Forest Ecology and Management 339 (2015) 96–104

Contents lists available at ScienceDirect

Forest Ecology and Management

journal homepage: www.elsevier .com/ locate/ foreco

Conservation of tropical forest tree species in a native timber plantationlandscape

http://dx.doi.org/10.1016/j.foreco.2014.11.0280378-1127/� 2014 Elsevier B.V. All rights reserved.

⇑ Corresponding author at: School of Botany, The University of Melbourne,Parkville, VIC 3010, Australia. Tel.: +61 3 418551570, +61 3 8344 3305.

E-mail addresses: elizabethpryde@gmail.com (E.C. Pryde), greg.holland@deakin.edu.au (G.J. Holland), S.Watson@latrobe.edu.au (S.J. Watson), steve.turton@jcu.edu.au (S.M. Turton), dale.nimmo@deakin.edu.au (D.G. Nimmo).

Elizabeth C. Pryde a,⇑, Greg J. Holland b, Simon J. Watson c, Stephen M. Turton a, Dale G. Nimmo b

a Centre for Tropical Environmental and Sustainability Sciences, School of Earth and Environmental Sciences, James Cook University, Cairns, QLD 4870, Australiab Landscape Ecology Research Group, School of Life and Environmental Sciences, Deakin University, Burwood, VIC 3125, Australiac Department of Zoology, La Trobe University, Bundoora, VIC 3086, Australia

a r t i c l e i n f o

Article history:Received 30 July 2014Received in revised form 26 November 2014Accepted 27 November 2014

Keywords:Native Eucalyptus plantationProduction landscapePapua New GuineaDispersal modeSuccessional stage

a b s t r a c t

Tropical terrestrial environments are becoming dominated by anthropogenic land-uses, making retentionof biodiversity in production landscapes of critical conservation importance. Native timber plantationsmay represent a land-use capable of balancing production and conservation by potentially supportingunderstorey plant and tree species otherwise restricted to old-growth forests, with little impact on yield.In this study we investigated the conservation value of native plantation forests in the lowlands of NewBritain, Papua New Guinea. We compared the composition of tree species (P10 cm DBH) of unlogged for-est to those of different aged native Eucalyptus deglupta plantations and intervening (historically logged)secondary forests. We found a high capacity for biodiversity conservation within plantations, with 70% offorest tree species persisting in mature plantations (13–15 years old). However, compositional analysesrevealed lower numbers of large individuals (P10 cm DBH) in both late-successional and non-verte-brate-dispersed species in the plantations, indicating the difficulty of retaining mature old-growth foresttrees in production land-uses. Secondary forest protected by conservation reserves was compositionallyindistinct to unlogged forest. Our results demonstrate the potential for tropical native timber plantationsto contribute to the retention of biodiversity. However, appropriate management is required to ensurethe persistence of source populations of old-growth forest tree species. With careful planning a balancebetween production and conservation can be achieved in lowland tropical regions.

� 2014 Elsevier B.V. All rights reserved.

1. Introduction

Deforestation and degradation of tropical forests has precipi-tated a change in the global composition of rainforest cover,whereby around half of remaining forest cover consists of second-ary regrowth and degraded old-growth forests (Chazdon et al.,2009b). Tropical lowland forests experience particularly high levelsof deforestation because they occur on flat and fertile soilcomparative to other tropical forests, making them valuable foragriculture, logging and agroforestry (Miettinen et al., 2011;Wright, 2010). The loss and degradation of these forests has broadramifications as they contain over half of the world’s terrestrialplant and animal species (Dirzo and Raven, 2003; Sodhi et al.,2010) and play a key role in maintaining global carbon andhydrological cycles (Bradshaw et al., 2007; Houghton, 2012).

Consequently, sustainable management of lowland forests hasbeen identified as a conservation priority (Bradshaw et al., 2009;Gibson et al., 2012).

The needs of local human populations and global demand forforest products means that the full protection of tropical lowlandforests is unlikely (Coad et al., 2009). For example, despite beingrecognised as one of the world’s most significant tropical wilder-ness areas (Myers et al., 2000), the Southeast Asia–Pacific regionhas one of the world’s highest rates of deforestation and degrada-tion (Shearman et al., 2012). As such, conservation priorities areshifting from using reserve-based systems to ones targeting sus-tainable management of multi-purpose landscapes, which attemptto balance biodiversity conservation with production land-uses(Melo et al., 2013). For such approaches to be successful, conserva-tion managers need to understand the capacity of different land-uses to support native biodiversity and the processes which allowpersistence of species in heterogeneous production landscapes(Perfecto and Vandermeer, 2010).

Native timber plantations may represent a land-use capable ofbalancing production and conservation in tropical forests. This is

E.C. Pryde et al. / Forest Ecology and Management 339 (2015) 96–104 97

because they can potentially support understorey plant and treespecies otherwise restricted to remnant forest (Bremer andFarley, 2010), which in-turn would provide for rainforest-dependent fauna (Brockerhoff et al., 2010). By contrast, mostagroforests and tree crops are comparatively limited in thiscapacity (Wilcove et al., 2012). However, the extent to whichnative timber plantations can support plant communities similarto natural forests is poorly understood, particularly outside of theNeotropics (Chazdon et al., 2009a; Stephens and Wagner, 2007).Given the expansion of timber plantations in the tropics (Carnuset al., 2006) and the push for greater representation of nativeplantations globally (Davis et al., 2012) it is crucial to gain a betterunderstanding of the contribution that such plantations can maketo biodiversity conservation.

Little is known regarding the functional breadth of forest plantspecies that can be supported in native plantations. For instance,do plantations support species from multiple forest successionalstages, or only early successional species? How is this affected byplantation age? The ecological mechanisms underpinning theability of species to colonise native plantations are also poorlyunderstood. Examining how species’ traits affect their ability tobecome established in plantations may provide valuable under-standing of the dynamics of understorey composition. For example,following clearing of tropical forests, recruitment of tree communi-ties is largely dependent on ex situ colonisation (Chazdon et al.,2007; Holl, 1999) because most of the seed bank and ‘‘seedlingbank’’ (pre-existing seeds stored in the soil and small seedlings)is destroyed during land clearing, particularly when fire is used(Mamede and de Araujo, 2008). Consequently, dispersal mode islikely to be a fundamental trait influencing assemblages post-clearing (Uhl et al., 1982).

Here, we investigate the role that native plantation forests canplay in biodiversity conservation in the lowlands of New BritainIsland, Papua New Guinea. We assess: (1) the relative ability ofdifferent aged Eucalyptus deglupta plantations and interveningsecondary forest (historically logged) elements to support treespecies of undisturbed forest; (2) the ability of plantations to sup-port a diverse range of successional tree species (e.g. early-, mid-,and late- successional species); and (3) the effects of dispersalmode on establishment in plantations. We include the non-planta-tion landscape elements in our study in order to comprehensively

Fig. 1. The study site on New Brita

assess the biodiversity value of this production landscape and togain insight into processes influencing the persistence of specieswithin plantations (Chazdon et al., 2009a; Gardner et al., 2009).Identifying the conservation value of native plantations willprovide vital information for management and the design ofproduction landscapes.

