What does species richness tell us about functional trait diversity? Predictions and evidence for responses of species and functional trait diversity to land-use change

Authors


M. M. Mayfield, The University of Queensland, School of Biological Sciences, Brisbane, 4072, Queensland, Australia.
E-mail: m.mayfield@uq.edu.au

ABSTRACT

In the conservation literature on land-use change, it is often assumed that land-use intensification drives species loss, driving a loss of functional trait diversity and ecosystem function. Modern research, however, does not support this cascade of loss for all natural systems. In this paper we explore the errors in this assumption and present a conceptual model taking a more mechanistic approach to the species–functional trait association in a context of land-use change. We provide empirical support for our model's predictions demonstrating that the association of species and functional trait diversity follows various trajectories in response to land-use change. The central premise of our model is that land-use change impacts upon processes of community assembly, not species per se. From the model, it is clear that community context (i.e. type of disturbance, species pool size) will affect the response trajectory of the relationship between species and functional trait diversity in communities undergoing land-use change. The maintenance of ecosystem function and of species diversity in the face of increasing land-use change are complementary goals. The use of a more ecologically realistic model of responses of species and functional traits will improve our ability to make wise management decisions to achieve both aims in specific at-risk systems.

INTRODUCTION

Concern is increasing about the loss of biodiversity under land-use change and how this will affect the ecosystem functions and services provided by natural communities (Chapin et al., 2000; Millennium Ecosystem Assessment 2005; Díaz et al., 2007b). The range of functions provided by a community is thought to largely depend on the diversity of functional trait states or values of key traits (referred to as functional trait diversity in this paper) and the diversity of species that express them (Díaz & Cabido, 2001; Prinzing et al., 2008). The relationship between species diversity and ecosystem function is known to be complex and context dependent (Naeem & Wright, 2003), yet it is often assumed in the conservation literature that land-use intensification simply causes species losses, resulting in a similar loss of functional trait diversity (e.g. Bowker et al., 2008; Christie & Hochuli, 2008; Milder et al., 2008). Such an assumption has motivated the use of biodiversity patterns as surrogates of ecosystem function, particularly when making conservation decisions (Srivastava & Vellend, 2005; Bowker et al., 2008; Smyth et al., 2009). However, this assumption exceeds empirical and theoretical support (Schwartz et al., 2000; Srivastava & Vellend, 2005).

Here, we draw upon ecological theory to detail changes in the diversity of species and functional traits (ΔDSF) that we should expect under land-use alteration in natural plant communities. Using three case studies, we provide evidence that the proposed ΔDSF response trajectories do occur in real ecosystems under anthropogenic land-use change. Finally, we recommend research directions aimed at developing a predictive framework for identifying systems where a loss of species and/or functional diversity is or is not expected following land-use change.

Current understanding of the DSF relationship

Previously proposed relationships between species and functional trait diversity (DSF) range from positive linear relationships with a range of slopes (Díaz & Cabido, 2001) to a variety of monotonically increasing curves (Naeem & Wright, 2003; Mayfield et al., 2005). These models all support the idea that if species diversity is lost due to human land-use change, functional trait diversity will be lost as well (Bowker et al., 2008; Christie & Hochuli, 2008; Milder et al., 2008). Flynn et al. (2009) recently examined ΔDSF along gradients of land-use change in natural communities and found no decrease in either the species or functional diversity of plants. Though an important step forward, the study by Flynn et al. (2009) examined multi-trait diversity indices (i.e. indices that incorporate multiple traits and their values, in this case Petchy & Gaston's, 2006, FD index) rather than the diversity of trait states or values (the number and type of trait states for a single trait). Most theoretical work on the DSF relationship has been based on the diversity of trait values for individual traits of potential importance to specific ecosystem functions (e.g. Díaz & Cabido, 2001; Naeem & Wright, 2003). Empirical studies of ΔDSF under land-use change based on functional trait values for individual traits are quite rare (Mayfield et al., 2005; Micheli & Halpern, 2005) and generally not based on a relevant theoretical framework.

