Long-term effect of sheep and goat grazing on plant diversity in a semi-natural dry grassland habitat.

Semi-natural dry grassland sites are of great importance for nature conservation because they support high species diversity and the abundance of "Red-List" species. Grazing has proved to be a successful management tool in terms of maintenance and restoration of biodiversity. For a deeper understanding of the effects of different grazers on species biodiversity in dry grasslands, it is necessary to study the long-term effects of major changes in grazing management. In a semi-natural dry grassland habitat, which was formerly grazed by cattle, we investigated the changes in plant species composition due to long term grazing by sheep and goats. Specifically we asked: a) How does long-term grazing by sheep and goats change the composition of all plant species and particularly those that are on the Red-List? Are changes caused mainly by species turnover? b) How does long-term grazing by sheep and goats influence the fertility and acidity of the soil? To address these questions, we compared the composition and diversity of plants as well deriving Ellenberg indicator values of the species. Long-term grazing by sheep and goats subsequent to a year-round cattle grazing changed the plant species composition of the dry grasslands resulting in a high species turnover rate. It did not, however, lead to an increase in plant species diversity even though Red-List species were considerably more abundant in 2013. Overall, the grazing regime studied positively influenced vegetation composition. The effects on local species composition due to species turnover might further be influenced by local factors like soil nitrogen availability.

Introduction

Most European grasslands are semi-natural and have often developed over centuries under extensive agricultural use or other disturbances (Ellenberg and Leuschner, 2010; Dengler et al., 2014). Without these disturbances, they would gradually decline due to succession (WallisDeVries et al., 2002). Dry grasslands are amongst the most endangered European habitats (Veen et al., 2009) because they are highly diverse, even though mesic and wet grasslands cover much larger areas (Rodwell et al., 2002). Dry grasslands are of great importance for nature conservation because they make a very substantial contribution to the biodiversity of Europe, supporting a diverse and often specialized flora and fauna, including many rare or threatened plant species (Wilson et al., 2012; Habel et al., 2013). Furthermore, rare animal species like beetles, butterflies and birds depend on dry grasslands (Goriup et al., 1991; van Swaay et al., 2006; Schirmel et al., 2015). The restoration and conservation of dry grasslands remains an important goal for nature conservation.

For the maintenance of biodiversity, these particular communities require low intensity, extensive management interventions such as grazing or mowing (Crofts and Jefferson, 1999; Török et al., 2016). Grazing is in many cases more useful than mowing (Tälle et al., 2016), but care is needed to optimize the timing, intensity, pattern and extent (e.g. Schönbach et al., 2011; Mavromihalis et al., 2013). Grazing tends to have profound effects on species richness; its effects depend strongly the type of grazing herbivores and on local conditions such as plot size (Dupré and Diekmann, 2001; Tóth et al., 2016).

Both sheep and goats influence the maintenance of species diversity in dry grasslands positively through epi- and endozoochorous seed dispersal (Fischer et al., 1996; Dostálek and Frantík, 2008; Benthien et al., 2016) although goats are most commonly used to reduce the woody coverage within the pastures (Elias and Tischew, 2016). Other studies suggest that cattle grazing might be more suitable than sheep in restoration of dry grasslands (e.g. Pykälä, 2003; Bakker et al., 2012). However, the degree and direction of changes likely depend on factors such as soil fertility. The influence of grazing on sites with fertile soils could differ from sites which were recently abandoned and never fertilized. Nitrogen availability influences plant species composition, including species turnover through gains and losses of species, changes in dominance and rarity (Cleland and Harpole, 2010). The restoration of former arable land and heavily fertilized grasslands to species-rich more natural grasslands has proven difficult because of the nutrient-rich soils (e.g. Poschlod et al., 1998; Smith et al., 2000).

To resolve these questions requires a longer term focus. Our study area was historically used for cattle grazing up until the late 1990s when sheep were introduced in an attempt to enhance threatened plant species and the overall biodiversity of this dry grassland. In this study we compared the vegetation composition after cattle grazing (ending in 1998) with the vegetation composition after 15 years of sheep/goat grazing. Specifically we asked the following questions:

How does long-term grazing by sheep and goats change the composition of all plant species and particularly those that are on the Red-List? Are changes caused mainly by species turnover?

