Community-level restoration profiles in Mediterranean vegetation: nurse-based vs. traditional reforestation

Authors


*Correspondence author. E-mail: prey@ujaen.es

Summary

  • 1 Many restoration projects aim to transform degraded vegetation into mature native plant communities. The recovery of the structure and properties of reference native plant communities is acknowledged as a landmark of restoration, and the success of restoration practices is measured in terms of community metrics. Nurse-assisted planting has been proposed as a promising technique for the restoration of Mediterranean and arid vegetation based on its success at seedling establishment rather than on the recovery of reference community properties. We investigate the adequacy of this restoration practice by shifting the focus of restoration from the species to the community level.
  • 2 In contrast to previous studies in the Mediterranean region, we assessed the efficacy of several management techniques, including nurse-based planting, in the recovery of the structure and properties of mature native reference communities. In a burned area of southern Spain, we applied widely used community analysis tools (rarefaction and species accumulation curves, Chao 2 richness estimator, species evenness and beta diversity) to compare the profiles of a reference native community and the community achieved through different management practices: traditional planting, nurse-based planting and controls where natural regeneration was allowed to occur.
  • 3 Our results showed that post-fire establishment of planted seedlings was higher under the canopy of nurses than in open ground. More importantly, nurse-based restoration increased the species richness and evenness of tall-shrubs and trees as well as the life-form diversity, with the restored community approaching the reference community in terms of both composition and structure of the bank of juveniles.
  • 4Synthesis and applications. We conclude that in contrast to traditional reforestation, where diversity is constrained by the elimination of pioneer and regenerated tall-shrubs, nurse-based restoration contributes to accelerated recovery of important community structure and properties. Thus, it should be used preferentially over traditional reforestation in Mediterranean mountains where restoration of native communities, and not merely tree establishment, is dictated under current habitat directives.

Introduction

Over centuries, major landscape changes have occurred in the Mediterranean basin in correspondence with major changes in land use and with the number, recurrence and size of fires (Pausas & Vallejo 1999; Blondel 2006). Mediterranean ecosystems have been transformed into a mosaic of landscapes. where remnant patches of natural vegetation appear in different successional stages and form part of different successional pathways. Their successional dynamics depend on the severity, size and recurrence of fire (Lloret & López-Soria 1993; Pausas & Vallejo 1999; Lloret, Pausas, & Vilà 2003), as well as on human management and the post-fire resilience of the vegetation present before the fire. Within these landscapes, natural vegetation units are considerably altered, so efforts aimed at restoring mature vegetation are needed.

Reforestation practices on degraded lands throughout the Mediterranean have long relied on procedures borrowed from commercial forestry without any attention to the species composition and diversity of native plant communities. Spontaneously regenerated vegetation (most frequently shrubs) has typically been removed to avoid competition with the planted, monospecific target trees. At a very slow pace, these practices are being replaced by techniques compatible with recent ecological views on restoration. A fundamental problem of traditional reforestation in Mediterranean areas is the high rate of early plant mortality due to severe summer drought and high rates of grazing (Meson & Montoya 1993; García-Salmeron 1995). It has been proposed that under stressful conditions, the negative effect of competition between plant species is compensated by protection against environmental stress, resulting in facilitation (Bertness & Callaway 1994; Callaway, Walker, & Abrams 1997). Subscribing to this view, research in the Mediterranean region has developed restoration practices that improve seedling survival and growth by planting under spontaneously established shrubs, which are used as nurse plants (Maestre et al. 2001; Castro et al. 2004; Castro, Zamora, & Hodar 2006; Gómez-Aparicio et al. 2004). It has been shown that compared with open areas, the canopy of nurses favours seedling performance by reducing radiation, temperature and herbivore damage and increasing soil humidity (Rey-Benayas 1998; Rey et al. 2004; Baraza, Zamora, & Hodar 2006; Gómez-Aparicio et al. 2008).

Current habitat conservation directives in many regions of the world consider the objective of reforestation to be the restoration of natural plant communities. According to this view, it is important not only to achieve the successful establishment of target species but also to reconstruct the diversity of the natural community in terms of species composition, growth forms and functional plant groups (Ruiz-Jaen & Aide 2005). The experimental studies in nurse-assisted restoration conducted to date (references above) have addressed the success of planting using different techniques, but they do not evaluate whether such techniques (e.g. traditional vs. nurse-based reforestation) achieve the goal of restoration in terms of diversity and other community properties.

