Mammalian community responses to a gradient of land-use ...

Mammalian community responses to a gradient of land-use intensity on the island of Borneo

Oliver Richard Wearn

Division of Ecology and Evolution Department of Life Sciences

Imperial College London

A thesis submitted for the degree of Doctor of Philosophy 2015

iii

Copyright declaration

The copyright of this thesis rests with the author and is made available under a Creative Commons

Attribution Non-Commercial No Derivatives licence. Researchers are free to copy, distribute or

transmit the thesis on the condition that they attribute it, that they do not use it for commercial

purposes and that they do not alter, transform or build upon it. For any reuse or distribution,

researchers must make clear to others the license terms of this work.

iv

Abstract

Southeast Asian rainforests have, in recent decades, experienced the highest rates of deforestation and

logging across the major tropical regions. This has left a vast estate of degraded forest in the region,

which is under threat from further degradation and conversion to agriculture, principally for the

expansion of oil palm (Elaeis guineensis) plantations. However, knowledge of the relative

conservation value of different land-uses in the region is still limited, and a robust quantitative basis

for resolving land-use tradeoffs, in particular between agricultural yield and biodiversity, is lacking. I

aimed to assess terrestrial mammal species richness, abundance and β-diversity across a gradient of

land-use intensity (old-growth forest, logged forest and oil palm) in Sabah, Malaysian Borneo.

Overall mammal species richness was conserved even in the intensively-logged forests that I sampled,

as were the majority of apparent old-growth specialists. Oil palm, on the other hand, harboured a

highly depauperate mammal community. These broad patterns were echoed for overall mammal

abundance, as well as in the potential ecosystem functions of mammals that I examined. However, I

found evidence that the fundamental drivers of community assembly were altered across the land-use

gradient and so, consequently, were patterns of β-diversity. Fine-grained β-diversity, in particular,

was highest across the land-uses in logged forest, reflecting the increased environmental heterogeneity

in this habitat. In addition, community composition and species abundance were not stable across

land-uses. Omnivores and herbivores were more abundant in logged forest compared to old-growth

forest, and only a limited number of carnivore species persisted in oil palm. Invasive species increased

in abundance with land-use intensity.

My findings underline the conservation importance of the large areas of degraded forest in Southeast

Asia, and the extremely limited value offered by oil palm. If sustainability of the palm oil industry is

to be achieved, new plantations should be diverted away from all remaining unfragmented forests,

both old-growth and degraded forests alike.

v

Acknowledgements

Inevitably, the acknowledgements that follow are but the tip of the iceberg of the many and varied

people and organisations that, in no small part, helped me at all stages in the formation of this thesis.

To all those people not mentioned here, rest assured I remain wholly indebted to you.

Firstly, I acknowledge my principal funder, the Sime Darby Foundation, as well as the Zoological

Society of London, for providing the resources needed to complete this work. I acknowledge the

Economic Planning Unit and, latterly, the Sabah Biodiversity Council, for allowing me to conduct

research in Malaysia. I also appreciate the agreement from the Sabah Forestry Department, Yayasan

Sabah, the Maliau Basin Management Committee, Benta Wawasan and Sabah Softwoods to allow

research to be done in their respective areas. I am grateful for the support I received from the Royal

Society South East Asia Rainforest Research Programme and am especially indebted to Glen

Reynolds and his team for the huge amount of logistical assistance I received in-country. I extend my

thanks also to Henry Bernard, my local collaborator, for kindly obliging with all my requests for

research or logistical support in Sabah.

An emphatic word of thanks must go to my supervisory team: Rob Ewers, Chris Carbone and Marcus

Rowcliffe. Thank-you for all of the ideas and counsel you gave me during my Ph. D. I couldn’t have

asked for a better team and your supervision was always absolutely spot-on. I’ve learnt a lot from you

all and will miss the sometimes confusing, sometimes enlightening, meetings we’ve had over the

years. I am also grateful to my broader supervisory panel, Tim Coulson and E. J. Milner-Gulland, for

the insightful feedback they provided during the earlier stages of my Ph. D.

Of everyone involved in this thesis, I am most deeply indebted to the research assistants who helped

me to collect the data out in Borneo, often shedding blood (if the leeches got their way), sweat and

tears in the process, with apparently little reward. To Leah Findlay, Matt Holmes, Faye Thompson,

vi

Jack Thorley and Jess Haysom, thank-you all for being superlative in the field, always rising to meet

whatever challenge was thrown at us and, quite obligingly, for putting up with me in the field. You

have all learned a life-time supply of free beer on me. Thank-you also to Jeremy Cusack and Robin

Loveridge for being great field companions and, of course, for sharing the heavy work-load on Team

Tikus. It was an absolute pleasure to work with you.

Thank-you so much, also, to the many and varied people who helped form the tapestry of social life at

the SAFE “temporary camp” that we called a home from home. It pains me that I am unable to name

everyone individually here in the limited space available, but thanks in particular to the rest of the

“trinity”, Sarah Luke and Claudia Gray, as well as Timm Döbert, Terhi Riutta, Jen Sheridan and

James Rice for making field life that much more enjoyable.

I could not have achieved the back-breaking task of collecting data on elusive mammals without the

tireless work of staff at the SAFE Project. In particular, Matiew bin Tarongak, James Loh, Mohd

Sabri bin Bationg and Aleks Warat Koban bin Lukas should all be applauded for working so tirelessly

and diligently in the field. Even when the day involved lugging 100 steel traps up the face of a

slippery hillside, it was a pleasure with you guys. The help of Ed Turner, MinSheng Khoo, Johnny

Larenus, Sarah Watson and Ryan Gray was also absolutely invaluable in making the fieldwork even

remotely possible.

During my, admittedly limited, time spent in the UK working in my two institutions – both at Silwood

Park and at the Institute of Zoology – I’ve also had the great pleasure of working with a fantastic

bunch of people over the years. I thank everyone I’ve ever shared an office with for making the write-

up just that bit more bearable and in particular all of the “FEC” lab group, both past and present. I am

especially grateful for the inordinate amount of help that Marion Pfeifer provided with all things

RapidEye.

vii

Lastly, I am deeply grateful to my family for their constant support and encouragement, and for their

unquestioning accommodation of my tropical wanderlust over the years before and during the making

of this thesis.

viii

Declaration of originality

R. M. Ewers, C. Carbone and J. M. Rowcliffe contributed at all stages of the thesis, including the

conception, design and analysis. All further assistance received is detailed below.

Chapter 1: R. M. Ewers and E. J. Milner-Gulland commented on an earlier draft of this chapter.

Chapter 2: R. M. Ewers, J. M. Rowcliffe, C. Carbone and H. Bernard provided extensive comments

on this chapter. M. Linkie and an anonymous reviewer also greatly helped to improve the manuscript

during submission to PLoS One.

Chapter 3: R. M. Ewers, C. Carbone, J. M. Rowcliffe and H. Bernard all provided comments on an

earlier draft of this chapter, which greatly assisted in improving the clarity of the presentation.

Chapter 4: R. M. Ewers and C. Carbone provided helpful comments on this chapter. M. Pfeifer pre-

processed the RapidEye satellite images and ran the models of above-ground live tree biomass, of

which the mapped outputs featured in my modelling of community assembly. These maps were

provided as part of an ongoing collaborative effort at the SAFE Project.

Chapter 5: R. M. Ewers, J. M. Rowcliffe and C. Carbone provided comments on this chapter. M.

Tobler provided helpful feedback on an earlier version of one of the hierarchical models used. The

Imperial College High Performance Computing facility was used to run the bulk of the analysis. M.

Pfeifer’s above-ground live tree biomass estimates were again used in this chapter.

Chapter 6: R. M. Ewers, J. M. Rowcliffe and C. Carbone all provided comments on this chapter.

I herewith certify that all other material in this thesis is solely the result of my own work.

ix

Table of contents

Copyright declaration .......................................................................................................................... iii

Abstract ................................................................................................................................................. iv

Acknowledgements ............................................................................................................................... v

Declaration of originality .................................................................................................................. viii

Table of contents .................................................................................................................................. ix

Chapter 1: Land-use change and its ecological impacts in Southeast Asia: A review .................... 1

1. 1. Introduction ................................................................................................................................. 1

1. 2. The impacts of land-use change on species and communities – what do we know? .................. 3

1. 2. 1. The influence of the land-use change process ..................................................................... 3

1. 2. 2. The influence of the spatial pattern and composition of land-uses in the landscape .......... 6

1. 3. Land-use change and fragmentation in a Southeast Asian context ........................................... 12

1. 4. Current understanding of the ecological impacts of selective logging in Southeast Asia ........ 16

1. 5. Biodiversity in plantation forests and croplands in Southeast Asia .......................................... 21

1. 6. Confounding factors in the assessment of ecological responses to land-use change ................ 25

1. 7. Previous studies of mammalian biodiversity in anthropogenic landscapes in Southeast Asia . 33

1. 8. Research objectives ................................................................................................................... 40

1. 9. Sampling methods, design and study site ................................................................................. 40

1. 10. Thesis chapter outline ............................................................................................................. 43

Chapter 2: Assessing the status of wild felids in a highly-disturbed commercial forest reserve in Borneo and the implications for camera trap survey design .......................................................... 47

Abstract ............................................................................................................................................. 47

2. 1. Introduction ............................................................................................................................... 48

2. 2. Methods ..................................................................................................................................... 51

2. 2. 1. Study area .......................................................................................................................... 51

2. 2. 2. Data collection .................................................................................................................. 52

2. 2. 3. Data analysis ..................................................................................................................... 54

2. 3. Results ....................................................................................................................................... 55

2. 4. Discussion ................................................................................................................................. 60

Appendices ........................................................................................................................................ 66

Appendix A – Camera-trapping data from previous studies (1998-2012) .................................... 66

Chapter 3: Grain-dependent responses of mammalian species richness and β-diversity to land-use and the implications for managing conservation values in tropical human-modified landscapes ............................................................................................................................................ 71

x

Abstract ............................................................................................................................................. 71

3. 1. Introduction ............................................................................................................................... 72

3. 2. Methods ..................................................................................................................................... 75

3. 2. 1. Study sites and sampling design ....................................................................................... 75

3. 2. 2. Mammal sampling ............................................................................................................ 77

3. 2. 3. Data analysis ..................................................................................................................... 78

3. 3. Results ....................................................................................................................................... 82

3. 4. Discussion ................................................................................................................................. 86

Appendices ........................................................................................................................................ 92

Appendix A – Detailed study site descriptions ............................................................................. 92

Appendix B - Quantifying β-diversity .......................................................................................... 94

Appendix C – Supplementary results ............................................................................................ 96

Appendix D – Mammal species list ............................................................................................ 100

Chapter 4: Anthropogenic land-use change alters the ecological processes assembling tropical rainforest mammal communities ..................................................................................................... 102

Abstract ........................................................................................................................................... 102

4. 1. Introduction ............................................................................................................................. 103

4. 2. Methods ................................................................................................................................... 107

4. 2. 1. Sampling design .............................................................................................................. 107

4. 2. 2. Field methods .................................................................................................................. 109

4. 2. 3. Data analysis ................................................................................................................... 110

4. 3. Results ..................................................................................................................................... 114

4. 3. Discussion ............................................................................................................................... 121

Appendices ...................................................................................................................................... 127

Appendix A – Detailed description of methods used to measure environmental variables ........ 127

Appendix B – Supplementary Results ........................................................................................ 132

Chapter 5: Species abundance across a gradient of tropical land-use intensity: a hierarchical multi-species approach applied to a Bornean mammal community ............................................ 137

Abstract ........................................................................................................................................... 137

5. 1. Introduction ............................................................................................................................. 138

5. 2. Methods ................................................................................................................................... 142

5. 2. 1. Sampling design .............................................................................................................. 142

5. 2. 2. Field methods .................................................................................................................. 143

5. 2. 3. Data analysis ................................................................................................................... 145

5. 3. Results ..................................................................................................................................... 151

5. 4. Discussion ............................................................................................................................... 160

xi

Appendices ...................................................................................................................................... 168

Appendix A – Supplementary figures ......................................................................................... 168

Appendix B – Supplementary tables ........................................................................................... 179

Appendix C – BUGS (Bayesian inference Using Gibbs Sampling) code used to obtain Markov Chain Monte Carlo (MCMC) samples of the joint posterior ...................................................... 181

Chapter 6: The conservation of terrestrial mammals in human-modified landscapes in Southeast Asia: richness, composition, abundance and heterogeneity ........................................ 185

6. 1. Key findings of this thesis ....................................................................................................... 185

6. 2. Putting the findings into context ............................................................................................. 187

6. 3. New research directions .......................................................................................................... 189

6. 4. Implications for conservation .................................................................................................. 194

6. 5. The scope of inference ............................................................................................................ 197

6. 6. Future work ............................................................................................................................. 199

6. 7. Avoiding Navjot Sodhi’s “Impending Disaster” ..................................................................... 200

Bibliography ...................................................................................................................................... 202

Appendices ......................................................................................................................................... 235

Appendix A – Species checklists for four study sites in south-east Sabah, Malaysia ................ 235

Appendix B – Supplementary figures of sampling effort over time ........................................... 240

Appendix C – Example camera trap images of Bornean mammal and bird species................... 242

Chapter 1: The impacts of land-use change on Southeast Asian biodiversity

1

Chapter 1:

Land-use change and its ecological impacts in Southeast Asia: A review

1. 1. Introduction

Global rates of habitat loss and degradation over the last half-century have been unprecedented in

human history (Ramankutty & Foley, 1999; Goldewijk, 2001; Gibbs et al., 2010). Most of the

changes in land-use wrought by man, both presently and historically, have been to convert natural

forests into agricultural land, including cropland, pasture and plantation forest (Ramankutty & Foley,

1999; Goldewijk, 2001). The natural forests of Europe, North America and China had largely

stabilized in area by 1960 (Ramankutty & Foley, 1999) and the recent upsurge in deforestation and

degradation is occurring with the expansion of agricultural land in the developing tropics (Green et

al., 2005; Gibbs et al., 2010). Land-use change is now entrenched as the leading cause of biodiversity

loss (Wilcove et al., 1998; Sala et al., 2000; Green et al., 2005; Schipper et al., 2008; Vié et al., 2009)

and is causing widespread population declines and local extinctions across many taxa (Ceballos &

Ehrlich, 2002; Vié et al., 2009; Collen et al., 2009; Hoffmann et al., 2010). Across terrestrial

vertebrates, habitat loss and degradation is implicated as an extinction threat process for 42% of

species assessed on the IUCN Red List (IUCN, 2014).

Of the three major tropical forest biomes, the fastest rates of deforestation and degradation are

occurring in Southeast Asia (Achard et al., 2002; Mayaux et al., 2005; Hansen et al., 2008). Over the

last decade, the islands of Sumatra and Borneo have experienced the fastest rates of forest loss in the

region, annually losing 2.7% and 1.3% of forest cover, respectively (Miettinen et al., 2011). For a

long time, accurate data on the extent of logging in the region was difficult to obtain, although it was

thought that the vast majority of remaining forest area had been designated for selective logging (e.g.

Meijaard & Sheil, 2007; Koh, 2007a; Giam et al., 2011), and logging was known to be occurring even

in protected areas (Curran et al., 2004). Recent mapping efforts have substantially improved


2

knowledge of the spatial extent and foci of past logging in the region (Bryan et al., 2013; Margono et

al., 2014; Gaveau et al., 2014). It is now estimated that 46% of Borneo’s remaining natural forests are

in a logged-over state (Gaveau et al., 2014), and 65% for the country of Indonesia (Margono et al.,

2014).

As well as playing host to widespread changes in land-use, Southeast Asia also harbours globally-

significant amounts of biodiversity, much of it endemic to the region (Myers et al., 2000; Sodhi et al.,

2004, 2009b); though it only covers 4% of the Earth’s land surface, it contains 20-25% of the animal

and plant species known to science (Woodruff, 2010). As a result of the high rates of deforestation

and logging, Southeast Asia is a focal region for species listed as threatened with extinction on the

IUCN Red List (Schipper et al., 2008; Sodhi et al., 2009b; IUCN, 2014), as well as for species which

are becoming increasingly threatened over time (Hoffmann et al., 2010). Indeed, much of the region

has been identified as one of the planet’s top three conservation priorities in the form of the Sundaland

biodiversity hotspot (Myers et al., 2000).

Given the current rates of habitat loss in the tropics and the ubiquity of anthropogenic landscapes, the

causal relationships between changes in land-use and declines in biodiversity might be assumed to be

well understood. Certainly, a disproportionately high amount of conservation research effort is

devoted to investigating this subject: 24% of papers from 3 high-impact conservation journals dealt

with habitat loss, degradation and fragmentation (Fazey et al., 2005). However, large gaps in our

understanding of the land-use impacts on biodiversity remain and, for many landscapes, we lack

information on even the most basic conservation biology of species and communities. This applies

especially to Southeast Asia, which is one of the world’s least known regions biologically. The island

of Borneo, especially, is a peak region of mammal species listed as Data Deficient on the IUCN Red

List (Schipper et al., 2008). With rates of tropical forest loss showing no signs of slowing (Bradshaw

et al., 2009), we urgently need better management and conservation tools for decision-makers in the

region if the “impending disaster” (Sodhi et al., 2004) is to be averted. An important step towards this


3

will be to bridge existing knowledge gaps and begin to construct empirically-supported models of

biodiversity at landscape scales. This will enable tradeoffs between biodiversity and yields in

production systems to be better reconciled, and also help identify any opportunities for biodiversity-

friendly practices which do not conflict with yields.

I here review the current state of land-use change research and identify 1) broad ecological trends that

have been empirically well-supported across landscapes and 2) neglected or newly-recognised aspects

to the land-use change process which could yield step-changes in our understanding of the ecological

impacts of land-use change. I take a forward-looking approach to the literature and attempt to avoid a

recapitulation of old debates that have largely been resolved or have been unproductive. Following

this, I review the overall state of knowledge on land-use change effects in Southeast Asia, ultimately

focussing on mammalian taxa. In doing so, I evaluated the available literature critically and gave less

weight to studies that were of short duration, limited sample size, or were obviously biased due to

study design. In light of this review, I present the aims of the current work, seeking to fill gaps in the

available knowledge. Finally, I briefly describe the study region and novel methods used in the

current thesis (more details can be found in the individual chapters) and outline the content of the

chapters to follow.

1. 2. The impacts of land-use change on species and communities – what do we know?

1. 2. 1. The influence of the land-use change process

It is useful to separate land-use impacts into those which are wrought by the change process itself (e.g.

log extraction or clearance of habitat) and those which are manifested by the properties of the new,

altered landscape (e.g. the areal extent and spatial configuration of different land-uses). This

distinction is rarely made, probably because the opportunities to study the change process itself are

relatively few: most studies commence after the altered landscape is already in existence (McGarigal

& Cushman, 2002). As a consequence of this, we know relatively less about the independent effects


4

of land-use change processes themselves, including how the speed and spatial pattern of the change

process may mitigate or aggravate impacts on biological communities.

The direct consequences of log extraction, clearance, enrichment planting or crop establishment are

more obvious for plants and, to some extent, for sedentary animal species, but are less obvious for

highly mobile species capable of escaping direct killing from the land-use change process. Where

land-use is altered, but not completely substituted for a new land-use type (e.g. during selective

logging), many species will be able to persist by temporarily emigrating from the focal areas of

disturbance. For example, the lar gibbon Hylobates lar and the banded leaf monkey Presbytis

melalophos were able to modify their ranging and adopt more cryptic behaviour (specifically, making

fewer calls) to avoid logging operations at Sungai Tekam, Peninsular Malaysia, all the while

remaining within their established home ranges (Johns, 1985). Bornean orangutans Pongo pygmaeus

generally do not maintain exclusive home-ranges and instead have been reported to temporarily

emigrate from active logging concessions altogether, only returning once the disturbance has ceased

(Davies, 1986; Morrogh-bernard et al., 2003; Ancrenaz et al., 2010). Individual-based simulation

models for sessile taxa have shown that the spatio-temporal pattern of logging disturbance may be

critical in determining population persistence for poorly-dispersing species with clumped

distributions, with higher probabilities of extinction if logging occurs in large contiguous blocks, as

opposed to at random in the landscape (Ramage et al., 2013a). However, equivalent simulations for

more mobile species, which may actively avoid the logging process itself, are currently lacking, in

part because of the paucity of suitable data on the movements of individual animals.

Where land-use type is completely substituted (e.g. forest to agricultural crop), the options available

are fewer and many small-bodied species (e.g. non-volant invertebrate taxa), as well as species which

rely on crypsis for avoiding predation, probably suffer large direct losses of individuals.

Hypothetically, more mobile species might avoid these initial population losses by: 1) seeking shelter

from the clearance process (e.g. in underground burrows or under deadwood); 2) fleeing to remnants


5

of the original habitat, or 3) dispersing from the landscape entirely. Which of these is applicable to

different species is almost entirely unknown, though basic life-history factors, such as the habitual use

of protected holes or burrows and a tolerance to open spaces, will probably increase survival rates

over this initial phase of land-use change. Many understorey-adapted forest birds rarely cross forest

gaps larger than 100 m (Laurance et al., 2002) and can become trapped in remnant habitat as it is

cleared; this resulted in a strong “crowding effect” of increased species richness and abundance in

remnant forest fragment during a clearance process observed in central Amazonia (Bierregaard &

Lovejoy, 1989; Ferraz et al., 2003). This influx of immigrants was associated with an increase in

territoriality in many of the species, implying increased competition for resources; this likely caused

the observed decreases in bird populations over the months that followed clearance (Bierregaard &

Lovejoy, 1989).

If we are to model and predict land-use change impacts with any reasonable accuracy, then species

responses to the land-use change process itself will need to be incorporated. Clearly, species which

disperse from the landscape will show an immediate, dramatic population crash at the landscape scale

(with potential knock-on effects on populations beyond the altered landscape), whilst species which

crowd into remnant habitat will show sudden population explosions at the scale of individual patches.

Species may show these responses facultatively, depending on the pattern and severity of the land-use

change process. Community-level parameters will likely also be biased by neglecting the impacts of

the change process itself. The most often used method of modelling species richness as a function of

land-use is to use the species-area relationship, which typical fits a power function of the form:

zcAS = , where S is the number of species, A is the area of habitat and z and c are constants (e.g.

Rosenzweig, 1995). This relationship will be wrongly parameterised if the clearance process has

given rise to temporary crowding effects. In addition, multiple lines of evidence, both theoretical

(Tilman et al., 1994; Ovaskainen & Hanski, 2002; Halley & Iwasa, 2011) and empirical (Ferraz et al.,

2003; Kuussaari et al., 2009; Wearn et al., 2012), now strongly support the idea that many

communities show gradual adjustments in species richness following changes in habitat area, giving


6

rise to an extinction debt (Tilman et al., 1994). Properly modelling this “relaxation” (Diamond, 1972)

process will necessitate characterising the initial state of the community after habitat change:

specifically, the degree to which communities are initially “supersaturated” due to crowding effects.

All other things being equal, a strong supersaturation effect will result in a larger initial extinction

debt, followed by an initially more rapid relaxation process.

1. 2. 2. The influence of the spatial pattern and composition of land-uses in the landscape

A more substantial corpus of knowledge has accumulated on the impacts of land-use change on

species and communities caused by the actual properties of the new, altered landscape. Much of this

has been accumulated by observational studies (McGarigal & Cushman, 2002), but some clear

patterns have nonetheless emerged. Most obviously, the nature of the alterations which occur to

natural habitat during land-use change will have a strong bearing on the species and communities that

are able to persist in a landscape. Typically, the complete conversion of natural habitat to an

agricultural crop will have a larger impact on biodiversity than partly converting natural habitat (such

as in some “jungle rubber” and shade-grown coffee systems) or introducing a new disturbance process

into natural habitat (as occurs during selective logging). This has been supported by global-scale

meta-analyses of land-use change effects, in which agricultural systems were significantly

impoverished in species richness compared to predominantly natural systems (Gibson et al., 2011;

Newbold et al., 2015). In addition, agricultural systems which mimic at least some of the structural

characteristics or microclimate conditions – such as incident light, humidity and temperature – of the

original habitat will support larger proportions of the original species pool (Jones et al., 2003;

Ranganathan et al., 2008; Gardner et al., 2009; Foster et al., 2011). This trend was supported when

plantation types were ranked by their perceived land-use intensity (Newbold et al., 2015), but was not

evident when shaded and un-shaded plantations were compared directly (Gibson et al., 2011).

When land-use alterations to a landscape are patchy, a complex mosaic of habitat types may result,

with fragments of natural habitat embedded in a matrix of one or more land-use types, such as


7

degraded forest, plantation forest, cropland, fallows or regrowth. In such landscapes, the areal extent

of land-use types in the new landscape has been the most consistently identified determinant of

species- and community-level properties (Hanski, 1998; Harrison & Bruna, 1999; Fahrig, 2003;

Radford et al., 2005; Ewers & Didham, 2006; Zurita & Bellocq, 2010). Such area effects have been

supported by both empirical (Bender et al., 1998; Connor et al., 2000; Debinski & Holt, 2000) and

theoretical (Macarthur & Wilson, 1967; Hanski, 1998; Hubbell, 2001; Flather & Bevers, 2002) studies

and at multiple scales, from individual patches (Turner, 1996; Debinski & Holt, 2000; Laurance et al.,

2002; Michalski & Peres, 2005; Watling & Donnelly, 2006), to local landscapes (Andrén, 1994;

Robinson et al., 1995; Mazerolle & Villard, 1999; Fahrig, 2002, 2003; Stephens et al., 2003; Radford

et al., 2005; Zurita & Bellocq, 2010). It is worth noting that, though not usually considered within the

remit of habitat fragmentation research, logged forests may be perceived by species as fragmented

landscapes when the intensity of logging disturbance is highly heterogeneous, as may be the case

when extraction rates reach very high levels in localised patches (Cannon et al., 1994; Berry et al.,

2008). Most studies of logging instead assume that logging disturbance is homogenous, in part

because of a paucity of data on the spatial variation in logging intensity (but see Asner et al., 2004).

Significant advances in our understanding of logging impacts may be gained by applying continuous

metrics akin to fragmentation metrics, such as the area of unlogged patches in a local landscape or

density of ‘soft’ edges (Broadbent et al., 2008), to these apparently uniform systems. Few studies of

this nature have been done to date, but in a forested system disturbed by fire Cushman & McGarigal

(2004) found that the composition of different habitat types surrounding a given location was an

important determinant of community composition.

The scale at which individual species respond most strongly to area effects will depend on their

mobility and distribution in a landscape: species which can move easily across a landscape will

exhibit little population sub-division and may respond to the total area of their preferred habitat

present in a landscape, whilst those with more limited exchange of individuals between sub-

populations, which are largely confined to remnants of intact habitat, may respond to the size


8

distribution of individual patches of preferred habitat. This has implications for the design of reserves

and conservation set-aside within production landscapes, which aim to maximise gains for

biodiversity in the smallest possible area. Simulation studies of the impacts of a patchy logging

process, albeit on tree communities, have shown that randomly distributed species communities (akin

to strongly-dispersing taxa) are most efficiently conserved simply by leaving a single large area

unlogged in the landscape, whilst aggregated species communities (akin to weakly-dispersing taxa)

are best conserved by setting aside a large number of small, intact patches of forest distributed across

the landscape (Potts & Vincent, 2008).

Negative effects of reductions in habitat area on species are usually hypothesised to be driven by: 1)

instantaneous sampling processes, especially for sedentary species; 2) the loss of key habitats or

resources (e.g. food or shelter), and 3) reductions in population size causing an increased stochastic

risk of extinction, for example due to environmental or demographic stochasticity. These drivers will

cause declines in population abundance or species richness over different time-scales: sampling

processes will result in immediate declines, whilst the reduced availability of resources (causing

deterministic population declines) and the effects of small population size (causing stochastic

declines) will both be manifested over the medium to long term. Although area effects are a

widespread finding, often observed in terms of a lower occupancy of smaller habitat areas, the relative

importance of the three main drivers of this pattern is poorly understood. This reflects a general

pattern that fragmentation studies are overwhelmingly focussed on patterns of change, rather than

mechanisms of change. In the few studies which have attempted to investigate these hypotheses

explicitly, the results have sometimes shown support for the resource-shortage (Burke & Nol, 1998;

Zanette et al., 2000) or the stochasticity hypothesis (Berger, 1990; Sekercioglu et al., 2002). Many

other studies have demonstrated that small populations are more prone to extinction (Schoener &

Spiller, 1987; Pimm et al., 1988; Hanski et al., 1995; Lima et al., 1996; Crooks & Suarez, 2001;

Krauss et al., 2003), but it is unclear in these cases if small populations went extinct because of

stochastic processes or because they were sink populations in deterministic decline, for example due


9

to resource shortages. It should also be noted that many other studies have not observed clear area

effects on individual species (Debinski & Holt, 2000; Brotons et al., 2003) or observed effects at odds

with the expectations (Dooley & Bowers, 1998; Brooker & Brooker, 2001), even though area effects

have been found to be marginally positive for species on average (Bender et al., 1998; Connor et al.,

2000; Haddad et al., 2015).

The aggregated area effects for individual species leave a more clearly observable (and far more

frequently studied) area effect on communities as a whole, in the form of the species-area relationship

(SAR). The SAR is one of the most widely-studied and most truly universal patterns in ecology, and

holds for a diversity of taxa and ecological systems (Macarthur & Wilson, 1967; Connor & McCoy,

1979; Rosenzweig, 1995; Drakare et al., 2006). When parameterised in continuous habitat, the SAR

may be useful for determining the number of species expected to exist in a given area of more-or-less

uniform habitat, such as a reserve or logging concession. It is also often applied in heterogeneous

landscapes, with parameterisation based on patches of preferred habitat, embedded in what is assumed

to be a more-or-less hostile matrix. Given that the SAR assumes a single habitable land-use type, its

application in heterogeneous landscapes is likely to be most successful when there is high “contrast”

between land-uses present in a landscape (e.g. forest fragments and an agricultural matrix habitat), or

for species which perceive the landscape in this way (Prugh et al., 2008). More recently, however, the

species-area model has been extended to include multiple land-uses, in order to incorporate the many

species which exhibit continuous, rather than binary, response functions across different types of land-

use (Tjørve, 2002; Triantis et al., 2003; Pereira & Daily, 2006; Koh & Ghazoul, 2010b). In addition to

this, Hanski et al. (2013) have modified the SAR to allow for metapopulation dynamics occurring in

fragmented landscapes, which may prove promising, in general, in the context of poorly-dispersing

taxa in heterogeneous environments. Indeed, in a meta-analysis of fragmentation effects on the

occurrence of species, Prugh et al. (2008) found that the strongest effects of patch isolation were in

forestry landscapes, rather than in agricultural or urban landscapes.


10

The influence of the spatial configuration of land-use types in a landscape, i.e. effects of habitat

fragmentation per se, have generally been found to be weaker than the effects of the areal extent of

different land-use types (Watling & Donnelly, 2006; Prugh et al., 2008), and are generally thought to

be restricted to landscapes with low levels of forest cover (Andrén, 1994; Pardini et al., 2010).

However, one notable exception to this is the effect of road networks introduced into continuous

habitat, a process which is expected to increase dramatically in the next half-century (Laurance et al.,

2014a). A rapidly-growing road ecology literature, albeit primarily in temperate and boreal systems,

has shown that many species avoid roads or suffer mortality due to collisions (Coffin, 2007; Fahrig &

Rytwinski, 2009), with bird and mammal occurrence near roads and other infrastructure being

depressed by approximately 30% across past studies (Benítez-López et al., 2010). If species habitually

avoid or cannot cross (for example due to fencing) roads, local population immigration rates will be

reduced, which may increase the risk of local population extinction (Hanski, 1998) and cause declines

in species richness (Chave & Norden, 2007). In addition, roads may introduce a broad range of edge

effects, including those mediated by abiotic mechanisms such as altered micro-climatic regimes

(Didham & Lawton, 1999; Grimbacher et al., 2006), as well as biotic mechanisms, such as the

changes in predation, competition and parasitism rates that have been observed in edge habitats more

generally (Paton, 1994; Donovan et al., 1997; Piper & Catterall, 2003; de Almeida et al., 2008). A

problematic aspect of road ecology studies, as in land-use change studies more broadly, has been the

suitability of control measurements (Fahrig & Rytwinski, 2009). The incidence of roads is often

correlated with other attributes of landscapes, such as habitat quality and extent, leading to uncertainty

over whether land managers and conservationists should prioritise the mitigation of road impacts or

not. In addition, roads interact in often complex ways with the societal context in which they are

introduced, leading to unforeseen indirect effects of roads, such as increased hunter access and

colonisation rates, which may be more substantial than the direct effects of fragmentation of habitat

(Wilkie et al., 2000; Laurance & Cochrane, 2001; Laurance et al., 2009).


11

Significant research effort has been devoted to elucidating the intrinsic characteristics of species

which make them more or less susceptible to population extinction under land-use change. Although

much of the research effort on this topic has focussed on birds (e.g. Castelletta et al., 2000; Lees &

Peres, 2008; Sodhi et al., 2010; Uezu & Metzger, 2011), and there is a persistent problem that many

species characteristics are highly correlated (Laurance, 1991; McKinney, 1997; Ewers & Didham,

2006), some broad generalisations are still possible. In particular, species which: 1) occur at low

density; 2) exhibit fluctuating populations; 3) are resource or micro-habitat specialists, or 4) have poor

dispersal abilities, are all likely to be more susceptible to extinction under land-use change

(McKinney, 1997; Tscharntke et al., 2002; Henle et al., 2004). If species responses to land-use change

are consistently determined by their intrinsic traits, this may give rise to “nestedness” patterns

(Patterson, 1987) along land-use gradients, i.e. species communities in progressively more species-

poor land-uses may be simple subsets of communities in species-rich land-uses. Nested patterns of

species occurrence have indeed been observed along gradients varying in one dimension, for example

in habitat fragments ordered by area (e.g. Lynam & Billick, 1999; Wethered & Lawes, 2003; Fischer

& Lindenmayer, 2005; Hill et al., 2011), suggesting that, at least in some cases, community change

can be modelled, and perhaps predicted, with reasonable accuracy (but see Mac Nally, 2007).

However, land use gradients often involve changes in many ecological properties of landscapes at

once. Moreover, there is often an interaction between the intrinsic characteristics of species with the

particular properties of a given landscape, such that no two landscapes function entirely consistently

(Wright et al., 1997; Henle et al., 2004). This means that models of land-use effects on communities

will need to sacrifice some generality if they are to be sufficiently accurate to be useful in a

management context and, ideally, to be parameterised on a landscape-by-landscape basis.

Much of the land-use change literature, thus far, has dealt with statistically confirming, or quantifying

effect sizes, of various landscape or species characteristics on biological responses. There is now a

pressing need to begin to use such statistical relationships to assess and predict the accumulated

biological responses over space and time. Available evidence suggests that it will be difficult to


12

construct simple models with high explanatory power for the multi-scale, multi-dimensional and

interdependent processes occurring in multiple-use landscapes, despite the promise of modified SARs,

for example. Instead, the way forward will likely require 1) empirically determining whether a

categorical description of land-use types is an allowable simplification for a given management

question in a given region, or whether additional complexity, such as habitat heterogeneity within

landscapes, is necessary (Cushman & McGarigal, 2004), 2) explicitly incorporating spatial scale into

models, including the potential for scale-dependent effects of land-use change (Wiens, 1989; Sax &

Gaines, 2003; Hill & Hamer, 2004), and 3) identifying groups of species which show similar

responses and which can therefore be modelled together (McIntyre & Hobbs, 1999; Vos et al., 2001;

Didham et al., 2012).

1. 3. Land-use change and fragmentation in a Southeast Asian context

The Southeast Asian region was almost entirely covered in closed tropical forest from at least the

early Holocene (Bird et al., 2005; Corlett, 2009a). Today, however, extensive areas of this original

forest have been cleared, degraded and fragmented (Stibig et al., 2003, 2004) due to agriculture,

plantation forestry and urban development. These large-scale changes in land-use have occurred in

relatively recent history compared to elsewhere in the world (Ramankutty & Foley, 1999), and the

history of land-use changes in the region may have important ramifications for the ecological patterns

and processes we observe today.

Small-scale clearance and settlement of forest areas had occurred ever since the first arrival of Homo

sapiens in Southeast Asia, 40,000-70,000 years before present (Corlett, 2007). Forest clearance

became more widespread with the spread of rice-cultivation in the last few thousand years (Corlett,

2009a) and, from the 15th Century, with the expansion of pepper grown for export to Europe and

China (Reid, 1995). However, it was not until the latter half of the 19th Century that changes in land-

use really began to define the ecology of large regions of Southeast Asia (Flint, 1994). This coincided

with the increasing political dominance of colonial powers, which resulted in the first commercial


13

logging (for a limited number of species, such as teak Tectona grandis, changal Neobalanocarpus

heimii, Bornean ironwood Eusideroxylon zwageri and Palaquium gutta for the production of “gutta

percha” latex) and also extensive conversion of natural forest to plantations, especially rubber Hevia

brasiliensis (Thang, 1987; Brookfield et al., 1995).

Logging at this time was, however, still dependent on human and animal power, and was largely

confined to the margins of navigable rivers in order to transport logs off-site (Johns, 1997). With the

dawn of mechanized logging in the 1950s, especially the development of the one-man chain-saw, as

well as soaring demand for hardwood logs (largely driven by Japan), commercial logging in its

modern form rapidly expanded in the region from the 1960s onwards (Brookfield et al., 1995). This

was also the first time that markets were developed for many of the dipterocarp species of Sundaland,

initially focussing on red merantis (Shorea spp.) but later expanding to other genera. Sundaland

lowland forests, and indeed Southeast Asian forest in general, are unusual in their high stocks of

saleable timber: most canopy trees belong to the Dipterocarpaceae family (Whitmore, 1984; Newbery

et al., 1992), of which the species can often be marketed together. These forests are, as a result, some

of the most productive in existence, with typical extraction rates of 14-24 trees ha-1 or a timber

volume of more than 100 m3/ha, and up to 72 trees ha-1 or 220 m3/ha possible in some areas; this

compares with 3-5 trees ha-1 in Amazonian terra firme forest or < 1 tree ha-1 in Central Africa, with

timber yields of usually less than 20 m3/ha (Johns, 1997; Putz et al., 2012). This high-stocking of

forests in the region fuelled a logging boom, with the volume of dipterocarp timber exported from just

the island of Borneo (East Malaysia and Kalimantan, Indonesia) becoming greater during this period

than the timber exports from all of the Neotropics and Afrotropics combined (Curran et al., 2004).

Reflecting this boom, the Malaysian state of Sabah saw primary forest cover fall from 55% of land

area in 1973 to just 25% a decade later (Dauvergne, 1997).

Logging operations also have substantial indirect effects on forests, even if parts of a concession

remain unlogged. They bring increasing numbers of people into an area either directly or indirectly


14

employed by the logging operation, and they increase the accessibility of forests by building roads,

which can facilitate hunting, mining and illegal logging, as well as spontaneous colonisation by slash-

and-burn farmers (Bennett & Gumal, 2001; Meijaard et al., 2005). Indeed, due to the high intensity of

logging in Southeast Asian dipterocarp forests, the density of logging roads is much higher than

elsewhere in the tropics: Gaveau et al. (2014) mapped 0.5 km of roads per km2 of forest logged in

Borneo (between 1973 and 2010), which is an order of magnitude higher than similar estimates for

Central Africa at 0.03 km/km2 (Laporte et al., 2007). Logging also substantially increases the risk of

fires, which can burn vast areas in El Nino years (Siegert et al., 2001). Perhaps most importantly,

logging decreases the perceived, as well as marketable, value of natural forest, and therefore much

reduces any opportunity costs of converting such areas to plantations or cropland (McMorrow &

Talip, 2001; Koh & Wilcove, 2008; Fitzherbert et al., 2008; Berry et al., 2010). Heavily-degraded

logged forests have been left out of recent efforts in Indonesia to control deforestation (Edwards &

Laurance, 2011) and, in general, are less likely to be deemed of high conservation value in landscape

planning exercises.

By the 1990s, log production had declined in many countries in the region due to unsustainable

harvesting practices (prompting logging bans and temporary moratoriums of various forms in

Thailand, Vietnam, Cambodia and the Philippines) and the role of plantation forestry increased,

principally using teak, rubber, Eucalyptus spp., Acacia spp., Paraserianthes (Albizia) falcataria,

white teak Gmelina arborea, Anthocephalus chinensis and Pinus spp.. Plantation forests, mostly at the

expense of logged forest, now dominate entire landscapes in many regions of Southeast Asia,

especially in Thailand and Vietnam, where they represent more than 20% of the forest estate (FAO,

2011). Few areas of continuous primary forest now remain in Southeast Asia, mostly in Indonesia,

Thailand and Malaysia (FAO, 2011).

In the last decade, a new threat to Southeast Asia’s forests has emerged in the form of oil palm Elaeis

guineensis agriculture. In the last two decades, the land area harvested for palm oil in Southeast Asia


15

has nearly quadrupled, from 2.7 million ha in 1991 to 10.7 million ha in 2011 (FAO, 2012). Indeed,

the world’s top three palm oil producers are found in Southeast Asia (Indonesia, Malaysia and

Thailand) and oil palm now dominates large areas of these countries; this is especially the case for

Malaysia, in which mature oil palm occupies 12.2% of total land area (FAO, 2012), and up to 13.4%

if the area of immature palms (0.4 million ha in 2005; Wicke et al., 2011) is included. Oil palm has

been a major driver of primary and logged forest loss: more than half of the expansion between 1990

and 2005 possibly occurred on previously forested land (Koh & Wilcove, 2008; but see Gibbs et al.,

2010), though this also includes plantation forests, such as Acacia mangium, and does not establish a

direct causal link between initial clearance and oil palm cultivation itself. Evidence at a local scale is

more conclusive, however (e.g. McMorrow & Talip, 2001; Abdullah & Nakagoshi, 2007; Ichikawa,

2007; Reynolds et al., 2011). The Indonesian province of Riau, on the island of Sumatra, is one such

example, in which ca. 85% of oil palm expansion between 1982 and 2007 occurred at the expense of

natural forest (Uryu et al., 2008). Plantation forestry is expanding rapidly in most countries in

Southeast Asia, especially Vietnam and Indonesia, but the expansion of oil palm has tempered this

and likely even caused a reduction in the plantation forest area reported for Malaysia (FAO, 2011;

Wicke et al., 2011).

Projections of population growth and per-capita intake of vegetable oil (whilst accounting for oil palm

yield improvements) suggest that the area of oil palm will need to increase more than two-fold, by 12

million ha, in order to meet demand by 2050, excluding any use in biodiesel production (medium

scenario; Corley, 2009). Much of this expansion of oil palm will occur in Indonesia, which is

predicted to increase its area under oil palm by 3-5 million ha in the next decade alone (Koh &

Ghazoul, 2010a), but an increasing proportion will also come from the agricultural power-houses of

Brazil and Colombia. The contribution of certain Southeast Asian nations which currently produce

only negligible amounts of palm oil is also rapidly increasing: Myanmar, Laos and Cambodia

currently do not report any land area under oil palm (FAO, 2012), but large-scale forest clearance

recently detected using satellite remote-sensing has already been linked to the establishment of new


16

plantations (Stibig et al., 2007). In contrast, Malaysia, which has spearheaded the growth in oil palm

production over the last three decades, will likely decrease in importance and is expected to reach its

maximum area of oil palm before 2020 (Shean, 2011).

1. 4. Current understanding of the ecological impacts of selective logging in Southeast Asia

The study of the impacts of logging has been the defining paradigm for ecological research in

Southeast Asia for at least the last three decades. The earliest ecological work done on the impacts of

logging was conducted within the context of forestry research, and was primarily concerned with the

management of timber stocks and the improvement of silvicultural systems (e.g. Wyatt-Smith, 1963;

Burgess, 1970; Tang, 1974; Whitmore, 1984; Appanah & Turnbull, 1998). Subsequently, the breadth

of research increased, extending to the impacts of logging on non-commercial, but nonetheless

ecologically important, species. This demonstrated the non-selective, random nature of “selective

logging”, in which trees species are killed by the logging process in proportion to their pre-logging

abundance and irrespective of tree size class (Johns, 1988). In addition, given that the disturbance

from this type of logging was in fact random with respect to species, high species richness is often

maintained, albeit at a larger spatial scale (Cannon et al., 1998). However, fig trees (Ficus spp.) may

experience high loss rates during logging: 75% of figs were lost at Sungai Tekam, Peninsular

Malaysia (Johns, 1987). This occurs since many species exhibit hemi-epiphytic growth forms (the

strangling figs, in the subgenus Urostigma) and host trees are commonly timber species. Figs,

particularly the hemi-epiphytic figs, are especially important for many frugivorous species, owing to

their large fruit crops, short fruiting intervals and weak inter- and intra-specific synchronicity in

fruiting, resulting in year-round fruit availability (Lambert & Marshall, 1991).

The high intensity of logging in Southeast Asian dipterocarp forests creates substantial habitat

heterogeneity at fine spatial scales, with observable vegetation phase changes occurring over tens of

metres (Cannon et al., 1994). Crown sizes of trees in Southeast Asia are relatively small compared to

the Neotropics (Putz & Appanah, 1987) and canopy gaps created by felling individual trees are


17

usually less than 600 m2. Nonetheless, “super-gaps” of > 2,000 m2 can form under the high-intensity

logging typical in Sundaland lowland dipterocarp forest (Sist et al., 2003). In logged forest in Sabah,

Malaysia, gaps were mostly less than 10 m wide (i.e. 100 m2) after 16 years of natural regeneration,

and very few were > 20 m across (Bebber et al., 2002). Habitat heterogeneity may also be created

over larger spatial scales, with ridgetops typically experiencing the greatest impacts (where skid trails,

access roads and log-landing areas are located) and less severe disturbance in valley bottoms (Johns,

1988). Very steep (> 35º) or rocky areas may avoid disturbance altogether, as may “islands” of habitat

surrounded by swamp forest (Cannon et al., 1994). In some cases, cable yarding may be employed on

steep slopes, in which case damage to vegetation may be even more severe than tractor-based log

extraction (Marsh & Greer, 1992). Overall, the amount of area directly affected by logging is usually

substantial: between 77 and 87% of logging concessions in West Kalimantan, Indonesia (Curran et al.,

1999), with as much as 17% of land area cleared just for transporting logs from felling site to sawmill

(Pinard et al., 2000a).

Immediately after logging, it has been suggested that fruiting and increased leaf flush may occur as a

response by the remaining trees to conditions resembling drought, as well as the increased light

availability (Chivers, 1974). Indeed, leaf production significantly increased following logging at the

Sungai Tekam study, as did fruit production along the edges of logging roads (Johns, 1988). In the

medium-term after logging, changes may occur due to the increased prevalence of canopy gaps,

which are otherwise rare in mature, closed canopy forest (ca. 1% of area; Newbery et al., 1992; Hill et

al., 2001; Sist et al., 2003). This leads to increased leaf production and tree growth rates at the

margins of gaps and also increased colonisation by pioneer species. In the high-intensity logging

disturbance typical in Southeast Asia, in which 62-80% of the canopy may be damaged (Johns, 1997),

such pioneer species can become dominant: for example, 25% of all trees (> 30 cm girth at breast

height) were of a single species, Macaranga hypoleuca, in logged areas in Sabah, Malaysia (Heydon

& Bulloh, 1997). Larger open areas created by logging, such as log landing areas and main haulage

roads, may be subjected to topsoil loss, erosion and compaction. In this case, regeneration may be


18

severely arrested and only grasses and herbaceous vegetation may be able to colonise in the medium-

term.

In the longer-term, further changes in tree species composition may occur with disrupted plant-animal

interactions, including herbivory, pollination, seed dispersal and seed predation (e.g. Ickes et al.,

2001; Brodie et al., 2009; Sethi & Howe, 2009; Harrison et al., 2013). For strangling fig trees, even if

dispersal agents persist, pre-logging densities are unlikely to be restored over the short- to medium-

term, since they require high canopy host trees upon which to germinate (Johns, 1987), which in turn

may take centuries to grow (Loader et al., 2011). Indeed, in general, logging may be detectable in the

floristic composition of tropical forests for centuries following the initial disturbance (Brearley et al.,

2004; Liebsch et al., 2008), which matches estimated rates of tropical forest recovery from other

disturbance processes evident in the palynological record (Cole et al., 2014). There is also evidence,

from Borneo’s lowland dipterocarp forests, that regional-scale logging and clearance of forest can

lead to the persistent failure of canopy tree recruitment in remaining protected forests, even if

protected areas are large (900 km2) and contiguous (Curran et al., 1999). Logging likely disrupts the

large-scale satiation of seed predators, which is necessary for recruitment in many dipterocarp species

(Curran & Leighton, 2000), and may also alter responses to the specific climatic conditions which are

ordinarily thought to induce flowering and mast fruiting (Curran et al., 1999).

Increasing concern for wildlife depletion in logged forests in the 1970s led to the first concerted

studies of logging impacts on wildlife in Southeast Asia, beginning with easily-observed, diurnal

vertebrates, including birds, primates and other large mammals (Davies & Payne, 1982; Wilson &

Johns, 1982; Johns, 1986a, 1986b, 1987; Wong, 1986; Lambert, 1992). These early studies

established that the species richness of logged areas was similar to old-growth forest and that few

species were extirpated by logging: for example, studies in lowland dipterocarp forest in Sabah,

Malaysia, showed that between 87 and 96% of bird species recorded in old-growth forest were also

recorded in logged forest (Johns, 1992; Lambert, 1992) and all six species of primate were still


19

present at the Sungai Tekam study site after it was logged six years earlier (Johns, 1986a). We now

have a broader taxonomic picture of logging impacts in Southeast Asia and these results have

generally been corroborated for a range of vertebrate and invertebrate groups, including bats (Fukuda

& Tisen, 2009; Furey et al., 2010), non-volant small mammals (Wells et al., 2007; Bernard et al.,

2009), butterflies (Willott et al., 2000; Ghazoul, 2002; Hamer et al., 2003; Dumbrell & Hill, 2005, but

see Hill et al., 1995), moths (Willott, 1999), dung beetles (Davis et al., 2001; Slade et al., 2011), ants

(Berry et al., 2010; Woodcock et al., 2011) and termites (Eggleton et al., 1997, 1999; Gathorne-Hardy

& Jones, 2002, but see Donovan et al., 2007). Recently, two meta-analyses of logging impacts on

biodiversity have extended the generality of this finding across all tropical forest regions (Gibson et

al., 2011; Putz et al., 2012).

However, this widely-repeated finding is a simplification in at least two respects. Firstly, logged

forests rarely retain all of the species occurring in old-growth forest and these species extirpations,

even if few in number, may have a disproportionate impact on global biodiversity if they occur mostly

amongst highly threatened, restricted-range species. It is, at least, a relatively robust macroecological

pattern that species occurring at low population density – which, all else being equal, will be more

susceptible to reductions in the availability of habitat or resources – often have a small geographic

range (Brown, 1984). There was no evidence that old-growth forest-dependents were also restricted-

range species for bird communities either in Central Kalimantan, Borneo (Cleary et al., 2007) or on

the island of Seram, Indonesia (Marsden, 1998). Some support for this, however, was found amongst

butterfly species in Borneo: although restricted-range species mostly persisted in logged forest, they

exhibited particularly depressed populations in this habitat compared to old-growth forest (Cleary &

Mooers, 2006). Indeed, evidence from a number of studies, albeit fragmentary, appears to support this

for butterfly communities, more generally, in undisturbed and disturbed habitats in Southeast Asia

(Koh, 2007b). Amongst Bornean mammals, there was also a significant trend for phylogenetically

older species, which typically also have small geographic ranges (as well as dietary specialisations),

to be more affected by logging (Meijaard et al., 2008).


20

Secondly, the widespread focus on univariate measures, such as species richness, in order to gauge the

severity of logging impacts, may be concealing important detrimental effects at the level of species

communities. The proliferation of gap dynamics in logged-over forests creates winners and losers out

of communities of species, since species vary in their pre-adaptations to exploit such gap resources or

in their ability to exhibit the phenotypic plasticity necessary to do so. As a result, logging-induced

changes in species composition and trophic structure can be significant (Eggleton et al., 1997;

Ghazoul, 2002; Akutsu et al., 2006; Cleary et al., 2007; Edwards et al., 2009, 2011), although

significant decreases in species diversity or evenness have not usually been found (but see Ghazoul,

2002). It is worth emphasising that, since logged areas typically retain most of the species occurring

in old-growth forest, these composition changes are driven almost entirely by abundance changes,

rather than the loss of species. Even so, changes in the relative abundance of different species may

have important effects on species interactions and, in turn, on ecosystem function (Tylianakis et al.,

2008), particularly if abundance changes are correlated between functionally-similar species. Species

occurring at very low density may even be functionally extinct, even if their populations persist.

When these species, often known only from single records, are removed from species lists, this may

result in a dramatically altered picture of the severity of logging impacts on biodiversity (Barlow et

al., 2010).

Current knowledge of the ecological impacts of logging is also deficient in one increasingly important

way: the overwhelming majority of studies have been done in forests that have undergone only one

rotation of logging. Although cutting cycles of 60 years or more were envisioned for lowland

dipterocarp forest (Whitmore, 1984; Reynolds et al., 2011), most forests in the region have not been

managed on a sustainable basis, with re-entry logging commonly occurring after 15-30 years in

Sabah, Malaysia (Fisher et al., 2011a) and after a prescribed 35 years in Indonesia (van Gardingen et

al., 2003). As a result, an increasing proportion of Southeast Asia’s production forest has now been

logged repeatedly (Reynolds et al., 2011), or soon will be (van Gardingen et al., 2003; Samsudin et

al., 2010). Ecological research has not kept pace with this evolving state of land-use in the region:


21

very few studies have specifically investigated the biological responses to repeated logging. To my

knowledge, only a handful of studies, all conducted in the landscape surrounding the Danum Valley

Conservation Area, have compared once-logged and repeatedly-logged forest (Slade et al., 2011;

Edwards et al., 2011, 2014a; Woodcock et al., 2011). More broadly, many forest areas in Southeast

Asia are highly disturbed due to small-scale logging, the collection of non-timber forest products,

mining, grazing, fire and shifting agriculture (e.g. Salafsky et al., 1993; Brookfield et al., 1995; de

Jong, 1997; McMorrow & Talip, 2001; Paoli et al., 2001; Rao et al., 2002; Pattanavibool & Dearden,

2002; Goldammer, 2006; Ichikawa, 2007). Unlike for the Neotropics, where these practices are also

widespread (Peres et al., 2010), there is scant information on the extent and scale of these disturbance

processes and very little knowledge about the ecological consequences for forests in the region. Given

the large variability in the nature of disturbed forests in the region, a way forward may be to move

away from categorical descriptions of land-use types and begin to quantify biodiversity responses

along continuous metrics of land-use intensity. This poses the challenge of identifying which

characteristics of disturbed landscapes species respond to, and at what scale, but may allow better

integration of findings across studies and more precise predictions of the effects of land-use change.

Support for this has recently been provided by a meta-analysis of the effects of tropical forest logging,

in which a continuous metric – logging intensity, measured as the volume of timber extracted –

proved to be the most important predictor of species richness responses to logging, despite the broad

range of studies included (Burivalova et al., 2014).

1. 5. Biodiversity in plantation forests and croplands in Southeast Asia

The study of land-use types beyond disturbed natural forest has lagged behind in Southeast Asia and

we therefore have poor knowledge of the relative biodiversity value of the huge variety of plantation

and cropland types prevalent in the region. Nonetheless, we would a priori expect greater ecological

responses, for example in terms of forest species loss and compositional changes, to the clearance and

replacement of forest as compared to its disturbance, for example under intense logging. The

relatively limited number of studies done in Southeast Asia across all three types of land-use – old-


22

growth forest, logged/disturbed forest and agriculture – broadly support this for a range of taxonomic

groups, including mammals (Wilson & Johns, 1982; Duff et al., 1984; Danielsen & Heegaard, 1995;

Bernard et al., 2009; Rustam et al., 2012), birds (Waltert et al., 2004; Sodhi et al., 2005; Peh et al.,

2005), amphibians (Wanger et al., 2010), termites (Gathorne-Hardy et al., 2002; Jones et al., 2003),

isopods (Hassall et al., 2006), beetles (Chung et al., 2000; Davis et al., 2001), butterflies (Schulze &

Waltert, 2004), moths (Beck et al., 2002; Chey, 2006) and all arthropods combined (Turner & Foster,

2008). Using data provided in a meta-analysis for Southeast Asia (Sodhi et al., 2009), it possible to

state that agriculture in the region, including plantation forestry, leads to an estimated average decline

in species richness of birds, mammals and invertebrates by 35.9% (SE = 4.7, n = 87 comparisons,

from 23 separate studies), which is 15% greater than that caused by logging and disturbance of forest

alone.

Some exceptions to this broad pattern have, however, been noted. The single study incorporating

reptiles (Wanger et al., 2010) showed equivocal responses with respect to land-use, as did one study

of small mammals (Nakagawa et al., 2006). Bees exhibited conflicting patterns of change across land-

use, with the species richness and diversity being higher in oil palm than in some secondary forests,

although the community composition changes from old-growth forest were more dramatic in oil palm

compared to secondary forest, and abundance was lower (Liow et al., 2001).

This effect size is also strongly dependent on the specific land-use investigated, most likely due to

differences across land-use in structural complexity, available resources and the properties (including

variability) of their microclimates. For example, one emerging pattern is that plantation forests

typically retain greater biodiversity than other types of plantation or cropland. For example,

Paraserianthes (Albizia) falcataria plantations in Sabah, Malaysia, were found to harbour 64% of the

avifauna of surrounding old-growth forests, most likely because the relatively open canopy of P.

falcataria allows the development of a complex and diverse under- and mid-storey plant community,

as well as the abundant food resource provided by insect pests of the tree itself (Mitra & Sheldon,


23

1993). Similarly, Acacia mangium plantations, which have been the subject of recent research effort

in the 5,000 km2 Planted Forest Zone in Sarawak, were found to contain 41% of old-growth forest

bird species (Sheldon et al., 2010) and 74% of the mammal taxa detected overall (McShea et al.,

2009). A. mangium plantations support a relatively diverse under-storey plant community (although

the canopy is lower and denser than in P. falcataria plantations) and additionally provide seeds on the

trees themselves, which are eaten by frugivorous species (Styring et al., 2011). Although the sample

sizes are small, this general pattern is evident in a recent meta-analysis (Gibson et al., 2011):

plantation forests in Asia (including oil palm) were associated with a standardised effect size of 0.32

(n = 27 comparisons, from 6 studies, for a range of biological metrics relative to old-growth forest),

whilst agricultural croplands had a much larger effect size of 2.59 (n = 33 comparisons, from 5

studies).

Oil palm plantations are a special case in that, although tree-based and perennial, they are intensively

managed for an agricultural crop, in the form of oil palm fruit. A relatively large amount of research

effort has been devoted to oil palm in the last decade and, even if the vast majority of the oil palm

literature is not relevant to biodiversity (Turner et al., 2008) or does not actually present new field

data (Foster et al., 2011), we now know comparatively more about its value for biodiversity than

many other important agricultural crops in the tropics, especially relative to its global land area

(Balmford et al., 2012). In almost all cases, oil palm habitat has been shown to support highly

impoverished species communities, even when compared to other degraded or plantation forest

habitats (Fitzherbert et al., 2008; Foster et al., 2011), and was shown to support just 10% of forest bird

species and 23% of forest vertebrate species overall (Danielsen et al., 2009; Sheldon et al., 2010).

This is apparently to do with the more intensive management practices associated with oil palm,

including frequent human disturbance, understorey weed clearing and herbicide application, as well as

its low and dense canopy (at least after closure at 15-20 years). As a result, the plant diversity and

structural complexity of oil palm habitat is substantially lower than in other tree-based plantations

(Chung et al., 2000; Danielsen et al., 2009; Sheldon et al., 2010). Interestingly, rubber plantations are


24

similarly depauperate in forest vertebrate species (Aratrakorn et al., 2006), for example containing

just 10-24% of forest bird species (Danielsen & Heegaard, 1995; Peh et al., 2006), whilst this is not

true of rubber agroforestry (Beukema et al., 2007). As in the case of oil palm, rubber plantations are

typically managed on an intensive basis, with frequent understorey clearing and human disturbance.

This may explain the high degree of similarity observed between species communities in oil palm and

rubber plantations (Aratrakorn et al., 2006). Both habitats share species which have high vagility,

large geographic ranges and a preference for open habitats, that is to say, species which are low

priorities for conservation attention (Danielsen & Heegaard, 1995; Aratrakorn et al., 2006).

Beyond oil palm, rubber, P. falcataria, and A. mangium, our knowledge of the biodiversity values of

Southeast Asia’s plantation and agricultural lands is currently sparse. For example, almost no studies

have been made in plantations composed of teak, Eucalyptus spp. (e.g. camaldulensis, deglupta and

pelita), Pinus spp. (e.g. merkusii, carabaea and oocarpa), or coconut Cocos nucifera, or in certain

agricultural landscapes which are highly dominant in some local areas, such as tea Camellia sinensis

monoculture. As a result, it is difficult to make any further generalisations beyond those tentatively

noted above. In particular, two generalisations, which might be expected to be relatively robust, have

not yet been supported by the limited evidence available. Firstly, agroforestry systems are a common

land-use type in Southeast Asia and typically harbour a diverse tree community (including, in “jungle

rubber” systems, some Dipterocarpaceae and other native canopy trees) and high structural diversity

(Thiollay, 1995); we would therefore expect agroforestry to contain greater biodiversity value than

monoculture plantation forests, but this has not been borne out by the data available so far: 59% and

27.1% of forest bird species were found in mixed and coffee Coffea canephora agroforestry systems

in Sumatra, respectively (Thiollay, 1995; Philpott et al., 2007), which is broadly similar to the

percentages retained by plantation forests (see above), with the exception of rubber and oil palm.

Secondly, open land-use types with no tree cover would be expected to contain significantly fewer

species than other land-use types, on the basis that their properties are the most dissimilar to native

vegetation types, but this was not supported conclusively in the case of bird, butterfly and dung beetle


25

communities compared between cacao Theobroma cacao agroforestry and maize Zea mays fields

(Schulze & Waltert, 2004), nor for amphibians and reptiles compared between cacao agroforestry and

open pasture (Wanger et al., 2010). Clearly, further studies are warranted if we are to reconcile the

various conservation strategies available to decision-makers in the face of inevitable expansions in

agricultural land area over the coming decades (Green et al., 2005; Balmford et al., 2012).

1. 6. Confounding factors in the assessment of ecological responses to land-use change

In reviewing current understanding of the impacts of land-use change in Southeast Asia I have largely

followed the current consensus, as represented in the literature, and attempted to avoid reproducing

the conclusions of studies which were based on obviously limited, biased or otherwise flawed

datasets. As a result, the overall generalisations are most likely robust in qualitative terms. Even so,

there is reason to be cautious with respect to the quantitative underpinnings of most studies reviewed

here, and this is especially important if quantitative assessments of biodiversity loss, and indeed

predictions into the future, are to be made on the basis of current knowledge. There are four main

confounding factors that are present in many land-use studies in the region: 1) the exchange of

transient and dispersing individuals between land-uses; 2) extinction debts (both deterministic and

stochastic); 3) inadequate statistical controls, and 4) imperfect detection of species. I will deal with

each of these confounding factors in turn.

