Study area
We conducted the study in the best-preserved stands of the Białowieża Forest, strictly protected within the Białowieża National Park (hereafter BNP; coordinates of Białowieża village: 52°42′N, 23°52′E). The extensive Białowieża Forest (c. 1500 km2) straddles the Polish-Belarusian border, where the climate is subcontinental with annual mean temperatures during May–July of 13–18 °C, and mean annual precipitation of 426–940 mm66,67.
The forest provides a unique opportunity to observe animals under conditions that likely prevailed across European lowlands before widespread deforestation and forest exploitation by humans66,68,69. The stands have retained a primeval character distinguished by a multi-layered structure, frequent fallen and standing dead trees, and a high species richness66,70. The stands are composed of about a dozen tree species of various ages, up to several hundred years old. The interspecific interactions and natural processes have been little affected by direct human activity.
We conducted observations mostly within the three permanent study plots (MS, N, W), totalling c. 130 ha, and in other nearby fragments of primeval oak-lime-hornbeam Tilio-Carpinetum or mixed deciduous-coniferous Pino-Quercetum stands. However, a small number of observations from adjacent managed deciduous forest stands were also included. For details of the study area see71,72,73.
Study species
Our study system focused on ground-nesting Wood Warblers Phylloscopus sibilatrix, blowflies Protocalliphora azurea, and Myrmica or Lasius ants, which occurred in the birds’ nests.
The Wood Warbler is a small (c. 10 g) insectivorous songbird that winters in equatorial Africa and breeds in temperate European forests, typically rearing one or two broods each year74. Wood Warblers build dome-shaped nests for each breeding attempt, composed of woven grass, leaves and moss, and lined with animal hair73. The nests are situated on the ground among moderately sparse vegetation, often under a tussock of vegetation or near a fallen tree-branch or log (see examples in Supplementary Fig. S2)53,75. The breeding season of Wood Warblers begins in late April–early May and ends in July–August, when nestlings from replacement clutches (after initial loss) or second broods leave the nest. The typical clutch size in BNP is 5–7 eggs, and the nestling stage lasts 12–13 days74,76.
Wood Warbler nests are inhabited by various arthropods, including Myrmica ruginodis or M. rubra ants, and less often Lasius platythorax, L. niger or L. brunneus. The ants foraged and/or raised their own broods within the Wood Warbler nests52. The Myrmica and Lasius ant species are common in Europe77,78. Their colonies contain from tens to thousands of workers, and can be found on the forest floor, e.g. in soil, within or under fallen dead wood, in patches of moss, or among fallen tree-leaves53,77,78. All of the ant species found in the Wood Warbler nests are predators of other arthropods77,79,80.
Blowflies, Protocalliphora spp., are obligatory blood-sucking (hematophagous) ectoparasites that reproduce within bird nests. The occurrence, abundance, and impact of blowflies on Wood Warbler offspring is largely unknown, similar to many other European songbirds that build dome-shaped nests. Adult blowflies emerge in late spring and summer to lay eggs on the birds’ nesting material or directly onto the skin of typically newly hatched nestlings14,26. The blowfly larvae hatch within two–three days, and develop in the structure of warm bird nests for another 6–15 days, during which they emerge intermittently to feed on host blood, before finally pupating within the nests14,25,26,27.
Data collection
Nest monitoring and measurements of nestlings
We searched for Wood Warbler nests daily from late April until mid-July in 2018–2020, by following birds mainly during nest-building. Nests were assigned to a deciduous or mixed deciduous-coniferous habitat type, depending on the tree stand where they were found. We inspected nests systematically, according to the protocol described in Wesołowski and Maziarz76. The number of observer visits was kept to a minimum to reduce disruptions for birds or potential risks of nest predation.
