Facilitation in plant communities: the past, the present, and the future

One important step to understanding the shifting balance between positive and negative interactions along environmental gradients is the design and implementation of experiments that move beyond the limitations of previous studies. These should, whenever possible, include at least one of the following aspects: (i) several co-occurring stress factors (e.g. temperature and water availability in arid areas and Mediterranean mountains), (ii) several levels within each stress factor, ideally accounting for a wide range in the variation of each factor, and (iii) an appropriate control of species-specific responses in competitive and facilitative ability and tolerance to stress (e.g. Tielbörger & Kadmon 1995; Callaway 1998; Choler et al. 2001; Gómez-Aparicio et al. 2004; Liancourt et al. 2005). Of particular interest in testing the SGH are experiments conducted in situations where benefactors may themselves be limited, such as at the extreme end of severity gradients. Experiments should also attempt to isolate the environmental factors affected by potential benefactors (to provide a mechanistic understanding of severity interaction relationships, e.g. Holmgren et al. (1997), Holmgren (2000)), evaluate different performance variables (e.g. survival, growth, physiological status and fitness), attempt to account for factors such as initial biomass effects (Gibson et al. 1999; Goldberg et al. 1999; Kikvidze et al. 2006a), provide a fine characterization of abiotic conditions, and be conducted over time periods long enough to cover different life stages in the studied species (e.g. Gómez-Aparicio et al. 2004; Miriti 2006; Schiffers & Tielbörger 2006). Taking into account these issues will be challenging and (comparatively) costly. However, we also recommend complementing them with observational studies conducted at the community level in multiple sites. When combined with appropriate statistical tools – such as structural equation models – these studies would allow us to assess multi-species interactions over broad geographical gradients, and to tease out the relative influence of different stress factors on them (Kikvidze et al. 2005).

In parallel with this rejuvenated research effort, we must also address the issue of the importance of facilitative interactions. A detectable process may not necessarily play a predominant role. Community composition is commonly seen as being regulated by filters, including chance biogeographical events, local abiotic conditions and interactions with other species (Grime 1998; Diaz et al. 1999; Lortie et al. 2004). The relative impact of each filter is not fixed, and there is a general assumption that filters operate to regulate distribution at different scales, e.g. chance biogeographical events and biotic interactions are considered most important for regulating distributions at the regional and local scales, respectively (but see Callaway & Ridenour 2004 and Valiente-Banuet et al. 2006 for examples of interactions determining ranges at a global scale). Central to these discussions, but frequently ignored, are the issues of the importance and intensity of interactions (Welden & Slauson 1986; Grace 1991; Corcket et al. 2003a; Brooker et al. 2005; Gaucherand et al. 2006; Grime 2007). Although the intensity of plant–plant interactions may remain unaltered along an environmental gradient, their impact relative to other processes (i.e. their importance) may vary. When and where, therefore, are facilitative interactions not only detectable but playing a key role, and does the importance of their role vary depending upon whether we are considering (for example) community composition or evolutionary processes?

The few studies to have considered the response of both the importance and the intensity of interactions demonstrate that they might not change in parallel along environmental gradients (Brooker et al. 2005; Gaucherand et al. 2006), and that a detectable effect on biomass need not necessarily translate into a fitness effect (Goldberg et al. 1999). Evaluating the effects of facilitation on the survival and fitness of individuals is therefore crucial if we want to understand when and where facilitation acts as a filtering process in plant communities. We predict that facilitation is likely to be important relative to other filters when climate stress (e.g. aridity) or disturbance (e.g. grazing) is high but not excessive so that well-fitted benefactor species (e.g. ‘nurses’) are able to grow and significantly ameliorate stress or provide protection from disturbances. Important facilitation would be indicated when these interactions enable a considerable number of beneficiary species to survive and reproduce when growing adjacent to the benefactor species (see Development of models incorporating plant facilitation for discussion of the related concept of a plant's interaction ‘kernel’). Low or extremely high environmental severity would prevent facilitation from having a significant role, either in regulating the composition of the community or as a selective force. Field studies need to test these simple predictions by considering explicitly the importance of interactions along environmental gradients (e.g. Gaucherand et al. 2006).

