Biodivers Conserv (2012) 21:323–342 DOI 10.1007/s10531-011-0185-y ORIGINAL PAPER

Plant phylogeny as a surrogate for turnover in beetle assemblages David A. Nipperess • Andrew J. Beattie • Daniel P. Faith Scott G. Ginn • Roger L. Kitching • Chris A. M. Reid • Tracey Russell • Lesley Hughes

•

Received: 10 January 2011 / Accepted: 31 October 2011 / Published online: 9 November 2011 Ó Springer Science+Business Media B.V. 2011

Abstract The ability to extrapolate from the known to the unknown is essential if we are to use the turnover of overall biodiversity, as opposed to a few well-known groups, to inform conservation planning. We investigated the usefulness of using evolutionary relationships of plants as a surrogate for the turnover of their associated beetle assemblages. If plant traits that are important to insects are phylogenetically conserved, it follows that there will be a positive relationship between insect faunal dissimilarity and plant evolutionary distance. We collected beetles using pyrethrum knock-down methods from 40 plant species belonging to four plant families in the Sydney region of Eastern Australia. We developed a novel approach for estimating variance in the dissimilarity of beetle assemblages, as D. A. Nipperess (&)
A. J. Beattie
L. Hughes Department of Biological Sciences, Macquarie University, Sydney, NSW 2109, Australia e-mail: [email protected] A. J. Beattie e-mail: [email protected] L. Hughes e-mail: [email protected] D. P. Faith
S. G. Ginn
C. A. M. Reid Australian Museum, 6 College Street, Sydney, NSW 2010, Australia e-mail: [email protected] S. G. Ginn e-mail: [email protected] C. A. M. Reid e-mail: [email protected] R. L. Kitching Environmental Futures Centre, Griffith School of the Environment, Griffith University, Nathan, QLD 4111, Australia e-mail: [email protected] T. Russell Faculty of Veterinary Science, University of Sydney, Sydney, NSW 2006, Australia e-mail: [email protected]

123

324

Biodivers Conserv (2012) 21:323–342

explained by plant phylogeny, by using phylogenetic eigenvectors as explanatory variables in a distance-based redundancy analysis. We found a highly significant relationship between faunal dissimilarity and plant evolutionary distance for the entire beetle assemblage, the herbivorous component, and the non-herbivorous component, indicating that beetles generally showed some preference for particular plant clades as habitat, regardless of feeding guild. When comparing observed dissimilarities with those predicted from 40 jack-knife replicates of a Generalised Dissimilarity Model, we were often able to predict beetle turnover from plant phylogenetic relationships, although the reliability of this result was highly variable. Nevertheless, the broad response of beetle assemblages to plant evolutionary relatedness indicates real potential for plant phylogenetic pattern to act as a useful surrogate for insect biodiversity, especially when supplemented with other environmental correlates. Keywords Australia
Biodiversity
Coleoptera
Conservation planning
Dissimilarity
Host specificity
Insect-plant interactions Abbreviations AIC Akaike information criterion BLADJ Branch length adjustment GDM Generalised dissimilarity modelling PCoA Principal coordinates analysis

Introduction With approximately a million described species, insects form a large proportion ([50%) of the known global biota (Chapman 2009). However, because the huge task of documenting the diversity and geographical distributions of insects is far from complete, with possibly 70% yet to be described (Hamilton et al. 2010), this ecologically important group (judged by numbers alone) remains under-utilised in conservation planning (Diniz-Filho et al. 2010). It is therefore important to determine whether insect turnover can be predicted from other, more readily available, variables. If so, reserve selection based on such variables, or model predictions derived from them, will increase our chances of adequately preserving insect diversity, given the likely absence of more direct information on the insects themselves. Underlying environmental variability is widely recognised as a good general surrogate for spatial patterns in biodiversity (Faith and Walker 1996; Faith et al. 2004; Ferrier et al. 2007). In the case of insects, a very important component of this variability is the diversity of plant species with which they are associated (Frenzel and Brandl 2001; Moir et al. 2010). Plants vary in a range of architectural and chemical characteristics that are important to the nutritional and microhabitat requirements of insects. A particular phytophagous insect species is more likely to be associated with a range of closely related host plants than a random collection of those locally available (Strong et al. 1984; Ødegaard et al. 2005; Weiblen et al. 2006). The likely reason for this is that characteristics of the host plants themselves are phylogenetically conserved—closely related host plants generally provide more similar environments for insects, both physically and chemically, than more distantly related species. If each plant species serves as a distinct habitat for insects, and the characteristics of these habitats are phylogenetically conserved, we might expect a positive relationship

123

Biodivers Conserv (2012) 21:323–342

325

between the evolutionary relatedness of host plant species and the similarity of their respective phytophagous assemblages. For example, Frenzel and Brandl (2001) found that the phytophagous assemblages on pairs of congeneric host plants were more similar in composition than assemblages on pairs of different host genera. Ødegaard et al. (2005) found a similar pattern with Panamanian beetle assemblages where congeneric host plant pairs had more similar assemblages than confamilials, which in turn had more similar assemblages than hosts of the same order. Several studies have taken a more explicitly phylogenetic approach where the measure of host plant relatedness was estimated as the patristic distance (minimum path length) between host plant species on a phylogenetic tree (Ødegaard et al. 2005; Novotny et al. 2006; Weiblen et al. 2006). By measuring evolutionary relatedness as a continuous variable, it becomes possible to examine the relationship as a linear model and thus to estimate how much variance in the turnover of insect assemblages can be explained from the phylogeny of their host plants. There is some evidence to indicate that not only is there phylogenetic clustering of host specificity in phytophagous insects, but more generally, that the habitat specificity of many insects, including non-herbivores, is also related to plant phylogeny. In a departure from the above studies, Kitching et al. (2003) compared plant relatedness to the dissimilarity of the total beetle assemblage sampled, rather than restricting the dataset to those insect species with a proven feeding relationship with particular plant species. They found a highly significant correlation between the dissimilarity of the total beetle assemblage and plant evolutionary relatedness. Interestingly, when the beetle dataset was reduced to include only unambiguously herbivorous families, the correlation was weaker though still apparent. This would not be an expected outcome if the overall relationship were being driven by herbivores alone. Non-herbivores may show preference for particular clades of plants because of shared microhabitat characteristics (e.g., peeling bark) or because they are specialised on feeding on organisms that are themselves specialised on particular plant clades. To date, studies investigating the relationship between insect turnover and plant phylogeny have been largely conducted in tropical rainforests. This bias can be attributed to high insect species richness in such environments, the drive to explore whether this diversity can be explained by plant-insect interactions, and cheaper costs associated with using local labour as parataxonomists (Basset et al. 2004; Novotny et al. 2002; Novotny et al. 2006; Ødegaard 2006). An exception to this trend is Novotny et al. (2006) who compared tropical New Guinea sites to temperate sites in Central Europe. A key finding of that study was that the relationship between insect turnover and plant phylogeny was not unique to tropical rainforests but appeared to be a more general pattern. The establishment of a general relationship between insect faunal dissimilarity and the evolutionary relatedness of their associated plants allows for the possibility of using plant phylogenetic pattern as a surrogate for turnover in insect biodiversity. In this study we test for generality of this relationship in two ways. First, we explore the relationship between the dissimilarity of total beetle assemblages, including non-herbivores, and the relatedness of their associated habitat-plants. Note that we refer to ‘‘habitat-plants’’ rather than ‘‘host plants’’ to avoid any automatic inference of a co-evolved host-phytophage interaction. Second, we establish whether the relationship occurs in sclerophyll vegetation in the temperate zone of Australia. Sclerophyll vegetation in Australia has evolved in conditions of rainfall unpredictability and significant soil nutrient poverty. Such conditions have been inferred to provide a particularly challenging environment for specialist herbivores (Orians and Milewski 2007) and we might, therefore, expect that this would lead to lower levels of habitat specificity and a generally weaker relationship between habitat plant phylogeny and

