sign in sign up
<<
Home , Biome, Ecoregion, Gorgeous Bushshrike, Kurrichane thrush, Land grabbing, Rattling cisticola

EFFECTS OF SHRUB ENCROACHMENT IN GRASS-DOMINATED ON VERTEBRATE COMMUNITIES

By

RICHARD A. STANTON, JR.

A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA

2017

© 2017 Richard A. Stanton, Jr.

To my parents

ACKNOWLEDGMENTS

I thank my advisors, Rob Fletcher and Bob McCleery, as well as my committee members, Ara Monadjem, Tom Frazer, and Todd Palmer, for their invaluable guidance and support at every stage of my doctoral studies. I would not have been successful without it. I also thank Muzi Sibiya, who devoted many hours in the field with me collecting data, translating

SisWati to English, and generally making it possible for me to work in Swazi communities. His cheerful demeanor, hard work, and outstanding talent as a naturalist were all essential to the success of this research, and I consider myself incredibly fortunate to have had his assistance.

Joe Soto-Shoender, Wes Boone, and Niels Blaum also made critical contributions that made my dissertation possible by providing or collecting data and offering input on one or more of my chapters. Likewise, Julie Shapiro, Jessica Laskowski, and Sarah Duncan provided more feedback on chapters, grant applications, and proposal drafts than I can properly measure much less repay.

Brian Reichert, Brad Udell, Dan Greene, and Rajeev Pillay were always willing to make time to discuss data analysis, coding, field methods, or logistics with me, and their unique knowledge was invaluable in helping me to make progress in my research.

The School of Natural Resources and the Environment and the Center for African Studies provided essential fellowship and grant support, respectively, for which I am very grateful. I am also deeply appreciative that Drs. Fletcher and McCleery provided funding to offset my travel costs. I want to extend warm thanks to the staff at the Research Center in Swaziland, who provided lodging, vehicles, meals, and field assistance for a reasonable price, as well as managers and staff at: Mbuluzi Game Reserve; Mlawula Reserve;

Mhlosinga Nature Reserve; Bar Circle Ranch; Mabuda Farm; and Nisela Safaris, who provided access, assistance, and in some, cases, complimentary or reduced-rate lodging. I thank the

4

management at Tongaat Hulett for permitting us to count on their sugar cane fields and the local chiefs that allowed us to do the same in Swazi homesteads and pastures.

Members of the Fletcher and McCleery Labs, the Natural Resource Management in working group, the Florida Cooperative Fish and Wildlife Research Unit, and the Oli Lab all provided camaraderie and an invaluable social setting for research that I truly appreciate. Finally, Hillary Carter and my family were an ongoing source of guidance and support in that is beyond measure, and their assistance during my doctoral studies was no exception.

5

TABLE OF CONTENTS

page

ACKNOWLEDGMENTS ...... 4

LIST OF TABLES ...... 9

LIST OF FIGURES ...... 10

ABSTRACT ...... 12

CHAPTER

1 INTRODUCTION ...... 14

2 SHRUB ENCROACHMENT AND VERTEBRATE DIVERSITY: A GLOBAL META-ANALYSIS ...... 22

Synopsis ...... 22 Background ...... 23 Materials and Methods ...... 26 Study Design ...... 26 Literature Search ...... 26 Inclusion Criteria ...... 27 Data Collection ...... 27 Analysis ...... 29 Effect size calculations ...... 29 Modeling approach ...... 30 Candidate models ...... 31 Model goodness-of-fit and possible publication bias ...... 32 Results...... 32 Literature Search ...... 32 Meta-analyses ...... 33 Discussion ...... 35 Shrub Encroachment Effects Across Climatic Gradients and Taxa ...... 35 Shrub Encroachment Effects Among ...... 36 Data Gaps and Limitations ...... 36 Implications for , Conservation, and Management ...... 37

3 SHRUB ENCROACHMENT AND -USE INTENSIFICATION IN AN AFRICAN SAVANNA HAVE DIFFERENT EFFECTS ON ALPHA AND BETA DIVERSITY ...... 44

Synopsis ...... 44 Background ...... 45 Materials and Methods ...... 48 Study Area ...... 48

6

Sampling Design ...... 49 Stratification by and shrub encroachment ...... 49 Bird surveys ...... 50 Patch structure ...... 50 Analytical Methods ...... 50 Estimating presence-absence and richness ...... 50 Generalized dissimilarity modeling ...... 52 Visualizing community change across environmental gradients ...... 53 Partitioning dissimilarity owing to turnover and nestedness...... 54 Results...... 54 Discussion ...... 55

4 DIET AND BODY SIZE EXPLAIN SHRUB ENCROACHMENT AND LAND-USE EFFECTS ON BIRD OCCUPANCY IN AN AFRICAN SAVANNA ...... 63

Synopsis ...... 63 Background ...... 64 Materials and Methods ...... 68 Study Area ...... 68 Sampling Design ...... 69 Stratification by land use and shrub-encroachment ...... 69 Bird surveys ...... 70 Patch vegetation structure ...... 70 Analytical Methods ...... 71 Species traits ...... 71 Quantifying occupancy ...... 71 Effects of shrub encroachment, land-use intensity, and interactions ...... 72 Results...... 74 Discussion ...... 76

5 PREDATORY BIRD CUES AFFECT PREY DETECTABILITY, NOT OCCUPANCY DYNAMICS, ACROSS A SHRUB COVER GRADIENT IN AN AFRICAN SAVANNA ...... 85

Synopsis ...... 85 Background ...... 86 Materials and Methods ...... 87 Study Area ...... 87 Experimental Design ...... 89 Data Collection ...... 91 Analytical Methods ...... 92 Dynamic occupancy modeling ...... 92 Poisson regression of counts ...... 94 Results...... 94 Discussion ...... 95

6 CONCLUSIONS ...... 104

7

APPENDIX

A SHRUB ENCROACHMENT ON ’S TERRESTRIAL BIOMES ...... 106

B DATA SOURCES USED IN A GLOBAL META-ANALYSIS OF SHRUB ENCROACHMENT EFFECTS ON VERTEBRATE DIVERSITY ...... 107

C DATABASES CONSULTED AND SEARCH TERMS USED IN META-ANALYSIS ...111

D ATTRIBUTES OF STUDIES USED IN THE META-ANALYSIS ...... 112

E META-REGRESSION RESULTS ...... 114

F SHRUB ENCROACHMENT AFFECTS VERTEBRATE COMMUNITY STRUCTURE AMONG BIOMES ...... 117

G REPRESENTATIVE FUNNEL PLOTS FROM THE META-ANALYSIS ...... 118

H DISTRIBUTION OF OBSERVED AND MODELED OCCUPANCY ...... 119

I MODELED AND OBSERVED SPECIES RICHNESS ...... 120

J TRAITS OF COMMON SPECIES ...... 121

K MODEL TUNING PARAMETERS FOR EACH SPECIES ...... 123

L DISTRIBUTION OF NAÏVE OCCUPANCY BY DIET AMONG LAND USES ...... 125

M PATTERNS OF DETECTION-NONDETECTION FOR FOUR SPECIES SUBJECTED TO AUDITORY CUES OF SEVERAL NON-RAPTOR PREDATORY BIRDS...... 126

REFERENCE LIST ...... 127

BIOGRAPHICAL SKETCH ...... 143

8

LIST OF TABLES

Table page 5-1 Models considered to test for non-raptor predatory bird cue treatment effects ...... 98

5-2 Parameter estimates (SE) from top-ranked dynamic occupancy models...... 99

5-3 Rankings of dynamic occupancy models ...... 100

A-1 Shrub encroachment is occurring in at least six of the ’s fourteen terrestrial biomes ...... 106

C-1 Databases consulted on 3 March 2016 and search terms used ...... 111

D-1 Studies used in a global meta-analysis of shrub encroachment effects ...... 112

E-1 Results from a set of meta-regressions fit to identify sources of heterogeneity ...... 114

J-1 Diet, predatory status, and mean mass (from Hockey et al. 2005) of 48 species ...... 122

K-1 Tuning parameters used when fitting Bayesian occupancy models ...... 124

9

LIST OF FIGURES

Figure page 2-1 Locations and mean annual associated with studies of shrub- encroachment ...... 39

2-2 Variation in vegetation structure across gradients...... 40

2-3 Shrub encroachment had no overall effect ...... 41

2-4 Shrub encroachment reduced vertebrate Shannon diversity...... 42

2-5 Shrub encroachment effects (r) on vertebrate community structure ...... 43

3-1 Species richness increased with shrub cover and decreased with land-use intensification...... 59

3-2 Bird species dissimilarity ...... 60

3-3 Differences in bird community composition...... 61

3-4 Bird community dissimilarity was mainly driven by species turnover...... 62

4-1 Geographic setting, sampling scheme, and distribution of shrub cover ...... 79

4-2 Bird occupancy was negatively associated with land-use intensification ...... 80

4-3 Bird occupancy had a weak overall association with shrub encroachment ...... 81

4-4 Predatory bird occupancy was negatively associated with community pastures ...... 82

4-5 Bird responses to land-use intensification differed among diet groups ...... 83

4-6 Body size was negatively associated with bird occupancy in protected areas ...... 84

5-1 A broadcast unit deployed in Swaziland ...... 102

5-2 Detection probability for chinspot batises increased with shrub cover ...... 103

F-1 Shrub encroachment effects (r) on vertebrate community structure among biomes...... 117

G-1 Representative funnel plots ...... 118

H-1 Observed and modeled proportion of area occupied...... 119

I-1 Bird species richness among local communities in the Lowveld savanna of Swaziland ...... 120

L-1 Naïve bird occupancy by diet across a gradient of land-use intensity ...... 125

10

M-1 Patterns of detection-nondetection for four species ...... 126

11

Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy

EFFECTS OF SHRUB ENCROACHMENT IN GRASS-DOMINATED BIOMES ON VERTEBRATE COMMUNITIES

By

Richard A. Stanton, Jr.

August 2017

Chair: Robert J, Fletcher, Jr. Cochair: Robert A. McCleery Major: Interdisciplinary Ecology

Grass-dominated biomes around the world are experiencing increases in shrub cover and land-use intensity. I examined how shrub encroachment and land-use intensification affected vertebrates at the community and species levels, determining: 1) whether global climatic gradients and vertebrate group explain community-level responses to encroachment; 2) the relative roles of encroachment and land-use intensification in explaining changes in bird community dissimilarity among locations; 3) if species traits explain species responses to encroachment, land-use intensification, and their interactions; and 4) whether predatory generalist birds affect the occupancy dynamics and detectability of prey. I quantified the effects of shrub encroachment and grazing on vertebrate species richness, Shannon diversity, and abundance across global variation in , , and land use, as well as among vertebrate groups using a meta-analysis, finding that net primary productivity was an important source of heterogeneity in shrub encroachment effects and that bird responses to encroachment (neutral to marginally positive) were distinct from those of and herpetofauna (negative). Through repeated sampling of bird communities in Swaziland I found that shrub cover rather than land- use intensity was the major correlate of bird community shifts away from their baseline

12

composition. I also quantified how changes in bird species occurrence varied with shrub encroachment and land-use intensity, discovering that diet and body size explained much of the variation in responses. Finally, I experimentally manipulated cues of predatory birds across a shrub gradient to quantify the interactive effects of predatory bird cues and shrub cover on the occurrence and detectability of several common savanna bird species. I found that some species responded to predatory bird cues by becoming more detectable, and this effect was moderated by shrub cover for one species. My findings provide a basis for: 1) focusing shrub encroachment mitigation on mammals, herpetofauna, and vertebrates in low net primary productivity; and 2) managing shrub encroachment rather than land-use intensification to maintain the composition of savanna bird communities. My results also suggest that predator- prey interactions can be an important driver prey behavior that varies with shrub cover.

13

CHAPTER 1 INTRODUCTION

We live in a world experiencing a panoply of transformations. In grass-dominated terrestrial biomes, for example, woody encroachment is widespread, land is being converted from pastoral use to intensive monocultures, and are shifting (Dieguez et al. 2017). One might surmise that biological communities subject to these conditions would be in an incessant state of flux, their composition and impossible to predict (Lawton 1999; Simberloff

2004). However, processes such as competition and contribute to community assembly in ways that should differ across environmental gradients and be predictable from species’ traits

(Simberloff 2004; McGill et al. 2006). Synergies among drivers of effects on biological communities can also occur, and should therefore be considered (Brook et al. 2008;

Oliver & Morecroft 2014; Darling et al. 2016).

This dissertation is a study of how vertebrate communities in grass-dominated biomes, e.g. African , are affected by shrub encroachment. I employed a combination of methods, including meta-analysis, field sampling of bird occurrence across multiple environmental gradients, and a field experiment in order to describe and explain the effects of shrub encroachment with consideration for species’ traits and possible synergies with land-use intensification, e.g. conversion of protected areas and pastoral to subsistence farming or . Consequently, I define and discuss shrub encroachment, outline how community responses to environmental change may be studied, and state the specific research objectives addressed in my dissertation below.

Shrub encroachment can be defined as the increase in number, cover, or of short- statured (< 5 m) woody plants relative to grasses and forbs in ecosystems where the predominant ground cover was historically grassy (Eldridge et al. 2011). By this fairly inclusive definition,

14

shrub encroachment is one of the most widespread types of land-cover change, occurring in at least 7 of the planet’s 14 terrestrial biomes (Olson et al. 2001; Chapter 2). Further, shrub encroachment has effects on a variety of processes, as well as variable effects on vertebrate species (Eldridge et al. 2011, Sirami and Monadjem 2012). Shrub encroachment has been attributed to: atmospheric CO2 enrichment; overgrazing; lack of grazing; invasive alien species; fire exclusion; and loss of native and mega- (Eldridge et al. 2011, Roques et al. 2001). Understanding how grasses and woody coexist can help establish how these multiple putative drivers may promote shrub encroachment.

Grasses and woody plants have a number of physiological and morphological differences that determine their responses to environmental variation. In savannas, the differences between grasses and woody plants have favored the presence of a near-continuous grassy understory while permitting co-existence of scattered woody plants, primarily . The main environmental drivers favoring grasses in savannas have been: 1) modest and temporally variable rainfall; 2) frequent fire; 3) low levels of atmospheric CO2; and 4) a diverse, highly-mobile assemblage of browsing and grazing (Bond et al. 2003; Lehmann et al. 2011; Stevens et al. 2016a). All four environmental conditions are changing rapidly throughout the world’s savannas, favoring woody encroachment by shrubs (Bond et al. 2005; Midgley & Bond 2015;

Stevens et al. 2016a). The physiological and morphological differences between grasses and woody plants that determine the effects of rainfall, atmospheric CO2, and herbivory warrant some discussion.

Modest and temporally-variable rainfall have separate effects that contribute to coexistence between grasses and woody plants, particularly trees. Modest average rainfall favors grasses whilst temporal variability in rainfall favors woody plants because woody plants have

15

more endogenous water reserves and perhaps because woody plants have deep roots that can exploit water unavailable to grasses (< 650 mm-y-1; Bond et al. 2003; Sankaran et al. 2004). The latter hypothesis seems reasonable but the empirical evidence for it is mixed (Scholes & Archer

1997). However, several researchers predict afforestation and shrub encroachment in savannas will increase based on projected changes in precipitation (Sankaran et al. 2005).

Frequent, low-intensity fire favors grasses over woody plants because differences between the two types facilitate different strategies for coping with fire (Sankaran et al.

2004). Grasses employ a high ratio of above-ground surface area to mass that encourages fire whereas woody plants have a much denser architecture that is less conducive to fire (Bond

2008). Further, woody plants indigenous to grass-dominated biomes frequently have that reduce the impact of fire on their above-ground biomass, such as thick bark and coppicing

(Parr et al. 2014). Grasses, on the other hand, defend their endogenous reserves by sequestering them below-ground, enabling rapid re-growth after fire. The dense, fibrous, and shallow roots of grasses also enable them to more readily take up the pulse of nutrients that become available after a fire (Bond et al. 2005). Several authors have pointed to such adaptations as evidence that grass-dominated ecosystems historically treated as anthropogenically-degraded are in fact savanna or , although the matter remains contentious (Parr et al. 2014; Ratnam et al.

2016).

Human-caused atmospheric CO2 enrichment is favoring woody plants in grass-dominated biomes because woody plants and grasses generally have different physiological mechanisms for fixing . Savanna grasses primarily employ C4 metabolism whereas woody plants use a C3 strategy (Scholes & Archer 1997). These differences evolved largely because carbon fixation requires air flow between plant organs and the atmosphere, which can exacerbate water loss in

16

dry environments. The C4 pathway reduces water loss but is less efficient at higher levels of

CO2, creating conditions which can lead to competitive displacement of grasses by woody plants.

This occurs under controlled conditions (Polley et al. 1994). There is also circumstantial empirical evidence that atmospheric CO2 is a real-world driver of shrub encroachment. For example, shrub encroachment has increased with atmospheric CO2 concentrations throughout southern Africa despite tremendous heterogeneity in rainfall, land use, and grazing activity, and only elephants appear to meaningfully counteract that trend (Wigley et al. 2009, 2010).

Most grass-dominated ecosystems have a diverse, highly mobile assemblage of grazing and browsing animals that contribute to coexistence between grasses and woody plants. An abundance of observational and experimental evidence shows that the composition of the community can determine the outcome of competition between grasses and woody plants (Young et al. 2005; Pringle et al. 2007; Sankaran et al. 2008). There is also evidence that the impairing the mobility of mega-herbivores can alter the spatial distribution and temporal dynamics of shrub and grass cover (Skarpe 1991).

Woody plant tissues are largely “recalcitrant,” i.e. they resist consumption by herbivores and are inefficiently assimilated into biomass (Mole 1994). The latter is important because it favors large body size in browsing animals— the major browsers in African savannas are megaherbivores such as elephants (Sankaran et al. 2008). The charismatic qualities of megaherbivores and their potential to come into conflict with people have led to massive range contractions and overall declines, perhaps exacerbating shrub encroachment (Wigley et al. 2010;

Ripple et al. 2015). Meanwhile, elephants have become abundant enough in some protected areas to raise concerns that they be pushing savannas toward a grassland state; at a minimum there is evidence that some species are at risk of widespread extirpation (Kiker et al. 2014).

17

However, there is also evidence suggesting tree cover can be maintained in savannas with unregulated elephant , at least for Vachellia drepanolobum (nee ), which is defended from elephants by ants (Goheen & Palmer 2010). Ultimately, both the abundance and mobility of mega-herbivores appear to be important drivers of woody plant dynamics. On the other hand, there are positive feedbacks between grazing and grass growth that contribute to the continued presence of grass cover in savannas. This occurs in part because grass tissues possess relatively modest immediate defenses against herbivory. Rather, plants sequester resources below ground while leaving leaves vulnerable.

However, relations between plants and grazers can change in absence of predation and when herbivore mobility is limited, leading to loss of grass cover and, ultimately, severe woody encroachment (Skarpe 1991; Roques et al. 2001). This is especially apparent where cattle have replaced wildlife and socioeconomic arrangements place few restraints on cattle stocking rates

(Roques et al. 2001). The combination of high numbers and inability to locate greener pastures can cause herbivores to destroy grasses, reducing the competition experienced by woody plants in the process. Human land-use practices in savannas include woody plant for fuel and building materials, which might locally counteract shrub encroachment, but several studies suggest this is generally not important because shrubs rapidly regenerate by coppicing, and therefore withstand harvest (Twine et al. 2016).

Shrub encroachment is a difficult process to manage because all of the above-mentioned factors are currently changing in ways that favor it throughout the world. Given that extensive shrub cover may soon become the “new normal” in many grass-dominated ecosystems, there is a need to understand how increased shrub cover is affecting biological communities, preferably grounded in the mechanisms that determine community composition.

18

Shrub encroachment can lead to restructuring of vertebrate communities via effects that vary widely among species (Eldridge et al. 2011, Sirami and Monadjem 2012). The influence of shrub encroachment on vertebrate community diversity has been the subject of approximately

188 studies (Chapter 2). However, these studies are almost exclusively concerned with community composition as characterized by species identity rather than more generalizable attributes such as species’ traits that could be targeted by natural selection. It is exceedingly difficult to synthesize the literature from such a perspective given the paucity of available studies, even for intensively-studied taxa such as birds.

Shrub encroachment studies as a whole have seldom been grounded in mechanism. Yet, species interactions such as antipredator behavior may explain a substantial part of community dynamics in response to rapid global change, particularly for shrub encroachment. Shrub encroachment in savannas can reduce temporal variability in cover and increase the total amount of cover by suppressing grasses and discouraging fire, and cover is central to the of vertebrates because it mediates predation risk (Mysterud & Ostbye 1999). The traits that work best in open savanna may be predictably different than those favored in shrub-encroached savanna (Lima 1992).

Predators exhibit strong consumptive and non-consumptive effects on prey dynamics that can drive community composition by selecting particular prey traits and promoting coexistence among prey species (Holt 1977; Lima 1992; Martin & Li 1992; Preisser et al. 2005). Further, species frequently exhibit competition and predator-prey interactions simultaneously, i.e. intraguild predation, that can have varied effects on population dynamics and coexistence (Arponen et al. 2008). Finally, communities typically include multiple predatory species; the complementary traits of multiple predators can reduce the amount predator-free

19

space whilst the presence of multiple predators can have emergent effects not predictable from the effects of individual species (Sih et al. 1998; Cresswell & Quinn 2013). Given the myriad ways predators can shape communities, any change in predator occurrence associated with an environmental gradient warrants close attention. This was important for the development of this dissertation research because I discovered that predatory bird occurrence increased with shrub encroachment in the Lowveld of Swaziland (Chapter 4).

In my dissertation, I undertook a comprehensive, multi-method investigation of shrub encroachment effects on vertebrate communities that: 1) tests whether the effects of shrub encroachment on vertebrate community structure are consistent or vary by taxon, among continents, or across global climatic gradients; 2) describes spatial differences in species richness and community composition, quantifying the associations between community composition and patterns of shrub encroachment and land-use intensity, respectively; 3) uses information about species traits to generate hypotheses about drivers of community assembly, while assessing whether patterns of species occurrence can be explained by synergies with other global change processes; and 4) tests whether changes in species’ occupancy dynamics and behavior caused by predator cues can be moderated by shrub cover. I used bird communities in the Lowveld savanna of Swaziland, and vertebrate communities in grass-dominated biomes worldwide, to meet the following research objectives:

1) Determine how variation in vertebrate species richness, Shannon diversity, and abundance responses to shrub encroachment varies across global variation in climate, across biomes, and among vertebrate groups (birds, mammals, and herpetofauna). Assess whether shrub thinning can reverse the effects of shrub encroachment on the structure of vertebrate communities.

20

2) Quantify associations among bird diversity (i.e. alpha diversity, nestedness, and turnover), shrub encroachment, and land-use intensification in a savanna land-use mosaic in order to compare and contrast the importance and consistency of their effects.

3) Determine whether shrub encroachment and land-use type interact to predict bird species occupancy in a savanna land-use mosaic and whether species’ responses to shrub encroachment and land-use intensification can be explained by species’ traits.

4) Determine whether cues of non-raptor predatory birds and shrub cover interact to predict changes in the abundance of several savanna bird species and whether species traits can explain those responses.

21

CHAPTER 2 SHRUB ENCROACHMENT AND VERTEBRATE DIVERSITY: A GLOBAL META- ANALYSIS

Synopsis

Across the planet, grass-dominated biomes are experiencing shrub encroachment driven by atmospheric CO2 enrichment and land-use change. By altering resource structure and availability, shrub encroachment may have important impacts on vertebrate communities. Yet the magnitude and variability of these effects across climatic gradients, continents, and taxa remain unknown, nor do we know whether shrub thinning restores the structure of vertebrate communities. I estimated relationships between percent shrub cover and the structure of terrestrial vertebrate communities (species richness, Shannon diversity, and community abundance) in experimentally-thinned and unmanipulated shrub-encroached grass-dominated biomes using systematic review and meta-analyses of 52 studies published worldwide from

1978-2016. I modeled the effects of , biome, mean annual precipitation, net primary productivity, and the normalized difference vegetation index (NDVI) on the relationship between shrub cover and vertebrate community structure. Species richness, Shannon diversity, and total abundance had no consistent relationship with shrub encroachment and experimental thinning did not reverse encroachment effects on vertebrate communities. However, some effects of shrub encroachment on vertebrate communities differed with net primary productivity, among vertebrate groups, and across continents. Encroachment had negative effects on vertebrate diversity at low net primary productivity, while mammalian and herpetofaunal diversity decreased with shrub encroachment. Shrub encroachment also had negative effects on species richness and total abundance in Africa but positive effects in . conservation and mitigation efforts responding to shrub-encroachment should focus on low- productivity locations, on mammals and herpetofauna, and in Africa. However, targeted research

22

in neglected such as central and will be needed to fill important gaps in our knowledge of shrub encroachment effects on vertebrates. Additionally, my findings provide an impetus for determining the mechanisms associated with changes in vertebrate diversity and abundance in shrub-encroached grass-dominated biomes.

Background

Across the planet, grass-dominated biomes are experiencing largely unprecedented increases in woody biomass attributable to short-statured plants (<6m, i.e., shrub encroachment;

Eldridge et al. 2011), driven by increasing atmospheric CO2 and land use (Van Auken 2000;

Cabral et al. 2003; Wigley et al. 2010; Stevens et al. 2016b). Land use, such as ranching, contributes to shrub encroachment by altering disturbance regimes whereas increasing atmospheric CO2 favors woody plants over grasses (Stevens et al. 2016b). Grass-dominated biomes experiencing shrub encroachment include and savannas, ,

Mediterranean dehesas, and in savanna maintained by seasonal flooding (e.g. the in

South America;Van Auken 2000; Assine & Soares 2004; Leuzinger et al. 2011; Naito &

2011; Parr et al. 2014; Stevens et al. 2016a). I define a biome as grass-dominated if is characterized by a near-continuous layer of grass and herbaceous plants (a physiognomic approach sensu Whittaker 1962). According to this definition, at least six of the world’s fourteen terrestrial biomes are grass-dominated, and shrub encroachment is occurring in all of them

(Olson et al. 2001; Parr et al. 2014; Appendix A). Shrub encroachment affects cover and other resources known to be critical to many vertebrates, which may lead to important and consistent effects of encroachment on communities across grass-dominated biomes (Ricklefs 2004; Pausas

& Keeley 2009; Parr et al. 2014). Indeed, shrub encroachment can change the structure and composition of terrestrial vertebrate communities (Brown et al. 1998; Chown 2010; Sirami &

23

Monadjem 2012). However, the effects of shrub encroachment on vertebrates have not been synthesized to interpret the consistency of effects and how those effects vary across climate conditions, continents, and vertebrate groups (i.e. birds, mammals, and herpetofauna) is unknown.