2. Methods

2.1. Study area

This study was conducted in the Open Bay Timber (OBT) opera-tion area on the Gazelle Peninsula, East New Britain, Papua NewGuinea (PNG) (Fig. 1). New Britain is an oceanic island, rich in ende-mism because of its evolutionary isolation (Mayr and Diamond,2001). The main vegetation type is wet tropical rainforest. Meanannual rainfall is 2000–3500 mm, with a noticeably wetter periodbetween December and March (McAlpine et al., 1983).

Open Bay Timber is one of only two plantation enterprises inPNG cultivating locally native tree species. The focal species, E.deglupta is an evergreen tree native to Indonesia, Timor-Leste,PNG and the Philippines (Ladiges et al., 2003). It is a fast-growing,light-demanding, wind- and water-dispersed species that formsdense, pure stands on river flats and after disturbances such aslandslips and volcanic activity (Paijmans, 1973). We have classifiedit here as an early-successional species (Section 2.4). Beforeforestry activity at Open Bay (�1950) E. deglupta stands occurrednaturally on river flats and disturbed patches of lowland forestand were eventually replaced by successional processes, culminat-ing in mixed alluvium forest (Paijmans, 1976).

E. deglupta plantations were first established in the 1980’sthrough conversion of selectively-logged secondary forest,between an elevation of 10–350 m. Plantation managementincludes clear-fell harvesting, after which remnant logs are left todecompose and fire used to clear weeds prior to seedling planting.Trees are planted on average at a density of �313 trees/ha (spacing4 m � 8 m). Manual weed tending occurs from six months-to-threeyears, and vine cutting from three-to-six years. There is no treethinning. Plantations are harvested at 15–17 years and Eucalyptustimber products are exported predominantly to Vietnam for usein construction and furniture (veneer).

in Island, Papua New Guinea.

98 E.C. Pryde et al. / Forest Ecology and Management 339 (2015) 96–104

At the time of the study E. deglupta plantations covered�14,900 ha (12,000 ha planted area, and �2900 ha of roads, riversand tributaries, and temporary employee housing). Plantationswere organised into management blocks of broadly different ages(young, intermediate, mature) which were separated into evenlyaged compartments. Within a given block, compartments coulddiffer by up to four years in age. The plantation landscape includedan area of �36,400 ha of which �13% was young plantations(2–6 years), �4% was intermediate-aged plantations (7–12 years),and �23% mature plantations (13–15 years), interspersed withsecondary forest (selectively-logged before 1992) and unloggedforest (�33% and 26% of the plantation landscape respectively).The broader landscape constituting the timber company’s produc-tion forest was comprised of a further �18,000 ha of unloggedforest and �24,000 ha of secondary forest (selectively-loggedbefore 1998). Riparian buffer zones occur around the two mainriver systems and their permanent tributaries (the Sai and ToriuRivers) and are embedded within the plantation matrix (we didnot have access to high enough resolution data to estimate theircoverage in the landscape). Riparian areas experienced someselective logging before the 1980s, and accessible edges are subjectto ongoing timber extraction by locals. Within the operation’sbounds a logged-over forest remnant of 382 ha was formallydesignated a watershed conservation area in 1991.

2.2. Study design

Areas within the OBT production landscape were classified intoone of four modified elements reflecting the main forest typespresent (Fig. 2.):

(1) Young plantations, 2–6 years after planting (20 sites).(2) Mature plantations, 13–15 years after planting (2–4 years

prior to harvest) (50 sites).

Fig. 2. Landscape elements surveyed: (a) young plantation, �3 years old; (b) mature pla(e) unlogged rainforest.

(3) Secondary riparian, vegetation within riparian buffer zones�20 years since selective logging (14 sites).

(4) Secondary remnant (conservation area), vegetation within a382 ha forest remnant 16–25 years since selective logging(10 sites).

These modified elements were compared to a fifth, referenceelement of unlogged lowland hill forest adjacent to the OBToperation area (unlogged forest; 48 sites). Intermediate-agedplantation blocks were not surveyed because those that werearrayed in large enough blocks for survey at the time of the studywere confounded by their proximity to transient villages.

2.3. Site selection

We used logging maps to select survey areas that had the high-est interior-to-edge ratios of even-aged vegetation to control forcontextual noise and edge effects (riparian buffer areas were nec-essarily linear and thus characterised by edges). Survey areasoccurred below 400 m a.s.l. to avoid variation associated with ele-vation gradients (Paijmans, 1976). Survey areas within a particularforest element were spread across the geographical extent of theelement to ensure sampling across the variation in topographyand soil type manifest in the region. Plantations, unlogged forest,and the secondary riparian element occurred in discrete patchesacross a large, accessible region within the landscape. This allowedus to sample elements which were interspersed amongst otherland use types, thus reducing the potential influence of spatialclustering driving results. However, this was not the case for sec-ondary remnant forest.

We established point transects in each landscape element,stratifying transect number according to the size of the element.Transects were between 1.2 and 2.4 km in length depending onlogistical constraints and element area, and were separated by a

ntation, �14 years old; (c) secondary riparian buffer; (d) secondary remnant forest;

E.C. Pryde et al. / Forest Ecology and Management 339 (2015) 96–104 99

minimum distance of 1 km. Where access was only possible onfoot, transects were manually cut into the forest. Survey sites werelocated 100 m perpendicular to such transects and sites were posi-tioned P200 m apart from one another to enhance independence.Elsewhere, transects were located along logging roads and surveysites were positioned at least 150 m from the road to reduce edgeeffects.

2.4. Data collection

At each survey site (n = 142), a 30 m line from the central surveypoint was marked in each of the cardinal directions resulting in asquare vegetation plot of 1800 m2. All trees P10 cm diameter atbreast height (DBH) were tallied, and species, size (DBH), mortalitystatus, and phenology (fruit/flower present) were recorded. Plots of1800 m2 are unlikely to capture all tree species occurring withindiverse tropical lowland forest. Consequently, we use these datato compare relative species density and composition betweentreatments (landscape elements) and discuss broader speciesrichness trends in regard to rarefaction curves.

The successional stage and mode of dispersal for each specieswas defined using data collected from several sources (Table S.1).Successional classes generally followed the description in vanValkenburg and Ketner (1994): (1) early-successional species(light-demanding (heliophilic), early maturing species (<5 years)with a relatively short life-span (<80 years) (e.g. Macarangaspecies)); (2) mid-successional species (heliophilic species butwith shade tolerant life-cycle stages (e.g. Octomeles sumatrana),and species which can tolerate shade in earlier developmentalstages (e.g. Canarium indicum)); and (3) late successional species(species that are relatively slow-growing, and tolerant of shadethrough all developmental stages (e.g. Celtis rigescens)).

