NATURAL RESOURCE USE IN MADAGASCAR
A Dissertation Submitted to the Temple University Graduate Board
In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY
by Kim E. Reuter July 2015
Examining Committee Members:
Brent J. Sewall, Advisory Chair, Biology Robert W. Sanders, Dissertation Examining Chair, Biology Erik E. Cordes, Biology Jacqueline C. Tanaka, Biology Hamil Pearsall, External Member, Temple University
© Copyright 2015
by
Kim E. Reuter All Rights Reserved
ii ABSTRACT
The anthropogenic use of natural resources has become a major cause of biodiversity loss and habitat degradation throughout the world. Deforestation - the conversion of forests to alternative land covers - has led to a decrease in local biodiversity directly through a decrease in habitat, and indirectly through habitat fragmentation. Likewise, defaunation – the loss of animals both directly through hunting and indirectly through deforestation – has led to the empty forest syndrome and subsequent deterioration of forest ecosystems. In many cases, areas where anthropogenic use of natural resources is high overlap with areas of high biodiversity value. Therefore, the present series of studies aims to better understand the impacts that different types of natural resources use and habitat degradation have on biodiversity. This dissertation details the results of five studies, which aimed to: 1) examine the effects of habitat degradation on plant-frugivore networks; 2), understand the live capture and extent of ownership of lemurs in Madagascar; 3) understand the micro- and macro-level drivers of wild meat consumption in Madagascar; 4) describe the capture, movement, and trade of wild meat in Madagascar; and 5) the impacts of habitat changes on the diets and vertical stratification of frugivorous bats.
For the first study, our objectives were to understand the effects of habitat degradation on (1) community structure, (2) network structure, and (3) seed dispersal services. We focused on fruit-bearing trees and frugivores (two lemur and five bird species) across a three-point gradient of habitat degradation in a tropical dry forest in
Madagascar. Our objectives were to understand the effects of habitat degradation on (1)
iii community structure, (2) network structure, and (3) seed dispersal services. We focused on fruit-bearing trees and frugivores (two lemur and five bird species) across a three- point gradient of habitat degradation in a tropical dry forest in Madagascar. Data on fruit consumption by frugivores were collected over 592 hours of observations at 13 fruiting tree species. We found that as habitat became more degraded: (1) the community structure of both frugivores and fruiting tree communities changed; (2) the mutualistic network structure became less complex and less connected; (3) the interaction strengths of pair-wise interactions changed and the asymmetries of these interactions shifted; and
(4) seed dispersal decreased by 91% in the secondary forest, compared to the primary forest. In addition, we show that frugivores: (1) sometimes stopped eating fruit in the degraded forest, even when they had consumed it in other forests; and (2) appeared to avoid some fruiting tree species while showing preference for others. The mutualistic network studied in this paper appeared sensitive to anthropogenic disturbance and a novel measure of effectiveness helped quantify these changes.
For the second study, our objectives were to provide the first quantitative estimates of the prevalence, spatial extent, correlates and timing of lemur ownership, procurement methods, within-country movements, and numbers and duration of ownership. Using semi-structured interviews of 1,093 households and 61 transporters, across 17 study sites, we found that lemur ownership was widespread and affected a variety of taxa. We estimate that 28,253 lemurs have been affected since 2010. Most lemurs were caught by owners and kept for either short (≤1 week) or long (≥3 years) periods. The live capture of lemurs in Madagascar is not highly organized but may threaten several endangered species.
iv For the third study, we investigated the role of wild meat in food security in
Madagascar, a country where wild meat consumption is poorly understood in urban areas and at regional scales. Using semi-structured interviews (n = 1339 heads-of-households,
21 towns), we aimed to: 1) quantify the amount and purpose of; 2) understand the drivers behind; and, 3) examine recent changes in wild meat consumption in Madagascar. Few respondents preferred wild meat (8 ± 3%) but most had eaten it at least once (78 ± 7%), and consumption occurred across ethnic groups, in urban and rural settings. More food insecure areas reported higher rates of recent consumption of wild meat. However, consumption was best explained by individual preferences and taboos. Few respondents
(<1 ± <1%) had increased rates of consumption during their lifetimes, and wild meat prices showed no change from 2005-2013. Most consumption involved wild pigs and small-bodied animals, though these animal groups and lemurs were consumed less in recent years. Given these data, wild meat is unlikely to enhance food security for most
Malagasy people in urban and well-connected rural areas.
For the fourth study, and to improve understanding of the wild meat trade in
Madagascar, our objectives were to: (1) quantify the volume of consumption, transport, and sale for different animal groups, compared to domestic meat; (2) describe the methods of capture and hunting for different animal groups; (3) analyze the patterns of movement of wild meat from the capture location to the final consumer, compared to domestic meat; and (4) examine how the prices of wild meat change depending on the venue through which the consumer purchases it. Data was collected in May-August 2013 using semi-structured interviews of consumers (n = 1343 households, 21 towns), meatsellers (n = 520 restaurants, open-air markets stalls, and supermarkets, 9 towns), and
v drivers of inter-city transit vehicles (n = 61, 5 towns). We found that: (1) a wide range of hunting methods were used, though their prevalence of use differed by animal group; (2) wild meat traveled distances of up to 166 km to reach consumers, though some animal groups were hunted locally (<10 km) in rural areas; (3) most wild meat was procured from free sources (hunting and receiving meat as a gift), though urban respondents who consumed bats and wild pigs were more likely to purchase those meats; and (4) wild meat was consumed at lower rates than domestic meat, though urban respondents consumed twice as much wild meat as rural respondents. We conclude that urban and rural respondents differ in how they interact with the wild meat commodity chain. We also believe that the consumption and trade of wild meat in Madagascar is likely more formalized that previously thought.
Finally, for our fifth study, we used stable isotope analysis to examine how foraging by three fruit bat species in Madagascar, Pteropus rufus, Eidolon dupreanum, and Rousettus madagascariensis, are impacted by habitat change across a large spatial scale. Our results indicated that the three species had broadly overlapping diets.
Differences in diet were nonetheless detectable and consistent between P. rufus and E. dupreanum, and these diets shifted when they co-occurred, suggesting resource partitioning across habitats and vertical strata within the canopy to avoid competition.
Changes in diet were also correlated with a decrease in forest cover, though at a larger spatial scale in P. rufus than in E. dupreanum. These results suggest fruit bat species exhibit differing foraging strategies in response to habitat change. They also highlight the key threats that fruit bats face from habitat change, and clarify the spatial scales at which
vi conservation efforts should be implemented to mitigate threats for these bat species in
Madagascar.
vii
“Not all those who wander are lost”
For my father, Peter, who is quite simply the best father anyone could ask for; it is from
him that I have learned a deep love for adventure, for challenges, and
for doing work that really matters.
For my sister, Ann, who brings sunshine to all those around her.
For my sister, Lauren, whose abilities to create and whose inner strength inspire me.
In memory of my mother, Christine, who was vibrant,
colorful, wonderful, and will always be missed.
viii ACKNOWLEDGMENTS
Many thanks to Brent Sewall for his guidance, training, and patience during my time at Temple University. I am especially thankful for his attention to detail, ability to think of the big picture, and for encouraging me to embrace research projects that I was interested in, even if they were outside his realm of expertise. Thanks also to Erik Cordes,
Jackie Tanaka, Bob Sanders, and Rob Kulathinal for their endless support, enthusiasm, guidance, knowledge, and advice. I also thank Hamil Pearsall for her unique perspective and for graciously serving as the external reviewer for this dissertation. I thank the faculty and staff of the Department of Biology, as well as my fellow graduate students; particularly Jay Lunden, Samuel Georgian, Danielle DeLeo, Craig Stanley Jr., Elizabeth
Reilly, Katherine Papacostas, and Sarah DeVaul.
I am deeply grateful to Andrea Gudiel, Shane Nieves, Haley Gilles, Melissa
Schaefer, and Abigail Wills for their assistance in the field. This dissertation would not have been possible without their strong work ethic, ability to overcome obstacles, and humor in the face of adversity. I am also indebted to my Malagasy assistants and translators, who made the research come to life and provided valuable insight into
Malagasy culture: Totozafy Eric Janvier, Sahondra Hanitriniaina, Jocelyn
Randrianarivelo, Tertius Rodriguez Belalahy, and Claudesse Barat. Many thanks to
Ismael Leandre, Raymond Raherindray, Olivier Raynaud, and Elodie Camprasse for translating English documents for me into Malagasy and French – usually on short notice.
I thank my host communities in Madagascar for their warmth and for their incredible willingness to host myself and my research teams. I am especially grateful towards the
ix people of Mahamasina for their assistance, including Laurent who took great care of my research team and who was always happy to chat in English.
Many other people provided advice on this research or collaborated on the dissertation in some way. In particular, I would like to thank Moritz Ritter (Temple
University) for assistance in analyzing historical wild meat price data. I also thank
Raymond Lee (Washington State University) for analyzing the hair samples collected from wild animals in Madagascar at no cost and for answering the many questions I had about the interpretation of stable isotope values.
Many thanks to my friends and family for their love and support. Thanks for
Micah Burkey for believing in me during my first years at Temple University and to
Toby Schaeffer for his never-ending cheerleading. Thanks to my father, Peter Reuter, for supporting me financially and emotionally when times got rough. Thanks to my sisters –
Ann Pankow and Lauren Reuter – for their ability to make me laugh and to my family in
Germany (especially my grandparents, Heide and Helmut Elsinger) for always welcoming me with open arms on my layovers to and from my field sites.
The research could not have been completed without assistance from: the Temple
University Internal Review Board, the Temple University Institutional Animal Care and
Use Committee, Madagascar National Parks, the Madagascar Ministry of Water and
Forests, the Madagascar Institute for the Conservation of Tropical Environments,
Community Centred Conservation Madagascar, Azimut, and the Ankarana National Park staff. Funding was provided by the National Science Foundation GRFP (Grant DGE-
1144462) to KER; National Science Foundation grant (DEB-1257916) to BJS; Explorers
Club grants to KER and SN; American Society of Primatology grant to KER; Ding
x Darling Scholarship to KER; and a Temple University Faculty Senate grant to BJS. All opinions, findings, conclusions and recommendations expressed are those of the authors and do not necessarily reflect the views of the National Science Foundation.
Chapter Three was originally published as Reuter et al. (2015) Live capture and ownership of lemurs in Madagascar: extent and conservation implications. Oryx. doi:10.1017/S003060531400074X and is reproduced here with permission of Cambridge
University Press.
xi TABLE OF CONTENTS
Page
ABSTRACT ...... iii
DEDICATION ...... viii
ACKNOWLEDGMENTS ...... ix
LIST OF TABLES ...... xviii
LIST OF FIGURES ...... xix
CHAPTER
1. INTRODUCTION ...... 1
Madagascar ...... 2
Natural resource use in Madagascar ...... 3
Implications of natural resource use on conservation initiatives in
Madagascar ...... 4
Focal research areas ...... 6
2. EFFECTS OF HABITAT DEGRADATION ON PLANT-FRUGIVORE
NETWORKS ...... 10
Abstract ...... 10
Introduction ...... 11
Methods ...... 17
Study site ...... 17
Forest types and transects ...... 18
xii Classification of forest types and transects ...... 18
Frugivores ...... 19
Habitat sampling ...... 19
Focal observations ...... 20
Analysis of frugivore and tree characteristics ...... 21
Binary network analysis ...... 22
Weight network analysis ...... 22
Results ...... 24
Effects of habitat degradation on community structure ...... 24
Effects of habitat degradation on network structure ...... 28
Effects of habitat degradation on seed dispersal ...... 36
Discussion ...... 36
Effects of habitat degradation on community and network structure ...... 36
Effects of habitat degradation on seed dispersaland conservation
implications ...... 42
3. LIVE CAPTURE AND OWNERSHIP OF LEMURS IN MADAGASCAR:
EXTENT AND CONSERVATION IMPLICATIONS ...... 45
Abstract ...... 45
Introduction ...... 45
Study Area ...... 50
Methods ...... 50
Data collection ...... 50
Analysis ...... 55
xiii Results ...... 56
Discussion ...... 62
Extent of ownership ...... 62
Patterns of ownership ...... 63
Animal welfare ...... 65
Conservation implications ...... 65
4. THE CONSUMPTION OF WILD MEATS IN MADAGASCAR: FOOD
SECURITY, DRIVERS OF CONSUMPTION, AND POPULARITY AS A
FOOD ITEM ...... 69
Abstract ...... 69
Introduction ...... 70
Methods ...... 74
Study site ...... 74
Research permissions ...... 76
Social surveys ...... 76
Analysis ...... 77
Results ...... 80
Amount and purpose of wild meat consumption ...... 80
Micro- and macro-level drivers of wild meat consumption ...... 88
Variation in wild meat consumption over time and space ...... 93
Discussion ...... 99
Amount and purpose of wild meat consumption ...... 99
Micro- and macro-level drivers of wild meat consumption ...... 100
xiv Variation in wild meat consumption over time and space ...... 101
Conservation implications ...... 103
Wild meat consumption and food security ...... 104
5. CAPTURE OF WILD ANIMALS AND MOVEMENT, TRADE, AND
CONSUMPTION OF WILD MEAT IN MADAGASCAR ...... 108
Abstract ...... 108
Introduction ...... 109
Methods ...... 119
Study site ...... 119
Legality of wild meat hunting and consumption in Madagascar ...... 119
Research permissions ...... 120
Social surveys ...... 122
Analysis ...... 126
Results ...... 129
Capture/hunting of wild meat animals (objective one) ...... 129
Movement of wild and domestic meat (objective two) ...... 137
Sale of wild and domestic meat (objective three) ...... 143
Volume of meat consumption (objective four) ...... 145
Discussion ...... 147
Hunting and capture of wild meat ...... 147
Movement of wild animals away from the capture location ...... 149
Exchange of wild meat via sale, barter, or gifting ...... 152
Consumption of wild meat ...... 153
xv Conservation implications ...... 154
6. IMPACTS OF HABITAT CHANGE ON THE DIETS AND VERTICAL
STRATIFICATION OF FRUGIVOROUS BATS ...... 161
Abstract ...... 161
Introduction ...... 162
Methods ...... 167
Study site ...... 167
Bat hair samples ...... 167
Stable isotope analysis ...... 172
Social surveys ...... 172
Ethical research statement ...... 173
Analysis ...... 173
Results ...... 177
Sampling effort ...... 177
Variation within individuals of each species ...... 177
Species differences in diet, diet breadth, and resource partitioning ...... 177
Regional and local differences and impact of forest cover and
climate on bat diets ...... 180
Discussion ...... 187
Diets and diet breadth of fruit bats in Madagascar ...... 188
Impact of forest cover on bat diets ...... 191
Conservation implications ...... 193
7. CONCLUSIONS ...... 195
xvi REFERENCES CITED ...... 204
APPENDICES
A. CHAPTER 2 SUPPLEMENTARY MATERIALS ...... 226
B. MEAT CONSUMPTION SURVEYS (HOUSEHOLDS) ...... 236
C. INTERVIEW QUESTIONS FOR MEAT SELLERS ...... 238
D. INTERVIEW QUESTIONS FOR TRANSPORTERS ...... 240
E. CHAPTER 5 SUPPLEMENTARY MATERIALS ...... 242
F. CHAPTER 6 SUPPLEMENTARY MATERIALS ...... 262
xvii LIST OF TABLES
Table Page
3.1: Study sites, with the sample size and the percentage of individuals who had either owned a lemur or seen someone else with a pet lemur ...... 53
3.2: Sightings of pet lemurs in Madagascar ...... 58
4.1: Towns included in the study and sample sizes of interviews ...... 75
4.2: Percent of individuals who had consumed wild meat once in their life and from 2010 to mid-2013 ...... 82
4.3: Range of mammalian groups consumed by town, number of wild animal groups consumed, and percent of people with meat-related taboos by town ...... 83
4.4: Results of model selection ...... 89
4.5: Changes in wild meat consumption, rates of recent consumption, and average duration since last consumption ...... 94
4.6: Shifts in wild meat consumption behavior among provinces ...... 99
5.1: Study sites included in project ...... 121
6.1: Towns where hair samples were collected ...... 170
6.2: Results of model selection ...... 186
S1: Length of time frugivores were observed ...... 227
S2: Information on fruiting trees observed ...... 228
S3: List of tree species included in mutualistic network ...... 230
S4: Density estimates of the diurnal frugivore guild ...... 231
S5: Number of visits of frugivores at fruiting trees ...... 232
S6: Scaled calculations of effectiveness ...... 233
xviii S7: Number of frugivores observed passing near canopy ...... 235
S8: Average distance traveled by consumer ...... 245
S9: Sale of meat in different venues ...... 246
S10: Transport of meat ...... 248
S11: Price paid by consumer ...... 249
S12: Percent of people who procured meat ...... 251
S13: Percent of people who procured meat from different sources ...... 252
S14: Volume of meat consumption ...... 254
S15: Stable isotope values by mammal species ...... 262
S16: Climate and forest cover characteristics ...... 263
S17: Stable isotope values for Mangifera indica ...... 264
xix LIST OF FIGURES
Figure Page
2.1: Changes in generality and seed dispersal services by frugivore species, across multiple forest types ...... 26
2.2: Effectiveness of fruiting trees in primary, secondary, and degraded forests ...... 29
2.3: Unequal pair-wise mutualisms by forest type ...... 32
2.4: Proportion of species with different species degree levels, by forest type ...... 33
2.5: Connectance and nestedness by forest type ...... 35
3.1: Maps of the study sites in Madagascar ...... 49
3.2: Locations in Madagascsar where respondents reported having seen pet lemurs during 2010-2013 and during 1960-2010 ...... 57
3.3: Distributions of the number of lemurs owned by all self-reported lemur owners ...... 61
3.4: Photographs of pet lemurs ...... 66
4.1: Percent of respondents who consumed different types of wild meat once in their lifetime and once following 2009 ...... 81
4.2: Consumer preference for wild and domestic meats ...... 84
4.3: Frequency of consumption of wild and domestic meats ...... 85
4.4: Reasons for wild meat consumption ...... 87
4.5: Impact of taboos against wild meat and a preference for wild meat at the respondent and town level ...... 91
4.6: Change in prices of one unit of wild meat over time ...... 95
4.7: Reasons why people decreased or stopped eating wild meat ...... 97
xx 5.1: Conceptual models of the commodity chain (A) and of the different sources from which a consumer can procure wild meat ...... 112
5.2: Hunting methods used to capture different types of wild mammals by urban and rural respondents ...... 131
5.3: Hunting methods used to capture different types of wild mammals in the past
(pre-2009) and more recently (2009-2013) ...... 133
5.4: Comparison of hunting methods used to capture wild mammals for sale and for personal consumption ...... 135
5.5: Percent of respondents who reported hunting wild mammal groups on a seasonal basis ...... 138
5.6: Time of year when respondents hunt for different types of wild mammals ...... 139
5.7: Distances that meat was transported or that urban and rural consumers traveled to procure wild and domestic meat ...... 140
5.8: Patterns of movement for wild and domestic meats as reported by urban and rural respondents ...... 146
6.1: Locations of cities where wild hair samples were collected ...... 169
6.2: Stable isotope values by species ...... 178
6.3: Comparison of δ15N and δ13C values for P. rufus and E. dupreanum at all towns where both were hunted ...... 182
6.4: Changins in diet (δ15N and δ13C values) when P. rufus and E. dupreanum were caught at hunting sites alone and alongside each other ...... 183
S1: Locations of transects ...... 226
S2: Sources of domestic meat ...... 256
xxi S3: Sources of commonly consumed wild meat ...... 258
S4: Sources of wild meat not commonly consumed ...... 260
xxii CHAPTER 1
INTRODUCTION
The anthropogenic use of natural resources has become a major cause of biodiversity loss and habitat degradation throughout the world (Dirzo and Raven 2003).
Anthropogenic habitat change has led to a decrease in local biodiversity directly through a decrease in habitat, and indirectly through habitat fragmentation and degradation
(reviewed by Fahrig 2003, Lamb et al. 2005, Wright 2005). Likewise, defaunation – the loss of animals directly through hunting and indirectly through deforestation – has often led to the empty forest syndrome (Wilkie et al. 2011) and subsequent deterioration of forest ecosystems (Gorchov et al. 1993, Medellin and Goano 1999, Kunz et al. 2011). In many cases, areas of high anthropogenic natural resource use overlap with areas of high biodiversity value (Myers et al. 2000). As such, it is understood that effective conservation of biodiversity requires an understanding of human behavior and how it drives natural resource use (Guerbois et al. 2013, Scales 2014). However, research on the drivers of natural resource use is a relatively recent endeavor (e.g. Mackenzie et al. 2011,
Pfeifer et al. 2012) and the motivations of natural resource use – both legal and illegal – and their interaction with conservation initiatives remain poorly understood (Guerbois et al. 2012). For example, in many cases, the unintended consequences of Integrated
Conservation and Development Projects (ICDPs) have actually resulted in an increase in natural resource use; ICDPs have unintentionally provided incentives (e.g. job security) for non-local people to immigrate closer to conservation projects. This can result in an increase in natural resource use both because of population increases but also because of
1 2 complex cultural and ethnic characteristics of the immigrant communities (Guerbois et al.
2013). Therefore, the present series of studies aims to better understand the impacts that natural resources use and habitat degradation have on biodiversity as well as the drivers and motivations behind this natural resource use. This dissertation details the results of five studies – with a regional focus on the island of Madagascar - which aimed to: 1) examine the effects of habitat degradation on plant-frugivore networks; 2), understand the live capture and extent of ownership of lemurs in Madagascar; 3) understand the micro- and macro-level drivers of wild meat consumption in Madagascar; 4) describe the capture, movement, and trade of wild meat in Madagascar; and 5) the impacts of habitat changes on the diets and vertical stratification of frugivorous bats.
Madagascar:
Madagascar, a global biodiversity hotspot containing 3.2% of the world’s plant species and 2.8% of the globe’s vertebrates (Myers et al. 2000) is the world’s fourth largest island. Home to a growing population (5.1 million in 1960 to 21.3 million people in 2011, World Bank 2013) the country has a turbulent political history, with the most recent coup d’état in 2009. In Madagascar, over 90% of the population lives on less than
2 USD per day (World Bank 2013), and political instability has exacerbated economic challenges; foreign aid decreased following the 2009 coup d’état, with losses of over
$400 million in foreign aid by 2012 (Ploch and Cook 2012). Further, 70% of the population regularly consume insufficient calories (< 2133 kcal/day) and an additional
8% and 3% of the rural and urban population, respectively, consume insufficient calories during seasonal rice shortages (Dostie et al. 2002). In part as a result, Madagascar has a 3 high rate (50%) of adolescent malnutrition compared with other African countries (rates typically <40%, Fotso 2007).
Natural resource use in Madagascar:
People in Madagascar regularly extract natural resources – wood (e.g. Innes
2010), non-timber forest products (e.g. Novy 1997), sapphires (e.g. Walsh 2005), and animals (e.g. Golden et al. 2011) – for personal use, cultural reasons, and economic gain.
This extraction occurs both legally and illegally and is practiced by people of different ethnicities, religions, and socioeconomic backgrounds (Innes 2010, Walsh 2005, Golden et al. 2011). Natural resource use behaviors are not static; for example, illegal logging increased in the years following the 2009 coup d’état (Innes 2010). However, these trends can be reversed; it has been noted that the paradigm of viewing rural Malagasy communities as too poor or uneducated to change their natural resource use behaviors is outdated (Scales 2014).
The reasons why local communities continue to use natural resources even when it is illegal are varied and diverse. In some cases, negative opinions and attitudes held by individuals may impact their resource use. For example, high levels of dissatisfaction and complex political dissonance with park managers or national governments may play a role in illegal sapphire mining in Madagascar (Walsh 2005). Alternatively, a lack of alternatives may drive some individuals to extract wood (for charcoal) from protected areas and to hunt animals (even when it is illegal) in some of the more rural areas of the country (Golden 2009). Finally, some extraction may also occur for economic reasons.
For example, the extraction of rosewood (Dalbergia sp.) timber is a lucrative trade (Innes 4 2010); in February 2015 The Guardian reported that 2-meter long logs of rosewood were selling for 125 USD (Caramel 2015). For comparison, 90% of the population lives on less than 2 USD per day (World Bank 2013).
Implications of natural resource use on conservation initiatives in Madagascar:
The extraction of natural resources in Madagascar has been labeled as unsustainable and could negatively impact biodiversity and tourism (Innes 2010, Golden et al. 2011, Walsh 2005). For example, only 9.9% of Madagascar’s primary vegetation remains (Myers et al. 2000), but extraction of wood for construction, timber, and for firewood continues (Innes 2010, Rakotondrasoa et al. 2012). In the past, the response to combat natural resource use has been to implement policies at the national and regional level that outlaw or curb use. Recently, however, several multimillion-dollar conservation initiatives have been proposed in addition to ongoing enforcement strategies (Schwitzer et al. 2014).
Undertaking conservation programming in Madagascar – where natural resources are often used by local communities – is not easy. For example, the Ranomafana Park in
Madagascar – the third-most visited national park in Madagascar (Christie and Crompton
2003) – excludes all natural resource use within the park boundaries. Nonetheless, despite extensive efforts at community development and outreach that exceed those near most other remaining natural areas in Madagasar, the park employs only 100 people. While this is a significant number, it is only a fraction of the area’s 27,000 residents. Further, while the park has provided infrastructural improvements to nearly a dozen villages, most of the 160 area villages have not seen such improvements (Scales 2014). In addition, this 5 park – which was featured in the Morgan Freeman narrated 2014 Island of Lemurs:
Madagascar film and is regularly highlighted on international television programs (most recently in season five of Anthony Bourdain’s Parts Unknown on CNN) – is used by the
Malagasy government and tourism industries to advertise Madagascar’s biodiversity and tourism appeal. On one hand, the park has been successful at conserving biodiversity and increasing levels of ecotourism at the national level and has provided benefits to some local people. On the other hand, it has not adequately addressed the needs of most of the local population; in this case, the costs of the park are felt at the household level through restrictions on land and resource use, but the benefits are realized at the national level
(through increased tourism and research income), by development projects at a village- level, and by select individuals locally (Scales 2014). As such, despite its status as one of the flagship efforts to combine conservation and development in Madagascar, the
Ranomafana National Park highlights the difficulties of expecting a national park to generate income capable of adequately replacing lost income from natural resource use
(Scales 2014).
It is likely that natural resource use in Madagascar will continue. If this is the case, the cost to local communities to procure natural resources legally will not be insignificant (Emerton 1999). In addition, individuals living in areas where natural resource use is outlawed will likely not possess the skills required to benefit from alternative livelihoods activities that would allow them to gain meaningful income from projects that do not require the illegal extraction of natural resources (Walsh 2005).
Therefore, to account for this ongoing use, conservation programs could: 1) increase the effectiveness of their alternate livelihoods programs by having a better understanding of 6 why natural resource use occurs; and 2) better target programs aimed at protecting biodiversity by increasing their understanding of how natural resource use is impacting ecosystem health.
Focal research areas:
Understanding natural resource use and how it impacts biodiversity in
Madagascar will inform conservation initiatives in-country and potentially across sub-
Saharan Africa more broadly. Therefore, the present series of studies aims to: 1) examine the effects of habitat degradation on plant-frugivore networks; 2), understand the live capture and extent of ownership of lemurs in Madagascar; 3) understand the micro- and macro-level drivers of wild meat consumption in Madagascar; 4) describe the capture, movement, and trade of wild meat in Madagascar; and 5) the impacts of habitat changes on the diets and vertical stratification of frugivorous bats. Broadly, these studies focus on forest use (deforestation and degradation) and defaunation (from hunting and live capture). These two types of natural resource use are considered to pose the greatest proximate threats to ecosystems in Madagascar. Only 10% (Myers et al. 2000) to 20%
(Schwitzer et al. 2014) of Madagascar’s primary vegetation remains while hunting continues to be cited as a threat to many of Madagascar’s mammals (e.g. Golden 2009,
Schwitzer et al. 2014).
For the first study – discussed in Chapter Two – our objectives were to understand the effects of habitat degradation on (1) community structure, (2) network structure, and
(3) seed dispersal services. We focused on fruit-bearing trees and frugivores (two lemur and five bird species) across a three-point gradient of habitat degradation in a tropical dry 7 forest in the Ankarana National Park, northern Madagascar. Using mutualistic networks – which model pairwise mutualistic interactions among suites of interacting partners – and a novel measure of effectiveness, this study aimed to clarify how habitat degradation affects mutualisms and seed dispersal. Data on fruit consumption by frugivores were collected over 592 hours of observations at 13 fruiting tree species. This is one of the first studies to examine the impacts of habitat degradation on mutualistic networks in areas where forest-based frugivores are limited in number and cannot be functionally replaced by non-forest generalists. In addition, the study provides a valuable case study for understanding how detailed measures of interaction strength can be used to understand ecosystem health.
For the second study, detailed in Chapter Three, our objectives were to provide the first quantitative estimates of the prevalence, spatial extent, correlates and timing of lemur ownership, procurement methods, within-country movements, and numbers and duration of ownership. Using semi-structured interviews (1,093 households and 61 transporters) across 17 study sites, this study was the first to quantify the ownership of pet lemurs in Madagascar. To our knowledge it was also the first to quantify the prevalence of primate ownership in a country in which they are endemic. As such, this study advanced our understanding of the threats that lemurs face in Madagascar from captivity and the threats that primates face more broadly in tropical, developing countries from the pet trade.
Chapter Four discussed the third study, where we investigated the role of wild meat in food security in Madagascar, a country where wild meat consumption is poorly understood in urban areas and at regional scales. Using semi-structured interviews (n = 8 1339 heads-of-households, 21 towns), we aimed to: 1) quantify the amount and purpose of; 2) understand the drivers behind; and, 3) examine recent changes in wild meat consumption in Madagascar. This study was important as – to our knowledge – no study had simultaneously examined the consumption of wild meat in urban and rural areas of
Madagascar. In addition, this study’s large sampling effort – 21 towns along a 1100 kilometer transect – allowed for a novel look at how wild meat consumption changed by ethnicity and region. By understanding the drivers of wild meat consumption in
Madagascar, this study advanced our understanding of why wild meat is consumed, how this varies between urban and rural areas, whether it is impacted by food security and ethnicity, and how policy and management initiatives can account for these realities.
In Chapter Five, we discuss our fourth study, which aimed to improve our understanding of the wild meat trade in Madagascar. Specifically, our objectives were to:
(1) quantify the volume of consumption, transport, and sale for different animal groups, compared to domestic meat; (2) describe the methods of capture and hunting for different animal groups; (3) analyze the patterns of movement of wild meat from the capture location to the final consumer, compared to domestic meat; and (4) examine how the prices of wild meat change depending on the venue through which the consumer purchases it. This study is the first to examine the wild meat commodity chain in
Madagascar and incorporates aspects of both the formal and informal trade. As such it advances our understanding of how different wild animals are hunted, moved, traded, and consumed across large areas of the country.
Finally, in Chapter Six for our fifth study, we used stable isotope analysis to examine how foraging by three fruit bat species in Madagascar, Pteropus rufus, Eidolon 9 dupreanum, and Rousettus madagascariensis, are impacted by habitat change. This study was novel in that we used stable isotopes – which provide new avenues to advance understanding of shifts in bat foraging resulting from habitat change (Dammhahn and
Goodman 2014) – to analyze diet over a relatively large geographic range. In contrast, many existing studies on bat diets tend to investigate foraging under artificial circumstances, over short time frames, or across small geographic scales. Given their status as important seed dispersers and pollinators (Baum 1995; Bollen and Van Elsacker
2002) and given that all three focal species have declining populations (IUCN 2013), this study was an important step in understanding how habitat loss is impacting these species. 10 CHAPTER 2
THE EFFECTS OF HABITAT DEGRADATION ON PLANT-FRUGIVORE
NETWORKS
Abstract:
Preserving areas of high biodiversity involves protecting species critical in forest regeneration efforts, including fruit eating animals (frugivores) who function as seed dispersers for fruiting plants. The mutualisms between frugivores and plants, can be impacted by habitat degradation. The use of mutualistic networks - which model pairwise mutualistic interactions among suites of interacting partners – and can clarify how habitat modification is affecting species interactions. In this study, we examined how plant- frugivore networks change along a gradient of habitat degradation in sub-tropical forests.
Our goal was to understand how a plant-frugivore network could withstand degradation in a region where forest-based frugivores were limited in number and could not be functionally replaced by non-forest generalists. Our objectives were to understand the effects of habitat degradation on (1) community structure, (2) network structure, and (3) seed dispersal services. We focused on fruit-bearing trees and frugivores (two lemur and five bird species) across a three-point gradient of habitat degradation in a tropical dry forest in Madagascar. Data on fruit consumption by frugivores were collected over 592 hours of observations at 13 fruiting tree species. We found that as habitat became more degraded: (1) the community structure of both frugivores and fruiting tree communities changed; (2) the mutualistic network structure became less complex and less connected;
(3) the interaction strengths of pair-wise interactions changed and the asymmetries of 11 these interactions shifted; and (4) seed dispersal decreased by 91% in the secondary forest, compared to the primary forest. In addition, we show that frugivores: (1) sometimes stopped eating fruit in the degraded forest, even when they had consumed it in other forests; and (2) appeared to avoid some fruiting tree species while showing preference for others. The mutualistic network studied in this paper appeared sensitive to anthropogenic disturbance and a novel measure of effectiveness helped quantify these changes.
Introduction:
Human-modified forests are expanding rapidly worldwide (Wright 2005) and 350 million hectares of forest have been removed in the tropics while an additional 500 million hectares have been degraded (Lamb et al. 2005). Areas of frequent or intense anthropogenic habitat modification and areas of high endemic biodiversity often overlap
(Myers et al. 2000, Brooks et al. 2002). Over 40% of the globe’s vascular plants and 35% of terrestrial vertebrates are endemic to just 25 biodiversity hotspots – defined as areas with at least 0.5% of the world’s flora (Brooks et al. 2002). However, these hotspots have already lost at least 70% of their habitat cover (Myers et al. 2000). Therefore, preserving these areas – many of which are highly degraded or surrounded by degraded habitat – involves protecting species critical in forest regeneration efforts, including fruit eating animals (frugivores) who function as seed dispersers.
The effects of habitat degradation on biodiversity and ecosystem function are evident in the changes to species composition and densities that occur following anthropogenic habitat modification (e.g. Tabarelli et al. 1999, Barlow et al. 2007, Gray et 12 al. 2007). These shifts in species richness and abundance can disrupt mutualisms, many of which are critical to an ecosystem’s larger community structure and function (Hawkins et al. 1990). A mutualism is a mutually beneficial interaction between two species
(Bronstein 1994); for example, fruit-bearing plants in plant-frugivore mutualistic relationships can account for 66% of seed dispersal in neotropical forests (reviewed by
Wunderle 1997). The critical importance of the plant-frugivore relationship is further illustrated by the reproductive failure observed to occur in fruit-bearing trees after the complete loss of frugivores from forest communities (Cordiero and Howe 2003). Plant- frugivore networks are key to the regeneration of sub-tropical forests (Chama et al.
2013), with large animals described as especially important in these regions (Wright et al.
2007).
One way to quantify the importance of plant-frugivore networks is through the use of mutualistic networks - which model pairwise mutualistic interactions among suites of interacting partners – and can clarify how habitat modification is affecting species interactions (Bascompte and Jordano 2007, Burgos et al. 2007, Tyliankis et al. 2010,
Wunderle 1997). Mutualistic networks, like predator-prey interactions, allow scientists to map which species interact and to what degree, in order to understand how species interactions are fractured by habitat degradation (Bascompte and Jordano 2007,
Tylianakis et al. 2010, Wunderle 1997). Understanding mutualistic networks can help conservation managers better plan regeneration programs and predict where ecosystems are most likely to experience secondary extinctions or species invasions (Ings et al 2009,
Tylianakis et al. 2010, Stachowicz 2001, Vazquez et al. 2009). They could be especially useful as conservation initiatives are now interested in conserving networks of species 13 across a variety of different habitat types (Evans et al. 2013). Nevertheless, studies on mutualistic networks are only recently expanding, and the application of the networks to understanding habitat degradation is new (Evans et al. 2013).
Several studies have recently addressed how plant-frugivore mutualistic networks change when forests become more degraded or resource-limited, though most studies focus on comparing forest interiors with habitat edges. These studies have found that generalist frugivores are not as affected by logging as specialist frugivores, are more abundant at forest edges than in forest interiors (Albrecht et al. 2012), and mutualistic networks at forest edges are more connected and more robust against species extinctions, compared to less disturbed interior forests (Menke et al. 2012). Likewise, a study on the plant-frugivore networks in anthropogenic landscapes suggested that while species richness changed with modification, the strength of the mutualistic network does not
(Plein et al. 2013). Though informative and promising in their conclusions that ecosystem functions and network structure are being retained despite habitat degradation, these studies tend to focus on overall network strength and functional diversity in regions where there are a large number of frugivores and a high number of non-specialist forest visitors (i.e. 39 frugivorous birds in Plein et al. 2013; 53 avian species in Chama et al.
2013; 51 bird species in Menke et al. 2012). Therefore, this may limit how applicable these results are to regions where frugivores and their ecosystem services cannot be functionally replaced.
In addition, studies on mutualisms have been mostly limited to pairwise observations (observations between two species), and 63% of 4500 articles surveyed by
Bronstein (1994) focused on only one-side of a mutualistic relationship. This is also true 14 for mutualistic network research where many published studies have considered the impact of habitat degradation only on one of the two interacting guilds (though Yang et al. 2013 examines two guilds and their changing interactions due to resource abundance).
To our knowledge, only one study has explicitly examined both plants and frugivores in modified landscapes (Chama et al. 2013). This study found that plants, in general, were more specialized than frugivorous birds but that overall mutualistic network structure was robust against changes in fruit abundance, plant species richness, or canopy cover
(Chama et al. 2013). The authors concluded that, in the tropics, frugivores tend to be generalists with high network redundancy, protecting them against degradation (Chama et al. 2013). However, given that this study focused on a large number of avian frugivores and did not directly consider, or measure, habitat characteristics beyond canopy cover, it is unclear how applicable these results are to gradients of habitat degradation and in areas where non-avian frugivores are present.
In addition, many studies do not simultaneously, empirically examine frugivores of multiple taxa, in particular most research focuses exclusively on birds (i.e. Chama et al. 2013, Plein et al. 2013), and in a review of network studies, only 40% of papers included frugivores with a body mass of larger than 1 kg (Vidal et al. 2013). Of the few studies that have included multiple taxa and studied networks across more than one forest type, one included both small mammals (3 species) and birds (29 species; Albrecht et al.
2013) while the other included primates (3 species) and birds (51 species; Menke et al.
2012). Neither of these studies explicitly compared between the different groups of frugivores and focused instead on the frugivores’ classifications in the literature as specialists or generalists, though Menke et al. (2012) did note that large-bodied 15 frugivores (3 species of monkeys) were less likely to provide seed dispersal benefits at forest edges than in forest interiors (Menke et al. 2012). In addition, it was concluded that, when primate species were lacking, birds could not functionally replace monkeys in dispersing large seeds (Menke et al. 2012). However is still unclear how important large- bodied frugivores are in seed-dispersal networks as they are typically not studied in this context (Vidal et al. 2013).
Finally, most network studies do not explicitly consider frugivore preference when examining mutualistic networks and most appear to make the assumption that whenever a frugivore has the ability to realize a mutualism, it will do so (Cantor et al.
2013). However, it is known that individuals can vary in their mutualistic interactions, independently of resource availability and other explanatory factors (Cantor et al. 2013).
Therefore network studies that account for individual variation or other avoidance and/or preference-based behavior may be more ecologically relevant (Cantor et al. 2013).
In this study, we examined how plant-frugivore networks change along a gradient of habitat degradation in a tropical forest. Our goal was to understand how a plant- frugivore network could withstand degradation in a region where forest-based frugivores were limited in number and could not be functionally replaced by non-forest generalists.
We examined a network that included both birds and primates, studied both guilds in the network, and accounted for frugivore avoidance and/or preference for fruit resources. We focus on frugivorous primates and birds due to the paucity of data relating to plant-seed disperser mutualistic networks, as compared to the plant-pollinator literature (Cordiero and Howe 2003). To undertake this work, a novel, data-inclusive measure of the effectiveness of the interaction (based on Sewall, Freestone, and Holland, unpublished) 16 was used to convey both the frequency of a mutualistic relationship and the relative importance of that relationship to mutualist partners. The development of such ecologically-relevant metrics has been recommended, especially when attempting to examine the importance of large-bodied frugivores quantitatively (Vidal et al. 2013).
Our objectives were to understand the effects of habitat degradation on (1) community structure, (2) network structure, and (3) seed dispersal services. We focused on fruit-bearing trees and frugivores (two lemur and five bird species) across a three- point gradient of habitat degradation in a tropical dry forest in Madagascar.
For objective one, and based on information regarding regional frugivore densities and habitat degradation (Sewall et al. 2013, Sewall and Andriamanarina unpublished), we hypothesized that (1A) compared to primary forests, more degraded forests would retain similar species richness of both fruiting trees and frugivores, but
(1B) the composition and density of mutualist species in each guild would shift significantly across the gradient. We also expected (1C) the primary forest to contain larger fruiting trees whereas more degraded forests would have smaller fruiting trees.
Finally, we hypothesized that (1D) more degraded forests would have lower fruit availability than primary forests.
For objective two, we hypothesized that increasing habitat degradation would be correlated with changes in the presence and strength of particular pairwise mutualistic interactions and in aggregate measures of mutualistic interactions at a whole-community scale. Specifically, we hypothesized that (2A) in more degraded habitats, mutualist species will compensate for the changes in available partners by (1) maintaining or strengthening pairwise mutualistic interactions with those partner species also found in 17 less degraded forests, (2) establishing new pairwise mutualistic interactions with partners with which they did not interact in less degraded forests, and (3) decreasing the total number of mutualist partners with which they interact. Because of these changes to pairwise interactions, we further hypothesized that mutualist community structure as a whole would become (2B) less complex, (2C) less connected, and (2D) more loosely organized. Finally, we predicted that frugivores would exhibit preference for some tree species, by (2E) avoiding some fruiting trees even when they are capable of consuming the fruit.
For objective three, we examined how small- and large-bodied frugivores differ in their provision of seed dispersal services, and hypothesize that (3A) large bodied frugivores would provide more seed dispersal services in primary forests while small bodied frugivores would be important dispersers in modified forests.
Methods:
Study site:
Research was conducted in Madagascar, a global biodiversity hotspot (Myers et al. 2000) in the semi-evergreen forests within and surrounding the Ankarana National
Park (Figure S1, Appendix A). This park contains a diversity of plant and animal taxa
(Cardiff and Befourouack 2003) and high primate densities (Hawkins et al. 1990). Low canopy height facilitates detection of canopy-feeding frugivores (Sewall et al. 2013).
Research was based at the main park entrance and conducted during the dry season
(early-June to early-August 2012), as food availability decreases and fruit-bearing trees become more important to frugivore survival (Sewall et al. 2013). 18
Forest types and transects:
Primary and secondary forests were identified based on previous research (Fowler et al. 1989, Hawkins et al. 1990, Sewall and Andriamanarina unpublished), and degraded forests were secondary forests that had undergone extensive recent anthropogenic disturbance (Gudiel, Nieves, Reuter, and Sewall, unpublished). Primary forest fragments in this study have been protected since 1956 from most degradation through eco-tourism and the park’s protected status (Cardiff and Befourouack 2003). The secondary forest, located both inside and outside the park, has regenerated following large-scale logging, grazing, and fires more than 60 years ago (Sewall and Andriamanarina unpublished), though it is now under some modest anthropogenic pressure. The degraded forest, located closest to area human settlements, undergoes continuous anthropogenic use and has both the highest mean densities of disturbance, and the highest within-forest variability in disturbance of all three forest types (Gudiel, Nieves, Reuter, and Sewall, unpublished).
Classification of forest types and transects:
To systematically collect data on the characteristics of fruiting trees, complete distance sampling for diurnal frugivores, and to identify trees for inclusion in the mutualistic network, transects were marked in all forest types (2.27 km in primary forest, n = 32; 1.5 km in secondary forest, n = 12; 2.27 km in degraded forest, n = 26). To avoid oversampling of habitats adjacent to trails or forest edges, all transects were placed perpendicular to these features and traversed interior habitat. Overall, transect length averaged 87 ± 15 m (mean ± 95% CI), and did not differ significantly by forest type 19 (Kruskal-Wallis Rank Sums Test, Chi square: 2.5572, p = 0.2784). The secondary forest was the least abundant forest type.
Frugivores:
The entire diurnal frugivore guild in the park was studied, including five frugivorous birds (Coracopsis vasa, Coracopsis nigra, Treron australis, Saroglossa aurata, and Hypsipetes madagascariensis) and two IUCN Endangered diurnal lemurs
(Eulemur coronatus and E. sanfordi, Andriaholinirina et al. 2014a,b). Common names are listed in Table S1 (Appendix A).
Frugivore densities were calculated for each forest type following established distance sampling protocols (Buckland et al. 2010; Thomas et al. 2010) along marked transects using the variable-width method. Sampling effort was equal across forest types
(12 km each), recording all visual encounters. Densities were calculated using Distance software (Thomas et al. 2010), using detection functions modeled for each frugivore species in each forest type.
Habitat sampling:
Fruiting trees were experimentably tractable, independent units. Species were not known a priori, and were identified through habitat sampling conducted twice (May and
June 2012) across 0.0227 km2 in both the primary and the degraded forests, and 0.015 km2 in the secondary forest. Thirty-three fruiting trees were identified across all forest types (Table S2, Appendix A). All were seasonally fruiting species except for two Ficus species that fruit year-round (Janzen 1979, Shanahan et al. 2001). Species identity; 20 diameter at breast height (DBH) greater than 10 cm; height; fruit crop size; fruit biomass; seed count data; canopy visibility; and location were recorded (Table S2, Appendix A).
Fruit crop size was measured by either counting every fruit, or estimated (following
Chapman et al. 1992). Only trees with ripe fruit were included in the study (fruit color, fallen fruit, and local botanical knowledge were used to determine ripeness). Most fruit biomass and seed count estimates were calculated from fruit samples collected at the study site in 2006/7, 2012, and 2013 (Table S3, Appendix A).
Due to time constraints, 13 species were chosen for inclusion (Table S1,
Appendix A). One of these 13 species (Burseraceae commiphora) finished fruiting before sufficient data could be collected and was excluded from further analysis. Fruiting tree species were selected for inclusion if they were found across more than one forest type and had high fruit biomass (relative to other species) in at least one forest type. Total fruit availability from populations of the tree species sampled under this selection method accounted for 97.94%, 99.60%, 94.19% of all fruit present along transects in the primary, secondary, and degraded forests.
