Discussion

Grazing by semi-feral horses and cattle under near-natural conditions was found to support higher plant species richness and higher prevalence of less widespread species (unicity) in grassland vegetation, as compared to annual mowing and to full exclosure of large herbivores. Also, we saw a tendency for the prevalence of forbs over graminoids to be higher, which may result in increased floral resource availability to anthophilous insects. At first glance, this result may seem trivial, since temperate grasslands as an ecosystem are inherently dependent on the disturbance regime maintained by the activities of megafauna (Janis et al. 2002; Pärtel et al. 2005; Kuneš et al. 2015). It is nevertheless, in our view, an important result, as it highlights the detrimental effect on grassland plant diversity of the now widespread grazing abandonment or mechanical biomass harvest. Since wild grazers, such as red deer, roe deer and hare, had free access to the fenced plots and were observed in the area, our full exclosure treatment mimics the standard conditions in most temperate European landscapes, where populations of wild herbivores are kept at very low densities, i.e. the maximum acceptable to agriculture and silviculture (Fløjgaard et al. 2022). In accordance, we found accumulation of litter in full exclosure plots. Thick layers of litter changes light and microclimate conditions at the soil surface, deteriorating conditions for plant recruitment and survival (Jensen & Gutekunst 2003) and reducing the richness of arthropod assemblages (van Klink et al. 2015). Although grazing and haymaking may share more similarities than dissimilarities (Pykälä 2000), our results indicate that annual autumn mowing does not support the same plant species richness and abundance of forbs as grazing.
One more reason not to over-emphasize differences in vegetation response between grazing treatments obtained from exclosure plots is that larger grazed landscapes are likely to encompass areas used more and less intensively by the large herbivores at different times of the year. Even if it would be possible to experimentally demonstrate plant community properties typical of either year-round, winter-only and summer-only grazing, it may be hypothesized that all these characteristics would be created or maintained in the landscape at large by such animal behaviour, thereby promoting overall habitat heterogeneity. More specifically, summer-only grazing might be a natural grazing regime in naturally nutrient-rich meadows, which are flooded during winter, and that winter-only grazing is a natural regime in grassland, heathland and open woodland on nutrient-poor higher grounds. We also acknowledge that our study only takes a snapshot of effects that may depend on legacy effects of previous land use  (Stroh et al. 2021) and may be transient in the dynamic interaction between herbivore population dynamics and inter-annual variation in weather conditions, such as the irregular occurrence of severe summer droughts  (Stampfli et al. 2018).
It should be emphasized that modern European grassland management as an agri-environmental practice often occurs as intense summer grazing at high stocking rates (Fløjgaard et al. 2022). This practice is very different from our summer-only grazing treatment, in which animal densities approach natural densities, which are set by winter forage carrying capacity.
Several recent studies have reported a positive effect of year-round grazing on plant species richness and – in particular – the prevalence of rare or threatened species (Köhler et al. 2016; Rupprecht et al. 2016). Temperate European landscapes have seen major habitat changes over recent decades and centuries affecting plant species and communities, including the effective disappearance of first, large wild herbivores and second, free-roaming livestock (Bruun & Fritzbøger 2002; Finderup Nielsen et al. 2021). Naturalistic grazing appears to be key in restoring habitat conditions lost in this long-term land-use change and intensification. Although, in the present study, we were not able to encompass the grassland-scrub-forest ecotone, the heterogeneity arising from naturalistic grazing is probably important to many species showing declining trends in human-dominated landscapes (e.g. Maes et al. 2014).
A tendency for naturalistic grazing to promote forbs over graminoids has recently been reported elsewhere (Henning et al. 2017; Dvorský et al. 2022). Despite graminoids lending their name to the grassland ecosystem, forbs are essential to the phylogenetic and functional diversity of grassland communities, and are particularly important to flower-seeking insects. It has been hypothesized that grass dominance in temperate open biomes is a relatively recent phenomenon  (Bråthen et al. 2021) and, no doubt, grass dominance has been promoted by agricultural grassland management  (Dengler et al. 2020). Possibly, the change of dominance towards forbs may be an effect of animal trampling as much as their forage preference (Striker et al. 2011). Although differences in forb:graminoid ratio between our treatments were not statistically significant given the variation between plots, we note that the highest levels of forb relative abundance was found in the three grazing treatments. We also note that certain forb species may be sensitive to grazing, but in a larger grazing landscape probably would find habitat in browsing shelter between thorny shrubs or fallen logs.
From a methodical standpoint, our results indicate that the application of the model derived biomass estimates in most cases do not change the overall patterns also revealed by the intercept based analyses. It would therefore seem that the supposed differences in species morphology were, in most cases, not large enough to generate significantly different results in terms of species composition. Furthermore, the application of the calibration models to the number of intercepts constitutes a complicating step in the analysis, due to the error propagation that results from using model estimates in a function. While this does not mean that the model derived results are invalid, it may be both more efficient and accurate to use the intercept based results. It can therefore be argued that the use of biomass calibration models may not be necessary for comparisons to be made between treatments, unless results from methodologically different studies are compared.
Grazing is a natural ecosystem process and, thus, restoring naturalistic grazing using feral animals may be seen as restoration goal in itself, notwithstanding the effects on biodiversity. Nevertheless, the present study demonstrates that naturalistic grazing as a management tool may indeed create habitat for regionally uncommon species of conservation concern and probably enhance resource availability to flower-seeking insects.