Effects of Long-Term Removal of Sheep Grazing on the Seedbanks of High-Level Grasslands and Blanket Bogs

Abstract

Many areas of vegetation in the British uplands have reduced species diversity as a result of sheep overgrazing. It has been suggested that abandonment or re-wilding strategies might be used to reverse this. A likely first step would be the removal or reduction of grazing livestock from upland areas, with a presumption that this would lead to a recovery in species richness. However, we do not know if this would work, or the timescales involved. One of the important areas where more knowledge is needed is information on the size and composition of soil seedbanks as regeneration from zseed is a likely pathway of recovery. Here, we compared seedbanks in both grazed and ungrazed plots in five experiments at Moor House NNR in the northern Pennines; these sheep grazing exclusion experiments were started 52 and 63/64 years ago. Soil samples (n=10) were collected from both grazed and ungrazed plots in each experiment, and seed emergence counted in glasshouse trials. We detected only seeds of common species and very few dicotyledonous species. This suggests that the soil seedbank is unlikely to be a reliable source of the less common species for ecological restoration in these upland communities, suggesting an extinction debt. Therefore, seed addition and the creation of suitable safe-sites for germination may be needed in conjunction with grazing controls to allow the establishment of plants that will increase the species richness of the vegetation. However, this interventionist restoration approach remains to be tested.

Introduction

The rationale for this work is based on the “wet desert” hypothesis put forward by Frank Fraser Darling back in the 1950s, which basically stated that high sheep grazing had reduced species diversity in upland plant communities (Darling, 1955). This has been highlighted recently by George Monbiot, who has described upland vegetation as “sheep-wrecked” (Monbiot, 2013). If the wet desert hypothesis is correct, we would expect that if sheep grazing is reduced then species diversity should increase again. This expectation is at the basis of recent suggestions that the agricultural use of the uplands should be reduced or re-wilded. The simplest approach to start this process would be the reduction or cessation of livestock grazing (Crumley, 2000; Marrs et al., 2018; 2020; Sandom et al., 2013; Stewart et al., 2005).

However, recent analysis of vegetation in long-term exclosure experiments at Moor House NNR in the northern Pennines has provided conflicting support for this hypothesis. On the one hand, a review of the properties of species that thrive once grazing is removed are the most nutritious, but they take a relatively long time to re-establish themselves, especially in grasslands (Marrs et al., 2020). Moreover, a recent analysis of the vegetation change in the continuously-grazed plots showed a general decline in many species groups over the timescale of the experiments (started between 1954 and 1972), but where sheep were removed there were both increases in some species (herbs, mosses, sedges) but faster reductions in grasses and liverworts (Milligan et al., 2016). These authors suggested that restoration to pre-treatment levels would probably require some form of intervention to create safe-sites (sensu Harper, 1977) for seed/spore germination (Milligan et al., 2016). However, a fundamental question before such work is undertaken is a requirement to know if there are either enough, or indeed any, seed/ spores of the target species in the soil seed/diaspore bank.

Here, we revisit five of the high-level experiments at Moor House, three grassland and two blanket bog sites. Detailed descriptions of vegetation change within these experiments have been published elsewhere (Marrs et al., 1988; Milligan et al., 2016; Rawes, 1981; 1983). Each site has a grazed plot and an ungrazed comparator (see Marrs et al., 1986 for detailed experimental methods). However, due to policy changes and grazing rights ownership over the long time-course of these experiments the sheep grazing intensity has been reduced. In the 1960s the overall grazing pressure was ca. 4.4 sheep ha-1 across all vegetation types, but in 1972, grazing density was more than halved to ca. 2.0 sheep ha-1, and it was halved again to ca. 1.0 sheep ha-1 in the early 2000s (Marrs et al., 2018; 2020; Milligan et al., 2016; 2018). In addition, the sheep distribution within the sites varies across the different plant communities with 11.6-23.2% greater densities on the most-grazed grassland communities compared to the least grazed blanket bog ones (Rawes & Welch, 1969). Thus, here our comparison is an ungrazed treatment versus a grazed treatment as the control, which was subject to the “business-as-usual” grazing pressure for that area.

