The study of interacting sets of species within ecosystems – community ecology – is generally less familiar to entomologists than the more species-centric studies of population and behavioural ecology. Here, I review some key areas of community ecology that should be of interest and concern to entomologists. The contrasting current models of community assembly – based, on the one hand, on ideas of neutral replacement and, on the other, of resource-based niche partitioning – are discussed. A synthetic version that identifies a stochastic/deterministic switch point based on a characteristic spatial scale for each community studied is presented as a possible way of combining these two approaches.
Insect ecologists, traditionally, pay little attention to developments in community ecology. Rather, much of the ‘mainstream’ of entomology is concerned with understanding and modifying the dynamics of populations of single species (Speight et al. 2008; Price et al. 2011). Where other species are considered, it is often as natural enemies, prey or competitors – i.e. as driving variables modifying the dynamics of the focal species. This species-centric approach is absolutely the right way to go when the control, modification, even restoration, of an insect species is the driver of the research. Ecologists in general, however, have a broader set of study objects examining, among other things, groups of interacting species within particular ecosystems – i.e. ecological ‘communities’ (Fauth et al. 1996). Relatively recently, there has been a wide promotion of the idea that it is the ecosystem processes driven by the species acting individually or in concert that keep nutrients cycling, water clean and plants pollinated, even maintaining pristine ecosystems so we can have a thriving ‘eco'tourism industry and agriculture (Millennium Ecosystem Assessment 2005; Schellhorn et al. 2008). It is the whole set of species organised into food webs and acting through feeding interactions that keep ecosystems healthy and ensure the provision of these services – which are often regarded as ‘free goods’ by environmental economists but, on any attempted replacement valuation, are worth trillions of dollars to the global economy (Constanza et al. 1997; Liu et al. 2010). Entomologists will be quick to point out that, above the plant level, most of the moving parts in these ecosystem machines are insects (following Wilson 1987).
Of course, when we study the many species of insects that occur within a particular ecosystem, we are not generally studying the whole community – indeed, if we fail to consider the micro-organisms, we are never actually studying the entire community. Even my beloved plant container food webs (e.g. Kitching 2000) may contain only arthropods as their multicelled living parts, yet they contain untold millions of highly diverse micro-organisms as well – generally ignored by we macro-biologists. So to speak of ‘insect communities’ (or bird communities, or lizard communities) is not strictly correct. We are actually studying taxonomically (sometimes trap- sometimes trophically-) defined assemblages of species (Fauth et al. 1996; Magurran 2004) and the interactions they encompass. Nevertheless, well-chosen, well-researched subsets of the entire arthropod assemblage within any terrestrial (or freshwater or marine) community can act as robust surrogates for the entire community and as such may be used as indicators of the health of the community and the ecosystem within which it occurs (Kitching et al. 2000; Davis et al. 2001; Andersen et al. 2004).
Global environments are experiencing a wide range of dramatic human-driven modifications including clearing, exotic invasion, inappropriate fire regimes and, not least, climate change. These stressors act separately but also interact significantly (Steffen et al. 2009; Lindenmayer et al. 2010; Sutherst et al. 2011). A crucial response to these changes on the part of both science and managers is the measurement and monitoring of these changes. Here, insect assemblages come into their own.
I contend, therefore, that affairs within community ecology are (or at least should be) of concern to entomologists. This means that a particular vigorous debate going on within community ecology is doubly of importance. This debate has turned into a dialectic in which a (relatively) new proposition (‘neutral theory’) has threatened the long-held but seldom-tested foundations of the whole subdiscipline (‘niche theory’).
This overview, then, has three goals. First, I have already gently reminded entomologists about the nature and importance of community ecology and will enlarge on that reminder by discussing, briefly, some of the more obvious emergent properties of communities relevant to understanding the patterns and roles of insects in ecosystems. Second, I shall review the current ‘hot’ debate in community ecology that contrasts deterministic (niche based) with neutral (randomly assembled) explanations for community structure. Last, I shall suggest ways in which these apparently contrasting approaches can be combined, especially for those, like me, interested in describing and explaining patterns in insect assemblages at the landscape level. A parallel, data-based development of some of these ideas surrounding a particular example is in Kitching et al. (in press).
