The Effects of Fragmentation on Foliar Defensive Traits and Insect Herbivory Rates in Dipterocarp Seedlings

The effects of fragmentation on foliar defensive traits and insect herbivory rates in dipterocarp seedlings

Author Kinneen, Lois

Published 2018-12

Thesis Type Thesis (PhD Doctorate)

School School of Environment and Sc

DOI https://doi.org/10.25904/1912/1101

Downloaded from http://hdl.handle.net/10072/385011

Griffith Research Online https://research-repository.griffith.edu.au The effects of fragmentation on foliar defensive traits and insect herbivory

rates in dipterocarp seedlings

Lois Karabel Kinneen B.A. (mod.), MSc.

Submitted in fulfilment of requirements of the degree of Doctor of Philosophy Environmental Futures Research Institute School of Environment and Science Griffith University December 2018

Synopsis

Human-modified landscapes are ubiquitous and often made up of remnant fragments of natural ecosystems nested within an agricultural or urban matrix. Understanding how species are affected by habitat degradation is a central issue in biodiversity research, yet investigations into the impacts on key ecological interactions have not kept pace.

Gaining insight into the responses of ecological processes is vital in order to maximise biodiversity conservation and develop sustainable management practices in a changing world. Herbivory is a fundamental ecosystem process as it mediates the transfer of energy between primary production and higher trophic levels. In tropical rainforests herbivory is primarily carried out by insects.

Here, I investigate how leaf damage changes over time by carrying out repeated measures of herbivory following fragmentation. In doing so, I build upon previous

‘snapshot’ studies which have primarily quantified leaf damage at single points in time.

An experiment was established within a large-scale manipulation experiment: the

Stability of Altered Forest Ecosystems (SAFE) project in the Malaysian state of Sabah, within Borneo. I used seedlings of two species of endemic Dipterocarpaceae as the study system for two main reasons. Firstly, seedlings represent the most vulnerable life stage in a tree’s life cycle, and therefore insect herbivory may be a major determinant of their growth and survival. Secondly, not only do members of the family

Dipterocarpaceae dominate in lowland tropical rainforests of Southeast Asia, most are economically valued for their timber and are consequently under pressure from logging which is leading to conservation concern. Five hundred and seventy-six seedlings were planted in 12 recently isolated one hectare fragments and 12 continuous forest control sites. Eight of these control sites were located in an area of continuous forest estimated to be over one million hectares north of the experimental landscape of the SAFE

i project. Four further control sites were established in a virgin jungle reserve which is over 2,200 hectares and located to the south-west of the experimental area. All leaves were scanned in situ on six occasions over two field seasons (May- October) in 2015 and 2016. At the end of the experiment, traits analyses were performed to quantify three common metrics of leaf defence: total phenolic content, acid detergent fibre and leaf strength.

Through this experiment I addressed the following aims; (1) to determine the initial effects of fragmentation on rates of leaf damage, (2) to investigate whether bottom-up control of herbivory was altered by fragmentation through changes in foliar defence, and (3) to assess whether responses were shared among study species or were instead individualistic.

I also use data from a key published monograph to create species interaction networks between lepidopteran caterpillars and their known host plants in tropical Asia, with the aim of exploring the importance of members of the Dipterocarpaceae for insect herbivores.

I found some evidence that herbivory is disrupted by fragmentation, with both species exhibiting lower levels of herbivory in fragments. Forest type was an important predictor of patterns of leaf area loss in this experiment, but overall differences between herbivory in fragments and control sites were not significant. Instead, herbivory was best explained by seedling traits, and predictors of leaf area loss varied between species indicating species-specific responses. Relaxation of phytochemical defensive traits was also detected in forest fragments, perhaps due to decreased levels of herbivory. A reduction in phytochemical defence may imply an increased vulnerability of seedlings in forest fragments which translates to differences in herbivore damage over time.

I highlight the importance of repeated measures experiments when investigating a complex and dynamic ecological process such as herbivory, and propose long term monitoring to fully understand the effects of forest fragmentation. The results of this thesis contribute to understanding the effects of fragmentation on insect herbivory, which remain uncertain, and provide evidence of the extent to which this key ecosystem process is disrupted due to anthropogenic habitat modification.

iii

Statement of Originality

I declare that this thesis is my original work and has not previously been submitted for a degree or diploma in any university. To the best of my knowledge and belief this thesis contains no material previously published or written by another person except where due reference is made in the thesis itself.

______

Lois Karabel Kinneen

Table of Contents

Synopsis ...... i Statement of Originality ...... iv Table of Contents ...... v List of Figures ...... vii List of Tables ...... xi List of Figures in Supplementary materials ...... xii List of Tables in Supplementary materials ...... xvi Journal articles arising from this thesis ...... xvii Acknowledgments ...... xviii Chapter 1 Introduction, aims and thesis outline ...... 1 1.1 The importance of phytophagous insects ...... 1 1.2 The effects of habitat modification on insect herbivory ...... 3 1.3 Aims ...... 8 1.4 Thesis Structure ...... 9 Chapter 2 Lepidopteran herbivory in the Dipterocarpaceae of Tropical Asia ...... 12 2.1 Abstract ...... 12 2.2 Introduction ...... 13 2.3 Materials and methods ...... 17 2.4 Results ...... 20 2.5 Discussion ...... 27 Chapter 3 General methods ...... 38 3.1 Introduction to the study region ...... 38 3.2 The Stability of Altered Forest Ecosystems (SAFE) project...... 43 3. 3 Seedling experimental design ...... 47 3.4 Study species ...... 48 3.5 Experimental set-up...... 53 3.6 Leaf scanning protocol and image processing ...... 55 3.7 Seedling traits ...... 60 3.8 Research permits ...... 62 S3 Supplementary material ...... 63 Chapter 4 The effects of fragmentation on insect herbivory rates of seedlings in a tropical rainforest ...... 65

4.1 Abstract ...... 65 4.2 Introduction ...... 66 4.3 Materials and methods ...... 68 4.4 Results ...... 75 4.5 Discussion ...... 82 S4 Supplementary material ...... 89 Forest structure at SAFE...... 89 Statistical analysis (continued) ...... 89 Chapter 5 Palatability and defence in two species of Dipterocarpaceae planted across a fragmented tropical landscape ...... 100 5.1 Abstract ...... 100 5.2 Introduction ...... 101 5.3 Materials and methods ...... 105 5.4 Results ...... 113 5.5 Discussion ...... 120 S5 Supplementary material ...... 128 Chapter 6 General discussion ...... 140 6.1 Introduction ...... 140 6.2 Revisiting research aims ...... 140 6.3 Conclusions ...... 149 References ...... 150

List of Figures

Figure 2.1 The proportion of species from each lepidopteran family recorded on Dipterocarpaceae from Robinson et al. 2001...... 21 Figure 2.2 Bipartite food web showing binary interactions between 287 species of Lepidoptera (right) and 85 species of Dipterocarpaceae (left). Interactions are coloured according to lepidopteran family. Width of bars for each dipterocarp species is proportional to the number of lepidopteran species with which they are recorded to interact with...... 22 Figure 2.3 The number of species that exhibit each mean degree value for each family of Lepidoptera recorded to feed on members of the Dipterocarpaceae. A higher degree value indicates a higher level of generalism across the Dipterocarpaceae. These values were calculated at species level for each herbivore species from a binary food web of lepidopteran caterpillars and their dipterocarp host plants. Most species exhibited a degree value of 1.Two tortricid species outliers had degree values of 23 and 35 these were Andrioplecta pulverula and Andrioplecta shoreae...... 25 Figure 2.4 Proportion of records for the 14 different categories of records for lepidopteran caterpillars using dipterocarp species as host plants according to Robinson et al 2001...... 26 Figure 3.1 The experimental design of the Stability of Altered Forest Ecosystems (SAFE) project in Sabah, Malaysian Borneo. The experimental landscape is made up of six replicate blocks (A-F), each containing four 1-ha, two 10-ha and one 100- ha fragments. Continuous logged forest sites are located north of the experimental area; Logged Forest Edge (LFE) and Logged Forest (LF1-3). The Brantian-Tatulit Virgin Jungle Reserve (VJR) is found to the south-west of the experimental area. Control old growth forest (OG1-3) plots are located in Maliau Basin Conservation Area. Source: Ewers et al. 2011...... 46 Figure 3.2 Schematic of hierarchical block design used for the seedling experiment to test the initial effects of forest fragmentation on insect herbivory rates at the Stability of Altered Forest Ecosystems (SAFE) project, Malaysia...... 48 Figure 3.3 Unprocessed scan from Dryobalanops lanceolata seedling (individual D1A01) taken before planting, showing lanceolate shape of lamina and fine secondary venation...... 51

vii

Figure 3.4 Unprocessed scan of Parashorea tomentella seedling (individual D4A12) taken before planting, showing broad leaf shape and distinct secondary venation. 52 Figure 3.5 Seedling experimental set-up. (a) shows planted seedlings of both study species; Dryobalanops lanceolata (top) and Parashorea tomentella (bottom two), (b) Research assistants Kiky and Mus planting a seedling of D. lanceolata, (c) constructing an enclosure and (d) a finished enclosure containing 12 planted seedlings...... 54 Figure 3.6 Stepwise method of image processing, top panel depicts Dryobalanops lanceolata and lower panel depicts Parashorea tomentella. From left to right for both rows; (a) raw image, (b) cleaned background and removal of petioles and stems, (c) grayscale image depicting chewing damage; (d) manipulated image showing “total” leaf area, (e) image with pathogen damage removed, (f) grayscale image showing pathogen and chewing damage, (g) example of a binary (black and white) image used to quantify leaf area metrics in ImageJ...... 57 Figure 3.7 Diagram of each enclosure showing spacing of planted seedlings and where environmental measurements were made. Canopy closure, Photosynthetically Active Radiation (PAR) and slope were measured at the centre of each enclosure. Locations of additional slope measurements are depicted by the orange diamond.60 Figure 4.1 Map of the study area and experimental design. A seedling experiment was established with the Stability of Altered Forest Ecosystems project in Sabah, Malaysian Borneo. Seedlings were planted in twelve one hectare fragments (Forest blocks C, D and F), and also in four continuous forest controls in each of three areas: Brantian-Tatulit Virgin Jungle Reserve (VJR), Logged Forest (LF) and Logged Forest Edge (LFE). The circular insert shows enclosure set-up; twelve seedlings were planted 50cm apart and two of these enclosure plots were established at each sampling site...... 74 Figure 4.2 Time series depicting changes in herbivory rates (% leaf area loss) at enclosure level, over the course of the experiment for seedlings of Dryobalanops lanceolata. Thin lines represent the relationship observed in each of the 48 established enclosures and the mean herbivory rate for each forest type is also shown using thicker lines. Enclosures in forest fragments are depicted with dotted yellow lines and those in continuous are shown with solid green lines...... 75 Figure 4.3 Time series depicting changes in herbivory rates (% leaf area loss) at enclosure level, over the course of the experiment for seedlings of Parashorea viii

tomentella. Thin lines represent the relationship observed in each of the 48 established enclosures and the mean herbivory rate for each forest type is also shown using thicker lines. Enclosures in forest fragments are depicted with dotted yellow lines and those in continuous are shown with solid green lines...... 76 Figure 4.4 Back-transformed predicted relationship between seedling height and herbivory for each enclosure following linear mixed effects modelling, for both species together (left), Dryobalanops lanceolata (middle), and Parashorea tomentella (right). Shaded areas represent the standard error surrounding estimated relationships...... 77 Figure 4.5 Back-transformed predicted relationships between the number of conspecific seedlings and herbivory for each enclosure following linear mixed effects modelling, for Dryobalanops lanceolata and Parashorea tomentella analysed together. Shaded areas represent the standard error surrounding estimated relationships. For seedlings in fragmented forest, the difference in herbivory between a plot with zero conspecifics and a plot with six surviving conspecifics is 3.50% vs. 6.25% leaf area loss...... 78 Figure 4.6 Back-transformed predicted relationship between the number of leaves and herbivory for each enclosure following linear mixed effects modelling, for both species together (left), Dryobalanops lanceolata (middle), and Parashorea tomentella (right). Shaded areas represent the standard error surrounding estimated relationships. The number of leaves was a significant predictor and remained in the top models for when both species were analysed together, and for D. lanceolata. 79 Figure 5.1 Map of study area. A seedling experiment was established within the Stability of Altered Forest Ecosystems (SAFE) project, located in Sabah Malaysian Borneo. One hectare fragments in ‘Forest blocks’ “C, D and F” were chosen for this experiment. Continuous forest control sites were chosen in Logged Forest Edge (LFE), Logged Forest (LF) and Virgin Jungle Reserve (VJR)...... 106 Figure 5.2 The relationship between (A) forest type and (B) herbivore damage with total foliar phenolic content expressed as mg Gallic Acid Equivalents (GAE) per gram dried leaf extract for seedlings of Dryobalanops lanceolata and Parashorea tomentella...... 115 Figure 5.3 The relationship between relative growth rate and total foliar phenolic content expressed as mg Gallic Acid Equivalents (GAE) per gram dried leaf extract for seedlings of Dryobalanops lanceolata and Parashorea tomentella...... 116 ix

Figure 5.4 The relationship between (A) forest type, and (B) herbivore damage with percentage foliar acid detergent fibre (ADF) content for seedlings of Dryobalanops lanceolata and Parashorea tomentella...... 117 Figure 5.5 The relationships between (A) relative growth rate and (B) canopy closure with leaf strength for seedlings of Dryobalanops lanceolata and Parashorea tomentella...... 119

List of Tables

Table 2.1 Summary information for lepidopteran caterpillars recorded to solely feed on Dipterocarpaceae. ‘No. Species’ refers to the number of Lepidoptera from each family that fed only on dipterocarps. ‘Total No. dipterocarp host plants’ represents the number of dipterocarp species that are host to these Lepidoptera...... 24 Table 2.2 Food web metrics for the dipterocarp-only binary network and the larger binary network which included all host plants for lepidopteran caterpillars recorded to feed on Dipterocarpaceae across tropical South Asia...... 27 Table 4.1 Results of mixed effects modelling for total herbivory per enclosure averaged across best-fitting models, for (A) both species analysed together, (B) Dryobalanops lanceolata and (C) Parashorea tomentella. β represent model- averaged coefficients, standard errors (SE), 95% confidence intervals (CIs) are also presented. Significant predictors are shown in bold...... 80 Table 5.1 Mean (x̅ ) ± standard error of each defensive trait for seedlings of Dryobalanops lanceolata and Parashorea tomentella. Bold p-values indicate significant differences between the two species when analysed together using mixed effects modelling...... 114 Table 5.2 Coefficients for mixed effects models for Total Phenolic Content (expressed as mgGAE/gExtract) for both species analysed together, and for Dryobalanops lanceolata and Parashorea tomentella analysed separately. These coefficients derived from the minimal adequate model following model reduction. The ΔAIC between the reduced and full models are provided...... 116 Table 5.3 Coefficients for mixed effects models for percentage Acid Detergent Fibre content, for both species analysed together, and for Dryobalanops lanceolata and Parashorea tomentella analysed separately. These coefficients are derived from the minimal adequate model following model reduction. The ΔAIC between the reduced and full models are provided...... 118 Table 5.4 Coefficients for mixed effects models for Leaf Strength (expressed as kg/cm2) for both species analysed together, and for Dryobalanops lanceolata and Parashorea tomentella analysed separately. These coefficients are derived from the minimal adequate model following model reduction. The ΔAIC between the reduced and full models are provided...... 120

List of Figures in Supplementary materials

Figure S2.1 Phylogeny of plant genera included in the larger food web. This tree was used to order plants in both food webs. Genera highlighted in green represent those of the Dipterocarpaceae. Using Phylomatic v3, the genera included in this study were grafted onto a mega tree of tropical plants created by Slik et al. 2018 ...... 34 Figure S2.2 Phylogeny of lepidopteran families taken from Breinholt et al. 2018, used to order the lepidopteran families for both food webs...... 35 Figure S2.3 Bipartite food web showing binary interactions for 287 dipterocarp-feeding lepidopteran species (right) and all recorded host plant genera for tropical South Asia (left). Interactions are coloured according to herbivore Family. Host plant data is based on 1175 species and pooled to genus level (557 genera). The 11 dipterocarp genera are highlighted in light green...... 37 Figure S4.1 Back-transformed predicted relationship between the number of dropped leaves and herbivory (%) in each enclosure following linear mixed effects modelling, for both species together (left), Dryobalanops lanceolata (middle), and Parashorea tomentella (right). Shaded areas represent the standard error surrounding estimated relationships...... 92 Figure S4.2 Back-transformed predicted relationship between canopy closure (%) and herbivory (%), following linear mixed effects modelling for Dryobalanops lanceolata. Shaded areas represent the standard error surrounding estimated relationships...... 93 Figure S4.3 Back-transformed predicted relationship between the number of new leaves and herbivory (%) following linear mixed effects modelling for Parashorea tomentella. Shaded areas represent the standard error surrounding estimated relationships...... 93 Figure S4.4 Back-transformed predicted relationship between counts of Curculionoidea and herbivory (%) following linear mixed effects modelling, for both species together (left), Dryobalanops lanceolata (middle), and Parashorea tomentella (right). Shaded areas represent the standard error surrounding estimated relationships. The left panel shows the predicted relationships without data overlaid; the right shows mean values at each enclosure...... 94 Figure S4.5 Back-transformed predicted relationship between slope (angle of inclination within each enclosure) and herbivory (%) following linear mixed effects

xii

modelling, for both species together (left), Dryobalanops lanceolata (middle), and Parashorea tomentella (right). Shaded areas represent the standard error surrounding estimated relationships...... 94 Figure S4.6 Pairplots of the response variable, leaf area loss (Herbivory) and explanatory variables for both species together. The upper panel contains scatterplots between each explanatory variable. Pearson’s correlation coefficients are provided in the lower panel, the font size of correlation coefficients is proportional to their value...... 95 Figure S4.7 Pairplots of explanatory variables for Dryobalanops lanceolata data versus the response variable leaf area loss (Herbivory). The upper panel contains scatterplots between each explanatory variable. Pearson’s correlation coefficients are provided in the lower panel, the font size of correlation coefficients is proportional to their value...... 96 Figure S4.8 Pairplots of explanatory variables for Parashorea tomentella data versus the response variable leaf area loss (Herbivory). The upper panel contains scatterplots between each explanatory variable. Pearson’s correlation coefficients are provided in the lower panel, the font size of correlation coefficients is proportional to their value...... 97 Figure S 5.1 Pairplot of the response variable, concentration of total phenolics expressed as mg gallic acid equivalents per g dried extract, and explanatory variables for both species together. The upper panel contains scatterplots between each explanatory variable. Pearson’s correlation coefficients are provided in the lower panel, the font size of correlation coefficients is proportional to their value...... 128 Figure S5.2 Pairplot of the response variable, concentration of total phenolics expressed as mg gallic acid equivalents per g dried extract, and explanatory variables for Dryobalanops lanceolata. The upper panel contains scatterplots between each explanatory variable. Pearson’s correlation coefficients are provided in the lower panel, the font size of correlation coefficients is proportional to their value...... 129 Figure S5.3 Pairplot of the response variable, concentration of total phenolics expressed as mg gallic acid equivalents per g dried extract, and explanatory variables for Parashorea tomentella. The upper panel contains scatterplots between each explanatory variable. Pearson’s correlation coefficients are provided in the lower panel, the font size of correlation coefficients is proportional to their value...... 130 xiii

Figure S5.4 Pairplot of the response variable, percentage acid detergent fibre (ADF) content, and explanatory variables for both species together. The upper panel contains scatterplots between each explanatory variable. Pearson’s correlation coefficients are provided in the lower panel, the font size of correlation coefficients is proportional to their value...... 131 Figure S5.5 Pairplot of the response variable, percentage acid detergent fibre (ADF) content, and explanatory variables for Dryobalanops lanceolata. The upper panel contains scatterplots between each explanatory variable. Pearson’s correlation coefficients are provided in the lower panel, the font size of correlation coefficients is proportional to their value...... 132 Figure S5.6 Pairplot of the response variable, percentage acid detergent fibre (ADF) content, and explanatory variables for Parashorea tomentella. The upper panel contains scatterplots between each explanatory variable. Pearson’s correlation coefficients are provided in the lower panel, the font size of correlation coefficients is proportional to their value...... 133 Figure S5.7 Pairplot of the response variable, leaf strength expressed as kgcm-2, and explanatory variables for both species together. The upper panel contains scatterplots between each explanatory variable. Pearson’s correlation coefficients are provided in the lower panel, the font size of correlation coefficients is proportional to their value...... 134 Figure S5.8 Pairplot of the response variable, leaf strength expressed as kgcm-2, and explanatory variables for Dryobalanops lanceolata. The upper panel contains scatterplots between each explanatory variable. Pearson’s correlation coefficients are provided in the lower panel, the font size of correlation coefficients is proportional to their value...... 135 Figure S5.9 Pairplot of the response variable, leaf strength expressed as kgcm-2, and explanatory variables for Dryobalanops lanceolata. The upper panel contains scatterplots between each explanatory variable. Pearson’s correlation coefficients are provided in the lower panel, the font size of correlation coefficients is proportional to their value...... 136 Figure S5.10 Relationships between (A) canopy closure (B) leaf age with the response variable total phenolic content, expressed as mg Gallic Acid Equivalents (GAE) per g dried extract for seedlings of D. lanceolata and P. tomentella...... 137

xiv

Figure S5.11 Relationships between (A) relative growth rate, (B) canopy closure and (C) leaf age with the response variable percentage acid detergent fibre content for seedlings of D. lanceolata and P. tomentella...... 138 Figure S5.12 Boxplot showing differences in leaf strength between (A) forest type (B) damage class for seedlings of D. lanceolata and P. tomentella...... 139 Figure S5.13 Relationship between leaf age with the response variable leaf strength for seedlings of D. lanceolata and P. tomentella...... 139

List of Tables in Supplementary materials

Table S 2.1 Summary of the number of species for each genus of Dipterocarpaceae for which there were records of lepidopteran herbivores in Robinson et al. 2001...... 36 Table S3.1 Locations and coordinates of sampling sites used for the seedling planting experiment; these include SAFE 1-ha sites in experimental fragment blocks C, D, F. Continuous forest controls were chosen in three locations; LF refers to "Logged Forest", LFE; "Logged Forest Edge" and VJR; "Virgin Jungle Reserve"...... 63 Table S4.1 Number of surviving seedlings and the number of leaves at each time point during the experiment...... 91 Table S4.2 Variance inflation factors (VIFs) for maximal models of both species analysed together, and each species analysed separately. A VIF of 4 was chosen as a cut-off above which a variable would be removed from maximal model structure, variables removed from maximal model structure are highlighted in bold...... 91 Table S4.3 Summary data showing mean herbivory (% leaf area loss) at enclosure level ± standard error in each forest type for each species across all time points. The difference in leaf area loss between continuous forest and fragments are provided for each time point...... 92

xvi

Journal articles arising from this thesis

Chapters 2, 4 and 5 of this thesis have been prepared as manuscripts to be submitted to peer-reviewed scientific journals. Details of my contribution and that of each co-author are provided at the beginning of each chapter. Those who contributed to the research but did not qualify for authorship are included in the Acknowledgements section at the end of each chapter. Chapter 4 has been submitted to a special issue of Biotropica. Chapter

5 will be formatted appropriately and submitted to Oecologia shortly after the submission of my thesis. The bibliographic details of each of these manuscripts are given below.

Chapter 2

Kinneen, Lois K., Fayle, Tom M., Maunsell, Sarah C., Stork, Nigel E., & Kitching,

Roger L. (prepared manuscript) Lepidopteran herbivory in the Dipterocarpaceae of

Tropical Asia.

Chapter 4

Kinneen, Lois K., Fayle, Tom M., Hardwick, Jane L., Maunsell, Sarah C., Stork, Nigel

E., Vairappan, Charles S., & Kitching, Roger L. (2019) The effects of fragmentation on insect herbivory rates of seedlings in tropical rainforest, Biotropica, [In Review]

Chapter 5

Kinneen, Lois K., Anilik, Junia, Hardwick, Jane L., Kitching, Roger L., Maunsell,

Sarah C., Palaniveloo, Kishneth, Stork, Nigel E., Vickneswaran, Mathavan, &

Vairappan, Charles S., (prepared manuscript) Palatability and defence of two species of

Dipterocarpaceae experimentally planted in a fragmented tropical landscape.

Signed: ______Date: December 14th 2018 Lois K. Kinneen

Co-Signed:______Date: December 5th 2018 Roger L. Kitching (Principal Supervisor)

xvii

Acknowledgments

Firstly, I thank my supervisory team, Professor Emeritus Roger Kitching, Professor

Emeritus Nigel Stork, Dr Tom Fayle and Dr Sarah Maunsell. Each brought a unique style and set of expertise to their supervisory role. I am grateful to Tom for steering me in the direction of this PhD and for his never-ending optimism. To Roger and Nigel, I enjoyed immensely our discussions on ecology and science in general. I benefitted from their long and successful careers and am thankful for the time they invested in me despite being formally retired. Sarah, the ever practical voice of reason, not only provided me with a great role model, but was a constant source of support throughout my PhD. I have benefitted greatly from the generosity she has shown me with her time and knowledge. I also thank Professor Hamish McCallum, who was acting primary supervisor during the final stages of my PhD.

I am grateful to have had the opportunity to work in Australia and Malaysia during my

PhD, and thank Griffith Graduate Research School for providing me with a scholarship which enabled me to do so. My project was funded in part by Griffith University’s

School of Environment and Science, and also by an Australian Research Council (ARC) grant (DP160102078). I extend my thanks to the Universiti Malaysia Sabah, and

Professor Charles Vairappan who provided me with laboratory facilities which permitted me to complete the chemistry component of this thesis. Dr Glen Reynolds also deserves special thanks for his logistical support and advice while working in

Sabah.

I have been incredibly fortunate to have met some brilliant people during my PhD.

Spending thirteen months in a remote field camp provided me with a unique and invaluable experience which I will never forget. There, I was surrounded by fun,

xviii passionate and ambitious scientists. Camp life would not have been as enjoyable without the tireless work by Ryan Gray, Risma and Basri.

I am especially grateful for the SAFE project and LOMBOK research assistants for all of their hard work and assistance in the forest. I will forever be appreciative of the strength and patience they afforded me, despite bursting into a fit of giggles and shouting “photocopy” the first time I scanned my seedlings in the forest.

To those I met at Griffith University: Marisa, Tamara, Erica, Claire, Maryam, Mehran,

Elliot, Legi, Casey and Louise, I thank you for your camaraderie. I have had a great time working with each of you and wish you all the best in your careers. I hope our paths cross again someday. I also thank Emma Window, who helped process some of my leaf scans. Special thanks go to Jane and Cheneal, the three of us began our PhD journeys together and our friendship will extend far beyond our time at Griffith, I am so grateful to have met you both.

Last but by no means least, I thank my family for their never-ending support of my dreams and desire to explore and understand this wonderful world. To my Dad,

Andrew, who nurtured my curiosity and passion for nature from a young age, and who constantly reminds me to stop and smell the roses as I go through life, I am forever thankful for your love and support. To my siblings, Ally, Simon and Simon’s partner

Saoirse, thank you for your positivity and encouragement. To my niece, Amber, you have brought so much joy during the final year of my PhD, I hope I instil in you a love for science and nature.

xix

Chapter 1 Introduction, aims and thesis outline

1.1 The importance of phytophagous insects

Tropical arthropods make up the majority of terrestrial, eukaryotic diversity (Basset et al. 2015) and consequently insects have been described as “the little things that run the world” (Wilson 1987). In no other ecosystem is this more evident than in tropical forests (Ewers et al. 2015). Here, not only do they form an important component of the hyper-diversity of the region, following similar trends to other taxa in that they increase in diversity and abundance towards the equator (Salazar and Marquis 2012), they are also responsible for a number of vital ecological processes. Invertebrate species play a dominant role in many aspects of ecosystem functioning in tropical forests: they are involved in pollination, decomposition and seed predation, they provide important prey species for higher trophic levels, and are argued to be the most important group of primary consumers (Coley and Barone 1996, Ewers et al. 2015). Despite their importance, there are large knowledge gaps that must be addressed. While an estimated one million species of insect have been described, it is thought that 80% of all insects remain undiscovered, the majority of which are predicted to occur in the tropics (Stork

2018).

Herbivory is a vital ecosystem process within green food webs as it provides the link and accordingly, the transfer of energy between primary production and higher trophic levels (Terborgh et al. 2001). On a global scale, plant-insect herbivore interactions form some of the most dominant species interactions and include over 40% of terrestrial metazoan biodiversity (Price 2002, Novotny et al. 2010). Phytophagous insects have a wide range of impacts on their environment, from being an important component of

1 food web structure, to contributing to carbon and nutrient cycling, to influencing individual plant fitness. They therefore have cascading effects at both population and community levels (Coley and Barone 1996, Ruiz-Guerra et al. 2010, Bagchi et al. 2014,

Metcalfe et al. 2014).

Insect herbivore damage has been shown to increase decomposition rates in leaf litter

(Moreno et al. 2017). Increased soil fertility has been found to correlate with outbreaks of herbivory (Maguire et al. 2015). Similarly, leaf-cutting ants make “refuse dumps” where they deposit decomposing leaf and flower fragments creating a mosaic of patches associated with higher soil fertility due to the deposition of decomposing leaf and flower fragments (Wirth et al. 2008).

Herbivore damage can directly increase the risk of mortality in plants by acting as a point of entry for pathogen infection. Higher incidence of fungal infection has been found in individuals which exhibit greater herbivory (Benítez-Malvido et al. 1999).

Some herbivores, of course, are themselves vectors of plant pathogens (Nakazawa et al.

2012). There is also evidence that insect herbivores disproportionately target certain plant species, leading to changes in the structure of plant communities (Bagchi et al.

2014). By preferentially targeting certain species, herbivores play an important role in consumer regulation, preventing populations of those species from increasing in numbers (Wirth et al. 2008).

Herbivorous insects comprise a diverse group of organisms which exhibit a wide variety of feeding strategies (Rossetti et al. 2017). Consequently, herbivory is often categorised according to three predominant feeding types; folivores consume leaf material, florivores consume flowers and granivores are seed predators (Chávez-Pesqueira et al.

2015). Folivory is the most common form of herbivory in tropical ecosystems and,

2 consequently, folivores have been proposed as the most important primary consumers in these diverse systems (Coley and Barone 1996, Suzuki et al. 2013). In an analysis of the amount of leaf damage attributed to each insect leaf feeding herbivore guild in the tropics, it was found that 93% was carried out by leaf chewers, with leaf mining accounting for 4% and galling and skeletonising insects causing just 1.3% of damage

(Adams et al. 2009). This is further supported by a peak in galling and leaf mining activity found in temperate zones during a global assessment of insect herbivory

(Kozlov et al. 2015). As a result, over 50% of studies on insect herbivory have dealt solely with leaf-chewing taxa, this focus has meant that leaf-chewers are relatively well known to science (Novotny et al. 2010). Less studied forms of herbivory include suckers who feed on xylem and phloem tissue (Novotny et al. 2010), those that feed on extra-floral nectaries (Hunt 2003). Further, more cryptic forms of herbivory exist such as root feeding which are underrepresented in the literature (Hunter 2001a).

Insect herbivores comprise a particularly useful functional group with which to study anthropogenic disturbance as their abundance and species richness respond relatively quickly compared to vertebrate groups (Fayle et al. 2015) due to their short generation times (Chávez-Pesqueira et al. 2015). They are also a good indicator of the trajectory of any given system, since species at higher trophic levels, which are potentially more vulnerable to habitat modification, are dependent on them (Rao et al. 2001). The antagonistic ecological interaction between plants and their herbivores should be a conservation priority, as disruptions in the balance of these interactions can have consequences for entire ecosystems (Chávez-Pesqueira et al. 2015).

1.2 The effects of habitat modification on insect herbivory

Habitat modification such as logging, fragmentation and increased edge creation alters the microclimate of forests enabling both light and wind to penetrate deeper, and as a

3 consequence modified forests are often hotter and drier than intact systems (Bierregaard et al. 1992, Newbery et al. 1996). Carbon sequestration and storage is reduced by the removal of woody biomass through logging (Berry et al. 2010). This may be particularly catastrophic in the lowland dipterocarp forests of South-east Asia, which have been identified as one of the world’s most important carbon sinks (Ashton and

Kettle 2012). Fragmentation not only alters ecosystem functioning through changes in hydrological and biogeochemical cycling, it can also affect species composition and species interactions (Laurance et al. 2011).

While some fragmented landscapes such as islands and high altitude habitats in mountain ranges do occur naturally, anthropogenic activity is the leading cause of fragmentation (Ewers and Didham 2006). Anthropogenic fragmentation results in habitat loss or change, which alters the spatial distribution of natural environments, resulting in smaller patches nested in a more or less modified “unnatural” matrix across a landscape (Didham et al. 2012). The detrimental effects of fragmentation on ecological dynamics are exacerbated because habitat conversion tends to occur in a non- random fashion, often in already degraded habitats. Consequently, forest fragments remain only in areas of low agricultural productivity; either in steep areas, at high elevations or in areas of poor soil fertility or poor drainage (Laurance 2008). Isolated habitat fragments are less likely to be recolonised following local extinction, and edge effects are likely to have much greater detrimental impacts in fragments compared with continuous habitats. The combination of these impacts all result in a reduction in forest area and can lead to the loss of vulnerable species (Murcia 1995).

Historically, fragmentation research has been primarily focused on its effects on species abundance and richness (Valiente-Banuet et al. 2015). While species loss is undoubtedly a concern, an additional issue is when particular species that provide

4 ecosystem services such as seed dispersal, pollination or nutrient cycling are lost (Sodhi et al. 2004). Moreover, ecosystem services which are reliant on complex interactions among multiple species can be disrupted before species loss occurs (Valiente-Banuet et al. 2015). Yet investigations into the effects of fragmentation on ecological interactions have not kept pace with those focussed on species loss. As of 2015, as few as 10% of studies on the effects of fragmentation examined the consequences of fragmentation on ecological interactions, and those that did investigate ecological interactions primarily focused on mutualistic ones such as pollination or seed dispersal (Chávez-Pesqueira et al. 2015). A geographical bias in studies of the effects of fragmentation on insect herbivory is also evident, with more than 50% of studies having been carried out in the

USA, Germany and Switzerland. Most of this research has been based in fragmented grassland and temperate forest rather than in any other habitat type (Rossetti et al.

2017). Consequently, there remains a relative paucity of information on how fragmentation affects insect herbivory in tropical regions.

The studies that have addressed the effects of fragmentation on insect herbivory have yielded inconsistent results. By 2010, at least 16 studies had been carried out investigating the effects of habitat fragmentation on insect herbivory; however, a general consensus could not be reached from these studies, with half suggesting an increase in leaf area loss in smaller fragments whilst the other half showing a decrease

(Ruiz-Guerra et al. 2010). This general uncertainty has given weight to the suggestion that research in the field of fragmentation should be individualistically driven and look at species-specific responses to habitat change (De La Vega et al. 2012, Didham and

Ewers 2012).

