Oecologia

, Volume 154, Issue 4, pp 773–783

Contrasting success in the restoration of plant and phytophagous beetle assemblages of species-rich mesotrophic grasslands

Authors

    • Centre for Agri-Environmental Research, Department of AgricultureThe University of Reading
  • A. R. Edwards
    • Centre for Agri-Environmental Research, Department of AgricultureThe University of Reading
  • C. S. Lawson
    • Centre for Agri-Environmental Research, Department of AgricultureThe University of Reading
  • D. B. Westbury
    • Centre for Agri-Environmental Research, Department of AgricultureThe University of Reading
  • A. J. Brook
    • Centre for Agri-Environmental Research, Department of AgricultureThe University of Reading
  • S. J. Harris
    • Centre for Agri-Environmental Research, Department of AgricultureThe University of Reading
  • V. K. Brown
    • Centre for Agri-Environmental Research, Department of AgricultureThe University of Reading
  • S. R. Mortimer
    • Centre for Agri-Environmental Research, Department of AgricultureThe University of Reading
Community Ecology - Original Paper

DOI: 10.1007/s00442-007-0872-2

Cite this article as:
Woodcock, B.A., Edwards, A.R., Lawson, C.S. et al. Oecologia (2008) 154: 773. doi:10.1007/s00442-007-0872-2

Abstract

Over the last 60 years changes to the management of species-rich mesotrophic grasslands have resulted in the large-scale loss and degradation of this habitat across Europe. Restoration of such grasslands on agriculturally improved pastures provides a potentially valuable approach to the conservation of these threatened areas. Over a four-year period a replicated block design was used to test the effects of seed addition (green hay spreading and brush harvest collection) and soil disturbance on the restoration of phytophagous beetle and plant communities. Patterns of increasing restoration success, particularly where hay spreading and soil disturbance were used in combination, were identified for the phytophagous beetles. In the case of the plants, however, initial differences in restoration success in response to these same treatments were not followed by subsequent temporal changes in plant community similarity to target mesotrophic grassland. It is possible that the long-term consequences of the management treatments would not be the establishment of beetle and plant communities characteristic of the targets for restoration. Restoration management to enhance plant establishment using hay spreading and soil disturbance techniques would, however, still increase community similarity in both taxa to that of species-rich mesotrophic grasslands, and so raise their conservation value.

Keywords

Cynosurus–CentaureaColeopteraGrassland restorationHay meadowResistanceSuccessional trajectories

Introduction

The agricultural intensification of pasture systems has meant that restoration management is now a key approach to the conservation of species-rich semi-natural grasslands across Europe (Bakker and Berendse 1999). The historical basis of restoration management, however, has remained focused principally on replicating target floral communities (Bakker and Berendse 1999; Young 2000; Walker et al. 2004). As a result, consideration of trophic levels other than the plants has remained relatively infrequent during ecosystem restoration (but see Davis et al. 2003; Wassenaar et al. 2005; Woodcock et al. 2006). This bias towards plants reflects the classification of most terrestrial ecosystems on the basis of their floral communities (e.g. Rodwell 1992). Given the importance of microbial, fungal and invertebrate communities, both as functional components of grasslands and in terms of their role in structuring plant assemblages (De Deyn et al. 2003; Walker et al. 2004), considering other trophic levels should be an integral component of all restoration programs (Young 2000).

Since the 1940s, the decline of agriculturally unimproved semi-natural grasslands across Europe has meant that as little as 1–2% of remaining lowland grasslands could be considered to be of high conservation value (e.g. Bakker and Berendse 1999; Blackstock et al. 1999). The causes of these declines include conversion to arable agriculture, inorganic fertilizers, herbicides, improved drainage and changes to cutting and grazing regimes (Duffey et al. 1974; Blackstock et al. 1999). The question of whether such agriculturally improved grasslands can be restored to a high conservation value state is of fundamental importance during any restoration attempt (Mitchell et al. 2000). This reflects a number of issues; principally, can restoration be achieved by simply re-establishing traditional management practices, or do more interventionist approaches need to be implemented (Bakker and Berendse 1999; Mitchell et al. 2000; Young 2000). Assessing the rate of assemblage convergence with target communities (e.g. remnant species-rich grasslands) across multiple trophic levels provides a basis for determining the relative success of such management practices (Mitchell et al. 2000; Hirst et al. 2005; Wassenaar et al. 2005; Woodcock et al. 2006).

The restoration of species-rich semi-natural grasslands is an objective of many European agri-environmental schemes (Walker et al. 2004). In most cases, however, the process has been limited, at least initially, by the availability of appropriate propagules characteristic of the target grassland plant communities (Pywell et al. 2003). In part, this reflects the local rarity of many remnant fragments of unimproved species-rich grassland (Bakker and Berendse 1999; Blackstock et al. 1999). For this reason, grassland restoration management has often focused on either the introduction of seeds, or on management intended to increase the establishment of species (Bakker and Berendse 1999; Pywell et al. 2003).

