Methods of enhancing botanical diversity within field margins of intensively managed grassland: a 7-year field experiment

Authors


Correspondence author. E-mail: rochelle.fritch@gmail.com

Summary

1. Increased intensification in agricultural grasslands has led to well-documented declines in the associated flora. Manipulation of field margins for biodiversity enhancement in arable systems has been extensively investigated. However, there is a paucity of corresponding long-term research within intensively managed grasslands.

2. We investigated a combination of establishment and management methods to enhance botanical diversity of newly established field margins in intensively managed grasslands. Three methods of field margin establishment were investigated including fencing, natural regeneration by rotavation, or seeding with a wildflower mixture. Subsequent sward management by either grazing or mowing was tested at three margin widths. Success of establishment was addressed in terms of persistence of species richness, plant community composition and incidence of noxious weeds.

3. Seeding with a wildflower mixture was the most successful establishment method to enhance plant species richness and this effect persisted throughout the 7 years of the experiment (inline image = 16·4 ± 0·43 SE plant species richness per 1 × 3 m2 quadrat). Mown (inline image = 6·01 ± 0·30 SE) and rotavated (inline image = 9·7 ± 0·34 SE) treatments contained significantly fewer plant species; grazed controls contained 9·83 ± 0·24 species.

4. Grazing led to a significant, but modest increase in species richness in fenced and rotavated plots compared to the mowing treatment, but had no effect in seeded plots. Grazing also led to an increased frequency and cover of competitive grasses in the seeded treatment.

5. Although margin width was not found to significantly influence species richness, there was increased herb cover and reduced abundance of noxious weeds in the wider seeded margins.

6.Synthesis and applications. The choice of establishment method and subsequent management of grassland field margins significantly affected their conservation value. The botanical diversity of margins within intensively managed pasture can be enhanced by sowing wildflower seed mixtures. This diversity can be maintained over time through appropriate management, i.e. either the reduction of high grazing pressure by seasonal fencing, or annual mowing. Management approaches that involve minimal change are currently adopted in many agri-environment schemes (such as fencing and/or the cessation of nutrient inputs) but did not produce swards of conservation value in this study.

Introduction

Changes in grassland management, such as increased use of inorganic fertilizer, increased stocking rates, frequent sward reseeding with grass monocultures and a move from hay to silage production, have led to dramatic decreases in the biodiversity associated with agricultural grasslands (Blackstock et al. 1999; Frame 2000). These losses have affected all aspects of farmland biodiversity, including plants (Vickery et al. 2001; Stehlik et al. 2007), invertebrates (Fenner & Palmer 1998; Benton et al. 2002) and birds (Donald et al. 2006).

Many floral and faunal species would have restricted ranges or be absent altogether from intensive farmland were it not for field boundary and field margin habitats (Marshall & Moonen 2002). Changes in field margin management, such as sowing wildflower mixtures or reducing pesticide inputs, have been shown to increase farmland biodiversity within arable ecosystems (Asteraki et al. 2004; Critchley et al. 2006). However, corresponding methods of field margin enhancement within grassland systems largely remain unexplored (but see Haysom et al. 2004; Cole et al. 2007; Sheridan et al. 2008).

Grassland conservation research primarily focuses on reduction of management intensity over entire fields or larger areas. While these aims are important, they do little to address the decline of diversity associated with intensively managed grasslands. Furthermore, uptake of agri-environment measures in intensively farmed areas has been low (Hynes et al. 2008). Intensive farms may be more likely to participate in conservation efforts if these are focused on contained, well-defined areas, and therefore do not interfere with overall farm production levels.

The creation of new habitats within intensively managed agricultural land can promote beneficial organisms for biological control, foster ecological resilience by increasing local alpha diversity (Duelli & Obrist 2003) and can act as ‘island’ refuges facilitating the movement of species between patches of semi-natural habitats (Albrecht et al. 2010). Furthermore, field margin diversity may be important for the maintenance of higher trophic level species, particularly farmland birds (Marshall & Moonen 2002).

