Data for: Early life conditions influence fledging success and subsequent local recruitment rates in a declining migratory songbird, the Whinchat Saxicola rubetra

Life history traits and environmental conditions influence reproductive success in animals, and consequences of these can influence subsequent survival and recruitment into breeding populations. Understanding influences on demographic rates is required to determine the causes of decline. Migratory species experience spatially and temporally variable conditions across their annual cycle, making identifying where the factors influencing demographic rates operate challenging. Here, we use the Whinchat Saxicola rubetra as a model declining long-distance migrant bird. We analyse 10 years of data from 247 nesting attempts and 2519 post-fledging observations of 1193 uniquely marked nestlings to examine the influence of life history traits, habitat characteristics and weather on survival of young from the nestling stage to local recruitment into the natal population. We detected potential silver spoon effects where conditions during the breeding stage influence subsequent apparent local recruitment rates, with higher recruitment for fledglings from larger broods, and recruitment rate negatively related to rainfall that chicks experienced in-nest. Additionally, extreme temperatures experienced pre- and post-fledging increased fledging success and recruitment rate. However, we could not determine whether this was driven by temperature influencing mortality during the post-fledging period or later in the annual cycle. Brood size declined with hatching date. In-nest survival increased with brood size and was highest at local temperature extremes. Furthermore, nest survival was highest at nests surrounded with 40–60% vegetation cover of Bracken Pteridium aquilinum within 50m of the nest. Our results show that breeding phenology and environmental factors may influence fledging success and recruitment in songbird populations, with conditions experienced during the nestling stage influencing local recruitment rates in Whinchats (i.e. silver spoon effect). Recruitment rates are key drivers of songbird population dynamics. Our results help identify some of the likely breeding season mechanisms that could be important population drivers.Microsoft Excel R RstudioFunding provided by: Natural Environment Research CouncilCrossref Funder Registry ID: http://dx.doi.org/10.13039/501100000270Award Number: NE/S00713X/1Study area Our study was conducted between 2013–2022 at RSPB Geltsdale nature reserve in the North Pennines in Cumbria, UK (54.9°N −2.6°S), which is jointly owned by the Royal Society for the Protection of Birds and the Weir Trust. The survey area is an ~11km2 sub-section of the reserve comprising blanket bog, heathland and acid grassland, with an altitude of 220–440m. Study species and field methods Whinchats are short-lived (<8 years) Afro-Palearctic migrants, breeding in grassland habitats throughout Europe and Western Asia and migrating annually to sub-Saharan Africa for the northern winter. Whinchats are ground nesting and usually lay a single clutch of 4–7 eggs. The incubation period is 12–14 days with young provisioned by both parents for ~13 days before fledging. Young are capable of flight 3–5 days after fledging, with a further 9–15 days spent close to their natal nest while they are still dependent on their parents for food (Collar 2005, Tome & Denac 2012). Post-independence, fledglings typically remain in their natal area for 1–2 months, when they undergo a partial moult prior to southerly migration (Collar 2005). We began searching for nests when Whinchats arrived at Geltsdale in May, with searches performed almost daily until nesting had ceased in July. Nests were located by observing adult behaviour (male singing, nest building, guarding, and incubating). Males are typically more conspicuous than females during the breeding season, so the male of a pair was usually identified first, but once a nest was located females were also identified. We visited each nest every 3–7 days, recording clutch size, brood size and to confirm number fledged (see Table 1 for definitions). We estimated first egg laying date through back-calculations from either observation of incomplete clutches assuming one egg is laid daily, or for nests found post-hatching, by back-calculating based on chick development stage assuming 14 days incubation and a clutch size equal to brood size plus the number of unhatched eggs. All chicks were ringed with a unique combination of three colour rings and a numbered metal ring 6–8 days post-hatching. Fledging success was usually determined from resighting of fledglings. Fledging date was estimated from chick development stage observed from nest visits. For our analyses, we used the nest visit data to determine how many chicks successfully hatched and whether a nest successfully fledged at least one chick. Nest visits and bird handling were undertaken by field workers with ringing permits granted by the British Trust for Ornithology, and to minimise nest disturbance, no active nest was intrusively monitored on more than four occasions. Colour-marked Whinchat fledglings were monitored in the year of fledging until autumn departure and were then searched for in the adult breeding population in subsequent years. Searches were made almost daily from late April until early September in all years 2013–2022. Typically, multiple observers independently surveyed the whole study area almost daily from May – July of each year. Late in the season fledglings and adults congregated to moult in certain areas, and these hotspots were surveyed more frequently in August and September. Despite rigorous and frequent searches of the field site, some nesting attempts would inevitably have been missed. Because failed nests are active for a shorter period than successful nests, they are more likely to be missed, so a direct estimate of nest success rates from failed vs fledged nests may overestimate fledging success. To account for this, we performed a nest survival analysis (Mayfield 1975, Dinsmore et al. 2002), which estimated the probability of nest survival from the total number of days that each nest survived (i.e. exposure days). Using this approach, we estimated a nest survival rate of 73.3%, which is slightly lower than the direct estimate from our data (80.6%). However, this is unlikely to affect our assessment of the systematic factors influencing nest survival unless these factors likewise influenced the likelihood of observers finding a nest. Given that nests were usually found by locating calling adults we find it unlikely that any of the key factors we investigate (e.g. vegetation, weather) should affect