Fast-growing pioneer tree stands as a rapid and effective strategy for bracken elimination in the Neotropics

Authors

  • David Douterlungne,

    Corresponding author
    • El Colegio de la Frontera Sur, San Cristóbal de Las Casas, Chiapas, Mexico
    Search for more papers by this author
  • Evert Thomas,

    1. Bioversity International, Regional Office for the Americas, Cali, Colombia
    Search for more papers by this author
  • Samuel I. Levy-Tacher

    1. El Colegio de la Frontera Sur, ECOSUR, Carr. Panamericana y Periférico Sur s/n Barrio Ma. Auxiliadora, San Cristóbal de Las Casas, Chiapas, México
    Search for more papers by this author
    • Correction added after online publication 15 July 2013; S. I. Levy-Tacher added as co-author.

Correspondence author. E-mail: daviddouter@hotmail.com

Summary

  1. Large areas of agricultural land around the world are degraded as a consequence of dominance by bracken fern (Pteridium spp.). Tropical production systems based on shifting cultivation and cattle breeding are particularly vulnerable to invasion of this species. In spite of this, effective methods for tropical bracken control are limited.
  2. Fast-growing tree species have been used successfully to out-compete aggressively colonizing heliophytes and trigger natural succession. Drawing on a traditional Mayan management technique, we evaluate the potential of the pioneer tree balsa (Ochroma pyramidale) to control Pteridium caudatum in Chiapas, Mexico. We tested different bracken cutting frequencies and balsa propagation methods in a factorial randomized block experiment. Eighteen months later, we quantified bracken biomass under the young balsa canopy.
  3. Living bracken rhizome biomass correlated significantly with balsa basal area, leaf litter biomass and understorey light intensity. While bracken rhizomes persisted in control plots, it was completely eradicated in plots with a minimum balsa basal area of 11 m2 ha−1. This threshold value was reached in less than 18 months with any of the tested propagation methods (seed broadcasting, direct sowing or nursery seedlings), on the condition of at least monthly bracken cutting during the first six months.
  4. The ability of fast-growing broad-leaved pioneer trees like balsa to quickly out-compete bracken fern offers opportunities for large-scale application in tropical rural areas where economic and technical resources are scarce.
  5. Synthesis and applications. Mayan subsistence farmers traditionally use balsa to out-compete invasive weeds, including bracken fern. Here, we highlight the usefulness of this method for quick and effective bracken control in southern Mexico. This approach, in combination with balsa's short rotation cycle, creates opportunities to rapidly convert bracken land into forest stands with commercial potential, thus providing local income and increasing the likelihood of adoption by rural people. We encourage further research to test the potential of balsa and other fast-growing pioneer trees species for controlling bracken and similar weeds.

Introduction

Bracken fern (Pteridium spp., Dennstaediaceae) is one of the most persistent weeds world-wide, due to a combination of an ancestral global distribution dating back to the Oligocene (~23·8 million years ago), and an effective dispersal and colonization capability (Der et al. 2009; Roos, Rödel & Beck 2011). Bracken's competitive advantage relates to aggressive rhizomatous and aerial growth (Marrs, Johnson & Duc 1998a), allelopathic and toxic effects on competing species and grazing animals, respectively (Marrs et al. 2000; Alonso-Amelot & Avendano 2002; Calderón Tobar et al. 2011), tolerance to fire (Gliessman 1978; Da Silva & Silva Matos 2006), extreme weather events (Roos et al. 2010) and the cutting of its fronds (Pakeman et al. 2002; Roos, Rödel & Beck 2011). Furthermore, the accumulation of slowly decomposing bracken fronds (Frankland 1976) depletes seed and seedling banks, constrains natural recruitment (Da Silva & Silva Matos 2006; Ghorbani et al. 2006) and can delay natural succession for decades to centuries (Marrs et al. 2000; Roos, Rödel & Beck 2011). Bracken typically colonizes open fields after disturbance events, resulting in loss of land for crop or animal husbandry (Schneider & Goeghegan 2006) and hence affecting people's economies and well-being. In Great Britain, bracken was estimated to cover 7·3% (c. 17 100 km2) of the land surface (Pakeman et al. 1996), representing an estimated reduction in the land's capital value of 5·5 billion USD (Robinson 2008). Prior to the introduction of control measures on the island of São Miguel (Azores), bracken-induced bladder cancer in one cattle breed was estimated to cause an annual economic loss of 16·2 million USD, as 18% of the slaughtered animals and 13·5 million litres of milk did not pass sanitary control (Pinto et al. 2001). In Chiapas, Mexico, where this study took place, at least 360 km² (about 0·5% of the state territory) is dominated by bracken (Suazo 1998).