123

326

Biodivers Conserv (2012) 21:323–342

insect faunal dissimilarity. Further, what specialisation does occur may be quite broad, being limited to particular plant families (or similar levels of evolutionary relatedness) rather than to specific, closely related, clades contained therein. To test these questions, we examine the turnover of beetle assemblages associated with plant species sampled in woodland vegetation of the Sydney region of Eastern Australia. We use frequency of occurrence as an indicator of the measure of association between an insect species and a particular plant species, without making any assumptions about the nature of the insect-plant interaction. We develop a novel technique for building an explanatory model to determine how much variance in beetle turnover can be explained by habitat-plant phylogeny, while controlling for variation in spatial separation and plant architectural traits. We compare results for total beetle assemblages (all adults), herbivorous families, and non-herbivorous families. For the total beetle assemblage, we also examine the strength of the relationship between faunal dissimilarity and habitat-plant relatedness within and between plant families. Finally, we assess the value of the relationship as a biodiversity surrogate by fitting a predictive model, using Generalised Dissimilarity Modelling (GDM), and comparing observed and predicted beetle turnover.

Methods Habitat plant selection and phylogeny We chose 40 habitat plant species found within the Hawkesbury-Nepean catchment of the Sydney basin, New South Wales, Australia. Each species was represented by a stand of ten individual plants spaced no more than 100 m apart. Locations of the stands are given in Fig. 1—more detailed location information is given in Table 2 in Appendix. All stands were located on shallow, sandy, nutrient-poor soils derived from Hawkesbury Sandstone. Climate is temperate with a mean annual minimum temperature of 11°C, mean annual maximum temperature of 22°C, and a mean annual rainfall of 1,069 mm (Australian Bureau of Meteorology 2010). Ten plant species were selected from each of four plant families—Ericaceae, Myrtaceae, Proteaceae, and Rutaceae. These families represent four of the five dominant plant families in the Sydney region (Robinson 2003) and are among the most speciose in the Australian flora as a whole (Orchard 1999). Within each plant family, species were chosen to represent a range of evolutionary relatedness from sister species to separate subfamilies. A full listing of plant species sampled is shown in Fig. 2. The evolutionary relationships between the plant species were determined using an informal supertree method (Bininda-Emonds 2004). The topology of family-level relationships was taken from the supertree of flowering plants compiled by Davies et al. (2004). For each of our targeted plant families, within-family phylogenies were then grafted onto the ‘backbone’ provided by the family-level supertree. Wherever possible, we used recent and comprehensive phylogenetic treatments to determine branching pattern within our target plant families. Where such treatments were not available for particular sections of the tree, we substituted the taxonomic classification as a reasonable surrogate for phylogeny (Crozier et al. 2005). Even in this case, we only recognised particular taxa as clades if there was evidence for monophyly, although this decision does lead to more unresolved polytomies in the tree than would have been the case if all available taxonomic ranks had been used. Further information on decisions regarding tree topology, including sources, is included in Tables 3, 4 in Appendix.

123

Biodivers Conserv (2012) 21:323–342

327

Fig. 1 Map showing the distribution of stands (represented as open circles) of ten individuals of each plant species sampled for beetles in the Sydney Basin, Australia. Each stand is no more than 100 m in diameter. Inset map shows the location of the study region within the Australian continent

To obtain meaningful estimates of the degree of evolutionary relatedness between host plant species, branch lengths were scaled to represent the time elapsed between divergence events and expressed in units of millions of years. To assign divergence dates to particular nodes on the tree, we selected dates from the literature on the basis that they were calculated under the assumption of a local molecular clock and were calibrated with fossil data. Ages of divergences between families were taken directly from those estimated by Davies et al. (2004). Ages of key within-family divergences were mostly taken from a recent review of the evolution of the Australian flora (Crisp et al. 2004). A small number of additional dates were derived from various sources (see Tables 3, 4 in Appendix). In all, 11 internal nodes out of a total of 33 were assigned cladogenic dates. Ages of all remaining nodes were estimated by interpolation using the BLADJ procedure of Phylocom (Webb et al. 2008). This procedure simply places all undated nodes as evenly spaced between dated nodes, thus minimising variance in branch length. Although not ideal, Webb (2000) found that the interpolation method is a considerable improvement over simpler node counts, even when relatively few nodes are given fixed dates. Additional details on the dates assigned to particular nodes are provided in Tables 3, 4 in Appendix. Beetle sampling and identification Insect sampling methods closely followed those of Andrew and Hughes (2004). Each stand of a particular plant species was sampled for eight consecutive seasons over a period from March 2003 to August 2007. For each plant species for each season, ten individuals were chosen haphazardly for insect collection. Given our haphazard sampling method, there was

123

328

Biodivers Conserv (2012) 21:323–342

Fig. 2 Chronogram of estimated evolutionary relationships of host plant species sampled for beetles in this study. Branch lengths are scaled to millions of years (Ma) since divergence. Scale bar indicates branch lengths

some potential for re-sampling the same individual plant over consecutive seasons. We avoided this as much as possible by recording the location of each plant and choosing stands where the target plant species was locally abundant. Beetles were collected using pyrethrum insecticide delivered from a backpack horticultural sprayer. This method allowed for targeted application to particular plants and was time-efficient relative to other methods (Moir et al. 2005). Pyrethrum spraying was conducted during the morning of days with low-wind conditions. All arthropods falling into two collecting trays (50 9 30 cm) placed under the sprayed foliage of each plant were preserved in 70% ethanol solution for later sorting. Trays were left under plants for a period of 20 min after spraying. By always using two collecting trays for all plants, variation in sample size was partially corrected for. In all, therefore, summing across all individuals and seasons, each plant species was sampled a total of 80 times, each time with two collecting trays. Adult beetles were sorted to morphospecies using the protocols of Oliver and Beattie (1993). Larvae were also collected but are not included in the analysis here because it was not possible to match larval morphospecies to corresponding adults. Specimens were further identified to the taxonomic level of family, and occasionally subfamily, to create reduced datasets of herbivorous and non-herbivorous adult beetles to which results from the total dataset could be compared. Adults were classified as herbivorous if the majority of members within a family or subfamily were described by Lawrence and Britton (1991) as feeding on living plant tissue. There are some problems with this method as family-level designations are fairly crude indicators of feeding guild (Stork 1987), even in the case of a simple division into herbivores and non-herbivores. We do, however, believe this to be an adequate approach for our purposes because: (1) we are confident that the majority of beetles so classified will in fact be herbivorous; and (2) reduction of the dataset to putative