High soil nutrient levels are often associated with reduced botanical diversity and the dominance of a few, highly competitive species in grasslands (Kleijn et al. 2009). A reduction in soil fertility may lead to swards of conservation value, but results vary and depend on soil type and previous management (Smith et al. 2000; Warren, Christal & Wilson 2002).

Natural regeneration is one method of establishment which preserves the local flora. If this method is employed, then the potential diversity of the field margin is a function of the seeds coming from local sites or from the seed bank (Asteraki et al. 2004). Options to promote the natural regeneration of newly established field margins include allowing the plant community to develop from the existing sward, as well as encouraging recolonization from the seed bank through the use of herbicide to remove the existing vegetation, followed by rotavation. Some studies suggest that local invertebrate taxa are equally well promoted by natural regeneration of field margin vegetation as by sown margins (Thomas & Marshall 1999; Anderson & Purvis 2008).

However, the success of grassland restoration is often seed-limited (Bakker & Berendse 1999) as the lower botanical diversity associated with intensively managed grasslands generally results in a less diverse soil seed bank. Most ‘desirable’ grassland species have shorter seed longevity than arable and ruderal species (Bossuyt & Hermy 2003), with only a few common species producing persistent seed banks (Bekker et al. 1997). Reintroduction of botanical diversity, through the use of seed mixtures, has been successful in the restoration of both arable land (Martin & Wilsey 2006) and intensively managed grasslands (Jefferson 2005).

Temporal persistence of floral diversity within sown arable field margins is a difficulty which has been attributed to lack of disturbance (Pywell et al. 2006). Within grassland systems, grazing herbivores can potentially increase levels of disturbance. However, success is largely dependent on the herbivore type (Vickery et al. 2001) and the intensity/timing of grazing (Bullock et al. 1994; Smith & Rushton 1994).

This study investigated whether botanical diversity within field margins on intensively managed lowland grasslands can be enhanced by: (i) method of establishment, (ii) modification of management, and (iii) increased margin width. The success of these methods, individually and in combination, is assessed in relation to species richness, species persistence, abundance of undesirable species, and the stability of the plant communities established. The results are discussed in the context of the practicality and effectiveness of different treatments in restoring and managing botanical diversity, and the implications for future agri-environment policies which focus on the creation and enhancement of grassland field margin habitats.

Materials and methods

Site location and description

The experiment was conducted on a lowland dairy farm at the Teagasc Research Centre, Co. Wexford (52°17′N, 6°30′W). The site is situated on clay-loam soil. All hedgerows were removed in the 1970s and paddocks separated by electric wire. The area was sown with Lolium perenne 4 years before the experiment commenced. Paddocks were grazed at a stocking rate of between 2·4 and 2·8 livestock units per hectare by a Friesian dairy herd on a 21-day rotation and cut for silage in alternate years. Between 200 and 375 kg ha−1 nitrogen (N), 0–50 kg ha−1 phosphorous (P) and 0–75 kg ha−1 potassium (K) were applied annually to the swards adjacent to the experimental plots from 2002 to 2008.

Experimental design

A stratified randomized factorial split-plot field margin experiment was established in spring 2002. New field margins were located and established alongside the internal subdivisions (achieved by wire fencing) of a larger field; thus, they were not located immediately adjacent to hedgerows. Nine 90 m long strips of grass sward were fenced off from the surrounding paddocks. One of three field margin widths (1·5, 2·5 and 3·5 m) was randomly assigned to each strip to provide three replicates of each width (see Fig. S1, Supporting information).