Community ecology has a long tradition, with its own set of widespread analytical techniques. The evaluation of restoration success should rely on these techniques, as they allow objective comparisons of the diversity and structure of the community obtained from different methods of restoration (Ruiz-Jaen & Aide 2005). Among these techniques, rarefaction is one of the most popular. Rarefaction controls for the effects of abundance and sampling effort when estimating the species richness and diversity of a community (Magurran 1988; Colwell, Mao, & Chang 2004). This technique, developed by Sanders (1968) and later corrected by Hulbert (1971), calculates the expected number of species if all the samples had the same number of captured individuals. That is, rarefaction estimates the pace at which species accumulate as sampling effort and the number of sampled individuals increase. Thus, species richness can be more objectively compared between several communities by standardizing it to a common sampling effort or number of specimens. However, to our knowledge, these techniques and their associated community metrics have yet to be adopted in studies of plant community restoration (but see Grant & Loneragan 2003).

Fundamental to evaluating the recovery of community structure and properties is knowledge of the structure and natural regeneration rate of the community that is being restored (reference community; Ruiz-Jaen & Aide 2005). Therefore, a preliminary step should be to typify the reference community, not only as a list of species but also in terms of diversity and the recruitment rate of target species. The reference community thus gives a pattern for comparisons of the success of our restoration efforts (Tucker & Murphy 1997; Shono, Davies, & Kheng 2006). Surprisingly, this issue is frequently neglected in restoration efforts using nurse-based planting (but see Siles et al. 2008).

This study aims to evaluate the success of restoration of native vegetation using four different management practices applied in a recently burned area in southern Spain: traditional reforestation, planting under nurses, spontaneous recovery protected from grazing, and spontaneous recovery exposed to grazing. Unlike previous studies on nurse-based restoration, we aim to measure restoration success in terms of recovery of the natural community structure and properties. By considering restoration success in such terms, we move the issue of nurse-assisted restoration from a focus on success in plant establishment to a more relevant holistic view of community structure. Specifically, we contrast community metrics (species richness and evenness, growth form diversities and plant cover obtained for recruits of late-successional woody species) of the undisturbed reference vegetation with these parameters in the burned areas under different management practices.

Materials and methods

Study area

The study site is located in the Natural Park of ‘Sierras de Cazorla, Segura y Las Villas’ (Jaén, SE Spain) and encompasses 836 ha affected by wildfire in August 2001. The potential native vegetation is a mixed forest of Pinus nigra, Quercus ilex and Quercus faginea (Valle 2003). However, as a result of traditional management for forestry and cattle raising, the vegetation was dominated by old afforestation stands of Pinus pinaster, Pinus nigra and Pinus halepensis, with a sparse spontaneous undergrowth of native shrubs and trees, and small patches of open native scrubland (Luque 1995). After the fire, the burned area became a mosaic of zones in different stages of regeneration: zones in initial state (woody species nonexistent), zones in pioneer state (dominated by early pioneer scrub vegetation), building state (dominated by pioneer and large woody species) and small remnants of unburned vegetation. Most of the burned area is formed by a limestone outcrop ranging from 760 to 1390 m elevation, with very shallow soils and steep slopes. Mean annual rainfall in the closest weather stations ranges between 771 and 1156 mm, concentrated in autumn and spring. Average annual temperature ranges from 11·6 to 14·2 °C.

Data collection

We chose four zones in the burned area with vegetation dominated by pioneer and resprouted shrubs. These zones were chosen to be distributed over all the burned area. They were 0·5–3 km from each other, with elevation ranging from 800 to 1100 m. Soil, slope and aspect were similar between zones. In each zone, we placed four 25 × 25 m adjacent experimental plots. Three plots were fenced to prevent ungulate grazing, and management practices were: traditional reforestation, planting under nurses, and a control plot for spontaneous recovery (no planting). The fourth plot was a control for spontaneous recovery (no planting) with exposure to ungulate grazing. Close to each zone, we chose a 25 × 25 m plot of unburned intact vegetation (avoiding previously reforested areas and/or logged vegetation) as representative of the reference community. Thus, our experiment conformed to a randomized-block design, with each block (zone) composed of the four treatments plus an adjacent patch of the reference community. Patches of reference community maintain vegetation comparable to the mature native plant community of the study area (Valle 2003) and have remained unburned over many decades.

