The presence of sheep leads to increases in plant diversity and reductions in the impact of deer on heather


Correspondence author. E-mail:


1. Grazing has been implicated in the decline of heather-dominated moorlands in Britain, but there has been little consideration of the effects of different species of herbivores on plant diversity. Sheep stocks have recently decreased in Scotland, and quantification of how this affects biodiversity is essential for understanding how different grazing regimes modify upland habitats.

2. We investigated the effects of grazing on plant biodiversity in heather/grass mosaics at 16 upland sites in Scotland. Red deer Cervus elaphus L. were present at all sites, but sheep Ovis aries L. had been removed from half of the sites. Our experimental design incorporated replication at three spatial scales, from a landscape level down to 10 × 10 m plots.

3. We quantified the relative effects of different herbivore species, vegetation structure and rainfall on heather utilisation, species richness and evenness (alpha diversity; Shannon–Wiener index) and beta diversity.

4. At all spatial scales, deer dung counts were higher and heather was shorter when sheep were absent. Furthermore, utilisation of heather was positively correlated with the amount of deer dung, the amount of grass present and smooth grass height.

5. Alpha diversity was consistently positively correlated with the relative amount of grass, but was also positively related to the amount of sheep dung at the largest spatial scale. At the finest scale, alpha diversity was negatively correlated with the amount of deer dung. Beta diversity was higher when sheep were present at all scales.

6.Synthesis and applications. Mixed grazing by sheep and deer appears to be beneficial for increasing both alpha and beta diversity and minimizing damage to heather in the uplands. The absence of sheep is likely to result in expanding deer populations and greater impact on heather. Management of grazing herbivores is an important tool for maintaining biodiversity in many ecosystems. Our results indicate that reducing livestock may alter the impacts of wild grazers on their habitats and drive changes in diversity, whereas mixed grazing can enhance habitat quality and maintain plant diversity. Therefore, this effect should be considered when devising policy recommendations.


Upland heather moorland is an internationally important resource, providing conservation and economic value, as well as food and habitat for many species (Thompson et al. 1995; Hester & Baillie 1998). The UK and Ireland possess the majority of Europe’s upland dwarf shrub heath, but coverage in Scotland has declined considerably over the past 60 years (Thompson et al. 1995). This has largely been attributed to changing agricultural practices, such as afforestation, increasing intensity of muirburn, expansion of deer populations and livestock farming (Hester et al. 1996; Hester & Baillie 1998; Albon et al. 2007). Consequently, upland dwarf shrub heath has been listed as a priority habitat under the UK Biodiversity Action Plan (UKBAP 2007) and the EC Habitats Directive (92/43/EEC), which recognizes the conservation significance of particular dwarf shrub and bryophyte species, as well as invertebrates and birds that use this habitat (Pakeman et al. 2003).

The importance of biodiversity in the functioning of ecosystems and the provision of ecosystem services is widely recognized, but not well quantified. Although the effects of large grazing herbivores on plant diversity have been studied in a range of habitats (Metzger et al. 2005; Dorrough et al. 2007; Kohyani et al. 2008; Ravolainen et al. 2010), moorland conservation has tended to focus specifically on damage to heather Calluna vulgaris (L.) Hull, largely ignoring plant diversity. Despite this, graminoids are relatively speciose and there is evidence of a role for grazing herbivores in maintaining this diversity (Olff & Ritchie 1998). Thus, to enhance our understanding of how grazing herbivores interact with and modify upland habitats, there is a need for information quantifying grazing impacts on plant diversity in addition to measures of heather condition.

Grazing presents a conundrum for ecologists and land managers, and there has been much discussion about the potential effects of under- or over-grazing on the economic value of land and biodiversity (Evans et al. 2006; Hanley et al. 2008; Fensham, Fairfax & Dwyer 2010). Although native herbivores and domesticated grazers at low densities can increase plant diversity, high densities of stock may be detrimental (Welch & Scott 1995; Olff & Ritchie 1998). There is clear evidence that ungulates directly impact on dwarf shrub heath, blanket bog and montane plants at high densities through grazing, trampling and dung (Clarke, Welch & Gordon 1995a; Hester & Baillie 1998; Albon et al. 2007) and facilitate the succession of heath to graminoid-dominated communities (Ratcliffe 1959; Welch & Scott 1995). Sheep Ovis aries L. and red deer Cervus elaphus L. are the predominant large grazing herbivores in the Scottish uplands. There has been a strong focus on competitive grazing interactions between these species, particularly their relative impacts on heather (Clarke, Welch & Gordon 1995b; Hester & Baillie 1998; Palmer & Hester 2000). Albon et al. (2007) demonstrated that sheep had a more detrimental impact on heather than did deer in the Scottish uplands, presumably because they aggregate in large flocks and trample heather. In comparison, smaller deer population sizes meant that their effects were more diffuse.