Emergent Properties of Communities
To understand any critique of community ecology, we need to have a clear picture of what the emergent features of ecological systems are that community ecology attempts to address. There are three principal interrelated sets of phenomena involved.
The idea of diversity at its most basic is the simple count of the number of species in a given place and time. A more sophisticated extension of this is provided by Hubbell (2001) as ‘the species richness and relative species abundance in space and time’ (see also Magurran 2004). These rather obvious emergent properties of community ecology are simply those that arise from sampling multi-species situations: the basic statistics, if you wish, of communities. Any community sample will contain a set of species represented by the differing numbers of individuals of each species. The earliest attempts at analysis simply extracted univariate measures from such data – richness, evenness, dominance – and their underlying variances – which could be processed using simple statistics. In fact, these were moments, or derivatives of moments, of underlying distributions to which existing statistical models could be fitted or new, empirical fits could be developed. After decades of competing claims, a general convergence on models not unlike the log-normal distribution appeared in the 1980s and 1990s (Magurran 1988), although there remained acknowledgement that empirical fits to concave simpler models (Hughes 1986) might well be adequate descriptions of simpler, less complete data sets.
Most recent analyses in this area have focused on the search for broad explanations of entire species/abundance distributions (Kitching et al. 2004; Hamilton et al. 2010) rather than naïve reliance on simple statistics such as richness and evenness. These, and the many ‘diversity indices’ derived from them, remain useful as summarising quantities but, for the diversity-oriented ecologist, probably conceal more than they reveal. Of course, once protocols for capturing the dimensions of diversity are developed, then studies using them can be designed on a variety of spatial and temporal scales across a variety of environmental mosaics. This has led to the widely used simplifications of alpha-, beta- and gamma-diversity (sensu Whittaker 1960). These terms relate to the diversity of organisms in a particular location (‘alpha'-diversity), contrasts in diversity between pairs of locations (‘beta-’diversity) and regional diversity across a set of habitats or ecosystems (‘gamma'- or landscape diversity). Of particular interest to insect ecologists interested in communities is the concept of beta-diversity, in which conclusions are drawn based on comparisons among two or more estimates of diversity from different locations reflecting environmental contrasts of one sort or another (for a full recent account, see Tuomisto 2010a,b).
One of the oldest observations in ecology is that species interact with each other through their feeding interactions to form more or less complex sets of connections along which flows energy from food – at least, potentially. These are food webs, and any single ‘vertical’ pathway through such a web is a ‘food chain’ (Gallopin 1972; Cohen 1978; Pimm 1982; Kitching 2000). Samples aimed at elucidating food webs require more information than the simple species/abundance quantities of diversity studies. In addition to knowing what is present in the target community and how many of each species there is, we also need to know what eats what. It may be sufficient to know, or deduce, the feeding level at which each species occurs; i.e. whether it is a photosynthesiser, decomposer, herbivore, predator or top predator. More sophisticated analyses seek actual dietary information on each participant in the food web under the study. As is the case with the ‘simple’ studies of diversity, food webs generate their own set of summarising statistics – the number of trophic levels, average food-chain length, connectance, guild distributions, predator : prey ratios and so forth. Very recently, the whole field of food-web ecology has been extended to incorporate more general network theory in which a wider range of interactions – not just ones based on feeding interactions – are included in the communities being analysed (Ings et al. 2009; Thompson et al. 2012).
Successions and seres
Often overlooked in summaries of community phenomena, especially by zoologists, are the sequence of so-called seral stages that make up primary or secondary successions – the accumulations of, and changes in, the communities that occur on newly available land or on recently disturbed land. Initially used to describe sequential changes in plant assemblages, these changes at a variety of spatial and temporal scales may be one of the most fundamental organising principles of ecological communities (Brown & Southwood 1987; Glenn-Lewin & van der Maarel 1992). Here, the essential sampling dimension is the time series that tracks diversity and other changes through time within ecosystems. Simple emergent properties include the recognition of species representative of early, middle and late seral stages, changes in the average life-history (even physiological) characteristics of emerging dominants within each seral stage and changes in the nature of species–species interactions, and so forth.