Three establishment treatments investigated were: (i) fenced off from the main part of the sward (‘fenced’), (ii) rotavated and allowed to regenerate naturally from the seed bank (‘rotavated’), and (iii) rotavated and seeded with a grass and wildflower mixture (‘seeded’). Vegetation was removed prior to rotavation and reseeding, using a glyphosate herbicide at recommended application rates. The seed mixture contained 10 grass and 31 herb species (Table 1) and the mixture was sown at a rate of 2·5 g m−2. Control plots consisted of existing pasture vegetation, which were grazed but had no subsequent application of nutrients or herbicide. Each 90 m strip was divided into three sections and an establishment method was randomly allocated to each 30 m section. Fencing was used to exclude grazing by dairy cattle from all treatment plots from February 2002 to June 2003.

Table 1.   Summary of changes in the frequency of plant species (with frequencies >5%) in permanent field margin quadrats between 2003 and 2008, ranked by most abundant species and mean percentage cover of the ten most frequent species in 2008
 % Frequency change 2003–2008Mean % cover in 2008
MCR + MR + GS + MS + GMCR + MR + GS + MS + G
  1. *Denotes species with cover <1%.

  2. †Species included in the seed mixture.

  3. Other species included within the seed mixture were: Alliaria petiolata, Angelica sylvestris, Anthyllis vulneraria, Arctium minus, Capsella bursa-pastoris, Dipsacus fullonum, Eupatorium cannabinum, Galium verum, Leontodon hispidus, Lythrum salicaria, Medicago lupulina, Origanum vulgare, Pedicularis palustris, Primula veris, Pulicaria dysenterica, Rhinanthus minor, Silene vulgaris, Succisa pratensis, and Vicia cracca.

  4. M, mown, C, grazed control, R + M, rotavated & mown, R + G, rotavated & grazed, S + M, seeded & mown, S + G, seeded & grazed.

  5. Categories: ‘−−’ is <−50%, ‘−’ is −50% to −11%, ‘0’ is −10% to +10%, ‘+’ is +11% to 50%, ‘++’ is >50%, blank spaces are absent from both years.

Agrostis spp.†000000292116211620
Holcus lanatus+++00322627241218
Lolium perenne000++133 15 14
Rumex spp.00++  43  
Ranunculus repens+++0++521186 
Rumex acetosa + +  1 42
Cynosurus cristatus0+ 00    24
Holcus mollis+00++021117   
Poa trivialis0++00+++ 5 7 2
Cirsium arvense++++4184  
Dactylis glomerata+000++42    
Anthoxanthum odoratum  + 0    3 
Arrhenatherum elatius+ 0 +4 11 23 
Plantago lanceolata 0 +0+    714
Phleum pratense  0 −−      
Daucus carota 000−−      
Cerastium fontanum0++++0++ 1 1 1
Alopecurus pratensis 000+0    6 
Leucanthemum vulgare    0     1
Epilobium spp. +00      
Senecio jacobaea 0+++  21  
Trifolium repens ++00+ 2 9 9
Festuca rubra    ++    4 
Juncus spp.  00++  *   
Taraxacum agg.  000      
Filipendula ulmaria    ++      
Poa annua + ++ +      
Urtica dioica  0+0       
Prunella vulgaris    0+      
Veronica serpyllifolia + 0 +      
Juncus bufonius+  0 +*     
Cirsium vulgare  0+00      
Ranunculus acris    +0      
Lychnis flos-cuculi    0+      
Digitalis purpurea          
Elytrigia repens+  + *     
Alopecurus geniculatus +   +      
Achillea millefolium     0      
Centaurea nigra    +       
Lotus corniculatus  0  0      
Quercus spp.0   0       
Trifolium pratense           

All treatment plots were mown and the harvested material removed in September 2002. During June 2003, plots were split and half of each (randomly selected) grazed on a 21-day rotation basis in conjunction with the main sward (‘grazed’ treatment). The ungrazed portion of each plot was mown annually in September and all vegetation removed (‘mown’ treatment). Nutrient and pesticide inputs were excluded from all plots over the duration of the experiment, although grazed plots, including the controls, received dung and urine inputs from the cattle.