Experimental plantings

The experimental community was planted with 1-year-old seedlings of 10 to 25 cm height, depending on the species. We planted five tree species (Q. ilex, Q. faginea, Prunus mahaleb, Sorbus torminalis and Fraxinus angustifolia) and seven tall-shrubs (Crataegus monogyna, Juniperus oxycedrus, Phillyrea latifolia, Pistacia lentiscus, Pistacia terebinthus, Quercus coccifera and Rosa sp.). In total, we planted 1044 specimens (October 2005), which were labelled and mapped in the plots.

Traditional reforestation.  In these plots, naturally regenerating pioneer vegetation was eliminated using a string trimmer. Subsequently, we planted 60 seedlings of different tree species in each plot, with species distributed according to the characteristics of each zone (Siles et al. 2009). In total, we planted 240 seedlings of five tree species in four zones. Only tree species were planted in these plots as traditional reforestation does not use shrub species. The distance between planted specimens was 3 × 3 m.

Nurse-based reforestation.  We planted seedlings of tall-shrubs and trees (the same tree species as in traditional reforestation) under plants spontaneously established in the plot (potential nurses). Species used as nurses were Cistus monspeliensis, Daphne gnidium, J. oxycedrus, Pistacia lentiscus, Q. faginea, Q. ilex, Rosmarinus officinalis, Thymus mastichina, Teucrium poleum and Ulex parviflorus. Distances between planted seedlings and the number of seedlings planted on each plot were determined by the number of established nurses (ranging between 162 and 257 per plot). At the time of planting, nurse plants were less than 1 m in height. Transplants were planted just below, or adjacent to (less than 10 cm apart), the canopy of nurses, always in a northern orientation to increase shading. In total, we planted 804 seedlings of five trees and seven tall-shrubs species.

Establishment success and seedling growth under different restoration techniques

The survival and height of planted seedlings in the planting treatments were recorded in March, June and September 2006, June and September 2007 and September 2008. These data were analysed as a randomized-block design using the procedure GLIMMIX of SAS/STAT (SAS Institute Inc. 2003). Each zone was considered a block (random factor), and treatment factor (traditional vs. nurse-based planting) was considered fixed. The statistical interaction between these two factors explored the consistency between zones in the treatment effect. Seedling establishment was analysed considering binomial error with a logit link function. Seedling height was analysed using a normal distribution and identity link function.

Community-level restoration profiles

We determined the abundance and species richness of tall-shrubs and trees, resulting both from the survival of planted seedlings and spontaneous regeneration. We counted all specimens for each species within each 25 × 25 m plot corresponding to each treatment. To quantify recruitment in the unburned reference vegetation, we counted the established seedlings and saplings (<1 m tall) of tall-shrubs and trees. The seedling community formed in each experimental treatment was evaluated in terms of: (i) tree species obtained only by planting, (ii) tall-shrub and tree species obtained only by planting and (iii) tall-shrub and tree species obtained both by planting and spontaneous regeneration. We grouped the information from the four plots corresponding to each treatment to obtain rarefaction estimates of species diversity. Differences between treatments were evaluated in June 2006 (soon after planting, to account for early seedling establishment levels previous to summer water stress) and in September of 2007 and 2008 (2 and 3 years after planting, when survival had largely stabilized).

