Overcoming resistance and resilience of an invaded community is necessary for effective restoration: a multi-site bracken control study

Authors


Correspondence author. E-mail: calluna@liv.ac.uk

Summary

  1.  The search for appropriate management strategies to control invasive plants is an important theme in environmental management. However, the recovery of the resident community species complement does not necessarily respond predictably to restoration efforts. Increasing restoration success requires an understanding of the resistance and resilience of the invaded community and the response of the newly developing community to management. Here, we used Pteridium-invaded heath and grass communities as a test system and investigated the effects of recommended Pteridium aquilinum control treatments on vegetation composition and diversity.
  2.  We evaluated seven field experiments in four regions of Great Britain designed to test five Pteridium control treatments, including ‘one-off’ (applied only at the start) and ‘repeated’ (applied regularly) treatments, against an untreated experimental control. The sites had context-dependent restoration targets, either a Calluna heathland or acid grassland. Species cover and diversity responses (higher plants, mosses plus lichens) to these treatments were monitored annually for 10 years.
  3.  Pteridium control treatments induced significant change in species composition compared to experimental controls in both vegetation types. On Calluna target sites, ‘repeated’ treatments overcame the resistance of the invaded community producing a gradual divergence in species composition and species diversity. In contrast, the ‘one-off’ treatments were ineffective.
  4.  At the acid grassland target sites, all treatments overcame the resistance of the Pteridium-dominated state producing changes in species composition in comparison with experimental controls.
  5.  Synthesis and applications. There are two important results for land managers: (i) where Calluna heathland is the target, ‘repeated’ treatments (cutting once or twice per year) were effective in overcoming the resistance of invaded community and moving species composition towards the target state, effectively creating an alternative state; (ii) where acid grassland is the target both ‘one-off’ and ‘repeated’ treatments overcame the invaded community resistance (‘one-off’ also overcame resilience) producing changes in species composition in the desired direction. The effectiveness of ‘one-off’ treatments was site dependent and produced alternative stable states within 10 years. In contrast, ‘repeated’ treatments were site independent but took longer to work and were more expensive.

Introduction

The search for appropriate management strategies to control invasive plants is an important theme in ecology and environmental management and of considerable applied relevance (Vitousek et al. 1996). Where native or non-native species invade an ecosystem, they can degrade the ecosystem by displacing vulnerable native species and influencing nutrient cycling, hydrology and energy budgets (Mack et al. 2000). Hence, conservation efforts are often targeted to restore these degraded systems; however, a major problem is that the recovery of the resident community species complement does not necessarily respond in a predictable manner to restoration efforts for controlling invasive plant species (Suding, Gross & Houseman 2004). Understanding the responses of invaded communities to restoration treatment is, therefore, essential for developing management strategies to conserve biodiversity and maintain ecosystem services (Seastedt, Hobbs & Suding 2008).

From a theoretical perspective, the aims of ecological restoration should be to remove the invasive species and recreate a more appropriate native community (SERI 2004). This involves overcoming the resistance (i.e. the ability of the ecosystem to avoid displacement) and resilience (i.e. the ability of the original ecosystem to return to its original state after disturbance) of the invaded community and moving it towards a new community with sufficient resistance and resilience to prevent a return to the original invaded state (Mitchell et al. 2000), essentially to create an alternative stable state (ASS; sensu Beisner, Haydon & Cuddington 2003). ASS theory predicts that ecosystems can exist in multiple states under the same external environmental conditions, moving from one stable state to another only by a large perturbation (Beisner, Haydon & Cuddington 2003). This can be performed by altering either the ecosystem state variables (e.g. species composition) or the ecosystem parameters (e.g. removal of grazing by fencing); the aim being to force the existing undesired ecosystem to change to an alternative desired state (Suding, Gross & Houseman 2004). Where a state is changed as a result of a repeated treatment application, the new state produced may be dependent on this treatment. Hence, the new alternative state (AS) produced may revert to the original if the treatment is stopped. If the resistance to change of the invaded community is low, the restoration treatment may be enough to allow the successional trajectory to reach a new set of state variables; but, where the treated community has a high resilience, the invaded community re-establishes rapidly (Marrs et al. 2000), thus wasting restoration effort. Therefore, knowledge of the resistance or resilience of an invaded community to restoration treatments can provide insights into both fundamental community dynamics and the effectiveness of restoration towards the desired state (Suding, Gross & Houseman 2004).

