Drainage of wetlands and agricultural intensification has resulted in serious biodiversity loss in Europe, not least in grasslands. Consequently, many meadow birds have drastically declined, and the habitats they select for breeding currently rely on land management. However, the selection of habitats maintained by agriculture may contribute to reduced fitness and thus remain maladaptive for individuals, which makes conservation challenging. An understanding of the relationships between species’ habitat selection, food supply and land management in the context of species’ behaviour is therefore crucial for conservation. Lowland populations of Great Snipe Gallinago media are currently declining at a moderate rate, causing a conservation concern. We examined the daytime site selection (assumed as foraging sites) and food supply of radiotracked Great Snipe males breeding on a floodplain in NE Poland. Foraging sites were classified at micro- and macro-scale levels using the logistic regression in a use–availability design. On the microscale level, males selected moderate sward height and density, and a large amount of bare ground patches, and the importance of these increased as the breeding season progressed. On the macro-scale level, these conditions were associated with (1) meadows mown twice per season and grazed thereafter (associated with the most abundant food resources – earthworms) and (2) extensively managed pastures, suggesting the importance of grazing. Abandoned or late-mown meadows under agri-environmental schemes (AES) were avoided by foraging males. However, parcels with delayed mowing offer safe breeding sites for females nesting close to leks, unlike land-use types preferred by foraging males, which may act as an ecological trap. Effective conservation of Great Snipes on floodplain meadows requires precisely targeted AES schemes that will provide a mosaic of intensive and extensive land-use patches in the vicinity of identified leks.

General study design

To examine habitat selection of the Great Snipe, we applied a use–availability design in two spatial scales (a micro-scale model and macro-scale model), where environmental characteristics at the male Great Snipe daytime presence sites (‘used’) were contrasted with characteristics at random sites (‘available’) within the study area. The study area was located in NE Poland in the upper Narew river valley. To localize foraging locations, Great Snipe males were trapped in May 2013 and 2014 on two leks (lek1 and lek2) and equipped with VHF transmitters. When an individual was approached, we locate the precise feeding site from where it was flushed and then sampled the attributes of the habitat. Some non-tagged birds were flushed by coincidence during fieldwork, and when their exact foraging site was detected we applied the same sampling protocol and included the data in our analyses (n = 41). To determine available resources within the study area in 2013 and 2014, we generated 200 and 250 random locations, respectively. Random sites in dense bushes and forest were omitted as non-potential foraging sites. In total, 383 random sites were sampled.

Invertebrates data

To assess subsurface invertebrates (different taxa, see README file) we collected soil samples (15x15x10 cm depth, cut with a steel frame) from the center of the plot. All potential prey (> 0.5 mm) were searched within those samples in the lab, when washed through a 1 mm mesh. Invertebrate abundance (total wet biomass of identified class, order or family weight, ±0,001 g) was then analyzed in the lab.

Statistical analyses

We coded sites where we found birds as 1 (“used”) and randomly selected sites from within our study domain as 0 (“available”), and used this as a binary response to model Great Snipe habitat selection as a function of candidate predictors measured at two spatial scales (see Micro-scale variables in Table 1, see Usage Notes). We used generalized linear models (GLMs) with a binomial error structure and logit link function to model the relationship between the presence/absence of birds and candidate predictors at the micro (GLM1) and the macro scale (GLM2). For more details go to published article.

Micro-scale model

For the micro-scale models, we considered nine candidate predictors (three of them in the squared form) as main terms and included also interactions between those variables and the date. For more details go to published article.

Macro-scale model

At the landscape scale, we considered seven candidate predictors (four of them in the squared form, see Macro-scale variables in Table 1 and Table 2, see Usage Notes) as main terms to build a global model. We used R environment (R Core Team 2019) to fit GLMs (for more details go to published article). Biomass of food resources was not included in model selection because these variables were not assessed on all sampling sites (for more details go to published article).

Variable codes in Table 1 and Table 2 refers to headings columns of uploaded file: Korniluk_et_al_2000_Great_Snipe_habitat_selection_data.

Additional column headers are described in uploaded README file.

Table 1. Variables at two levels – micro- and macro-scale.

Table 2. Land use types used in the macro-scale model.