Control of bracken is extremely difficult, and many strategies have been tested without achieving complete bracken elimination (Marrs, Johnson & Duc 1998b; Pakeman et al. 2002; Le Duc, Pakeman & Marrs 2003; Stewart et al. 2008; Roos, Rödel & Beck 2011). Most literature on restoration of bracken land concerns strategies targeted to temperate climates (but see Roos, Rödel & Beck 2011) and can be classified in three main groups. The first type is based on physically damaging the fronds or rhizomes, often through the use of heavy machinery, such as tractors and bulldozers (e.g. Le Duc, Pakeman & Marrs 2003; Ghorbani et al. 2006). A second group of treatments relies on chemical substances, such as ASULAM (e.g. Pakeman et al. 2002; Stewart et al. 2008). Both strategies are unlikely to be applied in poor rural contexts as they require substantial financial and/or labour investments during prolonged periods of time (often exceeding 10 years; Marrs, Johnson & Duc 1998b; Ghorbani et al. 2006). The third type of control measures takes advantage of ecological processes such as competition and natural succession (e.g. Pakeman et al. 2002; Slocum et al. 2006), and may be better tailored to the tropical rural context as they are less resource-intensive.

In tropical ecosystems, pioneer tree species are of particular interest for biological bracken control of bracken, owing to their ability to out-compete aggressively colonizing light-demanding weeds, and trigger forest succession (Lamb, Erskine & Parrotta 2005). The Lacandon and Ch'ol Mayan ethnic groups from the southern Mexican rainforest have longstanding experience with the use of key tree species to assist natural regeneration of tropical evergreen forest on abandoned slash and burn plots. They sow or tolerate balsa or Ochroma pyramidale (Cav. ex Lam.) Urb., Malvaceae, a neotropical broad-leaved pioneer tree species, on their fallows to accelerate soil recovery (Diemont & Martin 2009). Balsa thrives on abandoned agricultural soils with minimal maintenance labour. Survival rates of 90% or more and mean heights of 6–7 m have been registered after the first growth year (Douterlungne 2005). Furthermore, from the second growth year onwards, abundant balsa litter increases soil organic matter (Diemont et al. 2006), while understorey shade constrains the establishment and growth of light-demanding weeds (D. Douterlungne, unpublished data). When located in the vicinity of old-growth forest, balsa's big white flowers attract seed-dispersing fauna, especially bats (Tschapka 2005) and insects, which in turn attract small tree-climbing mammals (D. Douterlungne, personal observation). Both balsa and bracken are amply distributed across the Neotropics (Lamprecht 1989; Der et al. 2009) and share similar habitat preferences for recently burnt (Vazquez-Yañes 1974; Gliessman 1978) or disturbed vegetation on well-drained soils (Kammesheidt 2000).

Establishment success and growth performance of balsa in bracken land have been the topic of a previous publication (Douterlungne et al. 2010). Here, we focus on the potential of balsa plantations to effectively eliminate Pteridium caudatum (L.) Maxon (previously Pteridium aquilinum var. caudatum). We scrutinize the relationship between living bracken biomass and three different balsa attributes: basal area, leaf litter production and understorey light intensity.

Materials and methods

Site description

Our experiment was carried out in Lacanjá Chansayab, Chiapas, Mexico (16˚46′08″N, 91˚08′12″W). The study area borders the largest undisturbed tropical rainforest area in Mesoamerica and is located at 350 m above sea level, with an annual precipitation between 2300 and 2500 mm, and a mean annual temperature of 25 °C (INEGI 1988). Predominant soils are humic acrisols (Muench 1978). The landscape is characterized by a mosaic of tropical evergreen forest –the local climax vegetation (Miranda 1998)– interspersed with secondary forest, crop fields and fallows nearby settlements. The experimental plots were laid out within a bracken infested area of c. 0·85 ha. Plots were impenetrable without the use of a machete, with dead bracken fronds forming a 0·5–1 m thick layer and living fronds reaching up to more than two metres high. According to local Lacandon farmers, on site bracken domination dated back more than thirty years.