The source of such seed and the methods of harvesting them have potential implications for the success of grassland restoration. The use of local provenance seeds, rather than commercially available varieties from nonlocal sources, has generally been considered preferable (Jones and Hayes 1999). Harvesting local provenance seeds has been achieved by a variety of methods, including hand collection, vacuum harvesting (e.g. Stevenson et al. 1997), brush harvesting (Edwards et al. 2006) and the application of green hay directly to restoration sites (Manchester et al. 1999; Mortimer et al. 2002). When collected from local donor sites, such methods of harvesting seeds would be likely to introduce both desirable and non-commercially available plant species (Manchester et al. 1999). After the introduction of seeds to restoration sites a variety of methods have been used to increase seedling recruitment; these include severe soil disturbance (Coulson et al. 2001), frequent cutting and grazing (Pywell et al. 2003), as well as nutrient stripping (Walker et al. 2004).

This study aims to test the benefits of a subset of these management practices, represented by seed addition (hay spreading and brush harvesting) and soil disturbance, during the restoration of phytophagous beetle and plant communities of species-rich mesotrophic grasslands. Reflecting the relative paucity of information on the restoration of trophic levels other than plants, this study focuses on the phytophagous beetles. The success of restoration in replicating the phytophagous beetles of target grasslands was contrasted with that shown by the plants. Phytophagous beetles show considerable functional diversity and represent a major component of the total overabundance and species richness of grassland invertebrates (Hoffman 1950–1958; Woodcock et al. 2005a; Woodcock et al. 2006; Woodcock et al. 2007). By considering both plant and phytophagous beetle communities, the factors limiting success in the restoration of multiple trophic levels can be contrasted (Mortimer et al. 2002; Wassenaar et al. 2005; Woodcock et al. 2006).

Materials and methods

Study site and experimental design

The experiment was established during 2000 on an agriculturally improved mesotrophic grassland site at Rocks Farm, East Sussex, UK (Lat: 50°55′56′′N; Lon: 0°24′13′′E). The experimental site had previously received 375 kg ha−1 of NPK fertiliser (ratio 20:10:10) per annum until 1994, when fertilizer application ceased. At the start of the experiment soil pH was 5.0, total nitrogen content was 0.3% and extractable phosphorus and potassium was at 10 and 82 mg l−1, respectively. The site had been grazed by cattle annually in the autumn and spring at rates of ca. 1.5–2.5 livestock units ha−1. The species-poor sward was dominated by grasses (Poaceae), such as Cynosurus cristatus L., Holcus lanatus L. and Lolium perenne L. Forbs were generally scarce in the sward, although both Taraxacumofficinale agg. Weber (Asteraceae) and Trifolium repens L. (Fabaceae) were frequent. The vegetation corresponded to that typical of a Lolium perenneCynosuruscristatus grassland, as classified MG6a by Rodwell (1992). The area surrounding the study site comprised principally of species-poor mesotrophic grassland (similar to the experimental site), arable agriculture and large areas of coniferous and deciduous woodland. With the exception of one area (see below for details) there were no other species-rich lowland hay meadows within a 5 km radius of the study site. There were, however, populations of some forb species typical of species-rich lowland hay meadows that existed in small populations on road verges and hedge banks in the local area, e.g. Lotus corniculatus L. (Fabaceae), Trifolium pratense L. (Fabaceae), Plantago lanceolata L. (Plantaginaceae) and Hypochaeris radicata L. (Asteraceae).

The experiment was structured as a randomised complete block design. Four replicate blocks were established, each comprising eight experimental plots 10 × 10 m in dimension and separated from adjacent plots by 5 m. Within each replicate block experimental treatments were assigned to the eight plots at random on the basis of a 2 × 4 factorial design. In all cases experimental treatments were applied only in 2000, and not in subsequent years. The first experimental factor was the addition of seeds from a species-rich mesotrophic grassland (the donor site), and comprised: (1) control, with no seed addition; (2) hay spreading at a low application rate; (3) hay spreading at a high application rate; (4) brush-harvested seeds applied at a high rate. The low rate comprised the application of material from one unit area of the donor site to three times the area on the experimental site, whilst the high rate comprised a 1:1 ratio. For the hay-spreading treatments, hay was cut at the donor site on July 2000, turned into rows, collected using a forage harvester and applied directly to the plots using a manure spreader. Seeds were collected with a brush harvester at the same time and from the same site as those harvested by conventional hay cutting. Brush-harvested seeds were air-dried, cleaned and broadcast onto the plots in July 2000. The brush-harvesting method uses rotating brushes to strip seeds from the sward, without the need to harvest the bulk of the foliage as is the case with hay cutting. In all cases the hay- and brush-harvested seeds were thoroughly mixed prior to application to avoid systematic bias originating in the way seeds were collected at the donor site. All seeds were collected from a single donor site, Coach Road (Lat: 50°55′27′′N; Lon: 0°23′51′′E), which was separated by ca. 800 m from the experimental site of Rocks Farm. This donor site was a species-rich mesotrophic grassland, classified as a Cynosurus cristatus–Centaurea nigra MG5 grassland (Rodwell 1992). This is the main form of unimproved lowland mesotrophic grassland in England and is therefore of considerable conservation importance. Approximately 5,000 ha remain in pure form, representing ca. 70% of the remaining semi-natural neutral grassland in the UK (Blackstock et al. 1999).

