2.1 Study region

We selected three different regions, characterized by arid, semi-arid and sub-humid climatic regimes in northeastern Iran in the Khorassan-Kopet Dagh floristic province of the Irano-Turanian region (see Figure 1 and Appendix S1 for more details).

Khajeh Kalat (the driest region in our study system) is located in the eastern Kopet Dagh and classified as a subtropical semi-savanna (Karimi et al., 2009). The area has a dry climate (based on the de Martonne aridity index). The mean annual precipitation (20-year mean) is 255 mm. Most of the flora consists of ephemerals and/or short life-span perennial grasses and forbs. The dominant species is Pistacia vera, accompanied by Artemisia kopetdaghensis, Poa bulbosa, Amygdalus spinosissima, Bunium persicum, Ephedra foliata, Zygophyllum atriplicoides etc. (Ahani et al., 2013).

Baharkish rangeland is a semi-arid region in the eastern part of Khorassan-Kopet Dagh. The mean annual precipitation (20-year mean) is 354.4 mm. The main vegetation type is Artemisia steppe in mid elevations, thorn-cushion communities with Acantholimon erinaceum, Acantholimon raddeanum, Acanthophyllum glandulosum and Onobrychis cornuta, mixed with grasses of mountain steppes (Festuca valesiaca and Stipa sp.) and also with dwarf shrubs (Ephedra intermedia, Hymenocrater sp. div. and Rosa sp. div.; Memariani et al., 2016a).

Darkesh area represents the sub-humid zone, which is the wettest region within our study system. It is located in the western part of Khorassan-Kopet Dagh. The mean annual precipitation (20-year mean) is 600 mm. Because of topographic and climatic variations, several vegetation types have developed in the area: dense stands of Juniperus polycarpos and Quercus castaneifolia woodlands in high mountains, Acer monspessulanum subsp. turcomanicum and Crataegus sp. in open shrublands, early successional scrubs with Paliurus spina-christi in the foothills, and Festuca valesiaca and Stipa steppes — sometimes combined with other vegetation types such as grassy montane steppes (Memariani et al., 2016b).

2.2 Sampling design

The sampling areas were hierarchically arranged in a nested sampling design, with four hierarchical levels: three climatic regions; 6–7 sampling areas in each climatic region; one intensively grazed and one ungrazed site in each sampling area; and 5 HG and 5 LG plots in each of these sites. The size of the studied regions under each climate was 1,600 ha, 1,035 ha and 2,000 ha for the arid, semi-arid and sub-humid conditions, respectively. The sampling areas were of similar size (3–5 ha) across all three climatic regions. The mean distance between individual sampling areas within each of the three climatic regions was 1.63 (±0.51) km and the minimal distance between two sampling areas was 1 km. The HG plots were exposed to heavy grazing (therefore considered as overgrazed and termed “high-grazing plots”; 3 AUM per ha, 90% sheep and 10% goats – see Appendix S3), whereas the LG plots were located within exclosures (and termed “low-grazing plots”). These had been prevented from grazing for around 35 years in all three regions, even though occassional grazing by goats has been observed (0.5 AUM per ha), as goats can climb the fences and it is difficult to keep them away (for more information on the grazing history, see Appendix S3). In addition to the grazing status given by the history of the exclosure, we also measured grazing intensity as the median number of droppings (56 ± 12 in the HG and 12 ± 6 in the LG plots) and also as the width of horizontal paths of livestock (0.27 ± 0.09 m for the HG plots and 0.04 ± 0.04 m for the LG plots). Within each sampling area, the mean distance between the LG and HG plots was 20 m. The individual HG/LG plots of 1 m × 1 m were placed randomly, in a relatively homogeneous area in terms of topography, land use and vegetation. This sampling scheme certainly does not include the whole species pool of these habitats. However, our aim was not to describe either flora or vegetation of these regions, but rather to perform unbiased comparisons between the HG and LG analogues (see the species–area curves in Appendix S4). Each plot was characterized by its geographic coordinates and altitude. In 2017, the abundance (cover percentage) of all plant species was recorded between April and June, when the growing season peaks in these regions. The species names were standardized according to Flora Iranica.

2.3 Statistical analyses

3.1 Plant species diversity across different grazing intensities and climatic regions

In total, 337 plant species were recorded in all three climatic regions, with 192 species recorded in the sub-humid, 115 in the semi-arid and 137 species in the arid region. No significant differences in mean values of Shannon index (H) of species diversity were detected among the three climatic regions (Table 1). However, the HG plots had lower species diversity (expressed as Shannon H), compared to the LG plots across the three climatic regions (Figure 2a). Further, the climate × grazing interaction was statistically significant and the differences between the HG and LG plots were most pronounced in the sub-humid climatic region (Table 1, Figure 2a).