Traditional approaches to managing degraded ecosystems dominated by invasive plant species have consisted of controlling the invasive species (e.g. with herbicides) and mitigating any disturbance that is likely to maintain its dominance (i.e. modify grazing or fertilization), with the aim of altering the system state variables (e.g. species composition) (Firn, House & Buckley 2010). Intervention methods targeting the plant invader can be described as either an initial treatment or a set of treatments applied at the start of any programme (e.g. a single herbicide or cutting treatment, here termed ‘one-off’) or treatments applied regularly (e.g. cutting applied annually, here termed ‘repeated’ treatments). By killing or reducing the biomass of the invasive species, these treatments create a ‘window of opportunity’ for native species to fill the niche space left by the invader (Pickett, Cadenasso & Meiners 2009), facilitating change in community structure, composition and diversity (Folke et al. 2004). However, to achieve a persistent effect, the resistance threshold of the ecosystem needs to be crossed, and there has to be a movement into a different alternative state. Where this occurs, the probability of the ecosystem returning back to its degraded state will be reduced (Spiegelberger et al. 2006). Surprisingly, there have been few attempts to assess the effectiveness of ‘one-off’ and ‘repeated’ treatments in overcoming the resistance and/or resilience of invaded ecosystems and shifting their species composition towards a stable native ecosystem of enhanced conservation value.

Pteridium aquilinum is a serious invasive weed of upland and marginal land in many parts of the world (Marrs & Watt 2006), including Great Britain, where it often occurs in dense stands and considerably reduces the conservation value of upland heaths and acid grasslands. The current agricultural and conservation policies in the UK are to reduce P. aquilinum infestation and restore either Calluna heathland or acid grassland. However, P. aquilinum is difficult to control because it has high productivity producing a dense frond cover and deep litter, which combine to reduce understorey vegetation (Marrs et al. 2000) and an extensive underground rhizome system with large carbohydrate reserves (Le Duc, Pakeman & Marrs 2003). It is, therefore, important from a policy perspective to identify the optimal P. aquilinum control treatments, as well as the impacts of control measures on recovery of native species. Even though control measures reduce dominance and/or abundance of the invader, they do not necessarily result in native species recovery. Where the invasive species are persistent perennial weeds (Marrs & Watt 2006), relatively long-term monitoring is needed to assess impacts on both the invasive species and the developing understorey flora; clearly, the impacts of ‘one-off’ and ‘repeated’ treatments may have different outcomes at short-, medium- and long-term time-scales (Spiegelberger et al. 2006).

In this paper, we address these issues by testing a range of recommended Pteridium control treatments (including ‘one-off’ and ‘repeated’) in a series of seven experiments over a 10-year period, with the focus on restoring two different vegetation targets: Calluna heathland and acid grassland. The key question was how do the different plant communities invaded by P. aquilinum respond to ‘one-off’ and ‘repeated’ treatments? Previous analyses of data from these experiments have assessed the effectiveness of the control treatments on P. aquilinum performance (Cox et al. 2007; Stewart et al. 2008) and on the developing understorey flora at the individual species level (Cox et al. 2008). Here, we focus on the effects of the Pteridium control treatments on the developing plant community composition and its diversity, with the aim of providing a national view of the impact of Pteridium control treatments on the changing plant communities. We hypothesized that (i) the Pteridium control treatments would result in different successional trajectories (for each vegetation target type) with varying rates of success based on the perturbation intensity of each treatment; (ii) ‘one-off’ or ‘repeated’ treatments would produce different plant compositional responses influenced by the resistance or resilience of the invaded community; and (iii) the Pteridium control treatments would affect the response of vascular and lower plant community differently based on the differential dispersal and growth capacities of each group.

Materials and methods

Site location

Seven stand-alone, replicated experiments were set up in four regions of Great Britain (Table 1); Cannock in the English midlands; Hordron Edge in the Peak District; Carneddau in North Wales and Sourhope in the Scottish Borders. There is no information on the time that P. aquilinum has been a dominant on these sites, but the restoration targets based on the surrounding vegetation were for Cannock and Peak a Calluna heathland (H12, Rodwell 1991) and for Carneddau and Sourhope an acid grassland (U4 or U5; Rodwell 1992). Exact methodological details are available in Le Duc, Pakeman & Marrs (2003); only a brief summary is given here (Table 1).

Table 1. Detailed description of experiments designed to test a range of Pteridium control and vegetation restoration methods across Great Britain. The experimental designs were all nested designs (all split- or split-split-plot over randomized block designs). All = the five Pteridium control treatments (Cut1pa, Cut2pa, CutSpray, Spray, SprayCut); square brackets enclose treatment start date. All experimental treatments at all levels were applied in balanced designs with appropriate untreated contrasts (for clarity these are not shown)
SitesSourhope 1Sourhope 2PeakCarneddauCannock 1Cannock 2Cannock 3
Location & site details
Ordnance Survey map referenceNT861202NT846210SK213870SH690711SJ976200SJ987181SJ987178
Elevation (m)325285290350145165175
Aspect (°)280130275190140125175
Slope (°)162292020189
Target vegetationAcid grasslandAcid grasslandCalluna heathlandAcid grasslandCalluna heathlandCalluna heathlandCalluna heathland
Experimental design
Blocks (replicates)2233222
Main-treatment plots6666662
Subtreatment plots2223332
Sub-subtreatment plots  32  2
Smallest plot size10 × 18 m10 × 18 m10 × 5 m10 × 5 m10 × 12 m10 × 12 m6 × 5 m
Total plots in location2424108108363616
Samples/plot/year3322333
Main treatments (Pteridium control)

