood Thrushes were more likely to be excluded from large forest fragments and had a higher mortality as a result (Rappole et al. 1989).
The primary objective of this study was to test if natal forest fragment size is a strong predictor of survival in the first year of life for Wood Thrushes (Hylocichla mustelina) at 4 stages of the life cycle: (1) nestling survival within the nest, (2) fledgling survival (first 16 days post-fledging), (3) pre-migration survival of independent juveniles, and (4) migration/wintering survival.We hypothesized that small forest fragments are poor quality habitat for Wood Thrush nests, fledglings, and juveniles due to higher predation risk.First, we predicted that daily nest survival is lower in small forest fragments, near forest edges.Second, we tested whether forest fragment size, percent forest cover nestling body condition before fledging, breeding territory habitat, and timing of nesting are important predictors of fledgling survival to independence at 16 days post-fledging.Third, we tested the prediction that originating from small natal fragments delays the timing of fall migration of juvenile Wood Thrushes or reduces their pre-migration and migration survival.This could occur through negative carry-over effects such as poorer body condition and later fledge date.Juveniles from nests initiated later in the season are expected to depart on fall migration later (Evans et al. 2019), and poor body condition is expected to delay migration departure (Mitchell et al. 2011, Stutchbury et al. 2011).Fourth, we tested if small natal fragment size or late fall migration timing lowered annual juvenile survival.Birds in poorer body condition or that departed later on fall migration may experience negative carry-over effects to the wintering season, decreasing annual survival.

Study Area
This study took place from 2016 to 2019 in forest fragments in Norfolk County on the north shore of Lake Erie in Ontario Canada (42.7131°N,80.5372°W).Study sites (n = 29 forest fragments) were selected to represent a gradient of forest size, ranging from 11 to 500 ha (Figure 1).The percent forest cover of the Long Point Conservation Authority jurisdiction in which Norfolk County is part of, was 21% with predominant land uses being crop agriculture (LPRCA, 2019).Norfolk County lies within the Carolinian forest zone dominated by deciduous and mixed forest types consisting of species such as sassafras (Sassafras albidum), oak (Quercus spp.), hickory (Carya spp.), tulip tree (Liriodendron tulipfera), eastern hemlock (Tsuga canadensis), eastern white pine (Pinus strobus), and understory species such as spicebush (Lindera benzoin), witch hazel (Hamamelis virginiana), and prickly ash (Xanthoxylum americanum).This region was chosen because of the high-density coverage of the Motus automated telemetry system (see Supplementary Material S1 for study area Motus tower locations).Within the breeding range for Wood Thrushes in southeastern Canada and the eastern United States, there are > 300 Motus receiver towers that potentially could remotely detect juveniles during their first migration.

Nest Monitoring and Radio-tagging
Beginning in mid-May of each year, we searched for Wood Thrush nests by locating singing males on territories and subsequently observing nests or nesting behavior.Nest contents were checked every 4-6 days using a pole with attached cellphone set to video mode.Egg incubation lasted ~14 days and nestlings fledged the nest ~26 days after the first egg was laid (Donovan et al. 1995).Only 15 nests out of 419 were abandoned during the study, in all cases during egg-laying stage, and these were not included in analyses of nest survival.
Typically, multiple nests were located and monitored at each forest fragment over the breeding season; however, not all nests were selected for radio-tagging nestlings (tagging date ranged from early June to mid-August).Nests were selected to ensure representation of the gradient of forest fragment sizes during each of the first 3 field seasons.On day 10 after hatching, nestlings were banded with uniquely numbered aluminum bands (US Fish and Wildlife/ Canadian Wildlife Service), a unique color band combination and mass, tarsus length, and wing chord measurements were taken.Body condition was calculated using a scaled mass index with mass and tarsus length (Peig and Green 2009).The scaled mass index has been identified as the best way to estimate energy reserves in some passerines (Peig andGreen 2009, Nip et al. 2018) by adjusting mass to standardized body length measure that is positively correlated with mass on a log-log scale.
The largest nestling (by mass) in the nest had a blood sample (25 µL) taken for genetic sexing (HealthGene Corporation, Vaughan, Ontario) and was equipped with a uniquely coded radio transmitter (Lotek NTQB-6, 1.5-1.7 g; ~1-yr battery life; 12.7-s burst rate) using a figure-eight leg loop harness (Rappole and Tipton 1991).This unavoidable bias toward heavier nestlings to avoid potential negative impacts of radiotags could have overestimated average juvenile survival and underestimated the effects of nestling body condition (Naef-Daenzer and Grüebler 2016, Evans et al. 2019).Three slightly different tag models were deployed due to manufacturing limitations throughout the 3-yr study (tag models were to be equivalent and direct replacements to the original); however, tags remained ~5% of the total body weight of the 10-day old nestling.Only one nestling was tagged at 131 of the 160 nests (82%); however, 2 nestlings were tagged at 29 (18%) nests.Of the 189 tagged nestlings (2016, n = 47; 2017, n = 66; 2018, n = 76) there was an even sex ratio of males to females (95:94).

