J. Alexandre F. Diniz-Filho, Departamento de Ecologia, Instituto de Ciências Biológicas, Universidade Federal de Goiás, Campus II, 74001-970, Goiânia, GO, Brasil. E-mail: email@example.com
Abstract. 1. Despite the abundance, richness and ecological importance of insects, distribution patterns remain unknown for most groups, and this creates serious difficulties for the evaluation of macroecological patterns and the underlying drivers. Although the problem is real, we provide an optimistic perspective on insect macroecology and conservation biogeography.
2. Although data for macroecological analysis of insects are not as complete as for many other organisms (e.g., mammals and birds), at least for some insect groups they are equivalent to what existed 10 or 20 years ago for the charismatic megafauna, so initiatives to compile data for broad-scale analyses are feasible.
3. The primary constraint for studies in insect macroecology and conservation biogeography is not (only) poor data; part of the problem arises from a lack of knowledge on how macroecological patterns and processes can be analysed and interpreted.
4. Finally, we present an overview of recent papers using insects as model organisms in macroecology, including richness and diversity gradients, ecogeographical rules, inter-specific relationships, conservation planning and modelling species distributions. Although our list is not exhaustive, it may be useful as guidelines for future research and encourage ICD readers to develop analyses for other insect groups.
Insects are the most diverse group of multicellular organisms on Earth, with more than a million described species and a large number of non-described species. For the majority of these species, there are almost no data on geographic distributions, even in a broad sense of extents of occurrence. The lack of knowledge of taxonomy and biogeographical patterns has been called the Linnean and Wallacean shortfalls (see Whittaker et al., 2005). As these shortfalls are ubiquitous in many insect groups, clearly, there are even less data on biology, phylogenetic relationships and life-history. On the other hand, insects are a very important functional group, and thus their conservation is critical to the ecological and evolutionary processes that drive and maintain the diversity of terrestrial ecological systems everywhere (Thomas et al., 2008).
The lack of basic data makes the evaluation of broad-scale diversity patterns and their underlying ecological and evolutionary drivers difficult, although these are the core of the macroecology research program (Brown, 1995; Blackburn & Gaston, 2003). Macroecology has been growing rapidly in the last 20 years, and recently it has also been recognised that there may be advantages if conservation strategies are based on broad-scale data, stimulating the development of the new field of ‘conservation biogeography’ (Whittaker et al., 2005). In both macroecology and conservation biogeography, the basic units of analysis are usually the species’ geographic range (but see below), which explains why Wallacean and Linnean shortfalls are a major drawback for using insects as model organisms to understand broad-scale diversity patterns (see Basset et al., 2009). Of course, this is the case for many other organisms as well (e.g., Heger et al., 2009).
Although, it is impossible to deny the existence of Wallacean and Linnean shortfalls with respect to insects, we would like to provide an optimistic (and, hopefully, provocative) perspective on insect macroecology and conservation biogeography. Our main arguments are that: (i) although the data for macroecological analysis of insects are not as good as for some other organisms, at least for some insect groups they are equivalent what we had 10 or 20 years ago for the charismatic megafauna or woody plants, so initiatives to compile data for broad-scale analyses are feasible; (ii) the primary constraint on such initiatives and the number of studies in insect macroecology and conservation biogeography is not only poor data; there is a lack of knowledge by researchers on how macroecological patterns and processes can be analysed and interpreted; and (iii) even given that the data are not perfect, there are many approaches that can still be used (and have been done) with some insect groups. We develop each of these points in more detail below.
Challenging Wallacean and Linnean shortfalls
It is clear that the lack of basic knowledge makes it difficult to test some macroecological patterns, especially those that rely on knowing the geographical distributions of constituent species (e.g., diversity gradients). Even for better known groups of insects, such as butterflies, ants and bees, even crude data on species’ ranges are still lacking for the hyper-rich tropical regions in Asia, South America and Africa. Of course, having high quality distribution data would be ideal, but due to impacts of humans on the Earth and accelerating rates of habitat loss, waiting for perfect data may be not a viable choice for conservation biogeography. Thus, it is important to search for fresh research perspectives that minimise problems due to limited data.