The second experimental factor tested was soil disturbance. Several studies on the enhancement of the botanical diversity of agriculturally improved grassland have shown the importance of disturbance of the existing sward in promoting seedling establishment (Walker et al. 2004). Disturbance was created using a power harrow and applied at two levels: (1) control, with no disturbance; (2) disturbance by power harrowing. Power harrowing was performed prior to seed additions in July 2000, resulting in the creation of bare ground and a reduction in overall grass cover of around 40%.

Subsequent to the establishment of experimental treatments in 2000, the plots were grazed by cattle from late April to late October in the first year after the seed addition treatments. Subsequently, it was managed as a hay meadow, with no spring grazing and a late July hay cut. Aftermath grazing by cattle occurred between late July and late October, at a stocking density of 0.3–0.4 livestock units ha−1. This management was typical of that used on most species-rich mesotrophic grasslands (Duffey et al. 1974).

Plant sampling

For each experimental plot, the botanical composition was recorded during late May to early June for the years 2001–2004. Ten 0.5 × 0.5 m quadrats were randomly positioned within each plot, based on random coordinates that excluded areas 1 m from the plot edge. In each quadrat the presence or absence of vascular plant species was recorded. The presence or absence of individual species within each of the ten quadrats was then summed to provide a score ranging from 0 to 10. This method is suited to determining the occurrence of potentially patchily distributed species at low frequency in the sward. Following the methodology described above, 15 spatially separated sets of ten quadrats were also placed in the species-rich mesotrophic grassland donor site at Coach Road during 2001. Each quadrat was separated by at least 15 m. Restoration of the plant community would be considered to be successful if the experimental plots were to contain the same floral species with the same frequencies as those found at the donor site. Within the terminology of this paper the donor site is considered to be the “target” for successful restoration. Botanical nomenclature follows Stace (1997).

Beetle sampling

Beetle sampling was also carried out for four years from 2001 to 2004. During each year, experimental plots were sampled three times (May, July and September). Sampling was performed using a Vortis suction sampler (Burkhard Ltd., Rickmansworth, UK). On each sampling occasion, the Vortis sampler was placed in 15 positions, located randomly within the plot area. For each of these 15 positions the Vortis was held in place for 10 s. The total area sampled per plot for each sampling date was 0.3 m2. Suction sampling is a quantitative method suitable for the collection of adult invertebrates inhabiting short grassland swards (Woodcock et al. 2005b; Woodcock et al. 2007). The Vortis suction sampler is designed to prevent airflow loss as a result of dislodged vegetation. All beetle counts were summed for individual years. Samples were also taken within the target/donor Coach Road mesotrophic grassland. Fifteen samples were made at this site in May, July and September 2001. Each sample was separated by at least 10 m, matching the intersample distances used at the experimental site. Restoration of the phytophagous beetle assemblage would also be considered successful if the experimental plots were to contain the same beetle species with the same frequencies as those found at the Coach Road donor site. The phytophagous beetles of the families Chrysomelidae, Bruchidae, Apionidae and Curculionidae were identified to species. These species represent the majority of the phytophagous beetles found within mesotrophic grasslands. Nomenclature follows Strejcek (1993) and Morris (2003).

