AOBPreview originally published online on March 12, 2008
Annals of Botany 2008 101(7):957-970; doi:10.1093/aob/mcn020
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Control of Pteridium aquilinum: Meta-analysis of a Multi-site Study in the UK
1 Centre for Evidence-Based Conservation, School of the Environment & Natural Resources, Bangor University, Bangor, Gwynedd, LL57 2UW, UK
2 Applied Vegetation Dynamics Laboratory, School of Biological Sciences, University of Liverpool, PO Box 147, Liverpool, L69 7ZB, UK
3 Macaulay Land Use Research Institute, Craigiebuckler, Aberdeen, AB15 8QH, UK
* For correspondence. E-mail g.stewart{at}bangor.ac.uk
Received: 16 October 2007 Returned for revision: 3 January 2008 Accepted: 18 January 2008 Published electronically: 12 March 2008
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONACKNOWLEDGEMENTSLITERATURE CITED |
|---|
Background and Aims: A great deal of money is spent controlling invasive weeds as part of international and national policies. It is essential that the funded treatments work across the region in which the policies operate. We argue that experiments across multiple sites are required to validate these programs as results from single sites may be misleading. Here, the control of Pteridium aquilinum (bracken) is used as a test example to address the following four questions. (1) Does the effectiveness of P. aquilinum-control treatments vary across sites? (2) Is the best treatment identified in previous research (cutting twice per year) consistent at all sites, and if not why not? (3) Is treatment performance related to P. aquilinum rhizome mass, litter cover or litter depth at the various sites? (4) Does successful P. aquilinum control influence species richness?
Methods Pteridium aquilinum: -control treatments were monitored for 10 years using six replicated experiments and analysed using meta-analysis. Meta-regressions were used to explore heterogeneity between sites.
Key Results: The effectiveness of treatments varied between sites depending on the measure used to assess P. aquilinum performance. In general, cutting twice per year was the most successful treatment but on some sites other, less expensive treatments were as good. The effectiveness of treatments at different sites was not related to rhizome mass, but the effectiveness of most applied treatments were inversely related to post-control litter. Effective treatment was also associated with high species richness.
Conclusions: It is concluded that successful development of national weed control programs requires multi-site experimental approaches. Here, meta-analyses demonstrate that variation in effectiveness between sites could be explained in part by pre-specified variables. Reliance on data from a single site for policy formulation is therefore clearly dangerous.
Key words: Bracken, Pteridium aquilinum, weed, control, herbicide, litter dynamics, species diversity, meta-analysis
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONACKNOWLEDGEMENTSLITERATURE CITED |
|---|
A fundamental limitation in vegetation management and restoration ecology is the ability to predict, with some degree of precision, the likely outcome of a proposed treatment across a range of sites. This is essential to meet the objectives of both policy-makers and those responsible for implementating practical management schemes. However, it is well known that accurate prediction is difficult when treatments may have to be applied over wide environmental conditions, and where the restoration objectives may vary according to the ecosystem being treated. An obvious solution is to carry out multi-site experiments where the same treatments are applied at the same time; however, there are relatively few examples of such an approach (Pywell et al., 2002; Le Duc et al., 2003; Pakeman, 2004). This is most problematic when the treatments have to be delivered via an Integrated Land Management Approach (Milligan et al., 2004), where restoration includes the control of invasive species coupled with techniques to restore more appropriate vegetation, and is especially difficult where the invasive weed is problematic, and success is variable.
The case-study problem
Pteridium aquilinum (bracken) is a serious invasive weed of upland and marginal land in many parts of the world (Pakeman and Marrs, 1992). In the UK, P. aquilinum is an especially problematic weed, causing problems in upland heaths and acid grasslands, where it often occurs in dense stands (summarized in Smith and Taylor, 1995). Pteridium aquilinum is difficult to control for a range of reasons, including an extensive rhizome system (Le Duc et al., 2003) and a high productivity that produces a dense frond cover and deep litter, which combine to reduce understorey vegetation (Marrs et al., 2000).
The current agricultural and conservation policy in the UK for P. aquilinum infestation is delivered through Agri-Environment Schemes (MAFF, 1993, 1996) and Biodiversity Action Plans (Anon, 1995a, b). Essentially, the policy is to reduce P. aquilinum infestations where possible and commonly to restore Calluna vulgaris-dominated heathland or acid grassland. Pteridium aquilinum is considered a mid-successional plant, usually occupying a niche between plagio-climax heath/moor/grassland and woodland (Marrs et al., 2000, Marrs and Watt, 2006), although it can persist for long periods and appears in some cases to represent a highly resilient stable state (Marrs et al., 2000; Suding et al., 2004; Marrs and Watt, 2006).