Competing ideas of the causes of such changes have revolved around questions such as: do ‘later’ species simply outcompete earlier ones? Do earlier species actually open the way for the establishment of later ones? Or, indeed, are there any cross-ecosystem generalisations possible or likely (Glenn-Lewin et al. 1992)?
There is also wide recognition that these temporal changes interact in vitally important ways with surrounding landscape units (Connell & Slatyer 1977; Lawton 1987), leading to a patchwork of out-of-phase seral stages within established ecosystems; this ‘intermediate disturbance hypothesis’ remains the basis for much modern vegetation theory (e.g. Molino & Sabatier 2001; Roxburgh et al. 2004; Catford et al. 2012 but see, also, Fox in press).
Determinism vs. Neutrality?
Niches – a deterministic theory?
Elementary ecological theory has long accepted that the distribution of any given species in time and space defines its ecological niche. It was G. Evelyn Hutchinson who, long ago, in his enigmatically titled paper ‘Concluding Remarks’ (Hutchinson 1957), invited fellow scientists to look at the ecological niche as an n-dimensional hypervolume within the n-space defined by the many environmental dimensions, both biological and physio-chemical, which determine simply the presence or absence, or the well-being of the organism in a particular location. This was a more formal and general statement of the earlier niche concept defined by Joseph Grinnell and Charles Elton, who simply viewed the niche as the ‘job’ carried out by particular species within the environment (Grinnell 1917; Elton 1927). The only difference of any consequence in these two niche definitions is that the Hutchinsonian one does not allow for empty niches given that it is defined by the interaction between organism and environment on which the very existence of the niche depends. There may be vacant space in the n-dimensional space within which organisms occur, but these cannot be viewed as niches in the absence of the interactor! The Elton/Grinnell concept, on the other hand, can freely speak of the absence of, e.g., middle-sized top predators within a community defined, as they are, by the job they would do if they were present. In practice, these semantic differences make little difference in the utility of the term niche, although there is no doubt that the Hutchinsonian term has been more readily mathematised.
Various restatements of these ideas have been made recently. These place greater emphasis on spatial and temporal variability – both extrinsic and intrinsic – and the dynamics of inter-patch change (for reviews see, e.g. Chesson 2000; Chase & Leibold 2003; Mouquet & Loreau 2003). Nevertheless, they concur in regarding species as differing fundamentally in one form or the other, either as a result of directional selection over evolutionary time and/or active exclusion in space or time in the here and now.
Returning to niche basics, Hutchinson (1957, 1957,1959), Vandermeer (1972) and others noted that the range of niche dimensions could usefully be divided into two non-overlapping sets. The physical and chemical dimensions of the environment (or their surrogates such as altitude, latitude or substrate type) define the outer, physiologically defined envelope within which a species can exist. The success or otherwise of an organism's tolerance of a particular location along any one of these fundamental dimensions is determined by deep evolutionary responses, often at the cellular or tissue level. With very few exceptions, a formal location record for individuals of a species implies that location falls within this physiological envelope. A geographical extrapolation of this by climate matching will define the fundamental spatial niche of the species. But the distribution of a species is not solely defined by its physiological tolerances. Species in nature live within communities and participate in complex networks based on feeding and other interactions. The species–species interactions in which a focal species participates are the biological dimensions of its niche.
The biological dimensions of a species' niche will determine which portion of the outer fundamental niche will actually be occupied. So, as a simple example, the wanderer butterfly (Danaus plexippus L.) has had a fundamental niche space that encompassed most of Australia since at least the last Ice Age. It did not establish here, however, until the 1900s, by which time several species of its preferred food plant had established as agricultural weeds. The availability of appropriate food plants is a dimension of the species' realised niche. Similar examples reflecting the presence, absence or prevalence of parasites, parasitoids, predators, pathogens or competitors can be readily identified for this and other species in which we might be interested. The emerging, well-established point is that species actually occur within their realised niche, which is that subspace of their fundamental niche into which the biological dimensions have been added. Considering interactions among multicelled organisms, this will virtually always be a smaller volume. In a few instances where obligate mutualisms are involved (as with lichens, mycorrhizae or even butterfly–ant interactions), the realised niche may be larger than the fundamental niche. If, however, we include mutualistic micro-organisms that form a part of the endo-biota of organisms, particularly insects, this may be a much larger category (undoubtedly warranting a whole overview for themselves) (see, e.g. Gunduz & Douglas 2008; Klepzig et al. 2009), but these are unlikely to be relevant to situations involving insects.