There is increasing pressure to reduce livestock grazing either to benefit wild herbivores or to improve the habitats in which they graze. For example, livestock reduction in a wildlife reserve in India has led to massive expansions of wild herbivores (Madhusudan 2004). However, in many protected areas it is recognized that grazing plays an important role in maintaining plant diversity (English National Park Authorities Association 2007). Therefore, understanding the consequences of changing livestock densities for wild herbivores and their habitats is important when assessing the impact of current and future policy scenarios. The intensity of sheep farming in Scotland increased after the Second World War, reaching a peak in the 1990s (Albon et al. 2007). However, stocks declined by roughly 22 % from 1997 to 2008 and are predicted to continue decreasing (Thomson & Midgley 2009). This has been attributed to the 2001 foot-and-mouth outbreak and reforms to the EU Common Agricultural Policy. This decline in sheep numbers has not been uniform, and the most significant reductions have occurred in the north and west (Thomson & Midgley 2009). It has been demonstrated that the exclusion or removal of herbivores alters species composition and competitive relationships within plant communities, with a decreased persistence of graminoids and an increase in heather (Hope et al. 1996; Hartley & Mitchell 2005). However, it is unclear how the decline of upland sheep farming will affect the diversity of plant communities across the landscape.

The direction and magnitude of changes to vegetation communities and deer populations in response to the absence of sheep will depend on the recovery potential of degraded vegetation, as well as the behavioural responses of deer. It is possible that deer will move into preferred grassy patches vacated by sheep, thereby facilitating the spread of graminoids and preventing the recovery of surrounding heather (Albon et al. 2007). The effect of changing grazing regimes on moorland plant communities is also likely to vary across spatial scales in response to factors such as rainfall or soil quality (Hester & Baillie 1998; Olff & Ritchie 1998; Hartley & Mitchell 2005). The importance of considering spatial variation in vegetation structure has been emphasized when measuring grazing impacts (Palmer & Hester 2000; Palmer et al. 2003; Gordon, Hester & Festa-Bianchet 2004) but is also crucial to studies of biodiversity (Olff & Ritchie 1998; Dufour et al. 2006; Dorrough et al. 2007; Ravolainen et al. 2010) and must be considered in grazing management.

van der Wal (2006) argued that the impact of heavy grazing should not automatically be regarded negatively as habitat degradation, but instead may be viewed as an ecosystem shift from one dynamic state to another. Thus, our aim was to disentangle the relative impacts of sheep and deer on plant diversity, to determine the effects of past sheep removal on upland habitats. We investigated the variation in upland plant diversity in relation to both the assemblage and the abundance of large grazing herbivores across a range of spatial scales, to determine whether deer populations and plant biodiversity indices (alpha and beta diversity) differ depending on whether sheep are present or absent.

Materials and methods

Site selection

We studied eight pairs of sites (upland hunting estates), ranging in area from 148 to 1611 km2 (Fig. S1, Supporting Information). Six of the pairs were visited in 2007 and again in 2009, while two additional pairs of sites were included in 2009. All sites had previously been subjected to a Rapid Habitat Impact Assessment (RHIA) (Nolan et al. 2002). Sheep were removed from one site within each pair (Table 1), whereas the other was stocked with sheep. Prior to removal, sheep were maintained at commercial stocking rates (Table 1). Sheep were reduced at two of the mixed grazing sites, but remained at commercial densities. All sites supported populations of red deer (Table 1). During 2008, sheep were re-introduced and a deer-proof fence erected at one of the sites that was ‘sheep-free’ in 2007, so a replacement site was chosen for the 2009 survey and the 2007 data excluded from the analyses for this pair.

