Predicting the post-fire responses of animal assemblages: testing a trait-based approach using spiders

Authors

  • Peter R. Langlands,

    Corresponding author
    1. School of Animal Biology MO92, University of Western Australia, Crawley, WA 6009, Australia
      Correspondence author. E-mail: peter.langlands@grs.uwa.edu.au
    Search for more papers by this author
  • Karl E. C. Brennan,

    1. Western Australian Department of Environment and Conservation, PO Box 10173, Kalgoorlie, WA 6430, Australia
    Search for more papers by this author
  • Volker W. Framenau,

    1. School of Animal Biology MO92, University of Western Australia, Crawley, WA 6009, Australia
    2. Department of Terrestrial Zoology, Western Australian Museum, Welshpool DC, WA 6986, Australia
    Search for more papers by this author
  • Barbara Y. Main

    1. School of Animal Biology MO92, University of Western Australia, Crawley, WA 6009, Australia
    Search for more papers by this author

Correspondence author. E-mail: peter.langlands@grs.uwa.edu.au

Summary

1. Developing a predictive understanding of how species assemblages respond to fire is a key conservation goal. In moving from solely describing patterns following fire to predicting changes, plant ecologists have successfully elucidated generalizations based on functional traits. Using species traits might also allow better predictions for fauna, but there are few empirical tests of this approach.

2. We examined whether species traits changed with post-fire age for spiders in 27 sites, representing a chronosequence of 0–20 years post-fire. We predicted a priori whether spiders with ten traits associated with survival, dispersal, reproduction, resource-utilization and microhabitat occupation would increase or decrease with post-fire age. We then tested these predictions using a direct (fourth-corner on individual traits and composite traits) and an indirect (emergent groups) approach, comparing the benefits of each and also examining the degree to which traits were intercorrelated.

3. For the seven individual traits that were significant, three followed predictions (body size, abundance of burrow ambushers and burrowers was greater in recently burnt sites); two were opposite (species with heavy sclerotisation of the cephalothorax and longer time to maturity were in greater abundance in long unburnt and recently burnt sites respectively); and two displayed response patterns more complex than predicted (abdominal scutes displayed a U-shaped response and dispersal ability a hump shaped curve). However, within a given trait, there were few significant differences among post-fire ages.

4. Several traits were intercorrelated and scores based on composite traits used in a fourth-corner analysis found significant patterns, but slightly different to those using individual traits. Changes in abundance with post-fire age were significant for three of the five emergent groups. The fourth-corner analysis yielded more detailed results, but overall we consider the two approaches complementary.

5. While we found significant differences in traits with post-fire age, our results suggest that a trait-based approach may not increase predictive power, at least for the assemblages of spiders we studied. That said, there are many refinements to faunal traits that could increase predictive power.

Introduction

Ecologists have long sought to understand and predict the consequences of disturbances (flooding, grazing, habitat fragmentation, fire, etc.) on ecosystems and assemblages of plants and animals (Connell 1978; Lavorel et al. 1997). A formidable challenge is elucidating how individual species of plants and animals might respond to a particular disturbance, especially for highly diverse groups, where taxonomic knowledge is poor and/or there is an absence of knowledge on the autecology of individual species. Therefore, studying solely the responses of individual species is unlikely to lead to a rapid increase in predictive power for unstudied species because it only generates location and taxa-specific information (Weiher & Keddy 1995). However, an understanding of which traits determine a species’ response to disturbance may lead to generalizations, which can predict how other species with the same traits will respond (Lavorel & Garnier 2002; McGill et al. 2006). We define a trait following McGill et al. (2006) as ‘a well-defined, measurable property of organisms, usually measured at the individual level and used comparatively across species’. Ideally, we might hope to be able to collect an animal from any locality and from only its morphological and behavioural traits predict the species’ change in abundance following disturbance.

A species-traits approach has yielded impressive advances in increasing predictive power for plants (Shipley, Vile & Garnier 2006; Keith et al. 2007). A prominent example is the marked distinction in post-fire responses between seeders vs. resprouters (Noble & Slatyer 1980). However, correlations between traits cause most species to occur only within a limited portion of the total multidimensional trait-space (Hubbell 2001). Therefore, to develop a better mechanistic understanding of post-fire responses it is important to identify groups of traits that are intercorrelated and might be functional (strongly influence organismal performance).

The adoption of a trait-based approach for fauna and fire lags behind that for plants (notable exceptions are Jonas & Joern 2007; Moretti et al. 2009; Moretti & Legg 2009). Using previous approaches for fire, we adapt the vital attributes (Noble & Slatyer 1980) and critical life cycles approaches (Whelan et al. 2002) to classify traits under five categories: survival, dispersal, reproduction, resource-utilization and microhabitat occupation. The interest of our search is convergent traits (Grime 2006; Pillar et al. 2009). That is, because of environmental filters, the species within a community and the traits they possess are more similar or under-dispersed (Weiher & Keddy 1995). We use current theories and functional understandings (existing mechanisms) of traits to predict how each trait might respond with increasing post-fire age. We then compare these qualitative predictions to the observed results.