The complexity of studying the impacts of fragmentation on insect herbivory lies in the multifaceted factors that can affect this process. Firstly, fragmentation can affect insect

5 herbivore communities directly through a decline in the abundance and species richness of phytophagous insects associated with reduced habitat area and increased isolation

(Rossetti et al. 2017). Herbivory can be reduced in fragments due to local extinctions of key herbivores (Benítez-Malvido et al. 2016). This may lead to differences in the patterns of damage and activity as different guilds of phytophagous insects may be more susceptible (Novotny et al. 2012, Rossetti et al. 2014). Conversely, the loss of plant species has consequences for the herbivores that rely upon them. Specialist guilds such as galling insects are particularly vulnerable and negatively affected by fragmentation

(Souza et al. 2018). The dispersal ability of different insect species is also an important factor to consider (Fáveri et al. 2008). The degree of isolation of a forest fragment can mean insects with reduced vagility and those which persist in small populations may be more vulnerable to inbreeding depression and extinction (Tscharntke and Brandl 2004).

The extent to which metapopulation theory can be applied to insect populations in tropical forests remains uncertain. A combination of high levels of plant species diversity and host specificity means that herbivores often naturally exist in patchy populations dependent on the spatial structure of their hosts (Lewis and Basset 2009).

At its most basic, Levins model suggests that while turnover of local populations can be high due to a complex series of extinction and colonisation events, the metapopulation remains relatively stable due to dispersal between patches containing suitable host plants (Hunter 2001b, Tscharntke et al. 2002). More complex models consider the size and arrangement of suitable habitat patches and propose local extinctions are a function of patch size while colonization events are dependent on patch location (Tscharntke et al. 2002). Both of these concepts have consequences for the conservation of tropical rainforest and management of modified landscapes. Conserving insect herbivores and maintaining healthy rates of herbivory will rely on being able to determine both the

6 occupancy and extinction rates of insects in these hyper diverse systems (Lewis and

Basset 2009). At a landscape level, habitat degradation such as fragmentation and the removal of suitable host plants through logging could increase the isolation of local herbivore populations and could have implications for the long term persistence of herbivores in a given area (Lewis and Basset 2009).

Indirect effects can include complex species interactions such as competition. There is increasing evidence that disturbance such as logging alters the functional composition in degraded systems, even shifting the functional importance to vertebrate species in some instances (Ewers et al. 2015) which has indirect effects on insect herbivory. In fragmented landscapes, edge habitats are often heavily browsed by vertebrate herbivores leading to changes in vegetation structure which in turn reduces the availability of certain types of plants required by insect herbivores (Bagchi et al. 2018).

Due to their trophic position, phytophagous insects are governed by bottom-up (changes in the plant resource) and top-down (predation by natural enemies) controls, both of which may be altered through habitat degradation (Wirth et al. 2008). Similar to the responses shown by phytophagous insects, the effects of top-down controls in forest fragments have been found to be variable. In general, the number of predators is expected to decrease in fragments, as abundance is usually correlated with area and connectivity at higher trophic levels (Bagchi et al. 2018). The removal of area-sensitive predatory species, however, may not translate to an overall decrease into top-down control of herbivory. There is, however, some evidence that generalist predators are less affected by fragmentation and they may provide some functional redundancy (Bagchi et al. 2018). Increased abundance of generalist predators in edge environments may lead to higher predation pressure in disturbed habitats, resulting in lower numbers of insect

7 herbivores (Ruiz-Guerra et al. 2012, Murphy et al. 2016), which may translate to lower levels of herbivore damage in these environments.

The effects of bottom-up and top-down forces on insect herbivory are further complicated as they rarely occur in isolation. The abundance of generalist predators is positively correlated with plant diversity, and consequently it has been suggested that by increasing plant diversity in agricultural systems insect herbivores can be suppressed

(Dassou and Tixier 2016). This ‘enemies hypothesis’, however, has been found to be most effective in systems with reduced diversity such as in agricultural landscapes, and further research is needed to support it in more complex systems such as tropical rainforests (Zhang and Adams 2011).

Bottom-up forces are thought to be the predominate driver in changes in dietary specificity in fragmented forests (Bagchi et al. 2018). Bottom-up and top-down controls may act in concert with one another. For example, reduced predation by ants was recorded in forest fragments characterised by lower plant species richness, which could not be significantly predicted by other fragment metrics such as size and distance from forest edge (Leles et al. 2017). In this thesis I have focussed on bottom-up controls through changes in environmental conditions and foliar traits.

1.3 Aims

This thesis aims to improve our understanding of the ecological responses to anthropogenic forest degradation, with a particular focus on forest fragmentation. My research feeds into a larger project which investigates the effects of habitat degradation on the tri-trophic interactions among plants, insect herbivores and their predators. The primary focus of this thesis is to explore the plant-insect herbivore dynamics of this larger system. Through a large-scale seedling experiment where specimens of two

8 species of Dipterocarpaceae were planted in one-hectare forest fragments and continuous forest control sites across a tropical landscape, I have addressed the following specific aims:

1. to understand the initial effects of forest fragmentation on insect herbivory;

2. to test whether the rate at which leaf damage occurs is disrupted through this form

of habitat modification;

3. to identify whether forest fragmentation alters bottom-up controls of insect

herbivory through changes in foliar defensive traits; and,

4. to determine whether these responses are shared between study species or are

individualistic.

Using a key published source on the known food-plants of lepidopteran caterpillars, I have also produced a meta-analysis to address the following aims:

5. to review members of the Dipterocarpaceae as a food source for lepidopteran larvae

throughout tropical Asia; and,

6. to determine the level of specialist relationships that have been recorded between

lepidopteran herbivores and their dipterocarp host plants.

1.4 Thesis Structure

This thesis is organised as a series of manuscripts ready for submission to peer- reviewed journals, with the exception of the present chapter and Chapters 3 (General methods) and 6 (General discussion). Chapter 2 is a review of “Lepidopteran herbivory in the Dipterocarpaceae of Tropical Asia”, which examines the current state of research on members of this plant family as a food source for herbivorous insects, and includes a food web constructed for dipterocarps and Lepidoptera across the Oriental region.

Chapter 3 is a “General methods” chapter providing the context for this PhD thesis; it gives a general overview of the study region and an introduction to the Stability of

Altered Forest Ecosystems (SAFE) project in Sabah, Malaysian Borneo, where fieldwork was carried out. This chapter also provides details of the experimental design used in this project, and an overview of the general procedures carried out during this research.

Following this, is Chapter 4 entitled “The effects of fragmentation on insect herbivory rates of seedlings in a fragmented tropical rainforest”. This paper aims to determine whether human disturbance, specifically following forest fragmentation, disrupts the key ecosystem process of insect herbivory. It explicitly examines the effects of fragmentation on rates of damage rather than single snapshot measures of damage which previous research studies have focussed on.

Chapter 5 is entitled “Palatability and defence in two species of Dipterocarpaceae planted across a fragmented tropical landscape”. Here, I investigate whether fragmentation alters insect herbivory through bottom-up changes in the plant resource; specifically the chapter examines whether three key foliar defensive traits are altered following forest fragmentation.

Chapter 6 concludes this thesis and provides a General discussion of the key findings from this research. Suggestions for future directions in the field of insect herbivory and fragmentation research are proposed.

Pertinent Supplementary materials can be found at the end of each chapter. To avoid repetition, a combined reference list is provided at the end of the thesis.

Statement of contribution to co-authored paper

Chapter 2 is a co-authored paper which has been prepared for publication. The bibliographic details of this co-authored paper are as follows:

Lois K. Kinneen, Tom M. Fayle, Sarah C. Maunsell, Nigel E. Stork, & Roger L. Kitching (prepared manuscript) Lepidopteran herbivory in the Dipterocarpaceae of Tropical Asia.

My contribution to this paper includes data collation, construction of food webs, data analysis and writing of the draft manuscript. Professor Roger L. Kitching was responsible for the original concept of this paper. Dr Sarah C. Maunsell made a large contribution to the direction and scope of this paper, and created the plant phylogeny used to order host plants in both food webs. Dr Tom M. Fayle and Professor Nigel E.

Stork co-supervised this project and assisted with edits of the manuscript.

(Signed) ______

Lois K. Kinneen Date: 10/12/18

(Co-signed) ______

Dr Tom M. Fayle Date: 13/12/18

(Co-signed) ______

Dr Sarah C. Maunsell Date: 14/12/18

(Co-signed) ______

Prof. Nigel E. Stork Date: 14/12/18

(Co-signed) ______

Prof. Roger L. Kitching Date: 14/12/18

Chapter 2 Lepidopteran herbivory in the Dipterocarpaceae of Tropical

Asia

Lois K. Kinneen1, Tom M. Fayle2, 3, 5, Sarah C. Maunsell1, 4, Nigel E. Stork1, & Roger

L. Kitching1

1Environmental Futures Research Institute & School of Environment and Science, Griffith University, Nathan, QLD 4111, Australia 2Biology Centre of the Czech Academy of Sciences, Institute of Entomology, Ceske Budejovice, South Bohemia, Czech Republic 3Faculty of Science, University of South Bohemia, Ceske Budejovice, South Bohemia, Czech Republic 4Museum of Comparative Zoology, Harvard University, Cambridge, MA 02138, USA 5Institute of Tropical Biology and Conservation, Universiti Malaysia Sabah, Kota Kinabalu, Sabah, Malaysia

2.1 Abstract

The Dipterocarpaceae are an ecologically dominant and economically valuable family of trees. In Southeast Asia they dominate forest stands and form a large portion of both canopy and subcanopy environments. As such, they are an important component of the biodiversity in species-rich tropical forests, and it follows that they are host to a wide array of phytophagous insects. The extent to which their herbivores are dependent solely on them is unclear. The aim of this study was to summarise existing information on interactions between Lepidoptera and their dipterocarp host plants. In addition, the importance of dipterocarps to these dipterocarp-feeding herbivores was assessed by exploring on which other host plants they have been recorded. Data were collated from a key monograph that includes known interactions between lepidopteran caterpillars and their host plants across tropical Asia. Qualitative bipartite food webs were created; first depicting interactions between Lepidoptera and their dipterocarp host plants and second

12 depicting all other host plants of these lepidopteran caterpillars. In total, 287 species of

Lepidoptera were recorded to feed on the Dipterocarpaceae and 38% of these species solely fed on members of this family. A high proportion of these dipterocarp-only interactions involve single species records. This highlights the need for further investigations to determine whether these truly reflect a high level of specialisation within the Dipterocarpaceae. The majority of records involved folivorous species, highlighting the need to investigate less common forms of herbivory such as stem and root feeders.

Key words: Lepidoptera; food web; dipterocarps; plant-insect interactions; dipterocarp

2.2 Introduction

The Dipterocarpaceae are a pantropical family of trees, with members principally found in Asia, but also in Africa and South America (Ashton and Kettle 2012). In Asia, the geographic breadth of dipterocarp forests spans from southern India to as far east as

New Guinea (Brearley et al. 2017). The forests of Southeast Asia, in particular, are unique in their high level of diversity and dipterocarp species form an important component of this ecosystem (Sakai, Momose, Yumoto, Nagamitsu, et al. 1999).

Lowland tropical rainforests, <500 m above sea level, have been found to be particularly diverse (Slik et al. 2003). The Dipterocarpaceae includes some of the tallest trees found in the tropics and dipterocarps make up the majority of emergent canopy trees in the tropical rainforests of Southeast Asia (Sakai, Momose, Yumoto, Nagamitsu, et al. 1999, Ashton and Kettle 2012). The family, however, also includes smaller tree species and therefore additionally contributes to an important component of the subcanopy environment (Ashton and Kettle 2012). Due to their dominance, members of the Dipterocarpaceae play an important role in maintaining the structure and function of forest across Southeast Asia (Brearley et al. 2017).

Part of the order Malvales, the Dipterocarpaceae is highly speciose, with an estimated

695 species from 16 genera worldwide (Christenhusz and Byng 2016); 510 species from

13 of these genera are distributed across Asia (Kettle 2010). Multiple species are able to coexist at relatively small spatial scales compared with other closely related species in tropical rainforests (Ashton and Kettle 2012). Previously, species of dipterocarps were thought to be habitat generalists. However, recent research has found that niche partitioning is much more common than formerly thought (Kettle et al. 2012). Certain species have specific habitat requirements and are dependent on particular soil conditions (pH levels, nutrient content, moisture levels and aeration) and the topography of the landscape (Newbery et al. 1996, Ashton and Kettle 2012). This results in a fine- scale arrangement of different species and any change in environmental conditions may alter the floristic structure of dipterocarp forests (Newbery et al. 1996, Kettle et al.

2012).

Dipterocarps exhibit “general flowering” and subsequent “mast fruiting” a few months later, meaning that their reproduction is usually synchronised (Sakai, Momose, Yumoto,

Nagamitsu, et al. 1999). The intervals between masting events can be irregular but they generally occur every 3-10 years (Hancock et al. 2005). This has important ecological impacts at higher trophic levels (Curran et al. 2004). It is thought that the strategy of synchronised reproduction helps satiate seed predators, reducing their overall negative impacts on tree regeneration (Sakai, Momose, Yumoto, Nagamitsu, et al. 1999). An abundance of fruit at a particular time has often been found to coincide with increased numbers of frugivores and is linked to increased faunal reproduction (Curran et al.

2004). Mast fruiting within the Dipterocarpaceae is thought to be triggered by environmental cues such as sharp decreases in night-time temperatures (Hancock et al.

2005), which can be associated with natural disturbance regimes like El Niňo cycles

(Curran et al. 1999, Sakai, Momose, Yumoto, Nagamitsu, et al. 1999). Modified landscapes including logged or fragmented forests are more susceptible to the negative impacts of El Niňo, in particular they endure longer periods of drought and wildfires, meaning that disturbance that once led to regeneration is now highly destructive (Curran et al. 2004). Moreover, climate change is manifesting in increased frequency of extreme climatic events, including El Niňo, which has further consequences for mast fruiting

(Butt et al. 2015).

Apart from having an important ecological role within ecosystems, dipterocarps produce timber of high economic value (Kettle et al. 2012). Towards the end of the last century, over 80% of all tropical timber came from just Indonesia, Malaysia and the

Philippines (Berry et al. 2010). It is feared that high levels of logging in this region have resulted in the local extinction of some dipterocarp species (Maycock et al. 2012). The vast majority of dipterocarp forests are now restricted to highly fragmented landscapes

(Kettle et al. 2012). Deforestation has been most intense in lowland areas due to easy access, the high density of commercially valuable dipterocarp trees and also because these areas are ideal for conversion to oil palm (Maycock et al. 2012). By 2010, all dipterocarp forests on Borneo located outside of protected areas were predicted to have been logged at least once. This has led to increasing concern that these areas of degraded forest are more vulnerable to conversion to agricultural uses such as oil palm

(Berry et al. 2010).

One of the most speciose orders of phytophagous insects is the Lepidoptera (Zahiri et al.

2012). They are the target taxa for many ecological studies since they are relatively well known taxonomically compared to other groups (Ehrlich and Raven 1964).

Encompassing both moths and butterflies, over 157,000 species have been described

(Breinholt et al. 2018). The majority of Lepidoptera rely on plant species throughout their entire life cycle and feed on plant tissue as caterpillars and nectar as adults (Regier et al. 2009, Krenn 2010). Due to their strong dependence on plants, the abundance of phytophagous insects parallels that of their host plants (Basset et al. 2015). Members of the Lepidoptera adopt a number of feeding strategies. Lepidopteran caterpillars are a dominant group of externally feeding phytophagous insects (Dyer et al. 2007), and are thus important within the leaf-chewing guild (Novotny et al. 2004). They also feed internally as leaf miners (Connor and Taverner 1997, Maunsell et al. 2015) and galling insects (Gaston et al. 1992, Shorthouse and Rohfritsch 1992) of both leaves (Mani

1964) and stems (Dhileepan 2003, Bedetti et al. 2013). Further species are inquilines which feed on galls induced by other insects and may inhabit these by displacing the original hosts (Yamazaki and Sugiura 2016). A key compendium of the feeding preferences for Lepidoptera is the work of Robinson et al. (2001) who compile information from a large number of published studies. In this chapter we use this valuable resource for our analysis as described further below.

Interaction networks comprise a series of nodes that are connected by links. Food webs are an extension of network theory that describe feeing interactions (Tylianakis et al.

2007). Depending on the scale in question, nodes can represent individuals, different taxonomic groups (species, genera, families etc.), communities or ecosystems (Poisot et al. 2016). Food webs are useful tools as they present the components of a biological system and further the understanding of how it works (Pimm et al. 1991). Interactions can represent mutualistic relationships such as pollination or antagonistic relationships such as herbivory and parasitism (Poisot et al. 2016). Increasingly they are being used in conservation biology to determine the vulnerability of ecosystems (McDonald-

Madden et al. 2016). The LifeWebs project is a large-scale venture that aims to study

16 food webs in the context of climate change and habitat degradations at a global scale.

LifeWebs’ aim is to identify and understand global patterns of change in these complex ecological interaction networks and thus not only contribute to the fundamental understanding of ecology but also to help predict how these systems will respond to habitat modification (Fayle et al. 2016).

In this chapter we aim to investigate the importance of the Dipterocarpaceae as host plants for lepidopteran caterpillars. Using a data-collation approach, our goal is to determine how many lepidopteran herbivores that have been recorded to use dipterocarps solely depend on members of this family.

2.3 Materials and methods

Data source

The Hostplants of the Moth and Butterfly Caterpillars of the Oriental Region by

Robinson et al. (2001) was the primary data source used for this study. This text is a collection of recorded interactions between lepidopteran caterpillars and plant species across tropical Asia. The geographical breadth of this database spans tropical South to

Southeast Asia, from Pakistan to Peninsular Malaysia, and extends to Borneo, Sulawesi, the Philippines and the tropical regions of China and the Ryukyu islands of Japan

(Robinson et al. 2001). Since laboratory-based feeding trials may yield interactions that are possible for lepidopteran herbivores but that do not occur in natural systems (Ehrlich and Raven 1964), records of host plants that were obtained through laboratory-based feeding trials were excluded from this study so that data would best reflect known natural host plant-Lepidoptera interactions.

Robinson et al. 2001 is divided into two parts so that it can be searched by host plant species or lepidopteran species. First, all of the members of the Dipterocarpaceae were

17 searched to identify the Lepidoptera that used them as host plants. Each lepidopteran species was then searched individually to ascertain what other plant species, if any, that they use as host plants.

Construction of bipartite food webs

First, data on lepidopteran species recorded to feed on members of the Dipterocarpaceae were collated and used to construct a binary matrix of interactions. This matrix was binary because feeding was recorded only in binary manner in Robinson et al 2001 i.e. abundance/frequency of interactions was not reported. These data were used to construct a dipterocarp only food web, which was constructed using recorded interactions among lepidopteran species and individual dipterocarp species. In order to investigate how specific these dipterocarp-feeding Lepidoptera were to the

Dipterocarpaceae, a larger web was constructed using recorded interactions among all dipterocarp-feeding Lepidoptera species and all plant genera recorded for the Oriental region. Although in most cases the text used recorded interactions at the species level, for the larger food web, host plant data were pooled at the plant genus level. The plantlist was used to validate the spelling of each plant species and to check for synonymies (The Plant List 2013). Both interaction matrices were ordered according to a community phylogeny of all plant genera recorded as hosts in Robinson et al. (2001).

The phylogenetic tree was created using the online version of Phylomatic v.3 whereby plant genera where grafted onto a mega tree (Slik et al. 2018) which was subsequently

‘pruned’. Since not all genera were represented in the mega tree, polytomies often occurred at lower levels (e.g. among genera) of the phylogeny, but major clades were resolved (Figure S2.1). Herbivorous Lepidoptera were ordered in food webs according to a recent phylogeny of lepidopteran families (Figure S2.2, (Breinholt et al. 2018)).

The R package Bipartite (Dormann et al. 2008) was used to construct bipartite food webs of (1) dipterocarp species and their known lepidopteran herbivores and (2) the whole known list of plant genera that these herbivores are also reliant upon.

Food web metrics

Metrics were calculated using the bipartite package (Dormann et al. 2008) in R version

3.5.1. To obtain a measure of specialisation for each lepidopteran on species of

Dipterocarpaceae, degree values were calculated for each species of Lepidoptera. This metric counts the number of plant species each insect is recorded on (Dormann 2011).

Connectance refers to the proportion of possible interactions that are realised within a given food web (Dunne et al. 2002). This metric can be interpreted as a network-level measure of redundancy within the system (Blüthgen et al. 2008). It is a qualitative measure calculated from binary data that does not take interaction strength into account, and is therefore a useful measure to describe binary food webs (Blüthgen, Menzel, et al.

2006).

The number of compartments was also calculated, which identifies the number of subgroups within a network that show interactions and are completely disconnected from each other (Krause et al. 2003, Fabian et al. 2013). Greater numbers of compartments within a food web is thought to imply increased stability (Krause et al.

2003). Compartment diversity was also determined for both food webs, and this measures the extent to which compartments are similar in size within a food web

(Fabian et al. 2013). It is calculated as Shannon’s diversity of compartment size

(Dormann et al. 2009). Higher values are associated with compartments of varying sizes.

Nestedness is common in interaction networks and describes the degree to which specialists (or those who interact with few species) interact with species with many interacting partners (Menke et al. 2012). There are several metrics with which the degree of nestedness within a network can be quantified: Nestedness metric based on

Overlap and Decreasing Fill (NODF) was chosen as it is a reliable metric that can be applied to a diverse array of interaction networks (Almeida-Neto et al. 2008). Values can range from zero to 100 with higher values indicating increased levels of nestedness

(Dormann et al. 2009).

Web asymmetry describes the balance between the two trophic levels, positive values are indicative of networks with greater numbers of consumer species whereas negative values indicate higher numbers of species within the lower trophic level (Dormann et al.

2009). Inferences on specialisation within a food web may also be made based on this metric, as strongly asymmetric webs with positive values (i.e. more consumers) can indicate low levels of specialist behaviour amongst those taxa (Blüthgen et al. 2007).

Web asymmetry is equal to zero when both trophic levels are balanced (Blüthgen et al.

2007).

2.4 Results

Two hundred and eighty seven species from 38 families of Lepidoptera were recorded to feed on 85 species of Dipterocarpaceae across tropical South Asia. This represents phytophagous lepidopteran records on 17% of the Asian dipterocarps. These species of

Lepidoptera were also recorded on 1,090 other host plant species from 138 different plant families. Records of lepidopteran caterpillars on members of the Dipterocarpaceae produced a complex web of interactions (Figure 2.2). However, these relationships represented only a small proportion of interactions observed in the larger food web

(Figure S2.3). In total, the larger web contained 2785 recorded interactions (Figure

S2.3), the dipterocarp food web containing 503 (18% of all interactions) (Figure 2.2).

Figure 2.1 The proportion of species from each lepidopteran family recorded on Dipterocarpaceae from Robinson et al. 2001.

Figure 2.2 Bipartite food web showing binary interactions between 287 species of Lepidoptera (right) and 85 species of Dipterocarpaceae (left). Interactions are coloured according to lepidopteran family. Width of bars for each dipterocarp species is proportional to the number of lepidopteran species with which they are recorded to interact with.

Lepidopteran herbivory within in the Dipterocarpaceae

Of the 287 recorded lepidopteran herbivores, 110 species (38% of all species) from 28 families were recorded to solely use dipterocarps as host plants (Table 2.1). 196 of the

287 Lepidoptera presented degree values of 1, indicating that they were recorded on just one host plant.

Fourteen out of the fifteen species of Gracillariidae included in this study solely fed on the Dipterocarpaceae. Nine of these species had degree values of one, a further three had degree values of two and one had a degree of three. Only one species was also recorded to feed on plant species outside of the Dipterocarpaceae.

By far the most generalist species that solely fed on dipterocarps were Andrioplecta shoreae and A. pulverula which were recorded on 35 and 23 different hosts, respectively (Figure 2.3). A. shoreae was one of the 110 species that solely fed on dipterocarp species. Within the Dipterocarpaceae, it was recorded to feed on species from eight of the 11 genera included in this analysis. A. pulverula was recorded on three other genera; Ficus, Quercus and Castanea. These three plant genera were grouped closely together in the phylogeny, and ordered genus numbers 438, 456 and 459 within the web (Figure S2.1).

Compared with other lepidopteran families, Erebidae contained the highest number as dipterocarp-feeding species, constituting 16% of all recorded species in the dipterocarp food web (Figure 2.1). Six species of Erebidae fed solely on members of the

Dipterocarpaceae (Table 2.1). Of these six, all but one species had degree values of one

(Hydrillodes minor, H. toresalis, Lymantria semicincta, L. subrosea and Simplicia turpatalis), H. gravatalis was associated with two host species.

Forty-four of the 85 species of dipterocarp with records of lepidopteran herbivores were members of the genus Shorea. S. robusta was host to the highest number of lepidopteran species which were primarily species of Erebidae (Figure 2.3). The genus with the second highest number of records was Dipterocarpus, which had information for 16 species. Single species records were present for four genera; Monotes,

Neobalanocarpus, Pentacme and Vateria (Table S2.1).

Figure 2.3 The number of species that exhibit each mean degree value for each family of Lepidoptera recorded to feed on members of the Dipterocarpaceae. A higher degree value indicates a higher level of generalism across the Dipterocarpaceae. These values were calculated at species level for each herbivore species from a binary food web of lepidopteran caterpillars and their dipterocarp host plants. Most species exhibited a degree value of 1.Two tortricid species outliers had degree values of 23 and 35 these were Andrioplecta pulverula and Andrioplecta shoreae.

Feeding strategies of dipterocarp-feeding Lepidoptera

Most herbivory was recorded on leaves (37% of all records within Dipterocarpaceae,

Figure 2.4). Rarer feeding strategies included Lepidoptera recorded specifically on pith, saplings, stems, galls, twigs and shoots (all <1% of records each, Figure 2.4). A relatively large portion of the records did not define the plant part on which lepidopteran caterpillars were recorded (14%, Figure 2.4).

Figure 2.4 Proportion of records for the 14 different categories of records for lepidopteran caterpillars using dipterocarp species as host plants according to Robinson et al 2001. Network-level patterns of herbivory

Food web structure in terms of connectance, number of compartments and compartment diversity was more complex in the dipterocarp-only food web (Table 2.2). Both food webs displayed asymmetry (Table 2.2). The dipterocarp food web was associated with a

26 positive value due to the higher number of herbivores relative to host plants. The larger food web was associated with a negative value for asymmetry indicating that there were more host plants than lepidopteran caterpillars present in that web. Fewer mean links per species was obtained in the dipterocarp web compared to the larger web. Low levels of nestedness were associated with both food webs (Table 2.2). Nestedness within the dipterocarp food web was slightly higher than that of the larger web.

Table 2.2 Food web metrics for the dipterocarp-only binary network and the larger binary network which included all host plants for lepidopteran caterpillars recorded to feed on Dipterocarpaceae across tropical South Asia.

Metric Dipterocarp food web Larger food web Connectance 0.021 0.018 Web asymmetry 0.542 -0.320 Mean links per species 1.356 3.312 Number of compartments 10 2 Compartment diversity 1.369 1.017 NODF 6.452 8.884

2.5 Discussion

Structure of food webs

Compartmentalisation, whereby communities can be divided into distinct subgroups, is usually higher in antagonistic networks such as those describing herbivore – host plant relationships. This is related to the adaptations by herbivores to overcome plant defence which naturally result in groups of related herbivores that feed on host plants with similar traits (Wardhaugh et al. 2015). Increased levels of compartmentalisation in food webs are thought to be associated with a lower risk of species extinctions (Fabian et al.

2013). While a higher number of compartments was detected in the dipterocarp-only food web, this may simply reflect the higher number of lepidopterans with a single host present in this web. Single species interactions are more likely to lead the network to be

27 split into compartments (Fabian et al. 2013), and therefore this metric is highly dependent on sampling intensity (Lewis et al. 2002). Moreover, the larger web was only split into two compartments suggesting little resilience within the larger system when records were pooled according to genera.

Relatively low connectance values were obtained for both food webs with proportions of realised links out of all hypothetically possible links being 0.021 and 0.018 for the dipterocarp web and whole web, respectively. This is unsurprising as this metric is strongly influenced by the size of the network, and decays with increasing complexity of food webs (Blüthgen, Menzel, et al. 2006). Connectance is also highly dependent on sampling intensity (Blüthgen et al. 2008). The large number of single species interactions suggested that these records may benefit from further sampling. Further research may lead to an increase in connectance values for both food webs. Moreover, since the webs cover the entire Oriental region, it is highly possible that many species of dipterocarps and herbivorous Lepidoptera do not overlap in geographical range, meaning that interactions among these species are not possible in the wild (forbidden links) (Olesen et al. 2011, Morales-Castilla et al. 2015).

The dipterocarp food web had a web asymmetry value which was positive, highlighting the fact that more herbivore species occur in this network than host plants. Greater numbers of consumer species usually reflects generalist behaviour of those taxa

(Blüthgen et al. 2007). A negative value for asymmetry was recorded in the larger food web, indicating there are more host plants than herbivore species. Asymmetry in the larger web is unsurprising as it could be considered ‘incomplete’, since only lepidopteran herbivores that were also recorded on dipterocarp species were included.

While these results may suggest that the dipterocarp food web is more asymmetric than the larger web, the relative strength of these metrics may not be comparable as the

28 larger web contains data pooled at genus-level. A higher mean number of links per species was obtained in the larger web, this is unsurprising as this web was larger and had more trophic interactions than the dipterocarp network.

Binary food web networks provide useful qualitative information on species interactions however, drawing conclusions on the level of specialism within binary networks should be done cautiously (Dormann and Blüthgen 2017), as they are highly sensitive to changes in sampling intensity (Banašek-Richter et al. 2004, Tylianakis et al. 2007). The nature of presence/absence data means no information on strength or number of interactions is given, as a consequence specialist behaviour may be underestimated if an insect species is recorded to utilise multiple species but they utilise a single species most often (Blüthgen, Menzel, et al. 2006).

Nevertheless, a large number of species had low degree values within the dipterocarp food web presented in this chapter. This may reflect specialist relationships between

Lepidoptera and dipterocarp species or a lack a general lack of information on feeding behaviour within this important plant family.

The Hostplants of the Moth and Butterfly Caterpillars of the Oriental Region provides a large number of interactions between dipterocarp species, their lepidopteran caterpillar herbivores and the additional plant species on which these feed. However, the paucity of information on sampling intensity within the collated text makes it difficult to make any firm conclusions from these data. Despite this, our dataset provides a useful platform from which to develop hypotheses and identify gaps in our current understanding of ecological interactions within this ecologically important plant family.

Records for the Dipterocarpaceae only included 85 of the 510 known species to occur across Asia. Whether this reflects a general lack of information on lepidopteran

29 herbivory within dipterocarp species or whether it is actually rare for members of this order to feed upon dipterocarps remains uncertain. Alternatively, the unsampled dipterocarp species might themselves be rare, meaning that feeding records are less likely to have been made. This analysis has highlighted the need to identify the extent to which the remaining 425 species of dipterocarp located throughout Asia are host to phytophagous lepidopteran insects. Furthermore, Lepidoptera are just one order of phytophagous insects and so there is a dearth of information on other phytophagous groups.

Patterns of lepidopteran herbivory within the Dipterocarpaceae

Despite the aforementioned limitations a number of interesting results were obtained during this exercise. Among the large number of lepidopteran species that were recorded on a single dipterocarp host plant was a high proportion of the Gracillariidae species included in this study. In total fifteen species were recorded to feed on dipterocarp species and only one of these was recorded to feed on a genus outside of the

Dipterocarpaceae (Acrocercops resplendens on Ficus). Many species within this family are leaf miners in their larval stage (Sinclair and Hughes 2010, Novotny et al. 2012).

Leaf miners tend to exhibit more specialist behaviour ((Novotny et al. 2010), so the low degree values obtained for species of Gracillariidae in this study may truly reflect a higher level of specialisation within this group. Furthermore, specialists are more vulnerable to habitat modification as they are directly dependent on their host plants

(Wirth et al. 2008, Rossetti et al. 2014, Bagchi et al. 2018). If a large number of specialist leaf miners are dependent on dipterocarps as host plants, they may be particularly vulnerable to the effects of logging.

Conversely, the tortricid moths A. pulverula and A. shoreae were recorded on many more host plants than the other lepidopteran caterpillars that are known to feed on dipterocarps. While A. shoreae was recorded to solely feed on members of the

Dipterocarpaceae, it had 35 host plants from eight different genera within this family.

Despite being a dipterocarp-specialist, we can reliably conclude that this species of

Lepidoptera displays generalist behaviour within the family. By far the best represented genus of dipterocarp in this analysis was Shorea. This mirrors its taxonomic representation as this is the most speciose genus of the Dipterocarpaceae (Blundell and

Peart 1998).

The effects of habitat modification on food webs

Ecosystems are comprised of a number of species connected through complex interactions (Terborgh et al. 2001). Loss of species due to habitat degradation or conversion can alter the structure of these interactions which may manifest in extinction cascades whereby species at higher trophic levels are also lost (Terborgh et al. 2001,

Valladares et al. 2012). In a changing world it is becoming increasingly important that trophic interactions are properly documented and examined (Fayle et al. 2016). This is particularly pertinent in tropical rainforest where there is growing evidence that food web dynamics maintain ecosystem functioning through a crucial balance between bottom-up and top-down forces (Gardner et al. 2009).

Changes in temperature, associated with altitudinal gradients have previously been found to alter the structure of food webs manifesting in lower parasitism in colder environments (Maunsell et al. 2015). Shifts in species distributions caused by climate change may lead to an increase in parasitism at higher elevations than before (Maunsell et al. 2015). Previous investigations into fragmentation on food web structure have

31 found that less specialised species occur at the ecotone between forest and farmlands, but that interactions between species in these edge environments are less vulnerable to species extinction (Menke et al. 2012).

Types of herbivory recorded in the Dipterocarpaceae

The majority of dipterocarp-Lepidoptera interactions were recorded on or within leaf tissue. This supports previous research suggesting that leaf-eating is the most common form of herbivory (Coley and Barone 1996, Suzuki et al. 2013). Fruits and seeds were the second and third most common plant parts on which lepidopteran caterpillars were recorded, together accounting for 35% of records. Given the focus on mast fruiting within the Dipterocarpaceae (Cannon et al. 2007) this is unsurprising. 14% of records were concerned with seed predation. While vertebrates are the primary seed predator pre-dispersal, insects have been identified as dominant post-dispersal seed predators for members of the Dipterocarpaceae (Chong et al. 2016). Another study that looked at changes in seed predation following logging found a shift in the functional composition of seed predators with vertebrates dominating seed predation in logged forest (Curran and Webb 2000).

Host specificity amongst seed predators within the Dipterocarpaceae is thought to be relatively low (Ghazoul 2016). This is again related to the phenomenon of mast fruiting and seeding. From a herbivore perspective, it is advantageous to adapt to feeding on a wide range of host plants in order to exploit seeds that are in short supply and also to maximise feeding during general flowering and seeding events (Nakagawa et al. 2003).

