Introduction

Habitat loss and degradation are the major drivers of the current global biodiversity crisis (Sala et al. 2000; Tilman et al. 2001; Sodhi et al. 2010). This stems from the widespread logging of tropical rainforests and conversion to agriculture (Hansen et al. 2008; Gibbs et al. 2010). An estimated 2.3 million ha of forests were logged each year between 1990 and 1997 (Achard et al. 2002), with 60% of the remaining forest now classified as degraded primary forest or as secondary growth on abandoned land (ITTO 2002). In turn, 5.8 million ha of forest was cleared annually between 1990 and 1997 (Achard et al. 2002), and the conversion of forests to low biodiversity plantations is ongoing and possibly increasing in rate (Gibbs et al. 2010).

Whilst retaining degraded and secondary forests is no substitute for the protection of large tracts of primary forest, these human-impacted habitats nonetheless support surprisingly high levels of biodiversity (Barlow and Peres 2004; Dent and Wright 2009; Berry et al. 2010; Edwards et al. 2011; Woodcock et al. 2011). Moreover, degraded lands support much greater levels of biodiversity than the agricultural lands that threaten to replace them (Barlow et al. 2007; Edwards et al. 2010b; Styring et al. 2011). Such data highlight the essential role that degraded rainforests can play in a global conservation strategy (Wright 2005; Meijaard and Sheil 2007; Chazdon et al. 2009; Edwards et al. 2011), particularly given the rapidly dwindling area of primary forest and the often prohibitively high costs of paying for the conservation of unlogged habitat (Laurance 2007; Fisher et al. 2011).

Carbon stock enhancements under the United Nations Collaborative Programme on Reducing Emissions from Deforestation and Forest Degradation (REDD+) framework represent one potential mechanism for funding the protection of logged rainforests from conversion to agriculture (Chazdon 2008; Edwards et al. 2010a). Under this scheme, countries would be able to increase the net primary productivity of trees, and thus long-term carbon sequestration, via planting of tree saplings and release of competition by vine cutting (Kobayashi 2007), to yield saleable units of carbon. Such payments may well come in addition to the sale of existing carbon already stored on the land, if the area is directly threatened with conversion (a conventional REDD payment; Edwards et al. 2010a). Given that the global carbon market was already worth US$144 billion in 2009 (Kossoy and Ambrosi 2010), the advent of REDD+ will likely see significant drives towards protecting and improving carbon profiles on logged lands.

Recently, concern has been expressed that vigorous management to enhance rates of carbon sequestration might negatively impact on biodiversity within treated forests (Putz and Redford 2009; Sasaki and Putz 2009). Whilst this would likely be the case if logged or secondary forest is planted with or replaced by non-native trees or monoculture (Barlow et al. 2007), a recent pair of studies has shown that carbon enhancement with a diverse array of native tree species in logged forest had little to no negative effects on avifaunal biodiversity (Edwards et al. 2009; Ansell et al. 2011). Hence, these types of carbon enhancement projects may represent a biodiversity-friendly strategy under REDD+ (Bekessy and Wintle 2008). However, data are now required for other taxa, particularly invertebrates, which attain exceptional densities and diversity in rainforests (Basset et al. 2003; Wilson and Hölldobler 2005; Novotny et al. 2006) and play keystone roles in multiple ecological processes including nutrient cycling (Burghouts et al. 1992; Milton and Kaspari 2007), pollination (Lewis 2009), seed dispersal (Slade et al. 2007), predation (Carroll and Janzen 1973; Tylianakis et al. 2007), and as mutualistic symbionts (Davidson and McKey 1993; Heil and McKey 2003).

Determining invertebrate responses to habitat management across multiple taxa can be problematic in hyperdiverse ecosystems such as tropical rain forests, where typically only a small proportion of species have been formally identified (Basset et al. 2004; Norris et al. 2010). One possible solution to this problem is to examine responses at higher taxonomic levels such as Ordinal or Familial level (Wall et al. 2008; Nakamura et al. 2009; Silveira et al. 2010). For instance, Burghouts et al. (1992) recorded significant changes in the relative abundance of several different invertebrate taxa within leaf-litter after logging, with declines in carnivores (Pseudoscorpionidae) and detritivores (Isoptera) suggesting that broad patterns of feeding guild composition could be negatively impacted by logging. Such data are likely to have lower discriminatory ability than species-level responses (Basset et al. 2008), but studies in tropical forests have found that ordinal level data were sufficient both to predict local species richness (Prinzing et al. 2003) and to discern environmental gradients (Kitching et al. 2001; Vasconcelos et al. 2009; Silveira et al. 2010). Here, we investigate the impacts of selective logging and subsequent carbon stock enhancement by comparing the abundance of different invertebrate taxa within two separate rainforest strata (leaf-litter and understorey) across unlogged, naturally-regenerating and rehabilitated forest. We additionally examine changes in Familial abundance within the functionally diverse Coleoptera, and we assess broader changes in functional composition by explicitly examining responses of different feeding guilds.