Toward a model of ΔDSF in response to land-use change

There is a long history of using ecological processes to explain species diversity in communities without anthropogenic disturbances (e.g. Grinnell, 1917; Elton, 1946; Grime, 1973). Ecological assembly processes such as environmental filtering, competitive exclusion and facilitation as well as the size of the regional species pool are all important for determining the diversity of species and functional trait diversity in natural communities (Keddy, 1992; Díaz et al., 1998; Milbauer & Leach, 2007; Box 1). It is only recently that studies of human-induced land-use change have cited any of these processes (mostly environmental filtering) as important for the assembly of novel communities following human-induced land-use change (Roy et al., 1999; Chazdon et al., 2003; Temperton et al., 2004; Hobbs et al., 2006; Lososováet al., 2006; Mayfield et al., 2006; Funk et al., 2008; Knapp et al., 2008; Ricotta et al., 2008). Studies on the conservation value of communities affected by human activities more commonly invoke processes such as the sampling effect hypothesis (SEH) to motivate conservation (Srivastava & Vellend, 2005). The SEH is an explanation for the positive DSF relationships observed in experimental communities of differing species richness (Aarssen, 1997; Huston, 1997); it posits that trait states (or values) will accumulate with successive additive random draws from a regional species pool (Tilman et al., 1997). This theory is often assumed to operate in natural communities (as found by Schwartz et al., 2000), though few studies have specifically tested this hypothesis in non-experimental systems (e.g. Palmer et al., 2000; Field et al., 2008).

Box 1 Key concepts in community assembly

  • 1Species pool: The species pool concept is central to understanding how communities assemble (Zobel, 1997). The assembly of a community following land-use change will depend both on the species that survive the disturbance and the species that colonize the community – the latter species determined by the species pool. Local species diversity is thought to be positively related to the size of the species pool (e.g. Aarssen & Schamp, 2002). Land-use intensification can change the species pool through regional extinctions and dispersal limitation, although this is unlikely to be a major source of change unless very large regions are altered at the same time.
  • 2Environmental filters: Environmental filters are non-randomly distributed abiotic conditions and resources that exclude species with unviable physiological limits (defined by functional traits) from entering or persisting in a community. The presence of environmental filters means that even in the absence of other assembly processes, a community will be a subset of the total species in the species pool (Keddy, 1992; Díaz et al., 2007a; Cornwell & Ackerly, 2009). Land-use intensification can alter the strength or magnitude of these filters. For example, logging may relax light restrictions (decrease the importance of a filter) or increase the ambient temperate (increasing the strength or importance of a filter).
  • 3Competitive exclusion: Competitive exclusion structures many communities (Grime, 1973; Paine, 1974). Competitive interactions are complex and land-use change can greatly alter the role that these processes play in community assembly. Land-use changes such as selective logging may result in a competitive release of certain functional types, or an increase in competition among species for certain limiting resources (which themselves may change following disturbance).
  • 4Functional redundancy: The presence or addition of species to a community possessing the same functional traits, or of the same functional type as a species already residing in the community, does not necessarily add to the functional richness of the community; rather, it defines the community's functional redundancy. Functional redundancy is an important characteristic of community's resilience to environmental change (McCann, 2000). Importantly, functional redundancy is an illustration of how functional trait diversity and species diversity can be decoupled.

By considering ecological assembly processes in the study of human-altered systems, it becomes clear that the concomitant loss of species and functional trait diversity is just one of a range of potential responses to land-use change (Naeem & Wright, 2003). We suggest that this negative ΔDSF vector (consistent with SEH) is only one of a set of vectors observed in natural communities. It should not, however, be the default assumption for how land-use change influences this important relationship. This is because land-use change affects the processes of community assembly, not species diversity per se (Fig. 1). When land-use change occurs, assembly processes such as environmental filtering, competition and facilitation also change (Box 1, Fig. 1). This can be due to new or amplified competitive interactions, the release from competitive or facilitative dynamics, and shifts in abiotic conditions and resources (Mayfield et al., 2005, Fig. 1). Importantly, given all the evolutionary and ecological sorting processes that may influence community-wide processes (Prinzing et al., 2008; Cavender-Bares et al., 2009), changes in species diversity should be considered an outcome of ecological assembly processes, not a causal driver of functional trait diversity nor of ecosystem function (Fig. 1). Species diversity may be correlated with functional diversity and ecosystem function in some cases (dashed line, Fig. 1). Such a correlation is expected in cases when ΔDSF trajectories are consistent with the sampling effect (SEH).

