Does logging and forest conversion to oil palm agriculture alter functional diversity in a biodiversity hotspot?

Forests in Southeast Asia are rapidly being logged and converted to oil palm. These changes in land-use are known to affect species diversity but consequences for the functional diversity of species assemblages are poorly understood. Environmental filtering of species with similar traits could lead to disproportionate reductions in trait diversity in degraded habitats. Here, we focus on dung beetles, which play a key role in ecosystem processes such as nutrient recycling and seed dispersal. We use morphological and behavioural traits to calculate a variety of functional diversity measures across a gradient of disturbance from primary forest through intensively logged forest to oil palm. Logging caused significant shifts in community composition but had very little effect on functional diversity, even after a repeated timber harvest. These data provide evidence for functional redundancy of dung beetles within primary forest and emphasize the high value of logged forests as refugia for biodiversity. In contrast, conversion of forest to oil palm greatly reduced taxonomic and functional diversity, with a marked decrease in the abundance of nocturnal foragers, a higher proportion of species with small body sizes and the complete loss of telecoprid species (dung-rollers), all indicating a decrease in the functional capacity of dung beetles within plantations. These changes also highlight the vulnerability of community functioning within logged forests in the event of further environmental degradation.


Introduction
Land-use change is the major driver of ecosystem degradation and biodiversity loss globally (Nepstad et al., 1999;Brooks et al., 2002;Nelson et al., 2006;Laurance, 2007), with an ever-growing proportion of the world's natural habitats being altered by anthropogenic activities (Morris, 2010). Roughly 13 million hectares of forest were converted annually between 2000 and 2010, concentrated within the tropics and principally for agricultural expansion (Hansen et al., 2008;FAO, 2010). In addition, 403 million hectares of tropical forest are designated for logging (Blaser, Sarre & Johnson, 2011), with the rate of logging about 20 times that of forest clearance (Asner et al., 2009).
The impacts of land-use change on biodiversity are often examined, particularly in tropical ecosystems, using into potential resilience and recovery of species in response to land-use change (Koh, Sodhi & Brook, 2004;Hillebrand, Bennett & Cadotte, 2008;Ockinger et al., 2010;Mouillot et al., 2013).
Despite the value of functional diversity metrics in inferring ecosystem processes Naeem, Duffy & Zavaleta, 2012;Mouillot et al., 2013), the impacts of tropical land-use change on functional diversity are poorly understood. Examination of the literature identified just 12 studies that quantified the functional diversity of tropical communities across a terrestrial disturbance gradient (Table 1). Of these studies, only three compared the functional diversity of communities in logged forest with those in primary forest. They found that amphibian functional diversity was higher in primary forest (Ernst, Linsenmair & Rödel, 2006), but that arboreal and avian functional diversity were not different (Baraloto et al., 2012;Edwards et al., 2013).
In addition, only one previous study has investigated the impacts of oil palm agriculture on functional diversity (Table 1; Edwards et al., 2013), yet this is a widespread and rapidly expanding crop globally (Fitzherbert et al., 2008;Gibbs et al., 2010). In Southeast Asia, the conversion of forest (both primary and logged) to oil palm agriculture has been rampant, with several million hectares of oil palm plantation replacing forest over the last two decades (Koh & Wilcove, 2008;Gibbs et al., 2010;Reynolds et al., 2011). Dung beetles provide key ecosystem functions and services, including nutrient recycling, soil aeration, secondary seed dispersal and parasite suppression (Nichols et al., 2008), as their habit of breaking apart dung piles and distributing the material away from the source. However, only one previous study has examined impacts of land-use change on the functional diversity of dung beetles (in Mexico; Barragán et al., 2011), yet these are globally widespread, highly abundant (Hanski & Cambefort, 1991), sensitive to environmental changes (Larsen, Williams & Kremen, 2005;Nichols et al., 2007) and key indicators for other taxonomic groups, especially mammals (Nichols et al., 2009).
In this study, we address these key knowledge gaps by investigating the impacts of land-use change on the taxonomic and functional diversity of dung beetles in the global biodiversity hotspot of Sundaland, Southeast Asia (Myers et al., 2000). We examine a gradient of increasing habitat disturbance from primary forest through once-logged and twice-logged forest to oil palm. We test the hypothesis that disturbance acts as an environmental filter, selecting species more functionally similar than expected by chance and hence leading to lowered functional diversity in disturbed habitats.

