A biological community is a group of organisms that interact and share an environment. Within a community, organisms may compete for the same resources (competition), profit from the presence of other organisms (facilitation) 3839 or use other organisms as a food source (trophic interaction) 35. In stable communities, these interactions lead to predictable, directional changes in community structure known as ecological succession. Succession is an important guiding principle in the community approach to ecological restoration 40. The restoring forest is a dynamic ecosystem, with changing species composition and forest structure, but interventions and management steer the forest towards a desired climax or pre-disturbance community structure. These interventions are usually designed to accelerate natural succession or to bypass intermediate successional phases. Basically, the community approach is focussing on restoring forest biodiversity per se. The many studies that apply facilitation as a restoration tool of woody communities 41 are typical examples of the community approach to forest restoration. Planting late-successional tree species (protégé species) under early-successional shrubs (nurse species) has been shown to be an effective means of restoring forests under high abiotic stress 4243 (Figure 1).

Insights from alternative stable states theory have also been useful to guide restoration practices that focus on community structure 44. In severely degraded systems, alternative stable states may make efforts to restore pre-disturbance communities difficult, if not impossible 45. In such cases, a single intervention may not suffice to induce forest regrowth: succession fails and the community is blocked in a low diversity/low biomass state. Exclusion of grazing animals may be an effective means for woodland restoration in degraded drylands, but only when soil moisture conditions also improve. Wet pulses caused by climatic oscillations such as the El Niño Southern Oscillation may provide such necessary additional impulse to induce a regime shift that leads to forest restoration 46. Similarly, planting and sowing of late successional tree species (an intervention to overcome seed limitation) has been found effective for the restoration of highly complex forest on bauxite mined sites, but only after careful site preparation and topsoil handling or replacement (interventions to overcome survival limitation caused by soil compaction, decreased soil porosity and infiltration capacity, and the loss of soil biota) 4748.

Restoration of species richness and community structure over time implies increasing ecosystem complexity and functionality 40. In the ecosystem approach, restoration of ecosystem functions such as primary production, energy flows and nutrient cycles, is the guiding principle on which restoration efforts are based 35. Basically, this approach aims at restoring suitable abiotic conditions that allow (passive) recolonization of species. The ecosystem perspective typically starts from a landscape point of view, building on spatial heterogeneity and broad spatial scales 49. The connections or barriers between neighbouring ecosystems have an effect on the resource balances and set limits on the communities that can be restored 50. Reforestation of degraded sites with trees that alter the physical and chemical characteristics of the soil and that affect the biochemical cycles through litter fall or root activity presents a typical example of the ecosystem approach to forest restoration 5152.

One of the major functions of forest ecosystems is carbon fixation 60, which is directly related to the ecosystem services carbon sequestration and the provision of fire and construction wood. There is evidence that tree diversity has a positive effect on ecosystem production (see Thompson et al. 61 for an overview). Based on the largest data set ever analysed in this context to date (12.000 permanent forest plots in eastern Canada), Paquette & Messier 62 reported that, after controlling for environmental and climate differences between plots, tree productivity was positively related to stand biodiversity. These results confirm earlier work in 5000 permanent plots in Mediterranean forests across Catalonia (NE Spain) 63. In a reforestation context, Piotto et al. 64 found that mixed plantations in Costa Rica performed better than monocultures for all growth variables considered, including height, diameter at breast height, volume, and above-ground biomass. Also in natural stands of tropical forest with high environmental and spatial variation, positive effects of tree species diversity on tree carbon storage were found 65. Positive effects of tree diversity on above-ground productivity are certainly not an universal pattern, however 6166, and above-ground biomass production and soil carbon fixation may also respond differently to tree diversity in plantation forests 67. This corroborates the result of a meta-analysis of BEF experiments where it was found that high biodiversity treatments do not always outperform the best performing monoculture 55. In a forest restoration context, where fast growing tree species with strong global timber markets are readily available, this may suggest that monocultures are an option. However, evidence is accumulating that focussing on one single ecosystem function often overlooks an important aspect of biodiversity: the possibility of one species to contribute to different ecosystem functions at the same time 68. Because different species often influence different ecosystem functions, focussing on one function in isolation will strongly underestimate the biodiversity required for maintaining an ecosystem with multiple functions, at multiple times and places in a changing environment 69. Although the evidence only comes from grasslands and aquatic environments so far, it convincingly shows that species redundancy is unlikely to occur when several ecosystem functions and services are considered in combination 68697071.

Therefore, it is highly unlikely that species poor plantations will outperform species diverse tree assemblages for a combination of forest ecosystem functions 28, including above-ground biomass production, disease resistance, carbon fixation, nectar provision, erosion control, water capitation, N₂-fixation and fruit production. It is therefore of special importance that reforestation efforts clearly define the ecosystem services and functions that the restored forest is intended to deliver. Also, it is important to realize that ecosystem functions of restoring forests may change over time because of changes in tree sizes, forest structure and relative importance of functional groups, even if there are no changes in tree species composition 72. Finally, it should be noticed that although there is already some knowledge on the effects of tree diversity on forest productivity, it is not known how understory shrub diversity, and even herbaceous species, affect forest productivity or other ecosystem functions. This may, for example, happen through these species' impacts on litter decomposition, on water capture and on the diversity of soil biota 73.