In the UK, P. aquilinum occurs across a wide range of conditions and, given that succession can occur along many trajectories (Mitchell et al., 2000; Suding et al., 2004), there is potentially a large number of possible outcomes for P. aquilinum control/vegetation restoration schemes. In addition, P. aquilinum communities are known to have high resilience (Cox et al., 2007, 2008; Stewart et al., 2007) and P. aquilinum control often produces variable and conflicting results (Le Duc et al., 2000; Cox et al., 2007). It is, therefore, important from a national policy perspective to understand the response of P. aquilinum to control treatment in a variety of bio-geographical situations.
Approach to this study
There are a number of approaches that can be applied to analyse such multi-site studies, including univariate analysis of variance (Le Duc et al., 2000; Pywell et al., 2002; Pakeman, 2004; Cox et al., 2007, 2008) and multivariate analysis (Cox, 2007). Recently there has been an increasing interest in the use of formal meta-analysis in ecological studies, specifically through the development of evidence-based frameworks and systematic review (Pullin and Knight, 2001, 2003; Sutherland et al., 2004; Pullin and Stewart, 2006). Meta-analysis is a statistical approach to the analysis of a collection of results from individual studies or sites for the purpose of integrating the findings and exploring variation in effects (Hedges and Olkin, 1985; Arnqvist and Wooster, 1995; Gurevitch and Hedges, 1999, 2001; Osenberg et al., 1999). The term has been used to describe any form of quantitative research synthesis, but also refers more specifically to a range of techniques that combine effect sizes using a weighting method. Meta-analytical approaches allow combination of heteroscedastic effect sizes and allow separation of within- and between-site variance, in contrast to ANOVA-type analyses, with a concurrent increase in power (Gurevitch and Hedges, 1999). Furthermore, the use of effect sizes and confidence intervals provide estimates of the precision and magnitude of an effect, in contrast to null-hypothesis significance tests (Shinichi and Cuthill, 2007).
Often, meta-analysis involves the comparison of experiments carried out by different researchers, in different situations and with heterogeneity in application methods, for example the review of asulam control of P. aquilinum (Stewart et al., 2007). In such analyses it is often difficult to attribute variation in the success of treatments between studies to environmental variation because of confounding by the different methodologies employed by different researchers at different sites (Stewart et al., 2005). However, data from a 10-year study of P. aquilinum control where six replicated experiments were carried out more-or-less simultaneously, with the same treatments being applied and results collected by the same team of observers (Le Duc et al., 2000; Cox et al., 2007) does not have these same problems. This dataset provides an opportunity to assess the value of meta-analytical approaches in ecological management and to assess the importance of inter-site variability with respect to treatment performance.
Here, meta-analytical approaches are used to address the following specific questions.
- (1) Does the effectiveness of P. aquilinum-control treatments vary across sites?
- (2) Is the best treatment identified in previous research (cutting twice per year, Marrs et al., 1998; Cox et al., 2007) consistent at all sites, and if not why not?
- (3) Is treatment performance related to P. aquilinum rhizome mass, litter cover or litter depth at the various sites?
- (4) Does successful P. aquilinum control influence species' richness?
- (2) Is the best treatment identified in previous research (cutting twice per year, Marrs et al., 1998; Cox et al., 2007) consistent at all sites, and if not why not?
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONACKNOWLEDGEMENTSLITERATURE CITED |
|---|
Six experiments were available for meta-analysis, as they had identical Pteridium aquilinum (L.) Kuhn control treatments applied; the experiments were set up at four different regional locations in the UK: Cannock Chase in the English Midlands; Hordron Edge in the North Peak Environmentally Sensitive Area, in the Peak District National Park, northern England; the Carneddau Estate, Snowdonia National Park, North Wales; and Sourhope in the Cheviot Hills, on the England/Scotland border. Sites are referred to hereafter as Cannock, Carneddau, Peak and Sourhope. Exact methodological details have been published elsewhere (Le Duc et al., 2000, 2003) so only a brief summary is given here (see Table 1 for details).
|
Experimental design and P. aquilinum-control treatments
Use of an all-terrain vehicle to apply the treatments constrained experimental layout, and accordingly split-plot designs were used throughout. Each experiment had two or three replicated blocks with the six P. aquilinum-control treatments applied randomly at the main-plot level. The vegetation restoration treatments were applied randomly within the sub- and sub-sub-plot levels, and are not discussed further here. The experiment at Peak had a split-split-plot design from the start, and the one at Carneddau was changed to a split-split-plot design during its course to accommodate additional treatments (Table 1).
