Connectivity at the Large Carnivore Scale: the Kafue-Zambezi Interface

1 Connectivity at the Large Carnivore Scale:

2 The Kafue-Zambezi Interface

4 Robin Michael Lines

6 Durrell Institute of Conservation and Ecology

7 School of Anthropology and Conservation

8 University of Kent, UK

11 April 2020

12 Thesis submitted for the degree of

13 Doctor of Philosophy in Biodiversity Management

14 Acknowledgements

15 This thesis is long in the making. Indeed, many eyebrows will doubtless be raised when

16 word of its completion percolates through various networks.

17 The body of work can be traced through earlier studies of African wild dogs in Namibia

18 pre-2010. In a sense the support received during that study set the scene for ensuing

19 research in Zambia and subsequently to this PhD.

20 For project development, permissions and fieldwork support I owe gratitude to Prof.

21 Andrew Nambota, HRH Snr Chief Inyambo Yeta, (the late) HRH Chief Moomba, their

22 Indunas and Headmen and Headwomen who facilitated gracious and smooth passage

23 through their respective Chiefdoms, and whose subjects could not have been more

24 welcoming. Indeed, the cultural vibrancy experienced throughout this study reflects a

25 defining characteristic of the Kafue-Zambezi landscape.

26 Department of Wildlife and National Parks, formerly Zambian Wildlife Authority,

27 facilitated requisite permits and visas. Bwana Moses Mulimo, Charles Chibila and staff,

28 Department of Wildlife and National Parks, provided fieldwork support, and on the whole

29 dealt largely diplomatically with political and security issues that occasionally cropped up.

30 Tracking has long been a passion of mine and was incorporated as a key component of this

31 study. Gavin Kambole, Earnest Kashaye, Patrick Matape, Daniel Mundia and Martin

32 Samuntafu provided ample spoor tracking skills and insightful interpretation of field

33 conditions. Richard Kraljik kindly made his camps available for occasional accommodation

34 during fly-camping.

35 The late Peter Moss was an endless fountain of knowledge, support and interest in all

36 matters Kafue and Zambia. John Hanks and Ian Swingland provided much motivation and

37 context for the study as former ecologists in Kafue National Park. Mike Musgraves’ holistic

38 knowledge and insights of the study area remain invaluable. George Cloughborough kept

39 spirits buoyed from his camp near Sesheke, offering welcome hospitality and entertainment

40 during resupply and vehicle maintenance trips, providing realpolitik guidance and

41 indulging my passion of fly-fishing for Tigers around the Zambezi’s dramatic cataracts and

42 waterfalls.

43 Joseph Tzanopolous demonstrated rare patience, insight and support as main supervisor,

44 with Douglas MacMillan providing useful input when bottlenecks loomed. Shelley Malekia

45 and Nicola Kerry-Yoxall were always on hand to help with a myriad of University

46 questions, going back to the halcyon days of my Masters. Lucy Pieterse, Dimitris

47 Bormpoudakis and Pantelis Xofis brought technical skills to manuscripts.

48 Financial support was provided by the University of Kent 50th Anniversary Bursary, WWF

49 Namibia, Humane Society International (Australia) and Westwood.

50 Finally, I wish to thank Ana Maria Puerta Ocampo and our extended families for their

51 unflinching positive outlook, patience and sustenance throughout this process.

iii

56 Author’s Declaration

57 All chapters were written in full by R. Lines, with comments and editorial suggestions

58 provided by supervisors J. Tzanopolous (all chapters) and D. MacMillan (chapter 4).

59 Chapters 2 and 3 include collaborations with researchers external to University of Kent

60 who have provided additional comments and editorial suggestions for ongoing submission

61 to peer review Journals. All research was approved by the School of Anthropology and

62 Conservation Ethics Advisory Group, University of Kent a Canterbury, with supplementary

63 research permits and permissions provided by Department of Wildlife and National Parks,

64 Zambia, and local Traditional Authorities.

65 Chapter 2: R. Lines conceived the idea in collaboration with J. Tzanopolous. R. Lines

66 undertook all fieldwork and data collection with support of Department of Wildlife and

67 National Parks staff and additional local trackers. R. Lines conducted all data analyses and

68 wrote the manuscript with feedback from co-authors.

69 Chapter 3: R. Lines conceived the idea in collaboration with J. Tzanopolous and P. Xofis.

70 R. Lines undertook all fieldwork and data collection with support of Department of

71 Wildlife and National Parks staff and additional local staff. R. Lines conducted data

72 analyses with co-authors and wrote the manuscript with feedback from co-authors.

73 Chapter 4: R. Lines conceived the idea in collaboration with D. MacMillan and J.

74 Tzanopolous. R. Lines undertook all fieldwork and data collection with support of

75 Department of Wildlife and National Parks staff and additional local staff. R. Lines

76 conducted data analyses with co-authors and wrote the manuscript with feedback from co-

77 authors.

78 Acronyms

79 AUC Area Under Curve

80 DNPW Department National Park and Wildlife

81 FR Forest Reserve

82 GIS Geographic Information System

83 GKS Greater Kafue System

84 GMA Game Management Area

85 HFP Human footprint pressure

86 IUCN International Union for Conservation of Nature

87 KAZA Kavango-Zambezi

88 KNP Kafue National Park

89 MaxEnt Maximum Entropy

90 NGO Non-Government Organisation

91 NP National Park

92 TFCA Transfrontier Conservation Area

93 WDC Wildlife Dispersal Corridor

94 WMA Wildlife Managed Area

95 ZAWA Zambia Wildlife Authority

100 Abstract

101 The growth and expansion of human populations and resource demands is driving large

102 scale fragmentation and loss of wildlife habitat, isolating wildlife populations and pushing

103 many species towards extinction at local to global scales.

104 Attempts to promote connectivity between wildlife managed areas at transboundary scales

105 has been proposed as a solution to negative effects associated with population isolation.

106 Such approaches commonly require the maintenance of wildlife populations throughout

107 human-dominated landscapes subject to various degrees of effective protection.

108 The aims of this study are to (1) assess the status of the large carnivore guild throughout ten

109 wildlife managed areas comprising the Zambian component of Kavango-Zambezi

110 Transfrontier Conservation Area between Kafue National Park and the Simalaha Wildlife

111 Recovery Sanctuary on the Zambezi River; (2) model habitat suitability and connectivity in

112 this landscape for Lion, Leopard and Spotted hyena; and (3) develop a site-specific map of

113 human footprint pressure for the landscape and test if it can be used a proxy for determining

114 the occurrence of these three species. And further, explore if there are thresholds in human

115 footprint pressure beyond which species are likely extirpated from wildlife managed areas.

116 Methods included library studies to determine historical status of the large carnivore guild

117 and twenty-six common prey species, spoor tracking in conjunction with qualitative

118 surveys and supplemental data analysis to ascertain species current distribution, remote

119 sensing with ground-truthing to build landcover maps, Maximum Entropy and Current

120 Flow models, and extensive use of Geographic Information Systems.

121 The findings conclude that there have been large scale losses in species assemblages

122 throughout majority of southern wildlife managed areas, including the Simalaha Wildlife

123 Recovery Sanctuary. However, no detectable changes were evident in Kafue National Park

124 and surrounding Game Management Areas. Human activities are limiting habitat suitability

125 and scope for occurrence in central southern areas of the landscape, with the likelihood of a

126 connectivity bottleneck occurring. There is significant overlap in habitat requirements and

127 scope for species movement. Human footprint pressure models appear to demonstrate

128 utility as a proxy measure for occurrence of our large carnivore subset, though require some

129 refinements and supplemental data layers to increase predictive power. Human footprint

130 pressure at the wildlife managed area scale indicates threshold levels at which target

131 species occur or are locally extirpated.

132 Analyses have identified important additions to the existing wildlife managed area network

133 in Open communal land that could provide valuable habitat and connectivity for target

134 species given effective management and finance, including containment of negative human

135 disturbance variables modelled (agro-pastoralist activities and infrastructure development).

136 The effects of poaching are also hypothesized to be a significant driver limiting species

137 persistence.

138 Continued expansion of human population, settlement and agro-pastoralist activities will

139 limit scope for expansion of large carnivores and their principle prey throughout the Kafue-

140 Zambezi interface, effectively severing connectivity and isolating the Greater Kafue

141 System from adjacent wildlife managed areas in the Kavango-Zambezi Transfrontier

142 Conservation Area.

vii

143 Narratives surrounding the development of wildlife-based land uses and species-level

144 connectivity benefit from the application of conservation science and generation of

145 empirical data to guide management.

146

147

148

149

150

151

152

153

154

155

156

157

158

159

160

161

viii

162 Table of Contents

163 Acknowledgements ii

164 Authors Declaration iv

165 Acronyms v

166 Abstract vi-viii

167 Chapter 1: Introduction 1

168 1.1 Protected Areas and Protected Area Networks 1

169 1.2 The Kavango-Zambezi Transfrontier Conservation Area

170 and the Greater Kafue System 3

171 1.3 Study Area: The Kafue-Zambezi Interface 6

172 1.4 Connectivity and Large Carnivores in the KAZA TFCA 14

173 1.5 Focal Species: Large Carnivores 16

174 1.5.1 African Lion (Panthera leo) 17

175 1.5.2 Leopard (Panthera pardus) 19

176 1.5.3 Spotted Hyena (Crocuta crocuta) 21

177 1.6 Thesis Structure 23

178 1.7 References 25

179 Chapter 2: Status of Terrestrial Mammals at the Kafue-Zambezi Interface:

180 Implications for Transboundary Connectivity 39

181 Abstract 39

182 2.1 Introduction 41

183 2.2 Methods 44

184 2.2.1 Study Area 44

185 2.2.2 Data Sources 49

186 2.2.3 Data Analyses 53

187 2.3 Results 54

188 2.3.1 Changes to Species Occurrence and Distribution 54

189 2.4 Discussion 58

190 2.5 Conclusions 61

191 2.6 References 62

192 Chapter 3: Modelling multi-species connectivity at the Kafue-Zambezi Interface:

193 Implications for Transboundary Carnivore Conservation 69

194 Abstract 69

195 3.1. Introduction 71

196 3.2. Methods 73

197 3.2.1 Study Area 73

198 3.2.2 Landcover Mapping 76

199 3.2.3 Habitat Suitability Modelling 78

200 3.2.4 Connectivity Modelling 81

201 3.3 Results 82

202 3.3.1 Landcover Mapping 82

203 3.3.2 MaxEnt Habitat Suitability Models 83

204 3.3.3 Connectivity Modelling using Linkage Mapper 87

205 3.4 Discussion 91

206 3.5 Conclusion 95

207 3.6 References 96

208 Chapter 4: Utility of Human footprint pressure mapping for large carnivore

209 conservation: The Kafue-Zambezi Interface 104

210 Abstract 104

211 4.1 Introduction 106

212 4.2 Methods 109

213 4.2.1 Study Area 109

214 4.2.2 Generating human footprint pressure maps 105

215 4.2.3 MaxEnt Habitat Suitability Modelling 116

216 4.3 Results 118

217 4.3.1 Human footprint pressure Mapping 118

218 4.3.2 MaxEnt Habitat Suitability Modelling Outputs 120

219 4.3.3 Human footprint pressure thresholds and species persistence 123

220 4.4 Discussion 124

221 4.5 Conclusions 127

222 4.6 References 129

223 Final Discussion 136

224 5.1 Recapitulation of Study Scope & Key Findings 136

225 5.2 Contributions to conservation science 137

226 5.2.1 Chapter 2: Species Assemblages and Distribution 137

227 5.2.2 Chapter 3: Landcover Mapping, Habitat Suitability and Connectivity

228 Modelling 138

229 5.2.3 Chapter 4: Human footprint pressure 140

230 5.3 Methodological Insights / Limitations 141

231 5.4 Broader Contribution of the Study to Literature 143

232 5.4.1 Transboundary conservation 143

233 5.4.2 Protected areas 143

234 5.4.3 Understudied landscapes 144

235 5.4.4 Local issues 144

236 References 145

237 Appendices: Supplemental methodological explanations, analysis & data 149

238 1 Spoor Tracking Surveys 149

239 2 Landcover Mapping 152

240 3 Key Interventions & Outcomes in Promotion of Landscape Connectivity 155

241 3.1 Diagnosing challenges to biodiversity conservation 155

242 List of Figures

243 1.1 The Kavango-Zambezi Transfrontier Conservation Area landscape 5

244 1.2 Protected Areas, Kafue-Zambezi Interface 9

245 1.3. KAZA TFCA Wildlife Dispersal Corridors 15

246 1.4 African Lion (Panthera leo) 18

247 1.5 Leopard (Panthera pardus) 20

248 1.6 Spotted Hyena (Crocuta crocuta) 22

249 2.1 The Kavango-Zambezi Transfrontier Conservation Area landscape 45

250 2.2 Wildlife Managed Areas within study area 47

251 2.3 Historical Species Distribution from Ansell (1978) 51

252 2.4 Contemporary Data Sources 53

253 2.5 Changes to carnivore and herbivore composition by area, 1978-2014/15 56

254 2.6 Distribution of large carnivores at Kafue-Zambezi Interface, 2014/5 57

xii

255 3.1 The Kavango-Zambezi Transfrontier Conservation Area landscape 74

256 3.2 Wildlife managed areas within study area, indicating proposed

257 Simalaha-Kafue Wildlife Dispersal Corridor 76

258 3.3 Landcover map of the wider study area 83

259 3.4 Habitat Suitability Map for Spotted Hyena 84

260 3.5 Habitat Suitability Map for Spotted Lion 85

261 3.6 Habitat Suitability Map for Spotted Leopard 86

262 3.7 Connectivity modelling outputs by species 89

263 3.8 Composite species connectivity model 90

264 4.1 The Kavango-Zambezi Transfrontier Conservation Area landscape 110

265 4.2 Dissagregated Human Footprint layers / wildlife managed areas 119

266 4.3 Human footprint pressure, Kafue-Zambezi Interface 120

267 4.4 Human footprint pressure-based Habitat Suitability Models 122

268

269

270

271

272

273

274

275

276

277

278

xiii

279 List of tables

280 2.1 Summary results of species detection by source and area,

281 with distribution change, 1978-2014/5 55

282 3.1 Predictor variables used in the model building process 87

283 3.2 Variable importance for the Maxent models (AUC, single variable) 81

284 4.1 Modifications to Human Pressure Variables and Pressure Scoring 116

285 4.2 Mean Human footprint pressures/Wildlife Managed Areas

286 /Species Occurrence 124

287 5 Summary of exploratory analyses to determine sampling strategy 151

288 6 Sentinel 2 Spectral Bands and central wavelengths 153

289

290 7 Accuracy assessments by landcover classification 154

xiv

291 Chapter 1: Introduction

292 1.1 Protected Areas and Protected Area Networks

293 Protected Areas encompassing IUCN Ia-VI categories remain for many the bedrock of

294 biodiversity conservation. Management objectives are typically spilt between 1.

295 Maintaining natural and cultural resources for local communities relying on natural capital

296 within and surrounding Protected Areas, and 2. For the broader global community focused

297 on wider cultural, economic, leisure, aesthetic and environmental well-being values

298 (UNEP-WCMC, 2015; Barrett et al., 2018). However Protected Areas are increasingly

299 imperiled and failing to fulfil their broad mandates at multiple scales (Brandon et al., 1998;

300 Leverington et al., 2010).

301 Evolving knowledge surrounding limitations of the existing Protected Area networks to

302 fulfil broad ecological relevance, political resilience and social acceptance mandates has

303 resulted in novel approaches to increase the effective spatial scale of Protected Areas,

304 incorporating new and existing areas into larger Protected Area networks at both National

305 and Transboundary scales (Suich et al., 2012). Transfrontier Conservation Areas commonly

306 seek to integrate wider ecosystem-scale landscapes incorporating complex and dynamic

307 coupled socioecological management units subject to rising anthropogenic disturbance

308 (Andersson et al., 2017)

309 A major driver of biodiversity conservation is the establishment of many larger Protected

310 Area networks is the promotion of functional ecological linkages between Protected Areas

311 and the wildlife populations residing within them (Hanks, 2000; Cumming, 2008). This is

312 especially relevant for wide-ranging low-density species of conservation concern such as

313 the large carnivore guild requiring vast areas to secure viable populations (Woodroffe &

314 Ginsberg, 1998; Crooks et al., 2011).

315 The ecological effectiveness of linking and managing large-scale Protected Area networks

316 largely concerns benefits surrounding the connection of wildlife populations and reduction

317 of risks associated with isolation of core wildlife managed areas (Margules & Pressey,

318 2000; Newmark, 2008).

319 The effectiveness of expanding and linking Protected Areas into larger networks are multi-

320 faceted and tempered by significant challenges. These include chronic underfunding for

321 biodiversity conservation and management (Pringle, 2017), expanding human settlement

322 and intensifying agropastoralist activities within and surrounding Protected Areas

323 (Wittemyer et al., 2008; Jones et al., 2018), the widespread over-exploitation (legal and

324 illegal) of wildlife and natural resources (Maxwell et al., 2016), negative effects of

325 veterinary fencing on wildlife movement (Ferguson & Hanks, 2012) and political struggle

326 over control of key resources (Duffy, 2006).

327 In central Southern Africa there has been a proliferation of Transfrontier Conservation

328 Areas (TFCAs) and ‘Peace Parks’ since the turn of the Millennium. While the ‘Peace Park’

329 inferences to ameliorate prospects for renewed regional conflict has perhaps reasonably

330 been portrayed as ‘palliative rhetoric’ (Murphree, 2017), broader goals are likely worthy

331 and justified. These include an antidote to the multifaceted problems of Colonial-era

332 boundaries bisected ecosystems, wildlife populations and communities, and addressing

333 socio-economic marginalization of people living on the boundaries of Protected Areas

334 (Andersson et al., 2017).

