Thibaud Decaëns, Laboratoire d'Ecologie, EA 1293 ECODIV, Fédération de Recherche SCALE, Bâtiment IRESE A, UFR Sciences et Techniques, Université de Rouen, F-76821 Mont Saint Aignan Cedex, France. E-mail: firstname.lastname@example.org
Aim To review published evidence regarding the factors that influence the geographic variation in diversity of soil organisms at different spatial scales.
Methods A search of the relevant literature was conducted using the Web of Science and the author's personal scientific database as the major sources. Special attention was paid to include seminal studies, highly cited papers and/or studies highlighting novel results.
Results Despite their significant contribution to global biodiversity, our taxonomic knowledge of soil biota is still poor compared with that of most above-ground organisms. This is particularly evident for small-bodied taxa. Global patterns of soil biodiversity distribution have been poorly documented and are thought to differ significantly from what is reported above-ground. Based on existing data, it appears that microorganisms do not respond to large-scale environmental gradients in the same way as metazoans. Whereas soil microflora seem to be mainly represented by cosmopolitan species, soil animals respond to altitudinal, latitudinal or area gradients in the same way as described for above-ground organisms. At local scales, there is less evidence that local factors regulate above- and below-ground communities in the same way. Except for a few taxa, the humpbacked response to stress and disturbance gradients doesn't seem to apply underground. Soil communities thus appear weakly structured by competition, although competitive constraints may account for assembly rules within specific taxa. The main factor constraining local soil biodiversity is the compact and heterogeneous nature of soils, which provides unrivalled potential for niche partitioning, thus allowing high levels of local biodiversity. This heterogeneity is increased by the impact of ecosystem engineers that generate resource patchiness at a range of spatio-temporal scales.
Study of the distribution patterns of living organisms and of the factors that they are driven by is a central theme that has stimulated a profusion of scientific research during recent decades (Huston, 1994). The large number of paradigms and theories proposed to explain spatio-temporal patterns of biodiversity at different scales attests to this interest, while it also underlines the complexity of the ecological processes involved. Gaston & Spicer (1998) pointed out that although we remain far from an accurate mapping of the occurrence of life on Earth, many patterns of spatial variation in biodiversity have been documented, and they provide a ‘skeleton’ upon which a biodiversity atlas and concepts may be built. A large range of abiotic and biotic factors, from the effect of the physical environment to the multitude of biotic interactions that link living species with each other, ultimately affects biodiversity patterns. The strength and specificity of the mechanisms that control a given group of organisms may, however, vary to a large extent according to environmental conditions that operate at different scales of space and time (Bardgett et al., 2005b). This makes the description and interpretation of biodiversity patterns a very complex task. Other constraints to their study arise from the facts that: (1) we only know at best a tiny fraction of the existing species diversity, and (2) our knowledge of both the distribution and abundance of many known species remains very poor (Wilson, 2002).
In this context, soils probably represent one of the most complex and poorly studied habitats of terrestrial ecosystems. Biological communities inhabiting soils are among the more diversified, comprise a wide range of life-forms and functions, are involved in a large number of ecological processes and provide key ecosystem services for human populations (Lavelle et al., 2006). Paradoxically, below-ground taxa still suffer from a conspicuous deficit of interest from the scientific community, and as yet we know very little about their taxonomy, systematics and biogeography (Decaëns et al., 2006, 2008a). To date, only a few studies have reported global patterns in soil biodiversity distribution at different spatial and temporal scales (Wardle, 2002; Bardgett et al., 2005b). While these authors highlighted and interpreted some consistent patterns, they also pointed out the dearth of information that is available on the diversity of soil biota, especially at the species level. To date we are still unable to provide clear responses to some fundamental questions such as: How many species live in the soil? How does soil biodiversity respond to environmental gradients at specific scales of space and time? What ‘in-soil’ and ‘off-soil’ factors are driving soil biodiversity patterns and at what scales do they operate? Are biodiversity patterns the same above and below ground?
The aim of this review is to provide a synthesis of our knowledge with regard to soil biodiversity patterns. I mainly address the different questions listed above, and hope to give insights into the wide range of processes that control soil biodiversity at a range of spatial and temporal scales. This work is also motivated by the applied issue of the conservation and management of below-ground biodiversity, for which a clear understanding of the factors that drive soil communities represents a necessary prerequisite. After a brief survey of what we know about biodiversity levels in soil communities, I review the main patterns and determinants of this biodiversity from the global to the local scales. Finally, I discuss a number of hypotheses that have been proposed to explain these patterns and conclude with perspectives for future work on soil biodiversity.