123

Biodivers Conserv (2012) 21:323–342

329

herbivores will inevitably increase the proportion of specialised phytophages. Despite this, note that we make no assumption that putatively herbivorous morphospecies were necessarily feeding on the plant species upon which they were found. Vouchers for the morphospecies are lodged with the Australian Museum. Plant architectural traits Although samples are approximately standardised in terms of foliage volume sprayed, our target host plant species vary greatly in a number of basic architectural traits that might be important for determining the overall abundance of beetle assemblages (Kennedy and Southwood 1984; Basset 1996; Campos et al. 2006). We therefore quantified the more obvious sources of architectural variation likely to influence insect sample size in order to account for such sampling effects in a linear model. We measured the following traits for each individual plant: height, total volume of the foliage (not just the portion sampled for beetles), average area of an adult leaf, and density of the foliage within the sampled volume. Foliage density was calculated as the number of leaves per metre intercepted by three transects passing through the centre of the foliage volume. Transects were oriented to correspond respectively to the breadth, depth and height of the measured volume, and were averaged to provide an index of foliage density for the plant. Our foliage density index, in conjunction with average leaf area, gives an indication of the total leaf area available as a substrate for insects and other small animals. Measuring beetle turnover The Bray-Curtis dissimilarity index, calculated for all possible pairs of habitat plant species sampled in a locality, was used as a measure of beetle faunal turnover. Rather than use raw abundance, each morphospecies was weighted by ‘repeat incidence’. Repeat incidence was defined simply as the frequency of occurrence (no. of individual samples out of a total of 80) of a particular beetle morphospecies on a particular plant species. By this method, beetle morphospecies that were more consistently associated with a particular habitat plant species were given more weight than morphospecies found only occasionally. As is appropriate for subsequent analyses, repeat incidence is a measure of association (frequency with which particular insect and plant species are found together) and not specificity (no. of plant species with which an insect species is associated). This method helps correct for both incidental capture of beetles with little or no association with a particular plant species (i.e., tourists), and for localised variation of abundance over space and time. When comparing dissimilarities calculated using repeat incidence data with abundancebased dissimilarities, Pearson’s correlation was very high (r = 0.899 for adult beetles) and we found no major differences in the shape of the frequency distribution. Note that the calculation of repeat incidence necessarily involves summing across individual plants and seasons and thus the unit of replication for subsequent analyses was the plant species (40 replicates) and not individual plants. This is appropriate given that evolutionary distance is naturally calculated between plant species rather than individuals. Explanatory model We built an explanatory model of beetle turnover for each of three datasets—total adults, herbivorous adults, and non-herbivorous adults. Variance in the dissimilarity of beetle

123

330

Biodivers Conserv (2012) 21:323–342

assemblages was modelled using three sets of explanatory variables: phylogenetic relationships between plants, spatial separation of plants, and plant traits. One option with such data would be to calculate a matrix regression where the matrix of pairwise dissimilarities in beetle assemblages is a function of corresponding matrices of patristic distance, geographic distance, and trait distance, and where the number of objects in the model would equal the number of possible pairwise combinations of plant species. Such an approach was adopted by Ødegaard et al. (2005) and Novotny et al. (2006) for the relationship between insect faunal dissimilarity and plant patristic distance. However, when attempting to estimate the total variance attributable to host plant phylogeny, matrix regression is problematic. As recognised by Ødegaard et al. (2005), treating pairwise distances as objects in a model represents a significant inflation of the original number of replicates, and further, not all such distances are statistically independent. Moreover, when calculating an R2 statistic from matrix regression, no solution currently exists for adjusting the value to be an unbiased estimator of the variance explained, nor can that variance be adequately partitioned among explanatory variables (Legendre et al. 2008). Our approach to these problems was to transform both the faunal dissimilarity matrix and the patristic distance matrix to their respective principal axes by Principal Coordinates Analysis (PCoA), with corrections for negative eigenvalues as necessary. This process, in effect, takes a matrix of pairwise distances between replicates (plant species in our case) and calculates corresponding coordinates for each replicate in a PCoA ordination. When all axes of the ordination are retained, the Euclidean distance between replicates corresponds to the original distances in the untransformed matrix. The PCoA transformation of a faunal dissimilarity matrix prior to fitting an explanatory model was first described by Legendre and Anderson (1999) as part of their method (distance-based Redundancy Analysis) for testing multi-species responses in multiple factor linear models. The same method for transforming a matrix of patristic distances to principal coordinates was independently described by Diniz-Filho et al. (1998) as the first step in their method (Phylogenetic Eigenvector Regression) for testing for phylogenetic inertia in species traits. The novelty of our approach is to combine these methods by using PCoA transformations of both the faunal dissimilarity matrix and the plant patristic distance matrix in the same explanatory model. By this means, we avoid the problems of regressing distance matrices directly. For each explanatory model, we conducted a distance-based Redundancy Analysis of the faunal dissimilarity matrix, where the explanatory variables consist of three tables: a PCoA transformation of the patristic distance matrix (the ‘Phylogeny’ table), a table of latitudes and longitudes of the plant locations (the ‘Space’ table), and a table of log-transformed plant traits (the ‘Traits’ table). By grouping explanatory variables into tables, we were able to examine the proportion of total variance explained by Phylogeny, Space, and Traits both independently and in combination using variance partitioning (Anderson and Gribble 1998). The Phylogeny table did not include all the axes of the PCoA ordination as this would have inevitably led to spurious inflation of explained variance. We therefore adopted the method of Diniz-Filho et al. (1998) where ordination axes were retained only if they captured more variance in the original patristic distance matrix than expected under a broken stick model. We also trialled forward selection using the Akaike Information Criterion (AIC) but found the resulting best selection to be essentially the same as that produced by the broken stick method. We used simple lat/long coordinates for spatial location as distances between sites were sufficiently short that correction for curvature of the earth was considered unnecessary. We tested the following fractions of explained variance for statistical significance: (1) total variance explained by the model (Phylogeny ? Spatial ? Traits); (2) variance