Botanical and soil sampling

Botanical data were collected using permanent, nested quadrats. Two, four and six 1 × 3 m quadrats were taken from the 1·5, 2·5 and 3·5 m wide margins respectively (see Fig. S2, Supporting information). Presence/absence data were recorded for the entire 1 × 3 m quadrat. Percentage cover of each species in the central 1 m2 was visually estimated according to the Braun–Blanquet scale (Braun-Blanquet, Fuller & Conrad 1932). To avoid edge effects, a 9 m long strip between treatments was not sampled. Plots were sampled in July of 2002, 2003, 2007 and 2008. Here, data from 2003, 2007 and 2008 were analysed while plant growth in 2002 was treated as an establishment period and not analysed (see Sheridan et al. 2008). Species were identified according to Stace (1997).

Soil samples, consisting of twenty pooled 10-cm depth cores, were taken from each plot in February 2003 and 2008. Samples were analysed to investigate levels of Morgan’s available P, K, and Mg (Jackson 1958; Murphy & Riley 1962).

Data analyses

A repeated measures analysis of total species richness per 1 × 3 m quadrat was undertaken, using GLIMMIX (SAS 9.1.3; SAS Institute Inc., Cary, NC, USA) with a spatial covariance matrix, to account for permanency of quadrat location. The effects of establishment, grazing, width, and time on species richness were investigated with all interactions between factors included in the full model. Initial maximal models were refined by the sequential removal of all non-significant terms. Minimal adequate models were identified by a process of assessment before and after the removal of terms using Akaike Information Criterion (AIC) (Akaike 1974). All tests of significance were at the < 0·05 level.

For each plot, species turnover (which is equivalent to 1 − Sørensen’s similarity index) between 2 years was calculated as:

image

where bi is the number of species present in a plot that are unique to year i; cj is the number of species present in a plot that are unique to year j; Si is the total number of species present in a plot in year i; and Sj is the total number of species present in a plot in year j (Magurran 2004).

Plant community dynamics were investigated further by dividing species into three groups: (i) weed species (species listed on the Irish Noxious Species Act 1936 and including Senecio jacobaea, Rumex obtusifolius, R. crispus, Cirsium arvense and C. vulgare); (ii) herbaceous species (excluding weeds); and (iii) grass species (monocotyledonous species). Species richness and changes in abundance were analysed for each of these groups using a nonparametric factorial analysis as in Brunner & Puri (2001) using Proc Mixed (SAS 9.1.3) with repeated measures and a spatial covariance matrix. This method was used as the data were zero inflated and could not be analysed using parametric methods.

Multivariate analysis was used to investigate the main effects of establishment, grazing, width and their interactions, on the plant community composition (CANOCO 4.5; Biometris, Plant Research International, Wageningen, Netherlands). A partial redundancy analysis (RDA) was performed using species percentage cover data that were averaged across quadrat subsamples within each replicate plot and centred by species, using a Monte Carlo permutation test (reduced model, 9999 permutations restricted to six split-plots, freely exchangeable whole-plots, no permutation at the split-plot level). Partial RDA was used, as plots had a homogeneous composition and showed linear species responses (Leps & Smilauer 2003).

In addition, differences in end-point vegetation composition for each establishment × grazing management were tested in six separate RDA analyses of the null hypothesis that treatment X differed from treatment Y. Significance tests were performed by Monte Carlo permutation tests after 9999 unrestricted permutations, as only one environmental variable (treatment) was considered.

A principal response curve (PRC) graphically demonstrated the change in plant species composition for each establishment × grazing management interaction, over time, with the grazed control as the reference zero line. The PRCs were based on partial RDA with time as a co-variable and treatment × time interactions set as environmental variables. Within these analyses, the average temporal trend was removed by treating it as a continuous covariate. The PRC diagrams correspond to the first and second RDA axes with species highly correlated to each axis displayed to the right of the diagram.