MATERIALS AND METHODS
For each pattern described in this review, I looked for relevant literature in the ISI Web of Science (http://isiwebofknowledge.com/) and in my personal scientific literature database. When different papers dealing with the same patterns were available, I selected those papers that present seminal studies, papers with high numbers of citations in the Web of Science and/or recent publications highlighting new results. I also focused on selected papers that provide examples covering the widest possible taxonomic range (from microorganisms to soil fauna of different body sizes).
I assessed the representation of soil organisms in two traditional journals of taxonomy and systematics, the Zoological Journal of the Linnean Society and Journal of Zoological Systematics and Evolutionary Research, because of their high impact factor ranking among zoology journals in the ISI Web of Knowledge, and another more recent and specialized online journal (Zootaxa). I first assessed the total number of publications produced by each journal and performed a ‘result analysis’ in the Web of Science to obtain the number of publications per year. I then made the same analyses while using simultaneously all the scientific and English names of the main taxonomic groups of soil animals (as listed in Swift et al., 1979, and Decaëns et al., 2006).
I finally estimated the global representation of the same soil animal taxa in the scientific literature by counting the total number of publications obtained for each of them when using their scientific and English names as key words in the Web of Science (Wilson et al., 2007). The same analysis was performed by using an equivalent number of above-ground animal taxa. The interest of the general public in the same selection of above- and below-ground organisms was estimated by counting the number of web pages that contained their names as keywords in the Google scholar search engine (http://www.scholar.google.com/) (Wilson et al., 2007). The Google search was performed without any restriction for language or file type and I displayed 100 results per page. All searches were performed in May 2008.
KNOWN AND ESTIMATED LEVELS OF SOIL BIODIVERSITY
Soil biological communities comprise species from virtually all the principal taxonomic groups found in terrestrial ecosystems (Swift et al., 1979) (Fig. 1). Soil organisms probably represent as much as 25% of the 1.5 million described living species world-wide (Decaëns et al., 2006, 2008a), representing as much as five times the known biodiversity of forest canopy (Stork, 1988; May, 1990). At a local scale, below-ground species richness is also much higher than that found in vegetation or above-ground fauna. A single square metre of temperate forest soil contains more than a thousand species of invertebrates, mostly mesoinvertebrates (body length ranging between 0.2 and 2 mm) and microinvertebrates (body length < 0.2 mm) (Schaefer & Schauermann, 1990). A single gram of soil in a similar ecosystem may also contain as many as 4000 genotypes of bacteria and about 2000 species of fungi (Torsvik et al., 1994; Hawksworth, 2001).
In spite of this critical contribution to global biodiversity, soil organisms have only weakly engaged the attention of taxonomists compared with other groups like higher plants and vertebrates (Decaëns et al., 2006). The question of how many below-ground organisms exist world-wide remains open, and there is probably no soil where we are able to identify or even quantify all the resident invertebrates or microorganisms (Wall et al., 2005). For this reason soils have been described as the third biotic frontier after oceanic abysses and tropical forest canopies (Andréet al., 1994; Hågvar, 1998). Figure 2 shows, for the principal taxonomic groups of soil organisms, the extent of the taxonomic deficit, which is the estimated gap in our taxonomic knowledge for a given taxon (Blaxter, 2004; Rougerie et al., 2009). Taxonomic deficit for soil biota is on average 76%, usually above 90% for organisms smaller than 100 µm, and tends to be higher for small-bodied taxa (Fig. 3) (Andréet al., 2002; Blaxter et al., 2005; Decaëns et al., 2006). On the other hand, even in groups that have been intensively studied, such as earthworms of the family Lumbricidae, the use of molecular identification tools (DNA barcodes) reveals an unsuspected number of cryptic species that are often impossible to separate on a morphological basis (King et al., 2008; Richard, 2008; Rougerie et al., 2009).