Table 1 presents my interpretation of the situation for some insects on which I have worked. In each case, the fundamental niche is expressed in geographical terms (a surrogate for the synoptic climate). I provide a key introductory reference in each case. I hasten to add that my conclusions are not always those of the authors with whom I have worked.
Table 1. Niche dimensions of some well-worked species of insect
Of course, all of this dogma assumes species are at evolutionary and ecological equilibria. At times of dramatic environmental change, a species will need to respond to dramatic encounters with other sectors of both its fundamental and realised dimensions. Indeed, whole new dimensions may arise. In these cases, ‘adapt or perish’ will be the rule.
The neutral alternative
In 1967, Robert Macarthur and Edward O. Wilson published their book The Theory of Island Biogeography. This was one of the first theoretical contributions that allowed ecologists to make predictions about the nature of ecological communities. As every ecology student knows, Macarthur and Wilson defined two time-driven curves plotting the number of species along the y-axis against time. Where the ascending extinction curve intersected with the descending colonisation curve, there was defined an equilibrium species richness for the island in question. This was a ‘neutral’ theory inasmuch as it made no biological distinctions among species. All that defined the prediction was the size of the biota on the adjacent ‘mainland’, the size of the island and the distance between the two. Subsequent criticisms, experimental tests and extensions of the theory drew attention to shortcomings (Simberloff & Wilson 1969, 1970; Lomolino et al. 2010). Clearly, sustainable populations of predatory species on the island could not be established until their prey species have established; some species may be specialised in being island invaders; yet others established if and only if the island was topographically heterogeneous; and so on. Nevertheless, the original, simplistic neutral theory of Macarthur and Wilson represented a very powerful predictive tool subsequently used in both fundamental and applied justifications for either existing patterns in data or in achieving (or attempting to achieve) management goals (see, e.g. Losos & Ricklefs 2009).
The Theory of Island Biogeography predicts only the equilibrium species richness in a habitat ‘island’. As we have noted already, there are many other emergent features that community ecologists attempt to explain. The most obvious and all-pervasive of these emergent features of multi-species assemblages have been species/abundance relationships. The log-normal distributions that emerge as clear contenders for the ‘general’ model for such curves are, ultimately, nothing more or less than empirical fits to large data sets. In 2001, Stephen Hubbell published an extended version of what he called the Unified Neutral Theory of Biodiversity and Biogeography. Simply stated, this presented a model in which the species/abundance curves observed in data sets derived from large (50 ha) vegetation plots could be derived from nothing more than a process of random replacement following the simulated deaths of individual trees based on the composition of the tree assemblage surrounding the plot, plus a factor allowing for the appearance of evolutionary novelty. The models assume competitive neutrality which produces stochasticity. The model produced distributions virtually indistinguishable from the log-normal distributions of other very large ‘real’ data sets. Like the Macarthur and Wilson model, it too was ‘neutral’ inasmuch as it made no biological distinctions among species. Other authors added representations of dispersal distance (Chave & Leigh 2002; Morlon et al. 2008), again in a neutral fashion, to the model and were able to show that this allowed even more commonly observed community patterns to be simulated with no further incorporation of biological details. This included the expected relationship between diminishing similarity in species composition with distance.
Not surprisingly, Hubbell's theory produced a heated response, and the debate continues. For zoologists, this was a direct challenge given that behavioural, dietary, habitat and other differences observed in the field had seemed to lead to the idea of evolution-driven dividing up of ‘niche’ space. The neutral idea was perhaps less of a challenge for field botanists, especially those dealing with the great diversity of rainforest tree species for which differences in ecological ‘strategy’ are not immediately obvious. Some animal assemblages, such as coral reefs and ant assemblages, did behave much more like trees in rainforests in this regard.
A synthetic theory?
It is perhaps indisputable that the composition of an assemblage of rainforest trees within a 50 ha plot can be recreated using the neutral models of replacement already described. In a similar fashion, chance invasion and establishment events have been suggested as the best way of accounting for differences in animal food-web composition when single habitat units of natural microcosms are considered (Kitching 1987, 2000).
