Grazing intensity drives plant diversity but does not affect forage production in a natural grassland dominated by the tussock-forming grass Andropogon lateralis Nees

Andropogon lateralis is a tall and highly plastic tussock-forming grass native from southern South America. It is a frequent component of Campos and Subtropical highland grasslands that often becomes dominant under lax grazing regimes. The aim of this work was to analyze the response of species diversity and forage production of a natural grassland dominated by A. lateralis to a wide range of grazing intensity. We hypothesized that species diversity and forage production would both peak at the intermediate canopy heights determined by grazing regimes of moderate intensity. A grazing experiment was conducted in a highland grassland with mesothermal humid climate at 922 masl (Atlantic Forest biome, Santa Catarina state, Brazil) that comprised 87 species from 20 families but had 50% of its standing biomass accounted by A. lateralis. Four pre-/post-grazing canopy heights—12/7, 20/12, 28/17, and 36/22 cm (measured on A. lateralis)—were arranged in a complete randomized block design with four replications, and intermittently stocked with beef heifers from October 2015 to October 2017. Andropogon lateralis cover decreased (from 75 to 50%), and species richness increased (15–25 species m−2) as canopy height decreased. Grazing intensity did not affect annual forage production (4.2 Mg DM ha−1). This natural grassland dominated by A. lateralis had a high capacity to adjust to grazing regimes of contrasting intensity, maintaining forage production stable over a wide range of canopy heights. However, to prevent losses in floristic diversity, such grassland should not be grazed at canopy heights higher than 28 cm.

www.nature.com/scientificreports/ influences not only the botanical composition but also the harvested yield of pastures [5][6][7][8][9] . For this reason, agroecosystems based on the use of highly diverse natural grasslands aim to design grazing management strategies that couple high enough animal production so that economic sustainability is ensured with the preservation of native species that provides the basis for resilience 10 .
Andropogon lateralis Nees is a tall tussock-forming perennial C4 grass native to South America. It is well accepted by grazing animals, presents an extremely high phenotypic plasticity showing large variations in specific leaf area (SLA) and leaf dry matter content (LDMC), and thus being able to display either a competing or a resource-conservation functional strategies depending on environmental conditions 11 . For instance, under lax grazing it conserves resources by reducing SLA and increasing LDMC, which along its tall stature confers such a competitive advantage that it often becomes highly dominant in large areas [12][13][14] .
The effects of grazing intensity on community structure have been found to mainly depend on, first, the ability of species to regrowth after a grazing event, and secondly, on the selectivity of the grazing animal for patches or species, both processes strongly interacting with forage biomass availability 15 . The often used humped-back model suggests that plant diversity in natural grassland ecosystems peaks at intermediate biomass levels 16 . Likewise, the intermediate disturbance hypothesis postulates that the highest diversity is maintained at intermediate scales of disturbance 17 . At community level, the replacement of few dominants by several subordinate species often occurs due to moderate grazing intensity and alleviating light limitation 4 .
In spite of its extension and relevance, few studies have focused on the response of natural grasslands dominated by A. lateralis to various grazing managements. In consequence, little is known about the processes determining species composition and forage production in these communities. Further, the particularly mild and productive climate in which these natural grasslands occur (> 1000 mm annual precipitation, no dry season, rare snowfall) difficults to extrapolate inferences from other studies. For instance, in such a humid climate, the intensity of competition for nutrients (nitrogen in particular) and for light, as well as the adaptive responses of different native species to nutrient deficiency and shading, could be much more important in determining competitive outcomes than in rangelands with typically semiarid and more continental climates (e.g. [5][6][7][18][19][20]. Grazing intensity, with its strong effects on animal selectivity and plant growth 21 , and thus on (the spatial heterogeneity of both canopy height and nutrient (especially nitrogen) cycling, could therefore have an exacerbated role in modulating inter-specific relationships in these grasslands.
The aim of this work was to analyze the response of species diversity and herbage production of a natural grassland dominated by A. lateralis to a wide range of grazing intensities. We hypothesized that species diversity and forage production would both peak at the intermediate canopy heights determined by grazing regimes of moderate intensity, because under lax grazing intensity, species able to become tall and unpalatable, and conserve nutrients, would become dominant, whereas under too intense grazing only a few species with prostate habits would survive.

Results
Forage production. The annual production of forage was unaffected by grazing intensity and similar in both years (p > 0.05; Table 1), averaging 4236 kg DM ha −1 .

