AbstractQuestion:  How do naturalistic grazing in contrast to mowing and free succession affect plant community composition and species richness in a temperate grassland grazed by semi-feral cattle and horses?Location: Mols Laboratory, DenmarkMethods: We investigated grazing exclosures in the rewilding area of the Mols Laboratory, four years after its establishment. We focused on moist to dry grassland vegetation, i.e. excluding scrub and woodland. Each experimental block consisted of five 5 × 9 m plots, representing four fenced treatments, i.e. summer-only grazing, winter-only grazing, full exclosure with annual autumn mowing and full exclosure with passive succession. The matrix was grazed by large herbivores at close-to-natural densities, i.e. regulated bottom-up by the carrying capacity of the area. Hence, the seasonal grazing treatments were grazed at close-to-natural animal density. Quantitative plant community composition was assessed using the point-intercept method in 25 × 25 cm quadrats, supplemented with biomass calibration models based on additional quadrats, in which above-ground plant biomass was harvested after recording and the material sorted to species and weighed. Uniqueness was assessed as the sum of inverse range sizes for constituent species.Results: We found an appreciably higher plant species richness in grazing treatments than under annual mowing and full exclosure, but only minor differences between seasonal grazing treatments. Uniqueness was highest in year-round and winter-only grazing and lowest in summer-only grazing. The forb:graminoid ratio tended to be high in the winter-only grazing treatment, whereas annual mowing was associated with dominance of graminoids over forbs. Full exclosure plots had accumulation of litter and the lowest species richness. Initial heterogeneity between plots within blocks and a systematic differences between blocks in moist and dry grasslands seemed to swamp treatment effects at this early point after the establishment of the experiment. Data analysis using the biomass estimates derived from the calibration models yielded only minor differences in the patterns described above, when compared to the results obtained using the raw number of intercepts.Conclusions: Grazing under near-natural conditions is a goal in itself in ecological restoration, but also proposed as an efficient management tool to promote conservation of grassland plants and communities. We found both plant species richness and the prevalence of rarer species (unicity) to be higher with grazing than mowing/abandonment. Similarly, the tendency for forbs to prevail under grazing may translate into enhanced floral resources for anthophilous insects. Summer-only grazing at low density of large herbivores was not significantly different from winter-only and year-round grazing, but this treatment was much closer to natural grazing than intensive summer grazing typical of agri-environmental practices.Keywords: biomass estimation, disturbance regime, point-intercept method, rewilding, uniquenessIntroductionGrazing by large herbivorous mammals is a key process shaping vegetation structure and habitat conditions for plants and other organisms (Bakker et al. 2016; Malhi et al. 2016; Galetti et al. 2018). In European conservation management, there is a strong tradition of aiming at mimicking traditional practices in agriculture and livestock husbandry, e.g. extensive haymaking and summer grazing (Varga et al. 2016). In reality, however, actual conservation management is often strongly constrained by the opportunities compatible with modern high input-high output farm management and agri-environment support schemes (Newton et al. 2012). Either way, conservation management practice is not always rooted in ecological theory and often fails to deliver the desired outcomes for biodiversity (Maxwell et al. 2020; Kindvall et al. 2022). Attempts to apply first principles to grazing management can be comprised under the term “naturalistic grazing”, which may be characterized as landscape-scale conservation management, under which grazing as a natural process is seen as an aim in itself, and where human intervention therefore is reduced to a minimum and where herbivore density is not human-controlled, but left to be resource-regulated (Hodder et al. 2005). Although “naturalistic grazing” is considered open-ended with regard to effects on herbivore populations and vegetation, monitoring the effects is crucial to our understanding of how grazing as a natural process works and interacts with other natural conditions and processes.In large contiguous landscapes, habitat use by large herbivores often shows substantial variation in diurnal, seasonal and between-year patterns. Animal activity tend to be concentrated in certain areas, while large areas may be much more extensively used, e.g. wet areas may be avoided during winter, but preferred in spring and summer (Górecka-Bruzda et al. 2020). Traditional European livestock husbandry had, and continues to have, the growth and survival of domestic animals as its core purpose. Therefore, summer-only grazing on pastures and winter feeding of stabled animals was