Table 1.   Summary of estimated grazing regimes and changes to sheep stocks at all sites
SitePairGrazing regimeYears of studySheepDeer
No. pre-removalRemoval yearNo. post-removalDeer countYear
11Deer2009750  199509152006/2007
21Deer/sheep2009900 90020162006/2007
42Deer/sheep2007/2009400 4003022006/2007
53Deer2007/2009200Late 1990s011622005
63Deer/sheep2007/2009350 3501202005
74Deer2007/2009600  199301002005
84Deer/sheep2007/2009800 800142007
95Deer2007/2009500c. 1990011952005
105Deer/sheep2007/2009700  200150021002005
116Deer2007/2009860  20020N/A 
126Deer/sheep2007/2009400  2002250N/A 
137Deer20091000  19850N/A 
147Deer/sheep2009N/A 1000–1500N/A 
158Deer2009N/A  1960/196106452008
168Deer/sheep2009N/A  4506452008

Within each site, we selected between two and four 0·25 × 0·25 km square (6·25 ha each; referred to as ‘0·25 km square’), which overlapped with areas previously surveyed by Nolan et al. (2002). The square were stratified in relation to the heather to grass ratio, and we only selected those comprising at least 50 % heather. We established three 10 × 10 m plots in each square, except at one pair of sites which had five plots in each, giving between six and twelve 10 × 10 m plots per site. These 10 × 10 m plots were selected at random within the 0·25 × 0·25 km square and formed the sampling unit for the study. As far as possible, we selected sites to minimize overall variation in factors such as altitude, aspect, slope, geology, climate and deer and red grouse Lagopus lagopus scotica Latham management. We identified potential sites using information from the RHIA data, a digital elevation model, an index of topographic exposure (TOPEX) and density data obtained from the agricultural census and the Deer Commission for Scotland. We visited all plots twice in each year, during late winter/spring (March–May) and again in mid-summer (June–August).

We obtained precipitation data from the Met Office and used an interpolation method, taking into account factors such as altitude, to estimate local total monthly and annual precipitation at each plot.

Estimation of herbivore presence and abundance

We used standing-crop dung counts to provide an index of herbivore presence and an estimation of density. We systematically searched a 1 × 10 m strip around the outside of each edge of the 10 × 10 m plots (i.e. 40 m2 per plot) and recorded the type of dung and number of faecal pellet clusters. We defined a pellet cluster as being a group of six or more pellets from sheep, red deer or roe deer Capreolus capreolus L. or a single cowpat. Faeces from rabbits Oryctolagus cuniculus L., mountain hares Lepus timidus L., voles and red grouse were counted individually. We carried out the dung surveys during both sampling periods in each year.

Vegetation measures

We walked a rough ‘W’-shaped transect across each 10 × 10 m plot and measured vegetation heights to the nearest centimetre at 40 points, spaced roughly evenly along the transect. We recorded the vegetation type at each point and calculated the mean height of heather (ericaceous species) and grasses in each plot from these 40 measurements. We categorized the graminoids as being either ‘smooth’ or ‘coarse’, following the habitat descriptions given by MacDonald et al. (1998) and calculated the mean heights of each group separately. Coarse graminoids included highly lignified and low-quality tussocky grasses or sedges, e.g. Nardus stricta L., Molinia caerulea (L.) Moench and Eriophorum vaginatum L., whereas the smooth graminoids included softer, narrow-leaved and more nutritious species, such as Agrostis, Festuca and Holcus. We measured vegetation heights in the first sampling period of 2007 and both periods in 2009. For consistency, we only included the data from the first survey of 2009 in our final analyses; however, there was no difference in the effects of grass or heather height on alpha or beta diversity between the two 2009 surveys. We estimated the percentage cover of heather, smooth grass and coarse grass based on the number of times each vegetation class was measured out of the 40 sampled points, to give a representative indication of vegetation cover across the whole plot. Where multiple plant species were present at a sampling point, the maximum height of each vegetation class was recorded.

We calculated a heather utilisation score for each 10 × 10 m plot, using a 20 cm pin frame quadrat, with five pins spaced 4 cm apart. We threw the frame randomly, and when it landed on at least one heather plant, we recorded the level of browsing on each live shoot that a pin touched. We repeated this procedure up to 20 times, until at least 50 shoots had been sampled. We assigned the level of browsing on each shoot to one of four categories by comparing the length of browsed and unbrowsed shoots: N0 = not browsed, N1 = browsed to less than half its length, N2 = browsed to more than half its length, and N3 = browsed into the previous season’s growth (Grant, Hamilton & Souter 1981). We summed the number of shoots in each category within the 50 sampled shoots and calculated heather utilisation according to the following equation developed by Grant, Hamilton & Souter (1981):


We measured heather utilisation in the winter/spring sampling period only, so that the previous year’s utilisation could be measured before the next growing season.