Manual, Remote Radio-tracking and Aerial Telemetry
Manual tracking of fledglings was conducted using a handheld 3-or 5-element Yagi antennae connected to an SRX 600 or SRX 800 receiver (Lotek Wireless Inc.).Each radiotagged fledgling was relocated on or near its natal territory approximately every 4 days to determine if the bird was alive (moving and/or parents attending) or dead (stationary).If the signal appeared to be stationary, or could not be detected, a 1-hr search was made to locate the tag/carcass or the live bird.Manual tracking of fledglings continued until they either dispersed from the natal territory and could no longer be located, they died, or if still on their natal territory until the end of the field season in late August.
Remote tracking using the tower array of the Motus Wildlife Tracking System (Taylor et al. 2017) allowed us to detect the juveniles during the pre-migration period, early migration, and their return the following spring (see Supplementary Material S1 for study area Motus tower locations).The high density of automated towers within our study area, and surrounding regions in all directions allowed for the detection of juveniles up to the onset of fall migration and those that survived migration but did not return to their natal forest fragment.For juveniles that survived to begin migration, Motus towers in the U.S. detected 61% (50 of 82) after their fall departure flight and during their fall and/or spring migratory routes.
Aerial telemetry was conducted to independently locate juveniles that returned to the general study area each spring ( 2017, 2018, and 2019) as a supplement to Motus automated detection to determine annual survival and to estimate breeding location of birds that recruited to our study area.Approximately 5 hr was spent each year (18 July 2017, 2 June 2018, and 15 May 2019) following a grid pattern from north to south covering the study area, with each pass being ~2-10 km apart (extending ~20 km from the outside edge of the furthest site fragment).Smaller circles were made over forest fragments with the study area or when a detection was made to pinpoint the bird's location and record a GPS point.

Territory Habitat
Vegetation measurements were taken at each nest site where a nestling was radio-tagged to assess within-fragment differences in habitat among individuals that could potentially influence fledgling survival on the natal territory (e.g., Vernasco et al. 2018).The protocol followed was a revised version of the Roberts-Pichette and Gillespie (1999) terrestrial vegetation sampling protocols that uses different plot sizes to sample trees and shrubs.First, to measure trees, three 20-m × 20-m plots were established around each nest starting with a plot, 20 m and 225° from the nest, and the other two plots randomly selected within 80 m of the nest.We did not map territory boundaries of males due to time constraints and assumed that the home range was within this distance from the nest.Within each 20 × 20 m plot, all trees with > 10 cm diameter at breast height (DBH) were identified to tree species and DBH to the nearest centimeter measured.Second, five nested 2 × 2 m plots at the center and edges of each 20 × 20 m plot were used to document all shrub and sapling species and % cover of each stem.The number of large trees and total number of trees had a significant positive relationship with fragment size while number of shrubs, shrub cover, and ground cover had a significant negative relationship with fragment size (Boyd et al. 2023).

Nest survival
The percent forest cover was calculated using open access Wooded Area land cover data layer available through Land Information Ontario from Ministry of Natural Resources and Forestry (2018) and buffering each nest at the 3 spatial scales (500 m, 2 km, and 5 km from the nest to represent home ranges of different potential predators).Nest distance to forest edge and the presence/absence of Brown-headed Cowbird eggs/young were added as predictor variables.
Daily nest survival rate (DSR) was calculated with the package RMark 2.2.4 (Laake 2013) in R (R Core Team 2021, White and Burnham 1999).This model requires data input of the length of breeding season, age of the nest when first found, date when last checked and active (i.e.eggs/nestlings alive), and the fate of the nest (successful versus preyed upon).Nests were classed as successful if at least one nestling fledged.To test for effects of nest age (e.g., days since first egg was laid for each nest) on nest survival each nest's age on each day of the nesting season was provided, which allows for both time of season and nest age to be used as temporal predictors of nest survival (Rotella et al. 2004).
Models were run in a two-stage hierarchical modeling process.First, we fit models to determine if there was a change in DSR based on temporal sources of variation which included using nest age, year, and both linear and a quadratic term for time (Supplementary Material Table S2).The best-fit temporal model (nest age) was subsequently used as the base model for the predictor variables that included forest fragment size, percent forest cover at 3 different spatial scales (500 m, 2 km, and 5 km from the nest), nest distance to forest edge, and nest parasitism.The full model set included models with nest age plus one predictor variable, models with nest age plus two predictor variables (as both additive and interactive models) and models with nest age plus three additive predictor variables (Supplementary Material Table S3).
Model support was determined using the Akaike Information Criterion corrected for small sample size (AIC c ) and cumulative model weights (w i ) (Arnold 2010).Model parameter and beta estimates were averaged across all models and beta estimates were used to infer biological importance.The estimated overall nest survival was determined by raising the DSR derived from the constant model to the power of 26 (number of nesting days for Wood Thrushes).The delta method (Powell 2007) was used to calculate standard error for cumulative survival probabilities.