Interest in the effects of indirect interactions among species that occupy different trophic levels has been ongoing and has increased in recent years (Root 1973; Menge 1976; Hay 1986; Wooton 1994; Rousset & Lepart 2000; Corcket et al. 2003b; Seifan & Kadmon 2006). However, complex indirect interactions within a trophic level, or among competitors (Aarssen 1992; Miller 1994; Levine 1999; Callaway & Pennings 2000; Tielbörger & Kadmon 2000b; Callaway 2007) have received much less attention. The outcomes of pair-wise interactions therefore shape our thinking and thus community theory (Connolly et al. 2001), with pair-wise studies of competitive interactions leading to the perspective that competitive hierarchies in plant communities are transitive, or linear, with each species out-competing all those that are lower in the hierarchy. From this we might conclude that the only outcome of strong competition in communities at equilibrium is the exclusion of some or most members. However, this overlooks potential positive effects that occur in multi-species communities due to non-transitive ‘networks’ of interactions (Aarssen 1992). When interactions take place among many species at the same time, the competitive suppression of one species (B) by another (A) can decrease the latter's competitive effect on a third species (C) (Fig. 3a), thus leading to the phenomenon of indirect facilitation.

Few studies have considered that the occurrence of indirect facilitation may depend on complementarities in the interactions between species and therefore the identities or traits of the organisms involved (Huisman & Weissing 1999; Huisman et al. 2001). For example, in our model system (Fig. 3) indirect facilitation is most likely when pairs of plants (A–B, B–C) compete for different resources, for example light for the first species pair, and water or nutrients for the second (Levine 1999; Pagès & Michalet 2003; Siemann & Rogers 2003; Kunstler et al. 2006). Plant characteristics other than resource uptake ability (such as the production of allelopathic chemicals) may create such species-specific interactions, and thus non-transitive interaction networks and indirect interactions (Pagès et al. 2003; Callaway & Howard 2006; Kunstler et al. 2006; Callaway 2007). If the existence of multiple interaction mechanisms among coexisting species increases the potential for strong indirect interactions, indirect facilitation should be more common in communities where several limiting factors co-occur with similar strengths. In contrast, in environments with one dominant limiting factor, such as xeric, N-poor or low light conditions, indirect facilitation should be less important.

Surprisingly few studies have analysed how gradients of diversity will influence, and be influenced by, indirect facilitative interactions. Indirect facilitation should be more common in species-rich communities (Miller 1994; Dodds 1997). Indeed, it may have the potential to sustain the coexistence of high species diversity by reducing the potential for competitive exclusion (Czaran et al. 2002; Laird & Schamp 2006) and may be the dominant facilitative process in more productive environments (given the predicted reduction in direct facilitative effects). Indirect facilitation among competitors therefore challenges assumptions about competition consistently leading to exclusion.

There is therefore a notable dearth of studies exploring the role of indirect facilitative effects in plant communities, including their impacts on diversity and evolution. Future studies, using artificial manipulations of diversity or modelling approaches (see Development of models incorporating plant facilitation), should: (i) attempt to assess the conditions under which we might expect the greatest level of non-transitivity within plant communities; (ii) connect the degree of non-transitivity in plant competitive networks to community diversity and facilitative promotion of species coexistence; (iii) explore how the role of non-transitivity and indirect facilitation varies along gradients both of diversity and of system productivity, and in relation to direct facilitative effects; and (iv) explore, through multi-species evolutionary modelling, the potential for indirect facilitative effects to drive or limit niche differentiation.

Although a broad suite of modelling approaches is available, and despite their likely relevance and utility, they have not yet been widely employed in furthering our understanding of the role of facilitative plant interactions. This represents a significant missed opportunity for promoting the development of theory in this field.

Recent work illustrates the potential for relatively simple models to provide generic insights into the dynamics of communities structured with a mixture of positive and negative interactions. For example, Yamamura et al. (2004) used a spatially explicit patch occupancy model to study the evolution of mutualisms, highlighting the importance of local spatial structure in determining the outcome of interactions. Their modelling approach represented space as a regular grid, each cell of which can be in one of two states, occupied or unoccupied. Rules determine the way in which neighbouring plants interact, reproduce, disperse their propagules, evolve and die. Yamamura et al. highlighted the importance of local spatial structure in determining the outcome of interactions among a mutualistic and a cheater species. Although having great potential, spatially explicit simulation models of this type have only recently been applied to facilitation.