Our aims were to test whether (a) the seedbank in these long-term experiments could assist restoration efforts to improve plant species diversity, and (b) assuming so, did sheep removal affect seedbank quality. Hence, this study compared the plant species composition of the vegetation and seedbanks in the five high-altitude, sheepgrazing experiments on the Moor House National Nature Reserve (Table 1, Fig. 1). Each experiment had two plots, an ungrazed plot where sheep were fenced out and a plot subject to “business-as-usual” sheep grazing pressures (Milligan et al., 2016). The experiments can be viewed in two groups, grassland and high-level bog. The grassland group contained three experiments set up on Hard Hill, Little Dun Fell and Knock Fell, and the high-level bog group contained two experiments at Silverband and Troutbeck Head (Marrs et al., 1986; Milligan et al., 2016). The three grassland experiments were set up between 1954 and 1955 and the high-level bogs in 1966 (Marrs et al., 1986), thus there is consistency between experiments within each group.

Fig. 1. Locations of all long-term vegetation experiments on Moor House NNR in northern England.

Table 1. Locations and summary of experimental history of the five sites at Moor House NNR

In the 1960s Eddy et al. (1968) classified the communities on both Hard Hill and Little Dun Fell as a Festucetum, an Agrosto-Festucetum on Knock Fell, and an Eriophoretum on both Silverband and Troutbeck Head. Milligan et al. (2016) reclassified the Festucetum as H19a (Vaccinium myrtillus-Cladonia arbuscula heath: Festuca ovina-Galium saxatile sub-community, Rodwell, 1991), the Agrosto-Festucetum as CG10 (Festuca ovina-Agrostis capillaris-Thymus praecox grassland community, Rodwell, 1992) and the Eriophoretum as M20b (Eriophorum vaginatum blanket and raised mire: Calluna vulgaris-Cladonia spp. sub-community, Rodwell, 1991) under the UK National Vegetation Classification. Soil types also differed ranging from acidic deep peats on the high-level bog to brown-earth soils at Knock Fell (Eddy et al., 1968; Heal & Smith, 1978; Rawes & Welch, 1969).

Materials and Methods

Seedbank assessment

In April 2018, ten random positions were located in both the grazed and ungrazed plots at each of the five experiments. Sampling was done after a severe winter and hence it would be expected that any seeds in the seedbank would have experience substantive natural chilling. At each position, the surface vegetation was removed, and three soil cores sampled using an auger (5 cm diameter, 7 cm depth) and pooled. At the time of the sampling the intervention had been in place for 63/64 years for the grasslands and 52 years for the high-level bogs.

We used the general seedling emergence method described by Ghorbani et al. (2006) and Lee et al. (2013). The individual soil samples (n=100) were spread individually over a 1 cm deep layer of horticultural sand in a seed tray (21.5×16 cm); these trays were positioned randomly on benches fitted with permanently-saturated capillary matting in an unheated glasshouse (Ghorbani et al., 2006; Lee et al., 2013). Emerging seedlings were identified, counted and removed over four months. Where identification was uncertain at the time of counting, the seedling was transplanted into a separate pot and grown on for twelve months so that an accurate identification could be made. Seedling densities per quadrat were corrected to numbers m-2. Control trays containing horticultural sand were included to identify seed contamination in the substrate and species which may have gained access to the glasshouse. No other pre-treatment (gibberellin or smoke treatment) was given, following (Thompson & Grime, 1997). The few individuals that were detected in these control trays were non-moorland species and were discounted in subsequent analyses. Nomenclature follows Stace (2019).

Vegetation assessment

At each site the abundance of plant species has been monitored at approximately 10-year intervals using point quadrats (Marrs et al., 1986) and all data are available in Rose et al. (2020). The vegetation data for both the grazed and ungrazed plot from the five experiments were extracted from this database for the latest sampling date to be as comparable as possible with the seedbank sampling. At the time of vegetation sampling the intervention had been in place for between for between 59 and 61 years (grasslands) and 48/49 years (high-level bog).

Statistical assessment

Seedbanks are very difficult to assess in these upland habitats because often very few seeds germinate in tests (Lee et al., 2013) as a result of a combination of poor seed production because of low temperatures and wet conditions coupled with poor incorporation into the surface soils as a result of litter accumulation (Ghorbani et al., 2006). Accordingly, here we only analyse the total number of germinable seed detected in each sample. To do this we used a generalised linear model (function ‘glm, R Core Team, 2019) with a poisson error distribution (Crawley, 2013). Model reduction was tested but the full interaction model was selected as the Minimum Adequate Model (MAM). Thereafter, data on seed densities were viewed in two ways. Where species were present in the seedbank of both grazed and ungrazed treatments we compared the densities in each treatment using the nonparametric Wilcoxon rank-sum test (‘wilcox.test’ function, Crawley, 2013; R Core Team, 2019). Otherwise, we present qualitative information on the presence/absence of species in both vegetation and seedbanks in the two grazing treatments at each site.

