Warblers

The demographic history of the Seychelles warbler is outlined in Fig.1. The species was first described in 1878 by Oustalet (1878 ) from the island of Marianne (96 ha), and in the same account was said by Lantz to be ‘rare on Ile Cousine’. Subsequent studies found the warbler on Cousin, but not Cousine, and Lantz's account was presumed to be a mistake (Vesey-Fitzgerald 1940 ). By 1938, the warbler was extinct on Marianne, and Vesey-Fitzgerald (1940 ) remarked that it ‘must be the rarest [bird] in the world’. Expeditions to Cousin in 1959, 1965, 1967 and 1968 documented 30, 50, 26 and 50 individuals, respectively (Penny 1967 ; Loustau-Lalanne 1968 ). However, birds were not uniquely ringed during these trips, so these estimates of population size are unlikely to have been very precise. In 1967, Cousin was designated as a nature reserve, and efforts began to increase the populations of native bird species (Penny 1967 ). Habitat restoration, consisting of the removal of coconut palms (Cocos nucifera) to allow the succession of natural pisona (Pisonia grandis) woodland, was successful, and the Cousin warbler population quickly recovered; since the 1980s, it has been at a carrying capacity of approximately 320 adults (Brouwer et al. 2009 (link)). Between 1987 and 2011, four new warbler populations were successfully established by translocation to the islands of Aride, Cousine, Denis and Frégate (Komdeur 1994 ; Richardson et al. 2006 ; Wright et al. 2014 (link)).
Historical samples were obtained from all known Seychelles warbler museum specimens, collected from Cousin (n = 19) and Marianne (n = 7) in 1876–1940 (Table S1). Although the temporal range of sampling of the museum specimens was wide, structure analyses suggested that they grouped into two populations (see Results), enabling us to group them for population genetic analyses. A small (approximately 1.5 × 1.5 × 3.0 mm) piece of skin was excised from the ventral surface of the foot and stored at room temperature in a sterile microfuge tube. Contemporary samples were collected as part of an intensive, long-term study of Seychelles warblers on Cousin Island (Brouwer et al. 2010 (link)). Since 1988, the entire population has been extensively monitored, often in both the main (June–September) and minor (November–March) breeding seasons each year, during which birds are routinely caught with mist nets and audio lures. A blood sample (approximately 25 μL) was collected from each bird by brachial venipuncture and stored at room temperature in a screw-topped microfuge tube containing 1.5 mL absolute ethanol. Each bird was fitted with a unique combination of three colour rings and a metal British Trust for Ornithology (BTO) ring. Over 96% of adult birds on Cousin have been ringed since 1997 (Richardson et al. 2001 (link)), and a representative sampling of the population was achieved in each year. For the present analysis, 50 samples were randomly chosen from 1997 and 2011 (of 160 and 197 samples available from that year, respectively) to provide two temporally distinct contemporary population samples for comparison with the historical data.

The Seychelles warbler population on the isolated island of Cousin (29 ha; 4°20’ S, 55°40’ E) contains ca 320 adult individuals, nearly all of which are colour-banded (using a combination of three colour rings and a British Trust for Ornithology metal ring)^{54 (link)}. The warbler’s life history is characterized by high annual adult survival (84%), mostly single-egg clutches, and extended periods (up to three months) of post-fledging care^{24 (link),32 (link)}. Individuals that have acquired a dominant breeding position generally defend the same territory, with the same partner, until their death^{55 (link)}. The correlation between the age of the dominant male and female in a territory is, while significant, actually relatively weak (Pearson product-moment correlation: r = 0.16, t_1531 =6.53, P < 0.001, Supplementary Fig. 3). This is because the age at which an individual obtains a dominant position varies considerably, pairs of birds do not become dominant at the same age, and the age at which dominant individuals die (and one of the pair is replaced) varies. Previous studies have shown that male and female dominants have similar breeding tenure, annual survival probabilities and rates of actuarial senescence^{24 (link),25 (link)}. The vast majority of breeding activity occurs in June–September (hereafter: main breeding season), when food availability is highest (breeding occurs in 94% of territories in this period)^{56 (link)}. Seychelles