The six main-plot, P. aquilinum-control treatments were: (1) untreated (experimental control); (2) cut once per year, in June; (3) cut twice per year, in both June and August; (4) a single June cut in year one followed by herbicide (asulam) spraying in year two (‘cut and spray’); (5) asulam in year one only (‘spray’); (6) asulam in year one followed by a single June cut in year two (‘spray and cut’). Combination treatments had only one application of each treatment, applied in years one and two. Generally, single cuts took place in June, second cuts and herbicide application in August.
Cutting was applied using a petrol-driven flail mower (Logic MFG series 300), trailed by a quad four-wheel-drive ATV (Kawasaki KLF 300) up to and including 1999, and subsequently using a hand-operated AEBI model HC55 flail mower. Asulam (Asulox, Bayer CropScience PLC) was sprayed in late August or early September at 4·4 kg active ingredient ha–1 (11 L Asulox ha–1) in 400 L water ha–1 using a standard knapsack sprayer. In two experiments (Sourhope 1, Cannock 2; Table 1) plots treated with a single application of asulam were re-treated with asulam in 1996, 3 years after first treatment. On that occasion the herbicide was applied using a weed wiper (Rotowiper, Bisset Engineering International Ltd., New Zealand) trailed by the ATV. In June 1997 flooding prevented access to Sourhope 2, and only a single cut (August) was applied that year.
Monitoring
All experiments were monitored in June and August before the application of any P. aquilinum-control treatments. Quadrats (1 m2) were placed at pre-selected random co-ordinates on 1-m grids within each sub- (sub-) plot. In June the cover of all species was estimated. In August a visual estimate of P. aquilinum cover was made in this 1 m2 quadrat and frond density and frond length were assessed in the central 0·25 m2; here, all fronds were cut at ground level, counted and length measured. Two quadrats were used in each sub-plot when the experiment was split to sub-plot level; when further divided, only one quadrat was used per sub-sub-plot.
Data analysis and presentation
Mean values for each plot in each block were obtained after appropriate transformation of raw data; frond density using (Y + 0·5)0·5, frond length and cover using ln(Y + 1). The means were then back-transformed for meta-analysis. Change over time was averaged across years to avoid post hoc rationalization regarding the choice of temporal end point. Sub- and sub-sub-plot treatments were disregarded to prevent any confounding impacts. The comparative impacts of the applied treatments were analysed by meta-analysis and meta-regression (Cooper and Hedges, 1994; Deeks et al., 2001; Gurevitch and Hedges, 2001); these analyses were performed on three P. aquilinum variables: cover (%), mean frond density (number m–2) and frond length (cm). Cohen's D effect sizes (difference between population means divided by average population standard deviation; Deeks et al., 2001) were derived from the treatment and control means, standard deviations and sample sizes. Sensitivity analyses were used to verify the robustness of the effect size estimator in comparison to the Glass and Hedge's effect size estimates (Cooper and Hedges, 1994) and there were no major differences between the effect size metrics. Response ratios were not used for this analysis as we were concerned with absolute rather than relative differences, particularly in relation to the meta-regressions. As the aim of this study was to determine the variation in effect between sites, a random effects model was used (Gurevitch and Hedges, 1999). This method uses the standardized mean difference method to express the size of the treatment effect in each study relative to the variability observed in that study (Deeks et al., 2001). Here, the DerSimonian and Laird method was used to perform this analysis (DerSimonian and Laird, 1986; Cooper and Hedges, 1994). The usual DerSimonian and Laird estimates of the effects sizes (
i) and mean and variance (
2 ) are given by Deeks et al. (2001) as:
|
|
2 set to zero if Q < k – 1, and the weights (w) are calculated by inverse variance, i.e.: |
|
|
|
|
|
Questions 1 and 2 (treatment effectiveness across sites and assessing best treatment) were tested by inspection of Forrest plots of the estimated treatment effects from the studies along with their 95 % confidence intervals, and by formal tests of homogeneity undertaken prior to each meta-analysis (Thompson and Sharp, 1999). Questions 3 and 4 (testing relationships between treatment performance and explanatory variables, including species richness) were tested by examining the associations of treatment effects with environmental variables that had been collected at the individual treatment plot level (P. aquilinum rhizome mass, litter cover, litter depth and species' richness), previously postulated to affect both P. aquilinum control and restoration success (Marrs et al., 1998). These associations were examined using univariate random effects SMD meta-regression in Stata version 8·2 (Stata Corporation, USA, 2003) using the program Metareg (Sharp, 1998). The association of treatment effect derived from treatment–control comparisons (n = 5) with explanatory variables (n = 4) and outcome measures (n = 3) required 60 regression analyses, necessitating the use of Bonferroni correction for each outcome measure, which were treated as independent metrics (Sankoh et al., 1997). The multiple comparison approach was preferred over the inclusion of treatment as a factor in order to retain independence, and to minimize the already high heterogeneity within the analyses. This allowed the exploration of heterogeneity by considering factors in relation to each treatment comparison separately, the most parsimonious approach.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONACKNOWLEDGEMENTSLITERATURE CITED |