There are several reasons why shrub encroachment effects on vertebrate communities may vary across global climatic and disturbance gradients, among continents, and across taxa.

Grass-dominated biomes maintained by limited and variable precipitation are less productive than biomes maintained by fire and herbivory and often have lower vertebrate diversity (Bond et al. 2005; Murphy & Bowman 2012). Additionally, spatial and temporal differences in available cover, , and nesting substrates between arid and mesic conditions can be particularly striking

(Fig. 1; Radford & Andersen 2012; Collins et al. 2014). For example, vegetation in grasslands and arid savannas is temporally stable and patchy compared to mesic grasslands and savannas, which are temporally dynamic and characterized by continuous vegetative cover (Fig.

2; Briske et al. 2003; Parr et al. 2014). Shrub encroachment may create more novel cover dynamics in drier grass-dominated biomes, where shrub cover has historically been uncommon, resulting in greater reductions in vertebrate diversity (Scholes & Archer 1997; Morton et al.

2011). In contrast, mesic grass-dominated biomes have historically exhibited dynamic and variable shrub cover that predates recent encroachment, and therefore may contain species for which shrub cover is not novel (Scholes & Archer, 1997; Morton et al., 2011). Vertebrates within a continent also share common evolutionary and ecological histories that may lead to different responses to shrub encroachment among continents. For example, most savannas are maintained by either limited precipitation or disturbance, but some are maintained by seasonal flooding and infrequent fire, such as the Pantanal in (Assine & Soares 2004;

24

Beerling & Osborne 2006; Murphy & Bowman 2012). Vertebrate groups may also differ in their responses to shrub encroachment because of underlying differences that are globally consistent.

Birds build in specific locations and types of vegetative cover and simplified vegetation structure can limit their diversity (Slagsvold 2001). Therefore, birds could increase in richness and diversity with shrub encroachment as new nest substrates become available. Responses of mammals and herpetofauna to shrub encroachment are more difficult to predict. Shrub encroachment could reduce the diversity of mammals and herpetofauna by removing specialist taxa adapted to grassy cover or bare ground (Ceballos et al. 2010), or responses driven by different associations might lead to compositional shifts but no change in diversity

(Leynaud & Bucher 2005).

I conducted a global meta-analysis and systematic review of shrub encroachment effects on vertebrate communities (birds, mammals, and herpetofauna), focusing on whether impacts vary across global climatic gradients driving historical disturbance regimes, among continents, and across vertebrate groups (Hurlbert & Haskell 2003). My objectives were to (1) estimate the magnitude of relationships between shrub cover and vertebrate community structure, and (2) identify associations between global climatic gradients, continents, and the magnitude of shrub encroachment effects on vertebrates. I predicted that reductions in vertebrate diversity and abundance with shrub encroachment would be greatest in grass-dominated biomes with the lowest productivity. I also predicted that vertebrate groups would exhibit specific shrub- encroachment responses. I expected birds to increase in richness, diversity, and abundance with shrub encroachment owing to a wider of variety of nesting and foraging options, whilst mammals and herpetofauna would decrease in richness and diversity (Kutt & Martin 2010).

25

Finally, I predicted that mammals and herpetofauna would exhibit no change in abundance because food availability should not increase with shrub encroachment.

Materials and Methods

Study Design

Meta-analysis is an established method of synthesizing quantitative research by taking effect sizes from multiple studies weighted according to their respective variances (Gurevitch et al. 2001; Cooper et al. 2009). Meta-analyses are also useful for modeling variation in effect sizes among studies using covariates, and are consequently well suited for synthesizing research that is global in scope (Stewart & Schmid 2015). Further, meta-analysis can account for lack of independence among and within studies, and be structured to permit inference beyond the existing literature, by the use of appropriate random-effects models (Cooper et al. 2009).

Literature Search

I systematically searched the Web of Science, ProQuest, and Google Scholar databases to locate studies of terrestrial vertebrate diversity across shrub-encroachment gradients in space or time using multiple keyword combinations (see Appendix B for details; Eldridge et al., 1997). I located unpublished studies by searching grey literature such as dissertations, theses, and government reports indexed by Google Scholar and ProQuest, as well as contacting researchers directly, and including studies located by checking the reference sections of relevant articles. I contacted the corresponding authors of all relevant papers with incomplete data and removed those studies if I did not receive a positive response to my data request within three weeks.

Finally, I assessed the assembled studies for geographic gaps in coverage. I noted an absence of suitable studies from India, Asia, and , so I contacted researchers working in those regions if a Google Scholar search indicated they had recently (since 2006; Appendix B)

26

published on shrub encroachment or vertebrate diversity in a grass-dominated biome in the . Each e-mail inquiry included a request to provide contact information for other researchers that might have suitable expertise (Biernacki & Waldorf 1981).

Inclusion Criteria

When conducting my literature search, I included studies if the authors collected vertebrate community data across a standardized area, indicated shrub encroachment had occurred, and measured percent shrub cover or shrub density at one or more spatial scales.

Further, I required that all included studies either: 1) report one or more measures of species richness, Shannon diversity, or total abundance for ≥5 species of at least one vertebrate group, or

2) indicate that species richness, Shannon diversity or total abundance of each vertebrate group studied could be computed from the underlying data. In the latter case, I computed effect sizes if

I could extract or acquire sufficient data. I separated studies that measured the effects of experiments and before-after, control-impact studies of shrub reduction by mechanical thinning, fire, and herbicides from observational studies of shrub encroachment and analyzed them separately because I expected their respective effects to be of opposite sign.

Data Collection

I recorded data from each study to identify heterogeneity among effects of shrub encroachment on vertebrate communities. For each study, I recorded: geographic coordinates, continent, and vertebrate group considered (herpetofauna, small mammals, ungulates, mammalian carnivores, or birds). I also recorded study design, i.e., longitudinal or space-for-time substitution, because the space-for-time-substitution design could yield different, and possibly unreliable, results (Pickett, 1989). Likewise, few shrub encroachment studies cover the entire possible gradient of percent shrub cover (0-100%) and vertebrate responses to encroachment may not be linear, so I recorded ranges, means, and standard deviations of percent shrub cover,

27

study-wide for each article and between author-defined groups where relevant (Burkett et al.

2005; Brook et al. 2008). I recorded shrub density where the authors did not record percent shrub cover. Finally, I recorded whether livestock grazing occurred during the study because it is a putative driver of both shrub encroachment and vertebrate community structure in grass- dominated biomes (D'Odorico et al. 2012; Ricketts et al. 2016; Waters et al. 2016).

Reporting of climatic data is frequently incomplete and data sources used vary across studies. I therefore used global databases to quantify mean annual precipitation and net primary productivity at the geographic coordinates of each study I identified. Specifically, I extracted historic mean annual precipitation and net primary productivity from WorldClim Version 1.4

(1950-2000 and 1981-2000 for precipitation and net primary productivity, respectively; Hijmans et al. 2005), and historic normalized difference vegetation index (NDVI) values from the Global

Land Cover Facility’s Data Interface (2001-2006; http://glcfapp.umiacs.umd.edu:8080/esdi/index.jsp.; Pettorelli et al. 2005). I collected these measures because they provide similar but potentially distinct information about ecological conditions among sites and no measure is clearly superior (Pettorelli et al. 2005). For example, precipitation should regulate community structure at smaller spatial scales than productivity and

NDVI (Harrison & Grace 2007). Likewise, the correlation between precipitation and NDVI is lower at higher levels of precipitation and can be moderated by conditions (Nicholson &

Farrar 1994; Fabricante et al. 2009). I also used a global database to assign each study to a biome and ( are nested within biomes and delineated based on distinctive species assemblages; Olson et al. 2001) to facilitate placing the studies in a detailed biogeographic context and permit testing for differences in shrub encroachment effects among biomes (Olson et al., 2001).

28

Analysis

Effect size calculations

I calculated ’s z-transformed r (r hereafter), a transformation of the correlation coefficient that facilitates model fitting and treated it as the effect size for all analyses (Cooper et al., 2009; Koricheva et al., 2013). I chose r because it provides directly interpretable estimates of shrub cover effects on species richness, Shannon diversity, or total abundance in terms of standard deviations and it was readily estimated from studies where these shrub cover effects were modeled directly. I calculated r from model partial coefficients, unstandardized coefficients and their respective standard errors, sample sizes, t-statistics, R2 values, F-statistics, and χ2- square statistics, or directly from raw data (Nakagawa & Cuthill 2007; Cooper et al. 2009).

When authors compared two groups with different mean shrub cover, such as control and experimental plots in thinning experiments (Lipsey and Wilson, 2001; Koricheva et al., 2013), I first calculated Hedge’s g values (Lipsey & Wilson 2001) and then converted them to correlation coefficients for Fisher’s z-transform. I calculated Hedge’s g using group means (e.g., control and treatment), sample sizes, and standard deviations (Koricheva et al. 2013).

Studies of shrub encroachment effects have employed diverse designs and analytical methods, often comparing several models. I selected effect sizes from each study based upon the simplest relevant model, i.e. the univariate model if reported, and the best multivariate model otherwise, as determined by the author’s chosen information-theoretic or hypothesis-testing criteria. I collected multiple measures from the same study when data were not suitably aggregated, such as when reporting effects of mesquite, Prosopis glandulosa, and creosote bush,

Larrea tridentata, separately (Boeing et al. 2014). However, I only collected results from the most recent sampling period or the most contrasting group (i.e. maximum ∆푠ℎ푟푢푏) when authors disaggregated their data into ≥3 groups according to year, season, range condition or grazing

29

intensity but did not provide the F-statistics required to calculate Hedge’s g directly (Lipsey &

Wilson 2001).

Modeling approach

I determined the relationships between shrub encroachment and vertebrate diversity by treating species richness, Shannon diversity (Shannon 1948), and total abundance, i.e. the total number of birds, herpetofauna ( or ), or mammals reported, as response variables. I built separate models for observational and experimental studies to determine if removal of shrubs by fire, herbicide, or mechanical means reverses the effects of encroachment— i.e., were the effects of thinning on vertebrates of opposite sign and similar magnitude? I also fit random-effects meta-regressions in R’s metafor package to determine if methodological or ecological conditions explained heterogeneity in the effect of shrub encroachment on vertebrate communities (Viechtbauer 2010; R Core Team 2016). I treated study as a random effect because I extracted more than one effect from some studies and wanted to make inferences beyond the population of studies sampled (Viechtbauer 2010). There were a priori reasons to expect that shrub-cover-vertebrate-diversity relationships may be nonlinear.

Shrub encroachment effects could also vary according to differing percent shrub cover considered among studies. I therefore fit a model with a linear and a quadratic term for shrub cover for richness, Shannon diversity, and community abundance in the metafor package and 2- part piecewise regressions of Fisher’s z-transformed r on study-wide mean percent shrub cover in R’s ‘segmented’ package (Muggeo 2003). I weighted each study in a piecewise regression by its respective precision (1/variance), akin to a meta-regression. These tests complemented meta- regressions testing if mean shrub cover effects on species richness, Shannon diversity, and community abundance were different from zero (Viechtbauer, 2010; R Core Team, 2016). I

30

focused on the meta-regression results unless the piecewise regressions were significant, in which case I reported and drew inference from the piecewise regression results.

Candidate models

I assessed global effect sizes and confidence intervals without moderators or control variables to determine the overall effects of shrub encroachment on vertebrate community structure (Koricheva et al. 2013). I then fit models accounting for several potential sources of effect size heterogeneity from a candidate set (Appendix C). I initially planned to test for the effects of mean annual precipitation, net primary productivity, NDVI, vertebrate group, continent, biome, whether the dominant encroaching shrub species was native or introduced, land-use type, and study design (longitudinal or not). I collapsed some categorical variables into fewer groups and removed some variables from consideration because the available data were insufficient for my planned analyses. I dropped whether the dominant encroaching shrub species was native or introduced from consideration prior to modelling because 95% of studies involved encroachment by native shrubs. I collapsed land-use types into grazed or ungrazed by cattle during the study period because the sample was comprised of and protected areas. I also pooled all mammalian taxa into a single group because most mammalian orders were poorly represented in the data. I found articles with useable effect sizes distributed across eight biomes but only three biomes were studied in >5 articles, so I only considered those biomes when testing for differences among biomes. The biomes I considered included: 1) and xeric , 2) temperate grasslands, savannas, and , and 3) tropical and subtropical grasslands, savannas and shrublands (Olson et al. 2001). There were also cases where a planned model could not be fit because there was no variation in the sample, e.g., all observational studies reporting Shannon diversity also employed space-for time substitution designs. I therefore ultimately fit 43 models (Appendix C). I interpreted significance from the P values for

31

each parameter estimates from each respective model (P < 0.05) but note where I found weak evidence for an effect, i.e., 0.05 < P < 0.10.

Model goodness-of-fit and possible publication bias

I estimated the amount of unmodeled heterogeneity remaining for each model using the I2 statistic, which ranges from 0-100% (Higgins et al. 2003). The I2 statistic can provide a heuristic aid for judging whether sources of heterogeneity in effects have meaningful explanatory power

(Higgins et al. 2003). I also plotted effect size against precision overlain with a 95% distribution expected in the absence of bias (i.e. funnel plots), and inspected them for apparent asymmetry

(Egger et al., 1997; Cooper et al., 2009). Asymmetry in funnel plots can indicate bias attributable to selective publication and other factors (Sterne et al. 2011).

Results

Literature Search

My literature searches identified 1694 results comprised of 255 unique articles. I selected

188 articles for further scrutiny after scanning titles and abstracts. I ultimately reduced the 188 potentially relevant articles to 43 that met my inclusion criteria (see Appendix D for references).

The articles I included were published between 1978 and 2016. I extracted 114 r values from the

43 articles. I extracted 15 effects from 6 experiments and 99 effects from 38 observational studies. Common reasons I removed studies from consideration included: 1) ordination methods were used that rendered the data uninterpretable for my purposes; and 2) no relevant tests of shrub cover effects on vertebrate community structure were reported or could be computed even though suitable data were collected.

I found differences in apparent research effort among vertebrate groups, with 61, 31, and

22 effects pertaining to birds, mammals and herpetofauna, respectively. Among mammals, I identified 19, 8, and 4 articles allocated to , carnivores, and ungulates, respectively.

32

Likewise, the effects I included were concentrated in North America and Africa (27 and 43 effects, respectively), whereas I collected few effects from Australia, , and South America

(4, 13, and 12 effects, respectively) and no effects from Asia. The coverage of vertebrate groups differed across continents. I collected effects from all groups in North America, effects from birds and mammals in Africa, data from birds only in Europe and Australia, and data from herpetofauna alone in South America (Appendix E). Finally, I extracted effects from studies across eight biomes. Five biomes were the subject of 1-2 articles and the remaining three biomes were the subject of five or more articles (Appendix A).

Meta-analyses

Meta-analysis did not reveal general, consistent effects of shrub encroachment on the species richness, Shannon diversity, or total abundance of vertebrate communities in either observational or experimental studies (P > 0.05; Fig. 2-3). Likewise, neither adding a quadratic shrub cover term nor piecewise regression indicated any nonlinear effect of mean shrub cover among studies (P > 0.05). Shrub encroachment effects on species richness, Shannon diversity, and community abundance also did not differ among biomes (P > 0.05). Univariate meta- regressions did indicate, however, that shrub encroachment effects in observational studies varied with net primary productivity, among vertebrate groups, and across continents (Fig. 2-4).

The effect of shrub encroachment on vertebrate diversity was negative at the lowest net primary productivity in observational studies, with no effect in the most productive locations

(훽̂ = 0.44 ± 0.14 SE, P = 0.0019, k [number of effects] = 23; Fig. 2-4). I also found positive but non-significant relationships of shrub encroachment effects on diversity with mean annual precipitation and NDVI in observational studies (훽̂ = 0.23 ± 0.15 SE and 0.30 ± 0.16 SE, P =

0.12 and 0.07, respectively, k = 23; Appendix D) and a non-significant positive relationship with

33

species richness and mean annual precipitation in thinning experiments (훽̂ = 0.35 ± 0.19 SE, P

= 0.06, k = 6; Appendix D). I found no other relationships between productivity, precipitation and vertebrate richness or total abundance (P > 0.05; Appendix D).

Shrub encroachment effects varied among vertebrate groups. Shannon diversity of mammals and herpetofauna was negatively correlated with shrub encroachment (훽̂ = −0.75 ±

0.31 SE and − 0.90 ± 0.26 SE, P = 0.02 and 0.0007, respectively, k = 23; Appendix D). Further, total abundance was also negatively correlated with shrub encroachment among herpetofauna and mammals (훽̂ = −0.51 ± 0.23 SE and − 0.55 ± 0.22 SE, P = 0.02 and 0.01, respectively, k

= 33; Fig 2-2-2-5). I found weak evidence of shrub encroachment effects on bird Shannon diversity and community abundance which were of the predicted sign (훽̂ = 0.30 ± 0.16 SE, 푃 =

0.06, k = 23 for Shannon diversity and 0.13 ± 0.17 SE, 푃 = 0.45, k =

33 for total abundance). The effects of shrub encroachment on species richness did not differ among groups (minimum P = 0.17; Appendix D).

Distributions of r differed among continents but not among the three most-studied biomes across observational studies (P > 0.05; Appendix D). In Africa, relationships between shrub encroachment and vertebrate species richness were negative (훽̂ = −0.28 ± 014 SE, P = 0.04, k

= 43; Fig. 2-5). Studies in Australia, Europe, and North America, however, exhibited positive relationships between shrub encroachment and species richness (훽̂ = 1.48 ± 0.38 SE, 0.57 ±0.29

SE, and 0.81± 0.25 SE, P = 0.0001, 0.05, and 0.0012, for Australia, Europe, and North America, respectively, k = 43; Fig. 2-5). Similarly, shrub encroachment was associated with reduced total abundance in Africa but increased total abundance in North America (훽̂ = −0.36 ± 0.15 SE, and

0.53± 0.25 SE, P = 0.02, and 0.05, for Africa and North America, respectively, k = 33; Fig. 2-5).

However, I found no effects of shrub encroachment on Shannon diversity among continents (All

34

P > 0.05). There was evidence of unmodeled heterogeneity in all of the meta-regressions as determined by I2 (range: 0.62-0.96; Appendix E). Funnel plots also indicated substantial heterogeneity in effects across studies did not indicate any systematic bias (Appendix F).

Discussion

Shrub Encroachment Effects Across Climatic Gradients and Taxa

Using a global meta-analysis, I found no consistent effect of shrub encroachment on vertebrate diversity across the planet. Yet I identified variability in shrub encroachment effects on vertebrate communities among continents and vertebrate groups as well as with net primary productivity. Specifically, shrub encroachment effects on vertebrate communities were negative:

1) in arid environments where shrub encroachment is novel and can precipitate a biome switch

(Knapp et al. 2008); 2) among mammals and herpetofauna; and 3) in Africa. Collectively, these results provide a basis for focusing conservation efforts in arid grass-dominated biomes and an impetus for determining the mechanisms associated with reduced diversity and abundance of mammals and herpetofauna in shrub-encroached grass-dominated biomes.

My finding that the diversity and abundance of mammals and herpetofauna declined with shrub encroachment was surprising but explicable considering the traits of each vertebrate group.

Mammals and herpetofauna may exhibit less variation in use of cover and foraging modes than birds, for example, leading to increased sensitivity to shrub encroachment (Slagsvold, 2001;

Eldridge et al., 2011). Also, studies are generally limited to similar species, e.g. carnivores or rodents, which might magnify these effects. There may be encroachment effects among mammalian taxa along these lines that warrant further exploration. For example, the only available evidence indicated that carnivore richness and abundance was highest at intermediate shrub cover and lowest at high shrub cover, but this was based on studies from a single ecoregion

(Blaum et al., 2007a, b, Blaum et al., 2009).

35

Shrub Encroachment Effects Among Continents

I expected to find differences in shrub encroachment effects on vertebrates among continents but the pattern I observed did not conform to a common biogeographic theme. For example, I did not observe consistent effects between biomes, which were delineated largely based on coarse differences in precipitation and temperature (i.e. desert and xeric versus temperate or non-temperate grassland, savanna, and shrubland). I suspect this occurred in part because the prevailing biome classifications were not developed with a focus on vertebrate biology, but rather on abiotic controls driving plant distributions (Olson et al. 2001). Further, there were not consistent effects between the New World and the Old World, with their distinctive common , nor between the Northern and Southern Hemispheres, which have distinct climate histories (Hays et al., 1976). Rather, Africa was markedly different from North

America, Europe, and Australia. I expect that more recent events such as land-use intensification and species invasions have had different trajectories in Africa than the other continents

(Lonsdale 1999; Ellis & Ramankutty 2008), and that the increased diversity and abundance seen in North America, Europe, and Australia may therefore be driven by widespread, generalist species invading grass-dominated biomes. Understanding how species traits can explain variation in the effects of shrub encroachment could also help identify mechanisms for the variation I found.

Data Gaps and Limitations

There were striking geographic gaps in available studies of shrub encroachment effects on vertebrates in Asia, India, and Australia. Further, some biomes in these areas may be misclassified as despite extensive grass cover maintained by herbivory and fire (Parr et al.

2014; Ratnam et al. 2016). This collection of neglected “tropical grassy biomes” faces several threats yet I found no data about the effects of shrub encroachment on vertebrate diversity for

36

them. The extant studies have additional noteworthy imbalances in geographic and taxonomic coverage. For example, I found no data for Australia’s extensive arid grasslands (Fensham et al.

2005; Tassicker et al. 2006). Likewise, I found no data from the vast of continental Asia, and limited data on herpetofauna and mammals across continents. Altogether, these gaps limit our ability to discuss the effects of shrub encroachment on vertebrates across a substantial portion of the world’s grass-dominated biomes. Future research in these regions would help fill these important data gaps.

Implications for Ecology, Conservation, and Management

I found that increased net primary productivity was associated with reduced impact of shrub encroachment on vertebrate diversity. My results suggest that reducing factors that facilitate shrub encroachment, such as overgrazing, would be most effective in maintaining historic vertebrate diversity in desert grasslands and semi-arid savannas. My results also indicate that shrub thinning has been ineffective in reversing shrub encroachment effects on vertebrate communities, at least at the spatial and temporal scales studied to date. Therefore, prevention and mitigation measures may be more effective than restoration. Grass-dominated biomes maintained by human livelihoods, on the other hand, generally have high net primary productivity, so vertebrate community structure in such biomes should be less sensitive to shrub encroachment and a lower priority for mitigation efforts (e.g. Laiolo et al. 2004). Finally, since global CO2 concentration appears to be the most consistent driver of shrub encroachment (Wigley et al.,

2009; Stevens et al., 2016), conserving vertebrate biodiversity in grass-dominated biomes should be treated as a global-scale problem, and detailed prioritization schemes developed based on the available evidence.

My results show how climate and land-cover change can interact and contribute to community assembly in ways that could not be predicted from studying global change drivers in

37

isolation. I found no general effect of shrub encroachment encompassing all taxa and across the world, yet there were important differences in vertebrate responses to shrub encroachment that appear to be driven by climatic factors associated with different disturbance dynamics among the world’s diverse grass-dominated biomes. This result is interesting given that the strength and generality of interactions among global change drivers are hotly debated (Dukes & Mooney

1999; Foster et al. 2016). Surprisingly, I also uncovered evidence that the structure of bird communities has been resilient in the face of shrub encroachment while other taxa have not.

Thoughtful synthesis paired with experimentation will ultimately yield more refined answers about why these differences exist. My results provide a valuable heuristic basis for global-scale conservation prioritization of vertebrate communities in the interim.

38

Figure 2-1. Locations and mean annual precipitation associated with studies of shrub-encroachment effects on vertebrate community structure worldwide, 1978-2016.

39

Figure 2-2. Variation in vegetation structure across gradients of annual precipitation and shrub encroachment in a semi-arid savanna: the Kalahari rangelands of . The sites in each column are the same, photographed in either a wet year (top row) or a dry year (bottom row).

40

Figure 2-3. Shrub encroachment had no overall effect on the species richness, Shannon diversity, and abundance of vertebrate communities, nor does experimental thinning by fire, herbicides, or mechanical removal. Plotted values are standardized correlation coefficients, i.e., Fisher’s z-transformed r. Each value of r is reported under the “Fisher’s z” column with its corresponding 95% confidence interval in parentheses

41

Figure 2-4. Shrub encroachment reduced vertebrate Shannon diversity at low net primary productivity but had no effect at high net primary productivity across observational studies (P = 0.0019). I obtained similar but non-significant results for the effects of mean annual precipitation and the normalized difference vegetation index (NDVI; P = 0.1154 and 0.0657, respectively). Points are displayed at different sizes proportional to their respective precisions.

42

Figure 2-5. Shrub encroachment effects (r) on vertebrate community structure among groups and across continents. Bird (a) species richness, (b) Shannon diversity, and (c) total abundance (n = 28, 12, and 13, respectively) responses to encroachment were positive or neutral whereas those of herpetofauna (n = 8, 5, and 7), and mammals (n = 7, 5, and 13) were negative. The left panel is taxon and right panel is continent. Median shrub encroachment effects on vertebrates were negative in Africa (n = 18, 10, and 15) and South America (n = 3, 5, and 4) but generally positive in Australia (n = 3, 1, and 0), Europe (n = 7, 4, and 2), and North America (n = 12, 3, and 12).