Several methods explore potential links between species traits and environmental conditions. These can either test the link indirectly (e.g. emergent groups), or directly (e.g. fourth-corner or RLQ analysis). Emergent groups are suites of species selected for sharing similar traits (Lavorel et al. 1997). The fourth-corner is an approach that combines information on species traits and environmental characteristics via a matrix of species abundances. While RLQ analysis is complementary to the fourth-corner, it focuses primarily on ordination and interpretation, rather than testing the significance of relationships (Dolédec et al. 1996; Legendre, Galzin & Harmelin-Vivien 1997). Most applications use an indirect approach, partly because of the initial limitation of the fourth-corner to presence–absence data only. However, the fourth-corner approach now incorporates abundance data and may yield additional insights compared to emergent groups (Aubin et al. 2009). That said, emergent groups easily incorporate links between traits, but for the fourth-corner, this remains an outstanding problem (Dray & Legendre 2008). Inclusion of linked traits into fourth-corner analyses may increase the chance of type I error (Aubin et al. 2009) and ignores biological reality. Ultimately, a direct approach that can incorporate linked traits is required. In the interim, however, it is unclear whether researchers should follow a direct or indirect approach, as only a single study has explicitly compared these approaches for the same data (see Aubin et al. 2009).

Despite the usefulness of the approaches described earlier, some problems remain. Neither the fourth-corner or emergent groups can yet incorporate potential spatial autocorrelation of sites or phylogenetic signal between species (See Dray & Legendre 2008, p. 3411). For faunas that are poorly known taxonomically and where autecological information is scant, there will be a reliance on inferring behavioural traits at higher taxonomic levels using information gleaned from related taxa. Thus, greater caution will be required when considering behavioural traits compared to morphological traits measured directly from each taxon. Even with these unresolved issues, however, the fourth-corner and emergent groups approaches represent the best techniques currently available and offer new insights into trait–environment relationships. As such, and because the search for a responsive set of traits for fauna is still in its infancy, to assist future workers we have included both morphological and behavioural traits.

Spiders are an interesting group for considering linkages between post-fire environments and traits because they show marked changes in species composition with fire (Buddle, Spence & Langor 2000; Brennan et al. 2006; Langlands, Brennan & Pearson 2006) and possess a diverse array of potentially influential traits. Spiders represent a range of body sizes; our species differed by an order of magnitude (0·6–9·4 mm). While spiders are unable to fly, many families possess the ability to drift passively (‘ballooning’) by using extensions of silk (Bell et al. 2005). While spiders generally have soft abdomens, some species also possess morphological adaptations that reduce desiccation, such as the presence of hard plates (scutes) on the abdomen or heavy sclerotisation of the cephalothorax (Main 1984). While almost all spiders are obligate carnivores, there is a vast array of different hunting strategies, which have attracted interest as feeding guilds (Wise 1993; Uetz, Halaj & Cady 1999). Additionally, while many have a broad diet others show increasing specialization. For example, some species are specialist predators of ants (myrmecophagous) or other spiders (araneophagous) (Polis & Yamashita 1991).

In this study, we examine if the species traits of spiders can make better predictions of post-fire responses using both the fourth-corner and emergent groups approaches. We address four questions: 1) Are individual traits of spiders linked with post-fire age? In short, we hypothesized that the abundance of spiders with the following traits would decrease with increasing post-fire age: larger body size, heavy sclerotisation (cephalothorax plus abdominal scutes), greater dispersal ability (ballooning) and using burrows. We hypothesized the abundance of spiders with the traits of slow maturation, short seasonal activity, diet specialization and a flattened body would increase with post-fire age. In our methods, we describe how each of these traits might act to structure assemblages of spiders following disturbance. 2) If significant relationships with traits are found, are these traits intercorrelated? and if so, 3) Do combinations of these intercorrelated traits have significant associations with post-fire age? Finally, 4) Are emergent groups of spiders linked with post-fire age and what additional insights does the fourth-corner approach offer over this method?

Materials and methods

Study location and design

The study site was the proposed Lorna Glen (Matuwa) Conservation Park in central Western Australia (26°13′33′′S, 121°33′28′′E). Satellite images were used to select sites representing six treatments of post-fire ages: zero (four sites), 6 months (three sites), 3 years (five sites), 5 (five sites), 8 (five sites) and 20+ (five sites) years since the last fire. See Appendix S1 and P.R. Langlands, K.E.C. Brennan & B. Ward (unpublished) for further details of study location, sites and fires.