All [Aug. 93];

Weed-wipe as follow-up to spray only [Sep.96]

All [Jul. 94]All [Jul. 93]All [Aug. 93]

All [Jul. 93];

Repeat: Spray, Spray & cut, Cut & spray [Aug. 99]

All [Aug. 93];

Weed-wipe as follow-up to spray only [Sep. 96]

Cut2pa [Jun. 95]
SubtreatmentsGrass seed [Aug. 93]Grass seed [Aug. 94]Stock fence [Jan. 94]Fertilize [Apr. 95] & Grass seed [Jun. 95]

Harrow + fertilize [Mar. 94];

Refertilize [May. 96]

Harrow + fertilize [Mar. 94]Litter-burn [Mar. 95]
Sub-subtreatmentsCalluna seed (brash + litter) [Nov. 93]Spot-spray [Aug. 97]Calluna seed (brash) [Dec. 96]

Experimental design

In each experiment, before treatment application, two or three replicate blocks were randomly located depending on the available area (Table 1). In six of the experiments, a randomized block experimental design was used with six Pteridium control treatments applied at the main-plot level (10 × 40 m). The six Pteridium control treatments were as follows: (i) untreated (experimental control); (ii) cut once per year in June (Cut1pa); (iii) cut twice per year in both June and August (Cut2pa); (iv) a single June cut in year one followed by asulam spraying in year 2 (CutSpray); (v) asulam in year 1 only (Spray); (vi) asulam in year 1 followed by a single June cut in year 2 (SprayCut). Generally, single cuts took place in June and second cuts and asulam application in August. Herbicide application was by knapsack sprayer (as Asulox, Bayer CropScience PLC; 4·4 kg active ingredient ha−1; 11 L Asulox in 400 L water ha−1). The seventh experiment (Cannock 3) also had a randomized block experimental design but had only two main-plot treatments: control and cut twice per year (Cut2pa).

The Pteridium control treatments applied in this study at the main-plot level reflect a combination of ‘one-off’ (CutSpray, Spray, SprayCut) and ‘repeated’ treatments (Cut1pa, Cut2pa). Unlike the main treatments, the sub- and sub-sub-plot vegetation restoration treatments were site specific and were applied in complete factorial combinations within nested designs (split-split-plot) and were designed to match individual site characteristics and required target vegetation (Table 1).

Monitoring

In 1993, species composition was measured in selected random sub-plots in each experiment to ensure that the vegetation and bracken variables were similar; these data were not used here. The vegetation in all experiments was monitored in June from 1994 to 2003, that is, before the application of the ‘repeated’ treatments. Quadrats (1 × 1 m) were placed at two or three pre-selected random coordinates on 1-m grids within each sub-(sub-)plot (Table 1), and the cover of all vascular plant, bryophytes and lichen species recorded. For species nomenclature details, see Appendix S1 (Supporting information).

Data analysis

Statistical analyses were performed in the R software environment (v.2.12.2; R Development Core Team 2011), using the nlme package for linear mixed models (LMM; Pinheiro et al. 2011) and the vegan package for multivariate analyses (Oksanen et al. 2011).

The species data set was analysed using both multivariate and univariate methods. In the multivariate analysis, all species that only occurred in <5% of the quadrats were removed before analysis, and a log-transformation (loge(x+1)) applied. Analyses were performed on the entire data sets and two subsets: (i) vascular plants and (ii) bryophytes plus lichens.

An initial explanatory analysis was carried out to investigate the range of variation within the plant community composition using detrended correspondence analysis (DCA). To aid interpretation, the vegetation targets and sites were fitted onto the DCA ordination plot using the vegan envfit function and 1000 permutations (Oksanen et al. 2011). Standard deviational ellipses of each site and vegetation targets hulls were then used to illustrate their position on the biplots (Oksanen et al. 2011). Permutational multivariate analysis of variance (PMAV) using Bray and Curtis distance matrices was also used to examine and quantify the differences in floristic composition between vegetation target hulls (Oksanen et al. 2011).

Second, principal response curves (PRCs, van den Brink & ter Braak 1999), a special case of redundancy analysis (RDA) for multivariate responses in a design with repeated observation, were used to quantify and visualize the overall effects of Pteridium control treatments on species composition of the two target communities over time. Separate PRC analyses were then carried out at each site and for vascular plants and bryophytes plus lichens to identify whether the responses were similar to overall community responses. PRC plots the temporal changes in species composition for each treatment as deviations from the experimental control represented as a zero line. In the PRC analysis, the transformed species data were standardized, and environmental variables were coded as a partial RDA that allows for time-specific treatment effects (Pteridum control treatments × time interactions) whilst controlling for the overall temporal trend using time as a covariable. The first axis of each PRC was inspected with randomization tests using the reduced model and 9999 permutations stratified to account for split-plots (freely exchangeable of whole plots but no permutation at split-plot level). In addition, vegetation compositional differences between treatments and the experimental control in each sampling year were evaluated in sequential tests for each sampling date, permuting freely the plots within blocks or sites under the reduced model (9999 permutations). For each year, all pairwise comparisons were tested with respect to the experimental control; thereafter, Bonferroni correction was used to assess the significance level of each contrast (Sokal & Rohlf 1995); here, the critical probability level for detecting significance between contrasts was α = 0·01. The species weights (bk) represent the affinity of each species with the treatments analysed, and the sign indicates the direction of the changes in abundance. The species weights were used to compare how these new communities were moving towards the target communities (National Vegetation Classification (NVC) types, H12 for Calluna heathland and U4 or U5 for acid grassland; Rodwell 1991, 1992). For clarity, only most frequent species are shown in the PRC plots.