First, macroecological analyses in the early 1980s were based on much coarser data than are available today, even for mammals and birds (e.g., Brown, 1995). But despite the lack of detail, broad-scale macroecological patterns now evaluated using more refined data tend to remain similar and thus are not strongly affected by data resolution (Hawkins et al., 2008; Keil & Hawkins, 2009). It is also important to note that detailed and comprehensive distribution data at the global scale for vertebrate groups appeared only in the last 5–10 years, and that a fully resolved species-level phylogeny coupled with complete data on ecology and life-history is not available even for mammals and birds. Thus, although we are still a long way from having all of the data we need for any large group of insects, it is not impossible to further develop insect macroecology now.
Thinking of the historical development of macroecology over the last 20 years, perhaps a good avenue for insect macroecology is first to begin generating coarse scale distributional data for many groups rather than trying to detect patterns using local, detailed inventories. In the context of the Wallacean shortfall, using data derived from coarse range maps and extents of occurrences may be better than conducting analyses based on local data, especially in areas where sampling has not been intensive (see Hawkins et al., 2008). Also, although deriving macroecological patterns from local occurrence records is very tempting because of the rapid advances in niche modelling (NM) techniques (see Elith & Graham, 2009), these methods may be difficult to apply based on limited and usually biased occurrence data (and indeed many insect species are probably known from a single locality, or even from a single individual). Even so, they can be used in some circumstances to evaluate gaps in knowledge (see Almeida et al., 2010).
The Linnean shortfall is probably a more serious challenge for insects, since most species are not yet described. However, macroecology deals with species as ‘particles’ diffusing in space and time (see Brown et al., 2003), so a (large) sample of species may be sufficient to allow us to identify overall patterns and processes. One problem is that samples are usually not random, and lack of knowledge about species tends to be associated with macroecological traits, such as geographic range size, body size, colour patterns or behaviour, or with where species are located (Lobo et al., 2007; Ferro & Diniz, 2008; Jones et al., 2009; Stork et al., 2009; Baselga et al., 2010). But, there may be ways to ameliorate the serious problems raised by the Linnean shortfall.
One useful approach may be to focus on higher hierarchical levels (the higher-taxon approach) to avoid problems in alpha taxonomy and evaluate patterns based on generic or familial richness (Williams & Gaston, 1994; Gaston et al., 1995; Balmford et al., 1996). Theoretical support for this approach may be based on the widespread existence of niche conservatism for tolerances to environmental conditions and patterns of resource use in an insect taxon (see below). Under this process, we would, for example, expect congeneric species to have similar broad-scale responses to the environment, so that patterns based on a generic level analysis would be similar to one based on species.
Another avenue to avoid both Linnean and Wallacean shortfalls is to focus on better known groups, as is already being done. Most macroecological analyses with insects have been focused on a few families or superfamilies, e.g., butterflies, ants and some Coleoptera (see ‘Highlights in insect macroecology and conservation biogeography’ section). This is expected because it is essentially impossible to compile data for entire insect orders. Further, it is important to bear in mind that, because insects are a very old group, it may not be ‘fair’ to expect a research program in insect macroecology to mimic work with mammals and birds taxonomically. Because of their age, highly diversified insect groups at the family or even generic level may enough to understand macroecological patterns. In this case, it is possible to use previous knowledge on patterns and process based on vertebrates to better define research programs in insect macroecology. For example, Hawkins et al. (2007a) examined tropical niche conservation as an explanation for latitudinal diversity gradients (see ‘Highlights in insect macroecology and conservation biogeography’ section) by analysing richness gradients for basal and derived groups of birds (which did not require a fully resolved phylogeny). Thus, if one is interested in latitudinal diversity gradients, one strategy to start is to focus on insect groups (wherever the taxonomic level) that appeared at more-or-less the same time as mammal and bird families. Comparing patterns of basal and derived groups of insects could then be useful to test the generality of niche conservatism as an explanation of richness gradients.
Moving to a macroecological perspective in insect ecology will obviously require achieving the delicate balance between waiting until we have more reliable data and initiating tests of broad-scale hypotheses with the data we have. Given the scope of the task of getting data for any major insect group, knowing when we have sufficient data to attempt a deeper search for patterns and eventually processes underlying such patterns (see Hortal et al., 2010), will always be difficult. Even more importantly, it may be necessary to change how data are collected, extending sampling effort across much broader geographical areas instead of increasing investment in local sampling through time. This will require a change in researchers’ scientific attitudes, including both how we write research proposals and increasing collaboration networks both to expand sample range and to acquire new theoretical and methodological skills.