Experimental design

In October 2005, we set up a fully factorial randomized block experiment of 2304 m², divided in four equal blocks of 576 m². Each block was separated by a two metre access trail and included nine adjacent plots of 8 × 8 m², where three balsa establishment methods were crossed with three bracken cutting frequencies. Control plots (without balsa trees) for each of the three cutting frequencies were not included in the original design, but located in homogenous bracken land in the direct vicinity (<10 m) of the four blocks. For statistical analyses, we included each of the respective control plots in the nearest block.

Local Lacandon participants manually cleared (machete) and burned the experimental area prior to establishing balsa trees. As a first planting method, we mimicked the traditional Lacandon method of randomly broadcasting c. 5000 balsa seeds per plot, which resulted in patchy distributions of emerging seedlings. The second and third method consisted of directly sowing seeds and transplanting two-month-old nursery seedlings at 2 × 2 m spacing. No thinning was carried out during the experiment. The minimum cutting frequency entailed only initial clearing of the site prior to planting or sowing, without any posterior cutting. The intermediate- and high-intensity treatments consisted of cutting all bracken fronds every four and two weeks, respectively, during the first six months of the experiment. Survival and growth rates are reported in Douterlungne et al. (2010).

Eighteen months after experiment initiation, balsa dominated all plots where bracken was cut. We quantified remaining bracken as: (i) visually estimated cover; (ii) weight of air-dried fronds; and (iii) weight of air-dried rhizomes, considering living and dead rhizomes separately. We quantified bracken rhizome and frond biomass in three randomly assigned 0·25-m2 subplots in the central 36 m2 of each plot, excavating holes as deep as the deepest rhizome (>0·5 m deep). However, given the similar statistical patterns observed for all bracken variables, we focus on living rhizome biomass. The carbohydrate reserves in bracken rhizomes are the fern's primary means for regrowth (Watt 1940), even after elimination of all of its fronds (Walker & Boneta 1995). Therefore, living rhizome biomass is considered the most adequate predictor for assessing success of long-term bracken control (Marrs, Johnson & Duc 1998a). We measured light intensity with a linear photosynthetically active radiation (PAR) sensor or ceptometer (Decagon, Pullman Washington, US) at four random locations in the same central plot area, at noon of cloudless days, above the bracken vegetation and under the balsa canopy, where relevant. To quantify balsa performance, we weighed all air-dried balsa leaf litter available in the bracken rhizome subplots and calculated basal area based on the nine trees from the central 36 m2 of each plot according to the following formula:math formula. We measured tree diameter 10 cm above-ground level, as several trees were smaller than 130 cm and would be excluded in breast height-based basal area calculations.

Statistical analysis

We used an iterative modelling process (following Zuur et al. 2009) to identify variables that significantly contributed to explaining plot-averaged bracken rhizome biomass. We started with linear mixed effects (LME) models of different complexities to address the spatial autocorrelation in our dataset and compared these with a generalized least squares (GLS) model (fixed component containing up to second-order interaction terms). Aikake's information criteria (AIC) and likelihood ratio tests (Zuur et al. 2009) indicated no significant random (block or plot) effect, and therefore, we retained the more parsimonious GLS model. To resolve heterogeneity in the normalized residuals, we applied exponential variance structures for basal area and leaf litter, while an additional square root transformation of the response variable ensured normality of residuals (verified with Q-Q plot and Shapiro–Wilkinson normality test: W = 0·96, = 0·16). After model simplification and removal of one influential data point, we retained the minimum adequate model. We calculated an approximate r2-value by comparing the variation in data explained by the GLS model with that of a null model (Crawley 2007).

To compare uncorrelated effect sizes of the individual contributions of correlated balsa attributes to bracken reduction, we calculated a partial correlation matrix of the standardized covariance matrix. We tested multivariate normality with a Shapiro–Wilkinson test for multivariate samples (Looney 1995) and evaluated conditional independence using t-test statistics (Marchetti, Drton & Sadeghi 2012).

Next, we examined individual relations between balsa performance variables and living rhizomes biomass, applying a similar iterative approach as above and maximizing the restricted log-likelihood. Additionally, we interpreted the biological meaning of the curves of predicted model values before deciding whether and which transformation to apply. Reported model equations are based on standardized data to help biological interpretation (Grace & Bollen 2005). We included control plots to evaluate the contribution of different bracken cutting intensities to bracken control in plots with no balsa trees.