Focal observations:
After tree species were selected for inclusion in the mutualistic network, individual trees with high species-specific fruit crop size and canopy visibility were selected for focal observation. Single observer focal observations at fruiting trees took place over three-hour intervals at peak, diurnal, frugivore foraging times. Ficus grevei, with its high fruit crop sizes (>50,000 fruit) and large canopies, necessitated two observers. During focal observations at fruiting trees, we recorded: 1) length and 21 frequency of visitation of all frugivores entering the canopy; 2) frequency, amount, and timing of fruit consumption by representatives of each frugivore species in the focal tree;
3) and frugivore removal of fruit from the focal tree canopy horizontally across the canopy edge. Sampling effort was split equally across all observable fruiting tree individuals in a species, with a minimum of 20 hours per tree species/forest type (Table
S3, Appendix A), so as to decrease sampling bias caused by an individual tree’s unique characteristics. At all focal trees except for Ficus grevei, frugivore preference/avoidance of fruiting tree species was also quantified indirectly by recording all visual observations of frugivores passing within a 15 meter radius of the fruiting tree without entering it during the course of focal observations.
Analysis of frugivore and tree characteristics:
A Bray-Curtis Similarity Index (PRIMER statistical software, Anderson 2001) was used to analyze changes in the composition and density in fruiting trees and frugivore populations across forest types (transects as replicates; forest type as a covariate). This is a non-parametric multivariate analysis of variance, where the test statistic is calculated on the basis of a dissimilarity matrix derived both composition and abundance data for the community at each transect (Anderson 2001). Due to the non- conformity of data to the assumptions of normality in ANOVA, we used non-parametric
Kruskal Wallis Rank Sums Tests and Steel-Dwass Multiple Comparisons post hoc tests
(SAS Institute 2015) to determine differences in habitat characteristics (DBH, canopy height, and fruit crop size).
22 Binary network analysis:
Binary mutualistic network data were analyzed using one matrix per forest type, based on the presence of fruit consumption by a frugivore at a tree species. Matrices were analyzed for connectance, nestedness, and generality (also known as species degree,
Blüthgen et al. 2009). Connectance is the proportion of all possible mutualistic interactions that are observed to occur in a network (Dunne et al. 2002). Nestedness is a measure of the structure of mutualistic network, where specialists interact with a subset of species, which in turn interact with generalists (Bascompte and Jordano 2007,
Blüthgen et al. 2009). Nestedness analyses were completed via randomizations with
Monte Carlo probability over 10,000 runs using Nestedness Calculator software
(following Atmar and Patterson 1995).
Weighted network analysis:
Weighted network analysis involves matrices that include measures of the strength of a mutualism, also known as effectiveness (Sewall, Freestone, and Holland, unpublished). Recent plant-frugivore papers have examined the strength of interactions using interaction frequencies of relatively coarse data, such as the number of visits by a seed disperser to a plant (Albrecht et al. 2012) and the number of fruit-eating visitors per plant species (Chama et al. 2013, Menke et al. 2012, Plein et al. 2013), yet these approaches exclude important components of the ecological interaction (Sewall,
Freestone, and Holland unpublished). Thus, for this study, novel and more ecologically meaningful definitions of effectiveness, based on methods developed by Sewall,
Freestone, and Holland (unpublished), were used as proxies for measuring the nutritional 23 benefits that frugivores gain from trees, and the seed dispersal benefits that trees gain from frugivores. Specifically, the nutritional benefits received (per day, per hectare) by a frugivore species from a tree species, were defined as: average biomass consumed per hour at one tree multiplied by tree density. The seed dispersal benefits received (per day, per hectare) by a tree species from a frugivore species, were defined as: average number of seeds removed per hour at one tree multiplied by tree density. Calculations only considered frugivores that consumed or removed fruit from a tree canopy. In addition, total biomass accounted for differences in fruit size across species (Table S2, Appendix
A). Weighted network visualizations were created using R v. 3.0.2 (R Development Core
Team 2013) with the bipartite package (Dormann et al. 2008).
Weighted matrices were scaled from 0-1, relative to the strongest interaction of its type found in any forest type (Table S6, Appendix A), with a higher score indicating a stronger interaction. Scores were placed into one of four classes, from least to strongest interaction (Table S6, Appendix A). Interaction asymmetries (Blüthgen et al. 2009) were then estimated at a coarse level, by counting the number of mutualisms where the two measures of effectiveness fell into two different classes.
For objective three, estimates of seed dispersal were calculated using the measure of effectiveness, described above, for the seed dispersal benefits provided by frugivores to trees.
24 Results:
Effects of habitat degradation on community structure:
There was mixed support for hypothesis (1A). Although the species richness of frugivores (n = 7) did not change in different habitat types, the species richness of fruiting trees did (Table S2, Appendix A), and was lower in modified forests. During the two-month study period, primary forests had 21 fruiting tree species, while secondary and degraded forest had 9 and 12 species, respectively.
The density of both frugivores (Table S4) and trees (Table S2) are listed in
Appendix A. In accordance with hypothesis (1B), on a measure of similarity combining composition and density, forest types differed significantly in their frugivore communities (Bray-Curtis Similarity PERMANOVA, p = 0.0041; Table S4, Appendix
A), with the degraded forest being significantly different than the primary (p = 0.002) and the secondary forests (PERMANOVA pair-wise test, p = 0.025). Primary and secondary forests were not significantly different in their frugivore composition and density
(PERMANOVA pair-wise test, p = 0.825). Likewise, the composition and density of fruiting trees differed significantly across the three forest types, even when outliers (three of the sixty-nine replicates) were excluded (Bray-Curtis Similarity PERMANOVA, p =
0.0001), and all forests were significantly different from each other (P < 0.05, pair-wise
PERMANOVA).
There was mixed support for hypothesis (1C). Fruiting trees in the primary forest were larger in height than trees in both modified forests, but only wider in diameter than those in the degraded forest. The average (± 95% CI) Diameter at Breast Height (DBH) of fruiting trees above 10 cm in the primary, secondary, and degraded forests were 36.77 25 ± 18.74 cm, 46.16 ± 65.18 cm, and 24.17 ± 1.70 cm, respectively, and were significantly different (Kruskal-Wallis Rank Sums Test, Chi square = 26.93, DF = 2, p < 0.0001). The average DBH of fruiting trees in the secondary forest was significantly different from both the degraded (p = 0.0004) and the primary forest (p < 0.0001), but the degraded and primary forests were not significantly different from each other (Steel-Dwass nonparametric comparison, p = 0.1444). The fruiting tree canopy height was significantly different between the three forest types (Kruskal-Wallis Rank Sums Test, Chi Square =
228.8655, p < 0.0001). The mean canopy heights were 22.93 ± 1.38 m, 13.05 ± 1.91 m, and 8.51 ± 0.54 m, in the primary, secondary, and degraded forests, respectively. All three forest types were significantly different from each other (Steel-Dwaas nonparametric comparison, p < 0.0001).
There was mixed support for hypothesis (1D). The total number of fruit available per hectare over the two-month period was higher in the secondary (94,664 fruit/hectare) and primary (93,211 fruit/hectare) forests than in the degraded forest (65,076 fruit/hectare). However, on a per-tree basis, the fruit crop size was highest in the secondary forest. Average (± 95% CI) fruit crop size for an individual tree was 1048 ±
1242 in the primary forest, 2755 ± 2798 in the secondary forest, and 525 ± 132 in the degraded forest; these were significantly different (Kruskal-Wallis Rank Sums Test, Chi square = 96.92, p < 0.0001). Trees in the primary forest were significantly higher in mean fruit crop size as compared to the degraded forest (Steel-Dwass nonparametric comparisons, p < 0.0001) and lower than the secondary forests (p < 0.0001). However, the degraded and secondary forests were not different from each other (Steel-Dwass nonparametric comparisons, p = 0.52). 26
Figure 2.1: Change in generality (A) and seed dispersal services (B), by frugivore species, across multiple forest types. Generality represents the proportion of mutualistic partners that a frugivore species has, relative to the whole network. The seed dispersal services are the proportion all seeds that a species disperses in a forest type. Birds a shaded in blue and lemurs are shaded in red.
A) 0.8 C. nigra 0.7 C. vasa 0.6 H. madagascariensis 0.5 S. aurata 0.4 T. australis Generality 0.3 E. coronatus 0.2 E. sanfordi 0.1 0 Primary Secondary Degraded Forest type
27 B) 0.9 C. nigra 0.8 0.7 C. vasa 0.6 H. madagascariensis 0.5 S. aurata 0.4 T. australis 0.3 E. coronatus 0.2 E. sanfordi Propor on of seeds dispersed 0.1 0 Primary Secondary Degraded Forest type
28 Effects of habitat degradation on network structure:
In order to examine both pair-wise mutualistic interactions and mutualistic network properties, we conducted 592.2 hours of frugivore observations at 12 fruiting tree species. Observation effort by tree species is listed in the online appendix (Table S3,
Appendix A), as well as the length of time that frugivores were present in focal fruiting trees (Table S1, Appendix A), and the total number of frugivores observed visiting at fruiting trees (Table S5, Appendix A).
In accordance with hypothesis (2A), frugivores varied in how they compensated for changing mutualistic partners in degraded habitats (Figure 2.1A). For example, H. madagascariensis slightly increased its generality – the proportion of potential partners with which it interacted – by maintaining and establishing new pairwise mutualistic interactions (relative to the primary forest; Figure 2.2) the more degraded the habitat became (Figure 2.1A). In contrast, E. coronatus decreased its generality markedly from the primary to the degraded forest (Figure 2.1A). Some primate and bird species ceased almost all participation in the mutualistic network in the most degraded habitat type.
Fruiting trees and frugivores did not differ significantly in their generality (Wilcoxin
Test, Chisquare = 0.5528, p = 0.4572).
The effectiveness – or strength - of mutualistic interactions changed across forest types (Table S6, Appendix A), with the strongest interactions found in the primary and degraded forest, while the secondary forest was characterized by many weak interactions
(Figure 2.2). The primary forest network included mutualisms with a wide range of interaction strengths, while the degraded forest network was characterized by fewer, stronger interactions (Figure 2.2, Table S6, Appendix A). Figure 2.2: Effectiveness of fruiting trees for frugivores in (A) primary forest, (B) secondary forest, and (C) degraded forest. Frugivore species are on the top row, listed from left to right in order of increasing adult mean body size (from Sewall et al. 2013).
Labels for birds are in blue, and lemurs are in red. Trees are on bottom row, listed from left to right in order of increasing mean fruit crop biomass per individual tree (calculated from Tables S2 and S3, Appendix A). Labels for trees are green. Interactions are represented by connections between tree and frugivore species, and represent benefits from fruiting trees to frugivores at a species level (fruit biomass of a tree species consumed per hour by a frugivore species) across the study area for each forest type.
Width of an interaction represents greater amount of benefits provided to a frugivore.
Width of the bar for a tree species represents total amount of benefits provided by that tree species to all frugivore partners. Width of the bar for a frugivore species represents total amount of benefits obtained by that frugivore from all tree partners. The sum of the widths of all bars represent the relative amount of total benefits provided or obtained in a forest type; this was highest in degraded forest and lowest in secondary forest. Shading of frugivore bars represent the density of the species within a particular forest type, with white representing the lowest frugivore density and black representing the highest frugivore density overall in all forest types (highest frugivore density was of H. madagascariensis in degraded forest). Shading of tree bars represent the density of fruiting trees of a species within a particular forest type, with white representing the lowest fruiting tree density and black representing the highest fruiting tree density overall in all forest types (highest fruiting tree density was of Grewia sp. in degraded forest).
Abbreviations for species names are the first two letters of the genus and species.
29 30 Frugivores: SAAU = Saroglossa aurata, HYMA = Hypsipetes madagascariensis, TRAU
= Treron australis, CONI = Coracopsis nigra, COVA = Coracopsis vasa, EUCO =
Eulemur coronatus, EUSA = Eulemur sanfordi. Trees: GRSP = Grewia sp., UNAN =
Unknown (Anosaly), FILU = Ficus lutea, MASP = Mantalania sp., CRSP = Croton sp.,
BAFL = Baudouinia fluggeiformis, PISP = Pittosporum sp., STMA = Strychnos madagascariensis, POSP = Poupartia sp., FIGR = Ficus grevei
31
Mutualisms were often unequal in the benefits exchanged, relative to the strongest interactions found in the matrices (Table S6, Appendix A). Many partnerships (11 out of
31) were classified as unequal, and fell in different classes of interaction strength (Table
S6, Appendix A). Five out of 11 unequal partnerships benefited the tree species relatively more than the frugivore species. In the primary forest, some pair-wise interactions benefited trees, some were considered equal in their benefits, and some benefited frugivores. In the secondary forest, there were no unequal mutualisms while in the degraded forest, all mutualisms that were unequal benefited frugivores more than trees
(Figure 2.3).
32 Figure 2.3: Unequal pair-wise mutualisms by forest type. The percentage of pair-wise mutualisms in each forest type that: 1) benefit the fruiting tree partner more than the frugivore (green); 2) that benefit the frugivore more than the fruiting tree partner (blue); or 3) where the frugivore and the fruiting tree receive equal benefits from the mutualism
(red). Sample size are n = 16, n = 11, and n = 4 for the primary, secondary, and degraded forest, respectively.
100% 90% 80% 70% 60% Tree 50% Equal 40% 30% Frugivore
Percent of mutualisms 20% 10% 0% Primary Secondary Degraded Forest type In accordance with hypothesis (2B), the primary forest had the most complex mutualistic network with 14 interacting species (7 trees and 7 frugivores, Figure 2.2).
Modified forests had less complex mutualistic networks with fewer interacting species; the secondary forest had 11 interacting species (5 trees and 6 frugivores) and the degraded forest had 5 interacting species (2 trees and 3 frugivores). In addition, like in many other mutualistic networks, our networks had many specialists, some generalist, and few super-generalists (Figure 2.4). However, the degraded forest had fewer generalists and more specialists than the secondary or primary forest (Figure 2.4).
Figure 2.4: The proportion of species that have different levels of species degree (generality), by forest type. Species degree is the proportion of mutualistic partners a species has, relative to the whole network. A species with a low species degree interacts with few species and is more specialized than a species with a higher species degree; values closer to 0 are specialists and values closer to 1 are more generalist.
0.9 Primary forest 0.8 0.7 Secondary forest 0.6 Degraded forest 0.5 0.4 0.3
Propor on of all species 0.2 0.1 0 0 - 0.33 0.34 - 0.66 0.67 - 1.00 Species degree (generality)
33 34 The mutualistic network structure was less connected (hypothesis 2C) in degraded habitats (Figure 2.5A). However, the mutualistic network was not more loosely organized
(hypothesis 2D) as the forest became more degraded. The nestedness of mutualistic networks did not decrease as the habitat became more degraded (Figure 2.5B). The primary forest was marginally different than random (Nestedness Calculator, p = 0.0546, two-tailed, 10,000 runs) while the secondary and degraded forest were not different than random (Nestedness Calculator, p > 0.20, two-tailed, 10,000 runs).
In accordance with hypothesis (2E), frugivores exhibited preference for some tree species by avoiding non-preferred species. In the primary forest, an average of 1.21 ±
0.15 frugivores per hour passed within 15m of a focal fruiting tree, without entering the canopy. This increased to 1.53 ± 0.33 frugivores per hour in the secondary forest, but decreased to 0.88 ± 0.12 frugivores per hour in the degraded forest. Data calculated by tree species can be found in the online appendix (Table S7, Appendix A). Likewise, the likelihood that a visiting frugivore would consume fruit at a tree differed by forest type
(Pearson Chisquare, χ2 = 23.956, p < 0.0001) and decreased as the habitat became more degraded. In the primary forest 69.62% (165 of 237) visits by frugivores to fruiting trees resulted in fruit consumption. This decreased to 47.26% (95 of 201 visits) and 48.57%
(17 of 35) of visits in the secondary and degraded forests, respectively.
35 Figure 2.5: Connectance (the proportion of realized links in the mutualistic network, figure A) and nestedness (measure of network structure, figure B) by forest type. Network connectance and nestedness values tend to be lower in less robust mutualistic networks. Data calculated based on the presence/absence of mutualistic interactions resulting in fruit consumption. In the primary forest, the mutualistic network was marginally more nested than random (Nested Software, p = 0.0546, two-tailed, 10,000 runs). A) 0.35
0.3
0.25
0.2
0.15 Connectance 0.1
(propor on of realized links) 0.05
0 Primary Forest Secondary Forest Degraded Forest
B) 100 99 98 97 96 95
Nestedness 94 93 92 91 90 Primary Forest Secondary Forest Degraded Forest
36 Effects of habitat degradation on seed dispersal:
Our calculations estimate that during time period studied, frugivores dispersed
27,093 seeds per hour, per hectare in the primary forest during peak foraging times. This decreased to 2,447 seeds in the secondary forest and then again to 949 seeds in the degraded forest, or only 9% and 3.5% of the volume in the primary forest, respectively.
Contrary to hypothesis (3A), both small-bodied and large-bodied frugivores were important in seed dispersal in all forest types (Figure 2.1B) and there did not appear to be a shift as the habitat became degraded.
Discussion:
Effects of habitat degradation on community and network structure:
In accordance with past studies (e.g. Tabarelli et al. 1999, Barlow et al. 2007,
Gray et al. 2007), habitat degradation impacted the community structure of both frugivore and fruiting tree communities. In addition, fruiting trees in the primary forest were larger in height than modified forests and wider in their diameter than the degraded forest
(Table S2, Appendix A).
Similar to some studies (e.g. Albrecht et al. 2012) habitat degradation impacted network structure (Figure 2.2) though – unlike other studies (Menke et al. 2012, Plein et al. 2013) – we found that network structure was negatively impacted by habitat degradation. First, the primary forest had the most complex network with the highest connectance and the network was marginally more nested than random (Figure 2.5); other studies found the opposite (Menke et al. 2012). Second, we found the number of species that participated in the networks changed and decreased as the habitat became 37 more degraded. In contrast, a past study found that while species richness changed in modified habitats, the strength of the mutualistic network did not (Plein et al. 2013).
These discrepancies may be because, at our study site, there were far fewer frugivores (n
= 7) than in past studies (51 in Menke et al. 2012; 39 in Plein et al. 2013) and these could not be functionally replaced in more degraded habitats. Third, the effectiveness of the mutualistic interactions changed as the forest was degraded; the primary forest network included mutualisms with a wide range of interaction strengths, the secondary forest was characterized by several, weak interactions, and the degraded forest network was characterized by fewer, stronger interactions (Figure 2.2).
In the secondary forest, feeding rates dropped dramatically despite levels of fruit biomass availability similar to those found in the primary forest. Our results – which show large drops in fruit consumption in secondary forests despite overlapping species richness and high levels of fruit biomass, relative to the primary forest – are in contrast with past studies, which showed that fruit availability was an important determinant of network structure (Chama et al. 2013) and that frugivore visits in modified habitats increased with fruit abundances (Plein et al. 2013). This may be for several reasons: 1) while fruit biomass remained high, the species found in the secondary forest were generally found at lower densities than in the primary or degraded forest (Table S2,
Appendix A); 2) most frugivores still participated in the network, perhaps because their preferred fruiting tree species were still present (albeit at lower densities), but the primary forest was close enough (within 2 kilometers) that some frugivores could simply feed in the primary forest if needed; 3) frugivores changed their foraging behavior due to habitat degradation or other anthropogenic disturbance (the percent of visits to fruiting trees 38 resulting in consumption decreased from 69.62% to 47.26% and 48.57% in the primary, secondary, and degraded forests).
In the degraded forest, the network structure changed again, with most frugivores no longer participating in the network, despite the presence of some tree species which they were known to consume in the secondary and primary forest (Figure 2.2). The remaining strong interactions in the degraded forest (between Grewia sp. and three frugivores) may be because: 1) H. madagascariensis is a generalist frugivore that is found at its highest densities in the degraded forest; and because 2) T. australis and E. coronatus shift to eating Grewia sp. in the absence of Ficus grevei and other preferred fruits. This shift may occur in the degraded forest (as opposed to the secondary forest) because the degraded forest only had one fruiting tree species in common with the primary forest (Grewia sp.) during our two-month study; therefore, the frugivores in the degraded forest – which is not as close to the primary forest as the secondary forest (>2 kilometers) – may not have the ability to reach other fruiting trees as easily as those in the secondary forest.
Four frugivores (of the seven studied) were not observed foraging at fruiting trees in the degraded forest, despite the fact that they were observed there during the frugivore census (Table S4, Appendix A). Therefore, it is possible that these species are consuming fruit from other sources. This study, which focused on the most abundant fruiting tree species available during the dry season, accounted for most of the fruit available to frugivores during the research period from fruiting trees (>94% in all forest types).
However, fruit is also available during this time period from vines and bushes.
Anecdotally, we observed that several fruiting species of vines and bushes were present 39 during our study, though Lantana camara (a non-native plant) was the most abundant by far, and was found primarily in the degraded forest. We observed H. madagascariensis feeding and S. aurata visiting this species, and it has been reported to constitute up to
25% of the diets of both lemur species included in this study at a different site (Freed
2012), though we never observed any lemurs feeding or visiting it at Ankarana. It would be valuable to increase our understanding of whether this non-native plant is spreading or not, as it may become an increasingly important food source, at least for some frugivore species, as the areas around the Ankarana Park continue to be degraded.
In all three forest types, we often observed frugivores passing around and over fruiting trees without entering the canopy (>0.88 frugivores per hour across all forest types). Given the energy spent in foraging (Nagy et al. 1999), it is surprising that the frugivores did not stop to feed; especially lemurs that could easily have passed through the fruiting tree canopies instead of around them. It is unlikely that this behavior was caused by frugivores avoiding researchers, given steps taken to be unobtrusive and the lack of alarm calls by lemurs (Wilson et al. 1989) during visits or near passes. It is also unlikely this avoidance behavior was caused by fruit not being ripe since we made efforts to only observe at trees where fruit was clearly ripe (typically fruits had changed color, fruit pulp was soft to the touch, and fruit had fallen to the ground under the canopy). It is possible that some avoidance may have been because frugivores could not consume the fruit in question. Our research did not account for forbidden links, which may predominantly affect H. madagascariensis, S. aurata, and T. australis. For example, these three bird species likely do not have the capability to crack the hard shell of the B. fluggeformis fruit and would have difficulty handling large fruit from the S. 40 madagascariensis and Mantalania sp. trees (Table S3, Appendix A). However, this explanation does not explain why trees with small fruits that were consumed by frugivores in most instances, were avoided or observed and not consumed in some cases.
Therefore, it could be that frugivores prioritize feeding at trees or in areas with a high density of fruit biomass (such as in Ficus grevei trees in the primary and secondary forest or in patches with a higher density of Grewia sp.in the degraded forest), and only opportunistically feed at other fruiting trees in passing or when absolutely necessary. A preference by frugivorous lemurs in other areas for Ficus sp. has been noted before
(Freed 2012) and would not be surprising given that this genus has been described as the one of the most important for tropical frugivores worldwide (Shanahan et al. 2001). It is notable that during the study time period – which is the start of the dry season – there are actually a relatively high abundance of fruit (BJ Sewall, personal communication) so it is possible that frugivores were not experiencing food stress and therefore were able to avoid eating non-preferred fruit sources. Understanding frugivore preference for fruit- sources is critical to understanding why, or why not, frugivores feed at fruiting trees and how habitat degradation (which can decrease the densities of fruit biomass found in a given area) will negatively affect frugivores.
It is interesting that mutualisms were not always equal in the benefits exchanged between frugivores and fruiting trees. Asymmetries in interaction strength may promote community stability (Bascompte et al. 2003, 2006), which may explain why half of the pair-wise interactions in the primary forest were considered asymmetrical or unequal in the benefits exchanged. However, the degraded forest also had several asymmetrical pair- wise mutualisms, all of which benefited frugivores more than trees. These data, however, 41 should be interpreted with caution as the measures of effectiveness do not account for all benefits that a frugivore might gain from a fruiting tree and visa versa. For example, we acknowledge the limitations of using the number of seeds consumed or removed from the canopy as a metric for the seed dispersal benefits that frugivores provide fruiting trees; the different frugivores in our mutualistic network have been termed seed dispersers (e.g.
H. madagascariensis), seed predators (e.g. C. vasa and C. nigra), and may often simply remove fruit from the tree but drop it just below the tree canopy (e.g. Eulemur sp, Bollen et al. 2004; Eulemur coronatus; Wilson et al. 1989). However, the information needed to provide a finer estimate of effectiveness (e.g. data regarding frugivore seed processing, post-gut passage germination rates, or even the proportion of seeds in a fruit that are actually viable) is not available for our study site. Nevertheless, it would be valuable to incorporate those data into the metric for effectiveness, and examine how these data increase our understanding how a mutualistic network changes as a result of habitat degradation. Despite these methodological caveats, this study is relatively easy to replicate for future studies on primates and other vertebrate frugivores (since it is based on observational data alone), and will be especially useful for rapidly gaining insight into mutualistic interactions in ecosystems where we have relatively little existing natural history knowledge. The resulting methodology could potentially be applied to a variety of habitats, both terrestrial and marine, especially when the relationships involve mutualisms between sessile and mobile organisms (reviewed by Vazquez et al. 2009) or foundation species (Stachowicz 2001).
42 Effects of habitat degradation on seed dispersal and conservation implications:
Our estimates of seed dispersal during the study time period highlight the impact that habitat degradation has seed dispersal; the number of seed dispersed decreased 91% between the primary and the secondary forest and a further 61% between secondary and degraded forests. These data bring to mind the ongoing debate about the trade offs between protecting large areas of secondary forest versus small areas of primary forest
(e.g. Barlow et al. 2007, Chazdon et al. 2009); it has been argued that in countries where the amount of remnant primary forest is low, the conservation value of secondary forests is high (Chazdon et al. 2009). On one hand, our data show that in some cases, secondary and primary forest were more alike than secondary and degraded forests were; for example, unlike the degraded forests, primary and secondary forests had similar numbers of fruits availabile per hectare of habitat. However, secondary forests were different in their habitat structure and fruiting tree species composition than the primary forest, and these differences affected the mutualistic networks (Figure 2.2), where primary forests had a wide range of strong and weak interactions, but the secondary forest was characterized by many, weak interactions. In other words, frugivores were still participating in the secondary forest mutualistic network, but the benefits exchanged in those mutualisms were few, compared to the primary forest. This has implications for conservation initiatives, as ecosystem services – such as seed dispersal – may be greatly diminished even though frugivores are still visiting and consuming fruit at fruiting trees.
It also indicates that, for some species, primary forest should remain a coservation priority if possible. 43 The mutualistic network studied in this paper appeared sensitive to anthropogenic disturbance (Figure 2.2). As the forest become more degraded, mutualistic networks became less speciose (Fig. 2.2) and less connected (Figure 2.5) and the asymmetries of pair-wise interactions changed (Figure 2.3). The relationship between connectance and the resilience of forest ecosystems has been noted in past studies; networks that become less connected are less robust to species extinctions as a decrease in connectance may increase the proportion of dependent mutualisms (mutualisms in which at least one of the partners relies on the other partner for survival) within a network (Jordano 1987). In addition, mutualistic networks in more degraded forests were simpler, with fewer species participating, fewer mutualistic interactions occurring, and an increasing dominance of one or a few interactions in the network (Figure 2.2). Simpler networks may be more vulnerable than complex ones to secondary extinctions following an initial extinction
(Bascompte and Jordano 2007). Finally, it is possible that non-native species (such as
Lantana camara) are now providing substantial fruit resources to some frugivores in these areas.
This study also provided interesting insight into the impacts of habitat degradation on two Endangered lemur species: Eulemur coronatus and Eulemur sanfordi. Similar to findings in other areas of Madagascar (Eulemur fulvus collaris, Bollen et al. 2004) these two species appear to be the primary seed dispersers of several trees in the primary forest that other frugivores (birds) are incapable of consuming due to their hard shells (e.g.
Baudouinia fluggeiformis) or because of their large size (e.g. Strychnos madagascariensis). In addition, both species decrease their fruit consumption in the secondary forest and only E. coronatus appears to be able to compensate for the loss of 44 preferred fruit species in the degraded forest; this is somewhat surprising as past studies noted that E. coronatus was found more often in the primary forest while E. sanfordi was found more often in forest edges and in degraded forest (Wilson et al. 1989, Fowler et al.
1990). However, given the decrease in fruit consumption by these two species in modified forests, it is unclear how the species will be able to adapt to resource declines and changes; the ranges of these two species are expected to shrink dramatically by 2080 due to climate change; their habitat ranges are predicted to contract to 0.4% and 46.5% of their current areas, respectively (Brown and Yoder 2015). Furthermore, these two lemur species are being taken from their natural habitat, at least at low levels, through hunting
(Randrianarisoa et al. 1999) and live capture (Reuter et al. 2015). The reduction of population sizes could decrease the seed dispersal of the trees in the primary forest that appear to rely on lemurs for dispersal (Figure 2.2); other studies have shown that population reductions of mammals due to poaching reduces seed dispersal and the population growth rates of fruit-bearing trees (Brodie et al. 2009).
Our study involved almost 600 hours of observations at focal fruiting trees, and in this way, we attempted an assessment of the effects that habitat degradation has on the frugivore-fruit tree mutualistic network. Although additional studies are needed to adapt our methods to more complex systems, our mutualistic network analysis nevertheless provides a novel quantitative view into the effects of habitat degradation on a community.
45 CHAPTER 3
LIVE CAPTURE AND EXTENT OF OWNERSHIP OF LEMURS IN
MADAGASCAR: EXTENT AND CONSERVATION IMPLICATIONS
Abstract:
Overexploitation is a significant threat to biodiversity, with live capture of millions of animals annually. An improved understanding of live capture of primates is needed, especially for Madagascar’s threatened lemurs. Our objectives were to provide the first quantitative estimates of the prevalence, spatial extent, correlates and timing of lemur ownership, procurement methods, within-country movements, and numbers and duration of ownership. Using semi-structured interviews of 1,093 households and 61 transporters, across 17 study sites, we found that lemur ownership was widespread and affected a variety of taxa. We estimate that 28,253 lemurs have been affected since 2010.
Most lemurs were caught by owners and kept for either short (≤1 week) or long (≥3 years) periods. The live capture of lemurs in Madagascar is not highly organized but may threaten several endangered species.
Introduction:
Overexploitation is a significant threat to biodiversity (Baillie et al. 2004), with hunting and live capture recorded throughout the tropics (e.g. Fa et al. 1995, Corlett
2007). In tropical forests hunting is conducted on a small scale for subsistence, and as part of organized trade for domestic and international markets (Corlett 2007). The bushmeat trade may be increasing with the human population (Corlett 2007) and as rural 46 communities gain access to urban markets (Fa et al. 1995, Duarte-Quiroga and Estrada
2003, Corlett 2007). However, despite advances in understanding hunting, live capture of animals through informal and formal routes remains poorly understood (Duarte-Quiroga and Estrada 2003, Nekaris et al. 2010).
Live capture may affect up to 4 million birds, 640,000 reptiles and 40,000 primates annually and the animals are traded globally (Karesh et al. 2005), usually to more affluent or urban customers (Duarte-Quiroga and Estrada 2003, Corlett 2007). In many cases, living animals are caught as part of trade networks for bushmeat and body parts, sometimes involving professional hunters, transporters and markets (Fa et al. 1995,
Duarte-Quiroga and Estrada 2003, Corlett 2007, Nekaris et al. 2010). This suggests that live capture, like the bushmeat trade, may be widespread and increasing.
Live capture is causing increasing concern in Madagascar (Schwitzer et al. 2013), where amphibians and reptiles are captured, sometimes to the point of near-extinction
(e.g. Grenoble 2013), and transported internationally via organized trade networks
(Andreone et al. 2005) for pet or medical trades. In contrast, little is known about the live capture of the country’s 197 native mammal species, 92% of which are endemic (IUCN
2013). Most documented captures of mammals in Madagascar are related to the bushmeat trade (Golden 2009, Razafimanahaka et al. 2012), although the trade appears to be less organized than in other countries (Golden 2009). However, recent political instability may have resulted in increased trading of bushmeat (Schwitzer et al. 2014) and facilitated an increase in live captures. Effective conservation of Madagascar’s mammals therefore requires a better understanding of the prevalence and breadth of the live capture of animals, including frequency, temporal trends, associated factors, and the extent to which 47 movement of animals from the point of capture is facilitated by an established trade network.
In particular, this information is needed for Madagascar’s endemic primates, the lemurs, which are one of the most threatened groups of large vertebrates (Schwitzer et al.
2014). Similar to other mammals, studies on lemur capture have focused on the bushmeat trade (e.g. Golden 2009), which may be increasing following a 2009 coup d’état
(Schwitzer et al. 2014). However, lemurs are easy to habituate (Eppley et al. 2011) and thus may be attractive as pets. Furthermore, records of holding facilities for captive lemurs indicate that ownership of lemurs has been ongoing and may be common (Welch
1996, Schwitzer et al. 2013). Although a small-scale study from the Union of the
Comoros suggested that the pet trade is the primary anthropogenic threat to introduced mongoose lemurs Eulemur mongoz (Clark 1997), the extent of live capture of lemurs has not been quantified.
The capture and sale of lemurs is illegal both domestically (Petter 1969,
Mittermeier et al. 2010) and internationally (UN 1973), with punishments including confiscation (Welch 1996). Despite formal restrictions on the capture and sale of lemurs, anecdotal reports of lemur ownership across north-west (Andrews et al. 1998), north-east
(Goodman 1993, Hekkela et al. 2007), east (Welch 1996, Birkinshaw et al. 2007), south- east (Rajaonson et al. 2010), south (Jolly et al. 1982), south-west (Zinner et al. 2001,
Sauther et al. 2013) and central (Nievergelt et al. 2002) Madagascar suggest that lemur ownership may be common and that regulations limiting lemur ownership are not enforced consistently. 48 Details about lemur ownership in Madagascar are scant. Anecdotal reports indicate that lemurs are kept in villages near forested areas (Zinner et al. 2001,
Birkinshaw et al. 2007, Hekkela et al. 2007), potentially as a back-up source of meat for food security (Zinner et al. 2001). They are also kept by hotel owners (Goodman 1993) to attract tourists (Schwitzer et al. 2013). The concept of lemurs as pets has been documented in Malagasy culture (Andrews et al. 1998, Sauther et al. 2013). In general, however, it is not clear how captive lemurs are obtained, to what extent they are moved in-country, or what happens to them post-capture. Given their threatened status
(Schwitzer et al. 2014), such information is needed to inform conservation efforts
(Mittermeier et al. 2010).
Our objectives were to (1) quantify the prevalence, spatial extent, correlates and timing of ownership, and (2) evaluate methods of procurement, movement around the country, the numbers kept and the duration of ownership. Based on the literature we hypothesized that (1A) many individuals would have owned a lemur or had knowledge of lemur ownership by others; (1B) lemur ownership would be widespread geographically and across taxa; (1C) the rate of lemur ownership would inversely correlate with human population density; and (1D) reports of lemur ownership would span the past few decades. The data collected were used to estimate the recent impact of lemur ownership in urban areas of Madagascar.
For the second objective we hypothesized that (2A) lemurs would be procured by owners through direct capture, and (2B) long-distance relocation of lemurs by means of public transport would be low, perhaps because of the illegal nature of lemur ownership; and if lemur ownership was more prevalent in (often) poorer rural areas the cost of 49 keeping a lemur would result in (2C) most owners keeping only one or two individuals.
Finally, because reports indicated lemurs were kept as a food resource or by hotel owners for display, we hypothesized that (2D) durations of ownership would be relatively short
(<1 month) or relatively long (≥1 year).
Figure 3.1: Maps of the study sites in Madgascar. Cities (a) in central and northern
Madagascar, and villages (b) in the vicinity of the Ankarana National Park, where interviews were conducted. The rectangle on (a) shows the location of (b) in northern
Madagascar.
50 Study area:
We collected data in cities and villages in central and northern Madagascar, between the capital city of Antananarivo and the northern regional capital of Antsiranana
(Figure 3.1). The cities were located along a 1,092 km highway, and the villages were located around the perimeter of Ankarana National Park (18,220 ha), which supports a high density of primates (Hawkins et al. 1990) and may be a key source of pet lemurs.
Methods:
International standards of research ethics were followed and research was approved by an ethics oversight committee (Temple University Institutional Review
Board, Protocol Number: 21414, May 2013). All primary researchers completed ethics training through the Collaborative Institutional Training Initiative. Research was authorized by the Madagascar Ministry of Water and Forests, Madagascar National
Parks, and locally elected officials.
Data collection:
During June–August 2013 we visited households (n = 1,093) in 10 cities (> 5,000 inhabitants) and seven villages (≤ 5,000 inhabitants) in northern and central Madagascar
(Figure 3.1, Table 3.1). In villages we sampled every fifth household. In cities random sampling was stratified by administrative unit. To ensure independent sampling only one person was interviewed per household. Respondents were head-of-household (self- identified as having major buying power for household goods) adults (≥ 18 years). If an eligible individual refused to participate or if nobody was present, sampling continued at 51 the next household. Interviews were anonymous and no identifying information was collected. Interviewees were reminded that questions could remain unanswered, the interview could be terminated at any point, and participation was voluntary. Verbal informed consent was received and interviewees chose the place, time and language
(French or local Malagasy dialect) of the interview. Interviews were conducted by a two- person team comprising an international project leader trained in ethical data collection and a trained Malagasy translator. Malagasy translators were always members of the predominant ethnic group of a study site, always fluent in the local Malagasy dialect, and never known to the interviewee.
To investigate the in-country movement of captive lemurs, interviews (n = 61,
Table 3.1) were conducted in city bus stations and ports that specialized in intercity transport. These busses are the primary methods of intercity transportation in
Madagascar. Individuals identified as vehicle drivers were asked to participate, and anonymity and informed consent were ensured. Villages did not have permanent bus stations or ports, and therefore no transportation interviews were conducted at these sites.
During 20-minute semi-structured interviews (Rietbergen-McCracken and
Narayan 1998) we provided participants with broad definitions of raising, purchasing and catching wild animals (raising was defined as keeping an animal alive for any period of time). Interviewees were asked whether they had ever seen a pet or raised lemur, using locally appropriate words for raise (miompy in Merina dialect, mitarimigna in
Sakalava/Antakarana) and pet (animaux domestiques in French, biby fiompy in Merina, biby tarimiana in Antakarana). If the answer was yes, we asked: Where, when, and how many did you see? We did not ask individuals whether they had owned a lemur, because 52 of the illegal nature of the activity and the potential for increased interviewee discomfort.
However, some interviewees indicated voluntarily that they were current or former owners. In such cases we asked the following questions: How did you procure the lemur?
How long did you have your pet lemur? Why did you stop keeping a pet lemur? Drivers at transportation hubs were asked: Have you ever transported a lemur? If the answer was yes: Was the lemur alive? What distance did you transport it? How much did you charge to transport it?
When speaking with interviewees we defined a pet lemur as any lemur that was perceived as belonging to an individual or a business, regardless of the purpose of its captivity. This included habituated, restrained and/or unrestrained lemurs that were cared for by an owner, but did not include fully wild lemurs living on privately owned land.
The concept of owning a habituated pet (such as a domestic cat) is understood in
Madagascar; respondents differentiated between free-roaming lemurs and habituated lemurs with an owner. We explicitly excluded captive lemurs in zoos. Interviewees could rarely identify the species of a captive lemur, and early interviews showed that providing images of lemurs did not result in consistent identification.
We conducted a literature search, in English, of first-hand reports of captive lemurs (excluding hunting, bushmeat and poaching) in Madagascar, comprising a search of ISI Web of Science, using the search terms ‘pet* OR own* OR captive* AND lemur*
AND Madagascar’; a Google Scholar search, for its flexibility in surveying literature across a wide variety of disciplines (keywords ‘pet lemur Madagascar’); and a search of
Lemur News (keywords ‘pet’, ‘pets’, ‘captive’), where researchers commonly share anecdotal reports. Table 3.1: Study sites, with the sample size and the percentage of individuals who had either owned a lemur or seen someone else with a pet lemur. Rural data from seven villages are aggregated to protect respondents’ identities. Population estimates for cities were obtained from the Ilo Program (2003) and from officials in villages.
Study Site Population Number of Percent of drivers Number of Know someone Current/ driver who had transported household who owns/has former interviews pet lemurs interviews owned a lemur owner Cities: Ambanja 28,468 8 0 (0%) 55 28 (50.9%) 6 (10.9%) Ambilobe 56,427 4 0 (0%) 99 31 (31.3%) 3 (3.03%) Andrevorevo - - - 40 20 (50.0%) 1 (2.5%) Andriba 32,000 - - 74 3 (4.05%) 6 (8.11%) Aniverano Nord 6,622 - - 90 29 (32.22%) 0 (0%) Antsohihy 105,317 1 0 (0%) 60 37 (61.67%) 4 (6.67%) Antananarivo 1,054,649 18 0 (0%) 199 71 (35.68%) 2 (1.01%) Antsiafabositra 8,328 - - 69 2 (2.89%) 9 (13.04%) Antsiranana 87,569 30 3 (10%) 180 33 (18.33%) 6 (3.33%) (Diego Suarez) Tsararivotra - - - 32 10 (31.25%) 0 (0%) Total (Cities) 61 3 (2 ± 3.92%) 898 264 (31.83 ± 37 (4.86 ± 12.03%) 2.85%) Villages: Ankarana Park No bus No bus stations or 192 41 (17.84 ± 0 (0 ± %) Perimeter Zone stations or ports 17.08%) Ambondromifehy 5000 ports Andranankoho 2000 Ampasinbengy 1997 Lambondry 120 53 54
Marotaolana 175 Matzaborimanga 400 Tsarakibany 250 Total (Villages and 61 3 (2 ± 3.92%) 1093 305 (26.07 ± 37 (2.86 ± Cities) 10.21%) 2.02%) Analysis:
As there may be greater variation between than within study sites, interviewees were used as subsamples within each study site for most analyses, except for subsets of the data with low sample size (owners of lemurs, n = 37; bus drivers who had transported them, n = 3). Results are presented as mean values with 95% confidence intervals. For mixed effects logistic regressions 95% CI was set at twice the standard error.
We used a mixed effects logistic regression for hypothesis 1c and examined two alternate models. Each model had a random effect of study site (city or village) plus a single fixed effect: the continuous variable ln(population size) (i.e. natural logarithmically transformed population size) or the categorical variable type of study site
(city or village). Population size was natural log transformed to increase model stability.
The effect of ln(population size) on whether a respondent self-reported lemur ownership was also modelled. We did not examine the effect of type of study site on self-reported ownership because all self-reported owners were in cities. Analyses were completed using R v. 3.0.2 (R Development Core Team 2013) with the lme4 package (Bates et al.
2013).
We estimated the total number of captive lemurs held in urban households since
2010 by extrapolating the frequency of lemur ownership at our urban sites. We assumed conservatively that only one lemur was owned per individual, representing one household. Based on Madagascar’s urban population (7.27 million people, UNDP 2013) and mean urban household size (4.4 people per household, INSTAT and ORC Macro
2005), there are c. 1,652,272 urban households in the country.
