Introduction
Tropical forests contain much of the world’s terrestrial biodiversity and significant carbon stocks (Bunker et al.
2005). Particular interest centres on assessing the biodiversity value of modified and disturbed forest ecosystems and the ability of such systems to buffer biodiversity losses expected with the degradation or conversion of more pristine habitats (Wright and Muller-Landau
2006; Chazdon et al.
2009). A complete inventory of organisms is not feasible (Lawton et al.
1998), but conservation management can benefit from the identification of any surrogate that broadly predicts overall biodiversity by reflecting the major determinants of taxonomic variety and species richness (Meijaard and Sheil
2012). One approach is to find and use easily assessed indicators (partial measures or estimator surrogates, sensu Sarkar and Margules
2002). However, selection of such indicators remains predominantly intuitive rather than evidence-based (Howard et al.
1997; Lawton et al.
1998; Watt
1998; Noss
1999; Dudley et al.
2005; Kessler et al.
2011; Le et al.
2012) and there remains the challenge of distinguishing change that can be attributed to external anthropogenic factors from underlying natural processes (Magurran et al.
2010). Candidate indicators such as landscape metrics, remotely-sensed variables, multi-species indices and formulated measures of ecosystem complexity or genetic diversity have found wide application but are of limited practicality in forests (UNEP-CBD
1996; Kapos et al.
2001; Delbaere
2002; European Academies’ Science Advisory Council (ESAC)
2004; Gregory et al.
2005; Duraiappah and Naeem
2005). Thus forest biodiversity surveys still maintain a taxonomic focus even though the costs of obtaining sufficient sampling can be high and the utility of any one species, or another single taxon, as a predictor of others remains uncertain (Lawton et al.
1998; Watt et al.
1998; Dufrêne and Legendre
1997; UNEP/CBD
2003; Gregory et al.
2005, but see also Schulze et al.
2004). Further, at large spatial scales where within-region diversity is large, higher level taxa (up to family level) must often be used (Villaseñor et al.
2005), but even this is only justifiable where extensive species data are already available (Sarkar et al.
2005). Such difficulties in forests contrast with intensively sampled, relatively species-poor, temperate agricultural lands where single surrogate species may be indentified (e.g. MacNally and Fleishman
2004; Sauberer et al.
2004) or where easily determined land use parameters such as the extent of adjacent semi-natural habitats, or the incidence of fertilizer use, predict broad species richness (Billeter et al.
2008).
While simple, cost-effective indicators are required (UNEP-CBD
1996; Duraiappah and Naeem
2005), an evidence-based procedure for their evaluation remains elusive. To address this problem, and mindful that validation requires reference baselines based on comprehensive species inventories (Delbaere
2002; UNEP/CBD
2003), we hypothesize that the best indicators for forest or forest-derived ecosystems will be those fundamental characteristics of the plant community that are clearly linked to ecosystem performance. For this reason, both taxonomic and adaptive (functional) plant characteristics were used to sample gradient-based forested landscape mosaics in well-characterized sites in Sumatra, Indonesia and Mato Grosso, Brazil. This approach treats taxonomic and functional characteristics as complementary elements of biodiversity (Folke et al.
1996; Duckworth et al.
2000; Loreau et al.
2001; Kleyer
2002; Gillison
2000,
2006), and proposes that such a typology may be better suited than taxa alone for ecological comparison (Folke et al.
1996; Gillison
2013). The work described in the present paper examines pristine and modified forest systems, testing the hypothesis that vegetation structure and traits are predictive of plant and animal species diversity and abundance, and demonstrates that plant functional type (PFT) diversity, mean canopy height, woody basal area and litter depth have potential as indicators of biological diversity. We also show that the ratio spp.:PFTs might predict animal species richness. A preliminary study of plant functional traits and termite occurrence in Sumatra sites (only) was published by Gillison et al. (
2003). It is argued that forest biodiversity is best addressed within the context of landscape dynamics where ecosystem performance is driven by the interconnectivity of biota across forest and non-forest components of landscape mosaics, i.e. given that the future of much tropical forest is to become multiple land use sites in which some pristine stands remain as reservoirs, the design of the mosaic and the choice of the land uses will determine the extent to which the whole landscape can retain its biota. The present study shows that the indicators we have detected at local regional scale also apply across widely separated biogeographic zones.