warblers can breed successfully in socially monogamous pairs, but, because of a lack of suitable breeding opportunities, young individuals often delay independent breeding and become subordinates within a territory, where they then may help with providing alloparental care (incubation (female subordinates only); provisioning (male and female subordinates)), or not^{54 (link)}. Subordinates are often retained offspring from previous breeding attempts^{33 (link)}, although a very small number of subordinates disperse to a new subordinate position in a different territory⁵⁷ . Territory inheritance in the Seychelles warbler is rare (only 3.7% of dominant breeding positions are obtained via offspring inheriting this status on their natal territory^{58 (link)}), so it is unlikely that inheritance is the main benefit accrued by subordinates. Subordinates benefit from helping as they obtain breeding experience^{59 (link)} and often gain indirect (kin-selected) fitness benefits through helping related offspring^{46 (link)}. Further, older (≥2 year old) female subordinates often (ca 40% in any year) gain direct fitness benefits through co-breeding (laying an egg in the same nest as the dominant female)^{22 (link),35 (link)}. Co-breeding subordinates always provide alloparental care and do not discriminate between their own or the dominant female’s offspring (i.e. they help all offspring in the nest)^{46 (link),60 (link)}. Further, previous studies found no evidence for reproductive conflict caused by co-breeding females^{35 (link),40 ,61 (link),62 (link)}, except in extreme cases^{32 (link)}. Therefore, we considered all subordinates that helped with incubation or provisioning as helpers, irrespective of whether they co-bred or not. Male subordinates acquire fewer benefits than females because they do not appear to benefit through gaining breeding experience^{59 (link)} and very rarely gain direct paternity, which may explain why most helpers (88%;^{36 (link)} 77% (n = 310) in this study) are female^{35 (link)}. Apart from providing the opportunity to obtain indirect fitness benefits, the prolonged presence of the parents may be beneficial for subordinates because it facilitates the eventual acquisition of a dominant position elsewhere. This is because breeders are more likely to allow related subordinates to remain in the territory until they are able to disperse to a dominant position elsewhere, but will evict unrelated subordinates irrespective of such opportunities, resulting in higher mortality^{33 (link),63 (link)}.

We fitted male willow warblers with Intigeo–W30 geolocators (Migrate Technology LTD, 0.3 g, c. seven months capacity) in East Denmark (55.61°N, 12.57°E; catching range 500 m) from May to mid-June (n = 17 in 2014, n = 20 in 2015) using leg-loop harnesses [39 (link)] made of 1 mm braided nylon cord. 17 birds were recaptured the year after tagging (n = 11 in 2015, n = 6 in 2016). Two loggers from 2014 contained no data.
Positions were estimated using the GeoLight package [40 (link)] in R [41 ]. A threshold of 3 lx was used and sun elevation angles between −3° and 0° provided the best fit using Hill-Ekström calibration [38 (link)] (breeding area calibration produced similar spatiotemporal patterns, Additional file 1: Appendix S2-S4).
Periods of no overall change in longitude for ≥5 days were considered staging. We excluded latitude from positions within ten days of equinox. Position outliers >10° from median longitude/latitude at each staging site were excluded (Additional file 1: Appendix S1).
Normalized Difference Vegetation Index (NDVI) was used to estimate vegetation conditions [42 (link)]. NDVI was obtained from the MODIS satellite product MOD13C1 [43 ]. Mean NDVI within a radius of 50 km for each wintering site were extracted with the adehabitat R package [44 (link)].
Data were pooled for all analyses because t-tests revealed no differences between the two years in average latitude (p = 0.44), longitude (p = 0.79) or NDVI (p = 0.23).
The western ‘detour’ between the staging sites before and after the Sahara coincided with Equinox. We estimated average westernmost latitude projecting from the mean position of the last European staging sites assuming a speed of 300 km/day (daily migration speed of willow warblers ringed in Denmark [45 ]).
Longitudinal spread of birds in winter was estimated using the loxodromic distance between longitude of the centre of mass for all individuals and the latitude and longitude of the centre of mass for each individual in five-day intervals in R using SDMTools [46 ] and geosphere [47 ].