Trees often have more than one mode of seed dispersal (Nathanand Muller-Landau, 2000), so we classified species according totheir primary mode of dispersal: animal dispersed (birds, batsand some arboreal and ground mammals); or non-vertebratedispersed (wind, water or gravity) (Paijmans, 1976). Where infor-mation regarding dispersal mode was not available at species-level, dispersal mode was inferred from genus-level knowledge(Table S.1).

2.5. Data analysis

2.5.1. Tree species richness, density and compositionAll data analysis was performed in the statistical computing

program R (R Core Team, 2013). To observe the accumulation ofspecies richness within each landscape element we conductedrarefaction analyses using the function ‘specaccum’ in R package‘vegan’ (Oksanen et al., 2013). Accumulation curves were calcu-lated based on both (i) survey sites (using the ‘exact’ method)and (ii) individuals sampled (using the ‘rarefaction’ method).

We modelled species density (sensu Gotelli and Colwell, 2001)for each element using generalised linear mixed models (GLMMs)in the ‘lme4’ package (Bates et al., 2013). Our response variablewas the total number of species recorded at each site, and wasmodelled as a Poisson distribution. Each transect was specified asa random effect to account for non-independent error structuresassociated with potential clustering of study sites (Zuur et al.,2011), and the landscape element that sites were located within(i.e. young plantation, mature plantation, etc.) was specified as afixed effect. Landscape elements were considered an importantinfluence on species richness where 95% confidence intervals (CI)for parameter estimates did not overlap zero when compared tothe reference element (unlogged forest).

We also used Poisson GLMMs to compare the density of speciesgrouped according to (i) successional stage and (ii) dispersal mode

per site. Response variables represented the total number ofspecies belonging to each category that occurred at each site. Eachtransect was assigned as a random effect, and the landscapeelement as a fixed effect. Pairwise comparisons between elements(with unlogged forest as reference category, and secondary rem-nant as reference category) were corrected for family-wise errorusing a Bonferroni correction to adjust significance levels (Quinnand Keough, 2002). All mixed models were tested for overdisper-sion, and for autocorrelation using Moran’s I statistic (Bivand,2014).

To assess differences in tree species composition across ele-ments, we applied a novel model-fitting method of multivariategeneralised linear models (GLMs) to our tree species basal areaand count data (Wang et al., 2012). This method directly modelsthe underlying mean–variance relationship in the abundance datarather than using distance based measures of community dissimi-larity/similarity (Warton et al., 2012). As such, the approach isuseful for assessing treatment effects on community compositionbecause it better detects the influence of treatments on rarespecies (Warton et al., 2012). This was important for our studybecause the lowland rainforests are hyper diverse with patchydominance by any one species (Mueller-Dombois and Fosberg,1998; Paijmans, 1976).

Models were fitted using the ‘manyGLM’ function in package‘mvabund’ in R (Wang et al., 2012, 2013). We built models basedon the number of individuals of species at sites (counts), specifyinga negative binomial error distribution. Landscape element wasincluded as a single categorical predictor. To determine whetherlandscape elements had a significant effect on tree communityassemblage, the data were resampled using parametric bootstrap-ping, with likelihood ratio employed as the test statistic (Wanget al., 2012). To compare pairwise significance among landscapeelements we ran the resampling step twice, first with unloggedforest as the reference factor and second with the secondaryremnant forest as the reference factor. The significance of land-scape elements on individual species abundances was correctedfor family-wise error between species using Holm’s multiple test-ing procedure (Wang et al., 2012). These significance values wereused to identify which species were having a differentiating effectamong elements, and abundance plots comparing these speciesamong elements were used to distinguish the element(s) in whichthey were comparatively most localised (Warton, 2008). In thisway we were able to determine species characteristic of certainland-covers.

Multivariate data are difficult to test for autocorrelation(Ramage et al., 2013). Here we attempted to determine whetherany spatial autocorrelation remained unexplained by the categori-cal predictor by including an interaction term (transects �elements) in our models following Wang (2012).

3. Results

3.1. Tree species richness, density and composition

In total, 95 tree species with DBH P10 cm were recorded acrossall elements in the plantation landscape. Accumulation of speciesacross sites was similar for unlogged forest and secondary remnantand riparian elements as evidenced by their rarefaction curves andoverlapping confidence intervals (Fig. 3). Accumulation of speciesin mature plantations was lower than unlogged and secondary for-est (although still quite high), and young plantations were depau-perate compared to all other elements (Fig. 3). The slightly steeperaccumulation curves of both secondary forest elements comparedto the curve of unlogged forest (Fig. 3a) appears to be mostlycaused by an increased stem density in secondary forest elements

Spe

cies

Sites

Spe

cies

(a)

(b)

100 20 30 40

40001000 2000 3000

Individuals

020

4060

800

2040

6080

Fig. 3. Accumulation of species based on (a) sites and (b) individuals (rarefaction)for each landscape element. Landscape elements are: UF – unlogged forest,SRE – secondary remnant, SRI – secondary riparian, MP – mature plantation andYP – young plantation.

100 E.C. Pryde et al. / Forest Ecology and Management 339 (2015) 96–104

(Gotelli and Colwell, 2001) as exemplified by the greater similarityin rarefaction curves of these three forest elements (Fig. 3b).Comparisons between species densities among elements (alphadiversity) supported the overall species richness results, demon-strating that species densities were similarly high in both unlogged(Fig. 4, S = 27.60 ± 0.29) and secondary elements, but were signifi-cantly lower in mature and young plantations (Table 1, p < 0.05).

Tree species composition differed significantly among land-scape elements (multivariate GLM, p < 0.001). All modified ele-ments differed in composition from unlogged forest except thesecondary remnant for which there was no significant difference(although the p-value was close to significance, p = 0.052,

0.0

2.5

5.0

7.5

10.0

Spec

ies

dens

ity

Non-vertebrTotal species density

0.0

20.0

40.0

60.0

0.0

2.5

5.0

7.5

10.0

12.5

0.0

5.0

10.0

15.0

Landscap

Early successional Mid suc

UF SRE SRUF SRE SRI MP YP

Fig. 4. Species density of all trees, and species densities grouped according to successiomean richness and black lines the 95% confidence intervals. Landscape elements are: UFplantation and YP – young plantation.

Table S.2). Univariate analyses of species’ relative abundance found38 species were the main drivers of compositional differencesamong landscape elements (i.e. occurring in different abundanceamong elements, padj < 0.05, Table S.2). The majority of these treeswere late-successional, animal-dispersed species (Fig. 5). We didnot detect any significant interaction between landscape elementsand transects (p = 0.06). This indicates that the effect of landscapeelement on species composition was not strongly affected by thetransect that the site was on (i.e. spatial autocorrelation was unli-kely to significantly affect results, Wang et al., 2012). However, werecognise that p = 0.06 is close to the cutoff for significance, andthus results need to be interpreted in light of potentially some(relatively weak) spatial structuring in the data.