Third, the PRC results were used to assess whether the invaded communities were resistant or resilient to Pteridium control treatments. An invaded community was resistant when it did not change when a management disturbance was applied, that is, when the PRC of the Pteridium control treatment showed no significant temporal change in species composition in comparison with experimental controls. An invaded community was resilient when the managements produced a significant change relative to the experimental control but then it reverted to its original state, that is, here the PRC returned towards the experimental control line showing no significant differences with it at the end of the experiment. Evidence for an emerging alternative stable state (ASS) was based on two criteria proposed by Schröder, Persson & De Roos (2005): (i) random divergence, here the new state to which community composition has been moved, represented by PRCs, was significantly different from the experimental control and (ii) non-recovery, here there was no movement back to the starting state after ceasing the management. Resilience and ASS could only be considered for ‘one-off’ treatments. As ‘repeated’ treatments manipulated the community up to 1 year before the final monitoring, here it was only possible to describe it as an alternative state (AS), as it was potentially held there by continued treatment.

Fourth, LMM were used to determine the effect of Pteridum control treatments and time on diversity variables (total species richness = number of species per 1 m2, richness of vascular plants and bryophytes–lichens and evenness). In these analyses, Pteridum control treatments and sites were treated as categorical fixed factors and time was treated as a continuous fixed factor. Random effects were defined to account: (i) spatial correlation, using an hierarchical structure of plot, the different subtreatments and sub-subtreatments, block and site and (ii) temporal correlation, including time nested within the spatial variables, and modelling the autocorrelation structure using a first-order autoregressive structure [AR(1)] (Pinheiro & Bates 2000). Model simplification guidelines followed Pinheiro & Bates (2000) using the Akaike Information Criterion (AIC). Models were fitted using the lme function and the REML method, and species richness data were loge-transformed before analysis. All values are reported as the mean ± standard error of the fixed factors.

Results

136 vascular plant species, 55 bryophytes and 13 lichens were recorded during the 10 years. Sites with acid grasslands as restoration target were more species rich than sites with a Calluna heathland target (111 vs. 68 vascular plant species and 55 vs. 40 bryophytes plus lichens, respectively). The initial plant species richness of Pteridium-invaded stands targeted for restoration to acid grasslands was 14 vs. 5 species in stands targeted for restoration to Calluna heathland.

Exploratory multivariate analyses

Preliminary ordination of the vegetation using DCA produced eigenvalues (λ) of 0·53, 0·21, 0·11 and 0·10 and gradient lengths (GL) of 4·26, 2·33, 2·13 and 2·38 for the first four axes (Fig. 1). The passive fit of the sites and vegetation types over species and sites biplots was significant (< 0·001; Fig. 1a,b). The two vegetation types (grassland vs. Calluna targets) occupied different regions of the ordination plot with little overlap, indicating significant compositional differences between them, accounting for 36% of the variance in the species data (Fig. 1a; PMAV with 9999 permutations; < 0·01). Calluna heathlands appeared at the positive side of axis 1 (Fig. 1a), associated with heathland species such as Calluna vulgaris, Campylopus introflexus, Deschampsia flexuosa and Vaccinium myrtillus and showing some degree of variation along axis 1 in species composition between different Calluna target sites (Peak, Cannock 1, 2 and 3; Fig. 1b). In contrast, acid grassland target sites appeared at the negative end of axis 1 and associated with Agrostis capillaris, Anthoxanthum odoratum, Pleurozium schreberi and Poa trivialis. Grasslands showed some degree of variation along the axis 2 between the three sites (Sourhope 1, 2 and Carneddau; Fig. 1b). Thus, axes 1 and 2 represented a variation gradient across Calluna heathland and acid grassland vegetation types.

Figure 1.