123

Biodivers Conserv (2012) 21:323–342

331

explained by habitat-plant phylogeny (Phylogeny); and (3) variance explained by plant phylogeny while controlling for covariation with spatial and trait variables (Phylogeny | Spatial ? Traits). For each fraction, a pseudo-F statistic was calculated and its statistical significance determined by 999 permutations of the residuals of the model (Legendre and Anderson 1999). All analyses were performed with the R statistical environment version 2.8 (R Development Core Team 2008) and, in particular, the ‘varpart’ function of the ‘vegan’ package (Oksanen et al. 2008) and the ‘rdaTest’ function (Legendre and Durand 2010). To explore the importance of family-level divergences among habitat-plants in the relationship between insect dissimilarity and plant relatedness, we repeated the explanatory model for the total adult beetle assemblage within each of the four plant families. Each model was based on PCoA transformations of beetle faunal dissimilarity and plant patristic distances as described above. Neither spatial nor trait data were included. Because each model consisted of only ten replicates (ten plant species in each family), we did not compare our results against the complete model (based on 40 plant species). Instead, we generated an expected distribution of explained variance by randomly drawing ten plant species from across all four plant families, and calculating an adjusted R2 statistic. This procedure was repeated 1,000 times to generate a distribution of expected adjusted R2 values. We then compared explained variance from our within-family analyses against this expected distribution to test if within-family relationships explained significantly less variance in beetle faunal dissimilarity than expected by chance. Predictive model The goals of explaining and predicting ecological phenomena are not necessarily the same and it is therefore reasonable to use different methods to meet those competing goals (Legendre and Legendre 1998). We adopted GDM (Ferrier et al. 2007) as a method for predicting the expected faunal dissimilarity of a plant species, given information on its evolutionary relationships, spatial location and architectural traits. GDM is a matrix regression technique that fits smoothing functions to predictor variables such that Euclidean distances derived from those variables provide the best possible fit between predicted and observed faunal dissimilarities. Our predictor variables were the same as those used in our explanatory model. The response variables were the original pairwise Bray-Curtis dissimilarities calculated from repeat incidence of adult assemblages. We used the GDM package version 1.1 for the R statistical environment to build our predictive model (Ferrier and Manion 2007). To test the ability of the GDM to predict faunal dissimilarity, we adopted a jackknife approach where each plant species was removed from the model one at a time, the GDM fitted, and the dissimilarity of the removed plant species from those retained in the model estimated from the species’ predictor variables. The resulting vector of predicted dissimilarities was compared to those actually observed and the Spearman correlation coefficient calculated. The jackknife procedure was performed twice—once with all the descriptor variables, and once with habitat-plant phylogeny alone.

Results We collected a total of 3,148 adult beetles sorted into 305 morphospecies from 41 families. The most speciose families were Chrysomelidae (60 morphospecies), Curculionidae (57 morphospecies), and Coccinellidae (22 morphospecies).

123

332

Biodivers Conserv (2012) 21:323–342

Fig. 3 Scatter plot comparing Bray-Curtis dissimilarity of adult beetle assemblages with the evolutionary (patristic) distance between their habitat-plants. A simple linear model is fitted for illustrative purposes only

As a simple visual demonstration of the relationship between plant evolutionary relatedness and beetle turnover, plant patristic distance is plotted against the dissimilarity of adult beetle assemblages (Fig. 3). Mantel correlation between the distance matrices is highly significant (r = 0.364, P \ 0.001). In our explanatory model, a significant proportion of the variance in turnover among beetle assemblages was explained by the full combination of habitat plant descriptors (Phylogeny ? Spatial ? Traits) and, of that combination, habitat-plant phylogeny was the most important component in all cases (Table 1). Controlling for covariance with spatial separation and architectural traits (Phylogeny | Spatial ? Traits) had little effect on determining the effect of habitat-plant phylogeny, confirming that our results were not confounded by either spatial autocorrelation or sample size effects related to plant architecture. Additional components of the partitioning of variance are provided in Table 5 in Appendix. Reducing the dataset to herbivorous or non-herbivorous beetles generally reduced both the total amount of variance explained by the model and that portion attributed to habitatplant phylogeny. Note that this loss of explanatory power cannot be related to sample size because the number of replicates (40 plant species) was constant in all three cases. Nevertheless, all datasets (all adults, herbivores, non-herbivores) were highly significant when tested by permutation. Surprisingly, there was no substantial difference in either variance explained or statistical significance for herbivores and non-herbivores. When fitting the explanatory model to the adult dataset within single plant families, there was a marked decrease in the amount of variance explained by habitat-plant phylogeny. The mean adjusted R2 statistic of 1,000 random draws of ten plant species (0.171) was similar to that observed for the full model of 40 plant species (0.161), but was considerably greater than that observed for single plant families. Significant departure of the explanatory model for each plant family from that expected was determined by calculating the proportion of times, out of 1,000, that a random draw generated an adjusted R2 statistic equal to or less than what was observed. Myrtaceae was most similar to the expected distribution and therefore had the strongest relationship between beetle faunal

123

Biodivers Conserv (2012) 21:323–342

333

Table 1 Variance in the turnover of beetle assemblages (all adults, herbivorous adults and non-herbivorous adults) as explained by the phylogeny, spatial separation and architectural traits of their associated habitatplants Data All Adults

Herbivores

Non-herbivores

Partition

R2

adj. R2

F

P

Phylogeny ? Spatial ? Traits

0.413

0.237

2.342

0.001

Phylogeny

0.225

0.161

3.489

0.001

3.256

0.001

Phylogeny | Spatial ? Traits

0.191

0.157

Residuals

0.587

0.763

Phylogeny ? Spatial ? Traits

0.387

0.203

2.104

0.001

Phylogeny

0.196

0.130

2.934

0.001

2.753

0.001

Phylogeny | Spatial ? Traits

0.169

0.127

Residuals

0.613

0.797

Phylogeny ? Spatial ? Traits

0.372

0.183

1.975

0.001

Phylogeny

0.198

0.131

2.966

0.001

2.710

0.001

Phylogeny | Spatial ? Traits

0.170

0.127

Residuals

0.628

0.816

Variance is partitioned into: total explained variance in the model (Phylogeny ? Spatial ? Traits); variance explained by plant phylogeny (Phylogeny); variance explained by plant phylogeny while controlling for covariation with spatial and trait variables (Phylogeny | Spatial ? Traits); and unexplained variance (residuals) With the exception of the residual component, partitions are tested for significance by 1,000 permutations of the model

dissimilarity and habitat plant phylogeny (adj. R2 = 0.088, P = 0.086), followed by Proteaceae (adj. R2 = 0.0468, P = 0.012), Ericaceae (adj. R2 = 0.0331, P = 0.006), and Rutaceae (adj. R2 = 0.023, P = 0.002). We fitted 40 GDMs to the full beetle dataset, with each jackknife replicate of the model omitting one plant species as previously described. For our set of 40 predictions from the model, we examined the distribution of Spearman correlation coefficients of the observed versus predicted dissimilarities, for all descriptor variables and for phylogeny alone (Fig. 4). When predicting from all descriptors, correlations ranged from 0.01 to 0.88 with an average of 0.51 and a standard deviation of 0.20. When predicting from host phylogeny alone, correlations ranged from 0.23 to 0.83 with an average of 0.54 and a standard deviation of 0.16. Overall, rank-order correlations indicated moderate success in predicting dissimilarities (although this varied greatly between plant species) and that phylogeny alone was generally as effective as using all descriptive variables.