3.2. Successional stage and species composition among elements

The unlogged forest was primarily composed of late-succes-sional species (67%), followed by mid-successional (25%), andearly-successional (�10%) species across all sites. The compositionof the secondary remnant element, while not significantly differentfrom unlogged forest, did contain a higher proportion of early- andmid-successional differentiating species (Fig. 5). The secondaryriparian and secondary remnant elements differed in their compo-sition (multivariate GLM, p < 0.006, Table S.2), even though bothelements had a similar species density of trees belonging to eachsuccessional class (Fig. 4). The secondary riparian element wascomprised of a greater proportion of individuals belonging toearly- and mid-successional differentiating species, and the sec-ondary remnant had a greater proportion of late-successional indi-viduals (Fig. 5). Cumulatively across all mature plantation sites 70%of tree species found in the unlogged forest were recorded, how-ever, in mature plantations they were sparsely distributed: 24were found in significantly lower abundance compared tounlogged forest (Fig. 5). Fig. 5 shows the occurrence of recruitedtrees in mature plantations, with most individuals representingearly-successional and mid-successional, heliophilic classes. Youngplantations contained very few trees aside from E. deglupta (Fig. 3)leading to either an absence or lower abundance for all speciesexcept the invasive exotic, Mutingia calabura (cherry), including acomplete absence of any late-successional trees.

Animal dispersed ate dispersed

e element

cessional Late successional

0.0

10.0

20.0

30.0

40.0

50.0

0.0

5.0

10.0

15.0

20.0

25.0

I MP YP UF SRE SRI MP YP

nal type and dispersal mode. Circles represent the raw data, bars are the predicted– unlogged forest, SRE – secondary remnant, SRI – secondary riparian, MP – mature

Table 1Model results of species density for all tree species, and species densities grouped according to successional type and dispersal mode. Emboldened figures represent values whose95% confidence intervals do not overlap with zero with unlogged forest as reference group (intercept).

Response variable Landscape element Coef SE z value

All species Intercept 3.325 0.292 11.369Secondary Remnant 0.054 0.508 0.107Secondary Riparian �0.122 0.508 �0.24Mature Plantation �1.445 0.418 �3.456Young Plantation �3.323 0.448 �7.208

Early successional species Intercept 1.212 0.276 4.394Secondary Remnant 0.326 0.487 0.669Secondary Riparian 0.562 0.478 1.177Mature Plantation �0.145 0.393 �0.368Young Plantation �1.489 0.444 �3.354

Mid successional species Intercept 2.289 0.190 12.057Secondary Remnant 0.158 0.335 0.471Secondary Riparian 0.079 0.333 0.239Mature Plantation �1.142 0.279 �4.100Young Plantation �3.236 0.438 �7.380

Late successional species Intercept 1.742 0.171 16.064Secondary Remnant �0.099 0.303 �0.327Secondary Riparian �0.548 0.304 �1.802Mature Plantation �2.024 0.261 �7.757Young Plantation �2.761 0.320 �8.626

Animal dispersed species Intercept 3.202 0.293 10.942Secondary Remnant 0.065 0.509 0.128Secondary Riparian �0.136 0.509 �0.267Mature Plantation �1.463 0.419 �3.492Young Plantation �3.467 0.467 �7.431

Non-vertebrate dispersed species Intercept 1.159 0.081 14.339Secondary Remnant �0.028 0.197 �0.141Secondary Riparian �0.014 0.171 �0.082Mature Plantation �1.120 0.161 �6.977Young Plantation �2.076 0.363 �5.723

0%

20%

40%

60%

80%

100%

Unlogged Forest

Secondary Remnant

Secondary Riparian

Mature Plantations

Young Plantations

ESNVD

ESA

MSNVD

MSA

LSNVD

LSA

Fig. 5. Composition of differentiating species (found to occur in significantlydifferent number among elements), grouped by successional class and mode ofdispersal; Eucalyptus deglupta is not shown. LS = late successional; MS, mid-successional; ES = early-successional; NVD = non-vertebrate-dispersal; A = animal-dispersal.

E.C. Pryde et al. / Forest Ecology and Management 339 (2015) 96–104 101

3.3. Dispersal mode and species composition among elements

Typical of wet tropical forests where the majority of woodyplants are vertebrate-dispersed (Jansen and Zuidema, 2001 andrefs. therein), >93% of the species found in unlogged forest wereanimal-dispersed. The species densities of dispersal modes weresimilar for both secondary remnant and riparian elements (Fig. 4,Table 1), although the number of non-vertebrate-dispersed indi-viduals was proportionally greater in the riparian element(Fig. 5). The species density of animal-dispersed trees was signifi-cantly lower on average in mature plantation sites compared to

unlogged forest (Fig. 4, Table 1). This was because, although 75%of animal-dispersed trees were recorded across all sites in matureplantations, they occurred in comparatively sparse distribution.Similarly, species density of non-vertebrate-dispersed specieswas also significantly lower in mature plantations (Table 1),however this was caused by an even lower representation ofnon-vertebrate-dispersed species found in unlogged forest (only50%), also sparsely distributed. The eight tree species recordedwithin the young plantations were comprised of animal-dispersedspecies that were patchily recruited (e.g. Macaranga spp.,Endospermum medullosum) and non-vertebrate-dispersed species(e.g. O. sumatrana, Alstonia scholaris) that occurred as largeremnant individuals in a few sites (Figs. 4 and 5).

4. Discussion

The ability to conserve biodiversity in production landscapeswill be critical to biodiversity conservation in the 21st century.Here, we have shown that native timber plantations and secondarylogged forests support a substantial proportion of forest tree biodi-versity. Native plantations supported 70% of forest tree species,and the number of tree species in unlogged and secondary forestelements was similar. However, clear differences were evident inthe species composition of unlogged forest and most modified ele-ments, and these were related to the attributes of species (namely,species’ successional stages and modes of dispersal). While modi-fied landscape elements can be species-rich and play a role inmaintaining biodiversity in tropical regions, the retention ofsecondary and unlogged forests within production landscapesremains critical for preserving populations of the full suite of treespecies.

102 E.C. Pryde et al. / Forest Ecology and Management 339 (2015) 96–104

4.1. Species richness and composition of landscape elements

Both secondary forest elements, and the secondary remnantforest in particular, demonstrated a high conservation value, asevidenced by the presence of a similar number of tree species incommon with unlogged forest. This finding is consistent with thehigh species richness observed in other selectively-logged tropicalforests after relatively short fallow periods (Gibson et al., 2012;Hall et al., 2003; Putz et al., 2012). In some cases, this comparablyhigh richness has been the result of an increased number ofearly- and mid-successional species after selective-logging, at theexpense of late-successional species, therefore masking composi-tional differences between forest types (Sheil and Burslem,2003). By contrast, we found no significant difference in composi-tion between the secondary remnant and unlogged forest in thisstudy, although the secondary remnant displayed a trend towardshigher abundance of early- and mid-successional species.