335 1.2 The Kavango-Zambezi Transfrontier Conservation Area and Greater Kafue System

336 The Kavango-Zambezi Transfrontier Conservation Area (KAZA TFCA) spans >500,000

337 km2 at the interface of Angola, Botswana, Namibia, Zambia and Zimbabwe (KAZA,

338 2011a). The landscape broadly encompasses the basins of the region’s two largest river

339 systems, the Kavango and Zambezi, giving the KAZA TFCA its name.

340 Concurrent with many of the subcontinent’s other TFCAs, KAZA’s stated objectives focus

341 on integrating conservation and development, promoting cooperation and facilitating

342 connectivity of ecosystems and wildlife populations (KAZA, 2011a). And again, the

343 challenges facing this complex, dynamic and coupled socio-ecological system are manifest,

344 and broadly similar to other regional TFCAs - mounting anthropogenic pressure, poor land

345 use planning, institutional conflicts and stakeholder disenfranchisement (Andersson et al.,

346 2017) are driving human encroachment and disturbance around and into former wildlife

347 areas, resulting in habitat loss and fragmentation (Watson et al., 2015; Newmark, 2008;

348 Simukonda, 2008). Unsustainable harvesting of wildlife threatens many of the Kavango-

349 Zambezi TFCA’s iconic natural assets critical to the development of wildlife-based land

350 uses options and natural economies (Funston et al., 2013).

351 With the region’s human population expected to double by 2050 (UN, 2019), and likely

352 impacts of climate change exacerbating socioeconomic development challenges (Pachauri

353 et al., 2014; Bellard et al., 2012), even moderately optimistic scenarios imply regional

354 biodiversity loss will accelerate significantly this century without drastic increases in

355 funding and political support (Biggs et al., 2008).

356 Collectively these challenges raise important questions surrounding the scope, scale and

357 ambition of narratives promoting landscape-level linkages, the interventions required to

358 maintain or expand connectivity, and what purposes these proposed linkages may serve in

359 the long term (Cumming, 2008). A clear imperative thus exists to promote evidence-based

360 socioeconomic and environmental policies and interventions built around the application of

361 conservation science (Sutherland et al., 2004), including research and monitoring of

362 changes to site and system states and their response to factors driving connectivity at the

363 scale of interest. But the process of informed decision-making is data hungry. Local,

364 regional and transboundary data sources are disparate and inconsistent, undermining

365 attempts to understand complex socioecological systems such as the Kavango-Zambezi

366 TFCA (Cumming, 2008). Data deficiencies ultimately constrain effective decision making

367 and appropriate interventions to promote biodiversity conservation and development.

368 The Kavango-Zambezi TFCA’s boundaries are imprecise. However, Cumming (2008)

369 characterizes the TFCA as comprising a matrix of >70 wildlife managed areas in eleven

370 IUCN Categories covering strict National Parks under State control with no (legal) human

371 settlement, through various designated hunting area categories (some unsettled and others

372 under Communal tenure and subject to extensive subsistence agriculture), through to

373 multiple use Communal Conservancies under Customary management of local Traditional

374 Authorities in their respective Chiefdoms. In total some 76% of the KAZA TFCA is

375 categorized as wildlife managed areas, 22% of which falls within Protected Areas with no

376 human settlement (fully protected), 54% falls within settled hunting areas and Community

377 Conservancies (partially protected), and the remaining 34% is covered by Communal areas

378 (nominally protected), including small portions of urban and extra-urban development.

379 Collectively these wildlife managed areas fall into three main clusters centered on the

380 region’s major National Parks – Kafue, Chobe and Hwange (Figure 1.1), and five

381 peripheral sub-clusters, with Kafue National Park and surrounding wildlife managed areas

382 constituting the major northern cluster and focus of my study.

383 384 Figure 1.1. The Kavango-Zambezi Transfrontier Conservation Area landscape, indicating 385 study area, clusters of wildlife managed areas (WMAs) and their degrees of protection. 386 Protected = National Parks IUCN II; Partially Protected = IUCN III-VI; Nominally 387 Protected = IUCN Not Reported (adapted from PPF, 2011a).

388

389

390

391

392 While at a Regional scale Southern Africa scores fairly well against a global index of

393 megafauna conservation, conservation effectiveness throughout major wildlife managed

394 areas across the KAZA TFCA is broadly considered weak and underfunded, with data

395 showing five major Protected Areas receiving between 9-39% of estimated budgets

396 required for their effective operation (Lindsey et al., 2017). Given limits to information

397 systems and data access, even for key species of conservation concern, combined with

398 heavily constrained State funding for core National Parks, it is clear budgets and resources

399 for broader wildlife managed areas under Communal and Customary tenure are woefully

400 inadequate (Cumming, 2008).

401 1.3 Study Area - The Kafue-Zambezi Interface

402 Historical drivers impacting species distribution and connectivity:

403 In developing a holistic understanding of landscape connectivity in the Kafue-Zambezi

404 interface my research sought to identify significant historical events that have shaped the

405 contemporary landscape. This provides a foundation for interpreting current drivers

406 impacting resource demands, land use and development pathways that will shape longer

407 term prospects for large carnivore conservation and connectivity throughout this study area

408 and the wider KAZA TFCA.

409 The mosaic of land uses and ownership at the Kafue-Zambezi interface represent a

410 complex, dynamic, coupled human-nature system with a written history covering a mere

411 160 years, though oral history predates this to c.1700 at the arrival of the Bantu people from

412 the Congo-basin region (Gann, 1969). The Bantu migration largely displaced resident San

413 Bushmen around the late Nineteenth Century after occupying the broader landscape for at

414 least 20,000 (and possibly over 70,000 years), leaving only remnants of their rock art as

415 sole evidence of their historical relationship with the landscape and its natural resources

416 (Suzman, 2001). Following the Bantu migration into Zambia I have relied on historical

417 texts from early explorers, hunters, traders and missionaries that have left a trail of grey

418 literature and few formal texts on the areas’ history, wildlife, people, commerce and politics

419 to frame the study area (e.g. Martelli, 1970; Sampson, 1972; Ansell, 1978; MacKenzie,

420 1997, Calvert, 2005; Macmillan & Hugh, 2005).

421 Major events shaping wildlife diversity and distribution in the Kafue-Zambezi interface can

422 broadly be broken down into the effects of commercial hunting, especially ivory and rhino

423 horn, in the late Nineteenth Century (e.g. Sampson, 1972), followed by the Rinderpest

424 Panzootic of the early 19th century that decimated game and livestock populations (Roeder

425 et al., 2006). These events were followed in the early 20th century by increased pressures on

426 habitat and wildlife associated with the development of the Zambezi Sawmills, the

427 Livingstone-Mulobezi railways and additional transport and settlement infrastructure

428 (Calvert, 2005). Zambian Independence in 1964 saw the deprioritizing of wildlife

429 conservation versus other development priorities (Gibson, 1999) at around the time the

430 Angolan Bush War (1966-1989) expanded into southwestern Zambia. During this conflict

431 foreign combatants set up encampments in the broader Simalaha area and used it as a base

432 to exploit local wildlife for rations and profit. Following cessation of hostilities small arms

433 proliferated, and in conjunction with expanding human population and limited funding for

434 law enforcement and natural resource management, ongoing unsustainable harvesting of

435 wildlife continued (Inyambo-Yeta, pers comms).

436

437 Based primarily on the research of Ansell (1978), and notwithstanding likelihood of non-

438 detection error, we can be confident that prior to the Angolan Bush War focal species

439 existed throughout the Kafue-Zambezi interface, and the landscape should thus be

440 considered connected at the large carnivore scale.

441 Contemporary drivers impacting species distribution and connectivity:

442 Kafue National Park is Zambia’s oldest and largest Protected Area at 22,480 km2, the

443 largest IUCN Category II National Park in the Kavango-Zambezi TFCA and second largest

444 throughout Africa (UNEP-WCMC, 2015). Kafue National Park is surrounded by nine

445 IUCN category VI Game Management Areas and multiple Forest Reserves with no

446 bordering fences inhibiting wildlife movement, creating a single effective wildlife managed

447 area, variously termed the Greater Kafue Ecosystem or System covering c.68,000 km2,

448 marginally smaller than Scotland, or about half the area of England or Greece (Fig 1.2).

449 Formally unsettled (fully) Protected Areas cover 33% of this landscape, with the remaining

450 67% under Customary tenure of local Chiefdoms. Of this 35% is settled (partially

451 protected) Game Management Areas, 3% is registered Communal Conservancies and 10%

452 State Forest Reserves, all of which are considered potentially partially protected in light of

453 significant funding and management constraints over the long term. The remaining 19%

454 encompasses further Open Communal area and heavily settled former wildlife managed

455 areas, now converted to agro-pastoralism and degazetted from the Zambia’s formal wildlife

456 estate.

457 458 Figure 1.2. Protected Areas, Kafue-Zambezi Interface, indicating proposed Wildlife 459 Dispersal Corridor (WDC).

460

461 Broadly, according to Lindsey et al. (2014), Kafue National Park and surrounding Game

462 Management Areas are under-performing in socioecological and economic perspectives

463 due to effects related to a rapidly expanding human populations, stubborn poverty levels

464 and de facto open-access systems in Game Management Areas. This results in widespread

465 bushmeat poaching and habitat encroachment which underfunded and under-resourced

466 State Authorities struggle to manage. State Authorities also extract revenues from Game

467 Management Areas to cover shortfalls in operational costs, limiting sufficient devolution of

468 user-rights over wildlife to communities, marginalizing their legal benefits from wildlife

469 and disincentivizing wildlife conservation over other land uses and livelihood

470 opportunities. Additionally, the photo-tourism industry is under-developed, and

471 unfavorable terms combined with corruption discourage investment and good practice by

472 hunting operators. Finally, blurred responsibilities surrounding anti-poaching in Game

473 Management Areas drives under-investment by all stakeholders. In combination these

474 challenges have resulted in major wildlife reductions in Kafue National Park and

475 surrounding Game Management Areas, and a loss of wildlife habitat in the Game

476 Management Areas.

477 Most of the Greater Kafue System lies between 900-1100 m above sea level. Rainfall

478 averages 650 mm in the south and 1,050 mm in the north, falling predominantly from

479 November to April. Vegetation is characterized by the Zambezian Miombo woodland

480 Ecoregion, typical of large areas throughout southern and eastern Africa, dominated by

481 Brachystegia spp., Combretum spp., Mopane spp., Terminalia spp. and Baikaea spp.

482 characteristic of substrate and precipitation. Woodlands are interspersed by open floodplain

483 grasslands and drainage lines (ZAWA, 2010). Species records include 158 mammals, 481

484 birds, 69 reptiles, 35 amphibians and 58 fish, with the greatest antelope diversity of any

485 African National Park (21 species), an intact carnivore guild and a full complement of

486 Zambia’s large mammals with exception of Giraffe (Giraffa giraffa), Black Rhinoceros

487 (Diceros bicornis) and Tsessebe (Damaliscus lunatus) (Ansell, 1978; Moss, 2012).

488 The Greater Kafue System has been included as Zambia’s majority component within

489 Kavango-Zambezi TFCA (KAZA, 2014). Proposed connectivity to the broader Kavango-

490 Zambezi landscape is contingent on development or maintenance of a landscape level

491 linkage routing south-southwest from the Kafue National Park border through a mosaic of

492 nominally, potentially and possibly protected wildlife managed areas including Mulobezi

493 and Sichifulo Game Management Areas, Nachitwe, Martin and Machili Forest Reserves,

494 the Nyawa communal areas, and the recently proclaimed Simalaha Communal

495 Conservancy including the fenced Simalaha Wildlife Recovery Sanctuary. In concert these

496 wildlife managed areas extend the Greater Kafue System to around 7.3m ha.

497 A secondary (south-westerly) linkage passing through Mulobezi to Sioma National Park

498 (bordering Namibia and Angola) has been proposed, though our focus remains the linkage

499 broadly following the Machili stream catchment basin from the southern Kafue National

500 Park border (S16.1380, E25.3650) to the northern bank of the Zambezi River in the

501 Simalaha Wildlife Recovery Sanctuary, bordering Botswana and Namibia (S17.5550,

502 E24.9770) (KAZA, 2014).

503 The proposed landscape linkage varies in length from 140-170 km. The human population

504 is around 110,000 and growing at 2.5% pa, with a population density ≈4.0/km2 (CSO,

505 2019). Communities are centered on a few larger settlements of 5,000-10,000 residents, and

506 otherwise in clusters of scattered villages typically concentrated along water courses,

507 seasonal waterholes, and few majority hand-pumped ground water supplies. Subsistence

508 agro-pastoralists dominate this landscape, with residents largely dependent on exploiting a

509 wide range of the area’s natural resources in support of basic livelihood needs (Musgrave,

510 2016). Formal employment opportunities beyond the distant urban settlements of Sesheke

511 and Livingstone are negligible. Customary law within the Lozi, Nkoya, and Tonga

512 ethnolinguistic groups represent the de facto regional governance system under Chiefdoms

513 (Brelsford, 1965; Musgrave, 2016).

514 Biodiversity conservation funding varies widely around a low mean throughout this

515 landscape, as with the broader KAZA TFCA. While precise figures are unavailable,

516 Lindsey et al. (2017) suggests Kafue National Park operates with 10-15% of recommended

517 budgets for large African Protected Areas yet has the highest funding for biodiversity

518 conservation throughout our study area. This is followed by Mulobezi then Sichifulo Game

519 Management Areas which receive minor budget allocations from the Department of

520 National Parks and Wildlife, augmented by finance and in-kind operational support from

521 resident safari hunting operators and other conservation actors, including local and

522 international NGO’s. Nachitwe, Martin and Machili Forest Reserves have intermittently

523 received minor budgets from the Provincial Forestry Offices (ZAWA, 2010; Chifunte, pers

524 comms). The recently proclaimed Simalaha Communal Conservancy only started receiving

525 any formal wildlife resource protection as recently as 2013 following no formal

526 biodiversity conservation budgets since the Angolan Bush/South African Border War

527 spilled over into Zambia in the early 1970’s (Inyambo-Yeta, pers comms). As part of the

528 Simalaha Communal Conservancy initiative a 240 km2 fenced Wildlife Recovery Sanctuary

529 stocked with >600 head of game has been built, funded by an international NGO. This

530 significant investment is being promoted as part of a programme for developing wildlife-

531 based land uses throughout the broader Simalaha Communal Conservancy (PPF, 2019). It

532 is unclear if the Nyawa Communal areas receives any formal wildlife management budget

533 or support.

534 1.4 Connectivity and Large Carnivores in the KAZA TFCA

535 Fragmentation and loss and wildlife habitat has direct effects on landscape structure by

536 isolating habitat patches and restricting or severing organism-scale movements between

537 patches (Taylor et al., 1993). As habitat fragmentation and loss expands any remaining

538 natural areas are partitioned into ever smaller, more isolated patches within an expanding

539 human-modified matrix, creating increased barriers to movement (Crooks et al., 2011).

540 These barriers limit resource access for wildlife, including seasonal water sources, hunting

541 and foraging areas, refugia and breeding sites (Crooks & Sanjayan, 2006). At larger spatio-

542 temporal scales barriers can impact adaptive responses to broader drivers including climate

543 change (Heller & Zavalenta, 2009). Barriers limiting ecological processes surrounding

544 immigration, emigration and dispersal, constrains gene flow and genetic health within and

545 between populations (Brook et al., 2002).

546 Large carnivores are especially vulnerable to habitat loss and fragmentation due to intrinsic

547 biological traits - large body size and extensive home range requirements, low densities and

548 slow population growth rates (Cardillo et al., 2005), combined with external anthropogenic

549 threats including effects of legal and illegal persecution (Balme et al., 2009) and prey

550 depletion (Wolf & Ripple, 2016). Together these drivers create edge effects (Woodroffe &

551 Ginsberg, 1998) and ecological traps limiting effective refugia from direct and indirect

552 human disturbance (Pitman et al., 2015) and constraining species ability to persist in

553 landscapes without functional linkages or a permeable matrix between suitable habitat

554 (Crooks et al., 2011).

555 Consequently, connectivity between wildlife managed areas is widely considered

556 fundamental to the long-term survival of majority large carnivore populations beyond the

557 very largest or intensively managed Protected Areas (Woodroffe & Ginsberg, 1998; Bauer

558 et al., 2015). Promoting landscape-level connectivity has been identified as a key

559 management objective for wildlife managed area networks, including dominant narratives

560 behind the development of the KAZA TFCA Programme (Van der Meer et al., 2016;

561 KAZA, 2018).

562 A core objective of the KAZA TFCA is to:

563 ”…ensure connectivity between key wildlife areas, where necessary, join

564 fragmented wildlife habitats in order to form an interconnected mosaic of Protected Areas,

565 as well as restore transboundary wildlife migratory corridors between wildlife dispersal

566 areas. These corridors re-establish and conserve large-scale ecological processes that

567 extend beyond the boundaries of Protected Areas” (KAZA, 2014).

568 And in the Zambian component of the KAZA TFCA similar explicit connectivity

569 objectives are identified:

570 “to join fragmented wildlife habitats into an interconnected mosaic of Protected

571 Areas and transboundary wildlife corridors, which will facilitate and enhance the free

572 movement of animals across international boundaries.” (ZAWA, 2008)

573 Empirical movement and connectivity studies are a relatively novel addition to our

574 understanding of natural resource management in the KAZA TFCA, and mainly limited to

575 elephant and large herbivore movements (e.g. Naidoo et al., 2012 & 2018; Metcalfe &

576 Kepe, 2008; Gerhardt-Weber & Katharina, 2011). While much anecdotal evidence and grey

577 literature supports transboundary movement of large carnivore throughout the KAZA

578 TFCA (KAZA, 2018), published literature is scarce. The work of Elliot et al. (2014) and

579 Cushman et al. (2016 & 2018) provide a substantive body of work surrounding lion

580 movement and connectivity models at the broader KAZA TFCA scale based on intensive

581 studies from one population. In reality few data support broader connectivity and directed

582 movements maps for large carnivores across much of the KAZA TFCA landscape beyond

583 aspirational maps (Figure 1.3). The Kafue–Chobe Wildlife Dispersal Corridor is a case in

584 point with no existing evidence of wildlife movement, or species-level studies from this

585 landscape.