43

CHAPTER 3 SHRUB ENCROACHMENT AND LAND-USE INTENSIFICATION IN AN AFRICAN SAVANNA HAVE DIFFERENT EFFECTS ON BIRD ALPHA AND BETA DIVERSITY

Synopsis

Human use of natural resources has altered multiple environmental gradients at once, causing large-scale land-use and land-cover change throughout the world’s ecosystems.

Communities differ in species composition and richness across these gradients of land-use and land-cover change, and the relative power of these gradients to explain local species richness

(alpha diversity) and changes in community composition among locations (i.e. beta diversity) can provide guidance for conservation. Further, beta diversity across environmental gradients can be partitioned into nestedness and turnover components, respectively. I determined alpha and beta diversity of birds across gradients of shrub encroachment and land-use intensity in the

Lowveld savanna of Swaziland using a multi-species occupancy model to correct for imperfect species detection. I then quantified the role of environmental gradients and distance in explaining species richness and beta diversity using generalized dissimilarity models, variance partitioning, and linear mixed models. Species richness increased with shrub encroachment and decreased with land-use intensification, suggesting that land-use intensification removed species and shrub encroachment augmented them. Shrub encroachment explained 56.2% of variability in beta diversity among local communities, compared to only 2.7% and 1.8% deviance explained by land-use intensity and distance, respectively. Turnover was the main driver of beta diversity across gradients of land-use intensity and shrub encroachment alike, while nestedness explained only 0.8-12.9% of beta diversity. Dissimilarity from nestedness was greatest at the highest levels of land-use intensity and shrub encroachment, indicating that extremes of both gradients removed species from savanna bird communities. Collectively, my results indicate that

44

maintaining bird communities typical of intact savanna may best be achieved by managing shrub cover.

Background

Global environmental changes are a function of multiple agents with interrelationships that are poorly understood (Wilson et al. 2006; Brook et al. 2008). There is a pressing need to examine relationships among multiple agents of global change in order to conserve diverse biotic communities and their associated functions (Didham et al. 2007; Turner et al. 2007). Examining biodiversity across environmental gradients arising from multiple global-change agents can inform conservation planning and suggest which community assembly mechanisms are important (Tylianakis et al. 2007; Devictor et al. 2010).

One particularly salient example of multiple agents of environmental change occurs in savannas. These areas are experiencing woody encroachment and land-use intensification, both of which are altering the structure and function of animal communities (Egger et al. 1997; Flynn et al. 2009; Mac Nally et al. 2009). Land-use intensification and land-cover change are particularly evident in African savannas, creating challenges for biodiversity conservation

(Lambin et al. 2003). Relatively little land has been formally protected in African savannas whilst land use has been intensifying as communal pastures are converted to homesteads interspersed with -fed or high-input plantation agriculture (Musters et al. 2000;

Balehegn 2015). Meanwhile, land-cover change in African savannas has also been driven by shrub encroachment, which occurs across land uses when woody cover replaces grasses as a consequence of enriched atmospheric CO2 and changes in disturbance regimes maintained by fire and grazing (Egger et al. 1997; Stevens et al. 2016a).

Alpha and beta diversity provide important complementary information for understanding how land uses, shrub encroachment and other gradients structure communities

45

(Anderson et al. 2011; Muenkemueller et al. 2012; Socolar et al. 2016). Alpha diversity, also known as local species richness, is an indicator of ecosystem function and stability as well as a frequent target of conservation (Maestre et al. 2012; Gamfeldt et al. 2013). However, alpha diversity provides incomplete information about the effects of environmental change on biodiversity. For example, species richness can be high in human-altered despite little contrast in species composition among sites (i.e. beta diversity), such as where species invasions have occurred (Socolar et al. 2016). Beta diversity summarizes changes in species occurrence or abundance across space and environmental gradients (Socolar et al. 2016). Beta diversity can be partitioned into nestedness and turnover. Nestedness occurs when species in one site are a subset of species in another, whereas turnover occurs when species replace one another (Baselga &

Orme 2012). Nestedness across environmental gradients has been attributed to systematic species losses as conditions become increasingly stressful (Worthen et al. 1998; Hylander et al. 2005).

Turnover, on the other hand, has been treated as an indicator of species sorting into conditions suitable to their respective niches (Parmentier & Hardy 2009).

Land-use intensification should primarily lead to nestedness while land-cover change from shrub encroachment may be more likely to alter turnover, with increasing land-cover heterogeneity promoting turnover. Species richness has been negatively associated with land-use intensification (e.g. agricultural conversion) whereas the effects of land-cover change on species richness are less clear (Eldridge et al. 2011; Verhulst et al. 2004; Waltert et al. 2004). Land-use intensification can have broad homogenizing effects on communities, associated with reduced species richness and nested species losses (Benton et al. 2003; Flynn et al. 2016; Gossner et al.

2016). Shrub encroachment, on the other hand, can promote land-cover heterogeneity, leading to increased species turnover (Kerr et al. 2001). Shrub encroachment in African savannas can

46

increase land-cover heterogeneity, potentially promoting beta diversity, particularly turnover.

Both land-cover change and land-use intensification can lead to shifts in community composition away from the baseline conditions generally emphasized by conservation practitioners (Skowno

& Bond 2003; Flynn et al. 2016). Quantifying how biotic communities change in composition across gradients of land-cover change and land-use intensification can therefore inform conservation by making the expected magnitude and consistency of changes attributable to these two key types of change clear.

Birds form speciose, readily-observed communities that have demonstrated differences in abiotic tolerances that affect their patterns of occurrence and are capable of traversing areas encompassing multiple patches with different land-use and land-cover properties (Cockburn

2006). Quantifying alpha and beta diversity across environmental gradients is therefore particularly relevant for birds because we know little about which gradients contribute the most to community change.

I surveyed birds to determine alpha and beta diversity across gradients of shrub encroachment and land-use intensification in the Lowveld savannas of Swaziland, where protected areas and pastoral lands are being converted to subsistence agriculture and sugar (Bailey et al. 2016). My objectives were to: 1) determine the relative magnitude of shrub encroachment and land-use intensification effects on alpha and beta diversity; 2) estimate how much variation in alpha and beta diversity were attributable to shrub encroachment and land-use intensity whilst accounting for geographic distance; and 3) partition beta diversity into its nestedness and turnover components across gradients of land-use intensity and shrub encroachment to understand how beta diversity changes across these gradients. I made several predictions. First, I expected that species richness would increase with shrub encroachment and

47

decline with land-use intensification because adding woody cover creates additional nest sites and foraging opportunities whereas land-use intensification creates multiple biotic stresses

(Foley et al. 2005). Second, I expected that beta diversity would be most strongly associated with shrub encroachment because encroachment created opportunities for species turnover, whilst land-use intensification and geographic distance would play lesser roles because both should contribute to beta diversity primarily via nestedness. I also predicted that the importance of nestedness in describing beta diversity would be greatest at the extreme ends of both the shrub encroachment and land-use intensification (Worthen et al. 1998). Finally, I predicted that land- use intensification would be at least as important as shrub encroachment in explaining beta diversity because both gradients can have large effects on community structure and composition

(Kerr et al. 2001; Sirami & Monadjem 2012).

Materials and Methods

Study Area

I sampled birds across gradients of shrub encroachment and land-use intensity throughout the Lowveld of Swaziland. The Swazi Lowveld is primarily savanna with a continuous grassy understory and patchily-distributed woody plants; mean annual precipitation ranges from 550-

725 mm (Goudie & Price Williams 1983; Monadjem 2000). Several native woody plants are increasing and contributing to shrub encroachment, including prickly acacia, Vachellia nilotica sicklebush, , and buffalo thorn, Ziziphus mucronata, (Monadjem 2000;

Loffler & Loffler 2005). The region experiences a mean monthly temperature of 26°C in January when most bird species are breeding (Sirami & Monadjem 2012). The land-use mosaic in the

Swazi Lowveld is mainly comprised of community pastures where people keep cattle and other livestock, homesteads where people practice rain-fed subsistence agriculture, and sugar cane plantations, which have been present since the 1950s and are increasing in extent (Bailey et al.

48

2016). There are also several protected areas in the Swazi Lowveld where recreation and, in some cases, limited cattle grazing are the only permitted activities (Hurst et al. 2014). Urban areas are limited in extent and primarily associated with sugar cane operations (Bailey et al.

2016).

Sampling Design

Stratification by land use and shrub encroachment

I sampled across the four major land uses and the full range of shrub cover present within each land-use type by employing a stratified sampling design. I visited six protected areas

(range: 700 to 30,000 ha), seven community pastures, seven homesteads, and a large sugar cane plantation (3800 ha; http://www.huletts.co.za/ops/swaziland.asp). I followed an animal care and use protocol approved by the University of Florida (protocol #: 201509045) and obtained permission to access study sites from local chiefs, landowners, and protected area managers as appropriate.

I divided the study area according to land-use intensity, i.e. protected areas, communal rangelands, homesteads with rain-fed crops, and sugar plantation (ordered from least to most intensive). I then collected a stratified sample of diurnal bird communities across the gradient of shrub encroachment where present (protected, , and homestead) and without regard to shrub cover in sugar plantation. I chose starting points in patches of low, medium, or high shrub cover in those land-use types and delineated 3 x 3 square grids with 9 points ≥250 m apart wherever possible— the homestead grids followed footpaths or roads because each homestead is a fenced compound, but I approximated a grid shape inasmuch as I could. Points within a grid were generally >1.5 km from any point in the nearest neighboring grid to maximize independence among grids. I established multiple grids with similar shrub cover on a if no grid with markedly different average shrub cover could be identified in order to obtain more

49

samples. I ultimately sampled 371 points distributed across 42 grids divided among 20 locations, with each location defined as a distinct village, plantation, or protected area.

Bird surveys

I repeated10-minute point counts 3-4 times at each point (mean: 3.97), between 11

December 2014 and 15 March 2015, counting all birds detected within 50 m. I recorded the observer, date, time, temperature, wind speed on a Beaufort scale, and percent cloud cover twice each morning. I rotated two observers among grids and surveyed from 30 mins before sunrise to

≤5 hrs after, or ≤3 hrs before sunset. If it was raining or wind speeds were >20 km-hr-1, I did not survey. I visited grids in a different sequence from one visit to the next to ensure all points were visited at different times of day.

Patch vegetation structure

I sampled patch vegetation structure at each point by measuring grass, shrub, and tree cover. I defined trees as woody plants >3 m and shrubs as woody plants ≤ 3m after Sirami and

Monadjem (2012). I measured grass, shrub, and tree cover using the line intercept method with three replicates per point and a 50 m tape. I measured grass, shrub, and tree cover using the line intercept method with three replicates per point and a 50m tape (Tansley & Chipp 1926). I put the three lines at each point at 0, 120, and 240 degrees orientation using an iPhone 4 compass

(Apple Inc., Cupertino, , USA), calculated mean percent cover contacting a line for grass, shrub, and tree cover and treated them as estimates of percent cover for each cover type.

Analytical Methods

Estimating species presence-absence and richness

The reliability of any community description is dependent on data quality and should be corrected for detection error when possible (Mihaljevic et al. 2015). I fitted a hierarchical occupancy model with parameter-expanded data augmentation that accounts for imperfect

50

species detection, including species not detected on any occasion, i.e. underestimation of species richness (Dorazio and Royle 2005). The model uses the detection histories of species to produce corrected estimates of species richness among sites using jags called through the R package jagsUI (Plummer 2003; Dorazio & Royle 2005; "DR metacommunity model" hereafter ; Kellner

2016; R Core Team 2016). The DR metacommunity model also permits inference at the metacommunity, local community (“points” in my analyses), and individual species levels (Kery

& Royle 2015). I focused on modeling heterogeneity in species detection (p) keeping occupancy probability (휓) constant, fitting a single model with random effects of sampling grid and time of day, complemented by linear and quadratic effects of date that were allowed to vary among species, because species exhibit different breeding schedules that could have affected their detectability.

I used flat normal priors with mean 0 and precision 0.1 for fixed effects and a uniform prior with mean 10 and precision 0.01 for random effects (Appendix K; Gelman & Hill 2007). I fitted the DR metacommunity model twice as suggested by Kery and Royle (2015) to manage computer memory. I first ran four chains for 100,000 iterations, discarding the first 5,000 as burn-in, and thinned each chain by collecting every 20th observation, collecting predicted species richness for each local community from the results and checking all model parameters for convergence using the Gelman-Rubin statistic and visual inspection of chains (Gelman & Rubin

1992). I then ran the model again under the same conditions, this time collecting only the

Markov chain Monte Carlo (MCMC) samples needed to assemble the z-array (defined below) from the model results.

I extracted estimated species richness at each sampling point (“local species richness” hereafter), directly from the model. Species richness in the DR metacommunity model was

51

calculated by estimating the occupancy status of a number of hypothetical species that may have occurred but were never detected and adding those species to those which were detected (Kery and Royle 2015). I then generated a site by species predicted presence-absence matrix (“z matrix” hereafter) by reducing the z array in the DR metacommunity model, which consists of n slices of site by species predicted presence-absence matrices, where n is the number of MCMC samples. I took the median value across all samples in the z array to obtain a z matrix. The z matrix value is 1when a species is observed on ≥1 occasion (MacKenzie et al. 2003; Kery &

Royle 2015), and all other cell values of 1 in a z matrix are based on model predictions where detections did not occur. Finally, I estimated beta diversity using the Jaccard dissimilarity for each pair of local communities from the z matrix using the vegdist function in the vegan R package, and used it in all subsequent analyses (Oksanen et al. 2016). I chose Jaccard rather than

Bray-Curtis dissimilarity because the Jaccard index meets all of the conditions of a good dissimilarity index outlined by Oksanen et al. (2016).

I determined the associations between species richness, shrub encroachment, and land- use intensification by regressing estimated species richness on shrub cover and land-use type in separate linear regression models. I included both linear and quadratic percent shrub cover terms in the shrub encroachment model to permit a species richness to peak at intermediate shrub cover if consistent with the data.

Generalized dissimilarity modeling

I quantified the variance in bird beta diversity explained by shrub encroachment, land-use intensification, and geographic distance using generalized dissimilarity modeling (Ferrier et al.

2007). Generalized dissimilarity models are well-suited to quantifying relationships between environmental gradients and beta diversity because: 1) they allow for nonlinearities in covariate relationships by combining flexible spline functions with a generalized linear model; and 2)

52

generalized dissimilarity model results can be compared against appropriate null expectations using bootstrap methods for both significance testing and variance partitioning (Ferrier et al.

2007; Fitzpatrick & Lisk 2016). I used the gdm R package that supports these features

(Fitzpatrick & Lisk 2016). Specifically, I used the gdm function to fit a model with percent shrub cover, land-use intensity, and geographic distance as covariates and used the default spline and knot locations, i.e. at the minimum, median, and maximum values for each respective covariate

(Fitzpatrick & Lisk 2016). I scaled land-use intensity from 0-1 in equal increments in order to produce a continuous measure (Baselga 2010; Baselga & Orme 2012). In other words, I assigned protected areas, community pastures, homesteads, and sugar plantations land-use intensity values of 0, 0.33, 0.66, and 1, respectively, in a model that allowed for nonlinear responses that could vary with increasing land-use intensity using spline functions. I then used the gdm.varImp function to implement a matrix permutation procedure that permuted the data 100 times in order to determine percent deviance explained and test statistical significance for each covariate

(Ferrier et al. 2007, Baselga & Orme 2012). I plotted the model results using the plot.gdm function to display the shape and relative magnitude of each gradient’s effect on beta diversity.

Visualizing community change across environmental gradients

I wanted to provide a qualitative, visual representation of how environmental gradients contributed to changes in community composition relative to a protected, open savanna baseline.

I did this by using applying non-metric multi-dimensional scaling to Jaccard dissimilarities among sites computed from the z-matrix to represent differences in community composition among sampling points in two dimensions, indicating the communities associated with a given land-use type with 95% ellipses and the vector of shrub encroachment effects in ordination space

(Oksanen 2015). I only proceeded with plotting the ordination if I observed stress values < 0.3, indicating that differences in community composition could be adequately represented in two

53

dimensions (Oksanen 2015). I performed all ordinations in the vegan R package using the functions “metaMDS,” “ordiellipse,” and “envfit” (Oksanen 2015).

Partitioning dissimilarity owing to turnover and nestedness

I used the R package betapart to determine the proportion of beta diversity attributable to turnover and nestedness across gradients of shrub encroachment and land-use intensification

(Baselga & Orme 2012). I used the beta.sample function to perform matrix permutation and determine the amount of species dissimilarity attributable to turnover and nestedness for each land-use type, with an estimate of uncertainty. I randomly drew 25% of sites within each land use 1000 times in order to get these estimates and plotted the density function for each land use.

Results

I recorded a total of 209 species, which were detected in 1-314 points (median: 9;

Appendix H). Observed species richness (richness not adjusted for detection probabilities; see below) averaged 16.3±0.32 SE (range: 1-40) and decreased with land-use intensity (means:

18.2±0.32 SE, 17.3±0.26 SE, 14.7±0.26 SE, and 8.2±0.28 SE for protected areas, pastures,

2 homesteads, and sugar plantation, respectively; F3, 367 = 835.5, P < 0.0001; R = 0.90). Shrub cover differed among land uses, ranging from 0-100% in protected areas and community pastures (means 32±1.3% SE and 41±1.8% SE respectively) but 0-35% in homesteads and 0-

24% in the sugar estate (means 7±0.4% SE and 2±0.3% SE respectively (F3, 367 = 38.04, P <

0.0001). I found pairwise differences in mean shrub cover among all land uses except between homesteads and sugar plantation (Tukey’s HSD adjusted P = 0.76 for homestead-sugar plantation versus 0.00-0.03 for all other post-hoc comparisons).

The DR metacommunity model indicated that 26.7%±0.5 (SE) of species occurring in local communities were detected over all visits (Appendix I). Predicted point species richness averaged 60±0.20 SE (range: 5-78; Appendix I). Associations between species richness and

54

shrub encroachment were similar after correcting for species non-detection, as were the relationships between species richness and land-use intensification, which explained most of the variation in species richness (Fig. 3-1; 푠ℎ푟푢푏 푚표푑푒푙: 훽푠ℎ푟푢푏 = 0.11 ±

2 0.02 푆퐸 푎푛푑 훽푠ℎ푟푢푏−푠푞 = −0.0010 ± 0.0003 푆퐸; 푃 < 0.0001; 푅 = 0.07; 푙푎푛푑 −

2 푢푠푒 푚표푑푒푙: 퐹3,367 = 24840, 푃 < 0.0001; 푅 > 0.99).

The generalized dissimilarity model revealed that beta diversity was primarily associated with shrub encroachment and increased nonlinearly with shrub cover, whilst land-use intensification and geographic distance had smaller effects that were also nonlinear (percent deviance explained: 56.2, 2.7% and 1.8% for shrub encroachment, land-use intensification, and geographic distance, respectively, all P < 0.001; Fig. 3-2). Non-metric multi-dimensional scaling plots indicated that bird community composition varied among land-use type (Fig. 3-3).

Turnover was the main component of beta diversity across both the land-use intensification and shrub encroachment gradients (range: 43-98% of beta diversity; Fig. 3-4). The contribution of turnover to beta diversity was similar among land-use types, sugar plantation excluded, but declined with shrub encroachment from 92% (0.003 SD) to 30% (0.035 SD) in the sites with >

80% and < 20% shrub cover, respectively (Fig. 3-4). The importance of nestedness in explaining beta diversity increased with shrub encroachment and was markedly different in sugar plantation than in less-intensive land uses (range: 1-13% of beta diversity, 9% and 13% in plantation and >

80% shrub cover, respectively; Fig. 3-4), meaning that the maxima of both gradients were associated with a higher fraction of beta diversity explained by nestedness.

Discussion

I identified shrub encroachment as the most important and consistent correlate of increasing beta diversity in savannas despite previous research establishing that both shrub

55

encroachment and land-use intensification can have deleterious effects on biodiversity in grassy biomes (Ceballos et al. 2010; Sirami & Monadjem 2012; Buij et al. 2013). I also observed decreases in species richness with increased land-use intensification, contrasting with modest increases in richness with shrub encroachment. Such relationships suggest that increasing shrub cover might offset localized species losses resulting from land-use intensification, but the identity of the species gained would probably not be characteristic of intact savanna.

The prevalence of turnover relative to nestedness is an indication that localized protection of species-rich patches may not maximize alpha and beta diversity of birds in Africa’s changing savannas (Baselga et al. 2012). Landscapes characterized by high species turnover are likely comprised of environmental gradients with unrelated effects that are difficult to capture with reserves (Parmentier & Hardy 2009; Baselga et al. 2012). Landscapes characterized by nestedness, on the other hand, should be comprised of gradients that are more homogenous in their effects across space, making site prioritization more straightforward (Baselga et al. 2012).

Similarly, management efforts aimed at preventing localized extinctions in savanna might focus on limiting the number of shrub and intensive agriculture in the because both extremes were associated with increased contributions of nestedness to beta diversity. The nestedness I saw with extremes of shrub encroachment and land-use intensification presumably resulted from stressors that remain to be identified, such as simplified and reduced foraging opportunities resulting from structural homogeneity in both , and perhaps applications in plantations (Worthen et al. 1998; Hylander et al. 2005). Some of these stressors may be amenable to mitigation, so identifying them would be useful. However, conservation programs aimed at maximizing alpha and beta diversity by providing wildlife habitat on working lands (i.e. “land-sharing;” Phalan et al. 2011) would create bird communities that are

56

substantially different in composition than those found in extensive open savannas protected from human exploitation.

I was surprised to discover that shrub cover was a more important and consistent driver of beta diversity than land-use intensification. This conclusion should be regarded as tentative because the results may be sensitive the manner in which I converted land use from a categorical variable, that may not be ordinal, to a continuous variable in order to operate within an established framework for generalized dissimilarity modeling. Indeed, looking at the bird community dissimilarities plotted in ordination space, it appears that bird community composition is dramatically different among land uses in a way that may not have been captured by the generalized dissimilarity model. This has two important implications. First, my results might systematically underestimate the importance of land-use type as a driver of bird community composition. Second, the usefulness of land-use intensity as an explanatory variable requires critical examination, since it implicitly assumes that land-use intensity is an ordinal measure of stress on bird species. Land-use intensity may be much less useful in capturing idiosyncratic (species-specific) responses to particular land-use types.

My findings demonstrate that communities can exhibit opposing responses to different global change gradients, even when each change gradient has known deleterious effects on species and ecosystems (Bennett et al. 2004; Sirami & Monadjem 2012; Broms et al. 2014;

Ogada et al. 2016), which may be superficially surprising but was straightforward to predict from the established effects of land-cover and land-use change, i.e. land-cover change as a source of heterogeneity and land-use change as a stressor that removes species. Fortunately, such opposing responses may be common and can be leveraged toward two critical ends: 1) to guide land-use planning and land-cover management toward particular conservation outcomes; and 2) to

57

stimulate focused inquiry into community assembly processes. The first end is proximate and pragmatic; the second ultimate yet also useful, because our confidence in generalizing from results such as ours will be contingent on whether the specific mechanisms involved can be expected to occur across a variety of settings.

58

Figure 3-1. Species richness increased with shrub cover and decreased with land-use intensification. Colors in the scatterplot correspond to different land-uses and match the color scheme in the boxplot. Best-fit line on left is a regression of species richness on shrub cover with a quadratic term included and dashed lines indicating the upper and lower 95% confidence bounds.

59

Figure 3-2. Bird species dissimilarity (i.e. “partial ecological distance”) in Lowveld savanna increased rapidly with shrub cover across nearly all values whilst dissimilarity increased with geographic distance but quickly attained an asymptote at a level of turnover similar to that observed at maximum land use intensity, according to a generalized dissimilarity model fit to modeled species presences and absences derived from survey data collected in Swaziland, December 2014 – March 2015.

60

Figure 3-3. Differences in bird community composition, i.e. Jaccard dissimilarity based on predicted species occurrence, among sampling locations as rendered by non-metric multi-dimensional scaling, with communities characteristic of particular land-use types indicated by 95% ellipses, and the direction of community compositional shifts associated with shrub encroachment indicated by an arrow.

61

Figure 3-4. Bird community dissimilarity was mainly driven by species turnover, and this was consistent across land uses other than sugar plantation, where nestedness was a more important driver of species dissimilarity. Each curve is a resampling-based distribution obtained by randomly drawing 25% of the data within each land use 1000 times and partitioning turnover and nestedness, respectively.

62

CHAPTER 4 DIET AND BODY SIZE EXPLAIN SHRUB ENCROACHMENT AND LAND-USE EFFECTS ON BIRD OCCUPANCY IN AN AFRICAN SAVANNA

Synopsis

Land-cover change and land-use intensification are two of the most important drivers of vertebrate biodiversity loss, yet interactions between them, i.e. synergies, remain poorly understood. Predicting the effects of land-cover change and land-use intensification requires understanding how species vary in their responses to these drivers, and species traits are increasingly used in ecology and conservation to generalize and predict global change effects on . Savannas worldwide are experiencing simultaneous shrub encroachment and widespread land-use intensification that has affected the structure and composition of animal communities yet the role of species traits in determining these effects is largely unexamined.

Birds possess diverse traits and occur across broad environmental conditions, making them an excellent taxon for examining whether species traits can explain global change effects on animals. I sampled bird occurrence across gradients of shrub encroachment and land-use intensity throughout the Lowveld of Swaziland (~6900 km2) to determine whether diet and body size explained heterogeneity in species’ responses to shrub encroachment and land-use intensification. I combined hierarchical occupancy models for 48 species with a meta-analytic framework to make this determination. I found that interactions between land-use type and shrub encroachment were surprisingly absent. Rather, land-use intensification had negative effects on birds overall, among several diet groups, and on large body size. Shrub encroachment effects, on the other hand, were positive for frugivores and predatory birds, and varied little across body sizes. My results show that the effects of shrub encroachment on bird communities can be consistent across land uses, which suggests common mechanisms underpinning community dynamics in encroached savanna, such as cover, food, or predator-prey interactions. Further, the

63

differing effects of shrub encroachment and land-use intensification may provide opportunities to manage biodiversity in mosaic landscapes by managing shrub cover.