Compositional similarity between the selectively-logged, sec-ondary remnant element and unlogged forest may be explainedby the low density of commercial timber species in PNG’s forests,and hence, a history of lower intensity harvesting disturbancecompared to other tropical countries (Bonnell et al., 2011; Johns,1992). The secondary remnant had also been left fallow and freefrom human disturbance since its designation as a conservationreserve over 16 years prior to this study. Another factor may bethe evolution of these forests in conjunction with regular volcanicdisturbances, and one of the longest histories of anthropogenicmodification in the world (over 30,000 years) (Lentfer et al.,2010). These regular disturbances may have resulted in the filter-ing of more disturbance-tolerant plant assemblages—an observa-tion which has been made of forests on neighbouring Melanesianislands (Bayliss-Smith et al., 2003). Therefore, the unlogged forestsin this study may resemble secondary forest more closely thanstudies in other regions experiencing different biogeographicprocesses and disturbance histories.

Selectively-logged secondary forests can play a vital conserva-tion role as source pools of forest propagules in production land-scapes, if properly managed (Edwards et al., 2010). However, theselectively-logged, riparian elements that are subject to ongoinglocal timber extraction, had a lower abundance of late-successionalspecies compared to the secondary remnant, exhibiting a dimin-ished conservation value. Significant decline in the biodiversityvalue of secondary stands experiencing continuing disturbancehas been well documented (Chazdon, 2003; Gibson et al., 2012;Laurance, 1997), and it is equally likely here that without adequateprotection from further modification, the biodiversity value ofsecondary elements will continue to decline.

Mature plantations demonstrated high cumulative speciesrichness, recruiting 70% of forest species across all sites, althoughoccupation of sites by E. deglupta resulted in their sparse distribu-tion (low site-level richness) compared to non-plantation ele-ments. Comparison of tree species richness with other tropicalplantation studies is confounded by differences in the age of plan-tations studied, the life-stage measurement of woody plants (seed-ling, sapling, tree), and the study site’s baseline forest speciesrichness. Nevertheless, the mature E. deglupta plantations con-tained a similar or greater proportion of native rainforest tree spe-cies compared to studies in exotic (Chapman and Chapman, 1996;Lugo, 1992; Parrotta, 1995) and native timber plantations (Keenanet al., 1997; Wardell-Johnson et al., 2005). By contrast, youngplantations 2–6 years post-clearfell had a substantially reducedrichness of recruiting trees, which is not surprising given that onlytrees P10 cm DBH were sampled and few recruits would havegrown to that size within the time since plantations were estab-lished. This highlights the varying contribution of differently agedplantations to the functional composition of trees in a landscape.

4.2. Successional stage is limited by plantation age

Mature plantations were mainly comprised of early- and mid-successional individuals. Most late-successional trees were foundin low density, but cumulatively, across mature plantation sites,two-thirds of late-successional species found in unlogged forestwere capable of growing to 10 cm DBH in mature plantations. Itis possible that even more late-successional species, which were<10 cm DBH, were present in mature plantations. Late-succes-sional trees tend to grow slowly (Laurans et al., 2012), resultingin more individuals occurring as difficult to detect saplings andseedlings (<10 cm DBH). For example, in some other plantationstudies, richness of late-successional trees was found to be higherin juvenile stages compared to adult stages (Farwig et al., 2009;Keenan et al., 1997). The richness of late-successional species isbest assessed where juvenile tree species (<10 cm DBH) can beidentified and recorded. This was not possible in this study and itis therefore likely that we have underestimated richness of thesespecies here. This type of assessment would be especially instruc-tive for those plantations incorporating a restoration role. For theindustrial plantations of this study, plantation age is a more influ-ential inhibitor of the density of late-successional trees that cansurvive beyond juvenile stages.

The ability of E. deglupta plantations to foster a breadth of suc-cessional types may arise from its natural role in these forests as asuccessional catalyst, where it invades disturbed sites, quicklyforms mono-specific stands, suppressing weedy plants and creat-ing abiotic conditions resembling intact rainforest (Paijmans,1973). Additionally, facilitative germination conditions are likelyaugmented by the low intensity of stand-level management: thin-ning was unprofitable and rare, and manual weed tending ceasedat three years. In many production plantations, high intensity man-agement has inhibited recruitment of mid- and late-successionalspecies (Kanowski et al., 2005; Keenan et al., 1997).

4.3. Dispersal mode drives colonisation in plantations

Of the trees recorded in unlogged forest, mature plantationssupported 50% of the non-vertebrate-dispersed species and 75%of animal-dispersed species. This finding likely reflects the lowerdispersal capacity of non-vertebrate-dispersed species throughoutthe landscape (Willson and Crome, 1989). Distribution of animal-dispersed trees in this landscape does not appear to be as limited,and may be explained by the high permeability of the matrix tolocal animal vectors (e.g. birds, bats). Our contemporaneousstudies of forest birds supports this hypothesis: we found justtwo species restricted to unlogged forest (Pryde et al. unpublisheddata), and birds, along with bats, are considered the primary treedispersers in these lowlands (Mayr and Diamond, 2001). Perme-ability is thought to be enhanced in plantation landscapes becauseof the proportion of the matrix with a continuity of tree cover(Brockerhoff et al., 2010; Keenan et al., 1997).

The recruitment of most animal-dispersed species in the planta-tions (albeit, in low abundance) suggests that plantations them-selves attract visitation by a range of seed-dispersing species.This is most likely in more mature stages when plantations havea higher tree species richness, however, in young plantations westill observed birds exploiting the perching structure and nectarprovided by E. deglupta, a particularly important food resourcefor the island’s parrot species (Marsden and Pilgrim, 2003).Between three-to-six years of age, E. deglupta can flower and growto a height of 10 m (Francis, 1988). The fast-growth rates typical ofplantation species has been similarly observed to provide struc-tural resources from very early stages (McClanahan and Wolfe,1993; Parrotta, 1995).

E.C. Pryde et al. / Forest Ecology and Management 339 (2015) 96–104 103

4.4. Conservation and management implications

The results from this study indicate that the combination ofnative timber plantations set among older secondary forest ele-ments can support high levels of tree species biodiversity. Nativetimber plantations are rare in many tropical regions and shouldbe considered in preference to exotic plantations due to theirenhanced capacity to harbor native biodiversity. We found thatolder plantations held greater benefit for biodiversity, with youngplantations demonstrating a poor capacity to recruit all but ahomogeneous assemblage of early-successional tree species.Therefore, temporally varying harvesting cycles to ensure the high-est possible cover of mature plantations through time (and theirpresence at all times) would increase the biodiversity conservationvalue of the landscape.

Regenerating secondary forest elements were found to beimportant refugia for populations of rainforest-restricted species,such as slow-growing, late-successional and non-vertebrate-dispersed trees. Therefore, recognition of their high conservationvalue is vital, and their protection from future harvesting andfurther plantation establishment recommended. Open Bay’s highratio of unlogged forest to clearfelled area in the broader land-scape, and the extensive, old, regenerating secondary forest withinthe production landscape, have been identified as key contextualcharacteristics for biodiversity conservation in multi-purpose land-scapes (Gibson et al., 2012; Letcher and Chazdon, 2009). Therefore,any future expansion of plantations to meet increasing timberdemands should be located on degraded lands, rather than byencroaching into logged-over forest, as is current practice in NewGuinea.