Species composition, richness, diversity and evenness

To compare the different seedling communities established under each restoration practice and in the reference community, we analysed their community properties through measurements of species richness, diversity and evenness. We used rarefaction (which controls for different sample sizes) and the Chao 2 estimator (which controls for undetected species) to analyse species richness. These analyses were conducted with Estimate 8.0 (Colwell 2007), which is widely used in community analysis. In our case, the basic unit of study for rarefaction analysis was each planted specimen (including those that did not survive through the entire experiment). Rarefied species richness was measured as Mao Tau S (Colwell et al. 2004; Colwell 2007), determined by the calculation of rarefaction curves based on surviving individual specimens. We standardized the expected species richness (i.e. Mao Tau S) to a maximum common number of specimens surviving under the different treatments (i.e. the point for comparison was the number of seedlings recruited in the treatment with the lowest number of recruits) to compare among treatments. Expected species richness was considered to be different when the 95% confidence intervals did not overlap at the comparison point. At this point, we also evaluated Pielou’s evenness index. We further evaluated total richness through the bias-corrected Chao 2 estimator (and its 95% confidence intervals). This estimator uses data on the rare species collected in the samples to estimate the number of additional species that are expected to be present but were not detected in the samples, and it estimates total richness (Colwell & Coddington 1994).

Theoretically, two assemblages of entirely different species composition can have identical richness, evenness and distribution of abundances. Thus, comparisons between the experimental plantings and the reference community must also account for species composition (Grant & Loneragan 2001; Sluis 2002). As an index of similarity in species composition between treatments and the reference community (β-diversity), we calculated the quantitative Sørenson’s index of similarity (Magurran 2004).

Analyses of tall-shrub and tree cover, and diversity of life-forms

We used the cover represented by tall-shrubs and trees and the diversity of life-forms in each treatment as other descriptors of community complexity. Plant cover for each species was estimated as two-dimensional cover (from two perpendicular diameters), using a metric tape, and subsequently standardized to the surface of the plot. Species present in each treatment were grouped according to several life-forms: spiny tall-shrubs (SPYN), non-spiny tall-shrubs (NSTS), deciduous trees (D), evergreen trees (E) and vines (V). Life-form diversity was estimated using Shannon’s index. The cover obtained by planting and spontaneous regeneration was quantified 3 years after planting as the sum of the cover of every individual in the plot, and it was statistically compared between treatments using Friedman’s nonparametric anova (each zone being a block) and Kendall Coefficients of Concordance.

Results

Establishment success using different planting techniques

At the end of the experiment, there was a significant effect of planting treatment (nurse vs. traditional reforestation) on seedling survival (F = 13·64, < 0·05). This effect was evident from the first summer (i.e. from the first period of water deficit; Fig. 1a, Table 1) and was consistent among zones (treatment by zone interaction was not significant; Fig. 1b, Table 1). Three years after planting, the probability of survival was much higher under nurses (0·66 ± 0·25; least-square mean after controlling for block effect ± SE) than with traditional planting (0·01 ± 0·02). In fact, in three out of the four zones, seedling survival after 3 years was zero in the traditional planting (see Fig. 1b). We did not detect an effect of the planting treatment on seedling height after 2 years in four out of five tree species planted (treatment effect, > 0·25 in each species). Nevertheless, growth of Prunus mahaleb was slightly higher under nurses than in traditional planting during the first year (F1,48 = 10·92, < 0·01; estimate of differences between treatments = 4·34 ± 1·32 cm) and the second year (F1,32 = 3·9, < 0·05; estimate of differences between treatments = 2·80 ± 1·42 cm).

Figure 1.

 (a) Cumulative survival of planted seedlings in traditional and nurse-based restoration; (b) Observed survival ± 1 SE for each treatment and zone after 3 years.

Table 1.   Results of randomized-block anova for seedling survival at different times after planting
Effects200620072008
MayJuneSeptemberJuneSeptemberSeptember
  1. Treatment fixed effect (traditional vs. nurse-based plantings) is evaluated through F-statistic. Zone and zone by treatment (Z × T) are considered random effects (z-statistic). Effects significant at P < 0·05 l are indicated in bold type. Degrees of freedom (d.f.) are corrected by Satterthwaite approach.

Zone0·930·550·890·800·750·75
Treatment (d.f.)12·68 (1, 2·34)7·26 (1, 2·79)10·78 (1, 3·08)18·18 (1, 3·26)14·22 (1, 3·65)13·64 (1, 3·65)
Z × T0·821·091·121·050·990·99

Community-level diversity profiles using different restoration techniques

Diversity of trees obtained by planting

In June of 2006, there was no significant difference in species richness between planting treatments. For the common maximum value of established seedlings (130), the expected richness (Mao Tau S) was five species in both treatments. Two years after planting, 131 seedlings survived under nurses, while only 25 (the common maximum of established seedlings) survived in traditional reforestation. At this time, the expected richness showed significant differences between treatments, reaching values of 2 ± 2·15 species (Mao Tau S ± 95% confidence limits) in traditional reforestation and 4·96 ± 0·03 under nurse plants (Fig. 2a). These differences remained during the third year after planting, with 129 surviving seedlings under nurses and 25 in traditional reforestation. Therefore, 2 years after planting, the diversity of trees in traditional reforestation decreased to half the diversity obtained in plantings under nurses.