the likelihood that a nest failed prior to being located. For further details on survival analysis see supplementary material. Vegetation and habitat sampling The vegetation substrate on which a nest was built and the vegetation within 5m, 50m and 200m radii of each nest were recorded between May – July of each year 2013–2014 and 2017–2019 after the nest had concluded. Radii boundaries were first marked, then observers measured the total area occupied by each individual vegetation type within this area, as follows: Bracken (Pteridium sp.), Tree scrub (e.g. Crataegus sp.), Tufted hair grass (Deschampsia caespitosa), Purple moor grass (Molinia caerulea), Rush (Juncus sp.), Bilberry etc. (Vaccinium myrtillus), Heather (Calluna vulgaris) and other grass (e.g. Holcus lanatus). These measures were then used to calculate the relative percentage cover by each vegetation type within each area. The number of trees and presence or absence of key features (e.g. fence post, wall) was also recorded within each radius; for the 5m radius the number of trees was counted exactly, whereas for 50m and 200m radii, the number of trees was estimated and binned. For further details on habitat sampling, see Table 1 and Table S1 (supplementary material). Additionally, the elevation at which each nest was located was recorded. Weather data To determine the effect of environmental conditions pre- and post-fledging we estimated relevant time windows for in-nest and post-fledging periods for each nest. The in-nest period, when chicks are flightless and dependent on parents for food, was defined as the period between the day of hatching and day of fledging. To account for variation of in-nest stage duration such as from the shorter periods of failed nests, this period was standardised as the median 13-days post-hatching in all cases (Fig. S1, supplementary material). The post-fledging period covers from when initially flightless chicks leave the nest until migratory dispersal. Upon fledging, Whinchat remain within 5 – 10m of their natal nest for 3 – 5 days when mortality risk is most acute (Tome & Denac 2012), then typically spend a further 10–12 days within 50–75m of their nest before increasing their range to >200m for the remainder of the post-fledging period (Tome & Denac 2012). To estimate dispersal date, an analysis of the final observation dates of fledglings at Geltsdale in the year of fledging indicated large drop-offs at 40- and 50-days post-fledging, with few observed after 50 days, suggesting most have either dispersed or died 50 days after fledging (Fig. S2). To minimise estimation error, we used two different period lengths, 40 days and 50 days after fledging; however, due to high correlation between these data, no final model included both post-fledging period lengths. Whilst our analysis cannot distinguish post-fledging mortality from mortality during subsequent stages of the annual cycle, we aimed to investigate links between post-fledging conditions and survival to recruitment as these conditions may impact recruitment directly via mortality during the post-fledging period, and indirectly by influencing subsequent survival in later stages of the annual cycle. We downloaded data on daily interpolated maximum and minimum temperature and total precipitation for the 5x5 km square encompassing Geltsdale (Eastings: 360000–365000; Northings: 555000–560000) for June-September each year from CEDA (Hollis et al. 2018). We then averaged the daily maximum and minimum temperature and total rainfall values for the two respective survival stage periods. Descriptions of all fixed effects (Table 1a), response variables (Table 1b) and random effects (Table 1c) are available in Table 1. Analysis All analyses were performed using R 4.0.2 (R Core Development Team, 2020). Weather data were extracted and analysed using the packages raster (Hijmans 2022), ncdf4 (Pierce 2021) and rgdal (Bivand, Keitt & Rowlingson 2022). Data were visualised using the ggplot2 (Wickham 2016) and cowplot (Wilke 2020) packages. All generalised linear mixed effects models (GLMMs) were built using the lme4 package (Bates et al. 2015) and analysed using lmerTest (Kuznetsova et al. 2017). Because vegetation data was sampled during a subset of years (2013 – 2014 & 2017 – 2019), but weather and life history trait data were available for all years (2013 – 2021), we performed two analyses for each response variable, one using all 9 years data without vegetation variables ('all years'), and another for the 5-year subset with all variables including vegetation ('vegetation years'). Vegetation sampling produced >40 individual variables, therefore, to avoid over-fitting we performed a two-stage modelling approach first used and validated, by Pearce-Higgins et al. (2009) (used also in: Dormann et al. 2013, Stanbury et al. 2022) to determine which explanatory variables to include in the final stage of analysis for 'vegetation years' models. First, a series of single-term GLMMs were run and the effect of each term on a given response variable was assessed individually, with explanatory variables P ≤ 0.10 considered for inclusion in subsequent models (see Tables S2 – 4, supplementary material). Secondly, two-term GLMMs were used to assess which continuous variables should be included as quadratic variables. Any quadratic variables with P ≤ 0.10 were considered for inclusion in subsequent models as both quadratic and linear variables. Finally, these variables were checked for correlations, and where two variables were highly correlated with each other (Spearman rank correlation coefficient, |rs| ≥ 0.70, Dormann et al. 2013) the more significant predictor (i.e. the smaller P-value) was included in the full model (Tables S5 – 6). This process was repeated for each response variable. For the analysis of data from all years, all weather and life history variables were included in the full model as both linear and quadratic terms (Tables S7 – 9) unless highly correlated (|rs| ≥ 0.70) with a more significant term (Tables S10–11). Once the number of explanatory variables was reduced, we fitted GLMMs by performing stepwise elimination by iteratively removing the term with the largest P-value and then refitted the model until the Minimum Adequate Model (MAM) remained (all variables P < 0.05). In total, we ran GLMMs for four response variables with 'Year' as a random effect for both the full sample of all years and for the vegetation years, all using the reduced set of predictors as fixed effects and including year as a random effect. Full details of all model structures, their response variables, fixed effects (full model and MAM), error distributions and sample sizes are available in Table 2, full model outputs are available in Tables S12–17