Results
Biodiversity summary
In 32 transects in Mato Grosso 542 plant species (1,241 records) and 369 unique (869 species-weighted) PFTs were recorded. In 16 representative subsets of these transects we documented 73 species of vertebrate fauna (17 mammals, 56 birds) and 64 termite species in 11 transect subsets. In Sumatra 16 transects yielded 562 plant species (980 records) and 216 unique (459 species-weighted) PFTs, together with 194 species of vertebrate fauna (31 mammals, 163 birds) in 15 representative transect subsets and 53 termite species in seven representative transect subsets (Tables S4–S12, Online Resources).
Predictors
Plant species richness (number of species in a transect) was best predicted by unique PFT richness, then vegetation structure, cover-abundance of bryophytes, mean canopy height and woody basal area (Table
1). In both regions local plant species richness was also correlated with 16 unique PFT-weighted PFEs (Table
2). Of these, 8 were strong (
P < 0.0001) and consistent between the two regions and seven close to significant (
P < 0.015) though with some variation between Brazil and Sumatra. Some features of vegetation structure, including PFT and plant species diversity, the ratio of plant species diversity to PFT diversity (spp.:PFTs), plant litter depth, mean canopy height, woody basal area, canopy cover, percentage of woody plants and cover-abundance of bryophytes also predicted animal species richness, though somewhat less strongly, with the exception of woody basal area in Sumatra, which was strongly correlated with termite species richness (
P = 0.001). Termite abundance (i.e. encounters per transect) was linked with litter depth in both regions (
P ≈ 0.016, though interpreted as not significant following correction for false discovery rates) but less strongly with plant species diversity (
P ≈ 0.042). Figure
1a–d illustrates differing regional trend lines for bird species richness against litter depth (a, b) and termite species richness, also against litter depth (c, d). Divergent responses between plant litter depth and bird and termite species diversities, respectively, may reflect regional differences in habitat structure, vegetation type and biogeography. The Sumatran sites that are modified agroforests or plantations have no natural savanna or parkland nearby, and hence probably a reduced pool of organisms from which to occupy new niches created in the process. In Brazil, increased species turnover would be expected at forest margins (and hence high β-diversity over the gradsect as a whole). Many unique PFT-weighted PFEs were significantly correlated with faunal diversity, but species-weighted PFEs were more efficient predictors overall (Table
2; Fig.
1e, f, main text; Tables S13, S14, Online Resources).
Table 1
Correlative values (Pearson product-moment correlation) between taxonomic target groups and candidate plant-based indicators (vegetation structure) common to both Brazil and Sumatra, showing separate regional data
Plant species | Unique PFT diversity | 0.956 | 0.0001 | 0.900 | 0.0001 |
| Bryophytes | 0.642 | 0.0001 | 0.716 | 0.002 |