This ensures survival on host plants which provide intermittent and unpredictable seed supplies (Ghazoul 2016). Interestingly, as previously mentioned, the two herbivore species which exhibited the highest number of host plant species were two species of

32 tortricid moths; A. shoreae and A. pulverula, both of which were recorded to feed on the fruits and seeds of all of their hosts. Moths from the familes Tortricidae, Pyralidae,

Crambidae, Immidae, Sesiidae and Cosmopterigidae are known seed predators of dipterocarps. Adults of these moths oviposit on either flowers and fruits, the larvae mature internally as the fruit develops and emerge either just before or shortly after seed dispersal (Ghazoul 2016).

Few studies were concerned with other plant parts, for example there were no records of root feeders within the Dipterocarpaceae. However, this feeding strategy may be rare in lepidopteran caterpillars, it is more common for other groups such as some families of

Coleoptera and is rarely studied (Hunter 2001a). It would therefore be interesting to focus research efforts on this more cryptic form of herbivory. Of more concern is the relatively high incidence (14%) of studies that did not give information on which plant part the Lepidoptera was recorded.

Conclusion

The aim of this study was to determine the importance of the Dipterocarpaceae for lepidopteran caterpillars across tropical Asia. While 38% of the species known to feed on dipterocarps were not recorded to feed on any non-dipterocarp species, the lack of information on sampling intensity makes it difficult to ascertain whether these are true dipterocarp-specialists. Nevertheless, summarising the records of lepidopteran herbivores on the Dipterocarpaceae has been useful in relation to identifying important knowledge gaps. We hope that this dataset can act as a foundation for future data collation efforts relating to Lepidoptera-Dipterocarpaceae food webs. Finally, it is clear from results of this study that there is paucity of information on more cryptic guilds of herbivores.

S2 Supplementary material

Figure S2.2 Phylogeny of lepidopteran families taken from Breinholt et al. 2018, used to order the lepidopteran families for both food webs.

Table S 2.1 Summary of the number of species for each genus of Dipterocarpaceae for which there were records of lepidopteran herbivores in Robinson et al. 2001.

Genus No. Species Anisoptera 3 Dipterocarpus 16 Dryobalanops 4 Hopea 8 Monotes 1 Neobalanocarpus 1 Parashorea 2 Pentacme 1 Shorea 44 Vateria 1 Vatica 4

Figure S2.3 Bipartite food web showing binary interactions for 287 dipterocarp-feeding lepidopteran species (right) and all recorded host plant genera for tropical South Asia (left). Interactions are coloured according to herbivore Family. Host plant data is based on 1175 species and pooled to genus level (557 genera). The 11 dipterocarp genera are highlighted in light green.

Chapter 3 General methods

This chapter provides context and gives an overview of the design of a seedling experiment which forms the basis of the subsequent chapters. I describe the study region and techniques used for field data collection. Details of chemical analyses are provided in Chapter 5.

3.1 Introduction to the study region

The importance of South-east Asian rainforests

The forests of South-east Asia are of global importance for their high levels of diversity

(Sakai, Momose, Yumoto, Nagamitsu, et al. 1999). Recent estimates of the number of tropical tree species found similar species richness across the Indo-Pacific region compared to that in the neotropical forests of South America (Slik et al. 2015). Despite this, a meta-analysis on the effects of anthropogenic land-use change on biodiversity across the tropics found that forests in Southeast Asia are the most vulnerable (Gibson et al. 2011). Southeast Asia is experiencing the highest regional rate of deforestation across the tropics, and it is estimated that it will lose 75% of its original forests by 2100, which is predicted to translate to a loss of 40% of its biodiversity (Sodhi et al. 2004,

Edwards et al. 2011). This predicted loss is particularly important since the region holds a high number of endemic species and, if their habitats are not protected, they will become extinct (Sodhi et al. 2004). Less than 10% of the forest in South-east Asia is protected (Sodhi et al. 2010). Consequently, primary tropical rainforest in the region is restricted to a series of variously sized fragments, nested in highly disturbed landscapes

(Bruhl et al. 2003). Although it is widely believed that the protection of intact primary forest is vital for biodiversity conservation (Gibson et al. 2011), there has recently been a shift in research to determine conservation value of degraded systems such as those

38 that have been selectively logged (Berry et al. 2010). Modified landscapes are ubiquitous and so there is a current drive to find ways to maintain their biodiversity and ecosystem functioning. It is estimated that three quarters of Borneo’s iconic Pongo pygmaeus (Bornean Orangutan), for example, are found in degraded forests which are under the management of forestry departments (Meijaard and Sheil 2007).

Borneo

With an area of 746, 000 km2 (Wooster et al. 2012), Borneo is the world’s fourth largest island (Ashton 2010). Located on the Sunda shelf, it makes up the largest land mass of the region known as Sundaland, alongside the Malay Peninsula, Sumatra, Java and their smaller neighbouring islands (den Tex et al. 2010). The naturally occurring vegetation of Borneo is tropical forest, of which there are three dominant types; Mixed

Dipterocarp, Kerangas (heath) and Peat Swamp Forests (Ashton 2010). Borneo is the centre of distribution for the plant family Dipterocarpaceae (Ghazoul 2016) and Mixed

Dipterocarp forests comprise the highest tree species richness of any forest type on the island (Ashton 2010). Ten percent of the forests of Borneo are protected, however, approximately 50% the remaining forests are classed as forestry concessions and either have experienced or are due to undergo selective logging (Meijaard and Sheil 2007).

Politically, Borneo is governed by three nations: Malaysia, Brunei and Indonesia (Sodhi et al. 2004), which adds to the complexity of biodiversity management and conservation across the island. Although they share similar ecologies, threats to biodiversity such as fire have been found to differ among the three states (Langner and Siegert 2009,

Wooster et al. 2012). The forests of Indonesia have been found to suffer greater fire damage than those in Brunei or Malaysia suggesting that Malaysia and Brunei have more fire prevention and suppression measures which may be lacking in Indonesian

Borneo (Langner and Siegert 2009).

Habitat degradation through logging and land-use change in Malaysia

Logging and conversion to oil palm have been identified as the most critical threats to biodiversity in Malaysia. The country’s oil palm industry continues to expand rapidly, with an average annual growth of 6.6% between the years of 1996 and 2006, Malaysia also had the highest commercial logging rates in the tropics over the period of 2004-

2007 (Sodhi et al. 2010). Logging is a form of habitat degradation which involves the extraction of commercially valuable timber species and thus directly affects both the structure and composition of forests (Akutsu et al. 2007). The logging history of the region can largely be attributed to the economic importance of one single family, the

Dipterocarpaceae, Borneo’s dominant family of trees (Ghazoul 2016). With the exception of members of the genus Vatica, all species of dipterocarp are considered valuable timber resources (Sist et al. 2003).

Selective logging remains the main method of timber extraction in Malaysia

(Nussbaum, Anderson and Spencer, 1995) and was developed from more destructive methods of timber extraction such as clear-fell logging (Sist et al. 2003). In 1948, the

“Malayan Uniform System” (MUS) was first introduced whereby all trees over 45cm dbh (diameter at breast height) could be extracted. This policy included measures to produce a more desirable, even-aged forest stand which could subsequently produce higher densities of commercially valuable timber species for future logging cycles.

Poisoning and girdling of any trees that were deemed 'defective' and above 5cm dbh was commonly employed, as well as systematic understory clearance and liana cutting.

The implementation of MUS was so commercially successful that by the 1970s most

40 lowland forests in Malaysia had been exploited. Focus then shifted to upland habitats and the brief implementation of the “Modified Malayan Uniform System” (MMUS), under which enrichment planting of timber species was carried out in steeper areas where natural rates of regeneration is low. Results were inconsistent and MMUS was discontinued shortly after its introduction, leading to the development of “Selective

Management Systems” (SMS) (Sist et al. 2003).

Initially hailed as a more sustainable land-use practice than previous logging policies, selective logging, under SMS, was also seen as a beneficial alternative to direct conversion to agriculture. This stemmed from the discovery that preserving trees in situ benefitted such things as soil fertility (Meijer 1974). There is also no doubt that selective logging is less harmful to biodiversity than conversion to agriculture (Gibson et al. 2011), although this generally accepted statement is based on relatively few data on a highly selected set of taxa (e.g. mammals, birds, butterflies). This policy was also based on a minimum diameter above which trees could be felled over a 30-35 year logging rotation (Sist et al. 2003). These cut-offs were 45cm dbh for non-dipterocarp species and 50cm dbh for dipterocarp species (Sist et al. 2003). Whilst selective logging aims to remove large trees only, its sustainability is widely questioned as the extraction methods used cause high levels of collateral damage, with up to 50% of forest stands being damaged in the process (Johns 1988, Sist et al. 2003). This damage is primarily caused by the extraction methods used. Removal of large felled trees requires the creation of “skid trails” and often logs are stored nearby on 'log landings' before transporting them elsewhere producing infertile, weed-susceptible and highly compressed patches in consequence (Nussbaum et al. 1995). Forest regeneration is also generally stunted following selective logging, due to removal of the top soil which holds the seed bank, and the compaction of soils which can reduce water infiltration and

41 increase soil erosion (Nussbaum et al. 1995). Despite the frequent expression of these concerns, this logging method remains a pervasive threat to forest ecosystems with up to 50% of the world’s tropical forests are predicted to undergo selective logging in the future (Bicknell et al. 2015). Alternatives have been suggested such as Reduced Impact

Logging (RIL), which has a less harmful effect on biodiversity (Bicknell et al. 2015).

Sabah

The Malaysian state of Sabah located in North-eastern Borneo, has an area of 73,700 km2, comprising approximately one-tenth of the island’s land area (Hazebroek et al.

2012). The climate is wet tropical, experiencing 2000-3000 mm of rainfall annually with monthly averages rarely falling below 100mm. The driest months usually fall between July-September (Walsh 1996). Although rainfall regimes are very variable in space across Sabah, and there is evidence that drought frequency and intensity are increasing in the region (Walsh and Newbery 1999). As in the rest of Borneo, the dominant indigenous vegetation is forest, ranging from mangrove, swamp forest, lowland dipterocarp to montane forest (Marsh and Greer 1992). The principal forest type is lowland dipterocarp and Sabah is home to 70% of the species of dipterocarps found on Borneo, six of which are endemic to this state (Chung et al. 2013).

Consequently, Sabah has a substantial history of logging which began towards the end of the 19th century (Ibbotson 2014) but experienced significant expansion in the 1960s

(Phua et al. 2016). By the 1990s, most of the lowland dipterocarp forests had experienced degradation from at least one rotation of logging (Marsh and Greer 1992,

Phua et al. 2016). The state has now been identified as a global hotspot for tropical deforestation (Bryan et al. 2013). In 2010, it was estimated that 39.5% of the forests of

Sabah had been converted to other land-use types (Gaveau et al. 2014).

3.2 The Stability of Altered Forest Ecosystems (SAFE) project

Fieldwork for this thesis was carried out within the Stability of Altered Forest

Ecosystems project, hereafter referred to as the SAFE project, located in the South-east of Sabah. Due to the biological complexity which exists in tropical ecosystems, large- scale manipulative experiments are vital in order to understand the mechanisms which drive the ecology of these habitats (Fayle et al. 2015). Such studies are rare due to high costs and logistical constraints associated with whole ecosystem experiments (Fayle et al. 2015).

The SAFE project is built on a partnership between researchers and industry, whereby the effects of forest degradation on ecosystem functioning can be studied, with results having direct implications for policy and land-use management (Turner et al. 2011). By designing a landscape which includes forest fragments nested within oil palm plantations, the importance of heterogeneity within a human-managed system can be studied and the results should have major implications for the conservation of biodiversity and maintenance of key ecosystem functions (Turner et al. 2012). Previous large-scale manipulation studies focused on fragmentation include the Biological

Dynamics of Forest Fragments Project in Brazil which was established in 1979 and focussed on the clearing of primary Amazonian forest (Bierregaard et al. 1992). Before the establishment of these large-scale studies research at larger scales was mainly restricted to observational studies of the effects of fragmentation (Fayle et al. 2015).

These studies were carried out in fragmented landscapes that were not specifically designed for research and the presence of confounding factors within these systems makes inferences on the consequences of fragmentation particularly difficult (Ewers and

Didham 2006).

SAFE comprises a 7,200-hectare (ha) experimental area is located within the Kalabakan

Forest Reserve (4°42’N, 117°34’E) which has experienced multiple rotations of logging since 1978 (Ewers et al. 2011, Hardwick et al. 2015). The latest round involved

“salvage logging”, whereby trees of any size could be extracted (Ewers et al. 2015).

Within this experimental landscape a series of fragmented forest blocks (A-F, Figure

3.1) have now been established; each block contains four 1-ha, two 10-ha and one 100- ha circular fragments. Although not all fragments were isolated at time of this study, the matrix surrounding each will eventually be terraced and converted to oil palm plantation

(Turner et al. 2012). This experimental area is nested within a larger landscape containing different land-use types (Hardwick et al. 2015). The Brantian-Tatulit Virgin

Jungle Reserve is located to the south of the experimental area, and the Ulu Segama

Forest Reserve is to the north. SAFE also includes old growth forest sites within the

Maliau Basin Conservation Area, an IUCN category Ia reserve (Pfeifer et al. 2015).

However, these sites were not used in this study.

The forest at SAFE is extremely heterogeneous and heavily degraded in places due to its complicated logging history (Pfeifer et al. 2015). “Logged Forest Edge (LFE)” and

“Logged Forest (LF)” areas are known to have experienced two rounds of selective logging, carried out under a Modified Malaysian Uniform System; the first removing approximately 113m3ha-1, and the second 37m3ha-1 (Struebig et al. 2013). The fragmented forest blocks (A-F) underwent a similar extraction of 113m3ha-1 during its first round of logging, but during the second rotation the forest within these blocks were logged three times resulting in an average extraction rate of 66m3ha-1 (Struebig et al.

2013). The primary genera targeted during these logging regimes were Dryobalanops,

Dipterocarpus, Shorea and Parashorea (Pfeifer et al. 2015). The resulting landscape includes a high density of logging roads and skid trails (Struebig et al. 2013). While

44 protected, the Virgin Jungle Reserve (VJR) at SAFE has experienced some logging on its periphery which was carried out during the construction of access roads (Struebig et al. 2013). There is also evidence that further illegal logging may have taken place in the

VJR (Pfeifer et al. 2015).

A series of survey points have been established across the SAFE experiment, based on a hierarchical fractal sampling design. They can be categorised according to their spatial distribution; with first order points being spaced 56 m apart, second order points being

178m apart, third order points 560 m and fourth order being spaced 1780 m apart

(Ewers et al. 2011). Choice of which survey points to use is largely dependent on the taxa or processes being studied, with larger and more mobile animals requiring a larger scale (Marsh and Ewers 2013). The advantage of sampling at these established points across the SAFE project is that data from different studies can be linked (Ewers et al.

2011). Sites were also carefully selected to reduce the confounding effects of latitude, slope and elevation, so that the effects of land-use change can be determined (Struebig et al. 2013).

Figure 3.1 The experimental design of the Stability of Altered Forest Ecosystems (SAFE) project in Sabah, Malaysian Borneo. The experimental landscape is made up of six replicate blocks (A-F), each containing four 1-ha, two 10-ha and one 100-ha fragments. Continuous logged forest sites are located north of the experimental area; Logged Forest Edge (LFE) and Logged Forest (LF1-3). The Brantian-Tatulit Virgin Jungle Reserve (VJR) is found to the south-west of the experimental area. Control old growth forest (OG1-3) plots are located in Maliau Basin Conservation Area. Source: Ewers et al. 2011.

3. 3 Seedling experimental design

To address the aims of this thesis, a seedling transplantation experiment was carried out within the SAFE landscape. Seedlings were chosen as the study system for this project as their leaf material is readily accessible and repeated scans can be made in situ without the need for progressive destructive sampling. Previous research in tropical forest has found that 95% of seedlings suffer herbivory damage (Benítez-Malvido and

Lemus-Albor 2005). Seedlings also represent the most vulnerable stage in a plant's life cycle (Souza et al. 2013) and high levels of defoliation can cause premature death

(Chung et al. 2013). This has consequences for the entire forest community as increased seedling mortality reduces the ability of degraded forest to regenerate (Basset et al.

2001). Understories of intact tropical forest are relatively stable compared with canopy environments that are exposed to ever-changing levels of light, wind and moisture.

Forest disturbance such as logging and fragmentation, therefore, stands to cause perturbations to understory microclimates and can directly affect the foraging behaviour of insect herbivores (Basset et al. 2001).

Sites were selected based on those that had been fragmented at the start of the experiment in May 2015; these included SAFE project “second-order points” located at the centre of twelve 1-ha fragments in three replicate forest blocks C, D and F (see

Figure 3.1). At the time of this experiment, the chosen fragments were surrounded by a matrix which had been salvage-logged earlier that year. This matrix environment included regenerating secondary tree species, as well as vines and scrub. Twelve

“second order points” located in Logged Forest (LF), Logged Forest Edge (LFE) and the Brantian-Tatulit Virgin Jungle Reserve (VJR) (Fig. 3.1), were chosen as control sites in continuous forest. Although the VJR will eventually be isolated over the course of the SAFE project, its large area of 2,200 ha (Ewers et al. 2011), means it was treated

47 as a continuous forest site relative to the 1-ha fragmented sites for this experiment. See

Table 3.1 for details of sampling locations including coordinates.

Due to the hierarchical block design used in this experiment (Figure 3.2), a mixed modelling approach with nested random effects was used throughout this thesis. The details of the statistical analyses used for each chapter are presented separately within each chapter.

Figure 3.2 Schematic of hierarchical block design used for the seedling experiment to test the initial effects of forest fragmentation on insect herbivory rates at the Stability of Altered Forest Ecosystems (SAFE) project, Malaysia. 3.4 Study species

Two species of Dipterocarpaceae were used in this experiment: Dryobalanops lanceolata Burck. and Parashorea tomentella (Symington) Meijer. They were selected because both species are endemic to Borneo (Chung et al. 2013), are common in Sabah

(Meijer 1974) and were readily sourced from local nurseries. In addition, they are both commercially valuable timber crops (Meijer 1974). D. lanceolata in particular, has been described as the most important timber species of its genus in northern Borneo (Ashton

1964). This species is classed as endangered according to the IUCN red list, and is mainly threatened due to habitat loss (Ashton 1998).

The genus Dryobalanops comprises seven species, all of which are found in Borneo.

The distribution of species which are not endemic to Borneo extends throughout Malaya and Central Sumatra and intervening islands (Ghazoul 2016). Adults are tall emergent

48 trees which form a large proportion of the canopy, and consequently saplings often dominate understory communities (LaFrankie 2010). One of the main field diagnostics for identifying members of this genus is the arrangement of secondary venation on the leaves. Veins are straight, tightly packed and run parallel to each other from the midrib the lamina margin (Ghazoul 2016, Figure 3.3). D. lanceolata, known locally as Kapur paji (Ashton 1964), can reach a height of 70m. Its bark is characteristically grey in colour and, as the name suggests, it bears lanceolate leaves (Ashton 1964). In adult specimens, leaves vary in shape; thicker leaves contain additional stomata and shorter drip-tips are found at the top of adult trees compared to lower leaves on the same tree

(Ghazoul 2016). D. lanceolata is a medium hardwood (Brearley et al. 2003) and seedlings can tolerate a wide range of light environments (Scholes et al. 1997).

The genus Parashorea includes 14 species, six of which are found in Borneo. The genus has a wider distribution than Dryobalanops, with species found in South

Myanmar, Thailand, Indochina, South China, throughout the rest of Sundaland and the

Philippines (Ghazoul 2016). P. tomentella, or locally known as Urat mata beluda, is a light hardwood (Brearley et al. 2003). The species has broad leaves with leaf venation that is more distinct than in D. lanceolata (Ghazoul 2016, Figure 3.4).

The two study species contrast in their light requirements. P. tomentella is considered to be a relatively light-demanding species (Brearley et al. 2003) while D. lanceolata is described as intermediate (Yeong et al. 2016) to shade-tolerant (Brearley et al. 2003,

Brearley 2006). Despite D. lanceolata exhibiting a preference for more shaded environments (Tange et al. 1998), demonstrated by higher growth rates in shaded habitats (Itoh et al. 1999) members of this species are able to survive under high light conditions by dissipating excess light energy non-photochemically (meaning their photosynthetic rate does not increase after a certain point) (Scholes et al. 1997). In a

49 study of photosynthetic rates of dipterocarp seedlings, P. tomentella exhibited a lower

-2 -1 light-saturated photosynthetic rate of 6.12 μmolCO2m s compared with a rate of 7.29

-2 -1 μmolCO2m s in seedlings D. lanceolata (Eschenbach et al. 1998), lending further support that this species might have a wider ecological range with respect to light environments.

Figure 3.3 Unprocessed scan from Dryobalanops lanceolata seedling (individual D1A01) taken before planting, showing lanceolate shape of lamina and fine secondary venation.

Figure 3.4 Unprocessed scan of Parashorea tomentella seedling (individual D4A12) taken before planting, showing broad leaf shape and distinct secondary venation.

3.5 Experimental set-up

A total of 48 enclosures were established at 24 sites across the SAFE project (Figure

3.2). Twelve seedlings (six of each species) were planted in each enclosure, resulting in a total number of 576 seedlings being planted across the landscape. Chicken wire with a

1-cm mesh size, and 1.5-m Belian wood posts were used in the construction of each enclosure (Figure 3.5). The Belian wood used for these enclosures was sourced from the scraps left by logging activities carried out within the SAFE landscape. While this was not strictly an exclusion experiment, enclosures were constructed to provide evidence of access by mammals and to protect seedlings from footfall by other scientists at each sampling point. Over the course of the experiment, no enclosures were damaged by large mammals such as bearded pig and Asian elephant which are present in the area.

Two enclosures were destroyed by tree falls.

Seedlings were sourced from the nursery at the Sabah Biodiversity Experiment in the

Malua Forest Reserve (Hector et al. 2011). Seedlings were grown from seeds collected from the contiguous Ulu Segama and Malua Forest Reserves, following a single mast fruiting event and were germinated and grown under controlled shade conditions (Philip

Ulok, personal communication, November 2017). Individual metal tags embossed with unique identifying codes were constructed and attached to the base of each seedling using green or black wire. During planting, these tags were covered using the surrounding leaf litter, to prevent reflecting light which could act as a deterrent or attractant to organisms. Individuals were planted 50 cm apart, alternating individuals from the two species (Figure 3.7). All leaves on each seedling were numbered sequentially, using indelible ink, on the bottom-left of the midvein on the abaxial leaf surface.

Seedlings were monitored from May to October over two years (2015 and 2016). Visits to each site were made every six weeks during these field seasons, resulting in a total of six time points. This timeframe spans the dry “season” in Sabah, and was chosen as the most appropriate time for monitoring since peaks in insect herbivores and their associated leaf damage have previously been found to occur immediately following the wet season in the tropics, although a gradual decline in activity occurs over the dry season (Williams-Linera and Herrera 2003, Whitfeld et al. 2012).

Figure 3.5 Seedling experimental set-up. (a) shows planted seedlings of both study species; Dryobalanops lanceolata (top) and Parashorea tomentella (bottom two), (b) Research assistants Kiky and Mus planting a seedling of D. lanceolata, (c) constructing an enclosure and (d) a finished enclosure containing 12 planted seedlings.

3.6 Leaf scanning protocol and image processing

Leaf area scans were made using a Canon Canoscan Lide 210 flatbed scanner set at 600 dots per inch (DPI) resolution, powered through a laptop. Flatbed scanners have previously been proven to accurately measure leaf area and defoliation, and have the advantage of being less expensive than leaf area meters (O’Neal et al. 2002). A 30 cm ruler was attached to the scanning bed to provide scale for each image. Leaves were scanned before planting and then again in situ at five subsequent time points, taking care not to damage any seedlings. This resulted in a total of 3,591 scans of 3,506 unique leaves over the six sampling points. The number of leaves varied at each time point as some leaves were abscissed and new leaves appeared. In total, 10,180 data points

(individual leaf measurements) were collected over all time points. On the rare occasion that leaves were larger than the bed of the scanner, indelible ink was use to draw a line perpendicular to the midvein and multiple scans were made to ensure the entire leaf had been captured. During image processing, these scans were cropped along the drawn line and the sum of the areas of each portion was used for the area of the entire leaf.

Processing of each leaf scan was carried out using Adobe® Photoshop® CC software version 2015.0.1 20150722.r.168. Damage was demarcated using a combination of the

'paint bucket' (with tolerance set at 30) and 'paintbrush' (set at 10 pixels) tools. Leaf petioles, stems and any shadows were removed manually using the 'eraser' function (set at 10 pixels) to ensure the leaf area was precise. Following these initial steps the resulting image was converted to grayscale and saved according to the seedling ID, time point of scan and damage type (e.g. 'LFE1A01T1_chew').

To quantify the total estimated leaf area prior to any damage, the 'paintbrush' tool (set at

10 pixels), was used to mark the original estimated leaf outline if damage had occurred

55 along the leaf margin. This was also converted to grayscale and saved a separate image

(e.g. 'LFE1A01T1_total').

Leaf area was quantified using the software ImageJ. This involved adjusting the image threshold to produce a binary (black and white) form, whilst taking care to ensure leaf area is not altered. Scale was set by measuring the number of pixels along a 2cm portion of the ruler on each scan. Leaf area loss was calculated as the percentage of the total leaf area that had been damaged (Equation 3.1), i.e.:

% Leaf Area Loss = (Total Leaf Area-Area of Chewed Leaf)/Total Leaf Area x 100 (3.1)

Figure 3.6 Stepwise method of image processing, top panel depicts Dryobalanops lanceolata and lower panel depicts Parashorea tomentella. From left to right for both rows; (a) raw image, (b) cleaned background and removal of petioles and stems, (c) grayscale image depicting chewing damage; (d) manipulated image showing “total” leaf area, (e) image with pathogen damage removed, (f) grayscale image showing pathogen and chewing damage, (g) example of a binary (black and white) image used to quantify leaf area metrics in ImageJ.

Canopy closure

Light environment is reported to be one of the most important abiotic factors regulating seedling survival (Corlett 2014). Canopy closure, defined as the proportion of sky obscured by vegetation as observed from a single point (Jennings 1999), was used as a proxy for shade in this study and was measured using a concave spherical densiometer.

A spherical densiometer is a handheld instrument which includes a 24-cell reflective grid which is either concave or convex in order to reflect a larger portion of the sky

(Jennings 1999). In each square the observer envisages four smaller cells and counts the number of these smaller cells filled by the reflection of the canopy above. This is then multiplied by 1.04 to obtain the percentage canopy closure above the observer

(Lemmon 1956). A bubble-level is present on the densiometer to ensure it is held horizontally when measurements are made (Jennings 1999). This method is widely used as a quick and relatively accurate way of assessing the canopy closure during forest inventories and when ground-truthing remote sensing techniques (Paletto and Tosi

2009). Canopy closure is a useful measure when considering microclimates in forest systems as it is directly related to the light regime at a specific location (Jennings 1999).

Four readings were made at the centre of each enclosure, facing north, east, south and west, and the average was then calculated to gain a single enclosure-level value.

Slope of enclosure

Slope is known to affect soil chemistry and hydrology and in turn can influence the structure and function of forest stands (Jucker et al. 2018), which can have important consequences for the composition of plant communities, with higher diversity found on slopes than on plateau sites where plants must be more resilient to drought conditions

(Daws et al. 2005, Engelbrecht et al. 2007). In edge environments, slope

58 has a significant positive effect on seedling growth and production of leaves (Howe

1990). The amount of leaf litter present also varies according to slope (Daws et al.

2005). Differences in leaf litter have been found to alter insect herbivory either by directly providing habitat for insect herbivores or by promoting predators such as spiders that can supress herbivores (Benítez-Malvido and Kossmann-Ferraz 1999).

Since forest fragments often occur on steep slopes (Fayle et al. 2015), it is important to measure slope when considering the effects of fragmentation.

The “Clinometer and bubble level” (PlaincodeTM, accuracy ±0.1°) android application was used to measure the slope of inclination for each enclosure. Three readings were taken, 50 cm apart, along the x and y axes of each enclosure (Figure 3.7). The clinometer and bubble level application measures the angle of inclination of the phone when placed on a surface. A mean angle of inclination was taken for both directions.

Photosynthetically Active Radiation

Photosynthetically Active Radiation (PAR) refers to the portion of the light spectrum

(440-700 nm) which can be utilised by plants for photosynthesis (Anderson 1964). An

Apogee MQ-301 Line Quantum sensor was used to measure PAR. This instrument is fitted with ten individual quantum sensors along a bar and has a spirit level to ensure it is held horizontally during readings. PAR measurements were taken 1 m above the ground at the centre of each enclosure (Figure 3.7). Readings were made at each visit

(six occasions), and weather conditions were recorded, measurement means were calculated across recording made in sunny conditions to obtain a single enclosure value.

Mean values were obtained from PAR measurements made on sunny days to minimise the effects of variable weather on light incidence within each enclosure.

Figure 3.7 Diagram of each enclosure showing spacing of planted seedlings and where environmental measurements were made. Canopy closure, Photosynthetically Active Radiation (PAR) and slope were measured at the centre of each enclosure. Locations of additional slope measurements are depicted by the orange diamond. 3.7 Seedling traits

Measurement of a number of seedling traits was made during the course of the experiment and on the uprooted seedlings at its termination.

Seedling height

Height has previously been found to be an important predictor of herbivory, with taller seedlings with greater numbers of leaves having higher levels of leaf area loss than shorter ones (Benítez-Malvido and Lemus-Albor 2005). Vertical seedling height was measured as the distance from the ground to the apical growing point of each seedling, to the nearest mm. During Year 1 of the experiment, however, some seedlings exhibited damage to their apex and data suggested they were decreasing in height. In Year 2 of the experiment, the stem diameter 10 cm above the ground of each seedling was also measured using digital callipers.

Number of leaves and leaf age

Certain species may produce an excess of leaves in order to satiate herbivores and retain enough undamaged leaf to survive (Aide 1992). For each seedling, the number of total leaves, new leaves and dropped leaves was recorded at each visit.

Leaf age is a strong determinant of herbivory; older leaves are better defended as they are generally tougher and have higher concentrations of chemical defences such as tannins and resins (Coley 1980, Peeters et al. 2007). Younger leaves are typically more nutritious and contain lower structural defences making them particularly palatable

(Howlett and Davidson 2001, Kursar and Coley 2003). Leaves are often categorised into general age classes such as young, intermediate and old, based on their position on a plant (Lawrence et al. 2003, Blüthgen, Metzner, et al. 2006). Species-specific preferences have been found in phasmids, for example, with some preferring young foliage and others old foliage (Blüthgen, Metzner, et al. 2006). In previous studies on one of the focal species of this thesis, D. lanceolata, no difference in the feeding behaviour of generalist phasmids was found between young and old leaves (Blüthgen and Metzner 2007).

Since seedlings could not be visited daily due to logistical constraints and to avoid interfering with herbivore feeding behaviour, the minimum number of days since the first recorded appearance of a new leaf and the final time point of the experiment was used as an estimate of leaf age. Although this is not a precise measurement it should provide a reliable proxy for leaf age because time between visits was standardised across all sites.

Leaf strength

Leaf strength was measured at the final time point in situ before seedlings were removed. One leaf from each seedling was chosen at random and the force required to puncture the leaf lamina was quantified using a pocket penetrometer fitted with a 0.25” tip. To standardise the pressure across the entire leaf blade and prevent any part of it from bending, a frame was constructed to hold each leaf whilst measurements were taken. This frame had a series of holes (2 cm in diameter) through which the penetrometer could be pushed. Five measurements were taken per leaf, and the average, expressed in kg cm-2, was used as a measure of leaf strength. The midrib was avoided on all leaves of both species, and for specimens of P. tomentella care was taken to ensure measurements were made within the intercostal area, since secondary venation is more pronounced in this species. All measurements were taken by the same person (me) to ensure they were standardised across sites.

Leaf chemistry

Chemical analysis of leaf material was carried out at the end of the experiment following the removal of surviving seedlings. In collaboration with Professor Charles S.

Vairappan’s laboratory at Universiti Malaysia Sabah in Kota Kinabalu, total phenolics and acid detergent fibre were quantified from a subsample of leaf material. Details of these analyses can be found in Chapter 5.

3.8 Research permits

Permission to work within the Kalabakan Forest Reserve was obtained from the Sabah

Forestry Department. Data for this thesis were collected under a Sabah Biodiversity

Council Access Licence (Reference number: JKM/MBS.1000-2/2 JLD.4 (56)).

S3 Supplementary material

Table S3.1 Locations and coordinates of sampling sites used for the seedling planting experiment; these include SAFE 1-ha sites in experimental fragment blocks C, D, F. Continuous forest controls were chosen in three locations; LF refers to "Logged Forest", LFE; "Logged Forest Edge" and VJR; "Virgin Jungle Reserve".

Forest Type Forest Block SAFE 2nd Order Point Enclosur Latitude Longitude e (°N) (°E) Fragment C 616 C1A 4.71022 117.62454 Fragment C 616 C1B 4.71013 117.62458 Fragment C 617 C2A 4.71104 117.62563 Fragment C 617 C2B 4.71115 117.62560 Fragment C 619 C3A 4.71382 117.62888 Fragment C 619 C3B 4.71383 117.62893 Fragment C 618 C4A 4.71278 117.62791 Fragment C 618 C4B 4.71289 117.62778 Fragment D 635 D1A 4.71683 117.58304 Fragment D 635 D1B 4.71673 117.58304 Fragment D 634 D2A 4.71539 117.58413 Fragment D 634 D2B 4.71552 117.58404 Fragment D 633 D3A 4.71338 117.58553 Fragment D 633 D3B 4.71344 117.58544 Fragment D 632 D4A 4.71212 117.58609 Fragment D 632 D4B 4.71221 117.58617 Fragment F 667 F1A 4.68914 117.54138 Fragment F 667 F1B 4.68921 117.54133 Fragment F 666 F2A 4.68953 117.54134 Fragment F 666 F2B 4.68964 117.54137 Fragment F 664 F3A 4.69474 117.54085 Fragment F 664 F3B 4.69480 117.54085 Fragment F 665 F4A 4.69343 117.54095 Fragment F 665 F4B 4.69344 117.54090 Continuous LF 697 LF1A 4.75702 117.69357 Continuous LF 697 LF1B 4.75703 117.69352 Continuous LF 694 LF2A 4.75659 117.69227 Continuous LF 694 LF2B 4.75660 117.69234 Continuous LF 680 LF3A 4.77018 117.68607 Continuous LF 680 LF3B 4.77025 117.68610 Continuous LF 684 LF4A 4.77128 117.68555 Continuous LF 684 LF4B 4.77122 117.68549 Continuous LFE 710 LFE1A 4.74333 117.58793 Continuous LFE 710 LFE1B 4.74340 117.58793 Continuous LFE 709 LFE2A 4.74215 117.58842 Continuous LFE 709 LFE2B 4.74217 117.58853 Continuous LFE 705 LFE3A 4.72800 117.59548 Continuous LFE 705 LFE3B 4.72782 117.59541 Continuous LFE 706 LFE4A 4.72891 117.59510 Continuous LFE 706 LFE4B 4.72892 117.59512 Continuous VJR 766 VJR1A 4.68422 117.53983 Continuous VJR 766 VJR1B 4.68407 117.54007 Continuous VJR 765 VJR2A 4.68545 117.53968 Continuous VJR 765 VJR2B 4.68543 117.53963 Continuous VJR 767 VJR3A 4.68124 117.53948 Continuous VJR 767 VJR3B 4.68168 117.53957 Continuous VJR 768 VJR4A 4.68052 117.53947 Continuous VJR 768 VJR4B 4.68044 117.53951

Statement of contribution to co-authored paper

Chapter 4 is a co-authored paper which has been submitted to a special issue of

Biotropica on invertebrate ecology in tropical rainforest, which is to be published in

April 2019. The bibliographic details of this co-authored paper are as follows:

Lois K. Kinneen, Tom M. Fayle, Jane L. Hardwick, Sarah C. Maunsell, Nigel E. Stork, Charles S. Vairappan & Roger L. Kitching (2019) The effects of fragmentation on insect herbivory rates of seedlings in a tropical rainforest, Biotropica, [In Review].