traditionally the norm in Denmark, in particular for cattle, while some horses have traditionally been left on pastures year-round (Fritzbøger 2004). The pattern of summer-only grazing is strongly reinforced in modern North-European livestock husbandry, in which standard practice is to turn livestock out at very high density (e.g. 800-1000 kg·ha-1) during a short period of intensive grazing at the peak of the growing season (typically May through September or shorter). One way of investigating the resulting impact on vegetation of the annual timing of herbivore activity is to compare areas, to which animal access is restricted to certain parts of the year  (Bullock et al. 2001). Modern European grasslands are often highly grass dominated. The relatively low prevalence of forbs may, however, be a legacy effect of past megafauna extinctions. It has been hypothesized that megafauna once sustained much higher abundance of forbs in grasslands (Bråthen et al. 2021). The shift in dominant growth form has likely been exacerbated by the more recent demise of large herbivores from European landscapes at large, and natural areas in particular. The shift has probably propagated to higher trophic levels, i.e. mega-diverse consumer taxa, e.g. arthropods and fungi  (Brunbjerg et al. 2018). Flower-visiting insects have attracted particular attention, partly because this functional group is particularly threatened by both land-use intensification and abandonment, and partly because some anthophilous insect taxa have shown dramatic declines in species richness and abundance (e.g. Hallmann et al. 2017; Warren et al. 2021). The ratio in vegetation of forbs to graminoids has therefore been particularly highlighted, as most forbs have flowers offering resources to anthophilous insects, while graminoids all have wind-pollinated flowers.The response of vegetation structure to grazing regime will likely involve changes in quantitative plant community composition, with the activities of large herbivores promoting the abundance of certain species, while limiting others. We therefore applied the point-intercept method to quantitatively recording vegetation structure (Jonasson 1988; Godínez-Alvarez et al. 2009; Bonham 2013). Non-destructivity is a virtue of the method, which was desired in the current setup of long-term monitoring plots, also surveyed for other groups of organisms. However, because of differences in plant architecture, the intercept-based abundance does not translate directly to biomass-based abundance. We therefore made calibration models per species and/or functional groups, based on an additional set of quadrats, first subjected to point-intercept recording, next to total harvest and dry-mass estimation per species.Plant community species richness, or alpha diversity, is of core interest to evaluations of vegetation under contrasted grazing regime, although results may depend on the actual quadrat size applied. From the perspective of gamma diversity in the region or country, however, community unicity - the regional rarity of constituent species - is of higher relevance. One way to evaluate the contribution of individual communities to regional gamma diversity is the ‘Sum of inverse range-sizes’ (Guerin & Lowe 2015; Ejrnæs et al. 2018), in which constituent species are given decreasing weight with increasing regional occupancy. Also, from the perspective of biodiversity conservation, community unicity may be more relevant than alpha diversity, e.g. even locally species-poor communities may be of high regional conservation value, if they tend to consist of relatively rare species.Our overarching aim was to assess differences in grassland vegetation structure, community richness and unicity (the prevalence of less widespread species) as a snapshot after four years of naturalistic year-round grazing, as compared to seasonal grazing regimes, to mechanical mowing and to free succession after grazing abandonment and mowing regimes. Specifically, we aimed at investigating:1) Does plot-scale plant species richness vary between year-round grazing, seasonal grazing (all at naturalistic herbivore density), mowing and passive succession?2) Does forb to graminoid ratio vary between year-round grazing, seasonal grazing (all at naturalistic herbivore density), mowing and passive succession?3) Does plant community unicity vary between year-round grazing, seasonal grazing (all at naturalistic herbivore density), mowing and passive succession?A subordinate aim was the methodological issue of non-destructive assessment of quantitative plant community composition and the sufficiency of the point-intercept method as compared to biomass estimation. Materials and methodsStudy siteThe Mols Laboratory is an ecological research station, owned by the Natural History Museum, Aarhus. It is 120 ha located in the glacially shaped hilly landscape of Mols Bjerge at 56.22° latitude and 10.57° longitude. The area covers wide gradients in soil moisture, nutrient status and vegetation openness. Roughly half of the area is covered by open habitats, the other half by scrub and forests, with all types in a mosaic with gradients both between open and