Fledgling survival
Data from both the manual tracking and the Motus tower detections were used to model fledgling survival and the probability of detecting a bird given that it was alive (p).Fledgling survival was estimated using the Burnham model (Burnham 1993) with the RMark package (Laake 2013) in R (R Core Team 2017).This model incorporates both live detections on, and dead recoveries between set time intervals, which, in this study was every 4 days.The Burnham model includes a term for fidelity (F) to better estimate survival since individuals may permanently emigrate away from the study area where live encounters are taking place, and so would not be detectable even if still alive.For Wood Thrush fledgling survival (1-16 days old), F was set to 1 because all dead recoveries and live encounters occurred within the study area.Radio-tagged birds were detectable dead or alive by manual tracking on or near the natal territory and detectable alive by Motus prior to the onset of fall migration.The Burnham model also considers the probability that birds who died are encountered as dead ("r").Known fate models consider that all dead birds can be located via radio-telemetry (e.g., r = 1), but in practice fledglings not old enough to be independent are sometimes not detected either alive or dead so their fate is unknown (Vernasco et al. 2018, this study).
All fledglings that were determined to be alive ≥ 16 days after fledging by Motus automated telemetry detections were coded as such in the encounter history on the last live encounter day (16 days).There were 9 individuals for whom there were gaps in manual tracking data due to scheduling and weather conflicts, but all 9 were subsequently detected by Motus towers within the study area in August and September.In addition, there were 10 tagged fledglings (out of 189 tracked) that despite intensive manual tracking disappeared during the fledgling period (average of 6 days post-fledging),but were detected by Motus prior to migration indicating that they had survived the fledgling period.The encounter history formats for all 19 of these birds were coded as "1" for the last encounter to signify surviving to 16-day post-fledging.There were 13 other fledglings that went undetected manually and were never detected by Motus and so their fates were coded as unknown (e.g., "0").
Similar to nest survival, a two-stage modelling process was implemented following similar methods to Vernasco et al. (2018).The first stage of modelling determined the top temporal model for p while holding survival (S) and r constant.The probability of detecting a tagged fledgling given that it was alive (p) was modelled with year, linearly with age, fledge date, and two age groups (age2 = 0-8 days, 9-16 days; age3 = 0-8 days, 9-12 days, 13-16 days).Detection probability could vary by year due to variation in tag design, with fledge date as later fledged birds will have a greater propensity to disperse out of receiver range, and with fledgling age because older age classes are more likely to disperse early and evade detection by manual tracking.Each parameter was modeled separately and with an additive model of year and fledge date, and the constant model for a total of 13 models (Supplementary Material Table 4).Next, the top p model was used to model S. Predictor variables modelled for S included a change in survival between calendar year (Year), linearly changing with age (Age), non-linearly changing with age (Age 2 ), time dependent (Time) and 3 age categories of 2 (age2 = 0-8 days, 9-16 days), 3 (age3 = 0-8 days, 9-12 days, 13-16 days), and 4 (age4 = 0-4 days, 5-8 days, 9-12 days, 13-16 days) age groupings, and the constant model for a total of 26 models (Supplementary Material Table 5).A quadratic term for time (Age 2 ) was included to model S (and not p) as fledgling survival with respect to age may not be linear because flight capability increases rapidly during the first week after fledging.
The top temporal model (S (Age) p (age3 + Year) r (constant) F (=1)) was then used as the base model to model S with the set of covariates that related to the key predictions (also ran models using S (Age) p (age3) r (constant) F (=1) as the base model, however, did not result in substantive change to the results).The covariates in these models included forest fragment size, % forest cover at 3 different spatial scales (500 m, 2 km, and 5 km from the nest), body condition at the time of tagging, and 2 vegetation measurements that included mean number of trees with > 30 cm DBH, and % shrub and sapling cover in the understory as the predictor variables of survival.The full model set included models with Age plus one predictor variable, models with Age plus two predictor variables, and the constant model for a total of 38 models (as both additive and interactive models; Supplementary Material Table 6).Model support was determined using the Akaike information criterion corrected for small sample size (AIC c ) and cumulative model weights (Arnold 2010).