Sampling and identification of spiders

Ground active spiders were sampled continuously over a 9-month period (30 April 2006–18 January 2007) using five pitfall traps (2 L jars with a mouth of 82 mm) per site. Traps were exchanged every 3 months. Male spiders were identified to species level. Owing to taxonomic difficulties of correctly matching females and juveniles to males for many of the undescribed taxa, the results presented here are only for males. All specimens are deposited in the Western Australian Museum.

Traits

We scored morphological and behavioural traits where there was a potential mechanism to explain post-fire responses. As noted in the introduction, owing to limited autecological information, behavioural traits were inferred at the family or sub-family level. We included only behavioural traits that we were confident had sufficient information available for such inferences. As we measured morphological traits directly from the specimens, these data are robust and independent of whether they have been named by taxonomists.

We classified these traits into five key categories: survival, dispersal, reproduction, resource-utilization and microhabitat occupation. The placement of traits within these categories is not mutually exclusive because burrowing could be placed under survival or microhabitat, but we consider the categories a useful framework on which to build. Those working with spiders in other areas or indeed other groups may find that some traits such as flat bodies specialized to live under eucalypt bark are not relevant, but the categories may help to find other relevant traits. A summary of each trait follows with a full list of species and traits in the online supplementary material (Table 1, Appendix S2).

Table 1.   Test of the direct link between the species traits of spiders and increasing post-fire age from a combined fourth-corner analysis (permutation models 2 and 4) using 9999 permutations. For each trait are the states scored, the test statistic and the value of the test statistic for each post-fire age with significance indicated (*P < 0·05, **P < 0·01)
TraitsScored statesTest statisticPost-fire age (years)
00·535820
Survival
 BurrowingYes/noChi15·9**0·73·00·00·43·2
 Body size (length)Mean carapace length (mm)Pseudo-F21·6**0·04·50·10·12·4
 Heavy sclerotisation of the cephalothoraxNormal/sclerotisedChi2·00·12·41·41·97·0*
 Abdominal scutesAbsent/partial/whole surfaceChi4·01·03·78·6*0·57·7
Dispersal
 BallooningYes/noChi7·3*1·23·15·3*2·30·5
Reproduction
 Time to maturity<3 years/>3 yearsChi6·6*0·03·00·40·11·7
 Phenology3/6/9 monthsChi1·33·55·10·40·71·6
Resource-utilization
 Hunting strategyActive/burrow/sit & wait/aerial webs/terrestrial websChi16·1**1·65·10·01·23·4
 Diet specialization (ants)Yes/noChi0·21·30·00·31·30·1
Microhabitat
 Flattened bodyYes/noChi0·40·10·50·00·00·5

Survival

Body size (measured as carapace length): Having soft abdomens, spiders are vulnerable to desiccation, particularly small spiders (Main 1984). As body size increases, the surface area to volume ratio decreases. Therefore, the body size of arthropods was predicted to increase with aridity (Remmert 1981), as shown for spiders in Europe (Entling et al. 2010). We hypothesized that species with a larger body size and therefore a decreased surface area to volume ratio may be better able to cope with the greater variability in temperature and humidity associated with recently burnt habitats. Indeed, the large body size of lycosids might be an adaptation to withstand larger temperature variation in open habitats (Coyle 1981). As an estimate of body size for each species, we measured the mean carapace length of five individuals. Abdomen length was not included as spider abdomens are soft and can change greatly in size when preserved.

Burrowing: The intensity of heat from fires decreases rapidly with increasing soil depth (Whelan 1995). Some species of spiders can construct burrows in the ground, creating a microhabitat that shields them from direct disturbance by fire. Species exhibiting this behaviour might have a greater chance of surviving the passage of fire relative to non-burrowing species and therefore occur in greater abundance in recently burnt areas.

Degree of sclerotisation: The development of thick plates or shields on soft body parts is considered an adaptation which reduces water loss, particularly for small-bodied species which, when compared with larger-bodied species, have a higher surface area to volume ratio (Main 1984). We scored two traits for sclerotisation. Heavy sclerotisation of the cephalothorax (head-thorax) was scored as present (the cephalothorax was a rigid case, often pitted and dark red) or absent. Secondly, we scored the presence of scutes (hard plates) on either the dorsal or ventral side of the abdomen as absent, partial coverage or full coverage.

Dispersal

Ballooning: The ability to disperse would be advantageous in colonizing recently disturbed habitats. Previous studies of spiders and some insect groups show that dispersal ability is important for re-colonizing disturbed sites (Crawford, Sugg & Edwards 1995; Moir et al. 2005). Many species of spiders can disperse over large distances by ballooning, releasing silk threads as juveniles to catch wind currents, while others can disperse only by walking (Main 1984). Species were scored as ballooners if there was a published record of ballooning for the genus or family (Bell et al. 2005). While species may possess the ability to balloon, the distance can vary greatly along a continuum from only a few metres to many kilometres (Bell et al. 2005). To simplify the analysis, we treated ballooning as a binary response.