We correlated arrival date, body mass, wing length and NDVI with longitude to evaluate causes of winter spread using Pearson’s r (Note that a weak relationship with arrival date is expected because of extra travel time). We tested for consistent north-south or east-west directional changes and direction of change in NDVI between consecutive winter sites using Sign tests. Lastly, we investigated trends over time in NDVI within sites using Pearson’s r. Potential effects of variation in longitudinal distribution of NDVI in earlier years on termination of migration were investigated by correlating site-specific NDVI among the last three winters before capture.

Climatic selection and constraint on the evolution of the desired phenotype may imply several scenarios, depending on which periods are most critical for thermoregulatory performance⁶² . Some scenarios assume that phenotype is shaped by extreme temperature events (e.g. either the warmest or the coldest days or months), which cause severe mortality of organisms that can easily overheat or overcool during these critical timeframes^{16 (link),18 ,45 (link),62} . Alternative scenarios assume that the phenotype is selected by the average temperature across year, as animals spend less time on cooling or heating, and thereby performs better in foraging or reproduction. We therefore retrieved both average, upper and lower monthly temperatures measured within species ranges to test our hypotheses under these alternative scenarios.
We obtained temperature data for each species from spatial analyses within the ‘sf’ (version 1.0-8)⁶³ and ‘raster’ (version 3.5-15)⁶⁴ R packages, using global raster layers of temperatures available in World Clim database (version 2.1)³⁴ . These rasters (Tmin, Tavg and Tmax; see below) had a resolution of 30” and were consist of monthly averages from a period of 58 years (1960-2018). The temperature metrics have been calculated within polygons of species ranges available in form of multi-polygon vector layers extracted from the BirdLife International database (version 2020.1)³⁵ .
We first excluded polygons identified as uncertain species presence, uncertain season of presence, non-native presence or species extinct in a region, leaving us with 9962 species (out of 9,993 species) with complete geographic data. Second, having polygons with only a certain, native and extant species presence we grouped them by the species (according to the phylogenetic taxonomy^{32 (link)}) and the season of presence (either breeding season, winter or year-round presence) and then we aggregated them to obtain single polygons specific to species and season (Supplementary Fig. 19). Third, using breeding and year-round species ranges, where species live at hotter period of the year; we calculated their zonal means of monthly temperature maximums (Tmax) and took the largest monthly value for each species (maximum temperature of all months). We also calculated their zonal means of monthly temperature averages (Tavg) and took the largest monthly value for each species (average temperature of hottest month). Fourth, we analogously used winter and year-round species ranges, where species live at colder period of the year. Then, we calculated their zonal means of monthly temperature minimum (Tmin) and took the lowest monthly value for each species (minimum temperature of all months). We also calculated their zonal means of monthly temperature averages (Tavg) and took the lowest monthly value for each species (average temperature of coldest month). Fifth, we took all (breeding, winter and year-round) species ranges and we calculated their zonal means of monthly temperature averages (Tavg) and averaged all monthly values to obtain average temperature of all months for each species. We also retrieved absolute latitude from the centroids of the above species ranges (breeding, winter and year-round, summarized to a single polygon per species), which described a simple geographic variation across species. Sixth, the obtained temperature measures (minimum temperature of all months, average temperature of coldest month, average temperature of all months, average temperature of hottest month and maximum temperature of all months) were used in models predicting the phenotype. Where we used these measures as response variables when predicting the temperature within species range (i.e. the environmental temperature to which the species is adapted), we transformed variables with two different formulas to normalize left-shewed distribution (Supplementary Fig. 20).
The temperature measures reflected the full range of global thermal environments occupied by birds. For example, the maximum temperature of all months ranged from −3.8 °C (in the emperor penguin Aptenodytes forsteri), through 29.9 °C (median, in the Minas gerais tyrannulet Phylloscartes roquettei) to 43.8 °C (in the Basra reed-warbler Acrocephalus griseldis). In contrast, the minimum temperature of all months ranged from −35.3 °C (in black-billed capercaillie Tetrao urogalloides), through 14.1 °C (median, e.g. in Yellow-breasted apalis Apalis flavida) to 24.8 °C (in the Seychelles warbler Acrocephalus sechellensis). Notably, our multiple measures of temperature indicated distinct aspects of seasonality in thermal conditions that may require different phenotypic adaptations across avian lineages.