Methods

Study area

We focused on one of the largest and most biologically important areas of forest in Southeast Asia (Myers et al. 2000; Lambert and Collar 2002): the 1 million hectare Yayasan Sabah (YS) logging concession in Northeast Borneo (4° 58′N, 117° 48′E). These lowland forests are numerically dominated by large trees of the family Dipterocarpaceae (Whitmore 1984), which are valuable timber species (Jansen 1997; Fisher et al. 2011) and have been selectively logged intensively, yielding harvests of up to 175 m3 ha−1 (Pinard and Putz 1996; Laurance 2007). The area is typical of the moist tropics and experiences relatively little monthly variation in climate, with mean annual rainfall of 2,700 mm and temperatures of 27°C (Marsh and Greer 1992).

We examined the ‘Innoprise and Forest Absorbing CO2 Emissions (FACE) Foundation Rainforest Rehabilitation Project’ (INFAPRO; Moura-Costa 1996), within the 1 million hectare Yayasan Sabah (YS) logging concession, Malaysian Borneo (45° 8′N, 117° 48′E; see Ansell et al. 2011 for further details). This area was selectively logged in 1988–1989 following a modified uniform system, with commercial stems >0.6 m DBH removed using tractor and high lead cable extraction techniques, resulting in ca. 80–100 m3 of timber extracted per ha (Whitmore 1984). Since 1993, INFAPRO has applied carbon enhancement management over 11,000 ha of these selectively logged rainforests (Ansell et al. 2011). The INFAPRO area is surrounded by naturally-regenerating rainforest in the Ulu Segama-Malua Forest Reserve and is situated near to 45,200 ha of unlogged forest in the Danum Valley Conservation Area (DVCA) and Palum Tambun Watershed Reserve (Marsh and Greer 1992). The close proximity of these three forest types within a single contiguous forest provides a unique opportunity to investigate the effects of carbon stock enhancement of logged rainforest on biodiversity.

Carbon enhancement management within INFAPRO utilised enrichment planting and liberation cutting techniques in all study locations (see Ansell et al. 2011). Enrichment planting involved planting lines of native dipeterocarp (95%) and wild fruit tree (5%) species at densities of ≈200 seedlings per ha. Liberation cutting involved the removal of all climbing vines and bamboos 6 months before and 3 years after enrichment planting, whilst non-commercial understorey trees, shrubs and gingers were cut along planting lines immediately before and 3 months after seedlings were planted.

Invertebrate sampling

Fieldwork was conducted from June to August 2010, corresponding with the drier season, along 36 transects, each 180 m long, located across unlogged, naturally regenerating and rehabilitated forest (12 replicated transects in each forest type; see Edwards et al. 2009). Rehabilitated forests in the study area were planted between 1993 and 1995 (≈6 years after logging) and thus sampled ≈16 years after carbon enhancement. Transects within each habitat were ≥500 m apart, with 300 m–9 km between transects in different habitats, to ensure statistical independence of each transect (see Hill and Hamer 2004).

Following recent recommendations for sampling arthropods in tropical landscapes (Missa et al. 2009), we sampled invertebrates using two separate methods: pitfall traps for leaf-litter fauna and sapling beating for understorey fauna.

Pitfall traps

To sample leaf-litter invertebrates, we used standardised pitfall traps placed at five census points spaced at 25 m intervals along each transect (n = 180 traps in total). Pitfall traps are a reliable and consistent method for sampling leaf-litter invertebrates (Kitching et al. 2001; Missa et al. 2009; Silveira et al. 2010). Each census point was considered independent, because the foraging range of most leaf litter invertebrate species is ≤5 m (e.g., Brühl et al. 2003). Traps consisted of plastic pots (120 mm height × 100 mm diam.) that were sunk into the earth so that the pot’s lip was level with the leaf-litter layer and were then filled with 300 ml of soapy water. Traps were collected after 24 h (4,320 trap hours in total).