My contribution to this paper involved all data collection, through the establishment and execution of a seedling experiment, processing of all leaf scans and data, subsequent analysis, and writing of the draft manuscript. Jane L. Hardwick provided edits to the manuscript. Dr Tom M. Fayle, Dr Sarah C. Maunsell, Professor Nigel E. Stork and

Professor Roger L. Kitching supervised this project and assisted with edits of the manuscript. Professor Charles S. Vairappan acted as local collaborator to the project, assisted with obtaining research permits, and provided edits to this manuscript.

(Signed) ______(Co-signed) ______

Lois K. Kinneen Date:06/11/18 Prof. Nigel E. Stork Date: 15/11/18

(Co-signed) ______(Co-signed) ______

Dr Tom M. Fayle Date: 09/11/18 Prof. Roger L. Kitching Date: 05/12/18

(Co-signed) ______(Co-signed) ______

Jane L. Hardwick Date: 18/11/18 Prof. Charles S. Vairappan Date: 18/12/18

(Co-signed) ______

Dr Sarah C. Maunsell Date: 10/12/18

Chapter 4 The effects of fragmentation on insect herbivory rates of

seedlings in a tropical rainforest

Lois K. Kinneen1*†, Tom M. Fayle2,3,5, Jane L. Hardwick1, Sarah C. Maunsell1,4, Nigel

E. Stork1, Charles S. Vairappan5 & Roger L. Kitching1

4.1 Abstract

Human-modified landscapes frequently comprise remnant natural fragments within an agricultural or urban matrix. Understanding how habitat modification impacts ecological processes is essential for biodiversity conservation and development of sustainable management practices. Herbivory is a fundamental process mediating energy transfer between primary producers and higher trophic levels, but is often estimated using snapshots, without longitudinal measurements. This may not fully reveal anthropogenic impacts on the rates of this process. To determine the effects of fragmentation on insect herbivory rates, an experiment was carried out within the world’s largest tropical forest fragmentation experiment, the Stability of Altered Forest

Ecosystem (SAFE) project in Malaysian Borneo. Seedlings of two endemic species of

Dipterocarpaceae were planted (N=576) in recently experimentally fragmented forest

*Corresponding author

† E-mail: [email protected] 65 and in continuous forest sites, and repeatedly monitored. Leaf scans were made in situ to quantify leaf damage and determine rates of damage. Herbivory rates increased significantly over time, however the differences in leaf damage between forest fragments and continuous forest varied throughout the experiment. Forest type partially explained overall herbivore damage, with lower leaf damage occurring in forest fragments. Patterns of defoliation, however, were best predicted by seedling dynamics; larger seedlings and those located in plots with high survival of conspecifics experienced higher levels of foliar damage. Leaf damage of Dryobalanops lanceolata co-varied with canopy closure, supporting the hypothesis that habitat disturbance disrupts herbivory rates. Taken together, our results demonstrate the importance of longitudinal herbivory measurements for disentangling the complex manner in which forest fragmentation impacts this fundamental ecological process.

Key words: Habitat modification; insect herbivory; seedlings; tropical rainforest; Borneo;

Dipterocarpaceae

4.2 Introduction

Ecosystems around the globe, including tropical forests, are suffering biodiversity loss and disruptions to ecosystem processes (Millennium Ecosystem Assessment 2005,

Hooper et al. 2012). Identifying and understanding the ecological processes that occur within a tropical forest is particularly imperative as these ecosystems are experiencing high rates of deforestation and degradation, which will have major impacts on global biodiversity, biogeochemical cycling and climate change (Bagchi et al. 2014). Not only are the tropics particularly species-rich, they also hold a high proportion of the world’s endemic species and are thus highly important from a conservation perspective (Barlow et al. 2018). One of the most pervasive threats to biodiversity and ecosystem functioning is land-use change, which often results in fragmented landscapes (Haddad et

66 al. 2015). Investigating the effects of forest modification on ecological processes and species interactions, rather than solely focusing on changes in species distributions and abundance at single trophic levels, is potentially more useful in order to reveal the full effects of habitat conversion (Ruiz-Guerra et al. 2010, Valiente-Banuet et al. 2015). Not only can forest disturbance affect biodiversity through the loss of forest-dependent species, but it also affects the key ecological processes necessary to maintain ecosystem health (Morante-Filho et al. 2016).

Herbivory is a ubiquitous ecosystem process (Levey et al. 2016), which in tropical rainforest is primarily carried out by invertebrate species (Coley and Barone 1996). The relationship between plants and their associated insect herbivores is one of the most dominant species interactions on the planet and, together these two groups comprise over 40% of global biomass (Novotny et al. 2010, Agrawal et al. 2012). As an antagonistic ecological interaction, herbivory mediates the transfer of energy from primary productivity to higher trophic levels (Wirth et al. 2008). Through consumer regulation, herbivory can play a role in maintaining co-existing plant populations and overall floristic community composition (Rao et al. 2001, Wirth et al. 2008). Changes to herbivory levels can have cascading effects on a whole suite of ecosystem processes, such as carbon and nutrient cycling (Coley and Barone 1996, Bagchi et al. 2014,

Metcalfe et al. 2014). Disruption to herbivory has most often been investigated in relation to outbreaks of herbivores in agroecosystems, with a relative paucity of studies relating to natural levels of insect herbivory and whole ecosystem health (Maguire et al.

2015).

Previous investigations into the effects of fragmentation on this key ecosystem process have yielded variable results. Some have reported a significant increase in herbivory in forest fragments (Rao et al. 2001, Morante-Filho et al. 2016, Yeong et al. 2016), while

67 others have reported a significant decrease (Valladares et al. 2006, Fáveri et al. 2008,

Ruiz-Guerra et al. 2010, Chávez-Pesqueira et al. 2015) or no detectable changes (Souza et al. 2013, Rossetti et al. 2017). Determining the effects of fragmentation on such a complex ecological interaction is challenging as herbivory is governed by a multitude of factors which may include, or be confounded by: landscape metrics such as patch size, isolation or connectivity and spatial scale (Maguire et al. 2016). Herbivory can be affected by both spatial and temporal elements of fragmentation (Bagchi et al. 2018).

Studies carried out under controlled field experimental settings are therefore vital in order to identify the deterministic factors that affect herbivory (Levey et al. 2016).We hypothesise that if key herbivores are more susceptible than plant species to the initial effects of forest fragmentation, a reduction in herbivory rates would be detected.

Conversely if generalist herbivores are less affected by fragmentation, the resulting shift in functional composition may lead to resilience in this ecosystem process resulting in little change or even increased herbivory rates.

The primary aim of this study has been to investigate whether insect herbivory rates were affected by forest fragmentation. First, we aimed to determine whether the magnitude of insect herbivory damage, in the form of leaf area loss, differed between continuous and fragmented forest sites. Second, we aimed to explore whether the initial impacts of fragmentation were shared between the two focal species of

Dipterocarpaceae, or if different factors determined individualistic responses.

4.3 Materials and methods

Study sites and experiment design

Fieldwork was carried out within the Stability of Altered Forest Ecosystems (SAFE) project’s experimental area in the Malaysian state of Sabah, on the island of Borneo

(Ewers et al. 2011). In May 2015, two species of endemic Dipterocarpaceae:

Dryobalanops lanceolata Burck and Parashorea tomentella (Symington) Meijer were planted across a recently fragmented landscape which had been previously logged (see

Supporting Information). Seedlings were sourced from the nursery at the Sabah

Biodiversity Experiment in the Malua Forest Reserve (Hector et al. 2011). A total of

576 individuals were planted across the SAFE landscape in twenty-four sites (See

Figure 4.1). Twelve sites were situated in continuous forest control sites; four in Logged

Forest (LF), four in Logged Forest Edge (LFE) which are both located in contiguous forest over 1 million hectares in area, and four further sites were located in the Brantian-

Tatulit Virgin Jungle Reserve (VJR). The VJR at SAFE will eventually be isolated following the completion of the experimental fragmentation of the landscape and will be

2,200 ha in area. Another twelve sites were situated within newly isolated (in the six months prior to planting) 1-hectare forest fragments; four in each of the SAFE experimental blocks C, D and F.

Two enclosures were constructed at each site and twelve seedlings were planted in each

(six of each species) (Figure 4.1 insert). Seedlings were visited on six occasions over the course of the experiment: three visits were made in 2015, and a subsequent three in

2016. Visits were spaced six weeks apart during both years from May-September. This resulted in different sample sizes at each time point (Table S1) due to the production of new leaves, which were numbered and included in all subsequent measurements, the loss of abscised leaves and the death of some seedlings. Seedlings were not replaced if they died. The two enclosures were set up within ten metres of each other at the centre of each fragment. This was to ensure that all enclosures in fragmented sites were located equidistant from the forest edge, since distance from edge may alter herbivory levels in forest fragments (Levey et al. 2016).

Quantifying insect herbivory

We focus on leaf chewing as this is the most prominent form of herbivory in tropical rainforest (Coley and Barone 1996, Adams et al. 2009, Souza et al. 2013) and because the occurrence of other forms of herbivory such as leaf mining, galling and leaf rolling was low over the course of this experiment. We quantified how much leaf material was lost to insect herbivory using image-based techniques.

A flatbed Canoscan Lide 220 scanner, powered through a laptop, was used to scan every leaf on each seedling in situ with care taken to ensure no damage was made to seedlings. Preliminary image processing was performed in Adobe Photoshop CC software version 2015.0.1 20150722.r.168. This stage involved manually removing leaf petioles, seedling stems and any shadows caused by the movement of the scanner’s light source. The background of each scan was also converted to a pure white, so that any debris or soil present on scans would not interfere with area measurements. Any herbivore damage on the interior of each leaf was filled with white before images were converted to a binary, black and white, form. This image was used to measure the

“remaining” area of each leaf. When herbivore damage occurred along the edge of the lamina, margins were manually drawn based on leaf symmetry. This processed image was used to quantify the “whole” leaf area of each leaf. The software ImageJ version

1.51 was used to quantify the “whole” leaf and “remaining” leaf areas by calculating the area including and excluding holes, respectively.

The magnitude of herbivore damage, recorded as the total percentage leaf area loss for each enclosure at every sampling point, was used as our response variable. This was calculated by subtracting the area of the “remaining” leaf from the “whole” area,

70 dividing by the “whole” area and multiplying by 100 to obtain a percentage leaf area loss for each enclosure.

Explanatory variables

Forest type (continuous or fragment) and time measured as the number of days since planting were included as explanatory variables along with a number of seedling traits and environmental variables. Seedling height, measured as the distance from the ground to the apical growing point of each seedling was recorded at each visit. Both the number of new leaves and dropped leaves were also included as explanatory variables in analyses along with the total number of leaves present in each enclosure.

Photosynthetically Active Radiation (PAR) was also recorded at the centre of each enclosure using an Apogee MQ-301 Line Quantum sensor fitted with ten individual quantum sensors. Canopy closure was quantified using a concave spherical densiometer, also measured at the centre of each enclosure. The slope of the ground at each enclosure was calculated as the mean angle of inclination, measured using a clinometer, across three points measured along both the X and Y axes of each enclosure.

Data on Curculionoidea, provided by Dr Adam Sharp, were included as a predictor variable for herbivory damage as they comprise an abundant, primarily phytophagous superfamily of beetles (Gómez-Zurita and Galián 2005). Members of this group have previously been identified as important herbivores for seedlings of Dipterocarpaceae

(Eichhorn et al. 2008), and certain species have been recorded consuming leaf material of one of the focal plant species in this study, P. tomentella (Chung et al. 2013). The curculionid data were collected in June 2013 as part of the SAFE core data collection

(Sharp 2018). Both counts and species richness of Curculionoidea were calculated at forest block level (C, D, F, VJR, LF and LFE).

Statistical Analysis

Mixed effects models were used due to the hierarchical block design of the experiment and to avoid pseudoreplication associated with repeated measures (Wang and

Goonewardene 2004, Bolker et al. 2009, Benítez-Malvido et al. 2018). Forest block (C,

D, F, LF, LFE and VJR, Figure 4.1), site and enclosure ID were included as nested random effects for all models. This random effects structure controls for spatial dependence due to the hierarchical experimental design, at site and forest block-level, and allows for repeated measures on the same individuals at enclosure level. Linear mixed effects models were performed using the lme4 package (Bates et al. 2015) in R version 3.5.1 (R Core Team 2018).

Models run with untransformed leaf area loss data produced residuals which exhibited heteroscedasticity and non-normality, so all further analyses were carried out using the square root of percentage leaf area loss as the response variable. All continuous predictor variables were centred and scaled prior to analysis using the ‘scale’ function.

For each predictor value, this function subtracts the mean and divides by the standard deviation of that predictor variable (Schielzeth 2010) and ensures that each predictor is on the same scale enabling the relative importance of individual predictors to be readily compared in model outputs (Schielzeth 2010). All predictor variables were visually checked for collinearity by producing pair plots and calculating Pearson’s correlation coefficients (Zuur et al. 2009). Variance inflation factors (VIFs) supported the removal of the same predictor variables (see Supplementary material).

Once the global model structure had been chosen, a complete set of models was produced using the ‘dredge’ function in the R package “MuMIn” (Barton 2018). Since this study included a number of biologically relevant predictor variables which requires

72 multiple models to be considered, a model averaging approach was adopted (Symonds and Moussalli 2011). A subset of models was chosen based on a cut-off of ΔAICc <4, which is a commonly chosen threshold (Harrison et al. 2018). That is, all models for which the difference in AICc between that model and the best model was less than four were included in model averaging. Parameters and estimates were averaged across the resulting model set. Models were first run with both plant species analysed together and then for each species separately to investigate any individualistic responses.

Figure 4.1 Map of the study area and experimental design. A seedling experiment was established with the Stability of Altered Forest Ecosystems project in Sabah, Malaysian Borneo. Seedlings were planted in twelve one hectare fragments (Forest blocks C, D and F), and also in four continuous forest controls in each of three areas: Brantian-Tatulit Virgin Jungle Reserve (VJR), Logged Forest (LF) and Logged Forest Edge (LFE). The circular insert shows enclosure set-up; twelve seedlings were planted 50cm apart and two of these enclosure plots were established at each sampling site.

4.4 Results

Leaf damage was quantified on 3,508 unique leaves, resulting in 10,180 data points across all six sampling times. Insect herbivore damage significantly increased over time during this experiment and time remained as a predictor in the final models obtained for both focal species analysed together and when analysed separately (Table 4.1).

Figure 4.2 Time series depicting changes in herbivory rates (% leaf area loss) at enclosure level, over the course of the experiment for seedlings of Dryobalanops lanceolata. Thin lines represent the relationship observed in each of the 48 established enclosures and the mean herbivory rate for each forest type is also shown using thicker lines. Enclosures in forest fragments are depicted with dotted yellow lines and those in continuous are shown with solid green lines. Overall, lower levels of herbivory were recorded in fragmented forest sites (Figure 4.2

& Figure 4.3). While forest type was not a significant predictor of herbivory, it remained in the structure of top models obtained for both species analysed together and those of each species analysed separately. Forest type was associated with moderately high relative importance compared with other predictors (0.39 when species were analysed together, and 0.23 and 0.19 for D. lanceolata and P. tomentella, respectively,

Table 1). Despite this, differences in rate of herbivory between the two forest types was low, averaging -0.998% (ranging from differences of -0.304% to -2.001%, Table S3) for

D. lanceolata. P. tomentella experienced an even smaller mean difference across all time periods of -0.666% (variation ranged from -0.265% to 3.635% for this species,

Table S3).

Figure 4.3 Time series depicting changes in herbivory rates (% leaf area loss) at enclosure level, over the course of the experiment for seedlings of Parashorea tomentella. Thin lines represent the relationship observed in each of the 48 established enclosures and the mean herbivory rate for each forest type is also shown using thicker lines. Enclosures in forest fragments are depicted with dotted yellow lines and those in continuous are shown with solid green lines. For D. lanceolata, herbivory increased over time in a relatively consistent manner across sites within each forest type. A sharper overall increase in herbivory rates, however, was observed in forest fragments in the first 80 days of the experiment, but this pattern stabilised between day 360 and 460 (Figure 4.2). Seedlings of D. lanceolata planted in continuous forest experienced an initial sharp increase in defoliation in the first 40 days of the experiment, but herbivory rates slowed to a general increase for the remainder of the experiment (Figure 4.2).

Levels of defoliation recorded in seedlings of P. tomentella, however, showed a more variable pattern, especially within forest fragments (Figure 4.3). In the early stages of the experiment (days 0-80) herbivory rates increased faster in seedlings of this species

76 planted in forest fragments and a higher level of defoliation was recorded in fragments at final sampling event of the first year of the study. This relationship, however, was not maintained in the second year when lower herbivory levels were consistently recorded in forest fragments (days 360-40, Figure 4.3). Levels of defoliation in seedlings planted in continuous forest increased over the study period, but at a more gradual rate.

Seedling height was found to be an important predictor of herbivory in this experiment and remained in the top model structure for all three sets of analyses, indicating that larger seedlings experienced higher levels of herbivore damage (p<0.05, Table 1). This relationship was consistent between continuous forest sites and forest fragments (Figure

4.4).

Figure 4.4 Back-transformed predicted relationship between seedling height and herbivory for each enclosure following linear mixed effects modelling, for both species together (left), Dryobalanops lanceolata (middle), and Parashorea tomentella (right). Shaded areas represent the standard error surrounding estimated relationships. Herbivory in both species analysed together.

When both study species were analysed together, model subset selection resulted in five models and included seven of the original 13 predictor variables. Considering both species analysed together, a positive relationship was detected between the number of conspecific seedlings in an enclosure and herbivore damage (parameter estimate=

0.286, p<0.05, Table 1(A)). This positive relationship was likely driven by seedlings in forest fragments (Figure 4.5), as a sharper increase in herbivory damage is associated with the presence of more conspecific seedlings compared with continuous forest sites.

The number of leaves in each enclosure was also a significant predictor of herbivory damage across both species (parameter estimate -0.217, p<0.05, Table 1(A)). Overall, mixed effects modelling detected a negative relationship between the total number of leaves and herbivory. However opposite trends were observed between continuous forest and fragments. A steep decline in the amount of herbivory damage was associated with more leaves in enclosures located in continuous forest sites, yet a slight increase was observed in fragmented sites (Figure 4.6).

Figure 4.5 Back-transformed predicted relationships between the number of conspecific seedlings and herbivory for each enclosure following linear mixed effects modelling, for Dryobalanops lanceolata and Parashorea tomentella analysed together. Shaded areas represent the standard error surrounding estimated relationships. For seedlings in fragmented forest, the difference in herbivory between a plot with zero conspecifics and a plot with six surviving conspecifics is 3.50% vs. 6.25% leaf area loss.

The number of abscissed leaves and Curculionoidea abundance also remained in the top models when both species were analysed together. However, neither were significant predictors of herbivory damage and both had low relative importance values (0.20 and

0.10, respectively, Table 1(A)). These relative importance values represent the relative likelihood that the parameter would be present in a single top model structure (Symonds and Moussalli 2011). Nevertheless, these results suggest that high herbivory damage was associated with higher levels of leaf fall (Figure S4.1). A negative relationship was detected between counts of Curculionoidea and herbivory, suggesting this group may not be the primary herbivores for our study species. Opposite trends, however, were observed in continuous forest and fragments, suggesting seedlings from continuous forest were driving the overall observed pattern. In fragments, a positive relationship was observed between the number of Curculionoidea and herbivory damage, suggesting these phytophagous beetles may play a more active role in herbivory in forest fragments

(Figure S4.4).

Figure 4.6 Back-transformed predicted relationship between the number of leaves and herbivory for each enclosure following linear mixed effects modelling, for both species together (left), Dryobalanops lanceolata (middle), and Parashorea tomentella (right). Shaded areas represent the standard error surrounding estimated relationships. The number of leaves was a significant predictor and remained in the top models for when both species were analysed together, and for D. lanceolata.

Table 4.1 Results of mixed effects modelling for total herbivory per enclosure averaged across best-fitting models, for (A) both species analysed together, (B) Dryobalanops lanceolata and (C) Parashorea tomentella. β represent model-averaged coefficients, standard errors (SE), 95% confidence intervals (CIs) are also presented. Significant predictors are shown in bold. Importanc Predictor Β SE Z Value Lower CI Upper CI P e

Both Species

Intercept 1.281 0.165 7.764 0.958 1.61 < 2e-16 -

Seedling Height 0.423 0.045 9.578 0.337 0.510 < 2e-16 1.00

Time 0.246 0.028 8.819 0.191 0.300 < 2e-16 1.00

Total Leaves -0.217 0.053 4.148 -0.320 -0.115 3.36e-05 1.00

Number of Seedlings 0.286 0.072 3.977 0.145 0.426 7.00e-05 1.00

Forest type: Fragment -0.134 0.216 0.623 -0.770 0.083 0.533 0.39

Dropped Leaves 0.015 0.033 0.658 0.005 0.140 0.658 0.20

Invert. Abundance -0.018 0.062 0.290 -0.367 -0.005 0.772 0.10

Dryobalanops lanceolata Intercept 1.685 0.155 10.866 1.381 1.989 < 2e-16 -

Total Leaves -0.253 0.066 3.858 -0.381 -0.125 0.0001 1.00

Seedling Height 0.334 0.066 5.052 0.204 0.464 4e-07 1.00

Time 0.086 0.066 1.305 0.036 0.206 0.192 0.71

Canopy 0.136 0.118 1.153 0.060 0.360 0.249 0.65

Forest type: Fragment -0.061 0.175 0.348 -0.822 0.282 0.728 0.23

Slope -0.011 0.041 0.266 -0.262 0.026 0.791 0.09

Invert. Abundance -0.013 0.060 0.210 -0.458 0.117 0.834 0.07

Parashorea tomentella

Intercept 2.033 0.153 13.319 1.734 2.332 <2e-16 -

Seedling Height 0.724 0.07 10.755 0.592 0.856 <2e-16 1.00

Time 0.398 0.052 7.609 0.295 0.500 <2e-16 1.00

Forest type: Fragment -0.049 0.153 0.320 -0.770 0.261 0.749 0.19

Slope -0.016 0.055 0.285 -0.330 0.058 0.776 0.12

Dropped Leaves 0.008 0.029 0.264 -0.018 0.187 0.792 0.09

New Leaves 0.006 0.028 0.229 -0.032 0.202 0.819 0.07

Species-specific predictors of herbivory.

Analysis of each species separately resulted in a total of 11 models with seven of the original 11 predictors remaining in the subset of models predicting herbivory damage in seedlings of D. lanceolata, and five unique models with six predictors remaining for seedlings of P. tomentella.

Slope within enclosures remained as a predictor variable in the subsets of models obtained for both species when they were analysed individually and was negatively associated with herbivory (Table 1(B) &(C), Figure S4.5). The relative importance value for this predictor, however, was comparatively low, 0.09 and 0.12 for D. lanceolata and P. tomentella, respectively (Table 1(B) & (C)).

For seedlings of D. lanceolata, a significant negative relationship between the number of leaves and insect herbivory damage was detected at enclosure level (parameter estimate= -0.253, p=0.0001, Table 1(B)). Higher insect herbivory damage was recorded in seedlings of D. lanceolata with fewer leaves (Figure 4.6). Canopy closure, while not significant, was associated with a reasonably high importance value in the top models for this species (importance value= 0.65, Table 1(B)), indicating a high probability that this predictor would be present in a single top model. Higher levels of herbivory were associated with sites with more closed canopies (Figure S4.2).

Counts of Curculionoidea remained in the subset of models for seedlings of D. lanceolata, although this predictor was also associated with a lower relative importance value (importance value= 0.07, Table 1(B)). The overall parameter estimate for this predictor was negative, indicating that higher levels of herbivory damage are associated with lower counts of Curculionoidea, and this relationship was apparent in seedlings

81 within continuous forest. The opposite trend, however, was observed in forest fragments

(Figure S4.4).

The number of dropped leaves also remained in the structure of the top models for P. tomentella, but was absent from those of D. lanceolata. A higher level of herbivory was recorded in enclosures with greater numbers of dropped leaves for seedlings of P. tomentella (parameter estimate= 0.008, Table 1(C), Figure S4.1). This predictor had a low relative importance of 0.09 (Table 1(C)). The number of new leaves also remained as a predictor for herbivory damage in P. tomentella, with a lower parameter estimate

(0.006, Table 1(C)) and importance value (0.07, Table 1(C)), suggesting that producing new leaves is associated with lower overall herbivory damage in seedlings of P. tomentella (Figure 4.3).

4.5 Discussion

The effects of forest fragmentation on insect herbivory rates

The primary aim of this study has been to ascertain whether insect herbivory occurred at different rates in forest fragments compared with continuous forest sites. We discovered that while forest type alone did not fully explain the patterns of herbivory in experimentally planted seedlings across the SAFE landscape, it did remain as a relatively important predictor in explaining patterns of leaf damage. In general, lower levels of herbivory damage were recorded in seedlings planted in forest fragments compared with those planted in continuous forest sites. This is consistent with other studies that found reduced herbivory in forest fragments (Benítez-Malvido 1998, Fáveri et al. 2008, Chávez-Pesqueira et al. 2015). Forest type had a lower relative importance than other predictors following model averaging which suggests that insect herbivory is governed by a complex combination of biotic and abiotic factors. This is unsurprising,

82 as it has been suggested that levels of leaf damage may not be predicted by simple landscape metrics such as forest area (Maguire et al. 2016).

While reduced herbivory was consistently recorded in fragments for one study species,

D. lanceolata, a more complex relationship was detected in seedlings of P. tomentella, which exhibited more variation in their response to fragmentation indicating that species-level responses differ. The largest difference in the level of leaf area loss between fragments and continuous forest was observed at the third sampling time for seedlings of this species (a mean increase in leaf area loss of 3.635%, Table S3). This was also associated with an inverted relationship, whereby higher levels of defoliation were recorded in enclosures located in forest fragments compared with those in continuous forest sites; the only time that this was observed in this experiment. This peak in leaf area loss may reflect an outbreak in insect herbivores in fragmented forest sites.

Furthermore, this gives weight to the argument that “snapshots” of herbivory at single points in time cannot be relied upon (Lowman 1984, de la Cruz and Dirzo 1987, Anstett et al. 2016), since data from this sampling event would erroneously lead to the conclusion that herbivory occurs at higher rates in fragments for seedlings of P. tomentella. Quantifying herbivory through repeated measures, we argue, provides a more accurate depiction of this complex ecological process. Damage levels obtained at a single point in time have been found to underestimate herbivory levels by up to five times (Lowman 1984). Despite this, studies continue to rely heavily on snapshots of leaf damage (Ruiz-Guerra et al. 2010).

Our results were in contrast to findings from another recent study carried out in Sabah, which also investigated the effects of fragmentation on insect herbivory in three species

83 of dipterocarp seedlings, one of which was D. lanceolata, and which found that herbivory damage was negatively related to forest area (Yeong et al. 2016). However, our study differed in methodology, Yeong et al. measuring herbivory damage, by eye, once after an 18 month period. Results from the Biological Dynamics of Forest

Fragments Project (BDFFP) in Brazil have shown that despite sharing similarities in soil types, climate, fragment ages, land-use history and even being geographically close, fragmented landscapes can experience vastly different trajectories in species responses and changes in species interactions (Laurance et al. 2011). Our findings parallel a similar seedling experiment, carried out at the BDFF project which, after one year, recorded lower herbivory damage on seedlings planted in forest fragments (Benítez-

Malvido 1998). However, after six years a consistent significant response to fragmentation could not be detected, as leaf damage did not appear to relate reliably with forest area (Benítez-Malvido et al. 1999).

We found a relatively small difference in the magnitude of herbivory damage occurring between fragments and continuous forest (mean difference of 0.998% and 0.666% for

D. lanceolata and P. tomentella, respectively, Table S3). Given the mean herbivory damage across all time points was 3.989% for D. lanceolata and 5.931% for P. tomentella in continuous forest sites, this translates to a 25% and 11% reduction in their herbivory in fragmented forest, respectively. The consequences of even small shifts in this process are not fully understood. Little research has been carried out into tipping points and determining to what degree differences in herbivory levels have broader ecological consequences such as increased mortality. This could be investigated through experiments where different levels of herbivory are experimentally induced, and survival is monitored over the long term. Previous research has found that as little as

10% defoliation can reduce plant fitness (Coley and Barone 1996) and we recorded

84 higher leaf area loss than this threshold in a number of enclosures during this experiment, which may have consequences for future survival rates of seedlings across the entire SAFE landscape.

Seedling traits as predictors of insect herbivory

Seedling traits played an important role in predicting herbivory damage throughout this experiment. A clear result from our study showed that seedling height was the single best predictor and a strong determinant of leaf damage. As all seedlings planted in this experiment were of the same age and had been grown under the same environmental conditions at the Sabah Biodiversity Experiment’s nursery, it is unlikely that seedling age is a confounding factor in this result. Our findings seem logical, as larger plants are more conspicuous, and therefore may be more readily accessible to natural enemies such as herbivores (Schlinkert et al. 2015). This is in line with previous research which found that seedling height predicted levels of leaf damage (Benítez-Malvido and

Lemus-Albor 2005). Interestingly, previous research on members of the

Dipterocarpaceae found that an increase in herbivory was significantly associated with seedling height, only when one of our study species, D. lanceolata, was removed from analysis (Massey et al. 2006). Our findings, however, confirm that levels of defoliation in seedlings of this species can be predicted by height.

When both species were analysed together, a significant positive relationship was detected between the numbers of conspecific seedlings and total herbivory in each enclosure, indicating that there is no safety in numbers when it comes to levels of defoliation. This gives support to the Janzen-Connell hypothesis of negative density dependence, which predicts that seedlings found in higher concentrations of conspecifics act as reservoirs for natural enemies such as insect herbivores (Massey et

85 al. 2006). The fact that this effect operated even in fragmented forest is encouraging in that it indicates that diversity maintenance via a Janzen-Connell effect may continue to operate even in areas affected by anthropogenic disturbance.

Conversely, a negative relationship was found between the number of leaves within each enclosure and insect herbivory. Since young expanding leaves are most vulnerable to herbivory damage (Kursar and Coley 2003), plants may produce leaves when biotic pressure from herbivores is low, such as during the dry season, thus taking advantage of reduced herbivore pressure minimising the area of young leaf tissue lost (Aide 1992).

Leaf synchrony whereby different plants produce leaves at the same time, is another established defence against herbivory as herbivores are satiated more easily (Coley and

Kursor 1996). An important consideration, however, is that lags in leaf production and herbivory have been recorded, with herbivory closely following increased production of leaves (Lamarre et al. 2014). Thus, while a higher number of leaves may be associated with lower levels of defoliation, it may not reflect future herbivory damage. One possibility is that in the absence of biotic pressure from herbivores, seedlings were able to invest their resources in producing leaves.

Species-specific responses to forest fragmentation

A secondary aim of this study was to determine if the two focal plant species responded in a similar way to forest fragmentation, and to determine if their responses were governed by the same predictors. When analysed together, species identity did not remain as an important predictor of levels of defoliation indicating that herbivory did not vary significantly between the two focal species. While time, seedling height, forest type and slope remained in model structures when species were analysed separately, a

86 number of other predictors contributed to explaining patterns of defoliation in each of the two species.

For seedlings of D. lanceolata, canopy closure played a role in predicting herbivory damage suggesting that light might be an important factor governing herbivory levels in this species. In our study, lower leaf area loss was recorded in sites with more open canopies. This conflicts with the resource availability hypothesis which states that plants growing in environments where light is restricted will invest in defensive compounds to mitigate loss of valuable photosynthetic material (Coley et al. 1985). It also contrasts with previous reports that insect herbivore activity is reduced in closed canopy environments compared with forest gaps, with more open environments experiencing up to 50% more herbivory in tropical systems (Piper et al. 2018). Edge habitats, which are characterised by higher levels of light (Bierregaard et al. 1992,

Newbery et al. 1996) have also been associated with higher levels of herbivory performed by generalist species (Murphy et al. 2016). We investigated palatability and defence in the surviving seedlings in this experiment elsewhere, and found evidence that canopy closure predicted leaf strength in our study species (Chapter 5). It is possible that the higher tensile strength found in leaves under more open canopies explains this surprising pattern of defoliation.

An experiment testing the interaction between light and experimentally induced herbivory in seedlings of Dipterocarpaceae found that seedlings located in areas with more closed canopies were less capable of enduring herbivory damage than those in more open environments (Blundell and Peart 2001). This was further supported by research which found higher mortality with increasing herbivory in the understory and that herbivory limited growth did not affect survival in seedlings found in light gaps

(Norghauer et al. 2008). Since higher levels of leaf area loss were recorded in more

87 shaded environments in this study, this may have consequences for the survival of seedlings of D. lanceolata in closed canopy environments.

In conclusion, our study contributes to understanding the effects of forest fragmentation on insect herbivory which, nevertheless, remains complex and unpredictable. We provide evidence that this vital ecological process may be disrupted through this form of environmental degradation by demonstrating that forest fragments are associated with lower levels of insect herbivore damage. Interestingly, we found that larger seedlings found in areas with high numbers of conspecifics suffered increased leaf damage, which may have consequences for their survival. We have shown that herbivory is a highly dynamic process governed by a multiplicity of factors and, therefore, should be monitored over long periods of time if the effects of habitat modification are to be disentangled.

Acknowledgements

This research was funded by an Australian Research Council Discovery Grant

DP160102078 and LKK received additional support from a Griffith University

Postgraduate Research Scholarship. TMF was also funded by a European Research

Council advanced grant (669609). For permission to carry out this research, we thank the Sabah Biodiversity Council (Access Licence: JKM/MBS.1000-2/2 JLD.4 (56)) and

Sabah Forestry Department and the SAFE project. We give special thanks to all of the

SAFE research assistants for their help in the field and to Ryan Gray for field assistance and logistical support. We extend our thanks to Philip Ulok and Dr Glen Reynolds for their assistance in sourcing the seedlings used in this experiment and Emma Window who assisted with processing of some leaf scans.