Background

People are changing the world at an unprecedented pace, leading to biodiversity loss, altered species interactions, and the of ecosystem services (Mooney et al. 2009). Further, numerous anthropogenic stressors are working simultaneously and interacting to accelerate extinctions and alter ecosystem processes (Chapin et al. 2000; Brook et al. 2008). Land-use intensification and land-cover change are two of the world’s most widespread and important drivers of biodiversity loss (Sala et al. 2000; Lambin et al. 2003). Land-use intensification occurs when people increase material inputs or labor applied to a landscape in order to produce more products or services per unit area such as when land dedicated to rain-fed subsistence agriculture is converted to plantation monoculture (Ellis & Ramankutty 2008). Land-cover change, on the other hand, occurs when the prevailing vegetation structure in a landscape shifts to a different state, regardless of land use (Turner et al. 2007). For example, shrub encroachment is occurring in savannas worldwide across a variety of land uses (Egger et al. 1997; Ellis & Ramankutty

2008).

Savannas cover 20% of the Earth’s terrestrial land mass, provide livelihoods for >1 billion people, and include multiple biodiversity hotspots (Myers et al. 2000; Lehman & Parr

2016). However, savanna biodiversity appears to be under threat from widespread simultaneous land-use intensification and land-cover change. Land-use intensification in savannas is characterized by a combination of: 1) conversion of protected areas to pasture and rain-fed agriculture to directly support increasing human populations; and 2) conversion of protected areas, pastures, and rain-fed agriculture to plantation monocultures to support , i.e. jobs and foreign exchange (Holden & Otsuka 2014). Land-use intensification

64

can have strong and wide-ranging effects on animals by causing habitat fragmentation and loss, indirectly affecting cover, food resources, and myriad anthropogenic stressors to animals such as , poaching, and noise (Foley et al. 2005). Consequently, land-use intensification generally has negative effects on habitat specialists and species of large body size (Owens &

Bennett 2000; Lambin et al. 2003). Land-cover change in savannas is largely characterized by shrub encroachment, which occurs across multiple land uses as a result of atmospheric CO2 enrichment, overgrazing by livestock, loss of browsing by wild megaherbivores, and fire suppression (Stevens et al. 2016a). Shrub encroachment also alters animal occurrence through habitat fragmentation that reduces some species and promotes others by changing cover and food resources (Seymour & Dean 2010; Sirami & Monadjem 2012). However, we know little about how the effects of shrub encroachment on the occurrence of species vary with animal diets and body sizes. Shrub encroachment might result in more habitat gains than losses because shrubs provide food resources for animals such as fruit, nectar, and arthropods (Burkett et al. 2005;

Foley et al. 2005). On the other hand, shrub encroachment increases the amount of temporally stable cover in the environment, which might reduce predation risk across species, altering competitive outcomes among prey species (Hanski 1983; Turner et al. 2007).

The diversity of processes affected by land-use intensification and shrub encroachment suggests synergies may be occurring that affect animal communities, but we lack a conceptual framework for predicting such synergies. However, we have some understanding of how synergies between land-use intensification and shrub encroachment could be predicted from species traits based on previous global change studies (Hansen et al. 2001; Bohlen et al. 2004).

Certain species traits are correlated with extirpations and extinctions; these traits could be used to predict global change effects generally (Brook et al. 2008). For example, both body size and

65

dietary specialization are positively correlated with extinction risk across multiple vertebrate classes (Owens & Bennett 2000; Purvis et al. 2000). Similarly, land-use intensification and shrub encroachment may be shaping animal communities by favoring different traits, so that animals with combinations of traits favored or disfavored by both will exhibit synergistic responses to land-use and land-cover change. Among birds, for example, land-use intensification is associated with human persecution that disproportionately affects large-bodied species whilst habitat loss negatively affects specialist species (Owens & Bennett 2000). Shrub encroachment, on the other hand, may increase or reduce habitat depending on species and I know of no reason shrub encroachment effects should be explained by body size. Thus one might predict that shrub- encroached savanna subject to heavy human use will favor small-bodied diet generalists over large-bodied diet specialists. However, changes in community composition in response to shrub encroachment have seldom been evaluated in terms of species traits and across gradients of land- use intensity (Kutt & Martin 2010; Seymour & Dean 2010) and whether synergies occur is entirely untested.

Shrub encroachment can vary in quantity and extent among land uses in savannas.

Consequently, understanding the effects of shrub encroachment on animals in savanna requires consideration of land-use intensification as well as species traits. A typical southern Africa savanna is an interspersed land-use mosaic of protected areas, community pastures stocked primarily with cattle, homesteads where people practice subsistence farming, and, increasingly, sugar cane plane plantations (MacKenzie et al. 2002; Ferrier et al. 2007; Collier et al. 2011).

Shrub encroachment is minimal on plantations and often minor in homesteads, where bare ground and crops are the predominant land-cover types (Twine et al. 2016). The association between land-use intensification and land-cover conversion means that land-use intensification

66

should limit shrub encroachment. This may seem self-evident in agricultural monocultures but the evidence for such interactions in less-intensive land uses is mixed (MacKenzie et al. 2002;

Tredennick et al. 2015; Twine et al. 2016). Intensive cattle grazing and altered fire regimes are associated with shrub encroachment in community pastures because they favor increased woody biomass, including pastures that have been converted to shrub thickets (Grant et al. 2004; Twine et al. 2016). However, people in sub-Saharan Africa also collect shrubs for firewood and building material from rural homesteads and community pastures (Gaugris & Van Rooyen 2010;

Twine et al. 2016). Wood collection can locally reduce shrub cover yet harvest may not be able to reverse encroachment because of ongoing overgrazing, altered fire regimes, and rapid regeneration of cut shrubs by coppicing (Balehegn 2015; Twine et al. 2016).

Bird communities provide an ideal community for understanding possible synergies between land-use intensification and shrub encroachment effects because bird traits are well- described and vary considerably among species; birds are also conspicuous enough to be counted across suitably large spatial scales using surveys (Whelan et al. 2015). I surveyed bird communities in a savanna to determine relationships between bird occupancy and shrub encroachment in protected areas, community pastures, homesteads, and sugar cane plantation in southern Africa. My objectives were to: 1) determine the additive and interactive effects of shrub encroachment and land-use intensity on the occupancy of bird species and; 2) determine the explanatory power of diet and body size as predictors of those effects among species. I predicted that shrub encroachment would be associated with a shift away from granivorous species toward mixed-feeding frugivores and nectarivores whilst insectivore occupancy would remain the same and body size would be unaffected. I expected increasing land-use intensity to favor smaller body sizes because habitat loss and fragmentation accompanied by some mix of persecution,

67

disturbance, bioaccumulation of pesticides, and extraction of production for human use should disproportionately affect larger species (Owens & Bennett 2000). Further, I predicted that land use by shrub encroachment interactions would be present. Specifically, I predicted that 1) large- bodied granivores would exhibit the largest interactions; 2) small-bodied mixed-feeding birds would exhibit the smallest interactions; and 3) shrub encroachment effects on all bird diet groups would decline with increasing land-use intensity because factors associated with land use would take precedence.

Materials and Methods

Study Area

I sampled birds across gradients of shrub encroachment and land-use intensity throughout the Lowveld of Swaziland (~6900 km2; Fig. 4-1). Lowveld savanna is characterized by a grass layer interspersed with woody plants (Hanan & Sankaran 2003). Several native woody plants are increasing and contributing to shrub encroachment, including sicklebush, Dichrostachys cinerea, buffalo thorn, Ziziphus mucronata, and prickly acacia, Vachellia nilotica, (Monadjem 2000;

Loffler & Loffler 2005). The Swazi Lowveld experiences a mean monthly temperature of 26°C in January when most bird species are breeding and typically receives 550-725 mm of rain annually (Goudie & Price Williams 1983; Sirami & Monadjem 2012). The land-use mosaic in the Swazi Lowveld is mainly comprised of community pastures, where people rear livestock, and homesteads, where people keep their homes and practice rain-fed subsistence agriculture, but sugar cane plantations have been present since the 1950s and are increasing in extent (Bailey et al. 2016). There are also several protected areas in the Swazi Lowveld where recreation and, in some cases, limited cattle grazing are the only permitted activities (Hurst et al. 2014). Urban areas are limited in extent and primarily associated with sugar cane operations (Bailey et al.

2016).

68

Sampling Design

Stratification by land use and shrub-encroachment

I sampled across the four predominant land uses—protected area savannas, communal rangelands, homesteads, and sugarcane—and the full range of shrub cover present within each land-use type by employing a stratified sampling design. I visited six protected areas (range: 700 to 16,000 ha; United Nations Development Programme Global Environment Facility 2013), seven communal rangelands, seven adjoining homesteads, and a large sugar cane plantation

(3800 ha; http://www.huletts.co.za/ops/swaziland. asp). I sampled within protected areas in the

Swazi Lowveld with permission from the appropriate authorities and in neighboring villages with permission from local chiefs. I distinguished communal rangelands from homesteads and associated farm fields in each village by consulting with local chiefs and their designees, treating them separately because of large differences in human and livestock activity as well as vegetation structure. All research followed an animal care and use protocol approved by the

University of Florida (protocol #: 201509045).

I divided the study area into land-use types and collected a stratified sample of diurnal bird communities across a gradient of shrub encroachment where present (protected, rangeland, and homestead) by selecting patches of low, medium, or high shrub cover. I defined low, medium, and high shrub cover as <~20%, ~20-60%, and >~60% mean cover as determined by visual estimation of 2 perpendicular ~1km transects of candidate patches. I then established square grids with three rows and three columns of points ≥250 m apart wherever possible— the homestead grids followed footpaths or roads because each homestead was fenced, but I attempted to replicate a square shape whenever possible. Points within a grid were generally >1.5 km from any point in the nearest neighboring grid to maximize independence among grids. I created multiple grids with similar shrub cover within a property if no grid with visibly different

69

average shrub cover could be established in a community in order to obtain more samples. I ultimately sampled 42 grids containing 371 sampling points divided among 20 properties or homesteads.

Bird surveys

I conducted repeated 10-minute point counts at each point, counting all birds detected within 50 m. I recorded the observer, date, time, and wind speed on a Beaufort scale, twice each morning. I visited each point 3-4 times (mean: 3.97) during the breeding season, 11 December

2014 - 15 March 2015. I rotated two observers among grids and surveyed between 30 mins before sunrise and 5 hrs after, or within 3 hrs of sunset. I did not survey when it was raining or wind speeds were >20 km/hr. I ran grids in a different sequence each visit so points within a grid were visited at different times of day.

Patch vegetation structure

I sampled patch vegetation structure at each point by measuring grass, shrub, and tree cover. I defined trees as woody plants >3 m and shrubs as woody plants ≤ 3m after Monadjem

(2005) and Sirami and Monadjem (2012) . I measured percent grass, shrub, and tree cover using the line intercept method with three replicates per point and a 50 m tape (Tansley and Chip

1926). I placed the three lines at each point at 0, 120, and 240 degrees orientation using an iPhone 4 compass (Apple Inc., Cupertino, California, USA). I calculated the mean of the line intercept values recorded at each point for grass, shrubs, and trees and treated them as estimates of percent cover for each cover type. I also made boxplots of percent shrub cover for each land use to display the apparent direct relationship between land-use intensity and shrub encroachment.

70

Analytical Methods

Species traits

I assigned a mass and diet group to each species using Hockey et al. (2005; Appendix 1).

I binned each species into a diet group based on the most common food type described in

Hockey et al. (2005), e.g., I classed as frugivores any species that ate mainly fruits although most frugivores also consume arthropods. I used the mean female mass when reported and the mean of mixed or unknown sex samples otherwise. Finally, 28 species of predatory birds occur in the

Swazi Lowveld, including several common non-raptor species (Hockey et al. 2005). I determined whether each species I detected was a predator or not and treated predatory status as a trait distinct from diet because predation is an important mechanism that shapes communities

(Hairston et al. 1960; Holt 1977).

Quantifying occupancy

I estimated species occupancy using Bayesian hierarchical occupancy models

(MacKenzie et al. 2002; Royle & Dorazio 2008). Occupancy models account for imperfect detection using repeated observations such as species’ detection histories and can be used to model occupancy as a function of covariates (MacKenzie et al. 2002). However, occupancy models are biased at low proportions of sites detected (prevalence hereafter), <~10%, and occupancy models fitted using maximum likelihood methods cannot reliably estimate parameters when few or no observations span gradients of interest (Royle & Dorazio 2008). I therefore only modeled species observed at ≥10% of the points I visited and used a Bayesian model to estimate the parameters of interest by shrinking effects toward zero in cases where information was sparse

(Royle & Dorazio 2008).

I fitted two models for each species: one that included the additive and interactive effects of land use and shrub cover plus a random effect of sampling grid on occupancy (because points

71

within a grid cannot be assumed independent) using means parameterization (Kery & Royle,

2015), and another without a land use by shrub cover interaction. I excluded points on sugar plantations from all models with land use by shrub cover interactions to exclude irrelevant parameters and facilitate model fitting. I included date and time of surveys as covariates on detection in all models. I used flat normal priors with mean 0 and precision 0.01 for fixed effects and a uniform prior with mean 10 and precision 0.01 for random effects (Gelman & Hill 2007).

For each species and model, I ran four chains for 190,000 iterations, discarding the first 40,000 as burn-in, and thinned each chain by collecting every 50th observation. In order to keep Gelman-

Rubin r-hat statistics ≤1.1 and collect enough samples to produce a unimodal density plot for each parameter of interest (Gelman & Rubin 1992; Gelman & Hill 2007), I repeated some models for up to 4,500,000 iterations, up to 1,000,000 as burn-in samples, and collected as few as every 5,000th observation in order to achieve convergence (Appendix K). I ran all occupancy models in JAGS called through R version 3.3.2 using the jagsUI package (Plummer 2003;

Kellner 2016; R Core Team 2016).

Effects of shrub encroachment, land-use intensity, and interactions

I used a metaregression approach to determining whether there were consistent additive and interactive effects of shrub encroachment, land-use intensity, and their interactions on the occupancy of savanna birds across species. I chose this approach for several reasons. First, I evaluated a sample of species and those species’ responses could not be evaluated with equal precision, therefore filtering the sample by statistical significance would be inappropriate.

Instead, metaregression made it possible to identify sources of heterogeneity in effect sizes among species while accounting for differences in the precision among the estimated species’ responses, thus fully using the available information. Further, by using a random effects model I were able to make and evaluate inferences beyond those species for which occupancy modeling

72

was tractable (Cooper et al. 2009). Multi-species occupancy models have similar properties but make assumptions about the distribution of occupancy responses which may not be realistic

(Dorazio & Royle 2005). I therefore fit random-effects metaregression models testing for each effect, treating model coefficients as the response variable because occupancy model coefficients are equivalent to log odds ratios, which are a standard response variable used in meta-analysis of binary outcomes (Cooper et al. 2009). I determined the precision for each response variable by taking the square of each model coefficient’s standard deviation (Viechtbauer 2010).

Specifically, I tested whether interactions between shrub cover and each of the three relevant land-use categories (protected areas, community pastures, and homesteads) were present in the community. If I found no evidence of interactions, I proceeded to fit metaregressions for the additive effects of each land use and for shrub encroachment across all species. I ran all metaregression models using the metafor package in R version 3.3.1 (Viechtbauer 2010).

Finally, I evaluated the explanatory power of diet, predatory status, and mass as factors contributing to heterogeneity in the effects of shrub encroachment and land-use intensity on bird occupancy. I fit additive metaregression models within each land-use intensity category for diet, predatory status, and mass. I log-transformed masses before model fitting and included models with a quadratic term for mass to consider whether intermediate body size was favored by shrub encroachment, land-use intensification or both conditions. I treated diet, predatory status, or mass as important predictors of shrub encroachment effects on bird species occupancy if 95% confidence intervals for the appropriate parameters overlapped zero and not important otherwise, although I note as ‘weak evidence’ where a predictor’s 90% confidence interval did not overlap zero but the 95% interval did. Finally, I also report I2 for each metaregression, which is a

73

measure of a metaregression model’s power to explain effect-size heterogeneity that is analogous to R2 (Higgins et al. 2003; Viechtbauer 2010).

Results

I recorded 209 species, 48 of which I detected in ≥10% of points. The 48 species ranged from 7.5 g to 252 g (Appendix J). Apparent prevalence averaged 14, 8, 11, 3, and 10% among frugivorous, insectivorous, nectarivorous, omnivorous, and granivorous species, respectively.

Apparent prevalence of non-raptor predatory birds averaged 14% across all points (range: 3-

40%). There was also variation in apparent prevalence among bird diet groups across land uses, with all diets less prevalent in sugar plantation (Appendix L). Similarly, shrub cover differed among land uses, ranging from 0-100% in protected areas and community pastures (means

32±1.3% SE and 41±1.8% SE respectively) but 0-35% in homesteads and 0-24% in the sugar estate (means 7±0.4% SE and 2±0.3% SE respectively; Fig. 4-1; F3, 367= 38.04, P < 0.0001). The differences in mean shrub cover between land uses was significant except between homesteads and sugar plantation (Tukey’s HSD adjusted P = 0.76 for homestead-sugar plantation versus

0.00-0.03 for all other post-hoc comparisons). Occupancy modeling identified 34 species exhibiting positive occupancy responses to shrub encroachment, whereas 32, 16, 21, and 9 species exhibited positive occupancy responses to protected areas, community pastures, homesteads, and sugar plantation, respectively (Appendix L).

Metaregression revealed no evidence for land use by shrub cover interactions (minimum

P = 0.52), and there was weak evidence for a positive shrub cover effect on bird occupancy (훽̂ =

0.02±0.01 SE; P = 0.0580) whereas occupancy declined from the least to the most intensive land uses (Fig. 4-2; 훽̂ = 0.10±0.60 SE; P = 0.8601, 훽̂ = -1.12±0.47 SE; P = 0.0176, 훽̂ = -1.85±0.86

74

SE; P = 0.0318, and 훽̂ = -3.30±0.57 SE; P < 0.0001, for protected areas, community pastures, homesteads, and sugar plantation, respectively).

Both diet and body mass explained variation in species responses to shrub cover and land use. For example, frugivore occupancy was positively associated with shrub encroachment whilst granivore occupancy associations with shrub encroachment were negative (Fig. 4-3; ̂훽 =

0.08 ± 0.02, 푃 < 0.0001, and 훽̂ = −0.02 ± 0.01, 푃 = 0.0015, respectively). Predator occupancy was positively associated with shrub encroachment but negatively associated with community pastures and homesteads (Fig. 4-4; 훽̂ = 0.20 ± 0.06 SE; 푃 = 0.0014, 훽̂ = -

0.82±1.78 SE; P = 0.6455, 훽̂ = -5.14±2.18 SE; P = 0.0183, 훽̂ = -7.85±2.48 SE; P = 0.0016, and

훽̂ = -1.94±3.13 SE; P < 0.5358, for shrub encroachment, protected areas, community pastures, homesteads, and sugar plantation, respectively). Protected areas were positively associated with nectarivore occupancy whereas all other land-use effects on diet groups were negative and impacted either frugivores, insectivores or granivores (Fig. 4-5; 훽̂ = 5.12±2.47 SE; P = 0.0378 for nectarivores in protected areas, 훽̂ = -2.49±0.78 SE; P = 0.0014, and 훽̂ = -3.23±0.92 SE; P =

0.0005 for frugivores in protected areas and community pastures, respectively, 훽̂ = -1.11±0.42

SE; P = 0.0079 and 훽̂ = -2.67±1.02 SE; P < 0.0001 for insectivores in homesteads and sugar plantation respectively, and 훽̂ = -3.57±1.05 SE; P = 0.0007 for granivores in sugar plantation).

̂ Occupancy responses to shrub cover exhibited a quadratic relationship with bird mass ( 훽푚푎푠푠 =

̂ 0.10±0.02 SE; P <0.0001, 훽푚푎푠푠_푠푞 = −0.01 ± 0.00 SE, 푃 < 0.0001). Relationships between

̂ mass and land use were quadratic in protected areas and community pastures (Fig. 4-6; 훽푚푎푠푠 = -

̂ 5.09±1.45 SE; P = 0.0004, 훽푚푎푠푠_푠푞 = 0.58 ± 0.19 SE, 푃 = 0.0018 ) but negative and linear in sugar plantation (Fig. 4-6; 훽̂ = 0.10±0.60 SE; P = 0.8601).

75

Discussion

I found that the effects of shrub encroachment on bird communities can be consistent across land uses, suggesting that bird communities in encroached savanna may be determined by a common mechanism. This is encouraging because shrub encroachment and land-use intensification have contrasting effects that could therefore provide opportunities to promote biodiversity in mosaic landscapes. That two of the biggest drivers of biodiversity change in savannas would have complementary effects was unforeseen and fortuitous, indicating that land- use intensification effects might be amenable to mitigation efforts (Shaw et al. 2002). Finally, since I was able to uncover a diet signal in a vertebrate group where mixed diets are ubiquitous

(Şekercioğlu et al. 2004), I expect that shrub encroachment and land-use intensification effects on diets of other vertebrate taxa will be larger, and therefore generally quite useful in explaining species’ responses to global change.

Bird responses to shrub encroachment were consistent across land uses. This was contrary to my prediction and suggests that either the same processes are driving community assembly or different processes are having the same effects. However, this does not mean that we can understand shrub encroachment effects on animals without considering land use. Indeed, the effects of shrub encroachment and land-use intensification were frequently contrasting. Shrub encroachment had positive effects on bird occupancy, favoring frugivores, insectivores, and birds of intermediate body size whereas land-use intensification had negative effects on the occupancy of savanna bird species that cut across diets and was unfavorable to larger species.

My findings suggest that increasing shrub cover in intensive land uses may mitigate the effects of land-use intensification leading to increases in frugivores, predators, nectarivores and larger-bodied species, with concomitant decreases in granivores, and collectively, more functionally diverse bird communities (Nicholson et al. 2009; Dehling et al. 2016). Managing

76

shrub cover adjacent to intensive land uses may also be an effective mitigation measure for maintaining representation of species’ traits, but spatially explicit research will be needed to establish this.

Protection and land-use intensification can have surprising effects on common species that may be overlooked if conservation is focused exclusively on rare and charismatic taxa

(Naughton-Treves et al. 2005). For example, it is remarkable that nectarivores were the only group that increased in occupancy in protected areas, indicating that protection may have limited benefits for conserving common birds and their associated ecological functions. However, shrub cover in protected areas can be managed more easily than in other land uses, creating an opportunity to achieve biodiversity benefits through coordinated management. More intensive land uses may also contribute to bird traits represented in a landscape in ways that defy intuition.

For example, body size did not explain bird occupancy in homesteads. Rather, species of approximately the same size exhibited opposing occupancy relationships with homesteads, suggestive of strong species interactions in need of description and explanation. One striking example are red-faced and speckled mousebirds (Urocolius indicus and Colius striatus, both

~55g), which exhibit strong positive and negative occupancy relationships with homesteads, respectively. In all other land uses, however, bird occupancy had the expected negative relationship with body size (Fig. 4-6). Land-use intensification also had unexpected effects on insectivorous bird occupancy. The negative effects of homesteads and sugar plantation on insectivorous birds is particularly interesting, since insectivorous bird diversity can be reduced both by simplification of vegetation structure and by pesticides (Sekercioglu 2002; Kutt &

Martin 2010). Distinguishing which process is leading to lower insectivore occupancy will be important to determine: 1) if increasing woody cover is likely to be an effective mitigation

77

measure; and 2) if so, whether those measures should be implemented in protected areas or within more intensive land uses.

Establishing the mechanisms underpinning bird responses to shrub encroachment and land-use intensification in savannas will be challenging. I found that non-raptorial predatory bird occupancy increased with shrub encroachment but decreased in community pastures and homesteads. This raises two questions. First, might non-raptorial predatory birds be determining the structure of shrub-encroached savanna bird communities? Second, does the effect of predatory birds in shrub-encroached savanna, if any, differ among land uses? Admittedly, other mechanisms such as the availability of food and suitable nest sites are also likely to play important roles. However, synergies among these mechanisms and others may be occurring that will require time to untangle (Bender et al. 1984; Brown et al. 2001). Therefore, two simultaneous lines of research development should be prioritized. First, we need to formalize and test theory that explicitly links species occurrence to environmental gradients in ways that account for synergies. I attempted to do so in a simple way by incorporating species traits, but more development in this area is possible. Second, experimental work tied to careful measurement of vital rates is needed, and it is important that such work is tied to predicted synergies and partitions variation in vital rates among the processes that determine community composition: selection; dispersal; drift; and speciation (Vellend 2010).

78

Figure 4-1. Geographic setting, sampling scheme, and distribution of shrub cover in a study of shrub encroachment and land-use intensification effects on vertebrates. A) Swaziland is located in southern Africa (dark polygon); B) I sampled birds across 42 grids (colored circles) allocated among protected areas, community pastures, homesteads, and sugar estate across Lowveld savanna there from December 2014-March 2015. C) Shrub cover was highest and most variable in protected areas and pastures, but was lower in homesteads, and nearly absent from sugar plantation. The dotted horizontal line represents study-wide mean percent shrub cover whilst the solid horizontal lines indicate group-level medians, and boxes group-level inter-quartile ranges (25th and 75th). Bubbles represent individual values at each of the 371 points were shrub cover was measured and bird occurrence recorded.

79

Figure 4-2. Bird occupancy was negatively associated with land-use intensification, based on a meta-analysis of 48 Bayesian occupancy models. The dotted horizontal line is at zero where there is no effect on occupancy whilst the solid horizontal lines indicate group- level medians, and boxes group-level inter-quartile ranges (25th and 75th). Bubbles represent individual species’ occupancy responses to each land-use type in those models.

80

Figure 4-3. Bird occupancy had a weak overall association with shrub encroachment (훽̂ = 0.02±0.01 SE; P = 0.0580). However, frugivores were positively associated with shrub encroachment whilst granivores were negatively associated with shrub encroachment ( ̂훽 = 0.08 ± 0.02, 푃 < 0.0001, and ̂훽 = −0.02 ± 0.01, 푃 = 0.0015, respectively). The dotted horizontal line is at zero where there is no effect on occupancy whilst the solid horizontal lines indicate group-level medians, and boxes group-level inter-quartile ranges (25th and 75th). Bubbles represent individual species’ occupancy responses to each land-use type in those models.