Results

A total of 76 higher plant species were recorded during the experiment. This included 50 herb, 17 grass, five woody, three rush, and one sedge species. A summary of species recorded with >5% frequency can be seen in Table 1. For simplicity, only data from 2003 and 2008 are presented. In 2003, 16 of the 31 herb species and all of the 10 grass species included in the seed mixture were recorded. In 2008, the number of seeded herbs had reduced to 11 while all 10 grass species persisted (Table 1). A full list of species recorded within each treatment can be seen in the Table S1 (Supporting information).

The method of field margin establishment had the greatest influence on botanical species richness (< 0·0001, Tables 2 and 3, Fig. 1a), with an increase from 4·5 species per quadrat in controls to 15·3 species per quadrat in the seeded plots in 2003. This increase in species richness within seeded treatments was maintained over the full experimental period, with seeded plots having the greatest species richness, followed by rotavated plots, and then fenced plots (< 0·0001). Species richness of herbs showed a similar trend (< 0·0001, Fig. 1b). Total species richness also increased within the rotavated and fenced treatments; however, species richness was always significantly higher in rotavated treatment compared to the fenced treatment. This increase in species richness may be partially attributed to the movement of species from the seeded plots. For example, Cynosurus cristatus and Alopecurus pratensis were not present in the fenced and rotavated treatments during 2003, but were recorded there in 2008 (Table 1).

Table 2.   Effects of establishment, year, grazing and width and the interactions of these factors, on total species richness over three sampling periods (2003, 2007 and 2008) calculated using GLIMMIX
 d.f.Total species richness
FP
  1. *< 0·05, **< 0·01, ***< 0·001.

  2. NS, not significant.

Establishment(2, 68)366·89***
Grazing(1, 76)68·15***
Grazing × establishment(2, 76)6·93**
Width(2, 68)1·23NS
Width × establishment(4, 68)2·83*
Year × establishment(4, 68)9·18***
Year × grazing(2, 76)9·24***
Table 3.   Effects of establishment, year, grazing and width and the interactions of these factors on the herb species richness, herb percentage cover, and weed percentage cover of the experimental plots over three sampling periods (2003, 2007 and 2008) calculated using nonparametric methods.
 Herb species richnessHerb percentage coverWeed percentage cover
FPFPFP
  1. *< 0·05, **< 0·01, ***< 0·001.

  2. NS, not significant.

Year21·80***19·19***14·56***
Establishment217·36***126·32***19·76***
Grazing41·28***4·79*1·93NS
Grazing × establishment2·41NS0·42NS0·01NS
Width0·33NS1·53NS2·7NS
Width × establishment3·02NS2·89*6·09**
Width × grazing0·00NS1·64NS3·7*
Width × grazing × establishment0·28NS0·28NS0·82NS
Year × establishment12·61***12·62***4·49**
Year × grazing24·93***4·52*1·15NS
Year × width0·89NS0·71NS2·52*
Figure 1.

 Year-to-year changes in mean (a) total species richness, (b) herb species richness, (c) herb cover, and (d) weed cover with error bars denoting SE (n = 36).

Herb cover was also greatest in the seeded treatment during all sampling years (< 0·0001, Fig. 1c). There was a significant establishment × time interaction (< 0·0001, Table 3). The fenced and rotavated treatments showed significant increases in herb cover from 2003 to 2008 (< 0·0001 and = 0·0003 respectively), while seeded treatment did not.

The interaction between grazing and establishment was significant for total species richness (= 0·0017, Fig. 1a, Table 2). In seeded plots, grazing had no significant impact on species richness, while in fenced and rotavated plots grazing significantly increased species richness. Grazing increased the frequency of competitive grass species (Lolium perenne and Poa trivialis) in rotavated and seeded plots (Table 1).