Discussion Our results clearly indicate that there is a significant relationship between beetle species turnover and the evolutionary relatedness of their habitat-plants (Fig. 3; Table 1). This is the first time such a relationship has been demonstrated for sclerophyllous vegetation of Australia, and only the second time for the temperate climatic zone as a whole (after Novotny et al. 2006).

123

334

Biodivers Conserv (2012) 21:323–342

Fig. 4 Distribution of Spearman correlation coefficients of observed versus predicted dissimilarities for each of 40 jackknife replicates of a Generalised Dissimilarity Model of the adult beetle dataset

An interesting outcome of our study is that reducing the beetle dataset to include only herbivorous families does not improve the relationship between faunal turnover and habitat-plant phylogeny. If the relationship was driven exclusively by phytophages specific to the plant species (or higher clades) upon which they were found, we would expect that reducing the dataset to putatively herbivorous taxa would increase the fit of an explanatory model, as it would increase the proportion of the assemblage that are so specialised. Instead, we observe a reduction in the fit of the model although the relationship is still highly significant. Interestingly, reducing the dataset to non-herbivores explains just as much variance as does herbivores. Clearly, non-herbivores must themselves be showing habitat-plant specificity. A large proportion of the sampled non-herbivorous taxa are either predators or fungivores. It is possible that these species are secondarily specialised by preferentially feeding on insects or fungi that are themselves specialised for feeding on particular plant clades. For example, coccinellids (Ladybirds) are predaceous on small Hemiptera (Slipinski 2007) which are themselves commonly specialised for particular plant taxa (Carver et al. 1991). Secondary specialisations may therefore, in aggregate, make a measurable contribution to the relationship. Alternatively, non-herbivores may be showing a preference for particular phylogenetically conserved traits of a plant that are important for providing suitable habitat. In this study, we incorporated a small number of plant architectural traits in our models. This was done to correct for potential sampling effects (due to the large range of plant sizes and growth forms), rather than to elucidate important habitat traits for non-herbivorous beetles. Nevertheless, for our explanatory model, we did test the proportion of variance in beetle turnover explained by plant traits (Table 5 in Appendix). In the case of non-herbivores, variance explained by plant traits was small and non-significant (adj. R2 = 0.005, F = 1.050, P = 0.318). After correction for collinearity with spatial separation and phylogeny however, plant traits made a larger and marginally significant contribution to explained variance (adj. R2 = 0.023, F = 1.242, P = 0.035). While we do not consider these results to be a definitive test on the subject, we do believe that the role of

123

Biodivers Conserv (2012) 21:323–342

335

phylogenetically conserved plant traits as a driver of turnover in non-herbivorous insects deserves further exploration. Even among herbivores, there is no guarantee that a beetle encountered on a plant is necessarily feeding on it. Major et al. (2003) found that approximately two-thirds of chrysomelid and curculionid species found on Callitris spp. were non-feeding casuals. Our use of repeat incidence, effectively weighting dissimilarity towards those species persistently found on particular plant species, is likely to have offset some of the effect of nonfeeding casuals. However, given that many of the beetles collected were likely not specialised for feeding on the plant species on which they were encountered, our results suggest that plants provide habitat for beetles beyond that of a simple feeding substrate, and that these habitats are phylogenetically conserved. Although there are some difficulties with making direct comparisons due to different analytical approaches, it seems that we have explained less variance in insect faunal composition than has been recorded for tropical studies such as the beetles of Panama (Ødegaard et al. 2005) or the herbivorous insects of Papua New Guinea (Novotny et al. 2006). These studies found R2 values (unadjusted) ranging from 0.27 to 0.40 (compared to our result for all adults of 0.225—see Table 1). It is tempting to conclude that this is due to differences in insect niche breadth between the temperate and tropical climatic zones, with tropical insects being more specialised to particular plant lineages than is the case for temperate insects. However, Novotny et al. (2006) found no difference in the average number of host plant species per insect species between tropical and temperate sites. A more likely reason for our explanatory model having a weaker fit than other studies is that we attempted to model total beetle assemblages and not just those species with an established feeding relationship with the plants upon which they were encountered. Even if, as we suggest, non-herbivores and non-feeding herbivores show some habitat preference for particular plant clades, this preference is likely to be less taxonomically specific than what we might expect of specialised phytophages. Our analyses within plant families had much lower explained variance than either the full model or that expected from randomisation, indicating that the relationship between total beetle turnover and habitat-plant phylogeny in our study was largely driven by evolutionary divergences between plant families. Specialisation at such a coarse level is likely to result in a relatively weaker phylogenetic signal than would be the case if insect species were generally specialised at lower taxonomic levels. It is also possible that our study shows a weaker relationship between insect turnover and habitat plant phylogeny than is found in tropical sites because of the unusual features of sclerophyll vegetation in Australia. If low nutrient availability and rainfall unpredictability restricts the ecological viability of specialised folivory, most species will be generalists (Orians and Milewski 2007). Balanced against this is the need to be sufficiently specialised to overcome the strong chemical defences and low palatability of Australian sclerophyllous foliage (Ohmart and Edwards 1991). Closely related plants often have a similar array of secondary compounds and this similarity is strong at the level of plant family and below (Jones and Lawton 1991). It follows that phytophagous beetle species would commonly be specialised to the level of plant family rather than genus or species. This inference is confirmed for Australian Chrysomelinae (Leaf Beetles) where specialisation has been determined for particular plant families or large clades contained therein (Reid 2006; Jurado-Rivera et al. 2008). Given a significant and general relationship between insect turnover and habitat plant phylogeny, there is some scope for predicting the distinctiveness of insect assemblages associated with particular plant species. GDM, based on either all habitat-plant