Scientific perception and required conceptual and methodological advances
Once the Wallacean and Linnean shortfalls are minimised it is important to consider what conceptual and methodological advances are required to further develop insect macroecology. Definitely, the first challenge is to move from an empirical towards a theoretical approach in which available data are used to test general ideas. For local-oriented ecologists, the main issue is to understand the balance between collecting local data (the traditional activity) and compiling enough information for an effective macroecological analysis.
Despite the advances in geographical ecology and broad-scale diversity analyses started in the early 1960s and 1970s, ecology has focused on local scales and dealt with population and community processes, also adopting an experimental (rather than observational) framework. This applies to insect ecology as well. However, in the early 1990s, a more pluralistic approach began to emerge and was integrated into the macroecology research program, being further extended to solve problems in biodiversity conservation. Macroecology itself is expected to evolve from a pattern-to-process correlative discipline to a mechanism-to-pattern deductive approach as it becomes a more mature research area. Local insect ecologists swimming in the conceptual pool of deductive population biology may help this important challenge to macroecology.
Even so, a shift in the geographical scale of analysis is not simple because it is correlated with many other aspects of scientific research and epistemology. For example, it is now recognised that, although macroecological approaches are limited in terms of causal inference, they may be the only way to assess the ecological and evolutionary processes that drive and maintain diversity at broad-scales. Because of the complexity of the processes, their interactions and emergent properties, up-scaling processes known to act at local scales will usually fail to explain broad-scale patterns (Hortal et al., 2010).
It is interesting to note that knowledge of insect ecology accumulated in the last 50 years may reveal additional challenges compared to the macroecology of vertebrates. For example, analyses often indicate that general models, such as water-energy balance, may explain richness patterns in some insect groups, as suggested by the correlations between richness and broad-scale environmental surrogate variables (see below). However, at the same time, being small bodied, insects may be more dependent on local habitats and microclimates, as well as on biotic interactions (e.g., insect–plant relationships) than mammals and birds. Thus, perhaps understating broad-scale patterns in insect diversity is challenging not only because there is so much diversity, but also because if their patterns are driven by these more local effects and are less influenced by historical processes (due to faster evolution and life cycles). If so, macroecological patterns may be different from those found for mammals and birds.
In general, limited ecological or environmental data usually forces researchers to adopt one of two strategies when analysing insects: accumulate more local data hoping that they can eventually be extrapolated to larger scales, or use the available, limited data to try to make progress. Unfortunately, there is no objective way to decide which path to take, but it should be remembered that perfect data do not and cannot exist, but they may be good enough to allow us to test models describing how we think nature works. It is helpful to look at past work in insect macroecology to understand how the balance was achieved and evaluate if results are sensible. This leads to our final section highlighting some recent macroecological studies of insects.
Highlights in insect macroecology and conservation biogeography
Here, we present an overview of selected recent papers using insects as model organisms in macroecological research, including analyses of richness gradients, ecogeographical rules, inter-specific relationships, conservation planning and species distribution modelling. A search in ‘Web of Science’ using a limited range of keywords [body size gradient, species richness gradient (entomology subject area only), diversity gradient* (entomology subject area only), range size (entomology subject area only), macroecology, Bergmann* AND insect*, Bergmann’s rule, Rapoport*, bioclim* AND insect*, niche model* AND insect*, Species distribution model* (entomology subject area only), GARP, and MAXENT] and filtering local studies not directly related to insect macroecology provided 363 studies that can be classified into major research fields (bibliography available on request). The most striking feature is that the number of studies has non-linearly increased in the last 10 years, notably after 2005 (Fig. 1). Although we cannot review all of these studies, in the following sections we discuss a few of them. This is obviously not intended to be an exhaustive review of the development of insect macroecology, but hopefully our selection is useful as a guideline for future research and will encourage readers to develop analyses for other insect groups.
Regional richness gradients
Evaluation of broad-scale gradients of species richness is challenging because it requires knowledge, even at coarse scales, of geographic extents of occurrences of species. Overcoming the Wallacean shortfall, some studies not only described patterns for insect richness, but also tested some of the hypotheses explaining such patterns based on new statistical methods. For instance, Lobo et al. (2002) pioneered the use of multiple partial regression in a macroecological context and showed that species richness of Scarabaeidae in France was explained by geographic patterns of climatic variation whereas topography has a small effect (see also Muff et al., 2009 for a recent application at more local scales).