DCA ordination for the first two axes of floristic composition data from seven Pteridium control experiments in Great Britain. (a) Ordination biplot showing the two different vegetation targets encircled (Calluna heathland, acid grassland); points represented Pteridium control treatment plots for each block, site and year. (b) Species biplot with standard deviational ellipses of each experimental site (blue ellipses represents acid grasslands and red ellipses Calluna heathlands), only the most frequent species are shown. Species codes are as follows: Ac, Agrostis capillaris; Aca, Agrostis castellana; Ap, Aira praecox; Ao, Anthoxanthum odoratum; Bcr, Brachythecium rutabulum; Cca, Carex caryophyllea; Cpar, Cirsium palustre; Cpi, Campylopus introflexus; Cpil, Carex pilulifera; Cpy, Campylopus pyriformis; Cv, Calluna vulgaris; Df, Deschampsia flexuosa; Ds, Dicranum scoparius; Ea, Epilobium angustifolium; Ep, Eurhynchium praelongum; Fo, Festuca Ovina; Gs, Galium saxatile; Hj, Hypnum jutlandicum; Lb, Lophocolea bidentata; Lca, Luzula campestris; Lm, Lathyrus linifolius; Psh, Pleurozium schreberi; Pan, Poa annua; Pt, Poa tivialis; Pe, Potentilla erecta; Paq, Pteridium aquilinum; Rs, Rhytidiadelphus squarrosus; Vc, Veronica chamaedrys; Vm, Vaccinium myrtillus; Vv, Vaccinium vitis-idaea.

The high proportion of the variation explained by axis 1 (56%) and the significant differentiation of vegetation types along axis 1 suggested that the split of the data set for subsequent analysis, that is, Calluna heathland and acid grassland target sites was justified.

Effects of Pteridium control treatments on species compositional response

The Pteridium control treatments induced significant change in species composition compared to the experimental controls in both vegetation target types (Fig. 2a,b); indeed, the overall results for the two vegetation types were broadly comparable in terms of (i) the variance accounted (grassland model 29% and PRC1 9%, heathland model 30% and PRC1 10%), (ii) statistical significance of both models (FGrassland = 15·03 and FHeathland = 33·02, < 0·01) and (iii) significance of the first PRC axes (9999 permutations, < 0·01). However, the compositional changes produced by the ‘one-off’ and ‘repeated’ treatments differed between the vegetation types, being clearer in experimental sites where Calluna heathland was the restoration target.

Figure 2.

Principal response curves analysis of the changes in community composition during 10 years of Pteridium control treatments relative to experimental control plots. (a) Calluna heathland target community; (b) Acid grassland target community. Solid lines represent ‘repeated’ treatments and broken lines ‘one-off’ treatments. Species codes are given in Fig. 1.

PRC axis 1 for Calluna heathland target sites (Fig. 2a) showed that all treatments moved away from experimental control until 1998. After 1998, two responses were observed: (i) the ‘repeated’ treatments (Cut1pa, Cut2pa) continued increasing through time away from the experimental control reaching values of nearly 2·0 in 2003 when they were significantly different from the experimental control (< 0·001) and (ii) the ‘one-off’ treatments (CutSpray, Spray, SprayCut) remained at 1998 values for a further year or two, but after 2000 moved back towards the experimental control. In 2003, none of these ‘one-off’ treatments were significantly different from the experimental control (> 0·01).

The patterns derived from the overall PRC analysis of Calluna heathland were similar when each site was analysed independently (Peak, Cannock 1, 2 and 3; Table 2, Fig. S1a, Supporting information) and for both the vascular plant species and bryophytes plus lichens subset (Fig. 3a, Table 2), indicating that the compositional response patterns to ‘one-off’ and ‘repeated’ treatments were consistent between the different sites and species groups.

Table 2. Percentage of total variance in the PRC analysis explained by Pteridium control treatments (including treatment × time interaction) and time for each vegetation type and site. The variance accounted by the first axis is also shown
SiteTreatmentTimeAxis 1 F-ratio P (9999)
Calluna heathlands 17131014·340·005
Bryophytes & Lichens221217
Vascular plants1486
Cannock 143138
Cannock 24215·59
Cannock 3263515
Peak31128
Acid grasslands 151398·000·005
Bryophytes & Lichens1085
Vascular plants14117
Sourhope 138227
Sourhope 241178
Cardennau28125
Figure 3.

Principal response curves analysis of the changes in vascular plants and bryophytes plus lichens community composition during 10 years of Pteridium control treatments on sites identified as Calluna heathland and acid grassland target communities. (a) Calluna heathland target (bryophytes plus lichen, vascular plants) and (b) acid grassland target communities (bryophytes plus lichen, vascular plants). Solid lines represent ‘repeated’ treatments and broken lines ‘one-off’ treatments. Species codes are given in Fig. 1.

The Calluna heathland species that increased in abundance after Pteridium control treatments (positive weights) can be grouped as follows (Fig. 2a; 3a): (i) large effect (>0·5 bk) – A. capillaris, C. introflexus, D. flexuosa and Galium saxatile and (ii) moderate effect (0·2–0·5 bk) – C. vulgaris, Campylopus pyriformis, Dicranum scoparius, Festuca ovina, F. rubra, Hypnum jutlandicum, Potentilla erecta, Rhytidiadelphus squarrosus and Vaccinum vitis-idaea, indicating that the managed community is moving towards the target community. Species responding negatively to Pteridium control treatments (negative weights) included Brachythecium rutabulum, Eurhynchium praelongum, Lophocolea bidentata and P. aquilinum.