586 587 Figure 1.3. KAZA TFCA Wildlife Dispersal Corridors (KAZA 2014).

588

589

590

591 1.5 Focal Species: Large Carnivores

592 The large carnivore guild are an iconic group of apex predators (Wallach et al., 2015) and

593 an integral component of complex food webs, driving ecosystem function, structure and

594 resilience by exerting top-down regulatory pressures on prey and their habitats (Estes et al.,

595 2011; Ripple et al., 2014). They also represent a pivotal resource in the promotion of

596 wildlife-based land uses, including tourism (Funston et al., 2013), thus serving a multitude

597 of ecological, socio-economic and political functions (Chapron & Lopez-Bao, 2014).

598 As apex predators at the top of the food chain large carnivores exist at low densities, even

599 in ideal conditions (Gittleman & Harvey, 1982). Their wide-ranging behaviour and feeding

600 habits commonly bring them into conflict with human due largely to real or perceived

601 threats to human lives and their livelihoods, and broader competition over shared resources

602 (Treves & Karanth, 2003). Direct and indirect human disturbance pressures on the large

603 carnivore guild is driving loss and fragmentation of habitat (Crooks, 2002), reduction of

604 wild prey base (Wolf & Ripple, 2016) and widespread persecution (Woodroffe, 2000). In

605 concert, intrinsic biological traits and exposure to external human pressures is driving

606 dramatic declines in range and population for the majority of species and increasing

607 extinction risk, even within and surrounding formally Protected Areas designated to

608 conserve them in the long term (Cardillo et al., 2004; Woodroffe & Ginsberg, 1998).

609 The Kafue-Zambezi landscape is known from historical (Ansell, 1978) and contemporary

610 records (Lines et al., 2018) to maintain an intact large carnivore guild, incorporation Lion

611 (Panthera leo), Leopard (Panthera pardus), Cheetah (Acinonyx jubatus), African wild dog

612 (Lycaon pictus) and Spotted Hyena (Crocuta crocuta). Initially we attempted to model

613 habitat suitability and connectivity for all five species known from the landscape based on a

614 pilot study to determine optimal sampling effort to detect target species and cover the

615 landscape in a single field season within an occupancy framework (Lines & MacKenzie,

616 unpublished data), though detection probability for African wild dog and cheetah proved

617 too small to generate sufficient data beyond occurrence, and we decided to omit these two

618 species for further analyses.

619 1.5.1 African Lion (Fig 1.4) are amongst our most well-known and iconic large mammals,

620 characteristic of Savanna and woodland biomes to which most populations are now

621 restricted to (Riggio et al., 2013). Lion are the most social of the felid family with related

622 females forming the basis of prides and related and unrelated males, singularly or in

623 coalitions, competing over pride tenure. Pride size averages four to six adults plus sub-

624 adults and cubs (Schaller, 1976). While lions tend to live at higher densities than most other

625 felids, the density of animals in resident populations varies dramatically from 0.5

626 adults/100 km² in the semi-desert of northwest Namibia to 55/100 km² in parts of the

627 Serengeti (Sunquist & Sunquist 2017). Pride range varies from 266 km² to 4,532 km²

628 (Bauer et al., 2016). In the absence of significant human pressures both density and home

629 range is heavily influenced by availability of preferred prey- medium to large bodied

630 mammals (Hayward & Kerley, 2005).

631

632 Figure 1.4. Lions mating, Kafue National Park. Credit R. Lines.

633 Lions are one of few species known to prey on humans, though such events are rare (Packer

634 et al., 2007). Where lions and human coexist, interactions are typically characterized by

635 conflict (Sillero-Zubiri & Laurenson, 2001; Thirgood et al., 2005). Such conflict has

636 resulted in large scale declines in lion population size and distribution, and the species are

637 categorized as ‘Threatened/Vulnerable A2abcd’ and declining, numbering 23,000-39,000

638 adults and yearlings across Africa. This represents a 42% decline over 3 generations

639 (22.3yrs) based on longitudinal data from 47 well-studied sub-populations (Bauer et al.,

640 2016). But course-scale analyses mask significant variation within and between countries

641 and by sub-populations. The KAZA TFCA maintains 13 lion conservation units (Bauer et

642 al., 2016) and the Okavango-Hwange complex has one of Africa’s 10 remaining lion

643 ‘strongholds’ (Riggio et al., 2013). Data on lion in the Greater Kafue System is patchy,

644 with studies restricted to northern sections of Kafue National Park (Midlane, 2014). This

645 study represents the only systematic body of work for southern Kafue including the mosaic

646 of land uses through to the Zambezi River.

647 1.5.2 Leopard (Fig 1.5) represents the smallest of four species in the Panthera

648 genus. They are largely solitary and highly adaptable, with the widest range of all wild cats.

649 Populations are found throughout sub-Saharan Africa, the Middle East, and South Asia to

650 the Russian Far East (Stein & Hayssen, 2013). Leopard’s behavioral plasticity allows them

651 to exist across a wide variety of habitats including forest, woodland, mountain, savanna and

652 semiarid environments, with some the highest densities recorded in suburban and extra-

653 urban areas (Jacobson et al., 2016). Diet reflects the species behavioral plasticity with a

654 preference for small to medium sized ungulates, though leopard readily eat fish, reptiles,

655 insects and birds (Hayward et al., 2006). In human dominated landscapes domestic animals,

656 including dogs and livestock, can form a large part of leopard prey, causing much human-

657 leopard conflict (Mukherjee et al., 2001; Athreya et al., 2014). Attacks on humans,

658 provoked or otherwise, are considered rare in southern Africa (Dunham et al., 2010),

659 though extensive records of man-eaters and attacks on humans exist elsewhere (e.g. Corbett

660 & Gobetti, 1946).

661 As with lion, leopard population density broadly tracks preferred prey biomass and habitat

662 productivity (Hayward et al., 2007; Macdonald & Loveridge, 2010), varying 300-fold from

663 0.1 – 30.9 individuals/100 km2 (Boast & Houser, 2012; Edgaonkar, 2008), complicating

664 reliable estimation of global population numbers across their range (Jacobson et al., 2016).

665 Despite leopard’s wide distribution the IUCN lists them as “Vulnerable A2cd” and several

666 Asian subspecies are listed as Endangered (Stein et al., 2017), with many sub-species in

667 decline throughout much of their range (Jacobson et al., 2016). The causes of these declines

668 are almost entirely anthropogenic: direct persecution (Thorn et al., 2013), illegal wildlife

669 trade including skins for cultural regalia (Datta et al., 2008, Balme, unpublished data),

670 poorly managed trophy hunting (Balme et al., 2009), prey declines (e.g. Lindsey et

671 al., 2012; Selvan et al., 2014) and habitat fragmentation (Gerland et al., 2014).

672 Stein et al. (2016) estimates a >30% worldwide range loss for the species in the last three

673 generations (22.3 years), and 21% range loss in southern Africa.

674 675 Figure 1.5. Male leopard, Kafue National Park. Credit: R. Lines

676 Knowledge of leopard in and around Kafue is limited to historical studies (Mitchelle, 1965;

677 Ansell, 1979; Bertram, 1982), and in broader scale analyses studying the impact of the

678 bushmeat poaching crisis (Lindsey et al., 2013). While research suggests ungulate

679 populations have perhaps increased by 24% across southern Africa as a whole (Craigie et

680 al., 2010), finer scale analyses in Zambia indicates severe reductions in Leopard’s principle

681 prey inside and outside of Protected Areas (Lindsey et al., 2014; Watson et al., 2015).

682 Zambian Game Management Areas and National Parks maintain large mammal populations

683 at 93.7% and 74.1% below estimated carrying capacity (Lindsey et al., 2014), with severe

684 implications for all large carnivores.

685 1.5.3 The Spotted hyena (Fig 1.6) is a highly social and adaptable species, and the sole

686 extant representative of the Crocuta genus. Spotted hyena live in matrilineal clans and are

687 found in a wide range of habitats from desert, to mountain, savannah and woodland (East &

688 Hofer, 2013). Sub-populations can be found in close proximity to human settlements,

689 notably in Ethiopia where co-existence with humans has a strong cultural/spiritual

690 association (Yirga et al., 2016; Abay et al., 2011). Early misconceptions surrounding their

691 scavenging habits have long been dispelled, and Spotted hyena are now understood to be

692 flexible and successful hunters, preying on a wide range of species, alongside their

693 scavenging abilities (Kruuk, 1972; Höner et al., 2005). As with leopard they can also prey

694 on domestic animals, causing similar human-wildlife conflict, though attacks on humans

695 are rare (Bohm & Höner, 2015; Butler et al., 2013).

696 Clan size varies from 5-80 individuals with densities ranging from 0.6-240/100 km2 –

697 exhibiting even greater variability than leopard (Hofer & Mills 1998). Home ranges vary

698 from 13 km2 to 1,065 km2 from high prey areas in productive habitat, to semi-desert

699 (Holecamp & Dloniak, 2010). Clan size and density are considered related to food

700 availability and primary production (Hayward et al., 2006; Mills, 1990).

701 The IUCN provides a global population estimate of 27,000-47,000 adults, classifying the

702 species as ‘Least Concern’, though there is much uncertainty in status within and between

703 countries and sub-populations, and a reassessment is current underway incorporating data

704 from this study (Weise, pers comms). Nonetheless the majority of southern Africa’s larger

705 Protected Area populations, and several populations in eastern Africa, are considered stable

706 (Bohn & Honer, 2015). But beyond, and increasingly within Protected Areas, direct and

707 indirect persecution is common. Spotted hyena are notable bycatch in snares set for

708 ungulates, as with other large carnivores (Lindsey et al., 2013), and this can be the cause of

709 majority adult mortality (Hofer et al., 1996). Spotted Hyaenas are also susceptible to mass

710 poisoning events as both target species and as bycatch (Holekamp et al., 1993). While the

711 numbers of hyena killed by sport hunting is low, impacts of legal off take can be drastic as

712 hunters tend to target the largest individuals, commonly the alpha females, with severe

713 implications for clan structure and breeding (Cozzi et al., 2015). Hyena are also killed for

714 food, medicine and witchcraft in many countries (Hofer & Mills 1998). As with most large

715 carnivores hyenas are also threatened by prey reduction, habitat loss, overgrazing by

716 livestock and game-meat hunting by humans (Bohn & Honer, 2015).

717 718 Figure 1.6. Spotted Hyena, Mulobezi Game Management Area, Zambia. Credit: R. Lines.

719 There have been few formal studies on Spotted hyena in Zambia (Berentsen et al., 2012;

720 M’soka et al., 2016), and no formal research on Spotted hyena in the Greater Kafue

721 System, though reports and sightings are widespread from studies on sympatric carnivore

722 (e.g. Mitchelle, 1965; Midlane et al., 2015) and via trophy hunting records (PHAZ, 2019).

723 1.6 Thesis Structure

724 Here, in Chapter 1, the Introduction, I have provided a general overview of Protected Areas

725 and Protected Area Networks in context of the broader Kavango-Zambezi Transfrontier

726 Conservation Area, the Greater Kafue Ecosystem and the Kafue-Zambezi Interface,

727 including a summary outlining key social, economic, environmental and cultural drivers

728 shaping the Kafue-Zambezi landscape. I then summarized connectivity considerations for

729 this landscape, and within the broader KAZA TFCA, then reviewed the status, ecology and

730 relevance of target species for connectivity and for broader conservation outcomes.

731 Chapter 2 presents an exhaustive historical analysis of wildlife in the Kafue-Zambezi

732 interface, referencing data and anecdotal evidence from early hunters and missionaries

733 through to grey literature and available published records, including oral records from

734 meetings with Traditional Authorities. I identify key drivers of changes to species

735 composition and distribution by wildlife managed area prior to contemporary fieldwork,

736 generating a foundational understanding of the area’s wildlife. I then integrate multiple

737 datasets from Government anti-poaching operations, interviews with safari hunters and

738 extensive spoor tracking to provide up to date analyses on the current status of 31 terrestrial

739 mammals known from the landscape from our historical analyses. This chapter was

740 published as Lines et al. (2018).

741 Chapter 3 delves into the argument for single and multi-species connectivity across this

742 landscape using lion, leopard and spotted hyena as proxies for broader habitat suitability

743 and connectivity analyses. I employ high resolution Sentinel imagery (Copernicus Sentinel

744 data, 2018). and extensive ground truthing to generate a landcover map at 10m resolution

745 for 38,000km2 , achieving >91% classification accuracy. I then integrate empirical

746 occurrence data from Chapter 1 with the landcover map to model habitat suitability for each

747 species, then apply a further analytical step to model potential for single and multi-species

748 corridors across this landscape, identifying a novel area for protection to increase scope and

749 scale for connectivity. This chapter is in review with Oryx (Lines et al., 2020).

750 Chapter 4 develops an innovative use of fine-scale, site-specific human footprint pressure

751 mapping as a proxy for understanding species sensitivity to anthropogenic disturbance

752 throughout the Kafue-Zambezi interface. Seven human impact layers are integrated from

753 our landcover maps plus supplemental sources to generate a large-scale, fine resolution

754 human footprint map covering c.4 m ha. I then use this map to model effects of derived

755 Human footprint pressure on lion, leopard and spotted hyena occurrence. Finally, I

756 undertake preliminary exploratory analysis to investigate if model and map outputs can

757 determine mean human footprint pressure thresholds at which species occur at the wildlife

758 managed areas scale.

759 In Chapter 5, the Discussion, I synthesize the three data chapters and contextualize with

760 important supplemental information and insights. This provides a meaningful interpretation

761 of scope and scale for single and multi-species connectivity given historical data, current

762 analyses and prospects under various future management interventions. I identify gaps in

763 knowledge and additional research priorities for our core study area and the broader KAZA

764 TFCA landscape. This produces a holistic understanding of connectivity at the large

765 carnivore scale for this dynamic, heterogenous landscape is thus presented.

766

767

768 1.7 References

769 Abay, G. Y., et al. (2011). Peri-urban spotted hyena (Crocuta crocuta) in northern

770 Ethiopia: diet, economic impact, and abundance. European Journal of Wildlife

771 Research, 57(4), 759-765.

772 Andersson, J., et al. (2017). Transfrontier Conservation Areas: People Living on the Edge.

773 Taylor & Francis.

774 Ansell, W. F. H. (1978). The Mammals of Zambia. National Parks & Wildlife Service.

775 Athreya, V., et al. (2016). A cat among the dogs: Leopard (Panthera pardus) diet in a

776 human-dominated landscape in western Maharashtra, India. Oryx, 50(1), 156-162.

777 Balme, G. A., et al. (2009). Evaluating methods for counting cryptic carnivores. The

778 Journal of Wildlife Management, 73(3), 433-441.

779 Balme, G. A., et al. (2009). Impact of conservation interventions on the dynamics and

780 persistence of a persecuted leopard (Panthera pardus) population. Biological

781 Conservation, 142(11), 2681-2690.

782 Barrett, M., et al. (2018). Living planet report 2018: Aiming higher.

783 Bauer, H., et al. (2016). Panthera leo (errata version published in 2017). The IUCN Red

784 List of Threatened Species 2016: e.T15951A115130419.

785 http://dx.doi.org/10.2305/IUCN.UK.2016-3.RLTS.T15951A107265605.en. Accessed

786 04/12/2019.

787 Bauer, H., et al. (2015). Lion (Panthera leo) populations are declining rapidly across Africa,

788 except in intensively managed areas. Proceedings of the National Academy of

789 Sciences, 112(48), 14894-14899.

790 Bellard, C., et al. (2012). Impacts of climate change on the future of biodiversity. Ecology

791 Letters, 15(4), 365-377.

792 Berentsen, A. R., et al. (2012). Survey of gastrointestinal parasite infection in African lion

793 (Panthera leo), African wild dog (Lycaon pictus) and spotted hyaena (Crocuta

794 crocuta) in the Luangwa Valley, Zambia. African Zoology, 47(2), 363-368.

795 Biggs, R., et al. (2008). Scenarios of biodiversity loss in southern Africa in the 21st

796 century. Global Environmental Change, 18(2), 296-309.

797 Boast, L. K. and Houser, A. (2012). Density of large predators on commercial farmland in

798 Ghanzi, Botswana. African Journal of Wildlife Research, 42(2), 138-144.

799 Bohm, T. & Höner, O.R. (2015) Crocuta crocuta. The IUCN Red List of Threatened

800 Species: e.T5674A45194782. http://dx.doi.org/10.2305/IUCN.UK.2015-

801 2.RLTS.T5674A45194782.en. accessed 04/12/2019.

802 Brandon, K., Redford, K.H. and Sanderson, S. eds., (1998). Parks in peril: people, politics,

803 and protected areas. Island Press.

804 Brelsford, W. V. (1965). The Tribes of Zambia. Printed by the Govt. Printer.

805 Brook, B. W., et al. (2002). Contribution of inbreeding to extinction risk in threatened

806 species. Conservation Ecology, 6(1). 26

807 Butler, J. R., et al. (2013). Dog eat dog, cat eat dog: social-ecological dimensions of dog

808 predation by wild carnivores. Free-Ranging Dogs and Wildlife Conservation. Oxford

809 University Press, Oxford, 117-143.

810 Cardillo, M., et al. (2004). Human population density and extinction risk in the world's

811 carnivores. PLoS Biology, 2(7), e197.