Sapling beating

Understorey invertebrates were sampled using a sapling beating method on five saplings (between 0.50 and 2 m in height, <2 cm diameter) separated by ≥1 m at each of the five pitfall census points along each transect (n = 900 saplings in total). Saplings were beaten twice using a sturdy stick over a circular beating sheet (white, 1 m diam.) and after each beat, invertebrates were carefully brushed into a collecting pot (5 cm diam.) located at the centre of the sheet and containing soapy water. Invertebrates from each group of five saplings at a census point were aggregated into one sample (n = 180 samples in total). The beating method samples understory invertebrates effectively, with a higher rate of species accumulation than branch clipping or hand collection (Moir et al. 2005; Rohr et al. 2006).

Invertebrates were sorted to Order, with the exception of Sub-class Acarina (mites) and Classes Diplopoda (millipedes), Chilopoda (centipedes) and Gastropoda (snails and slugs). Coleoptera were then identified to Family level, using CSIRO (1991a, b). All sampled Hymenoptera were Formicidae (ants).

Data analysis

Abundance

We determined how the overall abundance of invertebrates per transect and the abundance of individual invertebrate taxa and Colepteran Families differed among unlogged, naturally-regenerating and rehabilitated forests, using generalized linear models (GLZs) with Poisson error and a log link in GLMStat v.5.7.7.

Trophic guild structure

We then assigned each taxon to one of four dietary guilds (carnivore, herbivore, detritivore and omnivore) on the basis of the type of food they consume (CSIRO 1991a, b; Jansen 1997; see Table S1). The Bornean leaf-litter ants are almost exclusively carnivores (Woodcock et al. 2011), whilst Bornean arboreal ants are about 65% herbivores (sap feeders) and 35% carnivores (Davidson et al. 2003), and guilds were thus assigned to understorey ants in these proportions. The Coleoptera were assigned to guilds according to the prevalent diet within each Family (CSIRO 1991a, b). We compared the abundance of individuals within each guild between forest types using GLZs with Poisson error and a log link in GLMStat v.5.7.7.

Results

Abundance

We recorded 20,159 individuals of 23 invertebrate Orders in the leaf-litter and 14,552 individuals of 28 invertebrate Orders in the understorey (Table S1). Total invertebrate abundance in both strata was >10% higher in naturally-regenerating forest than in either of the other two forest types, whilst abundance of leaf-litter invertebrates was 10% lower in rehabilitated forest than in unlogged forest and understorey invertebrate abundance was 20% higher than in unlogged forest (Table 1). Similarly, at transect-level, there was a significant difference in invertebrate abundance among forest types and in the two strata (Table 1; leaf-litter, χ2 = 176.4, df = 2.33, P < 0.0001; understorey, χ2 = 119.4, df = 2.33, P < 0.0001). Naturally-regenerating forest had the highest abundance per transect within both strata, followed by unlogged forest in the leaf-litter and rehabilitated forest in the understorey (Table 1).

Table 1 Total abundance and mean abundance per transect of different taxa of leaf-litter and understorey invertebrates in unlogged (UL), naturally-regenerating (NR), and rehabilitated (R) forests
Full size table

Comparing the abundances of individual taxa with those in unlogged forest, 17% of leaf-litter taxa declined significantly in abundance post-logging, increasing to 33% post-rehabilitation, whereas the proportion of taxa that increased significantly in abundance was 17% in both cases (Fig. 1). In contrast, the proportion of understorey taxa that declined significantly was 15% in both types of forest whereas 45% of taxa increased in abundance post-logging, declining to 33% post-rehabilitation (Fig. 1). There were three cases where logging and rehabilitation of forest had opposite effects: the abundance of Coleoptera and Diptera in leaf-litter and of Araneae in the understorey were all significantly higher in naturally-regenerating forest but significantly lower in rehabilitated forest compared to unlogged forest (Table 1). For the twelve taxa that occurred in sufficient numbers for analysis in both strata, there was little congruence between the leaf litter and understorey in responses to forest management: Orthoptera in both strata were equally abundant in all three types of forest but all other 11 taxa showed qualitatively different changes in abundance within the leaf litter and the understorey (Table 1).