Travis et al. (2005) extended the model of Yamamura et al. to consider plant facilitation specifically, demonstrating that positive interactions are most likely towards the harsh end of environmental gradients (although this depends on whether the gradient primarily influences reproduction or survival; Travis et al. 2006). Such models can therefore readily provide hypotheses suitable for testing with field experiments. In addition they can also be used to run experiments. Brooker et al. (2007) used the spatial model of Travis et al. (2005, 2006) to explore the impact of interactions on the capacity of species to track a moving ‘climate envelope’. They found that the spatial arrangement of species with particular traits (as determined initially by interactions), along with the rate of climate change and level of long-distance dispersal, interacted to regulate the capacity of species to track their climate envelope.

Such developments in facilitation modelling are at an early stage. Although these spatially explicit models are a move in the right direction, they do not yet properly represent facilitation as most empirical plant ecologists have described it. In the mutualism models adapted for facilitation research the interaction term is commonly +/+ (Odum 1968), i.e. reciprocal benefit. However, although mutualistic interactions between neighbouring plants are possible, and may be more common than currently expected (Pugnaire et al. 1996), the impact of the beneficiary on the benefactor may range from positive to zero (+/0 commensalism), and some interactions might even be classified as parasitic (+/–). Future modelling should therefore start to distinguish clearly between mutualism and facilitation. Such a distinction would, for example, be important in determining the outcome of evolutionary modelling studies – facilitative interactions may have very different evolutionary impacts from mutualistic interactions, and may also evolve under different circumstances.

There is also a need for modelling approaches that accommodate variation in the role of facilitative interactions along environmental gradients, both in space and time, and at a local as well as landscape scale. If the spatial extents of positive and negative effects differ, the net outcome of interactions will be highly dependent on the spatial relation of the individual plants. Appropriate models would include the key elements of neighbourhood models (currently used to examine competitive plant interactions), i.e. individuals with an explicit location and basal extension where no other plant can exist, and a zone of competitive influence in which negative interactions occur (Stoll & Weiner 2000). Facilitation could be included through a zone of facilitation, although such zones of influence typically assume that all neighbours within a certain distance of a focal plant experience the same negative impact from competition. This is clearly unrealistic. More sophisticated approaches using interaction kernels, which describe the relationship between the distance of a neighbour from a focal individual and the competitive impact of that individual, have recently been developed (e.g. Murrell & Law 2003). However, few, if any, experiments in the literature provide the parameters necessary for such models.

Models thus extended to incorporate facilitation have the potential for a wide range of applications. For example, by allowing evolution of competition and facilitation kernels we may systematically test under which circumstances (frequency of interactions, gene flow, etc.) facilitation might evolve as an evolutionary stable strategy. Modelling could also explore the potential for different life-history characteristics (such as dispersal, dormancy, time to first reproduction, self-compatibility) to evolve in response to the balance between facilitation and competition present within a local community. Furthermore, the need to investigate interactions in multi-species assemblages is increasingly being recognized (e.g. Weigelt et al. 2007) and two-species models (Travis et al. 2006; Brooker et al. 2007) could readily be extended to multi-species models. This would enable, for example, the dynamics of indirect interactions to be examined in detail, including predicting the conditions under which this type of effect is likely to be observed in natural systems, and how its community and evolutionary impacts might compare with those of direct facilitative effects.

Recent studies indicate that facilitation may act as an evolutionary force: Scheffer & van Nes (2006) predicted that indirect facilitation may drive convergence of species’ niches, and Valiente-Banuet et al. (2006) suggested that facilitation is a source of stabilizing selection for the regeneration niches of ancient Tertiary species within Mediterranean-climate ecosystems. However, although the possible relationships between facilitation and evolution have been discussed previously (Hunter & Aarssen 1988; Brooker & Callaghan 1998; Callaway 2007), an evolutionary focus is absent from almost all recent work in the facilitation field.