|---|
The use of meta-analysis is becoming increasingly prevalent in ecology (Hedges and Olkin, 1985; Arnqvist and Wooster, 1995; Gurevitch and Hedges, 1999, 2001; Osenberg et al., 1999) but not all ecologists are familiar with the output of such analyses. The results presented in Table 2 summarize the overall impact of different treatments across the six study sites, with the diagonal structure resulting from comparison of all permutations of five treatment comparisons. The pooled effect size (D) is significant when the confidence intervals do not cross zero (Table 2, Figs 1 and 2). Effect sizes smaller than 0·2 are considered small by convention (Cohen, 1988), although there is no empirical basis for appraising ecological effect sizes for D. Thus the majority of comparisons presented illustrate moderate-to-large statistically significant effects (Table 2). The statistical significance of variation between sites is tested using the Q statistic following a chi-square distribution under the null hypothesis that the true treatment effect is the same for all sites. The majority of the comparisons vary significantly between sites (see below). The Forrest plots (Figs 1, 2) illustrate the effectiveness of an applied treatment at each site, the overall pooled effect of treatment, and the variation in effect both within and between sites. Further information regarding interpretation of Forrest plots is available at (www.cebc.bangor.ac.uk, evidenced-based conservation, review guidelines).
|
|
|
2 measure of heterogeneity). Significant results are indicated in bold (P < 0.05).Question 1: does the effectiveness of P. aquilinum-control treatments vary across sites?
There was significant variation (P < 0·05) between sites for almost all treatments relative to their respective untreated control for all three P. aquilinum frond response variables (Table 2, Fig. 1); the two exceptions were (1) the cut twice per year treatment for frond length and (2) cut and spray for frond density. Cutting twice per year is the most effective treatment overall for reducing cover and height (biggest significant effect size, Table 2, Fig. 1), but spraying once is more effective in reducing frond density (Table 2, Fig. 1).
There was also significant variation (P < 0·05) between sites for other treatment comparisons but these varied with the outcome measurement. For cover (Table 2A) there was variation between sites when comparing the cut twice per year treatment versus spraying and cutting. For frond length, there was significant variation between sites for most treatment comparisons (Table 2B), the exceptions being cutting once or twice per year versus cutting and spraying. For frond density there were significant site effects for the untreated control versus all treatments except cut and spray (Table 2C). For the other frond density treatment comparisons the picture was variable, with the following being significant: (1) cutting once per year versus cut and spray and spraying, (2) cutting twice per year versus cut and spray, spray and cut versus spray and cut and spray.
Question 2: is the best treatment identified in previous research (cutting twice per year) consistent at all sites, and if not why not?
The effectiveness of cutting twice per year varies significantly (P < 0·05) between sites in comparison with uncut controls for cover and frond density, but not for frond length (Table 2, Fig. 2). The relative effectiveness of cutting twice per year does not generally vary significantly (P < 0·05) except in comparison with sprayed and cut (cover and frond length), sprayed (frond length) and cut and sprayed (frond density; Table 2, Fig. 2). Thus, although the effectiveness of cutting varies across sites, it remains the most effective treatment in nearly all cases.
Question 3: is treatment performance related to P. aquilinum rhizome mass, litter cover or litter depth at the various sites?
The effectiveness of P. aquilinum control was related to measured rhizome mass only for frond length in the cut-twice treatment (P < 0·05, Table 3A). This reduction in frond length associated with the cut-twice treatment produced an increase in effect size of 0·42 per unit increase in rhizome mass (Table 3A) but the result is not significant (at the alpha level of P < 0·0025) when the multiple comparisons are accounted for with Sidak's or Bonferroni adjustment. The relationships of the control treatments with litter cover and depth were varied, with significant negative results found in seven and eight comparisons, respectively, although the coefficients were small (Table 3). Adjustment for multiple comparisons reduces the number of significant relationships to one and three for litter cover and depth, respectively (Table 3). Where a significant regression coefficient was found, the P. aquilinum-control treatment reduced the litter variable relative to the untreated experimental control.
|
Question 4: does successful P. aquilinum control influence species richness?