812 Cardillo, M., et al. (2005). Multiple causes of high extinction risk in large mammal species.

813 Science (New York, N.Y.), 309(5738), 1239-1241.

814 Chapron, G. and Lopez-Bao, J. V. (2014). Conserving carnivores: politics in play. Science

815 (New York, N.Y.), 343(6176), 1199-1200.

816 Corbett, J. and Gobetti, J. (1946). Man-Eaters of Kumaon. Vol. 109. Oxford University

817 Press New York.

818 Cozzi, G., et al. (2015). Effects of trophy hunting leftovers on the ranging behaviour of

819 large carnivores: A case study on spotted hyenas. PloS One, 10(3), e0121471.

820 Craigie, I. D., et al. (2010). Large mammal population declines in Africa’s Protected Areas.

821 Biological Conservation, 143(9), 2221-2228.

822 Crooks, K. R. (2002). Relative sensitivities of mammalian carnivores to habitat

823 fragmentation. Conservation Biology, 16(2), 488-502.

824 Crooks, K. R. and Sanjayan, M. (2006). Connectivity Conservation. Cambridge University

825 Press.

826 Crooks, K. R., et al. (2011). Global patterns of fragmentation and connectivity of

827 mammalian carnivore habitat. Philosophical Transactions of the Royal Society B:

828 Biological Sciences, 366(1578), 2642-2651.

829 Copernicus Sentinel data (2018). Retrieved from ASF DAAC. Accessed 01/06/2018.

830 Cumming, D. H. (2008). Large scale conservation planning and priorities for the Kavango-

831 Zambezi Transfrontier Conservation Area. Unpublished Report Commissioned by

832 Conservation International.

833 Cushman, S. A., et al. (2016). A multi-scale assessment of population connectivity in

834 African lions (Panthera leo) in response to landscape change. Landscape Ecology,

835 31(6), 1337-1353.

836 Cushman, S.A., et al. (2018). Prioritizing core areas, corridors and conflict hotspots for lion

837 conservation in southern Africa. PloS one, 13(7), p.e0196213.

838 Datta, A., et al. (2008). Empty forests: Large carnivore and prey abundance in Namdapha

839 National Park, north-east India. Biological Conservation, 141(5), 1429-1435.

840 Duffy, R. (2006). Global governance and environmental management: the politics of

841 transfrontier conservation areas in Southern Africa. Political Geography, 25, 89-112.

842 East, M. and Hofer, H. (2013). Crocuta crocuta spotted hyena. Mammals of Africa, 5, 273-

843 281.

844 Edgaonkar, A. (2008). Ecology of the Leopard (Panthera pardus) in Bori Wildlife

845 Sanctuary and Satpura National Park, India. Citeseer. 28

846 Elliot, N. B., et al. (2014). Movements vary according to dispersal stage, group size, and

847 rainfall: the case of the African lion. Ecology, 95(10), 2860-2869.

848 Estes, J. A., et al. (2011). Trophic downgrading of planet Earth. Science (New York, N.Y.),

849 333(6040), 301-306.

850 Ferguson, K. and Hanks, J. (2012). The effects of Protected Area and veterinary fencing on

851 wildlife conservation in southern Africa. Parks, 18(1), 49-60.

852 Funston, P. J., et al. (2013). Insights into the management of large carnivores for profitable

853 wildlife-based land uses in African savannas. PloS One, 8(3), e59044.

854 Gerhardt-Weber, V. and Katharina, E. (2011). Elephant Movements and Human-Elephant

855 Conflict in a Transfrontier Conservation Area.

856 Gerland, P., et al. (2014). World population stabilization unlikely this century. Science

857 (New York, N.Y.), 346 (6206), 234-237.

858 Gittleman, J. L. and Harvey, P. H. (1982). Carnivore home-range size, metabolic needs and

859 ecology. Behavioral Ecology and Sociobiology, 10(1), 57-63.

860 Hanks, J. (2000). The role of Transfrontier Conservation Areas (TFCAs) in Southern Africa

861 in the conservation of mammalian biodiversity. Priorities for the Conservation of

862 Mammalian Diversity: Has the Panda had its Day? 3, 239.

863 Hayward, M. W. and Kerley, G. I. (2005). Prey preferences of the lion (Panthera leo).

864 Journal of Zoology, 267(3), 309-322.

865 Hayward, M., et al. (2006). Prey preferences of the leopard (Panthera pardus). Journal of

866 Zoology, 270(2), 298-313.

867 Hayward, M. W., et al. (2007). Carrying capacity of large African predators: predictions

868 and tests. Biological Conservation, 139(1), 219-229.

869 Heller, N. E. and Zavaleta, E. S. (2009). Biodiversity management in the face of climate

870 change: a review of 22 years of recommendations. Biological Conservation, 142(1),

871 14-32.

872 Hofer, H., et al. (1996). The impact of game meat hunting on target and non-target species

873 in the Serengeti. In: The Exploitation of Mammal Populations. Springer, 117-146.

874 Holekamp, K., et al. (1993). Fission of a spotted hyena clan: consequences of prolonged

875 female absenteeism and causes of female emigration. Ethology, 93(4), 285-299.

876 Holekamp, K. E. and Dloniak, S. M. (2010). Intraspecific variation in the behavioral

877 ecology of a tropical carnivore, the spotted hyena. In: Advances in the Study of

878 Behavior. Vol. 42. Elsevier, 189-229.

879 Höner, O. P., et al. (2005). The effect of prey abundance and foraging tactics on the

880 population dynamics of a social, territorial carnivore, the spotted hyena. Oikos, 108(3),

881 544-554.

882 Jacobson, A. P., et al. (2016). Leopard (Panthera pardus) status, distribution, and the

883 research efforts across its range. Peerj, 4, e1974.

884 Jones, K. R., et al. (2018). One-third of global protected land is under intense human

885 pressure. Science (New York, N.Y.), 360 (6390), 788-791.

886 KAZA (2018). The Kavango Zambezi Transfrontier Conservation Area Carnivore

887 Conservation Strategy. KAZA TFCA Secretariat, Hwange, Zimbabwe.

888 KAZA TFCA Secretariat (2014). Master Integrated Development Plan. KAZA Secretariat.

889 Victoria Falls.

890 KAZA TFCA SECRETARIAT (2011a). Treaty Between the Governments of Angola,

891 Botswana, Namibia, Zambia and Zimbabwe on the Establishment of the Kavango–

892 Zambezi Transfrontier Conservation Area. Kavango–Zambezi TFCA Secretariat,

893 Kasane, Botswana.

894 KAZA TFCA Secretariat Master Integrated Development Plan (2014). Kavango–Zambezi

895 TFCA Secretariat, Victoria Falls, Zambia.

896 Kruuk, H. (1972). The Spotted Hyena: A Study of Predation and Social Behavior. (No.

897 Sirsi) a102104.

898 Leverington, F., Costa, K.L., Pavese, H., Lisle, A. and Hockings, M., 2010. A global

899 analysis of protected area management effectiveness. Environmental management,

900 46(5), 685-698.

901 Lindsey, P., et al. (2012). Illegal hunting and the bush-meat trade in savanna Africa:

902 drivers, impacts and solutions to address the problem. Panthera/Zoological Society of

903 London. Wildlife Conservation Society Report, New York, 74.

904 Lindsey, P., et al. (2013). Benefits of wildlife-based land uses on private lands in Namibia

905 and limitations affecting their development. Oryx, 47(1), 41-53.

906 Lindsey, P. A., et al. (2013). The bushmeat trade in African savannas: Impacts, drivers, and

907 possible solutions. Biological Conservation, 160, 80-96.

908 Lindsey, P., et al. (2014). Underperformance of the Zambian Protected Area network: steps

909 needed to improve functionality and effectiveness. Plos One, 9(5), e94109.

910 Lindsey, P., et al. (2017). The performance of African Protected Areas for lions and their

911 prey. Biological Conservation, 209, 137-149.

912 Lines, R., et al. (2018). Status of terrestrial mammals at the Kafue–Zambezi interface:

913 implications for transboundary connectivity. Oryx, 1-10.

914 Macdonald, D. and Loveridge, A. (2010). The Biology and Conservation of Wild Felids.

915 Vol. 2. Oxford University Press.

916 Margules, C. R. and Pressey, R. L. (2000). Systematic conservation planning. Nature,

917 405(6783), 243-253.

918 Maxwell, S. L., et al. (2016). Biodiversity: The ravages of guns, nets and bulldozers.

919 Nature News, 536(7615), 143.

920 Metcalfe, S. and Kepe, T. (2008). “Your Elephant on Our Land” The Struggle to Manage

921 Wildlife Mobility on Zambian Communal Land in the Kavango-Zambezi Transfrontier

922 Conservation Area. The Journal of Environment & Development, 17(2), 99-117.

923 Midlane, N. (2014). The Conservation Status and Dynamics of a Protected African Lion

924 Panthera leo Population in Kafue National Park, Zambia. (Doctoral dissertation,

925 University of Cape Town).

926 Midlane, N., et al. (2015). To track or to call: comparing methods for estimating population

927 abundance of African lions Panthera leo in Kafue National Park. Biodiversity and

928 Conservation, 24(6), 1311-1327.

929 Mills, G. and Hofer, H. (1998). Hyaenas: Status Survey and Conservation Action Plan.

930 IUCN, Gland, Switzerland. SSC Hyaena Specialist Group.

931 Mills, M. G. (1990). Kalahari Hyaenas. Unwin Hyman.

932 Mitchelle, B. (1965). Predation on large mammals in the Kafue National Park, Zambia.

933 African Zoology, 1(2).

934 Moss, P. (2012). Handbook to Kafue National Park. Future Publishing (PTY) Ltd.

935 M'soka, J., et al. (2016). Spotted hyaena survival and density in a lion depleted ecosystem:

936 The effects of prey availability, humans and competition between large carnivores in

937 African savannahs. Biological Conservation, 201, 348-355.

938 Mukherjee, S. and Mishra, C. (2001). Predation by leopard (Panthera pardus) in Majhatal

939 Harsang Wildlife Sanctuary, W. Himalayas. Journal of the Bombay Natural History

940 Society, 98(2),7-2.

941 Murphree, M. (2017). in Andersson, J., et al. (2017). Transfrontier Conservation Areas:

942 People Living on the Edge. Taylor & Francis, 15. 33

943 Musgrave, M. (2016). Scale, Governance and Change in Zambezi Teak Forests:

944 Sustainable Development for Commodity and Community. Cambridge Scholars

945 Publishing.

946 Naidoo, R., et al. (2012). Home on the range: factors explaining partial migration of

947 African buffalo in a tropical environment. PloS One, 7(5), e36527.

948 Naidoo, R., et al. (2018). Evaluating the effectiveness of local- and regional-scale wildlife

949 corridors using quantitative metrics of functional connectivity. Biological

950 Conservation, 217, 96-103.

951 Newmark, W. D. (2008). Isolation of African Protected Areas. Frontiers in Ecology and the

952 Environment, 6(6), 321-328.

953 Pachauri, R. K., et al. (2014). Climate Change 2014: Synthesis Report. Contribution of

954 Working Groups I, II and III to the Fifth Assessment Report of the Intergovernmental

955 Panel on Climate Change.

956 Packer, C., et al. (2007). The ecology of man-eating lions in Tanzania. Accra (Ghana)

957 FAO.

958 PHAZ (2019). Professional Hunters Association of Zambia. Record Book

959 https://www.phazambia.com/zambia-record-book/. Accessed 02/11/2019

960 Pitman, R. T., et al. (2015). The importance of refugia, ecological traps and scale for large

961 carnivore management. Biodiversity and Conservation, 24(8), 1975-1987.

962 PPF (2019). The Kavango-Zambezi Transfrontier Conservation Area. Peace Park

963 Foundation, Stellenbosch, South Africa.

964 Pringle, R. M. (2017). Upgrading Protected Areas to conserve wild biodiversity. Nature,

965 546(7656), 91-99.

966 Riggio, J., et al. (2013). The size of savannah Africa: a lion’s (Panthera leo) view.

967 Biodiversity and Conservation, 22(1), 17-35.

968 Ripple, W. J., et al. (2014). Status and ecological effects of the world’s largest carnivores.

969 Science, 343(6167).

970 Sanderson, E.W., Jaiteh, M., Levy, M.A., Redford, K.H., Wannebo, A.V. and Woolmer, G.,

971 2002. The human footprint and the last of the wild: the human footprint is a global

972 map of human influence on the land surface, which suggests that human beings are

973 stewards of nature, whether we like it or not. BioScience, 52(10), 891-904.

974 Schaller, G. B. (1976). The Serengeti Lion: A Study of Predator-Prey Relations. University

975 of Chicago Press.

976 Selvan, K. M., et al. (2014). Population density and abundance of sympatric large

977 carnivores in the lowland tropical evergreen forest of Indian Eastern Himalayas.

978 Mammalian Biology, 79(4), 254-258.

979 Sillero-Zubiri, C. and Laurenson, M. K. (2001). Interactions between carnivores and local

980 communities: Conflict or co-existence? Conservation Biology Series-Cambridge-, 282-

981 312.

982 Simukonda, C. (2008). A Countrywide Survey of Large Mammals, Zambia. Zambia

983 Wildlife Authority, Lusaka, Zambia.

984 Stein, A., et al. (2017). Panthera pardus (errata version published in 2016). The IUCN Red

985 List of Threatened Species 2016: e. T15954A102421779. Accessed 14/11/2018.

986 Stein, A. B. and Hayssen, V. (2013). Panthera pardus (Carnivora: Felidae). Mammalian

987 Species, 45(900), 30-48.

988 Suich, H., et al. (2012). Evolution and Innovation in Wildlife Conservation: Parks and

989 Game Ranches to Transfrontier Conservation Areas. Earthscan.

990 Sunquist, M. and Sunquist, F. (2017). Wild Cats of the World. University of Chicago press.

991 Sutherland, W. J., et al. (2004). The need for evidence-based conservation. Trends in

992 Ecology & Evolution, 19(6), 305-308.

993 Taylor, P. D., et al. (1993). Connectivity is a vital element of landscape structure. Oikos,

994 571-573.

995 Thirgood, S., et al. (2005). The impact of human-wildlife conflict on human lives and

996 livelihoods. Conservation Biology Series, Cambridge, 9, 13.

997 Treves, A. and Karanth, K. U. (2003). Human‐carnivore conflict and perspectives on

998 carnivore management worldwide. Conservation Biology, 17(6), 1491-1499.

999 UNEP-WCMC (2015). United Nations Environment Programme's World Conservation

1000 Monitoring Centre. Protected Planet Database https://www.protectedplanet.net/.

1001 Accessed 01/05/2015.

1002 United Nations (2019). Department of Economic and Social Affairs, Population Division.

1003 World Population Prospects 2019: Ten Key Findings.

1004 https://population.un.org/wpp/Publications/Files/WPP2019_10KeyFindings.pdf.

1005 Accessed 4/12/2019.

1006 Van der Meer, E., et al. (2016). Large carnivores as tourism flagship species for the

1007 Zimbabwe component of the Kavango Zambezi Transfrontier Conservation Area.

1008 African Journal of Wildlife Research, 46(2), 121-134.

1009 Wallach, A. D., et al. (2015). What is an apex predator? Oikos, 124(11), 1453-1461.

1010 Watson, F. G., et al. (2015). Human encroachment into Protected Area networks in

1011 Zambia: implications for large carnivore conservation. Regional Environmental

1012 Change, 15(2), 415-429.

1013 Wittemyer, G., et al. (2008). Accelerated human population growth at Protected Area

1014 edges. Science (New York, N.Y.), 321(5885), 123-126.

1015 Wolf, C. and Ripple, W. J. (2016). Prey depletion as a threat to the world's large carnivores.

1016 Royal Society Open Science, 3(8), 160252.

1017 Woodroffe, R. (2000). Predators and People: Using Human Densities to Interpret Declines

1018 of Large Carnivores Animal Conservation Forum. Cambridge University Press, pp.

1019 165-173.

1020 Woodroffe, R. and Ginsberg, J. R. (1998). Edge effects and the extinction of populations

1021 inside Protected Areas. Science, 280(5372), 2126-2128.

1022 Yirga, G., et al. (2016). Spotted hyena (Crocuta crocuta) concentrate around urban waste

1023 dumps across Tigray, northern Ethiopia. Wildlife Research, 42(7), 563-569.

1024 ZAWA (2010). General Management Plan for Kafue National Park. Zambia Wildlife

1025 Authority. Chilanga, Zambia.

1026 ZAWA (2008). Integrated Development Plan for the Zambian Component of the Kavango-

1027 Zambezi Transfrontier Conservation Area. Chilanga, Zambia. Peace Parks

1028 Foundation. Stellenbosch, South Africa.

1029

1030

1031

1032

1033

1034

1035 Chapter 2: Status of Terrestrial Mammals at the Kafue-Zambezi Interface:

1036 Implications for Transboundary Connectivity.

1037 Lines, R. M.,* MacMillan, D. C. and Tzanopolous, J.

1038 Durrell Institute of Conservation and Ecology, Marlow Building, The University of Kent,

1039 Canterbury, Kent, CT2 7NR. U.K.

1040 *corresponding author [email protected]

1041

1042 Abstract

1043 The Kavango-Zambezi Transfrontier Conservation Area Programme promotes landscape-

1044 level connectivity between clusters of wildlife managed areas in five neighboring countries.

1045 However, declining regional biodiversity can undermine efforts to maintain, expand and

1046 link wildlife populations. Narratives promoting species connectivity should thus be founded

1047 on studies of system and state changes in key resources.