Reproduction

Time to maturity: Species can be divided into two broad life-history categories, r- or K- selected (Southwood 1977). Although not without limitations, this can provide a useful dichotomy of the life history of species (Polis & Yamashita 1991). One aspect of these categories is time to maturity, short for r- selected and long for K- selected species. Species with short maturity times would have an advantage in colonizing and reproducing in recently disturbed areas. The time taken to mature can vary widely between species of spiders from one to several years (Main 1984). To differentiate between species, we scored maturation time as less than or greater than 3 years. We did not attempt a finer separation in time to maturity ages because of the lack of natural history information.

Phenology: The reproductive activity of spiders can be highly seasonal with males sexually mature and actively searching for females for limited periods (Merrett 1967; Framenau & Elgar 2005). Fire may disadvantage those species with highly specific environmental triggers for mating activity compared with species that can mate anytime throughout the year. As a crude measure of phenology, species were scored according to the length of time over which they were collected: 3, 6 or 9 months.

Resource-utilization

Hunting strategy: Spiders are almost all obligate predators, but the strategies they use to catch prey differ substantially. Some are active hunters, others may sit-and-wait, while many use different types of webs to capture prey (Turnbull 1973; Wise 1993). We used taxonomic literature and expert opinion to allocate species to one of five hunting strategies: active hunters, sit-and-wait hunters, burrow-dwelling sedentary hunters, terrestrial webs and aerial webs.

Diet specialization: While as a group spiders are viewed as polyphagous predators, some have specialized feeding habits (Marc & Canard 1997). For example, zodariids are specialist ant predators (Jocqué 1991). Species with restricted diets may be limited to habitats that support suitable supply of the required food. Hence, diet specialization may increase with time since disturbance. We scored species that are known ant specialists as a binary trait. We did not score species that mimic ants as potential predators of ants, because mimicry can also evolve to evade predators (see Cushing 1997).

Microhabitat

Flattened body: Several groups of Australian spiders exhibit an extremely flattened body shape, which might have evolved with the rapid speciation of Eucalyptus trees and the crevice habitat provided under the peeling bark (Żabka 1991). In instances where the bark is thick and fire intensity is low, by sheltering under bark some spiders might survive the passage of fire (Brennan, Moir & Wittkuhn in press). However, for species of eucalypt with thin bark (or for trees with thicker bark but where fire intensity is high), most of the bark may be consumed by the fire or slough off in ensuing months. Thus, there may be limited microhabitat for crevice-dwelling spiders for several years. At our study site, fire resulted in a loss of bark crevices and so a flattened body might not confer a survival advantage during burning and the lack of suitable crevices post-fire may inhibit recolonization.

Statistical analysis

To test the link between environmental conditions and individual spider traits directly, we used the fourth-corner analysis (Legendre, Galzin & Harmelin-Vivien 1997; Dray & Legendre 2008). This analysis uses three data matrices (Matrix R: environment by sites, Matrix L: species by sites, Matrix Q: species by traits) to test directly the link between environmental characteristics and species traits via the observed abundance of species (Appendix S3). We used the two-stage permutation test recommended by Dray & Legendre (2008), which combines the results from permutation models 2 and 4 (see Dray & Legendre 2008; : Figs 2 and 3). Previous work indicated that the post-fire age of sites in years was a good predictor of the spider community (P.R. Langlands, K.E.C. Brennan & B. Ward, unpublished); we therefore chose this variable to represent the matrix R (environmental characteristics). Matrix R was coded as dummy variables following Aubin et al. (2009), which allows for nonlinear patterns with increasing post-fire age. Matrix L was represented by a matrix of abundances of spiders that were fourth root transformed to reduce the dominance of abundant species. A matrix of the ten traits scored for each species represented Matrix Q, and all traits were coded as qualitative variables except for body size, which was quantitative. Analyses were conducted using the ‘fourth-corner’ function with 9999 permutations in the R package ade4 (Dray & Dufour 2007).

Figure 2.

 Principal coordinate analysis of the similarity based on ten traits between 179 species of spiders (Table 1). Vectors represent the Spearman correlation of traits, with the length and direction indicating the relationship with PCoA axes. Abbreviations: PCoA = principal coordinate analysis, Phen = Phenology, AbScute = Abdominal Scutes, Hunt = Hunting strategy, Cephalothorax = Heavy sclerotisation of cephalothorax. Six dominant families display unique symbols with the remaining species pooled into one symbol for clarity. The third PCoA axis explained 21·9% of the total variation (91·7% for all three axes) and correlates strongly with Phenology (Appendix S4).

Figure 3.

 Mean abundance (4th root) ± SE for emergent groups which had a significant response with post-fire age (years). For the traits and species of each group see Table 3. Unlike letters indicate significant differences in pair-wise tests.