Ultimately, the capacity of plantations to contribute to biodiver-sity in terms of species population sizes and reproductive successwill be modest compared to that of unlogged and secondaryelements because of clearfell practices and competition from plan-tation trees (Catterrall et al., 2005). Thus, ensuring that old-growthrainforest is retained is critical, to provide refuge for rainforest-restricted species and source pools for species capable of existingin modified elements.

Acknowledgements

This research would not have been possible without the help offorestry workers employed by the Open Bay Timber Company, EastNew Britain Province, Papua New Guinea, particularly DionisioQuinones, the late (great) Francis Yendkao, Eddy Kaukia, StevenTaubuso, Hendry Bapo and Albert Keso. We are much indebted tothe Kol, Baia, Makolkol and Baining communities for allowing usaccess to their customary land, embracing us as guests, and for pro-viding field guidance and assistance. We are likewise grateful tothe Open Bay Timber Company for permitting us to conduct thisstudy. The field research for this study was supported financiallyby grants from James Cook University, Australia Pacific ScienceFoundation, Oregon Zoo Futures Fund and the Skyrail RainforestFoundation. We thank David Warton and Yi Wang for their assis-tance with data analysis and two anonymous reviewers for theirvaluable contribution to the manuscript.

Appendix A. Supplementary material

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.foreco.2014.11.028.

References

Bates, D., Maechler, M., Ben Bolker, Walker, S., 2013. lme4: Linear mixed-effectsmodels using Eigen and S4. R package version 1.1–6. <http://CRAN.R-project.org/package=lme4>.

Bayliss-Smith, T., Hviding, E., Whitmore, T.C., 2003. Rainforest composition andhistories of human disturbance in Solomon Islands. Ambio 32, 346–352. http://dx.doi.org/10.1579/0044-7447-32.5.346.

Bivand, R., 2014. spdep: Spatial dependence: weighting schemes, statistics andmodels. R package version 0.5-74. <http://CRAN.R-project.org/package=spdep>.

Bonnell, T.R., Chapman, C.A., Reyna-Hurtado, R., 2011. Post-logging recovery time islonger than expected in an East African tropical forest. For. Ecol. Manage. 261,855–864. http://dx.doi.org/10.1016/j.foreco.2010.12.016.

Bradshaw, C.J.A., Sodhi, N.S., Peh, K.S.-H., Brook, B.W., 2007. Global evidence thatdeforestation amplifies flood risk and severity in the developing world. GlobalChange Biol. 13, 2379–2395. http://dx.doi.org/10.1111/j.1365-2486.2007.01446.x.

Bradshaw, C.J.A., Sodhi, N.S., Brook, B.W., 2009. Tropical turmoil: a biodiversitytragedy in progress. Front. Ecol. Environ. 7, 79–87. http://dx.doi.org/10.1890/070193.

Bremer, L.L., Farley, K.A., 2010. Does plantation forestry restore biodiversity orcreate green deserts? a synthesis of the effects of land-use transitions on plantspecies richness. Biodivers. Conserv. 19, 3893–3915. http://dx.doi.org/10.1007/s10531-010-9936-4.

Brockerhoff, E.G., Jactel, H., Parrotta, J.A., Quine, C.P., Sayer, J.A., 2010. PlantationForests and Biodiversity: Oxymoron or Opportunity? Biodiv. Conserv. 17, 925–951. http://dx.doi.org/10.1007/s10531-008-9380-x.

Carnus, J.-M., Parrotta, J.A., Brockerhoff, E.G., Arbez, M., Jactel, H., Kremer, A., Lamb,D., O’Hara, K., Walters, B., 2006. Planted forests and biodiversity. J. Forest. 104,65–77.

Catterrall, C.P., Kanowski, J., Lamb, D., Killin, D., Erskine, P.D., Wardell-Johnson, G.W.,2005. Trade-offs between timber production and biodiversity in rainforestplantations: emerging issues and an ecological perspective. In: Erskine, P.D.,Bristow, M. (Eds.), Reforestation in the Tropics and Subtropics of Australia UsingRainforest Tree Species. Rural Industries Research and DevelopmentCorporation, pp. 162–182.

Chapman, L.J., Chapman, C.A., 1996. Exotic tree plantations and the regeneration ofnatural forests in Kibale National Park, Uganda. Biol. Conserv. 76, 253–257.http://dx.doi.org/10.1016/0006-3207(95)00124-7.

Chazdon, R.L., 2003. Tropical forest recovery: legacies of human impact and naturaldisturbances. Perspect. Plant Ecol. Evol. Syst. 6, 51–71. http://dx.doi.org/10.1078/1433-8319-00042.

Chazdon, R.L., Letcher, S.G., van Breugel, M., Martínez-Ramos, M., Bongers, F.,Finegan, B., 2007. Rates of change in tree communities of secondary Neotropicalforests following major disturbances. Philos. Trans. R. Soc. Lond., B Biol. Sci. 362,273–289. http://dx.doi.org/10.1098/rstb.2006.1990.

Chazdon, R.L., Harvey, C.A., Komar, O., Griffith, D.M., Ferguson, B.G., Martínez-Ramos, M., Morales, H., Nigh, R., Soto-Pinto, L., van Breugel, M., Philpott, S.M.,2009a. Beyond reserves: a research agenda for conserving biodiversity inhuman-modified tropical landscapes. Biotropica 41, 142–153. http://dx.doi.org/10.1111/j.1744-7429.2008.00471.x.

Chazdon, R.L., Peres, C.A., Dent, D., Sheil, D., Lugo, A.E., Lamb, D., Stork, N.E., Miller,S.E., 2009b. The potential for species conservation in tropical secondary forests.Conserv. Biol. 23, 1406–1417. http://dx.doi.org/10.1111/j.1523-1739.2009.01338.x.

Coad, L., Burgess, N.D., Bomhard, B., Besancon, C., 2009. Progress on the Conventionon Biological Diversity’s 2010 and 2012 Targets for Protected Area Coverage.United Nations Environment Programme World Conservation MonitoringCentre (UNEP-WCMC), Cambridge, UK.

Davis, A.S., Jacobs, D.F., Dumroese, R.K., 2012. Challenging a paradigm: towardintegrating indigenous species into tropical plantation forestry. In: Stanturf, J.,Madsen, P. (Eds.), Forest Landscape Restoration: Integrating Natural and SocialSciences. Springer, pp. 293–308.

Dirzo, R., Raven, P.H., 2003. Global state of biodiversity and loss. Annu. Rev. Environ.Resour. 28, 137–167. http://dx.doi.org/10.1146/annurev.energy.28.050302.105532.

Edwards, D.P., Larsen, T.H., Docherty, T.D.S., Ansell, F.A., Hsu, W.W., Derhe, M.A.,Hamer, K.C., Wilcove, D.S., 2010. Degraded lands worth protecting: thebiological importance of Southeast Asia’s repeatedly logged forests. Proc. R.Soc. B 278, 82–90. http://dx.doi.org/10.1098/rspb.2010.1062.