S4 Supplementary material

Forest structure at SAFE

The native vegetation at SAFE is lowland tropical rain forest dominated by members of the Dipterocarpaceae (Döbert et al. 2017). The forest at SAFE, however, is now extremely heterogeneous and heavily degraded in places due to its complicated logging history (Pfeifer et al. 2015). The forest in “Logged Forest Edge (LFE)” and “Logged

Forest (LF)” are known to have experienced two rounds of selective logging, carried out under a modified Malaysian uniform system, the first removing approximately

113m3ha-1, and the second 37m3ha-1 (Struebig et al. 2013). While strictly protected, the

Virgin Jungle Reserve at SAFE has experienced some logging on its periphery, carried out during the construction of access roads (Struebig et al. 2013). There is also evidence that further illegal logging may have taken place at VJR (Pfeifer et al. 2015). Within

SAFE’s experimental area, the fragmented forest blocks (A-F) underwent a similar extraction of 113m3ha-1 during its first round of logging in the 1970s, but during the second rotation the forest within these blocks were logged three times resulting in an average extraction rate of 66m3ha-1 (Struebig et al. 2013). The matrix surrounding the fragments used in this experiment was salvage logged, whereby trees of any size could be removed (Ewers et al. 2015), earlier in 2015, and was primarily composed of regenerating secondary species at the time of this study.

Statistical analysis (continued)

Prior to analysis, all predictor variables were checked for collinearity visually, by producing pair plots and calculating Pearson’s correlation coefficients (Zuur et al.

2009). A cut-off of R ≤ 0.7 was chosen as an acceptable amount of correlation between two variables above which the variable with the weakest correlation with the response

89 variable, leaf area loss, would be removed. For both plant species analysed together, abundance and species richness of Curculionoidea showed high correlation (R= 0.97).

Species richness showed a weaker association with the response variable and so was removed from the global model structure (R= -0.01). For D. lanceolata, a high correlation was detected between the total number of leaves and the total number of seedlings per enclosure (R= 0.76). The total number of seedlings showed a weaker association with the response variable and was excluded from the global model for this species (R= 0.1). For P. tomentella, the total number of leaves and total number of seedlings were also highly correlated (R= 0.8), and once more the number of seedlings showed a weaker correlation with leaf area loss and so was excluded from analysis (R=

0.07).

Variance inflation factors (VIFs) were also calculated for all explanatory variables and used as a complementary method of detecting multi-collinearity (Zuur et al. 2010). A

VIF of 5 was chosen as a cut-off, above which variables would be removed from global model structures, a relatively conservative cut-off within common thresholds of 3-10

(Craney and Surles 2002, Zuur et al. 2010). Generally VIFs above 10 are considered to indicate that a predictor is very strongly correlated with one or more other predictors in a model (Smith et al. 2009). Both species richness and counts of Curculionoidea were associated with high VIFs in all sets of analyses (Table S2). Once species richness had been removed, due to weaker correlations with the response variables, counts of

Curculionoidea exhibited acceptable VIFs. VIFs also confirmed that the total number of seedlings should be removed from the global models for both D. lanceolata and P. tomentella (Table S2; Total Seedlings VIF=6.11 and 5.74, respectively).

Table S4.1 Number of surviving seedlings and the number of leaves at each time point during the experiment.

Number of Seedlings Number of Leaves

Time point Fragment Continuous Total Fragment Continuous Total

0 288 288 576 1214 1189 2403

1 250 208 458 1058 718 1776

2 228 195 423 945 656 1601

3 161 137 298 834 631 1465

4 154 128 282 819 632 1451

5 148 121 269 863 621 1484

Table S4.2 Variance inflation factors (VIFs) for maximal models of both species analysed together, and each species analysed separately. A VIF of 4 was chosen as a cut-off above which a variable would be removed from maximal model structure, variables removed from maximal model structure are highlighted in bold.

Both species Dryobalanops lanceolata Parashorea tomentella Variable VIF Variable VIF Variable VIF Forest type 1.618 Forest type 1.805 Forest type 1.503 Time 2.169 Time 3.648 Time 2.919 Slope X 1.356 Slope X 1.457 Slope X 1.314 Slope Y 1.305 Slope Y 1.327 Slope Y 1.300 Canopy 1.511 Canopy 1.620 Canopy 1.459 PAR 1.534 PAR 1.630 PAR 1.509 Total Leaves 3.322 Total Leaves 3.828 Total Leaves 3.575 New Leaf 1.688 New Leaf 1.977 New Leaf 1.796 Leaf Drop 1.514 Leaf Drop 1.686 Leaf Drop 1.407 Total Seedlings 3.908 Total Seedlings 6.408 Total Seedlings 5.863 Seedling Height 1.519 Seedling Height 1.404 Seedling Height 1.543 Invert Abundance 1.600 Invert Abundance 1.571 Invert Abundance 1.534 Invert Richness 17.622 Invert Richness 18.746 Invert Richness 17.046 Species 2.056

Table S4.3 Summary data showing mean herbivory (% leaf area loss) at enclosure level ± standard error in each forest type for each species across all time points. The difference in leaf area loss between continuous forest and fragments are provided for each time point.

Sampling Point Forest Species Mean Type 0 1 2 3 4 5

Fragment 0.936 2.367 3.150 3.910 3.919 3.669 2.990 ±0.171 ±0.372 ±0.637 ±0.698 ±0.740 ±0.548 ±0.243 Dryobalanops 3.989 Continuous 1.240 3.886 4.186 4.305 4.644 5.670 lanceolata ±0.310 ±0.262 ±0.821 ±0.769 ±0.770 ±0.751 ±0.795 6

Difference -0.304 -1.519 -1.035 -0.395 -0.725 -2.001 -0.998

Fragment 1.550 4.079 9.412 4.904 5.286 6.533 5.265 ±0.336 ±0.852 ±1.963 ±1.017 ±1.005 ±1.323 ±0.428 Parashorea tomentella Continuous 2.138 4.344 5.777 7.198 7.901 8.227 5.931 ±0.365 ±0.727 ±1.003 ±1.166 ±1.038 ±1.234 ±0.428 Difference -0.588 -0.265 3.635 -2.295 -2.615 -1.694 -0.666

Figure S4.1 Back-transformed predicted relationship between the number of dropped leaves and herbivory (%) in each enclosure following linear mixed effects modelling, for both species together (left), Dryobalanops lanceolata (middle), and Parashorea tomentella (right). Shaded areas represent the standard error surrounding estimated relationships.

Figure S4.2 Back-transformed predicted relationship between canopy closure (%) and herbivory (%), following linear mixed effects modelling for Dryobalanops lanceolata. Shaded areas represent the standard error surrounding estimated relationships.

Figure S4.3 Back-transformed predicted relationship between the number of new leaves and herbivory (%) following linear mixed effects modelling for Parashorea tomentella. Shaded areas represent the standard error surrounding estimated relationships.

Figure S4.4 Back-transformed predicted relationship between counts of Curculionoidea and herbivory (%) following linear mixed effects modelling, for both species together (left), Dryobalanops lanceolata (middle), and Parashorea tomentella (right). Shaded areas represent the standard error surrounding estimated relationships. The left panel shows the predicted relationships without data overlaid; the right shows mean values at each enclosure.

Figure S4.5 Back-transformed predicted relationship between slope (angle of inclination within each enclosure) and herbivory (%) following linear mixed effects modelling, for both species together (left), Dryobalanops lanceolata (middle), and Parashorea tomentella (right). Shaded areas represent the standard error surrounding estimated relationships.

Figure S4.6 Pairplots of the response variable, leaf area loss (Herbivory) and explanatory variables for both species together. The upper panel contains scatterplots between each explanatory variable. Pearson’s correlation coefficients are provided in the lower panel, the font size of correlation coefficients is proportional to their value.

Figure S4.7 Pairplots of explanatory variables for Dryobalanops lanceolata data versus the response variable leaf area loss (Herbivory). The upper panel contains scatterplots between each explanatory variable. Pearson’s correlation coefficients are provided in the lower panel, the font size of correlation coefficients is proportional to their value.

Figure S4.8 Pairplots of explanatory variables for Parashorea tomentella data versus the response variable leaf area loss (Herbivory). The upper panel contains scatterplots between each explanatory variable. Pearson’s correlation coefficients are provided in the lower panel, the font size of correlation coefficients is proportional to their value.

Statement of contribution to co-authored paper

Chapter 5 is a co-authored paper which has been prepared for submission to Oecologia.

The bibliographic details of this co-authored paper are as follows:

Lois K. Kinneen, Junia Anilik, Jane L. Hardwick, Roger L. Kitching, Sarah C.

Maunsell, Kishneth Palaniveloo, Nigel E. Stork, Mathavan Vickneswaran, & Charles S.

Vairappan (prepared manuscript) Palatability of two species of Dipterocarpaceae experimentally planted in a fragmented tropical landscape

My contribution to this paper includes data collection, through the establishment and execution of a seedling experiment, laboratory work, all data processing and analysis, and writing of the manuscript. Junia Anilik and Mathavan Vickneswaran assisted with laboratory work. Jane L. Hardwick provided edits to the manuscript and field assistance.

Dr Kishneth Palaniveloo provided laboratory training, assisted with laboratory work and provided edits of the manuscript. Dr Sarah C. Maunsell, Professor Nigel E. Stork and

Universiti Malaysia Sabah and provided edits to this manuscript.

(Signed) ______(Co-signed) ______

Lois K. Kinneen Date: 25/07/18 Dr Sarah C. Maunsell Date: 05/09/18 (Co-signed) ______(Co-signed)______Junia Anilik Date: 09/08/18 Dr Kishneth Palaniveloo Date: 13/08/18 (Co-signed)______(Co-signed) ______Jane L. Hardwick Date: 21/08/18 Prof. Nigel E. Stork Date: 05/12/18 (Co-signed) ______(Co-signed)______Prof. Roger L. Kitching Date: 5/12/18 Mathavan Vickneswaran Date: 06/08/18

(Co-signed)______Prof. Charles Vairappan Date: 3/08/18

Chapter 5 Palatability and defence in two species of Dipterocarpaceae

planted across a fragmented tropical landscape

Lois K. Kinneen1, Junia Anilik2, Jane L. Hardwick1, Roger L. Kitching1, Sarah C.

Maunsell1,3, Kishneth Palaniveloo2,4, Nigel E. Stork1, Mathavan Vickneswaran2, &

Charles S. Vairappan2

1Environmental Futures Research Institute & School of Environment and Science, Griffith University, Nathan, QLD 4111, Australia 2Institute of Tropical Biology and Conservation, Universiti Malaysia Sabah, Kota Kinabalu, Sabah, Malaysia 3Museum of Comparative Zoology, Harvard University, Cambridge, MA 02138, USA 4Institute of Ocean and Earth Sciences, Institute for Graduate Studies Building, University of Malaya, 50603 Wilayah Persekutuan Kuala Lumpur, Malaysia

5.1 Abstract

Forest loss and fragmentation disrupt vital plant-animal interactions. Specifically, forest disturbance may impact on how well plants defend themselves against insect herbivory.

Faced with the challenge of keeping their leaves visible to optimise photosynthesis whilst concurrently minimising damage from natural enemies such as insect herbivores, plants employ two main lines of adaptive defence in response to their environment: antifeedant chemical and physical traits.

This study investigates the effects of fragmentation on bottom-up control of insect herbivory through changes in foliar defensive traits. Seedlings of two endemic

Dipterocarpaceae species, Dryobalanops lanceolata and Parashorea tomentella, were planted in twelve one hectare fragments and twelve continuous forest sites across a

100 fragmented landscape in Sabah, Malaysia. Three common indicators of leaf quality were quantified: total foliar phenolics, acid detergent fibre and leaf strength.

Total phenolics and fibre content varied significantly between species, with P. tomentella having higher polyphenols but lower fibre content, suggesting this species may have a greater reliance on chemical rather than mechanical defence compared to D. lanceolata, thus highlighting apparent species-specific defensive strategies. Lower foliar phenolics were recorded in seedlings in forest fragments although this relationship was only significant for individuals of P. tomentella. Further evidence suggests a reduction in acid detergent fibre present in fragments. These results suggest a relaxation in some phytochemical defences in forest fragments, and imply an increased vulnerability to insect herbivory for seedlings in fragmented habitats. Such changes may have cascading consequences at population and community levels with potential implications for regeneration success within forest fragments.

Key words: total phenolics; acid detergent fibre; leaf defence; insect herbivory; fragmentation

5.2 Introduction

Plants generally, and their leaves in particular, are fundamental to biogeochemical cycles, form the base of food chains and provide habitat for a suite of lifeforms (Wright et al. 2004). Leaves are vital plant organs necessary for photosynthesis and while maximising light interception are faced with the challenge of minimising damage from natural enemies such as herbivores (Corlett 2014). Often described as an “evolutionary arms race”, this antagonistic interaction between plants and herbivores can cause strong selective pressure favouring defensive traits in some plant species, and has been proposed as a mechanism to explain the co-diversification of plants and insects (Feeny

101

1976, Kursar et al. 2009). Plants employ a range of secondary chemical compounds to deter insect herbivory and, similarly, insect herbivores have evolved a variety of strategies to evade predation through the use of different secondary metabolites leading to specialist plant-insect herbivore relationships (Feeny 1976, Janzen 1984, Kessler and

Baldwin 2002).

Phytophagous insects, in parallel many other taxa, increase in both diversity and abundance towards the tropics (Salazar and Marquis 2012). Within tropical ecosystems they play a vital role as consumers, since the majority of herbivore damage is carried out by invertebrates (Coley and Barone 1996). Although plant species in the tropics experience a higher degree of herbivore pressure, this rarely corresponds with increased levels of defoliation suggesting that plants may be better defended at lower latitudes

(Salazar and Marquis 2012).

Plants exhibit two main lines of defence against herbivory; chemical and physical. A number of hypotheses have been proposed to explain the variation in defensive traits among plant species. The theory of plant apparency suggests plants which are more easily detected by herbivores, due to higher densities, slower growth rates or high leaf longevity, should rely more on chemical defences to act as deterrent to a broad range of enemies (Feeny 1976). The resource availability hypothesis proposes that environmental constraints which cause reduced growth rates favour larger investments in defensive traits (Coley et al. 1985). The carbon-nutrient balance hypothesis suggests that when plant growth is limited by nutrients such as nitrogen, rather than by light, plants will invest their carbon, produced through photosynthesis, in carbon-based defensive compounds such as polyphenols. This mitigates leaf loss as their replacement is expensive (Bryant et al. 1983).

102

Leaf traits are regularly used to explain patterns of herbivore abundance and their associated levels of damage (Whitfeld et al. 2012). Kursar and Coley (2003) proposed that plant species can be placed along a spectrum between two main strategies to explain insect herbivory patterns; defence or escape. The defence strategy is characterised by strong chemical defensive traits, and, typically, slow growth and slow expansion of leaves. Conversely, species which rely on the escape strategy do not invest in chemical defence in their younger leaves but instead delay greening of their leaves and adopt a faster rate of leaf expansion, thus minimising the amount of leaf material that can be lost to herbivores during the young and more vulnerable phase of leaf development.

Due to their sessile nature, it is beneficial for plants to exhibit plasticity in their leaf traits in response to changes in their environment so as to maintain functioning regardless of abiotic stresses (Karban et al. 1999, Rozendaal et al. 2006). Adaptations to leaf traits have been studied when comparing sun and shade leaves within the same individual and between different life-history strategies such as pioneer versus shade- tolerant species (Onoda et al. 2011, Ruiz-Guerra et al. 2016). It follows that species are expected to show phenotypic plasticity at a larger scale in response to habitat modification, although this is relatively understudied. In dry forest of the subtropics, edge effects have been found to alter leaf quality significantly with a higher carbon and nitrogen content found in leaf litter of edge habitats, which in turn disrupts decomposition rates (Moreno et al. 2017). Previous investigations into the effects of forest fragmentation on defensive leaf traits have yielded inconclusive results. While there is some evidence that fragmentation may lead to the relaxation of certain defensive traits, others have failed to detect a significant response (Fáveri et al. 2008,

Ruiz-Guerra et al. 2016).

103

A wealth of defensive traits affect herbivory (Coley 1983), but some are more readily quantifiable than others (Coley 1983). Foliar phenolics are toxic to most insect herbivores and as such, act as a deterrent to generalist species. They act by reducing the palatability of leaf tissue by decreasing the efficacy of the herbivore’s digestive enzymes (Waterman and Mole 1994). Some specialist herbivores invest in their own secondary compounds to cope with these toxins but they are commonly associated with reduced growth and overall fitness (Kessler and Baldwin 2002). Fibre and leaf strength, which generally correlate with one another, have been shown to predict herbivore damage in tropical forests (Coley 1983). It is important to note, however, the difference between the two defensive traits: fibre is a chemical component of the leaf which is harder to digest, whereas leaf strength and toughness refer to structural properties of the leaf which act as a barrier to herbivory before chemical defences can be encountered

(Sanson et al. 2001, Mali and Borges 2003, Caldwell et al. 2016). Toughness can affect the behaviour of insect herbivores directly as chewing insects must have mouthparts strong enough to fracture and tear a leaf’s surface (Sanson et al. 2001). It requires more energy to consume tough leaves and so toughness has been found to play a role in food preferences of herbivores (Choong et al. 1992, Caldwell et al. 2016).

Microclimate, specifically the light environment, is known to alter leaf traits and consequently affects insect herbivores through such changes to their host plant (Stiegel et al. 2017). Light plays a role in determining how much investment is made in foliar defensive traits, for example increased carbon-based defensive chemicals such as phenolics are found in more open environments (Ingersoll et al. 2010, Stiegel et al.

2017). Some phenolic compounds may even protect plants from damage caused by high light incidence (Close and McArthur 2002). In addition, higher fibre content and leaf strength have been found in shade leaves compared to sun leaves (Onoda et al.

104

2011). Leaf lifetime is an important factor to consider, as it is thought to determine what type of defence will be more advantageous: polyphenols and fibre are relatively expensive to produce and are therefore beneficial only if a leaf is long-lived (Coley et al. 1985).

Here we investigate whether forest fragmentation causes a shift in three foliar defensive traits: total phenolics, acid detergent fibre and leaf strength, and whether these in turn are associated with altered insect herbivory levels through bottom-up control. We hypothesise that more disturbed, and therefore more open, sites should be characterised by leaves with lower toughness and fibre content, but higher phenolic content.

Furthermore, we hypothesise that the two study species may exhibit differences in their foliar defensive traits and therefore responses to forest disturbance may be species- specific.

5.3 Materials and methods

Study site

Field work was carried out at the Stability of Altered Forest Ecosystems (SAFE) project in Sabah, Malaysian Borneo, a well replicated, large-scale manipulation experiment designed to test the effects of anthropogenic disturbance such as logging and fragmentation on a suite of forest functions. This heterogeneous landscape comprises a

7200 hectare (ha) experimental area adjacent to the 2200 ha Brantian-Tatulit Virgin

Jungle Reserve, within over one million hectares of continuous forest and agricultural plantation, predominantly oil palm (Ewers et al. 2011). The forest at SAFE is primarily composed of highly degraded lowland and hill dipterocarp forest which has experienced multiple logging regimes (Deere et al. 2018).

105

Figure 5.1 Map of study area. A seedling experiment was established within the Stability of Altered Forest Ecosystems (SAFE) project, located in Sabah Malaysian Borneo. One hectare fragments in ‘Forest blocks’ “C, D and F” were chosen for this experiment. Continuous forest control sites were chosen in Logged Forest Edge (LFE), Logged Forest (LF) and Virgin Jungle Reserve (VJR).

106

Leaf material

Leaf material was obtained from seedlings which were destructively sampled following the completion of an experiment to test the initial effects of forest fragmentation on rates of damage by insect herbivores which ran from May 2015 until October 2016.

Two species of Dipterocarpaceae were planted in this experiment; Dryobalanops lanceolata Burck and Parashorea tomentella (Symington) Meijer, both of which are endemic to Borneo and common in lowland dipterocarp rainforest (Baltzer et al. 2005,

Brearley 2006, Brearley, Press and Scholes 2003, Chung et al. 2011). Twelve seedlings of each species were initially planted and enclosed using chicken wire in twelve one hectare forest fragments and twelve continuous forest sites, resulting in 576 seedlings being planted across the landscape (Figure 5.1). Seedlings were planted 50 cm apart and the minimum distance between sites was 178 m. Further details on the experimental design and methods used over the course of the experiment can be found in Chapter 3 of this thesis.

At the end of the experiment, all 268 surviving seedlings were removed and their leaves were dried at 70°C for 72 hours before being transported to Universiti Malaysia Sabah, for chemical analysis. Since each analysis required a minimum amount of leaf material, leaves were weighed accurately and were randomly assigned to either phenolic or fibre analysis.

Defensive leaf traits

Total phenolic content

The phenolic content of both damaged and undamaged leaves was measured from a subsample of surviving seedlings (N=100 seedlings, approximately four from each site).

107

Leaf samples were weighed and ground using a mortar and pestle before extraction with

80% methanol for 72 hours. Crude extracts were then filtered and the solvent was evaporated at 40°C under reduced pressure using an Eyela rotary evaporator. Dried extracts were then resuspended in 1mL of analytical grade methanol (99.8% v/v) and stored at -20°C pending further analysis. Total phenolic content of each extract was determined using an adapted Folin-Ciocalteu procedure according to Zhang et al. 2006, using gallic acid as a standard. A serial dilution of gallic acid was prepared using distilled water to produce the following concentrations: 100%, 50%, 25%, 12.5% and

6.25%.

Assays were performed in triplicates for each sample using the following method: 10µL of the extract were placed in a well of a microplate, followed by 10µL of analytical grade methanol (99.8%v/v). This diluted each sample two-fold and prevented the occurrence of machine read errors which occur when the fluorescence of a sample is too high. 100µL of Folin-Ciocalteu reagent was added to each sample and this was mixed well and left for 30 minutes, before 80µL of 7.5% sodium carbonate solution was added. Once more, samples were mixed and then microplates were covered and left at room temperature for 2 hours. Triplicate blank samples were used as a control and prepared substituting analytical grade methanol (99.8% v/v) in place of leaf extract.

Absorbance of each sample was measured at 735nm using a Tecan Infinite M200 microplate reader and total phenolic content was calculated as in equation (1) and expressed as milligrams Gallic Acid Equivalents (GAE) per gram dry extract, adapted from (Zhang et al. 2006). Norm is the mean assay value of sample triplicates obtained from calibration curve (μgmL-1) minus the mean assay value of blank samples. V refers to the volume of plant extract in μL, D is the dilution factor and weight refers to the dry weight of sample extract in g.

108

푇표푡푎푙 푃ℎ푒푛표푙𝑖푐 퐶표푛푡푒푛푡 (푚𝑔퐺퐴퐸𝑔−1) = ([푁표푟푚 푥 푉 푥 퐷]/1000)/푤푒𝑖𝑔ℎ푡 (1)

Fibre content

Fibre content was measured in both damaged and undamaged leaves from a subsample of surviving seedlings (N=81 seedlings, 3-4 individuals per site). Acid Detergent Fibre

(ADF) was quantified following the protocol outlined by (Ankom Technology 2011) using an Ankom2000 Automated Fibre Analyser. This method uses cetyl trimethylammonium bromide (CTAB) and sulphuric acid to digest the proteins, starches, sugars, pectin and lipids in plant cells, leaving the less digestible fibrous components, cellulose and lignin (DeMilto et al. 2017).

Leaf strength

Measurements of leaf strength were carried out in situ on all 268 surviving seedlings before their removal and quantified as the force required to puncture a leaf using a pocket penetrometer fitted with a 0.25” tip. This measures the minimal leaf strength of the blade (Choong et al. 1992). To ensure the entire leaf blade experienced the same tension during recordings, a frame was constructed with a series of drilled 2-cm holes, through which the penetrometer could be pushed. Five measurements were taken per blade, and strength was expressed as the mean value in kgcm-2. The midrib and any venation were avoided and care was taken to ensure measurements were made within the intercostal lamina region.

109

Explanatory variables

Forest type

Since the primary aim of this study was to investigate whether foliar defensive traits are altered through forest fragmentation, forest type was included as a binary explanatory variable (fragment or continuous). The one hectare forest fragments used in this study were recently isolated in 2015 before the establishment of the seedling experiment.

Leaf Age

Seedlings were visited six times over the course of the experiment, and during each visit any new leaves were labelled using indelible ink. The minimum number of days since a leaf’s first noted appearance was used as a measure of leaf age and calculated as the number of days between a leaf’s first appearance and the final time point of the experiment. Seedlings could not be visited on a daily basis due to logistical constraints.

Leaves that were present on seedlings at time of planting were assigned the maximum number of days of the experiment (600 days) as a measure of their age. While this is not a precise measurement, as the specific day of leaf expansion was not recorded, it provides a consistent proxy for leaf age across seedlings in each site in this experiment.

Relative growth rate

Relative growth rate (RGR) of each seedling was calculated according to Bebber et al.,

2002, as the difference between the log10 vertical height of each seedling at the final time point (T5) and the initial time point (T0) divided by the number of days between these time points (see equation (2)). Seedling height was measured as the distance in centimetres to the nearest millimetre from the ground to the apical growing point.

110

RGR= (푙표𝑔10[퐻푇5 − 퐻푇0])/(푇5 − 푇0 ) (2)

Insect herbivore damage

A secondary aim of this study was to determine if herbivory damage induced any changes in foliar defence. All leaves were scanned before seedling removal at the end of the experiment, using a Canon Canoscan Lide 210 flatbed scanner. Leaf area loss was quantified using ImageJ version 1.48 (National Institutes of Health, USA). For these analyses, leaves were coded as “eaten” if they exhibited >1% damage, and “not eaten” if they had less than this. In order to obtain the minimum leaf mass required for phenolic and fibre analyses, multiple leaves from the same individual were often pooled and on these occasions, leaves were grouped according to whether they were damaged or undamaged. By assigning leaves to this binary system a balanced design of “eaten” or “not eaten” samples could be achieved for chemical analysis.

Canopy closure

A concave spherical densiometer was used to measure canopy closure, which was treated as a proxy for shade. Measurements were taken facing north, east, south and west and averaged to provide a mean percentage cover above each enclosure.

Statistical analyses

Mixed effects models were used to test the effect of forest fragmentation on the three separate defensive traits: total foliar phenolics, fibre content and leaf strength. All analyses were carried out using the nlme package (Pinheiro et al. 2018) in R version

3.4.4 (R Core Team 2018). Fixed covariates (canopy closure, leaf age and relative growth rate) were scaled (mean set to zero, standard deviation of one), to help with model convergence.

111

Pair plots of response variables versus all explanatory variables were produced in order to visually assess the presence of collinearity between predictors and interactions between explanatory variables (Zuur et al. 2011) (See Figures S5.1-S5.9). Pearson’s correlation coefficients were calculated and confirmed that collinearity was not an issue between variables (Figures S5.1-S5.9). To determine the optimal random effects structure, likelihood testing was performed using restricted maximum likelihood

(REML) estimation; the fixed structure of each model remained unchanged and included all predictor variables. During this stage, generalised least squares (GLS) was used as a null regression model, and mixed models were built first with a random intercept only, then with both random slope and intercept. Models were compared using

Akaike’s Information Criterion (AIC) (Zuur et al. 2011).

Preliminary data exploration revealed a nonlinear relationship between total phenolic content and relative growth rate for Dryobalanops lanceolata (Figure 5.3), a cubic expression for this covariate was included in the maximal model structure for this species (I(RGR.Dlan^3)). Similarly, polynomial relationships were observed between relative growth rate and acid detergent fibre, consequently a quadratic expression for growth rate was included for D. lanceolata (I(RGR.Dlan^2)), and a cubic expression was included for P. tomentella (I(RGR.Ptom^3)) (Figure S5.11 (A)). Finally, for analyses of leaf strength, a quadratic expression for canopy closure was included in the models for both species analysed together and for P. tomentella (I(Canopy.Ptom^2)

(Figure 5.5 (B)).

For each of the response variables, full models included forest type (fragment or continuous) and insect herbivore damage (eaten or not eaten) as fixed factors, and canopy closure, leaf age and relative growth rate were included as covariates, in addition to the polynomial expressions described above. To control for the clustered

112 sampling design, forest block (C, D, F, LF, LFE and VJR, see Figure 5.1) was included as a random effect. To obtain the minimal adequate model, full models were refitted with maximum likelihood (ML) estimation before backwards selection involving the sequential removal of insignificant predictors. Likelihood testing was then used to determine the optimal fixed structure. Once the minimal adequate model was determined, it was refitted using REML estimation. All models were validated visually to check for homogeneity of variance and normality of residuals (Zuur et al. 2011). To test for interspecific variation, species identity was included as a fixed factor in the first set of analyses. Models were then repeated for each species separately to investigate intraspecific variation in each of the three explanatory variables.

5.4 Results

Interspecific variation in leaf defence

Seedlings of D. lanceolata contained lower levels of foliar phenolics and had higher fibre content compared to those of P. tomentella (p<0.05, Table 5.1). No significant difference in leaf strength was found between the two species (p>0.05, Table 5.1).

Species identity and the presence of herbivore damage significantly predicted total phenolic content for seedlings in this experiment (p<0.05, Table 5.2). Undamaged leaves contained significantly lower concentrations of phenols compared with undamaged leaves (parameter estimate= -29.787, p=0.01, Table 5.2). The minimal adequate model for total phenolic content also included forest type and relative growth rate, although neither variable was significant, their presence in the reduced model suggests they partially explain phenolic content when both species are analysed together. Negative parameter estimates suggest lower phenolic content is present in leaves of seedlings located in forest fragments and in individuals with increased growth

113 rates (parameter estimate= -47.189 and -10.130 for forest type and relative growth rate, respectively, Table 5.2, Figure5.2(A) & Figure 5.3).

Seedlings of D. lanceolata contained higher proportions of acid detergent fibre in their leaves (Table 5.1, p<0.05) than those of P. tomentella. The minimal adequate model for fibre content contained species identity and herbivore damage. Undamaged leaves had significantly higher fibre content (parameter estimate= 1.823, p=0.02, Table 5.3).

Leaf strength did not vary significantly between species or forest type but instead was predicted by canopy closure where a significant negative linear relationship was detected (parameter estimate= -8.138, p=0.008, Table 5.4). A second order polynomial expression for canopy closure also remained in the minimal adequate model for this response variable.

Table 5.1 Mean (x̅ ) ± standard error of each defensive trait for seedlings of Dryobalanops lanceolata and Parashorea tomentella. Bold p-values indicate significant differences between the two species when analysed together using mixed effects modelling.

Dryobalanops lanceolata Parashorea tomentella p x̅ ± SE x̅ ± SE

Total phenolic content (mgGAEg-1) 140.202±9.594 165.575±7.664 0.048

Acid detergent fibre (%) 69.304±0.418 66.0650±0.4508 0.000

Leaf strength (kgcm-2) 134.601 ±2.360 134.502 ±2.786 0.550

Intraspecific variation in leaf defence

Total phenolic content

The minimal adequate model for total phenolic content, in seedlings of D. lanceolata, included insect herbivore damage, only, as a fixed effect. Undamaged leaves had significantly lower phenolic content (Figure 5.2(B), parameter estimate= -40.236,

114 p=0.019, Table 5.2). The minimal adequate model, for individuals of P. tomentella, included forest type as a fixed factor and leaf age and growth rate as fixed covariates.

Higher phenolic content was recorded in older leaves, and while this trend was not significant, the removal of leaf age reduced the overall fit of the model (Figure

S5.10(B)). Significantly lower levels of foliar phenolics were observed in P. tomentella seedlings located in forest fragments (p=0.039, Table 5.2, Figure 5.2(A)). A significant negative relationship was also observed between relative growth rate and phenolic content for seedlings of this species; individuals which exhibited higher growth rates had reduced foliar phenolic content (parameter estimate= -13.912, p=0.045, Table 5.2,

Figure 5.3).

Figure 5.2 The relationship between (A) forest type and (B) herbivore damage with total foliar phenolic content expressed as mg Gallic Acid Equivalents (GAE) per gram dried leaf extract for seedlings of Dryobalanops lanceolata and Parashorea tomentella.

115

Figure 5.3 The relationship between relative growth rate and total foliar phenolic content expressed as mg Gallic Acid Equivalents (GAE) per gram dried leaf extract for seedlings of Dryobalanops lanceolata and Parashorea tomentella.

Table 5.2 Coefficients for mixed effects models for Total Phenolic Content (expressed as mgGAE/gExtract) for both species analysed together, and for Dryobalanops lanceolata and Parashorea tomentella analysed separately. These coefficients derived from the minimal adequate model following model reduction. The ΔAIC between the reduced and full models are provided.

Species Term Coeff. SE T d.f P ΔAIC

Intercept 182.361 20.026 9.106 172 0.000

Forest type -47.189 26.036 -1.812 4 0.144 Both Species Species 24.060 12.048 1.997 172 0.048 -3.146 Together Insect herbivory -29.787 11.475 -2.596 172 0.010

Relative growth rate -10.130 6.055 -1.673 172 0.096

Intercept 160.596 22.542 7.124 93 0.000 Dryobalanops -6.162 lanceolata Insect herbivory -40.236 16.892 -2.382 93 0.019

Intercept 191.638 11.342 16.896 73 0.000

Forest type -49.415 16.335 -3.025 4 0.039 Parashorea -2.989 tomentella Leaf age 12.882 6.933 1.858 73 0.067

Relative growth rate -13.912 6.830 -2.037 73 0.045

116

Fibre content

The minimal adequate model for fibre content, for individuals of D. lanceolata, contained forest type and insect herbivore damage as fixed factors and a second order polynomial expression for relative growth rate as a fixed covariate. Leaves of seedlings planted in forest fragments exhibited lower fibre content (Figure 5.4(A)), and although this was not significant, removing forest type reduced model fit. Similarly, growth rate explained some of the variation in fibre content, whereby the highest fibre content was found in seedlings which exhibited medium growth rates (~0.00015 cm/day) for D. lanceolata (see Figure S5.11(A). Herbivore damage significantly predicted fibre content, with lower fibre content found in eaten leaves for this species (p=0.021, Table

5.3).

Figure 5.4 The relationship between (A) forest type, and (B) herbivore damage with percentage foliar acid detergent fibre (ADF) content for seedlings of Dryobalanops lanceolata and Parashorea tomentella. The minimal adequate model for seedlings of P. tomentella contained both damage class and leaf age as fixed effects. Similar to seedlings of D. lanceolata, significantly lower fibre content was found in eaten leaves (p=0.005, Table 5.3). Leaf age co-varied with fibre content, although this was not significant, older leaves were associated with

117 higher proportions of acid digested fibre (see Figure S5.11(C)). Fibre content was measured on a relatively small subsample of leaf material (N=81 seedlings).

Consequently, it is possible that with an increased sample size the nonsignificant trends observed would become significant.

Table 5.3 Coefficients for mixed effects models for percentage Acid Detergent Fibre content, for both species analysed together, and for Dryobalanops lanceolata and Parashorea tomentella analysed separately. These coefficients are derived from the minimal adequate model following model reduction. The ΔAIC between the reduced and full models are provided.