To examine which traits were linked, we calculated a similarity matrix between all species based on the ten traits. We used the Gower coefficient because it can incorporate both quantitative and qualitative variables. A principal coordinate analysis (PCoA) was then employed on the resulting matrix to simplify the intercorrelated traits to a reduced number of axes. We chose this method as it allows any similarity measure. To relate the resulting axes back to the original traits, we used vector overlays of the correlation between axes and traits.

To test the link between environment and combined traits directly, and address the inability of the fourth-corner to incorporate linked traits, we devised the following approach. We incorporated the axis scores for each species from the PCoA into a fourth-corner analysis as matrix Q (Appendix S3). We recognize that this method, although theoretically sound, is somewhat crude as it contains multiple steps. We again used the two-stage permutational test (models 2 and 4).

To test the link between environmental conditions and traits indirectly, we used emergent group analyses based on the similarity matrix constructed from the ten traits (Appendix S3). This method is indirect as it requires two steps (Dray & Legendre 2008). Traits and species abundances (matrices Q and L) are first linked, and then related to environmental conditions (Matrix R). Ward’s method was used to cluster species and the resulting dendrogram was arbitrarily split into five emergent groups. The abundance of species within each emergent group was then summed for each site. Permutational analysis of variance (permanova) was then used to test for significant differences in the abundances of groups between post-fire ages. permanova is highly robust as it uses permutation for significance testing and any dissimilarity measure may be used (Legendre & Anderson 1999). We used Euclidean distance with unrestricted permutation of the raw data, type III sums of squares and 9999 permutations. As argued by Moran (2003), we have not corrected for multiple analyses. While none of the above methods can currently incorporate spatial autocorrelation, it has been shown elsewhere that spatial location explained less than 4% of the variation in species richness or composition of our spider assemblages (Langlands 2010). Analyses were performed using permanova+ (Anderson, Gorley & Clarke 2008; version 1.01) and R (R Development Core Team 2009).

Results

Our data set comprised 4234 male specimens representing 179 species of spiders from 24 families, of which only 32 species (=17·9%) had scientific names. Previous analyses indicated significant differences in the species composition of spiders with increasing post-fire age (P.R. Langlands, K.E.C. Brennan & B. Ward, unpublished; Appendix S4). Matching these differences in species composition were shifts in the abundance of species with particular traits among post-fire ages. The fourth-corner analysis showed significant links between post-fire age and spiders for seven of the ten traits measured (Table 1). There were significant results within all of the groups of traits except microhabitat occupation. There were more large-bodied individuals at recently burnt sites (0-year-old) (Fig. 1a). Abundance of species with heavy sclerotisation of the cephalothorax increased in 8-year-old sites and was significantly higher in 20-year-old sites (Fig. 1b). Abdominal scutes displayed a U-shaped pattern with significantly fewer scutes in 5-year-old sites and high abundances in recently burnt and long unburnt sites (Fig. 1c). The abundance of spiders that can disperse via ballooning was lowest in 0 aged sites and highest in 5-year-old sites (Fig. 1d). There were significantly more long-lived individuals in 0-year-old sites, compared to all other ages (Fig. 1e). Likewise, both burrow-dwelling sedentary hunters and burrowing spiders were in significantly greater abundance in 0 aged sites compared to all other ages (Fig. 1f, g). However, within a particular trait there was usually only one post-fire age that differed significantly from all others.

Figure 1.

 Mean abundance (4th root) or ratio, ±SE, with post-fire age (years). Asterisk denotes significant ages based on fourth-corner analysis not graphs (Table 1).

The first three axes of the PCoA explained 91·7% of the variation between species based on the ten traits. High values along axis 1 were associated with the presence of partial abdominal scutes, an active hunting strategy and to a lesser extent, a more specialized diet (ants), and not using burrows (Fig. 2, Appendix S5). High values along axis 2 were strongly associated with high dispersal ability (ballooning) and to a lesser extent hunting from terrestrial webs. Axis 3 was associated with phenology. Plotting of unique vectors for each trait onto the PCoA ordination showed that ballooning and hunting using a terrestrial web were strongly intercorrelated (Fig. 2). Similarly strongly intercorrelated were heavy sclerotisation of the cephalothorax, diet specialization on ants, the presence of full abdominal scutes and active hunting. Correlated to a lesser extent were body length, time to maturity, and burrowing. Ordination also showed a clear influence of phylogeny as many species within a family clustered together in trait space. For example, five of the six species of Nemesiidae clustered very tightly in the lower left corner of the ordination (Fig. 2). Other families such as the Zodariidae clustered together more loosely in trait space.

