2.1.3. Balatonakali-Dörgicse

The site is between the villages of Balatonakali and Dörgicse, 145–240 m above sea level. The ca. 200 ha wood pasture is part of the Natura 2000 network. The western part of the area was cleared a few years

ago and sparse wooded vegetation was left behind. The open pasture with a sheep paddock stretches more or less continuously along the southern border of the village of Dörgicse. The southwestern and northwestern parts of the area are covered by continuous forests. In the sheep paddock near Dörgicse, about 1000 head of sheep are kept. They are grazed in smaller flocks mostly on the open pasture and the nearest parts of the wood pasture.

#### *2.2. Sampling Design and Dung Beetle Trapping*

The dung beetle assemblages have been sampled three times in 2016, in spring (May), summer (July), and autumn (October). Sampling was carried out simultaneously in all three sites. In total, 81 pitfall trap units baited with either cattle, horse, or sheep dung (three of each for every habitat in all three locations) were installed. The pitfall traps baited with di fferent dungs were placed as if they were vertices of a triangle with sides slightly more than 10 m. The location of the sampling plots was selected to meet two basic criteria: they should be at least 100 m away from each other [38] and the edge of the habitat type should be at least 50 m away. As far as the terrain allowed, traps were installed in locations sheltered by stumps, smaller trees, or shrubs to avoid damage caused by trampling.

The 1 L containers (diameter: 11 cm, height: 15 cm) were dug into the soil up to their rim. About 3 dl of diluted (50%) propylene-glycol was used as a preservative in each container. The mouth was covered by a hexagonal chicken wire (with a mesh diameter of 25 mm) that was fixed to the soil by U-shaped wires. The dung bait was wrapped in mosquito net and fixed to the wire mesh covering the mouth by bailing wire.

The dung used as bait was always collected freshly for each sampling period from animals grazing on the specific sites. The same batch of dung was used in each trap of a particular site. Livestock were not treated with anthelmintics prior to dung collection. The collected dung was portioned into 400 (cattle, horse) and 100 g (sheep) packs within laboratory conditions. Packs were sterilised by freezing at −20 °C for at least 72 h. The pitfall traps were emptied after one week. In one case, the container was dug up by animals; this sample was excluded from the analysis. Dung beetles were identified at the species level in the laboratory. Nomenclature follows the Catalogue of Palaearctic Coleoptera [39].

#### *2.3. Data Analysis*

We separately tested the e ffect of four explanatory variables (habitat, season, dung, and locality) on abundance and species richness by generalised linear mixed models (GLMMs; [40]) and on Shannon diversity using linear mixed models (LMMs), where the response variables were the number of species (with Poisson distribution), the abundance (with negative binomial distribution), or the calculated Shannon diversity (with Gaussian distribution) of dung beetles. Replicates (sampling plots with three pitfall traps) were included as random factors. We compared the models fitted with explanatory and random factors to null models (including the random factor only) by ANOVA [41].

Our main aim was to reveal the di fferences in dung beetle assemblages of forest, wood pasture, and pasture habitat types. Thus, first, we compared species richness, abundances, and Shannon diversity among habitat types by a pairwise comparison with Tukey corrections, based on models described above. Second, we created subsets of season and dung type (thereby creating 9 datasets both for abundances and for number of species) and separately analysed the e ffect of habitat type. We applied GLMMs and LMMs [40], where the response variables were the number of species, the abundance, or the diversity of dung beetles, while the explanatory variable was only the habitat type (forest, pasture, and wood pasture). Replicates within the locality were included as random factors, and we repeated these analyses for all 9 subsets (3 seasons × 3 dung types) of our dataset. We applied pairwise comparisons with Tukey corrections between habitat types based on model results.

We explored the differences between the composition of dung beetle assemblages with nonmetric multidimensional scaling (NMDS) using a Bray–Curtis similarity measure [42]. We applied a Permutational Multivariate Analysis of Variance (PERMANOVA) [42] to analyse the effect of habitat, season, locality, and dung type as explanatory variables on assemblage composition. We evaluated the association of dung beetle species to habitat, dung type, season, and locality separately, by indicator species analysis [42]. The indicator values of the species were tested via the Monte-Carlo simulation using 10,000 permutations. The accepted significance level was *p* < 0.05.

The statistical analyses were carried out using R 3.4.4 statistical environment [43]. We used "lme4" 1.1-17 and "MASS" 7.3-51.4 packages for models [44,45], "lsmeans" 2.27-62 and "multcomp" 1.4-10 packages for pairwise comparison [46,47], "vegan" 2.5-2. package for nonmetric multidimensional scaling [48], and "labdsv" ver. 1.8-0 for indicator species analysis [49].