81

Figure 4-4. Predatory bird occupancy was negatively associated with community pastures ( 훽̂ = −5.14 ± 2.18, 푃 = 0.0183) and homesteads ( ̂훽 = −7.85 ± 2.48, 푃 = 0.016), but positively associated with shrub encroachment ( ̂훽 = 0.20 ± 0.06, 푃 = 0.0014). The dotted horizontal line is at zero where there is no effect on occupancy whilst the solid horizontal lines indicate group-level medians, and boxes group-level inter-quartile ranges (25th and 75th). Bubbles represent individual species’ occupancy responses to each land-use type in those models.

82

Figure 4-5. Bird responses to land-use intensification differed among diet groups. Protected areas favored nectarivore occupancy (훽̂ = 5.12 ± 2.47, 푃 = 0.0378) and disfavored frugivore occupancy (훽̂ = −2.49 ± 0.78, 푃 = 0.0014) whilst invertebrate occupancy was negatively associated with the two most intensive land uses of homesteads and sugar plantation (훽̂ = −1.11 ± 0.42, 푃 = 0.0079 and (훽̂ = −2.67 ± 0.53, 푃 < 0.0001, respectively). Granivorous birds were also less prevalent in sugar plantation (훽̂ = −3.57 ± 1.05, 푃 = 0.0007). The dotted horizontal line is at zero where there is no effect on occupancy whilst the solid horizontal lines indicate group-level medians, and boxes group-level inter-quartile ranges (25th and 75th). Bubbles represent individual species’ occupancy responses to each land-use type in those models.

83

Figure 4-6. Body size was negatively associated with bird occupancy in protected areas ( ̂훽 = −0.61 ± 0.18, 푃 = 0.0006) and community pastures ( ̂훽 = −0.48 ± 0.21, 푃 = 0.0240) but the negative effect was most pronounced in sugar plantation ( ̂훽 = −1.04 ± 0.52, 푃 = 0.0440). Body size did not explain species responses to homesteads ( ̂훽 = −0.60 ± 0.48, 푃 = 0.2043). Bubbles in each plot represent individual species occupancy responses to a given land-use type; their relative sizes reflect that species’ weight in a metaregression (i.e., 1/variance). Solid lines are best fit lines; dashed lines are upper and lower 95% confidence intervals.

84

CHAPTER 5 PREDATORY BIRD CUES AFFECT PREY DETECTABILITY, NOT OCCUPANCY DYNAMICS, ACROSS A SHRUB COVER GRADIENT IN AN AFRICAN SAVANNA

Synopsis

Predators can affect prey occurrence, abundance, and behavior across gradients of cover.

Prey responses to predators should be contingent on available cover, which provides refugia from predators. Savannas are experiencing widespread increases in woody cover (shrub encroachment) that can be accompanied by increased occurrence of non-raptorial predatory birds. Testing how prey responses to increased predation risk vary across shrub cover gradients is needed to understand how shrub encroachment and related biotic interactions drive community dynamics. I conducted a predatory bird cue addition experiment across a shrub cover gradient in an African savanna. I addressed three questions: 1) do cues of increased predation risk change the occupancy dynamics, detectability, or breeding effort of savanna bird species? 2) Are species’ responses to predator cues moderated by shrub cover? And 3) do migratory species, which lack up-to-date knowledge of patch quality, respond more strongly to cues of predation risk? I surveyed bird occurrence and searched for active nests across a shrub cover gradient under treatment and control conditions, testing for effects on two resident and two migratory focal species using dynamic occupancy models and Poisson regression. I found that predator cues had effects on detectability for two species and no effect on occupancy dynamics or nesting for any species. A resident species, the chinspot batis (Batis molitor) exhibited a treatment by shrub cover interaction, becoming more detectable with increasing shrub cover when predator cues were added, implying that batises may have been more vigilant where there was adequate escape cover. A migrant species, the violet-backed starling (Cinnyricinclus leucogaster), was also more detectable where predator cues were added. These results show that while predator

85

activity effects on prey behavior can be cover-dependent, researchers have yet to fully understand which prey traits explain prey responses to predator cues.

Background

Predation risk can be a key driver of animal occurrence and behavior (Connell 1988;

Gurevitch et al. 2000). Prey responses to increased predation risk include changes in settlement patterns, as well as actions that reduce conspicuousness (e.g. less activity, fewer vocalizations, and lower breeding effort; Lima & Dill 1990; Lima & Bednekoff 1999; Fontaine and Martin

2000). However, species can also become more conspicuous in response to increased predation risk when they engage in behaviors such as perching in prominent locations, mobbing, and alarm calling (Lima 2009). Further, species responses to increased predation risk can be cover- dependent, because cover can affect the outcome of predation attempts (Lima and Dill 1990;

Lima 1992).

Altering predation risk or cues of predation risk can affect prey settlement patterns

(Fontaine & Martin 2006; Hua et al. 2013). However, predation risk can vary within a season and little is known about how adding cues of increased predation risk during the breeding season affects the occurrence, abundance and behavior of prey (Fontaine and Martin 2006; Cox et al.

2012). Prey should only quit a breeding patch with high predation risk if other available patches are of higher quality (Charnov 1976). Since prey generally settle in breeding sites in order of patch quality, a patch that remains vacant during the breeding season should generally be inferior to an occupied patch (Fretwell 1972). Consequently, prey should primarily change their behaviors in response to increased predation risk whilst remaining in an occupied patch.

However, prey may also vacate and colonize a new patch if it is of higher quality (Betts et al.

2008b). Further, factors such as migratory status should predict use of cues of predation risk to make settlement and breeding decisions because migrants may lack current information on

86

predation risk that should be available to residents (Mönkkönen et al. 1999; Seppänen et al.

2007; Betts et al. 2008a; Hua et al. 2014). However, few studies test whether prey response to predator cues differ with contrasting information about predation risk between migrant and resident species (Fontaine and Martin 2006).

Understanding the role of cover in moderating prey responses to increased predation risk is important because many biomes are experiencing shifts in cover regimes as woody shrubs replace grasses, increasing the quantity and temporal stability of cover (“shrub encroachment”;

Eldridge et al. 2011). Shrub encroachment is particularly evident in African savannas, and is associated with substantial changes in vertebrate community composition, including an increase in non-raptorial predatory birds (Sirami and Monadjem 2012; Chapters 2-3). Given the change in predatory bird occurrence, testing how prey respond to increased predation risk across shrub cover gradients is necessary to understand how shrub encroachment is driving prey community dynamics. Birds provide an ideal taxon to test for cover-dependent responses to predation risk in shrub-encroached savanna because they use acoustic cues to glean information about their environment and are readily counted using surveys (Mönkkönen et al. 1999).

I predicted that 1) cues of predatory birds would affect the behavior of prey, as indicated by changes in detectability, reduced breeding effort, and altered occupancy dynamics (i.e. increased probability of patch extinction and decreased probability of patch colonization); 2) migratory species would respond more strongly to cues of increased predation risk than residents; and 3) prey responses to predatory bird cues would be moderated by shrub cover.

Materials and Methods

Study Area

I studied the effects of predatory bird cues on avian community structure on the Mbuluzi and Mlawula game reserves in Swaziland’s Lowveld savanna (26.15 S, 31.78 E). The region is a

87

subtropical savanna (annual precipitation ranges from 550-725 mm; Monadjem 2000), and breeding activity peaks for most birds during a summer rainfall period (November-January;

Sirami and Monadjem 2012). Shrub encroachment is extensive in the region, and several native woody plants are increasing and contributing to shrub encroachment including sicklebush,

Dichrostachys cinerea, buffalo thorn, Ziziphus mucronata, and prickly acacia, Vachellia nilotica,

(Monadjem 2000; Loffler & Loffler 2005; Wigley et al. 2009). Shrub encroachment in African savanna leads to large changes in the composition of bird communities, including increases in the occurrence of several non-raptor predatory birds (Sirami & Monadjem 2012; Chapter 4). The

Mbuluzi and Mlawula reserves were embedded in a land-use mosaic mainly comprised of subsistence farming, communal rangelands, and commercial agriculture, subject to a general pattern of land-use intensification largely driven by land conversion to sugar cane plantation

(Sirami & Monadjem 2012; Bailey et al. 2016).

Focal species

I selected four common species with complementary migratory statuses to study. I selected species that varied in their response to shrub cover to determine if their responses were caused by predatory birds and to compare residents to migrants because they might differ in their use of predator cues. I chose two migratory species—the violet-backed starling (Cinnyricinclus leucogaster) and the spotted flycatcher (Muscicapa striata)–which increase with shrub encroachment (Sirami & Monadjem 2012; Chapter 4). Likewise, I selected two resident species– chinspot batis (Batis molitor) and white-browed scrub robin (Cercotrichus leucophrys)—which have negative and positive response to shrub encroachment, respectively (Chapter 4, but see

Sirami and Monadjem 2012).

88

Experimental Design

I used a regression experimental design with procedural controls (Williams 1970). I sampled bird occurrence on 24 plots of 50m radius from 21 October - 15 December 2015 when migrants were settling (settlement begins in September and October for violet-backed starlings and spotted flycatchers, respectively; Hockey et al. 2005) and resident breeders were selecting breeding locations, and searched for nests from 20 Dec 2015 - 12 Jan 2016, immediately after predator cue addition treatments ceased (Sirami and Monadjem 2012). I identified a set of 24 plots from 44 candidate sites that I visited 4 times from 4 December 2014 - 15 March 2015

(during the pre-treatment year), paired for similar shrub cover and proximity (< 7 km apart) that did not experience clearing or burning during the study. I randomly assigned one of each pair to a predatory bird auditory cue addition treatment or a procedural control where I placed dummy boxes that were manipulated in the same manner as treatment gear (Hua et al. 2013). Each plot was > 250m from any neighboring plot and each plot was > 100 m from visible edges, i.e. abrupt changes in shrub, tree, and grass cover. I did not attempt to remove predators or protect prey species from predation events.

I broadcast a mix of vocalizations from the 6 species that comprised > 95% of the non- raptor predatory birds detected in Swaziland’s Lowveld savanna (see Chapters 3-4) on all treatment plots: Burchell’s coucal, Centropus burchelli; fork-tailed drongo, Dicrurus adsimilis; grey-headed , Malaconotus blanchoti; , Laniarius ferrugineus; southern yellow-billed hornbill, Tockus leucomelas; and kingfisher, Halcyon senegalensis. Collectively, these species prey on other birds at all life stages, and most species have been documented killing both nestlings and adults (Hockey et al. 2005). I broadcast vocalizations of these species as a group because I was interested in the collective effects of

89

predatory birds on common savanna bird species (cf. Zanette et al. 2011) and whether those effects differed between migrants and residents or were moderated by shrub cover.

I rotated one playback station among three locations within each plot twice weekly to reduce habituation to playback conditions. Each playback station consisted of a functional station or dummy gear. Each functional station consisted of a SanDisk Clip Sport (Western Digital

Technologies, Inc., Irvine, California) MP3 player and a RACPower 3150 mAh lithium-ion battery inside an EcoExtreme model #GDI-AQCSE101 3 amp waterproof housing and speaker

(Grace Digital, Poway, California), powered by three 2100 mAh AA batteries (Xtech, ) suspended 0.5-2.3 m above ground level in the shade to prevent the electronics from overheating

(Fig. 5-1). I used plain black dry cases (S3 cases model T1000, Fort Collins, Colorado) as dummy stations, which I placed and moved in the same manner as treatment stations (Hua et al.

2013). I visited every station for < 2 mins once every 24 hours in order to maintain a continuous power supply for the MP3 players and speakers. These procedural controls account for the human activity required in plots for treatment manipulations and for the physical objects in treatment plots, but do not account for noise effects independent of predators per se. I focused on procedural controls that did not include sound because control vocalizations can have unknown effects on prey (Kroodsma 1989; Francis et al. 2009).

I prepared 3 playback files for use in predator cue treatments by drawing 2 - 3 recordings per playback file from a set of available exemplars (Kroodsma et al. 2001), totaling ≤4 individuals of any species. I listened to everything with the focal species in the foreground and rated “A” or “B” quality by xeno-canto.org users and selected all that I deemed useable as-is or with minimal editing, e.g., truncating a non-target species from the end of a recording in program

Audacity (www.audacityteam.org). I also added recordings of 3 southern boubou groups and 1

90

grey-headed bushshrike group that I recorded in Swaziland; I did not broadcast these in the same place they were recorded. The vocalizations were mostly songs and calls but I excluded alarm calls. I determined how many vocalizations to include in each file and day based on previous field experience and acoustic monitoring data (unpublished). I provided 564 total vocalizations per day in each playback file, comprised of 48, 216, 48, 180, 8, and 64 vocalizations of

Burchell’s coucal, fork-tailed drongo, grey-headed bushshrike, southern boubou, southern yellow-billed hornbill, and woodland kingfisher, respectively, representing their relative frequency of detection in shrub-encroached savanna in Swaziland (unpublished data). I played vocalizations each day from dawn to twilight because that is when non-raptor predator birds were active.

Data Collection

I collected data on plot-level bird community composition using fixed-radius 10 min stationary point count surveys (30 mins before sunrise to 5 hrs after, 21 October-15 December). I surveyed each plot 12 times— twice-daily during 6 periods where I assumed population closure at each site during a single day but not between periods (5-17 days; mean: 10.1 days). I alternated the sequence in which plots were visited in order to control for date and time-of-day effects on bird activity. I recorded each bird seen or heard and identified to species. I did not survey when winds were >20 km/hr or in the rain.

I sampled vegetation structure twice at each plot, first from 24 January – 23 February

2015 during the pre-treatment year, then from 11 – 12 January 2016 during the field experiment.

I described patch vegetation structure in each plot by measuring grass, shrub, and tree cover. I defined trees as woody plants > 3 m and shrubs as woody plants ≤ 3m after Sirami and

Monadjem (2012). I measured grass, shrub, and tree cover using the line intercept method with 3 replicates per point and a 50 m tape. I also recorded grass, shrub and tree height at 10 m intervals

91

along each line (Tansley & Chipp 1926). I placed the three lines at each point at 0, 120, and 240 degrees orientation using an iPhone 4 compass (Apple Inc., Cupertino, California, USA), calculated mean percent cover contacting a line for grass, shrub, and tree cover and treated them as estimates of percent cover for each cover type.

I located active nests using behavioral cues and random searches evenly distributed across time and throughout the day among the sampling stations (Martin & Geupel 1993). I did this by allocating 6 hours of nest searching per 50 m plot in 1, 2, or 3 hr stints, randomly starting each day with either a treatment plot or a control plot, and alternating between treatment and control plots on subsequent stints. I recorded the duration, date, and time of all nest-searching stints, as well as the species associated with each active nest I discovered and the stage at which I discovered each nest, i.e., nest building, egg laying, incubation, or chick rearing.

Analytical Methods

Dynamic occupancy modeling

Treatments could influence bird communities by: 1) causing birds that had already settled to exit the plots with the addition of predator cues (i.e., local extinction); 2) impeding the settlement of birds into plots (i.e., local colonization); or 3) simply altering behavior of birds without changes in distribution. For instance, detectability of birds could change because predators can cause species to either use more cryptic behaviors (e.g., a reduction in singing) or more conspicuous behaviors (e.g., an increase in mobbing or alarm calls). Each of these issues could vary by treatment and by shrub cover, since shrub cover may alter the perceived risk of predation by offering potential refugia or escape cover (Lima 1992).

To distinguish among these biological responses, I used dynamic occupancy models, which allow for estimation of colonization-extinction dynamics while accounting for variation in detectability (MacKenzie et al. 2003). I wanted to distinguish 1) whether treatment by shrub

92

cover interactions were present; 2) if there was a main effect of treatment in the absence of interactions; and 3) whether any treatment effects or treatment by shrub cover interactions that occurred were best explained by changes in detectability, colonization probability, or extinction probability. I estimated initial occupancy (휓), colonization (훾) and extinction (휖) rates, and detectability (p) for each species from the survey data using dynamic occupancy models in the R package unmarked (Fiske & Chandler 2011). In this context, occupancy (휓) refers to the probability a species occurs in a patch during the first sampling period, whereas colonization (훾) and extinction (휖) refer to changes in the probability of occupancy between sampling periods, and detectability (p) is the probability of recording a species during a survey provided it is present (Fiske & Chandler 2011). In this design, there were 6 primary survey periods where I estimated colonization-extinction dynamics, and 2 secondary periods for each primary period, in which I assumed the population was closed. I only included treatment effects or treatment by shrub cover interactions on one term in a given model, treating P < 0.05 as evidence effects were present and using AIC to establish whether more than one model had some support, i.e. Δ AIC ≤

4 (Table 5.1; Akaike 1973; Burnham and Anderson 2002). In some cases, models would not converge when parameter values were near the boundaries possible under the model, including when I used shrub cover as a predictor of initial occupancy rates. In these cases, I reduced model complexity by removing the shrub cover term from the model of initial occupancy probability. I did not test for treatment effects on initial occupancy per se because I had not yet implemented treatments. Finally, I fit null models for each species (i.e. constant 휓, 훾, 휖, and p) and plotted detection-nondetection histories for each species and plot (colored according to treatment status) in order to explore whether the presence of colonization-extinction dynamics was plausible because I could not be certain I was studying open populations and the magnitude of the

93

parameter estimates was necessary to interpret the biological relevance and potential biases of any treatment effects (Betts et al. 2008b; McKann et al. 2013). This is important in part because dynamic occupancy models can exhibit positively-biased estimates of 훾 and 휖 under a variety of common sampling conditions (McKann et al. 2013).

Poisson regression of nest counts

I tested for differences in the total number of nests of all breeding non-predatory species between treatment and control plots and for treatment by shrub cover interactions using Poisson regression, interpreting P values < 0.05 as evidence of an effect. I did not test for treatment effects on any individual species because sample sizes were small (≤ 4 nests per species).

Results

I observed variation in detection-nondetection among sampling periods for all species, indicating that colonization-extinction dynamics may have been present (Appendix M). Further, null models indicated that colonization-extinction dynamics were present for some species after accounting for imperfect detection (estimated 훾 [표푟𝑖𝑔𝑖푛푎푙 푠푐푎푙푒]: 0.217±0.127 SE, 0.036±0.047

SE, 0.182±0.159 SE, and 0.208±0.089 SE for chinspot batis, white-browed scrub robin, spotted flycatcher, and violet-backed starling, respectively; estimated 휖 [표푟𝑖𝑔𝑖푛푎푙 푠푐푎푙푒] 0.092±0.127

SE, 0.519±0.243 SE, 0.220±0.284SE, and 0.347±0.172SE for chinspot batis, white-browed scrub robin, spotted flycatcher, and violet-backed starling, respectively). I found that predator cues had effects on detectability for two species and no effect on colonization or extinction probabilities for any species (all P > 0.05; Table 5-1). The chinspot batis exhibited a treatment by shrub cover interaction, wherein the positive effect of predator cues increased with shrub cover (shrub, treatment, and interaction 훽s were -0.95±0.40 SE, 0.89±0.41 SE, and 1.37±0.45 SE, respectively, interaction P = 0.002; Fig. 5-1). Violet-backed starlings were also more detectable where predator cues had been added (treatment 훽 = 2.22±0.69 SE; P = 0.001). There was marginal

94

evidence that treatment reduced the detectability of white-browed scrub robins (훽 = −1.65 ±

0.92 SE, P = 0.07). However, the null model for the white-browed scrub robin had substantial support (Δ AIC = 0.04; Table 5.2).

I found 27 nests of 15 non-predator species, 10 of which were in treatment plots (means:

0.83±0.32 SE and 1.42±0.40 SE in treatment and control plots respectively). I found 2 chinspot batis nests, one in a treatment plot and the other in a control plot, no white-browed scrub robin nests, and one violet-backed starling nest, which was in a treatment plot. I found no evidence that the total number of nests of all species was explained by the predator cue treatment (P = 0.18, df

= 23, 22) or a treatment by shrub cover interaction (P = 0.98, df = 23, 20), although the model appeared to be overdispersed (휒2 test, P = 0.034). A quasipoisson model yielded nearly the same results (P = 0.53 and 0.99 for treatment and shrub  treatment interactions, respectively; overdispersion parameter: 1.42)

Discussion

My results demonstrate that cover can moderate the response of birds to cues of predation risk even when the cues do not affect occupancy dynamics. This implies that vertebrates are capable of adjusting their anti-predator behavior to suit local cover conditions. Further, my results are noteworthy in that two species became more detectable when predator cues were added. It is also interesting that the smaller, more maneuverable chinspot batis changed its conspicuousness depending on cover, likely because additional cover reduced the risk of antipredator behaviors such as mobbing and taking prominent perches (Hockey et al. 2005). The larger, less agile violet-backed starling, on the other hand, may not accrue any risk reduction by being proximal to shrub cover.

95

Prey can rely on habitat selection to avoid predators, or prey can make behavioral changes that influence how conspicuous they are to predators as they use habitat (Lima & Dill

1990; Lima 2009). It seems the species which responded to predatory bird cues in my study changed their behavior rather than selected different habitats, which should have affected colonization and extinction rates instead of detectability. There are several important reasons I may have seen this result, all of which could have reduced the predation risk perceived by prey.

For example, I used auditory cues and did not manipulate visual cues. Likewise, while the treatment species are known predators, the risk they present to prey is unknown and could be rather modest (Hockey et al 2005). Combining cues of multiple species also may have generated non-additive effects (Sih et al. 1998). This can occur, for example, when predators interfere with one another, reducing risk, which might reasonably occur between fork-tailed drongos and other species because drongos can exhibit sentinel behaviors (Radford et al. 2011). Consequently, the treatment may be an underestimate of the biological effects of non-raptor predatory birds associated with shrub encroachment on the prey species I studied. Finally, the increased detectability I found and attributed to behavioral changes could also occur if the treatment increased abundance (Guillera-Arroita et al. 2012). I cannot dismiss this possibility but deem it unlikely because I studied territorial species on plots that could not accommodate more than one entire territory of these species (Monadjem, unpublished data).

Migratory birds should make more use of heterospecific cues than residents, but several factors can mitigate such behaviors (Mönkkönen et al. 1999). For example, many migratory birds are long-lived and individuals often show fidelity to locations where they have bred or foraged successfully in the past (Switzer 1993). The value of this prior experience may trump immediate cues of predation risk in cases where up-to-date information is less valuable than historic

96

information (Mönkkönen et al. 1999). Likewise, auditory cues of predators may not present the most salient source of information about habitat quality. Migrants often rely on cues of heterospecific prey species with similar ecological requirements, which could be the most parsimonious cue of habitat quality in many cases (Seppänen et al. 2007).

My findings suggest that savanna birds may employ a variety of strategies to manage predation risk that differ among species. It would be valuable to summarize the distribution of those strategies in animal communities and better explain variation among species. A substantive body of literature indicates that species traits should explain which anti-predator strategies are employed among species (McGill et al. 2006; Cresswell 2008; Lima 2009). For example, species possess morphometric traits that can predict the value of cover in determining the outcome of a predation attempt (Lima 1992). Testing whether such traits can explain cover-dependent responses to cues of increased predation risk would improve our understanding of predator-prey interactions.

97

Table 5-1. Models considered to test for non-raptor predatory bird cue treatment effects and treatment by shrub cover interactions on the detectability (p) of four common bird species in the Lowveld of Swaziland. The same models were fit to all species, and the same set of models were fit to test for treatment and treatment by shrub cover interactions on site colonization (ϒ) and extinction (ϵ) probabilities during the course of the field experiment. Model Description Comment

ψ(.)ϒ(.)ϵ(.)p(.) Null model

ψ(shrub)ϒ(.)ϵ(.)p(shrub X treat) Initial occupancy ~shrub

cover, detection ~ shrub X

treat interaction

ψ(shrub)ϒ(.)ϵ(.)p(treat) Detection ~ treat Treat models fit when no

interactions present

ψ(.)ϒ(.)ϵ(.)p(shrub X treat) No shrub effect on initial Shrub removed from ψ

occupancy models if needed to

converge

ψ(.)ϒ(.)ϵ(.)p(treat)

The symbols ψ, ϒ, ϵ, and p refer to the probabilities of initial site occupancy, colonization, extinction, and detection, respectively.

98

Table 5-2. Parameter estimates (SE) from top-ranked dynamic occupancy models testing for effects of experimentally-added non-raptor predatory bird cues and cue by shrub cover interactions on the probability of colonization, extinction, and detection for four common species found in shrub-encroached Lowveld savanna of Swaziland.