Figure 1.

Proposed relationship between land-use change, species richness and ecosystem function. The model applies for a particular species pool and for an individual functional trait. Arrows imply the direction of cause and effect. Viable functional trait states are those trait states that are compatible with the environment of the community, not all of which may be represented in the community or necessarily represented in the species pool. Only represented trait states contribute to the observed diversity of functional trait states. The dashed line indicates a possible correlation, consistent with the sampling effect hypothesis.

Expected ΔDSF vectors

According to our model (Fig. 1), land-use intensification may result in various response trajectories of ΔDSF, depending on how changes in land use affect the assembly processes acting on the communities in question. Here we describe the possible ΔDSF response trajectories with reference to Fig. 2. We also provide ecologically sound explanations for each trajectory. These explanations should not be considered as an exhaustive list of mechanisms but rather as a list of feasible examples that should be directly tested in the future.

Figure 2.

Feasible ΔDSF vectors (changes in the diversity of species and functional traits) and possible underlying mechanisms. The filled circles indicate the starting DSF in undisturbed communities. Open circles indicate the relative position of DSF following extensive modification of the natural community, consistent with human activities such as logging or grazing. Positions of circles should be considered relative to each other and not as exact locations on axes. SEH refers to ‘the sampling effect hypothesis’.

In communities with high functional redundancy (Box 1) species richness may decline due to an increased importance of environmental filters (Box 1, Fig. 2), as might occur with increasing ambient temperatures following logging. If species losses are evenly (or randomly) distributed across functional trait groups, then no corresponding loss of functional trait diversity is expected (Fig. 2, vector A). This vector/pattern is unlikely if land-use change disproportionately removes species of certain functional groups, as is predicted for many systems (Flynn et al., 2009).

Land-use change may also lead to an increase in species diversity without a corresponding increase in functional trait diversity in communities with high functional redundancy (Fig. 2, vector C). We expect this relationship when land-use change relaxes competitive interactions, allowing more species from the regional species pool to coexist. We believe this relationship is particularly likely when all functional types from the species pool are represented in the community and none are excluded by environmental filters. Large species pools (such as those found in the tropics) may facilitate this pattern.

A loss of functional trait diversity while species diversity remains constant (Fig. 2, vector D) could occur if land-use change is followed by an immediate loss of species through extinction filters (the initial death of all individuals from particular species in response to a disturbance or to the environment created immediately following that disturbance) followed by an influx of functionally similar species. For example, pastoral areas that were originally forested may continue to support high species diversity accompanied by a replacement of diverse growth forms with numerous weedy species with the same herbaceous growth form (McIntyre & Lavorel, 1994). It is also possible for functional trait diversity to increase without a change in species diversity following land-use change (Fig. 2, vector B). This pattern is expected when species with novel functional trait values (native or exotic) replace functionally redundant species within the community.

The expected correlated shifts in species and functional trait diversity are also possible under changing land use (Fig. 2, vectors F and H). The most commonly assumed trajectory of ΔDSF is a correlated decrease, consistent with SEH (Fig. 2, vector H; Lawton & Brown, 1993; Hector et al., 2002). It is important to note, however, that this relationship could also occur from a non-random loss of species with particular trait values or states, as has been shown for some systems (Flynn et al., 2009).

The opposite pattern, an increase in both species and functional trait diversity (Fig. 2, vector F), might occur when land-use change releases environmentally compatible species found in a large species pool from strong or dominant competitors. For example, selective logging of a common species might release species from a dominant competitor (see Australia results below). The addition of species and trait states following such a disturbance is consistent with SEH and the intermediate disturbance hypothesis (Grime, 1973).