| Woody plants <2 m tall | 0.688 | 0.0001 | 0.614 | 0.011 |
| Mean canopy height | 0.558 | 0.001 | 0.894 | 0.0001 |
| Basal area all woody plants | 0.499 | 0.004 | 0.925 | 0.0001 |
| Litter depth | 0.359 | 0.043 | 0.674 | 0.004 |
Bird species | Litter depth | −0.695 | 0.003 | 0.619 | 0.032 |
Mammal species | Basal area of woody plants | 0.613 | 0.012 | 0.617 | 0.014 |
Mean canopy height | 0.597 | 0.015 | 0.615 | 0.015 |
Termite species | Litter depth | 0.710 | 0.014 | 0.847 | 0.016 |
Basal area all woody plants | 0.614 | 0.045 | 0.955 | 0.001 |
Termite abundance | Litter depth | 0.769 | 0.016 | 0.907 | 0.005 |
Plant species diversity | 0.620 | 0.042 | 0.847 | 0.016 |
Table 2
Correlative values (Pearson product-moment correlation) between taxonomic target groups and candidate plant functional element (PFE) traits common to both Brazil and Sumatra, showing separate regional data
Plant species | Dorsiventral ls. (do)b
| 0.958 | 0.0001 | 0.900 | 0.0001 |
| Mesophyll (me)b
| 0.818 | 0.0001 | 0.837 | 0.0001 |
| Phanerophyte (ph)b
| 0.816 | 0.0001 | 0.954 | 0.0001 |
| Lateral incl. ls.(la)b
| 0.789 | 0.0001 | 0.921 | 0.0001 |
| Platyphyll (pl)b
| 0.721 | 0.0001 | 0.840 | 0.0001 |
| Green p/s stem (ct)b
| 0.687 | 0.0001 | 0.908 | 0.0001 |
| Composite incl. ls. (co)b
| 0.507 | 0.003 | 0.838 | 0.0001 |
| Succulent (su)b
| 0.488 | 0.005 | 0.826 | 0.0001 |
| Rosulate ls.(ro)b
| 0.463 | 0.008 | 0.833 | 0.0001 |
| Lianoid life form (li)b
| 0.822 | 0.0001 | 0.744 | 0.001 |
| Graminoid (pv)b
| 0.578 | 0.001 | 0.734 | 0.001 |
| Notophyll (no)b
| 0.815 | 0.0001 | 0.712 | 0.002 |
| Epiphyte (ep)b
| 0.465 | 0.007 | 0.707 | 0.002 |
| Adventitious roots (ad)b
| 0.722 | 0.0001 | 0.593 | 0.015 |
| Microphyll (mi)b
| 0.399 | 0.024 | 0.503 | 0.047 |
| Hemicryptophyte (hc)b
| 0.668 | 0.0001 | 0.500 | 0.048 |
Mammal species | Succulent leaves (su)a
| 0.491 | 0.053 | 0.784 | 0.001 |
| Filicoid leaves (fi)a
| 0.625 | 0.010 | 0.569 | 0.027 |
| Filicoid leaves (fi)b
| 0.621 | 0.010 | 0.564 | 0.029 |
| Lateral incl. leaves (la)b
| 0.517 | 0.040 | 0.898 | 0.0001 |
| Adventitious roots (ad)b
| 0.616 | 0.011 | 0.537 | 0.039 |
Termite species | Lateral incl. leaves (la)a
| 0.669 | 0.024 | 0.838 | 0.019 |
Termite abundance | Lateral incl. leaves (la)a
| 0.721 | 0.012 | 0.839 | 0.018 |
| Lateral incl. leaves (la)b
| 0.606 | 0.048 | 0.763 | 0.046 |
| Dorsiventral leaves (do)a
| 0.623 | 0.040 | 0.839 | 0.018 |
| Mesophyll size leaves (me)a
| 0.735 | 0.010 | 0.765 | 0.045 |
Combining Brazilian and Sumatran data increased the number of significant generic predictors and the statistical significance of correlations between plant-based variables and species diversity in faunal groups (Tables
3,
4). Correlations of richness between faunal groups also improved substantially when Brazilian and Sumatran data were combined: bird and mammal species diversity (
r = 0.676,
P = 0.0001, highly significant), mammal and termite species diversity (
r = 0.550,
P ≈ 0.027, though not significant following correction for false discovery rates) and mammal species diversity and termite abundance (
r = 0.710,
P ≈ 0.002, significant) [data not tabulated].