Departure dates
Motus tower data records were accessed using the motus and motusData R packages and cleaned following guidelines provided by Birds Canada (2022) (see Supplementary Material 7 for determination of departure date).We examined juvenile migration departure dates using a linear regression (glm function, package lme4, Bates et al. 2015) and gamma distribution (link="log") in program R (v. 4.0.5;R Core Team 2021).Additive models were run with natal forest fragment size, fledge date, body condition at time of fledging, sex, and year as predictor variables.Models were ranked using AIC values adjusted for small sample size (Burnham and Anderson 2002) and top models with ∆ AIC c < 2 were considered equivalent to the best model.The full model was run using the dredge function (MuMIn package, Bartoń 2020), followed by model averaging.

Pre-migration, migration/wintering, and annual survival
The pre-migration survival period was defined as > 16 days after fledging to fall migration period, migration/wintering survival was the period after fall departure to spring arrival, and annual survival from fledging the nest to spring return.Pre-migration survival was determined using Motus tower detections that were made after August 25 (date of earliest fall migration departure) each year (n = 133).For migration/ wintering and apparent annual survival analyses, birds that were detected the following spring through Motus (plus one individual detected through aerial telemetry but not Motus) were coded as alive if they were detected within or near to the study area (up to 110 km; see Supplementary Material 8).
Survival was modelled using a linear regression with glm (family binomial with logit link) in the R package lme4 (Bates et al. 2015) rather than other survival analysis models (i.e.RMark) as we cannot estimate the encounter probability (e.g., birds that were alive but not detected) using the Motus tower detections.We used the probability of survival as the response variable with forest fragment size, fledge date (for pre-migration and annual survival only), departure date on fall migration (for migration/over-wintering survival only), nestling body condition, sex, and year as predictors.To determine if departure date had a carry-over effect on migration/ over-wintering survival, models included only the individuals that had a clear departure date (n = 82) based on time of day (departures after sunset), time of year (after August 25), and sequential Motus tower detections in a southerly direction (Supplementary Material 7).Model support was determined using AIC c and w i (Arnold 2010).The full models (for premigration, migration, and annual survival) were run using the dredge function (MuMIn package, Bartón 2020), followed by model averaging.

Nesting Success
Nest predation occurred at 54% (219 of 404) of nests.Nest age was overwhelming the top-supported temporal model with ~100% of the model weight (Supplementary Material Table 2), with nest survival decreasing with increased nest age (Figure 2).The amount of forest cover within 5 km of the nest had a cumulative model weighting of ~44%, while models including forest fragment size had only a cumulative weighting of 21% (Table 1, Supplementary Material 3).As there was little to no support for models using the 500-m or 2-km spatial cover for %forest cover, they were removed from the final model set.The top two models included nest age and percent forest cover at 5 km, with the second-best model also including cowbird parasitism.Nest survival tended to decrease with increasing forest cover and with the presence of cowbirds, but these were relatively weak predictors because the βeta coefficient estimates slightly overlapped zero.

Fledgling Survival
Observed fledgling apparent survival to 16 days after leaving nest was 70% (133 of 189) based on both manual tracking and Motus tower data.Of the 56 birds not known to have survived the fledgling period, 43 (77%) were confirmed to have been preyed upon, and 13 (23%) of the birds were never relocated alive or dead.
The top temporal model for fledgling survival was S(Age) p(age3 + Year) r(~1) F(=1) (Table 2) which had a w i of 0.26 and an increase AIC c of 2.0 with the second model S(Age + Fledge Date).Models that included S(Age) had a cumulative weight (∑w i = 0.43).Nestling body condition was a strong predictor of fledgling survival in the environmental model sets with the top models (<2 AIC) having 0.37 of the cumulative model weights (Table 3).Survival increased with age and nestlings in better body condition were more likely to survive to 16days old (Figure 3).Models that included percent forest cover at 500-m and 2-km spatial scale had similar low support (w i = 0.07-0.10)and the models with forest cover within 5 km and fragment size had no support.

Departure Dates
The average fall migration departure date for juvenile Wood Thrushes was September 19 (range: August 25 to October 15) with a tendency for females to depart before males (model estimate ± SE = 0.008 ± 0.006, P = 0.21).General linear regression indicated that the top models included fledge date, natal forest fragment size, body condition, sex, and year (∑w i = 0.79, <2 ∆ AIC, Table 4), with fledge date and year (2017 and 2018) as the only significant predictors of fall migration departure (0.0009 ± 0.0001, P < 0.001; 0.003 ± 0.0009, P = 0.0014 (2017); and 0.003 ± 0.0008, P = 0.004 ( 2018)).Juvenile Wood Thrushes that fledged earlier in the breeding season (by June 22) departed on autumn migration 9 days earlier than nestlings that fledged later in the season (between July 23 and August 7; Figure 4).