Study location
Our study is based within the Yayasan Sabah (YS) logging concession and adjacent oil palm plantations, in eastern Sabah, Malaysian Borneo (4°58′ N, 117°48′ E). Within the  (Fisher et al., 2011). Adjacent to this primary forest is the 238 000 ha Ulu Segama-Malua Forest Reserve (US-MFR) containing selectively logged forest, of which 97 000 ha (41%) has undergone a single rotation of timber extraction (once-logged forest) and the remaining area has undergone two rotations of logging (twice-logged forest). The first rotation of timber extraction took place between 1987 and 1991, with commercial stems > 0.6 m diameter removed to yield ≈ 115 m 3 of timber per ha (Fisher et al., 2011). Twice-logged locations were relogged between 2001 and 2007 with the minimum harvested tree diameter reduced to > 0.4 m, yielding an additional 15-72 m 3 of timber per ha Fisher et al., 2011). Surrounding the US-MFR are oil palm plantations, with sampling locations restricted to mature plantations (10-15 years old), with density of ≈ 100 palms ha −1 (Edwards et al., 2010).

Dung beetle sampling
Fieldwork was conducted between August and October 2009, and between February and September 2011. In each of our four habitats, we created four sampling sites that were widely spaced across the landscape. Sites within a habitat were separated by ≥ 2 km, and distances between sites in different habitats ranged from 1 to 92 km. We used standardized baited pitfall traps to sample dung beetles (Coleoptera: Scarabaeidae: Scarabaeinae) across the four habitat types. Within each site, we created two transects (separated by 500-900 m), and along each transect we placed five pitfall traps baited with human dung at 100 m intervals (total traps = 160; see Edwards et al., 2011 for further details), which was sufficient to ensure independence (Larsen & Forsyth, 2005). We left traps for 4 days and rebaited after 48 h, with beetles collected every 24 h . We used reference collections (T. Larsen) housed at the Forest Research Centre, Sandakan, Malaysia and Smithsonian Museum, Washington DC, USA for species determinations.

Species richness, diversity, evenness and composition
We compared dung-beetle species richness between forested habitats and oil palm using sample-based rarefaction curves with 95% confidence intervals and standardized by the total abundance of individuals in a particular habitat (Gotelli & Colwell, 2001). To assess the accuracy of our sampling, we calculated the average of four commonly used abundance based estimators of species richness (ACE, CHAO1, JACK1 and Bootstrap) using ESTIMATES v. 8.2 (University of Connecticut, Storrs, CT, USA). We measured species diversity using the Shannon-Wiener index and calculated species evenness using Pielou's evenness index in Vegan package (Oksanen et al., 2011).
To investigate the change in species composition between habitats, we used a non-metric multidimensional scaling ordination (Clarke & Warwick, 2001), using the isoMDS function with Bray-Curtis dissimilarity measure within the MASS package (Magurran, 2004). Communities were standardized as a proportion of the total number of individuals on each transect. To test for significant differences in composition, we used a permutational multivariate analysis of variance (ADONIS function in Vegan; Oksanen et al., 2011) with 1000 permutations.