55 56 Price-related data are presented in Malagasy ariary, with U.S. dollar equivalents in parentheses, based on the exchange rate of 1 June 2013 (MGA 2,197 to USD 1, United
Nations Treasury 2014). For comparison, 81.3% of the population lives on Village data are aggregated for anonymity. Interviewees did not always give mutually exclusive answers, or know or provide the information requested. Therefore, sample sizes vary but are clearly indicated. Results: In accordance with hypothesis (1A), > 25% of interviewees (26.07%, 95% CI = 15.86 – 36.28) had seen a captive lemur; in three locations most respondents had knowledge of lemur ownership (Table 3.1). However, the mean percentage of individuals at a study site who reported currently or formerly owning a lemur was low (2.86%, 95% CI = 0.91 – 4.81, Table 3.1). In accordance with hypothesis (1B), reports of lemur ownership were geographically widespread (Figure 3.2). Respondents had seen captive lemurs in many areas of Madagascar in 2009 and earlier (prior to the 2009 coup d’état) and in 2010 and later (following the 2009 coup d’état, Figure 3.2, Table 3.2). We and other authors (Table 3.2) identified several taxa in four families that have been kept in captivity but, given the difficulties in accurate species identification, it was not possible to determine which taxa were most commonly and least commonly held captive. In contrast to hypothesis (1C), the human population size at a study site did not affect whether a respondent had ever seen a captive lemur (mixed effects logistic 57 regression, ln(population size) = 0.25, 95% CI = 0.30, P = 0.097) or whether a respondent had owned a lemur (mixed effects logistic regression, ln(population size) = 0.40, 95% CI = 0.48, P = 0.098). However, our results supported hypothesis (1C) when population size was examined as a categorical effect. Specifically, respondents in cities were more likely to have seen a captive lemur than respondents in villages (mixed effects logistic regression, type of study site(city) = 1.64, 95% CI = 1.61, P = 0.041). All self-reported lemur owners were from cities. Figure 3.2: Locations in Madagascar where respondents reported having seen pet lemurs during 2010–2013 (circles) and during 1960–2010 (triangles). 58 Table 3.2: Sightings of pet lemurs in Madagascar for which species or genus could be determined, including location, year, and evidence for taxon identification. Includes lemur taxa observed as pets that were: (1) named by respondents in our study or described in sufficient detail to permit identification, (2) personally observed and identified by our research team (bold and italicized), and (3) recorded as pets in scientific literature (bold). This is likely an incomplete list given that most respondents did not know species names and were not able to describe qualitative features of the lemurs they had seen in enough detail to permit species identification. The number of locations and dates a lemur species was seen also indicates the number of times it was cited by respondents. Lemur Location Seen by Respondent Notes (year seen) Family Cheirogaleidae Cheirogaleus sp. or Microcebus Tsararivotra (2000) Respondent identified lemur by sp. local name. Genus identified “Tsitsihy” using Harcourt & Thornback (1990) Family Indriidae Avahi laniger Sahalanona (2009) Rajaonson et al. 2010 Propithecus sp. Andriba (2010, 2011) Respondents identified lemur “Tsibahaka” (“black and white species in a guidebook. lemur”) Family Lemuridae Eulemur collaris Berenty (1980) Jolly et al. 1982 Eulemur coronatus Antsiranana (2013) Personally observed by Ankarana Perimeter Zone (2013) authors. Anivorano Nord (2013) Eulemur fulvus Morondava region (1999) Zinner et al. 2001 Eulemur sanfordi Antsiranana (2013) Personally observed by Anivorano Nord (2013) authors. Ambilobe (2013) Eulemur macaco Antsiranana (2013) Personally observed by Ambilobe (2013) authors. Hapalemur alaotrensis Lac Alaotra Mittermeier et al. 2010 Hapalemur griseus Lac Alaotra (1997, 1999) Nievergelt et al. 2002; Mutschler et al. 2001 Toamasina (2005) 59 Birkinshaw et al. 2007 Hapalemur occidentalis Ambanja (1983) Respondents identified lemur Antsiranana (Diego Suarez, 2011) species in guidebook. Hapalemur sp. Antsohihy (2013) Respondent listed “bamboo “Bamboo Lemur” lemur” as the type of lemur seen in captivity. Lemur catta Ambohitsoabe (Miadanandriana, 1990) Non-bold data: data provided to “Ring-tailed lemur” Ambohimanarina (1980) us by respondents who Ambositra (2000) identified pets as “ring-tailed Ankazobe lemurs”. Befotaka (2010) Bold data: Zinner et al. 2001; Fianarantsoa (1985) Sauther et al. 2013 Mahajanga (1995) Morondava region (1999) Toamasina Tulear (2013, 2009, 2012) Votamandry (1980) Varecia variegata Ile Sainte Marie (1991) Goodman 1993, Near Maroantsetra (1998) Hekkela et al. 2007 Masoala Vasey & Tattersall 2002 Family Lepilemuridae Lepilemur sp. Antsiranana (2013) Personally observed by authors. Lepilemur dorsalis Nosy Be (1995) Andrews et al. 1998 60 Most people (92% of n = 305) and most owners (84% of n = 37) could recall the date of seeing or owning a captive lemur, and many of these encounters had occurred since 2010 (63%, 95% CI = 49 – 77; and 43%, respectively). In accordance with hypothesis (1D), respondents noted having seen pet lemurs in every decade since 1960, whereas respondents reported personal ownership of lemurs in every decade since 1980. In cities 1.71% (95% CI = 0.71 – 2.71) of interviewees had owned a lemur since 2010, although only 0.6 (95% CI = 0.3 – 0.9) of individuals owned a lemur for ≥ 3 years during this time. If our urban study sites are representative of cities throughout Madagascar we therefore estimate that 28,253 lemurs have been held in captivity since 2010, with 9,913 of these held for at least 3 years. Thirteen of 37 owners reported the method of procurement of their lemur(s). In accordance with hypothesis (2A), the majority of owners procured their lemur(s) by direct capture (46%), although some had bought lemurs (31%) or received them as gifts (23%). Owners who had captured their lemurs indicated that the lemurs came from forested areas (n = 4) or were captured using a rope trap (n = 1) or after being habituated by gold miners (n = 1). Of the owners who had purchased their lemurs, three reported prices: MGA 3,000 (USD 1.03) per lemur from the French Mountain Protected Area near Antsiranana; MGA 3,000 (USD 1.03) from a dalaly (travelling merchant) in Antsohihy; and MGA 30,000 (USD 13.65) from an individual in Antananarivo. In accordance with hypothesis (2B), few (n = 3) intercity drivers had ever transported a live lemur (Table 3.1). All three were interviewed in Antsiranana and had transported live lemurs as recently as 2012 or 2013. The mean distance travelled was 41.8 61 km (95% CI = 14.3 – 69.3; range 15.0–62.4 km). Two of the three drivers had only ever transported one lemur and did not charge for the service. In contrast, the third driver estimated transporting one lemur every 3 months. It was not clear whether this was the same lemur being moved repeatedly or different individuals. The driver charged MGA 2,000 (USD 0.91) to transport one lemur 63 km. Overall, drivers did not charge an extra fee for personal baggage, including animals and meat transported for personal use. Most lemur owners (97%) told us how many lemurs they had owned. In accordance with hypothesis (2C), most of these households (68%) had owned only one lemur and the mean number of lemurs owned was 2.14 per household (95% CI = 0.96 – 3.32; range 1–22; Figure 3.3). Figure 3.3: Distribution of the number of lemurs owned by all self-reported lemur owners (n = 37). 62 Ten owners reported the duration of ownership as 1.17 years (95% CI = 0.05 – 2.29). However, as predicted by hypothesis (2D), most owners reported having their lemurs for ≤1 week (50%) or ≥ 3 years (30%). Some (20%) had owned their lemur for 2– 6 months. Eleven owners provided twelve explanations why they stopped keeping lemurs (one individual had owned more than one lemur and gave two different answers). Some (45%) said that their lemurs escaped; two sold the lemurs for MGA 10,000 (USD 4.55) each. In other cases the lemurs died (n = 2), were given away (n = 1), were killed for misbehaving (n = 1), or were returned to the forest (n = 1). Discussion: Extent of ownership: It is difficult to acquire data on the in-country acquisition, trade and ownership of live primates (Nijman et al. 2011). To our knowledge no previous study has quantified the prevalence and distribution of in-country ownership of endemic primates, although studies have estimated the number of primates sold in local markets (Ceballos-Mago et al. 2010, Shepherd 2010), quantified the number of captive primates in religious settings (Eudey 1994), examined characteristics of primate owners (Jones-Engel et al. 2005), and quantified densities of urban, free-living primate populations (Kyes et al. 2011). Our data indicate that knowledge of captive lemurs is common in Madagascar and a small but not insubstantial percentage of respondents have owned a lemur. Furthermore, lemur ownership is ongoing and is geographically widespread in the country. 63 Patterns of ownership: The rates at which respondents reported having seen or owned captive lemurs was higher in cities than in villages, although increasingly larger cities did not have correspondingly higher rates. Our city–village comparisons should be interpreted with caution because some respondents referred to seeing and/or owning lemurs in a different location to where they were interviewed. Nonetheless, our results indicate lemur ownership is common in urban areas. Given the (presumably) greater access to lemurs’ natural habitat from villages, the reasons for lower rates of ownership in villages are unclear. It is possible that lemurs are captured in remnant habitats and transported to cities, where owners may have more resources to keep them. Alternatively, our results may have been influenced by local factors at our study sites, including environmental education efforts that may have convinced residents not to capture lemurs (Madagascar National Parks, pers. comm.) or caused them to conceal their participation in such activities. Local cultural beliefs, including an aversion to hunting lemurs (Cardiff and Befourouack 2003), may also have influenced behaviour. The reported durations of lemur ownership suggest a dichotomy in motivation, although we did not examine this explicitly. Short ownership may be attributable to lemurs being held temporarily in rural areas prior to consumption (Zinner et al. 2001) or sale, or the difficulties in keeping some species (such as folivores) alive for long periods of time (Junge et al. 2009, Mittermeier et al. 2010). In Indonesia the motivations for keeping primates as pets may relate to professional status or religion (Jones-Engel et al. 64 2005), and in Mexico City empathy and a desire to possess a primate, as well as social status, were important motivating factors (Duarte-Quiroga and Estrada 2003). Well-established, high-volume markets are often highly organized (Andreone et al. 2005, Shepherd 2010). In Madagascar the movement of domestic meat from rural areas to urban markets occurs through middlemen and can involve coordinated transfers of cash and meat across hundreds of kilometres (KER, unpublished data). In contrast, and aligned with Golden’s (2009) statements about the informal nature of bushmeat trade in Madagascar, we found no evidence that live capture of lemurs involved the use of dedicated transport mechanisms or middlemen in a consistent manner. Although information may have been concealed, respondents never mentioned sellers capturing and selling lemurs as a regular business; lemurs were sometimes transported using regular passenger buses and typically were owned by the same person who captured them. This contrasts with the trade of live-captured amphibians in Madagascar (Andreone et al. 2005) and the black-market trade of primates in other regions (Duarte-Quiroga and Estrada 2003, Shepherd 2010). Similar to the primate trade in Mexico City (Duarte-Quiroga and Estrada 2003) and Venezuela (Ceballos-Mago et al. 2010), lemur ownership in Madagascar was not limited to foreigners living in Madagascar; in contrast, foreigners living in the Union of the Comores have been hypothesized to be the primary owners of pet lemurs there (Tattersall 1998). People keep lemurs despite being aware that this is illegal, which may suggest that enforcement is limited. Captive lemurs may be prevalent because of the low cost of obtaining native species from local habitats (Duarte-Quiroga and Estrada 2003); our price data, similar to those of Sauther et al. (2013), indicate that the cost of 65 purchasing a lemur and transporting it to a city is less than a typical day’s income in Madagascar. Animal welfare: Captive lemurs were kept in a variety of settings (Figure 3.4) and were sometimes exploited as for-profit attractions (Schwitzer et al. 2013). Lemurs were often kept in cramped conditions and given food that was inconsistent with their natural diet, including rice and bananas. Respondents often described lemurs in positive terms, comparing their hands, eyes and size to those of children, and some owners expressed higher esteem for their pet lemurs than for other domestic animals (e.g. dogs). These responses may be associated with cultural beliefs that lemurs are closely related to, or represent, humans (KER pers. obs., Jones et al. 2008). The positive attitudes towards lemurs are similar to those expressed in a study of primates in Mexico (Duarte-Quiroga and Estrada 2003). Conservation implications: Our estimate that 28,253 lemurs may have been held in captivity in Madagascar since 2010 is significant given the threatened status of many species of lemurs as a result of habitat destruction and hunting (Mittermeier et al. 2010). This may be a conservative estimate because respondents may not have been forthcoming about their status as current or former lemur owners given that lemur ownership is illegal in Madagascar (Mittermeier et al. 2010) and given that they were not asked directly whether or not they had owned lemurs in the past. At least four species have been reduced to < 500 individuals, and a further nine species to < 10,000 individuals in total (IUCN 2013). The true number of 66 captive lemurs may be higher than our estimate but further study is needed to confirm this, as we extrapolated our estimates to regions outside the study area. Lemur ownership continues at these levels despite increases in conservation awareness, area of protected habitat, and spending on lemur-related outreach (Schwitzer et al. 2014). Figure 3.4: Photographs of pet lemurs, taken with interviewee consent: (a) Lepilemur sp., held in a cage at a private residence. (b) Eulemur sp., restrained with a rope at a restaurant. (c) Eulemur coronatus male with its owner in a city. This lemur was not restrained but was highly habituated (image courtesy of Elodie Camprasse). (d) Habituated lemur being held by a tourist at a higher-end hotel (image courtesy of Olivier Raynaud). 67 Although lemur capture is generally detrimental to wild populations, it could potentially play a role in conservation efforts (Schwitzer et al. 2013) through programs aimed at the reintroduction of threatened species, maintaining genetic diversity in species whose wild populations are disappearing, and serving as a focus of environmental education. For such outcomes to occur some form of legalization of lemur ownership may be required; there are indications that the government of Madagascar may regulate some captive facilities in the future (Schwitzer et al. 2013). In this scenario for-profit businesses could enter a permitting process that would require inspections, minimum captivity requirements to ensure animal health, and limiting ownership to species known to fare well in captivity (Schwitzer et al. 2013). The legalization and regulation of private ownership would be more difficult but could include more consistent enforcement, using fines and a citizens’ reporting process, and a strict registration and permitting process combined with regular monitoring. The potential of such efforts, however, is unlikely to be realized unless legal frameworks, enforcement mechanisms, and monitoring efforts are reinforced and implemented on a much broader scale, and substantially more funding and personnel are made available. Any legalization initiative should proceed with caution to avoid complicating enforcement efforts and weakening existing protection of lemurs. Lemur species are not all equally represented in captive ownership. For instance, there were no reports of ownership of the widely distributed aye-aye (Daubentonia madagascariensis), perhaps because it is associated with negative taboos throughout much of its range (Mittermeier et al. 2010), including our rural study sites (Cardiff and Befourouack 2003). However, even low rates of live capture may be a threat for species with few individuals or subpopulations remaining (e.g. Eulemur sanfordi and Eulemur 68 coronatus). Likewise, wild populations of large-bodied species (e.g. Varecia variegata), which tend to have low reproductive rates (Harvey and Clutton-Brock 1985), may have low capacity to respond to removal of individuals. Comparative studies are needed to ascertain how the impacts of live capture vary between lemur species, and to understand how such impacts compare to those of habitat change, hunting and other threats. Our work is a first step towards quantifying the live capture of lemurs, informing efforts to conserve lemur populations, and clarifying means to regulate lemur ownership effectively. Our findings indicate that lemur ownership is common and widespread, and that a large number of threatened or otherwise susceptible taxa may be affected. They also highlight the importance of quantifying ownership of endemic primates in other tropical countries, especially where they are facing additional anthropogenic threats such as hunting and habitat change. 69 CHAPTER 4 THE CONSUMPTION OF WILD MEAT IN MADAGASCAR: FOOD SECURITY, DRIVERS OF CONSUMPTION, AND POPULARITY AS A FOOD ITEM Abstract: The role of wild meat consumption is debated; some communities rely heavily on wild meat, yet others treat it as a luxury good. We investigated the role of wild meat in food security in Madagascar, a country where wild meat consumption is poorly understood in urban areas and at regional scales. Using semi-structured interviews (n = 1343 heads-of-households, 21 towns), we aimed to: 1) quantify the amount and purpose of; 2) understand the drivers behind; and, 3) examine recent changes in wild meat consumption in Madagascar. Few respondents preferred wild meat (8 ± 3%) but most had eaten it at least once (78 ± 7%), and consumption occurred across ethnic groups, in urban and rural settings. More food insecure areas reported higher rates of recent consumption of wild meat. However, consumption was best explained by individual preferences and taboos. Few respondents (<1 ± <1%) had increased rates of consumption during their lifetimes, and wild meat prices showed no change from 2005-2013. Most consumption involved wild pigs and small-bodied animals, though these animal groups and lemurs were consumed less in recent years. Given these data, wild meat is unlikely to enhance food security for most Malagasy people in urban and well-connected rural areas. 70 Introduction: Wild meat can be less expensive than domestic meat (Fa et al. 2003) and thus has the potential to improve food security in poor communities in developing countries (Golden et al. 2011). Wild meat can add important nutritional value to consumer diets (Golden et al. 2011), which is important because malnourishment is often caused by a lack of protein, rather than a lack of calories (de Boer and Aiking 2011). Wild meat also provides economic security or fallback income for the rural poor (Kumpel et al. 2010, Golden et al. 2014a) because participation in hunting requires low capitalization and has few barriers to entry (Brown and Williams 2003), enabling even marginalized groups to access protein options (Golden et al. 2014a). Many communities, and even whole countries, have been considered ‘dependent’ on wild meat due to limited alternate protein sources (Fa et al. 2003). Nonetheless, if a meat resource becomes unavailable, communities will switch to alternate sources and can do so within one generation (Bennett 2002). This begs the question of how many people truly depend on wild meat and would suffer if it were no longer available (Bennett 2002). In addition, wild meat is used not only for subsistence but is also traded as a luxury good (Bennet 2002, Kumpel et al. 2010), and this trade may be becoming more commercialized (Lindsey et al. 2013). Thus, the role of wild meat in food security in developing countries is debated (Bennett 2002, Fa et al. 2003, Kumpel et al. 2010). Understanding why wild meat consumption occurs, and to what extent it enhances food security, requires understanding several drivers and correlates of consumption. First, economic vulnerability can affect rates of wild meat consumption; it is sometimes the last resort of poor or marginalized people (Lindsey et al. 2013). Second, consumption 71 patterns can be influenced by cultural values; elevated consumption of wild meat can reinforce social status among wealthy or high-status individuals (deFrance 2009). Third, immigration can bring people with distinct food preferences to new areas, affecting patterns of wild meat consumption at large scales (Poulsen et al. 2009). Fourth, consumer behavior may be impacted by price and preference; increasing prices often lead consumers to eat less wild meat and/or substitute it with other foods (Dostie et al. 2002), except where wild meat is strongly preferred over cheaper alternatives (Kumpel et al. 2010). Finally, consumption of wild meat may be affected by legal context; effective enforcement of laws and regulations that limit or prohibit hunting of wild animals can reduce rates of illegal hunting, consumption, or trade of wild meat (i.e. “bushmeat”, Lindsey et al. 2013). Thus, patterns of consumption of wild meat are highly variable and context-dependent, and are driven by a suite of micro- and macro-level drivers. Key to the debate over the role of wild meat in food security is that consumption is occurring at unsustainable levels (Bennett 2002, Fa et al. 2003, Kumpel et al. 2010). By themselves, tropical forests can only provide enough meat protein for one person per square kilometer (Robinson and Bennett 2000). Yet wild meat consumption occurs at an enormous scale, reaching up to 3.4 million tonnes per year in Central Africa alone (Wilkie and Carpenter 1999, Fa and Peres 2001), and resulting in population declines in large-bodied animals (Lindsey et al. 2013). In Central Africa, wildlife is being extracted at a rate that is six times higher than is sustainable (Bennett 2002) and the supply of wild meat protein is expected to drop 81% by 2050 due to overhunting (Fa et al. 2003). In many areas, wild meat will simply not be available in the future regardless of need 72 (Bennett 2002). Where subsistence hunting is unsustainable, food insecurity will increase unless alternate food sources are found (Fa et al. 2003, Lindsey et al. 2013). Wild meat might be expected to enhance food security in areas where human poverty and malnutrition occur near wild habitats (Golden et al. 2011), such as in Madagascar, a country where over 90% of the population lives on less than 2 USD per day (World Bank 2013), and political instability has exacerbated economic challenges. In Madagascar, foreign aid decreased following a 2009 coup d’état, with losses of over $400 million in foreign aid by 2012 (Ploch and Cook 2012). Further, 70% of the population regularly consume insufficient calories (< 2133 kcal/day) and an additional 8% and 3% of the rural and urban population, respectively, consume insufficient calories during seasonal rice shortages (Dostie et al. 2002). In part as a result, Madagascar has a high rate (50%) of adolescent malnutrition compared with other African countries (rates typically < 40%, Fotso 2007). A wide range of wild animals are known to be consumed in Madagascar – including lemurs, tenrecs, bats, mongoose, civets, fossa, wild cats, and wild pigs (Garcia and Goodman 2003, Goodman and Raselimanana 2003, Golden 2009) – despite national laws limiting or prohibiting hunting of these animals (Rakotoarivelo et al. 2011). Anecdotal reports and regional studies suggest that consumption occurs for subsistence (Golden 2009, Jenkins et al. 2011), following human-wildlife conflict (interactions between people and wild animals resulting in negative impacts on people and their resources or on wild animals and their habitats; Goodman and Raselimanana 2003, Madden 2004), or for luxury reasons (Golden 2009). These reports further suggest variation by species (Golden 2009) and region (e.g., Golden 2009, Jenkins et al. 2011). 73 Thus, evidence is conflicting about the purpose of wild meat consumption, and its role in providing food security in Madagascar. In this study, we aimed to assess the contribution of wild meat consumption to food security in Madagascar over large spatial scales and across ethnic groups. Our first objective was to quantify the amount and clarify the purpose of wild meat consumption among Malagasy people. In accordance with patterns observed in previous studies of wild meat consumption in Madagascar (Golden 2009, Randrianandrianina et al. 2010, Jenkins et al. 2011), we hypothesized that: (1A) most Malagasy people would have eaten wild meat in their lifetime; (1B) a wide variety of animal groups would have been consumed; (1C) domestic meat (i.e., meat from animals that were farmed or ranched) would be preferred over and consumed more frequently than wild meat; (1D) the purpose of consumption would vary by animal group; and (1E) consumption of wild meat would be higher in areas with greater food insecurity. Our second objective was to understand the micro- and macro-level drivers of wild meat consumption in Madagascar. Since several variables (food security, ethnic beliefs, individual preferences, and demographic and geographic variables) have been correlated with wild meat consumption elsewhere (Poulsen et al. 2009, Kumpel et al. 2010, Lindsey et al. 2013), we hypothesized that (2A) each of these variables would affect wild meat consumption in Madagascar, but that (2B) these variables would vary in relative importance. Our third objective was to examine variation in wild meat consumption over time and space. Because of the long history of wild meat consumption in Madagascar (e.g., Perez et al. 2005, Jenkins et al. 2011), we hypothesized that (3A) the overall popularity of 74 wild meat consumption, as evidenced by frequency of consumption or price, would not have varied over time. Nonetheless, because the 2009 coup d’état may have led to accelerated hunting of some animals (Schwitzer et al. 2014), we further hypothesized that (3B) consumption of specific types of wild meat would have increased in the recent past, and that (3C) this increase would have resulted from lax enforcement following the coup d’état. Further, while the greater amount of meat from large-bodied species may make them more desirable to hunt, some large-bodied animals have been locally extirpated or even hunted to extinction (Perez et al. 2005). We therefore hypothesized that (3D) most recent consumption of wild meat would be focused on small-bodied animals. Finally, because regions or ethnic groups may differ in beliefs, preferences, and taboos (Lambek 1992), we hypothesized that (3E) patterns of wild meat consumption would vary with the regional and ethnic characteristics of the local population. Methods: Study site: Data were collected (May-August 2013) in twelve urban (range: 6,622 – 1,054,649 inhabitants) and nine rural towns (range: 120-5,000) in central and northern Madagascar (Table 4.1). Towns were located along the 1,100 km-long highway connecting the northern regional capital (Antsiranana) with the national capital (Antananarivo) in the center of the country; this highway crosses several habitat types (Goodman and Benstead 2003) and ethnic groups (CIA 1976). Rural towns were sampled around the perimeter of Ankarana National Park, an important (18,220 ha) protected area 75 in northern Madagascar. Rural towns were located along roads or walking trails within 20 km of the highway. Table 4.1: Towns included in the study and the sample size of interviews at each location. The national capital Antananarivo was the southernmost population center included in this study while the regional capital Antsiranana was the northernmost. Population estimates retrieved from the Ilo (2003) database are indicated by (*); other population estimates were retrieved from elected officials. Town Number of Population Distance from interviews Antananarivo (km) Antananarivo 199 1,054,649* 0 Ankazobe 63 13,085* 92 Mahatsinjo 58 15,000* 177 Andriba 122 32,000* 198 Antsiafabositra 70 8,328* 243 Tsararivotra 32 - 496 Andrevorevo 40 - 582 Antsohihy 60 105,317* 668 Ambanja 55 28,468* 865 Ambilobe 99 56,427* 962 Ankarana 996 National Park (Rural towns) Ambondromifehy 30 5,000 1013 Ampasinbengy 30 1,997 1043 Andranokoho 33 2,000 1005 Lambondry 34 120 1052 Mahamasina 28 650 997 Marovato 30 400 1047 Marotaolana 30 175 990 Matzaborimanga 30 400 1022 Tsarakibany 30 250 1040 Aniverano Nord 90 15,000* 1030 Antsiranana 180 87,569* 1100 Total 1343 - - 76 Research permissions: The research design was approved by an ethical review board and conducted following all applicable laws, with authorization from the Madagascar Ministry of Water and Forests, Madagascar National Parks, and the highest-ranking locally elected official. Social surveys: Data were collected using semi-structured interviews (Rietbergen-McCracken and Narayan 1998) of adult male and female heads-of-household. Sampling of households was systematic in rural towns and stratified random in urban towns. No identifying information was collected from respondents. Interviews were conducted in the interviewee’s language of choice (French or local Malagasy dialect) after verbal informed consent had been secured by an American/British researcher and a translator. Further details on sampling protocols are in Reuter et al. (2015). During interviews, which lasted 11 ± 0.53 minutes (mean ± 95% C.I.), we collected data on meat consumption, including wild and domestic meat. Fish and non- meat protein sources were excluded from the survey scope. Interview questions (Appendix B) were designed to obtain systematic data on the interviewee’s 3-day diet recall (i.e., acquisition and consumption of meat within the last three days; following Jenkins et al. 2011); preferred meat type; lifetime acquisition and consumption of wild meat (following Jenkins et al. 2011); meat-related food taboos; and changes in meat- eating habits over time. 77 Analysis: Unless otherwise noted, results are presented as means ± 95% confidence intervals. Only aggregate information is presented for rural towns to protect respondent anonymity. Sometimes, the results are delineated between lifetime wild meat habits and recent wild meat habits; unless otherwise noted, lifetime wild meat habits include all data while recent data are from 2013 only (i.e., the 6-8 month period prior to interviews). Sometimes changes were examined along a continuum of human population. Human population data were retrieved from Ilo (2003) or from elected officials and were natural- log transformed prior to analyses to meet assumptions of normality. In one case, percentage data was arcsine-transformed to meet assumptions of normality. Initial questions on (non-fish) meat consumption were open-ended. Responses on domestic meat included a wide diversity of species, though analyses focus on chicken, pig, and zebu (a subspecies of cattle adapted to warm climates, and found throughout Madagascar) because of respondents’ strong emphasis on these species. Detailed follow- up questions focused on consumption of wild mammals. Because respondents typically could not identify exact species, but could differentiate between broader animal groups, data were aggregated into commonly-recognized mammal groups at differing taxonomic levels: lemurs (Superfamily: Lemuroidea superfamily), bats (Order: Chiroptera), tenrecs (Family; Tenrecinae), fossa (Cryptoprocta ferox), mongoose (Family: Herpestidae), rats and mice (Order: Rodentia), civets (Family: Viverridae), wild cat (Felis silvestris), and wild pig (Potamochoerus larvatus). Hypotheses in objective one were tested with the Kruskal Wallis Rank Sums Test and the Steel-Dwass Multiple Comparisons post hoc test or with a Pearson Chi-squared 78 Test. Relationships between wild meat consumption and predictor variables were examined with regression. Hypotheses (1A, 1B, 1C) used towns as replicates. As noted in the results, hypotheses (1D, 1E) used respondents as replicates where sample sizes were small and towns as replicates otherwise. For objective two, a two-tier model estimation and selection approach was used to examine the relative importance of five variables in predicting recent wild meat consumption. The three predictor variables examined at the respondent level (representing micro-level drivers) were: (1) prevalence of taboos, the number of mammal groups against which a respondent had a consumption taboo (index with range of 0-9, one point for each group); (2) access to meat, whether meat was consumed in the immediate (three days prior to the interview) past (proxy for food security); and (3) meat preference (categorical variable: wild meat/domestic meat/no meat preference). The response variable, recent wild meat consumption, the number of wild animals consumed in 2013 prior to the interview (a 6-8 month time period) was log-transformed prior to analysis. Then, we examined how variables at the town level (representing macro-level drivers) impacted wild meat consumption. Predictor variables included those retained in the best model from respondent-level analyses, including: prevalence of taboos (averaged at the town level for this analysis); and meat preference (the percentage of people within a town with a preference for wild meat). Two additional predictor variables were included in the town-level analysis: (4) province (the political province in which a town was located; this is a proxy for regional and ethnic characteristics of a population); and (5) town population (size of the human population). Town population was log- 79 transformed prior to analysis. The response variable was recent wild meat consumption in 2013 (averaged at the town level and log-transformed prior to analysis). Correlation coefficients suggested no pairwise correlations (|r| < 0.20) among these five variables. We identified a set of candidate models with these variables and – to limit complexity – two pairwise interactions at most. Candidate models were then ranked with the small- sample-size corrected Aikake Information Criterion (AICc; Hurvich and Tsai 1989). The best model had the lowest AICc. ΔAICc (the difference in AICc values from the best model, where ΔAICc < 2 suggests substantial support for a model) and Aikake weights (wi, the “weight of evidence” of a model relative to the other candidate models considered) were used to evaluate relative support for alternate models (Burnham and Anderson 2002). For objective three, we determined wild meat prices from 452 dated purchase records of wild meat, reported during interviews across all urban and seven rural towns. The change of wild meat price over time is analyzed from 2005-2013 for bats and from 2008-2013 for tenrecs and wild pigs. Prices were adjusted for inflation using two consumer price indices (Food Index and General Index, where 2000 is the base year) from FAOSTAT (2013), and converted into base year prices (in Malagasy Ariary) using the average index number for 2005. Price indices for 2005-2012 represented annual averages, while the 2013 price indices were averages from January to August 2013 (the date of the last surveys). 80 Results: Amount and purpose of wild meat consumption: In accordance with hypotheses (1A, 1B), most individuals across all towns had consumed some type of wild meat at least once before (78 ± 7%, Figure 4.1, Table 4.2). A diversity of mammals were consumed in each town (Table 4.3); tenrecs and bats were the most commonly consumed groups. Individuals who had consumed wild meat at least once reported consuming meat from 3.06 ± 0.36 wild animal groups (Table 4.3), though many interviewees had only consumed meat from one group (36 ± 8%). Most respondents across all towns expressed a preference for at least one type of meat (97 ± 2%); the remainder expressed no preference (2.02 ± 2.02%) or were vegetarian (0.05 ± 0.10%). In accordance with hypothesis (1C), of those who expressed a preference, few listed any type of wild meat (8 ± 3%, Figure 4.2). All wild meats were listed as preferred less frequently than the three most-preferred domestic meats (Steel- Dwass Multiple Comparisons, p < 0.05, Figure 4.2). However, despite this strong overall preference for domestic meat, the frequency of consumption of domestic meat was not always detectably higher than that of bats or wild pigs (Figure 4.3). 81 100 Life me consump on 90 80 Consump on from 2010-2013 70 60 50 40 30 20 10 Percent of respondents (%) 0 All Wild Lemur Tenrec Bat Fossa Mongoose Rats & Civets Wild Cat Wild Pig Meat Mice Animal Group Figure 4.1: Percent of respondents who reported having consumed different types of wild meat at least once in their lifetime, and once in the time period following the 2009 political coup d’état. Error bars depict 95% confidence intervals and towns are replicates. Table 4.2: Percent of individuals, by town, who had consumed a type of wild meat at least once in their lifetime (“L”) and at least once in since the 2009 political coup d’état (between 2010 and mid-2013; “R”). Data indicate the mean ± 95% confidence interval. City All Wild meat Lemur Tenrec Bats Fossa Mongoose Rats & Mice Civets Wild Cat Wild Pig Urban L(%) R(%) L(%) R(%) L(%) R (%) L(%) R(%) L(%) R(%) L(%) R(%) L(%) R (%) L(%) R(%) L(%) R(%) L(%) R (%) Ambanja 78 62 31 5 49 33 69 40 0 0 0 0 0 0 18 5 0 0 44 25 Ambilobe 82 63 35 10 69 21 62 46 1 0 2 0 0 0 24 2 7 0 46 32 Andrevorevo 93 78 53 8 80 58 70 10 8 0 13 0 0 0 48 28 28 10 65 45 Andriba 93 70 19 5 70 28 88 58 2 1 0 0 0 0 43 13 39 15 17 6 Aniverano Nord 83 57 40 16 71 54 26 11 32 6 1 0 2 0 2 1 0 0 0 0 Ankazobe 60 2 13 2 57 0 22 0 0 0 0 0 0 0 8 2 3 0 - - Antananarivo 60 15 6 0 37 5 16 3 1 0 0 0 1 0 7 2 7 2 30 7 Antsiafabositra 94 83 37 13 80 51 77 39 11 4 0 0 0 0 61 31 41 21 84 63 Antsiranana 44 29 8 3 18 12 25 17 1 0 1 1 0 0 7 2 1 0 19 11 (Diego Suarez) Antsohihy 90 73 32 8 78 45 78 57 2 0 2 0 3 0 27 7 12 2 55 47 Mahatsinjo 86 17 5 2 60 5 69 9 0 0 0 0 0 0 40 5 28 2 - - Tsararivotra 100 84 16 0 81 59 72 22 3 0 9 3 3 0 34 19 16 3 53 41 63± 31± 56± 26± <1± 27± 41± 28± Urban 80±0 53±6 25±9 6±3 5±5 1±1 2±2 1±1 0±0 10±6 15±9 5±4 11 13 15 12 <1 11 14 12 Rural Ankarana 59 60 27 10 67 36 34 29 13 4 5 2 <1 <1 25 15 5 4 12 11 National Park 56± 33± 45± <1± <1± 26± 20± Total 78±7 25±6 8±4 65±8 27±8 9±5 2±2 4±2 1±1 26±7 13±5 11±6 5±4 12 13 10 1 <1 12 10 82 Table 4.3: The range of mammalian wild meat groups consumed (by town), average number of wild animal groups consumed, and the percent of people having meat- related taboos in a town. Average Range of number of wild Percent of number of meat groups consumed (% individuals with Town wild meat of individuals who had only meat-related groups eaten one species) taboo (%) consumed Urban Ambanja 1 to 7 (45) 2.65 89.1 Ambilobe 1 to 8 (34) 2.91 84.9 Andrevorevo 1 to 8 (15) 4.38 92.5 Andriba 1 to 7 (30) 3.02 81.9 Aniverano 1 to 9 (16) 3.92 96.7 Nord Ankazobe 1 to 6 (41) 2.15 46 Antananarivo 1 to 7 (59) 1.71 50.25 Antsiafabositra 1 to 9 (24) 4.07 85.7 Antsiranana 1 to 9 (64) 1.9 84.4 (Diego Suarez) Antsohihy 1 to 8 (29) 3.31 75 Mahatsinjo 1 to 7 (39) 2.34 68.9 Tsararivotra 1 to 9 (37) 2.93 75 Total Urban 1 to 9 (35 ± 9) 2.94 ± 0.49 78 ± 9 Rural Perimeter Zone Ankarana 1 to 9 (34) 3.21 ± 0.55 92 ± 6 National Park (aggregate of 9 towns) Total 1 to 9 (36 ± 8) 3.06 ± 0.36 83 ± 6 83 84 Figure 4.2: Consumer preference (% of individuals who listed an animal group as a preferred type of meat) of wild and domestic meat, with towns as replicates. Different letters indicate significant differences between meat types (Steel-Dwass Multiple Comparisons, p < 0.05). Boxes to the left of the dashed line are wild meats while those to the right are domestic meats. Types of wild meat not listed by any respondents as a preferred type of meat are not included. 85 Figure 4.3: Frequency of consumption of domestic and wild meat, with towns as replicates, for individuals with a stated preference for these meats. Boxes to the left of the dashed line are wild meats while those to the right are domestic meats. Data are from answers to question 2 of the interview (Appendix B). Only animals whose meat was preferred by more than 20 respondents are shown; others are excluded due to small sample sizes. Different letters indicate significant differences between meat types (Steel- Dwass Multiple Comparisons, p < 0.05). 86 In accordance with hypothesis (1D), mammals were hunted and consumed for differing reasons (Figure 4.4). Some mammal groups, including three of the four carnivore groups (civets, fossa, and wild cats but not mongooses) were eaten primarily due to human-wildlife conflict (all three carnivore groups ≥ 80% of respondents). Others were consumed primarily due to insufficient food resources (lemurs, mongoose, and tenrecs; all ≥ 64%). Still others were eaten primarily as a luxury item, purchased using discretionary income (bats and wild pigs; both ≥ 50%). No animal group was consumed primarily when respondents were on a vacation, hosting guests, or eating out at a restaurant, though a subset (> 11%) of bats, lemurs, tenrecs, and wild pigs were consumed on these occasions. The percent of individuals who reported consuming wild meat due to human-wildlife conflict differed significantly by animal group (Pearson Chisquare, DF = 6, χ2 = 108.636, p < 0.0001; rats/mice and mongoose removed from analysis due to small sample sizes; respondents as replicates). The percent of people who cited human-wildlife conflict as a reason for consuming any type of wild animal decreased significantly with the human population of a town (Regression, F (1,8) = 11.1232, p = 0.0103; towns as replicates). In contrast, the percent of people who cited a lack of sufficient food as the reason they consumed wild meat did not change significantly with the human population of a town (Regression, F (1,8) = 1.5679, p = 0.2459; towns as replicates, percentage data are arcsine-transformed). 87 Figure 4.4: Reasons why respondents had consumed wild meat. Respondents are replicates, due to small sample sizes. “Vacation/Guests/Restaurant” = eaten when respondent was on a vacation, hosting guests, or eating out at a restaurant; “Discretionary Spending” = eaten when respondent had money to purchase meat; “When respondent had insufficient food” = eaten when respondent had no other meat options; “Human-Wildlife Conflict” = eaten after animal was caught consuming human food or after capture in traps used to protect farm animals and crops. In accordance with hypothesis (1E), a town’s food security (proportion of a town that had consumed meat in the three days prior to our interview) was inversely related to its respondents’ recent (6-8 months prior to interview) wild meat consumption (Regression, F (1,19) = 5.2469, p = 0.0336; R2 = 0.216). The percentage of people who cited a lack of sufficient food as the reason behind the consumption of wild meat differed significantly by animal group (Pearson Chisquare, DF = 6, χ2 = 54.265, p < 0.0001; rats/mice and mongoose removed from analysis for small sample size; respondents as 88 replicates). Some wild animal groups – lemurs (Regression, F (1,19) = 6.9750, p = 0.0161; towns as replicates), tenrecs (Regression, F (1,19) = 8.8948, p = 0.0077), and wild pigs (Regression, F (1,19) = 6.6373, p = 0.0185) – were consumed at higher frequencies in 2013 in food-insecure towns, but all others were not (all p-values > 0.20). Micro- and macro-level drivers of wild meat consumption: In accordance with hypothesis (2B), recent wild meat consumption (in the 6-8 months prior to the interview) at the respondent level was best explained by a model with the predictor variables prevalence of taboos and meat preference (Table 4.4) but not the variable access to meat. Further examination of the best model suggested that both variables were significant predictors of recent wild meat consumption at the individual respondent level (multiple regression; R2 = 0.25; prevalence of taboos, p < 0.0001; meat preference [wild meat - no meat], p < 0.0001; meat preference [domestic meat - no meat], p = 0.018; Figure 4.5A, 4.5B). At the town level, recent wild meat consumption was best explained by the variables town population and meat preference (Table 4.4), but not the variables prevalence of taboos or province. However, a second model that also ranked highly (ΔAICc = 0.71) only retained a single variable: meat preference. Further examination of the best model suggested that only meat preference was a significant predictor of wild meat consumption at the town level (multiple regression; R2 = 0.42; town population, p = 0.7194; meat preference, p = 0.0049; Figure 4.5C). 89 Table 4.4: Results of model selection listing the corrected Aikake information criterion (AICc), the ΔAICc (difference from the best fit model), and the Aikake weight (wi). POP – town population; L – province (location); T – prevalence of taboos; A – access to meat; P – preference for meat. Model AICc ΔAICc wi Respondent-level variables T + P, Random effect: town 1503.6 0 0.97 T + A + P, Random effect: town 1510.36 6.76 0.033 T + A + P + T*A, Random effect: 1519.96 16.36 < 0.0001 town T + A + P + A*P, Random effect: 1522.83 19.23 < 0.0001 town T, Random effect: town 1526.58 22.98 < 0.0001 T + A + P + T*P, Random effect: 1527.57 23.97 < 0.0001 town T + A + P + T*P + A*P, Random 1529.86 26.26 < 0.0001 effect: town T + A + P + T*A + A*P, Random 1532.33 28.74 < 0.0001 effect: town P, Random effect: town 1532.82 29.22 < 0.0001 T + A, Random effect: town 1532.87 29.27 < 0.0001 T + A + P + T*A + T*P, Random 1537.11 33.51 < 0.0001 effect: town A + P, Random effect: town 1538.19 34.59 < 0.0001 T + A + P + T*A + T*P + A*P, 1549.43 45.83 < 0.0001 Random effect: town Intercept only, Random effect: 1564.15 60.55 < 0.0001 town A, Random effect: town 1568.69 65.09 < 0.0001 T + A + P + T*A + T*P + A*P 1700.72 197.12 < 0.0001 Town-level variables POP + P -1.83 0 0.48 P -1.12 0.71 0.34 POP + T + P 1.51 3.33 0.09 POP 4.61 6.44 0.02 Intercept only 5.6 7.42 0.01 POP + L + P 5.75 7.58 0.01 L + T + P 6.26 8.09 < 0.01 T + P 6.29 8.12 < 0.01 90 POP + T 6.29 8.12 < 0.01 L 7.32 9.14 < 0.01 T 7.43 9.25 < 0.01 L + T 9.04 10.87 < 0.01 POP + L 9.88 11.71 < 0.01 POP + L + T + P 10.26 12.08 < 0.01 POP + L + T 11.31 13.14 < 0.01 POP + L + T + P + POP*P 16.24 18.06 < 0.001 POP + L + T + P + POP*T 16.47 18.3 < 0.001 POP + L + T + P + L*T 19.28 21.11 < 0.001 POP + L + T + P + L*T 19.28 21.11 < 0.001 POP + L + T + P + POP*P + L*P 23.07 24.89 < 0.001 POP + L + T + P + POP*T + L*P 23.07 24.9 < 0.001 POP + L + T + P + POP*T + L*T 23.74 25.57 < 0.001 POP + L + T + P + POP*T + 23.84 25.66 < 0.001 POP*P POP + L + T + P + L*T + L*P 27.9 29.72 < 0.001 POP + L + T + P + POP*P + L*T 28.76 30.58 < 0.001 91 Figure 4.5: The impact of taboos against wild meat and a preference for wild meat at the respondent (A; B) and town-level (C). All bivariate plots are derived from the multiple regression analysis of the best model; (A) and (C) show leverage plots, and (B) shows least squares means of significant relationships between predictor variables (prevalence of taboos, meat preference) and the response variable (recent wild meat consumption). Town population, a variable in the town-level best model, was not significantly related to recent wild meat consumption and is not shown. A) 92 B) C) 93 Variation in wild meat consumption over time and space: Evidence of wild meat consumption was mostly but not completely consistent with hypothesis (3A). Specifically, respondents reported consuming wild meat in every decade since the 1940’s, and this consumption continues; many (31 ± 11%) respondents had consumed at least one type of wild meat in the 6-8 months prior to the interview (Table 4.5). Most respondents had not varied rates of wild meat consumption over their lifetimes (Table 4.5), though a substantial minority had decreased their consumption of wild meat (14 ± 5%). Virtually no respondents had increased their consumption of wild meat (<1 ± <1%, Table 4.5). On average, respondents had last eaten wild meat 4 ± 3 years ago (Table 4.5). Similarly, and also consistent with hypothesis (3A), prices for wild meat have not varied over the past 6-9 years. The price of bat meat did not change significantly from 2005 to 2013 when adjusted for inflation using either consumer price index, the General Index (price per whole bat; Mixed effects model, Regression, p = 0.9461) or the Food Index (Mixed effects model, Regression, p = 0.9141). Inflation-adjusted prices likewise did not vary for tenrec (price per whole tenrec; General Index: Mixed effects model, Regression, p = 0.6989; Food Index: Mixed effects model, Regression, p = 0.7311) or wild pig (General Index, Price per ‘piece’: Mixed effects model, p = 0.7890; Price per kg: p = 0.10; Food Index Price per ‘piece’: Mixed effects model, Regression, p = 0.7923; Price per kg: p = 0.1094) from 2008-2013 (Figure 4.6). Table 4.5: Changes in wild meat consumption, rates of recent consumption (6-8 months prior to the interview), and the average duration since last consumption. Body mass estimate for wild cats was retrieved from Brockman et al. (2008). All other body size estimates were calculated from Garbutt (2007), using the average of the maximum mass recorded for each species in an animal group. Averages ± 95% CI, with towns as replicates. Percent of respondents Percent of Last time Percent of respondents who who decreased the individuals who meat from increased the frequency of frequency of had consumed the animal Animal species Body mass consumption of meat from consumption of meat meat of the animal species or or group (kg) the animal species or group from the animal species species or group group was across their lifetime (%) or group across their within 6-8 months consumed lifetime (%) of interview (%) (years ago) Bat 0.05 ± 0.14 0 ± 0 2 ± 2* 11 ± 5* 8 ± 2* Civet 3.47 ± 1.38 0 ± 0 <1 ± <1 3 ± 2* 11 ± 4* Fossa 10 0 ± 0 0 ± 0 <1 ± <1* 19 ± 8* Lemur 1.65 ± 1.90 0 ± 0 1 ± <1* 5 ± 3* 11 ± 3* Mongoose 0.93 ± 0.34 0 ± 0 0 ± 0 <1 ± <1 19 ± 13* Rats and mice 0.19 ± 0.25 0 ± 0 0 ± 0 0 ± 0 17 ± 6* Tenrec 0.12 ± 0.36 <1 ± <1 3 ± 2* 19 ± 9* 6 ± 3* Wild Cat 5.44 0 ± 0 0 ± 0 2 ± 2* 11 ± 5* Wild Pig 70 <1 ± <1 8 ± 3* 12 ± 7* 5 ± 3* All Wild meat - <1 ± <1 14 ± 5* 31 ± 11* 8 ± 1* * Significant difference from zero, one sample t-test, p < 0.05 94 Figure 4.6: Change in the prices of one unit of wild meat, by animal group, over time (bat and tenrec unit = one animal; wild pig unit = 1 kg). Prices (Malagasy Ariary) have been adjusted for inflation using annual FAOSTAT (2013) Food Indices and then standardized at the 2005 base year. Prices for all three animal groups have not changed significantly in the recent past. In contrast to hypothesis (3B), the consumption of lemurs, bats, tenrecs, wild pig, and civets had declined in recent years (Table 4.5). Nonetheless, consumption of some of these groups was not uncommon; more than 10% of respondents had consumed meat from bats, tenrecs, and wild pig in the previous 6-8 months (Table 4.5) and more than 20% had consumed meat from these same animal groups during the 2010-2013 period (Figure 4.1, Table 4.2). 