The effectiveness of P. aquilinum-control treatment was significantly related to species' richness (species' richness ranged from 12–38 species per site) in 12 out of 15 comparisons (eight of 15 were significant following adjustment for multiple comparisons, Table 3D). Whilst the regression coefficients are low, these are statistically significant and indicate that effective P. aquilinum control is associated with species' rich sites.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONACKNOWLEDGEMENTSLITERATURE CITED |
|---|
The results presented here demonstrate the usefulness of the meta-analytical approach for investigating the significance of applied ecological restoration treatments in a multi-site study. Often, meta-analysis is used to compare treatment effects derived from different studies (Cooper and Hedges, 1994; Gurevitch and Hedges, 1999), with systematic reviews using meta-analysis to synthesize and summarize research findings to support evidence-based conservation (Stewart et al., 2005; Pullin and Stewart, 2006). Formal systematic reviews have recently been advocated in an attempt to avoid the biases associated with previous ecological meta-analyses (Leimu and Koricheva, 2004, 2005). Here, the approach was used to investigate hypotheses within a series of individual experiments at different sites within the UK, which had all been run with common treatments and methodologies. The results allowed cross-site comparisons of applied treatments using a formal statistical approach. There are of course some drawbacks to this approach, one of which is the time frame over which the study effect size is derived. Here, we used a 10-year period, partly because there were 10 years of data available, but also because the results reflected a longer-term impact. As a result, shorter-term success of some treatments may have been under-estimated, and this is likely in the treatments where asulam was applied because there is often a good initial effect followed by subsequent regrowth (Marrs et al., 1998).
The results from the inter-site comparison of all treatments against the untreated controls showed three important results. First, there were significant reductions in at least one P. aquilinum performance measure relative to untreated controls for all applied treatments on most sites. Second, there were significant differences in response between sites. Third, different responses were found for some site x treatment comparisons depending on the P. aquilinum measure used (frond length, frond density, frond cover). Thus, in answer to our first question, the effectiveness of P. aquilinum control varies between sites, confirming the predictions of other workers (Pottier et al., 2005).
Comparisons of all treatments at all sites revealed that cutting twice within a year was usually the most effective treatment, but there were a few site x treatment combinations where other treatments were more effective at some sites depending on the measure used. Where P. aquilinum had been cut twice per year, the effective control obtained was at the expense of applying a total of 20 treatments over the 10-year period of the study. However, where other treatments were successful, fewer applications (one or two) were used, with obvious financial implications. Unfortunately, we cannot at this point predict where these treatments are likely to be superior, although the meta-regression gives some indications. Thus, the answer to our second question is that cutting twice within a year is generally the most effective treatment of those tested, but further research is required to identify the situations when this is not the case.
We also investigated the relationships between applied treatments and a range of other P. aquilinum variables (rhizome mass, litter cover, litter depth) that have been shown to be important in both P. aquilinum control and the subsequent regeneration on semi-natural communities (Marrs et al., 1998). The relationship with the rhizome mass was studied because ultimately the rhizomes are the main part of the P. aquilinum-control problems (Pakeman et al., 1994; Pakeman and Marrs, 1996; Pottier et al., 2005). Here, there was a positive relationship between rhizome mass and the effectiveness of the cut twice per year treatment, supporting both the hypotheses underlying questions 2 and 3. Litter variables were studied because they represent a barrier to the regeneration of new plant communities (Lowday and Marrs, 1992; Marrs et al., 2007). Our results supports the hypothesis that most site x treatment combinations reduce the depth and cover of P. aquilinum litter, which indicated that the sites showed some reduction in litter and hence should be more amenable to colonization by other species.
We also showed that there were significant positive effects of P. aquilinum control on species' richness, which answers question 4. Here, the regression coefficients are small, but they are highly significant. This is an important result because it indicates that one of the major outcomes of the P. aquilinum-control strategy is the development of a plant community with greater plant species' diversity than under dense P. aquilinum cover. Details of the vegetation development of these experiments have been described elsewhere (Cox et al., 2007; Le Duc et al., 2007).