Pre-migration, Migration, and Annual Survival
Pre-migration Juvenile apparent pre-migration survival (from 16-days old to the onset of migration) was estimated from detection by Motus towers.Across all years, 89% (119 out of 133) of birds that survived the fledgling stage also survived the premigration period (duration of 52 days on average).General linear regression indicated that fledge date was not a significant predictor of pre-migration survival (0.004 ± 0.011, P = 0.69) and forest fragment size had little effect (-0.002 ± 0.002, Daily survival probability decreased with increasing nest age but was not predicted by natal fragment size (Table 1).
Although the top weighted model included forest fragment size (w i = 0.12; Table 5) there was little difference in AIC c with the second highest weighted model (w i = 0.10, ∆AIC = 0.41) which was the null model.

Migration and annual survival
Migration and over-winter survival were low, as only 26% (31 out of 119) of birds that began fall migration returned the following spring.Forest fragment size had no effect (-0.002 ± 0.002, P = 0.47) and body condition of nestlings did not predict migration survival (0.140 ± 0.153, P = 0.36).
Fall migration departure timing also did not predict migration survival (-0.009 ± 0.024, P = 0.70).Nestling body condition just before fledging was the top weighted parameter in the model set (w i = 0.46) but 95% CI overlapped zero (-0.036, 0.495).None of the modelled parameters were strong predictors of Wood Thrush juvenile migration survival (Table 6), as models < 2 AIC c included the null model (w i = 0.06).Similarly, natal forest fragment size, nestling body condition and nestling fledge date were not found to be good predictors of annual survival (Table 6, Figure 5).Across the annual cycle, from the time of fledging to returning the next spring to breed, estimated annual survival was 16%  -year period, 2016-2018 (n = 189 fledglings).For all models p (probability of detecting a live fledgling) was modelled as p(age3 + Year).Survival models included calendar year (Year), linearly changes with age (Age), non-linearly changes with age (Age 2 ), time dependent (Time), fledge date, and 3 age categories of 2 (age2), 3 (age3), and 4 (age4) age groupings.Models are ranked by Akaike's information criterion (AIC) with small sample size adjustment (AIC c ), with number of parameters (k), and model weight (w i ) given for each model.The full model set is shown in Supplementary Material Table 5 44) levels of body condition prior to fledging with standard error (SE).Fledgling survival was predicted by nestling body condition but not by natal fragment size (Table 3).
(31 189) despite the high density of Motus towers in the region that could have detected returning juveniles (Table 6) and the searches conducted with aerial telemetry.Of the 31 individuals detected in spring by Motus and aerial telemetry, 6 were last detected outside of our study area ranging in distance from 28 to 110 km away (see Supplementary Material 8 for tower locations).Aerial surveys detected 9 of 30 (30%) of individuals that Motus had detected remotely in early spring plus 1 returning individual that had not been detected by Motus.The other first-year birds were not detected with aerial surveys within our study area and must have bred elsewhere.

DISCUSSION
This is one of the first studies for a forest migratory songbird to directly estimate apparent juvenile survival at 3 life stages-fledgling (first 16-days post-fledging), pre-migration (>16 days to onset of fall migration), and migration/wintering periods.This was possible using manual and automated radio-telemetry techniques along with technological advancements in radio-tags (small tags with > 1-yr battery life).We show for Wood Thrushes, a species in chronic decline across its North American breeding range (Sauer et al. 2019), that small natal forest fragments did not reduce fledgling, premigration, or annual juvenile survival and did not delay onset of fall migration.Estimating survival probability at different stages throughout the year is a first step to identifying where mortality risk is highest and how to mitigate this through conservation actions and planning.Survival probability was lowest during the 16-day fledgling periods on their natal territory (70%, or 0.86 weekly) and the 6-month migratory/wintering period (26%, or 0.95 weekly probability), and very high for juveniles during the 6-8 weeks prior to onset of fall migration (Figure 6).