A commonly conceded problem with focussing on records of species presence and absence, and the

species lists they generate, is that the presence of a species does not necessarily equate with the

presence of a viable population in a particular land-use (Gardner et al., 2007; Gibson et al., 2011). In

other words, if the population growth rate (r) could be measured within the given land-use only (i.e.

assuming no migration between land-uses), it would not necessarily be greater than zero. This is

important, since the highest conservation priority should be assigned to land-uses with viable

populations. There are a number of ways that misleading species records in a particular land-use type

(i.e. from populations for which r would be less than 0) might arise. Species may be detected in a


26

plantation or other land-use type in which they derive no use simply because they are moving between

patches of habitat which are used (i.e. the problem of transient individuals). Species may also be

detected in a land-use type in which they may derive some use, but are unable to reproduce (i.e. the

problem of long-distance ranging). For example, flocks of birds were observed ranging into P.

falcataria plantations from surrounding forest to feed on the abundant caterpillars on plantation trees

(Mitra & Sheldon, 1993). In addition, bearded pigs and leopard cats are both known to forage outside

forest areas to exploit increased food resources in oil palm habitat, in the form of oil palm fruit and

small mammals, respectively (Ickes, 2001; Rajaratnam et al., 2007). Such species may not be able to

procure sufficient resources from the plantation forest alone and suitable conditions for reproduction

may not exist. Equally, populations in optimal habitat will produce a surplus of individuals which

would not survive due to competition for resources, but these individuals may disperse to surrounding

habitats even if they are sub-optimal (i.e. the problem of mass effects; Shmida & Wilson, 1985).

Each of these processes giving rise to misleading species records in a particular land-use type are a

by-product of being in proximity to a stable population source in optimal habitat, and the effects will

therefore likely decline with distance from this source. In this way they can be considered a type of

edge-effect, but in the opposite direction to that usually studied, i.e. from good-quality to poor-quality

habitat in this case. This reverse edge-effect, sometimes called a “spill-over” effect in the literature

(e.g. Barlow et al., 2010; Gibson et al., 2011), can be explicitly tested for and quantified. This can

start with the simple prescription that all species records are associated with data on the distance from

undisturbed, or otherwise optimal, habitat. Alternatively, land-use types should be studied in isolation,

where independence is judged relative to the ranging and dispersal abilities of focal taxa. This has

rarely been done in land-use studies thus far conducted in Southeast Asia, with most studies focussing

on plantations and cropland adjacent to large, intact forest areas. The extent of the bias caused by this

widespread practice is largely unknown, though in a global analysis of land-use effects, there was no

indication of larger effect sizes in studies conducted at study sites surrounded by natural habitats (i.e.

old-growth or logged forest; Gibson et al., 2011).


27

Aside from spill-over, species may also be detected in a land-use in which they are not viable due to

time-delays in extinction. If the ratio of births and deaths for a species is below replacement (i.e. r <

0), the population may temporarily exist in a “living dead” state (Janzen, 1986), until the population

gradually declines deterministically to extinction. Such species are said to represent a deterministic

“extinction debt” (Kuussaari et al., 2009). Strong evidence of long-term population declines in

specific land-uses is lacking, since most studies are short-term in length. However, populations of two

Cercopithecus primate species in logged forest in Uganda apparently showed continuing declines

nearly three decades after logging (Chapman et al., 2000), and there was evidence of reduced infant

recruitment in two Presbytis primate species at the Sungai Tekam study site even 12 years after

logging (Johns & Johns, 1995). Long-term extinction debts such as these may create a window of

opportunity to conserve species populations using management interventions which foster positive

population growth rates (e.g. by using artificial nest boxes, limiting the clearance of undergrowth in

managed forests and plantations or restoring patches of native habitat).

The final population trend is one which exhibits a stable or positive population growth rate (r ≥ 0),

which can be inferred from changes in abundance (and, ideally, survival and recruitment rates) over

time. In most cases, such populations represent the highest chance of survival in a landscape for a

given species and identifying them should be the focus, where possible, when assessing the

biodiversity value of a particular land-use. However, it is worth noting that even populations with r ≥

0 may not represent viable populations in a specific land-use, for two reasons. In the first case, this

can occur due to another type of reverse edge effect, and certainly the most cryptic, which is the

“rescue effect” (Brown & Kodric-Brown, 1977). The rescue effect is when individuals permanently

dispersing from a source population sustain a sink population indefinitely. This will manifest itself in

a deterministically declining population only once the source population is lost (unlike the sink

populations considered above, which were already declining), or it can be inferred from detailed

knowledge of demographic parameters and dispersal rates between land-uses. The second reason that

populations with r ≥ 0 may not be viable is really a re-statement of the fact that all populations go


28

extinct over an indefinite time horizon. However, this may have more immediate consequences for

population viability in a land-use when the effective population is small and therefore vulnerable to

extinction due to environmental and demographic stochasticity. Such populations represent a

stochastic, as opposed to demographic, extinction debt and may be pervasive in many land-uses.

Identifying the size of this particular debt in a given land-use is an ongoing challenge but, at the very

least, an attempt to formally estimate population size or density across land-uses would help to assess

its likely significance. Unfortunately this is still rarely done, with most studies relying on indices of

abundance (see below).

The next potential problem of confounding is associated with inadequate study design, arising either

from poor statistical practice or real-world constraints on sampling. Inadequate study design can take

many forms, such as low levels of treatment replication and low sample sizes within experimental

treatments, which may have an impact on the ability to discern treatment effects from sampling

variation (i.e. lead to Type II errors). However, potentially more serious is the lack of suitable

statistical controls, which may mean inferences about the effects of land-use change are entirely

confounded. Control measurements are necessarily taken at a different point in space or time to

measurements taken under the effect of a given treatment, in this case land-use change, but ideally

should be identical in all respects except the treatment itself. In practice, ecological systems vary in

space and time, for example due to differences in the prevailing edaphic properties, slope or elevation

across space, due to neutral turnover in species communities, or due to seasonal effects across time.

This means that laboratory-standard controls are rarely achievable in field ecology. This is especially

the case when the spatial and temporal design of experimental is highly constrained, as is most often

the case in studies of land-use change, in which opportunities to conduct manipulative experiments

are rare. In Southeast Asian landscapes, for example, land-uses are not randomly distributed with

respect to elevation, with plantations typically established below 400 m elevation, and remaining old-

growth forest typically occurring at higher elevations (Reynolds et al., 2011; Edwards et al., 2014b;

Scriven et al., 2015). Although every effort may be made at the design stage of a study to choose


29

treatment and control areas which are comparable, it may not always be possible in practice. In

addition, substantial a priori knowledge of an ecological system, often lacking in poorly-studied

tropical systems, may be required to identify all potential confounding factors at the design stage. This

underlines the importance of making strong a priori hypotheses about the ecological systems under

study.

The problem of confounding differences between treatment and control areas may be especially acute

in studies which are not replicated at the treatment level at all, because these differences will covary

perfectly with the treatment effect. This issue of “pseudoreplication” has been well known in

ecological field studies for several decades (Hurlbert, 1984; Heffner et al., 1996), and may be an even

more common feature in studies of the impacts of land-use change than in other areas of ecology. In

the most part, this is because land-uses within any given landscape are often highly segregated in

space, meaning that sampling locations within each land-use may not all be truly independent from a

statistical viewpoint. Recently, attention has been drawn to this in the context of studies assessing the

impacts of logging (Putz et al., 2012; Laufer et al., 2013; Ramage et al., 2013b). Indeed, just 7% of

studies on the effects of logging, across all tropical forests, were deemed to be definitely free of

pseudoreplication (Ramage et al., 2013b). For example, logging concessions are typically allocated in

contiguous blocks and managed separately to any adjacent old-growth forests, such as protected areas,

making the interspersion of experimental units difficult to achieve within a given landscape. Even if

interspersion were achievable, however, it is not always clear that it would be desirable within a given

study, given that it may increase the inferential problems associated with the exchange of individuals

among land-uses (see above). The scale of interspersion can be matched to the dispersal abilities of

focal taxonomic groups but, for highly vagile groups, this may mean interspersion is ultimately

required at scales which are impractical to sample at sufficient intensity and coverage within a single

study. It is also worth noting that it is clear that any impacts of logging which are perhaps

unjustifiably inferred from single, pseudoreplicated studies are nonetheless real: three large-scale

meta-analyses have confirmed that the ecological effects are significantly negative overall, albeit


30

modest compared to other forest disturbance processes (Sodhi et al., 2009; Gibson et al., 2011; Putz et

al., 2012)

There are no simple solutions to the challenges inherent in designing studies of land-use change. From

an analysis viewpoint, it may be possible to statistically control for some of the inherent differences

between experimental units, by explicitly including extra covariates during modelling. It may also be

possible to examine the residual spatial autocorrelation in the response variable, after land-use has

been accounted for, which may for example be indicative of a broad gradient in community

composition. Most importantly, however, the scope of inferences must be matched to the study design

in hand, and in particular to the specific nature of the land-use change process, the taxonomic groups

under study, as well as the spatial extent and scale of sampling. This may mean that single studies of

land-use change often cannot make abstract generalisations about the impacts of a land-use change

process per se – this is the realm of continental- and global-scale meta-analyses – but nonetheless are

able to make inferences which may be of sufficient scope to be relevant to political and management

decisions directly affecting land-use.

The fourth main confounding factor present throughout the land-use impacts literature for Southeast

Asia is the imperfect detection of species. Ecological studies typically collect data on the abundance

or occurrence of a species in count or binary form, respectively. However, for the majority of species

which cannot be completely censused, these data are negatively biased with respect to the actual

abundance or occurrence. The formal link between such count or binary data and absolute abundance

can be made by quantifying the extent of the imperfect detection problem, i.e. by statistically

estimating the probability of detecting a species or an individual of that species (Williams et al.,

2002). Crucially, detection probability may vary due to 1) the properties of the detector (which may

be a human observer, trap or electronic device), 2) the characteristics of the species or individual

(such as its behaviour, crypsis, movement rates, habitat use and interest in any baits or lures used),

and 3) spatial and temporal influences which interact with these properties of detectors and


31

species/individuals (for example, effects of weather patterns through time or spatial variation in

vegetation structure). Failing to control for detection probability can lead to quantitatively biased

results, which will compromise any ability to accurately assess and predict biodiversity losses, but

will often not affect the qualitative conclusions drawn; this depends, however, on a monotonic

relationship existing between the observed data and actual occurrence or abundance, and even

qualitative conclusions may be entirely misleading where this does not exist. In addition, use of count

or binary data in this way is only indicative of relative differences in abundance or occurrence of a

species, for example over space or time, and is not suited to making comparisons between species

with substantially different ecology, behaviour and morphology.

Many of the earliest studies in Southeast Asia are known to have various sampling design

shortcomings, most obviously with respect to controlling for imperfect detection (Lambert, 1992;

Sheldon et al., 2010). Despite this, some of these studies are the only such examples of work done on

a particular taxon or in a particular land-use and therefore the inferences drawn from them can have

disproportionate influence. For example, one of the only longitudinal studies to date of the impacts of

logging on bird and mammal communities (e.g. Johns, 1986a, 1989, 1992) was entirely conducted

along roads in the logged forest areas and along cut transects in the old-growth forest. Detection

probabilities along roads are very unlikely to be comparable with those made inside the forest itself

and such sampling will also bias detections in logged forest towards species which frequent the

specific vegetation characteristics found along the edges of roads (e.g. nectarivorous birds and

herbivorous mammals). Although the problem of imperfect detection is well-known in the vertebrate

ecology literature, it is more rarely considered in studies of invertebrates, though similar biases are

likely to be present. For example, moths are typically sampled using light traps, the effectiveness of

which is thought to vary depending on many factors, including vegetation density (Willott, 1999). As

a result, studies of the impacts of land-use on moths in Southeast Asia have been unable to make

unbiased comparisons of moth abundance, typically focussing instead on species richness or diversity

controlled for the varying sample sizes (e.g. Chey et al., 1997; Willott, 1999; Beck et al., 2002).


32

The problem of imperfect detection is a very active area of statistical research and a number of

methods are available which help control for this confounding factor (Iknayan et al., 2013; Dénes et

al., 2015). For species-level parameters, such as abundance and occurrence, robust estimation

methods exist which rely on replicate observations (e.g. over space or time, or by different observers)

to confront the seemingly intractable problem of estimating detection probability and abundance or

occurrence at the same time (Otis et al., 1978; MacKenzie et al., 2002). Although sometimes held up

as a statistical panacea, these models, which include capture-recapture- and occupancy-based

approaches, often suffer from estimation issues when sample sizes are low (Guillera-Arroita et al.,

2010; Marques et al., 2011) and can be sensitive to violations of their assumptions (Otis et al., 1978;

Kendall, 1999; Rota et al., 2009; Harmsen et al., 2010; Welsh et al., 2013), some of which require

further empirical validation (Rota et al., 2009; Efford & Dawson, 2012). Even so, it remains the case

that accounting for detectability much reduces the chances of making wrong inferences compared to

using naïve estimators such as abundance indices (Sollmann et al., 2013; Guillera-Arroita et al.,

2014), albeit at the cost of requiring further data collection. In their simplest form, these models

assume “closure” (i.e. no changes in abundance or occupancy occur) but, given suitable data, these

models can also be extended to the dynamic case, allowing insights to be made about the viability of

populations across land-use, for example by providing estimates of population extinction, colonisation

and turnover rates (MacKenzie et al., 2003), as well as individual vital rates (Pollock, 1982; Gardner

et al., 2010).

For species richness and diversity, estimates can be made at a common sample size using rarefaction

or, alternatively, asymptotic estimates can be obtained using non-parametric methods which also

attempt to correct for the number of unseen species (Bunge & Fitzpatrick, 1993; Boulinier et al.,

1998; Chao & Shen, 2003; Chao et al., 2005). In principle, these methods can correct for changes over

space or time in the overall efficiency of the detection process, but may not perform well if species-

level detection probabilities are strongly heterogeneous, especially if this heterogeneity is itself

manifested over space or time (Iknayan et al., 2013). Recently, a more process-based approach to


33

estimating species richness (or indeed diversity) has emerged, in which detection probabilities for

each species are modelled as a function of site- or survey-specific covariates, all within the framework

of a multi-species hierarchical model (Dorazio et al., 2006). These “metacommunity” models have

also been extended to the case when species richness is a dynamic balance of colonization and

extinction over time (Dorazio et al., 2010).

1. 7. Previous studies of mammalian biodiversity in anthropogenic landscapes in Southeast Asia

Mammals are a taxonomic group which foster a high level of popular and scientific attention and, as a

result, are often the focus of conservation and research efforts. In Southeast Asia, this has resulted in a

reasonable corpus of anecdotal information about mammalian species and community responses to

land-use change, but surprisingly has not yet translated into a corpus of robust quantitative

knowledge. This is a consequence of the difficulties associated with formally sampling a group which

is composed of highly mobile, rare, cryptic, and often nocturnal species inhabiting a dense tropical

forest environment. Nonetheless, a meta-analysis of key studies conducted in Southeast Asia between

1975 and 2007 identified mammals as the group most sensitive to land-use change, based on the

studies done so far, and the median effect size was more than twice that for birds, a much more well-

known group (Sodhi et al., 2009). Mammals also likely play important roles in ecosystems in the

region, as herbivores (Ickes et al., 2001; Harrison et al., 2013), ecosystem “engineers” (Campos-

Arceiz, 2009), seed dispersers (Corlett, 1998; Brodie et al., 2009; Nakashima et al., 2010; Campos-

Arceiz et al., 2012), granivores (Blate et al., 1998; Curran & Webb, 2000; Mcconkey, 2005; Wells &

Bagchi, 2005; Kitamura et al., 2008; Wells et al., 2009; Hautier et al., 2010; Bagchi et al., 2011) and

predators (Ross et al., 2013), roles which may not easily be replaced by smaller-bodied taxonomic

groups which operate at smaller spatial scales (Corlett, 2009; Campos-Arceiz et al., 2012). Given this,

we urgently need to quantitatively assess, and indeed begin to predict into the future, mammalian

biodiversity losses due to the rapid and large-scale changes in land-use ongoing in the region.


34

I collated information on all of the mammal-focussed, multi-species studies done in anthropogenic

land-uses (i.e. excluding those which only sampled in old-growth forest) and found 31 such studies,

conducted between 1969 and 2012 (Table 1). In reviewing these studies, the current body of

knowledge on mammalian responses to land-use change was found to be deficient in a number of

ways. Most obviously, studies have often only been able to obtain a sparse dataset with few species

records. This is principally to do with the difficulty of detecting most mammal species in tropical

forests, but is also a function of survey efforts; although the median study duration has been 12

months, with a median survey effort of > 7,000 trap nights or 270 km of transects (from n = 27, 23

and 9 studies, respectively), this has frequently been insufficient to enable robust estimation of species

richness, occupancy or abundance (81% of studies), with most studies reporting fewer than 10 records

for all but a few common species. In combination with this, many studies have expended their survey

efforts over a limited subset of the study area over which they aim to make inferences about and,

worse, have typically placed their sampling locations in a non-random and highly biased manner (e.g.

along roads and trails, or near field centres at the edge of the study area). Very few studies have

combined the basic tenets of proper study design: stratification of a study area, random placement of

sampling locations and sufficient replication therein. I scored the 31 studies for the extent of the study

areas that were sampled (biased sampling was also scored lower), as well as the temporal or spatial

replication that was achieved, and found that 39% of the studies were deficient in at least one of these

respects (“Coverage” or “Replication” < 2, Table 2).

In addition to these deficiencies in sampling design and effort, three of the four confounding factors

that are present in the land-use impacts literature in general (see above) are also especially acute in

mammal-focussed studies. Specifically, problems caused by transient and dispersing individuals are

perhaps relatively worse, given the large home range sizes and strong dispersal abilities of some

mammal species. The requisite details on study design in order to assess this (not to mention ranging

data on mammal species in the region) are often lacking in studies, but of the 31 multi-species studies

conducted so far in Southeast Asia, this problem likely affects the majority, and is very rarely


35

investigated (except in one bat-focussed study: Struebig et al., 2009). Secondly, mammals are often

long-lived, meaning responses to land-use change at a population level may often be delayed

compared to other groups, giving rise to substantial extinction debts. Thirdly, the low detection

probabilities inherent in sampling most mammal species mean that count- and occurrence-based

indices are likely to exhibit substantial negative bias compared to true abundance and occupancy.

Although relative indices may be instructive when investigating qualitative changes (if monotonically

related to the true parameter of interest), they are frequently used as if they are absolute quantities

(e.g. when converted directly to density, as in some of the live-trapping and line-transect studies in

Table 1) and this will lead to especially poor inference in the case of mammal species. There has, at

least, been a general trend of improvement in the statistical treatment of data through time, especially

since the emergence of occupancy methods (MacKenzie et al., 2006).

Of the 31 studies conducted to date, I found six to be of sufficient quality, in terms of study design

and statistical analysis (Inference Strength Score ≥ 0.9, Table 2), to allow strong inferences about

mammalian responses to land-use change. This included three long-term small mammal studies

(Lynam & Billick, 1999; Wells et al., 2007; Struebig et al., 2008) and three long-term camera-

trapping studies (McShea et al., 2009; Samejima et al., 2012; Brodie & Giordano, 2013). Of these,

three investigated community or species responses due to logging, two studied the distribution of

species among habitat fragments and one tested for the importance of logged forest set-aside habitat in

a large-scale A. mangium plantation.

At this stage, generalisations are difficult to make at a species level; contradictory responses to land-

use change have been reported for many species, likely, at least in part, due to insufficient sampling

efforts and the problem of imperfect detection (see above). At a community level, however, it does

appear to be a consistent finding that mammal species richness and overall abundance, although not

necessarily diversity or evenness, is maintained at the landscape scale when forests are disturbed by

logging (disturbance by fire or shifting agriculture is more poorly known). This is probably explained


36

by a high degree of dietary flexibility amongst many mammal species (e.g. Langham, 1983; Johns,

1986a; Meijaard & Sheil, 2008; Meijaard et al., 2010) and the fact that very few species directly

utilise harvested timber species to any substantial degree (Johns, 1988; Heydon & Bulloh, 1997).

When natural forest is converted to plantation, some species appear able to persist, but this is likely to

be dependent on the maintenance of sufficient ground cover (Maddox et al., 2007; Sunarto et al.,

2012), as well as patches of natural forest at the landscape scale (Nasi et al., 2007; McShea et al.,

2009). Intensively-managed plantations such as oil palm and rubber appear to support a more

depauperate community of mammal species than other tree-based plantation types, such as A.

mangium and P. falcataria, though the confounding influence of remnant natural habitat at the

landscape scale has not been investigated or controlled for. I identified six studies which have

investigated mammal abundance and diversity in oil palm: one made observations of primates and

squirrels and captured bats in mist-nets (Danielsen & Heegaard, 1995); four studies sampled only for

small mammals (Wood, 1984; Rajaratnam et al., 2007; Bernard et al., 2009; Puan et al., 2011), and

one included both large and small mammals (Scott & Gemita, 2004; Maddox et al., 2007). Across

these studies, only a small number of native mammal species were recorded inside oil palm habitat,

although invasive murid rodent species can reach very high densities, if not diversity, in some areas of

extensive oil palm monoculture in Peninsular Malaysia (Wood, 1984; Puan et al., 2011).

Despite the large amount of research attention, in general, given to mammals, surprisingly no study

conducted in Southeast Asia has yet assessed the responses of an entire mammal community when

forests are logged or converted to plantations. Those studies which have assessed responses by a

subset of mammal species have almost entirely been compromised either by inadequate sample sizes

or poor design and analysis. Community-level responses, besides species richness, have largely been

ignored and we have scant knowledge of changes in β-diversity and community structure, as well as

the emergent effects on ecosystem function.


37

Table 1. Multi-species mammal studies conducted across land-use types in Southeast Asia between 1969 and 2012. Studies which only sampled in old-growth forest were excluded.

Study number References Survey

dates

Study duration (months)

Methods

Sample size

Taxonomic focus Study locations

Land-uses surveyed

Trap nights

Transect length (km)

Old-growth Logged Fire-

disturbed Secondary Fragment

(< 100 km2)

Edge Non-natural land-uses

1 Wood, 1984; Wood & Liau, 1984a, 1984b

1969-1979 132 Live-trap 230,400 Terrestrial small mammals

Johor state, Peninsular Malaysia Mature oil palm

2 Wilson & Wilson, 1975; Wilson & Johns, 1982

1973 onwards

2+ Line-transect, sign survey

153 Large mammals East Kalimantan, Indonesian Borneo

Albizia

3 Johns, 1983, 1985, 1986a, 1992; Johns & Johns, 1995

1979-1993 ? Line-transect, censusing

? Primates Sungai Tekam Forest Reserve, Peninsular Malaysia and Ulu Segama Forest Reserve, Sabah, Malaysian Borneo

4 Kemper & Bell, 1985

1981 5 Live-trap 5,757 Terrestrial small mammals

Pasoh Forest Reserve, Peninsular Malaysia

5 Duff et al., 1984 1983 4 Sign survey 274 All non-volant mammals

Sabah Softwoods Brumas Estate, Sabah, Malaysian Borneo

Albizia, Eucalyptus, shade Cacao

6 Stuebing & Gasis, 1989


Sabah Softwoods Brumas Estate, Sabah, Malaysian Borneo

Albizia, Eucalyptus, shade Cacao

7 Bennett & Dahaban, 1995; Dahaban, 1996

1989-1992 24 Line-transect

548 All non-volant, diurnal mammals

Various sites (n = 11) in Sarawak and Brunei Darussalam, Borneo

8 Charles & Ang, 2010

1989 onwards

? Live-trap, camera trap

17,955 Terrestrial small mammals and carnivores

Brunei-Muara and Tutong districts, Brunei Darussalam, Borneo

9 Danielsen & Heegaard, 1995

1991 2 Line-transect, mist-netting

3,170 160 Primates, squirrels, treeshrews and bats

Riau and Jambi provinces, Sumatra, Indonesia

Rubber, mature oil palm

10 Laidlaw, 2000 1991-1992 13 Sign survey 360 All non-volant, diurnal mammals

Pahang and Selangor states, Peninsular Malaysia

Acacia

11 Heydon & Bulloh, 1996, 1997

1992-1993 19 Line-transect

122 Civets and mouse-deer Tragulus spp.

Ulu Segama Forest Reserve, Sabah, Malaysian Borneo

12 Lynam, 1997; Lynam & Billick, 1999


Chiew Larn reservoir islands and surrounding area, southern Thailand

13 Rajaratnam et al., 2007

1993-1994 12 Live-trap 38,445 Leopard cat Prionailurus bengalensis and small mammal prey

Tabin Wildlife Reserve and adjacent oil palm, Sabah, Malaysian Borneo

Mature oil palm


38

14 Pattanavibool & Dearden, 2002

1997-1998 10 Sign survey 56 Large mammals Om Koi and Mae Tuen Wildlife Sanctuaries, northern Thailand

15 Nasi et al., 2007 2000 3 Line-transect

? Primates Riau province, Sumatra, Indonesia Acacia, rubber

16 Mohd-Azlan, 2003, 2006; Mohd-Azlan & Sharma, 2006

2000-2001 21 Camera trap 5,972 Large mammals Jerangau Forest Reserve, Peninsular Malaysia

?

17 Bernard, 2004; Bernard et al., 2009


Tabin Wildlife Reserve, Sabah, Malaysian Borneo

Mixed-age oil palm

18 Scott & Gemita, 2004; Maddox et al., 2007

2001-2006 9+ Live-trap, camera trap, sign survey, car-based survey

9,925 422 All non-volant mammals

Jambi Province, Sumatra, Indonesia Mixed-age oil palm

19 Numata et al., 2005 2002 3 Camera trap 677 Terrestrial mammals

Pasoh Forest Reserve, Peninsular Malaysia

20 Wells et al., 2007 2002-2004 18 Live-trap 40,552 Terrestrial small mammals

Various sites (n = 6) in Sabah, Malaysian Borneo

21 Struebig et al., 2008 2002-2007 ? Live-trap 2,321 Insectivorous bats Krau landscape, Pahang state, Peninsular Malaysia

22 Nakagawa et al., 2006


Lambir Hills National Park and surrounding areas, Sarawak, Malaysian Borneo

Rubber

23 Kitamura & Thong-Aree, 2010

2004-2007 36 Camera trap 11,106 Terrestrial mammals

Hala-Bala Wildlife Sanctuary, southern Thailand

24 Belden et al., 2007a, 2007b; McShea et al., 2009


Planted Forest Zone, Sarawak, Malaysian Borneo

? Acacia

25 Rustam et al., 2012 2005-2010 ? Camera trap 1,017 Terrestrial mammals

Sungai Wain Protection Forest, Bukit Soeharto Grand Forest Park and surrounding areas, East Kalimantan, Indonesian Borneo

26 Onoguchi & Matsubayashi, 2008

2006 4 Camera trap 797 Terrestrial mammals

Deramakot Forest Reserve, Sabah, Malaysian Borneo

27 Puan et al., 2011 2007-2008 14 Live-trap 25,200 Terrestrial small mammals

Negeri Sembilan state, Peninsular Malaysia

Young Oil Palm

28 Imai et al., 2009; Samejima et al., 2012


Deramakot and Tangkulap-Pinangah forest reserves, Sabah, Malaysian Borneo

29 Mohamed et al., 2009, 2013; Wilting et al., 2010, 2012

2008-2010 17 Camera trap, car-based surveys

7,052 615 Carnivores Three forest reserves in Sabah, Malaysian Borneo: Deramakot, Tangkulap-Pinangah and Segaliud-Lokan

30 Bernard et al., 2012 2009-2010 16 Camera trap 3,733 Felids Tabin Wildlife Reserve and adjacent oil palm, Sabah, Malaysian Borneo

31 Brodie & Giordano, 2013

2010-2012 12+ Camera trap 16,608 Clouded leopard Neofelis diardi and four prey species

Various sites (n = 7) in Sabah and Sarawak, Malaysian Borneo


39

Table 2. Assessment of multi-species land-use studies of mammals conducted in Southeast Asia based on the study design and statistical techniques employed. Study Number References Coveragea Replicationb Robust statisticsc Inference strength scored

1 Wood (1984) 1 3 3 0.8

2 Wilson & Wilson (1975), Wilson & Johns (1982) 1 1 1 0.3

3 Johns (1983, 1985, 1986, 1992), Johns & Johns (1995) 1 2 1 0.4

4 Kemper & Bell (1985) 2 2 1 0.6

5 Duff et al. (1984) 2 1 1 0.4

6 Stuebing & Gasis (1989) 1 2 1 0.4

7 Dahaban et al. (1996), Bennett & Dahaban (1995) 1 2 1 0.4

8 Charles & Ang (2010) 2 3 2 0.8

9 Danielsen & Heegaard (1995) 2 1 1 0.4

10 Laidlaw (2000) 2 2 1 0.6

11 Heydon & Bulloh (1996, 1997) 2 2 2 0.7

12 Lynam & Billick (1999), Lynam (1997) 3 3 2 0.9

13 Rajaratnam et al. (2007) 1 2 2 0.6

14 Pattanavibool & Dearden (2002) 1 2 2 0.6

15 Nasi et al. (2007) 3 2 2 0.8

16 Mohd-Azlan (2003, 2006), Mohd-Azlan & Sharma(2006) 2 2 2 0.7

17 Bernard (2004), Bernard et al. (2009) 1 2 2 0.6

18 Scott & Gemita (2004), Scott et al. (2004), Maddox et al. (2007) 3 2 2 0.8

19 Numata et al. (2005) 1 1 1 0.3

20 Wells et al. (2007) 2 3 3 0.9

21 Struebig et al. (2008) 3 2 3 0.9

22 Nakagawa et al. (2006) 3 2 2 0.8

23 Kitamura et al. (2010) 2 2 2 0.7

24 Belden et al. (2007a, 2007b), McShea et al. (2009) 2 3 3 0.9

25 Rustam et al. (2012) 2 2 2 0.7

26 Onoguchi & Matsubayashi (2008) 1 1 1 0.3

27 Puan et al. (2011) 2 3 2 0.8

28 Imai et al. (2009), Samejima et al. (2012) 3 3 2 0.9

29 Mohamed et al. (2009, 2013), Wilting et al. (2010, 2012) 2 2 3 0.8

30 Bernard et al. (2012) 2 2 1 0.6

31 Brodie & Giordano (2013) 2 3 3 0.9 a1 = limited proportion of study site covered or highly-biased sample e.g. on trails or roads; 2 = reasonable extent, with limited bias; 3 = majority of study site with fully random sampling b1 = very limited replication in space or time; 2 = reasonable replication, though possibly pseudoreplicated; 3 = well replicated, with limited pseudoreplication c1 = limited number of species occurrences and/or no accounting for sampling effort or variance; 2 = relative abundance index or similar, based on reasonable sample size; 3 = robust estimation of species richness, abundance or occupancy based on relatively large sample size dCalculated as a proportion of the maximum potential score (i.e. [Coverage + Replication + Robust Statistics] / 9). Therefore all studies achieve a minimum score of 0.3 and a maximum of 1. This score is subjective and is not formally related to the probability of a study making poor inference, e.g. the probability of a Type I or Type II error.


40

1. 8. Research objectives

Four inter-linked objectives for the current work arise from reviewing current knowledge of

mammalian responses to land-use change in the region.

1. Robustly quantify the response of mammalian species richness, including both large and

small mammals, to the principal gradient of land-use intensity in Southeast Asia (old-growth

forest, logged forest and oil palm plantations).

2. Investigate the spatial patterns of species occurrence and, in turn, the effects of spatial grain

on estimates of mammalian species richness and β-diversity across land-uses

3. Characterise changes in community composition along the gradient of land-use intensity and

identify the local-scale drivers assembling communities.

4. Assess abundance responses, using robust estimation methods, of individual mammal

species, across both large and small mammals, to land-use change and the potential

implications for community structure and ecosystem functioning.

1. 9. Sampling methods, design and study site

To circumvent many of the problems that have proved challenging in previous studies of mammalian

responses to land-use, I employed a combination of field methods (covering both small mammals and

large mammals) with high sampling effort, in a nested plot design, with high levels of study site

coverage (see individual chapters for detailed information on the sampling design and field methods

used). Within each plot, sampling occurred at either systematic random locations (for small mammals)

or a random subset of locations (for large mammals). I sampled along the principal land-use trajectory

in the region: old-growth forest selectively logged forest oil palm plantation. Plots within the

forested land-uses were embedded within large continuous tracts of habitat, thereby minimising the

occurrence of transient and dispersing individuals from other land-uses, as well as the potential for

“rescue effects”. Oil palm plots, as is common for plantations in the region, were in the vicinity of

remnant logged forest areas. The potential for spill-over from these forested areas was assessed using

landscape-level metrics (distance from the nearest forest and percentage forest cover). For species


41

richness and abundance estimation, robust methods were used, which controlled for imperfect

detection. Considering these details together, this study fills important gaps in current knowledge of

mammalian species and community responses to land-use change in the region. Moreover, this is the

only study thus far which has characterised almost the entire non-volant mammalian community

across these land-uses, including both large and small mammals.

Fieldwork was carried out between October 2010 and July 2014 in south-eastern Sabah, Malaysian

Borneo, at the Maliau Basin Conservation Area, Stability of Altered Forest Ecosystems (SAFE)

Project experimental area, Brantian-Tatulit Virgin Jungle Reserve and Benta Wawasan and Sabah

Softwoods oil palm plantations. More detailed information on these study sites can be found in the

individual chapters. The SAFE Project area is also the focus of a long-term forest fragmentation

experiment (Ewers et al., 2011), in which ~63% of the landscape (9,400 ha, including the Virgin

Jungle Reserve, which abuts the experimental area) began to be gradually cleared in April 2013 for

the development of an oil palm plantation (as of April 2015, this process is yet to be completed). The

SAFE project has been given the opportunity to dictate the size and location of remnant forest

fragments totalling ~10% of the experimental area, and has opted to create fragments of 1, 10 and 100

ha size. The 2,200 ha Virgin Jungle Reserve will also be isolated by the clearance. My own sampling

was conducted with a view to this future change in land-use from logged forest to fragmented forest,

with plots placed to coincide with the planned forest fragments. The current study can therefore also

be viewed as a pre-fragmentation baseline for the longer-term investigation of fragmentation effects

on mammals.


42

Figure 1. Land-use types sampled in this thesis: (A) old-growth forest with an intact canopy, (B) logged forest with a discontinuous canopy and few mature trees remaining, and (C) oil palm plantations with remnant forest patches in the broader landscape.

A

B

C


43

1. 10. Thesis chapter outline

Chapter 2 reports on the richness and relative abundance of Felidae from the SAFE project

experimental area. This is of particular conservation importance since this study site is amongst the

most heavily-disturbed logged forest so far sampled within the range of these species. The finding that

all five species, including the rare Bornean bay cat Pardofelis badia, have so far persisted suggests

that this land-use type, much undervalued by land-use managers and policy-makers alike, needs to be

given greater priority for conservation.

I used fully random placement of camera traps in collecting data on these species of wild cat and, to

my knowledge, this has never been done before in Southeast Asia. I also show in Chapter 2 that

typical placement, for example along roads and trails, produces a biased account of the relative

abundance of these species, due to differences in the propensity of species to use different habitat

features. This is the case both within my own dataset, when locations on- and off-trail are compared,

and also across studies, when the results from my random sampling are compared to the

(bootstrapped) results from previous, non-random camera-trapping studies conducted in the region.

This first investigation of the utility of random camera trap sampling in the region demonstrates that

the method yields data on even the rarest of species (typically the focus and justification of the non-

random methods that have prevailed), provides insights into finer-scale habitat use than has been

possible before and also provides an alternative explanation for the extreme rarity of some species in

camera trap studies, such as the bay cat.

Chapter 3 explores the often-neglected spatial component of diversity and its relevance in the

assessment of the biodiversity value of land-uses, as well as in the design of conservation set-aside.

The spatial grain at which studies of land-use change are conducted is often not considered, and I

show that grain-dependent responses to land-use may be one reason for the conflicting results

reported in the literature. Using the nested sampling design that was employed (in order of increasing

grain: sampling points plots blocks land-use), I quantify the species richness of large and


44

small mammals at multiple spatial grains within each land-use, using measures which are robust to

variation in sampling effort. The grain of sampling affects the type of response inferred, with large

mammal richness at the sampling point level significantly lower in logged forest relative to old-

growth forest, but unchanged at the overall land-use level, due to higher turnover of species across

space. Small mammals, however, exhibit higher richness at all spatial grains in logged forest relative

to old-growth forest, highlighting the fact that patterns even within a single taxonomic group, in this

case terrestrial mammals, can be non-concordant.

The rate of increase in species richness with sampling grain can be explained by the spatial

distribution of individuals and species, and grain-dependent richness responses are symptomatic of an

alteration in these distributions across land-uses. Chapter 3 also characterises these changes using the

framework of β-diversity, i.e. the variance in community composition between two points in space. β-

diversity metrics are often not comparable across communities of different sizes and different

sampling designs, so I calculate β-diversity as the deviance from a null model of the occurrence of

species. This allows for meaningful comparisons between large and small mammals and controls for

both sampling effort and the spatial pattern of sampling locations. By calculating β-diversity in each

land-use at each spatial grain in the sampling design (points, plots and blocks), contrasting patterns

are again evident between large and small mammals. In particular, large mammal communities in old-

growth forest become more heterogeneous at coarser spatial grains and small mammal communities

become more homogeneous, whilst this pattern is reversed in logged forest. This finding suggests that

small forest reserves, such as High Conservation Value areas, set-aside for protection from either

logging or outright clearance within concessions, will capture different levels of large and small

mammal diversity depending on their size and distribution in a concession. However, both species

groups exhibited a strong β-diversity signal at the fine spatial grain of individual sampling points,

most likely an effect of logging-induced habitat heterogeneity.


45

Figure 2. Methods used in this thesis to sample terrestrial mammal communities: (A) camera-trapping at random locations, including poorly accessible areas such as steep slopes, and (B) live-trapping for smaller mammals, such as the large treeshrew Tupaia tana.

In order to build better predictive models of biodiversity under land-use change, it will be necessary to

have a deeper understanding of the land-use change process than the patterns of richness and β-

diversity can reveal alone. In a step towards this, in Chapter 4 I identify the drivers of the strongly

non-random patterns of species occurrence seen in the previous chapter, modelling community

composition as a function of environmental control and spatial processes, including dispersal

assembly. This analysis uncovers the novel set of mechanics assembling mammal communities in

landscapes which are logged and converted to oil palm plantations, which will have implications for

any attempt to restore these ecosystems towards a more intact state. In particular, there is a striking

shift from the importance of spatial processes towards the dominance of environmental control along

B

A


46

the gradient of land-use intensity. An analysis of species co-occurrence in parallel to this, also

indicates a potential weakening along the land-use gradient of inter-specific competition as an

influential process in community assembly. Analyses of this nature are rarely applied to mammalian

communities, but may prove to be a fertile ground of new hypotheses to test, ideally with

manipulative experiments, in the future.

In Chapter 5, I explore species-level abundance responses to the land-use gradient, complementing

previous chapters on the richness and composition of communities. I combine the live- and camera-

trapping data across all species into one framework, using a hierarchical modelling approach, and use

both continuous and categorical descriptors of the land-use gradient. Using a multi-species

hierarchical approach also allows for the estimation of abundance for even the rarest species in the

community by “borrowing strength” from the more well-sampled species. The analysis provides the

first robust estimates of abundance for any taxonomic group along the old-growth forest logged

forest oil palm land-use gradient. I find that the abundance of most species is conserved from old-

growth to logged forest (indeed, omnivorous and herbivorous species show large increases on

average), whilst dramatic declines are seen from forest to oil palm, in particular in species of high

conservation concern. The main exception to this overall pattern is for invasive species, which

increase along the land-use gradient. By constructing nine dietary functional effects groups, I also

show that the biomass of species carrying out key ecosystem functions such as leaf-eating and

invertebrate predation is also conserved in logged forests, but only vertebrate predation is maintained

in oil palm plantations.

Finally, Chapter 6 serves as a synthesis, by seeking the connections between the findings within

individual chapters and identifying the broader conclusions and implications of the work.

Chapter 2: Using random camera locations to survey felids in a logging concession

47

Chapter 2:

Assessing the status of wild felids in a highly-disturbed commercial forest reserve in

Borneo and the implications for camera trap survey design

This chapter was published at the following location:

Wearn, O.R., Rowcliffe, J.M., Carbone, C., Bernard, H. & Ewers, R.M. (2013) Assessing the

status of wild felids in a highly-disturbed commercial forest reserve in Borneo and the

implications for camera trap survey design. PLoS One, 8, e77598.

Abstract

The proliferation of camera-trapping studies has led to a spate of extensions in the known

distributions of many wild cat species, not least in Borneo. However, we still do not have a clear

picture of the spatial patterns of felid abundance in Southeast Asia, particularly with respect to the

large areas of highly-disturbed habitat. An important obstacle to increasing the usefulness of camera

trap data is the widespread practice of setting cameras at non-random locations. Non-random

deployment interacts with non-random space-use by animals, causing biases in our inferences about

relative abundance from detection frequencies alone. This may be a particular problem if surveys do

not adequately sample the full range of habitat features present in a study region. Using camera-

trapping records and incidental sightings from the Kalabakan Forest Reserve, Sabah, Malaysian

Borneo, we aimed to assess the relative abundance of felid species in highly-disturbed forest, as well

as investigate felid space-use and the potential for biases resulting from non-random sampling.

Although the area has been intensively logged over three decades, it was found to still retain the full

complement of Bornean felids, including the bay cat Pardofelis badia, a poorly known Bornean

endemic. Camera-trapping using strictly random locations detected four of the five Bornean felid

species and revealed inter- and intra-specific differences in space-use. We compare our results with an

extensive dataset of > 1,200 felid records from previous camera-trapping studies and show that the

relative abundance of the bay cat, in particular, may have previously been underestimated due to the


48

use of non-random survey locations. Further surveys for this species using random locations will be

crucial in determining its conservation status. We advocate the more wide-spread use of random

survey locations in future camera-trapping surveys in order to increase the robustness and generality

of inferences that can be made.

2. 1. Introduction

With rates of forest loss and degradation in Southeast Asia exceeding all other tropical regions

(Achard et al., 2002), and the majority of remaining forest existing in a highly disturbed state (Curran

et al., 2004; Laurance, 2007), there is now an urgent need for accurate assessments of the impacts on

wildlife in the region. This situation applies especially to Borneo and to the five species of felid

inhabiting the island: Sunda clouded leopard Neofelis diardi (Vulnerable on the IUCN Red List),

leopard cat Prionailurus bengalensis (Least Concern), flat-headed cat Prionailurus planiceps

(Endangered), marbled cat Pardofelis marmorata (Vulnerable) and bay cat Pardofelis badia

(Endangered). For all of these species, we still have a paucity of information on their distributions,

population statuses and responses to land-use changes. This is particularly the case for the bay cat, a

Bornean endemic which has been called “the world’s least known felid” (Sunquist & Sunquist, 2002).

Certainly, very few confirmed records of it exist (Mohd-Azlan & Sanderson, 2007) and it has

variously been suggested to be either tolerant (Kitchener et al., 2004; Hunter, 2011) or intolerant of

habitat disturbance (Wilting & Mohamed, 2010).

A number of targeted field studies of Borneo’s terrestrial fauna have recently been undertaken

(McShea et al., 2009; Mathai et al., 2010; Samejima et al., 2012), with some focussing on wild felids

(Mohd-Azlan & Sanderson, 2007; Mohamed et al., 2009; Cheyne & Macdonald, 2011; Brodie &

Giordano, 2012b). Importantly, there has been a rapid increase over the last decade in the use of

camera traps for conducting such studies (Rowcliffe & Carbone, 2008), allowing intensive surveys to

be made over larger areas with reduced effort in the field. This has led to significant extensions in the

known distributions and habitat tolerances of many species (Wilting et al., 2010b; Brodie & Giordano,


49

2011; Matsubayashi et al., 2011; Lhota et al., 2012), including Borneo’s wild cat species (Mohamed

et al., 2009; Wilting et al., 2010a; Bernard et al., 2012). However, it remains the case that few camera

trap surveys have been done beyond the boundaries of protected areas, in forests which are not

pristine and not sustainably managed (but see Scott & Gemita, 2004; McShea et al., 2009; Rustam et

al., 2012). Less than 6% of land area in Indonesia and Malaysia is protected (IUCN categories I-IV;

IUCN/UNEP, 2012) and most landscapes are now dominated by highly-disturbed forests which have

undergone multiple rounds of logging (Curran et al., 2004; Miettinen et al., 2011; Reynolds et al.,

2011). It is only these highly-disturbed forests that still occur over sufficiently large and contiguous

areas to potentially conserve viable populations of felid species occurring at very low densities, such

as the clouded leopard (Wilting et al., 2006, 2012; Brodie & Giordano, 2012b).

The proliferation of camera-trap studies has allowed more robust inference on the relative abundance

of highly cryptic species than has been possible before. This has led to a re-assessment of the

supposed rarity of some taxa, including the Asiatic golden cat Pardofelis temminckii (Mohd-Azlan &

Sharma, 2006; Johnson et al., 2009; Bashir et al., 2011). The bay cat, on the other hand, has remained

consistently rare in camera-trap surveys throughout its range, usually appearing at least one order of

magnitude less frequently than other Bornean felids (Hunter, 2011). Since detection frequencies are a

function of both abundance and detection probability (Williams et al., 2002), the rarity of bay cat

records could reflect low detection probability rather than low population densities. Low detection

probability in camera trap surveys can result from a range of factors, broadly categorised as factors

that reduce camera sensitivity, and factors that reduce the chances of animals encountering cameras.

An important species-specific correlate of camera-sensitivity is body size (Rowcliffe et al., 2011).

However, the bay cat is comparable in size to the other three small cats of Borneo: the leopard cat,

flat-headed cat and marbled cat (Sunquist & Sunquist, 2002). In terrestrial surveys, reduced detection

probabilities are obviously expected for arboreal species, but it also does not seem likely that the bay

cat is more arboreal than the other cat species: all direct sightings have been made on or very close to


50

the ground (Mohd-Azlan & Sanderson, 2007) and its morphology is consistent with terrestriality

(Kitchener et al., 2004).

Low detection probabilities can also result from avoidance of the particular habitat features on which

camera-trapping surveys typically focus. Ever since their early use in mark-recapture studies (Karanth

& Nichols, 1998), camera traps have been deployed preferentially where the presence of a focal

species is deemed most likely – usually on trails, roads, water points or mineral licks – in order to

increase individual capture probability. It is now common and accepted practice to use these non-

random deployment locations in general wildlife surveys and then calculate an index of relative

abundance (Williams et al., 2002). In some cases, researchers have stated that cameras were deployed

“randomly” but actually refer to a two-step process in which potential deployment zones (typically

squares of a grid overlain on the study area) are selected at random and then cameras are deployed

non-randomly within these zones. Given that deployment zones are typically much larger than the

area actually sampled by a camera trap – 2 km2 grid squares are often used (e.g. Ahumada et al.,

2011) compared to sensors with maximum detection zones mostly less than 2 x 10-4 km2 (Meek et al.,

2012) – species may be detected less frequently, or not at all, if they avoid certain habitat features

within the focal area.

Choosing ‘optimal’ locations for deploying cameras in this way violates a key assumption of

sampling theory – the random selection of sample units – and necessarily limits the scope of inference

of a study to the specific conditions found at the survey locations. Inferences made beyond this

limited subset of features of a habitat or landscape are likely to be biased, even though this is routinely

done. To our knowledge, no camera-trapping study conducted in Borneo, or indeed more broadly in

the Palaeotropics, has used strictly random locations (within 5-10 m of a pre-marked point, e.g. Kays

et al., 2011). This may have implications for the currently inferred abundance and understanding of

habitat use for all species, including felids. Owing to the prevailing use of non-random camera trap

surveys, up to now it has not been possible to explicitly test for these possibilities.


51

We aimed to assess the status of wild felids in a highly-disturbed commercial forest reserve, gathering

together both incidental sightings and camera-trapping records from strictly random locations. Given

that we used random camera locations, we also investigated the potential for non-random survey

designs to interact with non-random space-use by animals, which would cause biased inferences about

relative abundance. To do this, we investigated felid space-use patterns with respect to anthropogenic

habitat features, which have typically been the focus of camera trap surveys, and also compared our

relative abundance estimates to those from previous camera-trapping studies conducted in the region.

As a result, we suggest that prevailing camera trap methods have indeed confounded assessments of

felid species rarity. Our findings have implications for the conservation of our focal species, as well as

the study design of camera-trap surveys in general.

2. 2. Methods

2. 2. 1. Study area

This study was carried out in Kalabakan Forest Reserve (4º 33’ N, 117º 16’ E) in the state of Sabah,

Malaysia, and forms part of the Stability of Altered Forest Ecosystems (SAFE) Project (Ewers et al.,

2011). Kalabakan Forest Reserve lies within the Yayasan Sabah Forest Management Area and, as

such, has been subject to multiple, intense rounds of logging, beginning in 1978 and ongoing until the

early 2000s. This has led to a heterogeneous landscape composed of stands which have undergone

varying intensities and timings of log extraction, using both tractor-based and high-lead yarding

methods. During the logging, a network of regenerating skid trails, logging roads and log-landing

areas was also created (approximately 10% of land area; Pinard et al., 2000). As a result, there is a

range of habitat types currently exhibited in the reserve, from grassy open areas and low scrub

vegetation, to lightly logged forest on steep slopes and in rocky areas, but the timber volume

remaining in the area is mostly very low (below 10 m3 ha-1). In addition to the logged forest areas,

large portions of the reserve have been terraced and planted with oil palm, or have been salvage

logged in preparation (removing all trees above 25 cm diameter at breast height). Medium-resolution


52

(250 m) land cover maps for 2010 (Miettinen et al., 2011) therefore indicate that just 54% of the area

of the Kalabakan Forest Reserve (2,240 km2) still retains natural forest cover.

2. 2. 2. Data collection

We deployed remotely-operated digital cameras (Reconyx HC500, Holmen, Wisconsin, USA) in the

north-east of the Kalabakan Forest Reserve (4º 42’ N, 117º 34’ E), overlapping with the SAFE Project

experimental area (72 km2). We had full permission from the land-owners and concession holders,

Yayasan Sabah and Benta Wawasan Sdn Bhd, to conduct this study in the reserve. We also had an

access license in place from the Sabah Biodiversity Council for the use of camera traps at the study

site. Our study was approved by the Zoological Society of London Animal Ethics Committee and the

work detailed here did not involve any direct sampling methods or the collection of any specimens.

We sampled 135 locations between May and December 2011 for an average of 49 camera-trap nights

(CTNs), giving a total effort of 6650 CTNs. Camera traps were deployed inside 18 separate plots,

each covering 1.75 ha, which were clustered into three groups (Fig. 1). This design was chosen to

overlap with the sampling locations of the SAFE Project (Ewers et al., 2011). The SAFE Project has

attempted to control the confounding effects of elevation by stratifying the study site and only

sampling within a relatively narrow range centred at ~450 m; the elevation of our plots reflects this

stratification. For each plot, we established a 4 x 12 grid of points (23 m spacing) in the field using a

tape measure and GPS receiver (Garmin GPSMAP 60CSx, Olathe, Kansas, USA). There is a margin

of error associated with these methods, but we ensured that field teams marked the grids with no

consideration of the practicalities of whether or where a camera might be set at each grid point. Grid

points can therefore be considered to be truly random within plots. Cameras were deployed at a

random subset of points within each grid (mean = 8 points per grid), as close to the marked points as

possible. Necessarily, large obstructions in the camera’s detection zone (such as rock boulders or

large tree buttresses) were avoided, but cameras were always deployed strictly within 5 m of marked

points. We usually set cameras at a height of 30 cm, to maximise detection for a range of species, but


53

some cameras were set higher (and faced downwards), depending on the situation found at each

random location. No bait or lure was used and disturbance to vegetation was kept to a minimum.

Cameras were programmed to take 10 consecutive photographs for each trigger event (over

approximately 5 seconds). We noted if the detection zone contained a logging road (wide, heavily

compacted ground, sometimes with gravel remnants, no canopy cover), a skid trail (width of a tractor,

canopy cover, recruiting vegetation at ground-level, earth-banked sides), footpath or none of these

(which we term “off-trail”).

Figure 1. Locations sampled using camera traps within the Kalabakan Forest Reserve, Malaysian Borneo. Camera traps were deployed at random locations (black points) within 1.75 ha plots (white rectangles), clustered into three groups placed deliberately to control for elevational effects. Shaded areas lie outside the Kalabakan Forest Reserve and are composed of the Brantian-Tatulit Virgin Jungle Reserve (to the south) and Mount Louisa Forest Reserve (to the north). Inset shows the location of Kalabakan Forest Reserve (red outline) within the Malaysian state of Sabah, northern Borneo.

In parallel to the camera-trapping effort, we also recorded the location and time of all incidental

records of felids obtained during the course of the fieldwork detailed here and across all of the

research activities at the SAFE Project. These data are inherently biased towards less-cryptic, large-

bodied and diurnal species, and sampling was highly non-random in space and time. We excluded


54

periods for which consistent reporting from the SAFE Project was unavailable, leaving approximately

10 months of observations between August 2010 and August 2011.

We also conducted an extensive literature search for previous camera-trapping studies done on any of

the five Bornean felid species. We used the ISI Web of Science (www.isiknowledge.com, using

various searches on the vernacular and scientific species names, as well as “camera trap*” and the

names of Southeast Asian countries) to locate published and peer-reviewed studies, and supplemented

this with other published and unpublished sources we were aware of or which were cited in other

sources. For inclusion in our database, studies had to report the total number of CTNs conducted, as

well as the number of independent captures. Where data were not presented in a suitable form, we

attempted to contact authors directly for clarification.

2. 2. 3. Data analysis

Image sequences were judged to be independent capture events if they a) contained different

individuals or b) were separated by more than an arbitrary 1 hour. We present a detection frequency

(d) – often referred to in the literature as a relative abundance index – for each species, which is the

number of independent captures per 100 CTNs (accounting for camera failure).

We modelled the binary detection or non-detection of each species as a function of the habitat features

at camera locations, using a generalised linear model with binomial errors and a logit link function.

Factor-level simplification was done using chi-squared likelihood ratio tests. For the clouded leopard,

we also tested for an interaction between habitat features and the sex of individuals. We used Fisher’s

exact test with the null hypothesis that the number of detections of males and females is not

conditional on whether a camera is placed on a logging feature (road or skid trail) or not.

Using empirical data from our literature survey of camera-trap records, we constructed a probability

density function for the expected detection frequency of each of the five cat species. This was done by


55

taking bootstrap samples of the data (n = 10,000) and calculating the overall detection frequency each

time. We stratified samples by study site, to give each study site equal weight and to ensure that each

observation within randomisations was independent.

We then used the bootstrapped median detection frequencies to estimate the minimum survey efforts

required to detect each species with a given probability, incorporating the uncertainty in the

distributions of d using the 95% quantile values. Detections of a species D were modelled as a

Poisson process, with a rate parameter λ (detections per camera-trap night). For consistency with the

camera-trapping literature we used the detection frequency d, which has units per 100 CTNs, i.e. λ = d

/ 100, and for a survey conducted over n camera trap nights, the expected number of detections E(D)

= λn = nd / 100. Given Poisson distributed detections, we can use the cumulative exponential

distribution to calculate the probability p of obtaining at least one detection of a species after

surveying for a given number of camera trap nights. Using d instead of λ gives

)100

(1

dnep

−−=

(1)

Plotting this for a range of n gives a detectability curve (Wintle et al., 2005). We can determine the

number of camera trap nights required for a given cumulative probability or “confidence” either

graphically or by solving for n. For example, for 90% confidence (p = 0.9) in detecting a species, the

minimum sampling effort required is calculated by

d

n )9.01log(100 −−=

(2)

All analyses were done in R version 2.12.2 (R Development Core Team, 2014).

2. 3. Results

Camera-trapping yielded 504 photos of wild cats, consisting of 41 independent captures across 29

locations (21% of random locations sampled). It took 873 camera trap nights to detect four out of the

five Bornean felid species. The clouded leopard was detected most frequently and at the most

locations, followed by the leopard cat (Table 1). Both of these species were also detected by


56

incidental sightings. The two rarest felids from camera-trapping were the marbled cat and bay cat; the

marbled cat was detected more times than the bay cat but at fewer locations (Table 1). Neither of

these species were observed during incidental sightings. In contrast, the flat-headed cat was not

detected during camera-trapping but was directly sighted. This observation consisted of a single

individual crossing a narrow logging road (~ 5 m wide and bordered with ~ 2 m of dense scrub and

Coelorachis glandulosa grass) approximately 70 m from the nearest stream (~5 m width) and at 180

m elevation.

Table 1. Wild felid species recorded from the Kalabakan Forest Reserve, Sabah, Malaysia.

Common name, scientific name

Direct

sightingsa

Camera trapping

No.

photos

Independent

capturesb

Detection

frequency (d)c

Naive

occupancyd

Sunda clouded leopard Neofelis diardi 1 267 14 0.211 0.081

Marbled cat Pardofelis marmorata 0 89 9 0.135 0.052

Bay cat Pardofelis badia 0 64 8 0.120 0.059

Leopard cat Prionailurus bengalensis 13 84 10 0.150 0.067

Flat-headed cat Prionailurus planiceps 1 0 0 0 0

aIncidental records obtained during the course of fieldwork (August 2010 to August 2011). bDefined as image sequences of different individuals or sequences obtained more than 1 hour apart. cThe number of captures per 100 camera trap nights. dThe proportion of sampled locations at which the species was detected.

The minimum adequate models for site detection probabilities of each species revealed a significantly

higher probability of detection on logging features (z = 2.639, p = 0.008), i.e. logging roads and skid

trails, for clouded leopard and on skid trails only for marbled cat (z = 2.615, p = 0.009). The

probabilities of detection on the back-transformed scale were 0.195 (SE = 0.066) and 0.040 (SE =

0.020) for clouded leopard at locations on and off logging features, respectively, and 0.200 (SE =

0.103) and 0.025 (SE = 0.014) for marbled cat on and off skid trails, respectively. There was an

indication of sex-specific differences in habitat feature use in clouded leopard (p = 0.061): male

clouded leopards were only detected on logging features, whilst two-thirds of detections of females

were made off-trail. No habitat feature variables were retained in the minimum adequate models for

http://www.flickr.com/photos/haile/7922940438/


57

the leopard cat and bay cat, which means that the null hypothesis of random use of habitat features

was not rejected. Note, however, that 7 of 8 independent captures of bay cat were off-trail.

We were able to obtain useable data on previous detections of the five Bornean felid species from 34

separate camera trap studies across at least 27 study sites. This represents a combined effort of

approximately 62 years of fieldwork between 1998 and 2011, resulting in 1,212 felid detections over

142,672 CTNs. The amount of effort spent surveying for each species was unequal, being largely

dependent on the geographic range of the species: bay cat survey effort (60,914 CTNs) has been

approximately half that of marbled cat (120,231 CTNs). Without exception, these studies used non-

random survey locations.

We calculated the detection frequency (d) across all studies combined for each species. These showed

an order of magnitude difference between the relatively commonly detected leopard cat (d = 0.701)

and clouded leopard (d = 0.320) to the more rarely detected marbled cat (d = 0.079), flat-headed cat (d

= 0.021) and bay cat (d = 0.015). Once we accounted for the unbalanced and autocorrelated nature of

the dataset using a stratified bootstrap sampling approach, the expected detection frequencies were

lower in the case of the bay cat and leopard cat, but the rank order of detection frequencies amongst

the species was the same (Fig. 2). The detection frequencies observed in the current study lay within

the 2.5% and 97.5% quantiles of the bootstrapped distribution for the clouded leopard, marbled cat

and flat-headed cat, but were significantly higher and lower than expected for the bay cat and leopard

cat, respectively (Fig. 2). The detection frequency we obtained for bay cat from random survey

locations was more than 10 times larger than that expected from previous studies.


58

Figure 2. Probability density functions for bootstrapped values of detection frequency (d) derived from previous camera-trapping studies. Data for each of Borneo’s felid species were obtained from 34 studies conducted between 1998 and 2011 and bootstrap randomisations (n = 10,000) were stratified according to study site. Each panel shows the probability density function obtained by kernel density estimation, the median d from bootstrap samples (solid line) and d obtained in the current study, using strictly random survey locations (dashed line). Dotted lines for bay cat and leopard cat show d calculated after excluding off-trail survey locations. Note that the x-axis is not consistent across panels.

Owing to this significant difference for the bay cat and leopard cat, we decided post hoc to compare

our overall detection frequencies for these species with those we would have obtained at our study site

with a traditional trail-based survey, by excluding data obtained from off-trail cameras. We found the

same qualitative differences between random and non-random camera placement designs within our


59

study as we had found between our results and those found in the camera-trapping literature: detection

frequencies were 2.5 times larger and 0.8 times smaller for the bay cat and leopard cat, respectively,

for a survey design with off-trail cameras than one without (dotted lines, Fig. 2).

We calculated the minimum survey efforts required for each species based on the data from previous

studies (Fig. 2) and Eq. 2 (using p = 0.9). Huge disparities between species were revealed, ranging

from more than 26,000 CTNs required for bay cat to just 425 CTNs for leopard cat (Fig. 3). We also

calculated the worst-case scenario (using the 2.5% quantile for d) and this extended the survey effort

required substantially in all cases (Fig. 3): the requirement more than doubled for clouded leopard and

leopard cat and more than trebled for marbled cat. For the bay cat and flat-headed cat, the lower

bound did not exclude zero, so we could not rule out the possibility of never detecting these species

regardless of survey effort. For the bay cat, we compared the required effort suggested by previous

studies to that suggested by the current study using random locations: the required effort was reduced

by more than 24,000 CTNs for the detection frequency observed in our study (Fig. 3). This was also

associated with a comparatively small increase of 1,106 CTNs in the minimum effort required for the

leopard cat.


60

Figure 3. Detectability curves and minimum required survey efforts calculated using a Poisson model for detections. Detectability curves for each of Borneo’s felid species were plotted using Eq. 1 and minimum survey efforts calculated using Eq. 2 with a “confidence” of 90% and per-trial probability of success estimated using d (captures per 100 camera trap nights). Solid lines use median d from bootstrap samples of camera trap data obtained from previous studies (with shading corresponding to the 95% quantiles of d from bootstrap samples) and dashed lines use d obtained in the current study using random survey locations (except for flat-headed cat, which was not detected by camera-trapping in our study). For each detectability curve, survey efforts required for 90% confidence are indicated with dot-dash lines and annotated on the axes. Note that the x-axis is not consistent across panels.