Data analysis

Changes in the structure of the beetle assemblages were assessed using the linear ordination method of partial redundancy analysis (pRDA). This was chosen on the basis of the short gradient lengths obtained from a preliminary detrended correspondence analysis (gradient length = 1.87). Only phytophagous beetle species represented by more than one individual were included in the analysis. In all cases, abundances of individual species were summed within a particular year and log10-transformed. Following ter Braak and Šmilauer (2002), the temporal change in beetle assemblage structure was assessed based on interactions between treatment variables and sample years (e.g. Env.Var.×2001, Env.Var.×2002, Env.Var.×2003 and Env.Var.×2004). Sample year (2001, 2002, 2003 and 2004) and replicate block were also included as covariables, with the latter of these used as a blocking factor. Individual sample years were therefore treated as temporal split-plots within the analysis, where samples were permutated freely between whole plots factors only. The two management treatments of seed addition and soil disturbance were tested individually. For the soil disturbance treatment, with its two factor levels, this required only one test on the interaction between year and the presence of soil disturbance. For the seed addition treatments, which had four factor levels, separate tests for the interactions control×year, low hay×year, high hay×year and brush harvesting×year were required. To provide insight into the interaction between the two restoration-management treatments of seed addition (four levels) and soil disturbance (two levels), the eight separate interactions were coded for by dummy environmental variables. The interaction of each of these dummy environmental variables with year was then tested following the methodology described above. In all cases tests were based on Monte Carlo permutation of both canonical axes under a reduced model (1,000 permutations). Note that the F statistics for the pRDA were calculated by Monte Carlo permutation tests, and while similar, it is not the same as the conventional F statistic used with statistical methods such as ANOVA and GLM. As such, the degrees of freedom are, by convention, not presented (ter Braak and Šmilauer 2002). The analysis was carried out in CANOCO 4.5.

To assess the success of restoration for the plants and beetles (SBeetles and SPlants), the similarity between assemblages of experimental plots and the target mesotrophic grassland was determined based on Euclidean distance between assemblages pairs of plots (Krebs 1999). Euclidean distance was defined as
$$ {\text{ED}}_{{jk}} = {\sqrt {{\sum\limits_{i = 1}^n {{\left( {X_{{ij}} - X_{{ik}} } \right)}^{2} } }} } $$
where EDjk is the Euclidean distance between samples j and k; Xij is the number of individuals of species i in sample j; Xik is the number of individuals of species i in sample k; n is the total number of species. There was an inverse relationship between the Euclidean distance and the similarity of samples in terms of their species composition and species relative abundance. The Euclidean distance between target mesotrophic grassland and each of the experimental plots was determined for 2002, 2003 and 2004. In all cases the Euclidean distance was measured relative to the mean species abundance scores recorded at the target mesotrophic grassland at Coach Road during 2001.

While “S” represented a measure of the overall success of restoration, it was possible that beetle and plant assemblages may have changed from what was seen in the initial year of establishment (2001) without showing a concomitant increase in assemblage similar to that of the target grassland (Mitchell et al. 2000). A second measure was therefore considered, which defined the resistance of the grasslands (RBeetles and RPlants) to the restoration management treatments. This parameter “R” made no assumption of the success of restoration, as defined by the previous parameter “S”. Resistance was measured as the Euclidean distance of the assemblages of each experimental plot, as recorded in the first year of the study (2001), from the assemblages of the experimental plots in the subsequent sample years (2002, 2003 and 2004). The smaller the Euclidean distance so the greater was the resistance of the agriculturally improved grassland to the restoration management treatments.

The response of restoration success (SBeetles and SPlants) and resistance (RBeetles and RPlants) to the management treatments of seed addition (four levels) and soil disturbance (two levels) was assessed using generalised linear models (GLM) within SAS 9.01. Fixed effects were the seed addition treatment, soil disturbance treatment, sample year (treated as a categorical variable) and all possible interactions of these factors. Replicate block was also included as a blocking factor. A nested analysis was used to account for the nonindependence of repeated samples made in experimental plots for each sample year. All effects were tested on the basis of a Type III analysis (order of parameter inclusion was not important) and model simplification was by deletion of the least significant factors. Deletion of a single fixed effect parameter already part of an interaction was not permitted. Finally, to assess whether successful restoration, independent of management treatments, in the beetle assemblages (SBeetles) was related to that of the plants (SPlants) these two parameters were correlated within a GLM. This was based on mean values of restoration success (SBeetles and SPlants) for each of the eight treatment combinations (derived from the four replicate blocks) for the years 2002, 2003 and 2004.

Results

In total 4,106 beetles were identified to one of 71 species (Apionidae Abundance = 269, species richness = 15; Curculionidae N = 667, SR = 32; Chrysomelidae N = 3,267, SR = 23; Bruchidae N = 1, SR = 1). A total of 73 species of vascular plant were found within the experimental plots. The seed addition treatments resulted in the introduction of a range of plant species not previously present at the site, including Centaurea nigra L. (Asteraceae), L. corniculatus, T. pratense and P. lanceolata.

Beetle responses to restoration management

The effect of the two management treatments on the assemblage structure of the phytophagous beetles was assessed using pRDA. For the seed addition treatments, significant responses for the control×year (F = 2.58, P = 0.02, explained variance = 14.9%) and the high hay application×year (F = 3.59, P < 0.001, explained variance = 19.6%) interactions were found. For the low hay application×year (F = 1.98, P = 0.15) and brush harvesting×year (F = 2.13, P = 0.09) interactions, no overall significant effects on beetle assemblage structure were found for these seed addition treatments. An overall soil disturbance treatment effect was also found for the soil disturbance×year interaction (F = 2.65, P = 0.01, explained variance 27.8%).