Measuring functional diversity
We examined five traits: behavioural guild, diel activity, body size, diet breath and diet preference (Supporting Information Table S1). Behavioural guilds were categorized into three main groups: rollers (telecoprid), tunnellers (paracoprid) and dwellers (endocoprid) (for descriptions see Hanski & Cambefort, 1991). Information on species behavioural guilds and diel activity (diurnal or nocturnal foragers) were obtained from the literature (Davis, 1999;Krikken & Huijbregts, 2007;Slade et al., 2007;Qie et al., 2011;Slade, Mann & Lewis, 2011) and personal observations. We used a dial calliper to measure body length (pygidium to anterior margin of pronotum) and elytra width to the nearest 0.01 mm (n = 1-27 individuals per species). Body size was then calculated as the product of these two variables (Larsen, Lopera & Forsyth, 2008). We investigated diet breadth with alternative baited traps: rotting vertebrate carrion (n = 19 trap days), rotting fruit (n = 18 trap days) or rotting fungus (n = 16 trap days). Trap design was identical as for traps baited with dung, beetles were collected every 24 h but traps were left for 48 h. Following Larsen et al. (2008), we used the number of different baits a species was attracted to (range = 1-4) as a measure of dietary breadth, and the bait a species was most frequently recorded on as a measure of dietary preference, standardized by the number of trap days (abundance/number trap days) (Supporting Information Table S2). Functional traits were not highly correlated (Kendall correlation: τ < 0.54).
Having obtained trait data, we used the formulae of Villéger, Mason & Mouillot (2008, Villéger et al., 2010, Villéger, Novack-Gottshall & Mouillot, 2011 to calculate five complementary measures of functional diversity: (1) functional richness (FRic), which quantifies the volume of functional space that a set of species occupies; (2) functional evenness (FEve), which describes how species' abundances are distributed throughout the occupied functional space; (3) functional divergence (FDiv), which summarizes the variation in species abundances with respect to the centre of functional space; (4) functional specialization (FSpe), which describes how functionally unique a community is relative to the regional pool of species, and; (5) functional dissimilarity (FDis), which indicates the overlap of functional space between two or more communities. In these methods, traits act as coordinates in functional space, thus identifying a species' functional niche (Villéger et al., 2008). Traits were given equal weighting and species were weighted by their relative abundance. Because our functional traits were a mixture of variable types, we calculated a distance matrix using the Gower distance measure, before running a principal coordinates analysis (PCoA) to calculate a new trait matrix of transformed coordinates (Villéger et al., 2008). Four PCoA axes were used to calculate the functional measures using a multidimensional convex hull to position species in functional trait space.
FSpe was measured as the average distance of a set of species from the centre of functional space, relative to the regional pool of all species (Villéger et al., 2010). FDis was measured as the volume of functional space that two communities share (Villéger et al., 2011). When two communities overlap completely, FDis is equal to zero, and as the overlap in functional space is reduced so dissimilarity increases towards one.

Observed and expected functional diversity
To assess whether disturbance leads to the selection of more functionally similar species than expected by chance, we compared the standardized effect size (SES) of our four functional diversity metrics (FRic, FEve, FDiv and FSpe) across habitats. We defined SES as [(observed -mean expected)/standard deviation of expected]. Expected functional diversity metrics were calculated from 1000 random communities generated from the overall regional species pool. An independent swap algorithm was used to maintain species richness and species frequency occurrence in the random communities (picante package of R) (Gotelli, 2000;Kembel et al., 2010). We then used one-sample t-tests with μ = 0 to determine whether the SES of each functional diversity metric was significantly different from zero.

Comparing among habitats
To check that our results were independent of spatial scale (Hamer & Hill, 2000), each of our species and functional diversity measures were calculated and compared at a large scale (the overall habitat) and a small scale (individual transects). For the latter, we used linear mixed-effect models (LME), including site as a random effect to account for repeated measures. Species abundance was square-root transformed prior to analysis. We also used a Monte-Carlo permutation test for Moran's I statistic (moran.mc within the spdep package), using our model residuals with 1000 repetitions, to test whether or not our transect level results were influenced by spatial autocorrelation. All analyses were run in R v.2.13.2 (R Development Core Team, 2011).

Species richness, diversity and composition
Across the four habitats, we recorded 26 285 individual dung beetles of 65 species. The four common estimators of species richness suggest that we sampled ≥ 89% of species in each of the four habitats (Table 2). There was a decrease in the overall species richness, diversity, evenness and abundance of individuals in oil palm compared to forest, both at the habitat scale ( Fig. 1; Table 2) and on individual transects (Table 2; LME: species richness, F3,12 = 18.39, P < 0.001; abundance F3,12 = 12.51, P < 0.001; species diversity F3,12 = 16.14, P < 0.001; evenness F3,12 = 5.99, P = 0.01). In contrast, logged forest communities did not differ significantly from those in primary forest with respect to species richness, diversity, evenness or abundance (all P ≥ 0.1) ( Table 2). Species composition differed significantly between forest and oil palm ( Fig. 2; ADONIS: r 2 = 0.54, d.f. = 3, P = 0.0001), with the three most abundant species in each forest habitat (Paragymnopleurus sparsus, Sisyphus thoracicus and Onthophagus cervicapra) replaced in oil palm by three congeneric species (O. sp. B, O. obscurior, O. rugicollis). Additionally, 37 of 52 species recorded in primary forest (71%) did not occur in samples from oil palm, while a further nine species occurred in oil palm but not in forest. The species assemblage of primary forest was significantly different from that of both once-logged (r 2 = 0.20, d.f. = 1, P = 0.001) and twice-logged forest (r 2 = 0.20, d.f. = 1, P = 0.02), but the assemblages in the twologged forest treatments did not differ (r 2 = 0.08, d.f. = 1, P = 0.29).