95 96 In contrast to hypothesis (3C), reasons for changes in wild meat consumption seemed mostly unrelated to the 2009 coup d’état. In particular, respondents never mentioned decreased enforcement following the coup d’état as a reason for changed consumption of wild meat. Instead, several other reasons were mentioned, and these reasons varied by animal group (Figure 4.7). Changes in taste preferences were common reasons for a decrease or end to lemur and tenrec consumption. Lack of availability (either due to rarity of the species or due to a lack of acceptance in a new region for the consumption of that meat) was commonly cited as a reason why bat consumption declined or ceased. When respondents stopped eating wild meat for religious reasons, it was usually after adopting Muslim or Adventist beliefs (n = 21 of 34 respondents); some of these respondents (n = 7) stopped eating all wild meat while most (n = 26) just stopped eating wild pig and one respondent stopped eating tenrec. When respondents cited medical reasons as a motivation to decrease consumption, it often involved concerns about high cholesterol (n = 8 of 31) and the belief that a low-meat diet (of both domestic and wild meat) would mitigate health problems. Four of these respondents stopped eating all wild meat, while the other four just stopped eating wild pig (n = 2) or tenrecs and bats (n = 2). In some cases (n = 39), women (and their children) adopted their husbands’ taboos after marriage or childbirth. Very few respondents (n = 2) stopped eating wild meat because they felt it was bad for the environment; one who stopped eating lemur felt that lemurs were more scarce and needed to be preserved while a second who stopped eating tenrecs felt the health of tenrec populations were linked to local rainfall cycles (it seems to be a local belief that tenrec populations are indicative of how much rainfall will occur). 97 Of the 11 (of 1343) respondents who reported increasing their consumption of wild meat, five stopped following previously-held taboos, two remarried and no-longer followed ex-partners’ taboos, and four changed their religions to ones that permitted consumption of wild meat. There was some evidence for hypothesis (3D); a large majority (61 ± 5%) of records of wild meat consumption were of small-bodied animals (those < 0.5 kg; Table 4.5). This pattern was widespread; consumption of small-bodied animals was > 50% in 19 of 21 towns. The last consumption of large-bodied (≥ 0.5 kg) groups of wild animals by a respondent occurred on average ≥ 11 years ago, except in wild pigs (last consumed 5 ± 3 years ago, Table 4.5). 100% 90% Change in religion 80% 70% Change in taste 60% Medical reasons 50% 40% New fady due to 30% marriage/childbirth Percent of respondents 20% No longer available 10% 0% Environment Lemur Tenrecs Bats Wild Pig (n=10) (n=39) (n=30) (n=94) Type of wild meat Figure 4.7: Reasons why people decreased or stopped eating wild meat. Individuals are replicates due to small sample sizes. 98 Finally, there was some evidence for hypothesis (3E). While patterns of wild meat consumption were broadly similar across the three provinces, three variables differed. Specifically, the percentages of individuals who: 1) had taboos against wild meat; 2) had consumed wild meat following the 2009 coup d’état; 3) and had decreased the frequency of consumption of any type of wild meat across their lifetime were each considerably lower among respondents in the central highland province of Antananarivo (which contains the capital city) than among respondents in the more northern and coastal provinces of Mahajunga or Antsiranana (Table 4.6). 99 Table 4.6: Shifts in wild meat consumption behavior among provinces, with data limited to urban towns. Antananarivo is the southern-most province while Antsiranana is the northern-most province. Data show averages ± 95% Confidence Intervals, with towns as replicates within provinces. Provinces Wild meat consumption Mahajunga Antsiranana Antananarivo behavior (n = 6 (n = 4 (n = 2 towns) towns) towns) Percent of individuals with taboos against 18.76 ± 16.09 56.51 ± 12.66 52.28 ± 30.79 wild meat Average number of taboos held against the nine mammalian animal groups (of 0.63 ± 0.57 1.39 ± 0.46 2.23 ± 1.55 people who held any kind of meat-related taboo) Percent of individuals who had ever 60 ± 0 92.67 ± 3.68 71.75 ± 18.25 consumed wild meat Percent of individuals who had consumed wild meat following the 2009 political 8.5 ± 12.74 67.5 ± 20.27 52.7 ± 15.73 coup d’état (2010 to mid-2013) Average frequency of wild meat 10.82 ± 2.95 14.38 ± 2.51 12.74 ± 5.90 consumption per year Last time wild meat was consumed, on 8.02 ± 1.08 8.93 ± 1.11 9.97 ± 1.08 average (years ago) Species richness of wild meat consumed 1.93 ± 0.43 3.34 ± 0.61 2.84 ± 0.81 Percent of respondents who increased the frequency of consumption of any type of 0.51 ± 0.98 0.89 ± 0.94 0.51 ± 0.99 wild meat in their lifetime Percent of respondents who decreased the frequency of consumption of any type of 9.41 ± 9.58 19.64 ± 10.59 18.69 ± 11.99 wild meat across their lifetime Discussion: Amount and purpose of wild meat consumption: Most Malagasy respondents (78 ± 7%) had consumed wild meat in their lifetime (Figure 4.1). This consumption occurred in all towns examined, spanned political and ethnic boundaries across a wide geographic region, and took place in both urban and rural settings (Table 4.3). This accords with previous studies suggesting occasional 100 consumption is high; at least 70% of people in western Madagascar (Razafimanahaka et al. 2012) and 95% of households in eastern Madagascar (Golden 2009) consumed wild meat at least once per year. In addition, wild meat consumption impacted a wide range of mammalian taxa and individual respondents had on average consumed wild meat from more than three distinct animal groups (Table 4.3). Bats and tenrecs were the most commonly consumed groups of wild animals (Figure 4.1, Table 4.2), and people with a stated preference for bats and wild pigs consumed the animals at a yearly frequency similar to chicken and domestic pig (Figure 4.3). Nonetheless, consistent with Jenkins et al. (2011), few individuals (8 ± 3%) expressed a preference for wild meat, and all wild meat types were preferred less than the most popular types of domestic meat (Figure 4.2). Consistent with prior studies (e.g. Golden 2009, Jenkins et al. 2011), the purpose of wild meat consumption varied by animal group (Figure 4.4), and lemurs, tenrecs, and mongoose were consumed when insufficient food was available. In contrast, carnivores were mostly consumed following human-wildlife conflict suggesting opportunistic consumption, while bats (Jenkins et al. 2011) and wild pigs were consumed when discretionary income was available, suggesting a luxury purpose. Micro- and macro-level drivers of wild meat consumption: Drivers of wild meat consumption were evident at both the micro- (respondent scale) and the macro-level (town scale). Specifically, at the micro level, a person’s taboos against wild meat and meat preferences together predicted his/her consumption of wild meat. At the macro level, the meat preference of the residents of a town was the strongest 101 predictor of wild meat consumption (Table 4.4, Figure 4.5). There are clearly other factors that influence the consumption of wild meat (e.g. economic status, Lindsey et al. 2013; price of meat options, Dostie et al. 2002). However, it is notable that simple models explained such a high amount of the variation in the data (R2 = 0.25 at the respondent-level, R2 = 0.42 at the town level) for something as complex as the sometimes-illegal and occasional consumption of nine different wild mammalian groups in Madagascar. Our results, from urban and well-connected rural areas of Madagascar, showed preference and taboos, and not access to meat – a proxy for food security – are important drivers of consumption. These results are in line with much prior research from urban areas in sub-Saharan Africa (Lindsey et al. 2013), but are different from those focusing on poorer rural areas in Madagascsar where consumption was strongly linked with food security (Golden et al. 2011). We did not collect data on income levels of individual respondents in our study, though broadly speaking, poverty rates in Madagascar are lower in urban (52.1%) than in rural areas (76.7%, IFAD 2007). Taken together, our results and those of previous studies (Golden et al. 2010, Lindsey et al. 2013) suggest that in wealthier and more urban areas, preference and social reasons may drive consumption while in poorer and more rural areas, food security may still be a driving factor. Further studies are needed to clarify these drivers. Variation in wild meat consumption over time and space: Contrary to recent reports (Schwitzer et al. 2014), we detected little evidence that the rate of wild meat consumption has increased in Madagascar. While consumption of 102 wild meat is not uncommon (31 ± 11% of respondents had consumed it in the last 6-8 months, Table 4.5), most (85 ± 5%) people had not changed their rate of wild meat consumption over their lifetimes, and a substantial minority (14 ± 5%) had decreased it. In addition, meat prices did not vary over the past 6-9 years (Figure 4.6), which – while by itself is not definitive – is also suggestive that demand may not have changed substantially. Taboos were an important factor affecting wild meat consumption. Most (83 ± 6%) respondents had meat-related taboos (Table 4.3) and – consistent with previous studies (Jones et al. 2008) – respondents who had more taboos consumed less wild meat (Figure 4.5). Further, in contrast to recent suggestions that taboos might be breaking down in Madagascar (Jones et al. 2008, Jenkins et al. 2011), only five of the 1343 respondents indicated they increased wild meat consumption because they stopped following a taboo. It remains unclear whether taboos continue to influence wild meat consumption because respondents believe the taboo (and fear supernatural repercussions) or because, as suggested by Jones et al. (2008), they seek to avoid social disapproval. Regardless, taboos in Madagascar are complex; they may be regional phenomena, but may also be village- or family-based, or impact just one or a few individuals (Lambek 1992). The role that preference plays in wild meat consumption also cannot be ignored. Meat preferences were strongly linked to rates of wild meat consumption (Figure 4.5), and respondents citing a preference for wild meat were not limited to any one ethnicity, geographic area, income class, or religious affiliation. Respondents did not cite changes in enforcement following the 2009 coup d’état as being related to wild meat consumption. Rather, respondents cited changing taste 103 preferences (for declines in lemur and tenrec consumption); acquisition of new taboos following marriage, childbirth, or a change in religion (wild pigs); and reduced availability of animals to hunt (bats, Figure 4.7). These responses suggest taste preferences, taboos, and availability may be more important drivers of wild meat consumption than has previously been appreciated. Conservation implications: Our research adds to accumulating evidence (e.g., Golden et al. 2011) that wild meat is being hunted and consumed in Madagascar at levels that are not biologically sustainable. Although our data suggest wild meat consumption is not increasing overall (Table 4.5), apart from wild pigs, recent consumption has focused on small-bodied animals such as bats and tenrecs (Figure 4.1). Because large-bodied animals provide more meat and are thus presumably are more desirable to hunt where available (Garcia and Goodman 2003), the emphasis on small-bodied animals may suggest large-bodied animals other than wild pigs have become so rare they are no longer viable hunting targets. This is consistent with the unsustainable pattern, observed in forests across multiple continents, in which the largest-bodied species are progressively hunted to rarity (Wright et al. 2007). Hunting of small-bodied bats, however, is also unsustainable; bats have such low reproductive rates that their populations are typically unable to sustain even low levels of hunting over time (Pierson and Rainey 1992). That hunting of bats in Madagascar is unsustainable is further supported by one of the respondents’ principal reasons for decreasing recent consumption of bats: reduced availability (Figure 4.7). Since bats are 104 primarily consumed with discretionary income (Figure 4.4), this result suggests respondents would eat more bats for a luxury purpose if they were available, but that larger bats have become rare or absent in many areas. The unsustainable hunting of bats in Madagascar has been noted before (e.g. Goodman 2006), and our data may indicate that this is occurring at a larger scale than previously recognized. Given the context of several decades of conservation attention to Madagascar, along with recent, heightened conservation programming there (Schwitzer et al. 2014), it is notable that concerns for the environment or animals were almost never cited as a reason why wild meat consumption had declined (Figure 4.7). This result may indicate that past efforts for environmental education at regional and national levels have not been entirely effective. However, unsustainable hunting of key species could be addressed in the future through: 1) better policing of existing wildlife laws including the implementation of biologically relevant hunting seasons, quotas, and enforcement of hunting bans for endangered species (Rakotoarivelo et al. 2011); 2) awareness-raising among the public of wildlife laws (Keane et al. 2011); 3) using existing taboos as a tactic to increase community compliance (Westerman and Gardner 2013); and 4) the development of alternative sources of food security, especially in the most remote areas of Madagascar (Golden 2009, Golden et al. 2014b). Wild meat consumption and food security: The role of food security in wild meat consumption is complex. In accordance with recent studies suggesting wild meat consumption may play a role in the food security of isolated communities in Madagascar (Gardner and Davies 2014, Golden et al. 105 2014a), respondents in our study with lower access to any meat in the three days prior to the survey (a proxy for food security) reported greater recent (in the previous 6-8 months) consumption of wild meat overall – and greater consumption of wild meat from lemurs, tenrecs, and wild pigs in particular. While these results suggest a link between wild meat consumption and food security, the link was not strong; access to meat was not a principal driver of wild meat consumption in our respondent population from urban and well-connected rural areas (Table 4.4). These results together suggest that wild meat consumption does not strongly influence food security, though particular animal groups may occasionally serve as a fallback food when preferred meat or basic food sources are harder to procure. In these situations, larger-bodied animals (e.g., lemurs, wild pigs) or easily caught animals (estivating tenrecs) may be preferentially hunted and consumed. Although the majority of the Malagasy population is not consuming the recommended number of calories per day (Dostie et al. 2002), wild meat is poorly suited to enhancing long-term food security in Madagascar. Our results suggest large-bodied animals may be no longer be widely available for consumption, and even small-bodied animals such as bats that are heavily hunted may be disappearing (Goodman 2006). As such, wild animals that could contribute to food security are not being hunted sustainably in Madagascar (Golden et al. 2011). Further, the extensive and ongoing loss of forest cover (Schwitzer et al. 2014) along with the widespread exploitation of mammals (Jenkins et al. 2011) may be decreasing access to wild meat in both rural and urban areas of Madagascar. The low preference for wild meat relative to domestic meats (Figure 4.2), and the wide prevalence of taboos against consumption of meat from particular animal 106 groups (Table 4.3) could further limit wild meat consumption and complicate the use of wild meat to promote food security. Therefore, alternative food security programs should be instituted that do not rely on wild meat consumption and that fit within existing social norms. One potential solution would be to increase production of domestic animals as an alternative to wild meat (Golden et al. 2014b). Alternatively, the promotion of plant-based proteins that could be more sustainable to produce than domestic meat (Pimentel & Pimentel 2003) and a more culturally appropriate means to improve nutrition among Malagasy people with religious and ethnic backgrounds that favor meat-restricted diets (Walsh 2007). A third might be the development of agroforestry systems that promote productivity and complementarity in food production, even within small land areas (Styger et al. 1999). Because food insecurity occurs across Madagascar, any such food security programs should target both the urban and rural poor (Nasi et al. 2011). These programs could be instituted by utilizing existing conservation frameworks and governance approaches outlined in the literature (World Bank 2006), including Community-Based Conservation (CBC) and Integrated Conservation and Development Programs (ICDPs). CBC programs – which involve community-level governance systems in the management of natural resources and ecosystems (Clarke and Jupiter 2010) – may be an appropriate framework to consider in areas with well-organized local communities (e.g., women’s groups, fishing associations) and in areas where existing top-down conservation initiatives (e.g., many national parks) are not considered to be a viable approach to conservation and food security programming. It may be less well adapted to areas without strongly functioning social institutions at local and higher scales (Berkes 107 2004), such as in many domains in Madagascar (African Development Bank Group 2005). In contrast, ICDPs – which simultaneously conduct conservation and development initiatives (Guerbois et al. 2013) – may be more adapted to providing both conservation programming and alternative livelihoods objectives in such cases. However, ICDPs are not always successful, as the objectives of wildlife conservation and social development can sometimes be in conflict (Berkes 2004), depending on the local context (Guerbois et al. 2013). The trade-offs between these objectives must be considered if ICDPs are to be successful (Campbell et al. 2010). Finally, it should be noted that CBC programs, ICDPs, and many of the other existing frameworks for conservation programming almost always focus on small-scale, relatively rural study sites (Reuter, KER, Juhn, D., and Grantham H.S. unpublished) and typically do not explicitly consider urban populations that are removed from remnant habitats where conservation programming is actively occurring. Therefore, targeted programming in Madagascar that aims to decrease consumption of wild meat in urban areas while enhancing food security will need to modify or extend existing frameworks for action and governance approaches to these urban contexts. Explicit consideration of the scale of these programs, the context in which these programs will be implemented, the relevant stakeholders to involve and their roles in program implementation, and needs for supporting or developing institutional structures to oversee and maintain the programs over time will all also be key to selecting the correct framework approach for conservation programming and to increasing the likelihood of program success (Berkes 2004, Sewall et al. 2011). 108 CHAPTER 5 CAPTURE OF WILD ANIMALS AND MOVEMENT, TRADE, AND CONSUMPTION OF WILD MEAT IN MADAGASCAR Abstract: The consumption and trade of wild meat constitutes a threat to many animal species. Understanding how wild animals are hunted, and how wild meat is moved, traded, and consumed – the commodity chain for wild meat – can help target conservation initiatives. Past research on wild meat commodity chains has focused on the formal trade and less on informal enterprises. Informal enterprises, however may contribute to a large portion of the wild meat trade in sub-Saharan Africa, but such informal enterprises have not yet been incorporated into studies on commodity chains for wild meat. Thus, in this study, our aim was to provide a more comprehensive understanding of the commodity chain involving both formal and informal components; we used the mammalian wild meat trade in Madagascar as a case study. Our objectives were to: (1) identify hunting strategies used to capture different groups of wild animals; (2) analyze the patterns of movement of wild meat from the capture location to the final consumer; (3) examine wild meat prices, volumes, and venues of sale; and (4) estimate the volume of wild meat consumption. Data was collected in May-August 2013 using semi-structured interviews with consumers (n = 1343 households, 21 towns), meatsellers (n = 520 restaurants, open-air markets stalls, and supermarkets, 9 towns), and drivers of inter-city transit vehicles (n = 61, 5 towns). We found that: (1) a wide range of hunting methods were used, though prevalence of use differed by animal group; (2) wild meat 109 traveled distances of up to 166 km to reach consumers, though some animal groups were hunted locally (<10 km) in rural areas; (3) most wild meat was procured from free sources (hunting and receiving meat as a gift), though urban respondents who consumed bats and wild pigs were more likely to purchase those meats; and (4) wild meat was consumed at lower rates than domestic meat, though urban respondents consumed twice as much wild meat as rural respondents. Apart from the hunting stage, the consumption and trade of wild meat in Madagascar was also likely more formalized that previously thought. Introduction: Consumption of wild meat constitutes a threat to many animal species (Cowlishaw et al. 2005); up to 3.4 million tonnes of wild meat are consumed per year in Central Africa alone (Wilkie and Carpenter 1999, Fa and Peres 2001), with hunting occurring at a rate six times higher than is sustainable (Bennett 2002). A wide variety of taxa are impacted though mammals account for over two-thirds of wild animals sold through the formal trade (Bowen-Jones et al. 2003). High demand for wild meat across sub-Saharan Africa resulting from growing urban populations and increasing consumer affluence has depleted wild animal populations while causing increases in wild meat prices (Crookes et al. 2005). Attracted by higher sale prices and heightened market demand, hunters often switch to more efficient hunting methods (reviewed by Lindsey et al. 2013), begin hunting in areas more distant from urban markets (as nearby habitats become depleted of wildlife, Crookes et al. 2005), and partner with a series of formal and informal traders and other actors to transport and sell wild meat to consumers far 110 removed from the hunting location (Bennet 2002, Kumpel et al. 2010). An understanding of how wild animals are hunted, and how wild meat is moved, traded, and consumed using formal and informal enterprises –referred to as the commodity chain for wild meat – is needed to protect animals from increasing threats from unsustainable hunting (Cowlishaw et al. 2005, Kamins et al. 2011). Past commodity-chain research suggests that the links between hunters and consumers are often well developed (Bowen-Jones et al. 2003, Cowlishaw et al. 2005, Kamins et al. 2011), though the length of the chain – the number of actors who move the wild animal from its natural habitat to the consumer – can differ (Bowen-Jones et al. 2003). In general, there are four principal stages in the process of moving wild meat through the commodity chain: 1) hunting/capture; 2) transport away from the capture location; 3) exchange via sale, barter, or gifting; and 4) consumption (Figure 5.1). The first step – the hunting of wild animals or their live capture (followed by killing at a later stage) – requires low capitalization with low barriers to entry into the profession (Brown 2003, Brown and Williams 2003). Hunting for wild animals can be illegal, and can include hunting protected animals, hunting outside of game season, and/or hunting without a permit. Following capture, wild meat (body parts or the whole carcass of a wild animal) is transported from its capture location to the point of sale, barter, or gifting to the consumer. If not consumed locally, such as by the hunter or the hunter’s immediate family, wild meat can be transported hundreds of kilometers from the point of capture to the final consumer, with transportation and sale involving multiple actors at different points in the commodity chain (Cowlishaw et al. 2005, Kamins et al. 2011, Lindsey et al. 2013). For example, in Ghana, wild meat is moved from rural to urban areas using 111 hunters, wholesalers, markets traders, and café owners (Cowlishaw et al. 2005). The actors in the commodity chain may capture and move diverse animal taxa or may specialize on particular taxa (e.g., ungulates and rodents, Cowlishaw et al. 2005; fruit bats, Kamins et al. 2011). During and following transportation, an animal is sold, bartered (both for non-monetary goods and social favors), or provided as a gift to relatives or friends of those involved in the commodity chain. Sale of wild meat may be driven in part by price; wild meat is often a cheap protein source compared to domestic animals (Fa et al. 2003). However, the price of wild meat is not static or equal across taxa; it varies with the distance that it travels from its harvest location (Cowlishaw et al. 2005, Lindsey et al. 2013) and with taste preferences (Cowlishaw et al. 2005). The final step of the commodity chain – consumption of wild meat – is important for nutritional (Fa et al. 2003, Golden et al. 2011) and cultural reasons (Lindsey et al. 2013). Consumption is the most-studied stage of the commodity chain and has been examined across Africa (e.g. Wilkie and Carpenter 1999, Fa and Peres 2001), Asia (reviewed by Karesh et al. 2005), and South America (e.g. Nasi et al. 2011); the other three stages are less well understood. These steps of the commodity chain are not always sequential and may not always occur; for example, a hunter may consume his/her own catch or the sale of wild animals may occur without a transport stage (Figure 5.1). 112 Figure 5.1: Conceptual models of the commodity chain (A) and of the different sources from which a consumer can procure wild meat (B). The commodity chain (A) is not always linear and does not always include all actors; for example, in some cases, the consumer is also the hunter. In image (B) there are two tiers of arrows; the arrows that connect peripheral boxes to the “purchase” and “free” boxes represent direct quantities measured (black arrows) while the arrows connecting the “purchase” and “free” boxes to the consumer represent sums of the peripheral boxes of each type (gray arrows). A) B) 113 Past research on wild meat commodity chains have focused on Central and West Africa where chains are relatively formalized (involving a consistent set of actors and using established venues). Accordingly, this research has focused on the formal trade (Bowen-Jones et al. 2003) and has only included informal enterprises (small businesses that lack large capital investments, that may be ephemeral, or that exist outside government oversight; Benjamin and Mbaye 2014) to a limited degree (but see Solly 2007 and Kamins et al. 2011, who included “wandering vendors” and currently inactive or occasional hunters and bushmeat vendors). However, informal sectors account for over 50% of the GDP and employment in many African countries (Benjamin and Mbaya 2014); it is therefore likely that a large portion of the wild meat trade is utilizing informal enterprises. Most wild meat is illegally extracted (and in such cases is often referred to as bushmeat; Lindsey et al. 2013) and is usually part of the informal trade (Benjamin and Mbaye 2014). Similarly, the emphasis of past studies has been on measuring the flow of goods that have market value (Bowen-Jones et al. 2003), and these studies have generally excluded barter, gifts, and other non-market exchanges of wild meat. Thus, past analyses of the commodity chain for wild meat may not be fully comprehensive, given their exclusion of the informal aspects of wild meat trading. Excluding informal and non- market exchanges may lead to an incomplete understanding of the commodity chain in areas such as Madagascar that have a less organized wild meat trade, that often employ household-to-household sales (Golden 2009; Golden et al. 2014), and that barter wild meat for non-monetary goods or social favors. The prevalence of informal and non-market exchange of wild meat in Madagascar provides an opportunity to obtain a more inclusive understanding of commodity chains 114 for wild meat. Prior studies – which were based predominantly in rural areas of Madagascar or were limited to one geographic region – suggest a highly informal commodity chain lacking developed transport and exchange stages in which 98% of an individual’s wild meat is hunted by the consumer (Golden et al. 2014a) and in which hunting for wild meat is conducted for sustenance only (Gardner and Davies 2014). However, these studies have not explicitly examined the urban wild meat trade nor considered the interaction between informal and formal enterprises along different points of the commodity chain. Such an understanding could prove highly beneficial for conservation because of the global conservation value of Madagascar’s wildlife and of its mammals in particular; 92% Madagascar’s mammals are endemic (IUCN 2013) and current levels of hunting are considered unsustainable to several mammal species including lemurs (Golden 2009), fossa (Golden 2009), and bats (Cardiff et al. 2009). In addition, the political instability following a 2009 coup d’état may have increased the volume of trading of lemurs (Schwitzer et al. 2014), although it remains unclear whether this has changed the broader commodity chain for wild meat in Madagascar. Therefore, effective conservation of Madagascar’s mammals requires a better understanding of how wild mammals are captured, how the meat of wild mammals is moved and traded, and how much wild meat from mammals is consumed in Madagascar. In Madagascar – like in other countries (Bowen-Jones et al. 2003) – there are a lack of comprehensive data about the different stages in the commodity chain, and how these vary by animal group. Regarding the capture of animals, most data are anecdotal though there is evidence that diverse taxa are hunted throughout the year using a variety of methods (e.g. Vasey 1996, Goodman and Raselimanana 2003; Randrianandrianina et 115 al. 2010). There is also evidence that – despite legal protection – protected animals are being hunted (Vasey 1996) and game animals are being hunted out of season (Randrianandrianina et al. 2010). Likewise, data on the transportation of wild meat away from the hunting location are mostly anecdotal (Vasey 1996, Martinez 2008). Only one study has estimated the distance that hunters travel to hunt wild meat (4.40 ± 2.90 km; Golden 2009) while a second study approximated that in one village, tenrecs were transported 15 km from the hunting site to a market (Tucker 2007). Regarding the sale of wild meat, price estimates for wild meat range from < 1-7 USD/kg in rural areas of the country (Gardner and Davis 2014, Golden et al. 2014). Unlike in other sub-Saharan countries (Cowlishaw et al. 2005) wild meat in Madagascar is usually sold fresh and not smoked. The only study that has explicitly compared the prices of wild meat to domestic meat found that more preferred wild meats were more expensive than domestic meats and less preferred wild meats were less expensive (Randrianandrianina et al. 2010). The final step of the commodity chain – consumption – is the best understood. Past studies have shown that the consumption of wild meat is occasional but not regular. Most households (95%; Golden 2009) and most people (60%; Razafimanahaka et al. 2012) consume wild meat at least once a year though it appears in just 1.3% of meals (Jenkins et al. 2011). In all of these four stages, it is not clear how the commodity chain varies by animal taxa or between rural and urban areas in Madagascar. Our overall aim was to improve understanding of the formal and informal commodity chain for wild meat in Madagascar, including the capture of wild animals and the movement, sale, and consumption of wild meat. Our objectives were to: (1) identify the principal methods used to hunt wild animals of different groups and the timing of 116 hunting; (2) analyze the patterns of movement of different types of wild meat from the capture location to the final consumer; (3) examine the prices, volumes, and venues of sale of wild meats; and (4) estimate the volume of wild meat consumed. For objectives 2- 4, we used the movement, sale, and consumption of domestic meat – i.e., meat from animals like zebu (a subspecies of cattle found in Madagascar), chickens, and pigs that were raised by farmers or ranchers – as a point of reference for comparison with wild meat. For our first objective, we determined hunting methods and timing. We hypothesized that hunting methods would: (1A) differ between urban and rural respondents due to the different resources available to hunters in each setting; (1B) differ before and after 2009, due to reports of decreased enforcement on hunting following a political coup d’état; (1C) focus on slingshots, spears, and other less-efficient tools when wild meat was for personal consumption; (1D) focus on nets, firearms, and other highly- efficient tools when wild meat was for sale. In addition, we hypothesized that the timing of hunting would: (1E) be year-round for all animals except for tenrecs, which estivate for a portion of the year (Gardner and Davies 2014); and (1F) that rural respondents would be more likely than urban ones to report year-round hunting, given the potential of rural respondents for living in closer proximity to remnant animal habitats. For these same reasons, we also hypothesized that (1G) the proportion of consumers who directly hunted wild meat would be lower in larger towns. Our second objective was to describe the patterns of movement of wild meat from its capture location to the final consumer, including differences by location and animal group. Given that past studies from rural northeastern Madagascar indicate hunters travel 117 relatively short distances to hunt (4.4 km; Golden 2009), we hypothesized (2A) that wild meat would be sourced locally. We also hypothesized that (2B) some animal groups (such as civets and wild cats) that are attracted to human-dominated landscapes would be hunted primarily in close proximity to towns (Golden 2009), whereas (2C) other groups of animals (such as lemurs) that avoid human-dominated landscapes would only be sourced farther away from a consumer. However, because less forest habitat for wild animals may be present near large towns, we hypothesized that (2D) the distance from which wild meat would be sourced would increase with the population of a town. In addition, because domestic meat is often raised near human settlements and habitats of wild animals may be remote from towns, we hypothesized that (2E) domestic meat (either raised by interviewees, or purchased at local markets) would be sourced more locally than wild meat. Finally, given that most wild meat is caught illegally in Madagascar (Golden et al. 2014), we further hypothesized that to avoid drawing attention to illegal wild meat, (2F) the volume of wild meat transported on the inter-city transit system would be low. Our third objective was to examine the prices, volumes, and venues of sale of wild meats, relative to domestic meat. We hypothesized that due to increased distances and costs of transport of wild meat (Crookes et al. 2005) and increased costs of living in towns with higher populations, (3A) prices would be positively correlated with the population of a town. Based on the urban meat price estimates in Randrianandrianina et al. (2010), we hypothesized that (3B) bat would be cheaper than domestic meat, wild pig would be the same price as domestic meat, and that tenrecs would be more expensive than domestic meat. In addition, we hypothesized (3C) that the cost of wild meat would 118 differ depending on the source from which it was purchased. Specifically, due to profit- taking by middlemen, we hypothesized that (3D) the per-unit price paid by the consumer would increase when wild meat was not purchased directly from the hunter. In terms of the volume of sale, given that most wild meat is caught illegally in Madagascar (Golden et al. 2014), we hypothesized that to avoid drawing attention to illegal wild meat, (3E) the volume of sale of wild meat at restaurants and markets would be lower than for domestic meat and that (3F) the trade of legal domestic meat would take place more frequently through established venues than the trade of wild meat. In addition, because wild meat hunting may require specialized methods, tools, or knowledge, or access to remnant forest habitats, and because of past research suggesting substantial portions of wild meat are sold to markets and restaurants (Randrianandrianina et al. 2010), we hypothesized that (3G) most consumers of wild meat would not have hunted the wild animals themselves; rather for most animal groups and in most instances of wild meat consumption, the wild meat would be provided to the consumer by a third party. Finally, since larger towns might have less forest cover remaining nearby, we hypothesized (3H) that the proportion of wild meat purchased from a third party (rather than hunted by the consumer or purchased directly from the hunter) would increase with the population of a town. Our fourth objective was to estimate the relative volume of consumption of wild meat. Given that price (Golden et al. 2014) and preference (Randrianandrianina et al. 2010) differ by animal and impact the likelihood of the consumption, we hypothesized that (4A) the volume of consumption would differ by animal group. In addition, given that regular consumption of wild meat is low in Madagascar (2.1% of animal protein 119 procured, Randrianandrianina et al. 2010; 1.3% of meals, Jenkins et al. 2011), we hypothesized that (4B) the volume of wild meat consumption would always be lower than the consumption of domestic meat. Methods: Study site: Data were collected (May-August 2013) at twelve urban and nine rural towns in central and northern Madagascar (Table 5.1). Urban towns were located along the 1,092 km-long highway connecting the northern regional capital Antsiranana with the national capital Antananarivo, located in the center of the country. Rural towns were located around the perimeter of Ankarana National Park, a large protected area (18220 ha) that contains one of the world’s highest densities of primates (Hawkins et al. 1990). Legality of wild meat hunting and consumption in Madagascar: As in many African countries (Lindsey et al. 2013), hunting in Madagascar is regulated using legal instruments, with the right to hunt belonging to the state and allowed through a permit-based hunting system (Rakotoarivelo et al. 2011). Regulations restrict the species that can be hunted, hunting methods, and hunting seasons; they also prohibit the possession, transport, sale, and export of illegally acquired animals (Appendix E). Many people hunt without permits (Rakotoarivelo et al. 2011) and some restaurants openly sell wild meat (Jenkins et al. 2011) that may have been illegally hunted. Game animals are also hunted outside of season (Randrianandrianina et al. 2010). Violators of natural resource extraction laws sometimes avoid punishment by paying 120 bribes (Global Witness 2009), and this may also be the case with the wild meat trade. In rural communities in northeast Madagascar, 66% of wildlife biomass is hunted illegally (Golden et al. 2014). Research permissions: Research design was approved by an ethical review committee for human subjects research (Temple University Institutional Review Board), and field research was conducted under the authorization of the Madagascar Ministry of Water and Forests and Madagascar National Parks. Permission to conduct research was also gained in each town from the highest ranking, locally elected official. 121 Table 5.1: Table of study sites included in research project. The number of interviews completed at each study site are listed by interview type. The distance of population centers from Antananarivo are listed; Antananarivo was the southernmost population center included in this study while Antsiranana was the northernmost. Population estimates flagged with (*) are estimates retrieved from the Ilo (2003) database. All other estimates were provided by locally elected officials. MC: meat consumption interview; R: restaurants; M: open-air markets; SM: supermarkets; TI: transportation interviews. Number of interviews Distance from Town Meatsellers Population Antananarivo MC TI (kilometers) R M SM Antananarivo 199 1501 31 3 18 1,054,649* 0 km Ankazobe 63 - 26 - - 13,085* 92 km Mahatsinjo 58 - 4 - - 15,000* 177 km Andriba 122 - 12 - - 32,000* 198 km Antsiafabositra 70 - 5 - - 8,328* 243 km Tsararivotra 32 - 5 - - 496 km Andrevorevo 40 - - 582 km Antsohihy 60 2 17 - 1 105,317* 668 km Ambanja 55 2 23 - 8 28,468* 865 km Ambilobe 99 - 23 - 4 56,427* 962 km Ankarana National Park Ambondromifehy 30 5000 Ampasinbengy 30 1997 Andranokoho 33 N/A N/A 2000 996 km Lambondry 34 120 Mahamasina 28 650 Marotaolana 30 175 Marovato 30 400 Matzaborimanga 30 400 Tsarakibany 30 250 Aniverano Nord 90 - 15 - - 15,000* 1025 km Antsiranana 180 9 193 - 30 87,569* 1100 km Total: 1343 163 354 3 61 1 150 restaurants were approached (105 in person and 45 by phone) for information regarding their history of wild meat; 140 restaurants had never sold wild meat and provided no additional information and 10 restaurants were given a full meatseller interview. 