To provide insight into the interaction between the management treatments of seed addition (four levels) and soil disturbance (two levels), eight dummy environmental variables coding for this interaction were created. Separate pRDA analysis of the interactions with these environmental variables and year were then tested. Of these eight treatment interactions, the control×year effect, receiving neither seed addition nor soil disturbance, showed a significant year interaction (F = 2.48, P = 0.05, explained variance = 7.5%). The individual restoration management treatments of high hay application (F = 2.42, P = 0.04, explained variance = 4.4%) and high hay application in combination with soil disturbance (F = 4.14, < 0.001, explained variance = 12.0%) also showed significant year×treatment interactions. Both the low hay application rate in combination with soil disturbance (F = 2.62, P = 0.03, explained variance = 8.2%) and brush harvesting with soil disturbance (F = 2.62, P = 0.03, explained variance = 7.4%) showed significant interaction with year. Considered together, all the significant year×treatment level interactions explained 44.0% of the unexplained model variance. When the significant treatment level×year interactions were displayed on a biplot there were clear successional changes in the assemblages of the phytophagous beetles in response to the significant restoration management treatments (Fig. 1).
https://static-content.springer.com/image/art%3A10.1007%2Fs00442-007-0872-2/MediaObjects/442_2007_872_Fig1_HTML.gif
Fig. 1A–B

RDA ordination diagrams for the sample years 2001–2004 showing the temporal response of the beetle assemblages to the significant restoration management treatments. A Represents the companion beetle species scatter plot to B. The change with time of the beetle assemblages in response to the management treatments in B is emphasised by the connection of the centroids of year×treatment interaction with arrows, from 2001×treatment (start of first arrow) to 2004×treatment (end of final arrow). Only species with high axis 1 or 2 scores are shown (thin line vectors) are shown in A. Year, year; HL, low rate of hay spreading; HH, high rate of hay spreading; BH, brush-harvested seeds; PH, disturbance by power harrow. Specific names are given in full, abbreviations for generic names are: L. = Longitarsus (Chrysomelidae); C. = Chaetocnema (Chrysomelidae); I. = Ischnopterapion (Apionidae); Pe. = Perapion (Apionidae); P. = Protapion (Apionidae); A. = Alophus (Curculionidae); S. = Sitona (Curculionidae); H. = Hypera (Curculionidae). The fourarm asterisk indicates samples from the species-rich chalk grassland donor site, which were included as supplementary sites and therefore had no effect on the main RDA analysis

Success of restoration

The success of restoration was assessed based on the Euclidean distance between assemblages of each experimental plot in 2002, 2003 and 2004, and the assemblages at the target species-rich mesotrophic grassland. The smaller the Euclidean distance the greater the success of restoration. This was assessed separately for both the plant and beetle assemblages. In all cases significant effects of replicate block were found. For the beetles restoration success (SBeetles) was significantly influenced by soil disturbance, seed addition, year and the interactions seed addition×year and soil disturbance×year. There was, however, no overall seed addition×soil disturbance×year interaction. Restoration success was greatest for the beetles where the soil disturbance management treatment was applied. There was also a strong temporal trend from 2002 to 2004 for increasing success of restoration, particularly where soil disturbance was used in combination with high hay application or brush harvesting (Table 1, Fig. 2A,C). Importantly, the control treatments showed no temporal increase in restoration success from 2002 to 2004.
Table 1

The effect of grassland restoration management of seed addition (Seed) and soil disturbance (Dist.) on the success of restoration (SBeetles and SPlants) and resistance to restoration management (RBeetles and RPlants) for plant and beetle assemblages of degraded mesotrophic grasslands. In addition to a block effect, all interactions of the soil disturbance, seed addition treatments and sample year (Year) were tested

 

Dist.

Seed

Dist.×Seed

Year

Dist.×year

Hay×year

SBeetles

F(1,21) = 56.5***

F(3,21) = 20.0***

F(3,21) = 4.96**

F(2,54) = 20.29***

F(2,54) = 5.13**

F(6,54) = 4.19**

RBeetles

F(1,24) = 29.2***

F(3,24) = 12.2***

NS

F(2,54) = 19.7***

F(2,54) = 4.13*

F(6,54) = 3.22**

SPlants

F(1,21) = 16.5***

F(3,21) = 38.2***

F(3,21) = 8.56***

F(2,62) = 4.75*

NS

NS

RPlants

F(1,21) = 19.5***

F(3,21) = 5.71**

F(3,21) = 3.52*

F(2,62) = 55.2

NS

NS

No significant Dist.×Hay×Year interactions were found

< 0.05; ** < 0.01; *** < 0.001

https://static-content.springer.com/image/art%3A10.1007%2Fs00442-007-0872-2/MediaObjects/442_2007_872_Fig2_HTML.gif
Fig. 2A–D