How does long-term grazing by sheep and goats influence the fertility and acidity of the soil?

We compared the plant species composition and diversity as well as Ellenberg indicator values of plant species found in a semi-natural dry grassland in northern Germany before and 15 years after the implementation of a new grazing management regime.

Materials and methods

Study area

The study was carried out in a landscape typical of the lowlands of northern central Europe in the coastal nature conservation reserve “Dummersdorfer Ufer” near the city of Lübeck, Federal State of Schleswig-Holstein, Germany (53°55′ N, 10°51′ E). The climate is humid sub-oceanic. Mean annual temperature is 8.8 °C (January: 1.0 °C; July: 17.7 °C) and mean precipitation is 712 mm year−1 (averaged over 1981–2010 for the closest meteorological station “Lübeck-Blankensee”, German Meteorological Service). Major soil types are nutrient poor, slightly acidic podsolic luvisols. The nature reserve covers an area of 340 ha along a cliff coast of 5.5 km length. Dry grasslands, open slopes, sandy beach zones, natural forest and small wetland habitats are found in the reserve. The study site comprises about 15 ha of mainly dry grasslands (Koelerio-Corynephoretea, Festuco-Brometea and Calluno-Ulicetea) and, to a lesser extent, humid grassland (Molinio-Arrhenatheretea).

Historical background and management

Since 1822, the “Dummersdorfer Ufer” has been fenced, parceled and used as a “Dreesch-Wirtschaft”, a historical form of extensive agricultural management: Four years of clover-grass cultivation (for nitrogen fixation capacity) followed by four years of rye, oats and buckwheat cultivation. From the 1960s, agriculture shifted towards industrialization and higher profits led to far more intensive farming. Soils were fertilized, pesticides applied and perennial pastures for cattle were established. As a consequence, dense grasslands spread and the area was progressively isolated and disconnected from other ecosystems. In addition the soil became severely decalcified to a depth of 15–25 cm (Gulski, 1994). Prior to 1998, extensive year-round cattle grazing had occurred at the site for at least 20 years. Since then the grasslands at “Dummersdorfer Ufer” have been grazed by semi-free ranging sheep (mainly the local old domestic breed grey horned heath) and cross-breed goats (Boer goat × Valais blackthroat goat). The flock grazes two to three times between April and October in the study area. These grazing events are relatively short and intensive (14–28 days) and depend on the state of the vegetation; the flock is removed when the vegetation is thoroughly grazed. The flock, which is managed by a shepherd, comprised 630 sheep (400 females, 220 lambs, 10 males) and 20 adult goats (18 females, 2 males) in 2013.

Method

A vegetation survey of all vascular plants in the study site was conducted during spring and summer in 1995 and 2013. A total of 32 sample plots were selected randomly and distributed randomly over the whole grassland area in order to represent all local variations of community types according to the different plant communities (Braun et al., 1998). Plot size varied according to the degree of homogeneity and was established during the first investigations in 1995. We used the same community-based approach in 2013 to ensure a comparable assessment of the vegetation. Many studies maintain that a constant plot size is necessary to compare species diversity (e.g. Dierssen, 1990; Dengler et al., 2009) and so we allocated the plots among three different size classes: small (6–8 m2, N = 10), medium (10–12 m2, N = 9) and large (16–20 m2, N = 13). Furthermore, we represented all plant species featured on the “Red List” of the Federal State of Schleswig-Holstein, Germany (Mierwald and Romahn, 2006), as R (naturally extremely rare species of high ecological value), V (near threatened), 0 (extinct), 1 (critically endangered), 2 (endangered) or 3 (vulnerable). We also classified all plant species according to their Ellenberg indicator values in terms of soil acidity and soil fertility (Ellenberg et al., 1992).

Data analysis

We tested our data for differences in median or mean of total and Red-List species numbers as well as frequency of species (total and red-listed species) using Wilcoxon signed rank test and the student's t-test according to the data set. Differences in mean Ellenberg's indicator values per plot were calculated by the Mann Whitney U test.