We can readily visualize the ‘competitor’ syndrome that might evolve in response to strong and consistent negative interactions in productive environments: a high, dense canopy of wide-spreading leaves, rapid potential RGR (relative growth rate), high morphological plasticity, etc. (a classic C-strategy species; Grime 1977). However, it may not be straightforward to predict the evolutionary response to facilitative interactions, as both facilitative and competitive mechanisms may evolve in response to facilitation. For example, if species A is facilitated by species B, any negative effects of species A could select for those B individuals that are weaker facilitators (either by selecting against the key facilitation trait or in favour of an avoidance mechanism in species B). However, an alternative second selective force might favour individuals of species A with lower negative impacts on their benefactor.

One key trait that may come under such selection within the context of an assemblage comprising facilitators and facilitated is dispersal. Selection might favour dispersal kernels in the facilitated species that map onto those of the facilitator, for example through the convergence of dispersal agents. Conversely, selection of the facilitator might favour dispersal kernels enabling escape from the negative consequences of competition (unless the level of competition is itself reduced). Similarly, selection acting on characteristics such as the balance between reproduction by clones and seed may vary depending on the nature or balance of interactions – reproduction by seed may aid dispersal and be favoured by avoidance of competition, whereas clonal growth may be selected for by facilitation in the immediate vicinity of the parent plant. It is interesting to speculate that the reduced competitive effect or enhanced clonality of stress-tolerant species may in part result from selection in favour of enhanced facilitation. As stated by Brooker & Callaghan (1998) ‘It may be the case that we already have evidence of the evolutionary impact of positive plant–plant interactions, but have never examined it in the light of this possible interpretation’.

The outcome of the selective process will depend not only upon the frequency and relative strength of the interactions, but also upon evolutionary constraints on specific traits – is evolution of an avoidance mechanism more likely than reduced competitive impact? The evolution of life-history traits will play an important role in determining the spatial dynamics of an assemblage, resulting in a continuous interaction between evolutionary dynamics related to the interactions and the spatial ecology of the system. The extension of relatively simple evolutionary models (e.g. Travis & Dytham 1999; Travis 2003), similar to the individual-based models discussed above, could explore the interplay between ecological and evolutionary dynamics that might result from these types of effects. However, given the potential complexity involved, including variability in the two-way nature of facilitation, a simple starting point for evolutionary studies might be to ask whether all facilitative interactions provide common selective forces, from this predicting patterns of traits, and then to explore whether traits vary in the expected manner along environmental gradients in association with quantification of the importance and intensity of interactions. With current rapid technological advances in genetic methodologies capable of looking at quantitative traits (e.g. Barton & Keightley 2002), we should ultimately aim to develop evolutionary models that explicitly incorporate the genetic architectures involved and are able to separate out the heritable from the environmental components of the traits concerned. An improved understanding of their genetic basis would also permit robust estimation of the strength and direction of selection on the suites of traits involved, and would provide insights into correlations and potential pleiotropic effects between these traits.

In the face of current rapid degradation of terrestrial ecosystems worldwide, there is an increasing need for the development of novel, low-cost and efficient restoration techniques for maintaining ecosystem function and services (e.g. Ormerod et al. 2003; Hobbs et al. 2006). Because facilitation has been recognized as an important structuring force in natural plant communities, it is being increasingly discussed as an ecological mechanism which could be exploited for developing vegetation restoration tools, particularly for severe and highly disturbed environments (see reviews by Young et al. 2005; Padilla & Pugnaire 2006; Halpern et al. 2007).

Yet, except for arid environments, the role of positive plant interactions in terrestrial restoration is almost always overlooked. This is despite studies reporting strong facilitative effects during restoration in high mountain environments (Walker & Powell 1999; Aerts et al. 2007), tropical forests (Parrotta et al. 1997) or highly disturbed sites such as mine spoils (Densmore 2005; Frérot et al. 2006), and a range of examples from aquatic and intertidal ecosystems (reviewed by Bruno et al. 2003; Halpern et al. 2007).