123

336

Biodivers Conserv (2012) 21:323–342

characteristics or phylogeny alone, predicted dissimilarities that often correlated well with the observed dissimilarities of adult beetle assemblages, including non-herbivores. Given the difficulty of obtaining estimates of the true turnover of insect faunas, we believe that patterns of plant phylogenetic diversity can be a potentially useful surrogate for applied purposes such as conservation planning. However, given that we did not find consistently high correlations between observed and predicted dissimilarities, we would recommend that habitat-plant phylogeny should be only one of many potential correlates that should be considered when predicting insect turnover with methods such as GDM. Among those potential correlates should be the spatial location of insect samples, as used in our study, as well as aspects of the physical environment such as annual average temperature and soil type. To include plant phylogenetic pattern in such models, however, we need a measure of the overall phylogenetic dissimilarity of pairs of sites, rather than pairs of plant species. Ferrier et al. (2007) and Nipperess et al. (2010) describe a general method for calculating a phylogenetic dissimilarity measure based on the proportional branch length on a phylogenetic tree that is shared between two sites. We envisage our work could be usefully extended to investigate the relationship between insect species dissimilarity and plant phylogenetic dissimilarity of pairs of sites, with appropriate modifications of GDM, to essentially predict dissimilarities in one group from dissimilarities in another. More generally, our results and those of previous studies suggest that, all else being equal, phylogenetically diverse communities of plants will support more insect species than communities of more closely related plants, although this has rarely been tested directly (however see Proches et al. 2009). Even in the absence of methods such as GDM, reserves selected on the basis of maximising plant phylogenetic diversity should also conserve insects and other poorly known taxa dependent on plants for food and/or habitat. In the case of phytophagous insects, we have good a priori reasons to expect this to be true and, given the results from this study, such an approach would go some way to conserving non-herbivorous taxa as well. Of course, conservation planning on this basis would not be perfect, not only because the amount of variance in insect species composition that can be explained by plant phylogeny is relatively small, but also because the distribution of insect species and their preferred plant species do not necessarily completely overlap (e.g., Taylor and Moir 2009). Nevertheless, while it is certainly not our intention to replace insect inventory projects as important information sources for conservation planning, we believe that surrogates such as plant phylogenetic patterns can play an important role in filling the numerous information gaps in understanding the spatial organisation of biodiversity. Acknowledgments We thank B. Bowman (Macquarie University) for assistance in collecting and sorting specimens. P. Wilson (Macquarie University), R. Colwell (University of Connecticut) and A. Ives (University of Wisconsin) provided valuable advice. N. Stork and eight anonymous reviewers provided comments and suggested revisions for this manuscript. This project was funded from a grant (DP0665761) awarded to L. Hughes, A. Beattie, D. Faith and R. Kitching by the Australian Research Council, Commonwealth of Australia.

Appendix Tables 2, 3, 4 and 5.

123

Biodivers Conserv (2012) 21:323–342 Table 2 Latitude and longitude (in decimal degrees) of the centre of the stand of each plant species sampled in this study

Species

337

Latitude

Longitude

Angophora hispida

-33.6744

151.1354

Banksia aemula

-33.6403

150.6873

Banksia ericifolia

-33.6961

151.0699

Banksia integrifolia

-33.7108

151.2763

Banksia marginata

-33.6923

151.0714

Banksia oblongifolia

-33.6404

150.6860

Banksia robur

-33.6268

151.2588

Banksia serrata

-33.6961

151.0698

Banksia spinulosa

-33.6957

151.0696

Boronia ledifolia

-33.6269

151.2582

Boronia pinnata

-33.6763

151.1227

Boronia serrulata

-33.5930

151.2821

Boronia thujona

-33.6749

151.2804

Brachyloma daphnoides

-33.6924

151.0724

Calytrix tetragona

-33.5981

151.2876

Corymbia gummifera

-33.6924

151.0723

Darwinia fascicularis

-33.5934

151.2830

Epacris longiflora

-33.6449

151.2414

Epacris microphylla

-33.6925

151.0737

Epacris pulchella

-33.6922

151.0718

Epacris purpurascens

-33.7715

151.1140

Eriostemon australasius

-33.6268

151.2582

Eriostemon buxifolius

-33.5934

151.2830

Eucalyptus haemastoma

-33.6922

151.0718

Grevillea buxifolia

-33.6937

151.0699

Hakea dactyloides

-33.6924

151.0700

Kunzea ambigua

-33.6746

151.1354

Kunzea capitata

-33.6744

151.1353

Leptospermum parvifolium

-33.6745

151.1355

Leptospermum trinervium

-33.6925

151.0710 151.0714

Leucopogon ericoides

-33.6921

Leucopogon esquamatos

-33.6189

151.2653

Leucopogon juniperinus

-33.7714

151.1140

Leucopogon microphyllus

-33.6190

151.2653

Melaleuca quinquenervia

-33.6529

151.1547

Phebalium squamulosum

-33.5982

151.2874

Philotheca salsolifolia

-33.5934

151.2831

Woollsia pungens

-33.6188

151.2652

Zieria pilosa

-33.6748

151.2804

Zieria smithii

-33.7423

151.0375

123

123

Division

Family

Genus

Family

Tribe

Family

Tribe

Family

Genus

Magnoliophyta

Rutaceae

Boronia

Myrtaceae

Eucalypteae

Ericaceae

Styphelieae

Proteaceae

Banksia

Banksia serrata

Banksia spinulosa

Brachyloma daphnoides

Brachyloma daphnoides

Corymbia gummifera

Melaleuca quinqueneriva

Boronia thujona

Zieria smithii

Banksia marginata

Terminal 1

Banksia ericifolia

Grevillea buxifolia

Leucopogon ericoides

Epacris pulchella

Eucalyptus haemastoma

Leptospermum trinervium

Boronia ledifolia

Boronia ledifolia

Boronia ledifolia

Terminal 2

Mast and Givnish 2002

Weston and Barker 2006

Quinn et al. 2003

Crayn and Quinn 2000

Udovicic and Ladiges 2000

Wilson et al. 2005

Weston et al. 1984

Chase et al. 1999

Davies et al. 2004

Source

Subspecies

Genus

Species

Genus

Subgenus

Genus

Species-group

Family

Resolution

The root of each subtree is defined as the youngest common node of two terminal species, given as ‘‘Terminal 1’’ and ‘‘2’’

Each listed taxon represents a distinct subtree. Subtrees were joined together to form an informal supertree

Rank

Taxon

Table 3 Sources for tree topology (branching pattern) used in this study

Species placement within subgenera follows Brooker (2000)

Within-family relationships still largely unresolved with subfamilial classification in doubt. Therefore, topology reflects only genus-level classification

Notes

338 Biodivers Conserv (2012) 21:323–342

Family

Subfamily

Genus

Rutaceae

Grevilleoideae

Banksia

Banksia ericifolia Banksia aemula Banksia marginata

Banksia serrata

Banksia spinulosa

Grevillea robusta

Boronia ledifolia

Epacris pulchella

Angophora hispida

Eucalyptus haemastoma

Melaleuca quinquenervia

Banksia serrata

Banksia spinulosa

Zieria smithii

Brachyloma daphnoides

Boronia ledifolia Melastoma melabathricum

27

25

41

88

47

45

37

66

75

81

101

117

129

Age (Ma)