The hypothesis that larger species diversity leads to higher stability of ecosystem functioning has been a point of debate for half a century, and it has re-emerged within the BEF framework 537475. The main ideas behind the biodiversity vs. ecosystem stability concept are functional response diversity and functional compensation 6176. This occurs when positive changes in the level of functioning of one species (a species becoming functionally dominant) are associated with negative changes in the functioning of other species. This compensation drives the stabilization of ecosystem properties such as biomass production 77. Basically, the stability of the functioning of an ecosystem can be measured in three ways: i) the long term variability of an ecosystem property through time in relation to background environmental variation (variance); ii) the impact (resistance); and iii) the recovery (resilience) of ecosystem properties to discrete disturbances 6178. As it is expected that these discrete and extreme disturbances such as extreme climate events and pest and disease outbreaks will become more frequent under the predicted climate change 79, it is very important to incorporate insights from the relation between biodiversity and stability of ecosystem functioning into forest restoration projects. It is crucial to realize that, just as the degree of species redundancy decreases when multiple ecosystem functions are considered (see earlier), there is currently strong experimental evidence that in changing environments, more species are required to guarantee ecosystem functioning than in constant environments [e.g., 6980].

Evidence for the latter comes from studies that have related forest tree diversity with measures of stability of forest ecosystem functioning. Lloret et al. 81 used satellite imagery to estimate the impact of the extreme 2003 summer drought on canopy greenness of different forests types in Spain, by quantifying the NDVI (normalized difference vegetation index). NDVI correlates with ecosystem CO₂ fluxes. These authors reported a positive relation between woody species diversity and resistance of canopy greenness against drought in forests on dry locations, whereas no such relation was discovered in more moist forests. Similarly, DeClerck et al. 82 related stability in stand productivity across 64 years with conifer diversity in the Sierra Nevada, USA. They found a significant relation between species richness and resilience of stand productivity after recurrent severe droughts. Resistance to drought was, however, not related to species diversity. These studies partly support positive biodiversity effects on stability of biomass production, but they also show that patterns may be complex, vary across ecosystem types, and depend on the measures that are used to quantify stability. In any case, temporal stability of ecosystem functioning is an important consideration for projects aiming at forest restoration, especially under the current global change scenario. Again, it is not known whether understory shrubs and herbaceous species contribute to the stability of forest ecosystem functioning.

Whereas general biodiversity measures are based on taxonomy in the first place (species presence or absence), functional diversity measures relate to what organisms effectively do in an ecosystem, quantify the distribution of traits in a community or measure the relative magnitude of species similarities and differences. How to best measure functional diversity is a much debated question, but Cadotte et al. 83 summarize five useful multivariate functional diversity measures. Some authors have suggested that functional diversity measures are particularly suitable or even better to predict the interactions between biodiversity and ecosystem processes 838485. Using a tree diversity index based on among-species variation in seed mass, wood density and maximum height Paquet and Messier 62 showed that this measure outperformed a taxonomically based diversity index in explaining tree productivity. Bunker et al. 86 demonstrated that removing certain functional groups from a tropical forest had more important effects on the above-ground carbon pool than randomly removing species. Vila et al. 63, on the contrary, reported that functional group richness performed worse than tree species richness, but this was likely due to a rather rudimentary functional group delineation. Thus, when selecting tree species for forest restoration, these findings suggest focusing on functional groups based on relevant plant traits. While these traits are readily available for species from temperate regions by now, the establishment of plant trait databases for tropical tree species and the centralisation of all available data in a general database are important works in progress 8788. Maximizing functional diversity can be achieved by quantifying the functional diversity of the species mix used for restoration. This can be done by delineating emergent or functional groups (assemblages of species performing similar functional roles) [e.g. 6189], or by using more complex, continuous or non-grouping measures of functional diversity 90. The selection of relevant plant traits remains, however, crucial with respect to the forest ecosystem functions to be restored. Scherer-Lorenzen et al. 58 provide a comprehensive list of species traits that can be used to quantify functional diversity of tree mixtures used for reforestation of European temperate forests. Selected traits included nominal (e.g. leaf type, crown architecture), ordinal (e.g. adult light requirements, height growth vigour) and scale variables (e.g. leaf N concentration, litter C:N ratio). A better mechanistic understanding of how species traits and their interactions affect ecosystem functioning is also important, however, to be able to proactively analyse different reforestation scenarios and their impact on forest functioning. In this context, it is important to realize that relationships between functional traits and ecosystem functions such as carbon storage in natural populations are not always transferable to tree plantations and vice versa 57.