Species Model Shrub ϒ(shrub) ϒ(treat) ϒ(shrub  treat) p(shrub) p(treat) p(shrub  treat)

Chinspot batis ψ(shrub)ϒ(.)ϵ(.)p(shrub  treat) -0.240 (0.927) NA NA NA -0.953 (0.401) 0.889 (0.409) 1.369 (0.449)

Spotted flycatcher ψ(.)ϒ(shrub  treat)ϵ(.)p(.) NA 14.7 (38.8) 37.5 (75.2) -40.6 (80.6) NA NA NA

Violet-backed starling ψ(shrub)ϒ(.)ϵ(.)p(shrub  treat) -30.4 (78.9) NA NA NA 0.524 (0.734) 2.218(0.747) -0.755 (0.802)

White-browed scrub robin ψ(shrub)ϒ(.)ϵ(.)p(treat) 0.837 (1.18) NA NA NA NA -1.65 (0.916) NA

99

Table 5-3. Rankings of dynamic occupancy models testing for effects of experimental non-raptor predatory bird cues and cue by shrub cover interactions on the probability of colonization, extinction, and detection for four common species found in the Lowveld savanna of Swaziland. Model K AIC 횫퐀퐈퐂 w chinspot batis

ψ(shrub)ϒ(.)ϵ(.)p(shrub  treat) 8 229.73 0.00 0.53

ψ(shrub)ϒ(treat)ϵ(.)p(.) 6 231.17 1.44 0.26

ψ(.)ϒ(.)ϵ(.)p(.) 4 231.79 2.06 0.19

ψ(shrub)ϒ(.)ϵ(treat)p(.) 6 235.62 5.89 0.03 spotted flycatcher

ψ(.)ϒ(shrub  treat)ϵ(.)p(.) 7 125.85 0.00 0.64

ψ(shrub)ϒ(.)ϵ(.)p(shrub  treat) 8 128.60 2.74 0.16

ψ(.)ϒ(treat)ϵ(.)p(.) 5 130.33 4.48 0.07

ψ(.)ϒ(.)ϵ(.)p(.) 4 130.33 4.48 0.07

ψ(.)ϒ(.)ϵ(.)p(shrub  treat) 7 131.86 6.01 0.03

ψ(.)ϒ(.)ϵ(.)p(treat) 5 132.16 6.30 0.03 violet-backed starling

ψ(shrub)ϒ(.)ϵ(.)p(shrub  treat) 8 154.44 0.00 0.68

ψ(.)ϒ(.)ϵ(.)p(treat) 5 156.85 2.41 0.20

ψ(.)ϒ(.)ϵ(.)p(shrub  treat) 7 159.21 4.77 0.06

ψ(.)ϒ(.)ϵ(shrub  treat)p(.) 7 160.66 6.22 0.03

ψ(.)ϒ(treat)ϵ(.)p(.) 5 161.97 7.53 0.02

ψ(.)ϒ(.)ϵ(treat)p(.) 5 163.04 8.60 0.01

ψ(.)ϒ(.)ϵ(.)p(.) 4 167.16 12.72 0.00

100

Table 5-3. Continued Model K AIC 횫퐀퐈퐂 w white-browed scrub robin

ψ(shrub)ϒ(.)ϵ(.)p(treat) 6 86.40 0.00 0.32

ψ(.)ϒ(.)ϵ(.)p(.) 4 86.44 0.04 0.32

ψ(shrub)ϒ(.)ϵ(shrub  treat)p(.) 8 88.29 1.89 0.13

ψ(shrub)ϒ(.)ϵ(.)p(shrub  treat) 8 88.83 2.43 0.10

ψ(shrub)ϒ(.)ϵ(treat)p(.) 6 90.08 3.68 0.05

ψ(shrub)ϒ(treat)ϵ(.)p(.) 6 90.10 3.70 0.05

ψ(shrub)ϒ(shrub  treat)ϵ(.)p(.) 8 90.66 4.25 0.04

The symbols ψ, ϒ, ϵ, and p refer to the probabilities of initial site occupancy, colonization, extinction, and detection, respectively, whilst K is the number of parameters in the model, AIC is Akaike’s Information Criterion, 훥AIC is Akaike’s criterion scaled relative to the lowest AIC value, which is set to zero, and w is the proportion of model weight accorded to each respective model.

101

Figure 5-1. A broadcast unit deployed in Swaziland, 3 November 2015. Courtesy of author. Units provided audio cues of non-raptor predatory birds and contained: a SanDisk Clip MP3 player; a RACPower 3150 mAh lithium-ion battery; a 3 amp waterproof housing and speaker (EcoExtreme); and 3 X 2100 mAh AA batteries.

102

Figure 5-2. Detection probability for chinspot batises increased with shrub cover on plots where auditory cues on non-raptor predatory birds were added (dashed bold line), whereas batis detectability approached zero as shrub cover increased in control plots (solid bold line). Dashed and solid gray lines are 95% confidence intervals.

103

CHAPTER 6 CONCLUSIONS

I had the opportunity to do an empirical study of bird occurrence across gradients of shrub cover and land-use intensity in the Lowveld savanna of Swaziland. There, I chose 1) to focus on a taxon with well-described traits that varied among species, 2) to collect data in a scale-explicit sampling design, 3) to test for synergies (i.e. statistical interactions) between the effects of land-use intensity and shrub cover on bird species occurrence, and 4) to assess the relative roles of land-use, shrub cover, and geographic distance as correlates of community dissimilarity in savanna birds. I discovered that traits can explain species responses to shrub encroachment and land-use intensification (i.e., diet, predatory status, and body size), that there were no important synergies between land-use and shrub cover in determining species occurrence, and 3) surprisingly, shrub cover may have been the most important and reliable correlate of species dissimilarity in the bird community. These studies demonstrate that the trait- based community ecology paradigm and recent literature on synergies in global change biology can be useful framing tools for empirical studies.

A concern for understanding the drivers of community change provided the impetus for testing whether non-raptor predatory birds, which increased with shrub encroachment in my observational study, were responsible for some of the changes in the occurrence of other bird species I observed. I tested this by conducting an experiment wherein I added auditory cues of predatory birds across a shrub gradient. This idea, although surprisingly almost without precedent, has yielded ambiguous results that have much of the uncertain flavor of classic community ecology. It may, however, contribute to identifying some candidate species that may be key players in the bird community, which can be targeted for more intensive study. Further, the results suggest that shrubs per se may be the main driver of community assembly in African

104

savannas, mediated by either habitat selection or demographic processes— a matter that can be resolved with focused study.

From a rather different perspective, the data I collected can continue to yield relevant information beyond what I have elected to include in my dissertation. For example it would be fairly straightforward to assess whether my results are robust to increasing spatial scale, and to determine which of several patterns best describes the metacommunity I sampled from, which would more fully employ the insights offered by Vellend and others (Logue et al. 2011). This would be quite valuable as hypothesis-generating work and improve our understanding of

Lowveld birds, although I think researchers will also need to adopt suitable objective means of delineating community boundaries for the work to be informative for vertebrates.

Finally, research generally contributes to the generation of many more ideas for future research. Some of the ideas that naturally flow from my dissertation research may be addressed with data I have in hand, though others will require additional experimentation. Of these, the spatial scale of predator effects on prey is likely to be fairly important and remains to be fully explored, though data from the field experiment I conducted could likely be used to do so. This contrasts fairly strongly with a less tractable but rather critical question: do multiple predators have emergent effects on prey (sensu Sih et al. 1998)? By choosing to focus on the effect of avian predators sensu lato, I sacrificed the opportunity to isolate the effects of specific species because the number of treatments needed was prohibitive. Yet, the ubiquity and reticulate nature of predation among non-raptorial birds may be particularly conducive to the occurrence of emergent multi-predator effects that have generally been neglected in studies of bird communities.

105

APPENDIXA SHRUB ENCROACHMENT ON EARTH’S TERRESTRIAL BIOMES

Table A-1. Shrub encroachment is occurring in at least six of the world’s fourteen terrestrial biomes, as established by the following peer-reviewed articles. Biome classifications follow Olson et al. (2001). Articles shown contain reviews or meta-analyses, or longitudinal or cross-sectional studies of shrub cover dynamics, unless stated otherwise. Biome Articles establishing shrub encroachment Grass-dominated? Major driver(s) forests/ -- No Deserts and xeric shrublands Hastings & Turner 1965 Frequently Overgrazing Buffington & Herbel 1965 Loss of browsers van Auken 2000 Flooded grasslands and savannas Mayle et al. 2007 Frequently Fire exclusion Junk et al. 2012 Overgrazing Sharp & Whittaker 2003 -- No Mediterranean forests, woodlands, and shrublands Maestre et al. 2009 Managed woodlands Pastoral abandonment Montane grasslands and shrublands Molinillo et al. 1997 Frequently Pastoral abandonment Dullinger et al. 2003 Sanz-Elorza et al. 2003 Temperate broadleaf and mixed forests Brown & Archer 1999 savanna Temperate coniferous forests Gilliam & Platt 1999 savanna Fire exclusion Temperate grasslands, savannas, and shrublands Naito & Cairns 2011 Frequently Fire exclusion Overgrazing Tropical and subtropical coniferous forests Goldammer & Peñafiel 1990 Bahamian, Luzon, & Fire exclusion Parson 1955 Miskito pine savanna Tropical and subtropical dry broadleaf forests Parr et al. 2011 Frequently Ratnam et al. 2016 Tropical and subtropical grasslands, savannas, and shrublands Wigley & Bond 2010 Yes Fire, herbivory, rainfall Van Auken 2000 Tropical and subtropical moist broadleaf forests -- No Freezing Tundra Myers-Smith et al. 2011 Yes Naito & Cairns 2011

106

APPENDIX B DATA SOURCES USED IN A GLOBAL META-ANALYSIS OF SHRUB ENCROACHMENT EFFECTS ON VERTEBRATE DIVERSITY

Alford, AL, Hellgren EC, Limb R, Engle, DM. 2012. Experimental tree removal in tallgrass : variable responses of and fauna along a woody cover gradient. Ecological Applications 22: 947-958.

Arnold TW, Higgins KF. 1986. Effects of shrub coverages on birds of North Dakota mixed-grass . Canadian Field-Naturalist 100:10-14.

Bestelmeyer BT, Kalil NI. Peters DPC. 2007. Does shrub invasion indirectly limit grass establishment via seedling herbivory? A test at grassland-shrubland . Journal of Vegetation Science 18:363-370.

Blaum N, Rossmanith E, Schwager M., Jeltsch F. 2007. Responses of mammalian carnivores to land use in arid savanna rangelands. Basic and 8:552-564.

Chapman RN et al. 2004. Tree invasion constrains the influence of herbaceous structure in grassland bird habitats. Ecoscience 11:55-63.

Cosentino BJ et al. 2013. Response of lizard community structure to desert grassland restoration mediated by a keystone . Biodiversity and Conservation 22:921-935.

Dean WRJ et al. A. 2002. Avian assemblages in native Acacia and alien Prosopis drainage line woodland in the Kalahari, South Africa. Journal of Arid Environments 51:1-19.

Desmond M. 2000. Effects of grazing practices and fossorial rodents on a winter avian community in Chihuahua, Mexico. Biological Conservation 116:235-242.

Doerr VA et al. 2009. Managing invasive native scrublands for improved biodiversity outcomes in agricultural landscapes. Report to the Central West Catchment Management Authority, Dubbo, New South Wales. 78 pp.

Dorado-Rodrigues, et al. 2015. Effects of shrub encroachment on the anuran community in periodically flooded grasslands of the largest Neotropical . Austral Ecology 40:547- 557.

Dullinger S, Dirnböck T, Grabherr G. 2003. Patterns of shrub invasion into high mountain grasslands of the northern calcareous , . , Antarctic, and Alpine Research 35:434-441.

Fonderflick J et al. 2010. Avifauna trends following changes in a Mediterranean upland pastoral system. Agriculture Ecosystems & Environment 137:337-347.

107

Gilliam, FS, Platt WJ. 1999. Effects of long-term fire exclusion on tree species composition and stand structure in an old-growth Pinus palustris (longleaf pine) forest. 140:15- 26.

Gottschalk TK, Ekschmitt K, Bairlein F. 2007. Relationships between vegetation and bird community composition in grasslands of the Serengeti. African Journal of Ecology 45:557- 565.

Henden JA et al. 2013. How Spatial Variation in Areal Extent and Configuration of Labile Vegetation States Affect the Riparian Bird Community in Arctic Tundra. PloS ONE 8: e63312.

Hull KL. 2002. Range management and grassland bird diversity in the Cypress Hills. Thesis, University of Calgary, Calgary, .

Kaphengst T, Ward D. 2008. Effects of habitat structure and shrub encroachment on bird species diversity in arid savanna in Northern Cape province, South Africa. Ostrich 79:133-140.

Karuaera NAG. 2011. Assessing the effects of on species abundance, composition, and diversity of small mammals at the Neudamm Agricultural Farm, Khomas region, . Thesis, University of Namibia and Humboldt-Universitat zu Berlin, Namibia.

Lloyd J et al. 1998. The effects of mesquite invasion on a southeastern Arizona grassland bird community. Wilson Bulletin 110:403-408.

Long AM, Jensen WE, Matlack RS. 2014. Influence of prescribed burning on bird abundance and species assemblage in a semiarid Great Plains grassland. Western North American Naturalist 74:396-404.

Mayle FE, Langstroth RP, Fisher, RA, Meir P. 2007. Long-term forest–savannah dynamics in the Bolivian Amazon: implications for conservation. Philosophical Transactions of the Royal Society B: Biological Sciences 362:291-307.

Molinillo M, Lasanta T, García-Ruiz JM. 1997. Managing mountainous degraded landscapes after farmland abandonment in the central Spanish Pyrenees, Environmental Management 21:587– 598.

Monasmith TJ et al. 2010. Short-Term Fire Effects on Small Mammal Populations and Vegetation of the Northern . International Journal of Ecology 2010:1-9.

Myers-Smith IH et al. 2011. Shrub expansion in tundra ecosystems: dynamics, impacts and research priorities. Environmental Research Letters 6:045509.

Pelegrin N, Bucher EH. 2010. Long-term effects of a on a lizard assemblage in the Arid Chaco forest. Journal of Arid Environments 74:368-372.

108

Pelegrin N, Bucher EH. 2012. Effects of habitat degradation on the lizard assemblage in the Arid Chaco, central . Journal of Arid Environments 79:13-19.

Pidgeon AM et al. 2001. Response of avian communities to historic habitat change in the northern Chihuahuan Desert. 15:1772-1788.

Rosenstock SS, Van Riper C. 2001. Breeding bird responses to juniper woodland expansion. Journal of Range Management 54:226-232.

Santana J et al. 2011. Long-term understory recovery after mechanical fuel reduction in Mediterranean cork oak forests. and Management 261:447-459.

Sanz-Elorza M, ED Dana, A González, E Sobrino. 2003 Changes in the high-mountain vegetation of the central Iberian Peninsula as a probable sign of global warming. Annals of 93:273–280.

Seymour C. 2006. The influence of size and density of the Camelthorn (Acacia erioloba) on its keystone role in the Xeric Kalahari. Dissertation, University of Cape Town, Cape Town, South Africa.

Seymour CL, Dean WRJ. 2010. The influence of changes in habitat structure on the species composition of bird assemblages in the southern Kalahari. Austral Ecology 35:581-592.

Seymour CL et al. 2015. On Bird Functional Diversity: Species Richness and Functional Differentiation Show Contrasting Responses to Rainfall and Vegetation Structure in an Arid Landscape. Ecosystems 18:971-984.

Sharp BR, Whittaker RJ. 2003. The irreversible cattle‐driven transformation of a seasonally flooded Australian savanna. Journal of 30:783-802.

Sirami C et al. 2009. The impact of shrub encroachment on savanna bird diversity from local to regional scale. Diversity and Distributions 15:948-957.

Smit IPJ, Prins HHT. 2015. Predicting the Effects of Woody Encroachment on Mammal Communities, Grazing Biomass and Fire Frequency in African Savannas. PloS ONE, 10: e0137857

Söderström B et al. 2001. Plants, and birds in semi-natural pastures in relation to local habitat and landscape factors. Biodiversity and Conservation 10:1839-1863.

Soto-Shoender JR. 2016. Response of mid- and large-sized mammals to woody encroachment in a southern African savanna. Dissertation, University of Florida, Gainesville, Florida.

109

Wasiolka B. 2008. The Impact of overgrazing on diversity and population dynamics of Pedioplania l. lineoocellata in the southern Kalahari. Dissertation,University of Potsdam, Potsdam.

Wasiolka B, Blaum N. 2011. Comparing biodiversity between protected savanna and adjacent non- protected farmland in the southern Kalahari. Journal of Arid Environments 75:836-841.

110

APPENDIX C DATABASES CONSULTED AND SEARCH TERMS USED IN META-ANALYSIS

Table C-1. Databases consulted on 3 March 2016 and search terms used to identify studies of shrub encroachment effects on the species richness, Shannon diversity, and abundance of vertebrate community in grass-dominated biomes worldwide. Database Search string Web of Science ("shrub encroachment" OR "bush thickening" OR "woody encroachment" OR "woody thickening") AND ("bird diversity" OR "avian diversity") ("shrub encroachment" OR "bush thickening" OR "woody encroachment" OR "woody thickening") AND birds ("shrub encroachment" OR "bush thickening" OR "woody encroachment" OR "woody thickening") ("mammal diversity" OR "rodent diversity" OR "carnivore diversity" OR "ungulate diversity") ("shrub encroachment" OR "bush thickening" OR "woody encroachment" OR "woody thickening") ("reptile diversity" OR "lizard diversity" OR "turtle diversity" OR " diversity") ("shrub encroachment" OR "bush thickening" OR "woody encroachment" OR "woody thickening") AND "vertebrate diversity" ProQuest ("shrub encroachment" OR "bush thickening" OR "woody encroachment" OR "woody thickening") ("bird diversity" OR "avian diversity") ("shrub encroachment" OR "bush thickening" OR "woody encroachment" OR "woody thickening") ("mammal diversity" OR "rodent diversity" OR "carnivore diversity" OR "ungulate diversity") ("shrub encroachment" OR "bush thickening" OR "woody encroachment" OR "woody thickening") ("reptile diversity" OR "lizard diversity" OR "turtle diversity" OR "amphibian diversity") ("shrub encroachment" OR "bush thickening" OR "woody encroachment" OR "woody thickening") "vertebrate diversity" Google Scholar "shrub encroachment" birds ("shrub encroachment" OR "bush thickening" OR "woody encroachment" OR "woody thickening") ("bird diversity" OR "avian diversity") ("shrub encroachment" OR "bush thickening" OR "woody encroachment" OR "woody thickening") ("mammal diversity" OR "rodent diversity" OR "carnivore diversity" OR "ungulate diversity") ("shrub encroachment" OR "bush thickening" OR "woody encroachment" OR "woody thickening") ("reptile diversity" OR "lizard diversity" OR "turtle diversity" OR "amphibian diversity") ("shrub encroachment" OR "bush thickening" OR "woody encroachment" OR "woody thickening") "vertebrate diversity"

111

APPENDIX D ATTRIBUTES OF STUDIES USED IN THE META-ANALYSIS

Table D-1. Studies used in a global meta-analysis of shrub encroachment effects on terrestrial vertebrate community structure with corresponding study locations, net primary productivity in g C m−2 yr−1 (NPP), normalized difference vegetation index (NDVI), mean annual precipitation in mm (MAP), biome and ecoregion (after Olson et al. 2001), vertebrate group, and response variable(s) reported, i.e., species richness, Shannon diversity (Shannon), and total abundance. Author & year Continent NPP NDVI MAP Biome Ecoregion Group Richness Shannon Abundance Experiments and quasi-experiments

Alford et al. 2012 N. America 1623 145 892 Temperate grasslands savannas and shrublands Central and Southern mixed grasslands Mammals X X

Cosentino et al. 2013 N. America 615 88 244 Deserts and xeric shrublands Chihuahuan desert Herpetofauna X X

Kaphengst and Ward 2008 Africa 1010 137 369 Deserts and xeric shrublands Kalahari xeric savanna Birds X X Long et al. 2014 N. America 886 111 478 Temperate grasslands savannas and shrublands Western short grasslands Birds X X X Monasmith et al. 2010 N. America 558 72 205 Deserts and xeric shrublands Chihuahuan desert Mammals X X X Stanton et al. unpub. Africa 1241 229 587 Tropical and subtropical grasslands savannas and shrublands Zambezian and woodlands Birds X X X

Observational studies

Alberico 1978 N. America 615 88 244 Deserts and xeric shrublands Chihuahuan desert Mammals X X

Arnold and Higgins 1986 N. America 1053 124 455 Temperate grasslands savannas and shrublands Northern short grasslands Birds X X

Bestelmeyer et al. 2007 N. America 392 62 238 Deserts and xeric shrublands Chihuahuan desert Mammals X

Blaum et al. 2007a Africa 543 79 200 Deserts and xeric shrublands Kalahari xeric savanna Mammals X X Blaum et al. 2007b Africa 543 79 200 Deserts and xeric shrublands Kalahari xeric savanna Mammals X X X Blaum et al. 2007c Africa 543 79 200 Deserts and xeric shrublands Kalahari xeric savanna Mammals X X X Blaum et al. 2009 Africa 543 79 200 Deserts and xeric shrublands Kalahari xeric savanna Mammals X X X

Boeing et al. 2013 N. America 615 88 244 Deserts and xeric shrublands Chihuahuan desert Herpetofauna X

Ceballos et al. 2010 N. America 674 87 406 Deserts and xeric shrublands Chihuahuan desert Mammals X X

Ceballos et al. 2010 N. America 674 87 406 Deserts and xeric shrublands Chihuahuan desert Birds X

Ceballos et al. 2010 N. America 674 87 406 Deserts and xeric shrublands Chihuahuan desert Herpetofauna X X Chapman et al. 2004 N. America 1294 139 622 Temperate grasslands savannas and shrublands Central and Southern mixed grasslands Birds X X X Dean et al. 2002 Africa 788 126 337 Deserts and xeric shrublands Kalahari xeric savanna Birds X X X

Desmond 2004 N. America 935 102 420 Deserts and xeric shrublands Chihuahuan desert Birds X X

Doerr et al. 2009 Australia 1039 136 475 Temperate grasslands savannas and shrublands Southeast Australia temperate savanna Birds X X

Dorado-Rodrigues 2015 S. America 1485 222 1352 Flooded grasslands and savannas Pantanal Herpetofauna X X

Fonderflick et al. 2010 Europe 980 161 812 Mediterranean forests woodlands and scrub Northeastern and Southern Mediterranean forests Birds X

Gottschalk et al. 2007 Africa 1084 179 727 Tropical and subtropical grasslands savannas and shrublands Southern Acacia-Commiphora bushlands and thickets Birds X X

Henden et al. 2013 Europe 275 13 541 Tundra Tundra Birds X

Hull 2002 N. America 650 93 320 Temperate grasslands savannas and shrublands Northern mixed grasslands Birds X

Kaphengst and Ward 2008 Africa 1010 137 369 Deserts and xeric shrublands Kalahari xeric savanna Birds X X

Karuaera 2011 Africa 806 140 354 Deserts and xeric shrublands Kalahari xeric savanna Mammals X X

Laiolo et al. 2004 Europe 624 51 1248 Temperate forest Alps conifer and mixed forests Birds X

Leynaud and Bucher 2005 S. America 957 186 648 Tropical and subtropical grasslands savannas and shrublands Dry Chaco Herpetofauna X

Lloyd 1998 N. America 962 98 296 Deserts and xeric shrublands Chihuahuan desert Birds X X

Pelegrin and Bucher 2010 S. America 419 76 534 Tropical and subtropical grasslands savannas and shrublands Dry Chaco Herpetofauna X X Pelegrin and Bucher 2012 S. America 1110 181 557 Tropical and subtropical grasslands savannas and shrublands Dry Chaco Herpetofauna X X X

Pidgeon et al. 2001 N. America 558 72 205 Deserts and xeric shrublands Chihuahuan desert Birds X

112

Table D-1. Continued. Author & year Continent NPP NDVI MAP Biome Ecoregion Group Richness Shannon Abundance

Rosenstock and Van Riper 2001 N. America 241 63 213 Deserts and xeric shrublands Colorado Plateau shrublands Birds X Santana et al. 2011 Europe 1286 167 603 Mediterranean forests woodlands and scrub Southwest Iberian Mediterranean sclerophyllous and mixed forests Birds X X X Seymour 2006 Africa 963 131 398 Deserts and xeric shrublands Kalahari xeric savanna Birds X X X

Seymour and Dean 2010 Africa 963 131 398 Deserts and xeric shrublands Kalahari xeric savanna Birds X X

Seymour et al. 2015 Africa 648 109 354 Deserts and xeric shrublands Kalahari xeric savanna Birds X

Sirami and Monadjem 2012 Africa 1047 211 587 Tropical and subtropical grasslands savannas and shrublands Zambezian and Mopane woodlands Birds X

Sirami et al. 2009 Africa 977 142 369 Deserts and xeric shrublands Kalahari xeric savanna Birds X

Smit and Prins 2015 Africa 748 183 489 Tropical and subtropical grasslands savannas and shrublands Zambezian and Mopane woodlands Mammals X X

Söderström et al. 2001 Europe 629 114 507 Temperate broadleaf and mixed forests Birds X

Soto 2016 Africa 1241 229 587 Tropical and subtropical grasslands savannas and shrublands Zambezian and Mopane woodlands Mammals X X Stanton et al. unpub. Africa 1241 229 587 Tropical and subtropical grasslands savannas and shrublands Zambezian and Mopane woodlands Birds X X X

Tassicker et al. 2006 Australia 1067 157 505 Tropical and subtropical grasslands savannas and shrublands Mitchell grass downs Birds X

Wasiolka 2008 Africa 543 79 200 Deserts and xeric shrublands Kalahari xeric savanna Herpetofauna X X

Wasiolka and Blaum 2011 Africa 543 79 200 Deserts and xeric shrublands Kalahari xeric savanna Herpetofauna X X

113

APPENDIX E META-REGRESSION RESULTS

Table E-1. Results from a set of meta-regressions fit to identify sources of heterogeneity in shrub encroachment effects on the species richness, Shannon diversity, and abundance of vertebrate communities in the world’s grass-dominated biomes. Model Beta SE P value Lower CI Upper CI I2 Richness, observational studies No covariates 0.06 0.13 0.62 -0.19 0.32 0.96 Mean shrub cover 0.00 0.01 0.80 -0.01 0.01 0.96 Mean annual precipitation, mm 0.01 0.15 0.95 -0.28 0.3 0.95 Normalized difference vegetation index -0.03 0.13 0.80 -0.30 0.23 0.96 Net primary productivity, g C m−2 yr−1 0.01 0.14 0.92 -0.26 0.29 0.96 Africa -0.28 0.14 0.04 -0.55 -0.01 0.93 Australia 1.48 0.38 0.00 0.73 2.23 0.93 Europe 0.57 0.29 0.05 0.00 1.14 0.93 North America 0.81 0.25 0.00 0.32 1.30 0.93 South America -0.39 0.40 0.33 -1.18 0.40 0.93 Desert and xeric shrubland -0.08 0.21 0.68 -0.49 0.32 0.96 Temperate grassland 0.49 0.44 0.27 -0.38 1.36 0.96 Tropical and subtropical grassland 0.26 0.35 0.69 -0.45 0.94 0.96 Herpetofauna -0.08 0.11 0.46 -0.29 0.13 0.95 Mammals -0.15 0.11 0.17 -0.35 0.06 0.95 Birds 0.11 0.13 0.40 -0.14 0.36 0.95 Grazed -0.07 0.16 0.64 -0.39 0.24 0.96 Space for time substitution 0.09 0.43 0.83 -0.75 0.94 0.96 Richness, experiments No covariates 0.23 0.19 0.24 -0.15 0.60 0.73 Mean shrub cover 0.02 0.02 0.35 -0.02 0.06 0.78 Mean annual precipitation, mm 0.35 0.19 0.06 -0.01 0.72 0.68 Normalized difference vegetation index 0.15 0.27 0.57 -0.38 0.69 0.80 Net primary productivity, g C m−2 yr−1 0.08 0.10 0.40 -0.11 0.28 0.00