Floristic composition.
A total of 87 species distributed in 20 families were identified during the experiment. The largest number of species was in the Poaceae family (23 genera, 37 species), followed by the Asteraceae (16 genera, 19 species). The Fabaceae and Cyperaceae had five and four species in four and three genera, respectively. The Apiaceae and Convolvulaceae families were represented by three species, and Oxalidaceae and Rubiaceae by two species. All other families had only one species (Supplementary material).
On average, 50% of the pasture cover was A. lateralis. Ten species comprised over 90% of the forage mass ( Table 2). The only non-grass was Ulex europaeus, a cosmopolitan invasive legume shrub native from Europe 22 . Two of the most important grasses were from the C3 physiological group pathway (Anthoxanthum odoratum and Piptochaetium montevidense), and the others were C4. The remaining species represented less than 1% of the total forage mass.
The dead material component showed significant participation in the forage mass, especially at taller canopy heights, reaching just over 20% of the forage mass in the 36 cm canopy height.
Species-area curves. Species accumulation curves differed among seasons and grazing intensity treatments ( Fig. 1). There was a greater distance between accumulation curves in spring 2017, two years after the experiment started, with differences already apparent at the plot level (4 m 2 ) and on the 15 m 2 scale, yielding a difference of approximately 30 species. The number of species was also higher in spring than in autumn. tantly in response to grazing management, with some lesser variation due to season (interaction: P < 0.0001). Already 6 months after the beginning of the experiment, an increased A. lateralis cover was observed under the most lax grazing regime (pre-grazing height 36 cm), whereas cover was lowest in the most intensive grazing regime (12 cm) and intermediate in the other two treatments (Table 3). This trend was observed in all subsequent assessments, even though during the second year the cover of A. lateralis decreased somewhat in the more lax grazing regime. Tussock volume showed similar, correlative changes than cover, becoming larger the lesser the grazing intensity with some minor variation between seasons (interaction: P = 0.0344; Fig. 2). Over time, tussock volume decreased in the two most intense grazing treatments (12 and 20 cm), with the lowest values observed in spring 2017. The mean distance between tussocks varied only a few centimeters between treatments and seasons, although the effect was statistically significant (P < 0.0001). The average distance between tussocks decreased over time for the 12 and 20 cm treatments.

Indices for vegetation community.
There was an interaction between grazing intensity and season in species richness (P = 0.0011). Grazing intensity and season had significant effects on species diversity (P < 0.0001 and P < 0.0001, respectively), dominance (P < 0.0001 and P = 0.0066, respectively), and heterogeneity (P < 0.0001 and P = 0.0120, respectively) ( Table 4).
Species richness and structural heterogeneity. Canopy height of 12 cm was positively associated with both structural heterogeneity and species richness. Despite being more heterogeneous (Table 4), the 20 cm canopy height was associated with the 28 and 36 cm treatments, as shown by the overlapping ellipses (Fig. 3).
Species richness was highest in the most intense grazing treatment (12 cm) and lowest in the most lax (36 cm), with intermediate values in the intermediate grazing intensities (20 and 28 cm). The difference between grazing treatments was maximal at the end of the experiment: 25 and 15 species, respectively for the most intense and lax grazing treatments (Table 4). Species diversity showed a similar behavior, with highest values at 12 cm, lowest at 36 cm, and intermediate at 20 and 28 cm.