Beta diversity of the presence-absence data of the plant species was calculated in R using the betapart package (Baselga, 2010) as Jaccard and Sørensen dissimilarity, including turnover and nestedness components for all three plot size classes and pooled data (including all plot sizes). Betapart allows computing pair-wise dissimilarities (distance matrices) and multiple-site dissimilarities, separating the turnover and nestedness-resultant components of taxonomic (incidence and abundance based), functional and phylogenetic beta diversity.

We used non-metric multidimensional scaling (NMDS) to visualize and evaluate patterns of dissimilarity within the years 1995 and 2013 based on their species composition. We then performed permutational multivariate analyses of variance (perMANOVA) (Anderson, 2001; McArdle and Anderson, 2001) to test the hypothesis of differences in species composition between the years 1995 and 2013. We performed perMANOVA based on Bray-Curtis dissimilarities using species presence/absence data. P-values were obtained from 999 permutations. All statistical analyses were done using the free statistical software R (packages vegan and BiodiversityR in version 3.3.2) and SPSS (version 16.0).

Results

We identified plant species from 35 families in 1995 and from 34 families in 2013. The most common plant families in 1995 were Poaceae (19.3%), Fabaceae (14.8%), Rosaceae (11.1%) and Asteraceae (8.9%) and in 2013, Poaceae (20.0%), Fabaceae (14.1%), Caryophyllaceae (9.6%) and Rosaceae (9.6%). Thymus pulegioides L., Plantago lanceolata L. and Festuca ovina L. agg. were the most common plant species in 2013 being present in more than 50% of the plots. In 1995, Festuca lemanii B., Helictotrichon pratense L. and Luzula campestris L. dominated. In 2013, 50% and in 1995, 36% of the most frequently encountered species present in more than 50% of the plots were Red-List species (Appendix Table 1).

The total number of vascular plant species and Red-List species (categories 1, 2, 3, V) found in vegetation plots did not change. 135 vascular plant species including 55 Red-List species were recorded in 32 vegetation plots in 1995 as well as in 2013. We found no significant difference in the total number of plant species as well as Red-List plant species per plot (Table 1).

Table 1

Number of plant species and Red-List species (categories 1, 2, 3, V) observed in vegetation plots (N = 32). Plot size: S = 6–8 m2 (N = 10), M = 10–12 m2 (N = 9), L = 16–20 m2 (N = 13), Pooled = all plot sizes. Significant test results (two-tailed t-test for difference in means) are included.

(A) All species Ntot = 135
	N	1995		2013		tcrit	P
		Mean	SD	Mean	SD
S	10	30.3	5.2	28.9	5.8	2.101	0.597
M	9	30.2	4.8	32.3	6.1	2.131	0.450
L	13	30.5	6.0	28.2	6.2	2.064	0.349
Pooled	32	30.4	5.4	29.6	6.3	2.000	0.588
(B) Red-List species Ntot = 55
	N	1995		2013		tcrit	p
		Mean	SD	Mean	SD
S	10	12.2	2.7	12.4	3.2	2.101	0.888
M	9	12.0	2.7	13.1	2.0	2.120	0.371
L	13	11.8	5.4	10.5	3.8	2.047	0.526
Pooled	32	12.0	4.0	11.8	3.4	2.000	0.896

NMDS plots suggested a possible difference in species composition for both total and Red-List plant species before and 15 years after grazing by a mixed herd of sheep and goats (Figs. 1 and 2). PermANOVA confirmed this result and showed significant differences in the biodiversity of vegetation plots between 1995 and 2013 for both total plant species and Red-List species (Table 2). In terms of Red-List plant species, the stress value for NMDS of the pooled data was greater than 0.2 (Fig. 2) and, therefore, interpretation must proceed cautiously. The source and degree of changes were further analyzed by calculating beta dissimilarity in species composition before and after 15 years of grazing by sheep and goats. The beta dissimilarity observed originated almost completely from species turnover with no or little contribution from nestedness components (Table 3). These results were consistent for both total and Red-List plant species composition. In vegetation plots, 39 plant species were found only in 1995 and 38 plant species only in 2013. In both cases 16 plant species were Red-List species (Appendix Table 2). The mean Ellenberg indicator values per plot for soil nitrogen were significantly lower in 2013 compared to those observed in 1995. We found no significant difference in mean Ellenberg indicator values per plot for soil acidity (Fig. 3, Appendix Table 3). All plant species are listed in the Appendix Table 4.