At a much broader scale, Hawkins et al. (2003) showed that, for most groups of organisms, richness can be well explained by energy-water variables, especially AET or PET. Out of the 75 studies included in their analyses, 19 were for insects (ants, butterflies, termites and some beetles). Additional studies have since shown that insect richness is well explained by measures of water-energy dynamics, including butterflies (Hawkins & Porter, 2003a), hawkmoths (Beck et al., 2006a) and dragonflies (Keil et al., 2008). Tests of other climatic-based hypotheses, including the model derived from the metabolic theory of ecology, have been tested for ants (Kaspari et al., 2004; Kaspari, 2005). Hawkins et al. (2007b) performed an analysis of the power of the metabolic model to explain geographical patterns in species richness for many groups of insects, and Ulrich and Fiera (2009) analysed springtails (Colembola) in Europe and also found that seasonality in energy availability drives richness patterns in this group.
A few of the ‘evolutionary’ hypotheses for latitudinal gradients have been tested with insects. Kaspari et al. (2004) tested alternative mechanisms driving higher ant diversification rates (speciation – extinction) in the tropics, and found that richness from samples worldwide were better explained by a model in which energy increases speciation (rather than a model in which energy increases abundances and buffers extinction). Hawkins and DeVries (2009), on the other hand, recently explained patterns of richness in North American butterflies based on Wiens and Donoghue’s (2004) tropical conservation model, in which species in older, more basal subfamilies tend to retain more sub-tropical and tropical distributions due to presumed niche conservatism. This contrasts with more derived groups that have adapted and diversified in colder environments (see also Löwenberg-Neto et al., 2008, for a similar phylogenetic approach with muscid flies). Niche conservatism also seems to apply to butterflies in other regions as well as North America (Hawkins, 2010).
Other studies have dealt with more specific hypotheses for diversity gradients in the context of biotic interactions and local-habitat associations. For example, Hawkins and Porter (2003b) analysed the congruence of patterns between butterflies and host plants in California to evaluate the potential influence of host–plant richness on broad-scale herbivore richness. Others, such as Malo and Baonza (2002), Sobek et al. (2009) and Brandle et al. (2008) also tested similar pairs of diversity gradients to look for coupled explanations based on extrinsic (climate) and intrinsic (biotic) effects.
Because of improvements in data availability for some organisms and computational capacity, there is currently a tendency to generate and interpret global biodiversity patterns. This allows us to evaluate regional idiosyncrasies and explore global explanations, which is important both for understanding macroecological processes and for conservation biogeography. For example, Eggleton et al. (1994) and Price et al. (1998) pioneered this type of study and analysed global patterns of termites and insect galling species, respectively. Although not fully global, Hawkins (2010) analysed butterfly species richness gradients in North America (including Mexico), Chile, Europe, northwest Africa, South Africa, Transbaikal Siberia and Australia.
An alternative approach to macroecology is to treat species, rather than localities, as the unit of analysis. This was the initial approach to macroecology in the 1980s, which was mainly concerned with evaluating statistical distributions of macroecological traits (such as body size, local abundances and geographic range size) and the relationships between them. This approach may be easier in some aspects, since it does not require detailed knowledge of species ranges and may be based on a (hopefully unbiased) sample of species for which data are available.
Using this approach, Rundle et al. (2007) found that geographic range size of damselflies is correlated with wing size in North America, implying that evolution of this trait has played an important role of range expansion and contraction and, consequently, in faunal shifts tracking climate changes. Moreover, this type of analysis may shed light in the debate on the relative roles of stochastic (neutral) and deterministic (niche) processes structuring assemblages. Beck et al. (2006b) evaluated the effects of scale, grain size and phylogeny in the evaluation of geographic range properties of hawkmoths in Southeast Asia.
Other recent studies have examined the relationships between macroecological variables, including the relationship between geographic range size, occupancy and local abundances. Siqueira et al. (2008) showed that body size - abundance relationship for Neotropical chironomid assemblages, despite following the well-known constraint envelope pattern (see Brown, 1995), is temporarily unstable. Heino (2008), on the other hand, showed that the relationship between abundance and occupancy is positive for boreal aquatic insects and is stable. Additional work on more groups and other parts of the world is clearly needed for us to know which pattern is more widespread.