The weakness of our taxonomic knowledge is of particular concern since the profile of below-ground taxa remains very low in systematic and taxonomy journals (Decaëns et al., 2008a) (Fig. 4). The scarcity of taxonomic expertise and of standardized sampling methods and designs has often been proposed to explain this trend (Giller, 1996; Andréet al., 2002; Decaëns et al., 2006, 2008a). Another explanation is the lack of wider public interest in below-ground biota. This is highlighted by the relationship between the amount of scientific literature dedicated to a given zootaxon and its representation on the internet (which reflects public interest) (Wilson et al., 2007) (Fig. 5a). When compared with above-ground biota, soil animals are poorly represented both on the internet and in the scientific literature and, for a given number of websites, also tend to have a lower scientific presence. In comparison, the known species diversity influences only slightly the number of websites and scientific articles (Fig. 5b), but again at equivalent levels of species richness, soil fauna tends to have lower scientific and web presence than that found for above-ground biota. It thus seems that awareness by the general public, rather than scientific interest (as represented by species diversity), explains the amount of research devoted to a given taxonomic group. Consequently, the less fashionable soil organisms are receiving relatively less scientific attention than the high-profile above-ground animals. As highlighted by Wilson et al., (2007), the existence of such taxonomic gaps defeats ecologists' claimed objective to construct generalizations. We thus need to widen our research scope and both wildlife management and teaching should consciously include less-studied groups such as soil biota. Raising awareness of soil biota is a key strategy that must be encouraged in the international arena through diverse initiatives to raise the profile of soil biological diversity.
GLOBAL AND LANDSCAPE PATTERNS OF SOIL BIODIVERSITY DISTRIBUTION
Area–diversity relationships in soil communities
The principal pattern of spatial scaling of biodiversity is the area–diversity relationship (Gaston, 2000). For a given taxonomic group species richness usually increases with the size of the geographic area, and the shape of this relationship conventionally fits a power-law model (McArthur & Wilson, 1967; Gaston & Spicer, 1998; Gaston, 2000). The processes proposed to explain this pattern may vary markedly according to the scale of observation. Increasing area at small-scale domains may enhance the probability of detecting rare species as a result of the sampling of more individuals. At larger scales, increasing area may result in higher habitat diversity and environmental stability, thus leading to a higher ratio between immigration and local extinction processes. Similarly, species–area relationships among biogeographic regions may result from a greater ratio between speciation and extinction rates (Gaston, 2000). Larger observation scales also encompass a greater number of nested levels of organization, from habitat patches to successional stages and landscape units, thus resulting in a larger size of species pools through higher potential spatial and ecological segregation.
The understanding of area–species diversity relationships is of utmost and practical importance as they allow prediction of the negative impact of habitat fragmentation on biodiversity (Gaston & Spicer, 1998). The effect of habitat fragmentation on soil biodiversity has been directly addressed in a few studies focusing on microarthropods (Rantalainen et al., 2005), termites (Fonseca De Souza & Brown, 1994), ants (Suarez et al., 1998; Carvalho & Vasconcelos, 1999; Vasconcelos et al., 2006) and ground beetles (Barbosa & Marquet, 2002; Driscoll & Weir, 2005). Some recent studies have underpinned the complexity and the specificity of the responses within or between broad taxonomic groups. Davies (2002) has, for example, illustrated opposite responses of two termite functional groups to the fragmentation of Amazonian rain forest (positive response for litter and wood feeders, negative for geophagous species). Assuming that different soil organisms have different dispersion and colonization capacities, Hedlund et al. (2004) predict a relative resistance of bacterial-based communities, and a higher vulnerability of fungal-based organisms and of plant-symbiotic organisms like mycorrhizae.
Latitudinal gradients of soil biodiversity
The increase in species richness from high latitudes to the equator is probably the most prominent and extensively studied pattern of biodiversity (Huston, 1994; Brown & Lomolino, 1998; Gaston & Spicer, 1998; Gaston, 2000). Many factors have been proposed to explain this gradient, including the increase in geographic area towards the equator or of the heterogeneity, productivity or environmental stability (both past and actual) of habitats. A remarkably small number of studies have explored latitudinal variations in soil biotic diversity (Bardgett et al., 2005b). The existence of such patterns for soil microorganisms has even been questioned by Wardle (2002), who pointed to the cosmopolitan nature and high dispersal capacity of many taxa, and the fact that trophic resources are rather similar whatever the latitude considered.