Survival: From Egg to Independent Juvenile
The daily nest survival estimate in this study (mean ± SE = 0.96 ± 0.002) was similar to what others have found for Wood Thrushes in fragmented forests (0.95 to 0.96; Newell andKostalos 2007, Etterson et al. 2014) and less than what has been documented in landscapes with very high forest cover (0.98; Schlossberg et al. 2018).Forest fragmentation effects on breeding songbirds can result in lower reproductive success due to increased nest predators (Hethcoat and Chalfoun 2015) and increased nest parasitism by Brownheaded Cowbirds (Etterson et al. 2014).In this study, however, forest fragment size was not a predictor of nest survival, although the percent forest cover within 5 km of the nest and nest parasitism by Brown-headed Cowbirds had a weak negative effect.Friesen et al. (1999) conducted a study on Wood Thrushes in southern Ontario and found that nesting success was not lower in small fragments (3-14 ha) than in larger fragments (26-140 ha).Etterson et al. (2014) conducted a study in Virginia USA and found that nest survival was lower in forest fragments compared with very large, contiguous (>9,000 ha) forest.The relationship between fragment size and nest survival is likely influenced by factors beyond the forest patch such as regional forest cover and configuration (Driscoll et al. 2010, Richmond et al. 2012) and land use matrix (e.g., agriculture, rural, or urban) (Phillips et al. 2005, Richmond et al. 2011).Our study region is set in a predominately agricultural land use matrix with an overall forest cover of only 21% and even the largest fragments (500 ha) did not provide a refuge from nest predators or cowbirds.4).

S. M. Hayes et al. cycle for first-year Wood Thrushes
As with fledgling survival studies on forest birds (Jenkins et al. 2016, Vernasco et al. 2018, Raybuck et al. 2019), we found that the first few days out of the nest are the most critical for fledgling survival.During this time, fledglings have no tail and only partially grown flight feathers and can flutter/hop short distances but cannot sustain powered flight which leaves them highly susceptible to predation (Naef-Daenzer and Grüebler 2016).Fledgling survival probability (0.70) was higher than for Hooded Warblers (0.51) and Rosebreasted Grosbeaks (0.62) studied in forest fragments in the same region (Moore et al. 2010, Eng et al. 2009).Natal fragment size did not predict fledgling survival, but nestling body condition (just before fledging) had a strong effect on fledgling survival, as has been found in other studies of songbirds (summarized in Cox et al. 2014, Evans et al. 2020).As the highest mortality occurs during the first few days out of the nest, birds in better body condition are likely to be better able to withstand adverse weather conditions (Jones et al. 2017) and have more developed flight capabilities and thus lower their risk of predation (Vitz and Rodewald 2011).Ontario over, 2016Ontario over, -2018 (n = 133 fledglings) (n = 133 fledglings).Additive models for survival were run with forest fragment size (FS), nestling body condition (BC), fledge date (FD), sex, and year as predictor variables.Models are ranked by Akaike's Information Criterion (AIC) with small sample size adjustment (AIC c ), with number of parameters (df), log likelihood (LL), and model weight (w i ) given for each model.6).sites/territories for following spring (Nocera et al. 2006, Mitchell et al. 2010) or learning the landscape features for navigational purposes (Brown and Taylor 2015).Juvenile Wood Thrushes undertake repeated gap-crossing behaviors between forest patches to carry out this local dispersal, and thus could be exposed to higher predation risk.We found that 85% of independent juveniles were detected by Motus > 5 km from their natal site in the days and weeks prior to onset of fall migration (S.Hayes, personal observation).Nevertheless, pre-migration survival was very high (89%) in this severely fragmented landscape and natal fragment size was not an important predictor of premigration or annual survival.On average, ~52 days elapsed between fledgling independence and onset of fall migration, making the daily survival probability very high (DSR = ~0.998).This high survival rate pre-migration is surprising because juveniles are very inexperienced after the fledgling period and the landscape was highly fragmented requiring gap-crossing to disperse far.Adult Wood Thrushes routinely cross gaps during the breeding season, travelling several hundred meters between forest fragments (MacIntosh et al. 2011) and apparently naïve juveniles can do the same at low risk.Future studies could use intensive manual tracking to document the extent of natural gap-crossing behaviors in juveniles and conduct experimental translocation studies and release tagged juveniles into small fragments (Valente et al. 2019).Few studies, and none for forest songbirds to our knowledge, have been able to track fledglings to determine their timing of fall migration months later because of limitations in radio-tag battery life and manual tracking of juveniles after they have dispersed far from their natal territory (Evans et al. 2019).In our study, early-fledged (before 15 June) individuals departed earlier on fall migration, on average by 9 days, than late fledged individuals (late July/early August), but natal fragment size and sex did not predict fall migration departure (Figure 4).It is unknown whether early-fledged birds are better able to explore the landscape (e.g., prospecting for future breeding sites) or depart on migration in better condition than late-fledged birds.We found that fall migration departure time in juveniles does not predict migration/wintering survival.For adult Wood Thrushes, fall migration timing did not predict arrival time at wintering sites due to prolonged migratory stopovers en route (Stutchbury et al. 2011), and it is possible that juveniles have a similar slow migration strategy.In Belize, juvenile Wood Thrushes were not in poorer body condition prior to spring migration compared to adults (McKinnon et al. 2014) suggesting that juveniles were not occupying poorquality habitat.

