Study Site an Experimental Design
The re-creation experiment was undertaken at Aston Rowant, Oxfordshire, UK (Lat: 51:41:23N; Lon 0:56:35W) on land previously managed for cereal production under conventional arable agricultural practices. In 2001, this field was taken out of arable production and a replicated block design experiment was set up to investigate the effect of introducing a grass-only seed mixture (2 factor levels) vs. seeds of local provenance (5 factor levels). Forty experimental plots, each of 10 x 10 m, were established in four blocks of 10 plots, each separated from any adjacent plots by 5 m. Within each block the 2 x 5 replicated block design of treatments was established.
Treatments one had two factor levels represented by a control and a grass-only seed mix sown at a rate of 25 kg/ha. This was a commercially available grass seed mix (Emorsgate Seeds, Ltd., UK) typical of those used in agri-environment scheme options for chalk grassland creation in the UK (DEFRA 2005). These seeds were not of local provenance but originated from populations with southern England. By weight the seed mix contained the grass species Agrostis capillaris L. (5%), Cynosurus cristatus L. (20%), Dactylis glomerata L. (10%), Festuca ovina L. (15%), Festuca rubra L. (20%), Phleum pratense L. (10%), Poa pratensis L. (20%). The seed mix was applied only in August 2001, prior to any of the subsequent management practices applied to the experimental plots.
The second treatment was the introduction of local provenance seeds harvested from an agriculturally unimproved chalk grassland (Aston Rowant National Nature Reserve, Oxfordshire, UK) adjoining the experimental field. This grassland was classified as a CG2 Festuca ovina - Avenula pratensis grassland under the UK's National Vegetation Classification (Rodwell 1992), and had a forb rich sward, that included the species Campanula rotundifolia L. (Campanulaceae), Sanguisorba minor Scop. (Rosaceae) and Scabiosa columbaria L. (Dipsacaceae). To ensure full phenological development of the plants present within this donor site the grassland remained ungrazed in 2001 to allow for the effective harvesting of seeds.
Local provenance seeds were then harvested from the donor site using two methods, hay spreading and brush harvesting. The local provenance seed treatment had five factor levels: 1) a control; 2-3) hay spreading at high and low rates; 4-5) brush harvesting at high and low rates. Hay was harvesting using conventional grass cutting machinery in August and October of 2001 and applied to the experimental plots within 1 week of each harvest (Jones et al. 1999). Brush harvested seeds were collected again in August and October using specialised machinery that strips seeds from the sward using rotating brushes, therefore avoiding the need to harvest the bulk of the foliage (Morgan and Collicutt 1994). The seeds were then air-dried, cleaned and broadcast onto the plots. The low rate of application for both hay and brush harvesting 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. The August and October applications of hay or brush harvested seeds were applied respectively to only one half of each experimental plot. However, as all subsequent analyses are based on the whole plot this split plot effect has been ignored.
Throughout the duration of the study all experimental plots were grazed by sheep at rates of between 12 - 15 ewes ha-1. Grazing occurred in blocks of approximately 4-5 weeks, commencing in May of each year. Once the sward had been grazed to a height of 4-7 cm grazing was ended for a 3-4 month period until the re-growth of the sward had reached a height > 15-20 cm. No grazing was permitted over the winter period and the site was not cut for hay. Some additional horse grazing on the site occurred, although on an ad hoc basis and at a low intensity. This grazing rate was approximately equivalent to stocking densities of 0.25 livestock units ha-1 yr-1 that have been recommended for UK lowland calcareous grassland (Crofts and Jefferson 1999).
Plant community composition was recorded for all experimental plots in August from 2002 to 2004. In each split-plot, ten 0.5 x 0.5 m quadrats were randomly positioned at least 1 m from the edge of the split-plot and the presence of all vascular plant species recorded. Plant species were scored on a scale of zero to ten based on their frequency in the ten quadrats. The botanical composition of the donor site was assessed in 2004 using the same methodology in 15 randomly positioned 10 x 10 m 'dummy' plots in each of the two blocks of vegetation used. Plant nomenclature follows Stace (1997).