1048 By integrating and augmenting multiple data sources throughout eight wildlife managed

1049 areas covering 1.7 m ha, we report changes from 1978-2015 to the occurrence and

1050 distribution of 31 mammal species throughout a landscape linking the Greater Kafue

1051 System to adjacent wildlife managed areas in Namibia and Botswana. Results indicate

1052 species diversity was largely unchanged in Kafue National Park, Mulobezi and Sichifulo

1053 Game Management Areas. However, 100% of large carnivore and 64% of prey diversity

1054 have been lost in the Simalaha areas in Zambia. No evidence was found of migration

1055 behaviour or species recolonization from adjacent wildlife areas was established. While

1056 temporal sampling scales impacts the definition of species occupancy and distribution, and

1057 data cannot elaborate on population size or trends, findings indicate an emerging

1058 connectivity bottleneck within Simalaha. At current disturbance levels, evidence suggests

1059 the Greater Kafue System, Zambia’s majority component in the Kavango-Zambezi

1060 Transfrontier Conservation Area, is becoming increasingly isolated at the large mammal

1061 scale contrary to prevailing narratives.

1062 Further investigations of the site-specific, interacting drivers impacting wildlife distribution

1063 and occurrence are required to provide management with appropriate conservation

1064 interventions aimed at wildlife recovery in key areas identified to promote transboundary

1065 connectivity in the Kavango-Zambezi Transfrontier Conservation Area.

1066

1067 Keywords: Kavango-Zambezi Transfrontier Area, Kafue, connectivity, mammal loss.

1068 Word count: 5449

1069 Citation: Lines, R., Tzanopoulos, J. and MacMillan, D.C., 2019. Status of terrestrial

1070 mammals at the Kafue-Zambezi Interface: Implications for transboundary connectivity.

1071 Oryx, 53(4), 764-773.

1072

1073

1074

1075

1076

1077 2.1 Introduction

1078 Wildlife managed areas are frequently clustered along international borders, with arbitrarily

1079 drawn political boundaries dividing ecosystems in which these areas occupy (Zbicz, 1999a;

1080 Hanks, 2000). Where fences and physical barriers combined with expanding human

1081 settlement and intensified agropastoralist activities, over-exploitation and extreme wildlife

1082 population decline can occur (Ogutu et al., 2016). Additionally, invasion, disease, pollution

1083 and climate change (Maxwell et al., 2016; Pachauri et al., 2014) interact with intrinsic

1084 species traits (Cardillo et al., 2008) to inhibit or sever wildlife movement patterns, isolating

1085 core wildlife managed areas (Margules & Pressey, 2000; Newmark, 2008). In concert these

1086 drivers are exposing wildlife populations to escalating edge-effects and ecological traps,

1087 threatening species persistence within and outside protected areas (Woodroffe & Ginsberg,

1088 1998; Battin, 2004).

1089 Conversely, intact species assemblages have wide-ranging implications for sustainable and

1090 resilient social-ecological systems (Cummings, 2011). Heterogeneity and functional

1091 diversity drives system productivity and its capacity to absorb, resist and respond to shocks,

1092 perturbations and other stressors that negatively impact system structure and function

1093 (Fischer et al., 2006). Cumulatively, threats to species persistence undermine habitat

1094 integrity, ecosystem services, food security, the development of sustainable wildlife-based

1095 land uses and human wellbeing (Lindsay et al., 2013; WHO/MEE, 2005).

1096 Acknowledging the limitations imposed by these constraints, stakeholders in Southern

1097 Africa are increasingly embracing Transfrontier Conservation Areas (TFCAs) as a new

1098 conservation paradigm (Hanks, 2000), considered an evolution of previous Community

1099 Based Natural Resource Management (CBNRM) approaches that yielded mixed results

1100 (Andersson, 2016). Enticing narratives include the integration of biodiversity conservation

1101 with the promotion of sustainable socioeconomic development and a culture of peace and

1102 cooperation at the ecosystem level, linked to the removal of fences and other barriers

1103 inhibiting the free movement of wildlife across vast interconnected landscapes (Linde et al.,

1104 2002, Hanks, 2003).

1105 The Kavango-Zambezi Transfrontier Conservation Area (KAZA TFCA) is working to

1106 capitalize on the regions’ unique diversity and distribution of wildlife assets by advocating

1107 shared natural resource management and development goals across an immense network of

1108 protected areas spanning over 500,000 km2 at the interface of Angola, Botswana, Namibia,

1109 Zambia and Zimbabwe (KAZA, 2011b; Hanks & Myburgh, 2015). Stated objectives to

1110 integrate conservation and development, promote peace and cooperation, and facilitate

1111 connectivity of wildlife populations between clusters of wildlife managed areas have

1112 become popular and compelling programme narratives driving north-south finance

1113 initiatives, non-government organisation engagement, and energizing State buy-in (KAZA,

1114 2011a; PPF, 2008; WWF, 2011).

1115 Notwithstanding evolving conservation and development narratives, the KAZA TFCA

1116 landscape faces many existing and emerging challenges constraining programme success.

1117 Mounting anthropogenic pressures combined with poor land use planning, institutional

1118 conflicts and stakeholder disenfranchisement (Andersson, 2016), are driving encroachment

1119 into wildlife areas, habitat loss and fragmentation (Watson et al., 2015; Newmark, 2008;

1120 Simukonda, 2008), and unsustainable harvesting of wildlife, threatening many of the

1121 Kavango-Zambezi TFCA’s iconic natural assets (Lindsay et al., 2013). With the region’s

1122 human population expected to double by 2050 (UN, 2015) and likely impacts of climate

1123 change exacerbating socioeconomic development challenges (Pachauri, et al., 2014;

1124 Bellard et al., 2012), even moderately optimist scenarios imply regional biodiversity loss

1125 will accelerate significantly this century (Briggs et al., 2008).

1126 Collectively these challenges raise important questions surrounding the scope, scale and

1127 ambition of narratives promoting landscape-level linkages, the interventions required to

1128 maintain or expand connectivity, and what purposes these proposed linkages may serve in

1129 the long term (Cumming, 2008). A clear imperative thus exists to promote evidence-based

1130 socioeconomic and environmental policies and interventions built around the application of

1131 conservation science (Sutherland et al., 2004), including research and monitoring of

1132 changes to site and system states, and their response to factors driving connectivity at the

1133 scale of interest. But the process of informed decision-making is data hungry. Local,

1134 regional and transboundary data sources are disparate and inconsistent, undermining

1135 attempts to understand complex social ecological systems such as the Kavango-Zambezi

1136 TFCA. Data deficiencies ultimately constrain effective decision making and appropriate

1137 interventions to promote biodiversity conservation and development.

1138 In this paper we interrogate and synthesize existing data sources, and supplement with

1139 additional research to document the historical and contemporary status of the African

1140 Elephant (Loxodonta africana), five large carnivores, one mesopredator and twenty four

1141 prey species throughout eight wildlife managed areas between the Greater Kafue System

1142 and the Zambezi River. This landscape is promoted as a key linkage to the central cluster of

1143 wildlife managed areas in Namibia and Botswana, at the heart of the KAZA TFCA (KAZA,

1144 2014).

1145 Through integration, harmonization and triangulation of data we were able to determine 1146 changes to species occurrence and distribution by wildlife managed area and designation.

1147

1148 2.2 Methods

1149 2.2.1 Study Area

1150 The KAZA TFCA’s boundaries are imprecise (Andersson, 2016). However, Cummings

1151 (2008) characterizes the TFCA as comprising a matrix of over 70 wildlife managed areas

1152 from strict National Parks under State control to multiple use areas under

1153 Customary/communal management. These wildlife managed areas fall into three major

1154 clusters and five periphery sub-clusters, with Kafue National Park and surrounding wildlife

1155 managed areas constituting the major northern cluster (Fig. 2.1).

1156

1157

1158 1159 Figure 2.1. The Kavango-Zambezi Transfrontier Conservation Area landscape, indicating 1160 study area, clusters of wildlife managed areas (WMAs) and their degrees of protection. 1161 Protected = National Parks IUCN II; Partially Protected = IUCN III-VI; Nominally 1162 Protected = IUCN Not Reported (adapted from PPF, 2011a).

1163

1164 At 22,480 km2 Kafue National Park is Zambia’s oldest and largest protected area, the

1165 largest National Park in the Kavango-Zambezi TFCA and 2nd largest National Park in

1166 Africa (UNEP/WCMC, 2016). In concert with nine surrounding IUCN category VI Game

1167 Management Areas and multiple Forest Reserves, the effective unfenced wildlife managed

1168 area, termed variously as the Greater Kafue Landscape or System, covers 68,000 km2 – a

1169 vast undeveloped area approximately half the size of England, and representing 9% of

1170 Zambia’s land mass and over 13% of the KAZA TFCA estate.

1171 Most of the Greater Kafue System lies between 900-1100 m above sea level. Rainfall

1172 averages 650 mm in the south and 1,050 mm in the north, falling predominantly from

1173 November to April. Vegetation is characterized by the Zambezian Miombo woodland

1174 Ecoregion, typical of large areas throughout southern and eastern Africa, dominated by

1175 Brachystegia spp., Combretum spp., Mopane spp., Terminalia spp. and Baikaea spp.

1176 Woodlands are interspersed by open floodplain grasslands and dambos (ZAWA, 2010).

1177 Species records include 158 mammals, 481 birds, 69 reptiles, 35 amphibians and 58 fish,

1178 with the greatest antelope diversity of any African National Park (21 species), an intact

1179 carnivore guild and a full complement of Zambia’s large mammals with the exception of

1180 Giraffe (Giraffa giraffa) and Black Rhinoceros (Diceros bicornis), known historically from

1181 the reports, and Tsessebe (Damaliscus lunatus) which has not appeared in any historical

1182 records (Moss, 2012).

1183 The Greater Kafue System has been included as Zambia’s majority component within

1184 KAZA TFCA (KAZA, 2014), with connectivity to the broader KAZA landscape contingent

1185 on the maintenance of a landscape level linkage routing south-southwest through a mosaic

1186 of nominally, potentially and possibly protected wildlife managed areas including Mulobezi

1187 and Sichifulo Game Management Areas, Nachitwe, Martin and Machili Forest Reserves,

1188 the Nyawa communal areas, and the recently proclaimed Simalaha Communal

1189 Conservancy (Fig. 2.2). In concert these wildlife managed areas extend the Greater Kafue

1190 System to around 73,000 km2.

1191

1192 1193 Figure 2.2 Wildlife Managed Areas within study area. 1194 CC=communal conservancy, FR=forest reserve, NP=National Park, GMA=Game 1195 Management Areas 1196 1197 A secondary (south-westerly) linkage passing through Mulobezi to Sioma National Park

1198 (bordering Namibia and Angola) has been proposed, though our focus remains the linkage

1199 broadly following the Machili stream catchment basin from the Kafue National Park border

1200 (S16.1380, E25.3650) to the northern bank of the Zambezi River (S17.5550, E24.9770),

1201 adjacent to Kasika and Salambala Communal Conservancies of East Zambezi Region in

1202 Namibia, and through to Chobe National Park in Botswana.

1203 This proposed landscape linkage varies in length from 140-170 km and contains a human

1204 population of around 110,000, growing at 2.5% pa, with a population density ≈4.0/km2

1205 (CSO, 2010). Communities are centered on a few larger settlements of 5,000-10,000

1206 residents, and otherwise in clusters of scattered villages typically concentrated along water

1207 courses, seasonal waterholes, and few pumped ground water supplies. Subsistence agro-

1208 pastoralists dominate this landscape, with residents largely dependent on exploiting a wide

1209 range of the area’s natural resources in support of basic livelihood needs (Musgrave, 2016).

1210 Formal employment opportunities beyond few distant urban settlements are negligible.

1211 Customary law within the Lozi, Nkoya, and Tonga ethnolinguistic groups represent the de

1212 facto regional governance system (Brelsford, 1965; Musgrave, 2016).

1213 Biodiversity conservation budgets have varied dramatically throughout this landscape, both

1214 spatially and temporally. While precise figures are unavailable, sources indicate that Kafue

1215 National Park (although operating with 10-15% of recommended protected area budgets)

1216 has received the greatest level of long-term biodiversity conservation support throughout

1217 the study area. This is followed by Mulobezi then Sichifulo Game Management Areas

1218 which receive minor budget allocations from the State Wildlife Authority, augmented by

1219 finance and in-kind operational support from resident safari hunting operators and

1220 conservation NGOs. Nachitwe, Martin and Machili Forest Reserves have intermittently

1221 received minor budgets from the State Wildlife Authority and Forestry Department

1222 (ZAWA, 2010; Chifunte, pers comms). The recently proclaimed Simalaha Communal

1223 Conservancy only started receiving any formal wildlife resource protection as recently as

1224 2013 following no formal biodiversity conservation budgets since pre-1978 (Inyambo-Yeta,

1225 pers comms). We were unable to ascertain if the Nyawa Communal areas receives any

1226 formal wildlife management budget. In additional a 240 km2 fenced Wildlife Recovery

1227 Sanctuary at the south of Simalaha, with an extensive open border against the Zambezi

1228 River, has received >600 head of game from eight species since 2013, representing a

1229 significant investment promoted as a justification for restocking the broader Simalaha

1230 Communal Conservancy (PPF, 2015). No formal monitoring of these species appears to

1231 have been undertaken since reintroductions.

1232 2.2.2 Data Sources

1233 The earliest records of terrestrial mammal occurrence and distribution in the vicinity of the

1234 proposed Kafue-Zambezi linkage are limited to disparate notes and reports in the grey

1235 literature from early explorers, hunters, traders and missionaries dating back to the late 19th

1236 century (e.g. Holub, 1975; Sampson, 1972), with approximate location data variously

1237 reported in relation to key landscape features. The first published checklists for Zambia

1238 (Pitman, 1934; Lancaster, 1953; Ansell, 1957/59/60) indicate no changes to the large

1239 mammal assemblage in and around Kafue National Park prior to the notable Black

1240 Rhinoceros extirpation in the mid-1980’s, though unresolved questions surround anecdotal

1241 records of a relic Giraffe population (Moss & Fennessy, pers comms). Data for these

1242 checklists were ostensibly collected through ad hoc and opportunistic sightings from

1243 Government staff and ‘expert’ observers reporting from their travels throughout the

1244 country, augmented by trading records and hunting ledgers kept by District Commissioners.

1245 The first systematic collation of species occurrence and distribution data was published by

1246 Ansell (1978), superseding previous literature. Amalgamated checklist data were mapped

1247 within ¼ degree grid squares, based on 1:50,000 Ordnance Survey map sheets. While data

1248 reflects minimum regional species range given the absence of reports from many

1249 inaccessible and largely unmapped periphery areas, much of this study area can be

1250 considered well mapped due to the established network of access routes developed

1251 alongside the nascent Teak logging and safari hunting industries (Musgrave, 2016).

1252 While Ansell (1978) reports on 38 terrestrial mammals >10kg from 11 taxonomic families

1253 we restricted the contemporary list to 31 readily detected species from nine taxonomic

1254 families, omitting seven species considered either at the edge of known range and/or habitat

1255 specialists requiring species-specific survey techniques beyond the scope of this study.

1256 Boundaries of contemporary land use classifications (UNEP-WCMC, 2016) were projected

1257 over Ansell’s (1978) maps using QGIS (QGIS, 2017) (Fig 2.3) to allow for extraction of

1258 historical species distribution data at comparable spatial scales: Kafue National Park

1259 (Kaunga and Nanzilla management blocks at 5,700 km2), Mulobezi Game Management

1260 Area (hereafter Mulobezi, at 3,420 km2), Sichifulo Game Management Area including

1261 Nachitwe, Martin and Machili Forest Reserves (hereafter Sichifulo, at 4,090 km2), and

1262 finally the Nyawa/Simalaha areas (2,800 km2).

1263

1264

1265

1266 Figure 2.3. Historical Species Distribution from Ansell (1978) showing species known 1267 range (solid squares), possible range (hatched squares) and former range (unfilled squares), 1268 mapped here for Blue Wildebeest (Connochaetes taurinus). Boundary of contemporary 1269 wildlife managed areas in yellow, study area in red.

1270

1271 In compiling contemporary data sets (Fig 2.4) we constrained data gathering to three

1272 broadly comparable ground-based survey approaches. We omitted aerial survey data (e.g.

1273 DNPW, 2016) given limitations to detection rates for many species of primary interest in

1274 forested areas (Jachmann, 2002).

1275 Firstly, the resident safari hunting operator, operational throughout Mulobezi and Sichifulo

1276 Game Management Areas during the preceding decade, was asked to provide sightings

1277 reports for 31 terrestrial mammals of interest through a questionnaire survey following the

1278 2014 hunting season. Cumulatively, multiple groups of guides, hunters and skilled local

1279 trackers traversed both Mulobezi and Sichifulo Game Management Areas on and off road,

1280 by open vehicle and on foot, covering >10,000 km/dry season (Kraljik, pers comms). This

1281 was considered, in my expert opinion, to be sufficient survey effort and local expertise to

1282 provide a high probability of detecting target species.

1283 Secondly, we collected patrol data from the local State and Community Wildlife Police

1284 Officers responsible for wildlife protection in southern Kafue National Park, Mulobezi and

1285 Sichifulo. We amalgamated data for the Kafue National Park patrol blocks adjacent to

1286 Mulobezi and Sichifulo Game Management Areas to provide a single area covering the

1287 border north of both Mulobezi and Sichifulo Game Management Areas. These data

1288 provided 1,920 georeferenced wildlife sighting reports during 2014/5 from 46,170 man-

1289 days of foot patrols (ZAWA, unpublished data).