Fig. 1
figure 1

Responses of leaf-litter and understorey invertebrates to logging and rehabilitation of rainforest. The proportion of taxa in unlogged forest that declined significantly (black), revealed no significant change (light grey) and increased significantly (dark grey) after logging followed by natural regeneration (NR) and by forest restoration (R)

Full size image

Trophic guild structure

The responses of individual taxa within each feeding guild were highly idiosyncratic. For instance among leaf-litter carnivores, Formicidae (ants) had significantly higher abundance in rehabilitated forest than elsewhere, staphylinid beetles had lowest abundance in rehabilitated forest and other taxa (Araneae, Pseudoscorpionida) were unaffected (Tables 1, 2). Similarly within understorey herbivores, Blattaria (cockroaches) and Thysanoptera (thrips) had significantly lower abundance in unlogged forest than elsewhere, whereas Phasmatodea (stick-insects) and Lepidoptera (butterfly and moth caterpillars) had highest abundance in unlogged forest, and other taxa (Orthoptera, Hemiptera) were unaffected (Table 1).

Table 2 Mean abundance per transect for families of beetles with n ≥ 30 in unlogged (UL), naturally-regenerating (NR), and rehabilitated (R) forests
Full size table

Despite these idiosyncracies, there was a significant difference among forest types in the overall abundance of leaf-litter carnivores (GLZ: χ2 = 11.7, df = 2.33, P = 0.0029), detritivores (χ2 = 343.0, df = 2.33, P < 0.0001) and omnivores (χ2 = 38.2, df = 2.33, P < 0.0001) (Fig. 2a). The abundance of carnivores and omnivores declined in naturally-regenerating forest compared to unlogged forest (P = 0.001 and P < 0.0001, respectively), but did not decline further in rehabilitated forests compared to naturally-regenerating forest for carnivores (P = 0.6) and declined only slightly further for omnivores (P = 0.03). In contrast, the abundance of detritivores was higher in naturally-regenerating forest than in unlogged or rehabilitated forest (both P < 0.0001), and rehabilitated forest had fewer detritivores than unlogged forest (P = 0.0002). There was no change among forest types in the abundance of herbivores (χ2 = 3.7, df = 2.33, P = 0.16).

Fig. 2
figure 2

The abundance of invertebrates (mean number per transect ±1SE) that were in the carnivore, herbivore, detritivore or omnivore dietary guilds in each type of forest. Superscripts represent significant differences at the **P < 0.01 and ***P < 0.001 levels. a Leaf-litter, b understorey

Full size image

In the understorey, there was a significant difference among forest types in the abundance of carnivores (GLZ: χ2 = 133.5, df = 2.33, P < 0.0001), herbivores (χ2 = 329.1, df = 2.33, P < 0.0001) and omnivores (χ2 = 23.0, df = 2.33, P < 0.0001) (Fig. 2b). In contrast to the leaf-litter, the abundance of these guilds in the understorey increased in naturally-regenerating forest compared to unlogged forest (all P < 0.0001) and rehabilitated forest (P < 0.0001, except herbivores which were at lower abundances [P < 0.0001] and omnivores which did not differ [P = 0.16]). Both omnivores (P = 0.001) and herbivores (P < 0.0001) were at higher abundance in rehabilitated forest than in unlogged forest, but carnivores were at lower abundance (P < 0.0001). There was no change between forest types in the abundance of detritivores (χ2 = 2.6, df = 2.33, P = 0.28). Across both strata, there were consistently smaller differences between unlogged and rehabilitated forest than between unlogged and naturally regenerating forest in the abundances of different feeding guilds, with the exception of understorey herbivores (Fig. 2).

Responses at Family level among the Coleoptera were similar to those at ordinal level among a broad range of taxa: leaf-litter Staphylinidae (carnivores) declined significantly from unlogged to naturally regenerating to rehabilitated forest whereas leaf-litter Ptiliidae (detritivores) were much more abundant in naturally regenerating forest than elsewhere and understorey Chrysomelidae (herbivores) were twice as abundant in rehabilitated forest as elsewhere (Table 2).