Biodiversity indicators

We estimated the percentage cover of all species of vascular plants in a 2 × 2 m quadrat in the south-west corner of each 10 × 10 m plot in the second visit to each site. We determined alpha diversity according to the Shannon–Wiener index using the Vegan package (Oksanen et al. 2009) in R (R Core Development Team 2008).

We performed de-trended correspondence analysis using the decorana function in the Vegan package (Oksanen et al. 2009) to create a matrix of the vegetation distance between each pair of 10 × 10 m plots within each 0·25 km square, based on the presence and abundance of species in each plot. We then calculated the average distance in multivariate space between all plots to derive an index of beta diversity for each 0·25 km square.

Statistical analysis

We carried out all statistical analyses using the program R (R Core Development Team 2008). We summed the dung count data for the two surveys in each year and used mean vegetation height measurements and percentage cover data collected in the first sampling period of each year. We took the mean of all variables for each of the 10 × 10 m plots per 0·25 km square and, similarly, averaged all 0·25 km square per site. We calculated the heather utilisation and alpha and beta diversity scores from the data for each 10 × 10 m plot and took the mean for each of the larger spatial scales.

We used linear mixed effects models with pair as a random term in the nlme library (Pinheiro et al. 2009), to test the effects of herbivore, habitat and environmental variables on grazing impacts, alpha and beta diversity. We expected that measures of grazing impact and biodiversity were likely to show some degree of spatial autocorrelation within and between sites, so we included an exponential spatial correlation term in our models. At the site scale, the spatial term consisted of the site coordinates nested within year. At the 0·25 km square scale, the coordinates of the square were nested within a site by year term, while at the 10 × 10 m scale the coordinates of each plot were nested within a 0·25 km square by year term. Our models included the following ecological and environmental terms as fixed explanatory variables. Herbivores: presence/absence of sheep, herbivore dung counts and number of years since sheep removal; habitat: heather utilisation, grass height, heather height, percentage of heather, percentage of grass (total, smooth and coarse) and the ratio of smooth grass to coarse grass; environment: altitude and total annual precipitation.

We used data from both years (2007 and 2009) and included year as both a fixed term and an interaction term, but where these were not significant we dropped them from the final model. We chose a priori the explanatory variables that we considered appropriate and tested these first in univariate models. We then combined all significant terms in a model and sequentially dropped terms that were not significant at α = 0·05. We also tested for interactions between the remaining terms where appropriate. Where alternative models were produced, we selected the final model based on comparison of Akaike’s Information Criterion scores. We repeated this process separately for each of the three spatial scales.


For simplicity, we present only the models and figures for data at the site level, as this is the unit at which management occurs. Models and figures relating to the 0·25 km square and 10 × 10 m and scales are presented in Supporting Information.

Effects of removing sheep on deer density

There was more deer dung present when sheep were absent at the site scale (t17 = −3·28; P = 0·004; Fig. 1a), at the 0·25 km square (t62 = −4·20; P < 0·001; Fig. S2a, Supporting Information) and at the 10 × 10 m scale (t220 = −5·51; P < 0·001; Fig. S3a, Supporting Information). In addition, the amount of deer dung was positively correlated with the length of time since sheep removal at the site scale (t17 = 2·11 P = 0·05), the 0·25 km square scale (t62 = 2·85; P = 0·006) and the 10 × 10 m scale (t220 = 4·21; P < 0·001).

Figure 1.

 Linear mixed effects models of the effect of sheep presence on (a) deer dung count and (b) mean heather height (cm) at the site scale. Closed circles indicate sites with deer only; open circles indicate sites with both deer and sheep.

Grazing impacts

At the site level, heather utilisation was positively correlated with the amount of deer dung, the percentage of grass and mean smooth grass height (Table 2, Fig. 2). The results were essentially the same at the 0·25 km square (Table S1, Fig. S4, Supporting Information) and 10 × 10 m scales (Table S2, Fig. S5, Supporting Information), although the amount of sheep dung was also significantly positively correlated with heather utilisation. At the 10 × 10 m scale, heather utilisation was negatively correlated with the percentage of heather, but within the heather/grass mosaic, the amount of heather was inversely related to the amount of grass, consistent with the results from the larger spatial scales. There was also a year effect at the site and 10 × 10 m scales, with lower heather utilisation in 2009 than in 2007. Length of time since sheep removal had no effect on grazing impacts at any spatial scale.