Discussion
Primary rainforests in Southeast Asia are highly threatened by rampant logging and the expansion of large-scale oil palm agriculture (Sodhi et al., 2010;Wilcove et al., 2013), yet this study is among the first assessments of how land-use change affects functional diversity in the region. We found marked reductions in the taxonomic and functional diversity of dung beetles following the conversion of forest to oil palm. In contrast, however, there was very little evidence of any such loss within logged forests, despite significant changes in species composition in comparison to primary forest and even after repeated rotations of logging. These data provide evidence for functional redundancy of dung beetles within primary forest, as also suggested for birds in Amazonian forests (Hidasi-Neto, Barlow & Cianciaruso, 2012). Our results also emphasize the importance of Once−logged Twice−logged Oil palm 95% CI Figure 1 Observed species richness, calculated from sample-based rarefaction curves and scaled to show the number of individuals on the x-axis for dung beetles across a disturbance gradient. Grey shading represents 95% confidence interval (CI) of primary forest.

Figure 2
Non-metric multidimensional scaling (MDS) ordination of community assemblages between unlogged (primary) forest, oncelogged forest, twice-logged forest and oil palm at the transect scale.

F. A. Edwards et al.
Functional diversity and land-use change in Borneo degraded forests as refugia for biodiversity Woodcock et al., 2011), and highlight the potential consequences of biodiversity losses for the support of ecosystem processes within agricultural systems. Dung beetle communities in oil palm are compositionally distinct from those of primary and logged forest (Table 2), with a shift of numerically dominant species, a loss of numerous forest specialists and the addition of a much smaller number of new, presumably disturbance-tolerant species (Figs 2 and 3). These findings support previous work from western Africa that recorded lower species richness and diversity of dung beetles in oil palm plantations compared to logged and primary forests (Davis & Philips, 2005). However, we found significantly lower abundance of dung beetles in oil palm than in forest, whereas the opposite was found in Africa (Davis & Philips, 2005). This variability highlights the need for a more geographically complete understanding of the impacts of oil palm as it expands across tropical regions and replaces both forest and natural grasslands (Koh et al., 2011;Garcia-Ulloa et al., 2012). Assessing the ability of species to persist within remnant forest patches and disperse across the wider land-use matrix will also be Table 3 Habitat and transect (mean ± 1SE) scale functional diversity indices in primary forest, once-logged forest, twice-logged forest and oil palm plantations. FRic, FEve and FDiv are bounded between 0 and 1, and FRic was standardized by a theoretical community of all 65 species in the regional pool  Functional richness of dung beetle communities, described as the minimum convex hull enclosing all species of a community and represented as the volume of functional space filled (denoted here by the area of the grey polygon), in (a) primary forest, (b) once-logged forest, (c) twicelogged forest and (d) oil palm. The black circles are proportional to the relative abundance of species in an individual habitat. Functional richness was much lower in oil palm than elsewhere.
Functional diversity and land-use change in Borneo F. A. Edwards et al. critical to evaluating strategies to enhance biodiversity within agricultural landscapes (Edwards et al., 2010). The dramatic decline that we observed in FRic following conversion of forest to oil palm indicates that the loss of forest species (Fig. 3) was not counterbalanced by the addition of new, disturbance-tolerant species that could either fill vacant functional niches or occupy different functional roles (Table 3; Fig. 3, Supporting Information Figs S1 and S2a,d). The community changes in oil palm show strong evidence for environmental filtering (SES < 0, Fig. S2a,b). In particular, the absence of rollers within oil palm may have been due to altered microclimatic conditions including increased soil temperatures (Lucey & Hill, 2012) decreasing the survival of roller larvae, which typically occur at shallower depths within the soil (Sowig, 1995;Larsen, 2012). We also found a higher proportion of smallbodied species in oil palm, possibly because maximum temperatures in this habitat come closer to exceeding the thermoregulatory tolerance of larger-bodied species, again leading to reduced survival (Nichols et al., 2013). In addition, many dietary generalists (feeding on ≥3 bait types) and species feeding on dung plus carrion were absent from oil palm (Supporting Information Table S2), in contrast to previous work indicating that species with broader diets were less vulnerable to local extinctions (Qie et al., 2011).