2. 4. Discussion

Using camera traps and direct sightings, we confirmed the presence of all five Bornean wild felids in

the Kalabakan Forest Reserve. Moreover, the four species detected by camera-trapping were

estimated to have a relative abundance of the same, or higher, order of magnitude as previous studies


61

conducted elsewhere (Fig. 2). We also investigated possible biases in the relative abundances derived

from past camera-trapping efforts, caused by the non-random survey designs which have typically

been used. Importantly, we found both inter- and (for clouded leopard) intra-specific differences in

the use of habitat features. In addition, there were significant differences between the relative

abundances we obtained using random camera locations and those from previous studies, and we

found similar differences in comparing on- and off-trail locations within our own survey design. We

suggest these differences are evidence of biases, caused by an interaction between patterns of animal

space-use and the non-random deployment of camera traps at locations chosen by researchers.

Many book and journal pages have been devoted to exploring issues of survey design for monitoring

and assessment of populations (Thompson et al., 1998; Olsen et al., 1999; Yoccoz et al., 2001;

Williams et al., 2002) and we do not wish to recapitulate all of the design principles that have been

recommended. However, random selection of sample units is central to most sampling schemes

(Cochran, 1977). Given our findings, it is clear that this should also be central to the design of camera

trap surveys. We have shown that this allows small-scale habitat-use to be investigated, and provides

a stronger basis for inferences about relative abundance. There are, however, some important

instances where a non-random design might be preferred for species monitoring, such as when the

detectability of individuals can be explicitly modelled using mark-recapture methods. Although such

models require high capture, and indeed recapture, rates and employ stricter assumptions (Otis et al.,

1978), they allow inferences about absolute abundance for the limited subset of species which can be

individually identified from camera trap images. Occupancy methods, too, account for detectability

(of a species) and, although having similarly demanding data requirements (a large number of

independent sample locations may be required for anything other than common species), may also

provide a strong basis for inference about the status of a population, if not abundance per se

(Mackenzie & Nichols, 2004). Beyond monitoring, non-random designs might also be considered in

rapid, preliminary surveys which seek only to determine if a species is present in an area, rather than

its population status.


62

We found all five species of Bornean felid in the Kalabakan Forest Reserve. We are aware of only

three other sites which have confirmed records of all five species: Deramakot Forest Reserve

(Mohamed et al., 2009), Danum Valley Conservation Area (Hearn & Ross, 2009) and Tabin Wildlife

Reserve (Bernard et al., 2012). These sites range from pristine (Danum), through to sustainably

managed (Deramakot) and selectively logged until the late 1980s (Tabin). The addition to this list of

the Kalabakan Forest Reserve, a highly-disturbed commercial forest reserve which has undergone

decades of sustained logging until very recently, therefore extends this list to the full range of forest

disturbances present in Borneo.

Taken together, our results suggest that the large areas of highly-disturbed natural forest in the region

could play a greater role in the conservation of wild felids than is currently recognised. It does still

remain to be known if populations of these five species would be viable in disturbed forest in the

long-term and we therefore echo previous assertions of the importance of undisturbed forest (Didham,

2011; Gibson et al., 2011). However, we did obtain photographic evidence of breeding within our

highly-disturbed study site for clouded leopard (one female with cub) and calculated a relative

abundance that was similar to those from previous studies, mostly done in more intact sites (Fig. 2).

The habitat tolerances of the bay cat are poorly known, but our results using random survey locations

indicate that the relative abundance of this species may be of the same order of magnitude as the other

wild felid species in disturbed habitats.

We did not detect the flat-headed cat in the period of our camera-trapping survey. Given a total survey

effort of 6650 CTNs, and based on the detectability of this species in previous studies (Fig. 3), we had

a 25% chance of failing to detect this species. Most records of this species have been obtained within

3 km of large bodies of freshwater (including rivers and lakes) and below 100 m elevation (Wilting et

al., 2010a). None of our random camera trap locations were near large water bodies and, due to a

stratification inherent in the survey design, locations were at a mean elevation of 432 m (range: 278-


63

543 m). If our survey design had been random with respect to elevation, then it is possible that we

would have also detected this species with our camera traps.

We found significantly higher probabilities of detection along logging features and skid trails for

clouded leopard and marbled cat, respectively. In contrast, the leopard cat and bay cat were not found

to preferentially use logging features and apparently exhibited random use of habitat features. Habitat-

use patterns have rarely been investigated using camera-trapping data, due to the ubiquity of non-

random sampling and the narrow range of habitats this necessarily focuses upon. As a result, studies

have generally focussed on modelling detection rates as a function of the properties of the trail or road

itself (Maffei et al., 2004; Larrucea et al., 2007; Harmsen et al., 2009). The only other study that we

are aware of that has used strictly random locations found marked differences between on- and off-

trail trapping rates for a range of species on Barro Colorado Island, Panama, including a six-times

higher trapping rate on trails for ocelot, Leopardus pardalis (Kays et al., 2011). Our own results

support this for two other species of wild felid.

We also found evidence of sex-specific differences in the use of habitat-features, with female clouded

leopards avoiding logging-related features, possibly due to the risk of aggression or infanticide on the

part of males (Cheyne & Macdonald, 2011). Heterogeneity in capture probabilities between the sexes

has been previously noted in clouded leopards (Cheyne & Macdonald, 2011; Wilting et al., 2012) and

is an important source of bias in parameter estimation under a mark-recapture framework (Harmsen et

al., 2010). Our results suggest that females may be more likely to be recaptured, and heterogeneity

reduced, if traditional trail-based survey locations are supplemented with off-trail locations. For the

marbled cat, the finding that detection probabilities were eight-fold higher on skid trails relative to

other features including logging roads should be a point of further investigation, and may help to

explain the low detection frequencies of this species in previous studies (Fig. 2).


64

We obtained data from previous camera-trapping surveys carried out across Southeast Asia to

characterise for the first time the probability distribution of d, the detection frequency, for each

species of felid, and from this provide general recommendations for minimum survey efforts. Though

rarely available, this is vital information for the effective design of wildlife surveys. The detection

frequencies observed for our random survey locations deviated significantly from the expectation

based on previous studies in the case of the bay cat and leopard cat. Together with the differences

observed between off-trail and on-trail locations within our own study, this suggests that non-random

sampling regimes have resulted in biased inferences with respect to the relative abundance of these

species, especially for the bay cat.

The bay cat was listed as Endangered when it was last assessed under the IUCN Red List categories

and criteria (Hearn et al., 2008). This was on the basis of an estimated population size of less than

2500 mature individuals and a projected population decrease of more than 20% over 12 years. Since

this assessment was made, the proliferation of camera trap studies has yielded a number of new

records for the bay cat, both published (this study, Mohamed et al., 2009; Bernard et al., 2012; Brodie

& Giordano, 2012a; Samejima & Semiadi, 2012) and unpublished (Anonymous, 2011a, 2011b, 2012),

which has greatly expanded the known habitat tolerances of this species in terms of both disturbance

and maximum altitude (up to ~1500 m). It now seems likely that the bay cat can occur in highly-

disturbed forest, as well as the vast areas of upland forest (300 to 1,000 m elevation) and possibly

even montane forest (>1,000 m elevation) in the proposed Heart of Borneo transboundary

conservation area (WWF, 2012). Our finding that bay cat detection frequencies increase substantially

using random camera locations could also indicate a widespread underestimation of its relative

abundance. Considering these facts, a case could be made for reconsideration of the conservation

status of the bay cat during the next IUCN Red List cat assessments, due to be completed by 2015.

However, important uncertainties still remain in assessing the future for the bay cat, especially with

regards to land-use trends in the Yayasan Sabah Forest Management Area (Reynolds et al., 2011),


65

which is emerging as an apparent stronghold for the species, but also more broadly in the ongoing

land-use planning process for the Heart of Borneo area (Persoon & Osseweijer, 2008).

Camera traps are typically placed non-randomly in order to obtain a greater quantity of data per unit

of effort expended or money spent. We have shown here that, for certain species such as the bay cat,

this may not always be appropriate. Cameras and other wildlife sensors, such as sound recorders, are

rapidly improving in terms of sensitivity, battery life, data storage capabilities and robustness to

adverse environmental conditions, and are therefore producing more data per unit of effort or

monetary input than ever before. As a result, the traditional barrier to strictly random survey locations

– the paucity of data that may result – is rapidly being overcome. There will always be a role for non-

random placement in certain circumstances, such as when confirming the presence of a particular

species or using mark-recapture methods, but otherwise we advocate the adoption of random survey

locations and an emphasis on quality of data – as judged by the robustness and generality of

conclusions that can be drawn – rather than quantity of data per se. This will be especially productive

for study sites or study species which are poorly known, such as the bay cat, or for multi-species

surveys, as a means of controlling for differential use of habitat features across species or between

sexes within the same species.


66

Appendices

Appendix A – Camera-trapping data from previous studies (1998-2012)


67

Table A1. Previous camera-trapping studies conducted in Southeast Asia (1998 to 2012) and the number of captures obtained for Bornean felid species.

Reference Survey years

Location (* = sites with all 5 Bornean wild cat species)

Study site code

Borneo Wild cat captures [number of individuals] Number

of Bornean

cat species

Camera trap

nights Logged (L) / Old-growth

(OG)

Clouded leoparda

Marbled cat

Bay cat

Flat-headed

cat

Leopard cat

Lynam (2006) 1998 Ta Phraya NP, Thailand 1

OG

0.9/100 camera trap

nights 0 - - ≥1 3 Lynam (2006) & Jenks et al.

(2011) 1999-2000 Khao Yai NP, Thailand 2

OG 6 1 - - 10 3 1,226

Jenks et al. (2011) 2003-2007 Khao Yai NP, Thailand 2

OG 8 1 - - 4 3 6,260

Lynam (2003) 1999-2002 Myanmar (17 different sites) 3

OG 50 15 - - 80 3 15,560

Grassman (2003) 1999-2003 Phu Khieo Wildlife Sanctuary, Thailand 4

OG 2 0 - 0 4 2 1,224

Rao et al. (2005) 2002-2003 Hkakaborazi National Park buffer zone, Myanmar 5

OG & L 77 0 - - 49 2 1,238

Kawanishi & Sunquist (2004) 1999-2000 Merapoh, Taman Negara National Park, Peninsular Malaysia 6

OG 5 1 - 0 59 3 4,336

Kawanishi & Sunquist (2004) 2000-2001 Kuala Terengan, Taman Negara National Park, Peninsular Malaysia 6

OG 6 5 - 0 1 3 4,847

Kawanishi & Sunquist (2004) 2000-2001 Kuala Koh, Taman Negara National Park, Peninsular Malaysia 6

OG 5 10 - 0 2 3 4,871

Mohd-Azlan & Sharma(2006) & Mohd-Azlan (2006) 2000-2001

Jerangau FR, Ulu Terenganu, Peninsular Malaysia 7

L 13 1 - 0 86 3 5,972

Yasuda et al. (2007) 2000-2003 Tasek Merimbun Heritage Park, Brunei Darussalam 8 L 0 0 0 4 1 2 334

Imai et al. (2009) 2002 Tabin Wildlife Reserve, Sabah, Malaysia* 9 L 1 1 0 1 2 4 574

Yasuda et al. (2007) 2002-2003 Pasoh FR, Peninsular Malaysia 10

OG 0 0 0 0 2 1 3,659

Yasuda et al. (2007) 2003-2006 Deramakot FR, Sabah, Malaysia* 11 L 5 0 0 2 7 3 981 Yasuda et al. (2007) & Yasuda et al. (2009) 2004-2007

Sungai Wain Protection Forest, Kalimantan, Indonesia 12 OG 0 0 1 0 0 1 770

Yasuda et al. (2007) & Yasuda et al. (2009) 2004-2007

Bukit Soeharto Research & Education Forest, Kalimantan, Indonesia 13 L 0 0 0 0 1 1 840

Scott & Gemita (2004) 2003 Jambi Province, Sumatra, Indonesia 14

L 0 0 - 0 104 1 3,010

Scott & Gemita (2004) 2003 Jambi Province, Sumatra, Indonesia 14

Oil Palm 0 0 - 0 16 1 522 Johnson et al. (2009) & Johnson et al. (2006) 2003-2006

Nam Et-Phou Louey National Protected Area, Laos 15

OG ≥5 39 - - 24 3 8,499

Mohd-Azlan & Sanderson (2007) 2003-2006 Various in Sarawak, Malaysia 16 L?

1

5,034

Mohd-Azlan & Lading (2006) 2004 Lambir Hills NP, Sarawak, Malaysia 17 OG 3 0 0 0 0 1 1,127

Kitamura et al. (2010) 2004-2007 Hala-Bala Wildlife Sanctuary, 18

OG & L 4 0 - 0 13 2 11,106


68

Thailand

Giman et al. (2007) 2005 PFZ, Sarawak, Malaysia 19 L + Acacia 0 0 0 0 1 1 1,632

McShea et al. (2009) 2005-2007 PFZ, Sarawak, Malaysia 19 L + Acacia ≥1 ≥1 0 0 ≥4 3 5,679

Mathai et al. (2010) 2005-2008 Upper Baram, Sarawak, Malaysia – Logged 20 L 0 0 0 0 0 0 461

Mathai et al. (2010) 2005-2008 Upper Baram, Sarawak, Malaysia – Old-growth 20 OG 3 1 0 0 1 3 4,791

Datta et al. (2008) 2006-2007 Namdapha National Park, India 21

OG & L 2 ≥1 - - ≥1 3 1,537

Cheyne & Macdonald (2011) 2008-2009 Sabangau Forest, Kalimantan, Indonesia 22 L 29 [4] 4 0 7 25 4 5,777

Mohamed et al. (2009) 2008-2009 Deramakot FR, Sabah, Malaysia* 11 L 10 [1] 0 1 4 183 4 1,916

Samejima et al. (2012) 2008-2009 Pre-harvested plots, Deramakot FR, Sabah, Malaysia* 11 L ≥6 ≥4 ≥1 0 ≥2 4 11,550

Samejima et al. (2012) 2008-2009 2 years after RIL plot, Deramakot FR, Sabah, Malaysia* 11 L 0 0 3 1 0 2 770





Imai et al. (2009) 2008-2009 Tangkulap FR, Sabah, Malaysia 23 L ≥1 ≥1 0 0 ≥2 3 4,320

Bernard et al. (2012) 2009-2010 Tabin Wildlife Reserve, Sabah, Malaysia* 9 L 5 2 2 0 7 4 2,700

Bernard et al. (2012) 2009-2010 Fragments near Tabin Wildlife Reserve, Sabah, Malaysia 24

L (Fragments) 0 0 0 0 13 1 1,033

Wilting et al. (2012) & Wilting et al. (2009, unpublished report) 2009

Tangkulap-Pinangah FR, Sabah, Malaysia 23 L 29 [5] 1 0 ≥2 ≥1 4 2,688

Wilting et al. (2012) & Wilting & Mohamed (2010, unpublished) 2010 Segaliud-Lokan FR, Sabah, Malaysia 25 L 15 [5] 3 0 0 >15 3 2,640

Brodie & Giordano (2011 & 2012) 2010

Maliau Basin Conservation Area buffer, Sabah, Malaysia (and adjacent area) 26 L 1 [1] 0 0 0 0 1 1,165

Brodie & Giordano (2011 & 2012) 2010

Maliau Basin Conservation Area, Sabah, Malaysia 26 OG 59 [4] 3 0 1 20 4 1,747

van Berkel et al. (unpublished) [T. van Berkel, 2012, personal communication] 2010-2011 Mohot River, Kalimantan, Indonesia 27 OG 0 0 0 0 ≥1 1 546


69

Reference Study

duration (months)

Trap sites Camera trap nights per site

Study area size

(km2)

Taxonomic focus

Inter-trap distance (km)

Paired cameras?

Flash (white or infrared)

Lure/ Bait?

Relative abundance

index? Occupancy? Density?

Lynam (2006) 5 28 30-40

Tigers & prey 1-2 Some White

Lynam (2006) & Jenks et al. (2011) 19 34 30-40

Tigers & prey 1-2 Some White

Jenks et al. (2011) 42 217 ≥21

Mammals ~1 N White

Lynam (2003) 36 430 >24 5491 Tigers ~2 Some White

Grassman (2003) 48 <200 30 200 Mammals 0.5-1 N White

Rao et al. (2005) 12 64 30-35

Mammals & birds 0.5 N White

Kawanishi & Sunquist (2004) 14 47 92 165

Tigers & prey 4 Y White

Kawanishi & Sunquist (2004) 11 43 113 164 Tigers & prey 4 Y White

Kawanishi & Sunquist (2004) 11 45 108 151 Tigers & prey 4 Y White

Mohd-Azlan & Sharma(2006) & Mohd-Azlan (2006) 21 24 284

Tigers 1.8-2.2 N White

Yasuda et al. (2007) ~33

Mammals

N White Imai et al. (2009) 1

Mammals

N White

Yasuda et al. (2007)

Mammals

N White Yasuda et al. (2007) ~24

Mammals

N White

Yasuda et al. (2007) & Yasuda et al. (2009)

10

Mammals 0.2 N White

Yasuda et al. (2007) & Yasuda et al. (2009)

22

Mammals 0.2 N White

Scott & Gemita (2004) 9 >48 ≥14

Mammals 0.5 N White

Scott & Gemita (2004) 9 48 ≥14

Mammals 0.5 N White

Johnson et al. (2009) & Johnson et al. (2006) 37 273 >37 500

Tigers & prey ~2 Y White

Mohd-Azlan & Sanderson (2007) 37

Mammals

? White

Mohd-Azlan & Lading (2006) 8 7 161

Mammals

N White

Kitamura et al. (2010) 36 45 247

Mammals & birds 0.3-0.5 N White

Giman et al. (2007) 5 83 <30

Mammals 0.2 minimum N White McShea et al. (2009) 22 212 14-36 (mean= 27) 644 Mammals 0.2 N White

Mathai et al. (2010) 45

150 Mammals

N White


70

Mathai et al. (2010) 45

150 Mammals

N White Datta et al. (2008) 4 80 15 360 Mammals 0.2-0.5 N White

Cheyne & Macdonald (2011) 18 27 214 7 Clouded leopard

Y

White & infrared

Mohamed et al. (2009) 7 48 42 112 Felids

1.2-2.4 (mean = 1.7) Y White

Samejima et al. (2012) 19 >135 90-150 47

Mammals & birds

N White

Samejima et al. (2012) 19 ≥9 90-150 3

Mammals & birds

N White

Samejima et al. (2012) 19 ≥9 90-150 3

Mammals & birds

N White

Samejima et al. (2012) 19 ≥9 90-150 3

Mammals & birds

N White

Samejima et al. (2012) 19 ≥9 90-150 3

Mammals & birds

N White

Samejima et al. (2012) 19 ≥9 90-150 3

Mammals & birds

N White

Imai et al. (2009) 11 81 53 7

Mammals & birds

N White

Bernard et al. (2012) 16 >9

Mammals 0.2 N White Bernard et al. (2012) 16 >8

Mammals 0.05-0.1 N White

Wilting et al. (2012) & Wilting et al. (2009, unpublished report) 5 64 42 122 Felids 1.7 Y White

Wilting et al. (2012) & Wilting & Mohamed (2010, unpublished) 4 55 48 114 Felids 1.7 Y White

Brodie & Giordano (2011 & 2012) 5 26 81-128

Clouded leopard 1-2 Y Infrared

Brodie & Giordano (2011 & 2012) 5 26 123-126

Clouded leopard 1-2 Y Infrared

van Berkel et al. (unpublished) [T. van Berkel, 2012, personal communication] 1 ≥25 3-43 (mean = 22)

Mammals & birds 0.5 N

White & infrared

aBoth recognised species of clouded leopard are here included (Neofelis nebulosa and N. diardi).

Chapter 3: Mammalian species richness and β-diversity across a tropical land-use gradient

71

Chapter 3:

Grain-dependent responses of mammalian species richness and β-diversity to land-use

and the implications for managing conservation values in tropical human-modified

landscapes

Abstract

Diversity responses to land-use change are poorly understood at local scales, hindering our ability to

make forecasts and management recommendations at scales which are of practical relevance. A key

barrier in this has been the under-appreciation of grain-dependent diversity responses and the role that

β-diversity – variation in community composition across space – plays in this. Decisions about the

most effective spatial arrangement of conservation set-aside, for example High Conservation Value

areas, have also neglected β-diversity, despite its role in determining the complementarity of sites. We

investigated local-scale richness and β-diversity at multiple spatial grains for large and small

mammals, across old-growth forest, logged forest and oil palm plantations in Borneo, using intensive

camera- and live-trapping. β-diversity was quantified by comparing observed β-diversity with that

obtained under a null model, in order to control for sampling effects, and we refer to this as the β-

diversity signal. Community responses to land-use were grain-dependent, with large mammals

showing reduced richness in logged forest compared to old-growth forest at the grain of individual

sampling points, but no change at the overall land-use level. Responses varied with species group,

however, with small mammals increasing in richness at all grains in logged forest compared to old-

growth forest. Both species groups were significantly depauperate in oil palm. Large mammal

communities in old-growth forest became more heterogeneous at coarser spatial grains and small

mammal communities became more homogeneous, whilst this pattern was reversed in logged forest.

Both groups, however, showed a significant β-diversity signal at the finest grain in logged forest, most

likely due to logging-induced environmental heterogeneity. The β-diversity signal in oil palm was

weak, but heterogeneity at the coarsest spatial grain was still evident, likely due to variation in


72

landscape forest cover. Our findings suggest that the most effective spatial arrangement of

conservation set-aside will involve a trade-off for large and small mammals. Greater consideration in

the conservation and management of tropical landscapes needs to be given to β-diversity at a range of

spatial grains.

3. 1. Introduction

It is widely acknowledged that global biodiversity is in decline (Butchart et al., 2010), primarily due

to unprecedented rates of habitat loss and degradation (Hansen et al., 2013). Many attempts have been

made at quantifying this biodiversity loss due to land-use change at coarse scales and forecasting this

into the future (Sodhi et al., 2004; Koh & Ghazoul, 2010a; Wearn et al., 2012; Pimm et al., 2014),

with the aim of informing policy-making at the highest administrative levels. In reality, biodiversity

loss at coarse scales is a summation of the changes occurring at the local scale of landscapes, such as

forestry concessions or private landholdings. Local stakeholders often make management decisions

that have substantial impact on the outcomes for biodiversity in these landscapes. At this local scale,

however, there is little consensus about the community responses to land-use, which hinders our

ability to make management recommendations and biodiversity forecasts at scales relevant to local

stakeholders.

Much confusion surrounding local-scale biodiversity responses has arisen due to an under-

appreciation of spatial grain (Hamer & Hill, 2000; Whittaker et al., 2001; Sax & Gaines, 2003). At the

smallest scales (e.g. those of a quadrat or plot), species richness has been shown to be stable

(Dornelas et al., 2014) or even increasing in post-disturbance areas (Vellend et al., 2013). On the other

hand, a number of other meta-analyses focussing on overtly disturbed areas, and which did not

account for spatial grain, have shown the contrasting result of declines in species richness (Dunn

2004; Scales & Marsden 2008; Sodhi et al. 2009; Gibson et al. 2011; Burivalova et al. 2014; Newbold

et al. 2015). It is difficult to completely reconcile these two apparently conflicting findings, and make

firm conclusions with respect to local-scale biodiversity responses, due to the lumping together of


73

studies using vastly different spatial grains. For example, in a review of past studies, Hill & Hamer

(2004) found that the effects of disturbance on Lepidoptera (moths and butterflies) and birds were

strongly grain-dependent. Specifically, in response to disturbance, Lepidoptera richness often

increased at small scales (< 1 ha) and decreased at intermediate scales (1 – 25 ha), while bird richness

also decreased at intermediate scales but then increased at still larger scales (> 25 ha). Although

consideration of spatial grain has largely been neglected in global meta-analyses, it offers the potential

of uniting seemingly contradictory results and allowing better forecasting of biodiversity changes at

the local-scale. An essential component in this framework will be a better understanding of

community variance, or β-diversity, which largely determines the relationship between spatial grain

and richness (Scheiner, 2004; Drakare et al., 2006). Indeed, changes in β-diversity can potentially

explain how, in response to land-use change, species richness might remain constant or even increase

at the level of a sampling point, yet decline at the level of a study site.

β-diversity patterns are important in systematic conservation planning, as they determine the

complementarity of communities across sites (Ferrier, 2002). This also applies, at smaller scales, to

decisions about how to allocate conservation set-aside. Major certification schemes, including those

of the Forest Stewardship Council (FSC), Round-table on Responsible Soy (RTRS) and Round-table

on Sustainable Palm Oil (RSPO), require concession holders to identify and set-aside forest patches

with High Conservation Value (HCV), but do so without explicit consideration of local-scale β-

diversity. Although it is an open question whether β-diversity is itself a HCV that should be

maintained, it is nonetheless a crucial determinant of the values conserved within set-aside,

particularly in the context of deciding the spatial distribution of patches and how large each patch

should be (Nekola & White, 2002). This is relevant in the context both of the expansion of cropland

and tree plantations into forested landscapes, which is ongoing at a rapid rate (Phalan et al., 2013;

Wilcove et al., 2013), and of the increasing uptake of sustainability principles by logging companies,

as required under certification schemes such as the FSC, but also more broadly under the banner of

retention forestry (Gustafsson et al., 2012; Lindenmayer et al., 2012).


74

Selective logging is the main driver of tropical forest degradation worldwide (Asner et al., 2009;

Bryan et al., 2013) and, by modifying the structure (Cannon et al., 1994; Hill, 1999; Bischoff et al.,

2005), resources (Johns, 1988; Heydon & Bulloh, 1997; Munshi-South et al., 2007) and micro-climate

(Hardwick et al., 2015) of forests through space, may act as a strong environmental filter on the

occurrence patterns of species post-logging (Cleary et al., 2005; Kitching et al., 2013). Only a handful

of studies have investigated β-diversity in logged forests, but these support the notion that

environmental heterogeneity in logged forests increases β-diversity (Hill & Hamer 2004; Berry et al.

2008; Woodcock et al. 2011, but see Kitching et al. 2012). Plantation habitats, by contrast, may be

more homogeneous in space than natural forest, not only in terms of floral species composition, but

also in terms of structure, resources and micro-climate (Thiollay, 1995; Scales & Marsden, 2008).

This may be particularly true of oil palm (Elaeis guineensis) plantations (Luskin & Potts, 2011; Foster

et al., 2011), which are expanding across the tropics at a rapid rate, particularly in Southeast Asia

(Wilcove et al., 2013).

Across taxa, β-diversity may vary depending on dispersal capacity, as well as the home-range size of

individuals: all else being equal, poor dispersers with small home-ranges will both be more dispersal-

limited and less able to buffer spatial variation in habitat quality, leading to higher β-diversity.

Certainly, Soininen et al. (2007a) found evidence across past studies that larger-bodied organisms,

which have higher dispersal capacity and larger home-ranges, generally exhibited lower levels of β-

diversity. Despite the expected differences among taxa, very few studies have explored this at the

local scale using data collected simultaneously on multiple species groups at the same spatial

locations (but see Dormann et al. 2007; Kessler et al. 2009; Cabra-García et al. 2012; Gossner et al.

2013).

The primary aim of our study was to quantify the species richness and β-diversity of mammal

communities across a land-use gradient and investigate whether diversity responses to land-use were

dependent on spatial grain. In doing so, we used robust estimators and comparisons with null models


75

to control for the specific properties of our sampling design. As a secondary aim, we also investigated

differences in richness and β-diversity among large and small mammals across a range of spatial

grains. We chose mammals as our focus due to the fact that they are a high-profile group that are

often the targets of policy and land-use decisions, and are often given strong weighting in

conservation planning, especially the HCV assessment process.

We made three specific hypotheses with regard to β-diversity. We hypothesised that logged forest

areas would be more environmentally heterogeneous than old-growth forest, therefore giving rise to

higher levels of β-diversity (HI), whilst oil palm would be environmentally homogeneous, giving rise

to lower levels of β-diversity (HII). We also hypothesised that small mammals (< 1 kg) would be more

dispersal-limited than large mammals, owing to their smaller body size, and less able to buffer fine-

grained variation in habitat quality (HIII). We therefore expected small mammals to exhibit greater

levels of β-diversity than large mammals. To address these hypotheses, we gathered one of the most

comprehensive datasets on local-scale mammal occurrence from the tropics that we are aware of,

using multiple sampling methods to incorporate nearly the entire non-volant community, from the

smallest murid rodents (~0.03 kg) up to the Asian elephant Elephas maximus (~2,700 kg). Our

findings with respect to the importance of spatial grain and β-diversity have important implications

for the conservation and management of biodiversity in these systems and, in particular, with regard

to optimal designs for conservation set-aside.

3. 2. Methods

3. 2. 1. Study sites and sampling design

We sampled mammals in three different land-uses, taking advantage of the experimental design of the

Stability of Altered Forest Ecosystems (SAFE) Project in Sabah, Malaysian Borneo (Ewers et al.,

2011). This consists of old-growth forest within the Maliau Basin Conservation Area, repeatedly-

logged forest within the Kalabakan Forest Reserve and two adjacent oil palm plantations straddling

the Kalabakan Forest Reserve boundary (see Appendix A for detailed study site descriptions).


76

We used a hierarchical nested sampling design in order to explore β-diversity at three different spatial

grains (Fig. 1). We based this on the fractal sampling design of the SAFE Project (Ewers et al., 2011),

which is an especially efficient design for quantifying β-diversity (Marsh & Ewers, 2012). At the

lowest level in the hierarchy were individual sampling points. These were clustered into rectangular

sampling grids, which we call here plots, of (4 x 12 =) 48 points separated by 23 m. At the highest

level in the hierarchy, plots were clustered together into blocks. These were arranged differently in the

logged forest compared to the other two land-uses (Fig. 1), in order to overlay the locations of future

experimental fragments (Ewers et al., 2011), but separation distances between plots (170 to 290 m)

and between blocks (0.6 to 3 km) were similar across the land-uses. The spatial arrangement of

sampling points at the SAFE Project has been deliberately designed to minimise confounding factors

across the land-use gradient, including latitude, slope and elevation (Ewers et al., 2011), and this

applied equally to our sampling design for mammals.

Figure 1. Sampling design across (a) old-growth forest, (b) oil palm and (c) logged forest used in this study, illustrating the three spatial grains within each land-use: individual sampling points, 1.75 ha rectangular plots (consisting of clusters of points) and blocks (consisting of clusters of plots). Blocks were arranged identically in old-growth forest and oil palm, and were arranged to coincide with future experimental forest fragments in logged forest. Separation between points, plots and blocks was nonetheless similar across land-uses. Shaded areas lie outside the Kalabakan Forest Reserve, consisting of a 2,200 ha Virgin Jungle Reserve (Brantian-Tatulit) to the south and an extensive (>1 million ha) area of logged forest to the north (Mount Louisa Forest Reserve and other contiguous forest reserves). Insets show the location within insular Southeast Asia and the spatial proximity of the three land-uses within southeast Sabah, Malaysian Borneo.


77

3. 2. 2. Mammal sampling

Small mammal trapping was conducted at the level of the plot, with each session consisting of seven

consecutive days. Two locally-made steel-mesh traps (18 cm wide, 10-13 cm tall and 28 cm in

length), baited with oil palm fruit, were placed at or near ground level (0 - 1.5 m) within 10 m of each

grid point. Traps were checked each morning and captured individuals were anaesthetised using

diethyl ether (following Wells et al. 2007), measured, permanently marked using a subcutaneous

passive inductive transponder tag (Francis Scientific Instruments, Cambridge, UK), identified to

species using Payne et al. (2007) and released at the capture location. Trapping was carried out

between May 2011 and March 2014, during which there were no major mast-fruiting events. Some

plots (8 of 31) were sampled more than once over this period (mean effort per plot = 925 trap nights).

We deployed camera traps (Reconyx HC500, Holmen, Wisconsin, USA) at a random subset of grid

points within plots, setting the cameras as close to the points as possible and strictly within 5 m. The

random deployment of cameras in this manner has rarely been used before, though is essential for

revealing species-specific patterns of space-use (Chapter 2), which is a contributor towards β-

diversity. Cameras were fixed to trees or wooden poles, or placed within locally-made steel security

cases in areas of high human traffic, with the camera sensors positioned at a height to maximise

detection for a range of species (most often 30 cm, though this was flexible depending on the terrain

encountered at each location). No bait or lure was used and disturbance to vegetation was kept to a

minimum. Camera traps were active between May 2011 and April 2014, during which most plots (39

of 42) were sampled for multiple sessions (mean points per plot = 13; mean effort per plot = 625 trap

nights).

In total, 543 points were camera-trapped and 1,488 points were live-trapped, and we used these

datasets for estimating large and small mammal species richness, respectively. Both trapping

protocols were used at 430 points and we used only this subset of the data for the β-diversity analyses.


78

This subset included data from 31 plots nested in 8 blocks (9 plots in 3 blocks for old-growth forest;

16 plots in 3 blocks for logged forest, and 6 plots in 2 blocks for oil palm).


All analyses were ultimately derived from the separate community matrices from live-trapping and

camera-trapping, with trap nights forming rows of the matrices, species forming the columns and each

cell containing the number of independent capture events. Unlike live traps, camera traps are

continuous-time detectors, so we considered photographic capture events to be independent if they a)

contained different individuals or b) were separated by > 12 hours, which matches the approximate

minimum separation between live trap events.

Our hierarchical sampling design allowed us to partition species richness and β-diversity into multiple

spatial grains across the three land-uses, by aggregating the community matrices to the appropriate

grain. However, unequal levels of effort, replication and sample completeness across spatial grains

and across land-uses makes comparisons of richness and β-diversity problematic, an issue that has

often been neglected in past studies (Beck et al., 2013).

For species richness, there are non-parametric estimators which can be used to make richness values

more robust to sampling design variation. We used the Abundance-based Coverage Estimator (ACE)

for overall richness in each land-use, because we were confident that sufficient sampling had been

done to estimate the minimum asymptotic richness (Gotelli & Chao, 2013), whilst we standardised

point richness to 90% sample coverage (Colwell et al., 2012). We hereafter refer to overall richness in

each land-use and point richness as γ-diversity and α-diversity, respectively. For both γ- and α-

diversity, we used the full camera trap and live trap datasets to make estimates for large and small

mammals, respectively.


79

We modelled either large or small mammal α-diversity using a Poisson generalised linear mixed-

effects model, with the hierarchical sampling design specified in the random effects (points nested

within plots, in turn nested within blocks), as well as an observation-level random effect to account

for overdispersion (Harrison, 2014). We note that, given our relatively fine-scale sampling of

mammal communities, spatial variation in α-diversity is due to coarse-scale species occurrence

patterns, as well as finer-scale patterns of habitat-use by individual animals. The random effects

helped account for any spatial dependence between sampling points, caused by detections of the same

individuals at multiple points. We also made estimates of γ- and α-diversity across large and small

mammals for the subset of locations which had been sampled using both live traps and camera traps.

In this case, we were able to model α-diversity as a function of both land-use and species group (large

or small mammal), as well as their interaction.

Commonly used metrics of β-diversity are also sensitive to the specific sampling design employed

(see Appendix B for more information). Instead of using β-diversity values directly, we compared

observations with an appropriate null model (Crist et al., 2003; Kraft et al., 2011), an approach which

has been underexploited to date (Lessard et al., 2012). Differences from the null model, calculated

using simple subtraction (βobserved – βnull), can be interpreted as a measure of β-diversity due to

community assembly processes (including those of intraspecific aggregation, environmental filtering

and dispersal limitation), over and above that due to the vagaries of the sampling process itself. We

refer to this difference between observed and null β-diversity as the β-diversity signal (as opposed to

the random β-diversity noise). Observed β-diversities were calculated using Lande’s (1996) additive

formulation, in which β-diversity at a given level, i, in a hierarchy is the average richness at the given

level substracted from that in the level above: βi = αi+1 – αi. This was done for each combination of

land-use (old-growth forest, logged forest and oil palm) and species group (large mammals, small

mammals or both combined), for each of three spatial grains: points (camera detection zone = 0.02

ha), plots (1.75 ha in area) and blocks (covering an average minimum convex polygon of 25 ha;

range: 24.1 – 25.4). It follows from Lande’s (1996) additive diversity partitioning that overall


80

observed γ-diversity of each land-use is: αpoint + βpoint + βplot + βblock. We used additive partitioning

because β-diversity is in units of species richness in this framework, which means differences from

null models can also be calculated in units of species richness using simple subtraction (βobserved –

βnull), allowing more straightforward comparisons between land-uses, between species groups and

between hierarchical levels.

To estimate null β-diversities, we used null models based on the sample-based randomisations of Crist

et al. (2003). For each spatial grain i in the hierarchy, we randomly shuffled (without replacement) the

community samples at the level below (i – 1), whilst constraining the random placements to maintain

the integrity of any higher-level (i + 1) spatial nesting. For example, null β-diversity for the plot level

was derived by randomly shuffling point-level communities amongst plots, but only amongst plots

within the same block. By constraining the null model in this way, we were able to test for differences

from null at the specific spatial grain of interest. We extended this to the case of multiple sampling

methods, by keeping the matrices derived from live-trapping and camera-trapping separate and

conducting the randomisations in parallel, mimicking how the data were generated. This also allowed

us to specifically control for the different sampling efforts achieved during live-trapping and camera-

trapping.

By repeating the randomisation process, we obtained distributions of differences from the null. We

calculated the 95% quantiles of these distributions and deemed differences to be significant if the

quantiles did not overlap zero. Sample size necessarily declines at the higher spatial levels of a fractal

sampling design, causing a loss in the precision of β-diversity estimates (Marsh & Ewers, 2012). This

was also true of our null model approach, because we had fewer community samples to shuffle at

higher levels. We used 1000 randomisations in all cases, except for our oil palm sampling design, in

which there were few possible combinations of placing plots within blocks. In this case we restricted

the number of randomisations to the number of combinations (n = 40).


81

We modelled the differences from null using linear mixed-effects models in order to explore

differences across land-use, across spatial grains and across the two species groups. Since β-diversity

at a given hierarchical level is, in the additive framework, the mean of the number of “missing”

species in each sample (species which are absent from a sample but present at the level above), we

took advantage of this by extracting the un-averaged number of missing species for each sample. We

calculated the difference from null for each of these observed values and accounted for the lack of

independence between values by specifying the hierarchical sampling design in the random effects

structure. Point-level values were nested within plots and blocks, whilst plot-level values were nested

within blocks. For the block-level model, no random effects were specified because this was the

highest level in the hierarchy.

Finally, using the approach outlined by Baselga (2010a), we differentiated between the two broad

proximate causes of β-diversity – species turnover and nestedness (see Appendix B for more

information) – to investigate which was primarily responsible for β-diversity at each spatial grain in

the three land-uses. Given the sample size dependence of these measures, we calculated them over

100 random sub-samples of our data (Baselga, 2010a), taking the minimum sample sizes at each

hierarchical level across the whole dataset each time (8 points per plot, 3 plots per block and 2 blocks

per land-use). This would still not enable fair comparisons across spatial grains, so we calculated

values as a proportion of the total β-diversity (nestedness and turnover), as given by the Sørensen

index (Baselga, 2010a). We modelled the proportion of β-diversity in the nestedness component using

beta regression models with a log link and constant dispersion parameter, constructing separate

models with land-use, species group or spatial grain as the explanatory variable. We used the

combined live trap and camera trap dataset for this analysis, removing 12 points which had been

camera-trapped for less than 7 days.


82

All analyses were done in R version 3.1.0 (R Development Core Team, 2014), using the additional

packages vegan 2.0-10 (Oksanen et al., 2013), iNEXT 1.0 (Hsieh, 2013), plyr 1-29 (Wickham, 2010),

lme4 1.1-6 (Bates et al., 2014) and betareg 3.0-5 (Cribari-Neto & Zeileis, 2010).

3. 3. Results

Live-trapping resulted in a total of 4,046 captures of 25 mammal species over 28,681 trap nights,

whilst camera-trapping resulted in a total of 12,788 independent captures of 58 mammal species over

26,251 trap nights. This gave a total of 65 mammal species (Appendix D), of which 19 species were

captured using both protocols. Over the points sampled using both live traps and camera traps (n =

430), we obtained 11,579 captures of 61 species over a combined effort of 27,176 trap nights.

Live-trapping of 2,976 locations over 28,681 trap nights resulted in a total of 4,046 captures of 25

species, whilst camera-trapping of 543 locations over 26,251 trap nights resulted in a total of 12,788

independent captures of 58 species. This gave a total of 65 mammal species (Appendix D), of which

19 species were captured using both protocols. Over the 430 locations sampled using both live traps

and camera traps, we obtained 11,579 captures of 61 species over the combined effort of 27,176 trap

nights.

Species accumulation curves in each land-use were closely approaching an asymptote (Appendix C,

Fig. C1), all with an estimated sample coverage ˃ 98%. Logged forest had the highest observed and

estimated total mammal γ-diversity, though the 95% confidence intervals overlap with those for old-

growth forest (Fig. 2). Of the 44 species found in old-growth forest, 39 species (89%) were also

detected in the logged habitats. Oil palm plantations were a significantly depauperate habitat (Fig. 2),

harbouring just 22 of the 63 species (35%) found in the forest habitats, in addition to the plantain

squirrel Callosciurus notatus and invasive domestic dog Canis familiaris. We separately calculated oil

palm γ-diversity after excluding 6 grid points which fell within a 200 m wide margin of forest-scrub


83

habitat that was connected to a 45 km2 block of logged forest (Appendix C, Table C1), which left just

19 forest species (30%) that were found in the oil palm crop itself.

The overall γ-diversity differences between land-uses were in large part due to the small mammals.

Observed and estimated large mammal γ-diversities were very similar for old-growth and logged

habitats (Fig. 2) and, for the full camera trap dataset, the 95% confidence intervals for oil palm

overlap, albeit slightly, with those of old-growth forest (Appendix C, Table C1). In contrast, small

mammal γ-diversity was significantly different among all three land-use contrasts, except for a slight

overlap in 95% confidence intervals between old-growth forest and oil palm in the combined live trap

and camera trap dataset (Fig. 2; Appendix C, Table C1).

Figure 2. Diversity partitions for all mammals, large mammals and small mammals across a gradient of land-uses, including observed values (± SD) at four spatial grains and estimated α- and γ-diversities (± 95% CI). Estimates of α-diversity (standardised to 90% sample coverage) are predictions from a mixed-effects model which accounted for the hierarchical nested sampling design. Estimates of γ-diversity were calculated using the Abundance-based Coverage Estimator (ACE). Only data from sampling points which were both camera-trapped and live-trapped were used in this figure (see Appendix C for full results).


84

Mixed-effects models of α-diversity (standardised to 90% sample coverage) indicated a significant

effect of land-use, for both small mammals from the live trap data (χ2(2) = 119, p < 0.0001) and large

mammals from the camera trap data (χ2(2) = 21.7, p < 0.0001). As with γ-diversity, large and small

mammals showed markedly different responses for α-diversity (Fig. 2), which was confirmed by a

significant interaction term between land-use and species group using the combined live trap and

camera trap dataset (χ2(2) = 251, p < 0.0001). In this model, logged forest had a significantly higher α-

diversity than old-growth forest for small mammals (3.7 times higher, z = 14.2, p < 0.0001) and a

significantly lower α-diversity for large mammals (24% lower, z = -2.51, p = 0.01). This difference

between the two forest habitats was also significant for small mammal α-diversity with the full live

trap dataset (5 times higher in logged forest, z = 6.76, p < 0.0001), but was not significant for large

mammals when the full camera trap dataset was used (19% lower in logged forest, z = -1.54, p =

0.12). Oil palm was, again, highly depauperate compared to the forest habitats (either with or without

the locations in the forest-scrub boundary; Appendix C, Table C2), and this difference was significant

for both small mammals from the live trap data (compared to old-growth forest: z = 4.61, p < 0.0001)

and large mammals from the camera trap data (compared to logged forest: z = -3.47, p < 0.01).

Diversity partitioning suggested that the majority of the γ-diversity was contained in the β-diversity

components (Fig. 2): 83% in old-growth forest and 84% in both logged forest and oil palm. The

percentages for each of the spatial grains also appear broadly similar for overall mammal diversity

(Fig. 2): 38%, 38% and 30% as βpoint-diversity; 20%, 25% and 27% as βplot-diversity, and 25%, 20%

and 28% as βblock-diversity for old-growth forest, logged forest and oil palm, respectively. However,

the proportion of diversity contained within the β components across land-use, especially βplot and

βblock, is markedly different for large and small mammals (Fig. 2).

Null model comparisons demonstrated that most community samples had a significant signal of non-

random assembly processes (as evidenced by 95% confidence intervals which did not overlap zero;

Appendix C, Table C3). In old-growth forest, the β-diversity signal at large spatial grains was


85

increasingly strong for large mammals and increasingly weak for small mammals, whilst this pattern

was reversed in logged forest (Fig. 3). The β-diversity signal in oil palm was found to be much lower

overall, due in part to the depauperate nature of the mammal community that exists there, especially

for small mammals. However the β-diversity signal for large mammals in oil palm was still

comparable at the point level to that found in old-growth forest, and did not decline at the block level

as it did in logged forest (Fig. 3).

Figure 3. β-diversity differences from null models (± SE) with increasing spatial grain, for all mammals, large mammals and small mammals. Panels show results across a gradient of land-uses. The horizontal line at y=0 represents the case of no difference between observed β-diversity and expected β-diversity from null models. Dashed vertical lines show the three spatial grains of β-diversity sampling within each land-use (points, plots and blocks). Smoothed lines between data points are to aid interpretation. See Appendix C, Table C3 for 95% CIs.

Mixed-effects models of βpoint differences from null showed significant differences among the land-

uses (χ2(2) = 7.70, p = 0.02) and among the species groups (χ2

(1) = 13.94, p < 0.001). These were driven

by: an increase from old-growth to logged habitats (showing support for HI); a reduction in oil palm

(showing support for HII), and the consistently higher values, irrespective of land-use, for large

mammals (showing no support for HIII). The interaction between land-use and species group was not

significant at this spatial grain (χ2(2) = 3.31, p = 0.19). There were no consistent differences in βplot or

βblock differences from null, either between land-uses (plot-level: χ2(2) = 0.87, p = 0.65; block-level: F(2,

10) = 0.30, p = 0.75) or species groups (plot-level: χ2(1) = 0.92, p = 0.34; block-level: F(1, 10) = 1.17, p =


86

0.30), showing no support at this spatial grain for any of HI to HIII. The interaction terms in both

models were also not significant (plot-level: χ2(2) = 0.28, p = 0.87; block-level: F(2, 10) = 0.63, p = 0.55).

β-diversity was predominantly generated by species turnover rather than nestedness, with turnover

forming the larger component in all cases except for small mammals at the plot level in oil palm and

block level in logged forest (Fig. 4). Nestedness formed a larger component of β-diversity for small

mammals compared to large mammals (z = 2.09, p = 0.04). There was a trend for nestedness to be

more important in oil palm (compared to logged forest: z = 1.68, p = 0.093), but no obvious patterns

across spatial grains (χ2(2) = 2.28, p = 0.32).

Figure 4. Percentage of overall β-diversity generated by nestedness (variation in species richness without species composition changes) and species turnover (changes in species composition) across species groups and land-use types, calculated following the approach of Baselga (2010).

3. 4. Discussion

Our finding that the vast majority of old-growth species are retained in logged forest is in agreement

with the emerging consensus, from studies of a large variety of taxa, that logged forest has substantial

conservation value (Dunn, 2004; Berry et al., 2010; Putz et al., 2012; Edwards et al., 2014a). Logging

responses are strongly taxon- and continent-specific (Burivalova et al., 2014), and our study also adds

to a relatively small body of literature on Southeast Asian mammals, supporting the general notion


87

that large areas of logged forest in the region retain much of the terrestrial mammal diversity of old-

growth forest (Kemper & Bell, 1985; Numata et al., 2005; Wells et al., 2007; Bernard et al., 2009;

Kitamura et al., 2010; Brodie et al., 2015), despite timber extraction rates that may be an order of

magnitude higher than on other continents (Putz et al., 2012).

Whilst supporting this general notion, our study also offers a more comprehensive assessment of

mammal community responses to logging than has been possible before. For the first time, we were

able to examine mammal diversity responses at multiple spatial grains, and across the whole terrestrial

mammal community, including both large and small mammals. This revealed a more nuanced view of

community responses to logging: logged habitats had either a higher or lower richness of large

mammals depending on spatial grain, whilst small mammals were richer in logged forest across all

spatial grains. Moreover, large mammal communities became more heterogeneous at increasing

spatial grains in old-growth forest but more homogeneous in logged forest, whilst the reverse pattern

was seen in small mammal communities.

Large mammal richness at small spatial grains was reduced by 19-24% in logged forest, even though

species richness at larger spatial grains was maintained. Similarly, Brodie et al. (2015) found a

reduction in large mammal richness of 11% at the sampling point level in recently-logged (< 10 years)

areas, similar to our logged areas (last logged 3 to 6 years before data collection). Therefore, whilst

logged forests in the region do retain much of the mammal diversity of old-growth forest, logging

may in fact be having subtle but pervasive impacts on the diversity of mammals utilising resources

within any given forest patch, with unknown consequences for ecosystem functioning.

Small mammals, on the other hand, appeared to respond positively to logging, which is consistent

with the broader literature from across the tropics (Isabirye-Basuta & Kasenene, 1987; Lambert et al.,

2006). Small mammals may be resilient to logging due to their apparently high dietary flexibility

(Langham, 1983; Munshi-South et al., 2007) and to the greater availability of their preferred


88

microhabitats post-logging (Cusack et al., 2015). Small mammal communities in old-growth habitats

are also likely constrained by supra-annual cycles of mast-fruiting in dipterocarp forests, in contrast to

more consistent food resources in logged forests (Curran & Leighton, 2000; Brearley et al., 2007;

Munshi-South et al., 2007).

Oil palm mammal communities were highly depauperate for both large and small mammals at all

spatial grains, even when including non-native species and species occurring in plantation margins.

This finding agrees with studies of a range of other taxa (Foster et al., 2011), as well as a small

number of published and unpublished mammal studies (Scott & Gemita, 2004; Maddox et al., 2007;

Rajaratnam et al., 2007; Bernard et al., 2009; Puan et al., 2011), and underlines the grave threat to

wildlife populations that oil palm expansion represents (Koh & Wilcove, 2008; Wilcove et al., 2013).

This is especially the case given that our results likely represent something of a best-case scenario for

oil palm biodiversity: plantations were in close proximity to a large block of well-protected forest,

riparian forest margins existed in the broader landscape and hunting levels were relatively low (only

three incidences of hunting activity were photographed in 3,104 camera trap nights).

We hypothesised that logged forest would be more environmentally heterogeneous than old-growth

forest, giving rise to higher β-diversity (HI). We found that the β-diversity signal was more strongly

evident in logged forests compared to the other land-uses consistently only at the smallest grains,

though small mammal communities showed a stronger signal of β-diversity in logged forest compared

to the other land-uses at more coarse spatial grains as well. This appears to match with the spatial

grain of heterogeneity imparted by the logging process: felling of individual dipterocarp trees usually

creates initial canopy gaps of less than 600 m2 (Sist et al., 2003) and these gaps are mostly less than

10 m in length (i.e. 100 m2) after a decade or more of regeneration (Cannon et al., 1994; Bebber et al.,

2002). In contrast, gaps are rare in old-growth forest, typically occupying less than 1% of forest area

(Sist et al., 2003). Other forms of disturbance – e.g. the creation of skid-trails, roads and log landings

– also impart heterogeneity at a more coarse grain than the felling process, as does variation among


89

logging compartments in the intensity of extraction (Cannon et al., 1994), which may be of an order

of magnitude (Berry et al., 2008). For small mammals, which show strong preferences for specific

microhabitats (Cusack et al., 2015), this latter source of environmental heterogeneity may have driven

the strong signal of β-diversity we observed at larger spatial grains. Note, however, that small

mammal β-diversity at the block level was primarily driven by nestedness rather than turnover in

logged forest, which may suggest that the processes of local extinction and dispersal limitation may

also be important at this scale. For large mammals, communities may not respond as strongly to forest

structure per se, and the greater homogeneity at coarse grains may reflect the greater homogeneity of

tree communities in logged forest at coarse grains, overwhelmingly dominated by a single pioneer

species, Macaranga pearsonii, in this forest.

We also hypothesised that oil palm would be environmentally homogeneous, giving rise to lower β-

diversity (HII). Oil palm communities, overall, were more homogeneous than forest, but this was not

consistently the case: large mammal communities at the block level showed a stronger β-diversity

signal in oil palm compared to logged forest. This was likely due to the substantial differences in

management practices between blocks – for example in the year of planting and the extent of

undergrowth clearance – and, perhaps more crucially, due to differences in the proximity to forest

across blocks. β-diversity in oil palm was also generated comparatively more by nestedness than in

the other land-uses.

Our final hypothesis was that small mammal communities would be more dispersal-limited than large

mammal communities, and would therefore show higher levels of β-diversity (HIII). Support for this

hypothesis was only found at the block level in logged forest and large mammals otherwise showed a

stronger signal of β-diversity. Given the greater dispersal abilities expected of larger bodied mammals

(Sutherland et al., 2000), this does not suggest a primary role for local-scale dispersal limitation in the

assembly of communities in these systems, and niche-based assembly may prevail. We do note,

however, that a greater proportion of β-diversity was driven by nestedness in small mammals than


90

large mammals, especially at the largest spatial grain, which is consistent with the idea of greater

dispersal limitation in this group (albeit not sufficiently high to drive stronger β-diversity patterns in

small mammals overall).

Our findings have implications for the management and conservation of mammal biodiversity at local

scales. In the context of logging, our results point to the importance of spatial heterogeneity,

particularly at fine grains, in maintaining the diversity of mammal communities at similar levels to

old-growth forest. Small mammal diversity may also be increased by heterogeneity in forest structure

at larger spatial grains, but the high levels of nestedness at this scale also suggests that populations

could benefit from interventions to increase connectivity amongst populations. For large mammals,

heterogeneity in forest structure at larger spatial grains was apparently less important, and the

maintenance instead of floristically diverse areas of old-growth forest may have greater benefits for

large mammal diversity. In the context of plantation landscapes, our findings point to the key role that

the maintenance of heterogeneity could play in improving biodiversity values, for example by

deliberately varying the year of planting across coupes within a concession and, more importantly, by

retaining forested areas in the broader landscape.

An understanding of β-diversity patterns is essential for the effective identification of HCV set-aside

in forest landscapes. In Southeast Asia, these forest landscapes are overwhelmingly composed of

logged and degraded forest (Margono et al., 2012; Bryan et al., 2013), and HCV assessments are

made in the context of re-entry logging under sustainable certification or conversion to tree plantation.

Typically ~10% or more of a concession may be considered for set-aside (WWF-Malaysia, 2009), in

patches of approximately 30 ha (Tawatao et al., 2014) or more. Given this, our results suggest that the

specific placement of set-aside for the conservation of large mammal communities, which we have

shown are homogeneous in logged forest at spatial grains < 30 ha, will be less critical and we would

tentatively suggest an approach of maximising the size of set-aside patches. Such patches, even when

isolated from surrounding natural forest, may have considerable value for mammals (McShea et al.,


91

2009; Bernard et al., 2014). For small mammals on the other hand, logged forest communities showed

substantial heterogeneity at the scale of conservation set-aside, which may favour a distributed

network of patches. Although the long-term viability of these meta-populations is largely unknown,

patches would ideally be connected, for example by riparian margins, and positioned according to

robust HCV baseline surveys. Trade-offs in the most effective spatial arrangement of conservation

areas often exist between different species groups (Schwenk & Donovan, 2011), and our findings for

large and small mammals suggest that a diversified strategy including a small number of large patches

and a network of smaller stepping-stone patches would be necessary for the conservation of both

groups. These recommendation for large and small mammals are supported by simulation studies,

albeit of sessile taxa, of randomly-occurring and aggregated species communities undergoing logging,

in which a single large set-aside patch was optimal for maximising yield and biodiversity in the case

of homogeneous communities, but multiple smaller reserves were favoured for aggregated

communities (Potts & Vincent, 2008). We should underline that our results are relevant for set-aside

at the local-scale, for example of a single concession, and a different approach may be necessary at

the regional scale of large forest management units or other administrative regions.

We have shown that diversity responses are strongly grain-dependent and that patterns of β-diversity

at each spatial grain play a fundamental role in this. Better forecasting of local-scale responses to

land-use will require consideration of this grain-dependency. Our data also suggest that management

decisions taken at the local scale, including optimising the spatial arrangement of conservation set-

aside, may be made more effective by considering patterns of β-diversity. Given the increased uptake

of sustainable forestry principles, in particular FSC, in the management of logged forests in the region

(Dennis et al., 2008), as well as rising membership of the RSPO and other crop certification schemes

(Edwards et al., 2012), it is now critical that the scientific underpinnings of HCV are improved, and

this should include consideration of β-diversity at a range of spatial grains.


92

Appendices

Appendix A – Detailed study site descriptions

The Maliau Basin Conservation Area (1,054 km2 including the buffer zone), one of the last remaining

examples of lowland undisturbed habitat in the region (Reynolds et al., 2011), represented our old-

growth control site. The sampled area consisted mostly of pristine hill forest, dominated by

dipterocarps including Shorea johorensis, Dryobalanops lanceolata and Parashorea melaanonan.

One-third of the sampled sites (lying within the buffer zone) were in a water catchment that had been

subjected to low levels of ironwood (Eusideroxylon zwageri) extraction in the 1970s and 1990s; some

old skid-trails were present in the area, though the structure and community composition of the

canopy and understorey were comparable to the surrounding unlogged forest (Ewers et al., 2011).

Part of the Kalabakan Forest Reserve (2,240 km2), the SAFE Project experimental area (94 km2

including a Virgin Jungle Reserve) represented our logged forest site. This was connected to a large

(> 1 million ha) area of logged forest to the north and was otherwise surrounded by oil palm

plantations. Similar to our old-growth forest site, the SAFE Project experimental area was composed

of hill dipterocarp forest, but had been affected by multiple, intense rounds of extraction, beginning in

1978 (Chong et al., 2005). Logging ended as recently as 2008, by which time timber restrictions had

been lifted in anticipation of future clearance, and a total of 179 m3 ha-1 had been removed from the

area (Yayasan Sabah, unpublished data). This land-use history, in combination with topographical

constraints on access, means there was substantial spatial variability in the intensity and timing of

logging, as well as the methods used (tractor-based and cable yarding), creating a highly

heterogeneous forest landscape. There were few old-growth trees remaining, and pioneer species such

as Macaranga pearsonii, M. hypoleuca and Neolamarckia cadamba were dominant. Indeed, M.

pearsonii alone formed ~10% of tree basal area (SAFE Project, unpublished data). In addition,

logging created a network of regenerating skid trails, roads of varying width and heavily degraded


93

log-landing areas, which means there were also some areas of grassland and low scrub, often

containing non-native shrubs including Clidemia hirta and Chromolaena odorata.

Our oil palm sites were spread across two neighbouring plantation estates: Selangan Batu estate

(operated by Benta Wawasan Sendirian Berhad) and Mawang estate (operated by Sabah Softwoods

Sendirian Berhad). Oil palms within the Selangan Batu estate were mostly planted in 2006 and were

1-4 m in height, forming a discontinuous canopy. Oil palms within the Mawang estate were mostly

planted in 2000, with some trees up to 10 m in height and forming a continuous canopy layer. Palms

in both estates were planted with approximately 10 m separation. Understorey communities within the

plantations consisted of various grasses, ferns, other herbs such as Ageratum conyzoides, and vines,

including the highly invasive Mikania cordata (T. Döbert, personal communication). Herbicide use in

the plantations meant that the area directly around each palm was bare in most cases, but there was

otherwise substantial variation in the extent of understorey growth. In the older plantations, where the

canopy was unbroken, much of the ground was bare except for the oil palm fronds which are cut

during harvesting and stacked between the rows of palms. Some small areas within the younger

plantations had been planted with seedlings of subsistence crops, or had been burned in anticipation of

doing so. Small riparian buffers of degraded logged forest existed in the broader landscape, as well as

a 45 km2 block of logged forest (managed by the Sabah Forestry Department) immediately to the west

of the sampling points. Interviews with the estate managers indicated that there were no active rodent

control programmes operating in the plantations (W. Lojinin, personal communication; R. Hussein,

personal communication), with no recent use of rodenticide or biocontrol by barn owls (Tyto alba

javanica).


94

Appendix B - Quantifying β-diversity

β-diversity patterns remain poorly characterised at least in part because of uncertainty surrounding

how to define β-diversity (Tuomisto, 2010) and how best to measure it (Jost, 2007; Baselga, 2010a;

Chao et al., 2012; Legendre & De Cáceres, 2013). β-diversity may be separated into those

components which vary due to sampling effects, including the effects of sampling extent (Soininen et

al., 2007b), grain (Mac Nally & Fleishman, 2004; Steinbauer et al., 2012; Olivier & Aarde, 2014),

replication (Crist & Veech, 2006; Chao et al., 2012) and sample completeness (Cardoso et al., 2009;

Beck et al., 2013), and those components which vary depending on the assembly of communities,

including patterns of species abundance, occupancy, co-occurrence and intraspecific aggregation

(Veech et al., 2003; Veech, 2005). It is these latter components that are typically of interest to

researchers.

In the context of diversity partitioning, there is the additional problem that β-diversity is calculated

using values for α- and γ-diversity, and as a result there has been a recurring debate about whether β-

diversity calculated in this way is truly independent (Jost, 2007, 2010; Baselga, 2010b; Ricotta, 2010;

Veech & Crist, 2010a, 2010b). Chao et al. (2012) recently synthesised this debate, showing

conclusively that neither additive β-diversity (= γ-α) nor multiplicative β-diversity (= γ/α) are free of

this dependence, and recommended a normalisation to overcome this. However, it remains unclear

whether sampling effects on α- and γ-diversity are completely controlled for using this normalisation.

The effects of the sampling process, as well as the size of regional species pools (Lessard et al., 2012),

can be accounted for by comparing observations with a null model which specifically includes these

details (Crist et al., 2003; Kraft et al., 2011). Any differences between observations and the null model

which remain are taken to be indicative of non-random processes that were excluded from the null

model, for example community assembly processes. This is the approach we chose to use here (see

the Methods for more details).


95

β-diversity sensu lato includes community variance due both to the turnover of species and due to

variation in species richness independent of turnover, i.e. the nestedness of communities. Baselga

(2007, 2010a) and others (for a review see Legendre, 2014) have argued for a separation of turnover

and nestedness and for β-diversity sensu stricto to be measured independently of the effects of

nestedness. The predominance of turnover or nestedness in communities is related to the assembly

processes at work. For example, niche assembly and random community drift will often be

responsible for patterns of turnover at local scales, whilst differential dispersal capacities and selective

extinction are more likely to create nested communities across space. The distinction between

turnover and nestedness is particularly important in the context of conservation set-aside; if β-

diversity is driven by species turnover, a distributed network of set-aside patches would be required to

ensure representation of all species, whilst if β-diversity is completely driven by nestedness patterns,

the optimal solution would simply be to prioritise the conservation of the most diverse forest patches

(most often, these will be areas with highest habitat quality for the given taxonomic group overall).