Table 2.   Summary of linear mixed effects models explaining variation in heather utilisation alpha diversity and beta diversity at the site scale
ModelVariableEffect sizetd.f.PAIC
  1. AIC, Akaike’s Information Criterion.

Heather utilisation (%)Deer dung count0·176 ± 0·0642·77140·015170
Amount of grass (%)0·254 ± 0·1132·25140·041
Mean smooth grass height (cm)1·459 ± 0·3973·68140·003
Year−3·412 ± 0·941−3·62140·003
Alpha diversity
 Model 1Sheep dung count0·017 ± 0·0053·08160·00712·3
Amount of grass (%)0·016 ± 0·0044·6216<0·001
 Model 2Deer dung count−0·005 ± 0·002−2·10160·05216·97
Amount of grass (%)0·015 ± 0·0043·77160·002
Beta diversity
 Model 1Sheep presence0·184 ± 0·0712·61170·018−3·62
 Model 2Heather height (cm)0·023 ± 0·0063·57170·002−2·22
 Heather heightSheep presence4·952 ± 1·6872·93170·009143
 Smooth grass: coarse grass ratioSheep presence0·413 ± 0·1742·372170·03050·4
Sheep dung count0·055 ± 0·0163·516170·00350·4
 Smooth grass (%)Sheep presence4·360 ± 0·8305·2517<0·001130
Sheep dung count0·420 ± 0·9454·4417<0·001138
Figure 2.

 Linear mixed models showing the effects of (a) deer dung count, (b) the amount of grass (%) and (c) mean smooth grass height (cm) on heather utilisation at the site scale. Closed circles indicate sites with deer only; open circles indicate sites with both deer and sheep. Solid line indicates the model for 2007 data only; dashed line indicates 2009 data.

Heather was taller when sheep were present at the site scale (Table 2, Fig. 1b), at the 0.25 km square scale (Table S1, Fig. S2b, Supporting Information) and the 10 × 10 m scale (Table S2, Fig. S3b, Supporting Information), but we found no effect of herbivores on mean grass height at any scale.

We could only measure the ratio of smooth grass to coarse grass at the site level, as both grass types were not always present at the finer scales. This ratio was positively correlated with both the presence of sheep and the amount of sheep dung at the site scale (Table S1 Supporting Information). Similarly, the amount of smooth grass was positively correlated with the presence of sheep and with the amount of sheep dung at all three spatial scales (Table 2, Tables S1 and S2 Supporting Information), while there was a concomitant decrease in the amount of coarse grass when sheep were present at the 0·25 km square (Table S1 Supporting Information) and the 10 × 10 m scales (Table S2, Supporting Information). The amount of heather was negatively correlated with the amount of sheep dung at the finest spatial scale (Table S2, Supporting Information).

Alpha diversity (Shannon–Wiener index)

At the site level, alpha diversity was positively correlated with both the percentage of grass and the amount of sheep dung (Table 2, Fig. 3). We also observed a trend for alpha diversity to decrease in relation to the amount of deer dung in a combined model with percentage of grass. At the 0·25 km square scale, only the percentage of grass appeared important (Table S1, Fig. S6, Supporting Information), whereas the results at the 10 × 10 m scale (Table S2, Fig. S7, Supporting Information) were more similar to the site scale, except that a negative effect of the amount of deer dung replaced sheep dung. Length of time since sheep removal had no effect on grazing impacts at any spatial scale.

Figure 3.

 Linear mixed models showing the effects of (a) sheep dung count and (b) the amount of grass (%) on alpha diversity at the site scale. Closed circles indicate sites with deer only; open circles indicate sites with both deer and sheep.

Beta diversity

At the site level, beta diversity was higher at sites where sheep were present, although an alternative model showed that beta diversity was positively correlated with mean heather height (Table 2, Fig. 4). Similarly, at the 0.25 km square scale, alternative models showed that sheep presence or mean heather height were positively correlated with beta diversity (Table S1, Fig. S8, Supporting Information). The length of time since sheep removal had no effect on grazing impacts at any spatial scale.