Our results suggest that the transition from primary or logged forest to oil palm results in such environmental stresses, particularly due to microclimatic changes, that large subsets of forest species are driven to local extinction irrespective of their dietary breadth or specialization. The absence of rollers within oil palm is particularly important in functional terms, given that they are highly abundant in forests and are behaviourally distinct from tunnellers and dwellers, moving nutrients and seeds away from concentrated dung piles and burying dung balls at shallower depths. In addition, dung removal rate, tunnel depth and volume of dung buried are all positively related to body size, and so the smaller species occurring within oil palm are likely to bury less dung and at lower depths (Slade et al., 2007;Nichols et al., 2008). Changes in the diversity and abundance of nocturnal versus diurnal species may also lead to longer exposure of dung at the surface, resulting in higher gaseous losses of nitrogen (Yamada et al., 2007). Consequently, our results suggest that the functional ability of dung beetles in oil palm is likely to be compromised.
The much lower taxonomic and functional diversity of dung beetles in oil palm also highlights the potential losses that could arise from further degradation of logged forests, for instance through wildfires, which can also act as strong environmental filters and alter microclimatic conditions within the forest (Peres, Barlow & Haugaasen, 2003;Slik & Van Balen, 2006;Lindenmayer et al., 2009;Brodie, Post & Laurance, 2012). Measuring additional functional traits could help in predicting the longer-term impacts of logging and forest conversion. For instance, measures of endothermy and fecundity could aid our understanding of the impacts of microclimatic changes and the likelihood of extinction lags caused by disturbance.
In conclusion, we provide new data on the impacts of land-use change on tropical dung beetles. Contrary to our expectations, even repeated timber harvests did not simplify the functional structure of dung beetle assemblages in Bornean rainforests, despite significant changes in species composition, highlighting the importance of protecting these degraded, logged-over forests. However, conversion of forests to oil palm greatly reduced both species and functional diversity. We suggest ecosystem functioning will be negatively impacted in oil palm, but quantifying the precise consequences across all habitats remains a major knowledge gap. For instance, the retention of forest patches and riparian strips within oil palm estates could support ecosystem services such as nutrient recycling within plantations, but data are needed to address this issue. Our results support previous findings that traditional metrics such as species richness and composition can hide important information about the impacts of land-use change on species traits and functional ecology. The two approaches provide different but complementary mechanisms for understanding human impacts on biodiversity, which can contribute to future conservation and agricultural management decisions (Loyola et al., 2008;Vandewalle et al., 2010;Hidasi-Neto et al., 2012). and 4: primary, once-logged and twice-logged forest (light grey); oil palm (dark grey). Figure S2. The mean standardized effect size (SES) of functional diversity metrics: (a) functional richness; (b) functional evenness; (c) functional divergence; (d) functional specialization in each habitat. SES = [(Observed − mean Expected)/SD Expected]. Expected functional metrics are calculated from 1000 randomizations of the regional pool of species in which species frequency occurrence and species richness are maintained. SES > zero indicates greater functional diversity than the regional species pool. Table S1. Broad trait categories. Scale indicates the type of trait, functional trait shows how the trait is measured and functional importance suggests the impacts of the trait for ecosystem functioning. Table S2. Abundance of species in each habitat, abbreviations: primary forest (P), once-logged forest (1L), twicelogged forest (2L) and oil palm plantation (OP), the functional traits used: the abundance of species visiting dung, carrion, fruit and fungi bait types, the average body size (measured to the nearest mm), the guild and the diel activity of species.