Following the approach of Baselga (2010a), we therefore separated observed β-diversity into its

turnover and nestedness components (more details on this approach are given in the Methods).


96

Appendix C – Supplementary results

Figure C1. Sample-based species accumulation curves across land-use types, based on a Bernoulli product model (Colwell et al., 2012). Only data from sampling points which were both camera-trapped and live-trapped were used. Solid lines show interpolated values, whilst dashed lines show extrapolated values. Filled circles show the observed γ-diversities.


97

Table C1. Observed and estimated γ-diversity across land-use types. Species group Land-use Dataset γobserved γACE

a 95% CI Mammals Old-growth forest Locations both camera- and live-trapped 42 46.7 42.0 - 52.7

Logged forest Locations both camera- and live-trapped 51 52.4 51.0 - 59.1

Oil palm plantation Locations both camera- and live-trapped 21 33.9 28.2 - 39.5

Oil palm cropb Locations both camera- and live-trapped within oil palm crop 18 27.1 22.2 - 32.0

Large mammals Old-growth forest All camera trap locations 30 31.8 30.0 - 36.8

Logged forest All camera trap locations 32 33.0 32.0 - 38.2

Oil palm plantation All camera trap locations 18 27.8 22.6 - 32.9

Oil palm cropb All camera trap locations within oil palm crop 17 28.7 23.6 - 33.9

Old-growth forest Locations both camera- and live-trapped 30 31.8 30.0 - 36.8




Small mammals Old-growth forest All live trap locations 11 13.2 11.0 - 16.6

Logged forest All live trap locations 21 21.5 21.0 - 25.5

Oil palm plantation All live trap locations 4 4.8 4.0 - 6.8

Oil palm cropb All live trap locations within oil palm crop 2c

Old-growth forest Locations both camera- and live-trapped 12 16.7 13.0 - 20.3




aEstimates of minimum asymptotic richness calculated using the Abundance-based Coverage Estimator (ACE). See Gotelli & Chao (2013). bExcluding sampling locations within scrub habitat at the oil palm plantation margins. cInsufficient data to estimate asymptotic minimum richness.


98

Table C2. Observed α-diversity and model estimates of standardised α-diversity across land-use types. Species group Land-use Dataset αobserved

a αstandardisedb 95% CI

Mammals Old-growth forest Locations both camera- and live-trapped 7.21 7.60 6.53 - 8.83

Logged forest Locations both camera- and live-trapped 8.19 9.29 7.68 - 11.24

Oil palm plantation Locations both camera- and live-trapped 3.29 3.19 2.42 - 4.21

Oil palm cropc Locations both camera- and live-trapped within oil palm crop 3.22 3.16 2.39 - 4.17

Large mammals Old-growth forest All camera trap locations 5.99 6.07 4.94 - 7.45

Logged forest All camera trap locations 4.94 4.91 3.75 - 6.42

Oil palm plantation All camera trap locations 3.00 2.91 2.12 - 3.98

Oil palm cropc All camera trap locations within oil palm crop 2.97 2.91 2.12 - 3.99

Old-growth forest Locations both camera- and live-trapped 5.99 5.95 5.04 - 7.02




Small mammals Old-growth forest All live trap locations 0.53 0.45 0.30 - 0.69

Logged forest All live trap locations 2.66 2.72 1.62 - 4.57

Oil palm plantation All live trap locations 0.10 0.08 0.04 - 0.16

Oil palm cropc All live trap locations within oil palm crop 0.10 0.07 0.04 - 0.14

Old-growth forest Locations both camera- and live-trapped 1.21 1.14 0.95 - 1.36




aMean observed α-diversity across sampling points. bStandardised to 90% sample coverage (Chao & Jost, 2012). Estimates are from mixed-effects models and include shrinkage. cExcluding sampling locations within scrub habitat at the oil palm plantation margins.


99

Table C3. β-diversity differences from null models across land-use types and spatial grains.

Species group Land-use Spatial grain β-diversity difference

from null 95% CIa

Mammals Old-growth forest βpoint 1.54 1.40 - 1.69

βplot 1.64 1.06 - 2.49

βblock 1.41 0.49 - 2.82

Logged forest βpoint 2.26 2.12 - 2.41

βplot 1.45 0.76 - 2.08

βblock 1.75 0.50 - 2.84

Oil palm βpoint 0.82 0.64 - 0.99

βplot 0.60 0.20 - 1.04

βblock 0.50 -0.01 - 0.99

Large mammals Old-growth forest βpoint 0.67 0.60 - 0.75

βplot 0.58 0.16 - 1.05

βblock -0.17 -0.67 - 0.33


βplot 0.63 0.26 - 1.01

βblock 1.40 0.48 - 2.14

Oil palm βpoint -0.01 -0.01 - 0.01

βplot -0.06 -0.11 - 0.05

βblock 0.00b 0.00 - 0.00

Small mammals Old-growth forest βpoint 0.87 0.75 - 1.00

βplot 1.05 0.66 - 1.55

βblock 1.51 0.68 - 3.01


βplot 0.78 0.32 - 1.32

βblock 0.34 -0.65 - 1.35

Oil palm βpoint 0.83 0.68 - 0.99

βplot 0.65 0.12 - 1.12

βblock 0.52 0.04 - 1.04

aConfidence intervals calculated as the quantiles of the distribution of difference from null values across simulations. bAll null simulations returned the same β-diversity as the observed case, likely caused by a sparse dataset in this case.


100

Appendix D – Mammal species list

Table D1. Order, scientific name and common name of mammal species detected by camera-trapping and live-trapping surveys in the Maliau Basin Conservation Area (including the Maliau Basin Buffer Zone Forest Reserve), Kalabakan Forest Reserve and nearby oil palm plantations (Selangan Batu and Mawang estates). Species Common name Old-growth forest Logged forest Oil palm Erinaceomorpha

Echinosorex gymnura Moon rat Pholidota Manis javanica Sunda pangolin Carnivora Prionailurus bengalensis Leopard cat Pardofelis badia Bay cat Pardofelis marmorata Marbled cat Neofelis diardi Sunda clouded leopard Diplogale hosei Hose's civet Hemigalus derbyanus Banded civet Paguma larvata Masked palm civet Paradoxurus

h h di Common palm civet

Arctictis binturong Binturong Viverra tangalunga Malay civet Prionodon linsang Banded linsang Herpestes semitorquatus Collared mongoose Herpestes brachyurus Short-tailed mongoose Canis familiaris Domestic dog Helarctos malayanus Sun bear * Mydaus javanensis Sunda stink badger Martes flavigula Yellow-throated marten Mustela nudipes Malay weasel Aonyx cinereus Oriental small-clawed otter Cetartiodactyla Sus barbatus Bearded pig Tragulus napu Greater mouse-deer Tragulus kanchil Lesser mouse-deer Muntiacus atherodes Bornean yellow muntjac Muntiacus muntjak Red muntjac Rusa unicolor Sambar deer Bos javanicus Banteng Scandentia Ptilocercus lowii Pen-tailed treeshrew Tupaia minor Lesser treeshrew Tupaia gracilis Slender treeshrew Tupaia longipes Plain treeshrew Tupaia tana Large treeshrew * Tupaia dorsalis Striped treeshrew Primates Cephalopachus bancanus Western tarsier Presbytis rubicunda Maroon langur Macaca fascicularis Long-tailed macaque Macaca nemestrina Southern pig-tailed macaque Hylobates muelleri Bornean gibbon Pongo pygmaeus Orangutan Rodentia Lariscus hosei Four-striped ground squirrel Callosciurus prevostii Prevost's squirrel Callosciurus notatus Plantain squirrel * Callosciurus adamsi Ear-spot squirrel Exilisciurus exilis Least pygmy squirrel Sundasciurus lowi Low's squirrel


101

Sundasciurus tenuis Slender squirrel Sundasciurus brookei Brooke's squirrel Sundasciurus hippurus Horse-tailed squirrel Rheithrosciurus macrotis Tufted ground squirrel Aeromys thomasi Thomas's flying squirrel Trichys fasciculata Long-tailed porcupine Hystrix brachyura Malay porcupine

Hystrix crassispinis Thick-spined porcupine Leopoldamys sabanus Long-tailed giant rat Sundamys muelleri Müller's rat Niviventer cremoriventer Dark-tailed tree rat Maxomys whiteheadi Whitehead's rat

Maxomys surifer Red spiny rat * Maxomys rajah Brown spiny rat Maxomys baeodon Small spiny rat Maxomys ochraceiventer Chestnut-bellied spiny rat Rattus exulans Polynesian rat Rattus rattus Black rat Proboscidea Elephas maximus Asian elephant *Records from scrub habitat at the oil palm plantation margins and not inside the oil palm crop itself.

Chapter 4: Drivers of mammalian community assembly across a tropical land-use gradient

102

Chapter 4:

Anthropogenic land-use change alters the ecological processes assembling tropical

rainforest mammal communities

Abstract

Relatively little is known about how drivers of community assembly change along environmental

gradients, including gradients which are due to anthropogenic drivers such as land-use change. In

addition, most previous research in this area has focussed on sessile taxa and has neglected groups of

high conservation concern, such as the vertebrates. Using concurrent camera- and live-trapping, we

investigated the local-scale processes assembling mammal communities along an environmental

gradient representing the principal trajectory of land-use in Borneo, including old-growth forest,

logged forest and oil palm plantations. We found that communities were assembled by a mixture of

niche and spatial processes – likely including dispersal limitation and, at fine scales, the home-ranges

of individuals – as well as community drift. The relative balance of these processes was not stable

across the land-use gradient, supporting the idea of a niche-neutrality continuum. Old-growth forest

communities were strongly spatially-structured, with important roles played by dispersal and broad-

scale environmental control, including coarse-grain variations in forest structure and topography. In

contrast, anthropogenic environments were not strongly spatially-structured, with more important

roles played by drift and logging-induced microhabitat variation in logged forest, and fine-scale

environmental control due to management practices in oil palm. Analysis of species co-occurrences

indicated that interactions, including competition, never exerted a strong effect on communities, but

what effect there was weakened along the land-use gradient. This suggests that environmental control

was predominantly due to the filtering of species by their environmental niche. We conclude that

anthropogenic disturbance may lead to novel mechanics governing the local assembly of

communities. An understanding of these mechanics for a broad range of organisms – both sessile and


103

motile – and ecosystems will be essential for testing and refining assembly theory, as well as for the

more effective conservation and management of highly-threatened natural systems.

4. 1. Introduction

The turn of this century saw the beginning of a resurgence of interest in community assembly in

ecology (HilleRisLambers et al., 2012). This was driven by new conceptual and mathematical models

of the mechanisms of assembly, including better integration of scale (Leibold et al., 2004) and

dispersal (Chave et al., 2002; Tilman, 2004). Perhaps most importantly, the emergence of neutral

theory (Bell, 2000; Hubbell, 2001) has brought the principle of parsimony in community ecology to

the fore and has challenged the discipline’s propensity for model proliferation rather than integration

(Lawton, 1999; Vellend, 2010). In addition, the last decade has seen the development of new

analytical tools to investigate the mechanisms of community assembly (Ovaskainen et al., 2010;

Wiens et al., 2010; Chase & Myers, 2011; Dray et al., 2012; Gotelli & Ulrich, 2012), as well as new

technology to collect community-wide biodiversity data in higher spatial and temporal resolutions

than has been possible before, even in the most challenging environments (Rowcliffe & Carbone,

2008; Turner, 2014).

Despite these developments in community assembly theory and data generation, there have been

relatively few empirical investigations of community assembly mechanisms to date, particularly for

non-sessile organisms. It is essential that new theoretical developments in community assembly are

confronted with empirical data from a wide range of organisms and ecosystems in order to inform the

next generation of models. Importantly, natural habitats across the globe, and in particular in the

tropics, are being subject to unprecedented rates of clearance and disturbance (Asner et al., 2009) and

it is critical that the anthropogenic component to community assembly is better understood and

integrated into models. This will require investigating community assembly along anthropogenic

environmental gradients. In this context, land-use change can be thought of as a large-scale natural

“experiment” which can reveal insights into community assembly processes along environmental

gradients which would otherwise be difficult to obtain (Lomolino & Perault, 2000). In addition to


104

theoretical insights, this will lead to better forecasts of the impacts of land-use change (Matias et al.,

2014) and more effective recommendations to policy-makers, industry and other stakeholders in order

to conserve biodiversity.

Two contrasting processes are generally thought to be dominant in community assembly at the local

scale: niche assembly and dispersal assembly. Niche assembly mechanisms have a long pedigree in

ecology (Grinnell, 1917; Elton, 1927; Hutchinson, 1957; Hardin, 1960; MacArthur & Levins, 1967)

and involve selection of species according to their fundamental environmental niche (the “abiotic

filter”), as well as small-scale interactions with competitors, mutualists and consumers (the “biotic

filter”). Niche assembly is generally viewed as deterministic, although stochasticity can play a role in

the outcome of niche assembly, as exemplified by priority effects and the effects of environmental

stochasticity (Weiher et al., 2011). Dispersal assembly refers to the stochastic assembly of a local

community by dispersal, i.e. by the movement of organisms across space (Vellend, 2010). Although

dispersal was recognised in some of the earliest models of community assembly, most notably in

island biogeography theory (Macarthur & Wilson, 1967), it has since become associated with the

unified neutral theory (Hubbell, 2001), in which dispersal is conceptually from the metacommunity

(Leibold et al., 2004). Contemporary assembly theory recognises that niche and dispersal assembly

processes are not mutually exclusive and that both may operate concurrently (Gravel et al., 2006;

Mutshinda & O’Hara, 2011). The challenge, therefore, is to identify the relative importance of these

two assembly mechanisms and how the balance might change in the face of anthropogenic drivers.

An important advance in this regard has been the development of analytical tools for specifically

dissecting the site-to-site variation in community composition, or β-diversity, as a function of

environmental and spatial components, using canonical ordination (Borcard et al., 1992; Borcard &

Legendre, 2002; Legendre et al., 2005; Peres-Neto et al., 2006). In this framework, the importance of

niche assembly can be inferred from the community variance explained in the ordination by the

environmental control component. In order to infer the importance of spatial processes, including

dispersal assembly, surrogate spatial variables representing positive spatial correlation can be used


105

during ordination. In this case, the variance explained will partly be due to spatially-structured

environmental variables, causing induced spatial dependence (Peres-Neto & Legendre, 2010), and it is

therefore necessary to factor out the environmental control component using variation partitioning.

The pure spatial component that remains, assuming all the important spatially-structured

environmental variables have been accounted for, represents the relative importance of contagious

spatial processes such as dispersal, as well as the home-ranging of individuals and positive biotic

interactions at small scales. Variation left unexplained by environmental control and space is likely

due to random community drift, as well as any unmeasured environmental variables (which are not

structured in space) and measurement error. In common with all other spatial modelling approaches,

this analytical method is not a direct test of causal processes. Nonetheless, it can lead to strong

mechanistic inferences when combined with clear a priori hypotheses and sound ecological

knowledge of the system under study (McIntire & Fajardo, 2009; Dray et al., 2012).

Niche assembly is generally characterised as involving both environmental filtering and species

interactions. In addition, the relative fitness differences between interacting species will often be a

function of the environment, potentially leading to complicated feed-backs between the two

components of niche assembly (Chesson, 2000; HilleRisLambers et al., 2012). Although species

interactions are generally thought to operate at smaller spatial scales than environmental filtering

(Soberón, 2007), the two components to niche assembly are otherwise difficult to separate on the

basis of spatial modelling alone. However, different classes of species interactions leave differing

signatures of co-occurrence in space and time and can, in combination with knowledge about the

strength of environmental control, allow for the testing of a priori hypotheses. In particular, ongoing

competition will often leave a signature of negative co-occurrence relationships between pairs of

species due to compensatory dynamics; a lack of negative co-occurrence relationships, despite

evidence for niche assembly in general, may therefore suggest a primacy of environmental filtering

over competition. Indeed, this has been the basis for a number of recent such assertions, based either

on species co-occurrence (Veech, 2006) or covariance (Houlahan et al., 2007; Volkov et al., 2009;

Mutshinda et al., 2009). Weak interspecific competition may increase the long-term stability of such


106

communities (Chesson, 2000), but a predominance of environmental filtering may increase the

sensitivity of communities to environmental change (Mutshinda et al., 2009).

Land-use change represents the principal threat to biodiversity in the tropics (Laurance et al., 2014b),

but we know very little about how assembly processes might change across environmental gradients

in general, let alone newly-emerging anthropogenic gradients. There is evidence that patterns of β-

diversity are altered by selective logging and conversion of forest to plantations (Berry et al., 2008;

Woodcock et al., 2011; Kitching et al., 2013; Chapter 3), and one explanation for these patterns is an

alteration in the balance of the underlying drivers of assembly. Logging greatly increases the

frequency and area of gap habitat in forests and may lead to increases in environmental heterogeneity

at the small scales at which individual trees are felled (Cannon et al., 1994). In parallel, logging may

lead to greater floristic homogeneity at larger scales, for example between logging coupes, with

disturbed areas typically becoming dominated by just a few pioneer species (Heydon & Bulloh, 1997;

Brearley et al., 2004). These changes wrought by logging may act as a strong environmental filter on

the composition of communities. In addition, it might be expected that the strength of species

interactions would be affected by disturbance (Tylianakis et al., 2008). For mammals, it has

previously been shown that diversity at small scales is significantly reduced by the intensive logging

typical in Southeast Asia (Brodie et al., 2015; Chapter 3). This may mean communities at small scales

are further from saturation than in old-growth forest, leading to weaker compensatory dynamics. This

dominance of environmental filtering over interactions may be more pronounced still in plantation

habitats, which drastically differ from pristine forest – in terms of structure, species composition,

resources and microclimate – and typically only host a depauperate subset of species (Aratrakorn et

al., 2006; Styring et al., 2011; Foster et al., 2011; Chapter 3).

Here we investigate the processes assembling mammal communities along an environmental gradient

of land-use change in the tropical forests of Borneo, including primary forest, logged forest and oil

palm (Elaeis guineensis) plantations. This land-use gradient represents the principal trajectory of land-

use change in Southeast Asia (Wilcove et al., 2013), with the majority of remaining forest in the


107

region now existing in a logged and degraded state (Gaveau et al., 2014) and conversion to oil palm

likely continuing at a rapid rate over the coming decades (Koh & Ghazoul, 2010a). We focussed on

the mammals owing to their highly threatened status in Southeast Asia (Sodhi et al., 2009b) and the

strong weighting they are often given in policy decisions surrounding land-use. Mammals are a

cryptic and mostly nocturnal group which occur at low densities and actively avoid human observers,

often resulting in datasets too sparse to investigate community assembly processes. To circumvent

these problems and generate robust community data, we used concurrent networks of camera traps

and live traps to sample almost the entire terrestrial mammal community, and expended much higher

levels of sampling effort than would be typical for other taxonomic groups. We had three competing

hypotheses about mammalian community assembly processes across the land-use gradient. The null

hypothesis was that mammal communities would mostly be assembled by neutral processes, including

dispersal and random drift, giving rise to a non-significant environmental control model and, owing to

the importance of drift, a low explanatory power of the overall environment-space model. The first

alternative hypothesis was that both niche and dispersal assembly would be important across all land-

uses, reflected in important portions of the community variance explained by environmental and

spatial variables, and that environmental filtering would be the dominant niche assembly process,

resulting in only weak evidence of competitive interactions in the form of negative co-occurrence

patterns. Our final competing hypothesis was that environmental filtering would become increasingly

more important than both dispersal and competitive interactions along the land-use gradient, resulting

in a relative increase in the variance explained by environmental variables and reduced evidence of

negative co-occurrence patterns.

4. 2. Methods

4. 2. 1. Sampling design



2011). This consists of old-growth forest within the Maliau Basin Conservation Area, repeatedly-

logged forest within the Kalabakan Forest Reserve and two adjacent oil palm plantations straddling


108

the Kalabakan Forest Reserve boundary (see Chapter 3 for more detailed descriptions of the study

sites).

We employed a clustered hierarchical sampling design, with individual sampling points clustered

together into 1.75 ha plots, and 3 to 6 plots in turn clustered into blocks (Fig. 1). This multi-scale

approach allowed for the investigation of fine-scale assembly processes, such as competition between

species, whilst also allowing for the investigation of larger-scale gradients in community composition

within a study site. Plots consisted of (4 x 12 =) 48 potential sampling points, separated by 23 m, of

which a random subset were chosen for sampling (mean number of points per plot = 14). Plots were

arranged together into blocks differently in the logged forest compared to the other two land-uses

(Fig. 1), so that they would overlay the locations of future experimental fragments in this study site

(Ewers et al., 2011). Separation distances between plots (170 to 290 m) and between blocks (0.6 to 3

km) were similar across the land-uses. The SAFE Project has been deliberately designed to minimise

confounding factors across the land-use gradient, including latitude, slope and elevation (Ewers et al.,

2011), and this applied equally to our sampling design for mammals.

Across the study sites, 430 points were sampled using both camera-trapping and live-trapping. These

were nested within 31 plots and 8 blocks (9 plots in 3 blocks for old-growth forest; 16 plots in 3

blocks for logged forest, and 6 plots in 2 blocks for oil palm). We excluded 12 points which had been

camera-trapped for less than seven days, giving a total sampling effort of 9,430 live trap nights and

19,116 camera trap nights (after correcting for camera failures). The sampling intensity was similar

across land-uses (mean trap nights per sampling point: 60 in old-growth forest; 78 in logged forest,

and 58 in oil palm).


109

Figure 1. Sampling design across logged forest (A), oil palm (B) and old-growth forest (C), showing the sampling points (in red) sampled using both camera traps and live traps. Clusters of sampling plots, i.e. sampling blocks, were arranged in the same spatial configuration in old-growth forest and oil palm, and were arranged to coincide with future experimental forest fragments in logged forest. Separation between points, plots and blocks was nonetheless similar across land-uses. Shaded areas lie outside the Kalabakan Forest Reserve, consisting of a 2,200 ha Virgin Jungle Reserve (Brantian-Tatulit) to the south and an extensive (>1 million ha) area of logged forest to the north (Mount Louisa Forest Reserve and other connecting reserves). Insets show the location within insular Southeast Asia and the spatial proximity of the three land-uses within south-east Sabah, Malaysian Borneo.

4. 2. 2. Field methods

We deployed camera traps (Reconyx HC500, Holmen, Wisconsin, USA) strictly within 5 m of each

randomly chosen sampling point, aiming to get as close as possible to each point (marked in the field

on a previous occasion), whilst avoiding major obstructions in the camera’s field of view. The random

deployment of cameras in this manner has rarely been used before, though it is essential for reducing

biases in species detection rates caused by non-random space-use by animals (Chapter 2). No bait or

lure was used, cameras used an infrared flash during night-time hours, and disturbance to vegetation

was kept to a minimum. Camera traps are continuous-time detectors, so we considered photographic

capture events to be independent if they a) contained different individuals or b) were separated by >

12 hours, which matched the approximate minimum separation between live trap events. Camera traps


110

were active between May 2011 and April 2014, during which most plots (28 of 31) were sampled

multiple times (mean effort per plot = 617 trap nights).

Small mammal trapping was conducted at the plot level, with two locally-made steel-mesh traps (18

cm wide, 10-13 cm tall and 28 cm in length) placed at or near ground level (0 - 1.5 m) within 10 m of

each of the 48 sampling points and baited with oil palm fruit. Here, however, we only use data from

the 418 points which were also sufficiently sampled using camera traps. Each session consisted of

seven consecutive trapping days and some plots (15 of 31) were sampled for multiple sessions over

the course of the study (mean effort per plot for the subset of points used in this analysis = 304 trap

nights). Traps were checked each morning and captured individuals were anaesthetised using diethyl

ether (following Wells et al. 2007), measured, permanently marked using a subcutaneous passive

inductive transponder tag (Francis Scientific Instruments, Cambridge, UK), identified to species using

Payne et al. (2007) and released at the capture location. Although we marked individuals, we here

used the number of detections of each species, since this better matches the protocol used for camera-

trapping. Trapping was carried out between May 2011 and July 2014, during which there were no

major mast-fruiting events.


We measured a range of environmental variables for each sampling point, which fell into three

distinct sets: 1) fine-scale habitat structure variables; 2) topographical variables, and 3) broader-scale

characteristics of the local landscape, including percent forest cover and a remotely-sensed measure of

above-ground live tree biomass within a 0.5 km radius (see Appendix A for full details of each

environmental variable). We expected these sets of variables to be inter-correlated and were interested

in the relative importance of each of these sets of factors, both independently and together, in the

overall environmental control model. This involved creating parsimonious redundancy analysis

(RDA) sub-models for each set of variables and then applying variation partitioning to the composite

model Y = f[E] + R = f[H + T + L] + R, where Y is the community response matrix, E is the


111

environmental component, R is a matrix of residuals, and H, T and L are, respectively, the habitat

structure, topography and landscape context variables.

Rather than use all available variables in each RDA sub-model, we applied the modified forward-

selection method of Blanchet et al. (2008) to select a parsimonious set of variables. This method uses

an unbiased estimate of explained variation, the adjusted coefficient of multiple determination (R2adj).

Inflated Type I error rates were prevented by applying a permutation test of the significance of the

saturated RDA model before forward-selecting in each case (Blanchet et al., 2008). In addition to

creating a set of three environmental RDA sub-models for the dataset as a whole, we also repeated

these analysis steps for each step along the land-use gradient and for large and small mammals

separately.

We constructed two sets of surrogate spatial variables, in order to model both fine-scale and broad-

scale spatial processes. Fine-scale positive spatial correlations were modelled using Moran’s

eigenvector maps (MEMs; Borcard & Legendre, 2002; Dray et al., 2006) based on the geographic

distances among sampling points (i.e. distance-based MEMs, or db-MEMs). We selected the db-

MEMs with positive eigenvalues, which model positive spatial correlations. MEMs represent a

spectral decomposition of the distances among points, for example sine-waves of different periods for

the specific case of regular sampling along a transect, and can potentially model spatial processes at

all scales perceivable in the sampling design (Borcard et al., 2004). The finest scale that can be

modelled is determined by the minimum distance connecting all points to their nearest neighbour. For

clustered sampling designs with large minimum distances, it is possible to create separate db-MEM

variables for each cluster and then assemble all variables into a single staggered matrix (Declerck et

al., 2011). We therefore created separate db-MEM variables for each block in our sampling design,

and the largest spatial processes modelled using these variables are at the block scale. We also

augmented the set of geographic coordinates for each block with a small number of supplementary

points (between 5 and 10 per block; 12% of the total) to fill in the widest gaps between plots, which

were then removed before RDA modelling (Borcard & Legendre, 2002; Borcard et al., 2004). This


112

causes a slight loss of orthogonality between db-MEM variables, but allowed for the modelling of

fine-scale spatial processes, giving minimum distances of between 67 and 76 m.

Broader-scale spatial processes, such as those between blocks, were modelled using surrogate trend-

surfaces (Borcard et al., 1992). For each land-use, orthogonal trend-surface variables representing

linear gradients and saddle-shaped responses were created using 1st- and 2nd-order polynomial

functions of the geographic coordinates. As with the db-MEM variables, we assembled the trend-

surface variables for each land-use into a single staggered matrix for RDA modelling.

We treated the two sets of spatial variables – db-MEM and trend-surface variables – in the same way

as our sets of environmental variables, by submitting each to the modified forward selection method

to create parsimonious RDA sub-models. The selected variables were then entered into the model Y =

f[S] + R = f[F + B] + R, where S is the spatial component that is subjected to variation partitioning, to

calculate the independent and shared portions of the variation explained by the fine-scale (F) and

broad-scale (B) spatial processes.

To make inferences about the relative importance, overall, of environmental control and space in

assembling mammal communities, we defined a global environmental control RDA model and a

global spatial RDA model for input into variation partitioning. These global models contained the

variables selected in the respective sub-models, such that Y = f[E + S] + R = f[(H + T + L) + (F + B)]

+ R. For the environmental control model, we also tested if adding land-use as a categorical variable

(LU) explained a significant additional portion of the variance not captured by the selected

environmental variables. Although land-use explained just 1.8% of the variance independently,

suggesting that our measured environmental variables successfully captured the environmental

gradient, we found that this fraction was nonetheless significant using a permutation test (F(2, 397) =

5.82, p < 0.001) and therefore added land-use to the global environmental control model, such that Y

= f[(E + LU) + S] + R = f[((H + T + L) + LU) + S] + R. We also created environmental control and

spatial RDA models for each land-use and for large (> 1 kg body mass) and small mammals


113

separately, and subjected each of these to variation partitioning. This allowed us to explore the

relative importance of environment and space in each land-use and for the two species groups. We

tested for significant differences in the variation explained overall, explained by environment and

explained by space in each case, using a bootstrap procedure (Peres-Neto et al., 2006).

RDA models in each case were fitted to Hellinger-transformed community matrices, with mammal

detections per seven days summed over the camera- and live-trapping protocols. Raw species data is

inappropriate for use in RDA, and the commonly-used Hellinger transformation, which divides by the

total abundance at a site and then takes the square-root (therefore dampening the effect of extremely

abundant species), has previously been shown to have desirable properties in the context of RDA

(Legendre & Gallagher, 2001). Detection probabilities likely vary across species and across the two

protocols, but we do not expect that this will substantially affect the relative sizes of the variance

fractions explained by environment and space, although imperfect detection could inflate the

unexplained variance in all cases. The Hellinger transformation is asymmetrical, meaning that species

absences, which could be a result of non-detection rather than lack of presence, have a lower

influence on the coefficient than presences. This transformation also does not give undue weight to

rare species (Legendre & Gallagher, 2001). Our use of random sampling locations controls for the

biases due to trap placement which are common in camera-trapping studies (Chapter 2).

We investigated co-occurrence patterns among species within each land-use using new probabilistic

models (Veech, 2013), which avoid the need for much-debated data randomisation algorithms and

comparison with null distributions (Gotelli & Ulrich, 2012). Instead, the probabilistic approach uses

the hypergeometric distribution to calculate the probability that two species co-occur either less or

more often than expected based on their mean incidence probabilities (Griffith et al., 2014). Using the

observed co-occurrence frequencies and a specified alpha level (in this case, α = 0.05), species co-

occurrences were then classified as significantly positive or negative, or occurring at random. Random

co-occurrence patterns can potentially be generated both by genuine non-association between species

or by a lack of statistical power. We had a relatively large number of sites within each land-use (56 to


114

213) and high sampling effort, with simulations suggesting that in this case models would typically

have very high power to detect any deviations from random co-occurrence of approximately > 5% of

the total number of sites. Following Veech (2013), we did not analyse species pairs with expected co-

occurrence frequencies < 1. We did not divide the mammal community into guilds of putatively

interacting species, in part because there was scant prior information available on the interactions

among Bornean mammals, but also because we were interested in a community-wide assessment of

the co-occurrence patterns across land-uses.

All analyses were done in R version 3.1.0 (R Core Team 2014), using the additional packages vegan

2.0-10 (Oksanen et al., 2013), PCNM 2.1-2 (Legendre et al., 2013) and cooccur 1.0 (Griffith et al.,

2014).

4. 3. Results

We obtained 1,237 live captures of 20 species and 10,464 photo-captures of 56 species over the

locations sampled using both trapping methods. Fifteen species were captured using both methods,

giving a total of 61 mammal species detected. Occurrence rates for individual species ranged from 77

% of sampling points for the red muntjac (Muntiacus muntjak) to just one point for five species (mean

occurrence = 12%). Overall trapping rates were similar for old-growth and logged forest (0.47 and

0.38 captures per trap night, respectively), but lower in oil palm (0.19).

Environmental and spatial variables together explained 33% of the community variance in the overall

RDA model (Fig. 2). The two sets of variables explained a similar portion of the variance (95% CI of

the difference: -0.03 – 0.03, p = 0.776), though a large portion was shared between them (49% of the

total explained variation).


115

Figure 2. Overall variation partitioning of the mammal community composition data, separated into variation explained by a) local habitat characteristics only, and local habitat in combination with land-use, b) spatial processes (represented by spatial surrogate variables) and c) environmental control (local habitat and land-use) and spatial processes together. Percentage values represent the adjusted coefficient of multiple determination (R2

adj) calculated using redundancy analyses. Values lying outside the area of the Euler diagrams represent the percentage variation left unexplained in each case.

Community variation in absolute terms was broadly similar across the land-uses (Var[YOld-growth] =

0.57; Var[YLogged] = 0.58; Var[YOil palm ] = 0.53). In old-growth forest, space explained significantly

more of the variation than environment (95% CI of the difference: 0.05 – 0.14, p < 0.001), with 90%

of the variation explained by environment being spatially-structured (Fig. 3). Only a small, albeit

significant, fraction of the variation was explained by pure environmental control (F(12, 112) = 1.34, p <

0.01). In logged forest, just 14% (95% CI: 10 – 19%) of the total variance was explained, much lower

than for old-growth forest (33%, 95% CI: 27 – 39%) and oil palm (30%, 95% CI: 15 – 46%).

However, the independent environmental and spatial components were still significant (environment:

F(10, 179) = 1.90, p < 0.001; space: F(16, 179) = 1.91, p < 0.001). The variance explained by environmental

and spatial variables was not significantly different for logged forest communities (95% CI of the

difference: -0.06 – 0.01, p = 0.17). In oil palm, 86% of the variation was explained by environmental


116

control (79% independently) and the spatial component was significantly smaller (95% CI of the

difference: 0.02 – 0.30, p = 0.02).

Figure 3. Variation partitioning of mammal community composition data across a tropical land-use gradient. Community variation was partitioned using redundancy analyses (RDA) according to three sets of environmental control variables, broad- and fine-scale spatial processes, and environmental control and space overall. The environmental and spatial variables were chosen separately for each land-use, using a modified forward selection procedure. Percentage values represent the adjusted coefficient of multiple determination (R2

adj) and values lying outside the area of the Euler diagrams represent the percentage variation left unexplained in each case. The landscape context RDA for oil palm could not be represented in full using a Euler diagram, and a small fraction (1.2%) shared between habitat structure and topography was omitted in order to allow for plotting.

Large mammals were responsible for 70% of the overall community variance (Var[Ylarge] = 0.48;

Var[Ysmall] = 0.21) and, as a result, the environment-space variation partitioning for large mammals

was similar to the overall case, with similar fractions of the variation explained by environment and

space (95% CI of the difference: -0.05 – 0.02, p = 0.34) and a large overlapping fraction (Appendix B,

Fig. B1). For small mammals, however, environment explained much more of the community


117

variation than space (95% CI of the difference: 0.06 – 0.17, p < 0.001), and the majority of this (69%)

was not spatially-structured (Appendix B, Fig. B1).

The measured environmental variables successfully captured the environmental gradient across the

land-uses, with environmental and community separation between the land-uses evident in ordination

space (Fig. 4). The relative importance of each set of environmental variables for structuring

communities within land-uses differed markedly (Fig. 3). Fine-scale habitat structure was an

important independent component in all land-uses (old-growth forest: F(6, 143) = 2.35, p < 0.001;

logged forest: F(7, 195) = 2.83, p < 0.001; oil palm: F(6, 45) = 3.08, p < 0.001). Habitat in the broader

landscape was also important in old-growth forest, both in combination with fine-scale habitat

structure and also independently (F(6, 143) = 2.35, p < 0.001), but was only important in oil palm in

combination with topography (F(4, 45) = 1.72, p = 0.01) and not independently (F(3, 45) = 1.29, p = 0.16).

Landscape context was found to be entirely unimportant for logged forest communities (F(1, 204) =

1.25, p = 0.31). Topography was important mostly in combination with other variables, but was also

important in its own right in old-growth forest (F(5, 143) = 3.03, p < 0.001).


118

Figure 4. Ordination tri-plot depicting the position of sampling points (coloured circles), species (blue crosses) and environmental variables (black arrows) along the first two axes of a redundancy analysis (RDA) of the mammal community composition data. The first and second axes were both significant in permutation tests (1st axis: F(1, 399) = 65.1, p < 0.001; 2nd axis: F(1, 399) = 35.2, p < 0.001) and explained 7.2% and 3.9% (calculated using the adjusted coefficient of multiple determination) of the community variation, respectively. Environmental variables were chosen by a modified forward-selection procedure applied to three separate RDA sub-models for habitat structure, topography and landscape context (corresponding to Fig. 1A). This explains the strong correlations evident here between some environmental variables, making causality difficulty to ascribe in these cases. Land-use was not included as a variable in this model and sampling points are instead coloured post hoc, illustrating the differences in community composition across land-uses. Scores for each sampling point were calculated as weighted sums of the species scores. Four species which were characteristic of old-growth forest (greater mouse-deer T. napu), logged forest (bearded pig S. barbatus and red spiny rat M. surifer) and oil palm (Malay civet V. tangalunga) are individually-named. Species located close to the origin either show weak responses to the environmental variables or show highest occurrence at mean values of the variables. Note that ‘logging road or not’ was a binary variable, with the reference level, i.e. ‘not on a logging road’, located at the origin.


119

The two anthropogenic land-uses were less structured in space than old-growth forest (Fig. 3; 95% CI

of the variance explained for old-growth forest: 0.26 – 0.37; 95% CI for logged forest: 0.05 – 0.15;

95% CI for oil palm: -0.05 – 0.24). Broad- and fine-scale spatial processes were equally important in

old-growth forest (95% CI of the difference: -0.04 – 0.08, p = 0.43) and oil palm (95% CI of the

difference: -0.11 – 0.13, p = 0.90), but there was a trend for the dominance of fine-scale space in

logged forest (95% CI of the difference: -0.06 – 0.00, p = 0.09).

Co-occurrence patterns became increasingly random along the land-use gradient, with 26%, 14% and

7% of analysed species pairs classified as non-random in old-growth forest (n = 392 species pairs),

logged forest (n = 627 pairs) and oil palm (n = 43 pairs), respectively. Of these non-random

associations, most were positive (Fig. 5) and only in old-growth forest was there any substantial

evidence of negative co-occurrences (13% of non-random associations). However, it should be noted

that Type II errors are a possibility for oil palm, in which 25% of associations classified as random

represented deviations of > 3% of the total number of sites (this was the case for just 1% and 0% for

old-growth and logged forest, respectively).


120

Figure 5. Observed and expected species co-occurrences between species (excluding species species pairs for which expected co-occurrences were < 1) for each land-use type. Significant positive co-occurrences (blue points) lie above the 1:1 line and significant negative co-occurrences (orange points) lie below it. Effects sizes were calculated by standardising the difference between observed and expected co-occurrences by the number of sampling points in each land-use.


121

4. 3. Discussion

Our overall results showed a strong role for environmental control in assembling mammal

communities and therefore do not support our initial hypothesis of completely neutral communities.

Indeed our results most closely fitted the predictions arising from our final hypothesis, which posited

that both niche and dispersal would be important in assembling communities and that the relative

importance of each would not be stable along the environmental gradient created by land-use, with

dispersal and competition declining in importance and environmental filtering increasing in

importance. This was supported by an increasing dominance of environmental variables over spatial

variables along the land-use gradient, in particular an increase in the share of the explained variation

contained within the pure environmental component, and by increasingly weak evidence of negative

species co-occurrences along the land-use gradient.

We also found that the explanatory power of our models for logged forest communities was low

compared to the other land-uses. We interpret this to be consistent with an increase in the importance

of random drift, a term which we use here to refer to unexplained stochasticity in the broadest sense,

for example due to demographic processes, environmental stochasticity and historical effects. We do

not consider the decreased explanatory power to be caused by under-sampling of logged forest

communities, since the sampling intensity in logged forest was similar, if not higher, than in the other

land-uses, as were the overall trapping rates. An alternative explanation is that the environmental

variables we used were inappropriate for logged forest. We specifically designed the variables to

capture the structural dimensions affected by logging (such as the vertical stratification of vegetation

density and the degree of canopy closure), but we did not have fine-scale information on tree species

composition or direct measures of resource abundance, such as the availability of fruit. It remains the

case, however, that spatial variables performed poorly in logged forest, indicating that, if there were

key environmental variables missing from the analysis, they were not spatially-structured. Why,

therefore, might drift be more important in logged relative to old-growth forest? One possibility is that

the size of local communities is smaller in logged forest, which results in more rapid drift (Hubbell,

2001). In particular, it has been shown that the fragmentation of local communities into favourable


122

habitat patches, such as could occur within the heterogeneous logged forest, reduces local community

sizes and can allow drift to prevail over niche assembly processes (Orrock & Watling, 2010).

Consistent with this hypothesis, we have previously found that the fine-scaled aggregation of

individuals and species in logged forest is greater than in old-growth forest (Chapter 3).

For old-growth forest communities, both fine- and coarse-grained variation in habitat was involved in

environmental control, with microhabitat structure, topography and the broader landscape all playing

important roles. From old-growth forest to logged forest, there was a marked reduction in the variance

explained by the local landscape context and topographical variables, which meant that space played a

larger role than environment in structuring mammal communities. This could be due to a destruction

by logging of the heterogeneity in forest structure and species composition which is ordinarily present

across local landscapes and along topographical gradients in old-growth forest (Newbery et al., 1996).

The very high intensity of logging that Southeast Asian dipterocarp forests are typically subjected to

(Putz et al., 2012) often results in forests that are uniformly dominated by pioneer tree species, such as

Macaranga pearsonii in our logged forest sites (~10% of basal area; M. Khoo, personal

communication). Whilst elevation featured in the topography RDA model for logged forest, we

consider this more likely to be due to the variability in logging intensity, and indeed logging methods

employed, that occurs with elevation (Pinard et al., 2000b), rather than an effect of any natural

elevational gradients.

Of the three land-uses, environmental control was strongest in oil palm, particularly due to variation

in habitat at fine scales. Although evidence exists from a range of taxonomic groups that oil palm

plantations act as a strong environmental filter at the land-use scale (Foster et al., 2011), far less is

known about the effects of fine-scale habitat variation on the occurrence of species within plantations.

Oil palm plantations often exhibit substantial heterogeneity (Luskin & Potts, 2011), for example in the

age and height of palms, the amount of scrub vegetation in the understorey and the presence of access

roads, and indeed we found these to be important fine-scale filters of mammal communities. This

echoes findings for other taxa in oil palm (Chung et al., 2000; Aratrakorn et al., 2006; Peh et al., 2006;


123

Nájera & Simonetti, 2010), and more generally in other plantation types (Thiollay, 1995; Greenberg et

al., 1997; Wanger et al., 2010; Styring et al., 2011; Lantschner et al., 2012), and may indeed be a

robust pattern in these highly-modified habitats. This sensitivity to fine-scale environmental control

could be because species in plantations are persisting closer to their biological tolerances (e.g. for

food resources, microclimate or cover from predators), but further work is needed to test this

hypothesis. Besides fine-scale environmental control, we also found that the local landscape context,

in particular the proportion of forest cover remaining, played an important role in explaining

community variation in oil palm. This has not previously been demonstrated for mammal

communities in oil palm, and suggests a spill-over effect from forest to oil palm (e.g. Lucey et al.,

2014) or, at least, the importance of forest patches in providing key resources, such as resting and

breeding sites (Rajaratnam et al., 2007; Koh, 2008a). A large part of the variation explained by

landscape context was shared with topography, most likely because of the highly non-random nature

of plantation expansion; steep and high elevation forest areas are often the least likely to be converted

and therefore have a greater proportion of forest cover remaining in the landscape.

We found that broad- and fine-scale spatial variables each captured important components of

community variation, of a similar relative importance in each land-use. However, logged forest and

oil palm communities were much less spatially-structured than old-growth communities in absolute

terms, with spatial variables in both cases explaining approximately one-third the variance explained

in old-growth forest. In logged forest, this was partly caused by the reduction in the importance of

environmental control, especially landscape context, and consequent reduction in the

environmentally-induced spatial component compared to old-growth forest. However, we also

calculated post hoc the variation explained by broad- and fine-scale space once the environmental

control model had been factored out. This showed that broad-scale space, independently of

environment, explained 6%, 3% and 1% of the community variation in old-growth forest, logged

forest and oil palm, respectively. This is consistent with a reduction in the importance of dispersal

limitation along the land-use gradient.


124

Figure 6. Mantel correlograms of observed community composition data (black) compared to fitted and residual values from redundancy analyses (RDA), across three land-use types. Fitted values in each case are shown for the environmental control RDA (green) and spatial RDA (blue). Residuals (red) are shown for the overall RDA for each land-use, containing both environmental and spatial variables. Filled points indicate significant correlations, as deduced using permutation tests and progressive Holm correction for multiple testing (α = 0.05 for the first distance class and α < 0.05 thereafter).

The variation explained by fine-scale space independent of environment also declined from old-

growth forest to logged forest and oil palm (from 9%, to 4% and 3%, respectively). To explore this

further, we generated (again, post hoc) multivariate Mantel correlograms (Legendre & Legendre,

2012) of the mammal communities in each land-use (Fig. 6). Although Mantel correlations between

community distances and geographic distances were weak in all cases (r < 0.12), a stronger signal of

positive correlation was evident in old-growth than in logged forest at distances less than ~100m. One

explanation for this is the weaker positive correlation in the environment in logged forest (Fig. 6) –

perhaps caused by logging disturbance – and the weaker induced spatial dependence that results from

this. An alternative explanation, that we consider more likely, is that this positive correlation is true

autocorrelation (sensu Peres-Neto & Legendre, 2010), generated by a fine-scale spatial process, such

as home-ranging, which is not necessarily causally-related to the environment. Although the

movement ecology of Bornean mammals is largely unknown, their ranging patterns are likely to be

non-random to varying degrees, and not easily subsumable under dispersal or neutrality (which is

theoretically rooted in the study of sessile organisms). Overall, this means that a proportion of the

variation attributed to fine-scale space is not necessarily indicative of dispersal assembly. We do,


125

however, consider that broad-scale space is indicative of true dispersal limitation. Between-block

movements of individuals are likely to be infrequent, except for the most wide-ranging species in the

community, such as the Sunda clouded leopard (Neofelis diardi).

The vast majority of species pairs we examined were found to be co-occurring at random. Moreover,

there was only weak evidence of the negative co-occurrences expected under competition, and this

was only the case in old-growth forest. Most of the non-random co-occurrences were positive. We are

not aware of any documented examples of mutualism or commensalism among Bornean mammals,

and the best explanation for the positive co-occurrences is shared habitat preferences. The few

significant negative co-occurrences could equally have been generated by dissimilar fine scale habitat

preferences. This is supported by the fact that 69% of the negative co-occurrences involved the same

species (Appendix B, Fig B2), the greater mouse-deer (Tragulus napu), which our environmental

RDA identified as an old-growth forest specialist (Fig. 4). However, among the negatively co-

occurring species, were three congeneric pairs which may be suitable candidates for future

investigation into inter-specific competition: two morphologically very similar and abundant

Maxomys rats (M. surifer and M. rajah), the greater and lesser (T. kanchil) mouse-deer, and the thick-

spined and Malay porcupines (Hystrix crassispinis and H. brachyura). Overall, however, we maintain

that environmental filtering is likely a more dominant assembly process for mammal communities

than species interactions. This does not discount the possibility that species pools have been shaped

by competition in the evolutionary past (Connell, 1980).

Our results suggest that the assembly of mammal communities in logged forest may be driven by

dissimilar mechanisms to old-growth forest, which could represent a pervasive, but as-yet-

undocumented, legacy of logging. Besides some fine-scale environmental control, environmental

control was weak overall, possibly due to the homogeneous composition of logged forest at coarse

scales, which meant that community drift was a more important process in relative and possibly

absolute terms. If this is a general result for mammal communities in the region, it could call into

question the prospects for large-scale, unassisted restoration of ecosystem processes in logged-over


126

forests, and may mean that more active management interventions, targeted at species of particularly

concern, may be more appropriate (Lamb et al., 2005). Our finding that mammal communities in oil

palm were strongly assembled by environmental control, particularly at fine-scales, suggests that there

is at least the potential for ameliorating the impacts of oil palm by altering management practices.

Further detailed study of the factors determining species persistence in oil palm would be required in

order to identify recommendations that would be practical for industry to implement. Although we

found an important role for landscape context in oil palm, in particular remnant forest cover, the

potential benefits of advocating this conservation strategy for mammals would have to be reconciled

with potential losses of natural forest elsewhere due to agricultural extensification (Edwards et al.,

2010b).

We conclude that assembly processes are not robust to anthropogenic environmental gradients such as

land-use, and indeed that disturbance may lead to novel mechanics governing the local assembly of

mammal communities. Our overall results support the niche-neutrality continuum model of

community assembly, in which dispersal, drift and environmental control all combine variously to

create communities at the local scale (Gravel et al., 2006; Mutshinda & O’Hara, 2011). Our study

represents one of the first to investigate the local scale drivers of mammalian community assembly

and it remains unknown whether our findings might also hold for other non-sessile organisms,

including other vertebrate groups. New biodiversity sampling technologies offer an opportunity to

generate a more comprehensive picture of community assembly for a wider range of taxonomic

groups in the future, thus allowing the testing and refinement of ecological theory, as well as

facilitating better conservation and management of these imperilled natural systems.


127

Appendices

Appendix A – Detailed description of methods used to measure environmental variables

A. 1. Field-based measurement of environmental variables

All habitat structure variables were quantified directly in the field. Eight variables were measured

directly at sampling points by a single observer (O.R.W.), whilst a further five were measured in

intensively-sampled 25 m2 vegetation plots as part of ongoing monitoring at the Stability of Altered

Forest Ecosystems (SAFE) Project (Ewers et al., 2011). We hereafter refer to these vegetation plots as

‘quadrats’ to distinguish them from the larger plots used to sample mammal communities. Two

quadrats were located within each sampling plot, spaced evenly within the plot area and 150 m apart

(centre-to-centre distance). Quadrats were mostly < 45 m from individual sampling points (range: 0 to

75 m).

Vegetation cover within a 5 m radius of each sampling point was estimated in four height strata

(ground: 0-0.5 m; understorey: 0.5-3 m; mid-storey: 3-20 m, and canopy: above 20 m) and placed into

one of five broad classes for each stratum (1: 0-25%; 2: 25-50%; 3: 50-75%, and 4: 75-100%).

Canopy cover was quantified using a spherical densiometer (Lemmon, 1957), held at waist height and

recorded as an average percentage across four measurements (one for each cardinal direction). Before

analysis, this percentage was arcsine-transformed, owing to the strong negative skew apparent across

all measurements. The intensity of habitat disturbance (“habitat score”) within a 5 m radius of each

point was recorded on a 1 to 5 scale (definitions provided in Table A1), following a similar

methodology to previous studies (Ewers et al., 2011; Cusack et al., 2015). Higher habitat scores

represent more intact sites, and sites within the oil palm plantation crop itself never exceeded a score

of 2, though areas in the margins of plantations reached scores of 3 in some areas. We also noted

whether the sampling point was on a logging road or not, due to the strong influence these features

have on the occurrence of some species in our study sites (see Chapter 2). Old logging roads, often

following ridge-lines, were apparent due to their graded surface, the poor regeneration state of


128

understorey vegetation, and the lack of canopy cover. Habitat scores and vegetation cover estimates

were treated as ordered categorical variables during analysis. We also included in the analysis a 2nd-

order polynomial term for habitat disturbance (“habitat score2”), due to the non-linear responses to

this variable we have previously found for some species (Cusack et al., 2015).

Quadrat-based variables were measured during the course of long-term vegetation monitoring using

RAINFOR protocols (Malhi et al., 2002; Turner et al., 2012). This involved mapping, measuring and

tagging all trees ≥ 10 cm diameter at breast height (DBH) inside each quadrat. Tree heights were

estimated by field teams on the ground, and were not significantly different from model-based

estimates obtained using DBH measurements (M. Pfeifer, unpubl. data). We also included a 2nd-order

polynomial term for DBH during the analysis. This was because we expected hump-shaped responses

to this variable in at least some species, owing to the fact that the largest mean DBH values were

observed within oil palm (Elaeis guineensis) quadrats. Total deadwood volumes in each quadrat were

obtained by summing the volumes of all coarse woody debris pieces (≥ 10 cm diameter), including

standing, hanging and fallen deadwood. Volumes of each piece were estimated using the formula for a

truncated cone, following (Baker et al., 2007), which required measuring the diameter of each piece at

both ends, as well as the length. For standing deadwood, the top-most diameter was estimated using

the taper function (Chambers et al., 2000).

Table A1. Definition of habitat disturbance scale. Habitat score Definition

1 Open area. Dominated by grasses and small shrubs (< 1 m in height). Typically on logging roads or old log landing areas. 2 Herbaceous scrub. Dominated by herbs (typically Zingiberaceae), vines and shrubs, with no trees > 3 m in height (except

oil palm Elaeis guineensis). Typically representing secondary re-growth from clear-felling, or large gaps due to landslides.

3 Heavily-disturbed forest. High scrub or dense understorey layer (typically with vines and Dinochloa climbing bamboo species), with a low, heavily-broken canopy layer (< 20 m). Possibly some large isolated trees (> 20 m). Intensively-logged area or large gap disturbance.

4 Disturbed forest. Mostly pioneer tree species (typically Macaranga species), but some old-growth dipterocarp species may be present. Discontinuous canopy. Lower intensity of logging or natural disturbance.

5 Undisturbed forest. Dominated by old-growth dipterocarps. High, continuous canopy with sparsely-vegetated understorey. Unlogged, with little recent disturbance evident.


129

A. 2. Satellite-based measurement of environmental variables

Topographical variables were all derived from the Advanced Spaceborne Thermal Emission and

Reflection Radiometer Global Digital Elevation Model (DEM) version 2 (https://lpdaac.usgs.gov),

jointly developed by the Ministry of Economy, Trade and Industry of Japan and the United States

National Aeronautics and Space Administration, and which was provided at 30 m horizontal

resolution. We extracted elevation data from this DEM at our sampling points, applying bilinear

interpolation. Slope was estimated as the maximum rate of change from each cell in the DEM and was

calculated in degrees (taking values between 0 and 90). To calculate flow accumulation, we 1) created

a depressionless DEM by filling in ‘sink’ artefacts in the data, 2) created a flow direction map from

this corrected DEM, and then 3) for each cell, summed the number of upstream cells. Areas of zero or

low flow accumulation represented ridges, whilst high flow accumulation areas represented gullies,

stream heads, streams and rivers. For the purposes of analysis, we log-transformed flow accumulation,

owing to the strong positive skew in the values, with rivers otherwise represented by very large

values. All topographical variables were calculated in ArcMap version 10 (ESRI, Redlands,

California, USA). During analysis, we also included 2nd-order polynomial terms for each

topographical variable, owing to the non-linear, and possibly hump-shaped, species responses we

expected for these variables.

Ground-based measurements of above-ground live tree biomass (AGB) were taken from all

vegetation quadrats (n = 193) sampled across the SAFE Project, which amounts to intensive-sampling

of > 12 ha in total. AGB was calculated for each quadrat using Chave et al.'s (2014) pan-tropical

algorithm. Spectral data were derived from sensors onboard the RapidEye satellite (European Space

Agency Earth Observation Portal: https://earth.esa.int), which were provided at 5 m resolution. All

pre-processing and atmospheric correction of the images, taken in 2012 and 2013, followed the steps

outlined in Pfeifer et al. (2015). Within 20 m radius buffers centred on each quadrat, we extracted the

spectral intensity values for each of the five bands present in the RapidEye images (blue, green, red,

red-edge and near-infrared) and calculated a spectral vegetation index, the Modified Soil-Adjusted

Vegetation Index 2 (MSAVI2) from the red and near-infrared bands (Qi et al. 1994). Note, we did not


130

take the red and near-infrared band spectral intensities forward into modelling, since these were used

in the calculation of MSAVI2. We transformed MSAVI2 by taking its exponent, because of the

saturating response observed at high levels of AGB. We also calculated a measure of image texture

(dissimilarity), within 9 x 9 pixel moving windows, for each band. 58 quadrats were covered by cloud

or cloud shadow in our images and were excluded. Using linear models of AGB as a function of each

possible combination of the nine covariates (MSAVI2, three spectral intensity covariates and five

dissimilarity covariates), we then obtained a candidate set of “best” models based on information-

theoretic criteria (i.e. models for which ΔAICc < 4) and calculated model-averaged estimates for each

parameter (Table A2) based on the model selection weights in this set. The pseudo-R2 (explained

deviance) of this final model was 0.53.

Based on the model-averaged parameters, we made AGB predictions within 500 m, 1 km and 2 km

radius buffers surrounding each of our sampling points, at a resolution of 25 m2, which matched the

resolution of our ground-based measurements. Finally, we calculated the mean AGB (excluding

cloud-covered pixels) within each buffer size. All steps in the analysis of AGB were done in R

version 3.1.0 (R Development Core Team, 2014), using the packages raster 2.3-0 (Hijmans, 2014),

rgeos 0.3-8 (Bivand & Rundel, 2014), glcm 1.2 (Zvoleff, 2015) and MuMIn (Barton, 2015).

AGB values across buffer sizes were highly correlated (Pearson’s r > 0.98), so we fitted redundancy

analysis (RDA) models (with the vegan 2.0-10 package in R; Oksanen et al. 2013) for each buffer size

and used the buffer size explaining the largest share of the community variation (calculated using the

adjusted coefficient of multiple determination, R2adj; Blanchet et al. 2008) in further analyses. This

selected the buffer with a 500 m radius (R2adj = 11.1%), although there were not large differences in

the variation explained by the different buffer sizes (1 km: 10.8 %; 2 km: 10.7 %).

In order to calculate landscape forest cover and distances from forest for each sampling point, we first

created a digitised forest cover map in ArcMap using visual interpretation of RapidEye satellite

images, in combination with cloud-free Landsat 7 and 8 images (30 m resolution) released by Hansen


131

et al. (2013). We distinguished natural forest from mature oil palm and Acacia mangium plantations

by observing the dynamics of vegetation gain and loss over multiple years (1999-2013), as well as

using our detailed knowledge of the study sites, but it is possible that some older plantation areas may

have been included in our forest cover map (if they were already > 5 m in height before the year 2000

and were not harvested after that time). Euclidean distances from forest were calculated in ArcMap

and percentage forest cover was quantified in buffers with 500 m, 1 km and 2 km radii using the rgeos

package in R. As for AGB, we fit RDA models for each of the buffer sizes and selected the radius

which explained the largest percentage of the community variation. This resulted in the 500 m radius

being chosen (R2adj = 6.2%), though similar percentage variances were explained by the other buffer

sizes (1 km: 5.8%; 2 km: 5.3%).

Table A2. Model-averaged parameter estimates for linear models of field-based above-ground live tree biomass (AGB) measurements, as a function of satellite-derived measures of vegetation, spectral intensity and image texture.

Parameter Estimate Standard error

(adjusted) z-value p-value Relative variable

importancea

Intercept 145.03 38.21 3.80 < 0.001 -

Band 2 (green) intensity -17.44 3.56 4.91 < 0.001 1

Band 2 (green) dissimilarity 20.99 5.92 3.55 < 0.001 1

Band 4 (red-edge) intensity 8.28 2.61 3.18 0.001 1

exp(MSAVI2) -29.97 12.05 2.49 0.013 0.97

Band 3 (red) dissimilarity 6.41 5.00 1.28 0.20 0.42

Band 1 (blue) dissimilarity 5.69 5.11 1.12 0.27 0.34

Band 5 (near-infrared) dissimilarity -1.15 1.34 0.86 0.39 0.31

Band 4 (red-edge) dissimilarity -1.89 3.38 0.56 0.58 0.26

Band 1 (blue) intensity -0.13 2.81 0.05 0.96 0.19 aCalculated as the sum of the AICc weights for the models in which the given parameter appears.


132

Appendix B – Supplementary Results

Table B1. Environmental variables chosen by a modified forward-selection procedure based on the adjusted coefficient of multiple determination (R2

adj), for the global model and models specific to land-use types (old-growth forest, logged forest and oil palm) and species groups (large mammals and small mammals).

Environmental variable

Redundancy analysis (RDA) model

Global Old-growth forest

Logged forest

Oil palm

Large mammals

Small mammals

Habitat structurea

Ground cover

Understorey cover

Midstorey cover

Canopy cover

Canopy closure (arcsine-transformed %)

Habitat score

Habitat score2

Logging road (binary)

Maximum tree height (m) (quadrat-based)c

Tree density (quadrat-based)c

Mean DBH (cm) (quadrat-based)c

Mean DBH2 (quadrat-based)c

Deadwood volume (m3) (quadrat-based)c

Topographyb

Elevation (m)

Elevation2

Flow accumulation

Flow accumulation2

Slope (degrees)

Slope2

Local landscape contextb

Above-ground live tree biomass (Mg/ha) (500 m radius)

Forest cover (%) (500 m radius)

Distance from forest (m)

aAll habitat structure variables were measured in the field.

bAll topographical and landscape variables were derived from satellite data. c"Quadrat-based" variables were measured within 25 m2 quadrats, located 0 - 75 m from sampling points, at a density of two quadrats per plot.


133

Figure B1. Shared and independent percentages of community variation explained (as determined using the adjusted coefficient of multiple determination, R2

adj) by environmental and spatial surrogate variables, for large and small mammal communities separately.


134

Figure B2. Species co-occurrence matrices across land-use types. Negative and positive co-occurrences indicate where two species occurred together significantly less or more often than expected by chance, respectively. Species columns are arranged into three sets (from left to right): species involved in significant co-occurrences; species only ever occurring randomly with respect to other species, and species which were not detected frequently enough to be included in the analysis. For species involved in significant co-occurrences, column order is by the sum of their co-occurrences (where positive and negative co-occurrences are scored as 1 and -1, respectively), from species which are involved in the most negative co-occurrences to species involved in the most positive co-occurrences.


135

Figure B3. Distance-based Moran’s eigenvector map (db-MEM) variables chosen by a modified forward selection procedure during redundancy analysis of the community composition data. MEM variables represent a spectral decomposition of the spatial relationships among points and are ordered from the coarsest to the finest scale within each block. Of the 156 db-MEMs generated, 47 were identified as significant in forward selection and were subsequently used as spatial surrogate variables for modelling fine-scale spatial processes. Single-predictor R2

adj values were derived from separate redundancy analyses for each db-MEM.


136

Figure B4. Examples of distance-based Moran’s eigenvector maps (db-MEMs) identified as significant by a modified forward-selection procedure used in a redundancy analysis. These db-MEMs were used as spatial surrogate variables of fine-scale spatial processes. Broad-scale spatial processes were modelled using trend-surface variables.