There was a significant interaction between width and establishment (Table 2, Fig. 2a), with 2·5 m rotavated plots having higher species richness than 1·5 m and 3·5 m rotavated plots (= 0·013 and = 0·044 respectively). Seeded plots showed a trend of increasing species richness with width, although this was not significant. Fenced treatments showed the opposite trend, with species richness decreasing over increasing margin widths, i.e. species richness was significantly greater in the 1·5 m than in the 3·5 m plots (= 0·046). A significant interaction was found between herb cover and margin width, (= 0·04, Fig. 2b, Table 3), with 3·5 m seeded margins containing a higher herb cover than 1·5 m margins (= 0·031).

Figure 2.

 Responses of (a) total species richness (mean species per quadrat ± SE), (b) herb percentage cover (mean cover per quadrat ± SE), and (c) weed percentage cover (mean cover per quadrat ± SE) to establishment and margin widths (n = 36, 72, 108 for 1·5, 2·5 & 3·5 m widths respectively).

Weed cover was significantly greater in rotavated treatments compared to seeded (= <0·0001) or fenced treatments (< 0·0001; Fig. 1d, Table 3) and generally decreased over time (Fig. 1d). This can largely be attributed to reductions in cover of Rumex species and C. vulgare over time (Table 1). Grazing did not significantly influence weed cover, however, the interaction between margin width and establishment was significant (= 0·002, Table 2). In seeded plots, weed cover decreased as width increased (< 0·05; Fig. 2c).

There was a significant interaction between establishment and species turnover period (Table 4, < 0·0001). Relative species turnover was significantly higher in the seeded and rotavated treatments during the initial period compared with fenced plots (Fig. 3). This was primarily because of the loss of ruderal species in these plots following the establishment period. There was also a significant interaction between grazing and turnover period (= 0·012, Table 4). Grazing caused a significantly increased species turnover in the short term from 2007 to 2008 and in long term from 2003 to 2008. Turnover from 2007 to 2008 was similar in all plots (c. 20% per year) with the exception of those which were seeded and mown, where it was significantly lower, c. 12% per year and this indicated the most stable plant community (Table 4).

Table 4.   Effects of establishment, year, grazing and width and the interactions of these factors, on species turnover rates of the experimental plots over the establishment period (year 1–2) and the experimental end-point (year 5–6) and long-term duration (year 2–6)
 Species turnover rate
DFFP
  1. *< 0·05, **< 0·01, ***< 0·001.

  2. NS, not significant.

Establishment(2, 36)6·92**
Grazing(1, 36)9·58**
Grazing × establishment(2, 36)1·52NS
Width(2, 36)2·26NS
Width × establishment(4, 36)1·42NS
Width × grazing(2, 36)0·67NS
Width × grazing × establishment(4, 36)0·54NS
Year × establishment(4, 72)16·05***
Year × grazing(2, 72)4·69NS
Year × Width(4, 72)0·85NS
Year × grazing × establishment(4, 72)2·25NS
Year × width × establishment(8, 72)0·97NS
Year × width × grazing × establishment(12, 72)0·80NS
Figure 3.

 Species turnover rates (mean turnover per quadrat ± SE) in initial establishment period 2002–2003, endpoint 2007–2008 and long-term turnover from 2003 to 2008 of experiment in each establishment × grazing split-plot (n = 36). M = mown, C = grazed control, R + M = rotavated & mown, R + G = rotavated & grazed, S + M = seeded & mown, S + G = seeded & grazed.

The Monte Carlo test showed a significant effect of establishment, width, and grazing, as well as an interaction between these factors on the plant species composition (Table 5). By 2008 most plant communities were significantly different from each other, with the exception of those in the ‘rotavated & mown’ treatment and the ‘mown’ treatment, which had converged over time (Table 6).

Table 5. F-values and significance of six separate Monte Carlo tests for the null hypothesis that species composition is affected by specific treatments and their interactions in 2008
 FP
  1. *< 0·05, **< 0·01, ***< 0·001.