Figure 2.

 Rarefaction curves of species richness obtained through traditional vs. nurse-based restoration. (a) Tree species only; (b) trees and tall-shrubs. Rarefaction curves are shown for results obtained just prior to the first summer (Jn-06) and after the second and third summer (Sp-07 and Sp-08 respectively). Vertical lines indicate the point of comparison [i.e. the common maximum number of recruited seedlings = 130 in June of 2006 and 25 in September of 2007 and 2008 for (a and b)].

Diversity of tall-shrubs and trees obtained by planting

The differences between planting treatments in terms of the diversity of tall-shrubs and trees are illustrated in Fig. 2b. In June of 2006, maximum seedling establishment was 130 plants. The expected richness was then 5 ± 0·0 (Mao Tau S ± 95% confidence limits) in traditional reforestation and 11·96 ± 0·05 in reforestation under nurses, a significant difference (non-overlapping 95% confidence intervals). Differences between treatments increased 2 years after planting, when 484 seedlings were established under nurses and only 25 survived in traditional reforestation. At the point of comparison (25 established seedlings), the expected species richness differed between treatments, with values of 9·87 ± 0·89 under nurses and 2 ± 2·15 in traditional plantings. Thus, the diversity of large woody species (tall-shrubs and trees) increased almost fivefold using nurses compared with traditional plantings, in part because of improved seedling establishment, but also because tall-shrubs are not planted in traditional reforestation.

Diversity obtained by planting and natural regeneration

Considering the specimens derived both from planting and from spontaneous regeneration, at the end of the experiment we found 628 established seedlings or saplings of tall-shrubs and trees in plots with nurse-based restoration, 321 in control plots free of grazing, 94 in traditional reforestation and 54 in controls exposed to grazing. In reference community plots, we found 507 seedlings or saplings of tall-shrubs, trees and vines (see Table S1).

At the point of comparison, the expected species richness in the reference community was 12·01 ± 2·29 (Mao Tau S ± 95% confidence limits). Three years after planting (Fig. 3), nurse-based restoration had a richness of 11·84 ± 0·81, while lower values were obtained in controls exposed to grazing (9·0 ± 2·49), traditional reforestation (8·56 ± 1·57) and in controls free of grazing (6·4 ± 2·07). Based on the non-overlapping 95 % confidence intervals, there were significant differences in the expected species richness among restoration treatments. While reference communities and nurse-based restoration plots did not differ from each other statistically, they exhibited higher richness than traditional reforestation or control plots free of grazing. Species richness in controls exposed to grazing did not differ from any other treatment.

Figure 3.

 Rarefaction curves of species richness obtained 3 years after planting under different restoration practices, including established seedlings from both planting and natural regeneration (see Materials and methods). Species curves refer only to seedlings/saplings of tall-shrubs and trees. Vertical lines indicate the point of comparison (i.e. the common maximum number of recruited seedlings = 54 in September of 2008). Treatments are: traditional reforestation (T), nurse-based restoration (UN), controls of natural regeneration exposed to grazing (CE) and free of grazing (CI) and the reference community (RC).

The Chao 2 richness estimator showed that the highest richness occurred in the reference community (17 ± 0·0, Chao 2 ± confidence limits), followed by planting under nurses (13 ± 0·0), controls exposed to grazing (9·44 ± 0·45), traditional planting (9 ± 0·03) and controls free of grazing (8 ± 0·0). Based on the non-overlapping 95% confidence intervals, there were significant differences among reference communities and the other treatments. The reference community and nurse-planting treatments have an estimated richness higher than traditional planting and controls of natural regeneration.