Chapter 5: Mammalian species abundance responses to land-use change

137

Chapter 5:

Species abundance across a gradient of tropical land-use intensity: a hierarchical

multi-species approach applied to a Bornean mammal community

Abstract

Despite rapid rates of global land-use change, we still have a limited ability to make forecasts of

species abundance. Although most work has hitherto focussed on species richness, abundance is often

a more useful state variable for conservation and management. Consistency in reported abundance

responses to land-use change has remained elusive across different study regions or taxa, even when

similar types of land-use transition (e.g. forest to plantation) have been compared. In part, this is

because the detectability of species in a given study is rarely accounted for. In addition, continuous, as

opposed to categorical, metrics of land-use change are rarely used, but may allow better predictions in

heterogeneous landscapes, as well as provide more practical recommendations for management and

conservation. We applied a novel hierarchical multi-species model to data from two sampling

methods, in order to estimate the abundance (number of individuals using a given sampling point) of a

near-complete terrestrial mammal metacommunity in our study region in northern Borneo, using both

categorical and continuous metrics of land-use change in the model. We found that mammalian

abundance was resilient overall across the transition from old-growth to logged forest, but declined

substantially in oil palm. Abundance responses to above-ground live tree biomass (a continuous

measure of logging intensity) in a given local landscape were negative overall, although weakly

unimodal, whilst they were strongly positive for landscape forest cover. From old-growth to logged

forest, small mammals increased in abundance much more than large mammals. Similarly, omnivores,

insectivores and herbivores increased more than other trophic guilds. From forest to oil palm, species

of high conservation concern fared especially poorly. Invasive species consistently increased along

the gradient of land-use intensity. The functional effects of these abundance changes, as assessed

using nine species groups based on diet, were minimal from old-growth to logged forest, but only the


138

vertebrate predation function was maintained in oil palm. Our results complement previous assertions

of the high value of even the most degraded forests in Southeast Asia for conserving species richness,

extending this value to their role in conserving the abundance of individual species and their

functions. Our results also underline the very low value of the oil palm crop habitat for conserving

species and support the view that “wildlife-friendly” practices within plantations offer a low potential

for reducing biodiversity impacts.

5. 1. Introduction

Wilderness areas free of anthropogenic influence are now rare on planet Earth, and changes in land-

use due to biomass extraction and agriculture have created novel ecosystems over vast areas. In

particular, tropical forests, the most biodiverse of terrestrial biomes, are currently experiencing

historically unprecedented rates of selective logging and conversion to plantations, cropland and

pasture (Asner et al., 2009; Gibbs et al., 2010). In Southeast Asia land-use change has been especially

acute, with the vast majority of remaining forest now existing in a logged-over state (Margono et al.,

2014; Gaveau et al., 2014) and deforestation rates, in large part due to oil palm (Elaeis guineensis)

plantation expansion (Koh & Wilcove, 2008; Gunarso et al., 2013), the highest among the major

tropical forest regions (Asner et al., 2009). Land-use is well-known as a major driver of ecological

change, for example as a leading cause of species endangerment (Vié et al., 2009), but there remains a

limited capacity to make biodiversity forecasts, especially of species abundances, at scales which are

relevant to local stakeholders and policy-makers responsible for making land-use decisions.

Most previous research on the biodiversity impacts of land-use change has focussed on community-

level parameters, in particular species richness. The more subtle impacts on species abundances have

been quantified less frequently, and often only for single focal species or a limited subset of species.

In part, this reflects the additional challenge that abundance monitoring represents. The problems

caused by imperfect detection are well-known (Williams et al., 2002; Sollmann et al., 2013) and, for

species richness estimation, there is a relatively well-established toolbox of statistical methods


139

available to control for this (Gotelli & Colwell, 2001). However, this is not the case for abundance

estimation, except in the limited cases when individuals of a species can be recognised, or are

deliberately marked. This has led many researchers to avoid making inferences about abundance

using the, often very sparse, data they have available for each species in a community, especially for

the rarest or most cryptic species. Despite this, abundance, as opposed to species incidence, is often

the state variable of most use for conservation. Abundance estimates give a finer resolution of

information on species responses to environmental change, facilitating better decisions surrounding

trade-offs in land-use (Phalan et al., 2011) and potentially acting as an early-warning indicator before

species go extinct altogether. Importantly, species abundances may also be indicative of ecosystem

functioning (Ewers et al., 2015), as well as the trophic structure and interaction strengths present in an

ecosystem (Barnes et al., 2014).

There is a developing consensus about the impacts of land-use change on species richness, such as the

relatively lower impacts of selective logging relative to plantation forestry, which in turn often retain

more species than monoculture plantations (Barlow et al., 2007; Scales & Marsden, 2008; Gibson et

al., 2011; Barnes et al., 2014; Edwards et al., 2014a). For abundance responses, on the other hand,

declines under increasing disturbance are typically less marked and consistent patterns across land-use

types have remained elusive (Sodhi et al., 2009a; Gibson et al., 2011; Newbold et al., 2014; Supp &

Ernest, 2014). Moreover, abundance responses to land-use change often differ across taxonomic

groups, with some groups reaching their highest abundance in highly modified habitats (Foster et al.,

2011; Senior et al., 2012). In a global meta-analysis across taxonomic groups, available data indicated

that mammals, in particular, were a group which apparently often increase in abundance under

disturbance (Gibson et al., 2011).

The variability in abundance responses reported across studies means that there is still a limited

capacity to make robust predictions about the impacts of land-use change (Newbold et al., 2014). The

majority of past studies have based their inferences about abundance on sparse data, often on a biased


140

subset of species in a community, and without controlling for the potentially confounding influence of

detection probability, for example across species and across land-use types. These factors may, at

least in part, explain the large variability in reported abundance responses and, in the worst cases, may

be a source of systematic bias in inferences.

Land-use change sometimes involves dramatic and rapid changes to a natural habitat, for example

when a primary forest is converted to pasture. More often, land-use change manifests itself as a

gradient of disturbance intensity, rather than distinct land-use categories, and incorporating this into

models of abundance may be an opportunity to increase the predictive power and practical relevance

of forecasts for conservation and management. For example, the intensity of selective logging may

vary considerably across a landscape, due to access constraints and natural variability in marketable

timber volumes (Cannon et al., 1994; Berry et al., 2008). Similarly, plantations may vary in their

proximity to remaining forests and in their structural properties, such as height and canopy cover, as

they mature (Luskin & Potts, 2011; Foster et al., 2011). The propensity in past studies to use land-use

categories, rather than continuous metrics of disturbance, perhaps reflects the often categorical nature

of land-use decisions and is therefore politically expedient. However, the focus on categorical

measures of land-use intensity may, in many cases, also be due to a lack of suitable metrics available

for modelling. The burgeoning availability of aerial and satellite remotely-sensed data is rapidly

overcoming this problem (Pettorelli et al., 2014), offering the potential to provide continuous metrics

of land-use intensity at any location and at any spatial extent.

Statistical tools for robustly estimating abundance are also rapidly advancing. Importantly, the

statistical toolbox has recently been extended to include abundance estimation for species which are

not individually identifiable (Royle & Nichols, 2003; Royle, 2004; Rowcliffe et al., 2008). In

addition, computing power is now sufficiently advanced such that complex but flexible hierarchical

models, which can account for the specific observational and ecological processes at work in a dataset

(Royle & Dorazio, 2008), are now a practical option using Bayesian inference and Markov chain


141

Monte Carlo (MCMC) methods. The application of new statistical models, using advanced

computational tools, has opened up the possibility, for the first time, of multi-species models of

abundance for whole communities. Moreover, these methods are ideally suited to the large datasets

being generated by the increasing deployment of remotely-operated sensors in the field, such as

acoustic recorders and camera traps.

Here we investigate species abundances for a community of terrestrial mammals across a land-use

gradient in Borneo. We estimate abundance for all species using a novel hierarchical model of the

mammal metacommunity in our study region which accounts for 1) imperfect detection, 2) correlated

detections in group-living species, 3) multiple sampling methods (camera traps and live traps), 4)

clustered sampling designs, and 5) habitat filtering according to land-use and fine-scale habitat

disturbance. We used both categorical and continuous approaches to characterise the land-use

gradient. In the former case, we used three categories which match the major land-use options

available to forest concession holders in the region: old-growth forest, logged forest and oil palm

plantation. For our continuous metrics of land-use intensity, we employed satellite-derived measures

of above-ground live tree biomass (AGB) and local landscape forest cover. AGB is directly

proportional to carbon content (Martin & Thomas, 2011), and this metric is therefore relevant for

assessing the value of High Carbon Stock (HCS) set-aside areas – likely to be implemented by many

of the world’s leading palm oil producers (Poynton, 2014) – for mammal species. Landscape forest

cover is relevant to management decisions concerning the quantity of forest set-aside within a

concession, for example as High Conservation Value (HCV) areas or riparian reserves in oil palm

plantations (Koh et al., 2009; Edwards et al., 2012). We also partitioned the mammal community

according to four ecological response traits – body size, conservation status, native status and trophic

guild – as well as into functional effects groups based on diet, and present abundance and biomass

responses of these sub-groups. In particular, we were interested in whether: a) small and large

mammals respond similarly, given that current consensus surrounding mammal responses to

disturbance are based primarily on small mammal studies (Gibson et al., 2011); b) particular trophic


142

guilds are more susceptible to land-use change, as indicated recently for birds (Gray et al., 2007;

Edwards et al., 2009); c) species of high conservation concern, often the focus of HCV surveys and

assessments, are more sensitive to logging and conversion to oil palm, relative to native and invasive

species of lower concern, and d) the biomass of particular dietary functional groups is altered across

the land-use gradient, with potential implications for ecosystem function.

5. 2. Methods

5. 2. 1. Sampling design



2011). This consists of old-growth forest within the Maliau Basin Conservation Area and Brantian-

Tatulit Virgin Jungle Reserve (VJR), repeatedly-logged forest within the Kalabakan Forest Reserve

and two adjacent oil palm plantations straddling the Kalabakan Forest Reserve boundary. The VJR

(4,140 ha) was gazetted in 1984, and is managed by the Sabah Forestry Department for the purposes

of research and conservation. It is split into four main fragments, of which the largest (2,200 ha) lay

within the SAFE Project area. There was some evidence of illegal logging in the reserve, with old skid

trails (now with continuous canopy cover) present in the vicinity of some sampling points. However,

the majority of the reserve remains unlogged and, for the purposes of our land-use categorisation, we

class this as old-growth forest. Detailed descriptions of the other study sites can be found in Chapter

3.

We employed a clustered hierarchical sampling design, with 48 sampling points clustered together

into each of 46 sampling plots (each covering 1.75 ha), in turn clustered into 11 sampling blocks

distributed across the land-use gradient (Fig. 1). This included 13 plots (in 4 blocks) in old-growth

forest, 24 plots (in 4 blocks) in logged forest and 9 plots (in 3 blocks) in oil palm plantations. The

clustered design we used likely comes at the cost of reduced precision in our abundance estimates, but

also allowed us to investigate the fine-scale spatial patterns of species occurrence, which we have


143

explored elsewhere (see Chapters 3 and 4). Sampling plots overlapped the SAFE Project sampling

design, and therefore benefitted from the deliberate control of potentially confounding factors

(including latitude, slope and elevation) that was central to this project’s design (Ewers et al., 2011).

Figure 1. Sampling design across a gradient of land-use intensities in Borneo, showing the plots sampled using both camera traps and live traps (in red) and plots sampled only with camera traps (in orange). In logged forest, plots were arranged to coincide with future experimental forest fragments. The shaded area in panel C shows the edge of Mount Louisa Forest Reserve, which connects Kalabakan Forest Reserve to an extensive (>1 million ha) area of contiguous logged forest to the north. The area of old-growth forest in panel C is indicative of the boundaries of the Brantian-Tatulit Virgin Jungle Reserve only, and forest along the western and southern boundaries has been degraded by edge effects and illegal logging. Insets show the location within insular Southeast Asia and the spatial proximity of panels A to C within south-east Sabah, Malaysia.

5. 2. 2. Field methods

Of the 48 sampling points within each plot, a random subset of 13 points (range: 8 to 22) in each of

the 46 plots were selected for camera-trapping, giving 590 points sampled in total. We deployed

camera traps (Reconyx HC500, Holmen, Wisconsin, USA) as close to each random point (marked in

the field on a separate occasion) as possible, and strictly within 5 m. Cameras were fixed to trees or


144

wooden poles, or placed within locally-made steel security cases in areas of high human traffic, with

the camera sensors set at a height to maximise detection for a range of species (most often 30 cm,

though this was flexible depending on the terrain encountered at each location). No bait or lure was

used, cameras used an infrared flash during night-time hours, and disturbance to vegetation was kept

to a minimum. Cameras were programmed to take 10 consecutive photos on being triggered (1 image

per ~ 0.5 s), with the minimum possible delay on triggering (~ 0.2 s), and the sensor was set to

maximum sensitivity. Camera-trapping took place between May 2011 and April 2014, during which

most plots (40 of 46) were sampled in multiple years (mean effort per plot = 635 trap nights). We

excluded 18 points which had been camera-trapped for less than seven days, giving a total sampling

effort of 29,121 camera trap nights (after correcting for camera failures).

Of the 46 plots sampled using camera traps, 31 were also sampled using live traps. Two locally-made

steel-mesh traps (18 cm wide, 10 - 13 cm tall and 28 cm in length), baited with oil palm fruit, were

placed at or near ground level (0 - 1.5 m) within 10 m of all 48 points in a plot. Each trapping session

consisted of seven consecutive trapping days and some plots (14 of 31) were sampled for multiple

sessions across the study period (mean effort per plot = 1099 trap nights). Traps were checked each

morning and captured individuals were anaesthetised using diethyl ether (following Wells et al. 2007),

measured, permanently marked using a subcutaneous passive inductive transponder tag (Francis

Scientific Instruments, Cambridge, UK), identified to species using Payne et al. (2007) and released at

the capture location. Trapping, totalling 34,058 trap nights, was carried out between May 2011 and

July 2014, during which there were no major mast-fruiting events.

We scored the habitat disturbance in a 5 m radius around each sampling point on a 1-5 scale,

representing a scale of high to low disturbance intensity. For example, a score of 1 was used in open

areas, such as on roads or log-landing areas, whilst a score of 5 was used in intact, high canopy forest

(full definitions of the levels are provided in Chapter 4, Appendix A). This variable was used because


145

it has previously been shown to be important in explaining the fine-scale composition of communities

(Chapter 4; Cusack et al., 2015). All measurements were made by a single observer (O.R.W.).


We reviewed all camera trap images (~800,000) in Adobe Photoshop Lightroom versions 4 and 5

(Adobe Systems, San Jose, California, USA) and added keyword tags to record the content of each

image, including the onset of photo-trapping events (i.e. when an animal entered a camera’s field-of-

view) and the species present. This created a text string in the Extensible Metadata Platform keyword

field of each image, which we exported to a spreadsheet using ExifTool version 9.41 (Harvey, 2013)

and subsequently parsed in R version 3.1.0 (R Core Team, 2014) . We then filtered the photo-trapping

events, with an algorithm written in R, into independent captures, using two criteria: 1) events

consisting of different individuals, or 2) events separated by > 1 hour. Different individuals were

distinguished based on species, as well as accessory tags we applied during cataloguing, such as sex,

age class and other distinguishing features. In this way, our data are fully reproducible and did not

involve any manual data entry into a spreadsheet.

To estimate species abundance, we extended a class of multi-species occupancy models (Royle &

Dorazio, 2008; Dorazio et al., 2011; Yamaura et al., 2011; Tobler et al., 2015) to accommodate

multiple sampling methods and group-living species. These all have in common a necessity for

replicate samples in space and time, in order to separate the latent ecological processes of interest

from the observational processes by which the data are generated. We therefore transformed our data

to the required form of detections and non-detections within temporal replicates, or occasions, for

each site. Here we define an occasion, for live-trapping, as a single night’s trapping at a point (i.e. two

trap nights, given that two traps were deployed per point) or, for camera-trapping, as five consecutive

calendar days. Data from partially-observed days at the beginning and end of a camera’s deployment,

and any days remaining after division into occasions, were excluded.


146

For the observation component of our model, we posited that heterogeneity among sites (i) in the

abundance (Ni,j) of species j imparts heterogeneity in species detection probabilities (pi,j,k) for

sampling method k, extending Royle & Nichols' (2003) approach to allow for multiple methods.

Specifically, species detection probabilities were assumed to equal the probability of observing at

least one of the individuals present at a site, each with a method- and site-specific, individual-level

detection probability (ri,j,k):

jiNkjikji rp ,)1(1 ,,,, −−= (1)

In turn, the number of species detections Di,j,k were assumed to be a realisation of ni,k binomial trials,

one for each occasion sampled, so that:

Di,j,k ~ Binomial(ni,k, pi,j,k) (2)

Importantly, this model formulation assumes that 1) detections across occasions are independent and

2) the observation of individuals within a sampling occasion is independent (individual probabilities

are simply multiplied to obtain pi,j,k). For the first assumption, there is the potential for violation of

independence given that camera traps are continuous-time detectors, but our filtering of photo-

trapping events into independent detections helped guard against this. We relaxed the second

assumption for seven species that were group-living (Asian elephant Elephas maximum, banteng Bos

javanicus, bearded pig Sus barbatus, domestic dog Canis familiaris, long-tailed macaque Macaca

fascicularis, southern pig-tailed macaque M. nemestrina and Oriental small-clawed otter Aonyx

cinereus) using a quasi-binomial observation process by introducing a parameter jθ (Royle &

Dorazio, 2008) into Eq. 2, which allowed detections to accrue faster than binomial:

)1(,)1(1 ,,,,

jjiN

kjikji rpθ+

−−= (3)

We considered two site-specific covariates – land-use type (LU) and fine-scale habitat disturbance

(HD) – acting on individual-level detection probabilities, and used a logit link function to keep ri,j,k on

the probability scale. We also included a 2nd-degree polynomial term for habitat disturbance, to allow

for unimodal responses.


147

We did not expect large variation in detection probability due to changes in the effectiveness of either

live- or camera-trapping with land-use or habitat disturbance, so we interpret any differences to be

largely due to the responses of species themselves to these factors. In particular, range sizes and

movement speeds may change across land-use, affecting the probability that an individual passes a

sampling point. Habitat disturbance may affect this probability too, by filtering species according to

the fine-scale habitat preferences of each species, which we have previously shown to affect trapping

rates (Chapter 2) and community composition per se (Chapter 4).

We accounted for differences in detection probability between live-trapping and camera-trapping by

including a covariate for trap-type. Given the small size of our live traps, we set individual detection

probabilities to zero for species which could never be caught with this method (by multiplying ri,j,k by

a binary matrix describing the availability of each species to trapping for each observation Di,j,k), but

considered that all species could theoretically be caught by the highly sensitive camera traps we were

using (even though not all species were camera-trapped).

The linear predictor for the observation component of our model was therefore:

logit(ri,j,k) = ×+ −LUrlj

rkj

,1,, βα LUi,l-1 ×+ HDr

j,β HDi ×+

2,HDrjβ HD 2

i (4)

where LUi,l-1 was a two-column matrix of dummy variables coding for the logged forest and oil palm

land-use (l) types, and the parameter for old-growth forest was integrated into the intercept. We note

that our model for the observation process assumes that all heterogeneity in detection probability is

either induced by variation in abundance or else by the covariates we included in the model. In

particular, heterogeneity induced by the proximity of trap locations to the centres of activity of

individual animals is not included.

For the ecological component of our model, we introduced a land-use filter (Ωj) which determined

whether a given land-use (l) was occupied by the species (wj,l = 1) or not (wj,l = 0). Ωj can be thought


148

of as the occupancy probability at the level of land-uses, rather than the traditional point-level

occupancy probability, and provides a source of zero-inflation in our model.

wj,l = Bernoulli(Ωj) (5)

For species which were detected in a given land-use during sampling, wj,l will be equal to 1 in every

MCMC iteration. Otherwise, wj,l varies across iterations, and the mean of these values provides the

probability that the species occupies the given land-use, despite not being detected. If a species was

present, it was assumed that individuals were randomly distributed in space, meaning that the

expected abundance at a sampling point would follow a Poisson distribution with intensity λij, so that

Ni,j = Poisson(λi,j) × wj,l (6)

The Poisson intensity for a given species represents the number of individuals using a given sampling

point in a given sampling session. We stress that this is not equivalent to animal density and will be a

function of the effective trapping area for each species, as is also true of occupancy estimates (Efford

& Dawson, 2012). In order to make density comparisons across species, it would be necessary to

make further assumptions, in particular that species home range sizes are comparable. However, our

abundance estimates likely serve as a good index of density changes across the land-use gradient,

given that we controlled for detectability by land-use category and habitat disturbance. This means

that changes in the availability of animals for trapping, for example due to changes in the size of their

home range, will be accounted for. We also note that the number of individuals using a given point in

space is an ecologically-relevant state variable, with potential implications for ecosystem function.

We modelled species abundances as a function of site-specific covariates and both spatial and

temporal additive random effects for each species, using a log link function to constrain Nij to be

positive. Spatial random effects accounted for the clustered sampling design we used, with sampling

points nested within plots, in turn within blocks. A temporal random effect of year enabled us to

account for varying abundance across the multiple years of our study. We considered year to be a

random effect, rather than modelling abundance as a function of covariates or using an explicit

population model, since any temporal variation was not a focus of the analysis presented here.


149

The site-specific covariates for abundance were either 1) categorical land-use types or 2) continuous

metrics of above-ground live tree biomass (AGB) and percent forest cover (FCOV), both calculated as

means within 500 m radius buffers around each sampling point (further details on the mapping of

AGB and forest cover are provided in Chapter 4). We have previously used this buffer size to

investigate community composition responses in the same study sites (Chapter 4). To allow for

unimodal responses, we also included a 2nd-degree polynomial term for AGB. We did not include a

polynomial term for FCOV because we had insufficient coverage of the covariate’s full range, which

meant that we did not have sufficient information to resolve any particular non-linear form of the

response.

The linear predictor for the categorical model (CAT) was therefore

log(λij) = ×+ −LU

ljCAT

j,

1,, λλ βα LUi,l-1

CATji

CATji

CATji ,,, εδγ +++ (7)

where γi,j, δi,j and εi,j represent the plot, block and year random effects, respectively. The linear

predictor for the continuous metric model (CONT) was

log(λij) = ×+ AGBj

CONTj

,, λλ βα AGBi ×+2,AGB

jλβ AGB 2

i ×+ FCOVj

,λβ FCOVi (8)

CONTji

CONTji

CONTji ,,, εδγ +++

Species-specific slope and intercept parameters in the link functions for the observation and

ecological model components were drawn as random effects from a normal distribution defined by

hyperparameters for the overall mean and standard deviation for the metacommunity. The land-use

filter Ωj and spatial and temporal random effects terms were estimated separately for each species.

Considering species as random effects is the key advantage of this class of multi-species hierarchical

models, and allows for inference about the rarest and most infrequently detected members of a

community by “borrowing strength” from the rest of the data. Necessarily, this involves making a

trade-off for well-sampled species which could have been modelled independently, given that

parameter estimates for species are pulled towards the overall community mean by way of

“shrinkage”. One way to mitigate this is to divide a community into sub-groups of ecologically-


150

similar species which are hypothesised to respond to the modelled covariates in a similar way

(Dorazio et al., 2011). In our case, we did not have any strong a priori justifications for sub-groups of

species, and modelled all species together.

Hierarchical models with random effects, such as the model outlined here, are analytically and

computationally intractable in a maximum likelihood framework (Dorazio et al., 2011), but MCMC

methods, applied within a Bayesian inference framework, can now be readily applied in such cases.

We wrote the model in the BUGS (Bayesian inference Using Gibbs Sampling) language (provided in

Appendix C) and used JAGS (Just Another Gibbs Sampler) version 3.4.0 (Plummer, 2013), called

using the runjags 1.2.1-0 package (Denwood, 2015) in R, to obtain the MCMC samples of the joint

posterior distribution. We used standard uninformative priors for all parameters (Kéry & Schaub,

2012): a flat normal prior for intercept and slope parameters, and a wide uniform prior for variance

parameters (Appendix C). As an exception to this, we used a half-Cauchy prior for the temporal

random effects, which is recommended for random effects with fewer than five levels (Gelman,

2006). All continuous covariates were centred to a mean of zero and scaled to unit standard deviation

prior to analysis, to assist with model convergence (Kéry & Schaub, 2012). We ran three MCMC

chains for 200,000 iterations each, thinning by a factor of 10 and discarding the first 80,000 iterations

as “burn-in”. To judge convergence, we visually inspected chains for good mixing and non-

directionality and checked that the Gelman-Rubin statistics (Gelman & Rubin, 1992) for all

parameters were < 1.1. We also assessed the goodness-of-fit of our model with a posterior predictive

check, based on the χ2 discrepancy statistic for binomial data (Kéry & Schaub, 2012; Tobler et al.,

2015).

Derived parameters, which carry over all sources of estimation uncertainty, are relatively

straightforward to calculate in a Bayesian framework using the marginal posterior distributions of

parameters. In particular, we calculated the mean abundance, summed abundance and summed

biomass for each of the ecological trait groups we were interested in. We also estimated the


151

probability of each species and trait-defined group declining at each transition along the categorical

land-use gradient (i.e. from old-growth forest to logged forest, and from logged forest to oil palm

plantation) by calculating the proportion of times abundance declined across the MCMC iterations.

Functional effects groups based on diet were mutually-inclusive groups (i.e. all species implicated in

leaf-eating, fruit-eating, seed-eating, bark-eating, root-eating, fungi-eating, invertebrate predation or

vertebrate predation), and we focussed on the summed local biomass of each group in this case, rather

than abundance. We used the median to summarise across MCMC iterations in all cases, owing to the

skew that was evident in the marginal posterior distributions of some parameters.

Species body masses were obtained from the PanTHERIA database (Jones et al., 2009) and Payne et

al. (2007). If a range of masses was provided, we took the mid-point. Small mammals were defined as

species never exceeding 1 kg. Species of high conservation concern were defined as those listed as

threatened (Critically Endangered, Endangered or Vulnerable) or ‘data deficient’ (n = 4) on the IUCN

(2014) Red List. Information on mammal diets was based on Langham (1983), Emmons (2000),

Payne et al. (2007) and, for scavenging records, Edwards et al. (2014), as well as our own field

observations.

5. 3. Results

A total of 4,381 live trap captures and 15,148 independent camera trap captures were made, for 57

mammal species. After reducing these data into detections or non-detections within sampling

occasions (17,025 live trap occasions and 5,428 camera trap occasions), this translated into 4,284 live

trap detections of 23 species, and 7,772 camera trap detections of 53 species (19 species were

common to both sampling methods). We also had a limited number of captures (mostly ≤ 2 per

species) for nine additional mammal species which we classified as obligate arboreal species (listed in

Appendix B, Table B2) and which we did not include in our abundance models. We did not, however,

exclude all rare species, with four terrestrial or scansorial species present in our models which were


152

captured fewer than five times (Hose’s civet Diplogale hosei, Malay weasel Mustela nudipes, plantain

squirrel Callosciurus notatus and western tarsier Cephalopachus bancanus).

The posterior predictive checks indicated that our models were an adequate reflection of the

ecological and observational processes generating the data (categorical land-use model: Bayesian p-

value, )Pr( 22simulatedobserved χχ > = 0.66, lack-of-fit statistic = 1.08; continuous metric model: p = 0.64,

lack-of-fit = 1.08). We also calculated Bayesian p-values for each species within our model

(Appendix B, Table B1); this indicated poor fit for three of the 57 species (red spiny rat Maxomys

surifer, brown spiny rat Maxomys rajah and long-tailed giant rat Leopoldamys sabanus), suggesting

caution is needed with respect to inferences about the absolute abundances of these species. In

particular, we suspect the poor fit in these cases is due to substantial heterogeneity in the individual

detection probability r, likely induced by the fine-scale patterns of space-use in these species. By

analogy with heterogeneity in a capture-recapture framework, we would expect this to exert negative

bias (Otis et al., 1978), and our estimates are therefore likely to be conservative for these species.

Based on the hyperparameter estimates, mean abundance across the mammal community was

marginally higher in logged forest compared to old-growth forest (Pr )0( , >LULoggedλβ = 0.81), but much

lower in oil palm compared to either of the two forest land-uses (Pr =< )0( ,LUOilPalmλβ 1.00). These

overall trends, however, belie substantial differences among species groups (Fig. 2) and among

individual species (Fig. 3; Appendix A, Fig. A5). From old-growth to logged forest, large mammals

exhibited a modest (11%) increase in mean abundance, but small mammals increased substantially

(by 169%). The mean abundance of high conservation concern species was similar, or increased, in

logged forest compared to old-growth forest (Fig. 2), but dropped precipitously (by 83%) in oil palm.

In contrast, the abundance of low conservation concern species was largely robust to the land-use

gradient, whilst invasive species increased substantially along the land-use gradient (Fig. 2). The

mean abundance of all trophic guilds except frugivores increased from old-growth to logged forest,

whilst the abundance of all guilds except carnivores declined in oil palm (Fig. 2). The trends in


153

summed abundances and biomasses for each trait-defined group were largely similar to those for

mean abundance (Appendix A, Fig. A1). However, the relatively modest abundance increases in

herbivores (19%) and threatened species (26%) in logged forest were much more prominent in terms

of biomass (140% and 108%, respectively), due to increases in large-bodied species in these groups

(e.g. sambar deer Rusa unicolor, banteng and Asian elephant). Similarly, large changes in mean

abundance apparent in omnivores (100%) were not as strong in terms of biomass (51%), because

these abundance changes were partly driven by small-bodied murid rodent species. The summed

biomasses of functional effects groups were maintained, or increased, from old-growth to logged

forest, but from forest to oil palm substantial declines were evident in all cases except vertebrate

predation (Fig. 4). For individual species, the mean number of individuals per sampling point was

relatively low (< 1; Appendix A, Fig. A5) in most cases, reflecting the patchy occurrence of species in

space. There was large uncertainty about abundance for the rarest species, which properly reflects the

information present in the data and underlines the importance of accounting for all sources of

uncertainty during modelling (observational as well as ecological), but abundance estimates across

land-use were nonetheless biologically plausible in all cases.

Figure 2. Mean species abundance across land-use categories, partitioned by ecological response groups defined by body size (large and small mammals), conservation status (threatened, non-threatened), native status (only invasives shown) and trophic guild (five mutually-exclusive feeding guilds). Error bars indicate 90% credible intervals.


154

Figure 3. Probabilities of a decline in abundance (calculated across MCMC iterations) from old-growth to logged forest (orange) and from logged forest to oil palm (purple), for each ecological response group and each species. We did not calculate the probability of decline from logged forest to oil palm for four species which were not recorded in logged forest.


155

Figure 4. Summed local biomass across land-use categories, partitioned by functional effects species groups based on diet. Each group is mutually-inclusive such that all species engaging in the given function are included in the group, meaning that some species are included in multiple groups. Error bars indicate 90% credible intervals.

Abundance responses were broadly negative for AGB (Pr AGB,( λβ < 0) = 0.96) and broadly positive

for forest cover (Pr FCOV,( λβ > 0) = 1.00; Figs. 5-6; Appendix A, Figs. A2-3). The effect of forest

cover was stronger, albeit more uncertain, than the effect of AGB (standardised hyperparameter

estimates with 90% credible intervals: AGB,λβ = -0.18, 90% CI: -0.01 – -0.35; 2,AGBλβ = -0.10, 90%

CI: -0.003 – -0.22; FCOV,λβ = 0.68, 90% CI: 0.38 – 0.98), and this was also true at the level of

individual species in most cases (Appendix A, Figs. A6-7). There was weak evidence, overall, of

unimodal responses to AGB (2,AGBλβ overlapped zero, Pr

2,( AGBλβ < 0) = 0.90), and this was also

generally the case for individual species, although some species (e.g. long-tailed giant rat, Low’s


156

squirrel Sundasciurus lowii, orangutan Pongo pygmaeus, plain treeshrew Tupaia longipes and sambar

deer) exhibited evidence of a threshold response in which increases in abundance with decreasing

AGB were not maintained below ~ 90 Mg/ha (Appendix A, Fig. A6). All ecological response trait

groups showed increased mean abundance under the decreases in AGB which accompany logging,

with the exception of frugivores (Fig. 5A). However, the increases were most stark in omnivores,

small mammals and non-natives (Fig. 5A), all groups which are dominated by murid rodent species.

All ecological response trait groups showed large abundance reductions in response to reduced forest

cover, except carnivores and non-natives (Fig. 6A). In fact, mean carnivore abundance exhibited a

unimodal response curve, being lowest at ~ 70% forest cover. This reflected a shift from native, forest

predators such as the yellow-throated marten Martes flavigula and Sunda clouded leopard Neofelis

diardi to native and non-native carnivores tolerant of more open habitats, such as the leopard cat

Prionailurus bengalensis, Malay civet Viverra tangalunga and domestic dog (Appendix A, Fig. A7).

For the continuous metrics, we also calculated the mean across species of the percentage change in

abundance along the land-use gradient (effectively giving each species equal weight, irrespective of

their absolute abundance). The mean percentage changes exhibited similar trends to the mean

abundance of each species group (Figs. 5B and 6B), except there was stronger evidence in some

groups of lower rates of abundance increases, or even decreases in abundance, at lower values of

AGB (< 90 Mg/ha), and there was no evidence of a recovery in carnivore abundance at low forest

cover (because, for mean abundance, responses were driven largely by the abundance of three

carnivores in particular: leopard cat, Malay civet and domestic dog). The biomass responses of the

dietary functional effects groups to declines in AGB were largely positive, whilst they were largely

negative under declines in forest cover, except in the case of vertebrate predation (Appendix A, Fig.

A4).


157

Figure 5. Mean abundance (A) and mean percentage change (B) across species as a function of above-ground live tree biomass in a given local landscape, partitioned by ecological response groups defined by body size, conservation status, native status (only invasives shown) and trophic guild. Percentage change refers to the change relative to the abundance at AGB values typical of intact forest (400 Mg/ha). Forest cover was fixed at 100%. 90% credible intervals (in grey) indicate uncertainty surrounding median estimates across MCMC iterations (red line). Greater uncertainty in panel B reflects the uncertainty about species-specific responses.


158

Figure 6. Mean abundance (A) and mean percentage change (B) across species as a function of forest cover in a given local landscape, partitioned by ecological response groups defined by body size, conservation status, native status (only invasives shown) and trophic guild. Percentage change refers to the change in abundance as forest cover decreases from 100%. Above-ground live tree biomass was fixed at the average across oil palm locations. 90% credible intervals (in grey) indicate uncertainty surrounding median estimates across MCMC iterations (red line). Greater uncertainty in panel B reflects the uncertainty in individual species responses.


159

Individual detection probabilities were, based on the overall estimated hyperparameters, higher using

camera traps rCameraα( = 0.023 for 5 camera trap nights, 90% CI: 0.013 – 0.038) than live traps r

Liveα( =

0.017 for 2 live trap nights, 90% CI: 0.006 – 0.043). Whilst this was true overall, live traps resulted in

higher detectability for nine small mammal species (Appendix A, Fig. A8). Detection probabilities

were, overall, marginally highest in old-growth forest, followed by logged forest LUrLogged

,(β = -0.13,

90% CI: -0.32 – 0.05; Pr )0( , <LUrLoggedβ = 0.89) and oil palm LUr

OilPalm,(β = -0.34, 90% CI: -1.38 – 0.30; Pr

)0( , <LUrOilPalmβ = 0.80). The effect of fine-scale habitat disturbance on individual detection probabilities

was positive overall (i.e. negative for increases in the habitat score; HDr ,β = -0.09, 90% CI: -0.01 – -

0.17) and there was some evidence of a unimodal response 2,( HDrβ = -0.03, 90% CI: -0.06 – 0.01; Pr

2,( HDrβ < 0) = 0.90). However, there was substantial variation in the response of individual species

(Appendix A, Fig. A9), reflecting contrasting patterns of fine-scale habitat-use. In addition, the

absolute detectability of different species showed marked variation, as expected from the differences

across species in body size, movement rates and degree of scansorial behaviour. Correcting for

detectability resulted in substantial alterations in the abundance ranks of some species, compared to

ranks based only on the number of independent captures (Appendix A, Fig. A10). This was especially

true in old-growth forest, and in particular for the yellow muntjac Muntiacus atherodes, brown spiny

rat, lesser mouser-deer Tragulus kanchil, greater mouse-deer T. napu and thick-spined porcupine

Hystrix crassispinis. Nonetheless, abundance ranks calculated using model estimates and the number

of independent captures were strongly correlated (rOld-growth = 0.79, rLogged = 0.85, rOilPalm = 0.84) and

ranks were not significantly different (Wilcoxon signed-rank test, old-growth forest: V = 950, p =

0.33; logged forest: V = 708, p = 0.87; oil palm: V = 756, p = 0.91). This suggests that the uncorrected

number of captures provides a similar snapshot of community structure to that revealed by local

abundance. However, this may not be the case for comparisons over space or time if detectability also

varies.


160

5. 4. Discussion

Mammalian abundance was conserved, or increased, from old-growth to logged forest overall, whilst

it declined substantially from forest to oil palm plantations. This was true of mean mammal

abundance, summed abundance and total biomass. Mammalian abundance (mean and summed) and

biomass responses to decreases in local landscape AGB due to logging were positive (and weakly

unimodal), but were strongly negative for decreases in local landscape forest cover. Although this is

the first time, to our knowledge, that a robust assessment of Southeast Asian mammal abundance has

been made along the principal land-use gradient in the region, these findings at the community level

broadly agree with the consensus from past studies of abundance responses (all of which used

detection frequencies) in small mammal communities (Wells et al., 2007; Bernard et al., 2009) and

large mammal communities (Imai et al., 2009; McShea et al., 2009; Samejima et al., 2012).

There are strong a priori reasons to expect these responses in mammals. Selective logging in

Southeast Asia primarily targets members of the Dipterocarpaceae, which few mammals make use of

to any significant degree in their diet (Meijaard et al., 2005), whilst at the same time logging creates

canopy openings which stimulate the growth of understorey vegetation, thereby increasing the food

resources available to terrestrial grazer-browsers and frugivores (Davies et al., 2001), and increasing

the availability of favourable microhabitats for small mammals (Cusack et al., 2015). In contrast, the

conversion of forest to oil palm likely removes the food resources and foraging or resting

microhabitats suitable for most mammals, especially in intensively-managed industrial plantations,

which typically limit the growth of most non-commercial plant species (Luskin & Potts, 2011).

More broadly across other taxonomic groups in Southeast Asia, few studies have investigated

abundance responses (and rarely using robust estimation methods), but apparent trends have usually

been similar to our results for mammals. Abundance in logged areas has usually been found to be

maintained at a community level (ants: Luke et al., 2014; birds: Edwards et al., 2011; butterflies:

Willott et al., 2000; dung beetles: Slade et al., 2011, isopods: Hassall et al., 2006), but substantially

declines in oil palm plantations (e.g. birds: Edwards et al., 2010; arthropods: Turner & Foster, 2008).


161

In some cases, however, apparent declines in abundance have also been noted for some groups in

logged areas (e.g. birds: Edwards et al., 2009; termites: Jones et al., 2003 and Luke et al., 2014), as

well as increases in oil palm plantations (e.g. isopods: Hassall et al., 2006; dung beetles: Chung et al.,

2000), contrary to our findings for mammals. It is difficult to reconcile these reported changes with

the apparent consensus from other studies in part because detectability has not been accounted for, as

the researchers note in some cases (Edwards et al., 2009), but there are also mechanistic explanations

for the atypical responses seen in some specialised groups (e.g. Jones et al., 2003; Hassall et al.,

2006).

The evidence overall, taken together with our findings for mammals, increasingly supports the view

that large, contiguous areas of logged forest in Southeast Asia not only conserve similar levels of

species richness to old-growth forest (Edwards et al., 2014; Chapter 3), but they also conserve the

community-level abundance of many groups. This adds further emphasis to the calls for increasing

recognition of logged and degraded forests as an essential part of the conservation estate (Wilson et

al., 2010; Berry et al., 2010; Edwards & Laurance, 2011; Edwards et al., 2011; Struebig et al., 2015a).

Remaining old-growth forests, which are undoubtedly the absolute highest priority for conservation

(Gibson et al., 2011), have largely already been gazetted in protected areas and degraded forests have

been the primary source of new land for expanding plantations in the region (Gunarso et al., 2013).

Degraded forests could represent a relatively low opportunity-cost option for conservation, given that

much of their timber value has been extracted (Wilson et al., 2010; Fisher et al., 2011b; Edwards et

al., 2014a). Our study is also one of the few studies which has been undertaken in repeatedly-logged

forests (Edwards et al., 2011, 2014a; Woodcock et al., 2011; Struebig et al., 2013), and the finding

that mammal community richness and abundance is maintained even in these heavily-degraded

forests, further strengthens the argument for low-cost conservation in such areas. Our findings with

respect to the community richness (Chapter 3) and abundance of mammals in oil palm also support

previous assertions of the very limited conservation role played by this land-use type (Edwards et al.,

2010b, 2014a).


162

By assessing almost the whole non-volant mammal community, we were also able to go further than

previous studies in the region and assess the abundance responses of important sub-groups of

mammals defined by their traits, as well as the potential functional effects of changes in abundance

across the community. We found that, for almost all response trait groups, logged forests retained

similar or higher abundances (mean and summed across species) and summed biomasses compared to

old-growth forest. This was also true for the biomasses of functional effects groups we examined, a

finding which is consistent with other evidence that the functional role of vertebrates increases in

logged relative to old-growth forests (Ewers et al., 2015). Moreover, these group-level increases were

largely maintained even at very low levels of AGB in a local landscape, indicative of high levels of

logging disturbance. On the other hand, our results indicate that conversion to oil palm, and

reductions in forest cover, cause declines in the abundance (mean and summed) and summed biomass

of almost all the trait-defined sub-groups we examined (not carnivores and invasives), as well as in

the biomasses of almost all the functional effects groups (not vertebrate predation).

The increase in the mean and summed abundance of small mammals was substantial in logged forest

(and also in response to declining AGB), similar to findings elsewhere in tropical forests (Isabirye-

Basuta & Kasenene, 1987; Lambert et al., 2006), and was much more dramatic than in large

mammals. We suggest that this is driven, in part, by an expansion in the area covered by

microhabitats which small mammals, in particular murid rodents, use for foraging and concealment

from predators (Cusack et al., 2015). In addition, populations of murid rodents in undisturbed

dipterocarp forests of the region are likely constrained by the supra-annual pulses in fruit and seed

availability which are a defining feature of these systems (Curran & Leighton, 2000). However, we

also note that the biomass increase in large mammals was much greater than the modest change in

mean abundance suggested, and was two orders of magnitude larger in absolute terms compared to

the increase in small mammals. Much of this increase was driven by increases in the mean abundance,

and average body size, of the herbivore trophic guild. This may have much greater implications for

ecosystem functions, such as seedling recruitment rates (Howlett & Davidson, 2003; Harrison et al.,

2013) and nutrient cycling (Wardle & Bardgett, 2004; Nichols et al., 2008; Doughty et al., 2013) than


163

the increased abundance of small mammals, even though small mammals may be significant seed

predators in these forests on a per capita basis (Blate et al., 1998; Wells & Bagchi, 2005).

Among other trophic guilds, we found that omnivores increased markedly in logged compared to old-

growth forest. Wide dietary breadth has often been proposed as a trait conferring increased resistance

to extinction (Laurance, 1991; McKinney, 1997), but support for its advantages under disturbance has

often been mixed (e.g. Posa & Sodhi, 2006; Gray et al., 2007; Rickart et al., 2011), perhaps because

dietary flexibility, rather than breadth per se, may be the more critical trait. Dietary flexibility is more

difficult to measure, and is poorly known for Bornean mammals, but we would expect that most of the

omnivorous species in our dataset (i.e. murid rodents, bearded pig and sun bear Helarctos malayanus)

also exhibit high dietary flexibility (Caldecott et al., 1993; Fredriksson et al., 2006). For insectivores,

some studies on birds have shown a disproportionate sensitivity to logging (Lambert & Collar, 2002;

Edwards et al., 2009), and disturbance more generally (Gray et al., 2007), which we did not find for

mammals. The abundance responses of insects, and invertebrates more generally, to logging is poorly

known in Southeast Asia, but we note that, at our study sites, invertebrate biomass is apparently

higher in logged forest compared to old-growth forest (Ewers et al., 2015), potentially indicating that

food resources for insectivorous mammals are conserved. For carnivores, we would expect numerical

responses to the abundance of vertebrate prey species. Most of the carnivores we studied, and in

particular the felids, focus on mammal prey such as murid rodents (Grassman et al., 2005; Rajaratnam

et al., 2007; Shehzad et al., 2012), which we have shown here are conserved in logged forests.

Frugivory is a trait which has often been associated with an increased susceptibility to disturbance

(Johns & Skorupa, 1987; Gray et al., 2007), but it is not clear whether logging causes a consistent

decline in fruit availability across species or not (Wong, 1986; Heydon & Bulloh, 1997; Munshi-

South et al., 2007). Certainly, some key fruiting resources such as hemi-epiphytic figs are often much

reduced after logging (Lambert, 1991), but the availability of small fruit on lianas and understorey

shrubs might increase in gaps or along edges (Davies et al., 2001; Meijaard et al., 2005). Frugivores

exhibited no changes in abundance from old-growth to logged forest, but modelling using the

continuous AGB metric revealed a modest decline in abundance with increasing logging disturbance.


164

We note, however, that the summed biomass of all species engaging in fruit-eating did not decline,

suggesting frugivory as a function may be resilient to logging, even though specialist frugivores do

not fare as well as other groups. Finally, of crucial conservation relevance, we found that the

abundance of high conservation concern species was retained in logged forests, and that this group

was resilient even to high intensities of logging (low levels of AGB) in a given landscape. We should

emphasise, however, that this does not necessarily mean that high conservation concern species would

persist in hypothetical landscapes consisting of homogeneously low AGB areas; AGB values refer to

an average over a local landscape, and will contain some patches of less intensively logged forest, as

well as areas that are heavily-disturbed.

Overall, we have shown that no trait-defined group suffered substantial losses in logged forest, and

that the functional effects we examined were also resilient to logging. These findings further

strengthen our arguments about the high conservation importance that logged forests, even those that

are repeatedly-logged, should be assigned in regional- or landscape-level land-use planning exercises.

In a positive move in this direction, palm oil producers, traders and buyers are increasingly

recognising the reputational risk of being associated with deforestation in their supply chain, and

many have now made commitments to help curtail the conversion of forest to oil palm. Applying this

in practice requires a consistent definition of forested land, as opposed to degraded scrub land, and the

possibility of using a carbon-based definition has arisen repeatedly during these discussions, in

particular a threshold of ≥ 35 MgC/ha to define High Carbon Stock (HCS) forest (Greenpeace, 2013;

Poynton, 2014). This is equivalent to an AGB of ~ 74 Mg/ha (assuming that carbon constitutes 47%

of live tree biomass; Martin & Thomas, 2011), which could, if our findings apply more broadly in the

region, yield major conservation benefits for mammals over the business-as-usual.

The biggest caveat to this conclusion is that bushmeat hunting, which often accompanies logging

(Bennett & Gumal, 2001), is strictly controlled. Brodie et al. (2015) investigated the effects of logging

on large mammal occupancy in the context of much higher levels of hunting than was present in our

study (~ 3% of days at a given camera trap location with ≥ 1 hunter detections, compared to < 0.001


165

% across forest locations in this study), finding strongly negative effects of recent logging (≤ 10 years

prior to sampling, similar to our study), possibly due to the synergistic effects of logging and hunting

in combination. We emphasise that the huge conservation potential of HCS forests for mammals will

only be realized with additional investment in protection, as well as investment in education and

livelihoods for communities in the vicinity of HCS forests.

Our conclusions concerning the conservation potential of oil palm are less optimistic. Although the

plantations in which we sampled may represent something of a best-case scenario for oil palm, with

relatively high levels of landscape forest cover and strict enforcement of hunting bans, our modelling

of mammal abundance as a function of forest cover indicates only a very limited potential for

conservation gains by attempting a land-sharing, ‘wildlife-friendly’ approach (e.g. Bhagwat & Willis,

2008; Koh, 2008; Koh et al., 2009) to this land-use. Increases in local landscape forest cover from 0 to

30%, the likely range which could realistically be manipulated in oil palm landscapes, resulted in very

limited abundance increases across species groups and across most individual species within the oil

palm crop, suggesting only a limited degree of ‘spill-over’ from remnant forest patches. Among

trophic guilds, only carnivores showed some resilience to decreases in forest cover, but this was in

large part driven by increases in free-ranging domestic dogs, which are considered a detrimental

invasive species across Asian landscapes (Azhar et al., 2013; Hughes & Macdonald, 2013). We did

not sample remnant forest fragments within the oil palm, but it is unlikely that the abundance and

richness of mammals in these areas would approach that of contiguous forest (Bernard et al., 2014),

even if individuals present in the oil palm crop itself were also counted. Overall, this indicates that a

land-sparing approach might better serve mammal conservation in the region, in which companies

intending to invest in on-site conservation, for example by retaining small forest patches in their

concession, are instead encouraged to invest in the conservation of equivalent, or larger, areas of

contiguous forest off-site, for example by way of a regional ‘bio-bank’ (Edwards et al., 2010b). As a

caveat to this, there may be the potential for ‘win-win’ solutions for both conservation and oil palm

yield, such as in the bio-control of pest species, and in this case on-site conservation activities should

be encouraged (Foster et al., 2011). In particular, the high abundance of leopard cats we found within


166

the oil palm crop, and the low abundance of murid rodents, suggests a likely role for this species in

bio-control (Silmi et al., 2013), though the necessary habitat conditions for the persistence of this

species remain inadequately known (Rajaratnam et al., 2007; Jennings et al., 2015).

In common with studies of other taxonomic groups in oil palm (Senior et al., 2012), we found that a

small number of species became dominant in the plantations (even reaching abundances exceeding

those in natural forest), and that the resilient species could not necessarily be predicted from

abundances in natural forest or the abundance responses to logging. These ‘winner’ species were

primarily carnivorous species, preying upon murid rodents (Whitehead’s rat Maxomys whiteheadi and

black rat Rattus rattus), herptiles and birds, and possibly benefitting from increased hunting success in

the open oil palm habitats (Rajaratnam et al., 2007). This underlines the often idiosyncratic nature of

abundance responses to intense disturbance by individual species, and supports the need for studies

which are designed to address specific conservation and management questions at the local scale. In

contrast to the strong metacommunity filter that oil palm represented, logging resulted in a more

continuous gradient of abundance responses across species (Fig. 3), perhaps reflective of the fact that

selective logging, to a greater or lesser extent, exaggerates the natural dynamics of gap creation and

succession. Species of globally high conservation concern which were most affected by logging

included both of the insectivorous civets, the banded civet Hemigalus derbyanus and Hose’s civet, as

well the tufted ground squirrel Rheithrosciurus macrotis and binturong Arctictis binturong. Our

metacommunity model also provided information on the patterns of fine-scale habitat-use across

species, revealing that some species – including invasives (Polynesian rat Rattus exulans, black rat

and domestic dog), macaque species and grazing herbivores (Asian elephant, banteng and sambar

deer) – preferred to use highly-disturbed log-landing areas and roads, whilst others actively avoided

these features – including treeshrews (Tupaia tana and T. longipes), some murid rodents (Maxomys

surifer and Leopoldamys sabanus) and the bay cat Pardofelis badia (Appendix A, Fig. A9).

Using a novel hierarchical model for a Southeast Asian mammal metacommunity, applied to one of

the largest mammal datasets across land-use to date, we have shed light on the contrasting abundance


167

responses to logging and conversion to oil palm, as well as the responses to the continuous metrics of

logging intensity and forest cover loss, both of which are directly relevant to conservation and

management at local scales. Our approach, which can integrate data from multiple sources, could be

applied to other taxonomic groups, and allow for more robust inferences with respect to the impacts of

land-use change in Southeast Asia. We anticipate that this will help resolve apparently contradictory

responses reported by previous studies, by properly accounting for all sources of uncertainty

surrounding estimates, and pave the way for improved biodiversity forecasting and more effective

decision-making in the face of biodiversity trade-offs across land-use.


168

Appendices

Appendix A – Supplementary figures

Figure A1. Summed abundance (A) and local biomass (B) as a function of land-use categories, partitioned by ecological response groups defined by body size, conservation status, native status (only invasives shown) and trophic guild. Error bars indicate 90% credible intervals.


169

Figure A2. Summed abundance (A) and local biomass (B) as a function of the average above-ground live tree biomass within a given local landscape, with forest cover fixed at 100%. Estimates were derived for ecological response groups defined by body size, conservation status, native status (only invasives shown) and trophic guild. 90% credible intervals (in grey) indicate uncertainty surrounding median estimates across MCMC iterations (red line).


170

Figure A3. Summed abundance (A) and local biomass (B) as a function of forest cover within a given local landscape (defined using a 500 m radius buffer). Above-ground live tree biomass was fixed at the average across oil palm locations. Estimates were derived for ecological response groups defined by body size, conservation status, native status (only invasives shown) and trophic guild. 90% credible intervals (in grey) indicate uncertainty surrounding median estimates across MCMC iterations (red line).


171

Figure A4. Summed local biomass responses to above-ground live tree biomass (A) and forest cover (B) in a given landscape (with other parameters fixed in each case), for functional effects groups based on diet. Each group is mutually-inclusive, so that all species engaging in the given function are included in the group. 90% credible intervals (in grey) indicate uncertainty surrounding median estimates across MCMC iterations (red line).


172

Figure A5. Abundance estimates for each of 57 mammal species across categorical land-uses. Error bars indicate 90% credible intervals.


173

Figure A6. Abundance for each of 57 mammal species, as a function of above-ground live tree biomass in a given landscape, with forest cover fixed at 100%. 90% credible intervals (in grey) indicate uncertainty surrounding median estimates across MCMC iterations (red line).


174

Figure A7. Abundance for each of 57 mammal species, as a function of local landscape forest cover, with above-ground live tree biomass fixed at the average for oil palm. 90% credible intervals (in grey) indicate uncertainty surrounding median estimates across MCMC iterations (red line).


175

Figure A8. Individual detection probability estimates across species and land-uses, for two sampling methods (live- and camera-trapping). Probabilities are per sampling occasion (two and five trap nights for live-trapping and camera-trapping, respectively). Error bars indicate 90% credible intervals. Estimates were made at a habitat score of 3.


176

Figure A9. Individual detection probability for each species as a function of the habitat score at a sampling point, indicative of fine-scale disturbance (low habitat scores indicate high disturbance), and for each of the land-use types. Estimates relate to detectability using camera traps (per five trap nights of sampling).


177


178

Figure A10. Comparison of relative abundance across land-use types for a hierarchical state-space model and a naïve index based on detection frequencies. Relative abundance was calculated separately for the modelled estimates and the detection frequencies, and was obtained by standardising values for each species according to the maximum value obtained for any species in any land-use. Species are ranked by their modelled abundance estimates.


179

Appendix B – Supplementary tables

Table B1. Bayesian p-values for 57 mammal species modelled using a hierarchical state-space model of abundance. Values > 0.95 or < 0.05 are indicative of poor fit.

Order Species Common name Bayesian p-

value Significant lack of fit

Erinaceomorpha

Echinosorex gymnura Moon rat 0.593

Pholidota

Manis javanica Sunda pangolin 0.444

Carnivora

Prionailurus bengalensis Leopard cat 0.359

Pardofelis badia Bay cat 0.302

Pardofelis marmorata Marbled cat 0.270

Neofelis diardi Sunda clouded leopard 0.254

Diplogale hosei Hose's civet 0.456

Hemigalus derbyanus Banded civet 0.759

Paguma larvata Masked palm civet 0.425

Paradoxurus hermaphroditus Common palm civet 0.447

Arctictis binturong Binturong 0.393

Viverra tangalunga Malay civet 0.439

Prionodon linsang Banded linsang 0.823

Herpestes semitorquatus Collared mongoose 0.639

Herpestes brachyurus Short-tailed mongoose 0.631

Canis familiaris Domestic dog 0.471

Helarctos malayanus Sun bear 0.718

Mydaus javanensis Sunda stink badger 0.725

Martes flavigula Yellow-throated marten 0.424

Mustela nudipes Malay weasel 0.476

Aonyx cinereus Oriental small-clawed otter 0.343

Cetartiodactyla

Sus barbatus Bearded pig 0.493

Tragulus napu Greater mouse-deer 0.857

Tragulus kanchil Lesser mouse-deer 0.941

Muntiacus atherodes Bornean yellow muntjac 0.762

Muntiacus muntjak Red muntjac 0.544

Rusa unicolor Sambar deer 0.587

Bos javanicus Banteng 0.452

Scandentia

Tupaia minor Lesser treeshrew 0.406

Tupaia gracilis Slender treeshrew 0.532

Tupaia longipes Plain treeshrew 0.648

Tupaia tana Large treeshrew 0.507

Tupaia dorsalis Striped treeshrew 0.529

Primates

Cephalopachus bancanus Western tarsier 0.382

Macaca fascicularis Long-tailed macaque 0.406

Macaca nemestrina Southern pig-tailed macaque 0.524

Pongo pygmaeus Orangutan 0.339

Rodentia


180

Lariscus hosei Four-striped ground squirrel 0.545

Callosciurus notatus Plantain squirrel 0.604

Callosciurus adamsi Ear-spot squirrel 0.594

Sundasciurus lowi Low's squirrel 0.591

Sundasciurus hippurus Horse-tailed squirrel 0.556

Rheithrosciurus macrotis Tufted ground squirrel 0.578

Trichys fasciculata Long-tailed porcupine 0.864

Hystrix brachyura Malay porcupine 0.740

Hystrix crassispinis Thick-spined porcupine 0.473

Leopoldamys sabanus Long-tailed giant rat 0.980 *

Sundamys muelleri Müller's rat 0.611

Niviventer cremoriventer Dark-tailed tree rat 0.251

Maxomys whiteheadi Whitehead's rat 0.793

Maxomys surifer Red spiny rat 0.998 *

Maxomys rajah Brown spiny rat 0.968 *

Maxomys baeodon Small spiny rat 0.351

Maxomys ochraceiventer Chestnut-bellied spiny rat 0.864

Rattus exulans Polynesian rat 0.278

Rattus rattus Black rat 0.553

Proboscidea

Elephas maximus Asian elephant 0.479

Table B2. List of obligate arboreal species which are poorly sampled using live traps and camera traps, and were excluded from modelling. Captures represent the sum of the number of independent captures (as defined in the main text) from both live- and camera-trapping. Order Species Common name Captures

Scandentia

Ptilocercus lowii Pen-tailed treeshrew 1

Primates

Presbytis hosei Hose's langur 1

Presbytis rubicunda Maroon langur 12

Hylobates muelleri Bornean gibbon 1

Rodentia

Callosciurus prevostii Prevost's squirrel 2

Exilisciurus exilis Least pygmy squirrel 1

Sundasciurus tenuis Slender squirrel 2

Sundasciurus brookei Brooke's squirrel 1

Aeromys thomasi Thomas's flying squirrel 1


181

Appendix C – BUGS (Bayesian inference Using Gibbs Sampling) code used to obtain Markov Chain

Monte Carlo (MCMC) samples of the joint posterior

C. 1. Representation of the hierarchical model, including priors, used to obtain abundance

estimates as a function of continuous metrics of land-use change (the categorical land-use

model, not shown, is similar).

# A multi-species matrix of detection histories constitutes the main data used by the model. These data are 5-dimensional, i.e. site x species x occasion x trap-type x session, and also highly unbalanced (not all 5-way combinations were observed, giving rise to a large number of NA values). Further complications arise because the model is required, at certain points, to split sites according to land-use, and split species according to two different criteria (whether they can be detected using both sampling methods or not, and whether they are group-living). # Here, the data are ‘flattened’ to two dimensions, i.e. ‘observation’ x species, giving the matrix D. Each ‘observation’ (row) of D constitutes a different site x trap-type x session combination. Every row contains data, and no NA values are used. The ‘missing’ dimensions to the data are then re-created virtually, principally by using nested indexing (e.g. plot.counter[]) and offsets (e.g. site.startstop.i[]). # Note that different sessions are ‘stacked’ in D and treated as if they were separate sites, as is common in occupancy analyses when occupancy dynamics across sessions are not being modelled explicitly. # The following data inputs must be provided to the model: # -------------------------------------------------------- # D = a 2-D ‘observation’ x species matrix of detection counts, where each row constitutes a different site x trap-type x session combination # occasions = a vector providing the number of sampling occasions corresponding to each row of D # traptype = a vector identifying the traptype used for each row of D # avail = a 2-D matrix indicating the availability of each species to the sampling method used in each ‘observation’ row of D # Log/OP = indicator variables, identifying logged forest and oil palm sites # AGB/AGB2 = 1st- and 2nd-order polynomials of the above-ground live tree biomass (AGB) variable # ForestCov = percent forest cover variable # plot.counter/block.counter/session.counter = vectors identifying which plot, block and session each site corresponds to # HabitatScore/HabitatScore2 = 1st- and 2nd-order polynomials of the habitat disturbance variable # landuse.i = a vector, used as an offset, containing the beginning and end site indices for each land-use # site.startstop.i = a 2-D matrix containing, for each site, the beginning and end row indices of D # The following dimensions must also be provided as inputs to the model: # ---------------------------------------------------------------------- # nbothtrapspp = number of species which are detectable using both sampling methods # ntraptype = number of sampling methods or trap-types (here, ntraptype = 2) # nctrapspp = number of species detectable with camera traps only # ngroupspp = number of species which are group-living # nlanduse = number of land-uses # nplot/nblock/nsession = number of plots, blocks and sessions # Parameters named “.r” relate to the observation model (r = individual detection probability). # Parameters named “.lam” relate to the abundance model (lambda = estimated local abundance). # Parameters named “.mu”, “.tau” and “.sd” refer to mean, precision and standard deviation hyperparameters, respectively.