Model
For both adults and juveniles, it is not known if fall migration timing influences mortality during migration itself, as opposed to after they arrive on the wintering grounds.There are many factors that affect migration survival however little is known as to where, or why, mortality happens during migration.For adult Black-throated Blue Warblers (Setophaga caerulescens), it has been estimated that most annual mortality (85%) happens during migration as opposed to the stationary breeding and wintering periods (Sillett and Holmes 2002) and this may also be true for independent juveniles given their inexperience with migration and stopover sites along the way.For instance, stopover habitat quality is highly variable in adult Wood Thrushes (Stanley et al. 2021) but it is not known if this affects migration survival.Very few studies have estimated juvenile survival at different life stages across the first full year of life for a songbird, as has been done for adult Black-throated Blue Warblers (Sillet and Holmes 2002) and Purple Martins (Tarof et al.2011).Our study on Wood Thrushes estimates that after fledging 55% of all mortality in the first year occurs during the migration and wintering period (Figure 6).The latter result is consistent with estimates of high adult mortality during the migration period in other Neotropical migrants (Sillett andHolmes 2002, Klaassen et al. 2013).
Even with manual, aerial, and Motus telemetry methods, only 16% of birds monitored after fledging, and 26% of birds that started fall migration, were detected returning in spring.Several returning individuals were detected by Motus towers 110 km from their natal forest fragment, but many (29%) were detected with aerial telemetry within the study area and presumably breeding.It is impossible to know how many others survived without being detected by the Motus network in Ontario or elsewhere.However, a stable isotope study of Wood Thrush populations found that only 10-15% of recruits or fewer were long-distance immigrants (>100 km from natal site; Rushing et al. 2021).Boyd et al. (2023) used Motus to estimate apparent annual adult Wood Thrush survival probability as 0.39, meaning that on average females must produce 0.39 female recruits per year for the population to be self-sustaining.Boyd et al. (2023) also quantified individual female Wood Thrush productivity in our study area using radio-telemetry (e.g., by finding all of a female's nests in a given season) and estimated this to average 1.88 female fledglings per season.Because apparent annual survival of fledglings is 0.16, this would yield only 0.30 female returning recruits per year, suggesting forest fragments in this region are population sinks for this species (see also Moore et al. 2010, Eng et al. 2009).Our estimates of adult survival, female productivity, and fledgling/juvenile survival are consistent with the decline in Wood Thrush populations that has been documented in this region (Sauer et al. 2019, Heide et al. 2023).Our findings of high reproductive success and high fledgling and pre-migration juvenile survival, combined with low recent breeding habitat loss in this region (Heide et al. 2023), suggests that the drivers of population decline could be associated with threats to juvenile migration and over-wintering survival.
We found no evidence for long-term carry-over effects of small natal forest fragment size at any of the three life stages for juvenile Wood Thrush survival or on fall migration timing, and fragment size also had little effect on nest survival.Variables associated with the natal territory (e.g., nestling body condition, timing of fall migration, fragment size, regional forest cover) did not predict metrics of survival even though small forest fragments are generally considered to be poor quality habitat for nesting songbirds (Falk et al. 2011, Richmond et al. 2012).A parallel study on adult Wood Thrushes using the same radio-tag models and forest fragments found adult return rate to the study area was 39% (Boyd et al. 2023) and no effect of breeding forest fragment size on female body condition or apparent adult annual survival.Wood Thrushes appear to thrive in small forest fragments in this region and so these supposedly marginal habitats can play an important role in conservation of this declining migratory songbird.Wood Thrush are relatively tolerant to low regional forest cover and fragmentation (Torrenta et al. 2022) and so negative effects of small forest fragments on could be important for more area-sensitive species.studies on forest passerines could conduct a similar radio-tracking study in a landscape that includes large contiguous forest (assuming the Motus infrastructure exists) to determine if there is a threshold at which large forest area does benefit juvenile survival/migration timing.

FIGURE 1 .
FIGURE 1. Map of study sites in Norfolk County on the north shore of Lake Erie, near Port Rowan Ontario (42.7131°N, 80.5372°W).Twenty-nine study sites (black hatched boxes) ranging from 11 ha to 445 ha were used over the course of the 3 field seasons (2016, 2017, and 2018).Inset map showing study area in the broader context of eastern North America with black dots representing active Motus Wildlife Network towers in 2018.

FIGURE 2 .
FIGURE 2. Nest age (calculated as 1st egg laid = day 0 on a 26-day cycle) model for daily nest survival for Wood Thrush across 29 forest fragments in southern Ontario.Dotted lines indicate 95% CIs and the vertical dashed line at day 14 represents the approximate date of eggs hatching.Daily survival probability decreased with increasing nest age but was not predicted by natal fragment size (Table1).