Successes of restoration in response to management for beetles and plant assemblages. Restoration success is measured by the Euclidean distance (±SE) between the experimental treatments for 2002, 2003 and 2004 and the target Cynosurus–Centaurea mesotrophic grassland. Experimental treatments were the four methods of seed addition and soil disturbance. The levels of the seed addition treatment were: open diamonds, control; filled diamonds, low rate of hay spreading; open squares, high rate of hay spreading; multisymbol, brush-harvested seeds. The seed addition treatments are figured both with (A, B) and without (C, D) soil disturbance. Note that for clarity an offset of 0.05 years has been used for consecutive seed addition methods from 2002 to 2004

In the case of the plants, restoration success (SPlants) showed significant year, seed addition and soil disturbance effects. There was also a significant seed addition×soil disturbance treatment interaction. Greatest restoration success for the plants occurred where soil disturbance was applied in combination with high application rates of hay- or brush-harvested seeds. Although year had a significant effect on success of restoration for the plants, there was no evidence of a strong temporal increase in restoration success for any of the treatment interactions. There were also no year×management treatment interactions (Table 1, Fig. 2B,D).

Independent of the effects of restoration treatment and year, there was a significant relationship between the success of restoration of the beetles (SBeetles) and the plants (SPlants). SBeetles showed significant linear (F(1,93) = 27.3, P < 0.001) and polynomial responses (F(1,93) = 21.5, P < 0.001) to SPlants. The general shape of this relationship indicates that a threshold of success in the restoration of the plants must be reached before a positive relationship with the success of restoration of the beetles becomes evident (Fig. 3).
https://static-content.springer.com/image/art%3A10.1007%2Fs00442-007-0872-2/MediaObjects/442_2007_872_Fig3_HTML.gif
Fig. 3

Relationship between restoration success in plant and phytophagous beetle assemblages during the restoration of species-rich Cynosurus–Centaurea mesotrophic grassland (y = −0.35x2 + 22.8x − 297.4; R2 = 0.61). This relationship is considered independent of restoration management treatment and considers all plots from 2002 to 2004. Restoration success is measured by the Euclidean distance between the assemblages of experimental plots and the target Cynosurus–Centaurea mesotrophic grassland

Resistance to restoration management

The resistance of the plants (RPlants) and beetles (RBeetles) to changes in species composition as a result of management treatment was measured by the Euclidean distance between assemblages of experimental plot in 2002, 2003 and 2004 and those of the same experimental plot in 2001. Increasing Euclidean distance over time indicated a temporal decrease in the resistance of an assemblage in response to a management treatment. For the phytophagous beetles, resistance was significantly influenced by soil disturbance, seed addition, year and the interactions soil disturbance×year and seed addition×year. In all cases significant block effects were found. Resistance in the phytophagous beetles was seen to decrease over time, particularly where the seed addition treatments of high hay application and brush harvesting were applied. This temporal decrease in resistance was more pronounced where soil disturbance was used in combination with these seed addition treatments. In the case of the control, where no restoration management treatment was applied, resistance remained high from 2002 to 2004, and no temporal trend for the phytophagous beetles was found. Resistance to the restoration management treatments was always greater without soil disturbance (Table 1, Fig. 4A,C).
https://static-content.springer.com/image/art%3A10.1007%2Fs00442-007-0872-2/MediaObjects/442_2007_872_Fig4_HTML.gif
Fig. 4A–D

Resistance of degraded mesotrophic grasslands to restoration management between 2002, 2003 and 2004 for plant and beetle assemblages. Resistance is measured as the Euclidean distance (±SE) between the assemblages of experimental plots in 2002, 2003 and 2004 and the corresponding plots in the experiment establishment year of 2001. Experimental treatments were the four methods of seed addition and soil disturbance. The levels of the seed addition treatment were: open diamonds, control; filled diamonds, low rate of hay spreading; open squares, high rate of hay spreading; multisymbol, brush-harvested seeds. The seed addition treatments are figured both with (A, B) and without (C, D) soil disturbance. Note that for clarity an offset of 0.05 years has been used for consecutive seed addition methods from 2002 to 2004

For the plant community, RPlants responded significantly to soil disturbance, seed addition, the interaction between these two factors and sample year. In contrast to the beetles, where resistance in the control showed little evidence of a temporal trend, all plant treatments showed a decrease in resistance from 2002 to 2004 (Table 1, Fig. 4B,D). This temporal trend of decreasing resistance showed little difference between the treatments. Overall resistance, however, was greatest where no soil disturbance had been used in combination with the seed addition treatments of high rates of hay application and brush harvesting.