We used anova to evaluate the impact of cutting frequencies and balsa establishment methods on bracken rhizome biomass. Mixed models and ancova were discarded using the same iterative decision-making process described previously. Factor levels without in-between significant differences were merged to improve model fit and coefficient estimation. We ordered the different management treatments based upon Tukey honest significant differences with 95% confidence intervals. All analyses were performed in R, version 2.14.1 (R Development Core Team 2011) with associated packages ggm (Marchetti, Drton & Sadeghi 2012), ggplot2 (Wickham 2009a), gridExtra (Auguie 2012), nlme (Pinheiro et al. 2011), mvnromtest (Jarek 2012), plyr (Wickham 2009b) and reshape (Wickham 2007).

Results

Balsa performance and bracken control

All balsa attributes correlated significantly with biomass of living bracken rhizomes (Fig. 1), and our overall model explains 87% of the observed variation (see Table S1, Supporting Information). Also bracken frond biomass (r2 = −0·70; < 0·001; F1,46 = 106·6) and frond cover (r2 = −0·81; < 0·001, F1,46 = 201·6) correlated negatively with balsa basal area. We observed only dead rhizomes and fronds in plots with balsa basal areas above 11 m2 ha−1, with the exception of two plots that had less than 80 g m−2 of living, but clearly exhausted rhizomes (Fig. 1a). In the same plots, mean above-ground bracken cover dropped in 18 months from 100 ± 0% to 17 ± 15%, corresponding to 256 ± 292 g m−2 of mainly dehydrated and dead fronds. By contrast, in plots with lower balsa performance (basal area <10 m2 ha−1), living rhizomes remained abundant (2995 ± 1576 g m−2), maintaining 5728 g m−2of living frond biomass on average and covering 91 ± 21% of plot surface areas.

Figure 1.

Univariate regressions of living bracken rhizome biomass vs. (a) balsa basal area (F1,45 = 172·3); (b) understorey light intensity (measured as photosynthetically active radiation) (F1,45 = 78·04); and (c) balsa leaf litter biomass (F1,45 = 175·2). All points correspond to plot-averaged values sampled in 18-month-old balsa stands. The red line and shaded zone represent back-transformed fitted values and 95% confidence intervals, respectively. All models and coefficients are significant at a 5% level. Equations are based on standardized data centred on the mean. Shade projected by overhanging trees of adjacent plots produced the outlier in the extreme left part of (b). The same overhanging trees also deposited litter in plots with low values of balsa basal area, explaining the outliers on the right hand side of (c).

Threshold values for complete bracken elimination were less straightforward for light intensity or balsa leaf litter (Fig. 1b,c). However, living bracken rhizomes tended to completely disappear in intensely shaded plots (<500 μmol m−2 s−1) and plots containing more than 80 g m−2 of balsa leaf litter on average. Not surprisingly, all balsa performance variables were correlated, and plots with increasing balsa basal area had stronger understorey shade and accumulated more leaf litter (Fig. 2). Independent effect size of balsa basal area on living bracken rhizome biomass was twice as high as for balsa leaf litter or understorey light intensity, which yielded comparable effect sizes (Table 1).

Table 1. Partial correlation matrix of living bracken rhizome biomass and balsa attributes. Values express the expected change in a dependent variable associated with one standard deviation change in a given predictor while controlling for the correlated effects of other predictor variables, similar to path- or structuralized equation analysis. All pairwise comparisons fulfil conditional independence at 0·001 confidence levels (t-tests; d.f. = 43)
 Balsa basal areaUnderstorey light intensityBalsa leaf litterBracken rhizomes
Balsa basal area1·000−0·1140·169−0·591
Understorey light intensity 1·000−0·4030·241
Balsa leaf litter  1·000−0·296
Bracken rhizomes   1·000
Figure 2.

Univariate regressions of plot-averaged basal area of 18-month-old balsa stands vs. (a) balsa leaf litter biomass (F1,34 = 43·32); (b) understorey light intensity (photosynthetically active radiation), where higher values represent higher light intensity (F1,33 = 22·01.). The red line and shaded zone represent back-transformed fitted values and 95% confidence interval of the model, respectively. All models and coefficients are significant at a 0·01% level. Equations are based on standardized data centred on the mean.