Results

The total number of seeds detected was low (Fig. 2) but were in the same order of magnitude as those detected in similar studies from slightly lower elevations at Moor House, but an order of magnitude lower than those in blanket bogs at the southern end of the Pennines (Lee et al., 2013). For the analysis of the total number of germinable seeds the additive model was selected as the MAM for the grassland sites, and the interaction model for the high-level bogs (Table 2). For grasslands, the greatest seed densities were at Hard Hill and Knock Fell (no significant difference between them) and Little Dun Fell had the least (Fig. 2C); there was a marginally greater seed density (P<0.06) in the ungrazed treatment (Fig. 2F). In the high-level bogs, there was no significant site differences between the grazed treatment at the two sites, but completely opposing responses in the ungrazed treatment, Silverband having a much lower seed density and Troutbeck Head a much greater one than their grazed counterparts (Fig. 2E).

Fig. 2. Total number of seeds of all species in both grazed and ungrazed plots in long-term experiments in three grasslands and two bogs at Moor House NNR. E, enclosure; G, grazed; NS, not significant. Significance: *P<0.05; **P<0.01.

Table 2. Results for the minimum adequate model derived from mixed-effects modelling of total seed density in grazed and ungrazed plots

SE, standard error; Dev, deviance; AIC, akaike information criterion; NS, not significant.

The change in deviance (%) and AIC relative to the null model are presented and significance denoted as: *P<0.10; **P<0.01; ***P<0.001.

The intercept is denoted as (Int).

In the grasslands, 17 species were present in both vegetation and seedbank (Table 3), seven of these were only found at Knock Fell, and of these four were graminoids and three dicotyledons (Potentilla erecta, Rumex acetosa, and Sagina procumbens), note S. procumbens was also found in the seedbank at the Hard Hill ungrazed plot. Of this group Anthoxanthum odoratum, Luzula campestris/multiflora, Poa humilis and S. procumbens were not detected in the seedbank of the ungrazed plots and R. acetosa was only detected in the seedbank of the grazed plot, where it was not found in the vegetation. Ten common species were detected in the vegetation of at least two sites and they were all detected in the seedbank of at least one treatment at each site (Table 3); six were graminoids and four dicoytledons. Of particular interest were those species that were present in the vegetation but not in the seedbank; three species at Hard Hill (Empetrum nigrum, Eriophiorum vaginatum, and Vaccinium vitis-idaea), two species at Little Dun Fell (Chamaenerium angustifolium and Festuca vivipara) but twenty one species at Knock Fell, the site with the greatest species diversity (Milligan et al., 2016). Of these species missing from the Knock Fell seedbank eighteen were dicotyledons, two were graminoids and one was a clubmoss (Seliginella selaginoides [lesser clubmoss]) (Table 4). Only Juncus bulbosus and J. effusus were detected in the seedbank of grassland sites but were not present in the vegetation (Table 4).

Table 3. Species that were detected (+) in both the vegetation and the soil seedbank in the three grassland sites

TC, taxonomic class; G, grazed; UG, ungrazed; Gr, graminoid, D, dicotyledon.

The species are ranked by the number of sites in which they were detected in the vegetation. Species are ranked by TC.

Table 4. Species that were detected (+) either (1) in the vegetation, or (2) the soil seedbank in the three grassland sites

TC, taxonomic class; G, grazed; UG, ungrazed; Gr, graminoid, D, dicotyledon.

Species are ranked by TC.

In the high-level bogs six graminoids plus Calluna vulgaris were found in both vegetation and soil seedbank, but C. vulgaris was absent from the Silverband seedbank (Table 5). Seven species, one fern, two graminoids and four dicotyledons were detected in the vegetation but not in the soil seedbank and seven species, six graminoids and one dicotyledon (Galium saxatile) were detected in the soil seedbank, but not in the vegetation (Table 5).

Table 5. Species that were detected (+) in (1) both vegetation and soil seedbank, (2) only in the vegetation, or (3) only in the soil seedbank in the two high-level bog experiments

TC, taxonomic class; G, Grazed; UG, Ungrazed; Gr, Graminoid, D, Dicotyledon.