Table 3
Correlative values (Pearson product-moment correlation) between taxonomic target groups and candidate plant-based indicators (vegetation structure) common to both Brazil and Sumatra, showing combined data
Plant species | Unique PFT diversity | 0.829 | 0.0001 |
| PFC | 0.703 | 0.0001 |
Basal area all woody plants | 0.565 | 0.0001 |
Mean canopy height | 0.558 | 0.0001 |
Woody plants <2 m tall cov/abd | 0.533 | 0.0001 |
Bryophyte cover/abundance | 0.509 | 0.0001 |
Litter depth (cm) | 0.455 | 0.001 |
Bird species | Spp.:PFTs | 0.682 | 0.0001 |
| Plant species | 0.565 | 0.002 |
Mammal species | Plant species | 0.681 | 0.0001 |
| Spp.:PFTs | 0.598 | 0.0001 |
Basal area of woody plants | 0.479 | 0.006 |
Mean canopy height | 0.475 | 0.007 |
Unique PFT diversity | 0.470 | 0.008 |
Termite species | Spp.:PFTs | 0.847 | 0.0001 |
| Plant species | 0.785 | 0.0001 |
Litter depth | 0.669 | 0.002 |
| Furcation index woody plants | −0.551 | 0.018 |
Basal area all woody plants | 0.541 | 0.021 |
Unique PFT diversity | 0.519 | 0.027 |
Termite abundance | Spp.:PFTs | 0.922 | 0.0001 |
Plant species | 0.791 | 0.0001 |
Total fauna species | Spp.:PFTs | 0.816 | 0.0001 |
Plant species | 0.727 | 0.002 |
Table 4
Correlative values (Pearson product-moment correlation) between taxonomic target groups and candidate unique PFT-weighted PFE indicators common to both Brazil and Sumatra, showing combined data
Plant species | Phanerophyte (ph) | 0.885 | 0.0001 |
| Dorsiventral (do) | 0.833 | 0.0001 |
Lateral incl. (la) | 0.804 | 0.0001 |
Mesophyll (me) | 0.784 | 0.0001 |
Notophyll (no) | 0.751 | 0.0001 |
Photosynthetic stem (ct) | 0.719 | 0.0001 |
Rosulate (ro) | 0.716 | 0.0001 |
Lianoid (li) | 0.709 | 0.0001 |
Succulent (su) | 0.634 | 0.0001 |
Adventitious (ad) | 0.588 | 0.0001 |
Graminoid (pv) | 0.571 | 0.0001 |
Hemicryptophyte (hc) | 0.555 | 0.0001 |
| Filicoid (fi) | 0.536 | 0.0001 |
Platyphyll (pl) | 0.475 | 0.001 |
Epiphytic (ep) | 0.458 | 0.001 |
Composite incl. (co) | 0.441 | 0.002 |
Microphyll (mi) | 0.425 | 0.003 |
Macrophyll (ma) | 0.291 | 0.045 |
Bird species | Rosulate (ro) | 0.480 | 0.010 |
| Chamaephyte (ch) | −0.475 | 0.011 |
| Phanerophyte (ph) | 0.414 | 0.029 |
| Lateral incl (la) | 0.378 | 0.047 |
Mammal species | Lateral incl. (la) | 0.707 | 0.0001 |
| Phanerophyte (ph) | 0.599 | 0.0001 |
Filicoid (fi) | 0.591 | 0.0001 |
Succulent (su) | 0.589 | 0.0001 |
Notophyll (no) | 0.575 | 0.001 |
Mesophyll (me) | 0.537 | 0.002 |
Hemicryptophyte (hc) | 0.524 | 0.002 |
Dorsiventral (do) | 0.471 | 0.008 |
Adventitious | 0.458 | 0.010 |
Rosulate (ro) | 0.457 | 0.010 |
Lianoid (li) | 0.438 | 0.014 |
Graminoid (pv) | 0.433 | 0.015 |
Epiphytic (ep) | 0.430 | 0.016 |
Pendulous incl. (pe) | −0.375 | 0.038 |
Termite species | Phanerophyte (ph) | 0.739 | 0.001 |
| Lateral incl. (la) | 0.632 | 0.005 |
Mesophyll (me) | 0.594 | 0.009 |
Notophyll (no) | 0.593 | 0.009 |
Leptophyll (le) | −0.583 | 0.011 |
Dorsiventral (do) | 0.527 | 0.025 |
Rosulate (ro) | 0.525 | 0.025 |
Lianoid (li) | 0.494 | 0.037 |
Termite abundance | Phanerophyte (ph) | 0.692 | 0.001 |
| Mesophyll (me) | 0.597 | 0.009 |
Notophyll (no) | 0.552 | 0.018 |
Lateral incl. (la) | 0.477 | 0.045 |
All fauna speciesa
| Phanerophyte (ph) | 0.646 | 0.009 |
| Mesophyll (me) | 0.604 | 0.017 |
Lateral incl. (la) | 0.565 | 0.028 |
Filicoid (fi) | 0.539 | 0.038 |
Plant species diversity was closely correlated with PFE diversity (Table
3). Although more than one species can occur within a single PFT and vice versa, species richness and PFT richness usually tend to be highly correlated. That their statistical relationship can and does vary with environment is indicated by a significant difference in regression slopes between the two regions (Fig.
2). Variation in within-sample diversity along land use intensity gradients therefore appears to be distinct between Brazil and Sumatra (see Appendix S3, Online Resources).