We aimed to establish the dates of hatching (day 0 ± 1 day), nestlings vacating the nest (fledging; ± 1 day) or nest failure (± 1–2 days). When nestlings hatched asynchronously, the hatching date corresponded to the earliest record of nestling hatching. The dates of fledging or nest failure were the mid-dates between the last visit when the nestlings were present in the nest, and the following visit, when the nest was found empty. Nest failure was primarily due to predation, which is the main cause of the Wood Warbler nest losses in BNP76,81 and elsewhere in Europe82,83.
To assess fitness consequences for birds of variable weather conditions, blowfly abundance and/or ant presence, we measured nestling growth and determined brood reduction (i.e. the mortality of chicks in the nest) from hatching until fledging. To define brood reduction, we assessed the number of hatchlings (nestlings up to 4 days old) and the number of fledglings leaving the nests. To ensure accurate counting and avoid premature fledging of nestlings, we established the number of fledglings on the day of measurement, when all nestlings were temporarily extracted from the nest.
We measured nestling growth on a single occasion when they were 6–9 days old (median 8 days), almost fully developed but too young to leave the nest. The measurements lasted for less than 10–15 min at each nest to minimise any potential risk of attracting predators. For each nestling we measured (using a ruler) the emerged length of the longest (3rd) primary feather vane (± 0.5 mm) on the left wing84,85, and body mass to the nearest 0.1 g using an electronic balance. The length of the feather vane is closely linked to feather growth86 and is one of the characteristics of nestling growth85,87. We treated the length of the primary feather vane and body mass as indices of nestling growth rate under varying conditions of weather, blood-sucking ectoparasites, or ant presence.
Extraction of arthropods from bird nests
To assess the number of blowflies and to establish the presence of ants, we checked the contents of 129 nests (including 11 nests from the managed forest stands) at which Wood Warbler nestlings had been measured. The sample included 86 successful breeding attempts (where a minimum of one nestling successfully left the nest), 27 failed (predated) nests (remnants of nestlings were found, but the nest structure remained intact), and 16 nests with an unknown fate (nestlings were large, so were capable of leaving the nest, but no family were located or other signs indicating fledging).
Due to ethical reasons, we were unable to collect the Wood Warbler nests and extract the ectoparasites and ants from them while they were in use by the birds. Removing the nests and replacing them with dummy nests would cause unacceptable nest desertion by adults. Therefore, we assessed the occurrence and number of blowflies or ant presence after Wood Warbler nestlings fledged or the breeding attempts failed naturally. We retrospectively explored the changes in blowfly infestation14, including the effect of ant presence53 in the same nests.
We collected nests from the field as soon as a breeding attempt ended, within approximately five days (median 1 day) following fledging or nest failure (nest structure remained intact). The delay of nest collection would not bias the ectoparasite infestation, as blowfly larvae pupate within bird nests and stay there after the hosts abandon their nests; puparia can be still found in nests collected in autumn or winter14. As the likelihood of finding ant broods (larvae or pupae associated with workers) was rather stable with the delay of nest collection53, the method seemed reliable also for assessing the presence of ant broods (35 of all 71 Wood Warbler nests containing ants). Only the number of nests with lone foraging ant workers could be underestimated, potentially inflating the uncertainty of tested relationships. However, as ants usually re-use rich food resources88, foraging Myrmica or Lasius ant workers might regularly exploit warbler nests, increasing the chances of finding the insects in the collected nests.
Wood Warbler nests were collected in one piece, with each placed into a separate sealed and labelled plastic bag. We carefully inspected the leaf litter around the nests, and the soil surface under them, to make sure that all blowfly larvae or pupae were collected. We transported the collected nests to a laboratory, where we stored them in a fridge for up to 5–6 days before the arthropod extraction.
To establish the number of blowflies and the presence of ants, in 2018, we carefully pulled apart the nesting material and searched for the arthropods amongst it 52. We gathered all blowfly pupae or larvae and a sample of ant specimens into separate tubes, labelled and filled with 70–80% alcohol, for later species identification. For nests collected in 2019–2020, we extracted the arthropods with a Berlese-Tullgren funnel. During the extraction, which usually lasted for 72 h, each nest was covered with fine metal mesh and placed c. 15 cm under the heat of a 40 W electric lamp. The arthropods were caught in 100 ml plastic bottles containing 30 ml of 70–80% ethanol, installed under each funnel. After the arthropod extraction, we carefully inspected the nesting material in the same way as in 2018, to collect any blowflies that remained within the nests. The quality of information collected on the number of ectoparasites and ant presence should be comparable each year.