3.2 Species composition across different grazing intensities and climatic regions

As shown by the tests based on the Bray–Curtis distances (PERMANOVA), the compositional heterogeneity differed significantly between the climatic regions (p = 0.001), as well as between the HG–LG plots (p = 0.001). The climate × grazing interaction was also significant (p = 0.001; Table 2). The effect of grazing was most pronounced in the sub-humid region (Table 2). The ANOSIM also shows that there are differences in the variability explained by the individual terms within the model: the factor climatic region explains more variability in the data (p < 0.001; Figure 2b) than the factor grazing (p = 0.001; Figure 2b) and their interaction grazing × climatic region (arid: p = 0.01, semi-arid: p = 0.57, sub-humid: p = 0.001; Figure 2b — see Table 2). Moreover, comparison of the explanatory power of the terms within the Dissimilarity analysis shows that the overall effect of the climate × grazing interaction was not larger than the sum of the individual main effects of climate or grazing.

3.3 Indicator species of different levels of grazing intensity

As shown by the Indicator Species Analysis, Poa bulbosa, Salsola dendroides, Avena sterilis, Convolvulus arvensis, Polygonum arenastrum, Verbascum cheiranthifolium and Aegilops triuncialis were more likely to be found on the HG plots, and Vicia peregrina, Cymbolaena grifity, Allium rubellum, Thalictrum sultanabadense, Bromus kopetdagensis, Astragalus jolderensis, Dactylis glomerata, Crupina vulgaris, Viola odorata, Tanacetum parthenium, Vicia sativa, Galium tricornutum, Quercus castaneifolia and Lonicera nummulariifolia were typical for the LG plots in all three climatic regions (see Table 3 for details and indicator species specific for particular climatic regions).

4.1 Effects of grazing and climate on species diversity

We detected a negative effect of overgrazing on species diversity under the three different climates, that is, across the whole study system (see also Osem et al., 2002; Erfanzadeh et al., 2015; Herrero-Jáuregui & Oesterheld, 2018). The significant effect of grazing on species diversity across different climatic regions in northeastern Iran contradicts the results obtained by a number of field studies from southern African drylands, which found that grazing has little effect on species diversity (Todd & Hoffman, 2009; Rutherford & Powrie, 2013; Saayman et al., 2016). In a series of fence-line studies, Rutherford and Powrie (2010) reported that intense grazing may even increase species diversity of the grazed communities by increasing evenness. Rutherford et al. (2012) reported an increase in evenness due to the suppression of dominant species that resulted in higher species diversity of heavily grazed communities. However, in our study, the high grazing intensity decreased the species diversity, most likely by eliminating the palatable species. This means that the effect of grazing depends on the type of grazed community even within a single climatic situation and, most importantly, it depends on the intensity of grazing (Tóth et al., 2018). In our study, it is therefore likely that the detected suppressive effect of grazing is due to its high intensity, as overgrazing is known to strongly suppress species diversity. Overgrazing is known to cause a non-random, directional loss of species, as the most palatable species are lost first from the community (see also Milton et al., 1994). In addition to species loss, intensive grazing reduces the above-ground biomass. This is partly because the remaining species cannot fully compensate for the species loss due to grazing (Verón et al., 2011) and also due to the defoliation of even the remaining, grazing-tolerant plants. Furthermore, grazing early in the season represents a serious problem for the Iranian and Middle Eastern rangelands. On the HG sites, grazing starts about one month before the major forage plants have grown up enough to protect the soil, which then gets heavily trampled (Ajorlo, 2005; Haji-Rahimi & Ghaderzadah, 2008). Even though extreme arid conditions are known to suppress diversity (Arroyo et al., 1988), we did not detect any significant differences in species diversity between the three climatic regions (see also Ricklefs et al., 2004; Herrero-Jáuregui & Oesterheld, 2018). This may be due to the fact that the precipitation gradient within our study system does not really include extremely dry conditions, in which the species diversity would be limited by drought stress. Most importantly, even though there are no significant differences in species diversity between the climatic regions, we found a highly significant interaction between the effect of climate and grazing on species diversity, with the most pronounced differences between the HG and LG plots detected in the sub-humid region, representing the wettest area in our study system. Apparently, the suppressive effect of grazing was strongest in the sub-humid conditions, where the diversity is unlikely to be limited by drought. This may be partly due to the high proportion of endemics in this region, which mostly rank among well-palatable species. Further, it is possible that grazing represents the most important limiting factor in conditions that are rich in resources and, therefore, most species are adapted to competition, rather than to the stress and disturbances induced by intensive grazing and trampling (see also Stevens et al., 2012; Lezama & Paruelo, 2016). Given that sub-humid regions receive stable and sufficient precipitation, plant–herbivore interactions may be the main driver of vegetation dynamics (see also Lauenroth & Burke, 2008; Milchunas et al., 2008). In accordance with the non-equilibrium theory (a more dynamic and less predictable perception of ecosystem dynamics accounting for disturbances, spatial heterogeneity, and multiple stable states; Briske et al., 2017), steppes in the arid and semi-arid regions of northeastern Iran differ in their flora and vegetation from the sub-humid region, where most species are adapted to drought and herbivory stress by possessing toxic compounds or spiny leaves and stems. Therefore, the differences between the HG and LG plots are less pronounced in the arid and semi-arid region (Rahmanian et al., 2019).