Farwig, N., Sajita, N., Böhning-Gaese, K., 2009. High seedling recruitment ofindigenous tree species in forest plantations in Kakamega Forest, westernKenya. For. Ecol. Manage. 257, 143–150. http://dx.doi.org/10.1016/j.foreco.2008.08.022.

Francis, J.K., 1988. Kamarere. Department of Agriculture, Forest Service, SouthernForest Experiment Station, New Orleans.

Gardner, T.A., Barlow, J., Chazdon, R.L., Ewers, R.M., Harvey, C.A., Peres, C.A., Sodhi,N.J., 2009. Prospects for tropical forest biodiversity in a human-modified world.Ecol. Lett. 12, 561–582. http://dx.doi.org/10.1111/j.1461-0248.2009.01294.x.

Gibson, L., Lee, T.M., Koh, L.P., Brook, B.W., Gardner, T.A., Barlow, J., Peres, C.A.,Bradshaw, C.J.A., Laurance, W.F., Lovejoy, T.E., Sodhi, N.S., 2012. Primary forestsare irreplaceable for sustaining tropical biodiversity. Nature 478, 378–381.http://dx.doi.org/10.1038/nature10425.

104 E.C. Pryde et al. / Forest Ecology and Management 339 (2015) 96–104

Gotelli, N.J., Colwell, R.K., 2001. Quantifying biodiversity: procedures and pitfalls inthe measurement and comparison of species richness. Ecol. Lett. 4, 379–391.http://dx.doi.org/10.1046/j.1461-0248.2001.00230.x.

Hall, J.S., Harris, D.J., Medjibe, V., Ashton, P.M., 2003. The effects of selective loggingon forest structure and tree species composition in a Central African forest:implications for management of conservation areas. For. Ecol. Manage. 183,249–264. http://dx.doi.org/10.1016/S0378-1127(03)00107-5.

Holl, K.D., 1999. Factors limiting tropical rain forest regeneration in abandonedpasture: seed rain, seed germination, microclimate, and soil. Biotropica 31,229–242. http://dx.doi.org/10.1111/j.1744-7429.1999.tb00135.x.

Houghton, R., 2012. Historic changes in terrestrial carbon storage. In: Lal, R., Lorenz,K., Hüttl, R.F., Schneider, B.U., Braun von, J. (Eds.), Recarbonization of theBiosphere – Ecosystems and the Global Carbon Cycle. Springer, Netherlands,Dordrecht, pp. 59–82. http://dx.doi.org/10.1007/978-94-007-4159-1_4.

Jansen, P.A., Zuidema, P.A., 2001. Logging, seed dispersal by vertebrates, and naturalregeneration of tropical timber trees. In: Fimbel, R.A., Grajal, A., Robinson, G.(Eds.), The Cutting Edge: Conserving Wildlife in Logged Tropical Forests.Columbia University Press, pp. 35–59.

Johns, A.D., 1992. The influence of deforestation and selective logging operations onplant diversity in Papua New Guinea. In: Whitmore, T.C., Sayer, J.A. (Eds.),Tropical Deforestation and Species Extinction. Chapman and Hall, London, pp.143–147.

Kanowski, J., Catterrall, C.P., Proctor, H., Reis, T.M., Tucker, N.I.J., Wardell-Johnson,G.W., 2005. Biodiversity values of timber plantations and restoration plantingsfor rainforest fauna in tropical and subtropical Australia. In: Erskine, P.D.,Bristow, M. (Eds.), Reforestation in the Tropics and Subtropics of Australia UsingRainforest Tree Species. Rural Industries Research and DevelopmentCorporation, Canberra, Australia, Canberra, Australia, pp. 194–216.

Keenan, R.J., Lamb, D., Woldring, O., Irvine, T., Jensen, R., 1997. Restoration of plantbiodiversity beneath tropical tree plantations in Northern Australia. For. Ecol.Manage. 99, 117–131. http://dx.doi.org/10.1016/s0378-1127(97)00198-9.

Ladiges, P., Udovicic, F., Nelson, G., 2003. Australian biogeographical connectionsand the phylogeny of large genera in the plant family Myrtaceae. J. Biogeogr. 30,989–998. http://dx.doi.org/10.1046/j.1365-2699.2003.00881.x.

Laurance, W.F., 1997. Hyper-disturbed parks: edge effects and the ecology ofisolated rainforest reserves in Tropical Australia. In: Laurance, W.F., Bierregaard,R.O., Jr (Eds.), Tropical Forest Remnants: Ecology, Management, andConservation of Fragmented Communities. The University of Chicago Press,Chicago & London, pp. 71–84.

Laurans, M., Martin, O., Nicolini, E., Vincent, G., 2012. Functional traits and theirplasticity predict tropical trees regeneration niche even among species withintermediate light requirements. J. Ecol. 100, 1440–1452. http://dx.doi.org/10.1111/j.1365-2745.2012.02007.x.

Lentfer, C.J., Pavlides, C., Specht, J., 2010. Natural and human impacts in a 35000-year vegetation history in central New Britain, Papua New Guinea. Quatern. Sci.Rev. 29, 3750–3767. http://dx.doi.org/10.1016/j.quascirev.2010.08.009.

Letcher, S.G., Chazdon, R.L., 2009. Rapid recovery of biomass, species richness, andspecies composition in a forest chronosequence in northeastern Costa Rica.Biotropica 41, 608–617. http://dx.doi.org/10.1111/j.1744-7429.2009.00517.x.

Lugo, A.E., 1992. Tree plantations for rehabilitating damaged forest lands in thetropics. In: Wali, M.K. (Ed.), Ecosystem Rehabilitation Ecosystem Analysis andSynthesis. Academic Publishing, The Hague, pp. 247–255.

Mamede, M.A., de Araujo, F.S., 2008. Effects of slash and burn practices on a soil seedbank of caatinga vegetation in Northeastern Brazil. J. Arid Environ. 72, 458–470.http://dx.doi.org/10.1016/j.jaridenv.2007.07.014.

Marsden, S.J., Pilgrim, J., 2003. Factors influencing the abundance of parrots andhornbills in pristine and disturbed forests on New Britain, PNG. Ibis 145, 45–53.http://dx.doi.org/10.1046/j.1474-919X.2003.00107.x.

Mayr, E., Diamond, J.M., 2001. The Birds of Northern Melanesia. Oxford UniversityPress, Oxford.

McAlpine, J., Keig, G., Falls, R., 1983. Climate of Papua New Guinea. CSIRO.McClanahan, T.R., Wolfe, R.W., 1993. Accelerating forest succession in a fragmented

landscape: the role of birds and perches. Conserv. Biol. 7, 279–288. http://dx.doi.org/10.1046/j.1523-1739.1993.07020279.x.

Melo, F.P.L., Arroyo-Rodríguez, V., Fahrig, L., Martínez-Ramos, M., Tabarelli, M.,2013. On the hope for biodiversity-friendly tropical landscapes. Trends Ecol.Evol. 28, 462–468. http://dx.doi.org/10.1016/j.tree.2013.01.001.

