The former Milovice military training area harbours rich butterfly assemblages; the 55–60 species currently recorded per site is above average for nature reserves in the country (Slancarova et al. 2014; Bartonova et al. 2016). This richness was arguably preserved there owing to exclusion of intensive agriculture and forestry, combined with the past finely-grained disturbance-succession dynamics typical for military areas (Reif et al. 2011; Cizek et al. 2013; Busek and Reif 2017). Following the cessation of military use, several species were lost, while others were subsequently gained. The presence of aurochs, horses, and wisents increases per-plot butterfly abundances and contributes to maintaining their diversity, providing for multiple species of conservation concern, including the critically endangered obligatorily myrmecophilous Phengaris alcon (cf. Thomas and Settele 2014).
Changes since termination of military use
The termination of military activities was followed by the successional overgrowth of the disturbed sparsely vegetated surfaces, and gradual dominance of coarse grasses and tall forbs. We therefore expected (hypothesis H1) decrease of specialists associated with small competitively inferior forbs, which our analyses of life history traits did not support. The only life history trait responding to the past-present ordination was mobility. Poorly mobile species were associated with the past military use. Among European butterflies, high mobility is a generalist trait associated with broad trophic ranges, long flight period and other features facilitating survival in human-dominated landscapes (Dapporto and Dennis 2013; Bartonova et al. 2014; Habel et al. 2019), whereas poor mobility increases extinction risks (Birkhofer et al. 2017; Essens et al. 2017). Because mobility relates inversely to local population density (Bartonova et al. 2016), some poorly mobile species may need large habitat areas to sustain viable populations. The changes after cessation of military use probably led to shrinking habitats supply for poorly mobile specialists.
Associations of lost and gained species with climatic niche traits (H2) were more straightforward. In agreement with the warmer and drier climate in Central Europe during the last few decades (Stuhldreher and Fartmann 2018), the lost species shared broad oceanity or precipitation niches, whereas the newly gained species require higher temperatures. Also, in agreement with H3, the locally lost species display decreasing distribution trends in the Czech Republic and elsewhere in Central and Western Europe (cf. van Swaay et al. 2010).
A combination of restricted mobility and broad oceanity or precipitation niches applies to several locally lost and nationally threatened species (cf. Benes et al. 2002; Hejda et al. 2017): the hesperids Pyrgus armoricanus (currently re-expanding elsewhere in Central Europe: Benes et al. 2020; Kettermann et al. 2020) and Thymelicus acteon and the satyrines Hipparchia semele, Hyponephele lycaon, and Erebia aethiops. The latter is a sparse woodland species (Slamova et al. 2012) only loosely associated with grasslands, but its current occurrence in the area was safely excluded by concurrent targeted searches. The remaining four, all declining in Central Europe (van Swaay et al. 2010), require sparsely vegetated substrates and often colonise such landforms as disused quarries and post-industrial barrens (Bourn and Thomas 2002; Benes et al. 2003; Tropek et al. 2010; Tropek et al. 2017). Broad oceanity tolerance certainly applies to Hipparchia semele, distributed from Eastern Europe to Atlantic coastal dunes (Schirmel and Fartmann 2014), but also to Hyponephele lycaon and Pyrgus armoricanus, whose distribution follows maritime climates far north to southern Fennoscandia (Fourcade et al. 2017; Mikkola 1979). The species newly gained during the last two decades include Iphiclides podalirius, Satyrium acaciae, S. spini, Lycaena dispar, and Polyommatus bellargus, all currently (re)expanding in Central Europe. The first three are associated with shrubs (Benes et al. 2002), and the fourth with tall ruderal forbs (Strausz et al. 2012), hence they might have profited from concurrent effects of successional abandonment. Only the fifth, gained as late as 2018, develops on Securigera varia (L) host plants growing at sparsely vegetated surfaces (Benes et al. 2003), which apparently profits from the ungulates’ grazing. The gains and losses thus reflect the interacting forces of climate and land use change (Reif et al. 2008; Thomas et al. 2015).
Refaunation by large ungulates
Large ungulates did not demonstrably change per plot species richness while increasing butterfly abundances, only partly supporting our hypothesis H4. At the same time, refaunation affected the local assemblages’ composition. It favoured species developing on small forbs over those developing on large forbs, grasses or shrubs, supporting our hypothesis H5.
As in other studies (cf. Hennig et al. 2017; Cromsigt et al. 2018b; Zielke et al. 2019), the immediate effects of year-round ungulates’ presence included reduction of tall coarse grasses, slowing down scrub growth due to browsing and bark peeling, reduction of grass blooming by feeding on grass inflorescences, and exposing barren ground around tracks and wallows. As in experiments with feral horses (Garrido et al. 2019), some richly blooming forbs, including species that rarely bloomed in the years preceding the refaunation (unpublished data), became notably more abundant. The differences in butterfly assemblages composition between refaunated and neglected plots became apparent only after statistical control for the effect of years and to monitored plots position. Still, species benefitting from refaunation included the iconic Phengaris alcon f. rebeli, whose host plant, the poorly competitive (cf. Petanidou et al. 1995; Habel et al. 2007) and chemically protected (Popovic et al. 2019) perennial Gentiana cruciata (cf. Petanidou et al. 1995), boomed shortly after the establishment of grazing. This obligatorily myrmecophilous butterfly is likely host plant limited, because its females prefer oviposition on plants overtopping surrounding vegetation (Meyer-Hozak 2000; Habel et al. 2016; Vilbas et al. 2016).