A major outcome of this study is the clear need for management experiments to be repeated in different places in order to develop evidence-based policy decisions, especially when such information is to be used to develop national guidelines and funding strategies (Sutherland et al., 2004; Anon, 2005; Pullin and Stewart, 2006). Too often, management conclusions are extrapolated from limited numbers of sites or single sites (e.g. Viggers and Hearn, 2005; Chamaillé-Jammes et al., 2007; Larsen and Guillemette, 2007). Our results here show that there is considerable site variation, possibly caused by differences in climatic regime, substrate, and past and current management practices, which all influence the starting position before management is applied and the desired end-point achieved (Marrs et al., 2000; Marrs and Watt, 2006).
There is also an obvious need to carry out further work to ascertain why P. aquilinum is so variable, and why in some places it is difficult to control whereas in others it is apparently less difficult (Pakeman, 2004). This is particularly important for P. aquilinum in view of its potential health implications for livestock (Marrs and Watt, 2006), its predicted increase under future climate change scenarios, and its potential to increase in area and density as a result of land-use changes resulting from the reduced stocking rates encouraged in some agri-environment schemes (Pottier et al., 2005).
We re-emphasize the utility of meta-analyses in providing increased control of Type-II errors (i.e. the probability of accepting a wrong null hypothesis; Cohen, 1988), and in enabling assessment of the magnitude of an effect as well as its statistical significance (Glass et al., 1981; Rosenthal, 1984; Gurevitch and Hedges, 1999, 2001). In contrast to more traditional ANOVA-type analyses, meta-analysis allows the practical importance of an experimental effect to be assessed (Harris and Rosenthal, 1985). We suggest that the use of meta-analyses should not be restricted to research synthesis, but extended to the analysis of any applied multi-site study where inter-site variability is important.
|
ACKNOWLEDGEMENTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONACKNOWLEDGEMENTSLITERATURE CITED |
|---|
We thank DEFRA and NERC of the UK for funding this work, and the land managers who permitted access to our experimental sites: Cannock, Staffordshire County Council (Sue Sheppard); Peaks (Jeremy Archdale and Neil Taylor); Carneddau (National Trust); and Sourhope (Macaulay Institute). We also thank Julia Koricheva, David Causton and an anonymous reviewer for their valuable and constructive, critical appraisal of this work.
|
LITERATURE CITED |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONACKNOWLEDGEMENTSLITERATURE CITED |
|---|
-
Anon. Action plans (1995) a II. London: HMSO. Biodiversity UK Steering Group Report.
Anon. Biodiversity: meeting the Rio challenge (1995) b I. London: HMSO. UK Steering Group Report.
Anon. P. aquilinum control. In: vegetation restoration and land management (2005) London: RDS, HMSO. Rural Development Service Technical Advice, Note 23.
Arnqvist G, Wooster D. Meta-analysis – synthesizing research findings in ecology and evolution. Trends in Ecology and Evolution (1995) 10:236–240.[CrossRef]
Chamaillé-Jammes, Valeix M, Fritz H. Managing heterogeneity in elephant distribution: interactions between elephant population density and surface-water availability. Journal of Applied Ecology (2007) 44:625–633.[CrossRef][Web of Science]
Cohen J. Statistical power analysis for the behavioural sciences (1988) 2nd edn. Mahwah, NJ: Lawrence Erlbaum Associates.
Cooper H, Hedges LV. The handbook of research synthesis (1994) New York: Russell Sage Foundation.
Cox ES. The implications and effects of bracken (Pteridium aquilinum (L.) Kuhn) control on species diversity, re-vegetation and bracken performance (2007) Liverpool, UK: University of Liverpool. PhD thesis.
Cox ES, Marrs RH, Pakeman RJ, Le Duc MG. A multi-site assessment of Pteridium aquilinum control across the UK. Applied Vegetation Science (2007) 10:429–440.[Web of Science]
Cox ES, Marrs RH, Pakeman RJ, Le Duc MG. Factors affecting the restoration of heathland and acid grassland on Pteridium aquilinum-infested land across the United Kingdom: a multisite study. In: Restoration Ecology (2008) In press. doi: 10·1111/j.1526-100X.2007·00326.x.
Deeks JJ, Altman DG, Bradburn MJ. Statistical methods for examining heterogeneity and combining results from several studies in meta-analysis. In: Systematic reviews in health care: meta-analysis in context—Egger M, Smith GD, Altman DG, eds. (2001) London: British Medical Journal Publishing Group. 285–312.
DerSimonian R, Laird N. Meta-analysis in clinical trials. Controlled Clinical Trials (1986) 7:177–188.[CrossRef][Web of Science][Medline]
Glass GV, McGaw B, Smith ML. Meta-analysis in social research (1981) Beverly Hills, CA: Sage Publications.