Fig. 1

Non-metric multidimensional scaling (NMDS) using Bray-Curtis dissimilarity for binary (i.e. presence-absence) data showing dissimilarity of species composition before and after 15 years of grazing management by sheep and goats (2013: circle, 1995: triangle). Distances between transects in the two-dimensional NMDS plot represent dissimilarities in species composition: (A) All plot sizes as pooled data (N = 32; stress = 0.120), (B) 6–8 m2 plot size (N = 10; stress = 0.121), (C) 10–12 m2 plots size (N = 9; stress = 0.135), (D) 16–20 m2 plot size (N = 13; stress = 0.176).

Fig. 2

Non-metric multidimensional scaling (NMDS) using Bray-Curtis dissimilarity for binary (i.e presence-absence) data showing dissimilarity of species composition of all Red-List plant species (categories 1, 2, 3, V) before and after 15 years of grazing management by sheep and goats (2013: circle, 1995: triangle). Distances between transects in the two-dimensional NMDS plot represent dissimilarities in species composition: (A) All plot sizes as pooled data (N = 32; stress = 0.211), (B) 6–8 m2 plot size (N = 10; stress = 0.124), (C) 10–12 m2 plots size (N = 9; stress = 0.139), (D) 16–20 m2 plot size (N = 13; stress = 0.176).

Table 2

Results of a PermaNOVA analysis of differences in biodiversity of vegetation plots between 1995 and 2013. Number of permutations: 999. See also Figs. 1 and 2.

All plant species
		Df	F.Model	R2	p
N = 32	Year	1	13.161	0.175	0.001
Pooled	Residuals	62		0.825
	Total	63		1.000
N = 10	Year	1	4.738	0.208	0.001
Small	Residuals	18		0.792
	Total	19		1.000
N = 9	Year	1	5.066	0.240	0.001
Medium	Residuals	16		0.760
	Total	17		1.000
N = 13	Year	1	5.864	0.196	0.001
Large	Residuals	24		0.804
	Total	25		1.000
Red-List plant species (1,2,3,V)
		Df	F.Model	R2	p-value
N = 32	Year	1	6.273	0.092	0.001
Pooled	Residuals	62		0.908
	Total	63		1.000
N = 10	Year	1	3.631	0.168	0.002
Small	Residuals	18		0.832
	Total	19		1.000
N = 9	Year	1	3.242	0.168	0.007
Medium	Residuals	16		0.832
	Total	17		1.000
N = 13	Year	1	2.216	0.085	0.014
Large	Residuals	24		0.915
	Total	25		1.000

Table 3

Changes in beta diversity of plant species between 1995 and 2013, given as Sørensen and Jaccard dissimilarity indices including spatial turnover and nestedness components. Red-List species include the categories 1, 2, 3 and V (Federal State of Schleswig-Holstein, Northern Germany). Pooled = all plot sizes included; small = plot size of 6–8 m2; medium = plot size of 10–12 m2 and large = plot size of 16–20 m2.