Discussion

Restoration success was defined as the replication of the plant and phytophagous beetle assemblage’s characteristic of a species-rich mesotrophic grassland. Given these criteria there was a positive relationship between restoration success of the plants and phytophagous beetles, independent of the management treatments. This was not, however, a simple linear relationship, as there was evidence of a minimum threshold of successful restoration required by the plant communities before similar success in the beetle assemblages became apparent. In part, this may reflect contrasting responses to management between plants and phytophagous beetles. For example, cutting swards short during the restoration of other grassland types has been shown to be beneficial for plants, although not for invertebrates (Bakker and Berendse 1999; Mortimer et al. 2002). It is also possible that this relationship may reflect differences in how plants and beetles became established. Establishment of plants characteristic of the target grassland was promoted by the use of seed addition treatments. This contrasted with the phytophagous beetles where colonisation was by natural immigration only. It is likely that these differences in the mode of species establishment into the experimental plots resulted in this discrepancy in restoration success between the plants and the beetles (Mortimer et al. 2002; Wassenaar et al. 2005; Woodcock et al. 2006). Independent of this difference, the relationship between restoration successes of these two trophic levels reflects the strong interaction between host plants and herbivores. What was not be determined in this study was the relative importance of the phytophagous beetles in influencing the plant communities in terms of their impact on plant population dynamics (De Deyn et al. 2003; Pywell et al. 2003).

Beetle responses to restoration management

Assessing the direct effects of management practices on the phytophagous beetles, independent of restoration success, can provide a basis for developing future management practices during grassland restoration (Mortimer et al. 2002; Woodcock et al. 2006). Unexpectedly, the assemblage structure of the beetles in the control treatment showed evidence of temporal changes in species composition from 2001 to 2004. These temporal changes in beetle assemblage structure, however, were characterised by relatively minor shifts in the proportions of species typical of agriculturally improved mesotrophic grasslands, such as Sitona lineatus L. (Curculionidae) and Ischnopterapion virens (Herbst) (Apionidae). Such temporal changes in the assemblage structure for the control treatment were attributed to individual beetle species response to seasonal changes in weather conditions or other stochastic environmental variations (Tilman 1987; Davis et al. 2003; Wassenaar et al. 2005). In contrast, combinations of the treatments of soil disturbance and seed addition resulted in the establishment of phytophagous beetle assemblages that differed significantly from the control over this same period. This was most apparent for the seed addition treatments of hay spreading (low and high application rates) and brush harvesting when used in combination with soil disturbance. It is suggested that without soil disturbance creating germination niches for host plants (Coulson et al. 2001), the seed addition treatments alone would fail to establish plants that were both characteristic of the target grasslands and utilised by the beetles. The only exception to this was the high application rate of hay, where it seems likely that the mulching effect of the hay addition may have limited the competitive ability of existing plants, particularly grasses, within the pre-restoration sward (Cummings et al. 2005).

Although differing significantly, some similarity in the direction of the successional changes in assemblage structure in response to seed addition and soil disturbance treatments were seen for the phytophagous beetles. This was clearest for the low and high application rates of hay spreading, which showed similar successional trajectories from 2001 to 2004. This probably reflects the similar floral species introduced by these two methods, although the establishment of more patchily distributed plants was likely to be greater under high application rates of hay spreading (Manchester et al. 1998; Pywell et al. 2003). Where brush harvesting was used as the method of seed collection, the successional trajectory from 2001 to 2004 was characterised by beetle species that were either polyphagous or fed principally on clovers (Trifolium spp.: Fabaceae), e.g. the weevils Sitona lepidus Gyllenhal (Curculionidae) and Protapion dichroum (Bedel) (Apionidae). This contrasted with the hay-spreading methods that resulted, by 2004, in the establishment of phytophagous beetle assemblages feeding on a much wider variety of Fabaceae and other forbs. This reflects the relative efficiency of these different seed collection methods, with hay spreading introducing a wider variety of plant species (Manchester et al. 1999; Edwards et al. 2006).

Contrasting success of restoration for plants and beetles

To assess the success of restoration, it is necessary to compare the relative merits of different management treatments in replicating a target community (Mitchell et al. 2000; Hirst et al. 2005). While a considerable level of success has been achieved in restoring plant assemblages in a number of communities, this has not always been the case for invertebrates (e.g. Young 2000; Walker et al. 2004; Wassenaar et al. 2005; Woodcock et al. 2006). This failure to restore invertebrate assemblages can be attributed to a number of potential factors, including the absence of key floral species or the limited dispersal ability of many invertebrates (Mortimer et al. 2002; Woodcock et al. 2006). For the phytophagous beetles, restoration success increased from 2002 to 2004 and showed interactions with the restoration management treatments of soil disturbance and seed addition. Such temporal increases in restoration success have been reported in other studies during the restoration of invertebrate assemblages (Davis et al. 2003; Wassenaar et al. 2005). In particular, soil disturbance in combination with seed addition from either high rates of hay spreading or brush harvesting showed the greatest temporal increase in restoration success for the beetles. In response to these same treatment combinations, the phytophagous beetles also showed concomitant declines in assemblage resistance to restoration management. There was a clear advantage, therefore, of using these management treatments during the restoration of beetle assemblages typical of species-rich mesotrophic grasslands. This is especially true when these management treatments were contrasted with the control, where the absence of strong temporal changes in both restoration success and resistance to restoration management were found. Without the instigation of such interventionist management, successful restoration of the beetle assemblages would therefore either not occur or would take a considerable period of time.