Gurevitch J, Hedges LV. Statistical issues in ecological meta-analyses. Ecology (1999) 80:1142–1149.[CrossRef][Web of Science]
Gurevitch J, Hedges LV. Meta-analysis. Combining results of independent experiments. In: Design and analysis of ecological experiments—Scheiner SM, Gurevitch J, eds. (2001) Oxford: Oxford University Press. 347–369.
Harris MJ, Rosenthal R. Mediation of interpersonal expectancy effects: 31 meta-analyses. Psychological Bulletin (1985) 97:363–386.[CrossRef][Web of Science]
Hedges LV, Olkin I. Statistical methods for meta-analysis (1985) London: Academic Press.
Larsen JK, Guillemette M. Effects of wind turbines on flight behaviour of wintering common eiders: implications for habitat use and collision risk. Journal of Applied Ecology (2007) 44:516–522.[CrossRef][Web of Science]
Le Duc MG, Pakeman RJ, Putwain PD, Marrs RH. The variable responses of P. aquilinum fronds to control treatments in Great Britain. Annals of Botany (2000) 85:17–29. (Supp. B).
Le Duc MG, Pakeman RJ, Marrs RH. Changes in the P. aquilinum rhizome system subjected to long-term experimental treatment. Journal of Applied Ecology (2003) 40:508–522.[CrossRef][Web of Science]
Le Duc MG, Pakeman RJ, Marrs RH. A restoration experiment on moorland infested by Pteridium aquilinum: plant species responses. Agriculture, Ecosystems & Environment (2007) 119:53–59.[CrossRef][Web of Science]
Leimu R, Koricheva J. Cumulative meta-analysis: a new tool for detection of temporal trends and publication bias in ecology. Proceedings of the Royal Society of London, Series B (2004) 271:1961–1966.[Medline]
Leimu R, Koricheva J. What determines the citation frequency of ecological papers? Trends in Ecology and Evolution (2005) 20:28–32.[CrossRef]
Lowday JE, Marrs RH. Control of bracken and the restoration of heathland. III. Bracken litter disturbance and heathland. Journal of Applied Ecology (1992) 29:212–217.[CrossRef][Web of Science]
MAFF. Environment matters. Evironmentally Sensitive Areas (1993) London: MAFF.
MAFF. Envirnoment matters. The Countryside Stewardship Scheme: information for farmers and land managers (1996) London: MAFF.
Marrs RH, Watt AS. Biological flora of the British Isles: Pteridium aquilinum (L.) Kuhn. Journal of Ecology (2006) 94:1272–1321.[CrossRef][Web of Science]
Marrs RH, Johnson SW, Le Duc MG. Control of bracken and restoration of heathland. VII. The response of bracken to 18 years of control or six years of control and followed by recovery. Journal of Applied Ecology (1998) 35:748–757.[Web of Science]
Marrs RH, Le Duc MG, Mitchell RJ, Goddard D, Paterson S, Pakeman RJ. The ecology of P. aquilinum: its role in succession and implications for control. Annals of Botany (2000) 85:3–15. (Supp. B).
Marrs RH, Galtress K, Tong C, Cox ES, Blackbird SJ, Heyes TJ, Pakeman RJ, Le Duc MG. Competing conservation goals, biodiversity or ecosystem services: element losses and species recruitment in a managed moorland-bracken model system. Journal of Environmental Management (2007) 85:1034–1047.[CrossRef][Web of Science][Medline]
Milligan AL, Putwain PD, Cox ES, Ghorbani J, Le Duc MG, Marrs RH. Developing an integrated land management strategy for the restoration of moorland vegetation on Molinia caerulea-dominated vegetation for conservation purposes in upland Britain. Biological Conservation (2004) 119:371–387.[CrossRef][Web of Science]
Mitchell RJ, Auld MHD, Le Duc MG, Marrs RH. Ecosystem stability and resilience: a review of their relevance for the conservation management of lowland heaths. Perspectives in Plant Ecology, Evolution & Systematics (2000) 3:142–160.[CrossRef]
Osenberg CW, Sarnelle O, Cooper SD, Holt RD. Resolving ecological questions through meta-analysis: goals, metrics and models. Ecology (1999) 80:1105–1117.[Web of Science]
Pakeman RJ. Consistency of plant species and trait responses along a productivity gradient: a multi-site analysis. Journal of Ecology (2004) 92:893–905.[CrossRef][Web of Science]
Pakeman RJ, Marrs RH. The conservation of bracken, Pteridium aquilinum (L.) Kuhn -dominated communities in the UK, and an assessment of the ecological impact of bracken expansion or its removal. Biological Conservation (1992) 62:101–114.[CrossRef][Web of Science]
Pakeman RJ, Marrs RH. Modelling the effects of climate change on the growth of bracken (Pteridium aquilinum) in Britain. Journal of Applied Ecology (1996) 33:561–575.[CrossRef][Web of Science]
Pakeman RJ, Marrs RH, Jacob PJ. A model of bracken (Pteridium aquilinum) growth and the effects of control strategies and changing climate. Journal of Applied Ecology (1994) 31:145–154.[CrossRef][Web of Science]
Pottier J, Pakeman RJ, Le Duc MG, Marrs RH. Validation of a model for predicting Pteridium aquilinum (L.) Kuhn production and control across the UK. In: BCPC International Congress – Crop Science & Technology 2005—Anon, ed. (2005) Alton, Hampshire, UK: British Crop Protection Council. 987–994.