In the acid grassland targets (Fig. 2b), the ‘one-off’ treatments (Spray, SprayCut and CutSpray) produced an immediate significant strong compositional response during the first 3 years (1994–1996; < 0·01). Thereafter, the compositional differences of these treatments were maintained stable at values of 1·5–2 until 2002 when they started to decline towards the experimental control line. However, at the end of the experiment, these treatments were still significantly different from the experimental control (2003; < 0·01). In contrast, the ‘repeated’ treatments (Cut1pa, Cut2pa) were not significantly different from the experimental control in 1994 (> 0·01), but through time the compositional differences increased, reaching values of ca. 1·75 in 2003. By the end of the experiment, the ‘repeated’ treatments had produced greater deviations from the experimental controls than the ‘one-off’ treatments (< 0·001).

In contrast to the stable response patterns for Calluna heathland, the overall pattern of the PRC analysis for acid grasslands was only reproduced for vascular plants (Fig. 3b, Table 2). The compositional response of bryophytes plus lichens to Pteridium control treatments was different to the overall pattern, showing an immediate significant difference in 1994 (< 0·001), which disappeared in 1995 and for the rest of the sequence, except for occasional effects, for example, the effect of SprayCut treatment in 2001 (Fig. 3b). Moreover, the responses of each site analysed independently (Carneddau, Sourhope 1 and 2) showed that, although the effect of ‘repeated’ treatments (Cut1pa, Cut2pa) was similar to overall patterns, the success of ‘one-off’ treatments was site related (Fig. S1b, Supporting information). SprayCut produced the greatest compositional change from experimental controls on Carneddau, whereas in Sourhope 1 and 2 the greatest changes were produced by Spray and CutSpray.

The acid grassland species that increased in abundance after Pteridium control treatments (positive weights, Figs 2b and 3b) only showed a moderate effect (0·2 to 0·5 bk), these were as follows: Aira praecox, Carex caryophyllea, F. ovina, F. rubra, Luzula campestris, G. saxatile, P. erecta and Vaccinium myrtillus, indicating that managed community was moving towards the target community. Species that showed a reduced abundance with Pteridium control treatments (negative weights) were Cirsium palustre, Lathyrus linifolius, P. trivialis, P. aquilinum and Veronica chamaedrys.

Effects of Pteridium control treatments on plant community diversity (richness and evenness)

The LMM showed that species richness response to Pteridium control treatments was different for each vegetation type. In the Calluna heathland target vegetation, the treatment × time interaction was significant (F5,1107 = 18·0, < 0·001; Fig. 4a). In the acid grassland targets, it was not significant (F5,1057 = 1·85, = 0·102; Fig. 4b) and the richness decreased in a loglinear manner over time with a common negative slope for all treatments including experimental control and sites (time = −0·03 ± 0·006, F1,1057 = 101·27, < 0·001). However, there was a treatment effect (F5,1057 = 10·14, < 0·01), showing CutSpray a greater intercept than the experimental control (different intercepts; 17 vs. 13 respectively; t-value = 6·41, < 0·001).

Figure 4.

Linear mixed-effects models assessing the effects of Pteridium control treatments on total richness and evenness over a 10-year period on (a) Calluna heathland target sites and (b) Acid grassland target sites. Different lines represented the minimal adequate models for each site.

At the sites with a heathland target, species richness showed a log-linear decrease over time (negative slope, time = −0·03 ± 0·007, t-value = −3·80, < 0·001) in the experimental control and SprayCut treatments, reaching as low as 4 and 5 species, respectively, at 2003 (Fig. 4a). In contrast, the rest of the treatments showed a loglinear increase (positive slope), but with different slopes. Cutpa1 and Cutpa2 showed the steeper slopes reaching 9 and 8 species in 2003, respectively (time = 0·07 ± 0·01, < 0·001), compared to CutSpray and Spray that only reached seven species by 2003 (time = 0·055 ± 0·01, t-value = 5·28, < 0·001). There was no site × treatments interaction (F10,135 = 0·80, = 0·625), but there was a site effect (F3,24 = 37·0, < 0·001), Peak being richer than Cannock sites (Fig. 4a). Similar results were found for the species richness of vascular plants and bryophytes plus lichens.

In the acid grassland targets, species richness for vascular plants and bryophytes plus lichens showed the same patterns as total richness, reducing through time with a common negative slope for all the treatments (vascular plants: time = −0·02 ± 0·002, F1,1057 = 46·42, < 0·001; bryophytes–lichens: time = −0·04 ± 0·004, F1,1057 = 88·39, < 0·001). Also, there was a treatment effect on richness of vascular plants (F5,1057 = 16·21, < 0·001) and bryophytes plus lichens (F5,1057 = 4·02, P < 0·01), where spray treatments had greater intercepts than experimental control.