Evenness showed very high values in nurse-based restoration, traditional reforestation, controls exposed to grazing and reference communities (J = 0·92, 0·90, 0·85 and 0·77 respectively). These values indicate that the number of recruited seedlings (considering both planted and spontaneously regenerating individuals) was evenly distributed among species. Evenness in controls free of grazing was lower (J = 0·52), indicating that the community is dominated by a few species (Q. ilex represented up to 70% of the seedling recruitment). Sorenson’s Index between the reference community and the managed plots gives values of 0·70 for nurse-based restoration, 0·56 for grazing-free control, 0·22 for traditional planting and 0·19 for grazing-exposed control.

Diversity of life-forms obtained from planting and natural regeneration

Planting tall-shrubs and trees under nurses functionally enriched the naturally regenerated community, which lacked many deciduous trees and SPYN. The life-form diversity was greatest in nurse-based restoration (Shannon index = 0·56) and followed the order nurse-based > reference community (0·52) > traditional planting (0·47) > control exposed to grazing (0·41) > control free of grazing (0·34). Vines only occurred in the reference community. There was no spiny shrub recruitment in traditional reforestation, although they were recorded in other plots. In spite of these differences, traditional reforestation showed higher life-form diversity than controls (exposed or free of grazing) because of the over-dominance of evergreen trees (especially Q. ilex) in controls free of grazing and of NSTS in controls exposed to grazing (Fig. 4).

Figure 4.

 Partitioning of total seedling abundance by plant life-form obtained in the reference community and through restoration practices (3 years after implementation). Life-forms are spiny tall-shrubs (SPYN), non-spiny tall-shrubs (NSTS), deciduous trees (D), evergreen trees (E) and vines (V). Acronyms of treatments as in Fig. 3.

Cover of tall-shrub and trees obtained by planting and natural regeneration

Our results show that 2 years after planting, the highest cover was obtained by planting under nurses, and that grazing reduced cover growth. Treatments differed significantly in tall-shrub and tree cover 2 years after planting (Friedman anovaχ2 = 8·40, = 0·038, d.f. = 3), with a concordance between zones in the ranking of treatment plant cover (concordance of Kendall, W = 0·7, average rank order correlation, r = 0·938). The ranking of plant cover was led by planting under nurse plants (8·10 ± 4·26%; mean + SE), followed by controls free of grazing (7·07 ± 4·08%), controls exposed to grazing (5·93 ± 3·40%), and lastly by traditional reforestation (3·35 ± 2·53%).

Discussion

Current habitat conservation policies (e.g. European Union’s Habitats Directive 92/43/CEE) state that management in abandoned and degraded habitats should aim to re-establish the structure and function of native plant communities (SER 2002). However, most current ecological approaches to this issue in the Mediterranean region are relatively simplistic because they measure the success of restoration techniques in terms of juvenile establishment, without explicit consideration of recovery of species composition, structure and properties similar to natural reference communities. In contrast to other studies in the Mediterranean region, here we have shown that nurse-assisted planting is not merely successful in achieving seedling establishment of many species but, more importantly, in assisting the recovery of community structure and properties. Thus, it should be preferred to traditional reforestation or no intervention in degraded Mediterranean mountain areas.

Establishment success using different restoration techniques

Traditional techniques for reforestation frequently show low establishment success because of summer mortality through water stress (Meson & Montoya 1993). In contrast with these practices, many studies have shown that the regeneration niche of woody species in the Mediterranean mountains is positively associated with shrubs (Rousset & Lepart 1999; García et al. 2000; Rey & Alcántara 2000). These studies confirmed a facilitative effect of shrubs and grasses on the early establishment of planted woody species. New restoration techniques for Mediterranean landscapes have been developed using naturally established shrubs and herbs as microhabitat for planting seedlings of woody species (Maestre et al. 2001; Castro et al. 2004, 2006; Gómez-Aparicio et al. 2004). Our results support this view, showing that post-fire establishment of planted tree seedlings is several times higher under the canopy of nurses (spontaneously established vegetation) than in open interspaces. Studies in Mediterranean environments have concluded that summer drought occurring during the first years after planting is the most limiting factor for the establishment of planted seedlings (García-Salmeron 1995; Rey-Benayas 1998; Gómez-Aparicio et al. 2004; Rey et al. 2004). The facilitative effect of nurses is primarily due to micro-environmental amelioration of water stress and irradiance and, secondarily, to improvement of soil texture and nutrient availability or grazing avoidance (Pugnaire et al. 1996; Gómez-Aparicio, Gomez, & Zamora 2005; Padilla & Pugnaire 2006; Baraza et al. 2006). Our results, showing increased seedling establishment under nurses compared with open interspaces, support the idea that a shift from traditional to nurse-assisted planting techniques will enhance post-fire seedling establishment of Mediterranean vegetation (Castro et al. 2004, 2006; Gómez-Aparicio et al. 2004) through alleviation of drought stress.