122 Social surveys: Data were collected from the following actors of the commodity chain: 1) hunters (informal enterprises); 2) inter-city transporters (formal enterprises); 3) meatsellers (both formal and informal enterprises; open-air markets, restaurants, and supermarkets); 4) and consumers. No hunters reported obtaining a permit or regularly selling wild meat; thus all of the hunters interviewed were considered part of the informal trade. In this study, a ‘restaurant’ was any business that sold prepared foods for consumption. This included smaller Malagasy-run restaurants referred to as Gargottes or Hotely (similar to the cafés/chopbars, or places that “cook and sell bushmeat in stews” in Cowlishaw et al. 2005) as well as larger Malagasy and expatriate-owned restaurants targeting foreigners and wealthier individuals. Meatsellers involved in formal enterprises included the more expensive restaurants, some of the larger Hotely, the supermarkets, and the permanent open-air market stands, whereas meatsellers involved in informal enterprises included the smaller Hotely and the intermittent open-air market stands (e.g., those only open on some market days). Data were collected using three semi-structured interview protocols (Rietbergen- McCracken and Narayan 1998): 1) meat consumption interviews with members of the community; 2) meatseller interviews with employees at meat selling venues; and 3) transportation interviews with employees of inter-city transportation companies (bus and boat). Data from hunters were collected through the meat consumption interviews; all interviewees were asked how they sourced their wild meats (including hunting) and whether they ever sold wild meat (Questions 1, 1a, and 1c in Appendix B). All surveys were administered by research teams composed of two individuals, including one lead 123 researcher (KER, AW, HG) and one translator (TEJ, TRB). Surveys were administered orally in the interviewees’ language of choice (local Malagasy dialect or French). Face- to-face recruitment was used to enroll respondents into the study. Verbal informed consent was always received prior to the onset of data collection. No identifying information was collected to protect respondent anonymity. Meat consumption interviews were administered at select households in each town. In rural towns (range: 120-5,000 inhabitants) every 5th household was sampled. In urban towns (range: 6,622 – 1,054,649 inhabitants) random sampling was stratified by administrative unit. To ensure independence of samples, interviews were requested with one person per household. Respondents included both males and females, were 18 years or older, and considered themselves to have a major buying power for household goods. If an eligible individual refused or if no one was present, an adjacent household was approached. Meat consumption interviews lasted an average of 11 ± 0.27 minutes, and 1343 people were interviewed (Table 5.1); we had no prior knowledge of respondents’ histories of hunting and consuming wild meat. The meat consumption interviews were designed to collect data on meat consumption habits (questionnaire in Appendix B). In the first section, respondents were asked to discuss any type of meat consumption and provide information (including data on hunting, purchasing and other attainment methods) about recent (within the last three days) meat consumption and preferred meat types. In the second section, respondents were asked questions regarding their history of wild meat consumption (amount, species, preference), sources of wild meat (e.g. hunting or purchase), and the circumstances of 124 consumption. In the third section, respondents were asked about any changes in their wild and domestic meat consumption as well as meat-related taboos. Meatseller interviews were conducted in all (n = 11) urban towns with permanent food markets or restaurants (Table 5.1), and were designed to collect data on the volumes and types of meat being sold in open-air markets, supermarkets, and restaurants (Appendix C). Interviews were administered to one adult worker per establishment after face-to-face recruitment at the respondent’s place of business (e.g. at a restaurant or a food stand). In all but the two largest towns (Antsiranana and Antananarivo, see below), we interviewed all meatsellers who were available during our visit (~60-100% per town) in open-air markets at permanent stands (where meat was sold on most days) and at intermittent stands (where meat was typically sold only on weekly market days). Interviews were conducted with individuals in open-air markets without prior knowledge of their history of selling wild meat and regardless of whether they appeared to be selling wild meat at the time of the interview. In Antsiranana and Antananarivo, the large size of the towns required a modification of the sampling strategy. In Antsiranana, data was collected from randomly sampled meatsellers in every permanent open-air market in the city. In the largest town of Antananarivo, a few meatsellers were sampled randomly from major open-air markets in 14 quarters, spread evenly across the city. These quarters represented a wide range of communities, including areas of different ethnic, religious, and socioeconomic characteristics. We also opportunistically sampled three supermarkets (enclosed, upscale food stores with diverse merchandise) in Antananarivo. In all towns except Antananarivo, knowledge derived from meatseller interviews was used to determine restaurants with a prior history of selling wild meat. In 125 Antananarivo, the size of the city and number of restaurants precluded the use of interview responses as the sole source of information about restaurants selling wild meat. We therefore visited 105 restaurants across 10 quarters and made telephone contact with an additional 45 restaurants spread across the city. After obtaining informed consent, we initially asked respondents from these Antananarivo restaurants if wild meat had ever been sold at the restaurant. If the respondent indicated the restaurant had never sold wild meat (n = 140), we ended the interview. For all others (n = 10), we completed the standard meatseller interview. For all types of meatsellers in all towns, we ensured independence of samples by selecting only one worker per establishment for an interview; this was the proprietor or manager or their designee. Full meatseller interviews (including with individuals at open-air markets, supermarkets, and restaurants; excluding the phone calls and in-person visits to restaurants in Antananarivo that did not result in a full meatseller interview) lasted an average of 8 ± 0.26 minutes and we interviewed representatives from a total of 354 open-air markets, 3 supermarkets, and 163 restaurants (Table 5.1). Questions included: recent meat sales (species, volume, rate, and price of sale during the previous three days); lifetime sale of wild meat (including data on hunting, purchasing, and other attainment methods); meat-related food taboos; and changes in type or rate of meat sales over time (Appendix C). Finally, transportation interviews were conducted in a few (n = 5) of the larger towns, to understand how domestic and wild meats were moved between towns. Interviews were conducted with boat and bush taxi drivers at major bus stations and ports (Table 5.1). Transportation interviews lasted an average of 8 ± 0.01 minutes, and 61 people were interviewed (Table 5.1). Respondents were asked to quantify the amount of 126 meat transported within the last three days (including information about departure and destination points, as well as any fees associated with transportation) and their lifetime history of wild meat transportation (Appendix D). Analysis: Results are presented as means ± 95% confidence intervals and towns are replicates unless otherwise noted. Individuals were only used as replicates when the sample size of respondents was less than 20. All analyses were completed using the JMP statistical software (JMP®, Version 10. SAS Institute Inc., Cary, NC, 1989-2007). For objective one, individual respondents were used as replicates when examining the use of different hunting methods. In 16 cases, respondents were not clear whether they “shot” an animal with a slingshot or a firearm, and in other cases, some respondents used the words “gun” but described a slingshot. Therefore, to avoid mischaracterizing responses, these 16 data points were eliminated from analysis. Given the categorical nature of the data, Pearson’s Chi-squared Tests were used to test for differences among animal groups, between urban and rural respondents, before and after the 2009 coup d’état, and between commercial and subsistence hunting. For objective one, data on timing of hunting was also reported. Respondents were not asked directly about timing of hunting, but many volunteered this information in response to questions about hunting for special occasions (see questionnaire, Appendix B). Data are presented by referencing quarters of the year; each quarter of the year corresponds with a three-month time period (Quarter 1: January – March; Quarter 2: April – June; Quarter 3: July – September; Quarter 4: October – December). In most 127 towns visited, Quarters 1 and 4 usually receive the greatest amount of rain and have higher temperatures, while Quarter 3 is the coolest and has the least amount of rain, and Quarter 2 is intermediate. Only respondents that explicitly stated a time frame in months were included in these analyses. Some respondents indicated that hunting seasons spanned more than one quarter; in these cases, the quarter which was most dominant in the stated time frame was picked as the quarter in which hunting took place. For hypothesis (1G) towns were replicates and town population was natural-log-transformed prior to analysis. For objective two, when the source location of a meat was a known town or landmark located on a road, the distance between that location and a respondent’s town of residence (i.e. the place he/she was interviewed) was calculated using the shortest route along the country’s road network. When the source location of a meat was in a town and/or landmark that was not located on a road, the distance between the location and the respondent’s town of residence was calculated as the sum of the shortest route along the country’s road network to the closest point, and the direct overland distance from that point on the road to the source location (without considering walking trails that might be utilized by the respondent, because these could not be detected from the satellite images and maps used to measure source-to-town distances). When respondents listed more than one source location, distances were averaged among the different source locations for that respondent. For hypothesis (2D) a mixed effects model was used with ‘town population’ as the fixed effect, ‘town’ and ‘animal group’ as random effects, and distance (in km) as the response. For hypothesis (2E) a mixed effects model was used 128 with ‘type of meat’ (i.e. wild meat or domestic meat) as the fixed effect, ‘town’ and ‘animal group’ as random effects, and distance (in km) as the response. For objective three, prices are presented in Malagasy Ariary, where 2,197 Ariary are equivalent to 1 USD (United Nations Operational Rates of Exchange, June 1st 2013, http://treasury.un.org/operationalrates/OperationalRates.aspx). Due to low sample sizes of meatsellers selling wild meat (see results for hypothesis 3E below), we obtained few estimates of sale prices from meatsellers. Therefore, we used the prices reported by the consumers during meat consumption interviews for wild meat purchases to calculate wild meat prices at the time of the interviews (Table S4, Appendix E). Only price data provided by respondents referring to sale during 2012 and 2013 were used in analyses, to decrease inaccuracies from incorrect information recall by respondents over longer time periods. For hypothesis (3B) a mixed effects model was used with ‘type of meat’ (i.e. wild or domestic meat) as the fixed effect, ‘town’, ‘unit of purchase’ (i.e. whether the purchase was of a whole animal, kilogram of meat, or plate of meat at a restaurant), and ‘animal group’ as the random effects, and cost of meat (Malagasy Ariary) as the response. For hypothesis (3C) a mixed effects model was used with ‘purchase venue’ (i.e. open-air market, hunter, restaurant, or middleman) as the fixed effect, ‘town’, ‘animal group’, and ‘unit of purchase’ as random effects, and cost of meat (Malagasy Ariary) as the response. For hypotheses (3H) towns were replicates and town population was log- transformed prior to analysis. For objective four, hypothesis (4A) used a mixed effects model with ‘animal group’ as the fixed effect, ‘town’ as a random effect, and consumption frequency (number of times consumed per year) as the response. Hypothesis (4B) used a mixed 129 effects model with ‘type of meat’ (i.e. wild or domestic meat) as the fixed effect, ‘town’ and ‘animal group’ as random effects, and consumption frequency (number of times consumed per year) as the response. Results: Capture/hunting of wild animals (objective one): Respondents to the meat consumption interviews (n = 1343) across all towns reported 1463 records of hunting animals, of which 84% provided information about the hunting method(s) used. The most popular hunting method included using a dog (33%, individuals are replicates), following by: catching with a trap/net (21%); with their hands or with other people (17%); a slingshot/throwing a rock (11%); stabbing or hitting with a stick/spear (10%); using a firearm (3%); cutting down the tree in which the animal was hiding (2%); digging up the animal’s burrow (1%); and using a blow tube or spit tube with darts (< 1%). Hunting methods differed significantly by animal group (Pearson’s Chi-Squared Test, χ2 = 1257.331, p < 0.0001; Figure 5.2). Lemurs were the only animals that were hunted by cutting down the entire tree; this hunting strategy was often associated with nocturnal lemurs. Tenrecs were the only animals hunted by digging up their burrows. In accordance with hypothesis (1A), hunting method differed between urban and rural respondents (Pearson’s Chi-Squared Test, χ2 = 215.107, p < 0.0001; Figure 5.2). Some methods/tools were reported exclusively by respondents living in urban areas: blow tubes (n = 8), firearms (n = 32), cutting down trees (n = 29), and digging up burrows (n = 14). Some animal groups were hunted using different methods when urban respondents 130 hunted them, as opposed to rural respondents, including: 1) bats (Pearson’s Chi-Squared Test, χ2 = 100.833, p < 0.0001); 2) civets (Pearson’s Chi-Squared Test, χ2 = 14.190, p = 0.03); 3) lemurs (Pearson’s Chi-Squared Test, χ2 = 37.745, p < 0.0001); 4) tenrecs (Pearson’s Chi-Squared Test, χ2 = 40.512, p < 0.0001); 5) and wild pigs (Pearson’s Chi- Squared Test, χ2 = 55.224, p < 0.0001). Sample sizes were too small to test fossa and mongoose, and there was no difference in how wild cats were hunted. There was mixed support for hypothesis (1B). When all hunting reports were aggregated, the type of hunting method differed in the pre-2009 period and the 2009 to mid-2013 period (Pearson’s Chi-Squared Test, χ2 = 64.455, p < 0.0001). However, when examining changes within groups of people (e.g. urban respondents who hunted bats) – with a few exceptions - there was no significant shift between periods in hunting strategies for most groups (Figure 5.3). Urban respondents who had hunted lemurs did use a different array of hunting methods pre-2009 than those who hunted from 2009 to mid-2013 (Pearson’s Chi-Squared Test, χ2 = 16.014, p = 0.03; Figure 5.3) as did urban and rural respondents who hunted tenrecs (urban: Pearson’s Chi-Squared Test, χ2 = 12.761, p = 0.01; rural: Pearson’s Chi-Squared Test, χ2 = 17.449, p = 0.002). Figure 5.2: Hunting methods used to capture different types of wild mammals by urban (1A) and rural (1B) respondents. Individuals are replicates. Hands = catching with hands by one or multiple people. Trap/Net = includes nets, cable traps, bait with traps. Spit Tube = a blow tube or a device to shoot darts. Stick = includes hitting animal with stick or spearing animal with stick. Individuals are replicates. Other = includes other hunting methods (e.g. fire, using a cross bow). A) 100% Cut down tree Dig up burrow 80% Dog 60% Hands Firearm/"Shoot" 40% Slingshot/Stone Spit Tube 20% using hun ng method S ck Percentage of respondents 0% Trap/Net Bats Civets Fossa Lemurs Mongoose Tenrecs Wild Cats Wild Pigs Other 131 132 B) 100% Cut down tree Dig up burrow 80% Dog Hands 60% Firearm/"Shoot" 40% Slingshot/Stone Spit Tube 20% S ck using hun ng method Percentage of respondents Trap/Net 0% Other Bats Civets Fossa Lemurs Mongoose Tenrecs Wild Cats Wild Pigs 133 Figure 5.3: Hunting methods used to capture different types of wild mammals in the past (pre-2009) and more recently (2009- 2013) for (A) urban-based hunters and (B) rural hunters. Only animal groups with ≥5 respondents in each time class are presented. Individuals are replicates. Hands = catching with hands by one or multiple people. Trap/Net = includes nets, cable traps, bait with traps. Spit Tube = a blow tube or a device to shoot darts. Stick = includes hitting animal with stick or spearing animal with stick. Individuals are replicates. Other = includes other hunting methods (e.g. fire, using a cross bow). A) 100% Cut down tree Dig up burrow 80% Dog 60% Hands 40% Firearm/"Shoot" Slingshot/Stone 20% Spit Tube using hun ng method S ck Percentage of respondents 0% Bats Bats Civets Civets Lemurs Lemurs Tenrecs Tenrecs Trap/Net pre-2009 2009-2013 pre-2009 2009-2013 pre-2009 2009-2013 pre-2009 2009-2013 Other 134 B) 100% Cut down tree Dig up burrow 80% Dog 60% Hands 40% Firearm/"Shoot" Slingshot/Stone 20% Spit Tube using hun ng method S ck Percentage of respondents 0% Bats Bats Civets Civets Lemurs Lemurs Tenrecs Tenrecs Trap/Net pre-2009 2009-2013 pre-2009 2009-2013 pre-2009 2009-2013 pre-2009 2009-2013 Other 135 Figure 5.4: Comparison of hunting methods used to capture wild mammals for sale and for personal consumption. Individuals are replicates. Hands = catching with hands by one or multiple people. Trap/Net = includes nets, cable traps, bait with traps. Spit Tube = a blow tube or a device to shoot darts. Stick = includes hitting animal with stick or spearing animal with stick. Individuals are replicates. Other = includes all other hunting methods (e.g. slingshots, fire, using a cross bow). 100% Dog 90% 80% Hands 70% Firearm/"Shoot" 60% 50% S ck 40% Trap/Net hun ng method 30% 20% Other 10% Percentage of respondents using 0% Bats Bats Tenrecs Tenrecs Wild Pig Wild Pig (for sale) (personal) (for sale) (personal) (for sale) (personal) Forty-three individuals across ten urban and four rural towns had caught wild meat for commercial purposes (8 ± 3% of people in towns where at least one individual had caught wild meat for sale). Most individuals had only caught and sold one type of wild meat, though five (12% of n = 43) had caught and sold two different types of wild meat, thereby providing 48 records of wild meat hunting for commercial reasons. In accordance with hypotheses (1C, 1D) there was a difference between the hunting methods used when an animal was caught for personal consumption or opposed for sale (Pearson’s Chi-Squared Test, χ2 = 109.342, p < 0.0001; Figure 5.4). Though sample sizes were too small to examine the differences in hunting methods within individual animal groups, some animals captured for resale tended to be caught using more efficient methods than animals captured for personal consumption. Bats caught for personal consumption were caught using nets proportionately less (Figure 5.4), than bats caught for resale. Likewise, tenrecs captured for personal consumption were caught using a variety of methods including dogs, hands, traps, sticks, slingshots, and by digging up burrows, though tenrecs caught for sale were almost always captured using dogs (Figure 5.4). Of the 1463 hunting records, 259 referenced hunting during a specific time of the year (e.g. an informal or formal hunting season) for at least one type of mammalian animal group. In accordance with hypothesis (1E), tenrecs were the only group of animals where over 10% of both urban and rural respondents (towns as replicates) mentioned a hunting season (Figure 5.5). However, contrary to hypothesis (1F), both urban and rural respondents reported hunting seasonally at similar proportions for each animal group (Figure 5.5). Most respondents reported hunting animals in the second and 136 137 fourth quarter of the year (Figure 5.6). Although we did not ask meatsellers whether hunting seasons impacted the sale of meat, one meatseller at an open-air market in Antsiranana and three restaurants in Antananarivo noted that the months of June-August (3rd quarter of the year) were the season when wild pig meat was easier to procure and sell. Finally, contrary to hypothesis (1G), the percentage of consumers who hunted wild meat did not change with the population of a town (Regression, F1,16 = 0.71, Slope = -4.93, p = 0.41, R2 = 0.04, towns as replicates with population log-transformed). Movement of wild and domestic meat (objective two): The 1343 respondents of the meat consumption interview provided the necessary information on 1879 records of wild meat consumption and 1234 records of domestic meat consumption, to calculate the distances traveled during the procurement of meat by consumers. Contrary to hypothesis (2A), wild meat was not often sourced from areas near a respondent’s town of residence (Figure 5.7A), though animals were often hunted locally (<10 km distance) in rural regions (Table S8, Appendix E). Wild meat sold in urban markets and restaurants tended to be sourced from areas close to those establishments (within 55 kilometers; Table S9, Appendix E), with consumers often reporting longer travel distances to these types of establishments when purchasing it (Table S8, Appendix E). Regarding hypotheses (2B, 2C), we did not detect evidence that some animal groups were hunted in closer proximity to towns than others (Table S8, Appendix E). Figure 5.5: Percent of respondents who reported hunting wild mammal groups on a seasonal basis. Towns are replicates. 35 30 25 20 15 hunting seasons hunting 10 5 Percent of respondents reporting reporting Percentrespondents of 0 Bat Civet Fossa Lemur Mongoose Rat/Mice Tenrec Wild Cat Wild Pig Animal Group 138 Figure 5.6: Time of year when respondents hunt for different types of wild mammals. Individuals are replicates. Sample sizes are listed for each animal group. Quarter 1: January-March; Quarter 2: April-June; Quarter 3: July-September; Quarter 4: October-December. 1 Bat (n=23) 0.9 Civet (n=9) Lemur (n=7) 0.8 Tenrec (n=111) 0.7 Wild Cat (n=3) 0.6 Wild Pig (n=9) 0.5 0.4 0.3 Propor on of respondents 0.2 0.1 0 Quarter 1 Quarter 2 Quarter 3 Quarter 4 Time (quarters of a year) 139 Figure 5.7: Distances that meat was transported or that urban and rural consumers traveled: to (A) procure wild meat as a gift, from hunting, or purchase; and (B) to purchase domestic and wild meat from restaurants, markets, hunters, and middlemen. The whiskers extend to data points within 1.5 times the interquartile range; black circles indicate outliers. Towns are replicates. ND: No Data; UW: Urban seller of wild meat; UD: Urban seller of domestic meat; RW: Rural seller of wild meat; RD: Rural seller of domestic meat. A) 140 141 B) 142 However, in accordance with hypothesis (2D), the distance that consumers traveled to purchase wild meat from any source (including restaurants, markets, hunters, and middlemen) increased significantly with the population of a town (Mixed Effects 2 Model, F1, 18.35 = 7.42, population(log) = 89.77, p = 0.01, model R = 0.49), as did the distance that respondents would travel to hunt wild meat (Mixed Effects Model, F1, 20.99 = 23.78, population(log) = 52.45, p < 0.0001, model R2 = 0.13). In accordance with hypothesis (2E), domestic meat was purchased more locally than wild meat (Mixed Effects Model, F1, 4.621 = 50.12, Type of meat = 54.87, p = 0.001, model R2 = 0.17). This was true in both urban and rural regions and for different types of venues where wild meat would be purchased (Figure 5.7B). Most transporters (65 ± 25%, towns as replicates) had transported meat in the three days prior to their interview. Of these, 2.7 ± 5.2% – all of whom (n = 4) were interviewed in Antsiranana – had transported wild meat (including bats, tenrecs, and wild pigs) and in the three days prior to the interview and 34 ± 27% had transported wild meat before. All transporters who had moved wild meat in the past three days had also transported domestic meat as well. In accordance with hypothesis (2F), the percentage of drivers who had ever transported chicken (62 ± 24%) was higher than for four other types of wild meat that they had transported in the past (≤ 24% on average, Table S10, Appendix E). Though the frequency of transport was lower for wild meat than domestic meat, when wild meat was transported, the number of individuals transported per trip were similar to that of a shipment of domestic animals (Table S10, Appendix E). 143 Sale of wild and domestic meat (objective three): Prices of wild and domestic meat, as reported by meatsellers (markets, restaurants, and supermarkets) are listed in Table S9 and prices as reported by consumers are listed in Table S11 (Appendix E). In accordance with hypothesis (3A), in 7 out of 9 cases where comparable purchase were made in rural and in urban areas (e.g. the purchase of one unit of wild meat from the same type of source, as reported by consumers), the price paid in urban areas was higher than the price paid, on average, in rural areas (Table S11, Appendix E). Notably, the two instances where this was not the case were in purchases of wild pig (Table S11, Appendix E). In contrast to hypothesis (3B), the costs of wild and domestic meat reported by consumers did not differ for: wild pig (Mixed Effects Model, F1, 1.99 = 0.036, cost of meat 2 = 334.52, p = 0.87, model R = 0.13), tenrec (Mixed Effects Model, F1, 2.54 = 6.96, cost of 2 meat = -4490.70, p = 0.09, model R = 0.11), lemur (Mixed Effects Model, F1, 3.05 = 1.34, 2 cost of meat = -1914.91, p = 0.33, model R = 0.11), and bats (Mixed Effects Model, F1, 2 2.05 = 10.89, cost of meat = -4211.02, p = 0.08, model R = 0.11). In accordance with hypothesis (3C), the cost of wild meat differed depending on whether it was purchased in an open-air market, from the hunter, at a restaurant, or from 2 a middleman (Mixed Effects Model, F3, 44.69 = 2.89, p = 0.046, model R = 0.35). In accordance with hypothesis (3D), the price paid by a consumer for wild meat increased when it was not purchased directly from the hunter (Table S11, Appendix E). For tenrecs purchased in urban areas, the lowest prices were offered by hunters, with higher prices required from middlemen, markets, and restaurants for the same unit of purchase (Table S11). Likewise, for bats purchased in urban areas, the lowest price was offered by 144 hunters, middlemen offered an intermediate price, and markets/restaurants offered the highest price for the same unit of purchase (Table S11). In accordance with hypothesis (3E), in open-air markets (which were randomly sampled without prior knowledge of wild meat sales history), the percent of meatsellers who had sold one of three different types of domestic meat in the three days prior to the interview was higher (≥ 18%) than the percent who had sold any one type of mammalian wild meat (≤ 1%; Table S9, Appendix E). In accordance with hypothesis (3F), the patterns of movement of domestic and wild meat (Figures S1-S3) from the source to the consumer may reflect their status as a legal, semi-legal (can be legally hunted some of the time), or illegal substance (Figure 5.8). Domestic meat is often obtained through purchase or raised by the consumer rather than being provided as a gift (Table S12, Appendix E), though urban respondents tend to purchase domestic meat from markets while (apart from zebu meat) rural respondents tend to own and raise domestic meat (Figure S2). In contrast, the patterns of procurement for mammals that are commonly consumed as wild meat are less clear but include a variety of means including purchasing and hunting (Figure S3). In further contrast, animals that are not commonly consumed or which are typically killed through human- wildlife conflict (e.g. carnivores that killed livestock or wild pigs that trampled rice fields) were almost exclusively consumed after being killed by the consumer (Figure S4). There was little support for hypothesis (3G). For most animal groups, a higher percentage of respondents (town as replicates) procured meat from free sources (e.g. by hunting or receiving the meat as a gift) than through purchase from a third party (Table S13, Appendix E). The exception to this was urban respondents; a higher percentage of 145 urban respondents purchased bat and wild pig meat than procured it from free sources (Table S13, Appendix E). In contrast to hypothesis (3H), the population of a town had no effect on the percent of consumers who bought wild meat (Regression, F1,16 = 0.74, Slope = 4.33, p = 0.40, R2 = 0.04, towns as replicates with population log-transformed). Volume of meat consumption (objective four): The 1343 respondents to the meat consumption interviews provided 1046 estimates of yearly consumption rates for nine different types of wild meat and 1282 estimates of yearly consumption rates for three different types of domestic meat. Urban respondents estimated that they had consumed wild meat 86 ± 56 (mean ± 95% CI, towns as replicates) times in their lifetime and 10 ± 6 times per year in the recent past (from 2009-2013). Rural respondents estimated that they had consumed wild meat 117 ± 122 times in their lifetime and 5 ± 5 times per year in the recent past (Table S14, Appendix E). Contrary to hypothesis (4A), the frequency of wild meat consumption per year did 2 not differ by animal group (Mixed Effects Model, F7, 1034 = 1.75, p = 0.09, model R = 0.10), with all wild animal groups being reportedly eaten on average ≤18 times per year during the 2009-2013 period in both urban and rural areas (Table S14, Appendix E). In accordance with hypothesis (4B), domestic meat was consumed more frequently per year than wild meat in this same period (Mixed Effects Model, F1, 9.48 = 15.62, p = 0.003, model R2 = 0.30), with urban and rural regions consuming domestic meat an average of 51 ± 11 and 41 ± 17 times per year, respectively. 146 Figure 5.8: Patterns of movement for wild and domestic meats as reported by urban and rural respondents. The arrows point in the direction of the consumer and the thickness of the arrow is proportional to the percentage of consumers (towns are replicates) who procured domestic meat from that source. When no arrow is present, no respondents reported procuring meat from that source. There are two tiers of arrows; the arrows that connect peripheral boxes to the “purchase” and “free” boxes represent direct quantities measured (black arrows) while the arrows connecting the “purchase” and “free” boxes to the consumer represent sums of the peripheral boxes of each type (gray arrows). Urban Rural All wild meat All domestic meat Discussion: This is the first study to comprehensively examine the commodity chain for wild meat in Madagascar. It is also the first study to compare aspects of the Malagasy wild meat trade across rural and urban areas, between different types of actors (e.g. hunters, transporters, vendors, and consumers), including both formal and informal enterprises. We examined how the trade of wild meat differed by animal group and from domestic meats across the four stages of the commodity chain. As a result of these findings we hypothesize that there are strong differences between urban and rural respondents in their interactions with the wild meat commodity chain. Our findings also suggest that, apart from the hunting stage, which in our meat consumption interviews was fully informal, the commodity chain for wild meat in Madagascar is likely more formalized than previously thought. Our results further suggest that this commodity chain includes robust transport and exchange stages, also in contrast to previous understanding. The commodity chain also varied between wild meat and domestic meat, by wild meat type, and for urban versus rural consumers. These results and their implications are discussed in more detail, below. Hunting and capture of wild meat: Similar to previous reports (e.g. Vasey 1996, Irwin et al. 2000, Goodman and Raselimanana 2003) a wide range of hunting methods were used to capture wild animals. The use of dogs while hunting was reported by a third of respondents, though hunting methods differed by animal group and between urban and rural respondents (Figure 5.2). For example, only urban respondents reported using blow tubes, firearms, cutting down 147 148 trees to capture nocturnal lemurs, and digging up burrows to catch estivating tenrecs. These results are supported by anecdotal reports that the use of firearms in Madagascar to hunt animals (specifically, lemurs) is somewhat limited to those who can to afford to own them. For example, in two cases, it was reported that firearms were loaned to some hunters by third-party “entrepreneurs” and traders (Duckworth et al. 1995, Vasey 1996), while a third report noted that wealthier, urban individuals would go on weekend hunting trips in forests and would use their rifles to hunt (Rigamonti 1996). Our data – which show the wide range of hunting methods used by both subsistence and commercial hunters – expand on prior studies from the country. For example, in eastern Madagascar, Jenkins et al. (2011) estimated that most lemurs were hunted with firearms but noted that their survey design may not have detected other types of hunting strategies, such as traps. Our data provide further evidence that lemurs are hunted using firearms, but also suggest that the use of firearms is only one of methods used (Figure 5.2). We did not detect evidence that the 2009 political coup d’état has changed how people hunt most taxa of wild animals, despite anecdotal reports of increased natural resource extraction (of timber and animals, Barrett and Ratsimbazafy 2009; lemurs, Schwitzer et al. 2014). However, lemurs were the exception; following the 2009 coup d’état, urban respondents were more likely to capture lemurs using traps or nets, slingshots, and dogs. In addition, they were less likely to capture lemurs by cutting down the trees in which the lemurs were resting and blow tubes were no longer used (Figure 5.3). It is unclear why hunting strategies changed, though it may be linked to widespread awareness campaigns that were instituted specifically targeting lemurs following the coup (e.g. Schwitzer et al. 2014). It may also be because nocturnal lemurs – which are 149 typically being targeted when individuals cut down trees to capture lemurs – have become rare and are no longer as accessible to urban hunters that visit remnant habitats sporadically. Our data confirm that urban and rural respondents participate in the hunting of wild meat in markedly different ways. The hunting method utilized by individuals differed on the basis of the purpose of hunting and may differ between urban and rural respondents. For example, animals captured for sale were caught using more efficient methods than those captured for personal consumption (Figure 5.4) and animals captures for sale were also frequently caught by urban individuals (93% of self-reported commercial hunters were urban respondents). Therefore, it is reasonable to hypothesize that urban respondents might generally hunt for commercial reasons (although hunting for pleasure, such as in Rigamonti 1996, or when visiting family in rural areas may also be important motivators). In contrast, rural respondents might be more likely to hunt for subsistence (Golden et al. 2014) or to protect their agricultural lands and livestock from nuisance animals, such as wild pigs and carnivores (Golden 2009). However, there are exceptions to these generalizations and rural respondents also do occasionally use more efficient hunting methods (Vasey 1996, Goodman and Raselimanana 2003) and also sell their wild meat catches when they have caught surplus animals (Golden et al. 2014). Movement of wild animals away from the capture location: In agreement with Golden (2009) many animal groups were hunted locally (<10 km) in rural areas. However, we also found the distance that respondents traveled to hunt wild meat increased with the population of a respondent’s town of residence. Urban 150 individuals typically traveled farther than rural respondents to hunt wild meat (Figure 5.7A); the average distance traveled by urban individuals to hunt was always greater than 70 km from their town of residence, on average (Table S8, Appendix E). The distances that commercial wild meat (wild meat that was purchased by the end consumer) traveled from the capture location to the consumer depended on the vendor type (Figure 5.7B, Table S8). When the meat was delivered to the consumer by a middleman (who would bring the wild meat to a respondent’s town in a car or on a bicycle), the middleman traveled from 0 km to 166 km to reach the consumer (Table S8). When the meat was delivered to the consumer via an open-air market or a restaurant, the wild meat was typically sourced from within 50 km (Table S9) of the establishment. However, in these cases – and depending on the animal group – urban and rural consumers traveled between 119 and 477 km, on average, to purchase meat from restaurants and between 0 to 198 km to purchase meat from open-air markets (Table S8). The distance traveled by consumers to purchase wild meat increased with the population of a town. This makes sense given that consumers attempting to purchase wild meat might have to travel farther to areas where wild meat is sold (perhaps to regions closer to remnant habitat). Alternatively, this large travel distance may also be because some types of wild meat were consumed using expendable income and may have been consumed for a special occasion, such as during an overnight visit with family members or on a business trip (Reuter, K.E., Gilles, H., Wills, A.R., and Sewall, B.J., unpublished). While these distances are large – and while transporters reported moving wild meat up to 389 kilometers, on average – the commodity chain of Madagascar seems to cover the same 151 distances as those examined in other areas of sub-Saharan Africa (Cowlishaw et al. 2005, Kamins et al. 2011). Comparing the transport patterns of wild meat to domestic meat can help highlight how the legal status of different wild meats may affect their commodity chains. Domestic meat was purchased more locally than wild meat in both urban and rural areas and was transported more regularly using inter-city transit systems. In contrast, few drivers (2.7 ± 5.2%) had transported wild meat in the three days prior to the interview, and all who had, were from the northern-most town of Antsiranana. These low levels of wild meat transport cannot be explained by inter-city transit exclusively transporting only a few types of meats; drivers transported various meats regularly and only one driver mentioned not having the legal paperwork needed to move meat. Yet, not all wild animal groups were transported using this system. First, carnivores had never been transported by any of the drivers interviewed perhaps because they are generally caught due to human-wildlife conflict (interactions between people and wild animals resulting in negative impacts on people and their resources or on wild animals and their habitats; Goodman and Raselimanana 2003, Madden 2004; they are legal to catch if they are a threat to agriculture or livestock, Appendix E). Second, lemurs were very rarely transported, perhaps because they are illegal to capture and hunt, and the illegal nature of these activities are relatively well known (KER, pers. obs.). Third, bats, tenrecs, and wild pigs – which are legal to capture and to sell under certain conditions and which were the most frequently consumed animals in our respondent pool (Table S14) – were most frequently transported. These data suggests that: 1) animals like carnivores, that are typically caught as a byproduct of human-wildlife conflict, are not typically moved far 152 from the capture location; 2) some wild meats – like from lemurs, which are illegal to hunt – are probably being moved from the capture location but potentially in private vehicles due to the illicit nature of the trade; and 3) some wild meats – like from bats, tenrecs, and wild pigs – are commonly moved via the country’s intercity transit system (drivers estimated that they transported hundreds to thousands of wild animals per year). Exchange of wild meat via sale, barter, or gifting: With the exception of urban respondents who consumed bat and wild pig meat, more respondents procured wild meat from free sources (e.g. by hunting or receiving the meat as a gift) than through purchase from a third party. However, many respondents did still purchase meat, and town population size did not affect the percent of customers in a town who purchased wild meat. Wild meat is typically consumed when it is a cheaper protein option than domestic and farmed meats (Fa et al. 2003, van Vliet et al. 2012). We found that the cost of wild meat did not differ from domestic meat, even when accounting for portion sizes and the meat source (e.g. restaurant or market). This contrasts with prior studies in Madagascar, which reported that some types of wild meat were more expensive than domestic meat (Tucker 2007, Randrianandrianina et al. 2010). In addition, we found that wild meat was more expensive in urban areas than in rural areas and – similar to other wild meat trade markets in Africa (Cowlishaw et al. 2005) – increased in cost when it was purchased from a third-party and not directly from a hunter. In comparison to domestic meat, wild meat was sold at only a few open-air markets (Table S9) and restaurants (usually only a few in each city). Usually, where these 153 market stalls and restaurants were established enterprises, they were well known in the communities for their status as serving wild meat. Market stalls – regardless of whether they domestic or wild meat - usually specialized in the sale of one or two types (taxonomic groups) of meat. In contrast, restaurants – regardless of whether they sold wild meat or not - usually offered at least three different types of meat per week. Consumption of wild meat: In accordance with prior studies (Randrianandrianina et al. 2010, Jenkins et al. 2011, Gardner and Davies 2014) we found that wild meat was not consumed as frequently as domestic meat (Table S14). However, we found that there were marked differences between urban and rural respondents in the consumption of wild meat; urban respondents consumed 24% more domestic meat per year, but 100% more wild meat per year than their rural counterparts (Table S14). We hypothesize that the motivations for consuming wild meat differ between urban and rural populations. Given that urban consumers travel farther distances to hunt than rural respondents – and still hunt at similar levels to rural respondents – and given that they also purchase some meats (bat and wild pig) even though the prices are not cheaper than domestic meats (and are more expensive than those paid by rural respondents), we hypothesize that consumption of wild meat in urban areas of Madagascar is largely driven by a preference for those meats. In contrast, we hypothesize that consumption in rural areas is driven by a mixture of preference for wild meats (rural respondents also purchase wild meat, even though it is not cheaper than domestic meat), due to human-wildlife conflict (when animals, like carnivores, are captured in traps 154 aimed at protecting livestock), or for subsistence. Our conclusions are consistent with other studies that suggest that the hunting and consumption of wild meat in rural Madagascar provides substantial economic value and food security to households (Gardner and Davies 2014, Golden et al. 2014) while in urban Madagascar, personal preference impacts consumption and the most preferred types of wild meats are also those that are the most expensive (Randrianandrianina et al. 2010). Notably, it is conceivable that the urban poor would also consume wild meat to fulfill their food security needs – a trend that has been seen elsewhere in sub-Saharan Africa (van Vliet et al. 2012) – but in Madagascar, it is unlikely that these individuals would have the resources required to travel to rural areas (where wild animals might still exist in remnant habitats) or have the financial capabilities to purchase wild meat in urban markets. Conservation implications: Our data indicate that the transport and exchange of wild meat in Madagascar are more commercial and more formalized than previously thought. Our data also highlight how interconnected and interdependent the informal and formal aspects of the commodity chain are; hunting appears to be mostly informal while other actors involved in the commodity chain include both formal and informal enterprises. Prior studies – which were based predominantly in rural areas of Madagascar or were limited to one geographic region – suggested that 98% of an individual’s wild meat was hunted by the consumer (Golden et al. 2014) and that hunting for wild meat is conducted for sustenance only (Gardner and Davies 2014). However, our data indicate that while hunting is still an important source of wild meat, a large portion of the wild meat trade is moving through 155 third-party actors; for example, urban respondents purchased 56% and 62% of the bats and wild pigs prior to consumption, while rural respondents purchased 28% and 31% of the bats and lemurs that they consumed (Table S13). In addition, a significant portion of wild animals in Madagascar are moved across long distances using the intercity transit system and sold through venues known for their status as ‘wild meat selling establishments’. These venues include market stalls that specialize in selling wild meat (bats, tenrecs, and wild pigs) most days of the week. Given: 1) the distance traveled by consumers to purchase wild meat (Table S8); 2) the well-known status of individual restaurants as ‘wild meat-selling entities’ in each town where data was collected; and 3) that < 1% of meatsellers at open-air markets sold wild meat in the three days prior to our interviews (Table S9), it is reasonable to assume that the majority of the volume of wild meat trading in Madagascar is either hunted by a consumer (Gardner and Davies 2014, Golden et al. 2014) or sold through a small number of specialized market sellers and/or restaurants who are known to offer wild meat to consumers. The small number of dedicated wild meat sellers may explain why prior researchers hypothesized that the Malagasy wild meat trade is not as formalized as in other areas of Africa (Golden 2009). However, these few dedicated meat sellers may still be selling large amounts of wild meat, and similar to other wild meat commodity chains in sub-Saharan Africa (Cowlishaw et al. 2005) we find that the system of moving wild meat from rural areas to select urban venues using inter-city transport across large distances seems to be relatively organized. Together, the formal and informal commodity chain in Madagascar appears capable of moving thousands of wild meat carcasses per year at minimum from the point of capture to the final consumer (Table S10). 156 Our data show that some animal groups face different threats from the wild meat trade than others. For example, while the consumption of bats decreased in both urban and rural areas following the 2009 coup d’état, the meatsellers interviewed by us indicated that they likely sold hundreds of bats per year; the threat to bats from the wild meat trade comes from both informal and illegal trading and the formal and legal trade. It has been noted that game species – including bats – are poorly protected by the law, given that knowledge of wildlife laws protecting game species is lower than for other animals (Keane et al. 2011). In contrast, lemurs – which continue to be consumed in both rural and urban settings – are hardly ever traded through fixed venues (such as markets and restaurants) and it is therefore difficult to make any estimates regarding the volume of the trade of this animal group. We found little evidence for the luxury trade of lemurs which was widely reported following the coup d’état in 2009 (Barrett and Ratsimbazafy 2009); we could not locate a single restaurant that currently sold lemur meat and most of the consumer reports of lemur consumption at restaurants were at least five years old. Therefore, we conclude that the threats facing lemurs from the wild meat trade likely do not typically involve restaurants and markets; instead, the threats come from the informal and illegal trade. Understanding the commodity chain can help identify points along the chain for conservation interventions (Bowen-Jones et al. 2003, Cowlishaw et al. 2005). During the first stage of the commodity chain – capture and hunting – conservation initiatives in Madagascar would have to be rather broad in their target audiences as hunting is mostly informal (e.g. most hunters do not have hunting permits, Jenkins et al. 2011, KER pers. obs.) but can be legal and illegal, depending on the animal group (Appendix E). It is 157 unlikely that hunting bans or increased enforcement will result in the desired effect of a reduction in unsustainable hunting (Bowen-Jones et al. 2003) given that most hunting in Madagascar is done illegally already (Golden et al. 2014). However, voluntary monitoring or citizen science programs could be instituted. For example, locally based monitoring programs in sub-Saharan Africa that worked with wild meat hunters to collect monitoring data, were found to be a cost-effective and accurate method of data collection (Rist et al. 2010). At a minimum, these would help inform managers of current capture rates and help provide information for how programs could help influence the demand for wild meat at other points in the commodity chain. In addition, it can help inform community-based conservation programs (Berkes 2004) or Integrated Conservation and Development Programs (Guerbois et al. 2013); for example, information collected from hunters and communities in northeast Madagascar has led to the development of alternative livelihoods programs which aim to decrease natural resource use within the boundaries of a protected area while providing hunters with more sustainable economic opportunities (Golden 2009; Golden et al. 2011). The second stage of the commodity chain – transportation – provides some opportunities for conservation programming. Some animals groups – including bats, tenrecs, and wild pigs – were often transported by inter-city transit (Table S10) and were sometimes even delivered directly to restaurants for sale to the final consumer (both actors in the legal, formal trade of wild meat). These meats could be monitored using existing legal frameworks that require transporters and restaurants to have permits to move and sell wild meat (Appendix E), although enforcement and monitoring at transit hubs would need to be increased. Given that inter-city transit operations did not 158 exclusively move wild meat (and wild meat transport did not constitute the major source of income for any driver interviewed), and already did not transport some illegal meats (like lemurs), drivers would likely cooperate with such a program. However, conservation initiatives would need to be prepared for the possibility that increased enforcement but unchanging consumer demand could increase the informal transportation of wild meat. The third step of the commodity chain – the sale, barter, or gifting of wild meat – also presents some opportunities for conservation planning. Fixed establishments, such as these restaurants and markets – many located in urban areas – are prime opportunities for the implementation of various management initiatives (Cowlishaw et al. 2005), in part because in our study a high proportion of wild meat appears to have been sold through a few well-known vendors in each locality. Animal groups – like bats, tenrecs, and wild pigs – that are traded through formal enterprises including restaurants and markets could be effectively monitored by working with meatsellers using routine inspections. Food inspections are undertaken by the government in Madagascar, and may increase in the near future (Sarter et al. 2010); wild meat monitoring could be included in these inspections as a matter of course. Working with informal enterprises is more difficult but could involve restaurants and meat-sellers registering themselves as a wild meat-selling venue with different levels of voluntary and mandatory reporting requirements, depending on the types and quantity of wild meat being sold. Voluntary programs, where restaurants pledge not to sell certain types of foods (e.g. sustainable seafood; www.fish2fork.com) have been used by non-profit and multilateral groups in the past; these public pledges could be used in Madagascar to leverage the desire for tourists and 159 other entities to support environmentally-friendly businesses. In addition, monitoring systems could provide early warning signals for the potential declines of wild populations; a permit system could be used to keep legal harvests sustainable; and increased enforcement could be used to protect species that are not legal to sell. Finally, the last step of the commodity chain – consumption by the consumer – presents opportunities for conservation programs aimed at curbing demand for species that are traded illegally or through informal venues, or where most of the hunting is conducted by the consumer. For these animal groups – like lemurs – public education and outreach might be the most effective way to decrease wild meat hunting consumption (Breuer and Mavinga 2010). It has been noted that the Malagasy public has low knowledge of wildlife laws and only 43% of individuals could correctly classify different species into their legal categories (e.g. protected, game, and nuisance animals; Keane et al. 2011). Therefore, outreach to the public could include information about: 1) specific species and their legal status; 2) types of hunting methods that are prohibited; 3) seasons in which hunting is legal; and 4) permitting procedures. These outreach campaigns must consider that in urban areas, wild meat consumption in Madagascar may be driven by preference rather than by food security needs and that human-wildlife conflict may also be a factor contributing to hunting in rural areas (Reuter, K.E., Gilles, H., Wills, A.R., and Sewall, B.J., unpublished). Therefore, successful outreach campaigns could appeal to the consumer in a way that acknowledges and leverages cultural and social norms – a tactic that has been successful in increasing community compliance in a marine reserve in southwest Madagascar (Westerman and Gardner 2014) – while providing alternatives to individuals who are hunting mammals due to human-wildlife conflict (Breuer and 160 Mavinga 2010) or for food security reasons (Golden 2009). Since compensation for damages caused by wildlife does not always succeed at protecting biodiversity, the use of compensation should be carefully considered prior to implementation (Bulte and Rondeau 2007). Finally, outreach in urban areas must be especially cognizant that, relative to rural consumers, urban consumers are willing to pay higher prices for wild meat, travel farther to procure wild meat, and utilize different hunting methods (such as firearms). In some cases, there is no single best point along the commodity chain for conservation interventions (Cowlishaw et al. 2005), and when considering conservation interventions, it is important to consider all aspects of the commodity chain as targeted programming can have unintended consequences on other stages of the commodity chain (Bowen-Jones et al. 2003). For instance, the provision of alternative protein sources to communities can decrease hunter income and cause hunters to increase their hunting or switch hunting strategies to higher value and possibly rarer species (to make up for lost income; Bowen-Jones et al. 2003). The political ecology of the commodity chain – the relationships and political agendas of different stakeholders and their impact on conservation programming – should also be considered as this can greatly impact the success of management and policy interventions (Berkes 2004). To be effective, conservation programs will need to engage, empower, and support local communities (Berkes 2004); when the local context is carefully considered and when appropriate frameworks and governance approaches are selected, conservation programming can succeed without sacrificing social development programming (Campbell et al. 2010). 