Replicable patterns
Regionally distinguishable relationships were found between some soil textural properties and biota (Tables S15, S16; Online Resources). Mato Grosso soil properties were weakly correlated with plant and animal species diversities whereas Sumatran soil properties were strongly correlated with plant species diversity and mammals, and to a lesser degree birds and termites (Tables S17, S18, Online Resources). However, no single soil variable was significantly correlated with fauna in either region, and only one (Al saturation) with plants. In contrast, plant adaptive features represented by PFEs (functional traits) exhibited significant and consistent cross-regional responses to soil properties and in both regions species-weighted PFEs were correlated with pH, CEC, H, K, P and texture (% sand, silt, clay). PFEs which were components of unique PFTs exhibited highly significant correlations with soil bulk density, and % sand, silt, clay, as well as CEC and organic carbon (e.g. Table S19, Online Resources).
Biodiversity indicators and carbon sequestration
For logistical reasons carbon estimates were recorded only for the Sumatran baseline where both total and aboveground carbon correlated strongly with vegetation structure, plant species and PFT diversity and the spp.:PFTs ratio (Table S19, Online Resources). A significant statistical relationship between plant species composition and either total or aboveground carbon was not detected. However, a borderline correlation between PFC and aboveground carbon (r = 0.603, P ≈ 0.013) and total carbon (r = 0.640, P ≈ 0.008) suggests logging and forest conversion affect PFTs and carbon stocks in parallel, but differentially, as carbon stocks are dominated by the largest trees. Termite species diversity and abundance were linked with aboveground carbon (termite diversity r = 0.890, P ≈ 0.007; termite abundance r = 0.898, P ≈ 0.006) and total carbon (diversity r = 0.789, P ≈ 0.035; abundance r = 0.802, P ≈ 0.030).
Discussion
The results provide evidence that the use of readily observable plant functional morphologies and vegetation structure is a practical basis for comparative ecological studies of complex terrestrial environments, both within and between regions. The different strengths of relationships may reflect both complex multi-causality and differences in effective sampling effort relative to inherent variability of the parameters assessed.
The gradsect approach proved to be efficient in sampling major axes of environmental variability. Many biodiversity surveys either employ unstructured sampling or else randomized or purely systematic (usually grid-based) approaches. While these may satisfy statistical sampling theory, they are inefficient and costly to apply in complex habitats, or depending on the size of the window employed are inconsistent with the spatial scale and patch dimensions of tropical landscapes (Huising et al.
2008). Where the aim is to detect maximum diversity or richness among species and functional groups, habitat variation is more efficiently sampled through gradient-based, non-random approaches, for which theory and practice are now well established (Gillison and Brewer
1985; Wessels et al.
1998; Jones and Eggleton
2000; Gillison
2002; Knollová et al.
2005; Parker et al.
2011).
The areas sampled in our study, both in Sumatra and Brazil included definitive areas of several hectares of intermediate disturbance, notably ‘Jungle Rubber’ in Sumatra, and ‘Capoeira’ in Brazil. The questions that arise are whether increases in alpha diversity in these cases should be consistent with the intermediate disturbance hypothesis, and whether the relatively small samples represented by a 40 × 5 m transect would be able to disentangle plant structural traits representative of forest community types from those occurring in their gap succession. The sampling approach using 40 × 5 m transects showed high peaks of alpha diversity consistent with that hypothesis and with other studies in Indo-Malesia using the same methodology to address ridge lines, soil catenary sequences, riparian strips and forest margins (Gillison and Liswanti
2004; Gillison et al.
2004). This level of detection is frequently beyond the capacity of sampling strategies employing larger plot sizes (e.g. 50 × 10 m and above). The relatively small plot size (40 × 5 m) facilitates intensive recording of taxa and functional types and at the same time is logistically suited to additional sampling along environmental gradients and to reduction in observer fatigue.
Gillison (
2013) has shown the plant functional approach (VegClass system) used here is highly sensitive not only to disturbance and modification but also to variation within ‘primary’ forest due to soil nutrient status (cf. Condit et al.
2013). The extent to which faunal groups might respond to such variations within the baseline transect is unknown, though given the relationship between vascular plants and faunal groups detected in the gradsects, some effects due to host plant specificity (for instance on herbivorous insects) might be expected. However, the present study focuses on modified forest landscapes where biota are responding to multiple changes along disturbance gradients and differing patterns of modification (forest and non-forest). The study was not intended to examine how location and scale related influences—for example proximity to primary forests, size of habitat, and landscape connectivity—might be detected and understood.