Evenness was only affected by Pteridium control treatments in Calluna heathland target sites, producing significant increases in all treatments compared to experimental control (treatment × time interaction, F5,1107 = 12·50, < 0·001; Fig. 4a,b). The greatest increase in evenness was produced in CutSpray, Cutpa1 and Cutpa2 (steeper slope, time = 0·03 ± 0·004, < 0·001), reaching values near 0·85 at 2003, whereas Spray and SprayCut reached 0·82 and 0·73 at 2003, respectively (time = 0·02 ± 0·004, < 0·001). In contrast, evenness on acid grassland target sites increased significantly with time during the experiment with a common positive slope for all treatments including experimental control and sites (time = 0·003 ± 0·001, F1,1057 = 51·15, < 0·001), reaching 0·86–0·88 at 2003 (Fig. 4a,b). This increase in evenness in both vegetation types indicates a reduction in dominant species cover and an increase in nondominants.

Discussion

These long-term experiments revealed that the compositional responses to Pteridium control treatments differed considerably between the two target communities (Calluna heathland and acid grasslands). More specifically, only ‘repeated’ treatments produced compositional changes in Calluna heathland targets, whereas both ‘one-off’ and ‘repeated’ treatments produced strong compositional responses in the same direction on acid grassland targets. These differences in compositional response potentially indicate different resistance and resilience of the invaded communities that could be caused by initial differences in species composition and richness between invaded communities before management; it is well known that species availability and species performance are among the general causes of differences in vegetation dynamics (Pickett, Cadenasso & Meiners 2009).

Calluna heathland target sites

In the Calluna heathland target sites, which had the lowest species richness, there were strong differences between the effect of ‘one-off’ and ‘repeated’ Pteridium control treatments on species composition. The ‘repeated’ treatments (Cut1pa, Cut2pa) overcame the resistance of the invaded community and produced a gradual increase in compositional differences compared to the experimental control over the 10 years, which paralleled a significant increase in richness and evenness. These treatments were the most effective means of inducing a shift of species composition towards a Calluna heathland target, confirming results found elsewhere (Marrs, Johnson & Le Duc 1998; Måren, Vandvik & Ekelund 2008). The higher effectiveness of ‘repeated’ treatments in driving compositional shifts may be due to the more intense disturbance produced by annual cutting, which suppresses the dominant P. aquilinum (Stewart et al. 2008) and creates regeneration niches after disturbance (Grubb 1977). Our data indicate that ‘repeated’ treatments opened ‘windows of opportunity’ every year that allows resident and new colonist species to colonize and develop, facilitating a yearly change in species composition, richness and evenness in comparison with experimental controls (Figs 2 and 3).

In contrast, the ‘one-off’ treatments (CutSpray, Spray, SprayCut) were less effective than ‘repeated’ ones inducing changes in species composition, diversity and evenness. The compositional differences produced by ‘one-off’ treatments reached a maximum after 5–7 years (1998–2000), with slight increases in important species abundances such as C. vulgaris and D. flexuosa, but these reduced through time and by the last sampling year (2003) were not significantly different in species composition to the experimental controls. This implies that, although ‘one-off’ treatments displaced the system (resistance), they were not strong enough to overcome the resilience of the invaded community, and the system returned to its previous state (Schröder, Persson & De Roos 2005). Pickett, Cadenasso & Meiners (2009) argued that whilst ‘regeneration niche’ provision was a major factor in influencing plant community change, the differential availability and performance of individual species are also important. Here, it seems that, although the gaps left by reducing P. aquilinum were colonized by resident species, there was little subsequent increase in abundance and few new colonist species persisted (low species richness increases); as a consequence, the ability to resist against the re-invasion by the invader is reduced (Schröder, Persson & De Roos 2005).

Previous research on Pteridium control strategies has documented different levels of success in reducing P. aquilinum cover and biomass at different sites, possibly caused by variations in climatic regime, substrate and past or present management (Marrs & Watt 2006; Stewart et al. 2008). However, the results here showed that the general patterns produced by ‘one-off’ and ‘repeated’ Pteridum control treatments on species composition of Calluna target sites were similar for all sites and plant species groups considered. This may be because the sites have a similar vegetation composition and the ability of vegetation to respond to experimental disturbances is related to functional traits of the dominant species (Bernhardt-Römermann et al. 2011), here D. flexuosa, G. saxatile, V. mirtyllus, C. vulgaris, D. scoparius, A. capillaris and H. jutlandicum. It appears that resistance and resilience are related to plant functional traits allowing similar processes of vegetation development following disturbance events (Bernhardt-Römermann et al. 2011). Plant functional traits are therefore crucially important in predicting responses to disturbances.

Acid grassland target sites

In the acid grassland target sites, which had greater species richness than the Calluna targets, all Pteridium control treatments produced a significant change in species community composition in the same direction compared to experimental controls. Results suggest that an ASS has been produced with a resilience lasting longer than 10 years for ‘one-off’ treatments (Folke et al. 2004). This implies that the resistance or resilience of the initial Pteridium-dominated community has been overcome by ‘one-off’ treatments for at least this period.