1290 Finally, in 2015, we undertook a systematic randomized spoor and sightings survey of large

1291 carnivores and their principle prey throughout 10 x 400 km2 survey blocks in Mulobezi and

1292 Sichifulo Game Management Areas and the Nyawa/Simalaha areas (see Appendix 1).

1293 Detection probability and survey effort were optimized for large carnivores following

1294 Funston et al. (2010) and Thorn et al. (2010). In addition, a site-specific calibration process

1295 was undertaken from July to September 2014, conducted at varying spatiotemporal scales,

1296 to establish survey effort required to detect large carnivores and sample the landscape in a

1297 single season (MacKenzie & Royle, 2005, MacKenzie, pers comms). In total 102 x 4 km

1298 transects were walked three times by the principle investigator and two experienced local

1299 trackers from the safari hunting industry, cumulatively providing 1,224 km of spoor

1300 transects over six months fieldwork during the dry season from May and Oct 2015 (see

1301 Appendix 1 for further data and explanations on transect selection and criteria).

1302

1303 Figure 2.4: Contemporary Data Sources

1304 2.2.3 Data Analysis

1305 A confirmed sighting from any of the three expert contemporary sources was marked as a

1306 positive detection at the scale of interest. Given the atypical nature of ongoing ungulate

1307 reintroductions and management in the fenced Simalaha Wildlife Sanctuary, we restrict

1308 reporting to the detection of the carnivore guild for this subset of the Simalaha Communal

1309 Conservancy.

1310 Data for each of the four composite wildlife management area blocks and three data sources

1311 were compiled against historical data to determine if any changes in species occurrence and

1312 distribution had been detected throughout the intervening years (1975-2014). Outputs 53

1313 reflected species persistence, loss or colonization at the composite wildlife management

1314 area scale.

1315 Given survey methods were optimized for resident large carnivores and their principle prey

1316 species, elevated non-detection risks existed where species exhibited significant seasonal

1317 movement patterns (migration), non-resident movement patterns (emigration and

1318 immigration), or where surveys did not cover the restricted ranges of habitat specialists.

1319 Table 2.1 and subsequent analyses acknowledges these constraints.

1320 Finally, an amalgamated distribution map was generated for the five extant large

1321 carnivores, indicating historical range within the survey area, and current known range

1322 within studied wildlife managed areas.

1323

1324 2.3 Results

1325 2.3.1 Changes to Species Occurrence and Distribution

1326 Table 2.1 indicates few non-detections recorded against any data sources since 1978

1327 throughout southern Kafue National Park, Mulobezi or Sichifulo Game Management Areas.

1328 Notably Hippopotamus (Hippopotamus amphibius) appear no longer resident in any of the

1329 waterways along the Machili stream and catchment area. Klipspringer (Oreotragus

1330 oreotragus) appear absent from Mulobezi, though core habitat for this species went

1331 unsurveyed. Steenbok (Raphicerus campestris) are considered at the extent of their

1332 northeast range approaching Kafue National Park, with a single sighting recorded in

1333 Mulobezi.

1334 1335 Table 2.1. Summary results of species detection by source and area, with distribution 1336 change, 1978-2014/5. 1337

1338 The absence of confirmed Caracal (Caracal caracal) and Serval (Leptailurus serval)

1339 sightings by Wildlife Police Office patrols in southern Kafue National Park appears an

1340 anomaly given detection from adjacent Game Management Areas. While I hypothesize

1341 these anomalies represents non-detection error versus absence, I nonetheless omitted these

1342 species from the final check list.

1343

Species Change by Wildlife Managed Area 1978-2015 30

10 Species Species Diversity

0 1978 2015 1978 2015 1978 2015 1978 2015 KNP/south Mulobezi Sichifulo Simalaha

Carnivores Prey Species 1344 1345 Figure 2.5. Changes to carnivore and herbivore composition by area, 1978-2014/15.

1346

1347 Major losses have occurred in the newly registered Simalaha Communal Conservancy,

1348 whereby 21/31 terrestrial mammals went undetected (Fig 2.5). Side-striped Jackal (Canis

1349 adustus) remained the only widespread carnivore detected in Simalaha. Both Spotted

1350 hyaena (Crocuta crocuta) and Leopard (Panthera pardus) were the only large carnivores

1351 detected within 60km of the Zambezi River in the Nyawa Communal area (Fig 2.6). The

1352 remaining large carnivore guild appears extirpated from the Simalaha/Nyawa area along

1353 with all ungulates >20kg, excluding the Southern Reedbuck (Redunca arundinum) and

1354 Greater Kudu (Tragelaphus strepsiceros). Kudu were also the only herding ungulate to be

1355 detected in Simalaha, through no aggregations over three animals were detected. Notably

1356 both Warthog (Phacochoerus africanus) and Bush pig (Potamochoerus larvatus), habitat

1357 and feeding generalists with high reproductive rates, went undetected in Simalaha. While

1358 >600 head of game comprising seven species have been introduced into the 240 km2

1359 Simalaha Wildlife Recovery Sanctuary since 2013, only Side-Striped jackal were detected

1360 inside the (non-predator proof) area. There was no evidence of any species range extension

1361 or recolonization throughout any of the sampled areas.

1362 1363 Figure 2.6. Distribution of large carnivores at Kafue-Zambezi Interface, 2014/5.

1364

1365

1366

1367 Although no long term, comparable, or landscape-level survey programme is in place to

1368 systematically monitor changes in species occurrence, distribution or abundance, much

1369 existing expertise and anecdotal evidence implies large scale population declines

1370 throughout the Greater Kafue System and beyond since 1978 (Chifunte, Daka, Hanks,

1371 Moomba, Moss & Inyambo-Yeta, pers comms). Contemporary data indicates Kafue

1372 National Park, the regions’ prime wildlife area, is maintaining the majority of terrestrial

1373 mammals significantly below carrying capacity (Simukonda, 2008). Nonetheless, with few

1374 historical survey data available for direct comparison, we restricted our analyses to species

1375 diversity at the scale of interest, versus any interpretation of spatiotemporal changes to

1376 community structure and abundance, which is beyond the scope of this paper.

1377

1378 2.4 Discussion

1379 Formal historical records explaining species loss in Simalaha and Nyawa areas are

1380 unavailable, though local Traditional Authorities (Chiefs Inyambo-Yeta, Moomba, pers

1381 comms) emphasized the impact of the Angolan Bush War (1966-1989) as a key driver,

1382 describing the activities of foreign combatant encampments in Simalaha being used as a

1383 base to exploit the areas’ wildlife for rations and profit. Following cessation of hostilities

1384 much small arms proliferation occurred, and in conjunction with expanding human

1385 population and limited funding for law enforcement and natural resource management,

1386 ongoing unsustainable harvesting of wildlife continued. Given these circumstances the

1387 authors hypothesize that wildlife managed areas closer to Kafue National Park were spared

1388 much of these pressures, having also received elevated political and revenue support for

1389 wildlife management in the long term (Daka, pers comms).

1390 Existing surveys at the Kafue-Zambezi interface have employed a range of ad hoc

1391 methodological approaches that failed to detect the majority of resident species throughout

1392 this landscape. The absence of a reliable baseline undermines efforts at evaluating the

1393 effectiveness of large-scale conservation interventions required to deliver key programme

1394 objectives within and between clusters of wildlife management areas.

1395 Acknowledging non-detection error, we confirm that the terrestrial mammal (>10 kg)

1396 richness in southern Kafue National Park remains unchanged since 1978. Mulobezi and

1397 Sichifulo retain largely intact mammalian diversity, with the notable exception of resident

1398 Hippopotamus. No new data could be provided for the existence of free-ranging Giraffe in

1399 any of these wildlife managed areas.

1400 While a single season survey design increases non-detection error associated with species

1401 dispersal or seasonal wildlife movement patterns, widespread losses, including three of six

1402 carnivore species and 16 of 25 prey species, were detected in the Simalaha Communal

1403 Conservancy / Nyawa areas, collectively key linking wildlife managed areas at the interface

1404 of the Greater Kafue System and adjacent wildlife managed areas in Namibia and

1405 Botswana.

1406 These data emphasize the challenges surrounding scope and scale of conservation

1407 interventions required to limit factors driving species loss from seven of nine taxonomic

1408 families, representing a wide range of species traits. Significantly, if drivers of species loss

1409 continue to limit population recovery in Simalaha/Nyawa areas then source-sink dynamics

1410 and edge effects can negatively impact population viability of vulnerable species in

1411 periphery wildlife managed areas at local and transboundary scales.

1412 Wide-ranging species are particularly susceptible to source-sink dynamics and edge effects,

1413 so the absence of large carnivores from the Simalaha and the Simalaha Wildlife Recovery

1414 Sanctuary indicates the need for additional research to understand the status and drivers of

1415 wildlife occurrence and distribution south of the Zambezi River throughout the wildlife

1416 managed areas of eastern Zambezi Region in Namibia, and the effects that ecological

1417 traps/attractive sinks might pose at transboundary scales on wildlife management

1418 interventions in Simalaha and other neighboring wildlife managed areas of Zambia.

1419 Broader scale implications of species loss and ecological traps within the Kavango-

1420 Zambezi TFCA relate to dominant narratives surrounding wildlife managed area

1421 connectivity. The extent to which existing and emerging drivers of species loss are severing

1422 biological linkages between the Greater Kafue System and adjacent wildlife managed areas

1423 in the Kavango-Zambezi TFCA remain unquantified and subject to speculation. However,

1424 data suggests a connectivity bottleneck at the large mammal level in the Simalaha

1425 Communal Conservancy, with only 10 of 31 species known from historical records

1426 detected throughout this area in 2014/5.

1427 While the long distance dispersal capabilities of large carnivores implies scope for gene

1428 flow between the Greater Kafue System and adjacent wildlife managed areas in the

1429 Kavango-Zambezi TFCA, the extent to which connectivity bottlenecks impact processes of

1430 immigration and emigration in highly mobile species is an important area of priority

1431 research for regional connectivity conservation management.

1432

1433

1434 2.5 Conclusions

1435 The study focused on ascertaining changes to the occurrence and distribution of 38

1436 terrestrial mammals >10 kg known from four composite wildlife managed areas between

1437 the Greater Kafue System and central cluster of wildlife managed areas in the KAZA

1438 TFCA, and the methodological approach was successful for 31 species at the scale of

1439 interest.

1440 While these data cannot elaborate on population numbers and trends, it is apparent that

1441 ongoing attempts to maintain population viability of vulnerable species, wildlife

1442 connectivity between clusters of wildlife managed areas, and the promotion of wildlife-

1443 based land uses, will depend on diagnosing and treating the interacting ecological, socio-

1444 economic and political drivers of species loss within and between clusters of wildlife

1445 managed areas utilizing comparative studies at appropriate temporal and spatial scales.

1446 The limits to which sufficient political and economic capital can be leveraged to bridge

1447 these knowledge gaps, act accordingly on the findings, and be subject to monitoring,

1448 evaluation and feedback, will likely determine future connectivity for Zambia’s majority

1449 component within the KAZA TFCA.

1450

1451

1452

1453

1454

1455 2.6 References

1456 Andersson, J. (Ed.). (2017). Transfrontier conservation areas: People living on the edge.

1457 Taylor & Francis.

1458 Ansell, W.F.H (1957). Some mammals from Northern Rhodesia. Journal of Natural

1459 History, 10(115), 529-551.

1460 Ansell, W.F.H (1959). Further data on Northern Rhodesian ungulates. Mammalia, 23(3),

1461 332-349.

1462 Ansell, W.F.H (1960). Mammals of Northern Rhodesia. The Government Printer, Lusaka.

1463 Ansell, W.F.H (1978). The Mammals of Zambia. National Parks & Wildlife Service. The

1464 Government Printer, Lusaka.

1465 Battin, J (2004). When good animals love bad habitats: ecological traps and the

1466 conservation of animal populations. Conservation Biology, 18(6), 1482-1491.

1467 Bellard, C., et al., (2012). Impacts of climate change on the future of biodiversity. Ecology

1468 Letters, 15: 365–377.

1469 Biggs, R., et al., (2008). Scenarios of biodiversity loss in southern Africa in the 21st

1470 century. Global Environmental Change, 18(2), 296-309.

1471 Brelsford, W. V. (1965). The Tribes of Zambia. The Government Printer, Lusaka.

1472 Cardillo, et al., (2008). The predictability of extinction: biological and external correlates of

1473 decline in mammals. Proceedings of the Royal Society B: Biological Sciences.

1474 275(1641), 1441-1448.

1475 CSO (2010). Central Statistics Office of Zambia. Available at:

1476 http://www.zamstats.gov.zm/. Accessed 23/06/2014.

1477 Cumming, D. H. (2008). Large scale conservation planning and priorities for the Kavango-

1478 Zambezi Transfrontier Conservation Area. Unpublished Report Commissioned by

1479 Conservation International.

1480 Cumming, G. S. (2011). Spatial resilience in social-ecological systems. Springer Science &

1481 Business Media.

1482 DNPW (2016). Report on the 2015 aerial census of elephants and other large mammals in

1483 Zambia: Volume II, Population estimates for other large mammals and birds.

1484 Department of National Parks and Wildlife, Chilanga, Zambia.

1485 Fischer, J., et al., (2006). Biodiversity, ecosystem function, and resilience: ten guiding

1486 principles for commodity production landscapes. Frontiers in Ecology and the

1487 Environment, 4(2), 80-86.

1488 Funston, P.J., et al., (2010). Substrate and species constraints on the use of track incidences

1489 to estimate African large carnivore abundance. Journal of Zoology, 281(1), 56-65.

1490 Hanks, J. (2000). The role of Transfrontier Conservation Areas (TFCAs) in Southern

1491 Africa in the conservation of mammalian biodiversity. In: pp. 239–256. A. Entwistle

1492 and N. Dunstone (Eds.). Priorities for the Conservation of Mammalian Diversity.

1493 Has the Panda had its day? Conservation Biology 3. Cambridge: Cambridge

1494 University Press.

1495 Hanks, J. (2003). Transfrontier Conservation Areas (TFCAs) in Southern Africa: their role

1496 in conserving biodiversity, socioeconomic development and promoting a culture of

1497 peace. Journal of Sustainable Forestry, 17(1-2), 127-148.

1498 Hanks, J. & Myburgh, W. (2015). The Evolution and Progression of Transfrontier

1499 Conservation Areas in the Southern African Development Community. In:

1500 Institutional Arrangements for Conservation, Development and Tourism in Eastern

1501 and Southern Africa. Springer, 157-179.

1502 Holub, E. (1975). Emil Holub's Travels North of the Zambezi, 1885-6. Manchester

1503 University Press.

1504 Jachmann, H. (2002). Comparison of aerial counts with ground counts for large African

1505 herbivores. Journal of Applied Ecology, 39(5), 841-852.

1506 Kavango-Zambezi TFCA (2011a). Treaty between the Governments of Angola, Botswana,

1507 Namibia, Zambia and Zimbabwe on the establishment of the Kavango-Zambezi

1508 Transfrontier Conservation Area. Kavango-Zambezi TFCA Secretariat. Kasane,

1509 Botswana.

1510 Kavango-Zambezi TFCA (2011b). Strategic Action Plan. Kavango-Zambezi TFCA

1511 Secretariat. http://www.kavangozambezi.org/strategic-action-plan. Accessed

1512 01/06/2015.

1513 KAZA (2014). Master Integrated Development Plan. Kavango-Zambezi TFCA Secretariat.

1514 Victoria Falls.

1515 Lancaster, D. G. (1953). A Check List of the Mammals of Northern Rhodesia. The

1516 Government Printer, Lusaka. 64

1517 Linde, H., et al., (2002). Beyond Boundaries: Transboundary Natural Resource

1518 Management in Sub-Saharan Africa. Washington, D.C., U.S.A.: Biodiversity

1519 Support Program.

1520 Lindsey, P., et al., (2013). Illegal hunting and the bush-meat trade in savanna Africa:

1521 drivers, impacts and solutions to address the problem. Panthera/Zoological Society

1522 of London/Wildlife Conservation Society report, New York.

1523 Lindsey, P. A., et al., (2013). Benefits of wildlife-based land uses on private lands in

1524 Namibia and limitations affecting their development. Oryx, 47(1), 41-53.

1525 MacKenzie, D. I. & Royle, J. A. (2005). Designing occupancy studies: general advice and

1526 allocating survey effort. Journal of Applied Ecology, 42(6), 1105-1114.

1527 Margules, C. R. & Pressey, R. L. (2000). Systematic conservation planning. Nature,

1528 405(6783), 243-253.

1529 Maxwell, S.L., et al., (2016). Biodiversity: The ravages of guns, nets and bulldozers.

1530 Nature, 536, 143-145.

1531 Moss, P. (2012). Handbook to Kafue National Park. Future Publishing (PTY) Ltd. South

1532 Africa.

1533 Musgrave, M. (2016). Scale, Governance and Change in Zambezi Teak Forests:

1534 Sustainable Development for Commodity and Community. Cambridge Scholars

1535 Publishing.

1536 Newmark, W. D. (2008). Isolation of African protected areas. Frontiers in Ecology and the

1537 Environment, 6(6), 321-328.

1538 Ogutu, J.O., et al., (2016). Extreme wildlife declines and concurrent increase in livestock

1539 numbers in Kenya: What are the causes? PloS one, 11(9), p.e0163249.

1540 Pachauri, R.K., et al., (2014). Climate Change 2014: Synthesis Report. Contribution of

1541 Working Groups I, II and III to the Fifth Assessment Report of the

1542 Intergovernmental Panel on Climate Change. IPCC, Geneva, Switzerland, 151.

1543 Pitman, C. (1934). A report on a faunal survey of Northern Rhodesia. Government Printer,

1544 Livingstone.

1545 PPF (2008). Integrated Development Plan for the Zambian Component of the Kavango-

1546 Zambezi Transfrontier Conservation Area. Peace Parks Foundation, Stellenbosch,

1547 South Africa.

1548 PPF (2015). Restocking Simalaha Communal Conservancy. Peace Parks Foundation

1549 http://www.peaceparks.org/news.php?pid=1481&mid=1546&lid=1120. Accessed

1550 17/07/2017.