Whereas conservation biologists have acknowledged the negative fitness consequences of reduced genetic diversity for decades, forest restoration projects may still incorporate very few genotypes 91. There is evidence, however, that monoclonal populations are more vulnerable to pathogens than genetically diverse assemblages [e.g. 9293]. The point that we want to make here, however, is that the effects of stand genetic diversity can be expected to extend far beyond the fitness of the individual trees or stands. It is only recently that it has become clear that variation in population genetic diversity or in genotype composition can have far-reaching ecological effects. The ecological consequences of genetic diversity (coined 'community genetics') have been demonstrated at different levels of organization, from the population over the community to the ecosystem 949596. For example, plant genotypic diversity and genotype identity have been shown to affect biomass production and community invasibility, and also the invertebrate diversity of the higher trophic levels 9798. It was also shown that litter decomposition and nutrient release differed between different Populus genotypes, indicating that selection of tree genotypes may have profound and long lasting effects on ecosystem functioning of restored forests 99100. Although a discipline as community genetics is in its infancy, there is already some evidence to suggest that there are extended consequences of plant genetic variation, up to the level of the ecosystem properties 96. The selection of specific genotypes, and the genotypic diversity of tree assemblages, may therefore have major implications for the functioning and the resilience of forests 61.

The above-ground biodiversity of forests also comprises fauna with important ecosystem services that include pollination, pest control and seed dispersal. The ecosystem services of birds, for instance, have been well documented 101 and in the light of forest restoration birds have been shown essential for dispersing tree seeds into restoring areas and overcoming seed dispersal and germination limitation 102103. Far less is known about the role of below-ground biota and the linkages between trees and these biota. The study of soil microbial community structure and functioning has traditionally received little attention in ecology. But as with above-ground biodiversity, there is evidence that below-ground diversity has a significant impact on ecosystem functioning. In a series of simplified tropical forests, Lovelock and Ewel 104 found significant positive relationships between the diversity of arbuscular mycorrhizal fungi (AMF) and ecosystem net primary productivity, and between AM fungal community evenness and ecosystem phosphorus-use efficiency. The quick development and availability of molecular tools such as t-RFLP and next-generation sequencing to quantify microbial diversity [e.g. 105106], together with the strong focus of the BEF approach on the functionality of ecosystems, has resulted in an increased interest in the role of microbial soil community diversity in driving processes such as organic matter decomposition and plant nutrient uptake. Because ecological restoration is usually occurring on highly disturbed or degraded sites, it is important that above-ground-below-ground species linkages are permanently considered during the restoration process 34, and more specific, that there is a synchronisation between above- and below-ground species associations 107. Clearly, the crucial question is whether below-ground microbial community simply follows the introduced tree and shrub species, or whether some kind of inoculation is required [see e.g. 108]. Among the relevant soil micro-organisms, arbuscular mycorrhizal fungi (AMF) and ectomycorrhizal fungi (ECMF) can be expected to play a major role during restoration of degraded sites. Many tree and shrub species associate with AMF and ECMF, which provide nutrients in exchange for plant carbohydrates. Recent evidence has shown that at least ECMF are dispersal limited, and are less abundant on isolated trees 109. This finding may urge for some kind of active inoculation of degraded restoration sites. How to successfully apply soil microorganisms in particular restoration projects, however, is an almost empty research field. Whereas fundamental insights in the role of AMF in structuring grassland communities are growing [but see 110], it remains largely unknown how these fungi contribute to successful restoration and the few available reports on the effects of large scale inoculations in grasslands did result in contradictory conclusions (White et al. 111vs. Smith et al. 112). Also, the inoculation of tree roots with mycorrhiza has received some attention in forest restoration projects, but the results are not straightforward [e.g., 113114]. This leads to the conclusion that at present much remains to be understood about how below-ground microbial diversity contributes to successful restoration of forest functions. Newly available molecular tools to quantify microbial diversity combined with detailed measurements of forest functioning are likely to increase our insights in how to apply below-ground biodiversity for restoration purposes.

While restored forests may deliver similar ecosystem services and conserve levels of biodiversity comparable to the pre-disturbance vegetation, restored forests rarely match the composition and structure of the original forest cover 115. Large changes in ecosystems will usually result in novel systems, comprising different species, interactions and functions 116117. In this context, it is important to realize that both the recent tendency towards accepting perennial, global change driven changes to the environment and the increasing application of the BEF framework to ecological restoration may facilitate the acceptance of using non-native species in forest restoration. While many ecologists still consider autochthony of species a prerequisite for their use in ecological restoration [see e.g. 118], a focus on species' functions rather than on species' origins is already advocated by others 119 as being a "more dynamic and pragmatic approach to the conservation and management of species". In this sense, the BEF approach may be at the source of a paradigm shift in restoration ecology 120.