Although both ‘one-off’ and ‘repeated’ treatments produced significant directional changes in species composition, the ‘one-off’ treatments were faster (1 year) in achieving this change, whereas the ‘repeated’ treatments took 6 years to produce similar effect, corroborating results found in the Calluna heathlands. These differences are related to the type of treatments (e.g. spray vs. cutting) rather than whether treatments are repeated or not. The faster compositional response of spray treatments in comparison with cutting treatments has been previously appointed by Måren, Vandvik & Ekelund (2008). The greater species diversity in the acid grassland target sites may be one reason why all treatments shifted the species composition towards a stable ecosystem of enhanced conservation value. It has been shown that diverse communities can often be resilient to disturbance and invasion (Elmqvist et al. 2003), facilitating their recovery after invasive species control (Fargione & Tilman 2005). Thus, natural colonization processes will be enough to re-establish acid grasslands. These results are in agreement with Pakeman, Le Duc & Marrs (1997) who found that Pteridium control frequently resulted in grass-dominated communities in the UK.

However, in contrast to the Calluna targets, these generalized compositional response patterns were only reproduced by the ‘repeated’ treatments on all sites, whereas compositional effects of ‘one-off’ treatments were site dependent. This may be because there were between-site variations in Pteridium control success for ‘one-off’ treatments (internal site factors; Stewart et al. 2008). As a consequence, the open niches for re-colonization by native species were reduced (Kettenring & Adams 2011) and hence preventing native species recovery. These results support the theory that the community's ability to compete against the invasive species and, therefore, stop the system from returning to its original state is dependent on the level to which the invasive species is reduced (Schröder, Persson & De Roos 2005).

At the same time, the compositional response of bryophytes plus lichens to Pteridium control treatments was almost non-existent at the acid grassland target sites (high resistance). This may be because bryophytes and lichens were not able to compete with vascular plants for light and space on the newly created gaps (Hájková, Hájek & Kintrová 2009). Thus, Pteridium control treatments on their own had relative little influence on the development of a bryophyte and lichen flora in acid grassland.

The strong compositional responses in acid grassland target were paralleled by changes in species richness and evenness. Species richness reduced slightly through time for all the treatments and sites, and this conflicts to some extent with the PRC results. However, because the decline is produced simultaneously and in the same way in all the treatments including experimental controls, it seems that there are unknown processes that are reducing species richness in these communities. Nevertheless, Pteridium control treatment always maintained a greater richness relative to untreated controls, and this helped increase acid grassland recovery, firstly by allowing new assemblages of species to colonize gaps left by P. aquilinum (Pickett, Cadenasso & Meiners 2009) and secondly by increasing the species density and producing compositional differences relative to experimental controls. Evenness also increased in all the treatments through time, indicating a reduction in dominant species cover and an increase in nondominants. This is an important result because one of the major outcomes of Pteridium control strategies should be to develop a community with greater species diversity and evenness (Marrs, Johnson & Le Duc 1998). This was achieved in the acid grassland targets with an ASS by ‘one-off’ treatments persisting for at least 10 years.

Management implications

The present work demonstrates that the plant communities invaded by P. aquilinum with two different restoration targets (Calluna heathland and acid grasslands) have different diversity and compositional response pathways to Pteridium control treatment, suggesting that different management strategies are needed for each community. In the less diverse Calluna heathland target sites, ‘repeated’ treatments were most effective in shifting community composition towards the target community and inducing an AS. Here, because the ‘repeated’ treatments are applied annually, there is no evidence that the state will not revert to the Pteridium invaded when these treatments are removed, that is, the AS may well be treatment dependent and could be transient if the intervention ceased (Schröder, Persson & De Roos 2005). Although the ‘one-off’ treatments produced a community displacement overcoming the resistance of the invaded community, its high resilience induced a convergence with the experimental controls within 10 years; hence, they are not appropriate for restoring the target vegetation. Based on these results, the best approach for restoring Calluna target vegetation on sites with low species richness may be regular disturbance in time, that is, creating available sites every year that facilitated colonization of resident and new native species that can displace the invasive species.

On acid grassland target sites, with greater species diversity, both ‘one-off’ and ‘repeated’ treatments were effective in overcoming resistance, and ‘one-off’ treatments also overcame the resilience of the invaded community. However, although ‘one-off’ treatments produced an ASS sensu Schröder, Persson & De Roos (2005), their effectiveness was site dependent. In contrast, the ‘repeated’ treatments produced an AS on all sites, suggesting that cutting treatments are an efficient option for restoring native species composition, even though they took longer to take effect and were more expensive than ‘one-off’ treatments. Moreover, the AS may be transient if the intervention ceased. Finally, our results also showed that, on grasslands, the introduction of bryophytes and lichens may be required as they were not able to develop after Pteridium control treatments' application.

Acknowledgements

We thank the DEFRA (Project BD1226) and the Basque-Country Government (J.G.A. BFI-2010-245) for financial support. Dr. Hugh McAllister for assistance with species identification, Dr Ruth Mitchell for perceptive comments on an early draft and land owners/managers for site access (Sourhope, James Hutton Institute; Peak, Jeremy Archdale and Neil Taylor; Carneddau, Countryside Council for Wales; Cannock, Staffordshire County Council/Natural England).

Ancillary