On the other hand, the few data available for soil fauna suggest that classical diversity gradients may occur. This was exemplified for oribatid mites (Maraun et al., 2007), ants (Kusnezov, 1957), springtails (Ulrich & Fiera, 2009) and termites (Eggleton, 1994; Lavelle & Spain, 2001) (Fig. 7). Lavelle (1983) also reported an increasing ecological complexity of earthworm communities towards the equator. He pointed to an enhanced efficiency of mutualism under tropical climates as a possible reason for a latitudinal gradient in soil animal communities (Lavelle, 1986). For other groups such as nematodes, the absence of clearly identified gradients may be ascribed to a gross deficit of sampling in intertropical regions (Boag & Yeates, 1998). The same applies for earthworms, which present a very high level of endemism in tropical rain forests (Lavelle & Lapied, 2003), thus suggesting a high proportion of still undescribed species in these ecosystems. The absence of a clear latitudinal gradient for soil biodiversity may thus only be applied to the smaller-bodied organisms, and corresponds to the lack of biodiversity surveys in the tropical compared with temperate areas for many groups of soil animals.
Altitudinal gradients of soil biodiversity
It is usually accepted that terrestrial biodiversity decreases with altitude (Brown & Lomolino, 1998; Gaston, 2000). The details of this relationship, however, vary significantly among taxonomic groups or geographical areas, ranging from a monotonic decrease to a humpbacked pattern with a peak of taxonomic richness at intermediate altitudes (Brown & Lomolino, 1998; Gaston & Spicer, 1998; Gaston, 2000). The few studies that have explored altitudinal variation in soil biodiversity have focused on invertebrates. Some have described a continuous decrease in the total number of species known by altitudinal stratum in a given geographical area. This is the case for earthworms in France (Bouché, 1972; Dahmouche, 2007), for example, and for ants in the Smoky Mountains, USA (Cole, 1940) (Fig. 8a). Collins (1980) also found a decrease in the local species richness of termites along an altitudinal transect in Sarawak (Fig. 8b). Lower species richness at high elevation has been interpreted as the result of: (1) harsh abiotic conditions (in particular temperatures) that function as a strong environmental filter and reduce the number of species able to colonise these ecosystems; (2) low levels of available energy which reduces ecosystem carrying capacity; and (3) small habitat areas.
As demonstrated for many above-ground organisms (Brown & Lomolino, 1998), altitudinal variations in soil biodiversity may also present strong local and/or taxonomic specificity. Collins (1980) reported that the number of beetle families in Sarawak, while generally decreasing with elevation, was highest between 500 and 1200 m. Conversely, dipteran richness was lowest at low elevation and presented a peak between 1300 and 1700 m (Fig. 8b). Loranger et al. (2001) also illustrated a humpbacked variation of springtail species richness across an altitudinal transect from 950 to 2150 m in the French Alps, while González et al. (2007) found that earthworm species richness increased from 0 to 1000 m above sea level in Puerto Rico. These results may partly be explained by altitudinal variation in environmental factors such as rainfall, temperature, pH or quality of organic matter. An increased influence of dispersal barriers, which enhance long-term speciation mechanisms, has also been invoked to explain the high biodiversity levels observed for some taxa at medium elevation ranges (Gaston, 2000).
ECOSYSTEM SCALE CONSTRAINTS ON SOIL BIODIVERSITY
Soil biodiversity controlled by stress and perturbation
It is well supported that the local species diversity of a given group of organisms is mainly driven by stress (resource availability, temperature, pH, etc.) and perturbations (fires, grazing, cultivation, etc.) (Grime, 1973; Huston, 1994; Huston & Deangelis, 1994; Rosenzweig, 1995; Wardle, 2002). Along gradients of stress or perturbation, species richness theoretically follows a humpbacked pattern of variation. In ecosystems characterized by harsh environmental conditions (strongly stressed or perturbed), a reduction in diversity may occur because only a small number of species may have the requisite traits to survive under such constraints (Wardle, 2002). Lower levels of diversity under more productive or less disturbed conditions may also occur as a result of a higher intensity of interspecific competition (Grime, 1973; Huston, 1994; Huston & Deangelis, 1994), a decrease in the heterogeneity of resource spatial distribution (Tilman, 1982) or as a consequence of mechanisms not related to biotic interactions (Abrams, 1995).