Species Term Coeff. SE T d.f P ΔAIC

Intercept 68.690 0.778 0.778 106 0.000

Both species Species -3.059 0.564 -5.424 106 0.000 -4.739

Insect herbivory 1.823 0.559 3.262 106 0.002

Intercept 69.818 0.915 76.284 43 0.000

Dryobalanops Forest type -1.637 1.144 -1.431 4 0.226 -5.472 lanceolata Insect herbivory 1.767 0.737 2.399 43 0.021

(Relative growth rate)2 -0.378 0.247 -1.534 43 0.132

Intercept 65.298 0.784 83.316 55 0.000 Parashorea Insect herbivory 2.562 0.874 2.932 55 0.005 -5.465 tomentella Leaf age 0.782 0.436 1.794 55 0.078

Leaf strength

When considering leaf strength as a response variable, the minimal adequate model, for individuals of D. lanceolata, included canopy closure and relative growth rate. Canopy closure significantly predicted leaf strength for individuals of D. lanceolata, with increased leaf strength occurring in more open environments (parameter estimate -

5.003, p=0.035, Table 5.4). While not significant, relative growth rate remained in the

118 minimal adequate model and suggested an increase in growth in more closed environments (p=0.148, Table 5.4).

The minimal adequate model for seedlings of P. tomentella included canopy closure, a second order polynomial expression for canopy closure and leaf age. Leaf strength increased with leaf age, however this relationship was not significant (p=0.089, Table

5.4, Figure S5.13). A significant second order polynomial trend was identified in response to canopy closure, indicating a peak in leaf strength in sites with 80% canopy closure (p=0.025, Table 5.4, Figure 5.5(B)). A significant linear relationship was also recorded with canopy closure demonstrating an overall decrease in leaf strength as canopy closure increases (parameter estimate= -12.8513, p=0.008, Table 5.4).

Figure 5.5 The relationships between (A) relative growth rate and (B) canopy closure with leaf strength for seedlings of Dryobalanops lanceolata and Parashorea tomentella.

119

Table 5.4 Coefficients for mixed effects models for Leaf Strength (expressed as kg/cm2) for both species analysed together, and for Dryobalanops lanceolata and Parashorea tomentella analysed separately. These coefficients are derived from the minimal adequate model following model reduction. The ΔAIC between the reduced and full models are provided.

Species Term Coeff. SE T d.f P ΔAIC

Intercept 134.905 5.509 24.490 260 0.000

Both species Canopy -8.138 3.029 -2.686 260 0.008 -6.829 closure

(Canopy)2 -1.820 1.094 -1.664 260 0.097

Intercept 134.681 4.836 27.850 135 0.000

Dryobalanops Canopy -5.003 2.353 -2.126 135 0.0353 -5.161 lanceolata closure Relative growth 3.300 2.267 1.456 135 0.148 rate

Intercept 135.052 6.440 20.973 116 0.000

Canopy -12.851 4.755 -2.703 116 0.008 Parashorea closure -4.615 tomentella (Canopy -3.902 1.715 -2.276 116 0.025 closure)2

Leaf age 4.382 2.547 1.721 116 0.089

5.5 Discussion

The importance of a species-specific approach

When analysed together, species identity was a significant predictor of both total foliar phenolics and acid detergent fibre content. Although both species showed consistent trends in response to forest type and herbivore damage, they each showed different responses to the other environmental and seedling traits included in the models. Only when species were analysed separately could further insights into the driving forces behind the observed differences in the three defensive traits be detected.

120

Does forest fragmentation alter foliar defensive traits?

The central aim of this study was to determine whether forest fragmentation altered foliar defensive traits in experimentally planted dipterocarp seedlings. We found evidence to suggest there is a relaxation of certain defensive traits in forest fragments. A significant reduction in total phenolic content was found in leaves from seedlings planted in one hectare forest fragments compared to those planted in continuous forest sites for one of the study species, P. tomentella, and while not significant the same trend was also observed in seedlings of D. lanceolata. Moreover, a decrease in fibre content was found in seedlings located in forest fragments, although this was not statistically significant in our study, it is an important trend that may become significant if sample sizes for this trait were larger. Nevertheless, our findings are consistent with other authors, who found evidence that forest fragmentation may reduce foliar defensive traits

(Fáveri et al. 2008).

Inconsistent responses to fragmentation have been found in previous studies in relation to leaf toughness. Fáveri et al. (2008) recorded decreased toughness in forest fragments, while another study found increases (De La Vega et al. 2012), and others failed to detect a significant response (Ruiz-Guerra et al. 2016). In our study, leaf strength did not differ between fragments and continuous forest (Figure S5.12(A)). This may reflect a longer time needed for adaptations to be revealed. Leaf strength is dependent on a diverse suite of factors including cell wall components and composition, leaf water content and cuticle thickness (Read and Sanson 2003, Matsuda et al. 2017). The seedlings used in our study were experimentally planted and monitored over sixteen months, perhaps a longer study period is required for changes in this trait to be detected.

121

Canopy closure predicted leaf strength for both species in this experiment, with seedlings in more closed canopy sites having weaker leaves. This relationship was significant when both species were analysed together and individually. These results follow previous research which has shown that leaves in more open environments are typically tougher than those in closed canopy environments (Lowman and Box 1983,

Martínez-Garza and Howe 2005, Agosta et al. 2017). In our study, however, a more complex polynomial relationship was also detected between canopy closure and leaf strength for P. tomentella suggesting the optimal canopy closure for maximal leaf strength in this species is ~80%.

Surprisingly, in our study, leaf strength was not predicted by herbivore damage (Figure

S5.12 (B)). This contradicts previous research where toughness has been shown to successfully predict insect herbivory using the same technique (Coley 1983). More recently the application of measuring force to puncture using a penetrometer has been questioned, and it has been suggested that using the force required to tear or shear a leaf might be more ecologically relevant (Lucas et al. 2000). Quantifying the force required for tearing and shearing requires more complex instruments and is more easily carried out under laboratory conditions (Peeters et al. 2007). It is possible that we failed to detect any significant variation in leaf strength due to inadequacies associated with the use of a simple penetrometer. However this method is commonly implemented to record leaf tensile properties under field conditions (Wright and Fuller 1984, Choong et al.

1992, Agosta et al. 2017, Matsuda et al. 2017).

Insect herbivory and foliar defence

Increased concentrations of total phenolics were detected in damaged leaves from seedlings of both species, although species-specific analyses indicated that this was only

122 significant in individuals of D. lanceolata suggesting that members of this species were driving the overall trend. Herbivory has previously been found to correlate with higher concentrations of foliar phenolics (Silva et al. 2015). This likely reflects an induced chemical response due to herbivory, which has been recorded in other plant genera (War et al. 2012, Bixenmann et al. 2016, Peschiutta et al. 2017). Induced chemical responses such as increased phenolic content can have important consequences for plants, and recently have been found to reduce the photosynthetic ability of a plant (Visakorpi et al.

2018). Since herbivore damage was consistent across forest fragments and continuous forest sites (see Chapter 4) for leaves included in these analyses, the reduced phenolic content in forest fragments should not reflect a reduction in herbivore damage.

Undamaged leaves were also characterised by increased fibre content, suggesting that more fibrous leaves are less palatable to insect herbivores.

Leaf age

Leaf age was an important predictor of foliar palatability in this experiment and remained present in the reduced models for all three defensive traits in seedlings of P. tomentella. While none were statistically significant, these trends are consistent with the literature that states mature leaves are better defended (Peeters et al. 2007), and confirms that it is a valuable metric to include when studying foliar defence.

To grow or defend?

Plants are challenged with the dilemma of investing in growth or defensive properties

(Peschiutta et al. 2017). In our study, seedlings of P. tomentella that exhibited higher growth rates contained significantly lower foliar phenolics. This supports the resource availability hypothesis and is in line with Kursar and Coley’s (2003) theory that plants can be placed along a spectrum from “defence’ to “escape” syndromes. Individuals of

123 this species are investing their carbon into vertical growth rather than chemical defence in the form of secondary metabolites, such as total phenolics. A non-significant relationship was also discovered between fibre content and relative growth rate in individuals of D. lanceolata. This more complicated polynomial relationship suggests that seedlings with lower growth rates produce more fibrous components in their leaves but the leaves of individuals with faster vertical growth rates contain less fibre. The latter provides further support to the hypothesis that plants either invest in growth or defensive traits. Despite not being significant, an interesting observation was made in seedlings of D. lanceolata whereby faster growing individuals also exhibited stronger leaves.

Consequences of variation in defensive traits

Investment in chemical and structural defences diverts resources from growth and production of leaf material (Herms and Mattson 1992), and we found evidence to support this. This in turn has ecological consequences for the plant as a reduction in growth rate may translate to a competitive disadvantage compared to conspecifics and increase the risk of premature mortality (Herms and Mattson 1992).

Conversely, plants invest in defensive traits to mitigate against natural enemies such as insect herbivores and pathogens. Shifts in plant traits stand to have consequences for these complex ecological interactions, and may result in a mismatch in these relationships, but will depend on the ability of insects to adapt to changes and the availability of other plant species with similar traits. This may be particularly catastrophic in tropical ecosystems where higher levels of specialisation occur (Forister et al. 2015).

124

Biomechanical traits successfully predict densities of herbivore guilds; all insect herbivores can be predicted by force required to shear or tear leaves (Peeters et al.

2007). Force to puncture is negatively correlated with external leaf chewers (Malishev and Sanson 2015). Sessile phloem feeders, such as sap suckers overcome structural traits through the use of salivary enzymes which break down the surface of the lamina, and are therefore not easily predicted by structural properties, but are more closely correlated with foliar nutrition (Peeters et al. 2007). Variation in defensive traits in response to anthropogenic disturbance such as fragmentation thus has important consequences for the functional diversity of herbivores in fragments. Smaller chewing insect herbivores may be vulnerable to increases in leaf strength or fibre content as they may not have the strength necessary to fracture leaves and access the food resource

(Peeters et al. 2007).

Limitations of this study and future directions

Seedlings used in this study were propagated from seeds collected during a single mast fruiting event, from the same forest location and grown under the same light and humidity regimes. We assumed that leaf traits did not vary significantly between seedlings at point of planting and that variations in leaf traits detected in this study reflect adaptations that were made over the study period in relation to each individual’s environment. It is of course possible that variation recorded in this study was due to genetic variation between the seedlings. However, as seedlings were randomised prior to planting, it is unlikely that traits were biased between forest treatments and consequently the differences we observe at the end of this experiment likely reflect induced changes due to their environment.

125

Our study included seedlings of just two species and hence, drawing conclusions for wider communities would require additional species to be tested. Despite this, a significant difference in foliar defensive traits was detected between one hectare forest fragments and continuous forest site, and merits further investigation over longer study periods. Since, another previous attempt to investigate the effects of forest fragmentation on leaf defence failed to detect any significant responses (Ruiz-Guerra et al. 2016), it is encouraging that significant responses could be detected and highlights the importance of carrying out traits analyses whilst studying herbivory.

We have presented evidence that fragmentation alters key defensive traits which could result in a reduced ability of seedlings to protect themselves from natural enemies such as insect herbivores. Since previous research has shown that defoliation of as little as

10% can reduce plant fitness (Coley and Barone 1996), such changes in defensive traits may have cascading consequences at population and community levels for plants with potential implications for regeneration success within fragments. In turn, changes in vegetation may impact insect herbivore communities which are likely to have further consequences for forest ecosystems.

Acknowledgements

This research was funded by an Australian Research Council Discovery Grant

DP160102078 and LKK received additional support from a Griffith University

Postgraduate Research Scholarship. For permission to carry out research we thank the

Sabah Biodiversity Council (Access Licence: JKM/MBS.1000-2/2 JLD.4 (56)) and

Sabah Forestry Department and the SAFE project. We give special thanks to all of the

SAFE research assistants for their help in the field. We extend our thanks to Hoe Seng

126

Tin and Yatau, M. H. for their advice and assistance in the laboratory and to Dr Tom M.

Fayle for providing valuable feedback on this manuscript.

127

S5 Supplementary material

Figure S 5.1 Pairplot of the response variable, concentration of total phenolics expressed as mg gallic acid equivalents per g dried extract, and explanatory variables for both species together. The upper panel contains scatterplots between each explanatory variable. Pearson’s correlation coefficients are provided in the lower panel, the font size of correlation coefficients is proportional to their value.

128

Figure S5.2 Pairplot of the response variable, concentration of total phenolics expressed as mg gallic acid equivalents per g dried extract, and explanatory variables for Dryobalanops lanceolata. The upper panel contains scatterplots between each explanatory variable. Pearson’s correlation coefficients are provided in the lower panel, the font size of correlation coefficients is proportional to their value.

129

Figure S5.3 Pairplot of the response variable, concentration of total phenolics expressed as mg gallic acid equivalents per g dried extract, and explanatory variables for Parashorea tomentella. The upper panel contains scatterplots between each explanatory variable. Pearson’s correlation coefficients are provided in the lower panel, the font size of correlation coefficients is proportional to their value.

130

131

Figure S5.5 Pairplot of the response variable, percentage acid detergent fibre (ADF) content, and explanatory variables for Dryobalanops lanceolata. The upper panel contains scatterplots between each explanatory variable. Pearson’s correlation coefficients are provided in the lower panel, the font size of correlation coefficients is proportional to their value.

132

Figure S5.6 Pairplot of the response variable, percentage acid detergent fibre (ADF) content, and explanatory variables for Parashorea tomentella. The upper panel contains scatterplots between each explanatory variable. Pearson’s correlation coefficients are provided in the lower panel, the font size of correlation coefficients is proportional to their value.

133

Figure S5.7 Pairplot of the response variable, leaf strength expressed as kgcm-2, and explanatory variables for both species together. The upper panel contains scatterplots between each explanatory variable. Pearson’s correlation coefficients are provided in the lower panel, the font size of correlation coefficients is proportional to their value.

134

Figure S5.8 Pairplot of the response variable, leaf strength expressed as kgcm-2, and explanatory variables for Dryobalanops lanceolata. The upper panel contains scatterplots between each explanatory variable. Pearson’s correlation coefficients are provided in the lower panel, the font size of correlation coefficients is proportional to their value.

135

Figure S5.9 Pairplot of the response variable, leaf strength expressed as kgcm-2, and explanatory variables for Dryobalanops lanceolata. The upper panel contains scatterplots between each explanatory variable. Pearson’s correlation coefficients are provided in the lower panel, the font size of correlation coefficients is proportional to their value.

136

Figure S5.10 Relationships between (A) canopy closure (B) leaf age with the response variable total phenolic content, expressed as mg Gallic Acid Equivalents (GAE) per g dried extract for seedlings of D. lanceolata and P. tomentella.

137

138

Figure S5.12 Boxplot showing differences in leaf strength between (A) forest type (B) damage class for seedlings of D. lanceolata and P. tomentella.

Figure S5.13 Relationship between leaf age with the response variable leaf strength for seedlings of D. lanceolata and P. tomentella.

139

Chapter 6 General discussion

6.1 Introduction

Tropical forests are responsible for one-third of terrestrial productivity and evapotranspiration and hold over 50% of terrestrial biodiversity (Malhi et al. 2014).

Rainforests in the tropics comprise the majority of global hotspots of biodiversity, which are characterised by high levels of endemism and habitat loss (Myers et al. 2000).

They also play a key role in global carbon cycling (Riutta et al. 2018). Despite their importance, tropical forests remain highly threatened by anthropogenic disturbance such as logging, deforestation and agricultural expansion and these destructive forces show no signs of slowing due to increasing human populations (Laurance et al. 2014, Lewis et al. 2015). As a consequence, understanding the effects of degradation on tropical rainforest should be a conservation priority (Steffan-Dewenter et al. 2007, Barlow et al.

2018). The impacts on the ecology of dipterocarp dominated forests is particularly pressing as they have long-experienced degradation from logging due to the economic value of members of this plant family (Brearley et al. 2017). Members of the

Dipterocarpaceae display synchronised reproduction through a phenomenon called mast fruiting, masting events are dependent on complex environmental cues which may be disrupted by climate change (Cannon et al. 2007). Furthermore, given the disproportionate attention that has previously been given to species’ distributions and numbers, a focus on the effects of habitat degradation on ecological processes is of utmost importance (Ruiz-Guerra et al. 2012, Valiente-Banuet et al. 2015).

6.2 Revisiting research aims

The overarching aim of this thesis has been to investigate the effects of forest fragmentation on a vital antagonistic interaction (insect herbivory) and associated foliar

140 defensive traits, with particular focus on seedlings of two species from the

Dipterocarpaceae. Throughout this final chapter, I shall revisit the research goals outlined in Chapter 1 and expand my discussion in relation to the following points: (1) the contributions this research may have to understanding the ecology and conservation of dipterocarps, and (2) the complex spatiotemporal components of fragmentation research with reference to limitations of this study and suggestions for future directions.

Using a combination of literature review, primary data collection and phytochemical analysis, I have shown;

 members of the Dipterocarpaceae are important host plants for phytophagous

insects of the order Lepidoptera (Chapter 2),

 herbivory is a complex and dynamic process, the understanding of which

benefits from repeated measure studies (Chapter 4),

 rates of leaf damage in seedlings may be reduced in fragments even in the initial

phase following isolation (Chapter 4),

 larger seedlings experience higher levels of defoliation (Chapter 4),

 density-dependent effects on herbivory can persist in small fragments (Chapter

4),

 canopy closure is an important predictor of herbivore damage (Chapter 4) and

leaf strength (Chapter 5),

 fragmentation causes a relaxation in two key foliar defensive traits: total

phenolic content and acid detergent fibre (Chapter 5),

 phytochemical defence may be species-specific (Chapter 5), and

 leaves that experience higher levels of insect herbivory contain higher

concentrations of total phenolics but lower acid detergent fibre (Chapter 5).

141

Results of this thesis suggest the even initial effects of fragmentation may disrupt insect herbivory and cause changes to foliar defence (Aims 1, 2 and 3, Chapters 4 and 5).

Forest type partially explained fibre content in seedlings of Dryobalanops lanceolata with lower fibre content recorded in forest fragments (Chapter 5). While this could indicate increased vulnerability to insect herbivory in forest isolates, this did not translate to observed patterns of defoliation in this species. Reduced leaf area loss was observed in forest fragments throughout this experiment, with the exception of a single sampling event (out of six) for seedlings of Parashorea tomentella (Chapter 4). A significant decrease in total phenolics was recorded in fragments for P. tomentella and, when both species were analysed together. This defensive trait was also significantly predicted by insect herbivore damage in seedlings of D. lanceolata, and when both species were analysed together (Chapter 5). Taking these results in tandem with those of

Chapter 4 we can infer that increased phenolics in continuous forest is likely an induced response to the higher levels of leaf area loss that was also recorded in those sites. The reduction in acid detergent fibre in seedlings of D. lanceolata located in forest fragments (Chapter 5) further suggests a relaxation in foliar defence.

This thesis has provided evidence of important species-specific responses to fragmentation (Aim 4, Chapters 4 and 5).

Previously, species identity has been found to explain patterns of herbivory in the

Dipterocarpaceae (Eichhorn et al. 2006). While both my study species showed similar patterns of defoliation in both forest types throughout this study (Chapter 4), they varied significantly in their foliar defensive traits (Chapter 5). D. lanceolata had significantly lower phenolic content than P. tomentella. Conversely D. lanceolata contained

142 significantly higher fibre content (Chapter 5), suggesting that this species may rely on mechanical rather than phytochemical defences which P. tomentella may rely upon.

Despite a stronger reliance on foliar phenolics, a significant reduction in phenolic content of leaves in forest fragments was only observed in P. tomentella. Similarly, despite having higher fibre content generally, lower amounts were found in leaves of D. lanceolata in forest fragments. This may indicate increased vulnerability for both species in forest fragments through a relaxation in both of their usual defensive mechanisms.

This thesis has used information on known interactions between plants and phytophagous lepidopteran caterpillars across tropical Asia to determine the importance of members of the Dipterocarpaceae as host plants (aims 5 and 6,

Chapter 2)

Through the use of a key published resource on the host plants of lepidopteran caterpillars across tropical Asia (Robinson et al. 2001), I have found that species of dipterocarp are important host plants for a diverse array of phytophagous Lepidoptera species.

This review also highlighted that the majority of research has been carried out on three main plant parts: leaves, fruits and seeds. This is unsurprising given that most studies on herbivory focus on leaf chewers (Novotny et al. 2010). The Dipterocarpaceae are also renowned for their reproductive strategy of mast fruiting (Janzen 1974), which logically leads to research into the products of this phenomenon (fruits and seeds) as a food source.

Contributions of this thesis to dipterocarp ecology and conservation

143

The Dipterocarpaceae are an important family of plant which dominates tropical rainforest in the study region of this thesis (Slik et al. 2003, Brearley et al. 2017). Their ecological dominance and economic value as timber species have led to high levels of extraction (Maycock et al. 2012). As a consequence of the high levels of deforestation and degradation from logging in Southeast Asia, the majority of lowland dipterocarp forests are highly fragmented (Kettle et al. 2012). In Sabah, a recent analysis of the conservation status of species from the Dipterocarpaceae found that all but one species could be classed as threatened under the IUCN red list criterion of habitat loss

(Maycock et al. 2012). Understanding the fundamental ecology of this important plant family is therefore vital in order to ensure the structure and function of dipterocarp forest communities (Brearley et al. 2017). This thesis has focused on the antagonistic relationship between dipterocarp seedlings and phytophagous insects, and has yielded some useful results, which may have implications for restoration methods such as enrichment planting.

Enrichment planting has been proposed as a method of restoration which can help re- establish the biodiversity of dipterocarps in degraded or fragmented habitats (Yeong et al. 2016) with concurrent financial incentives when valuable timber species are planted

(Kettle 2010). Enrichment planting may play a role in countering the detrimental effects of fragmentation; a recent study recorded no difference in the survival of planted seedlings in forest isolates and in continuous forest sites (Yeong et al. 2016). The long- term success of such regimes, however, relies on a myriad of factors including ensuring genetic diversity of seedling stock is maintained, and appropriate aftercare is provided to ensure the growth and survival of seedlings (Kettle 2010).

While this thesis did not explicitly measure density-dependent effects, evidence that the density of conspecific seedlings affects herbivory rates even in highly degraded, one

144 hectare fragments was found (Chapter 4). This has important consequences for the patterns or density at which dipterocarp seedlings should be planted under enrichment planting regimes. High numbers of conspecifics translated to elevated herbivore damage when both species were analysed together (Chapter 4). Similarly, survival of planted dipterocarps is higher on slopes (Hattori et al. 2013) and this thesis found evidence that seedlings found on steeper slopes suffer less biotic pressure from insect herbivores

(Chapter 4).

Light availability has previously been identified as an important abiotic factor that dictates the growth and survival of dipterocarp seedlings (Bebber, Brown, Speight, et al.

2002). Consequently, it has been proposed that optimal light conditions should be maintained actively following restocking of degraded forests (Bebber, Brown, Speight, et al. 2002, Kettle 2010). My results lend support to this proposal and further highlight the importance of light. Canopy closure was influential in both patterns of herbivore damage (Chapter 4) and leaf traits (Chapter 5) in this study. Leaves were significantly stronger for both species (Chapter 5) and experienced lower levels of defoliation

(Chapter 4) in sites with more open canopies.

Soil type is an important factor affecting growth and survival of dipterocarps (Hattori et al. 2013), but was outside the scope of the current study. Light regimes, however, have been found to co-vary with soil type in lowland dipterocarp forests with both playing a role in predicting leaf traits (Dent and Burslem 2016).

Insects play an important role in dipterocarp ecology. Insects are required for pollination (Ghazoul 1997, Momose et al. 1998). In the early 1980s, however, pollination in dipterocarps was thought to be performed primarily by thrips, this was based on research carried out on the genus Shorea in Peninsular Malaysia and Sri Lanka

145

(Chan and Appanah 1980, Appanah 1993). As further research was carried out in different genera and geographic locations a number of pollination syndromes have been identified within this plant family (Momose et al. 1998). Rare interactions with

Hemiptera and Homoptera have been recorded (Ashton et al. 1988). In Borneo, the primary visitors to Shorea flowers have been found to be social bees from the genus

Apis, and others by beetles (members of Curculionoidea and Chrysomelidae)

(Nagamitsu et al. 1999, Sakai, Momose, Yumoto, Kato, et al. 1999), and genera with larger flowers such as Dipterocarpus by members of the Lepidoptera (Ghazoul 1997).

This relatively recent expansion of knowledge confirms diverse relationships with an array of insects and also highlights the overall dearth of information present on species interactions of dipterocarps.

The large compilation of Lepidoptera-Dipterocarpaceae interactions provided by

Robinson et al. 2001 has been useful in showing that dipterocarps are important host plants for herbivores from this Order of herbivorous insects (Chapter 2). This compilation underlines the utility of basic natural history information in meta-analyses.

Further research is required to determine the importance of dipterocarps for herbivores from other insect Orders (Orthoptera, Diptera and Coleoptera). This is particularly apposite given a recent analysis showed that Coleoptera were more sensitive to habitat degradation than species of Lepidoptera (Gibson et al. 2011).

Spatiotemporal components of investigating forest fragmentation

I used one-hectare forest fragments and continuous forest control sites within the SAFE project to investigate the initial effects of forest fragmentation on herbivory rates. I chose these spatial scales for two primary reasons: (1) previous research has found responses to fragmentation were strongest when studied between sites with large

146 differences in area (Rossetti et al. 2017) and (2) while the SAFE project experimental design also includes 10 and 100 hectare forest fragments, not enough of these fragments had been isolated at the time of this study to provide a balanced experimental design.

An obvious next step would be to expand this study to other fragment sizes within the

SAFE project which have now been created, and which have now been left isolated for a longer period. Inclusion of other fragment sizes may reveal further insights into this complex ecological interaction.

There is uncertainty as to the extent to which the effects of fragmentation on insect herbivory varies in space and time (Bagchi et al. 2018). However there is evidence to suggest that both are important. Moreover, multiple forms of habitat degradation often occur simultaneously- forest fragments are more vulnerable to fire and logging and more likely to be hunted and as such they are both spatially and temporally heterogeneous (Malhi et al. 2014).

First, the effects of fragmentation are dependent on the spatial scale at which this interaction is studied. The best predictors of insect herbivory at a local scale may differ from those that are important at larger scales (Tscharntke and Brandl 2004). This means that making global predictions of the effects of fragmentation is challenging, and means that field experiments, such as this one. These are invaluable for disentangling the driving forces behind changes in ecological processes (Bagchi et al. 2018).

There is also evidence for a strong temporal component to the effects of fragmentation.

Significant differences in biotic interactions may not manifest in the early stages following fragmentation (Chávez-Pesqueira et al. 2015). The effects of fragmentation are strongly dependent on the contrast between the forest that is fragmented and its surrounding matrix habitat (Laurance et al. 2011, Murphy et al. 2016), which likely

147 changes over time. Despite being carried out in the early stages of fragmentation at

SAFE, this thesis has revealed important differences in herbivory rates and phytochemical defence. The matrix environment at SAFE, which is currently naturally regenerating salvage-logged forest, is designated to become oil palm plantation.

However, planting had not begun during the time of this research. It is likely that further disruption to herbivory will occur once the plantations have been established and the process will continue to change as plantations mature.

This research has confirmed that herbivory is a complex and dynamic ecological interaction which changes over time (Chapter 4). I argue that it is inappropriate to rely upon snapshots of damage at single points in time and instead suggest that repeated measures be taken, whenever logistically feasible, when trying to disentangle effects of habitat degradation on this process. At the BDFFP in Brazil, insect herbivory in seedlings has been studied over twenty years and important spatiotemporal dynamics have been revealed, indicating that this process is affected by fragment size (Benítez-

Malvido 1998), but that these effects become more variable over time (Benítez-Malvido et al. 1999). Moreover, location within fragments has consequences for herbivory due to the prevalence of edge effects, whereby greater levels of herbivory are found at the centre of the fragment and they decrease as you reach the forest edge (Benítez-Malvido et al. 2018).

The SAFE project offers a rare opportunity for dynamic ecological interactions such as herbivory to be studied in a large-scale experimental framework over different spatial and temporal scales. A similar seedling experiment investigating the effects of fragmentation on insect herbivory in seedlings has been established at SAFE as of 2017.

This study is being led by Dr Cristina Banks-Leite in collaboration with Professor

148

Robert Ewers and offers the chance for future comparisons between results of this study and theirs.

6.3 Conclusions

A consequence of the hyperdiversity of tropical rainforests is that they are remarkably heterogeneous and dynamic. Understanding such complex ecosystems is challenging, and it is therefore useful to focus on smaller components of ecological importance.

Insect herbivory is a useful ecological interaction to study as any shifts can have cascading effects throughout the entire system. The multifaceted and complex array of factors that govern this process may be species specific and changes over time.

Although the relevance of this study at larger spatial scales awaits confirmation, nevertheless, the results of this thesis are of relevance to a wide readership, including those interested in insect herbivory, dipterocarp ecology, ecosystem functioning and forest fragmentation.

149

References

Adams, J. M., Zhang, Y., Basri, M., and Shukor, N., 2009. Do tropical forest leaves

suffer more insect herbivory? A comparison of tropical versus temperate

herbivory, estimated from leaf litter. Ecological Research, 24 (6), 1381–1392.

Agosta, S. J., Hulshof, C. M., and Staats, E. G., 2017. Organismal responses to habitat

change: herbivore performance, climate and leaf traits in regenerating tropical dry

forests. Journal of Animal Ecology, 86 (3), 590–604.

Agrawal, A. A., Hastings, A. P., Johnson, M. T. J., Maron, J. L., and Salminen, J.-P.,

2012. Insect Herbivores Drive Real-Time Ecological and Evolutionary Change in

Plant Populations. Science, 338 (6103), 113–116.

Aide, T. M., 1992. Dry season leaf production - an escape from herbivory. Biotropica,

24 (4), 532–537.

Akutsu, K., Khen, C. V., and Toda, M. J., 2007. Assessment of higher insect taxa as

bioindicators for different logging-disturbance regimes in lowland tropical rain

forest in Sabah, Malaysia. Ecological Research, 22, 542–550.

Almeida-Neto, M., Guimarães, P., Guimarães, P. R. J., Loyola, R. D., and Ulrich, W.,

2008. A consistent metric for nestedness analysis in ecological systems:

reconciling concept and measureement. Oikos, 117, 1227–1239.

Anderson, M. C., 1964. Studies of the Woodland Light Climate. Journal of Ecology, 52

(1), 27–41.

Ankom Technology, 2011. Acid Detergent Fiber in Feeds - Filter Bag Technique (for

A2000 and A20001) [online]. Method 6. Available from:

150

https://www.ankom.com/sites/default/files/document-

files/Method_13_NDF_Method_A2000_RevE_4_10_15.pdf [Accessed 20 Apr

2018].

Anstett, D. N., Nunes, K. A., Baskett, C., and Kotanen, P. M., 2016. Sources of

Controversy Surrounding Latitudinal Patterns in Herbivory and Defense. Trends in

Ecology and Evolution, 31 (10), 789–802.

Appanah, S., 1993. Mass flowering of dipterocarp forests in the aseasonal tropics.

Journal of Biosciences, 18 (4), 457–474.

Ashton, P., 1998. Dryobalanops lanceolata. The IUCN Red List of Threatened Species

[online]. Available from:

http://dx.doi.org/10.2305/IUCN.UK.1998.RLTS.T33164A9763445.en. [Accessed

13 Jun 2018].

Ashton, P. S., 1964. Manual of the Dipterocarp Trees of Brunei State. Oxford, United

Kingdom: Oxford University Press.

Ashton, P. S., 2010. Conservation of Borneo biodiversity: Do small lowland parks have

a role, or are big inland sanctuaries sufficient? Brunei as an example. Biodiversity

and Conservation, 19 (2), 343–356.

Ashton, P. S., Givnish, T. J. J., and Appanah, S., 1988. Staggered Flowering in the

Dipterocarpaceae: New Insights Into Floral Induction and the Evolution of Mast

Fruiting in the Aseasonal Tropics. The American Naturalist, 132 (1), 44–66.

Ashton, P. S. and Kettle, C. J., 2012. Dipterocarp biology as a window to the

understanding of tropical forest structure: Where are we looking now? Biotropica,

44 (5), 575–576.

151

Bagchi, R., Brown, L. M., Elphick, C. S., Wagner, D. L., and Singer, M. S., 2018.

Anthropogenic fragmentation of landscapes: mechanisms for eroding the

specificity of plant–herbivore interactions. Oecologia, 187 (2), 521–533.

Bagchi, R., Gallery, R. E., Gripenberg, S., Gurr, S. J., Narayan, L., Addis, C. E.,

Freckleton, R. P., and Lewis, O. T., 2014. Pathogens and insect herbivores drive

rainforest plant diversity and composition. Nature, 506 (7486), 85–8.

Baltzer, J. L., Thomas, S. C., Nilus, R., and Burslem, D. F. R. P., 2005. Edaphic

Specialization in Tropical Trees: Physiological Correlates and Responses to

Reciprocal Transplantation. Ecology, 86 (11), 3063–3077.

Banašek-Richter, C., Cattin, M., and Bersier, L. F., 2004. Sampling effects and the

robustness of quantitative and qualitative food-web descriptors. Journal of

Theoretical Biology, 226 (1), 23–32.

Barlow, J., França, F., Gardner, T., Hicks, C., Lennox, G., Berenguer, E., Castello, L.,

Economo, E., Ferreira, J., Guénard, B., Leal, C., Isaac, V., Lees, A., Paar, C.,

Wilson, S., Young, P., and Graham, N., 2018. The future of hyperdiverse tropical

ecosystems. Nature, 559, 517-.

Barton, K., 2018. MuMIn: Multi-Model Inference. R package version 1.42.1, (1).

Basset, Y., Charles, E., Hammond, D. S., and Brown, V. K., 2001. Short-term effects of

canopy openness on insect herbivores in a rain forest in Guyana. Journal of

Applied Ecology, 38, 1045–1058.

Basset, Y., Cizek, L., Cuénoud, P., Didham, R. K., Novotny, V., Ødegaard, F., Roslin,

T., Tishechkin, A. K., Schmidl, J., Winchester, N. N., Roubik, D. W., Aberlenc, H.

P., Bail, J., Barrios, H., Bridle, J. R., Castaño-Meneses, G., Corbara, B., Curletti,

152

G., Duarte Da Rocha, W., De Bakker, D., Delabie, J. H. C., Dejean, A., Fagan, L.

L., Floren, A., Kitching, R. L., Medianero, E., Gama De Oliveira, E., Orivel, J.,

Pollet, M., Rapp, M., Ribeiro, S. P., Roisin, Y., Schmidt, J. B., Sørensen, L.,

Lewinsohn, T. M., and Leponce, M., 2015. Arthropod distribution in a tropical

rainforest: Tackling a four dimensional puzzle. PLoS ONE, 10 (12), 1–22.

Bates, D., Mächler, M., Bolker, B., and Walker, S., 2015. Fitting Linear Mixed-Effects

Models Using lme4. J Stat Soft, 67 (1), 48.

Bebber, D., Brown, N., and Speight, M., 2002. Drought and Root Herbivory in

Understorey Parashorea Kurz (Dipterocarpaceae) Seedlings in Borneo. Journal of

Tropical Ecology, 18 (5), 795–804.