Discussion

Logging intensities in our study area are some of the highest on Earth (Laurance 2007; Edwards et al. 2011), and logging resulted in marked changes in the abundance of different invertebrate taxa (Table 1). However, these included both increases and decreases, and fewer than 20% of taxa in either stratum had significantly lower abundance in logged forest (Fig. 1). This proportion is in keeping with species-level data for a narrower range of taxa, supporting the notion that such degraded forests can make an important contribution to the conservation of biodiversity (Hamer et al. 2003; Wright 2005; Meijaard and Sheil 2007; Chazdon et al. 2009; Berry et al. 2010; Edwards et al. 2011; Woodcock et al. 2011). Nonetheless, there were also net differences in the abundances of different feeding guilds in the two habitats (Fig. 2), as also found by Burghouts et al. (1992), suggesting that degraded forest had different functional characteristics to unlogged forest. In particular, the much higher abundance of leaf-litter detritivores and of understorey herbivores and carnivores in naturally regenerating forest may have reflected higher net primary productivity (Berry et al. 2010) and consequent changes in food-web structure and composition.

Rehabilitation of logged forest had different effects on the leaf-litter and understorey fauna. In the understorey, rehabilitation had no effect on the proportion of taxa adversely affected whereas in the leaf-litter, this proportion almost doubled (from 17 to 33%; Fig. 1), corresponding with a marked decrease in total abundance across all taxa (Table 1). There were also two cases where leaf-litter taxa had significantly higher abundance in naturally-regenerating forest but significantly lower abundance in rehabilitated forest compared to unlogged forest (Table 1). Adverse effects of rehabilitation on leaf-litter biodiversity probably reflected changes in leaf-litter abundance, composition or nutrient content (McGlynn et al. 2007; Vasconcelos et al. 2009), corresponding with a 20–50% reduction in the abundance of tree and shrub seedlings and in the prevalence of woody climbers in rehabilitated forest (Ansell et al. 2011). Desiccation-prone taxa such as Collembola may also have been affected by changes in microclimate (Levings and Windsor 1984). There were, however, much smaller differences between unlogged and rehabilitated forest than between unlogged and naturally regenerating forest in both total invertebrate abundance (Table 1) and the abundances of different feeding guilds (Fig. 2). This applied to all feeding guilds in both strata, with the exception of understorey herbivores, suggesting that key elements of ecosystem functioning, such as leaf-litter mixing and nutrient cycling, may have been more similar within rehabilitated forest to those found in primary forest. These changes were still apparent 16 years after rehabilitation took place, suggesting that they were not transient. However, our three treatments were part of a large contiguous block of forest, and so the invertebrate fauna within rehabilitated forests may have been enhanced to some extent by immigration from logged forest.

Despite our recording a broad consensus in the responses of different feeding guilds to forest management, the responses of individual taxa within each guild were highly idiosyncratic (Tables 1, 2). Different species within each taxon are also likely to have responded individualistically but nonetheless, our data provide strong support to previous studies highlighting the idiosyncratic responses of tropical forest animal and plant communities to habitat disturbance (Lawton et al. 1998; Barlow et al. 2007; Basset et al. 2008; Hayes et al. 2009). In addition to different taxa showing idiosyncratic responses to forest management, the responses of most individual taxa also differed qualitatively between rainforest strata (Table 1). This supports previous studies emphasising the distinctiveness of faunal communities within adjacent strata of tropical forests (Hill et al. 2001; Beaulieu et al. 2010). Recent attention in this respect has focused mainly on the canopy (Basset et al. 2003; Elwood and Foster 2004; Dumbrell and Hill 2005), but our results show that many taxa can also have qualitatively different responses to forest management within adjacent strata close to ground level.

We present no information for individual species but our measured impacts on overall invertebrate abundance and on different feeding guilds are similar to patterns seen for bird species that were sampled on the same transects (Edwards et al. 2009). Carbon enhancement reduced avian abundance compared to naturally-regenerating forest but restored feeding guild composition in terms of the abundance of insectivores, but not frugivores. Our results support these previous data in suggesting that carbon stock enhancements carried out in these forests have only limited adverse effects on biodiversity but with some effects on abundance within particular guilds. Elsewhere, the size of the benefit from rehabilitation, in terms of carbon sequestration and biodiversity protection or recovery, will vary with the intensity of habitat disturbance (Kobayashi 2007). Timber yields from many selectively logged forests are much lower than those in our study area and in these cases, the cost-benefit ratio of rehabilitation may be more questionable (Losos 2001; Rey Benayas et al. 2009). Conversely, even greater benefits might arise in forests that have experienced higher timber extraction rates than our study area, for instance through a second round of logging, or have been burned or have regenerated on abandoned farmland. In addition, less vigorous rehabilitation involving lower densities of enrichment planting or less intensive liberation cutting may retain much of the benefit in terms of carbon stock enhancement whilst reducing both labour costs and impacts on biodiversity, and this requires further study.