While this peaked response of biodiversity to productivity/disturbance gradients has long been presented as ubiquitous in nature (Huston, 1994; Rosenzweig, 1995), a cursory examination of the available literature reveals that it may actually not be as prevalent as commonly believed (Mackey & Currie, 2000, 2001; Mittelbach et al., 2001). Rather, positive monotonic, negative monotonic, and unimodal relationships all appear, and non-significant relationships are also described in a large number of studies. This suggests that the commonly accepted humpbacked diversity–productivity/disturbance relationship may not be the rule (Mackey & Currie, 2001; Mittelbach et al., 2001). In fact, the shape of the biodiversity response to adversity gradients may vary according to, for example, the importance of competition in driving community assembly, the nature of disturbances (e.g. frequency, intensity, predictability, natural versus anthropogenic), the dispersal capacity of the focal taxon, the size of the species pool adapted to highly productive habitats, the efficiency of sampling in detecting rare species or the scale at which the study is conducted (Mackey & Currie, 2000, 2001; Mittelbach et al., 2001; Pärtel et al., 2007).
The results currently available for soil biota are rather inconsistent (Bardgett et al., 2005b; Wardle, 2006). As an example, the diversity of microbial communities (Derry et al., 1999; Degens et al., 2000) (Fig. 9a), chilopods or woodlice (Paoletti, 1988; cited in Wardle, 2006) seems to increase monotonically along a gradient of resource availability. In a synthesis of results from nine different studies, Wardle (2006) found little if any support for decomposer diversity decreasing at the most favourable extremity of stress gradients. He concluded that many soil taxa are not strongly regulated by competition, and that increasing resource availability is not likely to result in competitive exclusion. According to Abrams (1995), such a pattern may also reflect other mechanisms, such as an increase with resource availability in the abundance of rare species, the availability of rare resources, the occurrence of combinations of resources or conditions necessary for specialist species, or in the overall resource volume that allow the coexistence of a greater number of species. Additionally, the strong heterogeneity of the soil environment and the large number of environmental factors affected by this heterogeneity probably also account for the lack of unimodal diversity–productivity relationship observed for most soil organisms.
Not all soil taxa present monotonic responses to productivity gradients, and humpbacked patterns have also been described for some of them. Dahmouche (2007) (after Bouché, 1972), for instance, found that species richness of earthworm assemblages decreased along a gradient of carbon availability (Fig. 9b), although this variation may also be explained by some environmental constraints that are likely to covary with soil carbon levels (e.g. hydromorphy). In this group, however, the systematic limitation of community richness to 10–12 species, whatever the level of available trophic resource (Lavelle et al., 1995; Decaëns et al., 2008b), as well as the well-known cases of competitive exclusion described in some grassland ecosystems (Chauvel et al., 1999; Decaëns et al., 2004) support the idea of strong competitive pressure in the more productive soils. Competition within earthworm communities may be enhanced by the high adaptability and ecological plasticity between juveniles and adults, which lead to a rapid saturation of the ecological space and a relative uniformity in species richness among communities (Decaëns et al., 2008b). Mycorrhizal fungi is another group for which competition is known to limit diversity when nitrogen availability increases (Egerton-Warburton & Allen, 2000; Jonsson et al., 2000; Egerton-Warburton et al., 2007).
Variations in soil biodiversity along disturbance gradients have been extensively illustrated in the scientific literature, mainly through studies on the impact of soil cultivation (Wardle, 2002; Bardgett et al., 2005b). They usually provide little support for the existence of a humpbacked pattern, but rather suggest a negative and monotonic response of most groups to anthropogenic perturbations induced by soil management practices (Wardle, 1995, 2002; Bardgett et al., 2005b). This is the case for microbial communities (Degens et al., 2000), nematodes (Bloemers et al., 1997), springtails (Chauvat et al., 2007), earthworms (Fragoso et al., 1997; Decaëns & Jiménez, 2002; Decaëns et al., 2003), termites (Gillison et al., 2003) or macrofauna as a whole (Lavelle & Pashanasi, 1989; Decaëns et al., 1994; Mathieu et al., 2005) (Fig. 10b). A shared limitation of these studies is that they mainly compare systems that strongly differ not only in disturbance levels but also in a number of other environmental factors. Cultivation is for instance known to severely reduce food supply (and thus productivity) for soil food webs through intense mineralization of soil organic matter and reduction of plant detritus inputs. A decrease in species richness with agricultural intensification may thus not be considered as a sufficient feature to support a negative monotonic response of soil biodiversity to perturbation gradients (Bardgett et al., 2005b). We still need to explore this question through experimental approaches if we wish todisentangle the relative impact of stress and disturbance in structuring soil communities.