Management methods

Bracken rhizome biomass varied significantly between different combinations of balsa establishment methods, cutting frequencies and their interactions, whereas the control treatments (without balsa) held significantly higher levels of rhizome biomass under all cutting frequencies (S2). Biweekly and monthly cutting frequencies during the first six months did not result in differences in living bracken rhizome biomasses regardless of the balsa establishment method used (Fig. 3).

Figure 3.

Differences in living bracken rhizomes due to different bracken cutting frequencies and balsa establishment methods (sampled 18 months after experiment set-up). Panels refer to different balsa establishment methods; Control: without balsa trees; Seeds: manually introducing seeds in the ground; Traditional: manually broadcasting large quantities of seeds; and Seedlings: transplantation of nursery seedlings. All plots were cleared from all above-ground vegetation before experimental set-up. Different letters indicate significantly different bracken cutting treatment means within each balsa establishment method (post hoc Tukey test with a 95% confidence interval).

Discussion

Ecological traits underlying bracken control

As opposed to other studies where complete bracken eradication by above-ground treatments was considered unrealistic or even impossible (Pienkowski et al. 1998; Roos, Rödel & Beck 2011), our 18-months-old balsa stands completely out-competed the 30-year-old bracken vegetation in plots with basal areas above 11 m² ha−1. Several ecological traits underlie balsa's potential to weaken bracken's competitive dominance. Of the three measured balsa attributes, basal area presented the largest independent effect on bracken reduction, suggesting a strong competitive effect for available light, water and soil nutrients. Balsa is one of the fastest growing pioneer trees in the lowland Neotropics, and its canopy rapidly overtops aggressively competing heliophytes. Ochroma yielded the highest growth volumes out of more than 70 tested species on degraded lands in Costa Rica (Butterfield 1996). In our bracken land, one-year-old balsa trees reached mean heights of over six metres and closed their canopies (Douterlungne et al. 2010).

In the tropics, reduced light intensity alters bracken's balance between stored rhizome carbohydrates and newly assimilated frond photosynthates (Le Duc, Pakeman & Marrs 2003). Light competition as a bracken control mechanism has been suggested, but can be costly without guaranteeing success (Walker & Boneta 1995; Humphrey & Swaine 1997; Gaudio et al. 2011). Also in our plots, light competition alone was not sufficient for complete bracken control, as demonstrated by partial correlation analysis. Furthermore, bracken persisted in most plots with low balsa basal area, despite their relatively low light intensities compared with control plots. In Ecuador, Roos, Rödel & Beck (2011) were unable to completely eradicate bracken after a combination of cutting and covering vegetation with black plastic sheets during almost two years, exposing it not only to strong shade, but also water and heat stress. Similarly, bracken biomass remained constant after reducing incoming sunlight with an 80% black nylon mesh for one year at a site nearby the current experiment (Peñaolsa-Guerrero 2008).

Balsa leaf litter demonstrated a similar independent effect size as understorey shade. Leaf laminas can be extremely large in young balsa individuals (0·429 ± 0·070 m2, n = 50, Douterlungne 2005), and leaf litter accumulates rapidly and abundantly under young balsa stands (Diemont et al. 2006), blocking incoming light at ground level and possibly obstructing the development of young emerging fronds (Fig. 4). Probably, the combination of an extremely rapid growth and a high production and turnover rate of big leafs contributes to balsa's ability to successfully out-compete bracken.

Figure 4.

Pteridium caudatum control experiment using Ochroma pyramidale stands in southern Mexico. (a) Four-month-old balsa stands (back) and regrowth of bracken in control plots (front). (b) Ten-month-old balsa stands almost completely out-competed bracken, and balsa leaf litter starts to accumulate.

Cutting of fronds clearly has a negative impact on bracken performance. However, cutting alone was insufficient to eliminate bracken in our experiment, as none of the cutting frequencies suppressed bracken in our control plots. In British heathlands, Marrs, Johnson & Duc (1998a) cut fronds of P. aquilinum every six months during 18 consecutive years and suppressed bracken effectively during the first years but failed to further reduce and permanently eradicate the fern. Their results of shorter cutting treatments suggest also bracken regrowth in the long term. Le Duc, Pakeman & Marrs (2003) compared the results of several British bracken control experiments and concluded that cutting bracken twice a year was the most successful treatment. However, they considered it to be ineffective in highly productive bracken lands, such as tropical lowland bracken with an all-year growth season. In Ecuador, the most successful treatment out of series of tested strategies consisted of six consecutive cutting events of P. arachnoideum, reducing its above-ground fresh weight to c. 30% of reference levels (Roos, Rödel & Beck 2011). All mentioned studies failed to control bracken completely but recommend bracken cutting in combination with follow-up treatments, such as the establishment of a closed forest canopy.