Using the PCoA axes in a fourth-corner analysis found PCoA axes 1 and 2 to be highly significant with the 0-year-old sites (Table 2). The negative correlation of 0-year-old sites with axis 1 indicated that recently burnt sites had a greater abundance of species without abdominal scutes and a lesser abundance of species with partial abdominal scutes or an active hunting strategy. It also indicated to a lesser extent that 0-year-old sites had greater abundance of species using burrows and a lesser abundance of species with a specialized diet (ants) (Appendix S5). For axis 2, there was a positive correlation; hence, the 0-year-old sites had fewer individuals that disperse via ballooning.

Table 2.   Test of the direct link between combinations of traits and post-fire age. Principal coordinate analysis (PCoA) was used to reduce the ten species traits of spiders (with varying degrees of intercorrelation) to three axes, which were then used as matrix Q in a fourth-corner analysis
PCoA axesPost-fire age (years)Variation (%)Cumulative variation (%)
00·535820
  1. Values represent a pseudo-F statistic. For details of axis correlation with traits, see Fig. 2 and Appendix S4. **P < 0·01.

111·28**0·144·700·440·172·8237·537·5
29·48**1·664·754·142·190·4332·369·8
30·020·150·650·220·080·1521·991·7

For the indirect analysis, five emergent groups of species chosen from the cluster analysis contained 12–65 species each. Several families had species within multiple groups (Table 3, Appendix S6). Subsequently, permutational analysis of variance (permanova) found abundances of only three of the groups displayed significant responses to post-fire age (Groups III, IV and V, Fig. 3). Group III consisted of small to large, active hunters, which were mostly myrmecophageous and non-ballooning, with a complex pattern in abundances with post-fire age (d.f. 5·21; pseudo-F = 2·89; P = 0·04). Group IV were small, active hunters with full scutes, predominately Oonopidae and were more abundant in recently burnt and long unburnt sites (d.f. 5·21; pseudo-F = 5·17; P = 0·004). Group V consisted mainly of mygalomorphs and lycosids, which were more abundant in 0-year-old sites (d.f. 5·21; pseudo-F = 6·93; P = 0·002, Fig. 3).

Table 3.   The five groups of species selected from cluster analysis with the number of species, shared traits and taxa for each group
Cluster groupsNumber of speciesShared traitsTaxa
  1. Note: Species from some families are within multiple groups. For each family, superscripts denote the fidelity of species within a group, A: all species in a family, S: Some species, F: Few species.

I65Small to large, with no scutes, active, sit-&-wait hunters and terrestrial websClubionidaeA, DesidaeA, DictynidaeA, GnaphosidaeS, LinyphiidaeA, LycosidaeF, MiturgidaeA, OxyopidaeA, PhilodromidaeA, ProdidomidaeF, SalticidaeF, SparassidaeA, ThomisidaeA, ZoridaeA
II48Small to Medium, some with flattened bodies, ballooning, mostly active hunters with some terrestrial websCorinnidaeF, GnaphosidaeS, PholcidaeA, SalticidaeS, TheridiidaeS
III27Small to large, ant eaters, active hunters and mostly non-ballooningCorinnidaeF, LamponidaeS, ProdidomidaeF, ZodariidaeA
IV27Small, with full scutes and active huntersCorinnidaeF, GnaphosidaeF, LamponidaeF, LiocranidaeA, OonopidaeA, SalticidaeF, TheridiidaeF
V12Large, without sclerotisation, mostly slow maturing and burrowing behaviourNemesiidaeA, SegestriidaeA, LycosidaeS

Discussion

We found significant patterns in the species traits of spiders with post-fire age, namely recently burnt sites (0-years-old) had more individuals with a large body, limited dispersal ability (i.e. presumed unable to balloon), burrowing habit and an ambushing hunting strategy from burrows. Although a majority (7 of 10) of our traits differed among post-fire ages, within traits there were few significant differences between post-fire ages. In four instances of seven, only the recently burned sites differed from all other ages (Fig. 1). Furthermore, the patterns of difference did not consistently match our predictions. Of the seven spider traits that were significantly different among post-fire ages, three followed predictions (body size, hunting strategy and burrowing behaviour), two were opposite (heavy sclerotisation of cephalothorax & time to maturity) and two displayed more complicated response patterns than predicted (abdominal scutes & ballooning). Given the limited research effort into the trait-based approach for fauna and fire, it is not surprising that our results are not as clear as found recently for plants (e.g. Keith et al. 2007; Aubin et al. 2009).

Are individual traits linked with post-fire age?

Survival

All four traits within the survival category had significant links with post-fire age; however, as noted earlier, only two of these followed predictions. Body size was expected to influence survival via differences in desiccation rate, with larger bodied spiders having an advantage in recently burnt and open habitats. In Europe, the body size of female Linyphiidae increased with increasing disturbance from flooding; however, Lycosidae were larger in less disturbed sites (Lambeets et al. 2008). However, body size often correlates with other traits (Peters 1983; Gaston & Blackburn 2000), and in our study it was correlated with burrowing. It is therefore unclear whether body size or burrowing is functional.