Community-level restoration profiles

The aims of ecological restoration should move beyond merely enhancing seedling establishment. A successful restoration project must achieve a community with the native components (species) and properties (structure, dynamics and resilience) of natural mature communities. Restoration techniques in seriously degraded areas should thus promote necessary changes in species composition and dynamics to eventually achieve the mature community (SER 2002). However, traditional techniques of reforestation have paid little attention to community properties and species composition. More surprisingly, more ecologically based techniques of restoration in the Mediterranean focus mainly on the establishment of multiple species and most frequently assume, rather than demonstrate, the recovery of community properties (but see Tucker & Murphy 1997; Shono et al. 2006 for other ecosystems, and Siles et al. 2008 in the Mediterranean). This study has attempted to bridge this gap, applying widely used community structure analytical tools to compare the community profiles resulting from different restoration practices, with control plots in naturally regenerating forest and reference plots in remnants of well-preserved communities.

Rarefaction analysis showed increased diversity of both trees and tall-shrubs in nurse-based restoration compared with traditional reforestation. Seedling diversity of planted trees can be doubled using nurse plants. Because we planted the same number of species in each treatment, the bank of juvenile tree species was comparable among treatments, and the results suggest that facilitation is shaping the diversity of the juveniles. Moreover, if planted tall-shrubs are also considered, the species diversity is almost five times higher than in traditional plantings. It could be argued that this last difference is artificially inflated as no shrubs were planted in the traditional treatment. However, we think that our conclusion is not an artefact of experimental design. First, traditional reforestation tends to eliminate rather than support shrub species, while more ecologically based techniques, such as nurse-based restoration, implement planting of tall-shrubs as important components of succession (Gómez-Aparicio et al. 2004; Castro et al. 2006). Secondly, we have additional evidence corroborating that most species of tall-shrubs fail to establish in our study area when planted without nurses (traditional reforestation) and that the mechanisms acting in their establishment under nurses are, primarily, due to amelioration of abiotic stresses and, secondarily, protection against ungulate grazing (G. Siles, P.J. Rey & J.M. Alcántara, unpublished data). Therefore, we believe the inclusion of tall-shrubs is a valid comparison of these restoration practices.

More interestingly, nurse-based restoration substantially increased the overall diversity when both planted seedlings and naturally regenerated species were considered. The community of established juveniles became extremely similar to that of adjacent well-preserved vegetation (i.e. the natural reference community). If compared with control plots of post-fire natural regeneration and traditional reforestation, the expected number of species of the nurse-based restoration was significantly higher and more similar to the reference community, indicating that planting under nurses effectively enriches the post-fire regenerated community, approaching the reference community in terms of both species composition and structure.

Evenness was very high in all treatments (including the reference community), except in the grazing-free control. This means that recruited seedlings were evenly distributed among species in the natural community and that the high evenness initially generated by planting did not result in an artefact in the structure of the restored community. In spite of these similarities, the rarefaction curves for nurse-based restoration and traditional planting saturated at lower richness, and showed higher evenness, than the curves for the reference community (which showed significantly higher Chao 2 richness than any management practice). Apart from the effect on evenness created by an even planting (in both traditional and nurse-based reforestation), the lower evenness and highest Chao 2 in juveniles of the reference community reveal the lack of rare species in the restored community. Multiple factors (not considered in our restoration practices) determine differences in natural recruitment among species, including species-specific seed removal, dispersal and predation, germination rate and seedling survival (Rey et al. 2002, 2004; Ramírez et al. 2006; Acacio et al. 2007). We would expect that evenness in planting treatments will approach that of the reference community as differential mortality among planted species occurs over time. Interestingly, the lower evenness in grazing-free controls of spontaneous regeneration, compared with exposed controls, suggests that frequency-dependent ungulate grazing is controlling species that otherwise would dominate the regenerated community.