Long-term bracken elimination and successional pathways

Conventional invasive species control methods based on damaging plant tissue such as through cutting or burning, or applying herbicides often result in reinvasion of the target species, or colonization by novel invasive plants, jeopardizing return on investment (Kettenring & Adams 2011). In the UK, bracken recolonized sites 10 (Pakeman et al. 2000) to 18 years (Marrs et al. 2000) after control treatments. In contrast, under shaded balsa canopies, recolonization by bracken is very unlikely in the absence of living rhizomes, while leaf litter covering the soil prevents new spores from germinating (Gliessman 1978; Walker & Boneta 1995).

Initiating natural successional processes, such as the ecological interactions under our balsa stands, can be expected to be more effective to achieve long-term bracken control and in general to unblock arrested successional pathways in degraded lands (Marrs et al. 2000; Slocum et al. 2006). The impact of artificial interventions may on the contrary cease or even be reversed once the intervention ends. Although natural succession may not always be a desired outcome after bracken elimination and most conventional control methods often constrain forest succession (Pakeman et al. 2000). Blocking incoming light, applying chemical substances or removing and compacting soil with heavy machinery can severely reduce seed and seedling banks (Ghorbani et al. 2003; Roos, Rödel & Beck 2011) or deteriorate micro-conditions for plant growth (Tirada-Corbalá & Slater 2010). Instead, fast-growing pioneer trees like balsa typically jumpstart natural succession in clearings (Kammesheidt 2000; Van Breugel, Bongers & Martínez-Ramos 2007), reducing even more the physical space, nutrients and light available for bracken. Three years after balsa establishment on a corn fallow located at less than 2 km from our experimental area, recruited tree juveniles reached an average density of 3·4 stems m−2, representing 51 different species (D. Douterlungne, unpublished data). Slocum et al. (2006) obtained promising results to control montane Dicranopteris pectinata fern thickets in the Dominican Republic by a combination of manually clearing and planting seedlings of woody plants. Three years after initial clearing, fern cover was reduced to 16%, while 28% of the planted trees were taller than two metres, and spontaneously recruited tree juveniles (>20 cm tall) reached a mean density of 2·3 ± 1·5 stems m−2 (Slocum et al. 2006). Once bracken or other aggressive colonizing ferns are suppressed, trees can be recruited without human intervention. In Sri Lanka, seedlings of pioneer shrubs and trees were found in Dicranopteris linearis fernlands after soil disturbance (Cohen, Singhakumara & Ashton 1995). In Great Britain, P. aquilinum was cut to increase incoming sunlight for oak seedlings, which responded with an increase in total biomass and leaf area (Humphrey & Swaine 1997). Furthermore, Betula pendula and B. pubescens saplings invaded a 2-m wide pathway cut into dense P. aquilinum areas, and Pinus silvestris seedlings established in sparse bracken (Marrs et al. 2000).

Balsa establishment and management

Using balsa to control aggressively dominating species is not a magic bullet solution and further long-term research is necessary to fully assess the potential of this method for large-scale bracken control. Using pioneer plantations to out-compete bracken requires a successful tree establishment and growth. While balsa thrives on well-drained soils, it performs poorly in compacted soils such as in recently abandoned pastures (Douterlungne & Ferguson 2012). Furthermore, prolonged droughts can provoke massive balsa leaf senescence (D. Douterlungne, personal observation), resulting in increased light penetration and possible new emergence of bracken fronds. Therefore, the method may be less effective at the dry end of balsa's realized niche where similar broad-leaved fast-growing pioneer trees could be tested as alternative solutions for bracken control. Slocum et al. (2006) showed that tree growth and survival rates in fern thickets vary with different environmental conditions. Also, site-specific conditions like proximity of forest seeds, presence of remaining isolated trees, the age of the bracken population, herbivore activity, etc. may affect the effectiveness of bracken elimination and forest regeneration (Marrs, Johnson & Duc 1998b; Marrs et al. 2000).