Interestingly, the decrease in species diversity resulting from overgrazing is not necessarily associated with a decrease in the community functional diversity. However, the effect of overgrazing on species diversity ad functional diversity as well as on the representation of individual functional traits was consistently detected to be the most profound in the subhumid climatic region (Rahmanian et al., 2019).

4.2 Effects of grazing and climate on species composition

Our results confirm the importance of the precipitation gradient in structuring plant communities in northeastern Iran. The heterogeneity in species composition differs significantly between the climatic regions, as well as between the HG and LG plots. The compositional heterogeneity between the HG and LG plots was detected to be largest in the sub-humid region, which is in accordance with the results on species diversity. However, when species diversity decreased due to intensive grazing, compositional heterogeneity increased. This increase may result from the suppression of some species, possibly dominants, by heavy grazing. In a relatively affluent environment, where species can be presumed to be adapted to competition rather than to the stress of aridity, intensive grazing causes massive changes in species composition.

The effect of overgrazing on heterogeneity in species composition is in accordance with the results of Asner et al. (2004), who observed that the increased spatial heterogeneity of vegetation may be an overarching pattern in desertification processes across ecosystems. Intensive grazing may also increase the variation in community composition by promoting the vegetation’s patchiness and heterogeneity in resource distribution (de Bello et al., 2007). Spatial heterogeneity can also arise from localized urine and dung deposition, trampling paths, and erosion rills (Olff & Ritchie, 1998). Further, comparison of the explanatory power of the terms within the Dissimilarity analysis shows that the overall effect of the climate × grazing interaction was not larger than the sum of their individual main effects, suggesting that these two factors interact in antagonistically. Species diversity alone does not adequately reflect shifts in vegetation structure in response to intensive grazing in the northeast of Iran. Our results suggest that it is important to consider shifts in species composition as well to achieve a more complex assessment.

The Indicator Species Analysis revealed species typical for each of the three regions, defined by the climate and precipitation (see also Milchunas et al., 1988; Díaz et al., 2007). The significant main effect of grazing is more interesting compared to the main effect of climate (de Bello et al., 2005), as it suggests that there are species sensitive to grazing that are shared by the three climatic regions,. Quercus castaneifolia and Lonicera nummulariifolia were identified as indicator species of the LG plots, probably because their seedlings and saplings are consumed by grazing animals and can only survive in exclosures. The canopy of these woody species may create a suitable habitat for some herbaceous species, like Dactylis glomerata, Crupina vulgaris or Galium tricornutum. On the other hand, sun flecks and open spaces, caused by the severe grazing, provide an opportunity for ruderal species to establish. Of course, populations of palatable species decline rapidly in grazed habitats (Díaz et al., 2007), being replaced by weedy or ruderal plants (Bullock et al., 2001; Pakeman, 2004), like Poa bulbosa, Aegilops triuncialis and Avena sterilis. These species produce less biomass and are not very attractive for livestock, but compete successfully with other species in the area. Some of the indicator species of the HG plots prefer open areas, such as Convolvulus arvensis and Polygonum arenastrum. They can germinate and establish better on microsites with bare ground, usually during the early successional stages or after trampling by livestock.

The significant interaction of the effects of climate and grazing shows that specific species respond to grazing in each of the three climatic regions. In the arid and semi-arid regions, the species with a opportunistic strategy, like Poa bulbosa and Galium humifusum, were only found in the HG plots and palatabale species, like Vicia sativa and Astragalus jolderensisi, were only found in the LG plots. In the sub-humid region, most species may be competitive rather than tolerant to stress and disturbance and are therefore likely to disappear from the HG plots. This may be the reason that grazing has the most profound effect on species composition in the sub-humid region. It was mostly palatable species that were identified as significant indicators of the LG plots in the sub-humid region, such as Dactylis glomerata, Crupina vulgaris, Viola odorata or Tanacetum parthenium (see Table 3). Unpalatable species, like Verbascum cheiranthifolium, were exlusively found on the HG plots.

Protecting at least some parts of rangelands from overgrazing should help maintain high levels of biodiversity, as well as prevent future economic losses. Most discussions on rangeland degradation have focused on the effects of overgrazing (Eswaran et al., 2001; Asner et al., 2004; Steinfeld et al., 2006). The sustainable management of pastoral ecosystems should seek to find a balance between production and protection of diversity. Such an approach also gives the best chances for sustainable pastoral agriculture.