Figure 4.

 Linear mixed models showing the effects of (a) sheep removal and (b) mean heather height (cm) on beta diversity at the site scale. Closed circles indicate sites with deer only; open circles indicate sites with both deer and sheep.


Our results demonstrate differing impacts of deer and sheep on heather utilisation and vascular plant diversity in the Scottish uplands, which are summarized in Fig. 5. The absence of sheep was correlated with a higher density of deer, so reductions in sheep grazing may not necessarily facilitate regeneration in degraded areas as has been predicted (Pakeman et al. 2003). However, we also demonstrated that alpha diversity was positively correlated with sheep density at the site level and that beta diversity was higher when sheep were present at both site and 0.25 km square scales, although we found no effect of time since sheep removal. However, given that for the majority of pairs, sheep had been absent for more than 10 years, it is not surprising that we could not detect an incremental relationship. Thus, although sheep are often associated with the degradation of upland plant communities (Albon et al. 2007), our results suggest that a mixed grazing regime increases upland plant diversity, compared to deer-only grazing, and may keep deer populations at lower densities, at least in the short-term, potentially reducing heather utilisation.

Figure 5.

 Summary of the effects of different grazing regimes on heather utilisation and plant diversity.

The amount of deer dung was higher on sheep-free sites and was positively correlated with the time elapsed since sheep removal. Although dung counts can only provide an index of relative herbivore abundance and decay rates may differ between sites, this pattern supports suggestions that the absence of sheep is associated with larger wild deer populations (Clutton-Brock & Albon 1989; Albon et al. 2007). It appears that sheep competitively exclude deer within heather-grass mosaics and deer populations are more diffuse when sheep are present (Osborne 1984; Clarke, Welch & Gordon 1995a). Although both species preferentially feed on grass, deer tend to spend more time higher on the slopes and consequently in heather when grazing sympatrically with sheep (Osborne 1984; Hester et al. 1999). Thus, the short-term effect of changing agricultural practices on animal population dynamics is likely to have important consequences for the health of upland plant communities, as heather utilization scores were positively related to the amount of deer dung at all scales.

Heather utilisation was also independently positively correlated with both the percentage cover of grass and smooth grass height. This suggests that herbivores are attracted by the availability of preferred grass species and subsequently graze on the surrounding heather (Clarke, Welch & Gordon 1995a; Palmer et al. 2003). Even if deer spend more time feeding in grass patches when sheep are absent, the fragmented vegetation structure means that higher deer densities will lead to greater impacts on heather, particularly along the grass/heather boundary (Clarke, Welch & Gordon 1995b; Hester & Baillie 1998; Palmer et al. 2003; Gordon, Hester & Festa-Bianchet 2004). Although heather was taller when sheep were present, there was no effect of herbivore abundance or identity on grass height. This contradicts previous studies that have demonstrated an increase in vegetation height in response to the exclusion of herbivores, or the selective removal of sheep (Hope et al. 1996; Hartley & Mitchell 2005). Thus, it appears that deer may be able to replace sheep by maintaining short grass patches.

A key result was that alpha diversity was positively correlated with the amount of sheep dung at the site scale. Similarly, we found a trend for a reduction in alpha diversity with increasing amounts of deer dung at the site level, which was significant at the finest scale, suggesting that deer and sheep have opposing effects on alpha diversity. Nevertheless, a consistent result across all scales was a strong positive correlation between the percentage cover of grass and alpha diversity. This relationship was not unexpected, as grasslands are relatively more species rich than upland heath, which tends to be dominated by a single species, C. vulgaris. Large grazing herbivores are capable of manipulating the vegetation structure of their habitats by influencing the dominance of particular species or communities of plants (Gordon, Hester & Festa-Bianchet 2004). In the absence of grazing, coarse grasses such as M. caerulea often competitively encroach, resulting in a build-up of dead tussocks that excludes other plant species (Hill, Evans & Bell 1992; Hulme et al. 1999; Marrs et al. 2004). Differences in grazing behaviour mean that sheep are more efficient than deer at maintaining short grass, and indeed, we found that both the smooth grass to coarse grass ratio and the cover of smooth grass were positively correlated with the presence of sheep and with the amount of sheep dung, while coarse grass cover was lower when sheep were present at the finer scales. This suggests that sheep increase alpha diversity by reducing the biomass of dominant coarse grasses, thus enabling the persistence of a more species-rich mixture of other graminoids.