182

model{ # Priors for observation model

# ---------------------------- for (j in 1:nbothtrapspp) { # i.e. small mammals for (t in 1:ntraptype) {

# Intercept parameter for observation model, varying by trap-type # Reference land-use is old-growth forest alpha.r[j, t] ~ dnorm(alpha.r.mu[t], alpha.r.tau[t])

} } for (j in nbothtrapspp+1:nbothtrapspp+nctrapspp) { # i.e. large mammals

# Prior for trap-type 1 (live traps) is set to a constant for large mammals alpha.r[j, 1] <- 0

for (t in 2:ntraptype) { alpha.r[j, t] ~ dnorm(alpha.r.mu[t], alpha.r.tau[t])

} } for (j in 1:nbothtrapspp+nctrapspp) { # i.e. all mammal species

# Species-specific parameter for difference in r (on log scale) between old-growth and logged forest Beta1.r[j] ~ dnorm(Beta1.r.mu, Beta1.r.tau) # Species-specific parameter for difference in r (on log scale) between old-growth forest and oil palm Beta2.r[j] ~ dnorm(Beta2.r.mu, Beta2.r.tau) # Species-specific 1st-order polynomial term for habitat disturbance effect Beta3.r[j] ~ dnorm(Beta3.r.mu, Beta3.r.tau) # Species-specific 2nd-order polynomial term for habitat disturbance effect Beta4.r[j] ~ dnorm(Beta4.r.mu, Beta4.r.tau)

} for (j in 1:(nbothtrapspp+nctrapspp-ngroupspp)) {

# i.e. non-grouping-living species # Quasibinomial overdispersion parameter set to zero theta[j] <- 0

} for (j in (nbothtrapspp+nctrapspp-ngroupspp+1):nbothtrapspp+nctrapspp) {

# i.e. group-living species theta[j] ~ dnorm(theta.mu, theta.tau) } # Hyperpriors for observation model # --------------------------------- for (t in 1:ntraptype) {

alpha.r.mu[t] ~ dnorm(0, 0.01) alpha.r.tau[t] <- pow(alpha.r.sd[t], -2) alpha.r.sd[t] ~ dunif(0, 10)

} Beta1.r.mu ~ dnorm(0, 0.01) Beta1.r.tau <- pow(Beta1.r.sd, -2) Beta1.r.sd ~ dunif(0, 10) Beta2.r.mu ~ dnorm(0, 0.01) Beta2.r.tau <- pow(Beta2.r.sd, -2) Beta2.r.sd ~ dunif(0, 10) Beta3.r.mu ~ dnorm(0, 0.01) Beta3.r.tau <- pow(Beta3.r.sd, -2) Beta3.r.sd ~ dunif(0, 10) Beta4.r.mu ~ dnorm(0, 0.01) Beta4.r.tau <- pow(Beta4.r.sd, -2) Beta4.r.sd ~ dunif(0, 10) theta.mu ~ dnorm(0, 0.01) theta.tau <- pow(theta.sd, -2) theta.sd ~ dunif(0, 10) # Hyperpriors for abundance model # ------------------------------- # Hyperparameters for intercept of abundance model alpha.lam.mu ~ dnorm(0, 0.01)


183

alpha.lam.tau <- pow(alpha.lam.sd, -2) alpha.lam.sd ~ dunif(0, 10) # Hyperparameters for 1st–order polynomial term for AGB # (AGB = above-ground live tree biomass) Beta1.lam.mu ~ dnorm(0, 0.01) Beta1.lam.tau <- pow(Beta1.lam.sd, -2) Beta1.lam.sd ~ dunif(0, 10) # Hyperparameters for 2nd–order polynomial term for AGB Beta2.lam.mu ~ dnorm(0, 0.01) Beta2.lam.tau <- pow(Beta2.lam.sd, -2) Beta2.lam.sd ~ dunif(0, 10) # Hyperparameters for effect of landscape forest cover Beta3.lam.mu ~ dnorm(0, 0.01) Beta3.lam.tau <- pow(Beta3.lam.sd, -2) Beta3.lam.sd ~ dunif(0, 10) # Hyperpriors for half-Cauchy scale parameter # ------------------------------------------- xi.tau <- pow(xi.sd, -2) xi.sd ~ dunif(0, 10) for (j in 1:nbothtrapspp+nctrapspp) { # i.e. all mammal species # Priors for land-use environmental filter # ---------------------------------------- omega[j] ~ dunif(0, 1) # Priors for abundance model # --------------------------

# Poisson intensity for abundance at the centred covariate values alpha.lam[j] ~ dnorm(alpha.lam.mu, alpha.lam.tau)

# 1st-order polynomial term for AGB effect Beta1.lam[j] ~ dnorm(Beta1.lam.mu, Beta1.lam.tau)

# 2nd-order polynomial term for AGB effect Beta2.lam[j] ~ dnorm(Beta2.lam.mu, Beta2.lam.tau) # Parameter for effect of landscape forest cover Beta3.lam[j] ~ dnorm(Beta3.lam.mu, Beta3.lam.tau)

# Priors for random spatial and temporal effects # ---------------------------------------------- for (plot in 1:nplot) {

# Random sampling plot effects gamma[plot, j] ~ dnorm(0, gamma.tau[j])

} for (block in 1:nblock) {

# Random block effects delta[block, j] ~ dnorm(0, delta.tau[j])

} for (k in 1:nsession) {

# Random session effects (here, sampling year effects) eps[k, j] ~ dnorm(0, eps.tau[j])

} eps.tau[j] ~ dgamma(0.5, 0.5) # chi-squared with 1 d.f. xi[j] ~ dnorm (0, xi.tau) # priors on scale parameter of half-Cauchy

sigma.cauchy[j] <- abs(xi[j]) / sqrt(eps.tau[j]) # Cauchy = normal/sqrt(chi-squared)

# Hyperpriors for random spatial effects # -------------------------------------- gamma.tau[j] <- pow(gamma.sd[j], -2) gamma.sd[j] ~ dunif(0, 10) delta.tau[j] <- pow(delta.sd[j], -2) delta.sd[j] ~ dunif(0, 10) for (l in 1:nlanduse) {

# Land-use filter # --------------- # Occurrence of species j in land-use l


184

# w[j, l] = 1 if the species was detected w[j, l] ~ dbern(omega[j]) for (i in (landuse.i[l]+1):landuse.i[l+1]) {

# Abundance is zero-inflated by multiplying by the outcome of the land-use filter N[i, j] <- A[i, j] * w[j, l]

# Ecological model for latent abundance state (Poisson random effect) # ------------------------------------------------------------------- # Abundance, if present, assumed to be Poisson-distributed A[i, j] ~ dpois(lambda[i, j]) # Linear predictor for abundance log(lambda[i, j]) <- alpha.lam[j] + Beta1.lam[j] * AGB[i] + Beta2.lam[j] * AGB2[i] + Beta3.lam[j] * ForestCov[i] + gamma[plot.counter[i], j] + delta[block.counter[i], j] + xi[j] * eps[session.counter[i], j]

for (obs in site.startstop.i[i, 1]:site.startstop.i[i, 2]) {

# Observation model for detection history # --------------------------------------- # Capture history D modelled as the outcome of a binomial process D[obs, j] ~ dbin(p[obs, j], occasions[obs]) # Species-level detection probability p p[obs, j] <- 1 - (1 - r[obs, j])^(N[i, j]^(1 + theta[j])) # r is the inverse logit of the linear predictor for r # avail[obs, j] specifies if species j can be caught by the sampling method used for observation obs, i.e. is ‘available’ r[obs, j] <- (1 / (1 + exp(-logit.r[obs, j]))) * avail[obs, j] # Linear predictor for individual detection probability r logit.r[obs, j] <- alpha.r[j, traptype[obs]+1] + Beta1.r[j] * Log[i] + Beta2.r[j] * OP[i] + Beta3.r[j] * HabitatScore[i] + Beta4.r[j] * HabitatScore2[i] # Calculate Pearson chi-squared residuals to assess goodness-of-fit # ----------------------------------------------------------------- # Simulate an ‘ideal’ dataset D.new[obs, j] ~ dbin(p[obs, j], occasions[obs]) # Calculate the observed and ‘ideal’ residuals # (A small non-zero value, 1.0E-9, prevents division by zero) Res[obs, j] <- (D[obs, j] - p[obs, j] * occasions[obs]) / sqrt((p[obs, j]+1.0E-9) * occasions[obs] * abs(1-p[obs, j]-1.0E-9)) Res.new[obs, j] <- (D.new[obs, j] - p[obs, j] * occasions[obs]) / sqrt((p[obs, j]+1.0E-9) * occasions[obs] * abs(1-p[obs, j]-1.0E-9)) Res.2[obs, j] <- Res[obs, j]^2 Res.new2[obs, j] <- Res.new[obs, j]^2

} # obs trap-types within site i # Derived occupancy parameter (optional) # -------------------------------------- occ[i, j] <- step(N[i, j] - 1)

} # i sites (including stacked sessions) } # l land-uses # Calculate the Pearson chi-squared discrepancy for each species # -------------------------------------------------------------- # Sum the residuals Pears.fit[j] <- sum(Res.2[, j]) Pears.fitnew[j] <- sum(Res.new2[, j])

} # j species

# Calculate the overall Pearson chi-squared discrepancy # ----------------------------------------------------- # Sum the residuals across species Pears.fit.all <- sum(Pears.fit) Pears.fitnew.all <- sum(Pears.fitnew)

}

Chapter 6: Conserving mammals in human-modified landscapes in Southeast Asia

185

Chapter 6:

The conservation of terrestrial mammals in human-modified landscapes in Southeast

Asia: richness, composition, abundance and heterogeneity

6. 1. Key findings of this thesis

Over the preceding chapters, I have attempted to fulfil the four objectives I outlined at the outset. In

brief, these objectives were to: 1) robustly quantify mammalian species richness across the principal

land-use gradient in Southeast Asia, 2) investigate grain-dependency in species richness and β-

diversity, 3) identify how land-use change is altering the composition of communities and the drivers

of local-scale assembly, and 4) robustly quantify mammalian species abundance across the land-use

gradient. I will now outline my findings as they relate to these objectives.

I found that the richness of terrestrial mammals in old-growth forest was retained overall in logged

forest (Chapter 3), and that this was also true for the Felidae (Chapter 2), the mammal family with

more species of high conservation concern than any other in Borneo (IUCN, 2014). However, changes

in the fine-scale occurrence of species in logged forest meant that species richness at the finest spatial

grain was lower for large mammals than in old-growth forest (Chapter 3). Concomitant with this, fine-

grained β-diversity was higher in logged forest compared to old-growth forest (Chapter 3). At coarser

spatial grains, however, β-diversity in logged forest decreased for large mammals and increased for

small mammals, whilst the opposite patterns for both species groups were evident in old-growth

forest. These changes in β-diversity patterns in logged forest can be understood in terms of an

alteration in the underlying drivers of community assembly (Chapter 4). Whilst communities in old-

growth forest were strongly spatially-structured – a signature of highly autocorrelated movements by

animals, dispersal limitation and a spatially-correlated environment – this was not the case in logged

forest, with fine-grained environmental heterogeneity (caused by logging), as well as random

community drift at coarser spatial grains, likely playing much stronger roles. Community composition


186

was also markedly different between old-growth and logged forest (Chapter 4), a difference which

could be explained in large part by changes in habitat structure, at both fine and broad scales, as a

result of logging. Beyond composition per se, the abundance of species also changed from old-growth

to logged forest (Chapter 5), in many cases showing an increasing trend with logging-related

disturbance in a landscape or, in some cases, a weakly unimodal response.

In the transition from forest to oil palm, the changes in species richness, β-diversity, community

composition and abundance were even starker. Oil palm plantations did not conserve similar levels of

species richness to forest, at any spatial grain (Chapter 3). After controlling for the smaller size of this

species pool, I also found that β-diversity was generally lower than in forest, across both large and

small mammals and across spatial grains (Chapter 3). The composition of mammal communities in oil

palm was also markedly different compared to forest, except for those sampling points in the margins

of the plantation, which more closely resembled forest communities (Chapter 4). Even for those

species which were retained in the plantations, abundances were vastly different to those in forest,

with most species persisting at low abundance, but with a limited subset of mammal species (mostly

low conservation concern members of the Carnivora) apparently prospering in the open oil palm

habitat and reaching a higher abundance than in forest.

The ability to achieve the objectives of my thesis was in large part predicated on the success of

random spatial sampling of the mammal communities. In particular, this was essential for analyses

relying on the spatial patterns of species occurrence (i.e. Chapters 3 and 4). Random sampling for

small mammals is widely practiced, but this has rarely been the case for camera-trapping. Given this, I

wanted to know if indeed there were strong reasons to depart from the fundamental statistical bedrock

of randomisation, for example in the case of rapid surveys of high conservation concern species.

Whilst the overall photo-capture rate across species was likely lowered by the use of random

sampling, capture rates for threatened felid species were statistically indistinguishable from rates

obtained in previous non-random studies, and indeed were much higher for the poorly-known bay cat,

Pardofelis badia (Chapter 2). In addition, explicit modelling of detection probabilities showed that,


187

whilst some species were much more detectable in the open areas used by non-random sampling (e.g.

roads), many other species showed the opposite pattern (Chapter 5).

6. 2. Putting the findings into context

There is a large amount of agreement between my findings and those of past studies of land-use

change in Southeast Asia, in particular the high value of selectively-logged forests for conserving

species richness, and the very low value of oil palm. Most previous studies in the region have found

that the overall species richness (i.e. γ-diversity) of a broad range of taxonomic groups in logged

forests is statistically indistinguishable from old-growth forest (Cannon et al., 1998; Ghazoul, 2002;

Hamer et al., 2003; Peh et al., 2005; Cleary et al., 2007; Berry et al., 2008; Woodcock et al., 2011;

Struebig et al., 2013; Edwards et al., 2014a), as I also found for terrestrial mammals (Chapter 3).

Moreover, these logged forests retain > 70% of old-growth forest species for most invertebrate and

vertebrate taxonomic groups investigated (Ghazoul, 2002; Peh et al., 2005; Slade et al., 2011;

Woodcock et al., 2011; Edwards et al., 2014a), broadly agreeing with my finding of 89% for

terrestrial mammals (Chapter 3).

In oil palm, on the other hand, it has been an almost universal finding that species richness declines

overall compared to forest, whether logged or old-growth forest (Koh & Wilcove, 2008; Fitzherbert et

al., 2008; Foster et al., 2011; Edwards et al., 2014a), and that the overwhelming majority of forest

species – approximately 75-80% across studies (Koh & Wilcove, 2008; Fitzherbert et al., 2008),

comparable to my observation of 70% for mammals (Chapter 3) – do not persist in the plantations.

The notable exceptions to these species richness patterns across land-use have been in termites (Jones

et al., 2003; Donovan et al., 2007; but see Eggleton et al., 1997) and canopy-dwelling butterflies

(Dumbrell & Hill, 2005), both of which have shown significantly lower richness in logged relative to

old-growth forest, possibly due to the specialized niches of species in these groups, which may not be

supported in logged forest. In addition, bees have been shown to increase in richness in oil palm

(Liow et al., 2001), the only group reported to respond in this way.


188

My findings with respect to species richness across the land-use gradient also agree with previous

studies focussed on terrestrial mammals. As with other taxonomic groups, the species richness of

terrestrial mammals in logged forests has been found to be broadly similar in logged and old-growth

forests (Kemper & Bell, 1985; Numata et al., 2005; Wells et al., 2007; Bernard et al., 2009; Kitamura

et al., 2010; Brodie et al., 2015), and substantially lower in oil palm plantations (Danielsen &

Heegaard, 1995; Scott & Gemita, 2004; Maddox et al., 2007; Rajaratnam et al., 2007; Bernard et al.,

2009; Puan et al., 2011; Silmi et al., 2013; Jennings et al., 2015). Only three previous studies,

however, have robustly investigated terrestrial mammal richness across multiple land-uses, finding

that differences between logged and old-growth forest were not significant for large mammals

(Kitamura et al., 2010; Brodie et al., 2015) or small mammals (after excluding poorly-sampled species

> 1 kg; Wells et al., 2007); I found that, whilst large mammal richness was not significantly different,

small mammal richness was significantly higher in logged forest (Chapter 3).

Beyond species richness responses, there is rather less consensus on the biodiversity impacts of land-

use change in Southeast Asia. One important example which relates to my findings is the change in β-

diversity across land-use. Levels of β-diversity will partly determine the optimal spatial design of

conservation set-aside in logged forest or plantation landscapes, and may also lead to grain-dependent

species richness responses across land-use (Hill & Hamer, 2004). I hypothesised in Chapter 3 that

logged forests would be more environmentally heterogeneous than old-growth forests, as supported

by environmental measurements in a limited number of studies (e.g. frequency of gaps: Berry et al.,

2008, understorey ground cover variance: Ghazoul, 2002; leaf area index variance: Ewers et al., 2015;

canopy cover variance: Struebig et al., 2013), as well as coarser-grained spatial variation in the

intensity of logging itself (Cannon et al., 1994; Berry et al., 2008), and that this would lead to higher

β-diversity in logged forest. My own results for terrestrial mammals supported this overall, with a

stronger signal of β-diversity in logged forests (particularly at fine spatial grains), and this has also

apparently been the case in previous studies of ants (Woodcock et al., 2011) and trees (Berry et al.,

2008). However, there are also a number of studies which do not appear to show obvious differences

between old-growth and logged forest (birds and dung beetles: Edwards et al., 2011; butterflies:


189

Dumbrell et al., 2008), or indeed show the opposite pattern to that which I found (butterflies: Hamer

& Hill, 2000, Cleary, 2003; dung beetles: Davis et al., 2001). The reasons for these discrepancies are

not yet clear. It might indeed be the case that logged forests are not more heterogeneous than old-

growth forest when defined along alternative environmental dimensions, such as resource availability

rather than habitat structure per se. Alternatively, and as shown in my own results, the β-diversity

signal may often be grain-dependent, meaning that studies carried out at different grains may not be

comparable. This might even occur when studies are apparently sampling at a similar grain, because

the effective sampling area will depend on the mobility of the taxa under consideration (Hill &

Hamer, 2004). It may also be significant that previous studies in the region have not controlled for

differences across land-use in the sampling process and in species pools (e.g. using the null model

approach I used in Chapter 3), both of which are known to have strong effects on observed β-diversity

(Lessard et al., 2012; Beck et al., 2013).

6. 3. New research directions

In this thesis, I also explored some new research avenues which have been little investigated to date. I

presented evidence that the fundamental drivers of community assembly are altered along the most

important land-use gradient in Southeast Asia. This idea has been little explored for the vast areas of

degraded forest in the region, but could have consequences for how we view the prospects for

wholesale ecological restoration in these systems, whether through natural or assisted recovery.

Although, as I have outlined in this thesis, there is a building consensus that the impacts of logging on

species richness and abundance are often minimal, we know much less about the changing roles of

environmental control, species interactions and spatial processes (e.g. dispersal) in constructing

communities. This may have direct management implications in determining the most effective

conservation interventions. Simplistically, communities under environmental control will be most

malleable to interventions which restore habitat, whilst dispersal-limited communities may respond

most strongly to interventions which increase connectivity amongst populations. We might also

expect that where environmental control is strong relative to compensatory dynamics, for example

generated by inter- and intra-specific interactions, communities may be less resilient to environmental


190

change (Mutshinda et al., 2009), be it over space or time. Further studies investigating the processes

assembling communities are now needed in the region, ideally for a broad range of taxonomic groups,

in order to corroborate my findings for terrestrial mammals.

I also explored the utility of hierarchical state-space models of abundance, which had not previously

been done for any taxonomic group in Southeast Asia. Although there have been a considerable

number of studies of the impacts of land-use change on mammals in Southeast Asia (see Chapter 1),

very few of these have provided robust estimates of abundance. Whilst the abundance responses of

some species across land-use have become increasingly clear as more studies have been done (e.g.

orangutans: Ancrenaz et al., 2010, Meijaard et al., 2010; some tupaiid species: Emmons, 2000, Wells

et al., 2007; some murid rodent species: Nakagawa et al., 2006, Rajaratnam et al., 2007, Wells et al.,

2007, Bernard et al., 2009; large herbivores: Wilson & Johns, 1982, Duff et al., 1984, Davies et al.,

2001; felids: Rajaratnam et al., 2007, Mohamed et al., 2009, 2013, Brodie & Giordano, 2012) there

remains a large group of species for which responses have remained variable and inconsistent

(Meijaard & Sheil, 2008). For example, conflicting responses to logging by mouse-deer (Tragulus

kanchil and T. napu), sun bears (Helarctos malayanus) and civets (Viverridae) have been reported in

the literature (Meijaard et al., 2005; Meijaard & Sheil, 2008). Given the high conservation status of

some of these species, as well as their potential importance for example in seed dispersal (Meijaard et

al., 2005), further studies using robust methods are needed to allow for comparisons with my findings.

Beyond the conservation importance of accurately estimating the abundance of threatened species,

absolute measures of abundance also provide the prospect of more directly assessing the functional

effects of abundance changes, for example by re-scaling abundance in terms of biomass (as I did in

Chapter 5). There is also now an opportunity to build upon biomass as an index of ecosystem function

and explicitly incorporate metabolic and food web theory (e.g. Barnes et al., 2014). This will require

better information on mammal diets in particular, as well as on the potential for dietary shifts across

land-use (e.g. in sun bear: Fredriksson et al., 2006; primates: Johns, 1983; Malay civet: Colón, 2002;

tupaiids: Munshi-South et al., 2007; murid rodents: Nakagawa et al., 2007). Quantifying the

ecosystem functions carried out by mammals may be especially crucial if, as suggested by recent


191

studies (Barnes et al., 2014; Ewers et al., 2015), losses in the richness or abundance of other

taxonomic groups mean that functional redundancy is eroded under increasing land-use intensity.

A common theme running throughout this thesis, which has been much-neglected in models of

biodiversity change across land-use, has been the important role of heterogeneity in structuring

communities in space. Environmental heterogeneity, at both local and landscape scales, was seen to

be a prominent driver of community variation (Chapter 4), giving rise to aggregated distributions of

individuals and species (Chapter 3). I explored this at the species level for felids (Chapter 2), finding

that each species exhibited contrasting patterns of habitat feature use, as indicated by their

unconditional probabilities of detection. This result was bolstered with explicit modelling of

conditional probabilities of detection across 57 species (Chapter 5), in which contrasting patterns of

fine-scale habitat use were again evident across species. In addition, variation among local landscapes

in logging intensity and forest cover was seen to exert influence on the abundance of individual

species (Chapter 5). This leads to a number of hypotheses that could be tested in future work. In

particular, my results are consistent with greater aggregation of individuals and species in favourable

patches of habitat (Chapters 3 and 4). This leads to the hypothesis that individuals in logged forest

may, in some species, have larger home ranges (possibly also with a less regular shape, or with

fragmented core areas of use), in order to capture sufficient resources across spatially-separated

patches of favourable patches. This is supported by the breakdown of spatially-autocorrelated patterns

of movement (Chapter 4) and the lower detection probability, overall, of species in logged forest (i.e.

because individuals are spending a greater amount of time unavailable for capture; Chapter 5).

Available radio-tracking data for Malay civets (Colón, 2002) and treeshrews (Munshi-South et al.,

2007) has so far not indicated a clear difference in home range size between logged and unlogged

forests, though sample sizes are small and the intensity of logging in these studies was much lower

than in my study. There are two further corollaries of this hypothesis: 1) for territorial species,

changes in home range characteristics may have implications for daily ranging distances required, and

for the energetic costs associated with this; 2) increases in fine-grained habitat heterogeneity may lead

to greater levels of “habitat sampling”, in which species utilise resources in non-optimal patches


192

situated between preferred habitats, and a consequent reduction in the strength of species-environment

relationships (Debinski et al., 2001). Habitat sampling would be especially frequent in species for

which habitat heterogeneity is at a much finer grain relative to their scales of movement.

The importance of heterogeneity in disturbed landscapes I have identified also suggests that

significant improvements in biodiversity forecasting might be made by considering land-use change

as a continuum of disturbance occurring at multiple spatial grains, rather than discretizing the land-use

change process into homogeneous categories. To my knowledge, this has been little explored in the

land-use change literature, beyond the widespread use of ordination methods poorly suited for

biodiversity forecasting (but see: Koh, 2008; McShea et al., 2009; Struebig et al., 2013). There are

parallels between this suggestion and the idea of continuous response gradients in the habitat

fragmentation literature (e.g. the “variegation” model: Mcintyre & Barrett, 1992, McIntyre & Hobbs,

1999; the “continuum” model: Fischer & Lindenmayer, 2006). Indeed, an increasing recognition of

heterogeneity in the land-use change literature and a decreasing emphasis on the patch-based view of

landscapes in the fragmentation literature (e.g. Laurance, 2008) may offer the potential for these two

relatively discrete sub-disciplines of ecology to find significant common ground. Continuous response

models recognise that different species, or groups of species, often show contrasting responses to

heterogeneity (Fischer et al., 2004; Manning et al., 2004), for example due to their differing innate

abilities to perceive the grain of a landscape (With, 1994) or their differing requirements for

persistence (Fischer & Lindenmayer, 2006). Hierarchical metacommunity models, such as I used in

Chapter 5, potentially offer an opportunity to bring together species- or group-specific responses to

land-use change, across multiple scales, under one framework, whilst also allowing for the estimation

of emergent community parameters, such as species richness and β-diversity, at the same time

(Dorazio et al., 2011).


193

Figure 1. An example of heterogeneity within the logged forest landscape of the Kalabakan Forest Reserve: (A) a southern pig-tailed macaque Macaca nemestrina crossing an open area on an old primary logging road, (B) a red muntjac Muntiacus muntjak in a dense area of herbaceous scrub, dominated by wild ginger (Zingiberaceae), and (C) a lesser mouse-deer Tragulus kanchil in a lightly-disturbed forest patch with an intact canopy.

A

B

C


194

6. 4. Implications for conservation

The findings in this thesis have a number of key implications for conservation. Firstly, there are

implications for wildlife surveys in the region, for example those undertaken by conservationists,

protected area managers or as a requirement for sustainability certification. Sustainability certification

schemes, such as the Forest Stewardship Council or Round-table on Responsible Palm Oil, in

particular, require the identification, often with very limited time, of High Conservation Value (HCV)

areas or species (e.g. Persey et al., 2011). As I outlined in Chapter 1, I would advocate the random

sampling approach as the default survey design for such surveys, though there may be circumstances

in which it is justifiable to deviate from this (e.g. if surveys are targeting a limited number of focal

species which are strongly associated with specific habitats). For felids, often a target of conservation

surveys, I provided evidence that random sampling may substantially decrease the sampling efforts

required to detect (with 90% probability) some species of high conservation concern (e.g. the bay cat),

albeit with the cost of marginally increasing sampling efforts in other low conservation concern

species (e.g. the leopard cat Prionailurus bengalensis). In Chapter 3 I showed that, in the logged

forests that are the focus of most HCV surveys, high levels of coarse-grained β-diversity in small

mammals may mean that sampling designs will need to be sufficiently large in extent (e.g. with a

coverage > 35 km2) in order to capture the full diversity of species (results suggested that this was not

as important for large mammals). In Chapter 5, estimates of abundance for single species showed

large uncertainty once all of the sources of ecological and observational sampling variation were

properly accounted for, even though I used very large sampling efforts. In part, this is a reflection of

the highly clustered sampling design that I was constrained to use (both for the long-term value of the

study as a baseline, but also because I was interested in fine-scale patterns of occurrence), but there is

an even more important implication of this: short-term surveys of species with limited sampling effort

should, in most cases, avoid making inferences about the abundance of species (e.g. using relative

abundance indices) and should instead focus on documenting which species are present in the study

site of interest. In addition, Chapter 5 showed the very high value of conducting sampling using

multiple methods. This allowed for the estimation of abundance in a greater number of species but it

was also the case that simultaneous live-trapping in areas that were camera-trapped allowed for the


195

identification of a far greater number of small mammals from camera trap images than has been

possible in previous studies (e.g. McShea et al., 2009; Samejima et al., 2012; Brodie et al., 2015).

The finding that overall species richness (γ-diversity) and species abundance are conserved in

contiguous logged forest, including > 85% of old-growth species, gives strong emphasis to calls for

the increased recognition of the conservation value of extensive tracts of degraded forests in Southeast

Asia (as discussed in Chapters 3 and 5), albeit with the condition that hunting is also controlled

effectively (Bennett & Gumal, 2001). Moreover, the depauperate mammal communities found within

oil palm (and the limited responses in abundance to increases of forest even as high as 30%) is

consistent with recent assertions that it is the “greatest immediate threat to biodiversity in Southeast

Asia” (Wilcove & Koh, 2010), and is unlikely to contribute substantially to conservation in the

region, even if more “wildlife-friendly” management practices were implemented (e.g. Bhagwat &

Willis, 2008; Koh, 2008). In a coarse, qualitative sense, these findings aid conservationists in making

recommendations to industry and governments, and give land-use managers and policy-makers the

requisite information on the relative biodiversity value of different land-use types in order to make

more well-informed decisions. However, in order to more effectively assist decision-makers to

resolve tradeoffs between competing land-uses, it will be necessary to integrate the expanding corpus

of data on biodiversity responses in a quantitative, and ideally spatially-explicit, framework. Thus far,

attempts to do this have largely relied on imprecise modelling of biodiversity responses, for example

using generic species-area relationships (Koh & Ghazoul, 2010) or assigning species to broad

categories of sensitivity to land-use change, based on expert opinion (Wilson et al., 2010). Edwards et

al. (2014), using non-spatial simulations, explored the economic and biodiversity tradeoffs between

selective logging, protection and oil palm, using field data on a range of taxonomic groups surveyed

in Sabah, Malaysia. They found that land-use portfolios comprising mostly of logged forest were the

most efficient way to conserve the most species for the lowest opportunity cost (Edwards et al.,

2014a). It remains to be seen, however, how this would translate to real landscapes in a spatially-

explicit scenario, and this represents a key area of uncertainty about land-use change in the region.


196

Although I have argued for a spatial separation of forest and oil palm land-uses in this thesis (i.e.

through a practice of landscape-scale land-sparing; Edwards et al., 2010), important uncertainties

remain about the biodiversity and economic value of retaining forest fragments within oil palm

plantations. For example, although the biodiversity value of small fragments is likely relatively low

for terrestrial mammals (Bernard et al., 2014) and birds (Edwards et al., 2010b), which mostly have

large area requirements, they may have more value for insectivorous bats (Struebig et al., 2008,

2011), butterflies (Benedick et al., 2006), dung beetles (Gray et al., 2014), ants (Lucey et al., 2014)

and other taxa with smaller area requirements. In addition, within a larger landscape context, even

small fragments may have a value as “stepping-stone” habitats for individuals dispersing between

larger blocks of contiguous forest (Uezu et al., 2008). Remnant forest fragments also sometimes come

at no cost to yields, for example in areas in which it is not profitable to grow oil palm, such as on

steep slopes, high elevation areas or on poor soils.

Perhaps most importantly, it may be the case that forest fragments even increase yields by exporting

ecosystem services to the oil palm crop, for example if predators of pest species exhibit spill-over

from forest fragments into the plantation habitat (Koh, 2008b; Lucey et al., 2014; but see Edwards et

al., 2014b). I found that carnivores substantially increased in abundance in oil palm compared to

forest (Chapter 5). In part, this was due to increased abundance of free-ranging domestic dogs (Canis

familiaris), but it was also due to increases in a limited number of native carnivores – the leopard cat,

common palm civet (Paradoxurus hermaphroditus) and Malay civet (Viverra tangalunga) – which

might be expected to be dependent on remnant forest patches. These native carnivores all likely prey

upon murid rodents (Rajaratnam et al., 2007; Nakashima et al., 2013; Naim et al., 2014) – which can

be a cause of substantial losses in oil palm yield when occurring in high abundance (Wood & Chung,

2003) – and may therefore be good candidates for biological control. Further studies of the diets of

these species within oil palm plantations are needed, as well as mensurative or manipulative

experiments investigating their role in reducing murid rodent populations. Crucially, the habitat

requirements of these species are still poorly known, and in particular whether they require remnant

forest fragments in order to persist. Certainly, both the leopard cat and common palm civet are known


197

to use resting sites within oil palm plantations (Rajaratnam et al., 2007; Nakashima et al., 2013), but

radio-tracking studies of these species, which have thus far focussed on individuals occurring at the

plantation-forest edge (Rajaratnam et al., 2007; Nakashima et al., 2013), remain equivocal on whether

some individuals are able to persist entirely within the plantation. The Malay civet appears to be more

dependent on remnant forest than the leopard cat and common palm civet, based on the available

radio-tracking (Jennings et al., 2006) and camera-trapping (Jennings et al., 2015) data available to

date.

6. 5. The scope of inference

My thesis leaves a number of questions surrounding the impacts of land-use change on mammals still

open. Importantly, the long-term viability of mammal populations in logged forest is poorly known

and, whilst I observed evidence of breeding in a large number of species (e.g. murid rodents, Malay

porcupine, tupaiids, some squirrel species, macaque species, orangutan, all ungulate species, pangolin,

sun bear, banded civet, Sunda clouded leopard and Asian elephant), populations could still have a

negative growth rate overall. Information on population trends over time or, ideally, the survival and

recruitment rates of individuals would be required to address this uncertainty. An additional

complication over time comes from the effects of forest successional dynamics. The effects of forest

canopy recovery, for example, in heavily-degraded forests are poorly known for mammals, and

indeed other taxonomic groups, with few studies reporting the time since logging occurred (in part

because of the difficulty of obtaining spatially- and temporally-precise information). In logged forests

in Central Africa, Clark et al. (2009) found evidence of unimodal abundance responses over time in

some mammal species (with initial recovery in the decade after logging, but with a decline thereafter).

No equivalent studies have been done in Southeast Asia, but Brodie et al. (2015) found that mammal

occupancy was higher for many species in older logged forests in Malaysian Borneo (logged > 10

years previously) compared to recently logged forests. An added complication is that, in repeatedly-

logged forests, it may be more important to consider the full history of logging through time, rather

than just the time since last logging.


198

It should also be noted that I focussed on terrestrial mammals in this study, and the responses of other

mammalian groups may be dissimilar. Some groups, most especially the arboreal squirrels (and in

particular the flying squirrels, Pteromyini), remain very difficult to sample effectively and very little

is known about their responses to disturbance. Novel sampling methods, such as arboreal camera-

trapping or hair traps (Castro-Arellano et al., 2008) may provide the necessary data to estimate the

abundance of these species. The responses to land-use change of bats, which can form a substantial

proportion of species in Southeast Asia mammal communities (often up to 50%), are also poorly

known (but see: Danielsen & Heegaard, 1995; Furey et al., 2010; Struebig et al., 2011, 2013), despite

being important pollinators and seed dispersers (Meijaard et al., 2005).

Just as my findings need to be corroborated across the full spectrum of mammal groups, my findings

also need to be corroborated over a broader set of study sites. My findings are likely to be directly

applicable to much of the broader Yayasan Sabah Forest Management Area (~ 1 million ha) within

which my study was set. This area, which contains nearly a third of remaining forest cover in the

Malaysian state of Sabah, has undergone rapid rates of degradation and forest conversion in recent

decades, and is now comprised of a mosaic of multiple land-uses including protected areas,

sustainable forest management (in areas much degraded by past logging) and plantations of fast-

growing timber and oil palm (Reynolds et al., 2011). Yayasan Sabah, the para-statal organisation

created by a Malaysian government decree in order to manage this area, ultimately intends to convert

20% of the original forest cover for plantations, of which 13% will be under oil palm (Reynolds et al.,

2011). My findings suggest that this will have caused considerable local-scale loss of mammal

biodiversity by the time the plantations are fully developed, but that the 80% of remaining natural

forest, although mostly in a highly degraded state, will retain much of the regional-scale mammal

biodiversity if protected from conversion in perpetuity. Similar land-use change processes are

occurring in Indonesian Borneo, Sumatra, other islands in the Indonesian archipelago, and indeed

continental Southeast Asia, and further studies should therefore aim to corroborate my findings more

broadly across landscapes in these regions. Much of the Bornean mammal fauna I investigated occurs

more widely across the Sunda shelf islands and Malay peninsula, with some species extending further


199

north into continental Southeast Asia proper (Payne et al., 2007), which does at least suggest that the

mammalian responses to land-use change I have found in this thesis, particularly on a species-level

basis, may be more broadly applicable in the region.

6. 6. Future work

This thesis has exploited patterns of animal occurrence to uncover the impacts of land-use change,

often with the necessity of estimating latent parameters which are not directly observable, but this

approach often entails combining strong assumptions with inductive reasoning in order to make

inferences, and there are limits to the mechanistic insights that can be gained from such an approach

(widespread though it is in ecology). There are two natural extensions of the approaches used in this

thesis: namely, the addition of data on individually-identified animals and experimental

manipulations. The former will allow for greater insights into the movement ecology of many of the

poorly-known Bornean mammals I have focussed upon and will, in combination with the trapping

data I have presented, allow some key missing details to be filled in, most obviously with respect to

dispersal and home-ranging.

Experimental manipulations, crucially, often allow for much stronger inferences about the

mechanisms at work in producing observed patterns (McGarigal & Cushman, 2002). In the context of

land-use change, opportunities to conduct experiments at sufficiently large scales are rare (Fayle et al.,

2015), but just such an experiment is now being undertaken with the mammal communities

investigated in this thesis. All of the logged forest sampling sites investigated in this thesis are

undergoing (at the time of writing) isolation within forest fragments, surrounded by an oil palm

matrix. These fragments have been strategically designed as part of the Stability of Altered Forest

Ecosystems (SAFE) Project (Ewers et al., 2011), in order to investigate the effects on biodiversity of,

for example, habitat area, isolation from contiguous forest and landscape forest cover. Importantly,

the SAFE Project represents a rare, long-term opportunity to investigate the temporal dynamics of

biodiversity responses to land-use change, which is a key area of uncertainty in biodiversity

forecasting (Wearn et al., 2012; Gibson et al., 2013).


200

6. 7. Avoiding Navjot Sodhi’s “Impending Disaster”

What does the future hold for Southeast Asia’s biodiversity? The human population growth rate for

Southeast Asia currently stands at 1.2%, and by 2030 there will an additional 90 million people in the

region (United Nations Population Division, 2015). Although an increasing proportion of this

population is living in urban centres, the economies of most countries in the region are tied to

agriculture and the impacts of this sector on natural habitats are likely to grow in the coming decades,

most especially due to the expansion of oil palm and rubber plantations (Wilcove et al., 2013;

Warren-Thomas et al., 2015). At the same time, production of timber from natural forests has been

declining since the 1990s and is at the lowest level in decades in most areas (FAO, 2011; Reynolds et

al., 2011), with the emphasis for wood production having instead shifted to plantations of fast-

growing exotics. Few unlogged forest areas remain in the region, and vast areas of forest now exist in

a highly degraded state (Reynolds et al., 2011; Margono et al., 2014; Gaveau et al., 2014), mostly

under the management of commercial interests. Historical trends in land-use have no doubt resulted in

heavy losses in local biodiversity, but thus far have apparently resulted in only a limited number of

global extinctions (Sodhi et al., 2004; Clements et al., 2006; Duckworth et al., 2012; Giam et al.,

2012; Szabo et al., 2012). Therefore, whilst it will not be possible to emerge from the Anthropocene

“bottleneck” (Malhi et al., 2014) with predominantly intact systems in Southeast Asia (as, for

example, is still possible in Amazonia), there is still a tangible, but rapidly narrowing, window of

conservation opportunity in the region, especially for those species and ecosystem functions which

can be conserved in degraded forests. This will require significant investment in the protection and

restoration of these degraded areas, in particular to reconnect remaining intact forest areas and provide

potential migration corridors for species adapting to future climate change (Scriven et al., 2015;

Struebig et al., 2015b). For vertebrates, this must also include an effort to curtail hunting and

overharvesting of species (Duckworth et al., 2012). Increases in investment could be achieved, for

example, by a rapidly scaled-up market for the conservation and enhancement of carbon stocks

(Edwards et al., 2010a), and by governments and companies making the transition towards more

sustainable policies on land-use (there is already some limited evidence that this occurring; Gregory et

al., 2012; Poynton, 2014). Provided, therefore, that further agricultural expansion into forest is


201

avoided (or, at least, compensated for by off-site conservation restoration activities), and that tradeoffs

between forest product yields and biodiversity in degraded forests can be effectively and efficiently

reconciled through partnerships between the commercial and conservation sectors, there is still hope

that Navjot Sodhi’s business-as-usual scenario of an “impending disaster” (Sodhi et al., 2004) in

Southeast Asia can largely be averted.

Bibliography

202

Bibliography

Abdullah, S.A. & Nakagoshi, N. (2007) Forest fragmentation and its correlation to human land use

change in the state of Selangor, Peninsular Malaysia. For. Ecol. Manage., 241, 39–48. Achard, F., Eva, H.D., Stibig, H.-J., Mayaux, P., Gallego, J., Richards, T. & Malingreau, J.-P. (2002)

Determination of deforestation rates of the world’s humid tropical forests. Science, 297, 999–1002.

Ahumada, J.A., Silva, C.E.F., Gajapersad, K., Hallam, C., Hurtado, J., Martin, E., McWilliam, A., Mugerwa, B., O’Brien, T., Rovero, F., Sheil, D., Spironello, W.R., Winarni, N. & Andelman, S.J. (2011) Community structure and diversity of tropical forest mammals: data from a global camera trap network. Philos. Trans. R. Soc. Lond. B. Biol. Sci., 366, 2703–2711.

Akutsu, K., Khen, C.V. & Toda, M.J. (2006) Assessment of higher insect taxa as bioindicators for different logging-disturbance regimes in lowland tropical rain forest in Sabah, Malaysia. Ecol. Res., 22, 542–550.

de Almeida, W.R., Wirth, R. & Leal, I.R. (2008) Edge-mediated reduction of phorid parasitism on leaf-cutting ants in a Brazilian Atlantic forest. Entomol. Exp. Appl., 129, 251–257.

Ancrenaz, M., Ambu, L., Sunjoto, I., Ahmad, E., Manokaran, K., Meijaard, E. & Lackman, I. (2010) Recent surveys in the forests of Ulu Segama Malua, Sabah, Malaysia, show that orang-utans (P. p. morio) can be maintained in slightly logged forests. PLoS One, 5, e11510.

Andrén, H. (1994) Effects of habitat fragmentation on birds and mammals in landscapes with different proportions of suitable habitat: a review. Oikos, 71, 355–366.

Anonymous (2011a) Rare bay cat spotted in Pulong Tau national park. The Borneo Post. Available from http://www.theborneopost.com/2011/01/04/rare-bay-cat-spotted-in-pulong-tau-national-park/.

Anonymous (2011b) Bay cat captured on video by camera traps. The Borneo Post. Available from http://www.theborneopost.com/2011/08/26/bay-cat-captured-on-video-by-camera-traps/.

Anonymous (2012) Endangered bay cat recorded to be seen in Upper Baram. The Borneo Post. Available from http://www.theborneopost.com/2012/11/21/endangered-bay-cat-recorded-to-be-seen-in-upper-baram/.

Appanah, S. & Turnbull, J.M. (1998) A Review of Dipterocarps: Taxonomy, Ecology and Silviculture. Center for International Forestry Research, Bogor, Indonesia.

Aratrakorn, S., Thunhikorn, S. & Donald, P.F. (2006) Changes in bird communities following conversion of lowland forest to oil palm and rubber plantations in southern Thailand. Bird Conserv. Int., 16, 71–82.

Asner, G.P., Keller, M. & Silva, J.N.M. (2004) Spatial and temporal dynamics of forest canopy gaps following selective logging in the eastern Amazon. Glob. Chang. Biol., 10, 765–783.

Asner, G.P., Rudel, T.K., Aide, T.M., Defries, R. & Emerson, R. (2009) A contemporary assessment of change in humid tropical forests. Conserv. Biol., 23, 1386–1395.

Azhar, B., Lindenmayer, D., Wood, J., Fischer, J., Manning, A., McElhinny, C. & Zakaria, M. (2013) Contribution of illegal hunting, culling of pest species, road accidents and feral dogs to biodiversity loss in established oil-palm landscapes. Wildl. Res., 40, 1–9.

Bagchi, R., Philipson, C.D., Slade, E.M., Hector, A., Phillips, S., Villanueva, J.F., Lewis, O.T., Lyal, C.H.C., Nilus, R., Madran, A., Scholes, J.D. & Press, M.C. (2011) Impacts of logging on density-dependent predation of dipterocarp seeds in a South East Asian rainforest. Philos. Trans. R. Soc. Lond. B. Biol. Sci., 366, 3246–3255.

Bibliography

203

Baker, T.R., Honorio Coronado, E.N., Phillips, O.L., Martin, J., Van Der Heijden, G.M.F., Garcia, M. & Silva Espejo, J. (2007) Low stocks of coarse woody debris in a southwest Amazonian forest. Oecologia, 152, 495–504.

Balmford, A., Green, R. & Phalan, B. (2012) What conservationists need to know about farming. Proc. R. Soc. B, 279, 2714–2724.

Barlow, J., Gardner, T., Araujo, I.S., Ávila-Pires, T.C., Bonaldo, A.B., Costa, J.E., Esposito, M.C., Ferreira, L. V, Hawes, J., Hernandez, M.I.M., Hoogmoed, M.S., Leite, R.N., Lo-Man-Hung, N.F., Malcolm, J.R., Martins, M.B., Mestre, L.A.M., Miranda-Santos, R., Nunes-Gutjahr, A.L., Overal, W.L., et al. (2007) Quantifying the biodiversity value of tropical primary, secondary, and plantation forests. Proc. Natl. Acad. Sci. U. S. A., 104, 18555–18560.

Barlow, J., Gardner, T.A., Louzada, J. & Peres, C.A. (2010) Measuring the conservation value of tropical primary forests: the effect of occasional species on estimates of biodiversity uniqueness. PLoS One, 5, e9609.

Barnes, A.D., Jochum, M., Mumme, S., Haneda, N.F., Farajallah, A., Widarto, T.H. & Brose, U. (2014) Consequences of tropical land use for multitrophic biodiversity and ecosystem functioning. Nat. Commun., 5, 5351.

Barton, K. (2015) MuMIn: Multi-Model Inference. R package version 1.13-4. Available from http://cran.r-project.org/package=MuMIn.

Baselga, A. (2010a) Partitioning the turnover and nestedness components of beta diversity. Glob. Ecol. Biogeogr., 19, 134–143.

Baselga, A. (2010b) Multiplicative partition of true diversity yields independent alpha and beta components; additive partition does not. Ecology, 91, 1974–1981.

Baselga, A., Jiménez-Valverde, A. & Niccolini, G. (2007) A multiple-site similarity measure independent of richness. Biol. Lett., 3, 642–645.

Bashir, T., Bhattacharya, T., Poudyal, K. & Sathyakumar, S. (2011) Notable observations on the melanistic Asiatic golden cat (Pardofelis temminckii) of Sikkim, India. NeBIO, 2, 1–4.

Bates, D., Maechler, M., Bolker, B. & Walker, S. (2014) lme4: Linear mixed-effects models using Eigen and S4. R package version 1.1-6. Available from http://cran.r-project.org/package=lme4.

Bebber, D., Brown, N., Speight, M., Moura-Costa, P. & Wai, Y.S. (2002) Spatial structure of light and dipterocarp seedling growth in a tropical secondary forest. For. Ecol. Manage., 157, 65–75.

Beck, J., Holloway, J.D. & Schwanghart, W. (2013) Undersampling and the measurement of beta diversity. Methods Ecol. Evol., 4, 370–382.

Beck, J., Schulze, C.H., Linsenmair, K.E. & Fiedler, K. (2002) From forest to farmland: diversity of geometrid moths along two habitat gradients on Borneo. J. Trop. Ecol., 18, 33–51.

Belden, G., Stuebing, R., Megum, N., McShea, W.J. & Stewart, C. (2007a) A camera trapping inventory for mammals in a mixed use planted forest in Sarawak. Raffles Bull. Zool., 55, 209–215.

Belden, G., Stuebing, R. & Nyegang, M. (2007b) Small carnivores in mixed-use forest in Bintulu division, Sarawak, Malaysia. Small Carniv. Conserv., 36, 35–37.

Bell, G. (2000) The distribution of abundance in neutral communities. Am. Nat., 155, 606–617. Bender, D.J., Contreras, T.A. & Fahrig, L. (1998) Habitat loss and population decline: a meta-analysis

of the patch size effect. Ecology, 79, 517–533. Benedick, S., Hill, J.K., Mustaffa, N., Chey, V.K., Maryati, M., Searle, J.B., Schilthuizen, M. &

Hamer, K.C. (2006) Impacts of rain forest fragmentation on butterflies in northern Borneo: species richness, turnover and the value of small fragments. J. Appl. Ecol., 43, 967–977.

Benítez-López, A., Alkemade, R. & Verweij, P.A. (2010) The impacts of roads and other infrastructure on mammal and bird populations: A meta-analysis. Biol. Conserv., 143, 1307–

Bibliography

204

1316. Bennett, E. & Dahaban, Z. (1995) Wildlife responses to disturbances in Sarawak and their

implications for forest management. In Ecology, conservation and management of Southeast Asian rainforests (eds R. Primack & T. Lovejoy), pp. 66–86.

Bennett, E. & Gumal, M. (2001) The inter-relationships of commercial logging, hunting and wildlife in Sarawak, and recommendations for forest management. In The cutting edge: conserving wildlife in logged tropical forest (eds R. Fimbel, A. Grajal & J. Robinson), pp. 359–374.

Berger, J. (1990) Persistence of different-sized populations: an empirical assessment of rapid extinctions in bighorn sheep. Conserv. Biol., 4, 91–98.

Bernard, H. (2004) Effects of selective logging on the microhabitat-use patterns of non-volant small mammals in a Bornean tropical lowland mixed-dipterocarp forest. Nat. Hum. Act., 8, 1–11.

Bernard, H., Baking, E.L., Giordano, A.J., Wearn, O.R., Hamid, A. & Ahmad, A.H. (2014) Terrestrial mammal species richness and composition in three small forest patches within an oil palm landscape in Sabah, Malaysian Borneo. Mammal Study, 39, 141–154.

Bernard, H., Baking, E.L., Matsubayashi, H. & Ahmad, A.H. (2012) Records of Bornean felids in and around Tabin Wildlife Reserve, Sabah, Malaysia. Cat News, 56, 4–7.

Bernard, H., Fjeldså, J. & Mohamed, M. (2009) A case study on the effects of disturbance and conversion of tropical lowland rain forest on the non-volant small mammals in north Borneo: management implications. Mammal Study, 34, 85–96.

Berry, N.J., Phillips, O.L., Lewis, S.L., Hill, J.K., Edwards, D.P., Tawatao, N.B., Ahmad, N., Magintan, D., Khen, C. V., Maryati, M., Ong, R.C. & Hamer, K.C. (2010) The high value of logged tropical forests: lessons from northern Borneo. Biodivers. Conserv., 19, 985–997.

Berry, N.J., Phillips, O.L., Ong, R.C. & Hamer, K.C. (2008) Impacts of selective logging on tree diversity across a rainforest landscape: the importance of spatial scale. Landsc. Ecol., 23, 915–929.

Beukema, H., Danielsen, F., Vincent, G., Hardiwinoto, S. & Andel, J. (2007) Plant and bird diversity in rubber agroforests in the lowlands of Sumatra, Indonesia. Agrofor. Syst., 70, 217–242.

Bhagwat, S.A. & Willis, K.J. (2008) Agroforestry as a solution to the oil palm debate. Conserv. Biol., 22, 1368–1369.

Bierregaard, R. & Lovejoy, T. (1989) Effects of forest fragmentation on Amazonian understory bird communities. Acta Amaz., 19, 215–241.

Bird, M.I., Taylor, D. & Hunt, C. (2005) Palaeoenvironments of insular Southeast Asia during the Last Glacial Period: a savanna corridor in Sundaland? Quat. Sci. Rev., 24, 2228–2242.

Bischoff, W., Newbery, D.M., Lingenfelder, M., Schnaeckel, R., Petol, G.H., Madani, L. & Ridsdale, C.E. (2005) Secondary succession and dipterocarp recruitment in Bornean rain forest after logging. For. Ecol. Manage., 218, 174–192.

Bivand, R. & Rundel, C. (2014) rgeos: Interface to Geometry Engine - Open Source (GEOS). R package version 0.3-8. Available from http://cran.r-project.org/package=rgeos.

Blanchet, F., Legendre, P. & Borcard, D. (2008) Forward selection of explanatory variables. Ecology, 89, 2623–2632.

Blate, G., Peart, D. & Leighton, M. (1998) Post-dispersal predation on isolated seeds: a comparative study of 40 tree species in a Southeast Asian rainforest. Oikos, 82, 522–538.

Borcard, D. & Legendre, P. (2002) All-scale spatial analysis of ecological data by means of principal coordinates of neighbour matrices. Ecol. Modell., 153, 51–68.

Borcard, D., Legendre, P., Avois-Jacquet, C. & Tuomisto, H. (2004) Dissecting the spatial structure of ecological data at multiple scales. Ecology, 85, 1826–1832.

Bibliography

205

Borcard, D., Legendre, P. & Drapeau, P. (1992) Partialling out the spatial component of ecological variation. Ecology, 73, 1045–1055.

Boulinier, T., Nichols, J. & Sauer, J. (1998) Estimating species richness: the importance of heterogeneity in species detectability. Ecology, 79, 1018–1028.

Bradshaw, C.J., Sodhi, N.S. & Brook, B.W. (2009) Tropical turmoil: a biodiversity tragedy in progress. Front. Ecol. Environ., 7, 79–87.

Brearley, F.Q., Prajadinata, S., Kidd, P.S. & Proctor, J. (2004) Structure and floristics of an old secondary rain forest in Central Kalimantan, Indonesia, and a comparison with adjacent primary forest. For. Ecol. Manage., 195, 385–397.

Brearley, F.Q., Proctor, J., Nagy, L., Dalrymple, G. & Voysey, B.C. (2007) Reproductive phenology over a 10-year period in a lowland evergreen rain forest of central Borneo. J. Ecol., 95, 828–839.

Broadbent, E.N., Asner, G.P., Keller, M., Knapp, D.E., Oliveira, P.J.C. & Silva, J.N. (2008) Forest fragmentation and edge effects from deforestation and selective logging in the Brazilian Amazon. Biol. Conserv., 141, 1745–1757.

Brodie, J. & Giordano, A.J. (2011) Small carnivores of the Maliau Basin, Sabah, Borneo, including a new locality for Hose’s Civet Diplogale hosei. Small Carniv. Conserv., 44, 1–6.

Brodie, J. & Giordano, A.J. (2012a) New high elevation record of the bay cat from Malaysian Borneo. Cat News, 56, 8.

Brodie, J. & Giordano, A.J. (2012b) Density of the Vulnerable Sunda clouded leopard Neofelis diardi in a protected area in Sabah, Malaysian Borneo. Oryx, 46, 427–430.

Brodie, J.F. & Giordano, A. (2013) Lack of trophic release with large mammal predators and prey in Borneo. Biol. Conserv., 163, 58–67.

Brodie, J.F., Giordano, A.J., Zipkin, E.F., Bernard, H., Mohd-Azlan, J. & Ambu, L. (2015) Correlation and persistence of hunting and logging impacts on tropical rainforest mammals. Conserv. Biol., 29, 110–121.

Brodie, J.F., Helmy, O.E., Brockelman, W.Y. & Maron, J.L. (2009a) Functional differences within a guild of tropical mammalian frugivores. Ecology, 90, 688–698.

Brodie, J.F., Helmy, O.E., Brockelman, W.Y. & Maron, J.L. (2009b) Bushmeat poaching reduces the seed dispersal and population growth rate of a mammal-dispersed tree. Ecol. Appl., 19, 854–863.

Brooker, M. & Brooker, L. (2001) Breeding biology, reproductive success and survival of blue-breasted fairy-wrens in fragmented habitat in the Western Australian wheatbelt. Wildl. Res., 28, 205–214.

Brookfield, H., Potter, L. & Byron, Y. (1995) In Place of the Forest: Environmental and Socio-Economic Transformation in Borneo and the Eastern Malay Peninsula. United Nations University Press.

Brotons, L., Mönkkönen, M. & Martin, J.L. (2003) Are fragments islands? Landscape context and density-area relationships in boreal forest birds. Am. Nat., 162, 343–357.

Brown, J. (1984) On the relationship between abundance and distribution of species. Am. Nat., 124, 255–279.

Brown, J. & Kodric-Brown, A. (1977) Turnover rates in insular biogeography: effect of immigration on extinction. Ecology, 58, 445–449.

Bryan, J.E., Shearman, P.L., Asner, G.P., Knapp, D.E., Aoro, G. & Lokes, B. (2013) Extreme differences in forest degradation in Borneo: comparing practices in Sarawak, Sabah, and Brunei. PLoS One, 8, e69679.

Bunge, J. & Fitzpatrick, M. (1993) Estimating the number of species: a review. J. Am. Stat. Assoc., 88, 364–373.

Bibliography

206

Burgess, P.F. (1970) An approach towards a silvicultural system for the hill forests of the Malay Peninsula. Malaysian For., 33, 126–134.

Burivalova, Z., Şekercioğlu, Ç.H. & Koh, L.P. (2014) Thresholds of logging intensity to maintain tropical forest biodiversity. Curr. Biol., 24, 1893–1898.

Burke, D.M. & Nol, E. (1998) Influence of food abundance, nest-site habitat, and forest fragmentation on breeding ovenbirds. Auk, 115, 96–104.

Butchart, S.H.M., Walpole, M., Collen, B., van Strien, A., Scharlemann, J.P.W., Almond, R.E.A., Baillie, J.E.M., Bomhard, B., Brown, C., Bruno, J., Carpenter, K.E., Carr, G.M., Chanson, J., Chenery, A.M., Csirke, J., Davidson, N.C., Dentener, F., Foster, M., Galli, A., et al. (2010) Global biodiversity: indicators of recent declines. Science, 328, 1164–1168.

Cabra-García, J., Bermúdez-Rivas, C., Osorio, A.M. & Chacón, P. (2012) Cross-taxon congruence of α and β diversity among five leaf litter arthropod groups in Colombia. Biodivers. Conserv., 21, 1493–1508.

Caldecott, J.O., Blouch, R.A. & Macdonald, A.A. (1993) The Bearded Pig (Sus barbatus). In Pigs, Peccaries and Hippos Status Survey and Action Plan (ed W.L.R. Oliver), pp. 136–145.

Campos-Arceiz, A. (2009) Shit happens (to be useful)! Use of elephant dung as habitat by amphibians. Biotropica, 41, 406–407.

Campos-Arceiz, A., Traeholt, C., Jaffar, R., Santamaria, L. & Corlett, R.T. (2012) Asian tapirs are no elephants when it comes to seed dispersal. Biotropica, 44, 220–227.

Cannon, C.H., Peart, D. & Leighton, M. (1998) Tree species diversity in commercially logged Bornean rainforest. Science, 281, 1366–1368.

Cannon, C.H., Peart, D.R., Leighton, M. & Kartawinata, K. (1994) The structure of lowland rainforest after selective logging in West Kalimantan, Indonesia. For. Ecol. Manage., 67, 49–68.

Cardoso, P., Borges, P.A. V. & Veech, J.A. (2009) Testing the performance of beta diversity measures based on incidence data: the robustness to undersampling. Divers. Distrib., 15, 1081–1090.

Castelletta, M., Sodhi, N. & Subaraj, R. (2000) Heavy extinctions of forest avifauna in Singapore: lessons for biodiversity conservation in Southeast Asia. Conserv. Biol., 14, 1870–1880.

Castro-Arellano, I., Madrid-Luna, C., Lacher, T.E. & León-Paniagua, L. (2008) Hair-trap efficacy for detecting mammalian carnivores in the tropics. J. Wildl. Manage., 72, 1405–1412.

Ceballos, G. & Ehrlich, P.R. (2002) Mammal population losses and the extinction crisis. Science, 296, 904–907.

Chambers, J.Q., Higuchi, N. & Melack, J.M. (2000) Decomposition and carbon cycling of dead trees in tropical forests of the central Amazon. Oecologia, 122, 380–388.

Chao, A., Chazdon, R.L., Colwell, R.K. & Shen, T.-J. (2005) A new statistical approach for assessing similarity of species composition with incidence and abundance data. Ecol. Lett., 8, 148–159.

Chao, A., Chiu, C.-H. & Hsieh, T.C. (2012) Proposing a resolution to debates on diversity partitioning. Ecology, 93, 2037–2051.

Chao, A. & Jost, L. (2012) Coverage-based rarefaction and extrapolation: standardizing samples by completeness rather than size. Ecology, 93, 2533–2547.

Chao, A. & Shen, T. (2003) Nonparametric estimation of Shannon’s index of diversity when there are unseen species in sample. Environ. Ecol. Stat., 10, 429–443.

Chapman, C.A., Balcomb, S.R., Gillespie, T.R., Skorupa, J.P. & Struhsaker, T.T. (2000) Long-term effects of logging on African primate communities: a 28-year comparison from Kibale National Park, Uganda. Conserv. Biol., 14, 207–217.

Charles, J. & Ang, B. (2010) Non-volant small mammal community responses to fragmentation of

Bibliography

207

kerangas forests in Brunei Darussalam. Biodivers. Conserv., 19, 543–561. Chase, J.M. & Myers, J.A. (2011) Disentangling the importance of ecological niches from stochastic

processes across scales. Philos. Trans. R. Soc. Lond. B. Biol. Sci., 366, 2351–2363. Chave, J., Muller-Landau, H. & Levin, S. (2002) Comparing classical community models: theoretical

consequences for patterns of diversity. Am. Nat., 159, 1–23. Chave, J. & Norden, N. (2007) Changes of species diversity in a simulated fragmented neutral

landscape. Ecol. Modell., 207, 3–10. Chave, J., Réjou-Méchain, M., Búrquez, A., Chidumayo, E., Colgan, M.S., Delitti, W.B.C., Duque,

A., Eid, T., Fearnside, P.M., Goodman, R.C., Henry, M., Martínez-Yrízar, A., Mugasha, W.A., Muller-Landau, H.C., Mencuccini, M., Nelson, B.W., Ngomanda, A., Nogueira, E.M., Ortiz-Malavassi, E., et al. (2014) Improved allometric models to estimate the aboveground biomass of tropical trees. Glob. Chang. Biol., 20, 3177–3190.

Chesson, P. (2000) Mechanisms of maintenance of species diversity. Annu. Rev. Ecol. Syst., 31, 343–366.

Chey, V.K. (2006) Impacts of forest conversion on biodiversity as indicated by moths. Malayan Nat. J., 57, 383–418.

Chey, V.K., Holloway, J. & Speight, M. (1997) Diversity of moths in forest plantations and natural forests in Sabah. Bull. Entomol. Res., 87, 371–385.

Cheyne, S.M. & Macdonald, D.W. (2011) Wild felid diversity and activity patterns in Sabangau peat-swamp forest, Indonesian Borneo. Oryx, 45, 119–124.

Chivers, D.J. (1974) The siamang in Malaya: a field-study of a primate in tropical rain forest. Contrib. Primatol., 4, 1–335.

Chong et al., B.S.H. (2005) Special environmental impact assessment (SEIA) for the proposed oil palm plantation and industrial tree plantation development at Benta Wawasan I and Benta Wawasan IIC, Yayasan Sabah Forest Management Area, Kalabakan and Gunung Rara Forest Reserves, Tawau. Chemsains Konsultant Sdn Bhd, Kota Kinabalu, Sabah, Malaysia.

Chung, A.Y., Eggleton, P., Speight, M.R., Hammond, P.M. & Chey, V.K. (2000) The diversity of beetle assemblages in different habitat types in Sabah, Malaysia. Bull. Entomol. Res., 90, 475–496.

Clark, C.J., Poulsen, J.R., Malonga, R. & Elkan, P.W. (2009) Logging concessions can extend the conservation estate for central African tropical forests. Conserv. Biol., 23, 1281–1293.

Cleary, D., Boyle, T., Setyawati, T., Anggraeni, C.D., Van Loon, E.E. & Menken, S.B. (2007) Bird species and traits associated with logged and unlogged forest in Borneo. Ecol. Appl., 17, 1184–1197.

Cleary, D.F.R. (2003) An examination of scale of assessment, logging and ENSO-induced fires on butterfly diversity in Borneo. Oecologia, 135, 313–321.

Cleary, D.F.R., Genner, M.J., Boyle, T.J.B., Setyawati, T., Angraeti, C.D. & Menken, S.B.J. (2005) Associations of bird species richness and community composition with local and landscape-scale environmental factors in Borneo. Landsc. Ecol., 20, 989–1001.

Cleary, D.F.R. & Mooers, A.O. (2006) Burning and logging differentially affect endemic vs. widely distributed butterfly species in Borneo. Divers. Distrib., 12, 409–416.

Clements, R., Sodhi, N.S., Schilthuizen, M. & Ng, P.K.L. (2006) Limestone karsts of Southeast Asia: Imperiled arks of biodiversity. Bioscience, 56, 733–742.