In this study, we analyzed faecal microbiota, which has been shown to be a good proxy for avian GM (Videvall et al., 2018 (link); Berlow et al., 2020 (link)), with sample collection and storage as described in Kropáčková et al. (2017) (link). Faecal samples of temperate passerines (405 samples from 52 species), were obtained during the 2014 breeding season (April–July) at various sampling sites in the Czechia (Supplementary Table 1 and Supplementary Figure 1) and were previously presented in Kropáčková et al. (2017) (link). Tropical species (205 samples from 47 species) were sampled in upland forest habitats in Cameroon (Mount Cameroon; approximately 4°07′N, 9°04′E; Supplementary Table 1 and Supplementary Figure 1). Tropical samples were collected during both the rainy (September 2014 at 950 and 1 100 meters above mean sea level) and dry seasons (November/December 2014 at 650 and 2,280 meters above mean sea level), the latter corresponding to the breeding season for most tropical passerines included in our dataset. Furthermore, the migration and wintering period of temperate trans-Saharan migrants also largely overlaps with the dry season in Cameroon. During the dry season, we also collected 25 samples from two temperate-breeding trans-Saharan migrant species, the garden warbler (Sylvia borin) and willow warbler (Phylloscopus trochilus).

Using systematically collected nest content and location data on warbler and blackbird nests gathered from May through July across three field seasons (2019–2021) in central Illinois, USA, we analyzed whether blackbirds that placed nests closer to actively breeding warbler or blackbird pairs experienced lower probabilities of brood parasitism. Detailed study site descriptions, nest searching/monitoring methodology, and how we determined warbler breeding status are described in detail in Lawson et al. (2020 (link)); Lawson, Enos, Mendes, et al. (2021 ). Briefly, we searched six sites in Champaign and Vermilion County two to three times a week for blackbird and warbler nests. For each nest, we noted parasitism status (parasitized vs. nonparasitized, binary status) and took location points for nest sites using GPS units to 3 m accuracy (Garmin model eTrex 10).
For breeding warbler pairs whose nests we could not find, we took a location point for one of the male's singing posts within the territory and used these points as proxies for active warbler nest locations. We determined breeding status by spot‐mapping pairs several times a week and noting behavioral cues strongly associated with an active nest present on territory (see Lawson, Enos, Mendes, et al., 2021 for a detailed description of warbler breeding status methods). For blackbirds, we only used known nests as location points, as we needed to include the parasitism status for each nest in our analyses.
We used ArcGIS (ver. 10.8.1; Redlands, 2022 ) to calculate the distance (m) between each blackbird nest and the closest warbler and blackbird nest or breeding pair location, within each year of the study. For analyses, we only used data from sites with both warbler and blackbird nests within a given year (2019, n = 6 sites; 2020, n = 4; 2021, n = 5). Blackbirds are facultatively polygynous and can have multiple active breeding nests in their territory at a time (Searcy & Yasukawa, 2014 ), which introduces biological pseudoreplication as an issue. Thus, we focused on whether the male blackbird's territory experienced any parasitism in any of his nests, by assigning male territories as “non‐parasitized” if no nests were parasitized or “parasitized” if at least one nest was parasitized. Distance to nearest warbler nest/pair or blackbird nest was averaged for all nests on a male blackbird's territory to produce a single distance metric to warbler and to blackbird nest per blackbird territory. Distances between blackbird nests and nearest warbler nest/pair ranged from 1.69 m to 2581.0 m, with most data points (>80%) falling linearly under the 300 m range. Therefore, for statistical purposes, we only included warbler nests and blackbird territories within 300 m of each other to minimize skewness of data and to increase model fit.
We ran a binary logistic regression model to test if distance to nearest warbler nest/pair significantly predicted the probability of brood parasitism on blackbird territories, with distance to the nearest nest as a fixed effect. We did not include year and/or site as fixed or random effects due to small sample sizes. No other variables were included in the model such as the blackbird or warbler nests' breeding stage (sensu Massoni & Reboreda, 2001 ). We also ran a second binary logistic regression model, with averaged distance to the nearest blackbird nest as a fixed effect instead of warbler nests. We ran all the analyses using R version 4.2.2 and the lme4 package (Bates et al., 2014 ) and set α = 0.05.