The better community performance resulting from nurse-based restoration is also confirmed by the comparison of other community parameters, including woody cover, and number and diversity of growth forms. All these parameters are favoured under this restoration practice as it does not involve elimination of the naturally regenerated vegetation. The number of different growth forms is higher in the reference community, similar among the controls of natural regeneration and nurse-based restoration, but lower under traditional reforestation. The single type of life-form missing under nurse-based restoration is vines (which only appeared in the reference community), but these are expected to spontaneously enter the community some years after planting. Chao 2 richness of the nurse-based restoration community is then expected to approximate that of the reference community. Of particular importance is the lack of SPYN and vines, most frequently fleshy-fruited species, in traditional reforestation plots. We have evidence that the absence of fleshy fruits in burned areas decreases their attractiveness for frugivorous birds (Siles, Rey & Alcántara, unpublished; R. Zamora, personal communication), which are the major agents of seed dispersal in Mediterranean vegetation. Lack of fleshy fruits may thus cause succession to remain arrested in these zones (Debussche, Escarre, & Lepart 1982; McClanahan & Wolfe 1993; Acacio et al. 2007).

Restoration practices should favour species compositions and assemblages such that the successional pathways conducive to a defined reference community are promoted (SER 2002). It is too soon to draw these conclusions from our short-term study, and it is clear that long-term surveys using experimental designs like the ones presented here are needed to confirm such possibility. However, there is now some experimental and modelling evidence that suggests the capability of nurse-based restoration to launch secondary succession (see Gómez-Aparicio et al. 2004). In a parallel study in the same region, we simulated the post-fire dynamics of 20-year-old burned areas through Markovian models of species substitution (Siles et al. 2008). We corroborated that sites with frequent establishment of tall-shrub and tree seedlings under nurse plants projected a stable community of similar composition to a defined reference community (indicative of successful secondary succession). In contrast, sites with low seedling establishment under the established vegetation projected a pioneer community (indicative of a situation of arrested succession).

The modelling example above may also serve as an indicator of success in achieving another desired ecosystem property, the resilience or ability to quickly recover after a disturbance (SER 2002). Regardless of such findings, we believe that the much higher expected species richness obtained for nurse-based restoration in our experiments renders a quick increase in ecosystem resilience, as most of the naturally regenerated and planted species have resprouting capacity (Pausas & Verdú 2005). In contrast, the loss of tall-shrubs and trees in plots under traditional reforestation would imply a lowered resilience, as this technique causes an impoverishment of functionally redundant species, and a concomitant increase in the ecosystem’s sensitivity to disturbances (Fonseca & Ganade 2001).

To conclude, our study is the first to demonstrate that nurse-assisted restoration contributes to the recovery of important community properties. Many studies have shown that nurse-based planting enhances the establishment of woody species. We move beyond the species level to assess the efficacy of this technique at the community level. Compared with traditional reforestation and/or no intervention, the nurse-assisted restoration enhances: (i) species and life-form diversity in the community, (ii) late-successional species cover, (iii) the chance for secondary succession and, presumably, (iv) resilience in the restored community. In contrast, traditional techniques of reforestation in Mediterranean regions fall short of obtaining communities similar to reference communities. Therefore, nurse-assisted restoration should be implemented when the aim of management is the restoration of native communities.

Acknowledgements

We thank the Junta Rectora of Sierras de Cazorla, Segura y Las Villas Natural Park for the permission and facilities to work in the burned area. The staff of the Escuela de Capacitación Forestal de Vadillo and, particularly, Jose Luis Herreros and students helped with planting tasks. We thank José Manuel Ramírez, Jesús Bastida, Rafa Jaime and Eli Fernández for their help in the field. This paper considerably benefited from comments of three anonymous reviewers. Housing facilities at Sierra de Cazorla were provided by Estación Biologica de Doñana. GS was granted by an FPI fellowship of the University of Jaén (UJA). Funds came from ‘Convenio de asesoramiento de las actuaciones de restauración del Puerto de las Palomas’ between Consejería de Medio Ambiente (Junta de Andalucía) and UJA.

Ancillary