Crisp et al. 2004

Crisp et al. 2004

Crisp et al. 2004

Barker et al. 2007

Pfeil and Crisp 2008

Jordan and Hill 1996

Crisp et al. 2004

Crisp et al. 2004

Davies et al. 2004

Davies et al. 2004

Davies et al. 2004

Davies et al. 2004

Davies et al. 2004

Source

Maximum

Maximum

Fixed

Fixed

Fixed

Minimum

Fixed

Fixed

Fixed

Fixed

Fixed

Fixed

Fixed

Type

Age of crown group

Age of divergence of serrata / aemula clade

Age of crown group Rutaceae

Earliest fossil pollen is ‘‘mideocene’’

Age of crown group

Estimated median date between divergence of Melastomataceae and radiation of Eucalypteae

Notes

Node ages are ‘‘fixed’’ (a single figure from the source) or an inferred minimum, median or maximum age from a potential range (see notes)

Each node is identified as the youngest common node of two terminal species, given as ‘‘Terminal 1’’ and ‘‘2’’

Family

Ericaceae

Corymbia gummifera

Corymbia gummifera

Tribe

Genus

Eucalypteae

Corymbia gummifera

Family

Myrtaceae

Corymbia ? Angophora

Corymbia gummifera

Corymbia gummifera

Order

‘‘Rosid II’’

Myrtales

Boronia ledifolia

Boronia ledifolia

Banksia marginata

Brachyloma daphnoides

Terminal 2

Terminal 1

‘‘Eudicots’’

Rank

‘‘Core eudicots’’

Taxon

Table 4 Sources for assigning ages to nodes used in this study

Biodivers Conserv (2012) 21:323–342 339

123

340

Biodivers Conserv (2012) 21:323–342

Table 5 Additional components of the partitioning of variance in the turnover of beetle assemblages (all adults, herbivorous adults and non-herbivorous adults) Data All adults

Herbivores

Non-herbivores

Partition

R2

adj. R2

F

P

Spatial

0.103

0.055

2.126

0.001

Spatial | Phylogeny ? Traits

0.079

0.049

2.020

0.001

Traits

0.121

0.020

1.201

0.065

Traits | Phylogeny ? Spatial

0.106

0.031

1.348

0.002

Spatial

0.100

0.051

2.050

0.001

Spatial | Phylogeny ? Traits

0.080

0.048

1.966

0.001

Traits

0.125

0.025

1.245

0.023

Traits | Phylogeny ? Spatial

0.105

0.027

1.284

0.008

Spatial

0.090

0.041

1.836

0.002

Spatial | Phylogeny ? Traits

0.073

0.038

1.748

0.002

Traits

0.107

0.005

1.050

0.318

Traits | Phylogeny ? Spatial

0.104

0.023

1.242

0.035

Variance is partitioned into: variance explained by spatial separation (Spatial); variance explained by spatial separation while controlling for covariation with phylogenetic and trait variables (Spatial | Phylogeny ? Traits); variance explained by plant architectural traits (Traits); and variance explained by plant traits while controlling for covariation with phylogenetic and spatial variables (Traits | Phylogeny ? Spatial) Partitions are tested for significance by 1,000 permutations of the model

References Anderson MJ, Gribble NA (1998) Partitioning the variation among spatial, temporal and environmental components in a multivariate data set. Aust J Ecol 23:158–167 Andrew NR, Hughes L (2004) Species diversity and structure of phytophagous beetle assemblages along a latitudinal gradient: predicting the potential impacts of climate change. Ecol Entomol 29:527–542 Australian Bureau of Meteorology (2010) Climate statistics for Australian locations. Australian Government. http://www.bom.gov.au/climate/data/. Accessed 16 Dec 2010 Barker NP, Weston PH, Rutschmann F et al (2007) Molecular dating of the ‘Gondwanan’ plant family Proteaceae is only partially congruent with the timing of the break-up of Gondwana. J Biogeogr 34:2012–2027 Basset Y (1996) Local communities of arboreal herbivores in Papua New Guinea: predictors of insect variables. Ecology 77:1906–1919 Basset Y, Novotny V, Miller SE et al (2004) Conservation and biological monitoring of tropical forests: the role of parataxonomists. J Appl Ecol 41:163–174 Bininda-Emonds ORP (2004) The evolution of supertrees. Trends Ecol Evol 19:315–322 Brooker MIH (2000) A new classification of the genus Eucalyptus. Aust Syst Bot 13:79–148 Campos RI, Vasconcelos HL, Ribeiro SP et al (2006) Relationship between tree size and insect assemblages associated with Anadenanthera macrocarpa. Ecography 29:442–450 Carver M, Gross GF, Woodward TE (1991) Hemiptera. In: Naumann ID, Carne PB, Lawrence JF et al (eds) The insects of Australia volume i. Melbourne University Press, Melbourne Chapman AD (2009) Numbers of living species in Australia and the World 2nd edn. Australian Biological Resources Study, Canberra Chase MW, Morton CM, Kallunki JA (1999) Phylogenetic relationships of Rutaceae: a cladistic analysis of the subfamilies using evidence from rbcL and atpB sequence variation. Am J Bot 86:1191–1199 Crayn DM, Quinn CJ (2000) The evolution of the atpb-rbcL intergenic spacer in the Epacrids (Ericales) and its systematic and evolutionary implications. Mol Phylogenet Evol 16:238–252 Crisp M, Cook LG, Steane DA (2004) Radiation of the Australian flora: what can comparisons of molecular phylogenies across multiple taxa tell us about the evolution of diversity in present-day communities? Philos Trans R Soc Lond B Biol Sci 359:1551–1571 Crozier RH, Dunnett AgapowPM (2005) Phylogenetic biodiversity assessment based on systematic nomenclature. Evolutionary Bioinformatics 1:11–36