Weather data
We obtained the mean daily temperatures and rainfall sums from a meteorological station, operated by the Meteorology and Water Management National Research Institute in the Białowieża village, 1–7 km from the study areas.
Data analyses
Weather conditions affecting blowfly ectoparasites
To explore the impact of weather on blowfly ectoparasites, for each Wood Warbler nest we calculated average temperatures from daily means, and total sums of rainfall from daily sums, for the two time-windows in which we assumed the impact of weather would be of greatest importance:
- i.
the early nestling stage, when Wood Warbler nestlings were 1–4 days old. During this stage, female blowflies require a minimum temperature of c. 16 °C to become active and oviposit in bird nests27. Thus, cool and wet weather in the early nestling stage should reduce the activity of ovipositing blowflies, leading to less frequent ectoparasite infestation of Wood Warbler nests.
- ii.
The late nestling stage, when the warbler nestlings were aged between over four days old and until fledging or nest failure. During this stage, blowfly larvae grow and develop in bird nests after hatching a few days after oviposition14,25,26,27. As the temperature of bird nests strongly depends on ambient temperatures21, mortality of blowfly larvae should increase in cool weather, resulting in fewer ectoparasites in nests collected shortly after the fledging of birds29.
Weather conditions affecting Wood Warbler nestling growth
To explore the impact of weather on nestling growth, for each nest we calculated the average temperatures and total sums of rainfall for the period when nestlings were over four days old and until their measurement, usually on day 8 from hatching (see above). During this stage, nestlings are no longer brooded by a parent74, so must balance their energetic expenditure between growth (feather length and body mass) or thermoregulation89. Thus, we expected that the gain in body mass and the growth of flight feathers would be reduced in nestlings during cool and wet weather, when maintaining a stable body temperature would be costly90.
Statistical analyses
All statistical tests were two-tailed and performed in R version 4.1.091.
The changes in blowfly infestation of the Wood Warbler nests
To test the changes in blowfly infestation of warbler nests, we used zero-augmented negative binomial models (package pscl in R;92,93), which deal with the problem of overdispersion and excess of zeros92. In this study, hurdle and zero-inflated models fitted with the same covariates had an almost identical Akaike Information Criterion (AIC). Therefore, we presented only the results of hurdle models, which are easier to interpret than zero-inflated models. Hurdle models consisted of two parts: a left-truncated count with a negative binomial distribution representing the number of blowflies in infested nests, and a zero hurdle binomial estimating the probability of blowfly presence. We used models with a negative binomial distribution, which had a much lower AIC than with a Poisson distribution on a count part.
We designed the most complex (global) model that contained a response variable of the number of blowflies in each of the 129 Wood Warbler nests. The covariates were: mean ambient temperature, total sum of rainfall, presence (or absence) of ants in the same nests, habitat type (deciduous vs mixed deciduous-coniferous forest), study year (2018–2020), the number of nestlings hatched (brood size), and nest phenology (the relative hatching date of Wood Warbler nestlings, as days from the median hatching date in a season: 23 May in 2018, 25 May in 2019 and 29 May in 2020). The initial global model also contained the two-way interaction terms that we suspected to be important: between temperature and rainfall, temperature and presence of ants, and rainfall and presence of ants.
To explore all potentially meaningful subsets of models, we used the same covariates on both parts (count and binomial) of the global model. We performed automated model selection with the MuMIn package94, starting from the most complex (global) model and using all possible simpler models (i.e. all subsets)95. To attain the minimum sample size of c. 20 data points for each parameter96, we limited the maximum number of parameters to six in each part (count or binomial) of the candidate models.