The method of tree establishment can indirectly affect the success of bracken control. All tested methods had pros and cons. Direct seeding ensured efficient use of germplasm, but exposed seeds to seed predators and leaching out by heavy rains. The traditional method required large amounts of seeds and resulted in an irregular distribution of balsa trees with local unshaded patches where bracken persists as potential re-infestation nuclei. Planting nursery seedlings was more expensive but ensured higher survival rates (but not necessarily growth rates). Ninety-two per cent of transplanted seedlings survived their first growth year in bracken vegetation without any post-establishment cutting (Douterlungne et al. 2010).

Ideally, optimal number and frequency of bracken cutting should strike a balance between the cost of treatment and its effect. The lack of differences in rhizome biomass in our biweekly and monthly cutting treatments suggest that there is still margin to further reduce the minimum amount and frequency of bracken cutting events to achieve least-effort eradication of P. caudatum. Although cutting clearly improved establishment and growth of transplanted balsa seedlings in bracken land (Douterlungne et al. 2010), our results suggest that bracken could also be controlled without any additional cutting events after successful balsa establishment; it may only take longer to accomplish. The fact that bracken rhizome biomass was only slightly lower in uncut plots with surviving balsa trees as compared to the control plots is probably explained by the relatively short time between rhizome sampling and belated balsa canopy closure. Somewhat surprisingly, uncut control plots contained significantly higher rhizome biomass as uncut plots with no surviving balsa trees. This can be related to a nucleation effect whereby shade from overhanging branches of trees in successful plots limits bracken performance in uncut plots without balsa trees, in spite of the two- to four-metre wide bufferstrips between our sampling areas. As control plots were surrounded by healthy untreated bracken, they are not influenced by possible border effects. Future experimental set-ups should use broader buffer zones between sample areas to minimize potential border effects.

Socio-economic aspects and the importance of traditional ecological knowledge for ecological restoration

Success probability and adoption rate of ecosystem restoration strategies in rural tropical regions increases when initial (monetary) investments are kept minimal, and short or medium-term possibilities for return on investment are realistic (Aronson, Milton & Blignaut 2007). Establishing dense balsa stands in post-disturbance conditions is an affordable practice for contemporary small-scale farmers in southern Mexico as it is part of the local traditional swidden agriculture (Nulman, Levy & Montes de Oca 2005; see Douterlungne & Ferguson 2012 for detailed cost estimation). Investment in follow-up treatments is expected to be unnecessary after balsa canopy closure and initiation of natural tree recruitment. Commercial exploitation of balsa wood can provide income and can be combined with swidden cultivation as the tree′s rotation period of 5–8 years coincides with common fallow periods (Lamprecht 1989). Furthermore, the recovery of land suitable for agricultural or agroforestry use can boost local economies (Turner et al. 2003; Schneider & Goeghegan 2006), whereas alternative sources of income may be obtained from Payment for Ecosystem Services programs (Alexander et al. 2011).

Our research highlights the potential of traditional ecological knowledge (TEK) as an important inspiration source for efficient forest restoration strategies in rural areas. In spite of the growing call in literature to incorporate local knowledge in the design of reforestation and restoration programs (e.g. Higgs 2003; Ramakrishnan 2007; Martin et al. 2010), until today, few examples of their integration in mainstream restoration practice exist. The fulfilment of local people's basic needs largely depends on their ability to sustainably manage or restore available resources in their living environments. Restoration approaches based on naturally occurring processes with restoration potential are more likely to be recognized, adopted and adjusted by inhabitants of natural ecosystems. Future participatory field experiments with local people should therefore test the usefulness and applicability of TEK for designing or improving restoration.

Acknowledgements

The present study was financed by research grants of the Fondo Sectorial CONACYT-CONAFOR (2005-S0002-14647), the National Institute of Ecology of Mexico (INE), and the Consejo de Ciencia y Tecnología of Chiapas (CHIS-2006-C06-44603). Special thanks are due to the Lacandon people of Lacanhá Chansayab: Manuel Castellanos Chankín, his family and Adolfo Chankín for sharing their knowledge, experience and home. We are grateful to Antonio Sánchez González and Francisco Román Dañobeytia for collaboration during fieldwork. We are indebted to Duncan Golicher, Bruce Ferguson and Karen Holl for commenting on earlier drafts of this paper.

Ancillary