1551 QGIS Development Team (2016). Quantum GIS Geographic Information System. Open

1552 Source Geospatial Foundation Project. http://qgis.osgeo.org

1553 Sampson, R. (1972). The Man with a Toothbrush in His Hat: The Story and Times of

1554 George Copp Westbeech in Central Africa. Multimedia Publications.

1555 Simukonda, C. (2008). A Countrywide Survey of Large Mammals, Zambia. Zambia

1556 Wildlife Authority, Lusaka, Zambia.

1557 Sutherland, W.J., et al., (2004). The need for evidence-based conservation. Trends in

1558 Ecology & Evolution, 19(6), 305-308.

1559 Thorn, M., et al., (2010). Comparative efficacy of sign surveys, spotlighting and audio

1560 playbacks in a landscape-scale carnivore survey. South African Journal of Wildlife

1561 Research, 40(1), 77-86.

1562 UNEP-WCMC (2016). United Nations Environment Programme's World Conservation

1563 Monitoring Centre. Protected Planet Database www.Protectedplanet.Net. Accessed

1564 01/05/2015.

1565 United Nations Department of Economic and Social Affairs and Population Division

1566 (2015). World Population Prospects: The 2015 Revision, Key Findings and

1567 Advance Tables. Working Paper, no. ESA/P/WP.241.

1568 Watson, F.G.R., et al., (2015). Human encroachment into protected area networks in

1569 Zambia: implications for large carnivore conservation. Regional Environmental

1570 Change, 15(2), 415-429.

1571 WHO/MEE (2005). Ecosystems and Human Well-being. World Health Organisation /

1572 Millennium Ecosystem Assessment, 2000. Washington, DC: Island Press.

1573 Woodroffe, R. & Ginsberg J. R. (1998). Edge effects and the extinction of populations

1574 inside protected areas. Science (New York, N.Y.), 280(5372), 2126-2128.

1575 WWF (2011). World Wildlife Fund. A Brighter Future for Communities and Wildlife.

1576 http://www.worldwildlife.org/stories/a-brighter-future-for-communities-and-

1577 wildlife. Accessed 15/02/2016.

1578 ZAWA (2010). General Management Plan for Kafue National Park. Chilanga. Zambia.

1579 Zbicz, D. (1999). Transboundary Cooperation between Internationally Adjoining Protected

1580 Areas. In pp. 199-204, On the Frontiers of Conservation: Proceedings of the 10th

1581 Conference on Research and Resource Management in Parks and on Public Lands.

1582 The George Wright Society, Hancock, Michigan.

1583

1584

1585

1586

1587

1588

1589

1590

1591

1592

1593

1594

1595

1596

1597 Chapter 3: Modelling multi-species connectivity at the Kafue-Zambezi

1598 Interface: Implications for Transboundary Carnivore Conservation

1599 Lines, R.M.,1* Bormpoudakis, D.1., Xofis, P.2 & Tzanopoulos, J.1,3

1600 1Durrell Institute of Conservation and Ecology, School of Anthropology and

1601 Conservation, University of Kent, UK

1602 2 International Hellenic University, Department of Forestry and Natural Environment,

1603 Drama, Greece

1604 3Kent Interdisciplinary Centre for Spatial Studies, School of Anthropology and

1605 Conservation, University of Kent, UK

1606 *Corresponding author: [email protected]

1607

1608 Abstract

1609 Linking wildlife areas with corridors facilitating species dispersal between core habitats is a

1610 key intervention to reduce deleterious effects of population isolation. Large heterogeneous

1611 networks of areas managed for wildlife protection present site- and species-scale

1612 complexity underpinning the scope and performance of proposed corridors. In Southern

1613 Africa, the Kavango-Zambezi Transfrontier Conservation Area seeks to link Kafue

1614 National Park to a cluster of wildlife area centered on Namibia and Botswana. To assess

1615 and identify potential linkages on the Zambian side we generated a high-resolution land

1616 cover map and combined empirical occurrence data for lion (Panthera leo), leopard

1617 (Panthera pardus) and Spotted hyena (Crocuta crocuta) to build habitat suitability maps.

1618 We then developed four connectivity models to map potential single and multi-species

1619 corridors between Kafue and the Zambezi River border with Namibia. Single and multi-

1620 species connectivity models selected corridors follow broadly similar pathways narrowing

1621 significantly in central-southern areas of the Kafue-Zambezi interface, indicating a

1622 potential connectivity bottleneck.

1623 Capturing the full extent of human disturbance and barriers to connectivity remains

1624 challenging, suggesting increased risk to corridor integrity than modelled here.

1625 Notwithstanding model limitations, these data provide important results for land use

1626 planners at the Kafue-Zambezi Interface, removing much speculations from existing

1627 connectivity narratives. Failure to control human disturbance and secure corridors will

1628 leave Kafue National Park, Zambia’s majority component in the Kavango-Zambezi

1629 Transfrontier Conservation Area, isolated.

1630

1631 Keywords: Connectivity; Carnivores; KAZA; Transfrontier Conservation Area; Kafue;

1632 Zambia; Sentinel; Landcover; MaxEnt; Linkage Mapper.

1633 Word Count: 6680

1634 Citation: Lines, R.M., Bormpoudakis, D., Xofis, P. & Tzanopoulos, J. (2020). Modelling

1635 multi-species connectivity at the Kafue-Zambezi Interface: Implications for

1636 Transboundary Carnivore Conservation. Oryx in review

1637

1638

1639

1640 3.1 Introduction

1641 Expanding human pressures on natural resources are driving rapid fragmentation and

1642 conversion of wildlife habitat to farmlands and rangelands, isolating wildlife populations,

1643 and driving widespread declines in wildlife (Maxwell et al., 2016; Marco et al., 2014). The

1644 large carnivore guild is notable in its sensitivity to anthropogenic pressures and exhibits

1645 elevated extinction risk with population isolation (Purvis et al., 2000; Cardillo el al., 2005).

1646 Ripple et al. (2014) calculated some 77% of terrestrial large carnivore species are now

1647 experiencing declines in population size and range due to anthropogenic pressures related

1648 to habitat fragmentation and loss. Maintaining free-ranging large carnivore populations is

1649 contingent on provision of sufficient habitat and prey species and limiting deleterious

1650 effects of human disturbances (Wolf & Ripple, 2017).

1651 Given the importance of maintaining structural and functional connectivity between

1652 wildlife managed areas, large scale conservation networks, including Transfrontier

1653 Conservation Areas, are being developed to address issues surrounding fragmentation and

1654 isolation of wildlife populations (Farhig, 2003; Cushman et al., 2013). These initiatives

1655 have important implications for many large carnivore populations (Crooks et al., 2011), the

1656 ecological functionality of the landscapes within which they reside (Ripple et al., 2014),

1657 and the development of wildlife-based land uses and economies (Funston et al., 2013).

1658 Understanding and promoting species-level connectivity between wildlife managed areas is

1659 a key objective of the Kavango-Zambezi Transfrontier Conservation Area (hereafter KAZA

1660 TFCA) Programme that spans five neighboring countries and ~70 wildlife managed areas

1661 across ~520,00 km2 of central southern Africa (KAZA, 2011a). The KAZA TFCA

1662 represents the largest and arguably most important and ambitious conservation initiative in

1663 Africa, covering a wide range of threatened species, habitats and ecosystem services that

1664 provide critical resources and natural wealth to the region’s people. But much success with

1665 key objectives in this programme is dependent on maintaining and expanding landscape

1666 connectivity.

1667 Yet few of the proposed wildlife managed area linkages in the KAZA TFCA Programme

1668 have been sufficiently studied at the species and scale of interest to provide planners and

1669 managers with the empirical evidence necessary for the informed decision making. This is

1670 critical to effective natural resource management and the development of wildlife-based

1671 land uses (Cumming, 2008; KAZA, 2014).

1672 Connectivity between Kafue National Park in central western Zambia and the core cluster

1673 of wildlife managed areas centered on Chobe National Park in Botswana represents one of

1674 the largest questions (and challenges) surrounding natural resource management in the

1675 KAZA TFCA Programme. If connectivity cannot be established or developed, Zambia’s

1676 majority component in the programme will effectively be isolated from the broader

1677 landscape, with significant implications for low density, wide-ranging species in Kafue

1678 National Park, and the potential and promise of the KAZA TFCA Programme as a whole.

1679 But promoting connectivity and species-level movement at large spatial scales in

1680 heterogenous landscapes is dependent on high quality data of system states and species

1681 level response to key drivers (Worboys et al., 2010). The main aim of this study is to assess

1682 landscape connectivity at the large carnivore scale at the interface of Kafue National Park

1683 and adjacent wildlife managed areas, through the Kafue-Simalaha Wildlife Dispersal

1684 Corridor and surrounds. Specifically, the objectives of this study are: a) to investigate and

1685 model the effect of environmental and anthropogenic factors on the occurrence three large

1686 carnivores between Kafue National Park and the Simalaha Wildlife Recovery Sanctuary;

1687 and b) to assess the potential of the proposed Kafue-Simalaha Wildlife Dispersal Corridor

1688 for single and multi-species landscape-level connectivity planning.

1689 In order to achieve these objectives, we generated high resolution landcover imagery for

1690 3.8m ha using the latest high resolution satellite imagery combined with ground truth data,

1691 then incorporated empirical occurrence data for lion (Panthera leo), leopard (Panthera

1692 pardus) and Spotted hyena (Crocuta crocuta) throughout ten wildlife managed area

1693 covering ~10,000 km2 to produce species-level habitat suitability maps. Finally, we

1694 constructed connectivity models to investigate potential for single and multi-species

1695 corridors in support of conservation and land use planning for this vast, heterogenous and

1696 increasingly human-impacted landscape.

1697 Our final output has broad applied value for corridor planning and the development of

1698 wildlife-based land uses throughout the Kavango-Zambezi Transfrontier Conservation Area

1699 and beyond, supporting the knowledge base for long term persistence of large carnivores,

1700 the ecological integrity of systems in which they reside and the promotion of sustainable

1701 wildlife-based land uses for the landscapes’ rural poor.

1702 3.2 Methods

1703 3.2.1 Study Area

1704 The KAZA TFCA encompasses a matrix of over 70 wildlife managed areas in IUCN I-VI

1705 and Not Reported categories (UNEP-WCMC, 2015), spanning the borders of Angola,

1706 Botswana, Namibia, Zambia and Zimbabwe in central southern Africa, extending over

1707 c.520,000 km2 (KAZA, 2014). Spatially, these wildlife managed areas fall into three major

1708 clusters and five periphery clusters, with Kafue National Park and surrounding wildlife

1709 managed areas representing the major northern cluster (Fig. 3.1).

1710

1711 Figure 3.1. The Kavango-Zambezi Transfrontier Conservation Area landscape, indicating 1712 study area, clusters of wildlife managed areas (WMAs) and their degrees of protection. 1713 Protected = National Parks IUCN II; Partially Protected = IUCN III-VI; Nominally 1714 Protected = IUCN Not Reported (adapted from PPF, 2011a).

1715

1716 Kafue National Park is Zambia’s oldest and largest protected area, the largest National Park

1717 in the KAZA TFCA and the 2nd largest National Park in Africa at 22,480 km2

1718 (UNEP/WCMC, 2015). In concert with nine surrounding IUCN category VI Game

1719 Management Areas, the effective unfenced area known as the Greater Kafue System covers

1720 c.68,000 km2 – a vast undeveloped area approximately half the size of England,

1721 representing 9% of Zambia’s land area (ZAWA, 2010).

1722 The Greater Kafue System has been included as Zambia’s majority component in the

1723 KAZA TFCA programme, encompassing c.25% of the Programme’s wildlife estate

1724 (KAZA, 2014). Terrestrial connectivity to adjacent clusters of wildlife managed areas

1725 comprises a minor potential linkage routing west-southwest to Sioma National Park,

1726 spanning ≥210 km, and a major potential linkage routing south-southwest towards Chobe

1727 National Park, at the heart of the KAZA TFCA – the focus of this study. This linkage,

1728 termed the Kafue-Simalaha Wildlife Dispersal Corridor (KAZA, 2014), passes through a

1729 mosaic of nominally, potentially and possibly protected wildlife managed areas including

1730 Mulobezi and Sichifulo Game Management Areas, Nachitwe, Martin and Machili Forest

1731 Reserves, the Nyawa communal areas, and Simalaha Communal Conservancy (Fig. 3.2). In

1732 concert these wildlife managed areas extend the Greater Kafue System to around 7.3 m ha

1733 and provide a contiguous matrix of wildlife managed areas spanning 140-170 km from the

1734 Kafue National Park border south-southwest towards the Zambezi River at the confluence

1735 of Zambia, Namibia and Botswana, broadly following the Machili watershed (Lines et al.,

1736 2018). Peripheral areas, considered outside of any formal wildlife protection category, are

1737 known as ‘Open’ Communal Areas, and fall under statutory land tenure authority of the

1738 local Chiefdom (Machina, 2005).

1739

1740 Figure 3.2 Wildlife managed areas within study area, indicating proposed Simalaha-Kafue 1741 Wildlife Dispersal Corridor.

1742

1743 3.2.2 Landcover Mapping

1744 A landcover map of the study area was produced using a time-series of Sentinel 2 images in

1745 an object-oriented image analysis (OBIA) environment using the software eCognition

1746 (eCognition Developer, T, 2014). Sentinel 2 (Copernicus Sentinel data, 2018) is a

1747 constellation of two polar-orbiting satellites launched by ESA under the program

1748 Copernicus (formerly known as GMES; Drusch et al., 2012). Due to the large size of the

1749 study area (32,812 km2), and the fact that it fell within two paths of the Sentinel 2

1750 (Copernicus Sentinel data, 2018) orbit, the study area was split in two parts and analyzed

1751 separately.

1752 The eastern part was mapped using a time series of three images acquired in January, April

1753 and August 2016 while the western part was mapped using a time series of three images

1754 acquired in May, September and December 2016. The images were geometrically and

1755 atmospherically corrected prior to segmentation and classification. During the classification

1756 process various spectral ratios were calculated including Normalized Difference Vegetation

1757 Index (NDVI;), Normalized Difference Wetness Index (NDWI), Normalized Difference

1758 Moisture Index (NDMI), Near InfraRed/Red (NIR).

1759 The first step in any image analysis using the OBIA approach (eCognition Developer, T,

1760 2014) is the creation of meaningful image objects by using segmentation algorithms that

1761 group neighboring pixels according to certain criteria of homogeneity. The size and shape of

1762 the objects depend primarily on the selected scale parameter (which determines the degree

1763 of homogeneity in the resulted objects) as well as by the shape and compactness criterion.

1764 Various scale parameters were tested, with the aim of generating objects that have low

1765 internal variation and low spatial autocorrelation, and the results were visually inspected. The

1766 adopted scale parameter was set at 100 while the colour and homogeneity criterion were set

1767 at 0.3 and 0.5, respectively. Minimum mapping unit was set at 0.25 ha.

1768 The training data set for the semi-automatic image classification was provided by in-situ

1769 observations. Six classes were identified in the field resulting in a training data set consisting

1770 of 172 in-situ observations made by the lead researcher while active in the field. The

1771 identified classes were Arable land, Kalahari, Mopane and Teak woodland, Seasonally

1772 flooded grasslands and permanent water. A vector dataset containing the settlements

1773 (comprising areas encompassing borders of housing aggregations), which were digitized

1774 manually using Google Earth, was integrated in the final classification product in a GIS

1775 environment (ESRI, 2012).

1776 Random Forests was the classifier which was found to perform better in this study area and

1777 with the data used, compared to Classification Trees, Support Vector Machines and K

1778 Nearest Neighbor (see Appendix 2 for further explanations).

1779 The two land cover maps that resulted from the classification of the two separate parts of the

1780 study area were then merged in eCognition and the accuracy assessment was performed for

1781 the entire land cover map. The method proposed by Congalton (1991) was employed for the

1782 accuracy assessment which uses an error matrix based on a TTA mask (see Appendix 2 for

1783 further explanations).

1784 3.2.3 Habitat Suitability Modelling

1785 The habitat suitability maps for lion, leopard and Spotted hyaena were generated using

1786 MaxEnt (Phillips et al., 2008), as this software has been found to perform well compared to

1787 other modelling techniques that use presence only data (Elith et al., 2006), and has been

1788 repeatedly used to model the distribution of large carnivores in diverse geographical

1789 contexts (e.g. Jackson et al., 2016; Di Minin et al., 2013; Angelieri et al., 2016; Ahmadi et

1790 al., 2017).

1791 We incorporated empirically generated occurrence data from on-going research and

1792 monitoring in the area, from Lines et al. (2018). In total, 102 x 4 km transects, optimized

1793 for site conditions, were surveyed on foot three times by the author and two experienced

1794 local trackers from the safari hunting industry, amounting to 1,224 km of spoor transects

1795 during the dry season of May–October 2015 (see Appendix 2 for further explanations). To

1796 account for sampling bias problems relevant to prediction accuracy and model fitting, we

1797 spatially filtered occurrence records for all species by thinning records within 500 m of

1798 each other (incorporating a 500 x 500 m pixel-size grid of the area) using spThin package

1799 in R (Aiello-Lammens et al., 2015). In total, 43 occurrence records were used for lions, 84

1800 for leopards and 78 for spotted hyenas. While we sought to incorporate occurrence data for

1801 the entire extant large carnivore guild known from the Greater Kafue System, sample sizes

1802 were too small for cheetah (Acinonyx jubatus) and African wild dog (Lycaon pictus) to

1803 include in final analyses. For cheetah and African wild dog occurrence records see Lines et

1804 al. (2018).

1805 The full set of predictor variables utilized is presented in Table 3.1. All vegetation-related

1806 variables (except Enhanced Vegetation Index), as well as proportion of arable land,

1807 distance to and proportion of settlements, and distance to water were derived from the land

1808 cover map generated using the Sentinel imagery described above. Roads layers, provided

1809 by Peace Parks Foundation, were modified where inaccurate using ground truthing via 4x4

1810 vehicle during fieldwork planning, and distances to roads calculated in QGIS (2016).