Phytophagous beetle assemblages were sampled from the experimental plots using a Vortis suction sampler (Burkhard Ltd, Rickmansworth, UK). As with the plants, beetle sampling was carried out annually from 2002 to 2004. During each year, experimental plots were sampled on three occasions (May, July and September) to provide a comprehensive measure of the full seasonal variation in the beetle fauna. On each sampling occasion, the suction sampler was placed in 15 positions, representing a total sampled area of 0.3m2. For each of these individual positions the sampler was held in location for 10 seconds. Suction sampling is a quantitative method suitable for the collection of adult invertebrates inhabiting short grassland swards (Mortimer et al. 2002; Woodcock et al. 2005; Woodcock et al. 2006; Woodcock et al. 2008). All beetle counts were summed for individual years. Samples were also taken from the Aston Rowant NNR donor site based on a similar set of fifteen samples (each representing 15 individual suction samples) taken in May, July and September 2001. Each sample was separated by at least 15 m. The phytophagous beetles from the families Apionidae, Curculionidae, Chrysomelidae, and Bruchidae were identified to species following the nomenclature of Strejcek (1993) and Morris (2003).
Re-creation of the plant community and associated phytophagous beetle assemblages would be considered to be successful if the experimental plots were to contain the same species with the same frequencies as those found at the donor site. To provide a measure of the success of re-creation the species composition of the plants and beetles were compared to those of the donor site (Aston Rowant NNR) using Jaccard's similarity coefficient. Jaccard's similarity index (Sj) is a binary coefficient that considers only the presence or absence of species, and is defined as: Sj = a / (a + b + c) where: a = number of species shared by the donor site and the experimental plot; b = number of species in the donor site only; c = number of species in the experimental plot only (Krebs 1999).
For the beetles all analyses are based on summed values of either abundance or species richness for individual experimental plots for each year. For the plants mean frequency values were derived from the 10 quadrats taken within each experimental plot. The response of plant and beetle species richness to the treatments of 'grass-only seed mix' (GRASS-ONLY) and 'seeds of local provenance' (LOCAL SEEDS) were assessed using a temporal split-plot ANOVA. At the whole plot level the ANOVA tested the effects of block (4 levels), GRASS-ONLY (2 levels), LOCAL SEEDS (5 levels) and the interaction between GRASS-ONLY x LOCAL SEEDS. These whole plot factors were tested against the error term of block x GRASS-ONLY x LOCAL SEEDS. The temporal split-plot explanatory variables were year (3 levels), year x GRASS-ONLY, year x LOCAL SEEDS and year x GRASS-ONLY x LOCAL SEEDS. All species richness values were Loge N+1 transformed to normalise the data. These analyses were then repeated for the measures of Jaccard's similarity between the experimental plots and the donor site at Aston Rowant NNR for the plants and beetles. No transformations of Jaccard's similarity were used. Analyses were carried out in SAS 9.01. The post hoc comparisons of means presented on graphs were performed using the CONTRAST statement within SAS.
Using the linear ordination method, partial redundancy analysis (pRDA), changes in the structure of both the plant and phytophagous beetle assemblages were assessed in response to the re-creation management treatments. For the phytophagous beetles, species represented by only one individual were excluded from the analysis. For both beetle abundances and plant cover logi0 transformations of the data set were used. Following ter Braak and Smilauer (2002), temporal change in beetle assemblage structure was assessed based on interactions between management treatment variables and sample years (e.g. Env.Var.*2002, Env.Var.*2003 and Env.Var.*2004). Included as covariables within these analyses were sample year (2002, 2003 and 2004) and replicate block, with the later of these designated as a blocking factor. Monte Carlo permutation tests (999 permutations) were then used to assess for differences in assemblage structure in response to the fixed treatment effect interactions with sample year. To account for the repeated measures taken from the same experimental plots on subsequent years Monte Carlo permutations were restricted so that these repeated measures were treated as split-plots in time, in a similar fashion to the ANOVA model described above. Monte Carlo permutations were permitted only between whole plot factors. The two management treatments of GRASS-ONLY and LOCAL SEEDS were tested individually. For the GRASS-ONLY treatment, with its two factor levels, this required only one test on the interaction between year and the presence of the grass-only seed mix. For the LOCAL SEEDS treatments, which had five factor levels, separate tests for the interaction between 'control x year', 'low rate hay spreading x year', 'high rate hay spreading x year',
'low rate brush harvesting x year' and 'high rate brush harvesting x year' were required. To provide insight into the interaction between the two re-creation management treatments the ten separate interactions terms from the 2 x 5 factorial design of the experiment 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. The analysis was carried out in CANOCO 4.5.
Was this article helpful?