The simplest explanation of the higher butterfly numbers at refaunated plots, attraction to increased nectar, is unlikely, as covariable nectar had no effect in ordinations. A tempting explanation is the smaller size of the forbs-feeding specialists, related to higher local population densities and lower mobility (cf. Bartonova et al. 2014). In any case, it is intriguing that many plant groups avoided by horses (e.g., Rosaceae, Fabaceae, Polygonaceae, Orobanchaceae: Chodkiewicz 2020), once the dominant grazers of West-Palaearctic grasslands, are frequent in the larval diet of European butterflies. Possible coevolutionary relationships between mammalian megafauna and herbivorous insects, and their conservation implications, deserve further investigation.
The patterns revealed by ordinations relating species composition to refaunation were admittedly less convincing than in studies comparing starkly contrasting habitats, such as close woodlands vs. clearings (cf. Benes et al. 2006; Sebek et al. 2015). It appears that the refaunated and neglected plots were interconnected by individual movements. The distances among study plots were within the routine movement abilities of most butterflies (Fric et al. 2010; Stevens and Baguette 2010; Vodickova et al. 2019), although this may not apply for the least mobile species (Korosi et al. 2008). Also, the small-scale vegetation mosaic at the study sites (Jirku et al. 2020; Fig. 1) could blur potential effects to species community structures. It is likely that individual butterflies located some of their vital resources at both grazed and ungrazed sections of the area, in line with the resource-based understanding of (animal) habitats (Dennis et al. 2006; Turlure et al. 2019).
The setting of our study did not allow distinguishing between the effects of horses and big bovids, as both pastures contained combinations of these ungulates. The literature on refaunation in temperate (e.g., Vera 2000; Zielke et al. 2020) and northern boreal (Macias-Fauria et al. 2020) regions agrees that these two ungulate groups supplement each other in effects on vegetation, as well as seasonal and diurnal habitat use. Additionally, both horses and bovids acted as dominant grazers in late Quaternary European ecosystems, and both were present in traditional rural landscapes.
The mechanical disturbance by armoured vehicles (factor tanks) exhibited no separate effect, seemingly countering the claims (e.g. Heneberg et al. 2016) that it provides disturbed conditions beneficial for some insects. Presence of tanks, however, increased the explanatory power of models containing ungulates or refaunation effects (Table 4), suggesting a complementarity with large grazers for some butterfly species. This might be the case of Spialia sertorius, a skipper closely associated with tanks in ordination diagrams and developing on Sanguisorba minor, a competitively inferior forb preferring sparsely vegetated surfaces (cf. Gros 2002). It is tempting to postulate that on military lands, and in the current Milovice reserves, the heavy vehicles supplement yet another lost component of the megaherbivore fauna of interglacial Europe, proboscideans (van Kolfschoten 2000).
The effect of domestic cattle, grazed at two plots for two years of the project, was orthogonal to the ordination gradient distinguishing refaunation and neglect. The cattle were grazed with high stocking and supplementary feeding during the vegetation season and were not grazed in winter. Such grazing style suppresses forbs and fails to suppress coarse grasses. Grazing by domestic breeds in more biodiversity-friendly ways is possible (Enri et al. 2017; Henning et al. 2017; Hall and Bunce 2019), but this was not the case in our system.
While being demonstrably positive for butterflies associated with poorly competitive forbs, the refaunation did not detectably imperil species associated with coarse grasses or shrubs. In this respect, the Milovice situation differs from some projects with documented negative outcomes for insect assemblages (cf. Lorimer and Driessen 2014; van Klink and WallisDeVries 2018). It seems beneficial that contrary to some refaunation sites amidst urbanised landscapes (Lorimer and Driessen 2014), our study system is situated in a diverse rural setting, including ungrazed/neglected plots, which provide conditions contrasting with the grazed sites. This habitat diversity likely allows for resource compensation/supplementation by the butterflies (Ouin et al. 2004), enabling coexistence of species requiring different disturbance levels (Bergman et al. 2018). The current grazing pressure ≈ 0.5 grazers*ha− 1 does not deplete the sites of larval host plants or nectar. There is a potential long-term risk, as the whole operation is funded from the EU Agri-environmental scheme “grazing”, which requires maintaining stable grazing intensity. Flexibility may be necessary, as grazing levels appropriate for restoring overgrown sites may become too high once species-rich dry grasslands develop, as well as if accelerating climate change will decrease rainfall levels during the vegetation period.