Human-induced habitat modification has a major impact on biodiversity in both study areas (Sumatra and Mato Grosso). Although the literature is rich in methods for assessing disturbance and related land use intensity (Watt et al.
1998), unambiguous, quantitative units remain elusive (Jackson et al.
2012). The present study showed that subjectively determined land use intensity and disturbance gradients correspond closely with changes in plant species and PFT diversities. Pristine lowland forests supported more PFTs but also more plant species per PFT than secondary or more heavily disrupted forests, thus indicating higher levels of niche complementarity at the scale of our sample-units. As more ecological niches become available for different PFTs with increasing disturbance (here indicated mainly by changes in vegetation structure and aboveground carbon), this ratio decreases until in freshly opened agricultural land or in extreme (e.g. degraded) conditions, the ratio approaches unity (Gillison
2002). In the present study, when regional data were combined, the spp.:PFTs ratio became the strongest overall predictor of faunal species diversity thus suggesting a generally consistent response to disturbance across all biota, though with some exceptions at intermediate disturbance levels (cf. Watt et al.
1998; Sheil and Burslem
2003), for example termite diversity in Brazil. Habitat disturbance (measured here as loss of phytomass—see Appendices S1 and S2, Online Resources) corresponded closely with decreasing spp.:PFTs ratio, supporting the use of the latter as an effective indicator of biodiversity where disturbance is a major driver of ecosystem performance.
Combining regional data resulted in an almost two-fold increase in the overall number of significant or near-significant generic indicators and a three-fold increase in numbers of indicators significant at the
P ≤ 0.0001 level, supporting the conclusion that such indicators may be applied with relative confidence in similar lowland tropical forested regions and with minimum effort. Unlike the traits used in our study, logistically demanding measurements of many functional traits, e.g. leaf mass per area, seed mass and seed output (Westoby et al.
2002; Cornelissen et al.
2003; Wright et al.
2004) are impractical for rapid survey in complex tropical forests. The results also suggest that readily-observable traits common to all terrestrial vegetation allow comparison where environments may be similar but where species differ (Gillison and Carpenter
1997). Further, it is shown that the construction of PFTs from PFEs facilitates complementary assessment of diversity in both species and functional types. Where limited sampling restricts statistical analyses, these may be improved by disaggregating PFTs into their generic PFE components. In our studies (Tables
2,
4) PFEs provided a supplementary subset of statistically significant biodiversity surrogates across a wide range of land cover types and spatial scales. Along the broader-scale environmental gradients in Mato Grosso, transects in structurally simple, savanna-related vegetation on an upland sandstone plateau (nutrient-poor, shallow soils) were richer in fauna than most structurally complex, lowland forest transects on deep, more fertile, well drained soils. Although the inclusion of the savanna-related outliers improved the sample range of species habitat, the coupling of species data from very different biomes may have reduced the effectiveness of simple univariate analyses. By comparison the smaller scale, but less physically heterogenous and more biodiverse Sumatran baseline produced more statistically robust biodiversity indicators.
Landscapes at tropical forest margins usually include a mosaic of habitats with and without trees where many so-called ‘forest’ biota range well beyond forest boundaries (Sanchez et al.
2005). Yet biodiversity-related surveys in tropical forest biomes typically rely on tree-based assessment (Dallmeier and Comiskey
1996). The omission of non-tree components of vegetation and non-forest habitat can exclude information critical for effective conservation planning and management. The present study provides scientific support for a logistically cost-effective assessment of forest biodiversity that includes all vascular plants. Although empirical evidence for plant response to soil variables such as Al
3+ is difficult to establish because of variations in nutrient-cycling pathways, correlations between vegetation structure, plant functional features and soil physical properties (% silt and sand) are readily interpretable, as these are soil parameters not influenced by vegetation (Table S15, Online Resources). As increasing silt content generally improves the supply of plant-available water during drier periods, a favourable soil texture may support higher plant productivity. Soil physical conditions, including litter depth, can be linked with faunal habitat. Plant litter is a food and habitat resource with important structural properties (measured here by depth) for termites and other invertebrate biota. Although litter depth frequently exhibits seasonal variation around its mean value (litter fall divided by mean residence time; Hairiah et al.