Discussion
During both experimental years, no difference was observed in forage production across grazing intensity treatments (Table 1). Hence, forage production was buffered over a wide range of canopy heights. This response has already been reported for monospecific grasslands 8,21,23 and associated with a trade-off between gross growth and senescence fluxes results in a relatively stable net forage production (= harvested or consumed herbage), and by underlying tiller size/density compensation 24 . Although our study produced a large variation in canopy structure, reflected in heights ranging from 12/7 to 36/22 cm, A. lateralis remained dominant even at the highest grazing intensity (Table 2), confirming the high plasticity reported for this species 11 .
In spite of a stable forage production, grazing intensity did promote strong changes in the plant community. At the lowest grazing intensity (36 cm canopy height), there was almost no temporal or spatial variation in A. lateralis cover (maximal), nor in richness and diversity (minimal), over the two years of the study (Tables 3 and 4). But as grazing intensity increased, and canopy height became shorter, it increased the participation and number Table 2. Rank consistency index (Cr) 51 and proportion of species (%) in the mean forage mass of the final two assessments (autumn and spring 2017) undertaken in a Highland grassland dominated by Andropogon lateralis Nees and managed under different canopy heights and intermittent stocking grazing. The percentage of the most important species was calculated excluding the dead material component; however, as there was considerable participation of this in the forage mass, the percentage and Cr of this component based on the total forage mass were also calculated. *Exotic species. The five most important species at each management height are given in brackets. www.nature.com/scientificreports/  www.nature.com/scientificreports/ of other species, especially in the lower stratum (Table 3). This effect was evident in both autumn and spring, becoming more marked as the experiment progressed. Maximal richness and diversity were thus observed in the shortest canopies at the end of the experimental period. Since A. lateralis, like other tall tufted grass, is highly competitive for light 11,25 under lax grazing regime, this response probably occurred as a result of changes in the light environment brought about by the reduction in A. lateralis cover that favored the development of subordinate, more shaded species 26,27 . However, concomitant effects of a possible higher nutrient availability due to more animal excreta at higher grazing intensities can not be excluded. Both height and shoot nitrogen concentration are strong determinants of individual and species dominance in diverse communities 28,29 . Under high grazing intensity, the mean distance between A. lateralis tussocks became shorter (Fig. 2). This may reflect central tussock dieback, a phenomenon that results from the death of the tussock centre and its subsequent fragmentation 30 . Therefore, a higher light availability penetrating into the canopy and increasing the participation of species already present in the forage mass or the appearance of new species would have resulted not only from lower the tussock height but also from their smaller basal cover. Increases in the number of species in short canopies has been reported 27 , demonstrating that grazing can be used as a practical tool to maintain or even increase plant diversity.
Grasses are more competitive than other herbaceous species 5 ; thus, species richness can be affected by grass dominance level, especially tussock-forming species, which have canopy advantages in terms of light competition 7,31,32 . The greater the abundance of dominant species, the more severe the interspecific competition 33,34 but grazing can affect species distribution patterns in the canopy, influencing competition processes 35 . Therefore, it is important to understand the mechanisms that drive grazing-generated changes (Fig. 3) based on species composition characteristics 27,36 .
The humped-back model indicates that plant diversity is at its maximum at an intermediate biomass level 16 . Based on this, we expected to observe a succession of structures across the grazing intensity treatments, from shorter and homogeneous at high intensity, to bimodal and heterogeneous at moderate intensities, to taller and homogeneous at low intensities. However, structural heterogeneity was highest in the shorter sward (Fig. 3), as well as species richness and diversity because of the reduced dominance of A. lateralis and increased participation of prostate species. This supports the idea that as long as grazing intensity creates spatial heterogeneity, species diversity can increase. The greater grazing intensity used in our study was not sufficient to reduce heterogeneity, species richness or diversity.

Conclusions
The forage production of a natural grassland dominated by A. lateralis in the Highlands of the Atlantic Forest proved to be remarkably stable over two years across a range of grazing intensities. Conversely, botanical composition varied markedly, with lesser richness and diversity, and higher cover of A. lateralis, as the grazing intensity became more lax and thus the canopy taller. Maximal diversity, as well as structural heterogeneity, were found at the highest grazing intensity and thus shorter canopy (12 cm treatment). A. lateralis tussocks were smaller and somewhat closer apart in this treatment, representing less than half of the forage mass. Under lax grazing, A. lateralis tussocks were large, and had accounted for over 75% of the standing biomass. This evidences the great capacity of these natural grasslands to adapt their structure over a great range of grazing intensities by changing both its species composition and functionality of the dominant species. Thus, grazing regime can be managed to increase the diversity of plant species while maintaining forage production of these natural grasslands by keeping canopy heights below 28 cm.