Species are ranked by taxonomic class TC.

Discussion

There is no doubt that the plant species diversity has declined over the last century or so as a result of a combination of factors including high levels of sheep-grazing plus the impacts of atmospheric pollution including SO2 and N on the vegetation (Monteith et al., 2016; Rose et al., 2016), although there is some evidence that pollutant damage is reducing (Mitchell et al., 2017). In the experiments reported here, there has been a reduction in many species-groups within the grazed plots over time, but herbs, mosses, and sedges have expanded when sheepgrazing was stopped (Milligan et al., 2016). Indeed, Marrs et al. (2020) have recently shown that a few species have increased in the ungrazed plots, but it has taken between 10-20 years in blanket bog for this to occur, and up to 60 years in the grasslands. Given, these results, it has been suggested that any rapid increase in biodiversity will come from either exogenous or endogenous processes, or indeed both (White, 1979; Marrs, 1988). Due to the size of the areas of vegetation involved exogenous processes require that the propagules (seeds/spores) can disperse from sites some distance from the study area, and endogenous processes rely on seed production within the existing vegetation. Irrespective of which of these processes occur, once the seed arrives within the vegetation they have to travel through the existing vegetation and litter to reach the soil surface for immediate germination if safe-sites are available (sensu Harper, 1977) or incorporation into the soil propagule bank for future germination (Ghorbani et al., 2006; Thomson et al., 1997). The obvious way to assess the regeneration potential of the vegetation is to measure the size and composition of the soil seedbank (González-Alday et al., 2009; Putwain & Gillham, 1990, Wagner et al., 2018).

Here, we compared the seedbank in both grazed and ungrazed plots in five experiments at Moor House and the results were disappointing in that the seedbanks were relatively small, but in the same order of magnitude as other studies at this site (Lee et al., 2013), and for the most part contained common species that were present in the vegetation, or were abundant nearby (e.g. the Carex and Juncus species). There were very few dicotyledonous species in the seedbank, especially at Knock Fell, the site with the greatest vascular plant diversity. These results show that seedbanks cannot necessarily be relied upon to provide the diversity of seed germination to provide new species to increase plant species biodiversity. It also suggests that some plant species within the existing vegetation may not be providing sufficient viable seed on a regular basis for them to be incorporated into the soil seedbank. This is certainly possible in these relatively harsh conditions (Heal, 1978). Pigott (1968), for example, showed that low temperatures prevented viable fruit set in Cirsium acaule in upland communities. It is also possible that any seed that is produced is prevented from entering the seedbank by the vegetation/litter depth (Ghorbani et al., 2006). These results are in line with previous work on similar habitats in Belgium (Bossuyt & Hermy, 2003) which gives some indication of how different species may perform in grasslands, and Thompson and Grime (1997), Morin and Payette et al. (1988) and Vander-Kloet (1988) give information on the longevity of some species in the seedbank. Taken together, it appears these communities suffer may from an extinction debt (Gonzalez, 2013) as the species that persist do not have the means to regenerate without help. Recent research shows that extinction debts may last for centuries (Isbell et al., 2019).

Thus, it is unlikely that the diversity of these upland communities will be increased quickly in the future unless there are positive interventions to (a) increase the seed input into the communities and (b) create suitable germinations sites. Such safe-sites can be created through a combination of methods to create bare patches, for example through the use of herbicides, fire or mechanical damage by cutting, rotavating, screefing screefing or introduction of species such as Wild Boar (Sus scrofa L.) that forage by creating soil disturbance (Bueno et al., 2010, Critchley et al., 2013, Humphrey & Coombs, 1997; Lee et al., 2013; Miles, 1974; Milligan et al., 2004). However, these suggestions need further experimental testing to assess their feasibility.

Author Contributions

HL, KC and RHM designed the seedbank study and sampled the seedbanks, HL, KC and HMcA performed the seedbank assessments, RR and JO’R assessed the vegetation and along with MF and RHM produced the vegetation database, MF provided an applied perspective and all contributed to writing the manuscript.

Conflict of Interest

RHM is President of the Heather Trust, this is an ambassadorial role only.

Funding

This work was part-funded by (1) the Leverhulme Trust’s award of an Emeritus Fellowship to RHM (EM2018-073/2) and (2) the Heather Trust and (3) NIEA-2020-12. Ms S. Yee produced the figures.