161 CHAPTER 6 IMPACTS OF HABITAT CHANGE ON THE DIETS AND VERTICAL STRATIFICATION OF FRUGIVOROUS BATS Abstract: Human-modified habitats are expanding rapidly, and many tropical countries have highly fragmented and degraded forests. Preserving biodiversity in these areas involves protecting species – like frugivorous bats – that are critical to forest regeneration. Though fruit bats provide critical ecosystem services including seed dispersal and pollination, the study of how their diets are affected by habitat change has often been limited to localized studies. This study used stable isotope analysis to examine how foraging by three fruit bat species in Madagascar, Pteropus rufus, Eidolon dupreanum, and Rousettus madagascariensis, are impacted by habitat change across a large spatial scale. Our results indicated that the three species had broadly overlapping diets. Differences in diet were nonetheless detectable and consistent between P. rufus and E. dupreanum, and these diets shifted when they co-occurred, suggesting resource partitioning across habitats and vertical strata within the canopy to avoid competition. Changes in diet were also correlated with a decrease in forest cover, though at a larger spatial scale in P. rufus than in E. dupreanum. These results suggest fruit bat species exhibit differing foraging strategies in response to habitat change. They also highlight the key threats that fruit bats face from habitat change, and clarify the spatial scales at which conservation efforts should be implemented to mitigate threats for these bat species in Madagascar. 162 Introduction: Anthropogenic changes to tropical forests have been extensive, with more than 350 million hectares of forest removed and an additional 500 million hectares degraded globally (Lamb et al. 2005; Wright 2005). These changes have led to clear impacts on tropical biodiversity (reviewed by Fahrig 2003; Lambin et al. 2003), including population declines, local extirpation, and global extinction of native species of mammals (Kinnaird et al. 2003). Further, even where mammal populations persist, habitat change may have additional, more subtle effects. For instance, anthropogenic changes in tropical forest habitats may result in a changed abundance or distribution of food resources, leading to modification of mammal foraging behavior (Cosson et al. 1999). Anthropogenic habitat changes affecting the foraging behavior of fruit bats may have ecological consequences that are particularly important. Although intact forest patches are often preferred by fruit bats over degraded and agricultural ones (Mildenstein et al. 2005), fruit bats do consume non-native food sources (Sewall et al. 2003), and can forage outside intact forests (Mildenstein et al. 2005), sometimes at high feeding densities (Luskin 2010). Because of their key roles as pollinators and seed dispersers of many species of tropical trees (Kunz et al. 2011), changes in bat foraging patterns in response to anthropogenic habitat change may have a disproportionate influence on tree reproduction and regeneration in tropical degraded landscapes (Gorchov et al. 1993; Medellin and Goano 1999; Kunz et al. 2011). Studies of bat foraging behavior to date have primarily used a standard set of methods: cafeteria trials on captive bats (e.g., Bravo et al. 2010), pollen or seed sampling from captured bats (e.g., Soto-Centeno and Kurta 2006), field examination of feeding 163 refuse (e.g., Sewall et al. 2003), analyses of guano (e.g., Bollen and Van Elsacker 2002), direct field observations of free-ranging bats (e.g., Sewall et al. 2013), and radio telemetry (e.g., Mildenstein et al. 2005). Each of these methods have some limitations, however, because they: 1) investigate foraging under artificial circumstances (cafeteria trials); 2) provide data representing foraging over a short time frame (pollen or seed sampling from captured bats, field examination of feeding refuse, analyses of guano) or small spatial scale (direct field observations); or, 3) indicate where a bat traveled but not directly what was consumed there or even if consumption occurred (radio telemetry). Such limitations of standard methods for studying bat foraging behavior could result in an incomplete picture of bat foraging, especially since bats may readily shift their diets to different food sources when food availability changes (Jenkins et al. 2007). Analytical techniques employing stable isotopes can provide new avenues to advance understanding of shifts in bat foraging behavior resulting from habitat change (Dammhahn and Goodman 2014). For example, stable isotopes (δ15N and δ13C) have been used in bats to identify when diets are switched from primarily C3 plants to primarily C4 plants (δ15N values are lower in C3 than C4 plants; Voigt and Matt 2004), to hypothesize the presence of commercially grown fruits in diets (δ15N may be higher in agricultural soils due to nitrogen fertilizers; Herrera et al. 2008), and to determine the direction of seed dispersal in bats that forage in forested and degraded areas (δ13C is higher in fruit seeds in successional sites than the primary forest; Voigt et al. 2012). In addition, δ13C isotope values have been used in bats to assess vertical stratification – the tendency of different species to feed at different heights in the canopy – among species with overlapping ranges (δ13C values decrease the higher an animal forages in the 164 canopy; Rex et al. 2011). Finally, variation in stable isotopes can also be used to examine diet breadth (a wider range of stable isotope values in a sample of bats indicates a wider diet breadth at the population level; Scanlon et al. 2013). Stable isotope analyses are also particularly advantageous for studying fruit bats, whose diet may vary widely over time and space. First, many fruit bats may be sequential specialists, exhibiting a pattern of targeting one or a few key fruit or flower resources at a time, then shifting to new resources over time in accordance with tree phenology (Mickleburgh et al. 1992). Such foraging patterns have complicated the implementation of diet studies; data may need to be collected over an entire year or more to understand the complete annual diet using direct observational methods, for instance. Depending on the tissues examined, however, stable isotope analyses can provide insight into bat foraging over time periods long enough to encompass such temporal shifts. For instance, carbon isotopes represent foraging over weeks to years, depending on the species and the type of tissue sample being considered (Crawford et al. 2008). Second, fruit bats may travel quickly, forage in ranges over tens of kilometers, and may encounter and forage in diverse habitats (Garbutt 2007). As a result, diet may be incompletely sampled in studies that cover small areas. However, stable isotope analyses of a bat’s tissues can provide insights into its foraging across its range, not only in a single location. Further, when combined with a sampling regime that accounts for multiple individuals in different populations across large geographic areas, stable isotope analyses could be used to increase our knowledge of how dietary composition, vertical stratification, and diet breadth shift at the population level in response to large-scale habitat change. 165 Advances in understanding the role of large-scale habitat change on the foraging behavior of fruit bats could prove particularly valuable for conservation in Madagascar, a biodiversity hotspot where more than 80% of original forest cover has been lost (Schwitzer et al. 2014) and where much of the remaining forest habitat is fragmented or degraded (Irwin et al. 2010). Madagascar’s fruit bat community is comprised of three species – Pteropus rufus E. Geoffrey 1803, Eidolon dupreanum Schlegel and Pollen 1866, and Rousettus madagascariensis Grandidier 1928 – that have broad, overlapping ranges (Andriafidison et al. 2006) and often roost and forage in locations outside of protected areas where they encounter extensive habitat modification (MacKinnon et al. 2003; Jenkins et al. 2007; Long and Racey 2007; Cardiff et al. 2009). These species are capable of traveling up to an average of 30 km in one night (Olesky et al. 2015) and – in contrast to a few other Old World fruit bats that transit among several roosts over wide areas or migrate seasonally (Tidemann & Nelson 2004, Ossa et al. 2012) - these species are not known to travel exceptionally long distances (Olesky et al. 2015). These fruit bats are known to be important seed dispersers and pollinators (Baum 1995; Bollen and Van Elsacker 2002), and could therefore play an important role in maintaining remnant primary forests and in regenerating secondary forests on human-modified lands. However, all three species have declining populations and are of conservation concern as Vulnerable (P. rufus and E. dupreanum) or Near Threatened (R. madagascariensis) species (IUCN 2013). Despite their importance for forest conservation, however, these three bats remain poorly understood (Goodman et al. 2005), and to date no studies have compared their foraging habits across large spatial scales. 166 The aim of this study was to compare foraging among Malagasy fruit bats in light of the large-scale habitat changes occurring in Madagascar. The specific objectives were to examine how fruit bat diets varied (1) by species and (2) across space as a result of changes in forest cover. For objective one, we hypothesized that, although all three species are primarily frugivorous (Garbutt 2007), their diets would differ. Specifically, in accordance with previous evidence of differences in the composition and abundance of food items in the diet of the three species (Andriafidison et al. 2006, Sewall et al. 2013), we expected among-species differences in diet would be stronger than within-species differences. Similarly, we hypothesized that diet breadth would differ significantly between species; we expected that, in accordance with previous studies of single fruit bat species, P. rufus would have the most diverse diet (40 and 100 plant species, respectively, in Bollen and Van Elsacker 2002; Long and Racey 2007), R. madagascariensis would have the second most diverse diet (47 species, Andrianaivoarivelo et al. 2011), and E. dupreanum would have the least diverse diet (30 species, Picot et al. 2007). We further hypothesized that the bats would exhibit resource partitioning such as via vertical stratification when one or more species were present in the same area. In particular, based on prior field observations (Andriafidison et al. 2006), we expected that stable isotope values would be consistent with P. rufus feeding at higher vertical strata than E. dupreanum. For objective two, due to potential differences in resource availability among sites, we hypothesized that the bat diets would differ spatially. Habitat type and quality varied across our study sites and so, given that these species consume different types of fruit resources in degraded areas (Jenkins et al. 2007; Long and Racey 2007), we 167 expected that their diets would vary at both regional and local spatial scales. We further hypothesized that variation in the bat diets would be driven by forest loss in neighboring areas. Specifically, because of the different types of fruit resources consumed by bat species in degraded areas (Jenkins et al. 2007; Long and Racey 2007), and because of the long nightly distances they fly (Garbutt 2007), we hypothesized that their diets would correlate with levels of forest cover, especially when examined at spatial scales close to the maximum nightly flight distances of these bats. Methods: Study Site: Data were collected from June to early August 2013 in and near six towns in central and northern Madagascar, located along the Route National 4 and Route National 6 roads (Table 6.1, Figure 6.1). The towns ranged from 45 to 425 km apart, and varied in distance from the coast and from protected areas (Figure 6.1). Four of the towns were near known roosting sites for at least one of the bat species (MacKinnon et al. 2003). Bat Hair Samples: Samples for stable isotope analyses were small hair cuttings retrieved from bats. No bats were captured for this study. Rather, samples were collected from carcasses of bats that had already been captured and killed by Malagasy hunters and that were destined for sale. Hair was used for stable isotope analysis rather than tissue samples because the slow turn-over rate of hair results in stable isotopes that integrate diet 168 information over a long period (several months, Crawford et al. 2008), and because of the lack of reliable electricity or drying facilities at study sites. Samples were obtained from bat carcasses from hunters or from public or well- known locations of wild and domestic meat sale (Table 6.1). Some limited hunting of bats is legal in Madagascar (Jenkins et al. 2007), and all three bat species are sold through the wild meat trade (Jenkins and Racey 2008). We focused sampling efforts on the wild meat trade to facilitate efficient collection of a large number of samples across a large spatial scale. This sampling procedure could have introduced some bias if hunters targeted areas for hunting that were not representative of typical foraging areas or if hunters prioritized some bat species over others. Such bias is considered to be minimal here, however, since hunters have an economic incentive to capture bats where they are most abundant and because hunters often use indiscriminate hunting techniques such as large nets that capture any bat species present (Jenkins and Racey 2008). 169 Figure 6.1: Locations of cities, indicated by black circles, where wild hair samples were collected. Dark gray regions denote protected habitat, black lines indicate roads, and the inset indicates the location of the study region on the island of Madagascar. Table 6.1: Towns where hair samples were collected, with hunting sites for each town. Town (hunting Number of Town Number of individuals sampled sites listed under meat sellers populationb a P. rufus E. dupreanum R. madagascariensis each town) enrolled in study Andriba 1 32,000 Mangasoavina - 1 - Antsiafabositrac 2 8,328 4 - - Antsohihy 1 105,317 - - Ambaliha 12 - - Ambodimanany 1 - - Amboroho 2 - - Ambilobe 1 56,427 Ambakiarano - 11 - Amborondolo 30 - 1 Beramanja 15 - - Isesy 15 2 - Mahivoragno - 16 - Mamoro 3 - - Anivorano Nordd 6 15,000 6 20 - Antsiranana 4 87,569 Akonokono 15 2 - Andranofanjava 16 - - Daraina 7 - 5 French Mountain 7 - - Mangoaka 5 - 1 Total: 15 - 138 52 7 aHunting sites are those reported by hunters and meat sellers in each town; this may not be an exhaustive list of hunting sites for each town. bTown population estimates were taken from the Ilo (2003) database. cIn Antsiafabositra the hunting site or sites were unknown, so we do not differentiate by hunting site. dIn Anivorano Nord, bats were hunted only at one hunting site but the name of this hunting site is unknown. 170 171 Providers of wild meat hair included hunters, restaurant owners and merchants at food market stands (hereafter these are collectively referred to as ‘hunters and meat sellers’). Sampling was undertaken in a manner intended to provide neither cost nor benefit to meat sellers. Specifically, with one exception, meat sellers were not paid for their participation, though they were reimbursed for any phone credit used to contact the research team. In Antsiranana, one meat seller was reimbursed a nominal amount (between $0.50-1.00 per sample for 17 samples; the usual market price of one bat was $1.00-2.50) to compensate for a reduction in sale price of the bat meat due to removal of hair samples. In Antsiranana, Anivorano Nord, Andriba, and Ansiafabositra, hair samples were also collected from bats after direct contact with hunters. These hunters were identified through a related research project (Reuter et al. 2015) and volunteered to assist the research project without compensation. Samples of ventral fur were collected from each bat carcass by a trained member of the research team within 24 hours of bat capture. Samples were obtained by cutting ventral fur with sterilized scissors, then storing this fur in labeled 1.5 mL plastic, centrifuge tubes. Care was taken to avoid contamination by blood. Hair samples were also collected opportunistically from carcasses of other mammals being sold by the same meat sellers to provide the first, or some of the first, reported values of stable isotopes for those species (Table S15, Appendix F). 171 172 Stable Isotope Analysis: For each bat individual sampled, 0.3 - 0.7 mg of hair were washed in ethanol and packaged in tin capsules for mass spectrometry (following Yohannes et al. 2008). Samples were analyzed using a Costech elemental analyzer interfaced with a continuous flow Micromass (Manchester, UK) Isoprime isotope ratio mass spectrometer (EA-IRMS) for 15N/14N and 13C/12C ratios (following Yohannes et al. 2008). Measurements are reported in δ notation (per mil (‰) units) and were calculated using the following formulae: δ13C = [(13C/12C)/(13C/12C of Peedee Belemnite) – 1] * 1000 And: δ15N = [(15N/14N)/(15N/14N of atmospheric nitrogen) – 1] * 1000 Ovalbumin was used as a routine standard. Precision for δ13C and δ15N was generally ± 0.2 and ± 0.4 ‰. Social Surveys: Meat sellers and hunters provided information about the geographic area in which the bat was hunted (hereafter, ‘hunting site’). In some cases, meat sellers from one town sourced bats from more than one hunting site (Table 6.1). Precise location data for reported hunting sites were recorded by taking GPS coordinates during visits to hunting sites with hunters, or by extracting coordinates from maps or satellite images on the basis of locations or landmarks indicated by the hunter or meat seller. 172 173 Ethical Research Statement: Research design, including the recruitment of hunters and meat sellers into the study, was approved by an ethical review board (Temple University Institutional Review Board, Protocol Number: 21414, May 2013), as was the collection of hair samples from wild meat (exempted study, Temple University Institutional Animal Care and Use Committee). Research was conducted under the authorization of the Madagascar Ministry of Water and Forests (Permit number: 071/13/MEF/SG/DGF/DCB.SAP/SCB, May 18 2013). In addition, permission to conduct research was gained in each town from the highest ranking, locally elected official. Hair samples were exported from Madagascar (Permit number: 141N_EA07/MG13, August 9th 2013, Madagascar Ministry of Environment and Forests), and were declared to the U.S. Fish and Wildlife Service at the New York City (JFK airport) port of entry. Analysis: For objective one and the first part of objective two, dietary differences among species and across space were analyzed using a Permutational ANOVA and pairwise post-hoc tests (Primer statistical software, Anderson 2001) of Bray-Curtis similarity, a measure that examines relationships among replicates in combined δ13C and δ15N values. Pairwise differences in δ15N and in δ13C values were analyzed separately using a non- parametric Wilcoxon Test. A Wilcoxon Test was also used to test for pairwise differences in dietary breadth; standard deviations of stable isotope values at different hunting sites were used as proxies for diet breadth. Non-parametric Kruskal-Wallis Rank Sums Tests were used to examine spatial variation among towns (a proxy for regional variation) and 173 174 hunting sites (a proxy for local variation). Wilcoxin and Kruskal-Wallis Rank Sums Tests were completed with JMP statistical software (JMP®, Version 10. SAS Institute Inc., Cary, NC, 1989-2007). Unless otherwise noted, averages are shown as the mean ± standard deviation. Sample sizes for R. madagascariensis were too low to include in many of the analyses. The alpha level for significance was set at 0.05, but given the difficulties of detecting strong differences when sample sizes are small, marginally significant findings (p ≤ 0.1) are also presented. For the second part of objective two, the primary predictor variables of interest were those related to forest cover, which was calculated from satellite imagery from the USGS Earth Explorers database (US Dept. of Interior and US Geological Survey 2014), using images taken by the Landsat 8 satellite during August - October 2013. These dates were selected because they were close to the time frame when hair samples were collected, but also when landscapes in the images were not obscured by cloud cover (dry season period). To evaluate the influence of the spatial scale of local habitat change on bat diets, satellite images were clipped to three different radii (5 km, 15 km, and 30 km) using ArcGIS (ESRI 2011); 30 km has been estimated as the maximum likely foraging radius for P. rufus which has the largest nightly flight distance of the three species (Garbutt 2007, Oleksy et al. 2015). Clipped images were processed using the CLASlite program (version 3.1, Asner et al. 2009), which is a software program used to classify tropical landscapes into forest and non-forest cover (Allnutt et al. 2013). For these analyses, forest cover was conservatively defined as a pixel that had no more than 5% bare ground and no less than 85% live vegetation (30 meter spatial resolution; Asner et 174 175 al. 2009). Post-processing, raster files were analyzed for their percent forest cover using ArcGIS (ESRI 2011). To evaluate whether changing stable isotope values could result from habitat or geographic differences within the study area, we evaluated to what extent dominant vegetation near hunting sites varied and stable isotope values shifted across our study area. Through the use of countrywide vegetation maps developed between 2003 and 2006 (based on Landsat analysis and an extensive ground-truthing effort, Moat and Smith 2007), we determined dominant vegetation within a 30 km radius of each hunting site in our study. This evaluation indicated that one forest type - Western Dry Forest - was the dominant non-degraded forest type (Table S16, Appendix F) near all hunting sites from which bats were caught. In addition, to determine to what extent changes in stable isotope values across our study area are due to geographic differences alone, we compared stable isotope values from leaves of a fruiting tree species that occurs across this region, the mango Mangifera indica. We did not find evidence that the δ13C and δ15N stable isotope values changed in M. indica across the study area (Table S17, Appendix F). Together these findings support the idea that changes in stable isotope values from samples in our study area are due to factors other than changes in habitat type or simple geographic variation. To control for the sometimes-strong effect of climate on stable isotope values (Hobson 1999), we also examined the annual temperature range of each site. This variable was used as a proxy for local climate surrounding a hunting site due to its strong correlation with other key climate variables, including annual precipitation levels, 175 176 precipitation seasonality, and temperature seasonality (P < 0.05 and |r| > 0.8 for all pairwise comparisons at hunting sites). Annual temperature range was calculated by clipping the WorldClim (Hijmans et al. 2005) raster file of temperature data from 1950- 2000 to a 30 km radius of each hunting site, and averaging across all pixels included within the clipped radius. Forest cover and climate data are presented by hunting site in Table S16 (Appendix F). The importance of forest cover on bat diets was analyzed using an information theoretical approach employing model estimation and selection. For P. rufus, we identified a set of candidate models including forest cover, climate, and interactions between forest cover and climate (Table 6.2). Individual bats were replicates, with hunting site as a random effect. In addition, to facilitate comparison against E. dupreanum, we also separately evaluated models for P. rufus using hunting sites as replicates. Because of small sample sizes in E. dupreanum, only models with one predictor variable were considered, and hunting sites were treated as replicates. Models were then ranked separately for each species on the basis of the corrected Akaike’s Information Criterion (AICc), which is adjusted for small sample size (Hurvich and Tsai ® 1989). AICc was determined with JMP statistical software (JMP , Version 10. SAS Institute Inc., Cary, NC, 1989-2007). Delta AICc (ΔAICc) and Akaike weights (wi) were also calculated for each model for comparison purposes (Burnham and Anderson 2002). We identified the best model as the model with the lowest AICc and, given the conceptual similarity of candidate models, only considered others with substantial support (ΔAICc < 2). 176 177 Results: Sampling Effort: Hair samples from the three bat species (n = 138 for P. rufus, n = 52 for E. dupreanum, and n = 7 for R. madagascariensis) were taken from the six towns and linked via social surveys to 17 hunting sites (Table 6.1). Samples from Wild Cat (Felis silvestris), Wild Pig (Potamochoerus larvatus), Greater Tenrec (Setifer setosus), Common Tenrec (Tenrec ecaudatus), and Small Indian Civet (Viverricula indica) were also collected opportunistically; results for these species are provided in Table S15 (Appendix F). Fifteen meat sellers and hunters provided the hair samples (Table 6.1). Variation Within Individuals of Each Species: Individual bats varied in their δ13C and δ15N values when multiple hair samples were taken from the same individual. Individual variation in δ13C values (different hair samples from the same individual bat) was highest in P. rufus (range of standard deviations: 0.02 – 1.86 ‰, n = 5), followed by E. dupreanum (0.05 - 0.87 ‰, n = 5), and R. madagascariensis (0.09 – 0.27 ‰, n = 4). Similarly, δ15N values were also most variable in individual P. rufus, intermediate in E. dupreanum, and least variable in R. madagascariensis (P. rufus – 0.33 – 2.21 ‰, n = 5; E. dupreanum - 0.13 – 0.76 ‰, n = 5; R. madagascariensis - 0.10 – 0.41 ‰, n = 4). Species Differences in Diet, Diet Breadth, and Resource Partitioning: 177 178 Diets: In accordance with our hypothesis, diets (as indicated by Bray-Curtis similarity of δ13C and δ15N values) differed significantly between species (PERMANOVA, Pseudo-F = 12.995, df = 2, p = 0.001), with P. rufus having different stable isotope values than both E. dupreanum (Pairwise PERMANOVA, t = 4.6284, p = 0.001) and R. madagascariensis (Pairwise PERMANOVA, t = 2.4271, p = 0.015). Of the three bat species, E. dupreanum had the highest δ13C values while R. madagascariensis had the highest δ15N values (Figure 6.2). 11 P. rufus E. dupreanum 10 R. madagascariensis 9 N 15 8 δ 7 6 5 -23.5 -23 -22.5 -22 -21.5 -21 -20.5 -20 -19.5 δ13C Figure 6.2: Stable isotope values by species. Stable isotope values by species with darker points showing the means ± SD (towns as replicates) of the towns (lighter colors). Sample size for include: E. dupreanum (n = 4 towns); P. rufus (n = 5); R. 178 179 madagascariensis (n = 2). Samples sizes from within each town (e.g. number of individuals sampled in each town) can be found in Table 6.1. Diet breadth: In contrast to our hypothesis, diet breadth (variation in a local population’s stable isotope levels, as indicated by the standard deviation of both δ13C and δ15N values) did not differ between P. rufus and E. dupreanum. The standard deviations of both δ13C and δ15N values did not differ between the two species (δ13C values: Wilcoxin Test, DF = 1, Chi-square = 0.1968, p = 0.6573; δ15N values: DF = 1, Chi- square = 0.7020, p = 0.4021; hunting sites as replicates). Sample sizes were too small to include R. madagasariensis in analyses. Resource partitioning: In accordance with our hypothesis, stable isotope values were consistent with resource partitioning – such as by vertical stratification – by P. rufus and E. dupreanum. This resource partitioning was apparent in two lines of evidence. First, resource partitioning was evident in differences between the two species where they co-occurred within the same region, as evidenced by sale of both species in the same town. Specifically, within the three towns where both P. rufus and E. dupreanum were hunted and data were available, P. rufus had lower or marginally lower δ15N values than E. dupreanum in two towns (Wilcoxin Tests; Ambilobe: DF = 1, χ2 = 2.9825, p = 0.0842; Anivorano Nord: DF = 1, χ2 = 7.8370, p = 0.0051; Antsiranana: Wilcoxin Test, DF = 1, χ2 = 0.1449, p = 0.7035; individual bats as replicates; Figure 6.3A) and lower δ13C values than E. dupreanum in all three towns (Wilcoxin Tests; Ambilobe: DF = 1, χ2 = 16.5102, 179 180 p < 0.0001; Anivorano Nord: DF = 1, χ2 = 4.0333, p = 0.0446; Antsiranana: DF = 1, χ2 = 4.1864, p = 0.0407; Figure 6.3B). Second, resource partitioning was evident in the different diets of P. rufus and E. dupreanum when examined at a finer scale, namely at particular hunting sites where the two species were captured together (at same site by the same hunter), compared to where they were captured separately. At hunting sites where P. rufus was caught alongside E. dupreanum, δ13C values (Wilcoxon Test, Chi-square = 15.9890, p < 0.0001; hunting sites as replicates) and δ15N values (Wilcoxon Test, Chi-square = 5.8887, p = 0.0152) were significantly higher than in areas where they were caught alone (Figure 6.4). Likewise, E. dupreanum had significantly higher δ13C and δ15N values when caught alongside P. rufus than when caught alone (Wilcoxon Test, Chi-square = 7.2807, p = 0.0070 and Chi-square = 10.6752, p = 0.0011, respectively; Figure 6.4). R. madagascariensis was always caught alongside P. rufus, so it was not possible to determine whether it shifted its resource use in the presence of other species. Regional and Local Differences and Impact of Forest Cover and Climate on Bat Diets: Regional changes in diet: In accordance with our hypothesis, E. dupreanum diets (as indicated by Bray-Curtis similarity of δ13C and δ15N values) differed significantly between towns (PERMANOVA, Pseudo-F = 5.4108, df = 3, p = 0.02), though P. rufus diets did not (PERMANOVA, Pseudo-F = 2.2652, df = 4, p = 0.17). For P. rufus, the changes occurred only in δ13C values (Kruskal-Wallis Rank Sums Test, DF = 4, Chi- square = 24.9607, p < 0.0001), and not in δ15N values (Kruskal-Wallis Rank Sums Test, 180 181 DF = 4, Chi-square = 1.6031, p = 0.8082). In contrast, for E. dupreanum, changes in diet between towns occurred for both the δ13C and δ15N values (Kruskal-Wallis Rank Sums, DF = 3, Chi-square = 8.9802, p = 0.0296; Chi-square = 14.7172, p = 0.0021, respectively). 181 182 A) B) Figure 6.3: Comparison of (a) δ15N and (b) δ13C values for P. rufus and E. dupreanum at all towns where both were hunted. P. rufus had (A) lower or marginally lower δ15N values at two of three sites and (B) lower δ13C values at all three towns where E. dupreanum samples were also available. Individual bats are replicates. Significant (*, p < 0.05) and marginally significant (§, p < 0.10) differences between the two species at a town are indicated, as are box-and-whisker plots and outliers. 182 183 9 P. rufus (Overlap) 8.5 P. rufus (No overlap) 8 E. dupreanum (Overlap) 7.5 E. dupreanum (No overlap) N 7 15 δ 6.5 6 5.5 5 -24 -23.5 -23 -22.5 -22 -21.5 -21 -20.5 -20 -19.5 13 δ C Figure 6.4: Changes in diet (δ15N and δ13C values) when P. rufus and E. dupreanum were caught at hunting sites alone (start of arrow) and when they were caught alongside each other (end of arrow). When caught in the presence of another species, both P. rufus (squares) and E. dupreanum (triangles) significantly shifted in their δ15N and δ13C values. Hunting sites are replicates in this analysis and the values for individual hunting sites are shown on the graph for both species and at all hunting sites where they were either found alone (‘no overlap’) or where their ranges overlapped (‘overlap’). The two species were caught alongside each other at 3 hunting sites (n = 36 P. rufus individuals and n = 24 E. dupreanum individuals), E. dupreanum was caught alone at a further 3 sites (n = 28 individuals) and P. rufus was caught alone at 11 hunting sites (n = 102 P. rufus individuals). More information on sample sizes can be found in Table 6.1. Data show means ± SD. 183 184 Local changes in diet: Diets also differed by hunting site. P. rufus diets (PERMANOVA, Pseudo-F = 5.4141, df = 13, p = 0.001), as did E. dupreanum diets (as indicated by Bray-Curtis similarity, of δ13C and δ15N values) differed significantly by hunting site (PERMANOVA, Pseudo-F = 3.8804, df = 5, p = 0.037). For P. rufus, δ13C values differed significantly by hunting site (Kruskal-Wallis Rank Sums Test, DF = 15, Chi-square = 76.3385, p < 0.0001) as did the δ15N values (Kruskal-Wallis Rank Sums Test, DF = 15, Chi-square = 57.0826, p < 0.0001). Likewise, for E. dupreanum, these changes by hunting site occurred for both δ13C and δ15N values (Kruskal-Wallis Rank Sums Test, DF = 5, Chi-square = 12.4342, p = 0.0293; Chi-square = 15.3215, p = 0.0091, respectively). Influence of forest cover on diet: There was some support for our hypothesis, as the diets of both P. rufus and E. dupreanum were correlated with forest cover, albeit at different scales. After controlling for climate, the diet of P. rufus was best explained by forest cover at the largest spatial scale examined. Specifically, δ13C and δ15N values for P. rufus were best explained by a combination of climate (annual temperature range on average during 1950-2000) and remnant forest cover within 30 km of their hunting site; the best model for δ13C also contained the climate-forest cover interaction (Table 6.2). For δ15N, alternate models with different forest cover radii (5 and 15 km) also had some support. Climate was included in all top-ranked models for both isotopes in P. rufus. Further examination of the best models for P. rufus revealed that the influence of the forest cover-climate interaction was strong for δ13C values (mixed effects model; forest 184 185 cover within 30 km = 0.014, p = 0.42; annual temperature range = -0.022, p = 0.15; interaction = -0.004, p = 0.012; model R2 = 0.41), but that the forest cover did not influence δ15N values (mixed effects model; forest cover within 30 km = -0.026, p = 0.32; annual temperature range = -0.006, p = 0.68; model R2 = 0.27). When these models were analyzed using hunting sites as replicates, δ13C and δ15N values were best explained by models with a single predictor variable. Models with the single variable of annual temperature range explained stable isotope variables the best, although models containing the single predictor of forest cover at all three radii also received some support (ΔAICc < ~2; Table 6.2). Unlike in P. rufus, climate was a poor predictor of δ13C values in E. dupreanum. The diet of E. dupreanum was also correlated with forest cover, though generally at smaller spatial scales than P. rufus. E. dupreanum diets were best explained by forest cover within 15 km (δ13C values; Table 6.2) or 5 km (δ15N values) of the hunting site. For δ15N, alternate models with larger forest cover radii or climate also had some support. Further examination of the best models for E. dupreanum indicated forest cover had a significant influence on δ13C values (Regression; forest cover within 15 km, p = 0.0060; model R2 = 0.88) but not on δ15N values (Regression; forest cover within 5 km, p = 0.1185; model R2 = 0.50). Sample size was too low to analyze the impacts of climate and forest cover on R. madagascariensis diet. 185 186 Table 6.2: Results of model selection, listing the corrected Akaike’s information criterion (AICc), the ΔAICc, and the Akaike weight (wi). The predictor variables included: T = annual temperature range; FC5 = forest cover within 5 km radius of hunting site; FC15 = forest cover within 15 km radius of hunting site; and FC30 = forest cover within 30 km radius of hunting site. Fixed Effects AICc ΔAICc wi P. rufus – δ13C values (individuals as replicates, ‘hunting site’ as random effect) T + FC30 + T * FC30 355.58 0 0.45 T + FC30 357.95 2.38 0.14 T + FC15 358.11 2.53 0.13 T + FC5 358.75 3.17 0.09 T + FC15 + T * FC15 358.87 3.30 0.09 T 359.09 3.51 0.08 T + FC5 + T * FC5 362.56 6.98 0.01 FC30 366.15 10.57 0.00 FC15 366.25 10.68 0.00 None (intercept only) 367.07 11.49 0.00 FC5 367.07 11.49 0.00 P. rufus – δ15N values (individuals as replicates, ‘hunting site’ as random effect) T + FC30 506.71 0 0.36 T + FC5 508.18 1.47 0.17 T + FC30 + T * FC30 508.32 1.61 0.16 T + FC15 508.54 1.83 0.14 T 509.29 2.58 0.10 T + FC15 + T * FC15 511.68 4.97 0.03 T + FC5 + T * FC5 511.80 5.09 0.03 FC30 524.06 17.35 0.00 FC5 526.00 19.29 0.00 FC15 526.29 19.58 0.00 None (intercept only) 526.89 20.18 0.00 P. rufus – δ13C values (hunting sites as replicates, no random effects) T 36.39 0 0.28 FC15 37.10 0.71 0.20 FC30 37.86 1.47 0.14 FC5 37.91 1.52 0.13 T + FC30 + T * FC30 38.17 1.78 0.12 T + FC15 40.11 3.73 0.04 T + FC15 + T * FC15 40.46 4.07 0.04 T + FC5 41.04 4.65 0.03 186 187 T + FC30 41.09 4.70 0.03 T + FC5 + T * FC5 45.07 8.68 0.00 P. rufus – δ15N values (hunting sites as replicates, no random effects) T 39.03 0 0.35 FC30 39.79 0.77 0.24 FC5 40.91 1.87 0.14 FC15 41.16 2.13 0.12 T + FC30 42.66 3.64 0.06 T + FC5 43.48 4.46 0.04 T + FC15 43.72 4.70 0.03 T + FC30 + T * FC30 45.63 6.60 0.01 T + FC15 + T * FC15 46.79 7.76 0.01 T + FC5 + T * FC5 47.02 7.99 0.01 E. dupreanum – δ13C values (hunting sites as replicates, no random effects) FC15 18.49 0 0.75 FC30 21.03 2.54 0.21 T 24.73 6.24 0.03 FC5 30.96 12.47 0.00 E. dupreanum – δ15N values (hunting sites as replicates, no random effects) FC5 27.76 0 0.40 T 28.26 0.50 0.31 FC30 29.28 1.52 0.19 FC15 30.46 2.70 0.10 Discussion: Stable isotope data suggested diets of the three species broadly overlapped but had some subtle, but still detectable, differences. In particular, P. rufus and E. dupreanum showed evidence of vertical stratification, particularly where they occurred together. Our data further exhibited spatial variation in diets of P. rufus and E. dupreanum diets at both regional and local levels, and this dietary variation was correlated with forest cover. The relationship between forest cover and diet was evident at a larger spatial scale in P. rufus than in E. dupreanum, and the interaction between climate and forest cover was also important in P. rufus. These results are discussed in more detail below. 187 188 Diets and Diet Breadth of Fruit Bats in Madagascar: Given the three bats species’ exclusive reliance on plant resources and their overlapping ranges (MacKinnon et al. 2003, Andriafidison et al. 2006), we expected that the bats would show evidence of trophic niche differentiation due to resource competition. Although this was generally supported by our data, P. rufus and E. dupreanum did not differ in diet breadth (suggesting at least these two species consume a similar diversity of food resources), and – in accordance with Dammhahn & Goodman (2014) – the differences among the three species in mean δ13C and δ15N values were not large in magnitude, suggesting relatively little separation in their diets. For example, mean δ13C values for the three species differed by only ~1.1‰ (range: -21.2 to -22.29‰), whereas a similar dry season study of a guild of fruit bats in South America found a difference of 6.8‰ (range: -21.5 to -28.3‰; Rex et al. 2011). The small species-level differences in stable isotope values we observed in Madagascar cannot be explained by the existence of distinct and highly specialized diets in the three species, because our results suggest substantial overlap in their diets (Figure 6.2) and because prior studies also indicate that they consume many of the same plant species (MacKinnon et al. 2003, Andriafidison et al. 2006). Another possible reason for dietary overlap is a decrease in available fruit resources over the dry season in seasonal habitats (Sewall et al. 2013). Any such seasonal restriction in diet composition, however, would only be captured partially in our results. This is due to a combined effect of the timing of our data collection and the several-month period represented in stable isotope analyses of hair samples (Crawford et al. 2008), which means that the results encompassed both wet and 188 189 dry seasons in our study area. Thus, other causes of limited fruit availability, such as the replacement of diverse native forests with more homogeneous human-dominated landscapes, could be driving the overlap in diets we observed. Differences between diets of P. rufus and E. dupreanum were more obvious after segregating data from hunting sites where only one species was captured and sites where they were captured at the same hunting site by the same hunter. Notably, in both the single-species and the co-occurrence data, P. rufus had lower δ15N and δ13C values than E. dupreanum (Figure 6.3), which could indicate that P. rufus feeds at higher strata in the forest canopy than E. dupreanum (Voigt and Matt 2004; Rex et al. 2011). Though absolute differences in the stable isotope values were small, even relatively small differences in δ13C values can indicate differences in foraging height of a half-dozen meters or more, with body size positively correlated with feeding height (Rex et al. 2011). This is also consistent with prior observations that when the two species forage simultaneously in kapok trees (Ceiba pentandra), P. rufus forages in higher branches than E. dupreanum (Andriafidison et al. 2006). E. dupreanum’s higher δ15N values may also indicate it is feeding on more commercially grown fruits (Herrera et al. 2008) or in areas that are non-forested and dominated by grasslands and drought-resistant plants (Winter 1979). This interpretation accords well with past observations that E. dupreanum more readily uses degraded habitats than P. rufus (MacKinnon et al. 2003). Further, compared to when they were captured separately, the diets of both bats shifted similarly – toward higher δ13C and δ15N values – when captured together. Specifically, P. rufus’s diet when they co-occurred coincided with that of E. dupreanum’s 189 190 diet in the absence of other bats. Further, with P. rufus present, E. dupreanum’s values shifted to even higher δ13C and δ15N values (Figure 6.4). These results may point to a scenario in which the larger and likely more competitively-dominant P. rufus is able to exclude E. dupreanum from its primary foraging areas in the higher canopy levels of the primary forest, leaving E. dupreanum to forage at a lower canopy height (higher δ13C values) and in more degraded areas including savannahs and agricultural areas with remnant native trees or cultivated fruit trees (higher δ15N values). In this scenario, in locations where primary forest is unavailable, P. rufus may shift its foraging to more degraded habitats (higher δ15N values) that lack high canopies (higher δ13C values), displacing E. dupreanum to even more degraded habitats (Figure 6.4). A closer examination of diet at particular types of hunting site also supports this idea: sites where the two species were caught together had less remnant habitat nearby (9 ± 5% within 15 km and 7 ± 4% within 30 km) than those where they were caught separately (20 ± 11% and 16 ± 11%, respectively). Our analysis of diet shifts assumed that data on bat capture could be used to infer co-occurrence, but it is worth noting the difference in temporal scale between the time frame of the foraging behavior (the shared use of foraging sites was determined for the night captured) and length of time that the stable isotopic analysis of hair samples records diet (several months; Crawford et al. 2008). Thus, while spatial patterns of bat captures are likely to correlate with co-occurrence over time, bias could be introduced if patterns of bat occurrence on the night of capture were not representative of typical patterns of occurrence on a months-long time frame. Our hunting site data nonetheless provide a 190 191 fine-scale estimate of co-occurrence and are suggestive of long-term bat occurrence patterns. While our data did not permit us to examine whether R. madagascariensis changed its foraging behavior in the presence of another species, our data (Figure 6.1) and the data in Dammhahm and Goodman (2014) suggest that it does have notable dietary overlap with the other frugivorous bats. Ecological theory suggests extensive niche overlap is unlikely to be maintained over evolutionary time frames unless differences exist on a separate niche dimension (MacArthur 1958); here, such overlap could be maintained through the use of different foraging behavior by R. madagascariensis. Specifically, unlike the other two species, which land on a branch, then search for, locate, and process fruit while in the tree canopy, R. madagascariensis hovers and removes fruit in flight before bringing it to a nearby feeding roost (Sewall et al. 2013). The unique feeding behavior employed by R. madagascariensis may enable it to access fruits or flowers on some branches that would be inaccessible to the other two bat species, and thus avoid competition with them. Impact of Forest Cover on Bat Diets: Our analyses suggest that the loss of forest cover may impact P. rufus at larger spatial scales (30 km) than E. dupreanum (5-15 km). This result fits well with previous findings that P. rufus is able to forage across a wider radius than E. dupreanum (Garbutt 2007); this wider foraging radius may provide P. rufus with more flexibility in selecting foraging sites. A greater ability to reach remaining intact forest patches may also be 191 192 reflected in P. rufus’s lower δ15N values than those of E. dupreanum. These values increase in drought-resistant plants and in agricultural areas (Voigt and Matt 2004) where primary forests may have been heavily degraded. In such locations, E. dupreanum may have few options but to forage in degraded habitats, whereas P. rufus, with its greater flight radius, could have more flexibility to visit farther, more intact habitats. As noted above, P. rufus could also displace E. dupreanum from intact habitats where they are accessible to both species but limited in extent. Our dataset was not large enough to analyze the impact of deforestation on R. madagascariensis. However, their short foraging distances relative to the other two Malagasy fruit bat species (Andrianaivoarivelo et al. 2011) implies that R. madagascariensis may be responsive to habitat change across smaller spatial scales. The impact of the interaction between forest cover and climate on P. rufus δ13C values was greater than the impact of forest cover alone. This is interesting because it suggests that P. rufus is not necessarily vulnerable only to the loss of forest cover, but to the combined impacts of annual climate variability – which may change the availability of fruits and resources – and a decrease in remnant habitat. It may be that P. rufus’s ability to range over wide areas allows it to overcome the impacts of climate variability or the impacts of forest loss individually, but not the combination of the two. Two other factors could also have affected the results. First, although we focused primarily on anthropogenic changes to habitat, stable isotope values – and their change across different regions and hunting sites – could also be affected by changes in the dominant forest types across our study regions. However, given that all hunting sites in 192 193 our study had the same dominant forest type (Table S16, Appendix F) and that the stable isotope values do not change across the study region for at least one tree species (Table S17, Appendix F), it is likely that the observed changes in diet are due to a loss in forest cover, not a change in forest type or simple geographic variation. Second, in the spatial analysis of forest cover, we also used hunting sites as centroids, which could introduce bias if hunting sites were actually near the edge of a bat’s nightly foraging range and if forest cover differed substantially outside the foraging range. However, the possibility for such bias seems limited since hunting sites for bats in Madagascar are often at or near roost sites (MacKinnon et al. 2003; Cardiff et al. 2009). Conservation Implications: These data provide evidence of the scale at which conservation and protection may need to be implemented for three different species of bat in Madagascar. Past studies have indicated that hunting at roost locations may be the primary threat facing fruit bats on the island (MacKinnon et al. 2003, Cardiff et al. 2009) and have pointed out that these species are known to roost successfully in smaller habitat patches (Jenkins et al. 2007). Based on this evidence and the finding of the significant effect of forest cover on a 30 km scale, a conservation strategy aimed at protecting P. rufus might include a focus on restricting hunting at roosts while protecting sufficient patches of primary forest within ~30 km. Conversely, programs aimed at protecting R. madagascariensis and E. dupreanum, in addition to mitigating hunting, could focus on limiting habitat change at more local (5-15 km) scales. These recommendations reinforce and make more specific 193 194 those outlined in Jenkins et al. (2007), which suggested prioritizing the protection of roosts from hunting while also protecting forest habitat near roosts. We did not exhaustively sample in each town or hunting site for an equal period of time, and therefore our sample sizes should not be used as an index of relative, regional abundances. However, locations in which bats were sold confirms the continued presence of bats near known roost sites in Madagascar (MacKinnon et al. 2003) and suggests the likely presence of previously-unknown roosts (e.g., near to our two most southerly sample sites, Andriba and Antsiafabositra). These results highlight the potential conservation value of monitoring wild meat venues for understanding and tracking the local occurrence of bat populations. They also indicate that populations of these bat species persist, though the extensive threats from hunting and habitat change raise questions about how long they can continue to do so. Population trends in these fruit bat species have important consequences for habitat regeneration and the protection of other endemic biodiversity. Continued hunting and habitat degradation may cause additional decreases in the size and health of populations of these three species, resulting in reduced seed dispersal and pollination in local ecosystems. This effect may be felt even as small remnant populations remain, since fruit bats are often most effective as seed dispersers at higher population densities and thus their seed dispersal services may be largely eliminated well before local extirpation (McConkey and Drake 2006). These findings emphasize the importance of protecting and continuing to study these species to understand how resilient they are to habitat degradation. 