114

Table E-1. Continued Model Beta SE P value Lower CI Upper CI I2 Shannon diversity, observational studies No covariates -0.11 0.15 0.47 -0.41 0.19 0.91 Mean shrub cover -0.02 0.01 0.05 -0.05 0.00 0.92 Mean annual precipitation, mm 0.23 0.15 0.12 -0.06 0.53 0.92 Normalized difference vegetation index 0.30 0.16 0.07 -0.02 0.61 0.92 Net primary productivity, g C m−2 yr−1 0.44 0.14 0.00 0.16 0.72 0.91 Africa -0.25 0.21 0.23 -0.67 0.16 0.86 Australia 0.84 0.63 0.18 -0.40 2.08 0.86 Europe 0.46 0.48 0.33 -0.47 1.39 0.86 North America 0.47 0.45 0.29 -0.41 1.36 0.86 South America -0.19 0.42 0.66 -1.00 0.63 0.86 Desert and xeric shrubland -0.24 0.21 0.26 -0.66 0.18 0.91 Temperate grassland 0.81 0.51 0.12 -0.20 1.81 0.91 Tropical and subtropical grassland -0.05 0.38 0.90 -0.80 0.70 0.91 Herpetofauna -0.75 0.31 0.02 -1.36 -0.13 0.84 Mammals -0.90 0.26 0.00 -1.41 -0.38 0.84 Birds 0.30 0.16 0.06 -0.01 0.61 0.84 Grazed -0.03 0.25 0.74 -0.57 -0.40 0.91 Shannon diversity, experiments -0.19 0.12 0.12 -0.42 0.05 0.94 No covariates -0.13 0.08 0.08 -0.28 0.02 0.00 Mean shrub cover 0.01 0.01 0.38 -0.01 0.04 0.00 Mean annual precipitation, mm 0.09 0.12 0.42 -0.14 0.32 0.00 Normalized difference vegetation index 0.10 0.19 0.58 -0.26 0.47 0.00 Net primary productivity, g C m−2 yr−1 0.08 0.10 0.40 -0.11 0.28 0.00

115

Table E-1. Continued. Model Beta SE P value Lower CI Upper CI I2 Abundance, observational studies No covariates -0.19 0.12 0.12 -0.42 0.05 0.94 Mean shrub cover -0.01 0.01 0.28 -0.02 0.01 0.94 Mean annual precipitation, mm -0.07 0.13 0.56 -0.33 0.17 0.94 Normalized difference vegetation index -0.07 0.12 0.58 -0.30 0.17 0.94 Net primary productivity, g C m−2 yr−1 -0.03 0.13 0.80 -0.28 0.21 0.94 Africa -0.36 0.15 0.02 -0.67 -0.06 0.93 Europe 0.69 0.51 0.18 -0.31 1.68 0.93 North America 0.53 0.25 0.03 0.05 1.01 0.93 South America -0.18 0.33 0.60 -0.82 0.47 0.93 Desert and xeric shrubland -0.10 0.17 0.57 -0.42 0.23 0.95 Temperate grassland -0.30 0.50 0.54 -1.27 0.67 0.95 Tropical and subtropical grassland -0.23 0.29 0.42 -0.80 0.33 0.95 Herpetofauna -0.51 0.23 0.02 -0.96 -0.07 0.93 Mammals -0.55 0.22 0.01 -0.99 -0.11 0.93 Birds 0.13 0.17 0.45 -0.20 0.45 0.93 Grazed -0.12 0.17 0.49 -0.44 0.21 0.94 Space for time substitution -0.76 0.41 0.06 -1.56 0.04 0.94 Abundance, experiments No covariates 0.07 0.32 0.82 -0.55 0.70 0.90 Mean shrub cover -0.00 0.04 0.91 -2.51 2.98 0.89 Mean annual precipitation, mm -1.12 0.30 0.00 -1.70 -0.54 0.22 Normalized difference vegetation index -0.19 0.38 0.62 -0.93 0.56 0.90 Net primary productivity, g C m−2 yr−1 -0.39 0.46 0.39 -0.52 0.75 0.87

116

APPENDIX F SHRUB ENCROACHMENT AFFECTS VERTEBRATE COMMUNITY STRUCTURE AMONG BIOMES

Figure F-1. Shrub encroachment effects (r) on vertebrate community structure among biomes. Vertebrate (a) species richness, (b) Shannon diversity, and (c) total abundance (n = 30, 19, and 35, respectively) responses to encroachment were not different from zero among biomes (P > 0.05).

117

APPENDIX G REPRESENTATIVE FUNNEL PLOTS FROM THE META-ANALYSIS

Figure G-1. Representative funnel plots indicating heterogeneous but symmetrical distributions of effect sizes of shrub encroachment on vertebrate community species richness, Shannon diversity and abundance. The plots correspond to meta-analyses of observational studies on A) species richness across all taxa but accounting for continent, B) community abundance across all taxa and locations, and C) Shannon diversity across all taxa and continents but accounting for net primary productivity.

118

APPENDIX H DISTRIBUTION OF OBSERVED AND MODELED OCCUPANCY

Figure H-1. Observed and modeled proportion of area occupied for a suite of 209 savanna bird species detected in Lowveld of Swaziland during sampling (December 2014 – March 2015) were similarly distributed, but the occupancy model predicted fewer rare species and more ubiquitous species. Modeled and observed species richness frequencies are indicated by the grey and brick-colored bars, respectively. The vertical blue lines are medians.

119

APPENDIX I MODELED AND OBSERVED SPECIES RICHNESS

Figure I-1. Bird species richness among local communities in the Lowveld savanna of Swaziland sampled from December 2014 – March 2015 as estimated from a hierarchical Bayesian occupancy model with data augmentation was approximately quadruple the number of species observed, on average. Modeled and observed species richness frequencies are indicated by the grey and brick-colored bars, respectively.

120

APPENDIX J TRAITS OF COMMON SPECIES

Species Latin name diet predator mass, g Red-eyed Dove Streptopelia semitorquata seeds 252 Cape Turtle Dove Streptopelia capicola seeds 153 Emerald-spotted Wood Dove Turtur chalcospilos seeds 64 Speckled Mousebird Colius striatus fruit 55 Red-faced Mousebird Urocolius indicus fruit 56 Woodland Kingfisher Halcyon senegalensis invertebrates Yes 63 Acacia Pied Barbet Tricholaema leucomelas fruit 31 Golden-tailed Woodpecker Campethera abingoni invertebrates 70 Cardinal Woodpecker Dendropicos fuscescens invertebrates 30 Barn Swallow Hirundo rustica invertebrates 20 Fork-tailed Drongo Dicrurus adsimilis invertebrates Yes 44 Black-headed Oriole Oriolus larvatus invertebrates 65 Southern Black Tit Parus niger invertebrates 21 Dark-capped Bulbul Pycnonotus tricolor fruit 39 Terrestrial Brownbul Phyllastrephus terrestris invertebrates 36 Sombre Greenbul Andropadus importunus fruit 31 Eastern Nicator Nicator gularis invertebrates 47 Kurrichane Turdus libonyanus invertebrates 63 White-browed Scrub Robin Cercotrichas leucophrys invertebrates 20 Warbler Phylloscopus trochilus invertebrates 9 Yellow-breasted Apalis Apalis flavida invertebrates 8 Long-billed Crombec Sylvietta rufescens invertebrates 11 Green-backed Camaroptera Camaroptera brachyura invertebrates 11 Rattling Cisticola chiniana invertebrates 16 Red-faced Cisticola Cisticola erythrops invertebrates 15 Tawny-flanked Prinia subflava invertebrates 10

121

Table J-1. Continued Species Latin name diet predator mass, g Spotted Flycatcher Muscicapa striata invertebrates 15 Chinspot Batis Batis molitor invertebrates 12 African Paradise Flycatcher Terpsiphone viridis invertebrates 14 Southern Fiscal collaris invertebrates Yes 40 Red-backed Lanius collurio invertebrates Yes 29 Southern Boubou Laniarius ferrugineus invertebrates Yes 59 Black-backed Puffback Dryoscopus cubla invertebrates 26 Brown-crowned Tchagra Tchagra australis invertebrates 33 Black-crowned Tchagra Tchagra senegalus invertebrates 53 Gorgeous Bushshrike Telophorus viridis invertebrates 37 Violet-backed Starling Cinnyricinclus leucogaster fruit 45 Cape Glossy Starling Lamprotornis nitens fruit 90 White-bellied Sunbird Cinnyris talatala nectar 8 Scarlet-chested Sunbird Chalcomitra senegalensis nectar 14 Cape White-eye Zosterops virens invertebrates 11 House Sparrow Passer domesticus seeds 26 Southern Grey-headed Sparrow Passer diffusus seeds 24 Spectacled Weaver Ploceus ocularis invertebrates 29 Southern Masked Weaver Ploceus velatus invertebrates 34 Blue Waxbill Uraeginthus angolensis seeds 10 Yellow-fronted Canary Crithagra mozambicus seeds 13 Golden-breasted Bunting Emberiza flaviventris seeds 20 Table J-1. Diet, predatory status, and mean mass (from Hockey et al. 2005) of 48 species detected at ≥ 10% of survey points placed across gradients of shrub encroachment and land-use intensification in the Lowveld savanna of Swaziland, December 2014- March 2015.

122

APPENDIX K MODEL TUNING PARAMETERS FOR EACH SPECIES species Latin name iterations burn-in thin Red-eyed Dove Streptopelia semitorquata 1E+05 4E+04 50 Cape Turtle Dove Streptopelia capicola 1E+05 4E+04 50 Emerald-spotted Wood Dove Turtur chalcospilos 1E+05 4E+04 50 Speckled Mousebird Colius striatus 1E+05 4E+04 50 Red-faced Mousebird Urocolius indicus 2E+05 4E+04 50 Woodland Kingfisher Halcyon senegalensis 2E+05 4E+04 50 Acacia Pied Barbet Tricholaema leucomelas 1E+05 4E+04 50 Golden-tailed Woodpecker Campethera abingoni 1E+05 4E+04 50 Cardinal Woodpecker Dendropicos fuscescens 1E+05 4E+04 50 Barn Swallow Hirundo rustica 1E+05 4E+04 50 Fork-tailed Drongo Dicrurus adsimilis 2E+06 4E+04 50 Black-headed Oriole Oriolus larvatus 1E+05 4E+04 50 Southern Black Tit Parus niger 1E+05 4E+04 50 Dark-capped Bulbul Pycnonotus tricolor 1E+05 4E+04 50 Terrestrial Brownbul Phyllastrephus terrestris 1E+05 4E+04 50 Sombre Greenbul Andropadus importunus 1E+05 4E+04 50 Eastern Nicator Nicator gularis 5E+05 4E+04 50 Kurrichane Thrush Turdus libonyanus 1E+05 4E+04 50 White-browed Scrub Robin Cercotrichas leucophrys 5E+05 4E+04 50 Willow Warbler Phylloscopus trochilus 1E+05 4E+04 50 Yellow-breasted Apalis Apalis flavida 1E+05 4E+04 50 Long-billed Crombec Sylvietta rufescens 3E+05 4E+04 50 Green-backed Camaroptera Camaroptera brachyura 1E+05 4E+04 50 Rattling Cisticola Cisticola chiniana 1E+05 4E+04 50 Red-faced Cisticola Cisticola erythrops 1E+05 4E+04 50 Tawny-flanked Prinia Prinia subflava 1E+05 4E+04 50

123

Table K-1. Continued. species Latin name iterations burn-in thin Spotted Flycatcher Muscicapa striata 1E+05 4E+04 50 Chinspot Batis Batis molitor 1E+05 4E+04 50 African Paradise Flycatcher Terpsiphone viridis 1E+05 4E+04 50 Southern Fiscal Lanius collaris 1E+05 4E+04 50 Red-backed Shrike Lanius collurio 1E+05 4E+04 50 Southern Boubou Laniarius ferrugineus 5E+06 2E+06 5000 Black-backed Puffback Dryoscopus cubla 1E+05 4E+04 50 Brown-crowned Tchagra Tchagra australis 1E+05 4E+04 50 Black-crowned Tchagra Tchagra senegalus 5E+05 4E+04 50 Gorgeous Bushshrike Telophorus viridis 1E+05 4E+04 50 Violet-backed Starling Cinnyricinclus leucogaster 1E+06 4E+04 50 Cape Glossy Starling Lamprotornis nitens 1E+05 4E+04 50 White-bellied Sunbird Cinnyris talatala 3E+05 4E+04 50 Scarlet-chested Sunbird Chalcomitra senegalensis 1E+05 4E+04 50 Cape White-eye Zosterops virens 1E+05 4E+04 50 House Sparrow Passer domesticus 1E+05 4E+04 50 Southern Grey-headed Sparrow Passer diffusus 1E+05 4E+04 50 Spectacled Weaver Ploceus ocularis 1E+05 4E+04 50 Southern Masked Weaver Ploceus velatus 1E+05 4E+04 50 Blue Waxbill Uraeginthus angolensis 1E+05 4E+04 50 Yellow-fronted Canary Crithagra mozambicus 1E+05 4E+04 50 Golden-breasted Bunting Emberiza flaviventris 5E+06 2E+06 5000 Table K-1. Tuning parameters used when fitting Bayesian occupancy models to 48 species of birds detected at ≥ 10% of survey points placed across gradients of shrub encroachment and land-use intensification in the Lowveld savanna of Swaziland, December 2014- March 2015.

124

APPENDIX L DISTRIBUTION OF NAÏVE OCCUPANCY BY DIET AMONG LAND USES

Figure L-1. Naïve bird occupancy by diet across a gradient of land-use intensity in the Lowveld savanna of Swaziland, December 2014-March 2015. Dark horizontal lines are group medians, boxes encompass the interquartile ranges, and isolated points are outliers.

125

APPENDIX M PATTERNS OF DETECTION-NONDETECTION FOR FOUR SPECIES SUBJECTED TO AUDITORY CUES OF SEVERAL NON-RAPTOR PREDATORY BIRDS

Figure M-1. Patterns of detection-nondetection for four species subjected to auditory cues of several non-raptor predatory birds. Treatment plots are represented by red lines and control plots are shown in black. The focal species were (a) chinspot batis (Batis molitor), (b) violet-backed starling (Cinnyricinclus leucogaster), (c) spotted flycatcher (Muscicapa striata), and (d) white-browed scrub robin (Cercotrichus leucophrys).

126

REFERENCE LIST

Akaike H. 1973. Maximum likelihood identification of Gaussian autoregressive moving average models. Biometrika 60:255-265.

Anderson MJ, et al. 2011. Navigating the multiple meanings of beta diversity: a roadmap for the practicing ecologist. Ecology Letters 14:19-28.

Arponen A, Moilanen A, Ferrier S. 2008. A successful community‐level strategy for conservation prioritization. Journal of Applied Ecology 45:1436-1445.

Assine ML, Soares PC. 2004. Quaternary of the Pantanal, west-central . Quaternary International 114:23-34.

Bailey KM, McCleery RA, Binford MW, Zweig C. 2016. Land-cover change within and around protected areas in a . Journal of Land Use Science 11:154-176.

Balehegn M. 2015. Unintended Consequences: The Ecological Repercussions of Land Grabbing in Sub-Saharan Africa. Environment 57:4-21.

Baselga A,. 2010. Partitioning the turnover and nestedness components of beta diversity. Global Ecology and Biogeography 19:134-143.

Baselga A, Gomez-Rodriguez C, Lobo JM. 2012. Historical Legacies in World Amphibian Diversity Revealed by the Turnover and Nestedness Components of Beta Diversity. Plos ONE 7:e32341.

Baselga A, Orme CDL. 2012. betapart: an R package for the study of beta diversity. Methods in Ecology and Evolution 3:808-812.

Beerling DJ, Osborne CP. 2006. The origin of the savanna biome. Global Change Biology 12:2023-2031.

Bender EA, Case TJ, Gilpin ME. 1984. Perturbation experiments in community ecology - theory and practice. Ecology 65:1-13.

Bennett AF, Hinsley SA, Bellamy PE, Swetnam RD, Mac Nally R. 2004. Do regional gradients in land-use influence richness, composition and turnover of bird assemblages in small woods? Biological Conservation 119:191-206.

Benton TG, Vickery JA, Wilson JD. 2003. Farmland biodiversity: is habitat heterogeneity the key? Trends in Ecology & Evolution 18:182-188.

Betts MG, Hadley AS, Rodenhouse N, Nocera JJ. 2008a. Social information trumps vegetation structure in breeding-site selection by a migrant songbird. Proceedings of the Royal Society B-Biological Sciences 275:2257-2263.

127

Betts MG, Rodenhouse NL, Sillet TS, Doran PJ, Holmes RT. 2008b, Dynamic occupancy models reveal within-breeding season movement up a habitat quality gradient by a migratory songbird. Ecography 31:592-600. Biernacki P, Waldorf D. 1981. Snowball sampling - problems and techniques of chain referral sampling. Sociological Methods & Research 10:141-163.

Boeing WJ, Griffis-Kyle KL, Jungels JM. 2014. Anuran Habitat Associations in the Northern Chihuahuan Desert, USA. Journal of Herpetology 48:103-110.

Bohlen PJ, Scheu S, Hale CM, McLean MA, Migge S, Groffman PM, Parkinson D. 2004. Non- Native Invasive Earthworms as Agents of Change in Northern Temperate Forests. Frontiers in Ecology and the Environment 2:427-435.

Boitani L, Cowling RM, Dublin HT, Mace GM, Parrish J, Possingham HP, Pressey RL, Rondinini C, Wilson KA. 2008. Change the IUCN protected area categories to reflect biodiversity outcomes. PloS Biology 6:436-438.

Bond WJ, Midgley GF, Woodward FI. 2003. The importance of low atmospheric CO2 and fire in promoting the spread of grasslands and savannas. Global Change Biology 9:973-982.

Bond WJ, Woodward FI, Midgley GF. 2005. The global distribution of ecosystems in a world without fire. New Phytologist 165:525-537.

Buffington LC, Herbel CH. 1965. Vegetational changes on a semidesert grassland range from 1858 to 1963. Ecological monographs 35:139-164.

Burnham KP, Anderson DR. 2003. Model selection and multimodel inference: a practical information-theoretic approach. Springer Science & Business Media.

Briske DD, Fuhlendorf SD, Smeins FE. 2003. Vegetation dynamics on rangelands: a critique of the current paradigms. Journal of Applied Ecology 40:601-614.

Broms KM, Johnson DS, Altwegg R, Conquest LL. 2014. Spatial occupancy models applied to atlas data show Southern Ground Hornbills strongly depend on protected areas. Ecological Applications 24:363-374.

Brook BW, Sodhi NS, Bradshaw CJA. 2008. Synergies among extinction drivers under global change. Trends in Ecology & Evolution 23:453-460.

Brown DE, Reichenbacher F, Franson SE. 1998. A classification of North American biotic communties. University of Utah Press, Salt City, Utah.

Brown JH, Whitham TG, Ernest SKM, Gehring CA. 2001. Complex species interactions and the dynamics of ecological systems: Long-term experiments. Science 293:643-650.

Brown JR, Archer S. 1999. Shrub invasion of grassland: recruitment is continuous and not regulated by herbaceous biomass or density. Ecology 80:2385-2396.

128

Buij R, Croes BM, Gort G, Komdeur J. 2013. The role of breeding range, diet, mobility and body size in associations of raptor communities and land-use in a West African savanna. Biological Conservation 166:231-246.

Burkett VR, et al. 2005. Nonlinear dynamics in ecosystem response to climatic change: Case studies and policy implications. Ecological Complexity 2:357-394.

Cabral AC, De Miguel JM, Rescia AJ, Schmitz MF, Pineda FD. 2003. Shrub encroachment in Argentinean savannas. Journal of Vegetation Science 14:145-152.

Cardinale BJ, et al. 2012. Biodiversity loss and its impact on humanity. Nature 486:59-67.

Ceballos G, Davidson A, List R, Pacheco J, Manzano-Fischer P, Santos-Barrera G, Cruzado J. 2010. Rapid Decline of a Grassland System and Its Ecological and Conservation Implications. PloS ONE 5: e8562.

Chapin FS, et al. 2000. Consequences of changing biodiversity. Nature 405:234-242.

Charnov EL. 1976. Optimal foraging, the marginal value theorem. Theoretical Population Biology 9:129-136.

Chown SL. 2010. Temporal biodiversity change in transformed landscapes: a southern African perspective. Philosophical Transactions of the Royal Society B-Biological Sciences 365:3729-3742.

Cockburn A. 2006. Prevalence of different modes of parental care in birds. Proceedings of the Royal Society B-Biological Sciences 273:1375-1383.

Collier BA, McCleery RA, Calhoun KW, Roques KG, Monadjem A. 2011. Detection probabilities of ungulates in the eastern Swaziland lowveld. South African Journal of Wildlife Research 41:61-67.

Collins SL, et al. 2014. A Multiscale, Hierarchical Model of Pulse Dynamics in Arid-Land Ecosystems. Annual Review of Ecology, Evolution, and Systematics 45:397-419.

Connell JH. 1988. Citation classic - some mechanisms producing structure in natural communities - a model and evidence from field experiments. Current Contents/Agriculture Biology & Environmental Sciences 30:16-16.

Cooper H, Hedges LV, Valentine JC. 2009. The Handbook of Research Synthesis and Meta- analysis. Page 615. Russell Sage Foundation, York, New York.

Cox WA, Thompson III FR, Faaborg J. 2012. Species and temporal factors affect predator- specific rates of nest predation for forest songbirds in the Midwest. The Auk 129:147- 155.

Cresswell W. 2008. Non-lethal effects of predation in birds. Ibis 150:3-17.

129

Cresswell W, Quinn JL. 2013. Contrasting risks from different predators change the overall nonlethal effects of predation risk. Behavioral Ecology 24:871-876.

D'Odorico P, Okin GS, Bestelmeyer BT. 2012. A synthetic review of feedbacks and drivers of shrub encroachment in arid grasslands. 5:520-530.

Darling ES, Côté IM. 2016. Quantifying the evidence for ecological synergies. Ecology Letters 11:1278-1286.

Dehling DM, Jordano P, Schaefer HM, Bohning-Gaese K, Schleuning M. 2016. Morphology predicts species' functional roles and their degree of specialization in plant-frugivore interactions. Proceedings of the Royal Society B-Biological Sciences 283: 20152444.

Devictor V, Mouillot D, Meynard C, Jiguet F, Thuiller W, Mouquet N. 2010. Spatial mismatch and congruence between taxonomic, phylogenetic and functional diversity: the need for integrative conservation strategies in a changing world. Ecology Letters 13:1030-1040.

Didham RK, Tylianakis JM, Gemmell NJ, Rand TA, Ewers RM. 2007. Interactive effects of habitat modification and species invasion on native species decline. Trends in Ecology & Evolution 22:489-496.

Dieguez H. 2017. Disentangling the signal of climatic fluctuations from land use: changes in ecosystem functioning in South American protected areas (1982‐2012). Remote Sensing in Ecology and Conservation 3. DOI: 10.1002/rse2.39.

Dorazio RM, Royle JA. 2005. Estimating size and composition of biological communities by modeling the occurrence of species. Journal of the American Statistical Association 100:389-398.

Dukes JS, Mooney HA. 1999. Does global change increase the success of biological invaders? Trends in Ecology & Evolution 14:135-139.

Egger M, Smith GD, Schneider M, Minder C. 1997. Bias in meta-analysis detected by a simple, graphical test. British Medical Journal 315:629-634.

Eldridge DJ, Bowker MA, Maestre FT, Roger E, Reynolds, JF, Whitford WG. 2011. Impacts of shrub encroachment on ecosystem structure and functioning: towards a global synthesis. Ecology Letters 14:709-722.

Ellis EC, Ramankutty N. 2008. Putting people in the map: anthropogenic biomes of the world. Frontiers in Ecology and the Environment 6:439-447.

Fabricante I, Oesterheld M, Paruelo JM. 2009. Annual and seasonal variation of NDVI explained by current and previous precipitation across Northern Patagonia. Journal of Arid Environments 73:745-753.

Feldman MW, Aoki K, Kumm J. 1996. Individual versus social learning: Evolutionary analysis in a fluctuating environment. Anthropological Science 104:209-231.

130

Fensham RJ, Fairfax RJ, Archer SR. 2005. Rainfall, land use and woody vegetation cover change in semi-arid Australian savanna. Journal of Ecology 93:596-606.

Ferrier S, Manion G, Elith J, Richardson K. 2007. Using generalized dissimilarity modelling to analyse and predict patterns of beta diversity in regional biodiversity assessment. Diversity and Distributions 13:252-264.

Fiske IJ, Chandler RB. 2011. Unmarked: An R Package for Fitting Hierarchical Models of Wildlife Occurrence and Abundance. Journal of Statistical Software 43:1-23.

Fitzpatrick M, Lisk M. 2016. gdm: Functions for Generalized Dissimilarity Modeling v. R- package version 1.1-7.

Flynn DFB, et al. 2016. Loss of functional diversity under land use intensification across multiple taxa. Ecology Letters 12:22-33.

Foley JA, et al. 2005. Global Consequences of Land Use. Science 309:570-574.

Foster CN, Sato CF, Lindenmayer DB, Barton PS. 2016. Integrating theory into disturbance interaction experiments to better inform ecosystem management. Global Change Biology 22:1325-1335.