The final two possible ΔDSF patterns (Fig. 2, vectors E, G) are consistent with modifications to environmental filters (Box 1). Habitat alteration frequently changes the specifics of environmental filters. Land-use changes may shift filters from favouring one set of trait states to another (deforestation may change filters from favouring shade-tolerant species to gap species or species requiring high light, e.g. Suding & Goldberg, 2001). Alternatively, filters may become less or more restrictive. For example, when woodlands are converted to agriculture, the nutrient and water retention capacities of soils are often reduced, which generally restricts the set of species that can tolerate these areas (e.g. Cramer et al., 2008). We predict that the removal of dominant competitors under land-use change will often allow the entry of numerous species into the community as competition for space, light and nutrients is likely to be reduced. The impacts of these removals, however, may increase the importance of environmental filters in directing community composition, resulting in functionally similar species joining communities (Fig. 2, vector G).

Reduced species diversity with an increase in functional trait diversity (Fig. 2, vector E) seems relatively unlikely. This pattern is most likely to be observed when major species losses under land-use change are mitigated by the colonization of some species with diverse functional types. This might occur, for example, when a uniformly tall and species-rich forest is removed and replaced by a relatively species-poor community expressing a wide range of heights and growth forms.

CASE STUDIES

To demonstrate that the proposed trajectories of ΔDSF can occur in real systems, we present results from three forest systems. The forest systems we examine are semi-arid woodland in south-eastern Australia, temperate forest in Quebec, Canada, and tropical rainforest in Costa Rica. Methods of data collection for each forest system are provided in Appendix S1 in the Supporting Information and in the original studies (Mayfield & Daily, 2005; Thompson & Eldridge, 2005; Mayfield et al., 2006; Aubin et al., 2007, 2009). Results should be taken as an illustration that the predicted patterns depicted in Fig. 2 do occur in natural systems in response to increased land-use intensity, not as the only patterns likely to be seen in similar systems or for different traits.

For each forest system, we examined the response trajectory of ΔDSF from relatively undisturbed forest sites to those that assembled following human-induced land-use change. All three studies included a survey of plant species richness in forests largely undisturbed by humans (no logging, clearing or planting) and plant communities within the same region that formed following logging and subsequent land uses. For the Australian study, the forests examined were dominated by Callitris glaucophylla which grow in stands of up to 10,000 stems per hectare when undisturbed and up to 36,000 stems per hectare when disturbed. We compared 63 sites classified as never having been logged (termed ‘unmodified forest’) with 20 sites known to be extensively logged c. 50 years ago and subsequently allowed to regrow (termed ‘modified forest’). For the Quebec (Canada) study, we examined a landscape dominated by deciduous maple forests and agricultural lands. For this dataset, we compared six unmanaged/undisturbed forest sites (with no known history of harvesting; ‘unmodified forest’) and six sites in areas that had been converted to pasture more than 100 years ago (‘modified forest’). In Costa Rica, data were collected on the Osa Peninsula across a mosaic landscape dominated by lowland tropical rainforest and road verge vegetation maintained by light grazing and cutting. We compared tropical forest understorey (non-tree component of the forest) that had not been logged within at least the last 100 years (‘unmodified forest’) and 12 road verge sites that had been converted from rain forest approximately 60 years ago (‘modified forest’). It should be noted that in Quebec and Costa Rica, forest had not been allowed to regrow in ‘modified sites’. Because the land-use change examined in all three studies occurred at least 50 years ago, we can be relatively certain that the observed ‘logged’ communities result from assembly processes, rather than initial extinction filters (Appendix S1).

All three datasets also had data on three or four of the following functional traits: growth form, leaf area, height and dispersal mode. We selected these traits due to their accepted connections to ecosystem functions and community assembly (Weiher et al., 1998; Table 1). Functional trait diversity for height and leaf area (continuous traits) was calculated as the standard deviation of abundance-weighted, loge-transformed trait values. For categorical traits (dispersal mode and growth form), diversity was calculated as a Shannon index (Magurran, 1988) of the number of species with each trait state (see Table 1). Our metric of species diversity was species richness (see Appendix S1). ΔDSF was measured as the percentage change between mean unmodified and modified sites and by examining the overlap of 95% confidence intervals along both x- and y-axes in Fig. 3 (Cumming & Finch, 2005).