Studies examining plant facilitation in terrestrial ecosystem restoration usually focus on the positive effects of nurse plants. Water-limited ecosystems offer one of the best opportunities for exploring these effects, as well illustrated by the study of Gómez-Aparicio et al. (2004). They investigated the use of naturally occurring shrubs as nurse plants for reforestation in a Mediterranean environment, and found that shrubs had a consistent beneficial effect on tree seedling survival and growth during four consecutive years. Such results directly contradict traditional reforestation management practice, where shrubs are removed prior to tree planting due to their presumed competitive effects on tree seedlings (see also Castro et al. 2004). Unfortunately, most studies addressing the use of nurse plants for restoration in arid systems have usually been conducted over shorter time frames, i.e. fewer than three growing seasons (e.g. Maestre et al. 2001, 2003b, 2004; Barchuk et al. 2005). Given the temporal variability in facilitative interactions in these systems (Tielbörger & Kadmon 2000a; Miriti 2006), this may be an insufficient period over which to assess whether nurse plants have a net beneficial effect, and longer-term studies are clearly needed. Furthermore, these studies commonly plant beneficiary species under existing nurse plants. However, in extremely degraded ecosystems nurse plants may themselves be lost (Gibson & Brown 1991). In such cases management for nurse plant re-establishment may be necessary as a first step, although it is also important to consider whether the restoration of nurse plant cover would be the most effective way of driving beneficiary re-establishment.

Novel techniques currently being developed in agro-ecosystems and polluted areas demonstrate the wide range of possible uses of facilitative interactions for environmental management. Facilitation can operate via increased pollinator visits, which lead to greater crop yields (Ricketts et al. 2004), enhanced water status of crops growing with species capable of hydraulic lift (Pate & Dawson 1999; Sekiya & Yano 2004), the transfer of symbiotically fixed nitrogen from legume to non-legume species (Jensen 1996; see also Hauggaard-Nielsen & Jensen 2005 for a review of facilitative root interactions in agro-ecosystems), or phytostabilization by metallicolous nurse plants in heavy-metal-polluted environments (Frérot et al. 2006). Transgressive over-yielding has been reported for multi-species mixtures in intensively managed grassland systems, although the precise mechanism is unclear (Kirwan et al. 2007). There may therefore be a wide range of facilitative mechanisms, in addition to the classic nurse plant effect, that could act as the basis for restoration tools.

As strikingly illustrated by the work of Gómez-Aparicio et al. (2004) discussed above, Sutherland et al. (2004) proposed that ‘Much of current conservation practice is based upon anecdote and myth’. The acknowledgement of facilitation in studies of restoration, particularly within environments that are difficult to restore, would help to raise facilitation's profile as a tool for natural resource conservation and management. Furthermore, applied facilitation research also has significant unexploited potential for advancing the science of restoration ecology in general. The explicit consideration of facilitation when exploring key topics in this research area, such as stable alternative states (Suding et al. 2004), succession (Walker et al. 2007) or the links between ecosystem structure/function and restoration success (Young et al. 2005; Cortina et al. 2006), are likely further to help managers and practitioners to develop effective tools to achieve the desired restoration targets.

Climate change, nitrogen deposition, biological invasions and land-use change have been suggested as the current major threats to global biodiversity (Sala et al. 2000; CBD 2003), and significant evidence is accumulating that interactions, including facilitative ones, play a role in mediating the impact on natural communities of these environmental change drivers (Brooker 2006; Maestre & Reynolds 2006, 2007). However, the exact nature of that role, and how it will alter in response to environmental change, remains unclear.