Pullin AS, Knight TM. Effectiveness in conservation practice: pointers from medicine and public health. Conservation Biology (2001) 15:50–54.[CrossRef][Web of Science]
Pullin AS, Knight TM. Support for decision making in conservation practice: an evidence-based approach. Journal for Nature Conservation (2003) 11:83–90.[CrossRef]
Pullin AS, Stewart GB. Guidelines for systematic review in conservation and environmental management. Conservation Biology (2006) 20:1647–1656.[CrossRef][Web of Science][Medline]
Pywell RF, Bullock JM, Hopkins A, Walker KJ, Sparks TH, Burke MJW, Peel S. Restoration of species rich grassland: assessing the limiting processes using a multi-site experiment. Journal of Applied Ecology (2002) 39:294–309.[CrossRef][Web of Science]
Rodwell JS. British plant communities, Vol. 1. Woodlands and scrub (1991) a. Cambridge: Cambridge University Press.
Rodwell JS. British plant communities, Vol. 2. Heaths and mires (1991) b. Cambridge: Cambridge University Press.
Rodwell JS. British plant communities, Vol. 3. Grassland and montane communities (1992) Cambridge: Cambridge University Press.
Rosenthal R. Meta-analytical procedure for social research (1984) Beverly Hills, CA: Sage Publications.
Sankoh AJ, Huque MF, Dubey SD. Some comments on frequently used multiple endpoint adjustments methods in clinical trials. Statistics in Medicine (1997) 16:2529–2542.[CrossRef][Web of Science][Medline]
Sharp S. Meta-analysis regression: statistics, biostatistics, and epidemiology. Stata Technical Bulletin (1998) 42:16–22.
Shinichi N, Cuthill IC. Effect size, confidence interval and statistical significance: a practical guide for biologists. Biological Reviews (2007) 82:591–605.[Medline]
Smith RT, Taylor JA. Bracken: an environmental issue (1995) Leeds: University of Leeds.
Stace C. New flora of the British Isles (1997) Cambridge: Cambridge University Press.
Stata Corporation. Stata statistical software: release 8·2 (2003) College Station, TX: Stata Corporation.
Stewart GB, Coles CF, Pullin AS. Applying evidence-based practice in conservation management: lessons from the first systematic review and dissemination projects. Biological Conservation (2005) 126:270–278.[CrossRef][Web of Science]
Stewart GB, Pullin AS, Tyler C. The effectiveness of asulam for bracken (Pteridium aquilinum) control in the UK: A meta-analysis. Environmental Management (2007) 40:747–760.[Medline]
Suding KN, Gross KL, Houseman GR. Alternative states and positive feedbacks in restoration ecology. Trends in Ecology and Evolution (2004) 19:46–53.[CrossRef]
Sutherland WJ, Pullin AS, Dolman PM, Knight TM. The need for evidence-based conservation. Trends in Ecology and Evolution (2004) 19:305–308.[CrossRef]
Thompson SG, Sharp SJ. Explaining heterogeneity in meta-analysis: a comparison of methods. Statistics in Medicine (1999) 18:2693–2708.[CrossRef][Web of Science][Medline]
Viggers KL, Hearn JP. The kangaroo conundrum: home range studies and implications for land management. Journal of Applied Ecology (2005) 42:99–107.[CrossRef][Web of Science]
CiteULike 
Connotea 
Del.icio.us What's this?
Related articles in Ann Bot:
- ContentSnapshots
Ann Bot 2008 101: NP.[Extract] [Full Text]
This article has been cited by other articles:
|
|
 |
|
  G. Stewart Meta-analysis in applied ecology Biol Lett, September 23, 2009; (2009) rsbl.2009.0546v1. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||