As some of the interaction terms appeared insignificant in the initial model selection, to minimise the risk of over-parametrisation, we included only the significant interaction term on a count part of the final global model. As described above, we performed model selection again. We tested linear relationships, as the quadratic effects of weather variables (presuming temperature or rainfall optima) appeared insignificant.
To test whether blowfly infestation changed with weather in the early or late nestling stages, we twice repeated the procedure described above. The first global model included the mean ambient temperature and the total sum of rainfall for the early nestling stage, and the second global model contained weather variables for the late nestling stage. The remaining covariates were the same.
A practice of including the same sets of covariates on count and binomial parts has been previously questioned97. However, our approach allowed us to comply with these objections97, as we presented only the most parsimonious models (with ΔAICc < 2) which all contained different sets of covariates on count and binomial parts.
We considered models differing by less than two AICc units from the best fitting model (with smallest AICc) as equally informative98. We averaged the top models, always with different sets of covariates on count and binomial parts, by using conditional averaging, to obtain parameter estimates and their 95% CIs reflecting model selection uncertainty.
The analyses comprised successful (n = 86) and other nests that failed in the late nestling stage (nest structure remained intact; n = 27) or their fate was unknown (nestlings were large enough to escape a predator attack; n = 16). Including failed nests or those with the unknown status increased sample sizes, and was justified because the same patterns emerged when only the successful breeding attempts were analysed.
We calculated bootstrapped means and 95% CIs (10,000 random re-samplings), to present the number of blowflies in infested nests with and without ants, or in different years.
The growth of Wood Warbler nestlings
To explore the changes in the growth of Wood Warbler nestlings with varying weather conditions, blowfly abundance, ant presence, or other factors (relative hatching date, brood size, nestling age, habitat type, study year), we used linear mixed effect models, with package glmmTMB99. In the analysis we employed the data from 2018 to 2020 for the length of the third primary feather and body mass of 742 nestlings from 129 Wood Warbler nests. The models contained a random intercept for nest identity that accounted for non-independent growth rates of nestlings from the same nests.
We modelled the length of the third primary feather or body mass, both assessed c. 8 days post-hatching, as responses to nine predictors: mean ambient temperature and total sum of rainfall in the stage when nestlings were over four days old and until their measurement, blowfly abundance (i.e., the number of blowflies per nestling in nests with and without blowflies), presence or absence of ant workers, and also habitat type (deciduous vs mixed deciduous-coniferous), study year, relative hatching date of birds, brood size, and nestling age (number of days since hatching on day 0). In models, we included nestling age to correct for deviations from the intended measurement age at 8 days post-hatching.
For both measures of nestling growth (feather growth and body mass), global models initially contained main effects of all covariates mentioned above and the potentially important two-way interaction terms between temperature and rainfall, temperature and blowfly abundance, rainfall and blowfly abundance, and temperature and ant presence. However, as none of the interaction terms appeared significant, we tested only the main effects. All models included only linear relationships, as none of the quadratic effects (presuming temperature or rainfall optima) were significant.
To create a subset of models with variables best explaining the growth of Wood Warbler nestlings, we used an automated model selection in MuMIn94, according to an information-theoretic approach. This method allowed us to find the most parsimonious models among the candidate model set, which were drawn from the most complex (global) model98. To attain the minimum sample size of c. 20 data points for each estimated parameter96, we limited the maximum number of parameters to six in the candidate models. We considered models differing by less than two AICc units from the best fitting model (with smallest AICc) as equally informative98. We averaged these top models (using conditional averaging) to obtain parameter estimates and their 95% CIs reflecting model selection uncertainty.
Mortality of Wood Warbler nestlings
To test whether the nest infestation with blowflies increased the mortality of Wood Warbler nestlings, we compared brood reduction (i.e. the difference between the number of fledglings and hatchlings) in 86 successful nests with and without ectoparasites, based on bootstrapped means and 95% CIs (10,000 random resamplings).
Source: Ecology - nature.com