The response of spiders with sclerotisation of the cephalothorax was opposite to the prediction of greater sclerotisation immediately post-fire. Thus, the mechanism for heavy sclerotisation of the cephalothorax and abdomen also requires attention. It is unclear why there would be more spiders with this trait in long unburnt sites. The response of abdominal scutes was similar with high abundance in long unburnt sites, but there was also relatively high abundance in recently burnt sites. Most of the species with full scutes were from the Oonopidae, which are leaf-litter dwellers. At our study sites, the proportion of ground covered by leaf-litter followed a U-shaped pattern and remained high in the recently burnt sites as scorched trees shed leaves and bark (P. R. Langlands, unpublished). Oonopid spiders selecting post-fire ages with more leaf-litter may explain why spiders with full scutes were more common in recently burnt and long unburnt sites. If so, the reduction in desiccation that spiders receive from possessing full scutes may not act directly to influence changes in the composition of species of spiders with post-fire age.

Dispersal

Contrary to our predictions, and to previous studies of primary and secondary successions, we found spiders that had greater dispersal ability via ballooning most abundant in 5-year-old sites, rather than those burnt recently. Although clearly different to disturbance by fire, ballooners dominated early colonists during primary succession following glaciations and volcanic eruptions (Crawford, Sugg & Edwards 1995; Hodkinson, Webb & Coulson 2002), and for two spider families in areas frequently disturbed by flooding (Lambeets et al. 2008). Perhaps this inconsistency relates to two things. First, although we were able to discriminate between ballooning and non-ballooning species, we were unable to consider differences in ballooning or walking ability. Secondly, patch size might play an important role in determining the influence of traits associated with dispersal ability. Logistical constraints meant that our recently burnt sites were only 300 by 300 m, so dispersal by walking may have been sufficient for colonization. It may be beneficial to incorporate the walking abilities of non-ballooning species.

Reproduction

Time to reach maturity was also opposite to our predictions with a very high abundance of long-lived species in recently burnt sites. This might be because males from long-lived burrowing species (Nemesiidae) survived the fire in situ and then left to find mates and fell into our pitfall traps. This possibly indicates a degree of inertia immediately following burning (see P.R. Langlands, K.E.C. Brennan & B. Ward, unpublished). Another explanation that cannot be excluded is that large wandering spiders are more likely to be caught by pitfall traps in burnt areas with little ground cover (Topping & Sunderland 1992). This may partially explain the significant results found for body size and burrowing in the 0-year-old sites. However, inspection of the raw data indicated clearly that long-lived species did not solely drive the significant 0-year-old result. Two large lycosid species, one of which burrows [Hoggicosa bicolor (Hogg, 1905)] were abundant in the 0-year-old sites. The phenology of species was not significantly different among post-fire ages, although our measure of seasonal activity was rather crude, and data with finer temporal resolution may yet again provide better insights.

Resource-utilization

Although abundance of burrowing ambushers was significantly higher in recently burnt sites, a lack of significant differences between other post-fire ages suggests hunting strategy does not differ markedly with time since fire. However, many studies show that web-hunting spiders require specific habitat architecture and respond to changes in vegetation accordingly (Riechert & Gillespie 1986; Uetz 1991). Therefore, our results for web building spiders may be in part because of our sampling method. As a result of the remote location and low catch rates from hand collecting and vacuuming in this habitat, we were restricted to using pitfall traps (Owen 2004). Pitfall traps sample many of the foliage and web building species (Brennan, Moir & Majer 2004). However, these taxa will probably be underrepresented in our data set and thus constrain our ability to detect post-fire changes in abundance of species utilizing an aerial web for hunting. For our study, it seems that specialized food sources are not limiting with no detectable result for diet specialization on ants. This result is contrary to predictions that niche breadth decreases with increasing time since disturbance (Odum 1969; Brown & Southwood 1983).

Are traits intercorrelated?

Several traits were intercorrelated from different categories of traits (e.g. survival, dispersal, reproduction etc.). Uncovering the functional significance of each trait is critical to developing a better mechanistic understanding of post-fire response patterns.

Methodological issues also demand the consideration of links between traits. Aubin et al. (2009) suggest including highly correlated or linked traits in the fourth-corner analysis could increase the level of Type I errors. Their proposed solution is to test for correlated traits and exclude redundant traits. Our maximum correlation was only 0·69 (burrowing and time to maturity) which we considered moderately high. Our preferred solution, however, was to use PCoA to incorporate axes representing combinations of traits into a fourth-corner analysis.

Are combinations of traits linked with post-fire age?