Miettinen, J., Shi, C., Liew, S.C., 2011. Deforestation rates in insular Southeast Asiabetween 2000 and 2010. Global Change Biol. 17, 2261–2270. http://dx.doi.org/10.1111/j.1365-2486.2011.02398.x.

Mueller-Dombois, D., Fosberg, F.R., 1998. Vegetation of the Tropical Pacific Islands.Springer.

Myers, N., Mittermeier, R.A., Mittermeier, C.G., da Fonseca, G.A., Kent, J., 2000.Biodiversity hotspots for conservation priorities. Nature 403, 853–858. http://dx.doi.org/10.1038/35002501.

Nathan, R., Muller-Landau, H.C., 2000. Spatial patterns of seed dispersal, theirdeterminants and consequences for recruitment. Trends Ecol. Evol. 15, 278–285. http://dx.doi.org/10.1016/S0169-5347(00)01874-7.

Oksanen, Jari F. Guillaume Blanchet, Kindt, R., Legendre, Pierre, Minchin, Peter R.,O’Hara, R.B., Simpson, Gavin L., Sólymos, Péter, Henry, M., Stevens, H., Wagner,Helene, 2013. vegan: Community Ecology Package. R package version 2.0-10.<http://CRAN.R-project.org/package=vegan>.

Paijmans, K., 1973. Plant succession on Pago and Witori volcanoes, New Britain. Pac.Sci. 27, 260–268.

Paijmans, K., 1976. New Guinea Vegetation. Australian National University Press,Canberra.

Parrotta, J.A., 1995. Influence of overstory composition on understory colonizationby native species in plantations on a degraded tropical site. J. Veg. Sci. 6,627–636. http://dx.doi.org/10.2307/3236433.

Perfecto, I., Vandermeer, J., 2010. The agroecological matrix as alternative to theland-sparing/agriculture intensification model. PNAS 107, 5786–5791. http://dx.doi.org/10.1073/pnas.0905455107.

Putz, F.E., Zuidema, P.A., Synnott, T., Peña-Claros, M., Pinard, M.A., Sheil, D., Vanclay,J.K., Sist, P., Gourlet-Fleury, S., Griscom, B., Palmer, J., Zagt, R., 2012. Sustainingconservation values in selectively logged tropical forests: the attained and theattainable. Conserv. Lett. 5, 296–303. http://dx.doi.org/10.1111/j.1755-263x.2012.00242.x.

Quinn, G.P., Keough, M.J., 2002. Experimental Design and Data Analysis forBiologists. Cambridge University Press.

Ramage, B.S., Sheil, D., Salim, H.M.W., Fletcher, C., Mustafa, N.A., Luruthusamay, J.C.,Harrison, R.D., Butod, E., Dzulkiply, A.D., Kassim, A.R., Potts, M.D., 2013.Pseudoreplication in tropical forests and the resulting effects on biodiversityconservation. Conserv. Biol. 27, 364–372. http://dx.doi.org/10.1111/cobi.12004.

R Core Team, 2013. R: A Language and Environment for Statistical Computing.Shearman, P., Laurance, W.F., Bryan, J., 2012. Are we approaching ‘‘peak timber’’ in

the tropics? Biol. Conserv. 151, 17–21. http://dx.doi.org/10.1016/j.biocon.2011.10.036.

Sheil, D., Burslem, D.F.R.P., 2003. Disturbing hypotheses in tropical forests. TrendsEcol. Evol. 18, 18–26. http://dx.doi.org/10.1016/S0169-5347(02)00005-8.

Sodhi, N.S., Koh, L.P., Clements, R., Wanger, T.C., Hill, J.K., Hamer, K.C., Clough, Y.,Tscharntke, T., Posa, M.R.C., Lee, T.M., 2010. Conserving Southeast Asian forestbiodiversity in human-modified landscapes. Biol. Conserv. 143, 2375–2384.http://dx.doi.org/10.1016/j.biocon.2009.12.029.

Stephens, S.S., Wagner, M.R., 2007. Forest plantations and biodiversity: a freshperspective. J. Forest. 105, 307–313.

Uhl, C., Jordan, C., Clark, K., Clark, H., Herrera, R., 1982. Ecosystem recovery inAmazon caatinga forest after cutting, cutting and burning, and bulldozerclearing treatments. Oikos 38, 313–320. http://dx.doi.org/10.2307/3544671.

van Valkenburg, J., Ketner, P., 1994. Vegetation changes following humandisturbance of mid-montane forest in the Wau area, Papua New Guinea. J.Trop. Ecol. 10, 41–54. http://dx.doi.org/10.1017/S0266467400007707.

Wang, Y., Naumann, U., Wright, S.T., 2012. Mvabund–an R package for model-basedanalysis of multivariate abundance data. Methods Ecol. Evol. 3, 471–474. http://dx.doi.org/10.1111/j.2041-210X.2012.00190.x.

Wang, Y., Naumann, U., Wright, S.T., Warton, D.I., 2013. mvabund: statisticalmethods for analysing multivariate abundance data. R package version 3.8.4.<http://CRAN.R-project.org/package=mvabund>.

Wardell-Johnson, G.W., Kanowski, J., Catterrall, C.P., McKenna, S., Piper, S., Lamb, D.,2005. Rainforest timber plantations and the restoration of plant biodiversity intropical and subtropical Australia, in: Reforestation in the Tropics andSubtropics of Australia Using Rainforest Tree Species. Rural IndustriesResearch and Development Corporation, pp. 162–182.

Warton, D.I., 2008. Raw data graphing: an informative but under-utilized tool forthe analysis of multivariate abundances. Austral Ecol. 33, 290–300. http://dx.doi.org/10.1111/j.1442-9993.2007.01816.x.

Warton, D.I., Wright, S.T., Wang, Y., 2012. Distance-based multivariate analysesconfound location and dispersion effects. Methods Ecol. Evol. 3, 89–101. http://dx.doi.org/10.1111/j.2041-210X.2011.00127.x.

Wilcove, D.S., Edwards, D.P., Fisher, B., 2012. High Conservation Value or highconfusion value? Sustainable agriculture and biodiversity conservation in thetropics. Conserv. Lett. 5, 20–27. http://dx.doi.org/10.1111/j.1755-263X.2011.00209.x.

Willson, M.F., Crome, F., 1989. Patterns of seed rain at the edge of a tropicalQueensland rain forest. J. Trop. Ecol. 5, 301–308. http://dx.doi.org/10.1017/S0266467400003680.

Wright, S.J., 2010. The future of tropical forests. Ann. N. Y. Acad. Sci. 1195, 1–27.http://dx.doi.org/10.1111/j.1749-6632.2010.05455.x.

Zuur, A.F., Ieno, E.N., Walker, N., Saveliev, A.A., Smith, G.M., 2011. Mixed EffectsModels and Extensions in Ecology with R. Springer.