Bebber, D., Brown, N., Speight, M., Moura-Costa, P., and Wai, Y. S., 2002. Spatial

structure of light and dipterocarp seedling growth in a tropical secondary forest.

Forest Ecology and Management, 157, 65–75.

Bedetti, C. S., Ferreira, B. G., Castro, N. M. De, and Mary, R., 2013. The influence of

parasitoidism on the anatomical and histochemical profiles of the host leaves in a

galling Lepidoptera - Bauhinia ungulata system. Brazilian Journal of Biosciences,

11 (2), 242–249.

Benítez-Malvido, J., 1998. Impact of forest fragmentation on seedling abundance in a

tropical rain forest. Conservation Biology, 12 (2), 380–389.

Benítez-Malvido, J., Dáttilo, W., Martínez-Falcón, A. P., Durán-Barrón, C., Valenzuela,

J., López, S., and Lombera, R., 2016. The multiple impacts of tropical forest

fragmentation on arthropod biodiversity and on their patterns of interactions with

host plants. PLoS ONE, 11 (1), 1–15.

153

Benítez-Malvido, J., García-Guzmán, G., and Kossmann-Ferraz, I. D., 1999. Leaf-

fungal incidence and herbivory on tree seedlings in tropical rainforest fragments:

An experimental study. Biological Conservation, 91 (2–3), 143–150.

Benítez-Malvido, J. and Kossmann-Ferraz, I. D., 1999. Litter cover variability affects

seedling performance and herbivory. Biotropica, 31 (4), 598–606.

Benítez-Malvido, J., Lázaro, A., and Ferraz, I. D. K., 2018. Effect of distance to edge

and edge interaction on seedling regeneration and biotic damage in tropical

rainforest fragments: A long-term experiment. Journal of Ecology, (Accepted

Article), 1–14.

Benítez-Malvido, J. and Lemus-Albor, A., 2005. The seedling community of tropical

rain forest edges and its interaction with herbivores and pathogens. Biotropica, 37

(2), 301–313.

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., and Hamer, K.

C., 2010. The high value of logged tropical forests: lessons from northern Borneo.

Biodiversity and Conservation, 19 (4), 985–997.

Bicknell, J. E., Struebig, M. J., and Davies, Z. G., 2015. Reconciling timber extraction

with biodiversity conservation in tropical forests using reduced-impact logging.

Journal of Applied Ecology, 52 (2), 379–388.

Bierregaard, R. O., Lovejoy, T. E., Kapos, V., dos Santos, A. A., and Hutchings, R. W.,

1992. The biological dynamics of tropical rainforest fragments : a prospective

comparison of fragments and continuous forest . The biological dynamics of

tropical rainforest fragments : a prospective comparison of fragments and

154

continuous forest . BioScience, 42 (11), 859–866.

Bixenmann, R. J., Coley, P. D., Weinhold, A., and Kursar, T. A., 2016. High herbivore

pressure favors constitutive over induced defense. Ecology and Evolution, 6 (17),

6037–6049.

Blundell, A. G. and Peart, D. R., 1998. Distance-dependence in herbivory and foliar

condition for juvenile Shorea trees in Bornean dipterocarp rain forest. Oecologia,

117 (1–2), 151–160.

Blundell, A. G. and Peart, D. R., 2001. Growth strategies of a shade tolerant tropical

tree: the interactive effects of canopy and simulated herbivory. The Journal of

Ecology, 89, 608–615.

Blüthgen, N., Fründ, J., Vázquez, D. P., and Menzel, F., 2008. What Do Interaction

Network Metrics Tell Us About Specialization and Biological Traits? Ecology, 89

(12), 3387–3399.

Blüthgen, N., Menzel, F., and Blüthgen, N., 2006. Measuring specialization in species

interaction networks. BMC Ecology, 6 (9).

Blüthgen, N., Menzel, F., Hovestadt, T., Fiala, B., and Blüthgen, N., 2007.

Specialization, Constraints, and Conflicting Interests in Mutualistic Networks.

Current Biology, 17 (4), 341–346.

Blüthgen, N. and Metzner, A., 2007. Contrasting leaf age preferences of specialist and

generalist stick insects (Phasmida). Oikos, 116 (11), 1853–1862.

Blüthgen, N., Metzner, A., and Ruf, D., 2006. Food plant selection by stick insects

(Phasmida) in a Bornean rain forest. Journal of Tropical Ecology, 22 (1), 35–40.

155

Bolker, B. M., Brooks, M. E., Clark, C. J., Geange, S. W., Poulsen, J. R., Stevens, M.

H. H., and White, J.-S. S., 2009. Generalized linear mixed models: a practical

guide for ecology and evolution. Trends in Ecology and Evolution, 24 (3), 127–

135.

Brearley, F. Q., 2006. Differences in the growth and ectomycorrhizal community of

Dryobalanops lanceolata (Dipterocarpaceae) seedlings grown in ultramafic and

non-ultramafic soils. Soil Biology and Biochemistry, 38 (12), 3407–3410.

Brearley, F. Q., Banin, L. F., and Saner, P., 2017. The ecology of the Asian

dipterocarps. Plant Ecology and Diversity, 9 (5–6), 429–436.

Brearley, F. Q., Press, M. C., and Scholes, J. D., 2003. Nutrients obtained from leaf

litter can improve the growth of dipterocarp seedlings. New Phytologist, 160, 101–

110.

Breinholt, J. W., Earl, C., Lemmon, A. R., Lemmon, E. M., Xiao, L., and Kawahara, A.

Y., 2018. Resolving Relationships among the Megadiverse Butterflies and Moths

with a Novel Pipeline for Anchored Phylogenomics. Systematic Biology, 67 (1),

78–93.

Bruhl, C. a, Eltz, T., and Linsenmair, K. E., 2003. Size does matter - effects of tropical

rainforest fragmentation on the leaf litter ant community in Sabah, Malaysia.

Biodiversity and Conservation, 12 (7), 1371–1389.

Bryan, J. E., Shearman, P. L., Asner, G. P., Knapp, D. E., Aoro, G., and Lokes, B.,

2013. Extreme Differences in Forest Degradation in Borneo: Comparing Practices

in Sarawak, Sabah, and Brunei. PLoS ONE, 8 (7), e69679.

Bryant, J. P., Chapin, F. S., and Klein, D. R., 1983. Carbon/ Nutrient Balance of Boreal

156

Plants in Relation to Vertebrate Herbivory. Oikos, 40, 357–368.

Butt, N., Seabrook, L., Maron, M., Law, B. S., Dawson, T. P., Syktus, J., and Mcalpine,

C. A., 2015. Cascading effects of climate extremes on vertebrate fauna through

changes to low-latitude tree flowering and fruiting phenology. Global Change

Biology, 21 (9), 3267–3277.

Caldwell, E., Read, J., and Sanson, G. D., 2016. Which leaf mechanical traits correlate

with insect herbivory among feeding guilds? Annals of Botany, 117 (2), 349–361.

Cannon, C. H., Curran, L. M., Marshall, A. J., and Leighton, M., 2007. Beyond mast-

fruiting events: Community asynchrony and individual dormancy dominate woody

plant reproductive behavior across seven Bornean forest types. Current Science, 93

(11), 1558–1566.

Chan, H. T. and Appanah, S., 1980. Reproductive biology of some Malaysian

dipterocarps. I: flowering biology. Malaysian Forester, 43 (2), 132–143.

Chávez-Pesqueira, M., Carmona, D., Suárez-Montes, P., Núñez-Farfán, J., and Aguilar,

R., 2015. Synthesizing habitat fragmentation effects on plant-antagonist

interactions in a phylogenetic context. Biological Conservation, 192, 304–314.

Chong, K. Y., Chong, R., Tan, L. W. A., Yee, A. T. K., Chua, M. A. H., Wong, K. M.,

and Tan, H. T. W., 2016. Seed production and survival of four dipterocarp species

in degraded forests in Singapore. Plant Ecology and Diversity, 9 (5–6), 483–490.

Choong, M. F., Lucas, P. W., Ong, J. S. Y., Pereira, B., Tan, H. T. W., and Turner, I.

M., 1992. Leaf fracture toughness and sclerophylly: their correlations and

ecological implications. New Phytologist, 121 (4), 597–610.

157

Christenhusz, M. J. M. and Byng, J. W., 2016. The number of known plant species in

the world and its annual increase. Phytotaxa, 261 (3), 201–217.

Chung, A. Y. C., Maycock, C. R., Khoo, E., Hastie, A., Nilus, R., Majapun, R., Kimjus,

K., and Chey, V. K., 2013. New records of insects associated with Bornean

endemic dipterocarp seedlings. Journal of Tropical Forest Science, 25 (1), 5–11.

Chung, A. Y. C., Maycock, C. R., Khoo, E., Khen, C. V., and Kendrick, R. C., 2011.

New records of florivory on dipterocarp flowers. Malayan Nature Journal, 63 (3),

577–590.

Close, D. C. and McArthur, C., 2002. Rethinking the Role of Many Plant Phenolics:

Protection from Photodamage Not Herbivores? Oikos, 99 (1), 166–172.

Coley, P. D., 1980. Effects of leaf age and plant life history patterns on herbivory.

Nature.

Coley, P. D., 1983. Herbivory and Defensive Characteristics of Tree Species in a

Lowland Tropical Forest. Ecological Monographs, 53 (2), 209–233.

Coley, P. D. and Barone, J. A., 1996. Herbivory and plant defenses in tropical forests.

Annual Review of Ecology, Evolution, and Systematics, 27 (1996), 305–335.

Coley, P. D., Bryant, J. P., and Chapin, F. S., 1985. Resource Availability and Plant

Antiherbivore Defense. Science, 230 (4728), 895–899.

Coley, P. D. and Kursor, T. A., 1996. Anti-Herbivore Defenses of Young Tropical

Leaves: Physiological Constraints and Ecological Trade-offs. In: Mulkey, S. S.,

Chazdon, R. L., and Smith, A. P., eds. Tropical Forest Plant Ecophysiology.

Boston, MA: Springer, 305–336.

158

Connor, E. F. and Taverner, M. P., 1997. The Evolution and Adaptive Significance of

the Leaf-Mining Habit. Oikos, 79 (1), 6.

Corlett, R. T., 2014. The ecology of tropical east asia. 2nd Ed. Oxford: Oxford

University Press.

Craney, T. A. and Surles, J. G., 2002. Model-Dependent Variance Inflation Factor

Cutoff Values. Quality Engineering, 14 (3), 391–403.

Curran, L. M., Caniago, I., Paoli, G. D., Astianti, D., Kusneti, M., Leighton, M.,

Nirarita, C. E., and Haeruman, H., 1999. Impact of El Niňo and Logging on

Canopy Tree Recruitment in Borneo. Science, 286 (5447), 2184–2188.

Curran, L. M., Trigg, S. N., McDonald, A. K., Astiani, D., Hardiono, Y. M., Siregar, P.,

Caniago, I., and Kasischke, E., 2004. Lowland Forest Loss in Protected Areas of

Indonesian Borneo. Science, 303 (5660), 1000–1003.

Curran, L. M. and Webb, C. O., 2000. Experimental Tests of the Spatiotemporal Scale

of Seed Predation in Mast-Fruiting Dipterocarpaceae. Ecological Monographs, 70

(1), 129–148.

Dassou, A. G. and Tixier, P., 2016. Response of pest control by generalist predators to

local-scale plant diversity: A meta-analysis. Ecology and Evolution, 6 (4), 1143–

1153.

Daws, M. I., Pearson, T. R. H., Burslem, D. F. R. P., Mullins, C. E., and Dalling, J. W.,

2005. Effects of topographic position, leaf litter and seed size on seedling

demography in a semi-deciduous tropical forest in Panamá. Plant Ecology, 179,

93–105.

159

Deere, N. J., Guillera-Arroita, G., Baking, E. L., Bernard, H., Pfeifer, M., Reynolds, G.,

Wearn, O. R., Davies, Z. G., and Struebig, M. J., 2018. High Carbon Stock forests

provide co-benefits for tropical biodiversity. Journal of Applied Ecology, 55 (2),

997–1008.

DeMilto, A. M., Rouquette, M., Mueller, U. G., Kellner, K., and Seal, J. N., 2017.

Effects of substrate, ant and fungal species on plant fiber degradation in a fungus-

gardening ant symbiosis. Journal of Insect Physiology, 98, 301–308.

Dent, D. H. and Burslem, D. F. R. P., 2016. Leaf traits of dipterocarp species with

contrasting distributions across a gradient of nutrient and light availability. Plant

Ecology and Diversity, 9 (5–6), 521–533.

Dhileepan, K., 2003. Seasonal variation in the effectiveness of the leaf-feeding beetle

Zygogramma bicolorata (Coleoptera: Chrysomelidae) and stem-galling moth

Epiblema strenuana (Lepidoptera: Tortricidae) as biocontrol agents on the weed

Parthenium hysterophorus (Asteraceae). Bulletin of Entomological Research, 93

(05), 393–401.

Didham, R. K. and Ewers, R. M., 2012. Predicting the impacts of edge effects in

fragmented habitats: Laurance and Yensen’s core area model revisited. Biological

Conservation, 155, 104–110.

Didham, R. K., Kapos, V., and Ewers, R. M., 2012. Rethinking the conceptual

foundations of habitat fragmentation research. Oikos, 121 (2), 161–170.

Döbert, T. F., Webber, B. L., Sugau, J. B., Dickinson, K. J. M., and Didham, R. K.,

2017. Logging, exotic plant invasions, and native plant reassembly in a lowland

tropical rain forest. Biotropica, 50 (2), 254–265.

160

Dormann, C. F., 2011. How to be a specialist? Quantifying specialisation in pollination

networks. Network Biology, 1 (1), 1–20.

Dormann, C. F. and Blüthgen, N., 2017. Food webs versus interaction networks:

Principles, pitfalls, and perspectives. Adaptive Food Webs: Stability and

Transitions of Real and Model Ecosystems, 9–18.

Dormann, C. F., Fruend, J., Bluethgen, N., and Gruber, B., 2009. Indices, graphs and

null models: analyzing bipartite ecological networks. The Open Ecology Journal,

2, 7–24.

Dormann, C. F., Gruber, B., and Frund, J., 2008. Introducing the bipartite Package:

Analysing Ecological Networks. R News, 8 (2), 8–11.

Dunne, J. A., Williams, R. J., and Martinez, N. D., 2002. Food-web structure and

network theory: The role of connectance and size. Proceedings of the National

Academy of Sciences, 99 (20), 12917–12922.

Dyer, L. A., Singer, M. S., Till, J. T., Stireman, J. O., Gentry, G. L., Marquis, R. J.,

Ricklefs, R. E., Greeney, H. F., Wagner, D. L., Morais, H. C., Diniz, I. R., Kursar,

T. A., and Coley, P. D., 2007. Host specificity of Lepidoptera in tropical and

temperate forests. Nature, 448 (August), 696–700.

Edwards, D. P., Larsen, T. H., Docherty, T. D. S. S., Ansell, F. A., Hsu, W. W., Derhe,

M. A., Hamer, K. C., and Wilcove, D. S., 2011. Degraded lands worth protecting:

the biological importance of Southeast Asia’s repeatedly logged forests.

Proceedings of the Royal Society B: Biological Sciences, 278 (1702), 82–90.

Ehrlich, P. R. and Raven, P. H., 1964. Butterflies and Plants: a Study in Coevolution.

Evolution, 18 (4), 586–608.

161

Eichhorn, M. P., Compton, S. G., and Hartley, S. E., 2006. Seedling Species Determines

Rates of Leaf Herbivory in a Malaysian Rain Forest. Journal of Tropical Ecology,

22 (5), 513–519.

Eichhorn, M. P., Compton, S. G., and Hartley, S. E., 2008. The Influence of Soil Type

on Rain Forest Insect Herbivore Communities. Biotropica, 40 (6), 707–713.

Engelbrecht, B. M. J., Comita, L. S., Condit, R., Kursar, T. A., Tyree, M. T., Turner, B.

L., and Hubbell, S. P., 2007. Drought sensitivity shapes species distribution

patterns in tropical forests. Nature, 447 (7140), 80–82.

Eschenbach, C., Glauner, R., Kleine, M., and Kappen, L., 1998. Photosynthesis rates of

selected tree species in lowland dipterocarp rainforest of Sabah, Malaysia. Trees -

Structure and Function, 12, 356–365.

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., March, J. J., Nilus, R., Pfeifer, M., Rao,

S. V., Sharp, A. C., Snaddon, J. L., Stork, N. E., Struebig, M. J., Wearn, O. R.,

Yusah, K. M., and Turner, E. C., 2015. Logging cuts the functional importance of

invertebrates in tropical rainforest. Nature Communications, 6, 6836.

Ewers, R. M. and Didham, R. K., 2006. Confounding factors in the detection of species

responses to habitat fragmentation. Biological reviews of the Cambridge

Philosophical Society, 81 (1), 117–42.

Ewers, R. M., Didham, R. K., Fahrig, L., Ferraz, G., Hector, A., Holt, R. D., Kapos, V.,

Reynolds, G., Sinun, W., Snaddon, J. L., and Turner, E. C., 2011. A large-scale

162

forest fragmentation experiment: the Stability of Altered Forest Ecosystems

Project. Philosophical transactions of the Royal Society of London. Series B,

Biological sciences, 366 (1582), 3292–3302.

Fabian, Y., Sandau, N., Bruggisser, O. T., Aebi, A., Kehrli, P., Rohr, R. P., Naisbit, R.

E., and Bersier, L. F., 2013. The importance of landscape and spatial structure for

hymenopteran-based food webs in an agro-ecosystem. Journal of Animal Ecology,

82 (6), 1203–1214.

Fáveri, S. B., Vasconcelos, H. L., and Dirzo, R., 2008. Effects of Amazonian forest

fragmentation on the interaction between plants, insect herbivores, and their natural

enemies. Journal of Tropical Ecology, 24 (1), 57–64.

Fayle, T. M., Sam, K., Humlova, A., Cagnolo, L., and Novotny, V., 2016. The

LifeWebs project: A call for data describing plant-herbivore interaction networks.

Frontiers of Biogeography, 8 (4), 3–5.

Fayle, T. M., Turner, E. C., Basset, Y., Ewers, R. M., Reynolds, G., and Novotny, V.,

2015. Whole-ecosystem experimental manipulations of tropical forests. Trends in

Ecology & Evolution, 30 (6), 334–346.

Feeny, P., 1976. Plant apparency and chemical defense. Recent Advances in

Phytochemistry, 10, 1–40.

Forister, M. L., Novotny, V., Panorska, A. K., Baje, L., Basset, Y., Butterill, P. T.,

Cizek, L., Coley, P. D., Dem, F., Diniz, I. R., Drozd, P., Fox, M., Glassmire, A. E.,

Hazen, R., Hrcek, J., Jahner, J. P., Kaman, O., Kozubowski, T. J., Kursar, T. A.,

Lewis, O. T., Lill, J., Marquis, R. J., Miller, S. E., Morais, H. C., Murakami, M.,

Nickel, H., Pardikes, N. A., Ricklefs, R. E., Singer, M. S., Smilanich, A. M.,

163

Stireman, J. O., Villamarín-Cortez, S., Vodka, S., Volf, M., Wagner, D. L., Walla,

T., Weiblen, G. D., and Dyer, L. A., 2015. The global distribution of diet breadth

in insect herbivores. Proceedings of the National Academy of Sciences, 112 (2),

442–447.

Gardner, T. A., Barlow, J., Chazdon, R., Ewers, R. M., Harvey, C. A., Peres, C. A., and

Sodhi, N. S., 2009. Prospects for tropical forest biodiversity in a human-modified

world. Ecology Letters, 12 (6), 561–582.

Gaston, K. J., Reavey, D., and Valladares, G. R., 1992. Intimacy and fidelity: internal

and external feeding by the British microlepidoptera. Ecological Entomology, 17

(1), 86–88.

Gaveau, D. L. A., Sloan, S., Molidena, E., Yaen, H., Sheil, D., Abram, N. K., Ancrenaz,

M., Nasi, R., Quinones, M., Wielaard, N., and Meijaard, E., 2014. Four decades of

forest persistence, clearance and logging on Borneo. PLoS ONE, 9 (7), 1–11.

Ghazoul, J., 1997. The pollination and breeding system of dipterocarpus obtusifolius

(Dipterocarpaceae) in dry deciduous forests of Thailand. Journal of Natural

History, 31 (6), 901–916.

Ghazoul, J., 2016. Dipterocarp Biology, Ecology and Conservation. Oxford, United

Kingdom: Oxford University Press.

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., and Sodhi, N. S., 2011.

Primary forests are irreplaceable for sustaining tropical biodiversity. Nature, 478

(7369), 378–381.

Gómez-Zurita, J. and Galián, J., 2005. Current knowledge on genes and genomes of

164

phytophagous beetles (Coleoptera: Chrysomeloidea, Curculionoidea): A review.

European Journal of Entomology, 102 (4), 577–597.

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., Melbourne, B. A., Nicholls, A. O., Orrock, J.

L., Song, D.-X., and Townshend, J. R., 2015. Habitat fragmentation and its lasting

impact on Earth’s ecosystems. Science Advances, 1 (2), e1500052.

Hancock, P. A., Milner-Gulland, E. J., and Keeling, M. J., 2005. An individual based

model of bearded pig abundance. Ecological Modelling, 181 (2–3), 123–137.

Hardwick, S. R., Toumi, R., Pfeifer, M., Turner, E. C., Nilus, R., and 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.

Agricultural and Forest Meteorology, 201, 187–195.

Harrison, X. A., Donaldson, L., Correa-Cano, M. E., Evans, J., Fisher, D. N., Goodwin,

C. E. D., Robinson, B. S., Hodgson, D. J., and Inger, R., 2018. A brief introduction

to mixed effects modelling and multi-model inference in ecology. PeerJ, 6, e4794.

Hattori, D., Kenzo, T., Yamauchi, N., Irino, K. O., Kendawang, J. J., Ninomiya, I., and

Sakurai, K., 2013. Effects of environmental factors on growth and mortality of

Parashorea macrophylla (Dipterocarpaceae) planted on slopes and valleys in a

degraded tropical secondary forest in Sarawak, Malaysia. Soil Science and Plant

Nutrition, 59 (2), 218–228.

Hazebroek, H. P., Adlin, T. Z., and Sinun, W., 2012. Danum Valley: the rain forest.

165

Kota Kinabalu, Malaysia: Natural History Publications (Borneo).

Hector, A., Philipson, C., Saner, P., Chamagne, J., Dzulkifli, D., O’Brien, M., Snaddon,

J. L., Ulok, P., Weilenmann, M., Reynolds, G., and Godfray, H. C. J., 2011. The

Sabah Biodiversity Experiment: a long-term test of the role of tree diversity in

restoring tropical forest structure and functioning. Philosophical transactions of

the Royal Society of London. Series B, Biological sciences, 366 (1582), 3303–15.

Herms, D. A. and Mattson, W. J., 1992. The Dilemma of Plants: To Grow or Defend.

The Quarterly Review of Biology, 67 (3), 283–335.

Hooper, D. U., Adair, E. C., Cardinale, B. J., Byrnes, J. E. K., Hungate, B. A., Matulich,

K. L., Gonzalez, A., Duffy, J. E., Gamfeldt, L., and Connor, M. I., 2012. A global

synthesis reveals biodiversity loss as a major driver of ecosystem change. Nature,

486 (7401), 105–108.

Howe, H. F., 1990. Survival and Growth of Juvenile Virola surinamensis in Panama :

Effects of Herbivory and Canopy Closure. Journal of Tropical Ecology, 6 (3),

259–280.

Howlett, B. E. and Davidson, D. W., 2001. Herbivory on Planted Dipterocarp Seedlings

in Secondary Logged Forests and Primary Forests of Sabah, Malaysia. Journal of

Tropical Ecology, 17 (2), 285–302.

Hunt, J. H., 2003. Cryptic Herbivores of the Rainforet Canopy. Science, 300 (5621),

916–917.

Hunter, M. D., 2001a. Out of sight, out of mind: the impacts of root-feeding insects in

natural and managed systems. Agricultural and Forest Entomology, 3 (1), 3–9.

166

Hunter, M. D., 2001b. Insect population dynamics meets ecosystem ecology: effects of

herbivory on soil nutrient dynamics. Agricultural and Forest Entomology, 3 (2),

77–84.

Ibbotson, R., 2014. The History of Logging in North Borneo. Kota Kinabalu: Opus

Publications.

Ingersoll, C. M., Niesenbaum, R. A., Weigle, C. E., and Lehman, J. H., 2010. Total

phenolics and individual phenolic acids vary with light environment in Lindera

benzoin. Botany, 88 (11), 1007–1010.

Itoh, A., Yamakura, T., and Lee, H. S., 1999. Effects of light intensity on the growth

and allometry of two bornean dryobalanops (Dipterocarpaceae) seedlings. Journal

of Tropical Forest Science, 11 (3), 610–618.

Janzen, D., 1984. Two ways to be a tropical big moth: Santa Rosa saturniids and

sphingids. Oxford Surveys in Evolutionary Biology, 1.

Janzen, D. H., 1974. Tropical Blackwater Rivers, Animals, and Mast Fruiting by the

Dipterocarpaceae. Biotropica, 6 (2), 69–103.

Jennings, S., 1999. Assessing forest canopies and understorey illumination: canopy

closure, canopy cover and other measures. Forestry, 72 (1), 59–74.

Johns, A. D., 1988. Effects of ‘Selective’ Timber Extraction on Rain Forest Structure

and Composition and Some Consequences for Frugivores and Folivores.

Biotropica, 20 (1), 31–37.

Jucker, T., Boris, B., Burslem, D. F. R. P., Nilus, R., Dalponte, M., Lewis, S. L.,

Phillips, O. L., Qie, L., and Coomes, D. A., 2018. Topography shapes the structure,

167

composition and function of tropical forest landscapes. Ecology Letters.

Karban, R., Agrawal, A. A., Thaler, J. S., and Adler, L. S., 1999. Induced plant

responses and information content about risk of herbivory. Trends in Ecology and

Evolution, 14 (11), 443–447.

Kessler, A. and Baldwin, I. T., 2002. Plant Responses To Insect Herbivory: The

Emerging Molecular Analysis. Annual Review of Plant Biology, 53 (1), 299–328.

Kettle, C. J., 2010. Ecological considerations for using dipterocarps for restoration of

lowland rainforest in Southeast Asia. Biodiversity and Conservation, 19 (4), 1137–

1151.

Kettle, C. J., Maycock, C. R., and Burslem, D., 2012. New Directions in Dipterocarp

Biology and Conservation: A Synthesis. Biotropica, 44 (5), 658–660.

Kozlov, M. V., Lanta, V., Zverev, V., and Zvereva, E. L., 2015. Global patterns in

background losses of woody plant foliage to insects. Global Ecology and

Biogeography, 24 (10), 1126–1135.

Krause, A. E., Frank, K. A., Mason, D. M., Ulanowicz, R. E., and Taylor, W. W., 2003.

Compartments revealed in food-web structure. Nature, 426 (6964), 282–285.

Krenn, H. W., 2010. Feeding Mechanisms of Adult Lepidoptera: Structure, Function,

and Evolution of the Mouthparts. Annual Review of Entomology, 55 (1), 307–327.

Kursar, T. A. and Coley, P. D., 2003. Convergence in defense syndromes of young

leaves in tropical rainforests. Biochemical Systematics and Ecology, 31 (8), 929–

949.

Kursar, T. a, Dexter, K. G., Lokvam, J., Pennington, R. T., Richardson, J. E., Weber, M.

168

G., Murakami, E. T., Drake, C., McGregor, R., and Coley, P. D., 2009. The

evolution of antiherbivore defenses and their contribution to species coexistence in

the tropical tree genus Inga. Proceedings of the National Academy of Sciences of

the United States of America, 106 (43), 18073–18078. de la Cruz, M. and Dirzo, R., 1987. A survey of the standing levels of herbivory in

seedlings from a Mexican rain forest. Biotropica, 19 (2), 98–106.

De La Vega, X., Grez, A. a., and Simonetti, J. a., 2012. Is top-down control by

predators driving insect abundance and herbivory rates in fragmented forests?

Austral Ecology, 37 (7), 836–844.

LaFrankie, J. V., 2010. Trees of tropical Asia: an illustrated guide. Phillipines: Black

Tree Publications.

Lamarre, G. P. A., Mendoza, I., Fine, P. V. A., and Baraloto, C., 2014. Leaf synchrony

and insect herbivory among tropical tree habitat specialists. Plant Ecology, 215 (2),

209–220.

Langner, A. and Siegert, F., 2009. Spatiotemporal fire occurrence in Borneo over a

period of 10 years. Global Change Biology, 15 (1), 48–62.

Laurance, W. F., 2008. Theory meets reality: How habitat fragmentation research has

transcended island biogeographic theory. Biological Conservation, 141 (7), 1731–

1744.

Laurance, W. F., Camargo, J. L. C., Luizão, R. C. C., Laurance, S. G., Pimm, S. L.,

Bruna, E. M., Stouffer, P. C., Williamson, G. B., Benítez-malvido, J., Vasconcelos,

H. L., Van Houtan, K. S., Zartman, C. E., Boyle, S. A., Didham, R. K., Andrade,

A., and Lovejoy, T. E., 2011. The fate of Amazonian forest fragments: A 32-year

169

investigation. Biological Conservation, 144 (1), 56–67.

Laurance, W. F., Sayer, J., and Cassman, K. G., 2014. Agricultural expansion and its

impacts on tropical nature. Trends in Ecology and Evolution, 29 (2), 107–116.

Lawrence, R., Potts, B. M., and Whitham, T. G., 2003. Relative importance of plant

ontogeny, host genetic variation, and leaf age for a common herbivore. Ecology, 84

(5), 1171–1178.

Leles, B., Xiao, X., Pasion, B. O., Nakamura, A., and Tomlinson, K. W., 2017. Does

plant diversity increase top–down control of herbivorous insects in tropical forest?

Oikos, 126 (8), 1142–1149.

Lemmon, P. E., 1956. A Spherical Densiometer For Estimating Forest Overstory

Density. Forest Science, 2 (4), 314–320.

Levey, D. J., Caughlin, T. T., Brudvig, L. A., Haddad, N. M., Damschen, E. I.,

Tewksbury, J. J., and Evans, D. M., 2016. Disentangling fragmentation effects on

herbivory in understory plants of longleaf pine savanna. Ecology, 97 (9), 2248–

2258.

Lewis, O. T. and Basset, Y., 2009. Insect conservation in tropical forests. Insect

conservation biology: Proceedings of the Royal Entomological Society’s 23nd

Symposium. CABI.

Lewis, O. T., Memmott, J., Lasalle, J., Lyal, C. H. C., Whitefoord, C., and Godfray, H.

C. J., 2002. Structure of a diverse tropical forest insect – parasitoid. Journal of

Animal Ecology, 71, 855–873.

Lewis, S. L., Edwards, D. P., and Galbraith, D., 2015. Increasing human dominance of

170

Tropical Forests. Science, 349 (6250), 827–832.

Lowman, M. D., 1984. An Assessment of Techniques for Measuring Herbivory: Is

Rainforest Defoliation More Intense Than we Thought? Biotropica, 16 (4), 264–

268.

Lowman, M. D. and Box, J. D., 1983. Variation in leaf toughness and phenolic content

among five species of Australian rain forest trees. Australian Journal of Ecology, 8

(1), 17–25.

Lucas, P. W., Turner, I. M., Dominy, N. J., and Yamashita, N., 2000. Mechanical

defences to herbivory. Annals of Botany, 86 (5), 913–920.

Maguire, D. Y., Buddle, C. M., and Bennett, E. M., 2016. Within and among patch

variability in patterns of insect herbivory across a fragmented forest landscape.

PLoS ONE, 11 (3), e0150843.

Maguire, D. Y., James, P. M. A., Buddle, C. M., and Bennett, E. M., 2015. Landscape

connectivity and insect herbivory: A framework for understanding tradeoffs among

ecosystem services. Global Ecology and Conservation, 4, 73–84.

Malhi, Y., Gardner, T. A., Goldsmith, G. R., Silman, M. R., and Zelazowski, P., 2014.

Tropical Forests in the Anthropocene. Annual Review of Environment and

Resources, 39 (1), 125–159.

Mali, S. and Borges, R. M., 2003. Phenolics, fibre, alkaloids, saponins, and cyanogenic

glycosides in a seasonal cloud forest in India. Biochemical Systematics and

Ecology, 31 (11), 1221–1246.

Malishev, M. and Sanson, G. D., 2015. Leaf mechanics and herbivory defence : How

171

tough tissue along the leaf body deters growing insect herbivores. Austral Ecology,

1–9.

Mani, M. S., 1964. Ecology of plant galls. Dordrecht: Springer.

Marsh, C. J. and Ewers, R. M., 2013. A fractal-based sampling design for ecological

surveys quantifying β-diversity. Methods in Ecology and Evolution, 4 (1), 63–72.

Marsh, C. W. and Greer, A. G., 1992. Forest land-use in Sabah, Malaysia: an

introduction to Danum Valley. Philosophical Transactions of the Royal Society B:

Biological Sciences, 335 (1275), 331–339.

Martínez-Garza, C. and Howe, H. F., 2005. Developmental strategy or immediate

responses in leaf traits of tropical tree species? International journal of plant

sciences, 166 (1), 41–48.

Massey, F. P., Massey, K., Press, M. C., and Hartley, S. E., 2006. Neighbourhood

Composition Determines Growth, Architecture and Herbivory in Tropical Rain

Forest Tree Seedlings. Journal of Ecology, 94 (3), 646–655.

Matsuda, I., Clauss, M., Tuuga, A., Sugau, J., Hanya, G., Yumoto, T., Bernard, H., and

Hummel, J., 2017. Factors Affecting Leaf Selection by Foregut-fermenting

Proboscis Monkeys: New Insight from in vitro Digestibility and Toughness of

Leaves. Scientific Reports, 7 (February), 1–10.

Maunsell, S. C., Kitching, R. L., Burwell, C. J., and Morris, R. J., 2015. Changes in

host-parasitoid food web structure with elevation. Journal of animal ecology, 84

(2), 353–363.

Maycock, C. R., Kettle, C. J., Khoo, E., Pereira, J. T., Sugau, J. B., Nilus, R., Ong, R.

172

C., Amaludin, N. A., Newman, M. F., and Burslem, D. F. R. P., 2012. A Revised

Conservation Assessment of Dipterocarps in Sabah. Biotropica, 44 (5), 649–657.

McDonald-Madden, E., Sabbadin, R., Game, E. T., Baxter, P. W. J., Chadès, I., and

Possingham, H. P., 2016. Using food-web theory to conserve ecosystems. Nature

Communications, 7 (May 2015), 1–8.

Meijaard, E. and Sheil, D., 2007. A logged forest in Borneo is better than none at all.

Nature, 446 (7139), 974.

Meijer, W., 1974. Plant Geographic Studies on Dipterocarpaceae in Malesia. Annals of

the Missouri Botanical Garden, 61 (3), 806–818.

Menke, S., Böhning-Gaese, K., and Schleuning, M., 2012. Plant-frugivore networks are

less specialized and more robust at forest-farmland edges than in the interior of a

tropical forest. Oikos, 121 (10), 1553–1566.