2006), relative differences along gradsects were consistent across all sites in both countries, as indeed elsewhere (see Fig. S2, Appendix S2, Online Resources).
A linkage between aboveground carbon, total organic carbon (standing vegetation, dead wood, litter and soil combined) and diversity in tree plant and termite species in Sumatra (Table S19, Online resources) suggests these variables should be examined further as candidate generic indicators. In both regions variations in soil texture and soil physical features such as bulk density exert important indirect effects on faunal diversity through their influence on plant growth and therefore on faunal habitats for which plants are the keystone providers. The same plant-based indicators can be used in other lowland forest types (Fig. S2, Appendix S2, Online Resources) although faunal baseline data are needed for proper evaluation. The lack of evidence for species-based indicators of other species reported here is consistent with findings in African tropical forests (Lawton et al.
1998). Where plant species identification is problematic, plant functional traits can be used as independent biodiversity surrogates. However, surrogacy is improved when functional trait and species data are combined. For this reason we suggest that the inclusion of adaptive PFTs and their component PFEs should be used to complement rather than replace species-based biodiversity assessment. The characterization of photosynthetic tissue, organs and life form in the PFEs together with vegetation structure (mean canopy height, percent canopy cover, basal area) contrasts with the more traditional and functionally restrictive (Raunkiaerean) plant life-forms and indicates greater potential for remote-sensing applications and monitoring forest condition at varying scales of spatial resolution (Asner et al.
2005). The emergence of the spp.:PFTs ratio as one of the more robust biodiversity surrogates, in addition to its potential use as an indicator in disturbed habitats, is a novel finding requiring further investigation.
Variable patterns of land use and differing management scales suggest that any single indicator, even the species diversity of a target taxon, will be of limited value to policy-makers and managers where multiple indicators are required, for example in the selection and gazetting of forest reserves (van Teeffelen et al.
2006). Alternatively, offering a set of simple indicators for efficient biodiversity assessment (cf. Hill and Hamer
2004) may be helpful for conservation decisions where comparative analyses of ecosystems are frustrated by incompatibilities in both scale and the biophysical environment. In cases such as the central Amazon basin, uncertainties surround the correct identification of many plant species (Gomes et al.
2013). Such challenges prevent stakeholders, who are otherwise willing, from investing in practical conservation evaluations and management (Meijaard and Sheil
2012). The present study shows that, based on a detailed analysis of the relationship between plant taxa and plant functional and structural types there is a scientifically defensible alternative when there are difficulties in identifying plant or other taxa. One of the central issues defining the utility of biodiversity indicators is their application across different biogeographic scales. Here we have shown that the indicators we detected at local regional scale also apply across widely separate biogeographic zones. Recent data also demonstrate that at global scale the plant functional and structural types used in the present study exhibit close relationships with climate, thus lending weight to their potential application across biomes (Gillison
2013).
Acknowledgments
We acknowledge the logistical support provided by Instituto Pró-Natura and UNDP/Brasília, the State Environmental Foundation of Mato Grosso, the Rohden Lignea Timber Company in Juruena, the Peugeot/ONF/IPN Carbon Sequestration Project in Cotriguaçu and the Municipal Secretariat of Agriculture in Castanheira. The Research and Development Center for Biology of the Indonesian Institute of Sciences (LIPI) provided botanical and zoological facilities through the Herbarium Bogoriense and the Museum Zoologicum Bogoriense (A. Budiman). In Brazil, herbarium and zoological facilities were provided by the Instituto de Biociências Universidade Federal de Mato Grosso, Cuiabá and Departamento de Zoologia, Universidade de Brasília. We thank N. Liswanti, J.J. Afriastini, I. Arief-Rachman, R.C. de Arruda, M. Boer, E. Carvelho, R. Carvelho, V. Kleber, L.A. Neto, L.A.Y. Nunes, M.C. de Oliveira, C.A.M. Passos, E. Permana, A. Rangel, C.H.N. Rohmar, L.F.U. dos Santos, E.M. Schuster, L. Sell, M. Tomazi, A.M. Vilela and U.R. Wasrin for technical assistance and advice. T.H. Booth, D. P. Faith and J.E. Richey kindly commented on the manuscript.