194 195 CHAPTER 7 CONCLUSION In this dissertation, we aimed to increase our understanding of why natural resource use occurs and how it impacts biodiversity in Madagascar. Specifically, we presented a series of studies that examined: 1) the effects of habitat degradation on plant- frugivore networks; 2) the live capture and extent of ownership of lemurs in Madagascar; 3) the micro- and macro-level drivers of wild meat consumption in Madagascar; 4) the capture, movement, and trade of wild meat in Madagascar; and 5) the impacts of habitat changes on the diets and vertical stratification of frugivorous bats. For the first study – discussed in Chapter Two – our objectives were to understand the effects of habitat degradation on (1) community structure, (2) network structure, and (3) seed dispersal services. Based on 592 hours of observations at 13 fruiting tree species, we found that as habitat became more degraded: (1) the community structure of both frugivores and fruiting tree communities changed; (2) the mutualistic network structure became less complex and less connected; (3) the interaction strengths of pair-wise interactions changed and the asymmetries of these interactions shifted; and (4) seed dispersal decreased by 91% in the secondary forest, compared to the primary forest. In addition, we show that frugivores: (1) sometimes stopped eating fruit in the degraded forest, even when they had consumed it in other forests; and (2) appeared to avoid some fruiting tree species while showing preference for others. The mutualistic network studied 195 196 in this paper appeared sensitive to anthropogenic disturbance and a novel measure of effectiveness helped quantify these changes. This study brings to mind the ongoing debate about the trade offs between protecting large areas of secondary forest versus small areas of primary forest (e.g. Barlow et al. 2007, Chazdon et al. 2009); it has been argued that in countries where the amount of remnant primary forest is low, the conservation value of secondary forests is high (Chazdon et al. 2009). In our study, frugivores were still participating in the secondary forest mutualistic network, but the benefits exchanged in those mutualisms were few, compared to the primary forest. This has implications for conservation initiatives, as ecosystem services – such as seed dispersal – may be greatly diminished even though frugivores are still visiting and consuming fruit at fruiting trees. Our data therefore indicate that, for some species, primary forest should remain a conservation priority whereever possible. This is especially true as the mutualistic network was less connected and less complex in modified habitats, as compared to primary forests. For the second study, detailed in Chapter Three, our objectives were to provide the first quantitative estimates of the prevalence, spatial extent, correlates and timing of lemur ownership, procurement methods, within-country movements, and numbers and duration of ownership. Using semi-structured interviews (1,093 households and 61 transporters) across 17 study sites, we found that lemur ownership was widespread and affected a variety of taxa. We estimate that 28,253 lemurs have been affected since 2010. Most lemurs were caught by owners and kept for either short (≤1 week) or long (≥3 years) periods. The live capture of lemurs in Madagascar is not highly organized but may 196 197 threaten several endangered species. This study was an excellent case study in understanding how wildlife trafficking (the illegal trade of animals for monetary or non- monetary goods, Izzo 2010; some respondents purchased their lemurs) and the actions of individuals working alone (e.g. individuals capturing lemurs without organized extraction routes/mechanisms) can be difficult to detect and monitor when it is so informally organized and does not cross international borders. Therefore, this study highlights the importance of quantifying ownership of endemic primates in other tropical countries, where such ownership may have gone relatively unnoticed; this is especially the case where species are facing additional anthropogenic threats such as hunting and habitat change. In Chapter Four (the third study), we investigated the role of wild meat in food security in Madagascar, a country where wild meat consumption is poorly understood in urban areas and at regional scales. Using semi-structured interviews (n = 1339 heads-of- households, 21 towns), we aimed to: 1) quantify the amount and purpose of; 2) understand the drivers behind; and, 3) examine recent changes in wild meat consumption in Madagascar. We found that few respondents preferred wild meat (8 ± 3%) but most had eaten it at least once (78 ± 7%), and consumption occurred across ethnic groups, in urban and rural settings. More food-insecure areas reported higher rates of recent consumption of wild meat. However, consumption was best explained by individual preferences and taboos. Few respondents (<1 ± <1%) had increased rates of consumption during their lifetimes, and wild meat prices showed no change from 2005-2013. Most consumption involved wild pigs and small-bodied animals such as bats, though these 197 198 animal groups and lemurs were consumed less in recent years. Given these data, wild meat is unlikely to enhance food security for most Malagasy people in urban and well- connected rural areas. This study highlights the complexities of wild meat consumption and natural resource use and illustrates the difficulties in assuming that different sectors of the general public will interact with the wild meat trade in the same way. For example, Madagascar is currently developing a national strategy for tackling wild animal hunting, but this strategy is based – to a certain extent – on the assumption that wild meat is consumed because cultural taboos are eroding and because of food security issues (Jones and Razafimanahaka 2012). However, our data indicate that the urban population in Madagascar (which currently constitutes 34% of the country’s population; The World Bank 2015) does not consume wild meat for these same reasons (i.e. because of eroding cultural beliefs or for food security reasons), despite consuming wild meat at levels similar to their rural counterparts. Rather, urban consumption of wild meat is driven by a combination of prefereces for meat and taboos against meat; our study shows that these preferences and taboos are not changing at the respondent level. This finding highlights the need for comprehensive conservation and policy planning that accounts for the consumption of wild meat in urban areas and does so in a way that recognizes the different motivations behind wild meat consumption in these areas. In Chapter Five, we discuss our fourth study, which aimed to improve our understanding of the wild meat trade in Madagascar. Specifically, our objectives were to: (1) quantify the volume of consumption, transport, and sale for different animal groups, 198 199 compared to domestic meat; (2) describe the methods of capture and hunting for different animal groups; (3) analyze the patterns of movement of wild meat from the capture location to the final consumer, compared to domestic meat; and (4) examine how the prices of wild meat change depending on the venue through which the consumer purchases it. We found that: (1) a wide range of hunting methods were used, though their prevalence of use differed by animal group; (2) wild meat traveled distances of up to 166 km to reach consumers, though some animal groups were hunted locally (<10 km) in rural areas; (3) most wild meat was procured from free sources (hunting and receiving meat as a gift), though urban respondents who consumed bats and wild pigs were more likely to purchase those meats; and (4) wild meat was consumed at lower rates than domestic meat, though urban respondents consumed twice as much wild meat as rural respondents. We conclude that urban and rural respondents differ in how they interact with the wild meat commodity chain. We also believe that the consumption and trade of wild meat in Madagascar is likely more formalized that previously thought. This chapter is a case study in the difficulties of detecting, evaluating, and valuing the wild meat trade, and the potential for comprehensive studies – involving study at large scales, involving both formal and informal components, and covering all stages of the wild meat commodity chain – to clarify understanding of the wild meat trade. In the case of Madagascar – and prior to this study - the wild meat trade was generally considered to be small and unorganized; this was likely because much of the trade was being conducted illegally and via several different types of informal enterprises. Detecting and evaluating the trade required intensive on-the-ground data collection 199 200 efforts; similar efforts in other areas of the world would be limited to those groups who have the resources to undertake a multi-month, multi-regional study. In addition, quantifying the volume of trade and assigning economic value to the trade becomes more difficult as more of the trade is conducted through informal enterprises that are usually not included in official government statistics or that are operating outside the realm of government oversight. Therefore, more research and development in this field is needed in order to create robust rapid assessment techniques and guidelines for how non- government entities (e.g. nonprofits and academic groups) can accurately measure wild meat commodity chains (e.g. providing guidelines for evaluating economic value of wild meat) in areas like Madagascar. These rapid assessment efforts and/or continued monitoring efforts are critical to understanding how the wild meat trade is changing within a region and whether conservation and policy initiatives are having their intended effect. In addition, better data on these issues are critical to calling attention to the problem of wildlife trafficking, which is the third largest illegal trade in the world (following illegal drugs and arms sales) and surpasses 20 billion USD in trade per year (Giovanni 2006). Collecting data on wildlife trade – both informal and formal – can help increase understanding of the commodity chain and can help identify points along the chain for conservation interventions (Bowen-Jones et al. 2003, Cowlishaw et al. 2005). When considering conservation interventions, it is important to consider all aspects of the commodity chain as targeted programming can have unintended consequences on other stages of the commodity chain (Bowen-Jones et al. 2003). For example, the provision of 200 201 alternative protein sources to communities can decrease hunter income and cause hunters to increase their hunting or switch hunting strategies to higher value and possibly rarer species (to make up for lost income; Bowen-Jones et al. 2003). In some cases, this means that there is no single best point along the commodity chain for conservation interventions (Cowlishaw et al. 2005). However, given that funding is limited for conservation programming and given that the commodity chain in Madagascar does utilize many informal actors, understanding the commodity chain can highlight stage- specific opportunities for conservation planning. As reviewed in Chapter Five, the commodity chains of different animal groups in Madagascar provide different opportunities for intervention. We also highlight that the commodity chains differ between urban and rural respondents and that care should be taken to tailor outreach to those different groups. Finally, Chapter Six (our fifth study) examined how three fruit bat species were impacted by habitat loss. Our results indicated that the three species had broadly overlapping diets. Differences in diet were nonetheless detectable and consistent between P. rufus and E. dupreanum, and these diets shifted when they co-occurred, suggesting resource partitioning across habitats and vertical strata within the canopy to avoid competition. Changes in diet were also correlated with a decrease in forest cover, though at a larger spatial scale in P. rufus than in E. dupreanum. These results suggest fruit bat species exhibit differing foraging strategies in response to habitat change. They also highlight the key threats that fruit bats face from habitat change, and clarify the spatial 201 202 scales at which conservation efforts should be implemented to mitigate threats for these bat species in Madagascar. This study is an example of how stable isotope techniques can help provide evidence of the scale at which conservation and protection may need to be implemented for species that are quite similar. In the case of our study, the three bat species have overlapping ranges and overlapping diets; it would be difficult to provide an accurate quantitative estimate of the scale for conservation programming from field observations or by using information from one species to infer the needs of another species. Based on our data, and based on past studies (MacKinnon et al. 2003, Cardiff et al. 2009), we suggest a conservation strategy aimed at protecting P. rufus via restricted hunting at roosts while protecting sufficient patches of primary forest within ~30 km. Conversely, programs aimed at protecting R. madagascariensis and E. dupreanum, in addition to mitigating hunting, could focus on limiting habitat change at more local (5-15 km) scales. The support for the effect of climate in the P. rufus models and not in the E. dupreanum models might also suggest a greater potential vulnerability to climate change effects on the diet of P. rufus than on E. dupreanum. The present series of studies aimed to increasing understanding of the impacts that natural resources use and habitat degradation have on biodiversity as well as the drivers and motivations behind this natural resource use. As a result of these studies, we found that the impacts of habitat degradation can be seen in the complexity of ecological networks (Chapter Two) and in the diets of animals living in degraded regions (Chapter Six). Moreover, we found that this habitat degradation is being driven by a wide range of 202 203 micro- and macro-drivers (Chapter Four), can occur as a result of demand for wildlife products in places far removed from remnant habitat (Chapter Five), and can impact tens of thousands of animals per year in-country (Chapters Three, Four, Five). As such, these data could help inform reforestation efforts (Chapters Two, Six), mammalian conservation programs (Chapters Three, Four, Five), and food security initiatives (Chapter Four). These studies increase our understanding of why natural resource use occurs, how it impacts biodiversity, and how it could be reconciled with conservation priorities both in Madagascar and in sub-Saharan Africa more broadly. 203 204 REFERENCES CITED African Development Bank Group. 2005. Republic of Madagascar: Country Governance Profile. Available online: http://www.afdb.org/fileadmin/uploads/afdb/Documents/Project-and- Operations/ADF-BD-IF-2005-152-EN-MADAGASCAR-COUNTRY- GOVERNANCE-PROFILE.PDF (accessed 1 June 2015) Albrecht, J., D. G. Berens, N. Blüthgen, B. Jaroszewicz, N. Selva, N. Farwig, and P. Bellingham. 2012. Logging and forest edges reduce redundancy in plant-frugivore networks in an old-growth European forest. Journal of Ecology 101:990-999. Allnutt, T. F., G. P. Asner, C. D. Golden, and G. V. N. Powell. 2013. Mapping recent deforestation and forest disturbance in northeastern Madagascar. Tropical Conservation Science 6:1-15. Anderson, M. J. 2001. A new method for non-parametric multivariate analysis of variance. Austral Ecology 26:32-46. Andreone, F., J. E. Cadle, N. Cox, F. Glaw, R. A. Nussbaum, C. J. Raxworthy, S. N. Stuart, D. Vallan, and M. Vences. 2005. Species review of amphibian extinction risks in Madagascar: conclusions from the global amphibian assessment. Conservation Biology 19:1790-1802. Andrews, J., P. Antilahimena, and C. R. Birkinshaw. 1998. Use of a day testing box by a wild sportive lemur, Lepilemur dorsalis, on Nosy Be, North-Western Madagascar. Folia Primatolog 69(suppl 1):18-21. Andriafidison, D., R. A. Andrianaivoarivelo, O. R. Ramilijaona, M. R. Razanahoera, J. MacKinnon, R. K. B. Jenkins, and P. A. Racey. 2006. Nectarivory by endemic Malagasy fruit bats during the dry season. Biotropica 38:85-90. Andriaholinirina, N., A. Baden, M. Blanco, L. Chikhi, A. Cooke, N. Davies, R. Dolch, G. Donati, J. Ganzhorn, C. Golden, L. F. Groeneveld, A. Hapke, M. Irwin, S. Johnson, P. Kappeler, T. King, R. Lewis, E. E. Louis, M. Markolf, V. Mass, R. A. Mittermeier, R. Nichols, E. Patel, C. J. Rabarivola, B. Raharivololona, S. Rajaobelina, G. Rakotoarisoa, B. Rakotomanga, J. Rakotonanahary, H. Rakotondrainibe, G. Rakotondratsimba, M. Rakotondratsimba, L. Rakotonirina, F. B. Ralainasolo, J. Ralison, T. Ramahaleo, H. Razafindraibe, J. Razafindramanana, N. Rowe, J. Salmona, M. Seiler, S. Volampeno, P. Wright, J. Youssouf, J. Zaonarivelo, and A. Zaramody. 2014a. Eulemur coronatus. The 204 205 IUCN Red List of Threatened Species. Version 2014.3. Available online: www.iucnredlist.org. Accessed March 19 2015. Andriaholinirina, N., A. Baden, M. Blanco, L. Chikhi, A. Cooke, N. Davies, R. Dolch, G. Donati, J. Ganzhorn, C. Golden, L. F. Groeneveld, A. Hapke, M. Irwin, S. Johnson, P. Kappeler, T. King, R. Lewis, E. E. Louis, M. Markolf, V. Mass, R. A. Mittermeier, R. Nichols, E. Patel, C. J. Rabarivola, B. Raharivololona, S. Rajaobelina, G. Rakotoarisoa, B. Rakotomanga, J. Rakotonanahary, H. Rakotondrainibe, G. Rakotondratsimba, M. Rakotondratsimba, L. Rakotonirina, F. B. Ralainasolo, J. Ralison, T. Ramahaleo, H. Razafindraibe, J. Razafindramanana, N. Rowe, J. Salmona, M. Seiler, S. Volampeno, P. Wright, J. Youssouf, J. Zaonarivelo, and A. Zaramody. 2014b. Eulemur sanfordi. The IUCN Red List of Threatened Species. Version 2014.3. Available online: www.iucnredlist.org. Accessed March 19 2015. Andrianaivoarivelo, R. A., O. R. Ramilijaona, P. A. Racey, N. Razafindrakoto, and R. K. B. Jenkins. 2011. Feeding ecology, habitat use and reproduction of Rousettus madagascariensis Grandidier, 1928 (Chiroptera: Pteropodidae) in eastern Madagascar. Mammalia 75:69-78. Asner, G. P., D. E. Knapp, A. Balaji, and Paez-Acosta. 2009. Automated mapping of tropical deforestation and forest degradation: CLASlite. Journal of Applied Remote Sensing 3:33524-33543. Atmar, W., and B. D. Patterson. 1995. The nestedness temperature calculator: a visual basic program, including 294 presence and absence matrices. AICS Research, Inc., University Park, New Mexico, USA and The Field Museum, Chicago, Illinois, USA. Baillie, J.E.M., C. Hilton-Taylor, and S. N. Stuart (Eds.). 2004. 2004 IUCN Red List of Threatened Species. A Global Species Assessment. IUCN, Gland, Switzerland and Cambridge, UK. Barlow, J., T. A. Gardner, I. S. Araujo, T. C. Avila-Pires, A. B. Bonaldo, J. E. Costa, M. C. Esposito, L. V. Ferreira, J. Hawes, M. I. M. Hernandez, M. S. Hoogmoed, R. N. Leite, N. F. Lo-Man-Hung, J. R. Malcolm, M. B. Martins, L. A. M. Mestre, R. Miranda-Santos, A. L. Nunes-Gutjahr, W. L. Overal, L. Parry, M. A. Riberio- Junior, M. N. F. da Silva, C. da Silva Motta, and C. A. Peres. 2007. Quantifying the biodiversity value of tropical primary, secondary, and plantation forests. Proceedings of the National Academy of Sciences 104:18555-18560. Barrett, M. A., J. and Ratsimbazafy. 2009. Luxury bushmeat trade threatens lemur conservation. Nature 461: 470 205 206 Bascompte, J., P. Jordano, C. J. Melian, and J. M. Oleson. 2003. The nested assembly of plant-animal mutualistic networks. Proceedings of the National Academy of Sciences 100:9383-9387. Bascompte, J., and P. Jordano. 2007. Plant-animal mutualistic networks, the architecture of biodiversity. Annual Review of Ecology, Evolution, and Systematics 38:567- 593. Bates, D., M. Maechler, B. Bolker, and S. Walker. 2013. Lme4: Linear mixed-effects models using Eigen and S4. R package version 1.0-5. Available online: http://CRAN.R-project.org/package=lme4 (accessed 6 March 2014) Baum, D. A. 1995. The comparative pollination and floral biology of baobabs (Adansonia- Bombacaceae). Annals of the Missouri Botanical Garden 82:322- 348. Benjamin, N., and A. A. Mbaye. 2014. Informality, growth, and development in Africa: WIDER Working Paper 2014/052. World Institute for Development Economics Research, UNU-WIDER, Helsinki, Finland. Bennett, E. L. 2002. Is there a link between wild meat and food security? Conservation Biology 16:590-592. Berkes, F. 2004. Rethinking Community-based conservation. Conservation Biology 18:621-630. Birkinshaw, C., A. Ravoahangy, and H.E. Andriamaharoa. 2007. Lemur records at priority sites for plant conservation. Lemur News 12:35-36. Blüthgen, N., J. Frund, D. P. Vazquez, and F. Menzel. 2009. What do interaction network metrics tell us about specialization and biological traits? Ecology 89:3387-3399. Bollen, A. and L. Van Elsacker. 2002. Feeding ecology of Pteropus rufus (Pteropodidae) in the littoral forest of Sainte Luce, SE Madagascar. Acta Chiropterologica 4:33- 47. Bollen, A., L. van Elsacker, and J. U. Ganzhorn. 2004. Tree dispersal strategies in the littoral forest of Sainte Luce (SE-Madagascar). Oecologia 139:604-616. Bowen-Jones, E., D. Brown, and E. J. Z. Robinson. 2003. Economic commodity or environmental crisis? An interdisciplinary approach to analyzing the bushmeat trade in Central and West Africa. Area 35:390-402. 206 207 Bravo, A., K. E. Harms, and L. H. Emmons. 2010. Preference for Collpa water by frugivorous bats (Artibeus): an experimental approach. Biotropica 42:276-280. Breuer, T., and F. B. Mavinga. 2010. Education for the conservation of great apes and other wildlife in Northern Congo – the importance of nature clubs. American Journal of Primatology 72:454-461. Brockman, D. K., L. R. Godfrey, L. J. Dollar, and J. Ratsirarson. 2008. Evidence of invasive Felis silvestris predation on Propithecus verreauxi at Beza Mahafaly Special Reserve, Madagascar. International Journal of Primatology 29:135-152. Brodie, J. F., O. E. Helmy, W. Y. Brockelman, and J. L. Maron. 2009. Bushmeat poaching reduces the seed dispersal and population growth rate of a mammal- dispersed tree. Ecological Applications 19:854-863. Bronstein, J. L. 1994. Our current understanding of mutualism. The Quarterly Review of Biology 69:31-51. Brooks, T. M., R. A. Mittermeier, C. G. Mittermeier, G. A. B. da Fonseca, A. B. Rylands, W. R. Konstant, P. Flick, J. Pilgrim, S. Oldfield, G. Magin, and C. Hilton-Taylor. 2002. Habitat loss and extinction in the hotspots of biodiversity. Conservation Biology 16:909-923. Brown, D. 2003. Bushmeat and Poverty Alleviation: Implications for Development Policy. ODI Wildlife Policy Briefing 2. Available online: http://www.odi.org/publications/2427-bushmeat-poverty-alleviation-implications- development-policy (accessed 24 March 2015) Brown, D., and A. Williams. 2003. The Case for Bushmeat as a Component of Development Policy: Issues and Challenges. International Forestry Review 5:148–155. Brown, J. L., and A. D. Yoder. 2015. Shifting ranges and conservation challenges for lemurs in the face of climate change. Ecology and Evolution 5:1131-1142. Buckland, S. T., A. J. Plumptre, L. Thomas, and E. A. Rexstad. 2010. Design and analysis of line transect surveys for primates. International Journal of Primatology 31:833-847. Bulte, E., and D. Rondeau. 2007. Compensation for wildlife damages: habitat conversion, species preservation and local welfare. Journal of Environmental Economics and Management 54:311-322. 207 208 Burgos, E., H. Ceva, R. P. J. Perazzo, M. Devoto, D. Medan, D. Zimmermann, and A. M. Delbue. 2007. Why nestedness in mutualistic networks? Journal of Theoretical Biology 249:307-313. Burnham, K. P., and D. R. Anderson. 2002. Model selection and multimodel inference: A practical information-theoretic approach. Springer Science and Business Media, New York, USA. Campbell, B. M., J. A. Sayer, and B. Walker. 2010. Navigating trade-offs: working for conservation and development outcomes. Ecology and Society 15:16. Cantor, M., M. M. Pires, G. O. Longo, P. R. Guimarães, and E. Z. F. Setz. 2013. Individual variation in resource use by opossums leading to nested fruit consumption. Oikos 122:1085-1093. Caramel, L. 2015. China’s rosewood craving cuts deep into Madagascar rainforests. The Guardian. Available online: http://www.theguardian.com/environment/2015/feb/16/rosewood-madagascar- china-illegal-rainforest (accessed 24 March 2015) Cardiff, S., and J. Befourouack. 2003. The Reserve Speciale d’Ankarana. In: The Natural History of Madagascar (eds. S. Goodman and J.P. Benstead). University of Chicago Press, Chicago, USA. Cardiff, D. G., F. H. Ratrimomanarivo, G. Rembert, and S. M. Goodman. 2009. Hunting, disturbance and roost persistence of bats in caves at Ankarana, northern Madagascar. African Journal of Ecology 47:640-649. Ceballos-Mago, N., C. E. Gonzalez, and D. J. Chivers. 2010. Impact of the pet trade on the Margarita capuchin monkey Cebus apella margaritae. Endangered Species Research 12:57-68. Central Intelligence Agency [CIA]. 1976. Madagascar Population Density and Ethnic Groups Map. In: The Indian Ocean Atlas. Central Intelligence Agency, McLean, USA. Chama, L., D. G. Berens, C. T. Downs, and N. Farwig. 2013. Habitat characteristics of forest fragments determine specialization of plant-frugivore networks in a mosaic forest landscape. PLoS One 8:e54956. Chapman, C. A., L. J. Chapman, R. Wangham, K. Hunt, D. Gebo, and L. Gardner. 1992. Estimators of fruit abundance of tropical trees. Biotropica 24:527-531. 208 209 Chazdon, R. L., C. A. Peres, D. Dent, D. Sheil, A. E. Lugo, D. Limb, N. E. Stork, and S. E. Miller. 2009. The potential for species conservation in tropical secondary forests. Conservation Biology 25:1406-1417. Christie, I. T., and D. E. Crompton. 2003. Republic of Madagascar: Tourism Sector Study. Africa Regional Working Paper Series No. 63. World Bank, Washington, D.C., USA. Clark, M. 1997. The mongoose lemur Eulemur mongoz on Anjouan, Comores. Dodo The Journal of The Wildlife Preservation Trusts 33:36-44. Clarke, P., and S. D. Jupiter. 2010. Law, custom and community-based natural resource management in Kubulau District (Fiji). Environmental Conservation 37:98-106. Cordiero, N. J., and H. F. Howe. 2003. Forest fragmentation severs mutualisms between seed dispersers and an endemic African tree. Proceedings of the National Academy of Sciences 100:14052-14056. Corlett, R. 2007. The impact of hunting on the mammalian fauna of tropical Asian forests. Biotropica 39:292-303. Cosson, J. F., J. M. Pons, and D. Masson. 1999. Effects of forest fragmentation on frugivorous and nectarivorous bats in French Guinea. Journal of Tropical Ecology 15:515-534. Cowlishaw, G., S. Mendelson, and J. Marcus Rowcliffe. 2005. Structure and operation of a bushmeat commodity chain in Southwestern Ghana. Conservation Biology 19:139-149. Crawford, K., R. A. McDonald, and S. Bearhop. 2008. Applications of stable isotope techniques to the ecology of mammals. Mammal Review 38:87-107. Crookes, D. J., N. Ankudey, and E. J. Milner-Gulland. 2005. The value of a long-term bushmeat market dataset as an indicator of system dynamics. Enviromental Conservation 32:333-339. Dammhahn, M., and S. M. Goodman. 2014. Trophic niche differentiation and microhabitat utilization revealed by stable isotope analyses in a dry-forest bat assemblage at Ankarana, northern Madagascar. Journal of Tropical Ecology 30:97-109. de Boer, J., and H. Aiking. 2011. On the merits of plant-based proteins for global food security: Marrying macro and micro perspectives. Ecological Economics 70:1259-1265. 209 210 De Jong, W., D. Donovan, and K.-I. Abe (Eds.). 2007. Extreme conflict and tropical forests. Springer, Dordrecht, Netherlands. deFrance, S. D. 2009. Zooarchaeology in complex societies: political economy, status, and ideology. Journal of Archaeological Research 17:105–168. Dirzo, R. and Raven, P.H. 2003. Global state of biodiversity and loss. Annual Review of Environment and Resources 28:137-161. Dormann, C. F., B. Gruber, and J. Fruend, J. 2008. Introducing the bipartite Package: Analysing Ecological Networks. R news 8/2:8-11. Dostie, B., S. Haggblade, and J. Randriamamonjy. 2002. Seasonal poverty in Madagascar: magnitude and solutions. Food Policy 27:493-518. Duarte-Quiroga, A., and A. Estrada. 2003. Primates as pets in Mexico City: An assessment of the species involved, source of origin, and general aspects of treatment. American Journal of Primatology 61:53-60. Duckworth, J. W., M. I. Evans, A. F. A. Hawkins, R. J. Safford, and R. J. Wilkinson. 1995. The lemurs of Marojejy Strict Nature Reserve, Madagascar: A status overview with notes on ecology and threats. International Journal of Primatology 16:545-559. Dunne, J. A., R. J. Williams, and N. D. Martinez. 2002. Network structure and biodiversity loss in food webs: robustness increases with connectance. Ecology Letters 5:558-567. Emerton, L. 1999. Balancing the opportunity cost of wildlife conservation for communities around Lake Mburo National Park, Uganda. Evaluating Eden (IIED) and Community Conservation Research Project, Universities of Manchester, Zimbabwe, and Cambridge. Eppley, T. M., E. Verjans, and G. Donati. 2011. Coping with low-quality diets: a first account of the feeding ecology of the southern gentle lemur, Hapalemur meridionalis, in the Mandena littoral forest, southeast Madagascar. Primates 52:7- 13. ESRI. 2011. ArcGIS Desktop: Release 10. Environmental Systems Research Institute, Redlands, USA. Eudey, A. A. 1994. Temple and pet primates in Thailand. Revue d’ecologie: la terre et la vie 49:273-280. 210 211 Evans, D. M., M. J. O. Pocock, and J. Memmott. 2013. The robustness of a network of ecological networks to habitat loss. Ecology Letters 16:844-852. Fa, J. E., and C. Peres. 2001. Hunting in tropical forests. In: Conservation of exploited species (eds. J. D. Reynolds, G. M. Mace, K. H. Redford, and J. G. Robinson). Cambridge University Press, Cambridge, UK. Fa, J.E., J. Juste, J. Perez del Val, and J. Castroviejo. 1995. Impact of market hunting on mammal species in Equatorial Guinea. Conservation Biology 9:1107-1115. Fa, J. E., D. Currie, and J. Meeuwig. 2003. Bushmeat and food security in the Congo Basin: linkages between wildlife and people’s future. Environmental Conservation 30:71-78. Fahrig, L. 2003. Effects of Habitat Fragmentation on Biodiversity. Annual Review of Ecology, Evolution, and Systematics 34:487-515. Fotso, J. C. 2007. Urban–rural differentials in child malnutrition: trends and socioeconomic correlates in sub-Saharan Africa. Health and Place 13:205-223. Fowler, S. V., P. Chapman, D. Checkley, S. Hurd, M. McHale, G.-S. Ramangason, J.-E. Randriamasy, P. Stewart, R. Walters, and J. M. Wilson. 1989. Survey and management proposals for a tropical deciduous forest at Ankarana in northern Madagascar. Biological Conservation 47:297-313. Freed, B. Z. 2012. Abstract: Fallback food consumption and sympatry in Eulemur coronatus and Eulemur sanfordi. The 81st Annual Meeting of the American Association of Physical Anthropologists. Available online: http://meeting.physanth.org/program/2012/session23/freed-2012-fallback-food- consumption-and-sympatry-in-eulemur-coronatus-and-eulemur-sanfordi.html (accessed 24 March 2015) Garbutt, N. 2007. Mammals of Madagascar: A complete guide. Yale University Press, New Haven, USA. Garcia, G., and S. M. Goodman. 2003. Hunting of protected animals in the Parc National d’Ankarafantsika, north-western Madagascar. Oryx 37:115-118. Gardner, C. J., and Z. G. Davies. 2014. Rural bushmeat consumption within multiple-use protected areas: Qualitative evidence from southwest Madagascar. Human Ecology 42:21-34. Giovanni, D. 2006. Taking animal trafficking out of the shadows. Innovations 1: 25-35. 211 212 Global Witness and the Environmental Investigation Agency. 2009. Investigation into the illegal felling, transport and export of precious wood in SAVA region, Madagascar. Global Witness. Available online: www.parcs-madagascar. com/doc/report_vsfinal.pdf (accessed 24 March 2015) Golden, C. D. 2009. Bushmeat hunting and use in the Makira Forest, north-eastern Madagascar: a conservation and livelihoods issue. Oryx 43:386-392. Golden, C. D., L. C. H. Fernald, J. S. Brashares, B. J. Rodolph Rasolofoniaina, and C. Kremen. 2011. Benefits of wildlife consumption to child nutrition in a biodiversity hotspot. Proceedings of the National Academy of Sciences 108:19653-19656. Golden, C., M. H. Bonds, J. S. Brashares, B. J. Rodolph Rasolofoniaina, and C. Kremen. 2014a. Economic valuation of subsistence harvest of wildlife in Madagascar. Conservation Biology 28:234-243. Golden, C. D., J. G. C. Rabehatonina, A. Rakotosoa, and M. Moore. 2014b. Socio- ecological analysis of natural resource use in Betampona Strict Natural Reserve. Madagascar Conservation and Development 9:83-89. Goodman, S. M. 1993. A reconnaissance of Ile Sainte Marie, Madagascar: The status of the forest, avifauna, lemurs and fruit bats. Biological Conservation 65:205-212. Goodman, S. M. 2006. Hunting of Microchiroptera in south-western Madagascar. Oryx 40:225-228. Goodman, S. M., D. Andriafidison, R. Andrianaivoarivelo, S. G. Cardiff, E. Ifticene, R. K. B. Jenkins, A. Kofoky, T. Mbohoahy, D. Rakotondravony, J. Ranivo, F. Ratrimomanarivo, J. Razafimanahaka, and P. A. Racey. 2005. The distribution and conservation of bats in the dry regions of Madagascar. Animal Conservation 8:153-165. Goodman, S. M., and J. P. Benstead. 2003. The Natural History of Madagascar. University of Chicago Press, Chicago, USA. Goodman, S. M., and A. Raselimanana. 2003. Hunting of wild animals by Sakalava of the Menabe region: a field report from Kirindy-Mite. Lemur News 8:4–6. Gorchov, D. L., F. Cornejo, C. Ascorra, and M. Jaramillo. 1993. The role of seed dispersal in the natural regeneration of rain forest after strip-cutting in the Peruvian Amazon. Vegetatio 107/108:339-349. 212 213 Gray, A. M., S. L. Baldauf, P. J. Mayhew, and J. K. Hill. 2007. The response of avian feeding guilds to tropical forest disturbance. Conservation Biology 21:133-141. Grenoble, R. 2013. Ploughshare tortoise smugglers caught with 10 percent of entire species at Thailand airport. Available online: http://www.huffingtonpost.com/2013/03/28/ploughshare-tortoise-smuggle- thailand_n_2972673.html (accessed 6 March 2014) Guerbois, C., A.-B. Dufour, G. Mtare, and H. Fritz. 2013. Insights for integrated conservation from attitudes of people toward protected areas near Hwange National Park, Zimbabwe. Conservation Biology 27:844-855. Harcourt, C., and J. Thornback (Eds.). 1990. Lemurs of Madagascar and the Comoros. The IUCN Red Data Book. International Union for Conservation of Nature, Cambridge, UK. Harvey, P. H., and T. H. Clutton-Brock. 1985. Life history variation in primates. Evolution 39:559-581. Hawkins, A. F. A., P. Chapman, J. U. Ganzhorn, Q. M. C. Blocam, S. C. Barlow, and S. J. Tonge. 1990. Vertebrate conservation in Ankarana Special Reserve, northern Madagascar. Biological Conservation 54:83-110. Hekkela, E. R., M. Rakotondratsima, and N. Vasey. 2007. Habitat and distribution of the Ruffed Lemur, Varecia, north of the Bay of Antongil in Northeastern Madagascar. Primate Conservation 22:89-95. Herrera, L. G., C. Korine, T. H. Fleming, and Z. Arad. 2008. Dietary implications of intrapopulation variation in nitrogen isotope composition of an old world fruit bat. Journal of Mammalogy 89:1184-1990. Hijmans, R. J., S. E. Cameron, J. L. Parra, P. G. Jones and A. Jarvis. 2005. Very high resolution interpolated climate surfaces for global land areas. International Journal of Climatology 25:1965-1978. Hobson, K. A. 1999. Tracing origins and migration of wildlife using stable isotopes: a review. Oecologia 120:314-326. Hurvich, C.M., and C-L. Tsai. 1989. Regression and time series model selection in small samples. Biometrika 76:297-307. Ilo. 2003. Recensement des Communes 2001. Available online : http://www.ilo.cornell.edu/ilo/data.html (accessed 24 March 2015) 213 214 Innes, J.L. 2010. Madagascar rosewood, illegal logging and the tropical timber trade. Madagascar Conservation and Development 5:6-10. Institut National de la Statistique [INSTAT] and ORC Macro. 2005. Enquête Démographique et de Santé, Madagascar 2003–2004: Rapport de synthèse. INSTAT and ORC Macro, Calverton, Maryland, USA. International Fund for Agricultural Development [IFAD]. 2007. Rural poverty in Madagascar. IFAD, Rome, Italy. Irwin, M. T., T. M. Smith, and P. C. Wright. 2000. Census of three eastern rainforest sites north of Ranomafana National Park: preliminary results and implications for lemur conservation. Lemur News 5:20-22. Irwin, M. T., P. C. Wright, C. Birkinshaw, B. L. Fisher, C. J. Gardner, J. Glos, S. M. Goodman, P. Loiselle, P. Rabeson, J-L. Raharison, M. J. Raherilalao, D. Rakotondravony, A. Raselimanana, J. Ratsimbazafy, J. S. Sparks, L. Wilme, and J. U. Ganzhorn. 2010. Patterns of species change in anthropogenically disturbed forests of Madagascar. Biological Conservation 143:2351-2362. International Union for Conservation of Nature [IUCN]. 2013. The IUCN Red List of Threatened Species. Version 2013.2. Available online : http://www.iucnredlist.org (accessed 24 March 2015) Janzen, D. H. 1979. How to be a fig. Annual Review of Ecology and Systematics 10:13- 51. Jenkins, R. K. B., D. Andriafidison, H. J. Razafimanahaka, A. Rabearivelo, N. Razafindrakoto, Z. Ratsimandresy, R. H. Andrianandrasana, E. Razafimahatratra, and P. A. Racey. 2007. Not rare, but threatened: the endemic Madagascar flying fox Pteropus rufus in a fragmented landscape. Oryx 41:263-271. Jenkins, R. K. B., A. Keane, A. R. Rakotoarivelo, V. Rakotomboavonjy, F. H. Randrianandrianina, H. J. Razafimanahaka, S. R. Ralaiarimalala, and J. P. G. Jones. 2011. Analysis of patterns of bushmeat consumption reveals extensive exploitation of protected species in Eastern Madagascar. PLoS One 6:e27570. Jenkins, R. K. B., and P. A. Racey. 2008. Bats as bushmeat in Madagascar. Madagascar Conservation and Development 3:22-30. Jolly, A., W. L. R. Oliver, and S. M. O’Connor. 1982. Population and troop ranges of Lemur catta and Lemur fulvus at Berenty, Madagascar: 1980 census. Folia primatologica 39:115-123. 214 215 Jones, J. P. G., and J. H. Razafimanahaka. 2012. Madagascar develops a national strategy to tackle bushmeat hunting. Available online: http://www.iucn.org/about/union/commissions/sustainable_use_and_livelihoods_s pecialist_group/sulinews/issue_2/sn2_bushmeat/#.UDoi0tzsyRU.facebook (accessed 1 June 2015) Jones, J. P. G., M. M. Andriamarovololona, and N. Hockley. 2008. The importance of taboos and social norms to conservation in Madagacsar. Conservation Biology 22:976-986. Jones-Engel, L., M. A. Schillaci, G. Engel, U. Paputungan, and J. W. Froehlich. 2005. Characterizing primate pet ownership in Sulawesi: implications for disease transmission. In: Special Topics in Primatology Volume 4: Commensalim and Conflict The Human Primate Interface (eds. J.D. Patterson and J. Williams). The American Society of Primatologists, USA. Jordano, P. 1997. Patterns of mutualistic interactions in pollination and seed dispersal: connectance, dependence, asymmetries, and coevolution. The American Naturalist 129: 657-577. Junge, R. E., C. V. William, and J. Campbell. 2009. Nutrition and behavior of lemurs. Veterinary Clinics of North America: Exotic Animal Practice 12:339-348. Kamins, A. O., O. Restif, Y. Ntiamoa-Baidu, R. Suu-Ire, D. T. S. Hayman, A. A. Cunninham, J. L. N. Wood, and J. M. Rowcliffe. 2011. Uncovering the fruit bat bushmeat commodity chain and the true extent of fruit bat hunting in Ghana, West Africa. 144:3000-3008. Karesh, W. B., R. A. Cook, E. L. Bennett, and J. Newcomb. 2005. Wildlife trade and global disease emergence. Emerging Infectious Diseases 7:1000-1002. Keane, A., A. A. Ramarolahy, J. P. G. Jones, and E. J. Milner-Gulland. 2011. Evidence for the effects of environmental engagement and education on knowledge of wildlife laws in Madagascar. Conservation Letters 4:55-63. Kinnaird, M. F., E. W. Sanderson, T. G. O’Brien, H. T. Wibisono, and G. Woolmer. 2003. Deforestation trends in a tropical landscape and implications for endangered large mammals. Conservation Biology 17:245-257. Kumpel, N. F., E. J. Milner-Gulland, G. Cowlishaw, and J. M. Rowcliffe. 2010. Incentives for hunting: the role of bushmeat in the household economy in rural Equatorial Guinea. Human Ecology 38:251-264. 215 216 Kunz, T. H., E. Braun de Torrez, D. Bauer, T. Lobova, and T. H. Fleming. 2011. Ecosystem services provided by bats. Annals of the New York Academy of Sciences 1223:1-38. Kyes, R. C., M. M. Ferroz, M. A. Aziz, M. R. Alam, F. Akhtar, S. Ahmed, M. I. Nahid, B. H. Badal, G. Engel, and L. Jones-Engel. 2011. Preliminary population survey of the rhesus macaques in Barmi Bazaar, Bangladesh: Survey approaches for studying urban primates. American Journal of Primatology 73(Suppl. 1):88. Lamb, D., P. D. Erskine, and J. A. Parrotta. 2005. Restoration of degraded tropical forest landscapes. Science 310:1628-1632. Lambek, M. 1992. Taboo as cultural practice among Malagasy speakers. Man 27:245- 266. Lambin, E. F., H. J. Geist, and E. Lepers. 2003. Dynamics of land-use and land-cover change in tropical regions. Annual Review of Environment and Resources 28:205-241. Lindsey, P.A., G. Balme, M. Becker, C. Begg, C. Bento, C. Bocchino, A. Dickman, R. W. Diggle, H. Eves, P. Henschel, D. Lewis, K. Marnewick, J. Mattheus, J. Weldon McNutt, R. McRobb, N. Midlane, J. Milanzi, R. Morley, M. Murphree, V. Opyene, J. Phadima, G. Purchase, D. Rentsch, C. Roche, J. Shaw, H. van der Westhuizen, N. Van Vliet, and P. Zisadza-Gandiwa. 2013. The bushmeat trade in African savannas: impacts, drivers, and possible solutions. Biological Conservation 160:80-96. Long, E., and P. A. Racey. 2007. An exotic plantation crop as a keystone resource for an endemic megachiropteran, Pteropus rufus, in Madagascar. Journal of Tropical Ecology 23:397-407. Luskin, M. S. 2010. Flying foxes prefer to forage in farmland in a tropical dry forest landscape mosaic in Fiji. Biotropica 42:246-250. MacArthur, R. H. 1958. Ecology of some warblers of northeastern coniferous forests. Ecology 39:599-619. Mackenzie, C. A., C. A. Chapman, and R. Sengupta. 2011. Spatial patterns of illegal resource extraction in Kibale National Park, Uganda. Environmental Conservation 39:38-50. MacKinnon, J. L., C. E. Hawkins, and P. A. Racey. 2003. Pteropodidae. In: The natural history of Madagascar (eds. S. Goodman and J. P. Benstead). University of Chicago Press, Chicago, USA. 216 217 Madden, F. 2004. Creating coexistence between humans and wildlife: global perspectives on local efforts to address human-wildlife conflict. Human Dimensions of Wildlife 9:247-257. Martinez, B. 2008. Occurrence of Bamboo lemurs, Hapalemur griseus occidentalis, in an agricultural landscape on the Masoala Peninsula. Lemur News 13:11-14. McConkey, K. R. and D. R. Drake. 2006. Flying foxes cease to function as seed dispersers long before they become rare. Ecology 87:271-276. Medellin, R. A. and O. Gaona. 1999. Seed dispersal by bats and birds in forest and disturbed habitats of Chiapas, Mexico. Biotropica 31:478-485. Menke, S., K. Böhning-Gaese, and M. Schleuning. 2012. Plant-frugivore networks are less specialized and more robust at forest-farmland edges than in the interior of a tropical forest. Oikos 121:1553-1566. Mickleburgh, S. P., A. M. Hutson, and P. A. Racey (Eds.). 1992. Old World Fruit Bats: An Action Plan for their Conservation. IUCN/SSC Chiroptera Specialist Group, Gland, Switzerland. Mildenstein, T. L., S. C. Stier, C. E. Nuevo-Diego, and L. S. Mills. 2005. Habitat selection of endangered and endemic large flying-foxes in Subic Bay, Philippines. Biological Conservation 126:93-102. Mittermeier, R.A., F. Hawkins, E. E. Louis, M. Richardson, C. Schwitzer, O. Langrand, A. B. Rylands, F. Hawkins, S. Rajoabelina, J. Ratsimbazafy, R. Rasoloarison, C. Roos, P. M. Kappeler, and J. Mackinson. 2010. Lemurs of Madagascar, 3rd Edition. Conservation International, Arlington, USA. Moat, J., and P. P. Smith. 2007. Atlas of the vegetation of Madagascar/Atlas de la vegetation de Madagascar. Royal Botanic Gardens, Kew, London, United Kingdom. Mutschler, T., A. J. Randrianarisoa, and A. T. C. Feistner. 2001. Population status of the Alaotran gentle lemur Hapalemur griseus alaotrensis. Oryx 35:152-157. Myers, N., R. A. Mittermeier, C. G. Mittermeier, G. A. B. da Fonseca, and J. Kent. 2000. Biodiversity hotspots for conservation priorities. Nature 403:853-858. Nagy, K. A., I. A. Girard, and T. K. Brown. 1999. Energetics of free-ranging mammals, reptiles, and birds. Annual Review of Nutrition 19:247-277. 217 218 Nasi, R., A. Taber, and N. van Vliet. 2011. Empty forests, empty stomachs? Bushmeat and livelihoods in the Congo and Amazon Basins. International Forestry Review 13:355-368. Nekaris, K. A. I., C. R. Shepherd, C. R. Starr, and V. Nijman. 2010. Exploring cultural drviers for wildlife trade via an ethnoprimatological approach: a case study for slender and slow lorises (Loris and Nycticebus) in South and Southeast Asia. American Journal of Primatology 72:877-886. Nievergelt, C. M., J. Pastorini, and D. S. Woodruff. 2002. Genetic variability and phylogeography in the wild Alaotran Gentle Lemur population. Evolutionary Anthropology. 11(Suppl 1):175-179. Nijman, V., K. A. I. Nekaris, G. Donati, M. Bruford, and J. Fa. 2011. Primate conservation: measuring and mitigating trade in primates. Endangered Species Research 13:159-161. Novy, J.W. 1997. Medicinal plants of the eastern region of Madagascar. Journal of Ethnopharmacology 55:119-126. Okello, J. (2010) Morphological and nutritional characteristics of Tamarindus indica (Linn) fruits in Uganda. Dissertation. School of Graduate Studies, Agroforesty. Makerere University. Available online: http://news.mak.ac.ug/documents/Makfiles/theses/Okello_Jaspher.pdf (accessed 15 November 2012) Oleksy, R., P. A. Racey, and G. Jones. 2015. High-resolution GPS tracking reveals habitat selection and the potential for long-distance seed dispersal by Madagascan flying foxed Pteropus rufus. Global Ecology and Conservation 3: 678-692. Ossa G., S. Kramer-Schadt, A. J. Peel, A. K. Scharf, and C. C. Voigt. 2012. The Movement Ecology of the Straw-Colored Fruit Bat, Eidolon helvum, in Sub- Saharan Africa Assessed by Stable Isotope Ratios. PLoS ONE 7: e45729. Perez, V. R., L. R. Godfrey, M. Nowak-Kemp, D. A. Burney, J. Ratsimbazafy, and N. Vasey. 2005. Evidence of early butchery of giant lemurs in Madagascar. Journal of Human Evolution 49:722-742. Petter, J. J. 1969. Speciation of Madagascan lemurs. Biological Journal of the Linnean Society 1:77-84. Pfeifer, M., N. D. Burgess, R. D. Swetnam, P. J. Platts, S. Willcock, and R. Marchant. 2012. Protected areas: mixed success in conserving East Africa’s evergreen forests. PLoS ONE 7:e39337. 218 219 Picot, M., R. K. B. Jenkins, O. Ramilijaona, P. A. Racey, and S. M. Carrière. 2007. The feeding ecology of Eidolon dupreanum (Pteropodidae) in eastern Madagascar. African Journal of Ecology 45:645-650. Pierson, E. D., and W. E. Rainey. 1992. The biology of flying foxes of the genus Pteropus: a review. In: Pacific Island Flying Foxes: Proceedings of an International Conservation Conference (eds. D. E. Wilson and G. L. Graham). U.S. Fish and Wildlife Service, Washington, D.C., USA. Pimental, D., and M. Pimentel. 2003. Sustainability of meat-based and plant-based diets and the environment. The American Journal of Clinical Nutrition 78:660S-663S. Plein, M. L. Längsfeld, E. L. Neuschulz, C. Schultheiss, L. Ingmann, T. Töpfer, K. Böhning-Gaese, and M. Schleuning. 2013. Constant properties of plant-frugivore networks despite fluctuations in fruit and bird communities in space and time. Ecology 94:1296-1306. Ploch, L., and Cook, N. 2012. Madagascar's Political Crisis – R40448: Congressional Research Service. Available online : http://fas.org/sgp/crs/row/R40448.pdf (accessed 24 March 2015) Poulsen, J. R., C. J. Clark, G. Mavah, and P. W. Elkan. 2009. Bushmeat supply and consumption in a tropical logging concession in Northern Congo. Conservation Biology 23:1597-1608. R Core Team. 2013. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Rakotoarivelo, A. R., J. H. Razafimanahaka, S. Rabesihanaka, J. P. G. Jones, and R. K. B. Jenkins. 