  2. NS, not significant.

Establishment14·797***
Grazing14·797***
Width1·605NS
Grazing × establishment1·999**
Width × establishment0·790NS
Width × grazing × establishment0·94NS
Table 6. F-values and significance of six separate Monte Carlo tests for the null hypothesis that species composition (measured as percentage cover using the Braun–Blanquet scale) is the same in comparison to each other in 2008
 FP
  1. *< 0·05, **< 0·01, ***< 0·001.

  2. M, mown, C, grazed control, R + M, rotavated & mown, R + G, rotavated & grazed, S + M, seeded & mown, S + G, seeded & grazed; NS, not significant.

M vs. R + M1·728NS
M vs. S + M10·616***
R + M vs. S + M5·718**
C vs. R + G3·134**
C vs. S + G5·827***
R + G vs. S + G5·827***

The principal response curve for the first RDA axis showed a clear distinction in plant community structure between the ‘seeded & mown’ treatments compared with all others (Fig. 4a). These plots also displayed stability in community composition over time (as also indicated by turnover), while the composition of all other treatment plots moved towards the grazed control situation (zero line). The principal response curve for the second RDA axis (Fig. 4b) showed the effect of grazing, with community composition of the grazed plots clustering near the control (zero line), while the mown plots clustered together.

Figure 4.

 Principal response curves (PRC) corresponding to the first (a) and second (b) partial redundancy analysis (RDA) axis for plant community data (as percentage cover) change over time versus the grazed control, the zero line, with interactions between the treatments and time acting as environmental variables and sampling time indicators as co-variables. The one-dimensional diagram on the right shows the species scores on the RDA axis. Species which are highly associated with each axis are shown on the right of each panel.

Grazing led to an increase in available soil magnesium and potassium compared to the mown plots, but had no effect on soil phosphate levels in 2008 (see Tables S2 and S3, Supporting information).

Discussion

Establishment method

This study demonstrated that the use of seed mixtures produced the highest species richness and herbaceous cover in experimental field margins over the 7-year experiment. Within this intensively managed grassland system the seed bank and seed rain were not sufficient to improve species richness. These findings concur with many studies on grassland restoration which have found that restoration tends to be propagule-limited (Bakker & Berendse 1999; Pywell et al. 2002; Martin & Wilsey 2006) and therefore requires the addition of seed to increase species richness. Although species richness increased within rotavated and fenced plots over time, this may have been because of the migration of seeded species into adjacent plots through wind dispersal or while plots were being cut (through the movement of hay).

The use of a seed mixture was also found to increase herbaceous cover and decrease weed cover. Rotavation gave rise to increased weed cover and there was little successful herb establishment from the seed bank. These results support our assertion that use of a seed mixture is appropriate when margins have been degraded by intensification and there are limited seed resources available locally. However, the introduction of seed mixtures may not be suitable in extensively managed grasslands or where appropriate species are found locally in the landscape. Under such conditions, seed rain may be sufficient to enhance margin diversity when coupled with appropriate management, such as reduced nutrient inputs and moderate grazing and/or mowing. These methods have been used successfully in some grassland restoration projects where the management history was less intensive (Walker et al. 2004). Due to reduced level of harvesting and minimal application of farm yard manure and fertilizer, field margins at the edge of fields (and adjacent to hedgerows, for example) are likely to differ from newly established field margins within larger fields, in terms of their seed bank, soil nutrient status, conservation value of existing vegetation and exposure to dispersal of flora and fauna.

Field margin management

Plant community composition was significantly affected by management of the field margin. Paddock grazing at this site results in low, uniform sward height with few species capable of setting seed (Vickery et al. 2001), and is typical of many intensively managed pastoral systems. Over the experimental period, the plant community composition of all the grazed treatments became more similar to the control, because of the increased frequency of competitive grasses in the grazed treatments. On the other hand, plots that were mown once annually were subject to lower levels of disturbance. Where soil nutrient status is high, this lack of disturbance can lead to low levels of seedling recruitment, as established vegetation can quickly out-compete seedlings for light (Hautier, Niklaus & Hector 2009).