canopy-covered habitats and dry and moist habitats. The most frequent open habitat type, as categorized under the European Habitats Directive, was Species-rich Nardus grasslands (6230). Despite all research blocks but one (block 70 was not located in a designated habitat type) being assigned to this type of grassland, quite large variation in the species composition and topography between different parts of the area is evident, foremost between hilly glacial gravelly till and sandy marine foreland shaped by the higher sea-level of the Littorina transgression (Atlantic; 6800 - 3900 BCE). This contrast is presumably particularly linked to hydrology, with the marine foreland being somewhat impacted by exfiltration of groundwater from the hills.In 2016, as part of a rewilding project, 13 heads of Galloway cattle and 12 Exmoor ponies, supplemented 6 months later by a stallion, were released. Since then, the herds have lived there under near-natural conditions, which means that population size is determined by the carrying capacity, i.e. no supplementary feeding. As population size approaches carrying capacity, food shortage is likely to kick in during the late winter months, at which time weak individuals will die. In order to minimize possible suffering of the animals and to comply with Danish animal welfare legislation, the herds have been continually evaluated, following a scoring protocol based on body condition and behaviour. Individuals failing to meet a set threshold are removed from the area. This so-called “reactive” population management has led to a dynamic development in population size, with the population of large herbivores growing to a total of 44 cattle and 25 ponies in the summer of 2019, and subsequently decreasing to the current level of 12 cattle and 26 ponies. Assuming a standard body mass for cattle of 550 kg and 350 kg for horse, this is equivalent of a drop in large herbivore density from almost 300 kg/ha to 140 kg/ha.With the purpose of monitoring the effects of the naturalistic grazing regime, 22 randomly selected blocks for permanent vegetation monitoring were established in the spring of 2017. Each block contained four treatment plots, i.e. summer-only grazing (exclosure November through April), winter-only grazing (exclosure May through October), annual autumn mowing (full exclosure with one annual cut during September-October and the thatch removed) and passive succession (full exclosure), all embedded in the matrix of year-round grazing. Animal density was regulated by forage availability of the entire area, which means close to natural levels. Shrubs (but not trees) were initially removed from the mown plots in order to allow cutting by machinery, but not from the other treatments. Fencing did not prevent access to plots by herbivores such as red deer, roe deer and hare, only horse and cattle. Vegetation recordingIn order to select grassland blocks, all monitoring blocks were initially surveyed in order to establish the dominant vegetation type. Out of the initially established 22 blocks, one was discontinued, four were located in closed-canopy forest, four were almost entirely covered by dense scrub and four had scrub-grassland mosaics with too high scrub cover for the point-intercept method to be practically applicable in all treatment plots, leaving nine blocks with mostly open grassland vegetation (Fig. 1).Field work was carried out in two periods: September 1-16, 2020 and August 2-20, 2021. In the first period, sampling quadrats in treatment plots were put in the periphery of the established circular monitoring plots (see Supplementary materials, Appendix 1) and subjected to point-intercept vegetation recording, after which the above-ground biomass was cut as close to the soil surface as possible, immediately sorted into fractions by plant species (with standing litter as a separate fraction) and dried at 55 ˚C until constant weight. The resulting data were used to create calibration models per species or functional groups for the prediction of plant species above-ground biomass from non-destructive point-intercept counts. In the second time period, sampling quadrats were located within the circular monitoring plots and surveyed using the same point-intercept method as in the first round. The resulting records were 1) used as-is, 2) subjected to prediction of species’ biomass using the regression models of the first round.Due to time constraints, only six of the nine blocks were included in first round of field work and were thus included in the construction of calibration models (i.e. block numbers 60, 62 and 70 were not sampled). Similarly, mown plots were not sampled, as they were in the process of being mown while the vegetation surveys were carried out. Including newly mown plots would have compromised the reliability of calibration models. In the second year, the field work took place one month earlier, allowing point-intercept recordings in the annually mown plots.The total sample size for the two periods of fieldwork were: 6 blocks × 4 treatment plots = 24 quadrats in the first period (in 2020) and 9 blocks × 5 treatment plots = 45 quadrats in the second period (in 2021).