FIGURE 4 .
FIGURE 4. Relationship between autumn migration departure dates as detected by Motus automated telemetry and nest fledge dates for 82 (2016, n = 11; 2017, n = 29; 2018, n = 42) juvenile Wood Thrushes.Males represented with open triangles with dashed trend line and females with filled circles and solid trend line.Shading around trend lines representing 95% CIs.Departure date was strongly predicted by fledge date but not by natal fragment size (Table4).

FIGURE 5 .
FIGURE 5. Scatterplot of apparent annual survival for juvenile Wood Thrushes in relation to natal fragment size (n = 189).On y-axis "1" = survived, "0" = not detected.Females are represented with filled circles, males with open triangles.Points are jittered around the y-axis for clarity.Natal fragment size did not predict apparent annual survival of juveniles (Table6).

FIGURE 6 .
FIGURE 6. Apparent annual survival for 3-life stages during the first year of life of Wood Thrushes.Fledgling survival (to 16 days) probability was 0.70 as determined by both manual and automated radio telemetry detections.Pre-migration (0.89), migration and wintering (0.26), and annual survival (0.16) determined using Motus automated telemetry towers (see Supplementary Material 9 for sample sizes at each life stage).Typical migration route of adults tracked with geolocators shown on map with dashed lines(Stanley et al. 2015).Wood Thrush drawing by Roger Hall.

TABLE 1 .
Top models (∆AIC < 2) for daily nest survival rate (DSR) models for Wood Thrushes across 29 forest fragments (11 to 500 ha) in southernOntario, 2016Ontario,  -2018 (n = 404 nests) (n = 404 nests).Additive and interactive models with nest age (NA), distance to forest edge (DistFE), nest parasitism (BHCO), forest fragment size (FS), and % forest cover (FC) at 5 km as predictor variables (see Supplementary Material Table3for full model set with additional spatial scales).Models are ranked by Akaike's information criterion (AIC) with small sample size adjustment (AIC c ), with number of parameters (k), and model weight (w i ) given for each model.Model variable βeta coefficient estimates are given with 95% CI.
a AIC c value = 1097.30.life cycle for first-year Wood Thrushes

TABLE 2 .
Top weighted model set of time-dependent models for Wood Thrushes fledgling survival (S) across 29 forest fragments in southern Ontario over a 3

TABLE 3 .
. Environmental models for Wood Thrushes fledgling survival across 29 forest fragments in southernOntario over, 2016Ontario over,  -2018 (n = 189 fledglings) (n = 189 fledglings).Survival models used the base temporal model with nestling body condition (BC) included (S(Age) p(age3 + Year)) and additional covariates of forest fragment size (FS), % forest cover (FC) at 3 spatial scales.Models are ranked by Akaike's Information Criterion (AIC) with small sample size adjustment (AIC c ), with number of parameters (k), and model weight (w i ) given for each model.The full model set is shown in Supplementary Material Table6.

TABLE 4 .
General liner models for fall departure dates of juvenile Wood Thrushes across 29 forest fragments in southernOntario over, 2016Ontario over,  -2018  (n = 82 fledglings)  (n = 82 fledglings).Predictor variables included fledge date (FD), natal forest fragment size (FS), body condition at fledging (BC), sex, and year as additive variables.Models are ranked by Akaike's information criterion for small sample size adjustment (AIC c ), with degrees of freedom (df), log likelihood (LL), and model weight (w i ) given for each model.

TABLE 5 .
General liner models for pre-migration survival (16 days to fall migration departure) of juvenile Wood Thrushes across 29 forest fragments in southern

TABLE 6 .
General liner models for apparent migration survival (period after fall departure and spring arrival, n = 82) and annual survival (period from fledging the nest to spring return, n = 189 fledglings) of juvenile Wood Thrushes across 29 forest fragments in southern Ontario over, 2016-2018.Models for survival were run with forest fragment size (FS), body condition (BC), fledge date (FD) (for annual survival only), departure date (DD) (migration survival only), sex, and year as predictor variables.Models are ranked by Akaike's Information Criterion (AIC) with small sample size adjustment (AIC c ), with number of parameters (df), log likelihood (LL), and model weight (w i ) given for each model.
b AIC c = 170.0.S. M. Hayes et al. life for first-year Wood Thrushes 11 Survival: From Juvenile to Breeding Adult Our results provide the first estimate of juvenile pre-migration survival (89 %) for a forest songbird.Only two other studies we know of have estimated pre-migration survival in juvenile migratory songbirds, and both were on swallows (Tarof et al. 2011, Evans et al. 2019) which are not comparable because they can fly well at the time of fledging and likely face a different predator community.Upon independence, juvenile forest songbirds typically leave the natal territory and make local dispersal movements through the landscape prior to migration, which may be a prospecting behavior to assess breeding