Cochran, W.G. (1977) Sampling Techniques, 3rd edition. John Wiley & Sons. Coffin, A.W. (2007) From roadkill to road ecology: A review of the ecological effects of roads. J.

Transp. Geogr., 15, 396–406.

Bibliography

208

Cole, L.E.S., Bhagwat, S.A. & Willis, K.J. (2014) Recovery and resilience of tropical forests after disturbance. Nat. Commun., 5, 3906.

Collen, B., Loh, J., Whitmee, S., McRae, L., Amin, R. & Baillie, J.E.M. (2009) Monitoring change in vertebrate abundance: the living planet index. Conserv. Biol., 23, 317–327.

Colón, C.P. (2002) Ranging behaviour and activity of the Malay civet (Viverra tangalunga) in a logged and an unlogged forest in Danum Valley, East Malaysia. J. Zool., 257, 473–485.

Colwell, R.K., Chao, A., Gotelli, N.J., Lin, S.-Y., Mao, C.X., Chazdon, R.L. & Longino, J.T. (2012) Models and estimators linking individual-based and sample-based rarefaction, extrapolation and comparison of assemblages. J. Plant Ecol., 5, 3–21.

Connell, J.H. (1980) Diversity and the coevolution of competitors, or the ghost of competition past. Oikos, 35, 131–138.

Connor, E., Courtney, A. & Yoder, J. (2000) Individuals-area relationships: the relationship between animal population density and area. Ecology, 81, 734–748.

Connor, E. & McCoy, E. (1979) The statistics and biology of the species-area relationship. Am. Nat., 113, 791–833.

Corlett, R. (2007) The impact of hunting on the mammalian fauna of tropical Asian forests. Biotropica, 39, 292–303.

Corlett, R.T. (1998) Frugivory and seed dispersal by vertebrates in the Oriental (Indomalayan) Region. Biol. Rev. Camb. Philos. Soc., 73, 413–448.

Corlett, R.T. (2009a) The Ecology of Tropical East Asia. Oxford University Press, Oxford, UK. Corlett, R.T. (2009b) Seed dispersal distances and plant migration potential in tropical East Asia.

Biotropica, 41, 592–598. Corley, R. (2009) How much palm oil do we need? Environ. Sci. Policy, 12, 134–139. Cribari-Neto, F. & Zeileis, A. (2010) Beta Regression in R. J. Stat. Softw., 34, 1–24. Crist, T.O. & Veech, J.A. (2006) Additive partitioning of rarefaction curves and species-area

relationships: unifying alpha-, beta- and gamma-diversity with sample size and habitat area. Ecol. Lett., 9, 923–932.

Crist, T.O., Veech, J.A., Gering, J.C. & Summerville, K.S. (2003) Partitioning species diversity across landscapes and regions: a hierarchical analysis of alpha, beta, and gamma diversity. Am. Nat., 162, 734–743.

Crooks, K. & Suarez, A. (2001) Extinction and colonization of birds on habitat islands. Conserv. Biol., 15, 159–172.

Curran, L. & Leighton, M. (2000) Vertebrate responses to spatiotemporal variation in seed production of mast-fruiting Dipterocarpaceae. Ecol. Monogr., 70, 101–128.

Curran, L. & Webb, C. (2000) Experimental tests of the spatiotemporal scale of seed predation in mast-fruiting Dipterocarpaceae. Ecol. Monogr., 70, 129–148.

Curran, L.M., Caniago, I., Paoli, G.D., Astiani, D., Kusneti, M., Leighton, M., Nirarita, C. & Haeruman, H. (1999) Impact of El Niño and logging on canopy tree recruitment in Borneo. Science, 286, 2184–2188.

Curran, L.M., Trigg, S.N., McDonald, A.K., Astiani, D., Hardiono, Y.M., Siregar, P., Caniago, I. & Kasischke, E. (2004) Lowland forest loss in protected areas of Indonesian Borneo. Science, 303, 1000–1003.

Cusack, J.J., Wearn, O.R., Bernard, H. & Ewers, R.M. (2015) Influence of microhabitat structure and disturbance on detection of native and non-native murids in logged and unlogged forests of northern Borneo. J. Trop. Ecol., 31, 25–25.

Cushman, S.A. & McGarigal, K. (2004) Hierarchical analysis of forest bird species-environment

Bibliography

209

relationships in the Oregon Coast Range. Ecol. Appl., 14, 1090–1105. Dahaban, Z., Nordin, M. & Bennett, E. (1996) Immediate effects on wildlife of selective logging in a

hill dipterocarp forest in Sarawak: mammals. In Tropical Rainforest Research - Current Issues (eds D. Edwards, W. Booth & S. Choy), pp. 341–346.

Danielsen, F., Beukema, H., Burgess, N.D., Parish, F., Brühl, C.A., Donald, P.F., Murdiyarso, D., Phalan, B., Reijnders, L., Struebig, M. & Fitzherbert, E.B. (2009) Biofuel plantations on forested lands: double jeopardy for biodiversity and climate. Conserv. Biol., 23, 348–358.

Danielsen, F. & Heegaard, M. (1995) Impact of logging and plantation development on species diversity: a case study from Sumatra. In Management of tropical forests: towards an integrated perspective (ed Ø. Sandbukt), pp. 73–92.

Dauvergne, P. (1997) Shadows in the Forest: Japan and the Politics of Timber in the Southeast Asia. MIT Press, Cambridge, Massachusetts.

Davies, G. (1986) The orang-utan in Sabah. Oryx, 20, 40–45. Davies, G., Heydon, M., Leader-Williams, N., MacKinnon, J. & Newing, H. (2001) The effects of

logging on tropical forest ungulates. In The cutting edge: conserving wildlife in logged tropical forest (eds R.A. Fimbel, A. Grajal & J.G. Robinson), pp. 93–124.

Davies, G. & Payne, J. (1982) A faunal survey of Sabah. WWF-Malaysia, Kuala Lumpur. Davis, A.J., Holloway, J.D., Huijbregts, H., Krikken, J., Kirk-Spriggs, A.H. & Sutton, S.L. (2001)

Dung beetles as indicators of change in the forests of northern Borneo. J. Appl. Ecol., 38, 593–616.

Debinski, D. & Holt, R. (2000) A survey and overview of habitat fragmentation experiments. Conserv. Biol., 14, 342–355.

Debinski, D.M., Ray, C. & Saveraid, E.H. (2001) Species diversity and the scale of the landscape mosaic: Do scales of movement and patch size affect diversity? Biol. Conserv., 98, 179–190.

Declerck, S.A.J., Coronel, J.S., Legendre, P. & Brendonck, L. (2011) Scale dependency of processes structuring metacommunities of cladocerans in temporary pools of High-Andes wetlands. Ecography, 34, 296–305.

Dénes, F. V., Silveira, L.F. & Beissinger, S.R. (2015) Estimating abundance of unmarked animal populations: accounting for imperfect detection and other sources of zero inflation. Methods Ecol. Evol., [in press].

Dennis, R.A., Meijaard, E., Nasi, R. & Gustafsson, L. (2008) Biodiversity conservation in Southeast Asian timber concessions: a critical evaluation of policy mechanisms and guidelines. Ecol. Soc., 13, 25.

Denwood, M.J. (2015) runjags: An R package providing interface utilities, parallel computing methods and additional distributions for MCMC models in JAGS. Available from http://cran.r-project.org/package=runjags.

Diamond, J. (1972) Biogeographic kinetics: estimation of relaxation times for avifaunas of southwest Pacific islands. Proc. Natl. Acad. Sci. U. S. A., 69, 3199–3203.

Didham, R.K. (2011) Life after logging: strategic withdrawal from the Garden of Eden or tactical error for wilderness conservation? Biotropica, 43, 393–395.

Didham, R.K., Kapos, V. & Ewers, R.M. (2012) Rethinking the conceptual foundations of habitat fragmentation research. Oikos, 121, 161–170.

Didham, R.K. & Lawton, J.H. (1999) Edge structure determines the magnitude of changes in microclimate and vegetation structure in tropical forest fragments. Biotropica, 31, 17–30.

Donovan, S.E., Griffiths, G.J.K., Homathevi, R. & Winder, L. (2007) The spatial pattern of soil-dwelling termites in primary and logged forest in Sabah, Malaysia. Ecol. Entomol., 32, 1–10.

Bibliography

210

Donovan, T., Jones, P., Annand, E. & Thompson III, F. (1997) Variation in local-scale edge effects: mechanisms and landscape context. Ecology, 78, 2064–2075.

Dooley, J.J.L. & Bowers, M.A.M. (1998) Demographic responses to habitat fragmentation: experimental tests at the landscape and patch scale. Ecology, 79, 969–980.

Dorazio, R., Gotelli, N. & Ellison, A. (2011) Modern methods of estimating biodiversity from presence-absence surveys. In Biodiversity Loss in a Changing Planet (eds O. Grillo & G. Venora), pp. 277–302.

Dorazio, R.M., Kéry, M., Royle, J.A. & Plattner, M. (2010) Models for inference in dynamic metacommunity systems. Ecology, 91, 2466–2475.

Dorazio, R.R.M., Royle, J.A., Söderström, B. & Glimskär, A. (2006) Estimating species richness and accumulation by modeling species occurrence and detectability. Ecology, 87, 842–854.

Dormann, C.F., Schweiger, O., Augenstein, I., Bailey, D., Billeter, R., de Blust, G., DeFilippi, R., Frenzel, M., Hendrickx, F., Herzog, F., Klotz, S., Liira, J., Maelfait, J.-P., Schmidt, T., Speelmans, M., van Wingerden, W.K.R.E. & Zobel, M. (2007) Effects of landscape structure and land-use intensity on similarity of plant and animal communities. Glob. Ecol. Biogeogr., 16, 774–787.

Dornelas, M., Gotelli, N.J., McGill, B., Shimadzu, H., Moyes, F., Sievers, C. & Magurran, A.E. (2014) Assemblage time series reveal biodiversity change but not systematic loss. Science, 344, 296–299.

Doughty, C.E., Wolf, A. & Malhi, Y. (2013) The legacy of the Pleistocene megafauna extinctions on nutrient availability in Amazonia. Nat. Geosci., 6, 761–764.

Drakare, S., Lennon, J.J. & Hillebrand, H. (2006) The imprint of the geographical, evolutionary and ecological context on species-area relationships. Ecol. Lett., 9, 215–227.

Dray, S., Legendre, P. & Peres-Neto, P.R. (2006) Spatial modelling: a comprehensive framework for principal coordinate analysis of neighbour matrices (PCNM). Ecol. Modell., 196, 483–493.

Dray, S., Pélissier, R., Couteron, P., Fortin, M.-J., Legendre, P., Peres-Neto, P.R., Bellier, E., Bivand, R., Blanchet, F.G., De Caceres, M., Dufour, A.-B., Heegaard, M., Jombart, T., Munoz, F., Oksanen, J., Thioulouse, J. & Wagner, H.H. (2012) Community ecology in the age of multivariate multiscale spatial analysis. Ecol. Monogr., 82, 257–275.

Duckworth, J.W., Batters, G., Belant, J.L., Bennett, E.L., Brunner, J., Burton, J., Challender, D.W.S., Cowling, V., Duplaix, N., Harris, J.D., Hedges, S., Long, B., Mahood, S.P., McGowan, P.J.K., McShea, W.J., Oliver, W.L.R., Perkin, S., Rawson, B.M., Shepherd, C.R., et al. (2012) Why South-East Asia should be the world’s priority for averting imminent species extinctions, and a call to join a developing cross-institutional programme to tackle this urgent issue. Sapiens, 5, 76–95.

Duff, A.B., Hall, R.A. & Marsh, C.W. (1984) A survey of wildlife in and around a commercial tree plantation in Sabah. Malaysian For., 47, 197–213.

Dumbrell, A.J., Clark, E.J., Frost, G.A., Randell, T.E., Pitchford, J.W. & Hill, J.K. (2008) Changes in species diversity following habitat disturbance are dependent on spatial scale: theoretical and empirical evidence. J. Appl. Ecol., 45, 1531–1539.

Dumbrell, A.J. & Hill, J.K. (2005) Impacts of selective logging on canopy and ground assemblages of tropical forest butterflies: implications for sampling. Biol. Conserv., 125, 123–131.

Dunn, R.R. (2004) Managing the tropical landscape: a comparison of the effects of logging and forest conversion to agriculture on ants, birds, and lepidoptera. For. Ecol. Manage., 191, 215–224.

Edwards, D.P., Ansell, F.A., Ahmad, A.H., Nilus, R. & Hamer, K.C. (2009) The value of rehabilitating logged rainforest for birds. Conserv. Biol., 23, 1628–1633.

Edwards, D.P., Fisher, B. & Boyd, E. (2010a) Protecting degraded rainforests: enhancement of forest

Bibliography

211

carbon stocks under REDD+. Conserv. Lett., 3, 313–316. Edwards, D.P., Fisher, B. & Wilcove, D.S. (2012) High Conservation Value or high confusion value?

Sustainable agriculture and biodiversity conservation in the tropics. Conserv. Lett., 5, 20–27. Edwards, D.P., Hodgson, J.A., Hamer, K.C., Mitchell, S.L., Ahmad, A.H., Cornell, S.J. & Wilcove,

D.S. (2010b) Wildlife-friendly oil palm plantations fail to protect biodiversity effectively. Conserv. Lett., 3, 236–242.

Edwards, D.P., Larsen, T.H., Docherty, T.D.S., Ansell, F.A., Hsu, W.W., Derhé, M.A., Hamer, K.C. & Wilcove, D.S. (2011) Degraded lands worth protecting: the biological importance of Southeast Asia’s repeatedly logged forests. Proc. R. Soc. London Ser. B Biol. Sci., 278, 82–90.

Edwards, D.P. & Laurance, W.F. (2011) Loophole in forest plan for Indonesia. Nature, 477, 33. Edwards, D.P., Magrach, A., Woodcock, P., Ji, Y., Lim, N.T.-L., Edwards, F.A., Larsen, T.H., Hsu,

W.W., Benedick, S., Vun Khen, C., Chung, A.Y.C., Reynolds, G., Fisher, B., Laurance, W.F., Wilcove, D.S., Hamer, K.C. & Yu, D.W. (2014a) Selective-logging and oil palm: multitaxon impacts, biodiversity indicators, and trade-offs for conservation planning. Ecol. Appl., 24, 2029–2049.

Edwards, F.A., Edwards, D.P., Sloan, S. & Hamer, K.C. (2014b) Sustainable management in crop monocultures: the impact of retaining forest on oil palm yield. PLoS One, 9, e91695.

Efford, M. & Dawson, D. (2012) Occupancy in continuous habitat. Ecosphere, 3, 1–15. Eggleton, P., Homathevi, R., Jeeva, D., Jones, D.T., Davies, R.G. & Maryati, M. (1997) The species

richness and composition of termites (Isoptera) in primary and regenerating lowland dipterocarp forest in Sabah, East Malaysia. Ecotropica, 3, 119–128.

Eggleton, P., Homathevi, R., Jones, D.T., MacDonald, J.A., Jeeva, D., Bignell, D.E., Davies, R.G. & Maryati, M. (1999) Termite assemblages, forest disturbance and greenhouse gas fluxes in Sabah, East Malaysia. Philos. Trans. R. Soc. Lond. B. Biol. Sci., 354, 1791–1802.

Elton, C. (1927) Animal Ecology. Sedgwick and Jackson, London. Emmons, L.H. (2000) Tupai: A Field Study of Bornean Treeshrews. University of California Press,

Los Angeles. Ewers, R.M., Boyle, M.J.W., Gleave, R.A., Plowman, N.S., Benedick, S., Bernard, H., Bishop, T.R.,

Bakhtiar, E.Y., Chey, V.K., Chung, A.Y.C., Davies, R.G., Edwards, D.P., Eggleton, P., Fayle, T.M., Hardwick, S.R., Homathevi, R., Kitching, R.L., Khoo, M.S., Luke, S.H., et al. (2015) Logging cuts the functional importance of invertebrates in tropical rainforest. Nat. Commun., 6, 6836.

Ewers, R.M. & Didham, R.K. (2006) Confounding factors in the detection of species responses to habitat fragmentation. Biol. Rev. Camb. Philos. Soc., 81, 117–142.

Ewers, R.M., Didham, R.K., Fahrig, L., Ferraz, G., Hector, A., Holt, R.D., Kapos, V., Reynolds, G., Sinun, W., Snaddon, J.L. & Turner, E.C. (2011) A large-scale forest fragmentation experiment: the Stability of Altered Forest Ecosystems Project. Philos. Trans. R. Soc. Lond. B. Biol. Sci., 366, 3292–3302.

Fahrig, L. (2002) Effect of habitat fragmentation on the extinction threshold: a synthesis. Ecol. Appl., 12, 346–353.

Fahrig, L. (2003) Effects of habitat fragmentation on biodiversity. Annu. Rev. Ecol. Evol. Syst., 34, 487–515.

Fahrig, L. & Rytwinski, T. (2009) Effects of roads on animal abundance: an emperical review and synthesis. Ecol. Soc., 14, 21–41.

FAO (2011) Southeast Asia Subregional Report: Asia-Pacific Forestry Sector Outlook Study II. Food and Agriculture Organisation of the United Nations, Bangkok, Thailand.

FAO (2012) FAOSTAT database. Food and Agriculture Organisation of the United Nations.

Bibliography

212

Available from http://faostat.fao.org. Fayle, T.M., Turner, E.C., Basset, Y., Ewers, R.M., Reynolds, G. & Novotny, V. (2015) Whole-

ecosystem experimental manipulations of tropical forests. Trends Ecol. Evol., [in press]. Fazey, I., Fischer, J. & Lindenmayer, D.B. (2005) What do conservation biologists publish? Biol.

Conserv., 124, 63–73. Ferraz, G., Russell, G.J., Stouffer, P.C., Bierregaard, R.O., Pimm, S.L. & Lovejoy, T.E. (2003) Rates

of species loss from Amazonian forest fragments. Proc. Natl. Acad. Sci. U. S. A., 100, 14069–14073.

Ferrier, S. (2002) Mapping spatial pattern in biodiversity for regional conservation planning: where to from here? Syst. Biol., 51, 331–363.

Fischer, J. & Lindenmayer, D. (2005) Perfectly nested or significantly nested – an important difference for conservation management. Oikos, 109, 485–489.

Fischer, J. & Lindenmayer, D. (2006) Beyond fragmentation: the continuum model for fauna research and conservation in human-modified landscapes. Oikos, 2, 473–480.

Fischer, J., Lindenmayer, D.B. & Fazey, I. (2004) Appreciating ecological complexity: habitat contours as a conceptual landscape model. Conserv. Biol., 18, 1245–1253.

Fisher, B., Edwards, D.P., Giam, X. & Wilcove, D.S. (2011a) The high costs of conserving Southeast Asia’s lowland rainforests. Front. Ecol. Environ., 9, 329–334.

Fisher, B., Edwards, D.P., Larsen, T.H., Ansell, F.A., Hsu, W.W., Roberts, C.S. & Wilcove, D.S. (2011b) Cost-effective conservation: Calculating biodiversity and logging trade-offs in Southeast Asia. Conserv. Lett., 4, 443–450.

Fitzherbert, E.B., Struebig, M.J., Morel, A., Danielsen, F., Brühl, C.A., Donald, P.F. & Phalan, B. (2008) How will oil palm expansion affect biodiversity? Trends Ecol. Evol., 23, 538–545.

Flather, C.H. & Bevers, M. (2002) Patchy reaction-diffusion and population abundance: the relative importance of habitat amount and arrangement. Am. Nat., 159, 40–56.

Flint, E. (1994) Changes in land use in South and Southeast Asia from 1880 to 1980: a data base prepared as part of a coordinated research program on carbon fluxes in the tropics. Chemosphere, 29, 1015–1062.

Foster, W.A., Snaddon, J.L., Turner, E.C., Fayle, T.M., Cockerill, T.D., Ellwood, M.D.F., Broad, G.R., Chung, A.Y.C., Eggleton, P., Khen, C.V. & Yusah, K.M. (2011) Establishing the evidence base for maintaining biodiversity and ecosystem function in the oil palm landscapes of South East Asia. Philos. Trans. R. Soc. Lond. B. Biol. Sci., 366, 3277–3291.

Fredriksson, G.M., Wich, S.A. & Trisno (2006) Frugivory in sun bears (Helarctos malayanus) is linked to El Niño-related fluctuations in fruiting phenology, East Kalimantan, Indonesia. Biol. J. Linn. Soc., 89, 489–508.

Fukuda, D. & Tisen, O. (2009) Bat diversity in the vegetation mosaic around a lowland dipterocarp forest of Borneo. Raffles Bull. Zool., 57, 213–221.

Furey, N.M., Mackie, I.J. & Racey, P.A. (2010) Bat diversity in Vietnamese limestone karst areas and the implications of forest degradation. Biodivers. Conserv., 19, 1821–1838.

van Gardingen, P.R., McLeish, M.J., Phillips, P.D., Fadilah, D., Tyrie, G. & Yasman, I. (2003) Financial and ecological analysis of management options for logged-over Dipterocarp forests in Indonesian Borneo. For. Ecol. Manage., 183, 1–29.

Gardner, B., Reppucci, J., Lucherini, M. & Royle, J.A. (2010) Spatially explicit inference for open populations: estimating demographic parameters from camera-trap studies. Ecology, 91, 3376–3383.

Gardner, T.A., Barlow, J., Chazdon, R., Ewers, R.M., Harvey, C.A., Peres, C.A. & Sodhi, N.S. (2009) Prospects for tropical forest biodiversity in a human-modified world. Ecol. Lett., 12, 561–582.

Bibliography

213

Gardner, T.A., Barlow, J., Parry, L.W., Peres, C.A., Mpeg, G., Perimetral, A. & Firme, B.T. (2007) Predicting the uncertain future of tropical forest species in a data vacuum. Biotropica, 39, 25–30.

Gathorne-Hardy, F., Jones, D. & Syaukani (2002) A regional perspective on the effects of human disturbance on the termites of Sundaland. Biodivers. Conserv., 11, 1991–2006.

Gaveau, D.L.A., Sloan, S., Molidena, E., Yaen, H., Sheil, D., Abram, N.K., Ancrenaz, M., Nasi, R., Quinones, M., Wielaard, N. & Meijaard, E. (2014) Four decades of forest persistence, clearance and logging on Borneo. PLoS One, 9, e101654.

Gelman, A. (2006) Prior distributions for variance parameters in hierarchical models. Bayesian Anal., 1, 515–533.

Gelman, A. & Rubin, D. (1992) Inference from iterative simulation using multiple sequences. Stat. Sci., 7, 457–511.

Ghazoul, J. (2002) Impact of logging on the richness and diversity of forest butterflies in a tropical dry forest in Thailand. Biodivers. Conserv., 11, 521–541.

Giam, X., Clements, G.R., Aziz, S.A., Chong, K.Y. & Miettinen, J. (2011) Rethinking the ‘back to wilderness’ concept for Sundaland’s forests. Biol. Conserv., 144, 3149–3152.

Giam, X., Koh, L.P., Tan, H.H., Miettinen, J., Tan, H.T.W. & Ng, P.K.L. (2012) Global extinctions of freshwater fishes follow peatland conversion in Sundaland. Front. Ecol. Environ., 10, 465–470.

Gibbs, H.K., Ruesch, A.S., Achard, F., Clayton, M.K., Holmgren, P., Ramankutty, N. & Foley, J.A. (2010) Tropical forests were the primary sources of new agricultural land in the 1980s and 1990s. Proc. Natl. Acad. Sci. U. S. A., 107, 16732–16737.

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. (2011) Primary forests are irreplaceable for sustaining tropical biodiversity. Nature, 478, 378–381.

Gibson, L., Lynam, A.J., Bradshaw, C.J.A., He, F., Bickford, D.P., Woodruff, D.S., Bumrungsri, S. & Laurance, W.F. (2013) Near-complete extinction of native small mammal fauna 25 years after forest fragmentation. Science, 341, 1508–1510.

Goldammer, J.G. (2006) History of equatorial vegetation fires and fire research in Southeast Asia before the 1997–98 episode: a reconstruction of creeping environmental changes. Mitig. Adapt. Strateg. Glob. Chang., 12, 13–32.

Goldewijk, K.K. (2001) Estimating global land use change over the past 300 years: the HYDE Database. Global Biogeochem. Cycles, 15, 417–433.

Gossner, M.M., Getzin, S., Lange, M., Pašalić, E., Türke, M., Wiegand, K. & Weisser, W.W. (2013) The importance of heterogeneity revisited from a multiscale and multitaxa approach. Biol. Conserv., 166, 212–220.

Gotelli, N. & Colwell, R. (2001) Quantifying biodiversity: procedures and pitfalls in the measurement and comparison of species richness. Ecol. Lett., 4, 379–391.

Gotelli, N.J. & Chao, A. (2013) Measuring and estimating species richness, species diversity, and biotic similarity from sampling data. In Encyclopedia of Biodiversity. 2nd Edition. (ed S.A. Levin), pp. 195–211.

Gotelli, N.J. & Ulrich, W. (2012) Statistical challenges in null model analysis. Oikos, 121, 171–180. Grassman, L.I., Tewes, M.E., Silvy, N.J. & Kreetiyutanont, K. (2005) Ecology of three sympatric

felids in a mixed evergreen forest in north-central Thailand. J. Mammal., 86, 29–38. Gravel, D., Canham, C.D., Beaudet, M. & Messier, C. (2006) Reconciling niche and neutrality: the

continuum hypothesis. Ecol. Lett., 9, 399–409. Gray, C.L., Slade, E.M., Mann, D.J. & Lewis, O.T. (2014) Do riparian reserves support dung beetle

biodiversity and ecosystem services in oil palm-dominated tropical landscapes? Ecol. Evol., 4, 1049–1060.

Bibliography

214

Gray, M.A., Baldauf, S.L., Mayhew, P.J. & Hill, J.K. (2007) The response of avian feeding guilds to tropical forest disturbance. Conserv. Biol., 21, 133–141.

Green, R.E., Cornell, S.J., Scharlemann, J.P.W. & Balmford, A. (2005) Farming and the fate of wild nature. Science, 307, 550–555.

Greenberg, R., Bichier, P., Angon, A.C. & Reitsma, R. (1997) Bird populations in shade and sun coffee plantations in central Guatemala. Conserv. Biol., 11, 448–459.

Greenpeace (2013) Identifying High Carbon Stock (HCS) forest for protection. Available from http://www.greenpeace.org/international/Global/international/briefings/forests/2013/HCS-Briefing-2013.pdf.

Gregory, S.D., Brook, B.W., Goossens, B., Ancrenaz, M., Alfred, R., Ambu, L.N. & Fordham, D.A. (2012) Long-term field data and climate-habitat models show that orangutan persistence depends on effective forest management and greenhouse gas mitigation. PLoS One, 7, e343846.

Griffith, D.M., Veech, J.A. & Marsh, C.J. (2014) cooccur: Probabilistic species co-occurrence analysis in R. Available from http://cran.r-project.org/package=cooccur.

Grimbacher, P.S., Catterall, C.P. & Kitching, R.L. (2006) Beetle species’ responses suggest that microclimate mediates fragmentation effects in tropical Australian rainforest. Austral Ecol., 31, 458–470.

Grinnell, J. (1917) Field tests of theories concerning distributional control. Am. Nat., 51, 115–128. Guillera-Arroita, G., Lahoz-Monfort, J.J., MacKenzie, D.I., Wintle, B.A. & McCarthy, M.A. (2014)

Ignoring imperfect detection in biological surveys is dangerous: a response to ‘fitting and interpreting occupancy models’. PLoS One, 9, e99571.

Guillera-Arroita, G., Ridout, M.S. & Morgan, B.J.T. (2010) Design of occupancy studies with imperfect detection. Methods Ecol. Evol., 1, 131–139.

Gunarso, P., Hartoyo, M.E., Agus, F. & Killeen, T.J. (2013) Oil palm and land use change in Indonesia, Malaysia and Papua New Guinea. In Reports from the Technical Panels of the 2nd Greenhouse Gas Working Group, Roundtable on Sustainable Palm Oil (RSPO) pp. 29–64.

Gustafsson, L., Baker, S.C., Bauhus, J., Beese, W.J., Brodie, A., Kouki, J., Lindenmayer, D.B., Lõhmus, A., Pastur, G.M., Messier, C., Neyland, M., Palik, B., Sverdrup-thygeson, A., Volney, W.J.A., Wayne, A. & Franklin, J.F. (2012) Retention forestry to maintain multifunctional forests: A world perspective. Bioscience, 62, 633–645.

Haddad, N.M., Brudvig, L.A., Clobert, J., Davies, K.F., Gonzalez, A., Holt, R.D., Lovejoy, T.E., Sexton, J.O., Austin, M.P., Collins, C.D., Cook, W.M., Damschen, E.I., Ewers, R.M., Foster, B.L., Jenkins, C.N., King, A.J., Laurance, W.F., Levey, D.J., Margules, C.R., et al. (2015) Habitat fragmentation and its lasting impact on Earth’s ecosystems. Sci. Adv., 1, e1500052.

Halley, J.M. & Iwasa, Y. (2011) Neutral theory as a predictor of avifaunal extinctions after habitat loss. Proc. Natl. Acad. Sci. U. S. A., 108, 2316–2321.

Hamer, K.C. & Hill, J.K. (2000) Scale-dependent effects of habitat disturbance on species richness in tropical forests. Conserv. Biol., 14, 1435–1440.

Hamer, K.C., Hill, J.K., Benedick, S., Mustaffa, N., Sherratt, T.N., Maryati, M. & K., C. V. (2003) Ecology of butterflies in natural and selectively logged forests of northern Borneo: the importance of habitat heterogeneity. J. Appl. Ecol., 40, 150–162.

Hansen, M.C., Potapov, P. V, Moore, R., Hancher, M., Turubanova, S.A., Tyukavina, A., Thau, D., Stehman, S. V, Goetz, S.J., Loveland, T.R., Kommareddy, A., Egorov, A., Chini, L., Justice, C.O. & Townshend, J.R.G. (2013) High-resolution global maps of 21st-century forest cover change. Science, 342, 850–853.

Hansen, M.C., Stehman, S. V, Potapov, P. V, Loveland, T.R., Townshend, J.R.G., DeFries, R.S., Pittman, K.W., Arunarwati, B., Stolle, F., Steininger, M.K., Carroll, M. & Dimiceli, C. (2008)

Bibliography

215

Humid tropical forest clearing from 2000 to 2005 quantified by using multitemporal and multiresolution remotely sensed data. Proc. Natl. Acad. Sci. U. S. A., 105, 9439–9444.

Hanski, I. (1998) Metapopulation dynamics. Nature, 396, 41–49. Hanski, I., Pakkala, T., Kuussaari, M. & Lei, G. (1995) Metapopulation persistence of an endangered

butterfly in a fragmented landscape. Oikos, 72, 21–28. Hanski, I., Zurita, G.A., Bellocq, M.I. & Rybicki, J. (2013) Species-fragmented area relationship.

Proc. Natl. Acad. Sci. U. S. A., 110, 12715–12720. Hardin, G. (1960) The competitive exclusion principle. Science, 131, 1292–1297. Hardwick, S.R., Toumi, R., Pfeifer, M., Turner, E.C., Nilus, R. & Ewers, R.M. (2015) The

relationship between leaf area index and microclimate in tropical forest and oil palm plantation: forest disturbance drives changes in microclimate. Agric. For. Meteorol., 201, 187–195.

Harmsen, B., Foster, R. & Silver, S. (2009) Differential use of trails by forest mammals and the implications for camera-trap studies: a case study from Belize. Biotropica, 42, 126–133.

Harmsen, B.J., Foster, R.J. & Doncaster, C.P. (2010) Heterogeneous capture rates in low density populations and consequences for capture-recapture analysis of camera-trap data. Popul. Ecol., 53, 253–259.

Harrison, R.D., Tan, S., Plotkin, J.B., Slik, F., Detto, M., Brenes, T., Itoh, A. & Davies, S.J. (2013) Consequences of defaunation for a tropical tree community. Ecol. Lett., 16, 687–694.

Harrison, S. & Bruna, E. (1999) Habitat fragmentation and large-scale conservation: what do we know for sure? Ecography, 22, 225–232.

Harrison, X.A. (2014) Using observation-level random effects to model overdispersion in count data in ecology and evolution. PeerJ, 2, e616.

Harvey, P. (2013) ExifTool, version 9.41. Available from http://owl.phy.queensu.ca/~phil/exiftool/. Hassall, M., Jones, D.T., Taiti, S., Latipi, Z., Sutton, S.L. & Mohammed, M. (2006) Biodiversity

and abundance of terrestrial isopods along a gradient of disturbance in Sabah, East Malaysia. Eur. J. Soil Biol., 42, S197–S207.

Hautier, Y., Saner, P., Philipson, C., Bagchi, R., Ong, R.C. & Hector, A. (2010) Effects of seed predators of different body size on seed mortality in Bornean logged forest. PLoS One, 5, e11651.

Hearn, A., Sanderson, J., Ross, J., Wilting, A. & Sunarto, S. (2008) Pardofelis badia. IUCN 2012. IUCN Red List of Threatened Species. Version 2012.2. Available from www.iucnredlist.org.

Hearn, A.J. & Ross, J. (2009) Flat-headed cat photographed in Danum. Available from http://borneanwildcat.blogspot.co.uk/2009/04/flat-headed-cat-photographed-in-danum.html.

Heffner, R., Butler, M. & Reilly, C. (1996) Pseudoreplication revisited. Ecology, 77, 2558–2562. Henle, K., Davies, K.F., Kleyer, M., Margules, C. & Settele, J. (2004) Predictors of species sensitivity

to fragmentation. Biodivers. Conserv., 13, 207–251. Heydon, M. & Bulloh, P. (1997) Mousedeer densities in a tropical rainforest: the impact of selective

logging. J. Appl. Ecol., 34, 484–496. Heydon, M.J. & Bulloh, P. (1996) The impact of selective logging on sympatric civet species in

Borneo. Oryx, 30, 31–36. Hijmans, R.J. (2014) raster: Geographic data analysis and modeling. R package version 2.3-0.

Available from http://cran.r-project.org/package=raster. Hill, J. (1999) Butterfly spatial distribution and habitat requirements in a tropical forest: impacts of

selective logging. J. Appl. Ecol., 36, 564–572. Hill, J. & Hamer, K. (2004) Determining impacts of habitat modification on diversity of tropical

Bibliography

216

forest fauna: the importance of spatial scale. J. Appl. Ecol., 41, 744–754. Hill, J., Hamer, K., Lace, L. & Banham, W. (1995) Effects of selective logging on tropical forest

butterflies on Buru, Indonesia. J. Appl. Ecol., 32, 754–760. Hill, J., Hamer, K., Tangah, J. & Dawood, M. (2001) Ecology of tropical butterflies in rainforest gaps.

Oecologia, 128, 294–302. Hill, J.K., Gray, M.A., Khen, C.V., Benedick, S., Tawatao, N. & Hamer, K.C. (2011) Ecological

impacts of tropical forest fragmentation: how consistent are patterns in species richness and nestedness? Philos. Trans. R. Soc. Lond. B. Biol. Sci., 366, 3265–3276.

HilleRisLambers, J., Adler, P.B., Harpole, W.S., Levine, J.M. & Mayfield, M.M. (2012) Rethinking community assembly through the lens of coexistence theory. Annu. Rev. Ecol. Evol. Syst., 43, 227–248.

Hoffmann, M., Hilton-Taylor, C., Angulo, A., Böhm, M., Brooks, T.M., Butchart, S.H.M., Carpenter, K.E., Chanson, J., Collen, B., Cox, N.A., Darwall, W.R.T., Dulvy, N.K., Harrison, L.R., Katariya, V., Pollock, C.M., Quader, S., Richman, N.I., Rodrigues, A.S.L., Tognelli, M.F., et al. (2010) The impact of conservation on the status of the world’s vertebrates. Science, 330, 1503–1509.

Houlahan, J.E., Currie, D.J., Cottenie, K., Cumming, G.S., Ernest, S.K.M., Findlay, C.S., Fuhlendorf, S.D., Gaedke, U., Legendre, P., Magnuson, J.J., McArdle, B.H., Muldavin, E.H., Noble, D., Russell, R., Stevens, R.D., Willis, T.J., Woiwod, I.P. & Wondzell, S.M. (2007) Compensatory dynamics are rare in natural ecological communities. Proc. Natl. Acad. Sci. U. S. A., 104, 3273–3277.

Howlett, B.E. & Davidson, D.W. (2003) Effects of seed availability, site conditions, and herbivory on pioneer recruitment after logging in Sabah, Malaysia. For. Ecol. Manage., 184, 369–383.

Hsieh, T.C. (2013) iNEXT: iNterpolation and EXTrapolation for species diversity. R package version 1.0. Available from http://chao.stat.nthu.edu.tw/blog/software-download/.

Hubbell, S.P. (2001) The Unified Neutral Theory of Biodiversity and Biogeography. Princeton University Press.

Hughes, J. & Macdonald, D.W. (2013) A review of the interactions between free-roaming domestic dogs and wildlife. Biol. Conserv., 157, 341–351.

Hunter, L. (2011) Carnivores of the World. Princeton University Press. Hurlbert, S. (1984) Pseudoreplication and the design of ecological field experiments. Ecol. Monogr.,

54, 187–211. Hutchinson, G.E. (1957) Concluding remarks. Cold Spring Harb. Symp. Quant. Biol., 22, 415–427. Ichikawa, M. (2007) Degradation and loss of forest land and land-use changes in Sarawak, East

Malaysia: a study of native land use by the Iban. Ecol. Res., 22, 403–413. Ickes, K. (2001) Hyper-abundance of native wild pigs (Sus scrofa) in a lowland dipterocarp rain forest

of Peninsular Malaysia. Biotropica, 33, 682–690. Ickes, K., Dewalt, S.J. & Appanah, S. (2001) Effects of native pigs (Sus scrofa) on woody

understorey vegetation in a Malaysian lowland rain forest. J. Trop. Ecol., 17. Iknayan, K.J., Tingley, M.W., Furnas, B.J. & Beissinger, S.R. (2013) Detecting diversity: emerging

methods to estimate species diversity. Trends Ecol. Evol., 29, 97–106. Imai, N., Samejima, H., Langner, A., Ong, R.C., Kita, S., Titin, J., Chung, A.Y.C., Lagan, P., Lee,

Y.F. & Kitayama, K. (2009) Co-benefits of sustainable forest management in biodiversity conservation and carbon sequestration. PLoS One, 4, e8267.

Isabirye-Basuta, G. & Kasenene, J.M. (1987) Small rodent populations in selectively felled and mature tracts of Kibale Forest, Uganda. Biotropica, 19, 260–266.

Bibliography

217

IUCN (2014) The IUCN Red List of Threatened Species. Version 2014.3. Available from http://www.iucnredlist.org.

IUCN/UNEP (2012) The World Database on Protected Areas (WDPA). Cambridge, UK. Janzen, D.H. (1986) The future of tropical ecology. Annu. Rev. Ecol. Syst., 17, 305–324. Jennings, A.P., Naim, M., Advento, A.D., Aryawan, A.A.K., Ps, S., Caliman, J.-P., Verwilghen, A. &

Veron, G. (2015) Diversity and occupancy of small carnivores within oil palm plantations in central Sumatra, Indonesia. Mammal Res., 60, 181–188.

Jennings, A.P., Seymour, A.S. & Dunstone, N. (2006) Ranging behaviour, spatial organization and activity of the Malay civet (Viverra tangalunga) on Buton Island, Sulawesi. J. Zool., 268, 63–71.

Johns, A. (1985) Behavioral responses of two Malaysian primates (Hylobates lar and Presbytis melalophos) to selective logging: vocal behavior, territoriality, and nonemigration. Int. J. Primatol., 6, 423–433.

Johns, A. (1986a) Effects of selective logging on the behavioral ecology of West Malaysian primates. Ecology, 67, 684–694.

Johns, A. (1986b) Effects of selective logging on the ecological organization of a peninsular Malaysian rainforest avifauna. Forktail, 1, 65–79.

Johns, A. (1989) Recovery of a peninsular Malaysian rainforest avifauna following selective timber logging: the first twelve years. Forktail, 4, 89–105.

Johns, A. (1992) Vertebrate responses to selective logging: implications for the design of logging systems. Philos. Trans. R. Soc. B Biol. Sci., 335, 437–442.

Johns, A. & Skorupa, J. (1987) Responses of rain-forest primates to habitat disturbance: a review. Int. J. Primatol., 8, 157–191.

Johns, A.D. (1983) Tropical forest logging and primates - can they co-exist? Oryx, 17, 114–118. Johns, A.D. (1987) The use of primary and selectively logged rainforest by Malaysian hornbills

(Bucerotidae) and implications for their conservation. Biol. Conserv., 40, 179–190. Johns, A.D. (1988) Effects of ‘selective’ timber extraction on rain forest structure and composition

and some consequences for frugivores and folivores. Biotropica, 20, 31–37. Johns, A.G. (1997) Timber Production and Biodiversity Conservation in Tropical Rain Forests.

Cambridge University Press. Johns, A.G. & Johns, B.G. (1995) Tropical forest primates and logging: long-term coexistence? Oryx,

29, 205–211. Johnson, A., Vongkhamheng, C. & Saithongdam, T. (2009) The diversity, status and conservation of

small carnivores in a montane tropical forest in northern Laos. Oryx, 43, 626–633. Jones, D.T., Susilo, F.X., Bignell, D.E., Hardiwinoto, S., Gillison, A.N. & Eggleton, P. (2003)

Termite assemblage collapse along a land-use intensification gradient in lowland central Sumatra, Indonesia. J. Appl. Ecol., 40, 380–391.

Jones, K.E., Bielby, J., Cardillo, M., Fritz, S.A., O’Dell, J., Orme, C.D.L., Safi, K., Sechrest, W., Boakes, E.H., Carbone, C., Connolly, C., Cutts, M.J., Foster, J.K., Grenyer, R., Habib, M., Plaster, C.A., Price, S.A., Rigby, E.A., Rist, J., et al. (2009) PanTHERIA: a species-level database of life history, ecology, and geography of extant and recently extinct mammals. Ecology, 90, 2648–2648.

de Jong, W. (1997) Developing swidden agriculture and the threat of biodiversity loss. Agric. Ecosyst. Environ., 62, 187–197.

Jost, L. (2007) Partitioning diversity into independent alpha and beta components. Ecology, 88, 2427–2439.

Bibliography

218

Jost, L. (2010) Independence of alpha and beta diversities. Ecology, 91, 1969–1974. Karanth, K.U. & Nichols, J.D. (1998) Estimation of tiger densities in India using photographic

captures and recaptures. Ecology, 79, 2852–2862. Kays, R., Tilak, S., Kranstauber, B., Jansen, P.A., Carbone, C., Rowcliffe, M.J., Fountain, T., Eggert,

J. & He, Z. (2011) Monitoring wild animal communities with arrays of motion sensitive camera traps. Int. J. Res. Rev. Wirel. Sens. Networks, 1, 19–29.

Kemper, C. & Bell, D.T. (1985) Small mammals and habitat structure in lowland rain forest of Peninsular Malaysia. J. Trop. Ecol., 1, 5–22.

Kendall, W. (1999) Robustness of closed capture-recapture methods to violations of the closure assumption. Ecology, 80, 2517–2525.

Kéry, M. & Schaub, M. (2012) Bayesian Population Analysis Using WinBUGS: A Hierarchical Approach. Academic Press.

Kessler, M., Abrahamczyk, S., Bos, M., Buchori, D., Putra, D.D., Gradstein, S.R., Höhn, P., Kluge, J., Orend, F., Pitopang, R., Saleh, S., Schulze, C.H., Sporn, S.G., Steffan-Dewenter, I., Tjitrosoedirdjo, S.S. & Tscharntke, T. (2009) Alpha and beta diversity of plants and animals along a tropical land-use gradient. Ecol. Appl., 19, 2142–2156.

Kitamura, S., Thong-Aree, S., Madsri, S. & Poonswad, P. (2010) Mammal diversity and conservation in a small isolated forest of southern Thailand. Raffles Bull. Zool., 58, 145–156.

Kitamura, S., Yumoto, T., Poonswad, P., Suzuki, S. & Wohandee, P. (2008) Rare seed-predating mammals determine seed fate of Canarium euphyllum, a large-seeded tree species in a moist evergreen forest, Thailand. Ecol. Res., 23, 169–177.

Kitchener, A., Yasuma, S., Andau, M. & Quillen, P. (2004) Three bay cats (Catopuma badia) from Borneo. Mamm. Biol., 69, 349–353.

Kitching, R.L., Ashton, L.A., Nakamura, A., Whitaker, T. & Khen, C.V. (2013) Distance-driven species turnover in Bornean rainforests: homogeneity and heterogeneity in primary and post-logging forests. Ecography, 36, 675–682.

Koh, L.P. (2007a) Impending disaster or sliver of hope for Southeast Asian forests? The devil may lie in the details. Biodivers. Conserv., 16, 3935–3938.

Koh, L.P. (2007b) Impacts of land use change on South-east Asian forest butterflies: a review. J. Appl. Ecol., 44, 703–713.

Koh, L.P. (2008a) Can oil palm plantations be made more hospitable for forest butterflies and birds? J. Appl. Ecol., 45, 1002–1009.

Koh, L.P. (2008b) Birds defend oil palms from herbivorous insects. Ecol. Appl., 18, 821–825. Koh, L.P. & Ghazoul, J. (2010a) Spatially explicit scenario analysis for reconciling agricultural

expansion, forest protection, and carbon conservation in Indonesia. Proc. Natl. Acad. Sci. U. S. A., 107, 11140–11144.

Koh, L.P. & Ghazoul, J. (2010b) A matrix-calibrated species-area model for predicting biodiversity losses due to land-use change. Conserv. Biol., 24, 994–1001.

Koh, L.P., Levang, P. & Ghazoul, J. (2009) Designer landscapes for sustainable biofuels. Trends Ecol. Evol., 24, 431–438.

Koh, L.P. & Wilcove, D.S. (2008) Is oil palm agriculture really destroying tropical biodiversity? Conserv. Lett., 1, 60–64.

Kraft, N.J.B., Comita, L.S., Chase, J.M., Sanders, N.J., Swenson, N.G., Crist, T.O., Stegen, J.C., Vellend, M., Boyle, B., Anderson, M.J., Cornell, H. V, Davies, K.F., Freestone, A.L., Inouye, B.D., Harrison, S.P. & Myers, J.A. (2011) Disentangling the drivers of β diversity along latitudinal and elevational gradients. Science, 333, 1755–1758.

Bibliography

219

Krauss, J., Steffan-Dewenter, I. & Tscharntke, T. (2003) Local species immigration, extinction, and turnover of butterflies in relation to habitat area and habitat isolation. Oecologia, 137, 591–602.

Kuussaari, M., Bommarco, R., Heikkinen, R.K., Helm, A., Krauss, J., Lindborg, R., Ockinger, E., Pärtel, M., Pino, J., Rodà, F., Stefanescu, C., Teder, T., Zobel, M. & Steffan-Dewenter, I. (2009) Extinction debt: a challenge for biodiversity conservation. Trends Ecol. Evol., 24, 564–571.

Laidlaw, R. (2000) Effects of habitat disturbance and protected areas on mammals of Peninsular Malaysia. Conserv. Biol., 14, 1639–1648.

Lamb, D., Erskine, P.D. & Parrotta, J.A. (2005) Restoration of degraded tropical forest landscapes. Science, 310, 1628–1632.

Lambert, F. (1991) The conservation of fig-eating birds in Malaysia. Biol. Conserv., 58, 31–40. Lambert, F. (1992) The consequences of selective logging for Bornean lowland forest birds. Philos.

Trans. R. Soc. B Biol. Sci., 335, 443–457. Lambert, F. & Collar, N. (2002) The future for Sundaic lowland forest birds: long-term effects of

commercial logging and fragmentation. Forktail, 18, 127–146. Lambert, F. & Marshall, A. (1991) Keystone characteristics of bird-dispersed Ficus in a Malaysian

lowland rain forest. J. Ecol., 79, 793–809. Lambert, T.D., Malcolm, J.R. & Zimmerman, B.L. (2006) Amazonian small mammal abundances in

relation to habitat structure and resource abundance. J. Mammal., 87, 766–776. Lande, R. (1996) Statistics and partitioning of species diversity, and similarity among multiple

communities. Oikos, 76, 5–13. Langham, N. (1983) Distribution and ecology of small mammals in three rain forest localities of

Peninsula Malaysia with particular references to Kedah Peak. Biotropica, 15, 199–206. Lantschner, M.V., Rusch, V. & Hayes, J.P. (2012) Habitat use by carnivores at different spatial scales

in a plantation forest landscape in Patagonia, Argentina. For. Ecol. Manage., 269, 271–278. Laporte, N.T., Stabach, J.A., Grosch, R., Lin, T.S. & Goetz, S.J. (2007) Expansion of industrial

logging in Central Africa. Science, 316, 1451. Larrucea, E.S., Brussard, P.F., Jaeger, M.M. & Barrett, R.H. (2007) Cameras, coyotes, and the

assumption of equal detectability. J. Wildl. Manage., 71, 1682–1689. Laufer, J., Michalski, F. & Peres, C. (2013) Assessing sampling biases in logging impact studies in

tropical forests. Trop. Conserv. Sci., 6, 16–34. Laurance, W. (2007) Forest destruction in tropical Asia. Curr. Sci., 93, 1544–1550. Laurance, W. (2008) Theory meets reality: how habitat fragmentation research has transcended island

biogeographic theory. Biol. Conserv., 141, 1731–1744. Laurance, W.F. (1991) Ecological correlates of extinction proneness in Australian tropical rain forest

mammals. Conserv. Biol., 5, 79–89. Laurance, W.F., Clements, G.R., Sloan, S., Connell, C.S.O., Mueller, N.D., Goosem, M., Venter, O.,

Edwards, D.P., Phalan, B., Balmford, A., Ree, R. Van Der & Arrea, I.B. (2014a) A global strategy for road building. Nature, 513, 229–232.

Laurance, W.F. & Cochrane, M.A. (2001) Synergistic effects in fragmented landscapes. Conserv. Biol., 15, 1488–1489.

Laurance, W.F., Goosem, M. & Laurance, S.G.W. (2009) Impacts of roads and linear clearings on tropical forests. Trends Ecol. Evol., 24, 659–669.

Laurance, W.F., Lovejoy, T.E., Vasconcelos, H.L., Bruna, E.M., Didham, R.K., Stouffer, P.C., Gascon, C., Bierregaard, R.O., Laurance, S.G. & Sampaio, E. (2002) Ecosystem decay of Amazonian forest fragments: a 22-Year investigation. Conserv. Biol., 16, 605–618.

Bibliography

220

Laurance, W.F., Sayer, J. & Cassman, K.G. (2014b) Agricultural expansion and its impacts on tropical nature. Trends Ecol. Evol., 29, 107–116.

Lawton, J.H. (1999) Are there general laws in ecology? Oikos, 84, 177–192. Lees, A.C. & Peres, C.A. (2008) Avian life-history determinants of local extinction risk in a hyper-

fragmented Neotropical forest landscape. Anim. Conserv., 11, 128–137. Legendre, P. (2014) Interpreting the replacement and richness difference components of beta

diversity. Glob. Ecol. Biogeogr., 23, 1324–1334. Legendre, P., Borcard, D., Blanchet, F.G. & Dray, S. (2013) PCNM: MEM spatial eigenfunction and

principal coordinate analyses. R package version 2.1-2. Available from http://R-Forge.R-project.org/projects/sedar/.

Legendre, P., Borcard, D. & Peres-Neto, P.R. (2005) Analyzing beta diversity: partitioning the spatial variation of community composition data. Ecol. Monogr., 75, 435–450.

Legendre, P. & De Cáceres, M. (2013) Beta diversity as the variance of community data: dissimilarity coefficients and partitioning. Ecol. Lett., 16, 951–963.

Legendre, P. & Gallagher, E. (2001) Ecologically meaningful transformations for ordination of species data. Oecologia, 129, 271–280.

Legendre, P. & Legendre, L. (2012) Numerical Ecology, 3rd edition. Elsevier Science BV, Amsterdam.

Leibold, M.A., Holyoak, M., Mouquet, N., Amarasekare, P., Chase, J.M., Hoopes, M.F., Holt, R.D., Shurin, J.B., Law, R., Tilman, D., Loreau, M. & Gonzalez, A. (2004) The metacommunity concept: a framework for multi-scale community ecology. Ecol. Lett., 7, 601–613.

Lemmon, P.E. (1957) A new instrument for measuring forest overstory density. J. For., 55, 667–668. Lessard, J.-P., Belmaker, J., Myers, J.A., Chase, J.M. & Rahbek, C. (2012) Inferring local ecological

processes amid species pool influences. Trends Ecol. Evol., 27, 600–607. Lhota, S., Loken, B., Spehar, S., Fell, E., Pospěch, A. & Kasyanto, N. (2012) Discovery of Miller’s

grizzled langur (Presbytis hosei canicrus) in Wehea Forest confirms the continued existence and extends known geographical range of an endangered primate. Am. J. Primatol., 74, 193–198.

Liebsch, D., Marques, M.C.M. & Goldenberg, R. (2008) How long does the Atlantic Rain Forest take to recover after a disturbance? Changes in species composition and ecological features during secondary succession. Biol. Conserv., 141, 1717–1725.

Lima, M., Marquet, P.A. & Jaksic, F.M. (1996) Extinction and colonization processes in subpopulations of five neotropical small mammal species. Oecologia, 107, 197–203.

Lindenmayer, D.B., Franklin, J.F., Lõhmus, A., Baker, S.C., Bauhus, J., Beese, W., Brodie, A., Kiehl, B., Kouki, J., Pastur, G.M., Messier, C., Neyland, M., Palik, B., Sverdrup-Thygeson, A., Volney, J., Wayne, A. & Gustafsson, L. (2012) A major shift to the retention approach for forestry can help resolve some global forest sustainability issues. Conserv. Lett., 5, 421–431.

Liow, L., Sodhi, N. & Elmqvist, T. (2001) Bee diversity along a disturbance gradient in tropical lowland forests of south-east Asia. J. Appl. Ecol., 38, 180–192.

Loader, N.J., Walsh, R.P.D., Robertson, I., Bidin, K., Ong, R.C., Reynolds, G., McCarroll, D., Gagen, M. & Young, G.H.F. (2011) Recent trends in the intrinsic water-use efficiency of ringless rainforest trees in Borneo. Philos. Trans. R. Soc. Lond. B. Biol. Sci., 366, 3330–3339.

Lomolino, M. & Perault, D. (2000) Assembly and disassembly of mammal communities in a fragmented temperate rain forest. Ecology, 81, 1517–1532.

Lucey, J.M., Tawatao, N., Senior, M.J.M., Chey, V.K., Benedick, S., Hamer, K.C., Woodcock, P., Newton, R.J., Bottrell, S.H. & Hill, J.K. (2014) Tropical forest fragments contribute to species richness in adjacent oil palm plantations. Biol. Conserv., 169, 268–276.

Bibliography

221

Luke, S.H., Fayle, T.M., Eggleton, P., Turner, E.C. & Davies, R.G. (2014) Functional structure of ant and termite assemblages in old growth forest, logged forest and oil palm plantation in Malaysian Borneo. Biodivers. Conserv., 23, 2817–2832.

Luskin, M.S. & Potts, M.D. (2011) Microclimate and habitat heterogeneity through the oil palm lifecycle. Basic Appl. Ecol., 12, 540–551.

Lynam, A. (1997) Rapid decline of small mammal diversity in monsoon evergreen forest fragments in Thailand. In Tropical forest remnants: ecology, management, and conservation of fragmented communities (eds W.F. Laurance & R.O. Bierregaard), pp. 222–240.

Lynam, A. & Billick, I. (1999) Differential responses of small mammals to fragmentation in a Thailand tropical forest. Biol. Conserv., 91, 191–200.

MacArthur, R.H. & Levins, R. (1967) The limiting similarity, convergence, and divergence of coexisting species. Am. Nat., 101, 377–385.

Macarthur, R.H. & Wilson, E.O. (1967) The Theory of Island Biogeography. Princeton University Press.

MacKenzie, D., Nichols, J. & Hines, J. (2003) Estimating site occupancy, colonization, and local extinction when a species is detected imperfectly. Ecology, 84, 2200–2207.

MacKenzie, D., Nichols, J. & Lachman, G. (2002) Estimating site occupancy rates when detection probabilities are less than one. Ecology, 83, 2248–2255.

Mackenzie, D.I. & Nichols, J.D. (2004) Occupancy as a surrogate for abundance estimation. Anim. Biodivers. Conserv., 1, 461–467.

MacKenzie, D.I., Nichols, J.D., Royle, J.A., Pollock, K.P., Bailey, L.L. & Hines, J.E. (2006) Occupancy Estimation and Modeling: Inferring Patterns and Dynamics of Species Occurrence. Academic Press, New York.

Maddox, T., Priatna, D., Gemita, E. & Salampessy, A. (2007) The conservation of tigers and other wildlife in oil palm plantations. Zoological Society of London, London.

Maffei, L., Cullar, E. & Noss, A. (2004) One thousand jaguars (Panthera onca) in Bolivia’s Chaco? Camera trapping in the Kaa-Iya National Park. J. Zool., 262, 295–304.

Malhi, Y., Gardner, T.A., Goldsmith, G.R., Silman, M.R. & Zelazowski, P. (2014) Tropical forests in the Anthropocene. Annu. Rev. Environ. Resour., 39, 125–159.

Malhi, Y., Phillips, O.L., Lloyd, J., Baker, T., Wright, J., Almeida, S., Arroyo, L., Frederiksen, T., Grace, J., Higuchi, N., Killeen, T., Laurance, W.F., Leano, C., Lewis, S., Meir, P., Monteagudo, A., Neill, D., Nunez Vargas, P., Panfil, S.N., et al. (2002) An international network to monitor the structure, composition and dynamics of Amazonian forests (RAINFOR). J. Veg. Sci., 13, 439–450.

Manning, A.D., Lindenmayer, D.B., Nix, H.A., Resource, C., Studies, E. & Australian, T. (2004) Continua and Umwelt: novel perspectives on viewing landscapes. Oikos, 104, 621–628.

Margono, B.A., Potapov, P. V., Turubanova, S., Stolle, F. & Hansen, M.C. (2014) Primary forest cover loss in Indonesia over 2000–2012. Nat. Clim. Chang., 4, 730–735.

Margono, B.A., Turubanova, S., Zhuravleva, I., Potapov, P., Tyukavina, A., Baccini, A., Goetz, S. & Hansen, M.C. (2012) Mapping and monitoring deforestation and forest degradation in Sumatra (Indonesia) using Landsat time series data sets from 1990 to 2010. Environ. Res. Lett., 7, 034010.

Marques, T.A., Thomas, L. & Royle, J.A. (2011) A hierarchical model for spatial capture-recapture data: comment. Ecology, 92, 526–528.

Marsden, S.J. (1998) Changes in bird abundance following selective logging on Seram, Indonesia. Conserv. Biol., 12, 605–611.

Marsh, C.J. & Ewers, R.M. (2012) A fractal-based sampling design for ecological surveys quantifying

Bibliography

222

β-diversity. Methods Ecol. Evol., 4, 63–72. Marsh, C.W. & Greer, A.G. (1992) Forest land-use in Sabah, Malaysia: an introduction to Danum

Valley. Philos. Trans. R. Soc. B Biol. Sci., 335, 331–339. Martin, A.R. & Thomas, S.C. (2011) A reassessment of carbon content in tropical trees. PLoS One, 6,

e23533. Mathai, J., Hon, J. & Juat, N. (2010) Small carnivores in a logging concession in the Upper Baram,

Sarawak, Borneo. Small Carniv. Conserv., 42, 1–9. Matias, M.G., Gravel, D., Guilhaumon, F., Desjardins-Proulx, P., Loreau, M., Münkemüller, T. &

Mouquet, N. (2014) Estimates of species extinctions from species-area relationships strongly depend on ecological context. Ecography, 37, 431–442.

Matsubayashi, H., Bernard, H. & Ahmad, A.H. (2011) Small carnivores of the Imbak Canyon, Sabah, Malaysia, Borneo, including a new locality for the Hose’s Civet Diplogale hosei. Small Carniv. Conserv., 45, 18–22.

Mayaux, P., Holmgren, P., Achard, F., Eva, H., Stibig, H.-J. & Branthomme, A. (2005) Tropical forest cover change in the 1990s and options for future monitoring. Philos. Trans. R. Soc. Lond. B. Biol. Sci., 360, 373–384.

Mazerolle, M.J. & Villard, M.-A. (1999) Patch characteristics and landscape context as predictors of species presence and abundance: A review. Écoscience, 6, 117–124.

Mcconkey, K.R. (2005) Influence of faeces on seed removal from gibbon droppings in a dipterocarp forest in Central Borneo. J. Trop. Ecol., 21, 117–120.

McGarigal, K. & Cushman, S.A. (2002) Comparative evaluation of experimental approaches to the study of habitat fragmentation effects. Ecol. Appl., 12, 335–345.

McIntire, E. & Fajardo, A. (2009) Beyond description: the active and effective way to infer processes from spatial patterns. Ecology, 90, 46–56.

Mcintyre, S. & Barrett, G.W. (1992) Habitat variegation, an alternative to fragmentation. Conserv. Biol., 6, 146–147.

McIntyre, S. & Hobbs, R. (1999) A framework for conceptualizing human effects on landscapes and its relevance to management and research models. Conserv. Biol., 13, 1282–1292.

McKinney, M.L. (1997) Extinction vulnerability and selectivity: combining ecological and paleontological views. Annu. Rev. Ecol. Syst., 28, 495–516.

McMorrow, J. & Talip, M.A. (2001) Decline of forest area in Sabah, Malaysia: relationship to state policies, land code and land capability. Glob. Environ. Chang., 11, 217–230.

McShea, W.J., Stewart, C., Peterson, L., Erb, P., Stuebing, R. & Giman, B. (2009) The importance of secondary forest blocks for terrestrial mammals within an Acacia/secondary forest matrix in Sarawak, Malaysia. Biol. Conserv., 142, 3108–3119.

Meek, P., Ballard, G. & Fleming, P. (2012) An introduction to camera trapping for wildlife surveys in Australia. Invasive Animals Cooperative Research Centre, Canberra, Australia.

Meijaard, E., Albar, G., Nardiyono, Rayadin, Y., Ancrenaz, M. & Spehar, S. (2010) Unexpected ecological resilience in Bornean orangutans and implications for pulp and paper plantation management. PLoS One, 5, e12813.

Meijaard, E. & Sheil, D. (2007) A logged forest in Borneo is better than none at all. Nature, 446, 974. Meijaard, E. & Sheil, D. (2008) The persistence and conservation of Borneo’s mammals in lowland

rain forests managed for timber: observations, overviews and opportunities. Ecol. Res., 23, 21–34.

Meijaard, E., Sheil, D., Marshall, A. & Nasi, R. (2008) Phylogenetic age is positively correlated with sensitivity to timber harvest in Bornean mammals. Biotropica, 40, 76–85.