Table 1.  Traits examined in case studies (Trait), their roles in species persistence (Species function), the ecosystem functions to which they have been linked (Ecosystem function) and the assembly process or feature (Assembly factor) through which they may influence community assembly (Assembly factor).
TraitSpecies functionEcosystem functionAssembly factor
  • *

    Growth form categories that were included in analyses of all three case studies were: climber, erect leafy, long basal, short basal, short tree, shrub and tussock; tall tree and semi-basal were additional categories for the Quebec and Australia studies; dwarf shrub was an additional category in Australia; and canopy climber, epiphyte, hemi-epiphyte, palmoid, short succulent and tall succulent were additional categories for the Costa Rica study.

  • Dispersal mechanism categories included in analyses of all three case studies were: exozoochory, wind, passive, insect and hoarding or caching; for Quebec and Costa Rica, endozoochory was an additional category; and water was an additional category for the Costa Rica study.

Growth form*Plant longevity, competitionResistance to disturbanceCompetitive exclusion
Leaf areaResource allocation, water uptake strategy, stress toleranceProductivityEnvironmental filtering
HeightCompetitionResistance to disturbanceCompetitive exclusion
Dispersal modeDispersalStabilitySpecies pool
Figure 3.

ΔDSF (changes in the diversity of species and functional traits) patterns observed for three forest systems, considering four functional traits. Arrows in each plot illustrate the vector of ΔDSF from unmodified forest to human-modified forest within the same region. Human modification in all sites started with logging, but in Quebec and Costa Rica human-modified plots have been further maintained as non-forest land uses (pasture and road verge shrubby communities). Open symbols indicate the species richness and trait state diversity for individual sampled plots, circles for unmodified forest plots and triangles for human-modified plots. Closed symbols indicate the mean species diversity and trait state diversity for unmodified forest sites (circles) and human-modified sites (triangles). Error bars are 95% confidence intervals.

Case study results and discussion

Results from our three case studies illustrate that a variety of ΔDSF vectors are observed under land-use change. In the Australian forests, we observed almost no ΔDSF from unmodified forests to those with a logging history. For all traits in that dataset, we saw a subtle increase in species richness and functional trait diversity (c. 5–10%; Fig. 2, vector F, & Fig. 3). These patterns are consistent with the interpretation that environmental filters in these forests are not altered substantially by logging, while competition is lessened to a small extent by the removal of the dominant tree species, Callitris glaucophylla (see Appendix S1 for details of the system; Fig. 2, vector F). These results do demonstrate that not all species from the species pool are represented in these forest communities (Fig. 2, vector F). Importantly, trees were allowed to re-establish immediately following logging (50 years of development), and thus the observed pattern might also reflect a recovery towards original conditions over time. Significant changes along either axis may have been more evident soon after logging.

It is only in the temperate forests of Quebec that we see the correlated loss of species and functional trait diversity between unmodified and modified sites for any trait (Fig. 2, vector H, & Fig. 3). This pattern may result from the SEH or biased losses of specific functional groups. For height and dispersal mechanism, the loss in functional trait diversity was greater than in species richness (80–90% and c. 50%, respectively), leading to a relationship similar to vector D (Fig. 2). In these sites, functionally diverse forest species were replaced by functionally similar old-field species, a finding consistent with a repetitive strong environmental filter (grazing).

For each of the three traits examined for the Costa Rica dataset, we found that species richness doubled with land-use change, with varying responses in trait diversity (Fig. 3). For growth form, we see an approximately 40% decrease in trait state richness (Fig. 2, vector G). Trees were not permitted to re-establish on these sites (Appendix S1), and their permanent removal as competitive dominants may have released non-tree species from strong resource competition, and may also have increased the importance of environmental filters related to high light, heat and wind conditions. For dispersal mechanism, we see a minor c. 12% decrease in trait state richness (Fig. 2, vector C or G). The slight decline in dispersal mechanisms may indicate that certain environmental filters have been reduced in importance, but not sufficiently as to eliminate the compatibility of certain dispersal modes in deforested communities. It may also be possible that dispersal traits are loosely correlated with another trait under stronger influence of new filters (Fig. 2, vector G) or that no new dispersal mechanisms are available in the species pool to add to these communities.