Under these conditions, mowing led to a more stable composition of the sward community, with lower rates of species turnover than were recorded in grazed plots. It is likely that intermediate disturbance levels may promote enhanced seedling recruitment and sward stability. Hay meadows are usually managed through a combination of mowing and grazing. Minor alterations in management, such as the timing and frequency of both mowing and grazing, can result in significant implications for plant biodiversity (Coulson et al. 2001). There is a lack of data comparing the effectiveness of mowing versus grazing in the restoration of grasslands (Pykala 2000). Increases in biodiversity may be achieved through the manipulation of disturbance through the timing, intensity and frequency of grazing, to produce more micro-sites for seedling germination (Bullock et al. 1994). The use of grazing may be more appropriate than mowing in pasture field margins, as herbivores are readily available and it simplifies management for farmers. Making management options more practical should attract farmers to agri-environment schemes and may lead to wider participation (Morris, Mills & Crawford 2000).

Margin width

Although width seemed less important than seeding and sward management in determining species richness of plots, it appeared to be an essential factor in the successful establishment of sown plant communities. Wider margins facilitated increased cover of herb species in sown margins, while reducing the dominance of perennial weeds. Wider margins, with their increased area, should theoretically lead to decreased extinction risk and thus higher species richness (MacArthur & Wilson 1967). According to Joshi et al. (2006), specialist species tend to be lost when grassland patch size is small, whereas generalist plant species remain constant. The retention of less competitive perennial herb species, and thus the conservation quality of seeded grassland field margins, may be determined by margin width.

Within the Irish context, field sizes are relatively small (inline image = 3·93 ha) and over 70% of the fields are smaller than 4 ha (Deverell, McDonnell & Devlin 2009). Therefore, the establishment of a 6 m margin (which is recommended in many agri-environment schemes) along all field edges would be inappropriate as it could constitute 12% of the average field area. If field margins are an appropriate agri-environment measure for a farm, then a more rational approach would be to dedicate a percentage of the productive area (for example 1–4%) of the farm to expanded margins.

Implications for enhancing field margins within intensive grasslands

With investment in agri-environment schemes approaching €3·7 billion annually in the EU (OECD 2004), new policy measures must show clearly identifiable and measurable biodiversity benefits (Finn et al. 2009). This research shows that there are more effective measures for the enhancement of plant diversity within grassland systems than are sometimes currently adopted. Cessation of nutrient inputs alone was not sufficient to restore plant diversity. While this is essential to prevent further loss of diversity, it appears to be of very limited value in terms of enhancing botanical diversity over an agri-environment contract period, even when farm management is relatively extensive and there are sources of propagules present in the surrounding landscape (see Sheridan, Finn & O’Donovan 2009).

Our results imply that under intensively managed grazing systems, plant diversity can be enhanced through the introduction of seed mixtures. To maintain this diversity over time requires combinations of management approaches that include either appropriate levels of grazing pressure, or reduction of grazing pressure by seasonal fencing of field margins that is followed by either annual mowing or extensive grazing. Wider margins allow better establishment of herbaceous cover and should be used when creating seeded field margins. When designing seed mixtures, appropriate species should be chosen, ideally targeting suitable grassland, geographical and soil types. The use of a single standard mixture may lead to homogeneity in field margins and therefore will not promote diversity at the wider landscape scale. Seed of local provenance should be used to ensure the continued existence of local genotypes (Walker et al. 2004).

Acknowledgements

We thank Rioch Fox, John Murphy, Donncha Madden, Mairead Shore, Anna Bunyan, Oisin Murphy, Michelle Moran and Giles King-Salter for technical assistance. This project was funded by the Department of Agriculture, Fisheries and Food under the National Development Plan 2006 Research Stimulus Fund.

Ancillary