Responses of soil biodiversity to successional gradients
The process of vegetation succession is classically described as a succession of three distinct phases: a build-up phase during which plant biomass aggrades, a maximal biomass phase and a decline or retrogression phase (Wardle, 2002). While the build-up phase is usually followed by a marked accumulation of soil organic matter and an increase in nitrogen availability, the retrogression phase corresponds to a drop in primary productivity and in organic matter and phosphorus availability for soil food webs.
During the aggradation phase, the biomass and diversity of soil microorganisms increase and fungi tend to dominate communities (Ohtonen et al., 1999; Schipper et al., 2001; Bardgett et al., 2005a). On the other hand, soil animal diversity may present different non-synchronous peaks, making any generalisation difficult. This feature has been demonstrated by Doblas-Miranda et al. (2008) in a post-glacial primary succession in New Zealand (Fig. 11a) and by Scheu & Schulz (1996) and Decaëns et al. (1997, 1998) in a post-pastoral secondary succession in Germany and France, respectively (Fig. 11b). Bihn et al. (2008) found ant taxonomic richness (number of genera) to increase during secondary forest succession in the Atlantic Forest of southern Brazil, although this pattern was more consistent for epigeic than for hypogeic taxa. Conversely, Scheu (1992) and Decaëns et al. (1997) observed little change in earthworm species richness along secondary successional gradients under temperate climatic conditions (Fig. 11b), with, however, a slight peak in pre-forested stages and an increasing proportion of epigeics (i.e. surface-dwelling species) relative to anecics (i.e. deep-burrowing species). Ponge & Delhaye (1995) also reported a stable richness of earthworms among different stages in the natural cycle of a virgin beech forest in north-western France. This great variability of responses among soil biotic groups underlines different responses to top-down and bottom-up controls and to abiotic factors during the chronosequence (Doblas-Miranda et al., 2008).
Changes in soil biodiversity during the retrogression phase have seldom been addressed in scientific studies. Doblas-Miranda et al. (2008) found that nematode and macroinvertebrate species richness only weakly decreased, except for epigeic macroinvertebrates, during forest senescence (Fig. 11a). Bernier & Ponge (1994) also showed that vegetation senescence in old temperate forests leads to an increased litter palatability for soil fauna, which favours earthworms to the detriment of epigeic arthropods.
A number of agricultural or sylvicultural systems of production use management cycles that drive the system toward succession-like dynamics. In these anthroposystems, soil communities may present similar temporal patterns to those described for true successions. In a forest cycle in Germany, Chauvat et al. (2003) observed a steady drop in microbial diversity after tree harvesting, followed by a monotonic increase during the growing phase (Fig. 11c). In a similar system in France, Hedde et al. (2007) reported a relative stability of the macrodetritivore species richness in the first phases of the cycle, and then a decrease in the older stands (Fig. 11c). In grazed systems, the diversity of soil fauna usually increases during the first years after vegetation sowing and then stabilizes (Hedde, 2006; Chauvat et al., 2007) (Fig. 11d), which reflects a positive response to an increase in resource availability followed by a state of niche saturation in older plots.
SPATIAL PATTERNS AT THE COMMUNITY SCALE
Local spatial patterns of species assemblages
At the community scale, species assemblages are usually spatially structured in patches consisting of specific species associations. The specific spatial range of these structures depends greatly on the body size of the organisms concerned, but also on the ecosystem type and on the sampling spatial design of the study (Ettema & Wardle, 2002). Despite the nested and complex nature of these patterns, some general features may be identified when enough empirical data are available. For instance, soil microbial communities have aggregated distributions at scales ranging from a few tens of centimetres to a few metres (Lavelle & Spain, 2001; Ettema & Wardle, 2002). Soil animals also commonly have aggregated horizontal distributions, forming multispecies aggregates of different sizes depending on the group: from a few tens of centimetres for nematodes (Ettema et al., 2000) and enchytraeids (Lavelle & Spain, 2001) to a few tens of metres for earthworms (Decaëns & Rossi, 2001; Margerie et al., 2001; Rossi, 2003a,b) (Fig. 12). Conversely, ants' nests are often overdispersed (Traniello & Levings, 1986).