Francis CD, Ortega CP, Cruz A. 2009. Noise Changes Avian Communities and Species Interactions. Current Biology 19:1415-1419.

Fretwell SD. 1972. Populations in Seasonal Environments. Princeton University Press, Princeton, NJ.

Fontaine JJ, Martin TE. 2006. Habitat selection responses of parents to offspring predation risk: an experimental test. The American Naturalist 168:811-818.

Gamfeldt L, et al. 2013. Higher levels of multiple ecosystem services are found in forests with more tree species. Nature Communications 4:1340.

Gaugris JY, Van Rooyen MW. 2010. Woody vegetation structure in conserved versus in a biodiversity hotspot: A case study in Maputaland, South Africa. South African Journal of Botany 76:289-298.

Gelman A, Hill J. 2007. Data analysis using regression and multilevel/hierarchical models. Cambridge University Press.

Gelman A, Rubin DB. 1992. Inference from Iterative Simulation Using Multiple Sequences. Statistical Science 7:457-472.

Giraldeau LA, Valone TJ, Templeton JJ. 2002. Potential disadvantages of using socially acquired information. Philosophical Transactions of the Royal Society of London Series B- Biological Sciences 357:1559-1566.

131

Goheen JR, Palmer TM. 2010. Defensive Plant-Ants Stabilize Megaherbivore-Driven Landscape Change in an African Savanna. Current Biology 20:1768-1772.

Goldammer JG, Penafiel SR. 1990. Fire in the pine-grassland biomes of tropical and subtropical Asia. In Fire in the Tropical Biota (pp. 45-62). Springer Berlin Heidelberg.

Gossner MM, et al. 2016. Land-use intensification causes multitrophic homogenization of grassland communities. Nature 540:266-269.

Goudie AS, Price Williams D. 1983. The Atlas of Swaziland in Price Williams D, editor. Occasional Papers. Swaziland National Trust Commission, Swaziland.

Grant TA, Madden E, Berkey GB. 2004. Tree and shrub invasion in northern mixed-grass prairie: implications for breeding grassland birds. Wildlife Society Bulletin 32:807-818.

Guillera-Arroita G, Ridout MS, Morgan BJT, Linkie M. 2012. Models for species-detection data collected along transects in the presence of abundance-induced heterogeneity and clustering in the detection process. Methods in Ecology and Evolution 3:358-367.

Gurevitch J, Curtis PS, Jones MH. 2001. Meta-analysis in ecology. Advances in Ecological Research 32:199-247.

Gurevitch J, Morrison JA, Hedges LV. 2000. The interaction between competition and predation: A meta-analysis of field experiments. American Naturalist 155:435-453.

Hairston NG, Smith FE, Slobodkin LB. 1960. Community structure, population control, and competition. American Naturalist 94:421-425.

Hanan NP, Sankaran M. 2003. Structure and function in African savannas: A continental scale analysis of patterns. Ecological Society of America Annual Meeting Abstracts 88:138- 138.

Hansen AJ, Neilson RP, Dale VH, Flather CH, Iverson LR, Currie DJ, Shafer S, Cook R, Bartlein PJ. 2001. Global Change in Forests: Responses of Species, Communities, and Biomes: Interactions between climate change and land use are projected to cause large shifts in biodiversity. BioScience 51:765-779.

Hanski I. 1983. Predator mediated coexistence. Acta Entomologica Fennica 42:33-35.

Harrison S, Grace JB. 2007. Biogeographic affinity helps explain productivity-richness relationships at regional and local scales. American Naturalist 170:S5-S15.

Hastings JR, Turner RM. 1965. The changing mile. An ecological study of vegetation change with time in the lower mile of an arid and semiarid region. University of Arizona Press.

Higgins JPT, Thompson SG, Deeks JJ, Altman DG. 2003. Measuring inconsistency in meta- analyses. British Medical Journal 327:557-560.

132

Hockey PAR, Dean WRJ, Ryan PG. 2005. Roberts' birds of southern Africa. Seventh edition. John Voelcker Bird Book Fund Westville, South Africa.

Holden ST, Otsuka K. 2014. The roles of reforms and land markets in the context of and land use intensification in Africa. Food Policy 48:88-97.

Holt RD. 1977. Predation, apparent competition, and structure of prey communities. Theoretical Population Biology 12:197-229.

Hua FY, Fletcher RJ, Sieving KE, Dorazio RM. 2013. Too risky to settle: avian community structure changes in response to perceived predation risk on adults and offspring. Proceedings of the Royal Society B-Biological Sciences 280: 20130762.

Hua FY, Sieving KE, Fletcher RJ, Wright CA. 2014. Increased perception of predation risk to adults and offspring alters avian reproductive strategy and performance. Behavioral Ecology 25:509-519.

Hurlbert AH, Haskell JP. 2003. The effect of energy and seasonality on avian species richness and community composition. American Naturalist 161:83-97.

Hylander K, Nilsson C, Jonsson BG, Gothner T. 2005. Differences in habitat quality explain nestedness in a land snail meta-community. Oikos 108:351-361.

Jetz W, Wilcove DS, Dobson AP. 2007. Projected impacts of climate and land-use change on the global diversity of birds. PloS Biology 5:1211-1219.

Johnson DH. 1980. The comparison of usage and availability measurements for evaluating resource preference. Ecology 61:65-71.

Junk W et al. 2006. Biodiversity and its conservation in the Pantanal of Mato Grosso, Brazil. Aquatic Sciences-Research Across Boundaries 68:278-309.

Keil P, et al. 2012. Patterns of beta diversity in Europe: the role of climate, land cover and distance across scales. Journal of Biogeography 39:1473-1486.

Kellner K. 2016. jagsUI: A Wrapper Around 'rjags' to Streamline JAGS' Analyses. Available online at: https://cran.r-project.org/web/packages/jagsUI/jagsUI.pdf.

Kerr JT, Southwood TRE, Cihlar J. 2001. Remotely sensed habitat diversity predicts species richness and community similarity in . Proceedings of the National Academy of Sciences of the of America 98:11365-11370.

Kery M, Royle JA. 2015. Applied Hierarchical Modeling in Ecology: Analysis of distribution, abundance and species richness in R and BUGS: Volume 1: Prelude and Static Models. Academic Press.

133

Knapp AK, et al. 2008. Shrub encroachment in North American grasslands: shifts in growth form dominance rapidly alters control of ecosystem carbon inputs. Global Change Biology 14:615-623.

Kroodsma DE. 1989. Suggested experimental-designs for song playbacks. Animal Behaviour 37:600-609.

Kroodsma DE, Byers BE, Goodale E, Johnson S, Liu WC. 2001. Pseudoreplication in playback experiments, revisited a decade later. Animal Behaviour 61:1029-1033.

Kutt AS, Martin TG. 2010. Bird foraging height predicts bird species response to woody vegetation change. Biodiversity and Conservation 19:2247-2262.

Laiolo P, Dondero F, Ciliento E, Rolando A. 2004. Consequences of pastoral abandonment for the structure and diversity of the alpine avifauna. Journal of Applied Ecology 41:294- 304.

LaManna JA, Martin TE. 2016. Costs of fear: behavioural and life-history responses to risk and their demographic consequences vary across species. Ecology Letters 19:403-413.

Lambin EF, Geist HJ, Lepers E. 2003. Dynamics of land-use and land-cover change in tropical regions. Annual Review of Environment and Resources 28:205-241.

Lawton JH. 1999. Are there general laws in ecology? Oikos 84:177-192.

Lehman CER, Parr CL. 2016. Tropical grassy biomes: linking ecology, human use and conservation. Philosophical Transactions of the Royal Society B-Biological Sciences 371. DOI: 10.1098/rstb.2016.0329.

Lehmann CER, Archibald SA, Hoffmann WA, Bond WJ. 2011. Deciphering the distribution of the savanna biome. New Phytologist 191:197-209.

Leuzinger S, Luo YQ, Beier C, Dieleman W, Vicca S, Korner C. 2011. Do global change experiments overestimate impacts on terrestrial ecosystems? Trends in Ecology & Evolution 26:236-241.

Leynaud GC, Bucher EH. 2005. Restoration of degraded Chaco woodlands: Effects on reptile assemblages. Forest Ecology and Management 213:384-390.

Lima SL. 1986. Predation risk and unpredictable feeding conditions - determinants of body-mass in birds. Ecology 67:377-385.

Lima SL. 1992. Strong preferences for apparently dangerous habitats - a consequence of differential escape from predators. Oikos 64:597-600.

Lima SL. 2009. Predators and the breeding bird: behavioral and reproductive flexibility under the risk of predation. Biological Reviews 84:485-513.

134

Lima SL, Bednekoff PA. 1999. Temporal variation in danger drives antipredator behavior: The predation risk allocation hypothesis. American Naturalist 153:649-659.

Lima SL, Dill LM. 1990. Behavioral decisions made under the risk of predation - a review and prospectus. Canadian Journal of Zoology 68:619-640.

Lipsey MW, Wilson DB. 2001. Practical Meta-analysis. Sage.

Loffler L, Loffler P. 2005. Swaziland tree atlas; including selected shrubs and climbers, Southern African Botanical Diversity Network Report.

Logue JB, Mouquet N, Peter H, Hillebrand H, Metacommunity Working Group. 2011. Empirical approaches to metacommunities: a review and comparison with theory. Trends in Ecology & Evolution 26:482-491.

Lonsdale WM. 1999. Global patterns of plant invasions and the concept of invasibility. Ecology 80:1522-1536.

Mac Nally R, Horrocks G, Lada H, Lake PS, Thomson JR, Taylor AC. 2009. Distribution of anuran amphibians in massively altered landscapes in south-eastern Australia: effects of climate change in an aridifying region. Global Ecology and Biogeography 18:575-585.

MacKenzie DI, Nichols JD, Lachman GB, Droege S, Royle JA, Langtimm CA. 2002. Estimating site occupancy rates when detection probabilities are less than one. Ecology 83:2248- 2255.

MacKenzie DI, Nichols JD, Hines JE, Knutson MG, Franklin AB. 2003. Estimating site occupancy, colonization, and local extinction when a species is detected imperfectly. Ecology 84:2200-2207.

Maestre FT, et al. 2012. Plant Species Richness and Ecosystem Multifunctionality in Global Drylands. Science 335:214-218.

Martin TE, Geupel GR. 1993. Nest-monitoring plots - methods for locating nests and monitoring success. Journal of Field Ornithology 64:507-519.

Martin TE, Li PJ. 1992. Life-history traits of open-nesting vs cavity-nesting birds. Ecology 73:579-592.

McGill BJ, Enquist BJ, Weiher E, Westoby M. 2006. Rebuilding community ecology from functional traits. Trends in Ecology & Evolution 21:178-185.

Midgley GF, Bond WJ. 2015. Future of African terrestrial biodiversity and ecosystems under anthropogenic climate change. Nature Climate Change 5:823-829.

Mihaljevic JR, Joseph MB, Johnson PTJ. 2015. Using multispecies occupancy models to improve the characterization and understanding of metacommunity structure. Ecology 96:1783-1792.

135

Mole S. 1994. Trade-offs and constraints in plant-herbivore defense theory - a life-history perspective. Oikos 71:3-12.

Monadjem A. 2000. An avifaunal (Aves) survey of the lower Usuthu basin, Swaziland. Durban Museum Novitates 25:25-31.

Monadjem A. 2005. Association between avian communities and vegetation structure in a low- lying woodland-savanna ecosystem in Swaziland. Ostrich 76:45-55.

Mooney H, Larigauderie A, Cesario M, Elmquist T, Hoegh-Guldberg O, Lavorel S, Mace GM, Palmer M, Scholes R, Yahara T. 2009. Biodiversity, climate change, and ecosystem services. Current Opinion in Environmental 1:46-54.

Mönkkönen M, Härdling R, Forsman JT, Tuomi J. 1999. Evolution of heterospecific attraction: using other species as cues in habitat selection. Evolutionary Ecology 13: 91-104.

Morton SR, et al. 2011. A fresh framework for the ecology of arid Australia. Journal of Arid Environments 75:313-329.

Muenkemueller T, de Bello F, Meynard CN, Gravel D, Lavergne S, Mouillot D, Mouquet N, Thuiller W. 2012. From diversity indices to community assembly processes: a test with simulated data. Ecography 35:468-480.

Muggeo VMR. 2003. Estimating regression models with unknown break-points. Statistics in Medicine 22:3055-3071.

Murphy BP, Bowman D. 2012. What controls the distribution of and savanna? Ecology Letters 15:748-758.

Musters CJM, de Graaf HJ, ter Keurs WJ. 2000. Ecology - Can protected areas be expanded in Africa? Science 287:1759-1760.

Myers N, Mittermeier RA, Mittermeier CG, da Fonseca GAB, Kent J. 2000. Biodiversity hotspots for conservation priorities. Nature 403:853-858.

Mysterud A, Ostbye E. 1999. Cover as a habitat element for temperate ungulates: effects on habitat selection and demography. Wildlife Society Bulletin 27:385-394.

Naito AT, Cairns DM. 2011. Patterns and processes of global shrub expansion. Progress in Physical 35:423-442.

Nakagawa S, Cuthill IC. 2007. Effect size, confidence interval and statistical significance: a practical guide for biologists. Biological Reviews 82:591-605.

Naughton-Treves L, Holland MB, Brandon K. 2005. The role of protected areas in conserving biodiversity and sustaining local livelihoods. Annual Review of Environment and Resources 30:219-252.

136

Newbold T, et al. 2016. Global patterns of terrestrial assemblage turnover within and among land uses. Ecography 39. DOI: 10.1111/ecog.01932.

Nicholson E, et al. 2009. Priority research areas for ecosystem services in a changing world. Journal of Applied Ecology 46:1139-1144.

Nicholson SE, Farrar TJ. 1994. The influence of soil type on the relationships between ndvi, rainfall, and soil-moisture in semiarid .1. NDVI response to rainfall. Remote Sensing of Environment 50:107-120.

Ogada D, et al. 2016. Another Continental Vulture Crisis: Africa's Vultures Collapsing toward Extinction. Conservation Letters 9:89-97.

Oksanen J, et al. 2016. vegan:Community Ecology Package. R package version 2.3- 5. http://CRAN.R-project.org/package=vegan

Oliver TH, Morecroft MD. 2014. Interactions between climate change and land use change on biodiversity: attribution problems, risks, and opportunities. Wiley Interdisciplinary Reviews-Climate Change 5:317-335.

Olson DM, et al. 2001. Terrestrial ecoregions of the : A new map of life on Earth. Bioscience 51:933-938.

Owens IPF, Bennett PM. 2000. Ecological basis of extinction risk in birds: Habitat loss versus human persecution and introduced predators. Proceedings of the National Academy of Sciences 97:12144-12148.

Parmentier I, Hardy OJ. 2009. The impact of ecological differentiation and dispersal limitation on species turnover and phylogenetic structure of inselberg's plant communities. Ecography 32:613-622.

Parr CL, Lehmann CER, Bond WJ, Hoffmann WA, Andersen AN. 2014. Tropical grassy biomes: misunderstood, neglected, and under threat. Trends in Ecology & Evolution 29:205-213.

Parsons JJ. 1955. The Miskito pine savanna of Nicaragua and Honduras. Annals of the Association of American Geographers 45:36-63.

Pausas JG, Keeley JE. 2009. A Burning Story: The Role of Fire in the . BioScience 59:593-601.

Pettorelli N, Vik JO, Mysterud A, Gaillard JM, Tucker CJ, Stenseth NC. 2005. Using the satellite-derived NDVI to assess ecological responses to environmental change. Trends in Ecology & Evolution 20:503-510.

Phalan B, Onial M, Balmford A, Green RE. 2011. Reconciling Food Production and Biodiversity Conservation: Land Sharing and Land Sparing Compared. Science 333:1289-1291.

137

Plummer M. 2003. JAGS: A program for analysis of Bayesian graphical models using Gibbs sampling. Proceedings of the 3rd international workshop on distributed statistical computing.

Polley HW, Johnson HB, Mayeux HS. 1994. Increasing CO2 - comparative responses of the C-4 grass Schizachyrium and grassland invader Prosopis. Ecology 75:976-988.

Preisser EL, Bolnick DI, Benard MF. 2005. Scared to death? The effects of intimidation and consumption in predator-prey interactions. Ecology 86:501-509.

Preisser EL, Orrock JL. 2012. The allometry of fear: interspecific relationships between body size and response to predation risk. Ecosphere 3. DOI: 10.1890/ES12-00084.1.

Pringle RM, Young TP, Rubenstein DI, McCauley DJ. 2007. Herbivore-initiated interaction and their modulation by productivity in an African savanna. Proceedings of the National Academy of Sciences of the United States of America 104:193-197.

Purvis A, Gittleman JL, Cowlishaw G, Mace GM. 2000. Predicting extinction risk in declining species. Proceedings of the Royal Society of London B: Biological Sciences 267:1947- 1952.

R Core Team. 2016. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL: http://www.R-project.org/. Radford AN, Bell MB, Hollén LI, Ridley AR. 2011. Singing for your supper: sentinel calling by kleptoparasites can mitigate the cost to victims. Evolution 65:900-906.

Radford IJ, Andersen AN. 2012. Effects of fire on grass-layer savanna macroinvertebrates as key food resources for insectivorous vertebrates in northern Australia. Austral Ecology 37:733-742.

Ratnam J, Tomlinson KW, Rasquinha DN, Sankaran M. 2016. Savannahs of Asia: antiquity, biogeography, and an uncertain future. Philosophical Transactions of the Royal Society B-Biological Sciences 371. DOI: 10.1098/rstb.2015.0305.

Ricketts AM, Sandercock BK. 2016. Patch‐burn grazing increases habitat heterogeneity and biodiversity of small mammals in managed rangelands. Ecosphere 7. DOI: 10.1002/ecs2.1431.

Ricklefs RE. 2004. A comprehensive framework for global patterns in biodiversity. Ecology Letters 7:1-15.

Ripple WJ, et al. 2015. Collapse of the world’s largest herbivores. Science Advances 1: e1400103.

Royle JA, Dorazio RM. 2008. Hierarchical modeling and inference in ecology: the analysis of data from populations, metapopulations and communities. Academic Press.

138

Sala OE, et al. 2000. Biodiversity - Global biodiversity scenarios for the year 2100. Science 287:1770-1774.

Sankaran M, et al. 2005. Determinants of woody cover in African savannas. Nature 438:846-849.

Sankaran M, Ratnam J, Hanan N. 2008. Woody cover in African savannas: the role of resources, fire and herbivory. Global Ecology and Biogeography 17:236-245.

Sankaran M, Ratnam J, Hanan NP. 2004. Tree-grass coexistence in savannas revisited - insights from an examination of assumptions and mechanisms invoked in existing models. Ecology Letters 7:480-490.

Scholes RJ, Archer SR. 1997. Tree-grass interactions in savannas. Annual Review of Ecology and Systematics 28:517-544.

Şekercioğlu CH. 2002. Forest fragmentation hits insectivorous birds hard. TheScientificWorld Journal 1:62-64.

Şekercioğlu ÇH, Daily GC, Ehrlich PR. 2004. Ecosystem consequences of bird declines. Proceedings of the National Academy of Sciences 101:18042-18047.

Seppänen J, Forsman J, Mönkkönen M, Thomson RL. 2007. Social information use is a process across time and ecology, reaching heterospecifics. Ecology 88: 1622-1633.

Seymour CL, Dean WRJ. 2010. The influence of changes in habitat structure on the species composition of bird assemblages in the southern Kalahari. Austral Ecology 35:581-592.

Shannon CE. 1948. A mathematical theory of communication. Bell System Technical Journal 27:623-656.

Shaw MR, Zavaleta ES, Chiariello NR, Cleland EE, Mooney HA, Field CB. 2002. Grassland responses to global environmental changes suppressed by elevated CO2. Science 298:1987-1990.

Sih A, Englund G, Wooster D. 1998. Emergent impacts of multiple predators on prey. Trends in Ecology & Evolution 13:350-355.

Simberloff D. 2004. Community ecology: Is it time to move on? American Naturalist 163:787- 799.

Sirami C, Monadjem A. 2012. Changes in bird communities in Swaziland savannas between 1998 and 2008 owing to shrub encroachment. Diversity and Distributions 18:390-400.

Sirami C, Seymour C, Midgley G, Barnard P. 2009. The impact of shrub encroachment on savanna bird diversity from local to regional scale. Diversity and Distributions 15:948- 957.

Skarpe C. 1991. Impact of grazing in savanna ecosystems. Ambio 20:351-356.

139

Skowno AL, Bond WJ. 2003. Bird community composition in an actively managed savanna reserve, importance of vegetation structure and vegetation composition. Biodiversity and Conservation 12:2279-2294.

Slagsvold T. 2001. Bird nests and construction behaviour. Nature 409:18-19.

Socolar JB, Gilroy JJ, Kunin WE, Edwards DP. 2016. How Should Beta-Diversity Inform Biodiversity Conservation? Trends in Ecology & Evolution 31:67-80.

Sterne JAC, et al. 2011. Recommendations for examining and interpreting funnel plot asymmetry in meta-analyses of randomised controlled trials. British Medical Journal 343. DOI: 10.1136/bmj.d4002.

Stevens N, Erasmus BFN, Archibald S, Bond WJ. 2016a. Woody encroachment over 70 years in South African savannahs: overgrazing, global change or extinction aftershock? Proceedings of the Royal Society B-Biological Sciences 371: DOI: 10.1098/rstb.2015.043.

Stevens N, Lehmann C, Murphy B, Durigan G. 2016b. Savanna woody encroachment is widespread across three continents. Global Change Biology 23:235-244.

Stewart GB, Schmid CH. 2015. Lessons from meta-analysis in ecology and evolution: the need for trans-disciplinary evidence synthesis methodologies. Research Synthesis Methods 6:109-110.

Switzer PV. 1993. Site fidelity in predictable and unpredictable habitats. Evolutionary Ecology 7:533-555.

Tansley AG, Chipp TF.1926. Aims and methods in the study of vegetation. British Empire Vegetation Committee. 384 pp. Tassicker AL, Kutt AS, Vanderduys E, Mangru S. 2006. The effects of vegetation structure on the birds in a tropical savanna woodland in north-eastern Australia. Rangeland Journal 28:139-152.

Tredennick AT, Hanan NP, Associate Editor: Egbert HVN, Editor: Susan K. 2015. Effects of Tree Harvest on the Stable-State Dynamics of Savanna and Forest. The American Naturalist 185:E153-E165.

Turner BL, II, Lambin EF, Reenberg A. 2007. The emergence of land change science for global environmental change and sustainability. Proceedings of the National Academy of Sciences of the United States of America 104:20666-20671.

Turner MG, Romme WH, Gardner RH, Oneill RV, Kratz TK. 1993. A revised concept of landscape equilibrium - disturbance and stability on scaled landscapes. 8:213-227.

140

Twine WC, Holdo RM. 2016. Fuelwood sustainability revisited: integrating size structure and resprouting into a spatially realistic fuelshed model. Journal of Applied Ecology 53:1766- 1776.

Tylianakis JM, Tscharntke T, Lewis OT. 2007. Habitat modification alters the structure of tropical host-parasitoid food webs. Nature 445:202-205.

Van Auken OW. 2000. Shrub invasions of North American semiarid grasslands. Annual Review of Ecology and Systematics 31:197-215.

Vellend M. 2010. Conceptual synthesis in community ecology. Quarterly Review of Biology 85:183-206.

Verhulst J, Baldi A, Kleijn D. 2004. Relationship between land-use intensity and species richness and abundance of birds in Hungary. Agriculture Ecosystems & Environment 104:465-473.

Viechtbauer W. 2010. Conducting Meta-Analyses in R with the metafor Package. Journal of Statistical Software 36:1-48.

Waltert M, Mardiastuti A, Muhlenberg M. 2004. Effects of land use on bird species richness in Sulawesi, . Conservation Biology 18:1339-1346.

Waters CM, Melville GJ, Toole ID, Smith WJ. 2016. Management of grazing intensity in the semi‐arid rangelands of southern Australia – effects on soil and biodiversity. & Development 28:1363-1375.

Whelan CJ, Sekercioglu CH, Wenny DG. 2015. Why birds matter: from economic ornithology to ecosystem services. Journal of Ornithology 156:S227-S238.

Whittaker RH. 1962. Classification of natural communities. Botanical Review 28:1-239.

Wigley BJ, Bond WJ, Hoffman MT. 2009. Bush encroachment under three contrasting land-use practices in a mesic South African savanna. African Journal of Ecology 47:62-70.

Wigley BJ, Bond WJ, Hoffman MT. 2010. expansion in a South African savanna under divergent land use: local vs. global drivers? Global Change Biology 16:964-976.

Williams JD. 1970. A regression approach to experimental design. The Journal of Experimental Education 39:83-90.

Wilson SK, Graham NAJ, Pratchett MS, Jones GP, Polunin NVC. 2006. Multiple disturbances and the global degradation of reefs: are reef fishes at risk or resilient? Global Change Biology 12:2220-2234.

Worthen WB, Jones MT, Jetton RM. 1998. Community structure and environmental stress: desiccation promotes nestedness in mycophagous fly communities. Oikos 81:45-54.

141

Yamaura Y, Kery M, Royle JA. 2016. Study of biological communities subject to imperfect detection: bias and precision of community N-mixture abundance models in small-sample situations. Ecological Research 31:289-305.

Young TP, Palmer TM, Gadd ME. 2005. Competition and compensation among cattle, zebras, and elephants in a semi-arid savanna in Laikipia, . Biological Conservation 122:351-359.

Zanette LY, White AF, Allen MC, Clinchy M. 2011. Perceived Predation Risk Reduces the Number of Offspring Songbirds Produce per Year. Science 334:1398-1401.

142

BIOGRAPHICAL SKETCH

Richard Stanton graduated from the State University of New York at Cobleskill in spring

2009 where he received his Bachelor of Technology degree in Wildlife Management. In spring

2013, Rich received his Master of Science degree in Fisheries and Wildlife Sciences from the

University of Missouri-Columbia. He received his Doctor of Philosophy in Interdisciplinary

Ecology from the School of Natural Resources and the Environment at the University of Florida in August 2017.

143