All plant species	C-score (species mean)	Sørensen		Jaccard
Pooled	0.101	βsor	0.289	βjac	0.448
N = 174		βsim	0.289	βjtu	0.448
		βnes	0.000	βjne	0.000
Small	0.162	βsor	0.397	βjac	0.568
N = 131		βsim	0.380	βjtu	0.551
		βnes	0.016	βjne	0.017
Medium	0.151	βsor	0.376	βjac	0.547
N = 127		βsim	0.376	βjtu	0.547
		βnes	0.000	βjne	0.000
Large	0.161	βsor	0.399	βjac	0.570
N = 141		βsim	0.365	βjtu	0.534
		βnes	0.034	βjne	0.036
RLS	C-score (species mean)	Sørensen		Jaccard
Pooled	0.106	βsor	0.296	βjac	0.457
N = 70		βsim	0.296	βjtu	0.457
		βnes	0.000	βjne	0.000
Small	0.131	βsor	0.341	βjac	0.509
N = 54		βsim	0.325	βjtu	0.491
		βnes	0.016	βjne	0.019
Medium	0.136	βsor	0.351	βjac	0.519
N = 51		βsim	0.324	βjtu	0.490
		βnes	0.026	βjne	0.029
Large	0.145	βsor	0.366	βjac	0.536
N = 55		βsim	0.350	βjtu	0.519
		βnes	0.016	βjne	0.017

ßsor = dissimilarity matrix accounting for beta diversity, measured as Sørensen pair-wise dissimilarity; βsim = dissimilarity matrix accounting for spatial turnover, measured as Simpson pair-wise dissimilarity; βnes = dissimilarity matrix accounting for nestedness-resultant dissimilarity, measured as the nestedness-fraction of Sørensen pair-wise dissimilarity.

βjac = dissimilarity matrix accounting for beta diversity, measured as Jaccard pair-wise dissimilarity; βjtu = dissimilarity matrix accounting for spatial turnover, measured as the turnover-fraction of Jaccard pair-wise dissimilarity; βjne = dissimilarity matrix accounting for nestedness-resultant dissimilarity, measured as the nestedness-fraction of Jaccard pair-wise dissimilarity.

Fig. 3

Box-whisker plots showing mean Ellenberg's indicator values (EIV) for all plant species per plot for soil fertility (A) and soil acidity (B) in 1995 and 2013. N = 32 (pooled data). Significant results from Mann-Whitney U test are included in the graphic: *p = 0.031, z = −2.155, r = −0.381.

Discussion

In our study area, long-term grazing by sheep and goats following year-round cattle grazing changed the plant species composition of the dry grasslands significantly. Although there was no increase in overall diversity of plants, Red-List species were encountered more frequently in 2013. Several studies have already shown that grazing by sheep resulted in significant changes in species composition in dry grasslands including an increase in plant species diversity (e.g. Krahulec et al., 2001; Hellström et al., 2003). That this did not occur in our study might be because grazing was not imposed on an abandoned site but instead on one that had already been grazed by cattle. The direction of changes and the magnitude of the effects on plant species composition highly depend on the timing and intensity of grazing (e.g. Schönbach et al., 2011; Mavromihalis et al., 2013). In our study, both the timing and intensity of grazing as well as the species of grazer changed with year-round cattle grazing being replaced by several, relatively short and intense grazing bouts by sheep and goats during the growing season.

An earlier study at the current study site which has a long history of fertilizer use found that extensive grazing by cattle has had negative effects on the floristic composition and species diversity of the grasslands (Braun et al., 1998). Similar to this, a study in Hungarian grasslands found no significant effect of cattle grazing on species richness in dry grasslands (Török et al., 2014). On the other hand in the Netherlands, cattle grazing has been shown to be successful in reducing the grass Deschampsia flexuosa L. (Bokdam and Gleichman, 2000) and in North America, following grazing by bison there was a positive correlation between species turnover and species richness (Bakker et al., 2003). A study in abandoned semi-natural grassland sites showed that cattle grazing resulted in a positive effect on species composition and diversity (Pykälä, 2003). Overall, these studies indicate an inconsistent effect of grazing by cattle. The striking changes in plant composition before and after 15 years of grazing by sheep and goats were almost completely due to species turnover. These results were mirrored by the large number of species present only in vegetation plots in 1995 or 2013; species such as Saxifraga tridactylites L., Sclerantus perennis L., Aira praecox L., Corynephorus canescens L., Festuca ovina L. agg., Ononis repens L. and Saxifraga tridactylites L. which are typical of dry grassland communities (Berg et al., 2004) were present in 2013 but not in 1995. The selective pressure of different types of herbivores on the vegetation might lead to the development of different pools of species if those species were variably responsive to low or high grazing intensities. Systems with a long evolutionary history of grazing have developed resilience mechanisms such as greater capacities for regrowth following herbivory and favoring prostrate growth form that allow reversible shifts in floristic composition with changes in grazing intensity (Milchunas and Lauenroth, 1993) and could be another reason why we observed such high rates of species turnover. Species composition might further be influenced by the combined effects of feeding behavior and seed dispersal mechanisms of the grazing animals. The old domestic breed grey horned heath is a generalist feeder (Benthien & Stolter unpublished) in this study area. This sheep breed has been shown to be adapted to more ‘poor quality’ feeds (Weyreter and Engelhardt, 1984). The effects on species composition might therefore be different for other sheep breeds that are more specialized. Both, endo- and epizoochorous seed dispersal affect local species composition by changing the local soil pool of seeds and in our study both sheep and goats contribute to local species diversity through their effects on seed dispersal (Benthien et al., 2016).