Different factors have been proposed to explain the spatial patterning of soil communities. First, each species has specific population dynamics features (mainly related to dispersion ability and demography) that may generate patchiness without any contribution from external factors (Decaëns & Rossi, 2001; Ettema & Wardle, 2002). Beyond this intrinsic population structuring, vegetation and soil properties strongly constrain the relative spatial distribution of the different species that comprise soil communities (Kuzmin, 1976; Ettema & Wardle, 2002). The horizontal patterns of microbial communities are, for instance, mainly determined by pedological factors, litter inputs and rhizospheric effects (Lavelle & Spain, 2001). The spatial distribution of nematode populations is also largely driven by plant and microhabitat distribution and by the spatial heterogeneity in soil physico-chemical properties at scales of a few centimetres (Kuzmin, 1976; Rossi et al., 1996; Lavelle & Spain, 2001). In a chalky grassland in north-western France, Margerie et al. (2001) demonstrated that earthworm assemblage distribution greatly depended on vegetation structure and soil depth. Similarly, spatial patterning of litter biota in forest ecosystems often reflects the influence zone and the relative position of individual trees at scales of a few tens of metres (Ettema & Wardle, 2002). This was for example described for microbial and fungal communities (Bruckner et al., 1999; Klironomos et al., 1999; Saetre & Bååth, 2000), nematodes (Görres et al., 1998; Klironomos et al., 1999) and microarthropods (Klironomos et al., 1999). On the other hand, Aubert et al. (2003) found that a higher heterogeneity in tree cover (i.e. the introduction of hornbeams into a beech plantation) paradoxically ends in a weak spatial structure of soil macroinvertebrates. In other situations, the correlation observed between soil properties and earthworm community patchiness was considered as a consequence of their engineering impact on soil structure rather than as their response to environmental heterogeneity (Decaëns & Rossi, 2001; Rossi, 2003b).
Finally, the importance of competitive exclusion and ecological complementarity in explaining the patchy distribution of species assemblages has been pointed out in the case of earthworm communities (Jiménez & Rossi, 2006; Jiménez et al., 2006; Decaëns et al., 2009). Both intra- and inter-specific competition for food and other resources are also assumed to be the main factors responsible for the regular distribution observed for ants' nests (Traniello & Levings, 1986). Actually, spatial segregation is an important mechanism allowing greater species coexistence and the maintenance of a higher local diversity within soil communities (Ettema & Wardle, 2002). Spatial distribution of soil biodiversity is thus driven by different factors related to soil properties, vegetation characteristics and interspecific interactions. The multiple scales at which these factors influence soil biota and interact with each other result in complex and nested patterns which still require more studies to be fully understood.
Species coexistence and competitive interactions of soil biota
Interspecific competition is recognized as a central factor of ecological community structuring (Diamond, 1975; Connell, 1983). Its implication in community assembly may be assessed through the use of null models that compare observed community patterns with random patterns obtained from null communities that simulate the absence of competition (Gotelli & Graves, 1996). For instance, in a set of competitively structured communities, species co-occurrence is lower than expected by chance because some species pairs are prevented by competitive exclusion (Diamond, 1975). On the other hand, species that coexist in a competitively structured community have to differ substantially in resource utilization, and thus usually present a lower niche overlap or a higher morphological differentiation than expected by chance (Schoener, 1974).
Microsite patterns and the importance of soil engineers
At smaller spatial scales (a few centimetres), the distribution of soil organisms is influenced by the heterogeneity induced in resource distribution by physical ecosystem engineers (Jones et al., 1994). For instance, Tiwari & Mishra (1993) reported higher fungal diversity in earthworm casts than in bulk soil (Fig. 13a), Loranger et al. (1998) found that arthropod species richness was enhanced in patches of high earthworm density in a Martinique vertisol and Decaëns et al. (1999) documented a higher taxonomic richness of epigeic macrofauna within andbelow the surface casts produced by a large anecic species in the Oriental Colombian savannas (Fig. 13b). These results are interpreted as a consequence of higher availability of the trophic resource (organic matter concentration and availability in casts) and microhabitats (surface roughness and burrows) for soil biota due to earthworm casting and burrowing activities. Boulton & Amberman (2006) also documented that bacterial diversity (phospholipid fatty acid (PLFA) profiles) is enhanced in soil experimentally processed by ant activities (Fig. 13c). They attributed their results to higher levels of nutrients available in ants' nests compared with bulk soil. On the other hand, Dauber & Wolters (2000) found that microbial functional diversity was increased or decreased in ants' nests, depending on the ant species. They concluded that species-specific differences in the effect on the soil microflora are related to feeding strategy and nest architecture.