123

Biodivers Conserv (2012) 21:323–342

341

Davies TJ, Barraclough TG, Chase MW et al (2004) Darwin’s abominable mystery: insights from a supertree of the angiosperms. Proc Natl Acad Sci U S A 101:1904–1909 Diniz-Filho JAF, De Sant’ana CER, Bini LM (1998) An eigenvector method for estimating phylogenetic inertia. Evolution 52:1247–1262 Diniz-Filho JAF, De Marco Jr P, Hawkins BA (2010) Defying the curse of ignorance: perspectives in insect macroecology and conservation biogeography. Insect Conserv Diver 3:172–179 Faith DP, Walker PA (1996) Environmental diversity: on the best-possible use of surrogate data for assessing the relative biodiversity of sets of areas. Biodivers Conserv 5:399–415 Faith DP, Ferrier S, Walker PA (2004) The ED strategy: how species-level surrogates indicate general biodiversity patterns through an ‘environmental diversity’ perspective. J Biogeogr 31:1207–1217 Ferrier S, Manion G (2007) GDM R Distribution Pack version 1.1. http://www.biomaps.net.au/gdm/. Accessed 16 Dec 2010 Ferrier S, Manion G, Elith J et al (2007) Using generalized dissimilarity modelling to analyse and predict patterns of beta diversity in regional biodiversity assessment. Divers Distrib 13:252–264 Frenzel M, Brandl R (2001) Hosts as habitats: faunal similarity of phytophagous insects between host plants. Ecol Entomol 26:594–601 Hamilton AJ, Basset Y, Benke KK et al (2010) Quantifying uncertainty in estimation of tropical arthropod species richness. Am Nat 176:90–95 Jones C, Lawton J (1991) Plant chemistry and insect species richness of British Umbellifers. J Anim Ecol 60:767–777 Jordan GJ, Hill RS (1996) The fossil record of the Epacridaceae. Ann Bot 77:341–346 Jurado-Rivera JA, Vogler AP, Reid CAM et al (2008) DNA barcoding insect–host plant associations. Proc R Soc Lond B Biol Sci 276:639–648 Kennedy CEJ, Southwood TRE (1984) The number of species of insects associated with British trees: a reanalysis. J Anim Ecol 53:455–478 Kitching RL, Hurley KL, Thalib L et al (2003) Tree relatedness and the similarity of insect assemblages: pushing the limits? In: Basset Y, Kitching RL, Miller S et al (eds) Arthropods of tropical forests: spatio-temporal dynamics and resource use in the canopy. Cambridge University Press, Cambridge Lawrence JF, Britton EB (1991) Coleoptera. In: Naumann ID, Carne PB, Lawrence JF et al (eds) The insects of Australia volume II. Melbourne University Press, Melbourne Legendre P, Anderson MJ (1999) Distance-based redundancy analysis: testing multispecies responses in multifactorial ecological experiments. Ecol Monogr 69:1–24 Legendre P, Durand S (2010) rdaTest: R library to compute simple and partial canonical redundancy analysis (RDA) with permutation test. http://www.bio.umontreal.ca/legendre/indexEn.html. Accessed 16 Dec 2010 Legendre P, Legendre L (1998) Numerical ecology. Elsevier, Amsterdam Legendre P, Borcard D, Peres-Neto PR (2008) Analyzing or explaining beta diversity? comment. Ecology 89:3238–3244 Major RE, Christie FJ, Gowing G et al (2003) The effect of habitat configuration on arboreal insects in fragmented woodlands of south-eastern Australia. Biol Conserv 113:35–48 Mast AR, Givnish T (2002) Historical biogeography and the origin of stomatal distributions in Banksia and Dryandra (Proteaceae) based on their cpDNA phylogeny. Am J Bot 89:1311–1323 Moir ML, Brennan KEC, Majer JD et al (2005) Toward an optimal sampling protocol for Hemiptera on understorey plants. J Insect Conserv 9:3–20 Moir ML, Brennan KEC, Majer JD et al (2010) Plant species redundancy and the restoration of fauna habitat: lessons from plant-dwelling bugs. Rest Ecol 18:136–147 Nipperess DA, Faith DP, Barton K (2010) Resemblance in phylogenetic diversity among ecological samples. J Veg Sci 21:809–820 Novotny V, Basset Y, Miller SE et al (2002) Low host specificity of herbivorous insects in a tropical forest. Nature 416:841–844 Novotny V, Drozd P, Miller SE et al (2006) Why are there so many species of herbivorous insects in tropical rainforests? Science 313:1115–1118 Ødegaard F (2006) Host specificity, alpha- and beta-diversity of phytophagous beetles in two tropical forests in Panama. Biodivers Conserv 15:83–105 Ødegaard F, Diserud OH, Ostbye K (2005) The importance of plant relatedness for host utilization among phytophagous insects. Ecol Lett 8:612–617 Ohmart CP, Edwards PB (1991) Insect herbivory on Eucalyptus. Annu Rev Entomol 36:637–657 Oksanen J, Kindt R, Legendre P et al. (2008) vegan: Community Ecology Package. R package version 1.150

123

342

Biodivers Conserv (2012) 21:323–342

Oliver I, Beattie AJ (1993) A possible method for the rapid assessment of biodiversity. Conserv Biol 7:562–568 Orchard AE (1999) Introduction. In: Orchard AE (ed) Flora of Australia volume 1–introduction, 2nd edn edn. CSIRO Publishing, Collingwood Orians GH, Milewski AV (2007) Ecology of Australia: the effects of nutrient-poor soils and intense fires. Biol Rev 82:393–423 Pfeil BE, Crisp MD (2008) The age and biogeography of Citrus and the orange subfamily (Rutaceae: Aurantioideae) in Australasia and New Caledonia. Am J Bot 95:1621–1631 Proches S, Forest F, Veldtman R et al (2009) Dissecting the plant-insect diversity relationship in the Cape. Mol Phylogenet Evol 51:94–99 Quinn CJ, Crayn DM, Heslewood M et al (2003) A molecular estimate of the phylogeny of Styphelieae (Ericaceae). Aust Syst Bot 16:581–594 R Development Core Team 2008 R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing. http://www.r-project.org/. Accessed 17 Dec 2010 Reid CAM (2006) A taxonomic revision of the Australian Chrysomelinae, with a key to the genera (Coleoptera : Chrysomelidae). Zootaxa 1292:1–119 Robinson L (2003) Field guide to the native plants of Sydney. Kangaroo Press, Kenthurst Slipinski SA (2007) Australian ladybird beetles (Coleoptera: Coccinellidae): their biology and classification. Australian Biological Resources Study, Canberra Stork NE (1987) Guild structure of arthropods from Bornean rain forest trees. Ecol Entomol 12:69–80 Strong DR, Lawton JH, Southwood SR (1984) Insects on plants. Harvard University Press, Cambridge Taylor GS, Moir ML (2009) In threat of co-extinction: two new species of Acizzia Heslop-Harrison (Hempitera: Psyllidae) from vulnerable species of Acacia and Pultenaea. Zootaxa 2249:20–32 Udovicic F, Ladiges PY (2000) Informativeness of nuclear and chloroplast DNA regions and the phylogeny of the eucalypts and related genera (Myrtaceae). Kew Bull 55:633–645 Webb CO (2000) Exploring the phylogenetic structure of ecological communities: an example for rain forest trees. Am Natur 156:145–155 Webb CO, Ackerly DD, Kembel SW (2008) Phylocom: software for the analysis of phylogenetic community structure and trait evolution. Version 4.0. http://www.phylodiversity.net/phylocom/. Accessed 17 Dec 2010 Weiblen GD, Webb CO, Novotny V et al (2006) Phylogenetic dispersion of host use in a tropical insect herbivore community. Ecology 87:S62–S75 Weston PH, Barker NP (2006) A new suprageneric classification of the Proteaceae, with an annotated checklist of genera. Telopea 11:314–344 Weston PH, Carolin RC, Armstrong JA (1984) A cladistic analysis of Boronia Sm. and Boronella Baill. (Rutaceae). Aust J Bot 32:187–203 Wilson PG, O’Brien MM, Heslewood M et al (2005) Relationships within Myrtaceae sensu lato based on a matK phylogeny. Plant Syst Evol 251:3–19

123