Bibliography

223

Meijaard, E., Sheil, D., Nasi, R., Augeri, D., Rosenbaum, B., Iskandar, D., Setyawati, T., Lammertink, M., Rachmatika, I., Wong, A., Soehartono, T., Stanley, S. & O’Brien, T. (2005) Life after Logging: Reconciling Wildlife Conservation and Production Forestry in Indonesian Borneo. Center for International Forestry Research, Bogor, Indonesia.

Michalski, F. & Peres, C. (2005) Anthropogenic determinants of primate and carnivore local extinctions in a fragmented forest landscape of southern Amazonia. Biol. Conserv., 124, 383–396.

Miettinen, J., Shi, C. & Liew, S.C. (2011) Deforestation rates in insular Southeast Asia between 2000 and 2010. Glob. Chang. Biol., 17, 2261–2270.

Mitra, S. & Sheldon, F. (1993) Use of an exotic tree plantation by Bornean lowland forest birds. Auk, 110, 529–540.

Mohamed, A., Samejima, H. & Wilting, A. (2009) Records of five Bornean cat species from Deramakot Forest Reserve in Sabah, Malaysia. Cat News, 51, 12–15.

Mohamed, A., Sollmann, R., Bernard, H., Ambu, L.N., Lagan, P., Mannan, S., Hofer, H. & Wilting, A. (2013) Density and habitat use of the leopard cat (Prionailurus bengalensis) in three commercial forest reserves in Sabah, Malaysian Borneo. J. Mammal., 94, 82–89.

Mohd-Azlan, J. (2003) The diversity and conservation of mustelids, viverrids, and herpestids in a disturbed forest in Peninsular Malaysia. Small Carniv. Conserv., 29, 8–9.

Mohd-Azlan, J. (2006) Mammal diversity and conservation in a secondary forest in Peninsular Malaysia. Biodivers. Conserv., 15, 1013–1025.

Mohd-Azlan, J. & Sanderson, J. (2007) Geographic distribution and conservation status of the bay cat Catopuma badia, a Bornean endemic. Oryx, 41, 394–397.

Mohd-Azlan, J. & Sharma, D. (2006) The diversity and activity patterns of wild felids in a secondary forest in Peninsular Malaysia. Oryx, 40, 36–41.

Morrogh-bernard, H., Husson, S., Page, S.E. & Rieley, J.O. (2003) Population status of the Bornean orang-utan (Pongo pygmaeus ) in the Sebangau peat swamp forest, Central Kalimantan, Indonesia. Biol. Conserv., 110, 141–152.

Munshi-South, J., Emmons, L.H. & Bernard, H. (2007) Behavioral monogamy and fruit availability in the large treeshrew (Tupaia tana) in Sabah, Malaysia. J. Mammal., 88, 1427–1438.

Mutshinda, C.M. & O’Hara, R.B. (2011) Integrating the niche and neutral perspectives on community structure and dynamics. Oecologia, 166, 241–251.

Mutshinda, C.M., O’Hara, R.B. & Woiwod, I.P. (2009) What drives community dynamics? Proc. R. Soc. B Biol. Sci., 276, 2923–2929.

Myers, N., Mittermeier, R.A., Mittermeier, C.G., da Fonseca, G.A. & Kent, J. (2000) Biodiversity hotspots for conservation priorities. Nature, 403, 853–858.

Naim, M., Ketut Aryawan, A., Sudharto, P., Jennings, A.P., Veron, G., Verwilghen, A., Turner, E.C., Aulia, P. & Caliman, J.-P. (2014) Understanding the relationship between rat populations and small carnivores in oil palm plantations: Outputs for sustainable control of rats. In Proceedings of the 4th International Conference on Oil Palm And Environment (ICOPE).

Nájera, A. & Simonetti, J.A. (2010) Can oil palm plantations become bird friendly? Agrofor. Syst., 80, 203–209.

Nakagawa, M., Hyodo, F. & Nakashizuka, T. (2007) Effect of forest use on trophic levels of small mammals: an analysis using stable isotopes. Can. J. Zool., 85, 472–478.

Nakagawa, M., Miguchi, H. & Nakashizuka, T. (2006) The effects of various forest uses on small mammal communities in Sarawak, Malaysia. For. Ecol. Manage., 231, 55–62.

Nakashima, Y., Inoue, E., Inoue-Murayama, M. & Sukor, J.R.A. (2010) Functional uniqueness of a small carnivore as seed dispersal agents: A case study of the common palm civets in the Tabin

Bibliography

224

Wildlife Reserve, Sabah, Malaysia. Oecologia, 164, 721–730. Nakashima, Y., Nakabayashi, M. & Sukor, J.A. (2013) Space use, habitat selection, and day-beds of

the common palm civet (Paradoxurus hermaphroditus) in human-modified habitats in Sabah, Borneo. J. Mammal., 94, 1169–1178.

Mac Nally, R. (2007) Use of the abundance spectrum and relative-abundance distributions to analyze assemblage change in massively altered landscapes. Am. Nat., 170, 319–330.

Mac Nally, R. & Fleishman, E. (2004) Comparative influence of spatial scale on beta diversity within regional assemblages of birds and butterflies. J. Biogeogr., 31, 917–929.

Nasi, R., Koponen, P., Poulsen, J.G., Buitenzorgy, M. & Rusmantoro, W. (2007) Impact of landscape and corridor design on primates in a large-scale industrial tropical plantation landscape. Biodivers. Conserv., 17, 1105–1126.

Nekola, J. & White, P. (2002) Conservation, the two pillars of ecological explanation, and the paradigm of distance. Nat. Areas J., 22, 305–310.

Newbery, D.M., Campbell, E.J.F., Lee, Y.F., Ridsdale, C.E. & Still, M.J. (1992) Primary lowland dipterocarp forest at Danum Valley, Sabah, Malaysia: structure, relative abundance and family composition. Philos. Trans. R. Soc. Lond. B. Biol. Sci., 335, 341–356.

Newbery, D.M., Campbell, E.J.F., Proctor, J. & Still, M.J. (1996) Primary lowland dipterocarp forest at Danum Valley, Sabah, Malaysia. Species composition and patterns in the understorey. Vegetatio, 122, 193–220.

Newbold, T., Hudson, L.N., Hill, S.L.L., Contu, S., Lysenko, I., Senior, R.A., Börger, L., Bennett, D.J., Choimes, A., Collen, B., Day, J., Palma, A. De, Díaz, S., Edgar, M.J., Feldman, A., Garon, M., Harrison, M.L.K., Alhusseini, T., Echeverria-Londoño, S., et al. (2015) Global effects of land use on local terrestrial biodiversity. Nature, 520, 45–50.

Newbold, T., Hudson, L.N., Phillips, H.R.P., Hill, S.L.L., Contu, S., Lysenko, I., Blandon, A., Butchart, S.H.M., Booth, H.L., Day, J., Palma, A. De, Harrison, M.L.K., Kirkpatrick, L., Pynegar, E., Robinson, A., Simpson, J., Mace, G.M., Scharlemann, J.P.W. & Purvis, A. (2014) A global model of the response of tropical and sub-tropical forest biodiversity to anthropogenic pressures. Proc. R. Soc. London Ser. B Biol. Sci., 281, 20141371.

Nichols, E., Spector, S., Louzada, J., Larsen, T., Amezquita, S. & Favila, M.E. (2008) Ecological functions and ecosystem services provided by Scarabaeinae dung beetles. Biol. Conserv., 141, 1461–1474.

Numata, S., Okuda, T., Sugimoto, T., Nishimura, S., Yoshida, K., Quah, E.S., Yasuda, M., Muangkhum, K. & Noor, N. (2005) Camera trapping: a non-invasive approach as an additional tool in the study of mammals in Pasoh Forest Reserve and adjacent fragmented areas in Peninsular Malaysia. Malayan Nat. J., 57, 29–45.

Oksanen, J., Blanchet, F.G., Kindt, R., Legendre, P., Minchin, P.R., O’Hara, R.B., Simpson, G.L., Solymos, P., Henry, M., Stevens, H. & Wagner, H. (2013) vegan: Community ecology package. R package version 2.0-10. Available from http://cran.r-project.org/package=vegan.

Olivier, P.I. & Aarde, R.J. Van (2014) Multi-scale sampling boosts inferences from beta diversity patterns in coastal forests of South Africa. J. Biogeogr., 41, 1428–1439.

Olsen, A.R., Sedransk, J., Edwards, D., Gotway, C.A., Liggett, W., Rathbun, S., Reckhow, K.H. & Young, L.J. (1999) Statistical issues for monitoring ecological and natural resources in the United States. Environ. Monit. Assess., 54, 1–45.

Onoguchi, G. & Matsubayashi, H. (2008) Comparative study on mammalian fauna in different harvesting intensities with reduced-impact and conventional logging in Sabah, Malaysia. In Sustainability and biodiversity assessment on forest utilization options (eds M. Ichikawa, S. Yamashita & T. Nakasizuka), pp. 133–140.

Orrock, J.L. & Watling, J.I. (2010) Local community size mediates ecological drift and competition in

Bibliography

225

metacommunities. Proc. Biol. Sci., 277, 2185–2191. Otis, D., Burnham, K., White, G. & Anderson, D. (1978) Statistical inference from capture data on

closed animal populations. Wildl. Monogr., 62, 1–135. Ovaskainen, O. & Hanski, I. (2002) Transient dynamics in metapopulation response to perturbation.

Theor. Popul. Biol., 61, 285–295. Ovaskainen, O., Hottola, J. & Siitonen, J. (2010) Modeling species co-occurrence by multivariate

logistic regression generates new hypotheses on fungal interactions. Ecology, 91, 2514–2521. Paoli, G.D., Peart, D.R., Leighton, M. & Samsoedin, I. (2001) An ecological and economic

assessment of the nontimber forest product gaharu wood in Gunung Palung National Park, West Kalimantan, Indonesia. Conserv. Biol., 15, 1721–1732.

Pardini, R., Bueno, A.D.A., Gardner, T.A., Prado, P.I. & Metzger, J.P. (2010) Beyond the fragmentation threshold hypothesis: regime shifts in biodiversity across fragmented landscapes. PLoS One, 5, e13666.

Paton, P. (1994) The effect of edge on avian nest success: how strong is the evidence? Conserv. Biol., 8, 17–26.

Pattanavibool, A. & Dearden, P. (2002) Fragmentation and wildlife in montane evergreen forests, northern Thailand. Biol. Conserv., 107, 155–164.

Patterson, B.D. (1987) The principle of nested subsets and its implications for biological conservation. Conserv. Biol., 1, 323–334.

Payne, J., Francis, C.M. & Phillipps, K. (2007) A Field Guide to the Mammals of Borneo, 4th edition. The Sabah Society, Kota Kinabalu.

Peh, K.S.-H., Jong, J. De, Sodhi, N.S., Lim, S.L.-H. & Yap, C.A.-M. (2005) Lowland rainforest avifauna and human disturbance: persistence of primary forest birds in selectively logged forests and mixed-rural habitats of southern Peninsular Malaysia. Biol. Conserv., 123, 489–505.

Peh, K.S.-H., Sodhi, N.S., de Jong, J., Sekercioglu, C.H., Yap, C. a.-M. & Lim, S.L.-H. (2006) Conservation value of degraded habitats for forest birds in southern Peninsular Malaysia. Divers. Distrib., 12, 572–581.

Pereira, H. & Daily, G. (2006) Modeling biodiversity dynamics in countryside landscapes. Ecology, 87, 1877–1885.

Peres, C.A., Gardner, T.A., Barlow, J., Zuanon, J., Michalski, F., Lees, A.C., Vieira, I.C.G., Moreira, F.M.S. & Feeley, K.J. (2010) Biodiversity conservation in human-modified Amazonian forest landscapes. Biol. Conserv., 143, 2314–2327.

Peres-Neto, P., Legendre, P., Dray, S. & Borcard, D. (2006) Variation partitioning of species data matrices: estimation and comparison of fractions. Ecology, 87, 2614–2625.

Peres-Neto, P.R. & Legendre, P. (2010) Estimating and controlling for spatial structure in the study of ecological communities. Glob. Ecol. Biogeogr., 19, 174–184.

Persey, S., Imanuddin & Sadiki, L. (2011) A Practical Handbook for Conserving High Conservation Value Species and Habitats within oil palm landscapes. Zoological Society of London, London.

Persoon, G.A. & Osseweijer, M. (2008) Reflections on the Heart of Borneo. Tropenbos International, Wageningen, the Netherlands.

Pettorelli, N., Laurance, W.F., O’Brien, T.G., Wegmann, M., Nagendra, H. & Turner, W. (2014) Satellite remote sensing for applied ecologists: Opportunities and challenges. J. Appl. Ecol., 51, 839–848.

Pfeifer, M., Kor, L., Nilus, R., Turner, E.C., Khen, C.V., Cusack, J.J., Lysenko, I., Chung, A.Y.C. & Ewers, R.M. (2015) Mapping the structure of Borneo’s tropical forests across a degradation gradient. Remote Sens. Environ., [submitted].

Bibliography

226

Phalan, B., Bertzky, M., Butchart, S.H.M., Donald, P.F., Scharlemann, J.P.W., Stattersfield, A.J. & Balmford, A. (2013) Crop expansion and conservation priorities in tropical countries. PLoS One, 8, e51759.

Phalan, B., Onial, M., Balmford, A. & Green, R.E. (2011) Reconciling food production and biodiversity conservation: land sharing and land sparing compared. Science, 333, 1289–1291.

Philpott, S.M., Bichier, P., Rice, R.A. & Greenberg, R. (2007) Biodiversity conservation, yield, and alternative products in coffee agroecosystems in Sumatra, Indonesia. Biodivers. Conserv., 17, 1805–1820.

Pimm, S., Jones, H. & Diamond, J. (1988) On the risk of extinction. Am. Nat., 132, 757–785. Pimm, S.L., Jenkins, C.N., Abell, R., Brooks, T.M., Gittleman, J.L., Joppa, L.N., Raven, P.H.,

Roberts, C.M. & Sexton, J.O. (2014) The biodiversity of species and their rates of extinction, distribution, and protection. Science, 344, 1246752.

Pinard, M.., Barker, M.. & Tay, J. (2000a) Soil disturbance and post-logging forest recovery on bulldozer paths in Sabah, Malaysia. For. Ecol. Manage., 130, 213–225.

Pinard, M.A., Putz, F.E. & Tay, J. (2000b) Lessons learned from the implementation of reduced-impact logging in hilly terrain in Sabah, Malaysia. Int. For. Rev., 2, 33–39.

Piper, S.D. & Catterall, C.P. (2003) A particular case and a general pattern: hyperaggressive behaviour by one species may mediate avifaunal decreases in fragmented Australian forests. Oikos, 101, 602–614.

Plummer, M. (2013) JAGS: Just Another Gibbs Sampler, version 3.4.0. Available from http://mcmc-jags.sourceforge.net/.

Pollock, K.H. (1982) A capture–recapture design robust to unequal probability of capture. J. Wildl. Manage., 46, 752–757.

Posa, M.R.C. & Sodhi, N.S. (2006) Effects of anthropogenic land use on forest birds and butterflies in Subic Bay, Philippines. Biol. Conserv., 129, 256–270.

Potts, M.D. & Vincent, J.R. (2008) Spatial distribution of species populations, relative economic values, and the optimal size and number of reserves. Environ. Resour. Econ., 39, 91–112.

Poynton, S. (2014) The history of the contentious number behind zero deforestation commitments for palm oil. Available from http://news.mongabay.com/2014/0714-poynton-commentary-35-tons-c.html.

Prugh, L.R., Hodges, K.E., Sinclair, A.R.E. & Brashares, J.S. (2008) Effect of habitat area and isolation on fragmented animal populations. Proc. Natl. Acad. Sci. U. S. A., 105, 20770–20775.

Puan, C.C.L., Goldizen, A.W.A., Zakaria, M., Hafidzi, M.N. & Baxter, G.S. (2011) Relationships among rat numbers, abundance of oil palm fruit and damage levels to fruit in an oil palm plantation. Integr. Zool., 6, 130–139.

Putz, F. & Appanah, S. (1987) Buried seeds, newly dispersed seeds, and the dynamics of a lowland forest in Malaysia. Biotropica, 19, 326–333.

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) Sustaining conservation values in selectively logged tropical forests: the attained and the attainable. Conserv. Lett., 5, 296–303.

Qi, J., Chehbouni, A., Huete, A.R., Kerr, Y.H. & Sorooshian, S. (1994) A modified soil adjusted vegetation index. Remote Sens. Environ., 48, 119–126.

R Development Core Team (2014) R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Available from http://www.r-project.org/.

Radford, J.Q., Bennett, A.F. & Cheers, G.J. (2005) Landscape-level thresholds of habitat cover for woodland-dependent birds. Biol. Conserv., 124, 317–337.

Bibliography

227

Rajaratnam, R., Sunquist, M., Rajaratnam, L. & Ambu, L. (2007) Diet and habitat selection of the leopard cat (Prionailurus bengalensis borneoensis) in an agricultural landscape in Sabah, Malaysian Borneo. J. Trop. Ecol., 23, 209–217.

Ramage, B.S., Marshalek, E.C., Kitzes, J. & Potts, M.D. (2013a) Conserving tropical biodiversity via strategic spatiotemporal harvest planning. J. Appl. Ecol., 50, 1301–1310.

Ramage, B.S., Sheil, D., Salim, H.M.W., Fletcher, C., Mustafa, N.-Z.A., Luruthusamay, J.C., Harrison, R.D., Butod, E., Dzulkiply, A.D., Kassim, A.R. & Potts, M.D. (2013b) Pseudoreplication in tropical forests and the resulting effects on biodiversity conservation. Conserv. Biol., 27, 364–372.

Ramankutty, N. & Foley, J. (1999) Estimating historical changes in global land cover: Croplands from 1700 to 1992. Global Biogeochem. Cycles, 13, 997–1027.

Ranganathan, J., Daniels, R.J.R., Chandran, M.D.S., Ehrlich, P.R. & Daily, G.C. (2008) Sustaining biodiversity in ancient tropical countryside. Proc. Natl. Acad. Sci. U. S. A., 105, 17852–17854.

Rao, M., Rabinowitz, A. & Khaing, S.T. (2002) Status review of the protected-area system in Myanmar, with recommendations for conservation planning. Conserv. Biol., 16, 360–368.

Reid, A. (1995) Humans and forests in pre-colonial Southeast Asia. Environ. Hist. Camb., 1, 93–110. Reynolds, G., Payne, J., Sinun, W., Mosigil, G. & Walsh, R.P.D. (2011) Changes in forest land use

and management in Sabah, Malaysian Borneo, 1990-2010, with a focus on the Danum Valley region. Philos. Trans. R. Soc. Lond. B. Biol. Sci., 366, 3168–3176.

Rickart, E.A., Balete, D.S., Rowe, R.J. & Heaney, L.R. (2011) Mammals of the northern Philippines: tolerance for habitat disturbance and resistance to invasive species in an endemic insular fauna. Divers. Distrib., 17, 530–541.

Ricotta, C. (2010) On beta diversity decomposition: trouble shared is not trouble halved. Ecology, 91, 1981–1983.

Robinson, S.K., Thompson, F.R., Donovan, T.M., Whitehead, D.R. & Faaborg, J. (1995) Regional forest fragmentation and the nesting success of migratory birds. Science, 267, 1987–1990.

Rosenzweig, M.L. (1995) Species Diversity in Space and Time. Cambridge University Press. Ross, J., Hearn, A.J., Johnson, P.J. & Macdonald, D.W. (2013) Activity patterns and temporal

avoidance by prey in response to Sunda clouded leopard predation risk. J. Zool., 290, 96–106. Rota, C.T., Fletcher Jr, R.J., Dorazio, R.M. & Betts, M.G. (2009) Occupancy estimation and the

closure assumption. J. Appl. Ecol., 46, 1173–1181. Rowcliffe, J.M. & Carbone, C. (2008) Surveys using camera traps: are we looking to a brighter

future? Anim. Conserv., 11, 185–186. Rowcliffe, J.M., Carbone, C., Jansen, P.A., Kays, R. & Kranstauber, B. (2011) Quantifying the

sensitivity of camera traps: an adapted distance sampling approach. Methods Ecol. Evol., 2, 464–476.

Rowcliffe, J.M., Field, J., Turvey, S.T. & Carbone, C. (2008) Estimating animal density using camera traps without the need for individual recognition. J. Appl. Ecol., 45, 1228–1236.

Royle, J. & Nichols, J. (2003) Estimating abundance from repeated presence-absence data or point counts. Ecology, 84, 777–790.

Royle, J.A. (2004) N-Mixture Models for Estimating Population Size from Spatially Replicated Counts. Biometrics, 60, 108–115.

Royle, J.A. & Dorazio, R.M. (2008) Hierarchical Modeling and Inference in Ecology: The Analysis of Data from Populations, Metapopulations and Communities. Academic Press.

Rustam, Yasuda, M. & Tsuyuki, S. (2012) Comparison of mammalian communities in a human-disturbed tropical landscape in East Kalimantan, Indonesia. Mammal Study, 37, 299–311.

Bibliography

228

Sala, O.E., Iii, F.S.C., Armesto, J.J., Berlow, E., Bloomfield, J., Dirzo, R., Huber-sanwald, E., Huenneke, L.F., Jackson, R.B., Kinzig, A., Leemans, R., Lodge, D.M., Mooney, H.A., Poff, N.L., Sykes, M.T., Walker, B.H. & Walker, M. (2000) Global biodiversity scenarios for the year 2100. Science, 287, 1770–1774.

Salafsky, N., Dugelby, B. & Terborgh, J. (1993) Can extractive reserves save the rain forest? An ecological and socioeconomic comparison of nontimber forest product extraction systems in Peten, Guatemala, and West Kalimantan, Indonesia. Conserv. Biol., 7, 39–52.

Samejima, H., Ong, R., Lagan, P. & Kitayama, K. (2012) Camera-trapping rates of mammals and birds in a Bornean tropical rainforest under sustainable forest management. For. Ecol. Manage., 270, 248–256.

Samejima, H. & Semiadi, G. (2012) First record of Hose’s Civet Diplogale hosei from Indonesia, and records of other carnivores in the Schwaner Mountains, Central Kalimantan, Indonesia. Small Carniv. Conserv., 46, 1–7.

Samsudin, M., Khairul Najwan, A.J., Jalil, M.S., Abd Rahman, K., Mohd Nizam, M.S., Wan Mohd Shukri, W.A., Ismail, H., Shamsudin, I. & Wan Razali, W.M. (2010) Stocking and species composition of second growth forests in Peninsular Malaysia. Malaysian For., 73, 213–225.

Sax, D.F. & Gaines, S.D. (2003) Species diversity: from global decreases to local increases. Trends Ecol. Evol., 18, 561–566.

Scales, B.R. & Marsden, S.J. (2008) Biodiversity in small-scale tropical agroforests: a review of species richness and abundance shifts and the factors influencing them. Environ. Conserv., 35, 160–172.

Scheiner, S. (2004) A mélange of curves–further dialogue about species–area relationships. Glob. Ecol. Biogeogr., 13, 479–484.

Schipper, J., Chanson, J.S., Chiozza, F., Cox, N.A., Hoffmann, M., Katariya, V., Lamoreux, J., Rodrigues, A.S.L., Stuart, S.N., Temple, H.J., Baillie, J., Boitani, L., Lacher, T.E., Mittermeier, R.A., Smith, A.T., Absolon, D., Aguiar, J.M., Amori, G., Bakkour, N., et al. (2008) The status of the world’s land and marine mammals: diversity, threat, and knowledge. Science, 322, 225–230.

Schoener, T. & Spiller, D. (1987) High population persistence in a system with high turnover. Nature, 330, 474–477.

Schulze, C. & Waltert, M. (2004) Biodiversity indicator groups of tropical land-use systems: comparing plants, birds, and insects. Ecol. Appl., 14, 1321–1333.

Schwenk, W.S. & Donovan, T.M. (2011) A multispecies framework for landscape conservation planning. Conserv. Biol., 25, 1010–1021.

Scott, D.M. & Gemita, E. (2004) Can human-modified landscapes in Sumatra provide suitable habitat for threatened cat species? Zoological Society of London, London.

Scriven, S.A., Hodgson, J.A., McClean, C.J. & Hill, J.K. (2015) Protected areas in Borneo may fail to conserve tropical forest biodiversity under climate change. Biol. Conserv., 184, 414–423.

Sekercioglu, C.H., Ehrlich, P.R., Daily, G.C., Aygen, D., Goehring, D. & Sandi, R.F. (2002) Disappearance of insectivorous birds from tropical forest fragments. Proc. Natl. Acad. Sci. U. S. A., 99, 263–267.

Senior, M.J.M., Hamer, K.C., Bottrell, S., Edwards, D.P., Fayle, T.M., Lucey, J.M., Mayhew, P.J., Newton, R., Peh, K.S.-H., Sheldon, F.H., Stewart, C., Styring, A.R., Thom, M.D.F., Woodcock, P. & Hill, J.K. (2012) Trait-dependent declines of species following conversion of rain forest to oil palm plantations. Biodivers. Conserv., 22, 253–268.

Sethi, P. & Howe, H.F. (2009) Recruitment of hornbill-dispersed trees in hunted and logged forests of the Indian Eastern Himalaya. Conserv. Biol., 23, 710–718.

Shean, M. (2011) Commodity intelligent report: obstacles may reduce future palm oil production

Bibliography

229

growth. Foreign Agriculture Service, United States Department of Agriculture. Shehzad, W., Riaz, T., Nawaz, M.A., Miquel, C., Poillot, C., Shah, S.A., Pompanon, F., Coissac, E. &

Taberlet, P. (2012) Carnivore diet analysis based on next-generation sequencing: application to the leopard cat (Prionailurus bengalensis) in Pakistan. Mol. Ecol., 21, 1951–1965.

Sheldon, F.H., Styring, A. & Hosner, P.A. (2010) Bird species richness in a Bornean exotic tree plantation: a long-term perspective. Biol. Conserv., 143, 399–407.

Shmida, A. & Wilson, M. V. (1985) Biological determinants of species diversity. J. Biogeogr., 12, 1–20.

Siegert, F., Ruecker, G., Hinrichs, A. & Hoffmann, A.A. (2001) Increased damage from fires in logged forests during droughts caused by El Niño. Nature, 414, 437–440.

Silmi, M., Mislan, S. & Dahlen, B. (2013) Using leopard cats (Prionailurus bengalensis) as biological pest control of rats in a palm oil plantation. Indones. Nat. Hist., 1, 31–36.

Sist, P., Fimbel, R., Sheil, D., Nasi, R. & Chevallier, M.-H. (2003) Towards sustainable management of mixed dipterocarp forests of South-east Asia: moving beyond minimum diameter cutting limits. Environ. Conserv., 30, 364–374.

Slade, E.M., Mann, D.J. & Lewis, O.T. (2011) Biodiversity and ecosystem function of tropical forest dung beetles under contrasting logging regimes. Biol. Conserv., 144, 166–174.

Soberón, J. (2007) Grinnellian and Eltonian niches and geographic distributions of species. Ecol. Lett., 10, 1115–1123.

Sodhi, N., Koh, L., Prawiradilaga, D., Tinulele, I., Putra, D. & Tongtan, T. (2005) Land use and conservation value for forest birds in Central Sulawesi (Indonesia). Biol. Conserv., 122, 547–558.

Sodhi, N.S., Koh, L.P., Brook, B.W. & Ng, P.K.L. (2004) Southeast Asian biodiversity: an impending disaster. Trends Ecol. Evol., 19, 654–660.

Sodhi, N.S., Lee, T.M., Koh, L.P. & Brook, B.W. (2009a) A meta-analysis of the impact of anthropogenic forest disturbance on Southeast Asia’s biotas. Biotropica, 41, 103–109.

Sodhi, N.S., Posa, M.R.C., Lee, T.M., Bickford, D., Koh, L.P. & Brook, B.W. (2009b) The state and conservation of Southeast Asian biodiversity. Biodivers. Conserv., 19, 317–328.

Sodhi, N.S., Wilcove, D.S., Lee, T.M., Sekercioglu, C.H., Subaraj, R., Bernard, H., Yong, D.L., Lim, S.L.H., Prawiradilaga, D.M. & Brook, B.W. (2010) Deforestation and avian extinction on tropical landbridge islands. Conserv. Biol., 24, 1290–1298.

Soininen, J., Lennon, J. & Hillebrand, H. (2007a) A multivariate analysis of beta diversity across organisms and environments. Ecology, 88, 2830–2838.

Soininen, J., McDonald, R. & Hillebrand, H. (2007b) The distance decay of similarity in ecological communities. Ecography, 30, 3–12.

Sollmann, R., Mohamed, A., Samejima, H. & Wilting, A. (2013) Risky business or simple solution – Relative abundance indices from camera-trapping. Biol. Conserv., 159, 405–412.

Steinbauer, M.J., Dolos, K., Reineking, B. & Beierkuhnlein, C. (2012) Current measures for distance decay in similarity of species composition are influenced by study extent and grain size. Glob. Ecol. Biogeogr., 21, 1203–1212.

Stephens, S., Noons, D.N., Rotella, J.J. & Willey, D.W. (2003) Effects of habitat fragmentation on avian nesting success: a review of the evidence at multiple spatial scales. Biol. Conserv., 115, 101–110.

Stibig, H., Achard, F. & Fritz, S. (2004) A new forest cover map of continental southeast Asia derived from SPOT-VEGETATION satellite imagery. Appl. Veg. Sci., 7, 153–162.

Stibig, H., Beuchle, R. & Achard, F. (2003) Mapping of the tropical forest cover of insular Southeast

Bibliography

230

Asia from SPOT4-Vegetation images. Int. J. Remote Sens., 24, 3651–3662. Stibig, H., Stolle, F., Dennis, R. & Feldkotter, C. (2007) Forest cover change in Southeast Asia–the

regional pattern. European Commission JRC Scientific and Technical Reports. Struebig, M.J., Fischer, M. & Gaveau, D.L.A. (2015a) Anticipated climate and land-cover changes

reveal refuge areas for Borneo’s orang-utans. Glob. Chang. Biol., [in press]. Struebig, M.J., Kingston, T., Petit, E.J., Le Comber, S.C., Zubaid, A., Mohd-Adnan, A. & Rossiter,

S.J. (2011) Parallel declines in species and genetic diversity in tropical forest fragments. Ecol. Lett., 14, 582–590.

Struebig, M.J., Kingston, T., Zubaid, A., Le Comber, S.C., Mohd-Adnan, A., Turner, A., Kelly, J., Bożek, M. & Rossiter, S.J. (2009) Conservation importance of limestone karst outcrops for Palaeotropical bats in a fragmented landscape. Biol. Conserv., 142, 2089–2096.

Struebig, M.J., Kingston, T., Zubaid, A., Mohd-Adnan, A. & Rossiter, S.J. (2008) Conservation value of forest fragments to Palaeotropical bats. Biol. Conserv., 141, 2112–2126.

Struebig, M.J., Turner, A., Giles, E., Lasmana, F., Tollington, S., Bernard, H. & Bell, D. (2013) Quantifying the biodiversity value of repeatedly logged rainforests: gradient and comparative approaches from Borneo. Adv. Ecol. Res., 48, 183–224.

Struebig, M.J., Wilting, A., Gaveau, D.L.A., Meijaard, E., Smith, R.J., Fischer, M., Metcalfe, K. & Kramer-Schadt, S. (2015b) Targeted conservation to safeguard a biodiversity hotspot from climate and land-Cover Change. Curr. Biol., 25, 372–378.

Stuebing, R.B. & Gasis, J. (1989) A survey of small mammals within a Sabah tree plantation in Malaysia. J. Trop. Ecol., 5, 203–214.

Styring, A.R., Ragai, R., Unggang, J., Stuebing, R., Hosner, P.A. & Sheldon, F.H. (2011) Bird community assembly in Bornean industrial tree plantations: effects of forest age and structure. For. Ecol. Manage., 261, 531–544.

Sunarto, S., Kelly, M.J., Parakkasi, K., Klenzendorf, S., Septayuda, E. & Kurniawan, H. (2012) Tigers need cover: multi-scale occupancy study of the big cat in Sumatran forest and plantation landscapes. PLoS One, 7, e30859.

Sunquist, M. & Sunquist, F. (2002) Wild Cats of the World. University of Chicago Press. Supp, S.R. & Ernest, S.K.M. (2014) Species-level and community-level responses to disturbance: a

cross-community analysis. Ecology, 95, 1717–1723. Sutherland, G., Harestad, A., Price, K. & Lertzman, K.P. (2000) Scaling of natal dispersal distances in

terrestrial birds and mammals. Conserv. Ecol., 4, 16. Szabo, J.K., Khwaja, N., Garnett, S.T. & Butchart, S.H.M. (2012) Global patterns and drivers of avian

extinctions at the species and subspecies level. PLoS One, 7, e47080. Tang, H.T. (1974) A brief assessment of the regeneration systems for hill forests in Peninsular

Malaysia. Malaysian For., 37, 263–270. Tawatao, N., Lucey, J.M., Senior, M., Benedick, S., Vun Khen, C., Hill, J.K. & Hamer, K.C. (2014)

Biodiversity of leaf-litter ants in fragmented tropical rainforests of Borneo: the value of publically and privately managed forest fragments. Biodivers. Conserv., 23, 3113–3126.

Thang, H.C. (1987) Forest management systems for tropical high forest, with special reference to Peninsular Malaysia. For. Ecol. Manage., 21, 3–20.

Thiollay, J. (1995) The role of traditional agroforests in the conservation of rain forest bird diversity in Sumatra. Conserv. Biol., 9, 335–353.

Thompson, W.L., White, G.C. & Gowan, C. (1998) Monitoring Vertebrate Populations. Academic Press, San Diego, California.

Tilman, D. (2004) Niche tradeoffs, neutrality, and community structure: a stochastic theory of

Bibliography

231

resource competition, invasion, and community assembly. Proc. Natl. Acad. Sci. U. S. A., 101, 10854–10861.

Tilman, D., May, R.M., Lehman, C.L. & Nowak, M.A. (1994) Habitat destruction and the extinction debt. Nature, 371, 65–66.

Tjørve, E. (2002) Habitat size and number in multi-habitat landscapes: a model approach based on species-area curves. Ecography, 1, 17–24.

Tobler, M.W., Hartley, A.Z., Carrillo-Percastegui, S.E. & Powell, G.V.N. (2015) Spatiotemporal hierarchical modelling of species richness and occupancy using camera trap data. J. Appl. Ecol., 52, 413–421.

Triantis, K.A., Mylonas, M., Lika, K., Natural, K.V. & Museum, H. (2003) A model for the species-area-habitat relationship. J. Biogeogr., 30, 19–27.

Tscharntke, T., Steffan-Dewenter, I., Kruess, A. & Thies, C. (2002) Characteristics of insect populations on habitat fragments: a mini review. Ecol. Res., 17, 229–239.

Tuomisto, H. (2010) A diversity of beta diversities: straightening up a concept gone awry. Part 1. Defining beta diversity as a function of alpha and gamma diversity. Ecography, 33, 2–22.

Turner, E.C., Abidin, Y.Z., Barlow, H., Tom, M., Nainar, A., Reynolds, G., Yusof, Y.B.I.N., Khoo, M.I.N.S. & Ewers, R.M. (2012) The Stability of Altered Forest Ecosystems Project: Investigating the Design of Human-Modified Landscapes for Productivity and Conservation. Plant., 88, 453–468.

Turner, E.C. & Foster, W.A. (2008) The impact of forest conversion to oil palm on arthropod abundance and biomass in Sabah, Malaysia. J. Trop. Ecol., 25, 23–30.

Turner, E.C., Snaddon, J.L., Fayle, T.M. & Foster, W.A. (2008) Oil palm research in context: identifying the need for biodiversity assessment. PLoS One, 3, e1572.

Turner, I. (1996) Species loss in fragments of tropical rain forest: a review of the evidence. J. Appl. Ecol., 33, 200–209.

Turner, W. (2014) Sensing biodiversity. Science, 346, 301–302. Tylianakis, J.M., Didham, R.K., Bascompte, J. & Wardle, D.A. (2008) Global change and species

interactions in terrestrial ecosystems. Ecol. Lett., 11, 1351–1363. Uezu, A., Beyer, D.D. & Metzger, J.P. (2008) Can agroforest woodlots work as stepping stones for

birds in the Atlantic forest region? Biodivers. Conserv., 17, 1907–1922. Uezu, A. & Metzger, J.P. (2011) Vanishing bird species in the Atlantic Forest: relative importance of

landscape configuration, forest structure and species characteristics. Biodivers. Conserv., 20, 3627–3643.

United Nations Population Division (2015) Population, consumption and the environment. Available from http://www.un.org/en/development/desa/population/publications/pdf/environment/PopulationConsumptionEnvironment2015.pdf.

Uryu, Y., Mott, C., Foead, N., Yulianto, K., Budiman, A., Setiabudi, Takakai, F., Nursamsu, Sunarto, Purastuti, E., Fadhli, N., Hutajulu, C.M.B., Jaenicke, J., Hatano, R., Siegert, F. & Stüwe, M. (2008) Deforestation, forest degradation, biodiversity loss and CO2 emissions in Riau, Sumatra, Indonesia. WWF-Indonesia.

Veech, J., Crist, T. & Summerville, K. (2003) Intraspecific aggregation decreases local species diversity of arthropods. Ecology, 84, 3376–3383.

Veech, J.A. (2005) Analyzing patterns of species diversity as departures from random expectations. Oikos, 108, 149–155.

Veech, J.A. (2006) A probability-based analysis of temporal and spatial co-occurrence in grassland birds. J. Biogeogr., 33, 2145–2153.

Bibliography

232

Veech, J.A. (2013) A probabilistic model for analysing species co-occurrence. Glob. Ecol. Biogeogr., 22, 252–260.

Veech, J.A. & Crist, T.O. (2010a) Diversity partitioning without statistical independence of alpha and beta. Ecology, 91, 1964–1969.

Veech, J.A. & Crist, T.O. (2010b) Toward a unified view of diversity partitioning. Ecology, 91, 1988–1992.

Vellend, M. (2010) Conceptual synthesis in community ecology. Q. Rev. Biol., 85, 183–206. Vellend, M., Baeten, L., Myers-Smith, I.H., Elmendorf, S.C., Beauséjour, R., Brown, C.D., De

Frenne, P., Verheyen, K. & Wipf, S. (2013) Global meta-analysis reveals no net change in local-scale plant biodiversity over time. Proc. Natl. Acad. Sci. U. S. A., 110, 19456–19459.

Vié, J., Hilton-Taylor, C. & Stuart, S. (2009) Wildlife in a Changing World - An Analysis of the 2008 IUCN Red List of Threatened Species. IUCN, Gland, Switzerland.

Volkov, I., Banavar, J.R., Hubbell, S.P. & Maritan, A. (2009) Inferring species interactions in tropical forests. Proc. Natl. Acad. Sci. U. S. A., 106, 13854–13859.

Vos, C.C., Verboom, J., Opdam, P.F. & Ter Braak, C.J. (2001) Toward ecologically scaled landscape indices. Am. Nat., 157, 24–41.

Waltert, M., Mardiastuti, A. & Mühlenberg, M. (2004) Effects of land use on bird species richness in Sulawesi, Indonesia. Conserv. Biol., 18, 1339–1346.

Wanger, T.C., Iskandar, D.T., Motzke, I., Brook, B.W., Sodhi, N.S., Clough, Y. & Tscharntke, T. (2010) Effects of land-use change on community composition of tropical amphibians and reptiles in Sulawesi, Indonesia. Conserv. Biol., 24, 795–802.

Wardle, D.A. & Bardgett, R.D. (2004) Human-induced changes in large herbivorous mammal density: the consequences for decomposers. Front. Ecol. Environ., 2, 145–153.

Warren-Thomas, E., Dolman, P.M. & Edwards, D.P. (2015) Increasing demand for natural rubber necessitates a robust sustainability initiative to mitigate impacts on tropical biodiversity. Conserv. Lett., [in press].

Watling, J.I. & Donnelly, M.A. (2006) Fragments as islands: a synthesis of faunal responses to habitat patchiness. Conserv. Biol., 20, 1016–1025.

Wearn, O.R., Reuman, D.C. & Ewers, R.M. (2012) Extinction debt and windows of conservation opportunity in the Brazilian Amazon. Science, 337, 228–232.

Weiher, E., Freund, D., Bunton, T., Stefanski, A., Lee, T. & Bentivenga, S. (2011) Advances, challenges and a developing synthesis of ecological community assembly theory. Philos. Trans. R. Soc. Lond. B. Biol. Sci., 366, 2403–2413.

Wells, K. & Bagchi, R. (2005) Eat in or take away - seed predation and removal by rats (Muridae) during a fruiting event in a dipterocarp rainforest. Raffles Bull. Zool., 53, 281–286.

Wells, K., Corlett, R.T., Lakim, M.B., Kalko, E.K. V. & Pfeiffer, M. (2009) Seed consumption by small mammals from Borneo. J. Trop. Ecol., 25, 555–558.

Wells, K., Kalko, E.K.V., Lakim, M.B. & Pfeiffer, M. (2007) Effects of rain forest logging on species richness and assemblage composition of small mammals in Southeast Asia. J. Biogeogr., 34, 1087–1099.

Welsh, A.H., Lindenmayer, D.B. & Donnelly, C.F. (2013) Fitting and interpreting occupancy models. PLoS One, 8, e52015.

Wethered, R. & Lawes, M.J. (2003) Matrix effects on bird assemblages in fragmented Afromontane forests in South Africa. Biol. Conserv., 114, 327–340.

Whitmore, T.C. (1984) Tropical Rain Forests of the Far East, 2nd edition. Clarendon Press, Oxford, UK.

Bibliography

233

Whittaker, R.J., Willis, K.J. & Field, R. (2001) Scale and species richness: towards a general, hierarchical theory of species diversity. J. Biogeogr., 28, 453–470.

Wicke, B., Sikkema, R., Dornburg, V. & Faaij, A. (2011) Exploring land use changes and the role of palm oil production in Indonesia and Malaysia. Land use policy, 28, 193–206.

Wickham, H. (2010) The split-apply-combine strategy for data analysis. J. Stat. Softw., 40, 1–29. Wiens, J. (1989) Spatial scaling in ecology. Funct. Ecol., 3, 385–397. Wiens, J.J., Ackerly, D.D., Allen, A.P., Anacker, B.L., Buckley, L.B., Cornell, H. V, Damschen, E.I.,

Jonathan Davies, T., Grytnes, J.-A., Harrison, S.P., Hawkins, B. a, Holt, R.D., McCain, C.M. & Stephens, P.R. (2010) Niche conservatism as an emerging principle in ecology and conservation biology. Ecol. Lett., 13, 1310–1324.

Wilcove, D., Rothstein, D., Dubow, J., Phillips, A. & Losos, E. (1998) Quantifying threats to imperiled species in the United States. Bioscience, 48, 607–615.

Wilcove, D.S., Giam, X., Edwards, D.P., Fisher, B. & Koh, L.P. (2013) Navjot’s nightmare revisited: logging, agriculture, and biodiversity in Southeast Asia. Trends Ecol. Evol., 28, 531–540.

Wilcove, D.S. & Koh, L.P. (2010) Addressing the threats to biodiversity from oil-palm agriculture. Biodivers. Conserv., 19, 999–1007.

Wilkie, D., Shaw, E., Rotberg, F., Morelli, G. & Auzel, P. (2000) Roads, development, and conservation in the Congo basin. Conserv. Biol., 14, 1614–1622.

Williams, B.K., Nichols, J.D. & Conroy, M.J. (2002) Analysis and Management of Animal Populations: Modeling, Estimation, and Decision Making. Academic Press, San Diego.

Willott, S., Lim, D., Compton, S.G. & Sutton, S.L. (2000) Effects of selective logging on the butterflies of a Bornean rainforest. Conserv. Biol., 14, 1055–1065.

Willott, S.J. (1999) The effects of selective logging on the distribution of moths in a Bornean rainforest. Philos. Trans. R. Soc. Lond. B. Biol. Sci., 354, 1783–1790.

Wilson, C.C. & Wilson, W.L. (1975) The influence of selective logging on primates and some other animals in East Kalimantan. Folia Primatol., 23, 245–274.

Wilson, K.A., Meuaard, E., Drummond, S., Grantham, H.S., Boitani, L., Catullo, G., Christie, L., Dennis, R., Dutton, I., Falcucci, A., Maiorano, L., Possingham, H.P., Rondinini, C., Turner, W.R., Venter, O. & Watts, M. (2010) Conserving biodiversity in production landscapes. Ecol. Appl., 20, 1721–1732.

Wilson, W.L. & Johns, A.D. (1982) Diversity and abundance of selected animal species in undisturbed forest, selectively logged forest and plantations in East Kalimantan, Indonesia. Biol. Conserv., 24, 205–218.

Wilting, A., Cord, A., Hearn, A.J., Hesse, D., Mohamed, A., Traeholt, C., Cheyne, S.M., Sunarto, S., Mohd-Azlan, J., Ross, J., Shapiro, A.C., Sebastian, A., Dech, S., Breitenmoser, C., Sanderson, J., Duckworth, J.W. & Hofer, H. (2010a) Modelling the species distribution of flat-headed cats (Prionailurus planiceps), an endangered South-East Asian small felid. PLoS One, 5, e9612.

Wilting, A., Fischer, F., Abu Bakar, S. & Linsenmair, K.E. (2006) Clouded leopards, the secretive top-carnivore of South-East Asian rainforests: their distribution, status and conservation needs in Sabah, Malaysia. BMC Ecol., 6, 16.

Wilting, A., Mohamed, A., Ambu, L.N., Lagan, P., Mannan, S., Hofer, H. & Sollmann, R. (2012) Density of the Vulnerable Sunda clouded leopard Neofelis diardi in two commercial forest reserves in Sabah, Malaysian Borneo. Oryx, 46, 423–426.

Wilting, A. & Mohamed, A. bin (2010) Wildlife surveys in Segaliud Lokan Forest Reserve KTS Plantations - Final Report. Available from http://www.forest.sabah.gov.my/tangkulap/Pages/KTS_Final_report_ConCaSa.pdf.

Wilting, A., Samejima, H. & Mohamed, A. (2010b) Diversity of Bornean viverrids and other small

Bibliography

234

carnivores in Deramakot Forest Reserve, Sabah, Malaysia. Small Carniv. Conserv., 42, 10–13. Wintle, B., Kavanagh, R., McCarthy, M.A. & Burgman, M.A. (2005) Estimating and dealing with

detectability in occupancy surveys for forest owls and arboreal marsupials. J. Wildl. Manage., 69, 905–917.

With, K.A. (1994) Using fractal analysis to assess how species perceive landscape structure. Landsc. Ecol., 9, 25–36.

Wong, M. (1986) Trophic organization of understory birds in a Malaysian dipterocarp forest. Auk, 103, 100–116.

Wood, B. (1984) A long-term study of Rattus tiomanicus populations in an oil palm plantation in Johore, Malaysia: I. Study methods and population size without control. J. Appl. Ecol., 21, 445–464.

Wood, B. & Liau, S. (1984a) A long-term study of Rattus tiomanicus populations in an oil palm plantation in Johore, Malaysia: II. Recovery from control and economic aspects. J. Appl. Ecol., 21, 465–472.

Wood, B. & Liau, S. (1984b) A long-term study of Rattus tiomanicus populations in an oil palm plantation in Johore, Malaysia: III. Bionomics and natural regulation. J. Appl. Ecol., 21, 473–495.

Wood, B.J. & Chung, G.F. (2003) A critical review of the development of rat control in Malaysian agriculture since the 1960s. Crop Prot., 22, 445–461.

Woodcock, P., Edwards, D.P., Fayle, T.M., Newton, R.J., Khen, C.V., Bottrell, S.H. & Hamer, K.C. (2011) The conservation value of South East Asia’s highly degraded forests: evidence from leaf-litter ants. Philos. Trans. R. Soc. Lond. B. Biol. Sci., 366, 3256–3264.

Woodruff, D.S. (2010) Biogeography and conservation in Southeast Asia: how 2.7 million years of repeated environmental fluctuations affect today’s patterns and the future of the remaining refugial-phase biodiversity. Biodivers. Conserv., 19, 919–941.

Wright, D., Patterson, B., Mikkelson, G., Cutler, A. & Atmar, W. (1997) A comparative analysis of nested subset patterns of species composition. Oecologia, 113, 1–20.

WWF (2012) Heart of Borneo: A Natural Priority for a Green Economy. Heart of Borneo Initiative, WWF Malaysia.

WWF-Malaysia (2009) High Conservation Value Forest (HCVF) toolkit for Malaysia: A national guide for identifying, managing and monitoring High Conservation Value Forests. Kuala Lumpur, Malaysia.

Wyatt-Smith, J. (1963) Manual of Malayan silviculture for inland forest. Malayan Forest Record 23, Forest Department, Kuala Lumpur, Malaysia.

Yamaura, Y., Andrew Royle, J., Kuboi, K., Tada, T., Ikeno, S. & Makino, S. (2011) Modelling community dynamics based on species-level abundance models from detection/nondetection data. J. Appl. Ecol., 48, 67–75.

Yoccoz, N., Nichols, J. & Boulinier, T. (2001) Monitoring of biological diversity in space and time. Trends Ecol. Evol., 16, 446–453.

Zanette, L., Doyle, P. & Trémont, S. (2000) Food shortage in small fragments: evidence from an area-sensitive passerine. Ecology, 81, 1654–1666.

Zurita, G.A. & Bellocq, M.I. (2010) Spatial patterns of bird community similarity: bird responses to landscape composition and configuration in the Atlantic forest. Landsc. Ecol., 25, 147–158.

Zvoleff, A. (2015) glcm: Calculate textures from grey-level co-occurrence matrices (GLCMs) in R. R package version 1.2. Available from http://cran.r-project.org/package=glcm

Appendices

235

Appendices

Appendix A – Species checklists for four study sites in south-east Sabah, Malaysia

Appendices

236

Table A1. Mammal species checklist for two protected areas (Maliau Basin and Brantian-Tatulit) and two unprotected areas (Kalabakan and oil palm plantations). Records were obtained from 2010 to 2015, using camera traps, live traps and direct sightings. Three levels of confidence were assigned to records: 3 (high confidence) = confirmed record, e.g. from a camera trap image; 2 (medium confidence) = possible record, i.e. unconfirmed but likely correct; 1 (low confidence) = unconfirmed record with limited evidence. Note, no live-trapping was done in the Brantian-Tatulit Virgin Jungle Reserve. Order Family Species Common name IUCN

statusa Maliau Basin

Conservation Areab Brantian-Tatulit

Virgin Jungle Reserve SAFE Project experimental area

(Kalabakan Forest Reserve) Oil palm

plantationsc Pholidota Manidae Manis javanica Sunda pangolin EN 3 3 3 3 Eulipotyphla Erinaceidae Echinosorex gymnura Moon rat LC 3 3 3 1 Soricidae Crocidura monticola Sunda shrew LC 1 Scandentia Ptilocercidae Ptilocercus lowii Pen-tailed treeshrew,

Feather-tailed treeshrew LC 3

Tupaiidae Tupaia minor Lesser treeshrew LC 2 3 Tupaia gracilis Slender treeshrew LC 2 3 3 Tupaia longipes Plain treeshrew,

Long-footed treeshrew LC 3 3 3 3

Tupaia tana Large treeshrew LC 3 3 3 3 Tupaia picta Painted treeshrew LC 1 Tupaia dorsalis Striped treeshrew DD 3 3 Dermoptera Cynocephalidae Galeopterus variegatus Colugo, Flying lemur LC 3 3 3 Primates Lorisidae Nycticebus menagensis Bornean slow loris VU 3 3 3 Tarsiidae Cephalopachus bancanus Western tarsier VU 3 3 3 Cercopithecidae Presbytis rubicunda Maroon langur,

Red leaf monkey LC 3 3 3

Presbytis hosei Hose's langur, Grey leaf monkey

VU 3 3 2

Trachypithecus cristatus Silvered langur, Silvery lutung

NT 1

Nasalis larvatus Proboscis monkey EN 1 Macaca fascicularis Long-tailed macaque,

Crab-eating macaque LC 3 3 3 3

Macaca nemestrina Southern pig-tailed macaque VU 3 3 3 3 Hylobatidae Hylobates muelleri Bornean gibbon EN 3 3 3 Pongidae Pongo pygmaeus Orangutan EN 1 3 3 Carnivora Canidae Canis familiaris Domestic dog LC 3 3 3 Ursidae Helarctos malayanus Sun bear VU 3 3 3 3* Mustelidae Martes flavigula Yellow-throated marten LC 3 3 3 3 Mustela nudipes Malay weasel LC 3 3 Mydaus javanensis Sunda stink badger, Malay

badger, Teledu LC 3 3 3

Aonyx cinereus Oriental small-clawed otter VU 3 3 3 Lutrogale perspicillata Smooth otter,

Smooth-coated otter VU 1

Viverridae Viverra tangalunga Malay civet LC 3 3 3 3

Appendices

237

Prionodon linsang Banded linsang LC 3 3 Paradoxurus hermaphroditus Common palm civet LC 3 2 3 3 Paguma larvata Masked palm civet LC 3 3 3 3 Arctictis binturong Binturong, Bear cat VU 3 3 3 Arctogalidia trivirgata Small-toothed palm civet LC 3 3 3 Hemigalus derbyanus Banded civet VU 3 3 3 3 Diplogale hosei Hose's civet VU 3 Herpestidae Herpestes brachyurus Short-tailed mongoose LC 3 3 3 3 Herpestes semitorquatus Collared mongoose DD 3 3 3 3 Felidae Neofelis diardi Sunda clouded leopard VU 3 3 3 Pardofelis marmorata Marbled cat VU 3 3 3 Pardofelis badia Bay cat EN 3 3 Prionailurus planiceps Flat-headed cat EN 3 Prionailurus bengalensis Leopard cat LC 3 3 3 3 Proboscidea Elephantidae Elephas maximus Asian elephant EN 3 3 Cetartiodactyla Suidae Sus barbatus Bearded pig VU 3 3 3 3 Tragulidae Tragulus napu Greater mouse-deer LC 3 3 3 Tragulus kanchil Lesser mouse-deer LC 3 3 3 Cervidae Muntiacus muntjak Red muntjac,

Common barking deer LC 3 3 3 3

Muntiacus atherodes Bornean yellow muntjac LC 3 3 3 Rusa unicolor Sambar deer VU 3 3 3 3 Bovidae Bos javanicus Banteng, Tembedau EN 3 3 Rodentia Sciuridae Lariscus hosei Four-striped ground squirrel NT 3 3 3 Callosciurus prevostii Prevost's squirrel LC 3 3 3 3 Callosciurus notatus Plantain squirrel LC 3 3 3 Callosciurus adamsi Ear-spot squirrel VU 3 Exilisciurus exilis Least pygmy squirrel DD 3 3 Sundasciurus lowi Low's squirrel LC 3 3 3 3 Sundasciurus tenuis Slender squirrel LC 3 Sundasciurus brookei Brooke's squirrel LC 3 Sundasciurus hippurus Horse-tailed squirrel NT 3 3 Ratufa affinis Giant squirrel NT 3 3 3 Rheithrosciurus macrotis Tufted ground squirrel VU 3 3 Petaurista petaurista Red giant flying squirrel LC 3 3 3 Aeromys thomasi Thomas's flying squirrel DD 3 3 3 Aeromys tephromelas Black flying squirrel DD 3 Petaurillus hosei/emiliae Pygmy flying squirrel DD 3 Muridae Rattus exulans Polynesian rat LC 3 3 Rattus rattus (=tanezumi) Black rat, House rat LC 3 3 Rattus tiomanicus Malaysian field rat LC 1 1 Sundamys muelleri Muller's rat LC 3 3 3*

Appendices

238

Maxomys surifer Red spiny rat LC 3 3 3 3* Maxomys rajah Brown spiny rat VU 3 3 3 3 Niviventer cremoriventer Dark-tailed tree rat VU 3 3 3 Maxomys whiteheadi Whitehead's rat VU 3 3 3 Maxomys baeodon Small spiny rat DD 3 Leopoldamys sabanus Long-tailed giant rat LC 3 3 3 3 Maxomys ochraceiventer Chestnut-bellied spiny rat DD 3 3 Haeromys

margarettae/pusillus Ranee mouse / Lesser ranee mouse

DD/VU 3

Hystricidae Hystrix brachyura Malay porcupine, Common porcupine

LC 3 3 3 3

Trichys fasciculata Long-tailed porcupine LC 3 3 3 Hystrix crassispinis Thick-spined porcupine LC 3 3 aThreat categories follow the IUCN Red List categories and criteria version 3.1: CR = critically endangered; EN = endangered; VU = vulnerable; NT = near threatened; LC = least concern, DD = data deficient. bRecords from the Maliau Basin Conservation Area do not include records from the logged-over buffer zone, which is known, for example, to harbour Asian elephant, banteng, Sunda clouded leopard, bay cat and flat-headed cat (O. R. Wearn, unpublished data). cSelengan Batu and Mawang oil palm estates within the SAFE landscape, managed by Benta Wawasan Sdn Bhd and Sabah Softwoods Sdn Bhd, respectively. Asterisks indicate records from scrub habitat at the oil palm plantation margins and not inside the oil palm crop itself.

Appendices

239

Table A2. Bird and reptile species checklist for two protected areas (Maliau Basin and Brantian-Tatulit) and two unprotected areas (Kalabakan, and the Mawang and Selangan Batu oil palm plantations). Records were obtained using camera traps (except where indicated), from 2010 to 2014.

Class Order Family Species Common name IUCN statusa

Maliau Basin Conservation Area

Kalabakan Forest Reserve

Brantian-Tatulit Virgin Jungle Reserve

Oil palm plantations

Reptilia Squamata Pythonidae Python reticulatus Reticulated pyton NE

Varanidae Varanus rudicollis Roughneck monitor NE

Varanus salvator Water monitor LC

Testudines Testudinae Manouria emys Asian brown tortoise EN b b

Aves Accipitriformes Accipitridae Nisaetus nanus Wallace's hawk-eagle VU

Spilornis cheela Crested serpent eagle LC

Columbiformes Columbidae Chalcophaps indica Emerald dove LC

Spilopelia chinensis Spotted dove LC

Cuculiformes Cuculidae Carpococcyx radiceus Bornean ground-cuckoo NT

Centropus sinensis Greater coucal LC

Galliformes Phasianidae Arborophila charltonii Chestnut-necklaced hill partridge VU

Argusianus argus Great argus NT

Coturnix chinensis Blue-breasted quail LC

Gallus gallus Red junglefowl LC

Lophura bulweri Bulwer's pheasant VU

Lophura ignita Crested fireback NT

Rollulus rouloul Crested partridge NT

Gruiformes Rallidae Amaurornis phoenicurus White-breasted waterhen LC

Rallina fasciata Red-legged crake LC

Passeriformes Estrildidae Lonchura atricapilla Chestnut munia LC

Muscicapidae Copsychus saularis Oriental magpie-robin LC

Copsychus stricklandii White-crowned shama NE

Nectariniidae Arachnothera longirostra Little spiderhunter LC

Pellorneidae Malacocincla malaccensis Short-tailed babbler NT

Malacopteron magnirostre Moustached babbler LC

Pellorneum capistratum Black-capped babbler LC

Ptilocichla leucogrammica Bornean ground-babbler VU

Pittidae Erythropitta arquata Blue-banded pitta LC c

Erythropitta ussheri Black-and-crimson pitta NT

Hydrornis baudii Blue-headed pitta VU

Hydrornis caeruleus Giant pitta NT

Hydrornis schwaneri Bornean banded pitta LC

Pitta sordida Hooded pitta LC

Pycnonotidae Pycnonotus goiavier Yellow-vented bulbul LC

Pelecaniformes Ardeidae Ardea alba Great egret LC

aLC = Least Concern, NT = Near Threatened, VU = Vulnerable, EN = Endangered, DD = Data Deficient, NE = Not Evaluated. bDirectly observed, not trapped. cRecord derived from live-trapping, not camera-trapping.

Appendices

240

Appendix B – Supplementary figures of sampling effort over time

Figure B1. Number of camera trap nights sampled in each of 46 sampling plots (after correcting for camera failure) across the four years of the study.

Appendices

241

Figure B2. Number of live trap nights of sampling in each of 31 plots, across the four years of the study.

Appendices

242

Appendix C – Example camera trap images of Bornean mammal and bird species

Plate C1. Marbled cat Pardofelis badia (image taken in the Maliau Basin Conservation Area).

Plate C2. Moon rat Echinosorex gymnura (image taken in the Maliau Basin Conservation Area).

Appendices

243

Plate C3. Sunda pangolin Manis javanica (image taken in the Maliau Basin Conservation Area). Although in decline globally due to hunting, this species was found in all land-use types.

Plate C4. Tufted ground squirrel Rheithrosciurus macrotis (image taken in the Maliau Basin Conservation Area). The function of the bushy tail is unknown, but may be used to distract predators.

Appendices

244

Plate C5. Bearded pig Sus barbatus (image taken in the Maliau Basin Conservation Area).

Plate C6. Banded civet Hemigalus derbyanus (image taken in the Maliau Basin Conservation Area). This species is usually solitary, but this image shows a chance encounter between conspecifics.

Appendices

245

Plate C7. Sunda clouded leopard Neofelis diardi (image taken in the Brantian-Tatulit Virgin Jungle Reserve). In areas with a denser understorey than in this image, males preferred to move along roads.

Plate C8. Bornean orangutan Pongo pygmaeus (image taken in the Kalabakan Forest Reserve). Orangutans were often seen, heard, or detected by their nests in the Kalabakan Forest Reserve.

Appendices

246

Plate C9. Southern pig-tailed macaque Macaca nemestrina (image taken in the Kalabakan Forest Reserve). This adaptable species was found to occur in old-growth and logged habitats alike.

Plate C10. Collared mongoose Herpestes semitorquatus (image taken in the Kalabakan Forest Reserve). This species was primarily found near streams and rivers, as in this image.

Appendices

247

Plate C11. Malay civet Viverra tangalunga (image taken in the Kalabakan Forest Reserve). This image is an unusual capture of this nocturnal civet during daylight hours.

Plate C12. Bay cat Pardofelis badia (image taken in the Kalabakan Forest Reserve). Few records of this Bornean endemic exist, but the Kalabakan Forest Reserve may be one its last strongholds.

Appendices

248

Plate C13. Banteng Bos javanicus (image taken in the Kalabakan Forest Reserve). Only a handful of banteng herds were recorded, and in all cases we found evidence of hunters operating in the vicinity.

Plate C14. Leopard cat Prionailurus bengalensis (image taken in the Selangan Batu oil palm plantation estate). This species was found to be more abundant in oil palm than in forest habitats.

Appendices

249

Plate C15. Striped treeshrew Tupaia dorsalis (image taken in the Maliau Basin Conservation Area). This species, which had not been recorded in Sabah for 25 years, was detected in two new study sites.

Plate C16. Bulwer’s pheasant Lophura bulweri (image taken in the Kalabakan Forest Reserve). A new population of this rare and threatened Bornean endemic was discovered during this study.

Mammalian community responses to a gradient of land-use ...

Documents

Transcript of Mammalian community responses to a gradient of land-use ...