Several studies have demonstrated the importance of facilitation for the maintenance of biodiversity in arctic and alpine habitats (e.g. Choler et al. 2001; Callaway et al. 2002), and these habitats are predicted to be amongst the most sensitive to global warming (Guisan & Theurillat 2000; ACIA 2004). Based on the original SGH, some authors have predicted a general shift in species interactions from more strongly positive to more strongly negative as these environments warm (e.g. Klanderud & Totland 2005; Klanderud 2005). In support, experiments that evaluated the impact of enhanced nitrogen availability in these ecosystems (simulating the expected effect of climate change on nutrient mineralization rates) mostly reported increased competitive interactions (e.g. Chapin et al. 1995; Bret-Harte et al. 2004; Klanderud & Totland 2005). Similarly, in a North American arctic tundra system, where artificially advanced snowmelt (a phenomenon expected under global warming) created a more severe environment through increased early-season frost events, neighbours increasingly facilitated the survival, growth and reproduction of Empetrum nigrum (Wipf et al. 2006). However, Shevtsova et al. (1997) reported that the positive effects of elevated temperature on the growth and reproduction of Vaccinium vitis-idaea and Empetrum nigrum in subarctic Finland were amplified by the presence of neighbours – in this instance a reduced level of environmental severity (warming) led to an increased impact of facilitative interactions.

These findings support the argument that the relationship between gradients of environmental severity and the relative role of interactions is more complicated than set out in the original SGH (e.g. Bertness & Ewanchuk 2002). At first glance we might conclude that such a pattern supports the recently proposed humped-back relationship (Michalet et al. 2006). However, it is notable that the increased role of facilitation with decreasing severity found by Shevtsova et al. (1997) occurs within an environment that is apparently less severe than, for example, the open tundra systems studied by Chapin et al. (1995) and Wipf et al. (2006), and in which the opposite pattern was observed. Such apparently contradictory findings indicate that the perception of environmental severity is to some extent species-specific, and that the response of interactions at the individual species level may not reflect the trend of average interactions within a community. Improving our knowledge of the variation in interactions under different environmental conditions between different species, growth-forms or strategies (e.g. competitive vs. stress-tolerant strategies), and how these differ from trends at the community level, will help in the development of general theory and in explaining such apparently contradictory results.

Invasive species are another critical global change driver. In the study of biological invasions, interactions are considered crucial in determining the success of exotics (e.g. Daehler 2003; Simberloff et al. 2003; Vilàet al. 2004). Although most studies have focused on negative interactions as the main drivers of invasive success (e.g. Callaway & Aschehoug 2000; Colautti et al. 2004; Mitchell et al. 2006), a sizable number have also reported that facilitation by other exotics (Simberloff & Von Holle 1999) or by natives (Maron & Connors 1996; Richardson et al. 2000; Lenz & Facelli 2003; Cavieres et al. 2005, 2007) can promote exotic invasion. For instance, Maron & Connors (1996) reported that the nitrogen-fixing native shrub Lupinus arboreus facilitates the establishment of exotic species by ameliorating soil nutrient shortage. Cavieres et al. (2005) found that the cushion plant Azorella monantha facilitates the establishment of the exotic species Taraxacum officinale in the high-alpine zone of the Andes of central Chile by providing microsites with milder microclimatic conditions (see also Cavieres et al. 2007). Reinhart et al. (2006) showed how seedlings of the invasive tree Acer platanoides had higher densities, recruitment and survival, and less photoinhibition and water stress when beneath conspecific canopies than when growing under adjacent native Pseudotsuga menziesii trees; they related these differences to the environmental modification created by the invaders. When exotic species facilitate the establishment of other exotics, accelerated invasion can occur. Simberloff & Von Holle (1999) coined the term ‘invasional meltdown’ to describe such a process. Invasional meltdown demonstrates that facilitation can be an important driver of plant invasion (Simberloff 2006) but, interestingly, invasion can also be ‘resisted’ by greater diversity of native species (e.g. Zavaleta & Hulvey 2004; Fargione & Tilman 2005), a mechanism by which facilitation of natives may help limit the influx of invasives.

Improving our understanding of facilitative interactions is therefore of direct relevance to understanding the impacts of environmental change drivers on biodiversity (Callaway 2007). Resolving the debate concerning the SGH (difficult though this may be) may, for example, enable us more accurately to predict changes in the role of interactions in response to environmental change, or the conditions under which interactions might restrict or enhance biodiversity change, for example through the influx of invasives. However, it is not necessarily the case that radical new experiments are needed to explore these issues. As with restoration ecology, simply recognizing the potential role of facilitation, and including it within both experimental designs and the interpretation of ecosystem responses, could provide us with valuable insights into facilitation both as a mediator of global change and as a fundamental ecological process.