Metcalfe, D. B., Asner, G. P., Martin, R. E., Silva Espejo, J. E., Huasco, W. H., Farfán

Amézquita, F. F., Carranza-Jimenez, L., Galiano Cabrera, D. F., Baca, L. D.,

Sinca, F., Huaraca Quispe, L. P., Taype, I. A., Mora, L. E., Dávila, A. R.,

Solórzano, M. M., Puma Vilca, B. L., Laupa Román, J. M., Guerra Bustios, P. C.,

Revilla, N. S., Tupayachi, R., Girardin, C. A. J., Doughty, C. E., and Malhi, Y.,

2014. Herbivory makes major contributions to ecosystem carbon and nutrient

cycling in tropical forests. Ecology Letters, 17 (3), 324–332.

Millennium Ecosystem Assessment, 2005. Millennium Ecosystem Assessment: Current

state and trends assessment. Washington DC: Island Press.

Momose, K., Yumoto, T., Nagamitsu, T., Kato, M., Momose, K., Yumoto, T.,

Nagamitsu, T., Kato, M., Nagamasu, H., Sakai, S., Rhett, D., Itioka, T., Hamid, A.

173

A., and Inoue, T., 1998. Pollination Biology in a Lowland Dipterocarp Forest in

Sarawak , Malaysia . I . Characteristics of the Plant-Pollinator Community in a

Lowland Dipterocarp Forest Hidetoshi Nagamasu , Shoko Sakai , Rhett D .

Harrison , Takao Itioka , Abang A . Hamid and Ta. American Journal of Botany,

85 (10), 1477–1501.

Morales-Castilla, I., Matias, M. G., Gravel, D., and Araújo, M. B., 2015. Inferring biotic

interactions from proxies. Trends in Ecology and Evolution, 30 (6), 347–356.

Morante-Filho, J. C., Arroyo-Rodríguez, V., Lohbeck, M., Tscharntke, T., and Faria, D.,

2016. Tropical forest loss and its multitrophic effects on insect herbivory. Ecology,

97 (12), 3315–3325.

Moreno, M. L., Rossetti, M. R., Pérez-Harguindeguy, N., and Valladares, G. R., 2017.

Edge and herbivory effects on leaf litter decomposability in a subtropical dry

forest. Ecological Research, 32 (3), 341–346.

Murcia, C., 1995. Edge effects in fragmented forests: implications for conservation.

Trends in ecology & evolution, 10 (2), 58–62.

Murphy, S. M., Battocletti, A. H., Tinghitella, R. M., Wimp, G. M., and Ries, L., 2016.

Complex community and evolutionary responses to habitat fragmentation and

habitat edges: What can we learn from insect science? Current Opinion in Insect

Science, 14, 61–65.

Myers, N., Mittermeier, R. A., Mittermeier, C. G., Fonseca, G. A. B., and Kent, J.,

2000. Biodiversity hotspots for conservation priorities. Nature, 403, 853–858.

Nagamitsu, T., Harrison, R. D., and Tamiji, I., 1999. Beetle Pollination of Vatica

parvifolia (Dipterocarpaceae) in Sarawak, Malaysia. Gardens’ Bulletin Singapore,

174

51, 43–54.

Nakagawa, M., Itioka, T., Momose, K., Yumoto, T., Komai, F., Morimoto, K., Jordal,

B. H., Kato, M., Kaliang, H., Hamid, A. A., Inoue, T., and Nakashizuka, T., 2003.

Resource use of insect seed predators during general flowering and seeding events

in a Bornean dipterocarp rain forest. Bulletin of Entomological Research, 93 (05),

455–466.

Nakazawa, T., Yamanaka, T., and Urano, S., 2012. Model analysis for plant disease

dynamics co-mediated by herbivory and herbivore-borne phytopathogens. Biology

Letters, 8 (4), 685–688.

Newbery, D. M., Campbell, E. J. F., Proctor, J., and Still, M. J., 1996. Primary lowland

dipterocarp forest at Danum Valley, Sabah, Malaysia. Species composition and

pattern in the understorey. Vegetatio, 122 (2), 193–220.

Norghauer, J. M., Malcolm, J. R., and Zimmerman, B. L., 2008. Canopy cover mediates

interactions between a specialist caterpillar and seedlings of a neotropical tree.

Journal of Ecology, 96 (1), 103–113.

Novotny, V., Basset, Y., Miller, S. E., Kitching, R. L., Laidlaw, M., Drozd, P., and

Cizek, L., 2004. Local Species Richness of Leaf-Chewing Insects Feeding on

Woody Plants from One Hectare of a Lowland Rainforest. Conservation Biology,

18 (1), 227–237.

Novotny, V., Miller, S. E., Baje, L., Balagawi, S., Basset, Y., Cizek, L., Craft, K. J.,

Dem, F., Drew, R. A. I., Hulcr, J., Leps, J., Lewis, O. T., Pokon, R., Stewart, A. J.

A., Allan Samuelson, G., and Weiblen, G. D., 2010. Guild-specific patterns of

species richness and host specialization in plant-herbivore food webs from a

175

tropical forest. Journal of Animal Ecology, 79 (6), 1193–1203.

Novotny, V., Miller, S. E., Hrcek, J., Baje, L., Basset, Y., Lewis, O. T., Stewart, A. J. a,

and Weiblen, G. D., 2012. Insects on plants: explaining the paradox of low

diversity within specialist herbivore guilds. The American naturalist, 179 (3), 351–

62.

Nussbaum, R., Anderson, J., and Spencer, T., 1995. Factors limiting the growth of

indigenous tree seedlings planted on degraded rainforest soils in Sabah, Malaysia.

Forest Ecology and Management, 74 (1–3), 149–159.

O’Neal, M. E., Landis, D. a, and Isaacs, R., 2002. An inexpensive, accurate method for

measuring leaf area and defoliation through digital image analysis. Journal of

economic entomology, 95 (6), 1190–1194.

Olesen, J. M., Bascompte, J., Dupont, Y. L., Elberling, H., Rasmussen, C., and Jordano,

P., 2011. Missing and forbidden links in mutualistic networks. Proceedings of the

Royal Society B: Biological Sciences, 278 (1706), 725–732.

Onoda, Y., Westoby, M., Adler, P. B., Choong, A. M. F., Clissold, F. J., Cornelissen, J.

H. C., Díaz, S., Dominy, N. J., Elgart, A., Enrico, L., Fine, P. V. a, Howard, J. J.,

Jalili, A., Kitajima, K., Kurokawa, H., McArthur, C., Lucas, P. W., Markesteijn,

L., Pérez-Harguindeguy, N., Poorter, L., Richards, L., Santiago, L. S., Sosinski, E.

E., Van Bael, S. a., Warton, D. I., Wright, I. J., Joseph Wright, S., and Yamashita,

N., 2011. Global patterns of leaf mechanical properties. Ecology Letters, 14 (3),

301–312.

Paletto, A. and Tosi, V., 2009. Forest canopy cover and canopy closure: Comparison of

assessment techniques. European Journal of Forest Research, 128 (3), 265–272.

176

Peeters, P. J., Sanson, G., and Read, J., 2007. Leaf biomechanical properties and the

densities of herbivorous insect guilds. Functional Ecology, 21 (2), 246–255.

Peschiutta, M. L., Scholz, F. G., and Goldstein, G., 2017. Herbivory alters plant carbon

assimilation, patterns of biomass allocation and nitrogen use efficiency. Acta

Oecologica, 86, 9–16.

Pfeifer, M., Lefebvre, V., Turner, E., Cusack, J., Khoo, M., Chey, V. K., Peni, M., and

Ewers, R. M., 2015. Deadwood biomass: an underestimated carbon stock in

degraded tropical forests? Environmental Research Letters, 10, 044019.

Phua, M.-H., Su Wah, H., Ioki, K., Hashim, M., Bidin, K., Musta, B., Suleiman, M.,

Yap, S. W., and Maycock, C. R., 2016. Estimating Logged-Over Lowland

Rainforest Aboveground Biomass in Sabah, Malaysia Using Airborne LiDAR

Data. Terrestrial Atmospheric and Oceanic Sciences, 27 (4), 481–489.

Pimm, S. L., Lawton, J. H., and Cohen, J. E., 1991. Food web patterns and their

consequences. Nature, 350, 669–674.

Pinheiro, J., Bates, D., DebRoy, S., Sarkar, D., and Team, R. C., 2018. nlme: Linear

and Nonlinear Mixed Effects Models. [online]. R package version 3.1-137,.

Available from: https://cran.r-project.org/package=nlme.

Piper, F. I., Altmann, S. H., and Lusk, C. H., 2018. Global patterns of insect herbivory

in gap and understory environments, and their implications for woody plant carbon

storage. Oikos, 127 (4), 483–496.

Poisot, T., Stouffer, D. B., and Kéfi, S., 2016. Describe, understand and predict: why do

we need networks in ecology? Functional Ecology, 30 (12), 1878–1882.

177

Price, P. W., 2002. Resource-driven terrestrial interaction webs. Ecological Research,

17 (2), 241–247.

R Core Team, 2018. R: A language and environment for statistical computing. Vienna:

R Foundation for Statistical Computing. [online]. Available from: https://www.r-

project.org/.

Rao, M., Terborgh, J., and Nuňez, P., 2001. Increased Herbivory in Forest Isolates:

Implications for Plant Community Structure and Composition. Conservation

Biology, 15 (3), 624–633.

Read, J. and Sanson, G. D., 2003. Characterizing sclerophylly: The mechanical

properties of a diverse range of leaf types. New Phytologist, 160 (1), 81–99.

Regier, J. C., Zwick, A., Cummings, M. P., Kawahara, A. Y., Cho, S., Weller, S., Roe,

A., Baixeras, J., Brown, J. W., Parr, C., Davis, D. R., Epstein, M., Hallwachs, W.,

Hausmann, A., Janzen, D. H., Kitching, I. J., Solis, M. A., Yen, S. H., Bazinet, A.

L., and Mitter, C., 2009. Toward reconstructing the evolution of advanced moths

and butterflies (Lepidoptera: Ditrysia): An initial molecular study. BMC

Evolutionary Biology, 9 (280), 1–21.

Riutta, T., Malhi, Y., Kho, L. K., Marthews, T. R., Huaraca Huasco, W., Khoo, M. S.,

Tan, S., Turner, E., Reynolds, G., Both, S., Burslem, D. F. R. P., Teh, Y. A.,

Vairappan, C. S., Majalap, N., and Ewers, R. M., 2018. Logging disturbance shifts

net primary productivity and its allocation in Bornean tropical forests. Global

Change Biology, 24 (7), 2913–2928.

Robinson, G. S., Ackery, P. R., Kitching, I. J., Beccaloni, G. W., and Hernandez, L. M.,

2001. Hostplants of the Moth and Butterfly Caterpillars of the Oriental Region.

178

London, UK: Natural History Museum.

Rossetti, M. R., González, E., Salvo, A., and Valladares, G., 2014. Not all in the same

boat: trends and mechanisms in herbivory responses to forest fragmentation differ

among insect guilds. Arthropod-Plant Interactions, 8 (6), 593–603.

Rossetti, M. R., Tscharntke, T., Aguilar, R., and Batáry, P., 2017. Responses of insect

herbivores and herbivory to habitat fragmentation: a hierarchical meta-analysis.

Ecology Letters, 20 (2), 264–272.

Rozendaal, D. M. A., Hurtado, V. H., and Poorter, L., 2006. Plasticity in leaf traits of 38

tropical tree species in response to light; relationships with light demand and adult

stature. Functional Ecology, 20 (2), 207–216.

Ruiz-Guerra, B., Guevara, R., Mariano, N. A., and Dirzo, R., 2010. Insect herbivory

declines with forest fragmentation and covaries with plant regeneration mode:

evidence from a Mexican tropical rain forest. Oikos, 119 (2), 317–325.

Ruiz-Guerra, B., Guevara, R., Velázquez-Rosas, N., and Dirzo, R., 2016. Does tropical

forest fragmentation affect plant anti-herbivore defensive and nutritional traits?

Journal of Tropical Ecology, 32 (2), 162–164.

Ruiz-Guerra, B., Renton, K., and Dirzo, R., 2012. Consequences of Fragmentation of

Tropical Moist Forest for Birds and Their Role in Predation of Herbivorous

Insects. Biotropica, 44 (2), 228–236.

Sakai, S., Momose, K., Yumoto, T., Kato, M., and Inoue, T., 1999. Beetle pollination of

Shorea parvifolia (section Mutica, Dipterocarpaceae) in a general flowering period

in Sarawak, Malaysia. American Journal of Botany, 86 (1), 62–69.

179

Sakai, S., Momose, K., Yumoto, T., Nagamitsu, T., Nagamasu, H., Hamid, A. A., and

Nakashizuka, T., 1999. Plant reproductive phenology over four years including an

episode of general flowering in a lowland dipterocarp forest, Sarawak, Malaysia.

American Journal of Botany, 86 (10), 1414–1436.

Salazar, D. and Marquis, R. J., 2012. Herbivore pressure increases toward the equator.

Proceedings of the National Academy of Sciences, 109 (31), 12616–12620.

Sanson, G., Read, J., Aranwela, N., Clissold, F., and Peeters, P., 2001. Measurement of

leaf biomechanical properties in studies of herbivory: Opportunities, problems and

procedures. Austral Ecology, 26, 535–546.

Schielzeth, H., 2010. Simple means to improve the interpretability of regression

coefficients. Methods in Ecology and Evolution, 1, 103–113.

Schlinkert, H., Westphal, C., Clough, Y., László, Z., Ludwig, M., and Tscharntke, T.,

2015. Plant size as determinant of species richness of herbivores, natural enemies

and pollinators across 21 brassicaceae species. PLoS ONE, 10 (8), 1–18.

Scholes, J. D., Press, M. C., and Zipperlen, S. W., 1997. Differences in light energy

utilisation and dissipation between dipterocarp rain forest tree seedlings.

Oecologia, 109 (1), 41–48.

Sharp, A. C., 2018. Diversity of beetles across a land-use gradient in Sabah, Malaysia.

PhD Dissertation. Imperial College London.

Shorthouse, J. D. and Rohfritsch, O., 1992. Biology of Insect-Induced Galls. New York:

Oxford University Press.

Silva, J. O., Espírito-Santo, M. M., and Morais, H. C., 2015. Leaf traits and herbivory

180

on deciduous and evergreen trees in a tropical dry forest. Basic and Applied

Ecology, 16 (3), 210–219.

Sinclair, R. J. and Hughes, L., 2010. Leaf miners: The hidden herbivores. Austral

Ecology, 35 (3), 300–313.

Sist, P., Fimbel, R., Sheil, D., Nasi, R., and Chevallier, M. H., 2003. Towards

sustainable management of mixed dipterocarp forests of Southeast Asia: Moving

beyond minimum diameter cutting limits. Environmental Conservation, 30 (4),

364–374.

Slik, J. W. F., Arroyo-Rodríguez, V., Aiba, S.-I., Alvarez-Loayza, P., Alves, L. F.,

Ashton, P., Balvanera, P., Bastian, M. L., Bellingham, P. J., van den Berg, E.,

Bernacci, L., da Conceição Bispo, P., Blanc, L., Böhning-Gaese, K., Boeckx, P.,

Bongers, F., Boyle, B., Bradford, M., Brearley, F. Q., Breuer-Ndoundou

Hockemba, M., Bunyavejchewin, S., Calderado Leal Matos, D., Castillo-Santiago,

M., Catharino, E. L. M., Chai, S.-L., Chen, Y., Colwell, R. K., Chazdon, R. L.,

Clark, C., Clark, D. B., Clark, D. A., Culmsee, H., Damas, K., Dattaraja, H. S.,

Dauby, G., Davidar, P., DeWalt, S. J., Doucet, J.-L., Duque, A., Durigan, G.,

Eichhorn, K. A. O., Eisenlohr, P. V., Eler, E., Ewango, C., Farwig, N., Feeley, K.

J., Ferreira, L., Field, R., de Oliveira Filho, A. T., Fletcher, C., Forshed, O.,

Franco, G., Fredriksson, G., Gillespie, T., Gillet, J.-F., Amarnath, G., Griffith, D.

M., Grogan, J., Gunatilleke, N., Harris, D., Harrison, R., Hector, A., Homeier, J.,

Imai, N., Itoh, A., Jansen, P. A., Joly, C. A., de Jong, B. H. J., Kartawinata, K.,

Kearsley, E., Kelly, D. L., Kenfack, D., Kessler, M., Kitayama, K., Kooyman, R.,

Larney, E., Laumonier, Y., Laurance, S., Laurance, W. F., Lawes, M. J., Amaral, I.

L. do, Letcher, S. G., Lindsell, J., Lu, X., Mansor, A., Marjokorpi, A., Martin, E.

181

H., Meilby, H., Melo, F. P. L., Metcalfe, D. J., Medjibe, V. P., Metzger, J. P.,

Millet, J., Mohandass, D., Montero, J. C., de Morisson Valeriano, M., Mugerwa,

B., Nagamasu, H., Nilus, R., Ochoa-Gaona, S., Onrizal, Page, N., Parolin, P.,

Parren, M., Parthasarathy, N., Paudel, E., Permana, A., Piedade, M. T. F., Pitman,

N. C. A., Poorter, L., Poulsen, A. D., Poulsen, J., Powers, J., Prasad, R. C.,

Puyravaud, J.-P., Razafimahaimodison, J.-C., Reitsma, J., dos Santos, J. R.,

Roberto Spironello, W., Romero-Saltos, H., Rovero, F., Rozak, A. H.,

Ruokolainen, K., Rutishauser, E., Saiter, F., Saner, P., Santos, B. A., Santos, F.,

Sarker, S. K., Satdichanh, M., Schmitt, C. B., Schöngart, J., Schulze, M.,

Suganuma, M. S., Sheil, D., da Silva Pinheiro, E., Sist, P., Stevart, T., Sukumar,

R., Sun, I.-F., Sunderland, T., Suresh, H. S., Suzuki, E., Tabarelli, M., Tang, J.,

Targhetta, N., Theilade, I., Thomas, D. W., Tchouto, P., Hurtado, J., Valencia, R.,

van Valkenburg, J. L. C. H., Van Do, T., Vasquez, R., Verbeeck, H., Adekunle, V.,

Vieira, S. A., Webb, C. O., Whitfeld, T., Wich, S. A., Williams, J., Wittmann, F.,

Wöll, H., Yang, X., Adou Yao, C. Y., Yap, S. L., Yoneda, T., Zahawi, R. A.,

Zakaria, R., Zang, R., de Assis, R. L., Garcia Luize, B., and Venticinque, E. M.,

2015. An estimate of the number of tropical tree species. Proceedings of the

National Academy of Sciences, 112 (24), 7472–7477.

Slik, J. W. F., Franklin, J., Arroyo-Rodríguez, V., Field, R., Aguilar, S., Aguirre, N.,

Ahumada, J., Aiba, S.-I., Alves, L. F., Anitha, K., Avella, A., Mora, F., Aymard

C., G. A., Báez, S., Balvanera, P., Bastian, M. L., Bastin, J.-F., Bellingham, P. J.,

van den Berg, E., da Conceição, P., Boeckx, P., Boehning-Gaese, K., Bongers, F.,

Boyle, B., Brambach, F., Brearley, F. Q., Brown, S., Chai, S.-L., Chazdon, R. L.,

Chen, S., Chhang, P., Chuyong, G., Ewango, C., Coronado, I. M., Cristóbal-

Azkarate, J., Culmsee, H., Damas, K., Dattaraja, H. S., Davidar, P., Dewalt, S. J.,

182

Din, H., Drake, D. R., Duque, A., Durigan, G., Eichhorn, K., Eler, E. S., Enoki, T.,

Ensslin, A., Fandohan, A. B., Farwig, N., Feeley, K. J., Fischer, M., Forshed, O.,

Souza Garcia, Q., Garkoti, S. C., Gillespie, T. W., Gillet, J.-F., Gonmadje, C.,

Granzow-de la Cerda, I., Griffith, D. M., Grogan, J., Rehman, K., Harris, D. J.,

Harrison, R. D., Hector, A., Hemp, A., Homeier, J., Hussain, M. S., Ibarra-

Manríquez, G., Hanum, I. F., Imai, N., Jansen, P. A., Joly, C. A., Joseph, S.,

Kartawinata, K., Kearsley, E., Kelly, D. L., Kessler, M., Killeen, T. J., Kooyman,

R. M., Laumonier, Y., Laurance, S. G., Laurance, W. F., Lawes, M. J., Letcher, S.

G., Lindsell, J., Lovett, J., Lozada, J., Lu, X., Lykke, A. M., Mahmud, K. Bin,

Diana, N. P., Mansor, A., Marshall, A. R., Martin, E. H., Calderado Leal Matos,

D., Meave, J. A., Melo, F. P. L., Huberto, Z., Mendoza, A., Metali, F., Medjibe, V.

P., Metzger, J. P., Metzker, T., Mohandass, D., Munguía-Rosas, M. A., Muñoz, R.,

Nurtjahy, E., de Oliveira, E. L., Parolin, P., Parren, M., Parthasarathy, N., Paudel,

E., Perez, R., Pérez-garcía, E. A., Pommer, U., Poorter, L., Qie, L., Piedade, M. T.

F., Roberto, J., Pinto, R., Dalberg Poulsen, A., Poulsen, J. R., Powers, J. S.,

Chandra Prasad, R., Puyravaud, J.-P., Rangel, O., Reitsma, J., Rocha, D. S. B.,

Rolim, S., Rovero, F., Rozak, A., Ruokolainen, K., Rutishauser, E., Rutten, G.,

Mohd. Said, M. N., Saiter, F. Z., Saner, P., Santos, B., do Santos, R. J., Sarker, S.

K., Schmitt, C. B., Schoengart, J., Schulze, M., Sheil, D., Sist, P., Souza, A. F.,

Spironello, W. R., Sposito, T., Steinmetz, R., Stevart, T., Suganuma, M. S., Sukri,

R., Sultana, A., Sukumar, R., Sunderland, T., Suresh, S. H. S., Suzuki, E.,

Tabarelli, M., Tang, J., Tanner, E. V. J., Targhetta, N., Theilade, I., Thomas, D.,

Timberlake, J., de Morisson Valeriano, M., van Valkenburg, J., Van Do, T., Van

Sam, H., Vandermeer, J. H., Verbeeck, H., Vetaas, O. R., Adekunle, V., Vieira, S.

A., Webb, C. O., Webb, E. L., Whitfeld, T., Wich, S., Williams, J., Wiser, S.,

183

Wittmann, F., Yang, X., Yao, C. Y. A., and Yap, S. L., 2018. Phylogenetic

classification of the world’s tropical forests. Proceedings of the National Academy

of Sciences, 115 (13), 1837–1842.

Slik, J. W. F., Poulsen, A. D., Ashton, P. S., Cannon, C. H., Eichhorn, K. A. O.,

Kartawinata, K., Lanniari, I., Nagamasu, H., Nakagawa, M., Van Nieuwstadt, M.

G. L., Payne, J., Purwaningsih, Saridan, A., Sidiyasa, K., Verburg, R. W., Webb,

C. O., and Wilkie, P., 2003. A Floristic Analysis of the Lowland Dipterocarp

Forests of Borneo. Journal of Biogeography, 30 (10), 1517–1531.

Smith, A. C., Koper, N., Francis, C. M., and Fahrig, L., 2009. Confronting collinearity:

Comparing methods for disentangling the effects of habitat loss and fragmentation.

Landscape Ecology, 24 (10), 1271–1285.

Sodhi, N. S., Koh, L. P., Brook, B. W., and Ng, P. K. L., 2004. Southeast Asian

biodiversity: an impending disaster. Trends in Ecology & Evolution, 19 (12), 654–

660.

Sodhi, N. S., Posa, M. R. C., Lee, T. M., Bickford, D., Koh, L. P., and Brook, B. W.,

2010. The state and conservation of Southeast Asian biodiversity. Biodiversity and

Conservation, 19 (2), 317–328.

Souza, D. G., Santos, B. A., Wirth, R., Leal, I. R., and Tabarelli, M., 2013. Community-

Level Patterns of Insect Herbivory in a Fragmented Atlantic Forest Landscape.

Environmental Entomology, 42 (3), 430–437.

Souza, D. G., Sfair, J. C., Arroyo-Rodríguez, V., Santos, J. C., and Tabarelli, M., 2018.

Depauperation and divergence of plant-specialist herbivore assemblages in a

fragmented tropical landscape. Ecological Entomology.

184

Steffan-Dewenter, I., Kessler, M., Barkmann, J., Bos, M. M., Buchori, D., Erasmi, S.,

Faust, H., Gerold, G., Glenk, K., Gradstein, S. R., Guhardja, E., Harteveld, M.,

Hertel, D., Hohn, P., Kappas, M., Kohler, S., Leuschner, C., Maertens, M.,

Marggraf, R., Migge-Kleian, S., Mogea, J., Pitopang, R., Schaefer, M., Schwarze,

S., Sporn, S. G., Steingrebe, A., Tjitrosoedirdjo, S. S., Tjitrosoemito, S., Twele, A.,

Weber, R., Woltmann, L., Zeller, M., and Tscharntke, T., 2007. Tradeoffs between

income, biodiversity, and ecosystem functioning during tropical rainforest

conversion and agroforestry intensification. Proceedings of the National Academy

of Sciences, 104 (12), 4973–4978.

Stiegel, S., Entling, M. H., and Mantilla-Contreras, J., 2017. Reading the leaves’ palm:

Leaf traits and herbivory along the microclimatic gradient of forest layers. PLoS

ONE, 12 (1), 1–17.

Stork, N. E., 2018. How Many Species of Insects and Other Terrestrial Arthropods Are

There on Earth? Annual Review of Entomology, 63 (1), 31–45.

Struebig, M. J., Turner, A., Giles, E., Lasmana, F., Tollington, S., Bernard, H., and Bell,

D., 2013. Quantifying the Biodiversity Value of Repeatedly Logged Rainforests.

Gradient and Comparative Approaches from Borneo. 1st ed. Advances in

Ecological Research. Elsevier Ltd.

Suzuki, S., Kitayama, K., Aiba, S.-I., Takyu, M., and Kikuzawa, K., 2013. Annual leaf

loss caused by folivorous insects in tropical rain forests on Mt. Kinabalu, Borneo.

Journal of Forest Research, 18 (4), 353–360.

Symonds, M. R. E. and Moussalli, A., 2011. A brief guide to model selection,

multimodel inference and model averaging in behavioural ecology using Akaike’s

185

information criterion. Behavioral Ecology and Sociobiology, 65, 13–21.

Tange, T., Harada, K., Kojima, K., and Sasaki, S., 1998. Responses of three dipterocarp

species to light regime. Proceedings of the Japan Academy. Ser. B: Physical and

Biological Sciences, 74 (B), 206–209.

Terborgh, J., Lopez, L., Nuñez, P., Rao, M., Shahabuddin, G., Orihuela, G., Riveros,

M., Ascanio, R., Adler, G. H., Lambert, T. D., and Balbas, L., 2001. Ecological

meltdown in predator-free forest fragments. Science, 294 (5548), 1923–1926. den Tex, R.-J., Thorington, R., Maldonado, J. E., and Leonard, J. A., 2010. Speciation

dynamics in the SE Asian tropics: Putting a time perspective on the phylogeny and

biogeography of Sundaland tree squirrels, Sundasciurus. Molecular Phylogenetics

and Evolution, 55 (2), 711–720.

The Plant List, 2013. Version 1.1. [online]. Available from: http://www.theplantlist.org/

[Accessed 11 Nov 2018].

Tscharntke, T. and Brandl, R., 2004. Plant-Insect Interactions in Fragmented

Landscapes. Annual Review of Entomology, 49 (1), 405–430.

Tscharntke, T., Steffan-dewenter, I., Kruess, A., and Thies, C., 2002. Characteristics of

insect population on habitat fragments: A mini reviw. Ecological Research, 17,

229–239.

Turner, E. C., Abidin, Y. Z., Barlow, H., Fayle, T. M., Jaafar, M. H. H., Khen, C. V.,

Larenus, J., Nainar, A., Reynolds, G., Yusof, Y. bin, Khoo, M. S., and Ewers, R.

M., 2012. The Stability of Altered Forest Ecosystems Project: Investigating the

Design of Human-Modified Landscapes for Productivity and Conservation. The

Planter, 88 (1036), 453–468.

186

Turner, E. C., Snaddon, J. L., Ewers, R. M., Fayle, T. M., and Foster, W. A., 2011. The

Impact of Oil Palm Expansion on Environmental Change: Putting Conservation

Research in Context. In: M.A. Dos Santos Bernardes, ed. Environmental Impact of

Biofuels. InTech, 20–40.

Tylianakis, J. M., Tscharntke, T., and Lewis, O. T., 2007. Habitat modification alters

the structure of tropical host-parasitoid food webs. Nature, 445 (7124), 202–205.

Valiente-Banuet, A., Aizen, M. A., Alcántara, J. M., Arroyo, J., Cocucci, A., Galetti,

M., García, M. B., García, D., Gómez, J. M., Jordano, P., Medel, R., Navarro, L.,

Obeso, J. R., Oviedo, R., Ramírez, N., Rey, P. J., Traveset, A., Verdú, M., and

Zamora, R., 2015. Beyond species loss: The extinction of ecological interactions in

a changing world. Functional Ecology, 29 (3), 299–307.

Valladares, G., Cagnolo, L., and Salvo, A., 2012. Forest fragmentation leads to food

web contraction. Oikos, 121 (2), 299–305.

Valladares, G. R., Salvo, A., and Cagnolo, L., 2006. Habitat Fragmentation Effects on

Trophic Processes of Insect-Plant Food Webs. Conservation Biology, 20 (1), 212–

217.

Visakorpi, K., Gripenberg, S., Malhi, Y., Bolas, C., Oliveras, I., Harris, N., Rifai, S.,

and Riutta, T., 2018. Small-scale indirect plant responses to insect herbivory could

have major impacts on canopy photosynthesis and isoprene emission. New

Phytologist, 220, 799–810.

Walsh, R. P. D., 1996. Drought Frequency Changes in Sabah and Adjacent Parts of

Northern Borneo Since the Late Nineteenth Century and Possible Implications for

Tropical Rain Forest Dynamics Stable URL : http://www.jstor.org/stable/2560059

187

Accessed : 29-02-2016 11 : 19 UTC Your. Journal of Tropical Ecology, 12 (3),

385–407.

Walsh, R. P. D. and Newbery, D. M., 1999. The ecoclimatology of Danum, Sabah, in

the context of the world’s rainforest regions, with particular reference to dry

periods and their impact. Philosophical Transactions of the Royal Society B:

Biological Sciences, 354 (1391), 1869–1883.

Wang, Z. and Goonewardene, L. A., 2004. The use of MIXED models in the analysis of

animal experiments with repeated measures data. Canadian Journal of Animal

Science, 84 (1), 1–11.

War, A. R., Paulraj, M. G., Ahmad, T., Buhroo, A. A., Hussain, B., Ignacimuthu, S.,

and Sharma, H. C., 2012. Mechanisms of plant defense against insect herbivores.

Plant signaling & behavior, 7 (10), 1306–1320.

Wardhaugh, C. W., Edwards, W., and Stork, N. E., 2015. The specialization and

structure of antagonistic and mutualistic networks of beetles on rainforest canopy

trees. Biological Journal of the Linnean Society, 114 (2), 287–295.

Waterman, P. G. and Mole, S., 1994. Analysis of Phenolic Plant Metabolites. Oxford,

United Kingdom: Blackwell Scientific Publications.

Whitfeld, T. J. S., Novotny, V., Miller, S. E., Hrcek, J., Klimes, P., and Weiblen, G. D.,

2012. Predicting tropical insect herbivore abundance from host plant traits and

phylogeny. Ecology, 93 (8), 211–222.

Williams-Linera, G. and Herrera, F., 2003. Folivory, herbivores, and environment in the

understory of a tropical montane cloud forest. Biotropica, 35 (1), 67–73.

188

Wilson, E. O., 1987. The Little Things That Run the World (The Importance and

Conservation of Invertebrates). Conservation Biology, 1 (4), 344–346.

Wirth, R., Meyer, S. T., Leal, I. R., and Tabarelli, M., 2008. Plant Herbivore

Interactions at the Forest Edge. Progress in Botany, 68, 423–448.

Wooster, M. J., Perry, G. L. W., and Zoumas, A., 2012. Fire, drought and El Niño

relationships on Borneo (Southeast Asia) in the pre-MODIS era (1980-2000).

Biogeosciences, 9 (1), 317–340.

Wright, A. D. and Fuller, S. C., 1984. A simple penetrometer for laboratory and field

use. Australian Entomologist, 11 (1), 13–15.

Wright, I. J., Reich, P. B., Westoby, M., Ackerly, D. D., Baruch, Z., Bongers, F.,

Cavender-Bares, J., Chapin, T., Cornellssen, J. H. C., Diemer, M., Flexas, J.,

Garnier, E., Groom, P. K., Gulias, J., Hikosaka, K., Lamont, B. B., Lee, T., Lee,

W., Lusk, C., Midgley, J. J., Navas, M. L., Niinemets, Ü., Oleksyn, J., Osada, H.,

Poorter, H., Poot, P., Prior, L., Pyankov, V. I., Roumet, C., Thomas, S. C.,

Tjoelker, M. G., Veneklaas, E. J., and Villar, R., 2004. The worldwide leaf

economics spectrum. Nature, 428 (6985), 821–827.

Yamazaki, K. and Sugiura, S., 2016. Stem-galling moths provide cetoniine beetles with

feeding sites via sap exudation of invasive alien plants. Entomological Science, 19

(2), 142–146.

Yeong, K. L., Reynolds, G., and Hill, J. K., 2016. Enrichment planting to improve

habitat quality and conservation value of tropical rainforest fragments. Biodiversity

and Conservation, 25 (5), 957–973.

Zahiri, R., Holloway, J. D., Kitching, I. J., Lafontaine, J. D., Mutanen, M., and

189

Wahlberg, N., 2012. Molecular phylogenetics of Erebidae (Lepidoptera,

Noctuoidea). Systematic Entomology, 37 (1), 102–124.

Zhang, Q., Zhang, J., Shen, J., Silva, A., Dennis, D. A., and Barrow, C. J., 2006. A

simple 96-well microplate method for estimation of total polyphenol content in

seaweeds. Journal of Applied Phycology, 18 (3–5), 445–450.

Zhang, Y. and Adams, J., 2011. Top-down control of herbivores varies with ecosystem

types. Journal of Ecology, 99 (2), 370–372.

Zuur, A. F., Ieno, E. N., and Elphick, C. S., 2010. A protocol for data exploration to

avoid common statistical problems. Methods in Ecology and Evolution, 1 (1), 3–

14.

Zuur, A. F., Ieno, E. N., Walker, N. J., Saveliev, A. A., and Smith, G. M., 2009. Mixed

Effects Models and Extensions in Ecology with R. New York: Springer.

Zuur, A. F., Leno, E. N., Walker, N. J., Saveliev, A. A., and Smith, G. M., 2011. Mixed

effects models and extensions in ecology with R. Public Health.

190