Altitudinal gradients parallel broad-scale latitudinal patterns in many aspects, but in some sense they are easier to evaluate because they are based on shorter geographic spans (although collecting along altitudinal bands presents its own challenges). Some recent analyses also merit mention in the context of inter-specific relationships, especially in the context of ecogeographical rules in an altitudinal gradients. Although these rules have also been investigated in a geographical context (see Hawkins & Lawton, 1995; Diniz-Filho & Fowler, 1998), it is possible to use the equivalence between latitudinal and altitudinal patterns to investigate them. For example, Kubota et al. (2007) and Brehm and Fiddler (2004) analysed variation in body size (Bergmann’s rule), whereas Bektov and Mikhail (2009) showed that range sizes of mayflies expand towards higher altitudes, supporting Rapoport’s rule.
When geographic ranges or related metrics are known for a set of species, and conservation targets and goals can be defined, it is possible to use a wide range of techniques for systematic conservation planning (Margules & Pressey, 2000). Insects are important for many ecosystem functions as part of interaction networks (as pollinators, for example), but they are rarely included as conservation targets, even when more appealing and larger plant species that depend on them are. A few examples, such as Proaches and Cowling (2006), show how insects can be useful for such analyses. Fattorini (2009) used butterfly data to evaluate conservation priorities in Europe and discussed the problem of establishing conservation hotspots based on endangered species. Dos Santos et al. (2008) showed how evaluation of distribution patterns can be used to establish which areas lack data and should be considered as research priorities. The relationship with patterns of human occupation must also be taken into account, and Pautasso and Polwell (2009) recently showed that aphid richness in Europe is positively correlated with human population density.
Niche modelling and their applications
Niche modelling is becoming a very popular approach for generating models of geographic range and is being used to address a large set of ecological and evolutionary questions (e.g., Pearson & Dawson, 2003; Jiménez-Valverde et al., 2008). Insects have been used in many cases, including studies focused on the effects of climate change on species distributions (e.g., Hering et al., 2009; Diniz-Filho et al., 2010) and achieving specific conservation goals (e.g., De Groot et al., 2009). However, although it is tempting to quickly generate maps describing species distributions, this can be criticised on many grounds, including the fact that models did not take into account the particular ecological characteristics of the taxa, historical barriers to dispersal, or the problem of biased geographical sampling (Hortal et al., 2008, 2010). Even so, NM can also be a powerful tool, and further developing this approach will be a popular research avenue over the next decade. In this context, both conceptual and methodological improvements to NM can be tested using insects. An initial potential improvement is to use more specific predictor variables, rather than general climate (temperature and precipitation), as done by Lippitt et al. (2008); see also Klok et al., 2003).
One of the most important discussions of NM is how the assumption of a species’ current niche requirements can be used in modelling and, consequently, the reliability of the projections of future distribution patterns under climate change. Initial reports recorded observed range shifts tracking climate change for many groups of organisms (see Thomas, 2001; Parmesan, 2006). NM may also reveal that climatically suitable areas may change even when distributional limits do not shift, as recently demonstrated by Diniz-Filho et al. (2010) for the Neotropical grasshopper Tropidacris cristata.
More interestingly, insects are an excellent group for testing local adaptation to changing climates and niche shifts because of their relatively rapid response to environmental changes. Indeed, one of the first analyses showing local adaptations to climate changes was done with Drosophila (Rodríguez-Trelles et al., 1998). Niche dynamics may also be evaluated during biological invasions, although the topic is still under strong debate. Fitzpatrick et al. (2007) reported that fire ants shifted their range after invading North America, although Peterson and Nakazawa (2008) attributed this to a methodological artefact (but see Fitzpatrick et al. (2008) for a reply, and Roura-Pascal et al. (2009) for an independent analysis). A recent analysis also found evidence that the drosophilid Zaprionus indicus shifted its climate niche after invading India from Africa (Da Mata et al., 2010).
We hope the above essay stimulates the use of insects as model organisms in macroecology and conservation biogeography. Beyond testing previously described patterns and underlying processes, insects may be ideal organisms to develop new approaches in macroecology. Among the many possibilities, one of the most attractive is that some insect groups may be quite useful for combining observational and experimental approaches (because they can be relatively easily manipulated under laboratorial conditions), which may be important for testing evolutionary influences on macroecological patterns in response to climate change. These can be extrapolated to other groups of organisms more difficult to manipulate, or at least used to generate new sets of complementary predictions. In sum, although the effort required to generate high quality data for many groups of insects may be great, we believe the potential rewards make it worth the effort.
We thank Joaquin Hortal and an anonymous reviewer for suggestions that improved the manuscript. Work by J. A. F. Diniz-Filho and P. De Marco Jr. has been continuously supported by CNPq productivity fellowships.