Vegetation, through litter production, also has a strong impact on levels of both stress and productivity in the soil system, with important consequences for the diversity of soil biota (Wardle, 2002). It has been documented that plants have specific effects on the diversity of soil communities, and that vegetation species richness is related to the richness of different groups of soil organisms. For instance, Gillison et al. (2003) found a direct correlation between the species richness of plants and termites in Sumatran forests (Fig. 14a). Other studies have demonstrated the influence of microhabitat diversity or litter species richness on the diversity of soil animals like mites (Anderson, 1978; Hansen & Coleman, 1998; Kaneko & Salamanca, 1999) (Fig. 14b & c). However, these patterns do not distinguish and establish a hierarchy between the impact of a modification of habitat structure by vegetation and the impact of vegetation through changes in the level of productivity or stress for soil communities.
CONCLUSIONS AND PERSPECTIVES
Despite the fact that soil communities may represent the main part of terrestrial biodiversity, a large proportion of their species still remain undescribed. Soil biologists thus have to deal with communities that may contain millions of species that are very difficult if not impossible to distinguish from each other (Decaëns et al., 2008a). This restriction results in an important taxonomic deficit for most groups of soil biota, and in an inadequate understanding of spatio-temporal biodiversity patterns that have been more accurately described for above-ground taxa. Understanding the factors driving soil biodiversity across different spatial and temporal scales is of primary importance if we wish to predict the response of soil communities to global changes and the impact these changes will have on the delivery of ecosystem services. It is thus of prime necessity to address this taxonomic deficit through actions at different levels in order to stimulate studies on soil biodiversity. These include the development and standardization of new taxonomic approaches, as well as educational and editorial strategies to improve both the popularity of soil taxa among the general public and their representation in the scientific literature.
Despite the dearth of information available on soil biodiversity patterns, the results synthesized in this review allow some conclusions to be drawn. First, at the local scale, most soil microorganisms and animals present similar responses to habitat constraints and biotic interactions that drive community assembly rules. Existing data also provide little evidence that driving factors may regulate above- and below-ground communities in the same way (Bardgett et al., 2005b). Except for a few taxa, stress and disturbance seem to restrict soil biodiversity levels only at the adverse extremity of gradients, suggesting that the humpback model may not apply underground. Soil communities thus seem to be weakly structured by competition, although strong competitive constraints may account for assembly rules within specific taxonomic groups and ecosystems. The main factor constraining soil biodiversity is the compact and heterogeneous nature of the soil matrix, which provides unrivalled potential for niche partitioning, thus allowing high levels of local diversity. This heterogeneity is otherwise strongly increased by the omnipresent activity of ecosystem engineers that generate patchiness at a range of spatio-temporal scales.
At larger scales, it seems that microorganisms do not respond to the main environmental gradients in the same way as metazoans. As proposed by Wardle (2002), soil microflora are mainly represented by cosmopolitan organisms with high ability for passive dispersal, a high proportion of which may present a widespread distribution which is made possible by the small differences that exist between environmental characteristics in different soil types. Conversely, the few studies currently available on large-scale patterns of soil animal diversity suggest biodiversity gradients that mainly agree with those classically described for above-ground organisms (Gaston, 2000). Although these major differences among taxa of contrasting body size may be explained by differences in the adaptive strategies and dispersal abilities of species, the paucity of available studies necessitates caution when building generalizations.
I would like to thanks Juan José Jiménez and Patrick Lavelle for all the fruitful discussion that greatly inspired this review and for making useful comments on an early version of the manuscript, and to Samuel James and James Curry for editing the English language.
Thibaud Decaëns is full professor and head of the Laboratoire ECODIV of the Université de Rouen. His research interests include various aspects of invertebrate ecology and biology, including community and functional ecology, taxonomy and conservation biology. He is particularly interested in soil fauna and macrolepidoptera and has been involved in biodiversity studies in a range of temperate and tropical ecosystems in Europe, Latin America and Equatorial Africa.