Fourth-corner analysis showed the composite scores of traits from the PCoA had some significant links with the 0-year-old sites. This more biologically realistic approach detected some of the pattern from the analysis on individual traits; there was lower abundance of heavy sclerotisation of the cephalothorax and higher abundance of burrowers and burrow hunters in 0-year-old sites. However, it did not detect significant links between other post-fire ages and suggested some different patterns for traits. A strong loading of abdominal scutes for axis 1 suggested this trait is significantly less abundant in the 0-year-old sites, compared with 5-year-old sites as identified by the analysis on individual traits. It also suggests diet specialization on ants is less abundant in 0-year-old sites, a finding not found by analysing traits separately. A major problem with this approach however is that a large number of traits correlated with axis 1 made it difficult to interpret the significant result. Interpretation of the second axis is easy, with only ballooning strongly correlated; however, this approach did not detect greater abundance of ballooners in 5-year-old sites.

Are emergent groups linked with post-fire age?

The indirect method of using emergent groups found links between post-fire age and the post-fire response of spiders. Of the five groups defined, three had significant changes in abundance. However, the post-hoc tests found these changes with post-fire age were not linear or simple. The responses of Group IV appear strongly related to abdominal scutes, while group V are related to body size, burrowing and time to maturity. Therefore, the results from the emergent group approach, particularly for these traits, were largely congruous with those of the fourth-corner. However, the significance of traits such as dispersal ability and heavy sclerotisation of the cephalothorax were less clear compared to the fourth-corner approach with individual traits.

A key advantage of emergent groups is that it generates response groups ultimately leading to the creation of functional types (e.g. plant functional types). We strived to select a limited number of groups that were biologically meaningful. However, there is a trade-off inherent with the size and number of emergent groups selected. While selecting more groups of smaller size may increase predictive accuracy and generate more statistically significant responses, this ultimately reduces practicality and the ability to generalize (Noble & Gitay 1996; Keith et al. 2007). Indeed, we note that the three response groups that had significant responses had the fewest species.

Current constraints and future directions

Our natural history knowledge and taxonomic understanding of Australian spiders is still poor for many families (Brennan, Moir & Majer 2004). This means that while we can have strong confidence in morphological traits measured directly from specimens, the quality of data available for behavioural traits may continue to hinder a trait-based approach. Indeed, we have spent much of the discussion of this paper wondering if a finer resolution of detail for behavioural traits would have distinguished more of the post-fire ages. This is somewhat disappointing. It is for faunas where autecological knowledge at species level is poor that the trait-based approach might offer the greatest benefit to develop rapidly a predictive understanding of faunal responses to disturbance. Two other potentially contributing factors also deserve mention. First, Moretti et al. (2009) have questioned whether the response of traits to fire can be generalized between regions where fire history differs at evolutionary time scales. They suggest that for areas with a long history of fire, community level changes in traits may exhibit greater inertia/resilience. Thus, the long history of fire in arid Australia may have contributed to the limited functional responses we observed. Secondly, while searching for convergent traits, the suite of traits we measured may contain divergent traits (limiting similarity). These function to allow species to coexist under competition, rather than under similar environmental conditions (Weiher & Keddy 1995).

As noted in our introduction, our interest was the ability of morphological and behavioural traits to predict changes in the abundance of species following disturbance. Therefore, we focused on traits at the individual level, and chose not to include community level traits such as rarity (e.g. geographical range, habitat specificity). These traits can be powerful for well-known faunas (e.g., Bonte, Lens & Maelfait 2006; Lambeets et al. 2008), but will often be unavailable for poorly known faunas.

Finally, we echo previous calls for statistical advances. The most important of these are the ability to incorporate phylogeny and spatial autocorrelation of sites into the fourth-corner approach (Dray & Legendre 2008; Aubin et al. 2009).

Conclusions

We set out to test, using spiders, if the species traits of fauna might increase the predictive power of post-fire responses. Ideally, we might hope to be able to collect a spider from a location and predict its change in abundance following fire from only morphological traits measured from specimens and inferred behavioural traits. We found significant relationships between species traits of spiders associated with survival, dispersal, reproduction and resource-utilization and post-fire age. Both a direct and an indirect method of assessing links between traits and the environment showed significant links, and we consider the combination of the two approaches complementary. Overall, while our analysis suggests caution over the predictive power of a trait-based approach of faunal responses post-fire, it is still too early to assess its full potential. We look forward to further studies using novel data sets, refined traits and experimental studies that test underlying mechanisms.

Acknowledgements

We thank R. Raven, B. Baehr and J. Waldock for useful insights into the biology of Australian spiders. Thanks to R. Black, D. Reed, two anonymous reviewers and the associate editor for useful comments that improved the manuscript. Fauna were collected under licenses and with ethics approval (WA Department of Environment and Conservation, AEC 2005/18). The WA Department of Environment and Conservation and the University of Western Australia provided funding.

Ancillary