Other local environmental factors including the availability of soil nitrogen and soil pH might also have had an influence on species composition and turnover rates. For example, species tolerant of acid soils might become dominant at the expense of rarer less tolerant species (e.g. Bobbink et al., 1998; Roem and Berendse, 2000). Furthermore, nitrogen enrichment is known to lead to competitive exclusion of some characteristic species by more nitrophilic species (Cleland and Harpole, 2010). We expected an increase in plant species with a preference to non-fertile soils due to reduction of soil nitrogen and indeed, Ellenberg's indicator values suggest a trend towards a decrease in soil nitrogen in 2013 compared to 1995. The total number of plant species indicative of nitrogen-poor soil was considerably higher in 2013 and, further, the mean Ellenberg's indicator value for soil fertility per plot showed a significant decrease in 2013. This pattern might be linked to the practice of removing sheep and goats from the study area during the night and thus limiting the input of nitrogen from excreta over 15 years. Although this might be only a minor part of the N cycling process in the study area, it should still be considered as an explanation of the changes in the local species composition.

The successful long-term restoration of biodiversity in dry grasslands requires close consideration of optimal local soil conditions for desired plant species. In dry grasslands relatively low soil nitrogen content is advantageous (e.g. Stevens et al., 2004; Löbel et al., 2006). This can be achieved by seasonal grazing using a mixed flock of sheep and goats. The type and intensity of grazing plays an important role in nitrogen cycling because of the larger volume of excreta produced by cattle (42–64 kg per head of cattle vs only 1 to 4 for sheep (Smith and Frost, 2000)). Overall annual nitrogen excretion was estimated at 76–116 kg per adult cow and 9 kg per adult ewe (Smith and Frost, 2000). Changing both the type of grazing and removing part of the herd at night over a prolonged period could have a significant impact on N cycling in the study site.

Conclusion

Overall, long-term grazing by sheep and goats, following long-term, year-round cattle grazing, changed the vegetation composition and diversity of the studied semi-natural dry grassland considerably. The effects on local species composition due to species turnover might further be influenced by local factors like soil nitrogen availability. Radical changes in grazing management might therefore have a large impact on processes leading to the restoration and maintenance of biodiversity. Understanding the grazing history and soil characteristics are important for evaluating the success of any new regime. In this study we were able to demonstrate the positive influence of seasonal sheep and goat grazing following year-round cattle grazing on the local species composition of a dry grassland habitat. These results will be beneficial for restoration of similar grasslands elsewhere.

Declarations

Author contribution statement

Oda Benthien: Conceived and designed the experiments; Performed the experiments; Analyzed and interpreted the data; Contributed reagents, materials, analysis tools or data; Wrote the paper.

Matthias Braun: Conceived and designed the experiments; Performed the experiments.

Jana C. Riemann: Analyzed and interpreted the data; Contributed reagents, materials, analysis tools or data.

Caroline Stolter: Analyzed and interpreted the data; Contributed reagents, materials, analysis tools or data; Wrote the paper.

Funding statement

This work was supported by funding from the Centre of Cultural Research Lübeck (ZKFL) and the POSSEHL foundation.

Competing interest statement

The authors declare no conflict of interest.

Additional information

No additional information is available for this paper.