CONCLUSIONS

The role of biodiversity in ecosystem function is considered one of the most important questions in conservation biology today. As human activities impact ever more intensively upon natural communities, ecological theory concerning this relationship is drawn on increasingly by studies aiming to conserve both biodiversity and ecosystem functions and services (Srivastava & Vellend, 2005). The maintenance of ecosystem function and species conservation are complementary aims for which we must understand the effects of land use. However, current theory does not support the argument that land use always leads to a loss of ecosystem function mediated through a loss of species and functional trait diversity.

The framework we have presented provides a basis for understanding how human activities probably affect ecological processes, and thus the diversity of natural communities. We have shown that species richness and functional trait diversity may follow numerous response trajectories after land-use change. The conceptual model we present in Fig. 1 provides a sound ecological explanation for all feasible patterns (Fig. 2) for natural communities experiencing land-use change, based on the growing appreciation that traits, not species, are at the centre of community assembly. Recent evidence suggests that SEH may be a common factor of importance to functional responses of plant communities to land-use change (Lalibertéet al., 2010). In such cases, we do expect a correlation between species and functional trait diversity (Fig. 2, F and H) and would find it acceptable to use species diversity as a proxy for functional diversity. Our study illustrates, however, that the assumption of SEH-compatible response trajectories, can be completely incorrect. Thus, rather than simply assuming a SEH-compatible DSF relationship for all conservation studies of human land-use change, scientists should carefully consider the details of their system and the most likely ΔDSF trajectories.

We hope that this paper spurs more focused research into the factors involved in determining which response trajectory a given community is likely to follow. In particular, we believe that the size of the local species pool, the productivity of the system, the type of disturbance and the traits examined are all likely to be key factors in determining the trajectory of ΔDSF. Information on the size of regional and global species pools is increasingly available, as is an understanding of how the strength of biotic interactions (competition and facilitation) changes across regional productivity gradients and latitude. We hope that increasing interest in the relationship between species and functional trait diversity, as well as the increasing availability of key data, will allow the development of a ΔDSF predictive framework that can be used to prioritize biodiversity and ecosystem function-based conservation.

ACKNOWLEDGEMENTS

This study results from Working Group 31 of the ARC-NZ Network for Vegetation Function. We thank network staff for logistical support and the network committee for supporting this project, which has been fun. We thank Wendy Thompson for providing data for the Australian woodlands and André Bouchard and Christian Messier for the Québec dataset. We would also like to thank Fabrice DeClerck, Dan Metcalfe, Etienne Laliberté, Jessie Wells, Yi Ding, Cibele Queiros and Carla Catterall for helpful discussions and feedback in the development of this paper. We also thank Shahid Naeem and two anonymous referees for their helpful comments.

BIOSKETCH

The work presented in this manuscript was completed as part of an ARC–NZ Research Network for Vegetation Function (http://www.vegfunction.net/) working group entitled: ‘Human-influenced country sides and plant traits.’ The overall aim of this group was to explore whether or not there are generalizations that can be made about how plant trait diversity responds to land-use intensification. The working group was lead by Dr Margaret Mayfield of the University of Queensland, who is also the lead author on this paper. She is a plant ecologist specializing on the assembly of plant communities following human-induced disturbance.

M.M.M., S.P.B., J.W.M. and P.A.V. originally formulated the ideas presented in this paper. M.M.M., S.P.B. and I.A. provided the data for the presented empirical studies. S.M. and P.A.V. were in charge of data analysis. M.M.M., S.P.B. and J.W.M. wrote the first draft of this manuscript and all authors contributed extensively to preparation of the final manuscript.

Editor: Navin Ramankutty

Ancillary