2011. Lois et règlements sur la faune sauvage à Madagascar: Progrès accomplis et besoins du future. Madagascar Conservation and Development 6:37- 44. Rakotondrasoa, O. L., F. Malaisse, G. L. Rajoelison, T. M. Razafimanantsoa, M. R. Rabearisoa, B. S. Ramamonjisoa, N. Raminosoa, F. J. Verheggen, M. Poncelet, É. Haubruge, and J. Bogaert. 2012. Tapia forest, endemic ecosystem to Madagascar: ecology, functions, causes of degradation and transformation: a review. Biotechnologie, Agronomie, Société et Environnement 16:541-552. Randrianandrianina, F. H., P. A. Racey, and R. K. B. Jenkins. 2010. Hunting and consumption of mammals and birds by people in urban areas of western Madagascar. Oryx 44:411-415. 219 220 Randrianarisoa, P.M., A. A. Rasamison, and L. Rakotozafy. 1999. Les Lémuriens de la région de Daraina: forêt d’Analamazabaza, forêt de Bekaraoka et forêt de Sahaka. Lemur News 4:19–21. Razafimanahaka, J. H., R. K. B. Jenkins, D. Andriafidison, F. Randriananrianina, V. Rakotomboavonjy, A. Keane, and J. P. G. Jones. 2012. Novel approach for quantifying illegal bushmeat consumption reveals high consumption of protected species in Madagascar. Oryx 46:584-592. Rajaonson, A., M. Ratolojanahary, J. Ratsimbazafy, A. Feistner, and T. King. 2010. Enquête préliminaire de la distribution des lémuriens de bambou dans et autour du Corridor forestier Fandriana-Vondrozo, Madagascar. Lemur News 15:34-39. Reuter, K. E., H. Gilles, A. R. Wills, and B. J. Sewall. 2015. Live capture and ownership of lemurs in Madagascar: extent and conservation implications. Oryx doi:10.1017/S003060531400074X Rex, K., R. Michener, T. H. Kunz, and C. C. Voigt. 2011. Vertical stratification of neotropical leaf-nosed bats (Chiroptera: Phyllostomidae) revealed by stable carbon isotopes. Journal of Tropical Ecology 27:211-222. Rietbergen-McCracken, J., and D. Narayan. 1998. Participation and social assessment: tools and techniques. The World Bank, Washington, D.C., USA. Rigamonti, M.M. 1996. Red ruffed lemur (Varecia variegate rubra): A rare species from the Masoala Rain Forests. Lemur News 2:9-11. Rist, J., E. J. Milner-Gulland, G. Cowlishaw, and M. Rowcliffe. 2010. Hunter reporting of catch per unit effort as a monitoring tool in a bushmeat-harvesting system. Conservation Biology 24:489-499. Robinson, J. G., and E. L. Bennett. 2000. Carrying capacity limits to sustainable hunting in tropical forests. In: Hunting for sustainability in tropical forests (eds. J. G. Robinson and E. L. Bennett). Columbia University Press, New York, USA. Sarter, S., G. Sarter, and P. Gilabert. 2010. A swot analysis of HACCP implementation in Madagascar. Food Control 21:253-259. SAS Institute. 2015. JMP 12 Online Documentation. Nonparametric Analysis. Available online: http://www.jmp.com/support/help/Nonparametric.shtml (accessed 1 June 2015) Sauther, M. L., F. P. Cuozzo, I. A. Youssouf Jacky, K. D. Fish, M. LaFleur, L. A. L. Ravelohasindrazana, and J. F. Ravoavy. 2013. Limestone cliff-face and cave use 220 221 by wild ring-tailed lemurs (Lemur catta) in southwestern Madagascar. Madagascar Conservation and Development 8:73-80. Scales, I. R. 2014. The future of conservation and development in Madagascar: time for a new paradigm? Madagascar Conservation and Development 9:5-12. Scanlon, A., T. P. Sophie, and L. da S. Stenberg. 2013. Insectivory in Fijian flying foxes (Pteropodidae). Australian Journal of Zoology 61:342-349. Schwitzer C., R. A. Mittermeier, N. Davies, S. Johnson, J. Ratsimbazafy, J. Razafindramanana, E. E. Louis Jr., and S. Rajaobelina (Eds). 2013. Lemurs of Madagascar: A Strategy for Their Conservation 2013–2016. IUCN SSC Primate Specialist Group, Bristol Conservation and Science Foundation, and Conservation International, Bristol, United Kingdom. Schwitzer, C., R. A. Mittermeier, S. E. Johnson, G. Donati, M. Irwin, H. Peacock, J. Ratsimbazafy, J. Razafindramanana, E. E. Louis Jr., L. Chikhi, I. C. Colquhoun, J. Tinsman, R. Dolch, M. LaFleur, S. Nash, E. Patel, B. Randrianambinina, T. Rasolofoharivelo, and P. C. Wright. 2014. Averting lemur extinctions amid Madagascar’s political crisis. Science 343:842-843. Sewall, B. J., E. F. Granek, and W. J. Trewhella. 2003. The endemic Comores Islands fruit bat Rousettus obliviosus: ecology, conservation, and Red List status. Oryx 37:344-352. Sewall, B. J., A. L. Freestone, M. F. E. Moutui, N. Toilibou, I. Said, S. M. Toumani, D. Attoumane, and C. M. Iboura. 2011. Reorienting systematic conservation assessment for effective conservation planning. Conservation Biology 25:688- 696. Sewall B. J., A. L. Freestone, J. E. Hawes, and E. Andriamanarina. 2013. Size-Energy Relationships in Ecological Communities. PLoS ONE 8:e68657. Shanahan, M., S. So, S. G. Compton, and R. Corlett. 2001. Fig-eating by vertebrate frugivores: a global review. Biological Reviews 76:529-572. Shepherd, C. R. 2010. Illegal primate trade in Indonesia exemplified by surveys carried out over a decade in North Sumatra. Endangered Species Research 11:201-205. Solly, H. 2007. Cameroon: from free gift to valued commodity – the bushmeat commodity chain around the Dja Reserve. In: Bushmeat and Livelihoods: Wildlife Management and Poverty Reduction (eds. G. Davies and D. Brown). Wiley-Blackwell, Hoboken, NJ, USA. 221 222 Soto-Centeno, J. A. and A. Kurta. 2006. Diet of two nectarivorous bats, Erophylla sezekorni and Monophyllus redmani (Phyllostomidae), on Puerto Rico. Journal of Mammalogy 87:19-26. Stiles, D. 2011. Elephant meat trade in Central Africa: Democratic Republic of Congo case study. IUCN, Gland, Switzerland. Styger, E., J. E. M. Rakotoarimanana, R. Rabevohitra, and E. C. M. Fernandes. 1999. Indigenous fruit trees of Madagascar: potential components of agroforestry systems to improve human nutrition and restore biological diversity. Agroforestry Systems 46:289-310. Tabarelli, M., W. Mantovani, and C. A. Perez. 1999. Effects of habitat fragmentation on plant guild structure in the montane Altantic forest of southeastern Brazil. Biological Conservation 91:119-127. Tattersall, I. 1998. Lemurs of the Comoro Archipelago: Status of Eulemur mongoz on Moheli and Anjouan, and of Eulemur fulvus on Mayotte. Lemur News 3:15-17. Thomas, L., S. T. Buckland, E. A. Rexstad, J. L. Lakke, S. Strindberg, S. L. Hedley, J. R. B. Bishop, T. A. Marques, and K. P. Bumham. 2010. Distance survey: design and analysis of distance sampling surveys for estimating population size. Journal of Applied Ecology 47:5-14. Tidemann, C. R., and J. E. Nelson. 2004. Long-distance movements of the grey-headed flying-fox (Pteropus polioce phallus). Journal of Zoology 263: 141-146. Tucker, B. 2007. Applying behavioral ecology and behavioral economics to conservation and development planning: An example from the Mikea Forest, Madagascar. Human Nature 18:190-208. Tylianakis, J. M., E. Laliberte, A. Nielsen, and J. Bascompte. 2010. Conservation of species interaction networks. Biological Conservation 143:2270-2279. United Nations Treasury. 2014. Madagascar: UN Operational Rates of Exchange. Available online: http://treasury.un.org/operationalrates/OperationalRates.aspx (accessed 4 June 2014) United Nations [UN]. 1973. Convention on International Trade in Endangered Species of Wild Fauna and Flora. United Nations. Washington, D.C., USA. United Nations Development Programme [UNDP]. 2013. Human development report 2013: The rise of the south: Human progress in a diverse world. United Nations Development Programme, New York, USA. 222 223 U.S. Department of the Interior and U.S. Geological Survey. 2014. The USGS Earth Explorer. Available online: http://earthexplorer.usgs.gov/ (accessed 10 January 2015) van Vliet, N., C. Nebesse, S. Gambalemoke, D. Akaibe, and R. Nasi. 2012. The bushmeat market in Kisangani, Democratic Republic of Congo: implications for conservation and food security. Oryx 46:196-203. Vasey, N. 1996. Clinging to life: Varevia variegate rubra and the Masoala Coastal Forests. Lemur News 2:7-9. Vasey, N., and I. Tattersall. 2002. Do ruffed lemurs form a hybrid zone? Distribution and discovery of Varecia, with systematic and conservation implications. American Museum Novitates 3376:1-26. Vazquez, D. P., N. P. Chacoff, and L. Cagnolo. 2009. Evaluating multiple determinants of the structure of plant-animal mutualistic networks. Ecology 90:2039-2046. Vidal, M., M. M. Pires, and P. R. Guimaraes. 2013. Large vertebrates as the missing components of seed dispersal networks. Biological Conservation 163:42-48. Voigt, C. C., and F. Matt. 2004. Nitrogen stress caused unpredictable enrichments of 15N in two nectar-feeding bat species. The Journal of Experimental Biology 207:1741- 1748. Voigt, C. C., S. L. Voigt-Heucke, and A. S. Kretzschmar. 2012. Isotopic evidence for seed transfer from successional areas into forests by short-tailed fruit bats (Carollia spp.; Phyllostomidae). Journal of Tropical Ecology 28:181-186. Walsh, A. 2005. The obvious aspects of ecological underprivilege in Ankarana, Northern Madagascar. American Anthropologist 107:654-665. Walsh, M. T. 2007. Island subsistence: hunting, trapping and the translocation of wildlife in the Western Indian Ocean. Azania: Archaeological Research in Africa 42:83- 113. Welch, C. 1996. Projects at Ivoloina and Betampona in Eastern Madagascar. Lemur News 2:11. Westerman, K., and C. J. Gardner. 2013. Adoption of socio-cultural norms to increase community compliance in permanent marine reserves in southwest Madagascar. Conservation Evidence 10:4-9. 223 224 Wilkie, D. S., E. L. Bennett, C. A. Peres, and A. A. Cunningham. 2011. The empty forest revisited. Annals of the New York Academy of Sciences 1223:120–128. Wilkie, D. S., and J. Carpenter. 1999. Bushmeat hunting in the Congo Basin: an as- sessment of impacts and options for mitigation. Biodiversity and Conservation 8:927–955. Wilson, J. M., P. D. Stewart, G.-Z. Ramangason, A. M. Denning, and M. S. Hutchings. 1989. Ecology and conservation of the crowned lemur, Lemur coronatus, at Ankarana, N. Madagascar. Folia Primatol 59:1-26. Winter, K. 1979. δ13C values of some succulent plants from Madagascar. Oecologia 40:103-112. World Bank. 2006. Scaling up marine management – the role of marine protected areas: Report No. 36636-GLB. The World Bank, Washington, DC, USA. World Bank. 2013. Madagascar: measuring the impact of the political crisis. Available online : http://www.worldbank.org/en/news/feature/2013/06/05/madagascar- measuring-the-impact-of-the-political-crisis (accessed 24 March 2015) World Bank. 2015. Urban population (% of total). Available online: http://data.worldbank.org/indicator/SP.URB.TOTL.IN.ZS (accessed 1 June 2015) Wright, S. J. 2005. Tropical Forests in a Changing Environment. Trends in Ecology and Evolution 20:553-560. Wright, S. J., K. E. Stoner, N. Beckman, R. T. Corlett, R. Dirzo, H.C. Muller-Landau, G. Nunez-Iturri, C. A. Peres, and B. C. Wang. 2007. The plight of large animals in tropical forests and the consequences for plant regeneration. Biotropica 39:289- 291. Wunderle, J. M. J. 1997. The role of animal seed dispersal in accelerating native forest regeneration on degraded tropical lands. Forest Ecology and Management 99:223- 235. Yang, S., R. Albert, and T. A. Carlo. 2013. Transience and constancy of interactions in a plant-frugivore network. Ecosphere 4:A147. Yohannes, E., B. Hansson, R.W. Lee, J. Waldenstrom, H. Westerdahl, M. Akesson, D. Hasselquist, and S. Bensch. 2008. Isotope signatures in winter moulted feathers predict malaria prevalence in a breeding avian host. Oecologia 158:299-306. 224 225 Zinner, D., J. Ostner, A. Dill, L. Razafimanantsoa, and R. Rasoloarison. 2001. Results of a reconnaissance expedition in the western dry forests between Morondava and Morombe. Lemur News 6:16-18. 225 226 APPENDIX A CHAPTER 2 SUPPLEMENTARY MATERIALS Figure S1: The locations of transects in the degraded (red), secondary (yellow), and primary forests (green) relative to the location of Mahamasina (lower right-hand corner), one of the three villages with a designated park entrance. 226 227 Table S1: The total length of time that frugivores were observed spending time in fruiting trees, across the different forest types. Time is listed in hours:minutes:seconds (HH:MM:SS). Frugivore Common Primary Secondary Degraded Total species Name forest forest forest Lesser Coracopsis nigra 0:12:14 4:15:31 - 4:27:45 Vasa Parrot Greater Coracopsis vasa 4:27:25 1:37:44 - 6:05:09 Vasa Parrot Eulemur Crowned 11:02:31 3:12:39 0:53:38 15:08:48 coronatus Lemur Brown Eulemur sanfordi 3:05:22 17:59:15 - 21:04:37 Lemur Hypsipetes Madagascar 21:53:17 5:23:08 0:56:02 27:32:27 madagascariensis Bulbul Saroglossa Madagascar 0:06:47 0:18:12 - 0:24:59 aurata Starling Madagascar Treron australis Green 20:39:30 8:08:40 0:49:10 29:37:20 Pigeon 227 228 Table S2: Information on fruiting trees identified across three forest types. All fruiting tree individuals with a diameter at breast height larger than 10 cm were recorded. Average DBH and height estimates are rounded to the nearest centimeter and meter, respectively. The numbers 1°, 2°, 3° refer to the primary, secondary, and degraded forest, respectively. The average fruit crop size (number of fruits per individual tree) is rounded to the nearest whole fruit (mean ± st. dev.). Species in bold were selected for inclusion in our sampling, and (apart from B. commiphora, due to completion of fruiting prior to sampling) were the target of focal observations and included in our mutualistic network analysis. Average fruit crop Species Density (per Number detected Average DBH (cm) Average height (m) size per individual Scientific name Local name hectare) tree 1° 2° 3° 1° 2° 3° 1° 2° 3° 1° 2° 3° Unknown Anosaly - 3 3 - 2 1.32 - 15 24 - 10 8 15 (+/-13) Baudouinia Baudouinia 73 - - 32.16 - - 22 - - 23 - - 84 (+/-85) fluggeiformis Croton sp. Croton A - 9 5 - 2.2 6 - 13 15 - 10 5 243 (+/-) 479 Croton sp. Croton B - - 1 - - 0.44 - - 13 - - 3 512 (+/ 448) Dichrostachys sp. Famoha 2 - - 0.88 - - 26 - - 40 - - 187 (+/- 45) Ficus grevei Ficus A 5 2 - 2.2 1.33 - 119 974 - 21 36 - 48,657 (+/- 45,511) Ficus lutea Ficus Lutea 2 - - 0.88 - - - - - 10 - - 346 (+/- 154) Neobeguea sp. Gavoala 1 - - 0.44 - - 39 - - 45 - - 384 Grewia sp. Grewia 4 24 143 1.76 16.67 62.99 16 20 24 26 13 9 692 (+/- 1519) Unknown Joby Ampototra 1 - - 0.44 - - 17 - - 10 - - Few fruit Capuronia sp. Kitata - 1 2 - 0.67 0.88 - 15 14 - 25 6 410 (+/- 395) Unknown Lamonty Ala 1 - - 0.44 - - 13 - - 20 - - 192 (+/- 81) Flacourtia indica Lamonty Gomo Gomo - - 2 - - 0.88 - - 17 - - 8 20 (+/- 11) Unknown Mafay 1 - - 0.44 - - 17 - - 40 - - 42 Mangifera indica Mango - - 4 - - 1.76 - - 36 - - 12 Not fully ripe Mantalania sp. Mantalany 9 1 - 3.97 0.67 - 14 11 - 16 15 - 13 (+/- 12) Burseraceae Mantamo 1 - - 0.44 - - 16 - - 15 - - 384 commiphora 228 229 Unknown Mapingo 2 - - 0.88 - - 36 - - 40 - - 229 (+/- 69) Poupartia sp. Paessoala 1 - - 0.44 - - 95 - - 40 - - 75 Unknown Rubaiceaea - - 1 - - 0.44 ------10 (Rubiaceae) Unknown Sarangaravatsy - - 7 - - - - 17 - - 4 52 (+/- 34) Majidea Soopinolaeea/ 1 - - 0.44 - - 40 - - 45 - - 100 zanguebarica Magiadea Unknown Taimbarika 3 - - 1.32 - - 17 - - 16 - - 31 (+/- 23) Tamarindus Tamarind - 5 5 - 2.2 3.33 - 27 30 - 17 9 44 (+-/48) indica Unknown Tapinengo 1 - - 0.44 - - 14 - - 30 - - 4 (end of fruiting) Tranonomboko Sely Unknown 3 - - 1.32 - - 18 - - 24 - - 76 (+/- 78) Malaly Unknown Unknown A 1 - - 0.44 - - 14 - - 25 - - 950 Unknown Unknown B 1 - - 0.44 - - 27 - - 30 - - 300 Strychnos Vacacoa 35 - - 15.41 - - 42 - - 22 - - 57 (+/- 55) madagascariensis Pittosporum sp. Vitaka - 18 14 - 12 6.17 - 15 15 - 9 7 835 (+/- 930) Vitex sp. Vitex - 3 2 - 1.32 0.88 - 19 21 - 14 15 19 (+/- 18) Unknown Wart Trunk 5 - - 2.2 - - 13 - - 10 - - 66 (+/- 26) (Tilaceae) 229 230 Table S3: A list of tree species (alphabetical by common name) included in the mutualistic network, identified by their scientific and common names, with average fruit biomass (g) and average fruit seed count. Given that there is a strong correlation between fruit mass and seed count among the Ficus species (BJS, unpublished data), the seed count data for Ficus grevei was calculated from the measured mass of this species and a regression formula of the relationship between Ficus fruit mass and seed count. The Tamarind seed count (denoted by ^) came from Okello 2010. The numbers 1°, 2°, 3° refer to the primary, secondary, and degraded forest, respectively. Average Average fruit Time spent observing for Common/ fruit seed Species frugivore feeding (hrs) Field Name biomass count (g) (seeds per fruit) 1° 2° 3° Unknown Anosaly 14.4 12.5 - 33 21 Baudouinia Baudouinia 5.72 0.63 45 - - fluggeiformis Croton sp. Croton 1.79 2.4 - 39 41 Ficus grevei Ficus grevei 1.72 268 21.55 41 - Ficus lutea Ficus lutea 0.9 323 20.57 - - Grewia sp. Grewia 0.3 3.6 20.5 25.38 24.36 Mantalania sp. Mantalany 33.2 7 21 21 - Poupartia sp. Paessoala 50.67 0.5 20.65 - - Tamarindus Tamarind 10.43 5.1^ - 24 24 indica Tranonomboko Unknown 7.85 0.57 21 - - Sely Malaly Strychnos Vacacoa 24.93 5.69 39.8 - - madagascariensis Pittosporum sp. Vitaka 1.51 3.5 - 55.39 33 230 231 Table S4: Density estimates per square kilometer for the diurnal frugivore guild at Ankarana National Park. Under the density column, the number in the parentheses is the Coefficient of Variation. The density estimates represent the calculated lower and upper bound density estimates using Distance software (Thomas et al. 2010). Frugivore Forest Number Density (per Confidence Species Type Observed km2) Interval Eulemur Primary 62 208.2 (41.3) 151; 284 coronatus Secondary 52 276 (33.9) 217; 354 Crowned lemur Degraded 17 105 (47.4) 98; 201 Primary 29 109 (36.7) 32; 197 Eulemur sanfordi Secondary 42 232.8 (36.2) 169; 309 Sanford’s lemur Degraded 15 163.1 (81.3) 154; 265 Coracopsis nigra Primary 13 44.7 (43.2) 20; 102 Lesser vasa Secondary 2* 8.5 (71)* 2; 30* parrot Degraded 9* 32.4 (78.4)* 8; 130* Coracopsis vasa Primary 31 86.6 (25) 53; 141 Greater vasa Secondary 12 91.9 (55.3) 33; 256 parrot Degraded 10 100 (43.5) 44; 231 Hypsipetes Primary 64 609.9 (27.8) 356; 1045 madagascariensis Secondary 41 235 (26.4) 141; 393 Bulbul Degraded 108 1204 (23.7) 758; 1915 Saroglossa Primary present** present** present** aurata Secondary 1 present** present** Starling Degraded 5 present** present** Primary 17 83.3 (37.7) 41; 171 Treron australis Secondary 21 210.6 (52.9) 77; 565 Green pigeon Degraded 17 483.7 (47) 196; 1193 *The Coracopsis nigra density estimates for the secondary and degraded forests in 2012 were calculated using detection function for the Coracopsis vasa populations in the same forest types. ** The Saroglossa aurata populations in 2012 were too small to calculate densities, and detection functions could not be borrowed from other bird species. We did not observe this species in the primary forest during our distance sampling work, but it was observed in the primary forest while doing unrelated work in 2012. 231 232 Table S5: The number of visits observed at fruiting trees per frugivore species, across all forest types. Coracopsis Coracopsis Eulemur Eulemur Hypsipetes Saroglossa Treron Tree species nigra vasa coronatus sanfordi madagascariensis aurata australis Degraded forest Anosaly ------Croton sp. - - - - 6 - - Grewia sp. - - 3 - 19 - 2 Tamarindus indica - - - - 2 - 12 Pittosporum sp. - - - - 3 - - Secondary forest Anosaly - - 1 - 4 - - Croton sp. - - 8 3 - - - Ficus grevei 7 3 42 46 13 1 22 Grewia sp. - - 3 - 6 - - Mantalania sp. - 2 - - - - - Tamarindus indica - - 2 - 6 1 2 Pittosporum sp. - - 5 4 18 1 - Primary forest Baudouinia fluggeiformis - - 9 1 - - - Ficus grevei 2 6 33 13 44 2 68 Ficus lutea - - 2 - 30 - - Grewia sp. - - 5 - 1 - - Mantalania sp. - - 9 - 1 - - Poupartia sp. - - - - - 1 - Tranonomboko Sely - - 1 - - - - Malaly Strychnos - 1 3 - 5 - - madagascariensis 232 233 Table S6: Scaled calculations (0 to 1) of effectiveness measures, showing the potential benefits received by frugivores from trees (fruit biomass consumption) and by trees from frugivores (potential seed dispersal). See methods for description of interaction frequency calculations. Zero indicates no benefits were exchanged between partners, and a one indicates that the pairwise mutualism provides the highest guild-specific benefit to a mutualistic partner. Where mutualisms were realized, boxes are colored on a four-point orange-scale, with colors become progressively darker as the interaction effectiveness increases. The four classes of interaction effectiveness were defined as: 0 – 0.10; 0.11 – 0.49; 0.50 – 0.75; and 0.76 – 1.00. Boxes shaded in blue indicate that the frugivore was observed visiting the tree, but never observed consuming or removing fruit. Green boxes indicate frugivore was observed eating fruit in the tree but not during focal observations. “FB” (Frugivore Benefits) columns, while interaction frequencies for benefits received by trees are listed under the “TB” (Tree Benefits) columns. Greater Lesser Brown Bulbul Green Pigeon Crowned Starling Vasa Vasa lemur Degraded Forest FB TB FB TB FB TB FB TB FB TB FB TB FB TB Anosaly 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Croton sp. >0 >0 0 0 0 0 0 0 0 0 0 0 0 0 Grewia sp. 1 0.05 0.33 0.02 0 0 0 0 0.73 0.04 0 0 0 0 Tamarindus indica 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Vitex sp. 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Secondary Forest Anosaly >0 >0 0 0 0 0 0 0 0 0 0 0 0 0 233 234 Croton sp. 0 0 0 0 0 0 0 0 0 0 >0 >0 0 0 Ficus grevei >0 0.03 >0 0.03 >0 >0 >0 0.03 0.02 0.08 0.02 0.09 0 0 Grewia sp. >0 >0 0 0 0 0 0 0 0 0 0 0 0 0 Mantalania sp. 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Tamarindus indica 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Vitex sp. 0.07 >0 0 0 0 0 0 0 0 0 0.07 >0 0 0 Primary Forest Baudouinia fluggeiformis 0 0 0 0 0 0 0 0 0.23 >0 0.09 >0 0 0 Ficus grevei 0.08 0.32 0.27 1 0.04 0.16 >0 0.01 0.25 0.95 0.14 0.52 >0 >0 Ficus lutea >0 >0 0 0 0 0 0 0 >0 >0 0 0 0 0 Grewia sp. >0 >0 0 0 0 0 0 0 >0 >0 0 0 0 0 Mantalania sp. 0 0 0 0 0 0 0 0 0.29 >0 0 0 0 0 Poupartia sp. 0 0 0 0 0 0 0 0 0 0 0 0 >0 >0 Strychnos 0 0 0 0 0 0 0 0 0.4 >0 0 0 0 0 madagascariensis Tranonomboko Sely 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Malaly 234 235 Table S7: Number of frugivores visually observed to pass within 15 meters of a focal fruiting tree per hour, without entering the tree’s canopy. These data were collected during focal observations. Data were not collected for Ficus grevei due to the high feeding preference exhibited by frugivores for this species; in other words, frugivore activity was high enough in Ficus grevei trees that observers could not simultaneously record observations of frugivores passing near trees without entering the canopy. Primary Secondary Degraded Forest Forest Forest Anosaly - 1 0.72 Boudouinia 1.6 - - fluggeiformis Croton sp. - 0.22 1 Ficus lutea 2.52 - - Mantalania sp. Not measured 3.89 - Grewia sp. 0 2 1.25 Pittosporum sp. - 1.17 1.05 Poupartia sp. 0.67 - - Tamarindus indica - 0.92 0.33 Tranonomboko Sely 1.22 - - Malaly Strychnos 1.28 - - madagascariensis 235 236 APPENDIX B MEAT CONSUMPTION SURVEYS (HOUSEHOLDS) Section One: In this section, respondents were asked about their wild meat and domestic meat consumption. Interviewers reminded respondents that we were interested in all kinds of meat consumed, including gifts, wild meat, purchased meat, and farmed/domestic meat. 1. Please list all of the different meats, not including fish, which you have eaten in the past three days. The following follow-up questions were asked for each meat named: a. How did you get this meat? (Answer choices: Purchase, Caught, Gift, Raised) b. If purchased: How much did it cost? Where did you purchase it? c. If caught: Where and when did you catch it? 2. Please list your top five favorite types of meat, not including fish. The following follow-up questions were asked for each meat listed: a. How often do you get to eat it? (We encouraged respondents to provide weekly or monthly estimates of consumption frequencies) b. Do you usually buy it? For what price? Where do you buy it? c. Do you usually catch it? Where do you catch it? d. Is this something you eat for a special occasion? Why do you consume this meat? Section Two: 3. Have you ever eaten a tenrec before? If yes, the follow-up questions were asked. a. How often do you get to eat it per year? How often have you eaten it in your lifetime? (We encouraged respondents to provide weekly or monthly estimations of consumption frequencies if they could not provide a yearly estimate) b. When did you last eat it? c. Do you usually buy it? For what price? Where do you buy it? d. Do you usually catch it? Where do you catch it? e. Is this something you eat for a special occasion? Why do you consume this meat? 4. Have you ever eaten bat before? (See question 3 for follow-up questions) 236 237 5. Have you ever eaten fossa before? (See question 3 for follow-up questions) 6. Have you ever eaten mongoose before? (See question 3 for follow-up questions) 7. Have you ever eaten rats and mice before? (See question 3 for follow-up questions) 8. Have you ever eaten civets before? (See question 3 for follow-up questions) 9. Have you ever eaten wild cats before? (See question 3 for follow-up questions) 10. Have you ever eaten wild pigs before? (See question 3 for follow-up questions) 11. Have you ever eaten lemur before? (See question 3 for follow-up questions) Section Three: 12. Have you changed the type of meat that you eat? If yes, how? If yes, why? Example: did you formerly eat a lot of zebu, but now you eat fish? Or did you formerly eat a lot of wild meat, but now you eat chicken? 13. Do you have any taboos (fady) about eating any animals in your area? Who follows these fadys (an individual family, ethnicity, or the whole village?)? What are the taboos? 237 238 APPENDIX C INTERVIEW QUESTIONS FOR MEAT SELLERS Section One (meat sales in the past three days): In this section, respondents were asked about their wild meat and domestic meat sales in the past three days. 1. Please list all of the different meats, not including fish, which you have sold in the past three days. The following follow-up questions were asked for each meat named: a. How many animals/kilograms have you sold in the past three days? b. Do you sell it dead or alive? c. How did you get this meat? (Answer choices: Purchase, Caught, Gift, Raised) d. If purchased: How much did it cost? Where did you purchase it? e. If caught: Where and when did you catch it? Section Two (history of selling wild meat): 2. Have you ever sold a tenrec before? If yes, the follow-up questions were asked. a. When you sell it, how many do you sell (per day or per week)? b. When did you last sell it? c. Do you usually buy it? For what price? Where do you buy it? d. Do you usually catch it? Where do you catch it? e. What price do you sell tenrec for? 3. Have you ever sold bat before? (See question 2 for follow-up questions) 4. Have you ever sold fossa before? (See question 2 for follow-up questions) 5. Have you ever sold mongoose before? (See question 2 for follow-up questions) 6. Have you ever sold rats and mice before? (See question 2 for follow-up questions) 7. Have you ever sold civets before? (See question 2 for follow-up questions) 8. Have you ever sold wild cats before? (See question 2 for follow-up questions) 9. Have you ever sold wild pigs before? (See question 2 for follow-up questions) 10. Have you ever sold lemur before? (See question 2 for follow-up questions) Section Three: 11. Have you changed the type of meat that you sell? If yes, how? If yes, why? 238 239 Example: did you formerly sell a lot of zebu, but now you sell a lot of fish? Or did you formerly sell a lot of wild meat, but now you sell more chicken? 12. Do you have any taboos (fady) about eating any animals in your area? Who follows these fadys (an individual family, ethnicity, or the whole village)? What are the taboos? 239 240 APPENDIX D INTERVIEW QUESTIONS FOR TRANSPORTERS Section One (transport of meat in the past three days): In this section, respondents were asked about their wild meat and domestic meat transport in the past three days. 1. Please list all of the different meats, not including fish, which you have transported in the past three days. The following follow-up questions were asked for each meat named: a. How many animals/kilograms have you transported in the past three days? b. Do you transport it dead or alive? c. How much do you charge for the transport of the animals? d. What is the transportation route (what is the name of the place where the meat is picked up and what is the name of the place where the meat is dropped off)? Section Two (history of transporting wild meat): 2. Have you ever transported a tenrec before? If yes, the follow-up questions were asked. a. When did you last transport it? b. How many did you transport (animals or kilograms) per trip? c. Did you transport it/them alive or dead? d. How much did you charge for the transport of the animals? e. What was the transportation route (what is the name of the place where the meat is picked up and what is the name of the place where the meat is dropped off)? 3. Have you ever transported bat before? (See question 2 for follow-up questions) 4. Have you ever transported fossa before? (See question 2 for follow-up questions) 5. Have you ever transported mongoose before? (See question 2 for follow-up questions) 6. Have you ever transported rats and mice before? (See question 2 for follow-up questions) 7. Have you ever transported civets before? (See question 2 for follow-up questions) 8. Have you ever transported wild cats before? (See question 2 for follow-up questions) 240 241 9. Have you ever transported wild pigs before? (See question 2 for follow-up questions) 10. Have you ever transported lemur before? (See question 2 for follow-up questions) 241 242 APPENDIX E CHAPTER 5 SUPPLEMENTARY MATERIALS Supplementary Information: The legality of hunting wild animals in Madagascar: In Madagascar, the legality of hunting wild animals is dictated through a three-tier classification system. Animals are classified into three categories: 1) protected (can never be hunted), 2) harmful and nuisance animals (can be hunted year-round; animals which damage crops or kill domestic animals); and 3) game animals (hunted during a multi- month hunting season that is set yearly by the government; Rakotoarivelo et al. 2011). Some methods of hunting are prohibited in Madagascar. For example, it is never legal to hunt at night, to use poisoned bait or sedatives, explosives, with fire (Rakotoarivelo et al. 2011), or by using nets and pit traps (Ordonnance nº 60-126). Hunting with “weapons of local manufacture” including spears, bows, and blowguns are authorized, but only for non-commercial purposes (Ordonnance nº 60-126) and when permitted (Rakotoarivelo et al. 2011). Extra permits are needed to for the use of firearms and for commercial and scientific hunting (Rakotoarivelo et al. 2011). As in many African countries (Lindsey et al. 2013), hunting in Madagascar is regulated using legal instruments, with the right to hunt belonging to the state and allowed through a permit-based hunting system (Rakotoarivelo et al. 2011). The same regulations which apply to hunters also apply to the “transport, peddling, sale, purchase, and release for consumption (of wild meat in) inns or restaurants” (Ordonnance nº 60- 242 243 128). Government officials, district-level officials, army officers, judicial police, customs officers, service agents of livestock, and managers of food halls and markets, all have the right to, “search and recognize offenses” of illegally acquired wild animals (Ordonnance nº 60-128). These individuals have the authority to enter storage rooms, kitchens, offices, and other private or public areas relevant to an innkeeper’s or restaurant’s trade, as it relates to the trade of animals and they can: 1) seize gear and firearms used in the hunting offense; and 2) sequester motor vehicles (cars and boats) used to hunt or transport animals (Ordonnance nº 60-128). Punishment for being caught with illegal wild meat includes: 1) a fine of 10,000 to 200,000 Ariary (5 to 100 USD; for comparison, 81.3% of the population lives on < 1.25 USD per day, UNDP 2013); 2) imprisonment for one month to two years; and 3) loss of hunting license and/or authorization to hunt commercially (Ordonnance nº 60-128). Offenders will always have weapons, gear, and vehicles confiscated and will always be imprisoned if it is the second offense within five years (Ordonnance nº 60-128). Individuals who transport, peddle, sell, export, or store animals in violation of regulations are also liable to the same punishments as the individuals who hunted the animal; this includes restaurants that store the meat and serve it to customers (Ordonnance nº 60-128). Defendants cannot claim “ignorance of zoological matters” to justify violating the law (Ordonnance nº 60-128). Literature cited: Lindsey, P.A., G. Balme, M. Becker, C. Begg, C. Bento, C. Bocchino, A. Dickman, R. W. Diggle, H. Eves, P. Henschel, D. Lewis, K. Marnewick, J. Mattheus, J. Weldon McNutt, R. McRobb, N. Midlane, J. Milanzi, R. Morley, M. Murphree, V. Opyene, J. Phadima, G. Purchase, D. Rentsch, C. Roche, J. Shaw, H. van der 243 244 Westhuizen, N. Van Vliet, and P. Zisadza-Gandiwa. 2013. The bushmeat trade in African savannas: impacts, drivers, and possible solutions. Biological Conservation 160:80-96. Ordonnance nº 60-126: fixant le régime de la chasse, de la pêche et de la protection de la faune. October 3 1960. Republic of Madagascar. Available online via the Food and Agricultural Organization of the United Nations Legal Office: http://faolex.fao.org/docs/pdf/mad4214.pdf Accessed March 24 2015 Ordonnance nº 60-128: fixant la procédure applicable à la répression des infractions à la législation forestière, de la chasse, de la pêche et de la protection de la nature, modifiée par l'ordonnance nº 62-085. September 29 1962. Available online via the Food and Agricultural Organization of the United Nations Legal Office: http://faolex.fao.org/docs/pdf/mad2819.pdf Accessed March 24 2015 Rakotoarivelo, A. R., J. H. Razafimanahaka, S. Rabesihanaka, J. P. G. Jones, and R. K. B. Jenkins. 2011. Lois et règlements sur la faune sauvage à Madagascar: Progrès accomplis et besoins du future. Madagascar Conservation and Development 6:37- 44. United Nations Development Programme [UNDP]. 2013. Human development report 2013: The rise of the south: Human progress in a diverse world. United Nations Development Programme, New York, USA. 244 245 Table S8: Average distances (in kilometers ± 95% CI, towns are replicates) traveled by the consumer to procure wild meat (no data from Rats/Mice). Free Purchased Animal Hunted From From Roadkill/Raised Gift From restaurant From market Group animal hunter middleman** Bat Urban 135 ± 92 --- 126 ± 94 171 ± 334 0 (10 ± 10) 147 ± 117 35 ± 26 Rural N.D. --- 122 (n=1 town) --- 0 (5, n =1 town) 222 ± 100 --- Civet Urban 71 ± 9 0 ± 0 69 ± 102 --- 0 ------Rural 3 ± 3 --- 113 ± 221 --- 0 (0 ± 0) ------Fossa Urban 177 ± 303 --- 16 ± 33 ------354 (n=1 town) --- Rural 52 ± 75 --- 5 (n=1 town) --- 0 ------Lemurs Urban 105 ± 78 0 ± 0 137 ± 105 N.D. 0 (101± 110) 453 ± 545 0 (n=1 town) Rural 7 ± 5 --- 96 ± 188 --- 0 (7 ± 6) 165 (n=1 town) --- Mongoose Urban 92 ± 105 --- 5 (n=1 town) ------Rural 160 ± 77 --- 250 ± 489 --- 0 (0, n=1 town) ------Tenrec Urban 76 ± 31 0 ± 0 98 ± 57 0 ± 0 0 (22 ± 28) 477 ± 359 129 ± 173 Rural 5 ± 4 --- 2 ± 3 --- 0 (0 ± 0) 155 ± 108 100 km Wild Cat Urban 89 ± 38 0 ± 0 117 ± 199 N.D. N.D. N.D. N.D. Rural 0 ± 0 --- N.D. ------Wild Pig Urban 112 ± 159 --- 102 ± 90 --- 0 (166 ± 271) 119 ± 197 198 ± 333 Rural <1 ± 1 --- 0 (n=1 town) --- 0 (4 ± 4) ------**Middlemen typically travel to a consumer’s town of residence, and the travel distances of the middleman to the consumer – as reported by the consumers – are listed in parentheses when available. 245 246 Table S9: Sale of meat at open-air markets, restaurants, and supermarkets. Data are shown as the mean ± 95% CI (towns as replicates). Information taken from meatsellers interviews. Percent of Percent of Length of respondents respondents time since Amount sold per who hunted Price of Had ever Sold last 3 who bought Price of Sale Animal Group last sale day when selling meat Purchase sold (%) days (%) meat (distance (Ariary) (years type of meat (distance (Ariary) traveled for ago) traveled to purchase) hunt) Wild meat Bats Markets 1 ± 2 1 ± 2 2 ± 4 11 ± 14 animals 100 (0 ± 0) ND 2812/Animal 4666/Plate Restaurants 66 ± 36 54 ± 33 <1 ± <1 11 ± 5 plates 100 (14 ± 27) 0 4076/Animal 5791/Plate Supermarkets 0 ± 0 0 ± 0 ------Lemurs Markets 0 ± 0 0 ± 0 ------Restaurants 2 ± 4 0 ± 0 0 (n=1) ND ND ND 5000/Animal 2000/Plate Supermarkets 0 ± 0 0 ± 0 ------Tenrecs Markets <1 ± <1 0 ± 0 1 ± 0 19 ± 25 animals 100 (10 ± 20) 0 5000/Animal 6000/Animal Restaurants 29 ± 39 0 ± 0 <1 ± <1 7 ± 4 plates 100 (32 ± 42) 0 6800/Animal 8400/Plate Supermarkets 0 ± 0 0 ± 0 ------Wild Pigs Markets 3 ± 4 1 ± 2 2 ± 3 ND 75 (42 ± 40) 25 (20 ± 0) 42500/Animal 4125/kg 4000/kg Restaurants 18 ± 20 27 ± 27 1 ± 2 7 ± 6 plates 91 (0 ± 0) 8 (92, n = 1) 3500/kg 13500/kg 15357/Plate 33 (552, n Supermarkets 100 66% <1 ± 1 ND 67 (0 ± 0) ND 14716/kg =1) Domestic % Raised Meat animals Chicken 246 247 Markets ND 18 ± 10 0 ± 0 7 ± 5 animals 15±10 (54±42) 4 ± 4 6361/Animal 12250/Animal Restaurants 81 ± 26 0 ± 0 5 ± 1 plates 53±9 (0±0) 0 ± 0 10,792/Animal ND 8575/kg Pig Markets ND 28 ± 12 0 ± 0 42 ± 37 kilograms 24±11 (11±9) 4 ± 4 499,899/Animal <1 ± <1 animals 6575/kg 6466/kg Restaurants 18 ± 23 0 ± 0 4 ± 1 plates 18±23 (0±0) 0 ± 0 19,500/kg 10,000/Plate Zebu Markets ND 54 ± 13 0 ± 0 41 ± 14 kilograms 51±13 (42±22) 2 ± 4 621,363/Animal 140,000/Animal <1 ± <1 animals 6009/kg 8373/kg Restaurants 46 ± 32 0 ± 0 11 ± 10 plates 43±32 (0±0) 0 ± 0 8479/kg 9923/Plate 247 248 Table S10: Transport of meat on the in-country bus and boat transport system (61 respondents across 5 urban towns). Data are shown as the mean ± 95% CI. Towns were used as replicates when sample sizes were higher (towns as replicates = TR) and individuals were used as replicates when sample sizes were low (individuals as replicates = IR). When Percent of drivers How recently transporting, how How many do you Transport who had did you Cost of transporting Animal Group many are transport per distance transported before transport meat transported per year? (km) (%) (years ago) trip? Wild meat TR IR IR IR IR IR 4083 ± 2061 Bats 24 ± 27 <1 ± <1 32 ± 4 1542 ± 954 389 ± 303 Ariary/Bag* 1000 ± 1959 Lemurs 2 ± 4 <1 ± <1 ND ND 36 ± 41 Ariary/Animal Tenrecs 12 ± 11 1 ± 2 10 ± 8 118 ± 73 100 ± 196 Ariary/Trip 106 ± 79 0 (n = 1) Wild Cat <1 ± 1 1 (n = 1) ND ND ND Ariary/Animal 4 ± 4 Animals 141 ± 269 Animals 13928 ± 8544 Wild Pigs 17 ± 18 <1 ± <1 30 ± 5 kg 2495 ± 1865 kg Ariary/Animal 128 ± 85 Domestic Meat TR TR TR TR TR TR 444 ± 109 Ariary/Animal Chicken 62 ± 24 ND 27 ± 16 ND 223 ± 117 7650 ± 5433 Ariary/Basket** * One bag reportedly holds anywhere from 10-60 bats. ** One basket can hold anywhere from 10-60 chickens, depending on the size. 248 249 Table S11: Price paid by consumer for wild meat from different sellers (average, towns are replicates). All price estimates are from 2012 or 2013 unless otherwise noted. Body size estimate for wild cats was retrieved from Brockman et al. (2008). All other body size estimates were calculated from Garbutt (2007) as the mean of all species in the taxon, using the maximum weight recorded for each species. Purchase Price (Ariary) Animal Group All Sources Hunter Middleman Market Restaurant Bats (0.12 ± 030 lbs) 3042/animal 2500/animal 3044/animal 3550/animal 4000/animal 2908/plate ------2761/plate Urban ------1798/animal 1950/animal 3136/animal --- Rural 1750/plate ------2625/plate --- 2500/serving 2500/serving ------Civet (7.64 ± 3.04 lbs) 2150/animal --- Urban ------4000/serving 4000/serving Rural ------Fossa (price estimate from 2010) (22 lbs) Urban 1000/plate ------1000/plate Rural ------Lemurs (3.64 ± 4.18 lbs) 3000/animal 4625/animal --- 3000/plate --- 3000/plate Urban ------2125/serving 2125/serving --- 3000/kg 3000/kg --- 4667/animal 1000/animal Rural ------249 250 1500/serving 1500/serving ------Mongoose (undated price estimate) (2.04 ± 0.71 lbs) Urban ------Rural 3000/animal ------Tenrecs (0.26 ± 0.80 lbs) 3298/animal 2000/animal 3358/animal 4300/animal 3000/animal 2000/plate --- 2000/plate ------Urban ------6000/kg ------3000/animal ------Rural ------Wild Cat (12 lbs) Urban 2000/animal ------Rural ------Wild Pig (155 lbs) ------7444/plate ------7444/plate Urban --- 3081/serving 2279/serving 2417/serving --- 4251/kg 3466/kg 4313/kg --- 30,000/animal 30,000/animal ------Rural ------2375/serving 2816/serving 4000/kg 4000/kg 250 251 Table S12: Percent of people who procured domestic meat from different sources in urban and rural regions. Averages ± 95% CI are shown with towns as replicates. Free (%) Purchased (%) Consumer Consumer From Consumer raised Animal received From all From From From market/ hunted animal/ animal as sources farmer middleman restaurant Group animal roadkill gift supermarket Chicken Urban NA 32 ± 9 3 ± 4 69 ± 10 0 ± 0 5 ± 5 2 ± 2 36 ± 16 Rural NA 60 ± 15 3 ± 3 45 ± 13 0 ± 0 16 ± 12 2 ± 3 1 ± 2 Pig Urban NA <1 ± <1 0 ± 0 99 ± 1 9 ± 17 3 ± 6 1 ± 1 58 ± 25 Rural NA 39 ± 30 22 ± 20 44 ± 31 0 ± 0 13 ± 17 0 ± 0 0 ± 0 Zebu Urban NA 0 ± 0 4 ± 3 96 ± 3 0 ± 0 <1 ± 1 <1 ± <1 62 ± 24 Rural NA 2 ± 3 6 ± 3 94 ± 3 0 ± 0 14 ± 8 2 ± 3 5 ± 5 251 252 Table S13: Percent of people who procured wild meat from different sources in urban and rural regions. Averages ± 95% CI are shown and towns are replicates. Free (%) Purchased (%) Consumer Consumer Consumer Animal raised received From all From From From From hunted animal/ animal as sources hunter middleman restaurant market Group animal roadkill gift Bat Urban 28 ± 8 0 ± 0 21 ± 9 56 ± 13 2 ± 2 17 ± 13 10 ± 6 10 ± 9 Rural 46 ± 25 0 ± 0 11 ± 11 28 ± 16 <1 ± <1 8 ± 7 4 ± 6 0 ± 0 Civet Urban 67 ± 9 13 ± 6 18 ± 7 3 ± 2 0 ± 0 <1 ± <1 0 ± 0 0 ± 0 Rural 59 ± 25 0 ± 0 18 ± 15 6 ± 5 0 ± 0 6 ± 8 0 ± 0 0 ± 0 Fossa Urban 83 ± 18 0 ± 0 10 ± 14 5 ± 10 0 ± 0 0 ± 0 5 ± 10 0 ± 0 Rural 50 ± 31 0 ± 0 26 ± 23 7 ± 13 0 ± 0 8 ± 14 0 ± 0 0 ± 0 Lemurs Urban 43 ± 13 <1 ± <1 37 ± 10 15 ± 9 0 ± 0 9 ± 9 2 ± 2 <1 ± <1 Rural 44 ± 22 0 ± 0 14 ± 12 31 ± 17 0 ± 0 19 ± 13 1 ± 2 0 ± 0 Mongoose Urban 100 ± 0 0 ± 0 31 ± 23 0 ± 0 0 ± 0 0 ± 0 0 ± 0 0 ± 0 Rural 86 ± 28 0 ± 0 26 ± 30 14 ± 28 0 ± 0 17 ± 28 0 ± 0 0 ± 0 Rats/Mice Urban 33% (n=1) 0 ± 0 33% (n=1) 0 ± 0 0 ± 0 0 ± 0 0 ± 0 0 ± 0 Rural 75 ± 28 0 ± 0 25 ± 28 0 ± 0 0 ± 0 0 ± 0 0 ± 0 0 ± 0 Tenrec 252 253 Urban 71 ± 6 <1 ± <1 23 ± 4 17 ± 6 <1 ± <1 9 ± 6 2 ± 2 1 ± 1 Rural 62 ± 18 0 ± 0 12 ± 9 12 ± 13 0 ± 0 7 ± 11 1 ± 2 <1 ± <1 Wild Cat Urban 74 ± 11 3 ± 3 11 ± 8 1 ± 2 N.D. N.D. N.D. N.D. Rural 77 ± 28 0 ± 0 20 ± 30 0 ± 0 0 ± 0 0 ± 0 0 ± 0 0 ± 0 Wild Pig Urban 32 ± 20 0 ± 0 15 ± 8 62 ± 17 0 ± 0 29 ± 19 4 ± 3 11 ± 9 Rural 87 ± 16 0 ± 0 4 ± 5 23 ± 30 0 ± 0 19 ± 23 0 ± 0 0 ± 0 253 254 Table S14: The volume/rate of wild meat consumption, for individuals who had consumed wild meat at least once in a lifetime and the volume per year. Significant differences between towns are noted. Columns note whether towns or individual respondents are treated as replicates and depict the mean ± 95% Confidence Interval. Where individuals were used as replicates, sample sizes are listed behind each mean in parentheses. ND: No Data. Average number of times Average number of times Average number of times consumed per year (2009- consumed in a lifetime consumed per year (pre-2009) Animal 2013) (towns as replicates) (individuals as replicates) (individuals as replicates) Urban Rural Urban Rural Urban Rural Bat 53 ± 48* 47 ± 47** 38 ± 24 (51) 3 ± 4 (5) 14 ± 4 (162) 7 ± 4 (32) Civet 12 ± 8* 11 ± 6 19 ± 24 (12) 1 ± <1 (6) 9 ± 6 (19) 3 ± 4 (16) Fossa 2 ± <1 4 ± 3 2 ± 2 (6) <1 (1) 1 ± 1 (5) 2 ± 2 (8) Lemur 13 ± 7* 30 ± 13 15 ± 10 (35) 4 ± 2 (9) 14 ± 16 (40) 5 ± 2 (32) Mongoose 19 ± 18 9 ± 12* 4 ± 3 (5) 2 (1) ND 1 ± 2 (2) Rats and mice 3 ± 4 3 ± 0 ND Tenrec 49 ± 29* 92 ± 77* 17 ± 8 (51) 2 ± <1 (3) 13 ± 4 (185) 9 ± 3 (90) Wild Cat 6 ± 4** 17 ± 17 2 ± 2 (6) ND 1 ± <1 (8) 1 ± <1 (9) Wild Pig 53 ± 24* 135 ± 102 14 ± 8 (21) ND 18 ± 6 (111) 8 ± 4 (29) All Wild 86 ± 56* 117 ± 122* 6 ± 4* <1 ± <1* 10 ± 6* 5 ± 5* 254 255 Meat (towns as (towns as (towns as (towns as replicates) replicates) replicates) replicates) Zebu (cattle) ND 88 ± 20* 67 ± 30* Chicken ND 30 ± 9* 27 ± 12* Pig ND 54 ± 17 18 ± 8** All Domestic ND 51 ± 11* 41 ± 17* Meat * Significant differences between towns, Analysis of Variance, p < 0.05 ** Marginal difference between towns, Analysis of Variance, 0.05 < p < 0.1 255 256 Figure S2: Sources of domestic meat in urban (left column) and rural regions (right column). The arrows point in the direction of the consumer and the thickness of the arrow is proportional to the percentage of consumers (towns are replicates) who procured domestic meat from that source. When no arrow is present, no respondents reported procuring meat from that source. There are two tiers of arrows; the arrows that connect peripheral boxes to the “purchase” and “free” boxes represent direct quantities measured (black arrows) while the arrows connecting the “purchase” and “free” boxes to the consumer represent sums of the peripheral boxes of each type (gray arrows). Urban Rural Chicken Pig 256 257 Zebu (Cattle) 257 258 Figure S3: Sources of commonly consumed wild mammals in urban (left column) and rural regions (right column). The arrows point in the direction of the consumer and the thickness of the arrow is proportional to the percentage of consumers (towns are replicates) who procured wild meat from that source. When no arrow is present, no respondents reported procuring wild meat from that source. There are two tiers of arrows; the arrows that connect peripheral boxes to the “purchase” and “free” boxes represent direct quantities measured (black arrows) while the arrows connecting the “purchase” and “free” boxes to the consumer represent sums of the peripheral boxes of each type (gray arrows). Urban Rural Lemurs Tenrecs 258 259 Bats Wild Pig 259 260 Figure S4: Sources of meat for wild mammals that are not commonly consumed or are typically killed through human-wildlife conflict in urban (left column) and rural regions (right column). The arrows point in the direction of the consumer and the thickness of the arrow is proportional to the percentage of consumers (towns are replicates) who procured domestic meat from that source. When no arrow is present, no respondents reported procuring meat from that source. There are two tiers of arrows; the arrows that connect peripheral boxes to the “purchase” and “free” boxes represent direct quantities measured (black arrows) while the arrows connecting the “purchase” and “free” boxes to the consumer represent sums of the peripheral boxes of each type (gray arrows). Urban Rural Wild Cat Fossa 260 261 Mongoose Civets Rats & Mice 261 APPENDIX F CHAPTER 6 SUPPLEMENTARY MATERIALS Table S15: Table of stable isotope values by mammal species, with means ± SD (towns are replicates for bats; individuals are replicates for other species). Bat Species (number of δ13C (‰) δ15N (‰) towns) E. dupreanum (n = 4) Average -21.20 ± 0.92 7.20 ± 1.02 Minimum Value -21.66 ± 1.24 5.07 ± 2.45 Maximum Value -20.58 ± 0.52 8.22 ± 1.69 P. rufus (n = 5) Average -22.29 ± 0.78 6.59 ± 0.26 Minimum Value -23.49 ± 1.36 3.74 ± 1.84 Maximum Value -20.20 ± 2.14 9.45 ± 2.34 R. madagascariensis (n = 2) Average -21.93 ± 0.85 8.78 ± 1.38 Minimum Value -22.08 ± 0.84 8.33 ± 2.02 Maximum Value -21.64 ± 1.27 9.27 ± 0.69 Other Mammal Species δ13C (‰) δ15N (‰) Felis silvestris (n=1) -20.39 11.52 Wild Cat Potamochoerus larvatus (n=3) -22.77 ± 0.61 8.81 ± 0.62 Wild Pig/Bushpig Setifer setosus (n=1) -21.81 10.70 Greater Tenrec Tenrec ecaudatus (n=2) -19.12 ± 2.73 7.14 ± 1.91 Common Tenrec Viverricula indica (n=4) -18.96 ± 0.95 9.09 ± 0.78 Small Indian Civet 262 Table S16: Climate and forest cover characteristics of each hunting site. WDF = Western Dry Forest/Deciduous Seasonally Dry Forest. W = Wetlands/Marshlands. HF = Humid Forest/Evergreen Forests. M = Mangroves. Town (hunting sites Annual Temperature Percent forest cover (radius)b Percent of native remnant listed under each town) Range (Mean)a primary vegetation by forest type (30 km radius) in 2003e 5 km 15km 30km WDF W HF M Andriba Mangasoavina 190 13% 18% 27% 59% 0% 41% 0% Antsiafabositrac 183 11% 11% 4% 90% 0% <1% 0% Antsohihy Ambaliha 146 25% 24% 11% 68% 4% 6% 23% Amboroho 172 6% 2% 5% 69% 3% 5% 24% Ambilobe Ambakiarano 148 19% 1% 5% 96% 2% 2% <1% Amborondolo 155 53% 56% 24% 74% 5% 1% 19% Beramanja 149 33% 30% 35% 55% 10% <1% 35% Isesy 146 6% 14% 5% 67% 12% <1% 21% Mahivoragno 135 2% 3% 5% 75% 10% 1% 14% Mamoro 144 23% 25% 40% 66% 5% 10% 20% Anivorano Nordd 132 48% 7% 11% 70% <1% 29% 0% Antsiranana Akonokono 119 19% 6% 6% 73% 10% 10% 6% Andranofanjava 125 53% 5% 11% 48% 5% 44% 2% Daraina 122 51% 54% 27% 55% 4% 34% 7% Mangoaka 113 5% 13% 7% 60% 10% 26% 4% aData derived from WorldClim database (Hijmans et al. 2005). bData extracted from satellite images downloaded from the USGS Earth Explorer database (US Dept. of Interior & US Geological Survey 2014). cIt was not possible to determine where the bats were hunted so the center of the town was used as the ‘hunting site’. dBats were hunted only at one hunting site, but the name of this hunting site is unknown. eData derived from Moat and Smith (2007). 263 264 Table S17: Table of stable isotope values for the leaves of Mangifera indica (Common name: Mango), with means ± SD (individuals are replicates) at sites spread across a distance of 440 km. Samples were collected within 30 km of the towns; leaves were collected directly from trees and were dried prior to storage in sealed plastic containers. Samples (100 mg per tree) were analyzed using the same procedures as for the hair analysis. Values did not differ by town for δ13C (ANOVA, F-ratio = 2.0780, DF = 3, P = 0.1735) or for δ15N (ANOVA, F-ratio = 0.0785, DF = 3, P = 0.9701). Town δ13C (‰) δ15N (‰) Antsiranana (n=3) -29.643 ± 0.72 3.92 ± 1.12 Anivorano Nord (n=3) -30.107 ± 0.71 3.63 ± 1.53 Ambilobe (n=2) -28.611 ± 0.09 4.47 ± 1.66 Antsohihy (n=5) -30.101 ± 0.88 3.78 ± 2.56 264