The absence of sheep was also negatively correlated with beta diversity at both spatial scales, lending further support to our hypothesis that sheep reduce the dominance of coarse grasses, thereby increasing the diversity. We found no direct relationship between grass height and beta diversity, which supports trends observed in other ecosystems (Metzger et al. 2005; Fensham, Fairfax & Dwyer 2010). In contrast, we did find that heather height was positively correlated with beta diversity at both spatial scales, which suggests a role for heather in restricting the dominance of grasses. However, we also found that heather was shorter on the sheep-free sites, which is likely to be due to greater utilisation associated with larger populations of deer. This suggests that the effects of sheep and heather height on beta diversity may be inter-related and provides further evidence for the role of sheep in maintaining beta diversity in the uplands.

Previous studies investigating the impacts of sheep and deer grazing in the Scottish uplands have tended to focus at either a very fine scale or more crudely across the landscape. However, there is often a discrepancy between the scale at which diversity is quantified and the scale at which herbivores are managed (Ravolainen et al. 2010). A strength of our study was that although we focussed primarily on the scale at which grazing herbivores are managed, we considered a range of spatial scales, replicated both within and between sites. Therefore, we have considerable statistical power to detect the differing effects of herbivory, habitat and environmental variables on plant biodiversity at both a local and a regional scale. Palmer et al. (2004) stressed the importance of spatial scale and reported that variance components of grazing impacts on heather decrease as spatial scale increases (Palmer & Hester 2000). Our results confirm this, but we found less variation in the determinants of biodiversity indicators. Nevertheless, our results emphasize the importance of spatial scale and indicate that the application of blanket management policies may not be appropriate throughout the uplands.

Most studies that have investigated changes in livestock numbers have focussed on the consequences for the wild herbivore populations (e.g. Madhusudan 2004). Where the effect on plant communities has been investigated, such as in the Australian desert (Fensham, Fairfax & Dwyer 2010) or the short grass plains of the Serengeti (Metzger et al. 2005), no overall effects of grazing on plant diversity have been detected. However, the key to our findings was the demonstration that different species of grazing herbivore can have differing impacts on plant diversity. Grazing by sheep is often considered to be in opposition to moorland conservation, and it may be argued that the positive effects of sheep on alpha and beta diversity are undesirable, as they reflect an increasing dominance of grass species at the expense of heather. However, the observed reduction in beta diversity when sheep were absent indicates that sheep may play an important role in maintaining species diversity in upland ecosystems by preventing the build-up of rank coarse grasses. Moreover, we present strong evidence that red deer abundances were higher on sites without sheep, and furthermore, this was correlated with a greater utilisation of heather on sheep-free sites. Thus, our results clearly suggest that the absence of sheep across large areas of the uplands may be detrimental for both the vascular plant diversity and the conservation of heather. These results have practical implications for the conservation of plant diversity (Fig. 5). In situations where deer are the main grazer, heather condition may be compromised by their increasing impact, while graminoid areas may become under-grazed, reducing diversity. Sheep-only grazing may have the reverse effect, leading to rank heather and increased graminoid diversity. However, under mixed grazing, deer numbers will be lower, thereby reducing the impact on heather, yet sheep could maintain plant diversity. In protected areas, managers recognize the importance of appropriate grazing pressure as a tool to maintain biodiversity. Our work shows that the effect of reducing livestock numbers needs to take into account the resulting changes in wild herbivores and the impact they might have. Furthermore, mixed grazing can be beneficial depending on the policy objectives for the habitat and its plant diversity. Extending the scope of this study to determine how changing grazing regimes affect both the abundance and the assemblages of other animals such as invertebrates, small mammals and birds and their potential interactions with plant communities is an obvious next step.


We thank Baptiste Algoet, Martin Duncan, Ingolf Kittel, Marie Mourtier and Andrew Thompson for assistance with fieldwork; Mark Brewer, Colin Beale and Thomas Cornulier for statistical advice and Jim McLeod for providing the rainfall data. We are grateful to the estate owners and managers for providing information and access. Ruth Mitchell, Steve Palmer and Ben Moore provided advice and valuable comments on earlier versions of this manuscript. This study was supported by funding from the Scottish Government.