The plants assemblages did show some parallels with the beetles, as for both groups soil disturbance in combination with either high rates of hay spreading or brush harvesting resulted in communities with greater similarity to that of the target mesotrophic grassland. There was little evidence, however, of a temporal increase in restoration success for the plant assemblages from 2002 to 2004. It was thought that for the plants these management treatments rapidly influenced species composition by increasing similarity with the target grasslands principally during the first year after restoration management began. Later temporal changes in restoration success were, however, minimal and showed no interaction with the management treatments. It is probable that the degree of restoration success achieved by the plants during the first year of management far exceeded that seen for the phytophagous beetles (Woodcock et al. 2006). Subsequent temporal increases in restoration success for the beetles may therefore have represented a time lag relative to the plants, as the success of restoration achieved by the beetles approaches that already reached by the plants. If during subsequent years no temporal increase in restoration success is shown by the plants, then the long-term restoration success for the phytophagous beetles may also be expected to plateau.

In contrast to the beetles, resistance to restoration management in the plants was seen to decrease from 2002 to 2004 for all treatments, although this trend was less pronounced for the control. This meant that while plant assemblages were changing in response to management, the absence of subsequent temporal increases in restoration success for the plants suggested that floral communities not characteristic of the target mesotrophic grassland were becoming established (Mitchell et al. 2000). Such establishment of alternative stable states is a potential problem in the restoration of any habitat type (Mitchell et al. 2000; Hirst et al. 2005). Possible management to increase the long-term success of restoration for the plants may include the repeated application of management treatments, in particular hay spreading.

Implications for future management

As with most semi-natural grasslands in Europe, the target for restoration in this study was not a climax community, but instead a sub-seral state maintained by intervention management in the form of hay cutting and grazing (Mitchell et al. 2000). Using such communities as targets during restoration may be questionable, given the potential temporal variability in their assemblage structure. A snapshot view of plant and beetle assemblages, however, provides a useful reference point for assessing restoration success (Mitchell et al. 2000). Plant assemblages have historically been used to define different grassland types and so have been the targets used during restoration (Bakker and Berendse 1999; Young 2000). Considering restoration from the context of other taxa, however, provides a multitrophic context within which to define restoration success (Mortimer et al. 2002; Walker et al. 2004; Wassenaar et al. 2005; Woodcock et al. 2006). There was importantly some level of concordance with the result between the beetle and plant assemblages in terms of their restoration success. The control, where addition of seed would be by natural processes such as seed rain, was demonstrated to be of limited value during restoration, even when used in combination with soil disturbance. The potential for restoring species-rich mesotrophic grasslands without the use of soil disturbance, even where seed addition methods such as hay spreading were used, was also of questionable value. Even where soil disturbance and seed addition methods were used, there was evidence that alternative stable states may develop that would be uncharacteristic of the target grassland used to define restoration success (Mitchell et al. 2000; Hirst et al. 2005). The establishment of such alternative stable states would represent failure to restore the grassland given the criteria for success laid out in this study. Such new grassland communities may still, however, have a conservation value that is much greater than that of pre-restoration agriculturally improved mesotrophic grasslands. Enhancement of the biodiversity value of these improved grasslands may therefore be a much more realistic goal for these management practices, particularly over the duration of this study.

Acknowledgments

This project was funded by the UK Department for the Environment, Food and Rural Affairs (Defra) as part of project BD1441. We thank Keith Datchler of the Beech Estate for permission to use the experimental sites and his support and enthusiasm throughout the project, Chris and Roland Davis (Agrifactors Southern Ltd.) for assistance with installing the treatments, and Dawn Brickwood of the High Weald Meadows Initiative for continued support. Victoria Chapman, Tracy Gray, Katherine Robertson and Corin Wilkins assisted with sampling. Special thanks to Darren Mann, James Hogan and George McGavin at the Hope Entomological Collections, University Museum Oxford.

Supplementary material

442_2007_872_MOESM1_ESM.doc (233 kb)
ESM (DOC 233 kb)

Copyright information

© Springer-Verlag 2007