I suggest that in both instances the adequacy of models based wholly on stochastic (i.e. chance) processes is a reflection of the spatial scale at which the community is being examined. So for the microcosms I described in 2000 (animal communities in water-filled tree holes), there are plausible deterministic explanations for observed structures when data are aggregated above the level of the individual habitat unit. In the case of these microcosms, the scale at which such deterministic explanations become useful is as little as a few hundreds of metres. Combine data from all the habitat units within, say, a few hectares, contrast this with similar combinations from other comparably sized patches, and a range of biologically driven explanations based on ideas of the niche become sensible. I discuss this at length in Food Webs and Container Habitats (Kitching 2000). Transferring these ideas to the much larger scale at which rainforest tree assemblages exist, I suggest, then, that a single 50 ha is comparable to a single unit of microcosm.
Viewed at a larger scale, deterministic explanations become again sensible – perhaps at a scale of hundreds of kilometres. Comparing swathes of rainforest vegetation on adjacent limestone and non-limestone substrates, for example, shows up dramatic differences based on the soil tolerances of the tree species concerned – a quintessentially niche-based differentiation. A similar argument can be put forward for high-elevation ‘elfin’ forests compared with their lower-elevation neighbours, for coastal forests in contrast to those in adjacent hinterlands, even for post-logging forests in recovery phase compared with pristine remnants as they once were. Within each of the forest types mentioned, neutral models would likely recreate assemblages that are realistic, yet accounting for differences between types demands niche-based explanations. At an intermediate scale, and based on nothing more than ecological commonsense, perhaps vertebrate assemblages may be reasonably explained with stochastic models up to scales of a few kilometres but become amenable to deterministically based explanations based on limiting values along realised niche dimensions when the spatial scale is in the tens of kilometres or more.
Figure 1 attempts to capture these thoughts intuitively in a diagram in which just three of the continuum of biological assemblages that may be studied are viewed against a spatial scale from small to large. In each case, I propose a switch distance at which deterministic (niche-based) levels of explanation are feasible but below which stochastic models will recreate observations adequately. This ‘stochastic-deterministic switch line’ (SDSL) describes a set of critical values that become larger as the characteristic scale of the assemblage being examined increases. In summary, changes in community structure will likely be amenable to deterministic explanation above a certain spatial scale – and the switch point will vary from community type to community type. The figure, of course, is nothing more than a hypothesis about how things work in nature and, like all such hypotheses, needs extensive testing against a range of data sets.
One such test of these ideas occurs when examining two sets of data on taxa that essentially operate at different spatial scales. Essentially, an adequate theoretical explanation demands a combination of neutral and niche ideas. In a recent study, we examined (among other things) the changes that occurred in assemblages of moths over a set of increasing distances from 100 m to 80 km within more or less continuous primary rainforest in Sabah (Kitching et al. in press). In that paper, we identified a clear distance–decay relationship over the spatial scale of our sampling. Curiously, this clear relationship is absent in logged-over forest – but that is a different story. In attempting to explain the pattern in the primary forest, we suggested that the underlying changes in vegetation over this spatial scale are well explained by neutral ideas, as modified to include dispersal distances (Chave & Leigh 2002; Morlon et al. 2008). The associated changes in the moth assemblages, however, we suggested, are driven by locally changing availability of larval food plants, and these represent key dimensions of the niches of moth species concerned. Of course, pedantically, any process with an underlying random element makes all the associated derivative processes in turn random. This, however, is not the way neutral ideas in community ecology should be interpreted. In terms of the present argument, this example strongly supports the assertion that both neutral and niche ideas are required to explain patterns in nature. Others have proposed various models in which the two approaches can (and should) be joined (see, e.g. Leibold & McPeek 2006). The current scale-based conceptual model adds to this ongoing debate.
The ideas presented in this essay have been incubating for several years. They finally came to fruition during a period I spent as visiting professor in the Department of Diversity and Evolutionary Biology at the Université Paul Sabatier in Toulouse. I thank the university for this privilege and Professor Christophe Thébaud, who was my primary host during that visit. Conversations with Professor Jerome Chave of that department crystallised many of my ideas on neutrality in ecology. The article has been much improved by the comments of four referees. Needless to say, errors of commission and omission remain my own.