Material and methods
This manuscript reports experiments with plants and animals and complies with national and institutional guidelines. All procedures involving animals in this study were approved by the Animal Ethics Committee of the Santa Catarina State University (6,241,030,918). We state also that all plant species used in this experiment is native and endemic in the region of study.
Area, experimental design, and management. The 37 . The experimental area was homogeneous in relation to the type of soil, with a slope of 7.5% + − 2.3, and the soil is classified as Humudept (with an umbric epipedon) (Soil Taxonomy). Soil fertility parameters were measured using samples taken from the upper 0-10 cm layer and were as follows: pH (SMP) = 4.8, P = 3.4 mg L −1 , K = 114 mg L −1 , organic matter (OM) = 5%, Al = 3.1 cmol c L −1 , Ca = 4.3 cmol c L −1 , and Mg = 2.1 cmol c L −1 . Historical climatic data and data from the experimental period were obtained from a weather station located 300 m from the study area (Fig. 4).
The experimental area was also homogeneous in relation to vegetation cover, which consisted of natural grassland predominated by A. lateralis 38 , with no history of anthropogenic use (ploughing, soil fertilization, fire, or species introductions) in the previous 40 years. Prior to the study, the experimental site was grazed by beef cattle, with no defined canopy height management target. The 14,000 m 2 experimental area was divided into 16 paddocks, each 875 m 2 .
First grazing was carried out on 19 August 2015 for homogenization; the animals remained in the paddocks for three days and reduced the canopy height to 11.7 cm. This date was set as the start of the experimental period and lasted until October 2017. Flanders Red (La Rouge Flamande) dry cows were used for grazing, each weighing approximately 613 ± 92 kg.
The experiment was conducted in a randomized complete block design with four treatments and four replications. To test our hypothesis four pre-grazing canopy heights were chosen (12,20,28, and 36 cm), based on measuring only the predominant species (A. lateralis). Such canopy heights were defined to create contrasting grazing intensities where taller canopy height (36 cm) corresponded to a less intense or a lax grazing regime and the lower one (12 cm) a very intense or hard grazing regime. The intermittent stocking method was used 39 , with irregular rest periods (determined by canopy height). The criterion for grazing interruption was defined as a reduction in the pre-grazing canopy height by 40% (7.2, 12, 16.8, and 21.6 cm). This management target was set to allow grazing animals to access almost exclusively leaves, since approximately 90% of all stem components are concentrated in the lower half-stratum of the canopy 40,41 . Vegetation cover and tussock dynamics of A. lateralis. The percentages of A. lateralis vegetation cover and tussock dynamics were evaluated in pre-grazing conditions at the beginning (October to November) and end (April to May) of the growing season (spring and autumn, respectively) for two consecutive years. The first of these, in October-November 2015, was characterized as the initial evaluation because no treatment effect had yet been observed. To assess the percentage of A. lateralis vegetation cover, four transects of 25 m (total 100 m) per plot were established, along which tussock size and gaps (the spaces between tussocks or tussock groups) were measured 44 . Twelve tussocks from the four transects were marked so that changes in base circumference, height, and crown projection diameter could be monitored, enabling the tussock volume to be calculated according to the following equation 45 : where V is volume, h is distance from soil to the top of tussock, r is radius of basal area occupied by tussock, and R is radius of area at top of tussock canopy. The average distance between tussocks was also evaluated according to the Point-Centred Quarter Method 46 , by measuring the distances from the four nearest tussocks to the reference tussock, one in each quadrant, thus allowing A. lateralis space occupancy to be more accurately estimated.
Floristic composition. A floristic composition survey was carried out for two consecutive years, at the beginning and end of the growing seasons under pre-grazing conditions, to facilitate the identification of spring and autumn flowering species (winter and summer species, respectively). This procedure was adopted V = 1/3�.h r 2 + rR + R 2  www.nature.com/scientificreports/ since the type of grassland studied here is characterized by a rare association of C3 and C4 species coexisting simultaneously 2 . Four squares of 0.25 m 2 (0.5 × 0.5 m) were established at the centre of each plot and same transects as used in the tussock evaluations, totalling 4 m 2 and 16 samples per plot (sampling unit). Floristic surveys were repeated in the same sampling units for each evaluation to verify the vegetation patterns and dynamics of the grassland community, using the Botanical analysis method (Botanal) 47 ; this method represents the relationship between the composition of species in the area and their participation in the forage mass, expressed on a dry weight basis. Each sample was ranked according to participation of the most frequent species in the forage mass. Then, other species present within the sample squares were identified and attributed a notional percentage (1%), since they made no significant contribution to forage mass. Forage mass in the Botanal samples was estimated visually by two trained evaluators and then calibrated through a regression equation using forage mass samples (kg DM ha −1 estimated per kg DM −1 obtained from cutting samples) from locations adjacent to the sampling plots after each evaluation period 48 . Five pasture height readings were recorded per sample using a sward stick 42 .
Vegetation community and heterogeneity indices. Plant community indices were calculated using the vegan 49 and BiodiversityR 50 packages in R statistical software 51 . The species accumulation curve was generated using the Specaccum function of the Vegan package and normalized by considering the number of species identified in the first evaluation. The rank consistency index (Cr) refers to the relative variation of species between treatments in the final two surveys 52 . Richness was estimated by counting the number of species, and it was estimated diversity (Shannon index, H′), and dominance (1-Simpson × 100 index) based on the proportion of species in the forage mass. Structural heterogeneity was calculated based on the pasture height of Botanal samples, which was expressed as a percentage variation of the mean height (coefficient of variation for canopy height).
Statistical data analysis. Data analysis was performed using R environment 51 . The variables were adjusted by linear mixed models using the lme4 package 53 , considering block as a random effect, the canopy height, grazin condition (pré and post-grazing) and season or year of evaluation as fixed effects and different variance and covariance structures. The first measurement was used as a covariable for A. lateralis vegetation cover and tussock dynamics as well as for vegetation indices. The model selection was based on Akaike Information Criterion (AIC). Principal coordinate analysis (PCoA) was performed with the vegdist function of the vegan 49 package using the Bray-Curtis method, the classification of species richness and structural heterogeneity at plot level (4 m 2 ) were fitted and ellipses for the treatments were added.

Data availability
The datasets generated during this study are available from figshare 54