1811 Bioclimatic parameters exhibit little spatial variability across the study area, so we did not

1812 use them as predictor variables (Fitzpatrick et al., 2013). To account for collinearity, we

1813 retained all uncorrelated variables (using r, < 0.7), indicating weak relationships. In our

1814 case, only variables related to anthropogenic pressures and elevation were correlated (Table

1815 3.2).

1816 In terms of variable selection, to keep models as simple as possible to aid in the

1817 interpretation of the species’ distribution and its environment drivers, we strived for

1818 “parsimonious and interpretable models” (Merow et al., 2013). In all models, through a

1819 backwards stepwise process, we excluded all variables that when added in the models did

1820 not produce any increase in Area Under Curve, a measure of model fit. For the same

1821 reasons (parsimony and interpretability), we ran models using only linear and quadratic

1822 features (as suggested by Merow et al., 2013). 30% of the occurrence data was left out as a

1823 testing set, and the models were trained with the remaining 70%. Fifty model iterations

1824 were ran, with 10,000 background points and default pseudoabsence generation. The

1825 resulting habitat suitability map provides a habitat suitability value between 0 and 1 for

1826 each grid cell/pixel.

1827 Table 3.1. Predictor variables used in the model building process. 1828 Source key: 1Peace Parks Foundation (2016), 2Expert knowledge/ground truthing, 1829 3Landcover map (this paper), 4ASTER Global DEM, 5MODIS Vegetation Index Products Predictor Variables 1 Proportion of arable land (3) 10 Distance to minor roads (unpaved) (1,2) 2 Proportion of Kalahari woodland (3) 11 Distance to major and minor roads, combined (1,2) Distance to major, minor roads & rail, combined 3 Proportion of Mopane woodland (3) 12 (1,2) 4 Proportion of Teak woodland (3) 13 Distance to train tracks (1) 5 Proportion of grassland (3) 14 Altitude (Digital elevation model) (4) Proportion of settlements (at 1, 2 and 3 km 6 15 Topographic Position Index (4) radius) (3) 7 Distance to water (4) 16 Slope (5) 8 Distance to settlements (4) 17 Enhanced Vegetation Index (5) 9 Distance to major roads (paved)* (2) 1830

1831

1832

1833

1834

1835 3.2.4 Connectivity Modelling

1836 To provide insights into scope, scale and any species-level overlap of potential ecological

1837 flows in the Kafue-Simalaha Wildlife Dispersal Corridor we developed three individual

1838 species, and one multi-species model using Linkage Mapper (McRea & Kavanagh, 2011),

1839 as this software provides robust estimates of connectivity for large carnivores in

1840 heterogenous landscapes (Wolf & Ripple, 2018; Castilho et al., 2015; Carroll et al., 2012).

1841 The Pathway Tool was chosen to identify and model potential linkages between core areas,

1842 graphically displaying a least-cost corridor where species would encounter more features

1843 facilitating, and less features impeding movement between core areas based on

1844 predetermined site and species-specific resistance layers - the inversed habitat suitability

1845 maps generated by MaxEnt (McRae & Kavanagh, 2011). Parameterizing Linkage Mapper

1846 requires the identification of paired nodes between which patch-based graphic analyses of

1847 connectivity can be undertaken. Given Kafue National Park represents a single source

1848 population of all three target species, with a border extending c.120 km within the study

1849 area, we tested the effect of multiple source nodes along the Kafue border on least-cost path

1850 outcomes, finally settling on 3 roughly equidistant nodes at the northwest, central and

1851 southwestern points bordering a representative cross-section of adjacent Game

1852 Management Areas, implying target species could move between any of these nodes to the

1853 southern node at the Simalaha Wildlife Recovery Sanctuary.

1854

1855

1856

1857

1858 3.3 Results

1859 3.3.1 Landcover Mapping

1860 The final landcover map (Fig. 3) had an overall classification accuracy of 91.6% and a

1861 Kappa Statistic of 0.88, indicating an excellent classification performance identifying our

1862 seven landcover variables. The dominant land cover is Kalahari woodland (56.2%),

1863 followed by Mopane woodland (15.6%) characteristic of south-central floodplains.

1864 Seasonally flooded grassland (13.0%) dominate the floodplains and drainage lines and are

1865 closely associated with water (0.2% and arable land (4.6%). Teak woodland (10.3%),

1866 formerly representing extensive closed-canopy forest tracts, has been heavily denuded by

1867 extractive activities, and further suppressed by fire (Musgrave, 2016). Settlements (0.1%)

1868 are closely associated with drainage lines (available surface water), grasslands and

1869 agricultural areas. Surface water is depicted here for dry season (May-Nov). After heavy

1870 rains most all grassland areas are commonly inundated for many months.

1871 1872 Figure 3.3. Landcover map of the wider study area, indicating wildlife managed areas and 1873 larger settlements.

1874

1875 3.3.2 MaxEnt Habitat Suitability Models

1876 Three habitat suitability models were built, one for each species (Fig. 3.4). The models for

1877 lion had the best AUC (0.81), followed by leopards (0.80) and spotted hyenas (0.79). All

1878 models identified Kafue National Park as core habitat with extensive areas of Mulobezi and

1879 Sichifulo Game Management Areas, and to decreasing extent the Forest reserves of

1880 Nachitwe and Martin. Machili Forest Reserve has been identified as largely unsuitable for

1881 all three species. Further south increased variability was exhibited between wildlife

1882 managed areas and species, though western areas of Simalaha present poor suitability.

1883 Interestingly east of Simalaha was identified as common suitable habitat.

1884

1885 1886 Figure 3.4: Habitat Suitability Map for Spotted Hyena

1887 1888 Figure 3.5: Habitat Suitability Map for Lion

1889 1890 Figure 3.6: Habitat Suitability Map for Leopard 1891

1892 According to model output, the three species distributions are affected by different

1893 variables (see Table 3.2). Notably the proportion of settlement, a proxy of human

1894 population density, had a negative effect on all species. Proportion of arable land had strong

1895 negative effects on leopard and hyena. All three species exhibited strong relationships to

1896 road and/or rail infrastructure indicating increased probability of finding all three species

1897 the further from roads and/or rail. Proportion of Kalahari woodland cover was positively

1898 associated with all species, with Lion positively associated with proportion of all woodland

1899 landcover classifications. Proportion of grasslands (lion and hyena) and distance to water

1900 (leopard and hyena) were also notable predictor variables. 86

1901 Table 3.1. Variable importance for the Maxent models, based on AUC using a single 1902 variable. Light grey are the variables with a negative contribution to relative habitat 1903 suitability. Spotted Predictor variable Lion Leopard hyena Proportion of arable land - 0.67 0.64 Proportion of settlements 0.57 0.53 0.52 Distance to settlements - - - Distance to major roads (paved) - - 0.65 Distance to minor roads (unpaved) - - - Distance to major and minor roads (combined) 0.72 - - Distance to major and minor roads and rail (combined) 0.70 - Distance to train tracks - - - Proportion of Kalahari woodland -0.64 -0.56 -0.52 Proportion of Mopane woodland -0.64 - - Proportion of Teak woodland -0.55 - - Proportion of grassland -0.58 - -0.60 Distance to water - -0.60 -0.59 Topographic Position Index -0.64 - - Slope - -0.65 - 1904

1905

1906 3.3.3 Connectivity Modelling using Linkage Mapper

1907 Pathways analyses indicates broadly similar spatial routing for all three species, with least-

1908 cost corridors consistently following the Kafue National Park border, indicating optimal

1909 (core) habitat within Kafue National Park (Fig. 3.5). Irrespective of source nodes, all

1910 models produced a least-cost corridor exiting Kafue National Park at the southern border, at

1911 the closest spatial point to the southern node. This followed the route through eastern

1912 Mulobezi Game Management Area, Sichifulo Game Management Area, the border areas of

1913 Martin Forest Reserve and Nyawa Communal Areas, then through the eastern section of

1914 Simalaha Communal Conservancy, extending notably east of Simalaha in the case of lion

1915 into Open Communal Areas. The Spotted hyena model provided the largest spatial extent of

1916 moderate to highly suitable corridor area (depicted as yellow/green/blue shading), followed

1917 by leopard then lion.

1918 Minimum corridor width, identified as moderately to highly suitable areas, was 5.1km for

1919 lion, 4.3km for leopard and 3.2km Spotted hyena. All three species’ paths narrowed

1920 noticeably in southern areas, and all reached the Simalaha Wildlife Recovery Sanctuary at

1921 the northeast corner.

1922

1923 1924 Figure 3.7. Connectivity modelling outputs by species. Habitat suitability following Fig 1925 3.4-3.6. Darker monochrome areas indicate increased habitat suitability.

1926

1927

1928 The derived multi-species model (Fig. 3.8) exhibits broadly similar corridor routing and

1929 spatial extent through Kafue National Park and eastern Mulobezi Game Management Area

1930 into Sichifulo Game Management Area, before displaying greater variability and a

1931 reduction in common corridor routing and current flow strength in central southern areas. A

1932 ~25 km long stretch of low current flow, starting at the interface of the Nyawa West

1933 Communal Area and the adjacent Open Communal Area, indicate potential constraints to

1934 species movement. Moderate-highly suitable current flow, though narrow (<2 km width),

1935 converges again within ~10 km from the northeast corner of the Simalaha Wildlife

1936 Recovery Sanctuary.

1937 1938 Figure 3.8. Composite connectivity model for 3 species, indicating area with potential 1939 connectivity constraint.

1940 3.4 Discussion

1941 Connectivity is an increasingly threatened ecological process and predicted to deteriorate

1942 due primarily to issues surrounding human population growth and encroachment around

1943 core wildlife areas (Wittemyer et al., 2008), with severe implications for many wide-

1944 ranging, low-density species (Ripple et al., 2014; Watson et al., 2015). Maintaining and

1945 promoting connectivity in wildlife managed area networks is thus considered an important

1946 intervention for the conservation of free-ranging large carnivore populations, elevating the

1947 value of the proposed Kafue-Simalaha Wildlife Dispersal Corridor, and providing a test

1948 case for explicit goals and outcomes within the KAZA-TFCA Programme.

1949 Developing linkages in heterogenous landscapes presents practical and theoretical

1950 challenges surrounding the identification of land that will best meet the requirements for

1951 focal species to move between core wildlife managed areas, and how to translate resource

1952 selection measurements into resistance (Beier et al., 2008). By choosing a multi-species

1953 approach that focuses on wildlife with high susceptibility to human disturbance and

1954 resource competition, and those driving broader ecological processes, our approach serves

1955 as a model for a collective umbrella under which a wider range of species and processes are

1956 factored in, creating a broadly applicable approach pertinent to the wider KAZA-TFCA

1957 landscape and beyond. Considering that our approach is based on often available species

1958 occurrence data and free to use Sentinel images, it could serve as a general approach in

1959 other data-poor regions in Central Southern Africa and beyond.

1960 Our findings indicate that in the area of the proposed Kafue-Simalaha Wildlife Dispersal

1961 Corridor, anthropogenic pressures are affecting the distribution of large carnivore fauna,

1962 with important implications for landscape and wildlife management in the area. Our results

1963 indicate that settlements, arable land and road and rail infrastructure have clear negative

1964 effects on all species, albeit to different extents across the species (see Table 2 for response

1965 curves). While well documented case studies highlight the occurrence of resident Spotted

1966 hyena and leopard populations in and around human settlements, in our case the area under

1967 study is not heavily urbanized nor populated as in other cases in the literature (e.g. Yirga et

1968 al., 2013; Athreya et al., 2013). Our findings indicate that the effects of settlements are

1969 context dependent, demonstrating the importance of understanding broader area and

1970 management characteristics (Woodroffe, 2000; Linnell et al., 2001). Road and rail

1971 infrastructure in various combinations impacts all species. While literature indicates wide-

1972 ranging impacts of transport infrastructure on vertebrate populations (e.g. Trombulak &

1973 Frissell, 2002; Coffin, 2007), there are significant limits to our understanding of how

1974 transport links affect carnivore populations, though by providing people access to remote

1975 areas there are likely broader issues surrounding negative synergistic effects of access at

1976 larger scales (Forman et al., 2003).

1977 All species in the study area were habitat generalists, and we therefore predicted strong

1978 relationship to a cross-section of habitat types. Lion exhibit a strong preference for

1979 woodland cover, a localised characteristic found by Elliot et al. (2014), and likely a

1980 response partly driven by spatial avoidance of humans (Oriol et al., 2015; Loveridge et al.,

1981 2017) and preferred prey availability (Schaller, 1976). The more catholic diet of leopard

1982 and hyena allows both species to exploit a wider range of habitats and food sources

1983 (Hayward & Kerley, 2008). The relationship of leopard and hyena to dry season surface

1984 water poses questions surrounding disturbance competition by seasonal shifting pastoralist

1985 activities at water points, including elevated poaching pressure (Lindsay et al., 2013),

1986 charcoal production (Munthali et al., 2018) and impacts of fire (Musgrave, 2016).

1987 Our modelled corridor routing for single and multi-species models broadly compliments

1988 existing outputs (Cushman et al., 2018 & 2016), indicating limited habitat suitability in

1989 southern and central areas, increasing habitat suitability towards Kafue National Park, and

1990 also preferential corridor routing along eastern areas of the proposed Kafue-Simalaha

1991 Wildlife Dispersal Corridor. At the wildlife managed area scale our output indicated

1992 elevated habitat suitability within Kafue National Park which is intuitively encouraging

1993 given Kafue National Park is a significant core wildlife area, and specifically for the

1994 region’s large carnivores (IUCN, 2006b; Jacobson et al., 2016; RWCP & IUCN/SSC,

1995 2015). Given extensive areas of northern and eastern Mulobezi GMA are ostensibly free of

1996 direct and much indirect human pressure (Lines, expert opinion) there was an expectation

1997 for elevated current flow in these areas, though evidently least-cost path models prioritised

1998 Kafue National Park and proximity to the destination node, though this does not preclude

1999 availability of suitable habitat and prey in remote areas of Mulobezi GMA (Lines et al.,

2000 2018). Further south, beyond the Forest Reserves and into the Communal Areas, current

2001 flow closely matched expectations, with constrained current flow for all species, and the

2002 possibility of a connectivity bottleneck south of Bombwe, representing an area where

2003 movement is funneled, curtailed or severed, and connectivity diminished or lost (Berger et

2004 al., 2006), in support of the hypothesis offered by Lines et al (2018).

2005 Notwithstanding expectations, and given likely compounded error through occurrence, land

2006 cover and habitat suitability analyses, the corridor modelling, both at species and multi-

2007 species level, clearly reflects reduced connectivity moving south from Kafue National Park

2008 into more human disturbed areas. From a connectivity management perspective this is a

2009 particularly important area to keep intact, as loss of this area might disproportionally

2010 compromise connectivity (Castilho et al., 2015). A greater understanding of implications

2011 concerning species-level demography on movement would be a valuable addition to these

2012 outputs.

2013 Our integrated approach has extended the analytical framework to provide higher resolution

2014 understanding of localized corridor characteristics – notably overlap in multi-species

2015 corridor routing and spatial constraints in habitat suitability for central-southern extents of

2016 the proposed Kafue-Simalaha Wildlife Dispersal Corridor. This has important implications

2017 for practical corridor planning and interventions to counter effects limiting movement.

2018 While landscape dynamics will affect individual vs population level movements (Muller et

2019 al., 2011), and species-scale demographics will likely influence dispersal mechanisms

2020 (Elliot et al., 2014), our approach has clearly provided valuable supplemental detail and

2021 understanding to existing broader-scale, species-specific models.

2022 What is clear from the land cover map is the extensive settlement and agricultural

2023 development along the majority of watercourses, even deep into Game Management Areas

2024 where settlement and agriculture should be precluded from core wildlife zones. The eastern

2025 areas of Sichifulo Game Management Area, almost all of Machili Forest Reserve and

2026 eastern Nyawa areas have been heavily affected by slash and burn agricultural development

2027 as with the central and western sections of Simalaha. Intensive settlement and agricultural

2028 development are occurring in all periphery areas bordering the proposed Kafue-Simalaha

2029 Wildlife Dispersal Corridor with the exception of the western extents of the Open

2030 Communal area bordering Nyawa East and Simalaha. We urge additional research and

2031 support in this area as a priority addition to wildlife-based land use planning for the Kafue-

2032 Simalaha Wildlife Dispersal Corridor.

2033

2034 3.5 Conclusion

2035 Integration of land cover, habitat suitability and corridor modelling at the landscape level

2036 produced an output broadly in line with existing literature and implying constriction of

2037 connectivity for all three species approaching the southern areas of the Kafue-Simalaha

2038 Wildlife Dispersal Corridor, but identified a potentially important new area in which to

2039 focus conservation efforts. Additionally, at the KAZA TFCA scale, the fencing of the

2040 Simalaha Wildlife Recovery Sanctuary and effects of the Zambezi River on wildlife

2041 movement south to the central cluster of wildlife managed areas centered on Zambezi

2042 Region in Namibia and Chobe National Park in Botswana have not been modelled.

2043 With the human population increasing, and expanding conversion of wildlife habitat for

2044 agriculture, any existing bottlenecks hindering wildlife movement, both for large carnivores

2045 and their prey, will likely increase in the absence of effective land use planning and

2046 implementation, including control of poaching and human-wildlife conflict. Nonetheless,

2047 our analyses have produced the first empirical data for evidence-based corridor planning for

2048 Zambia’s key linkage in the KAZA-TFCA programme.

2049

2050

2051

2052

2053 3.6 References

2054 Ahmadi, M., et al., (2017). Combining landscape suitability and habitat connectivity to

2055 conserve the last surviving population of cheetah in Asia. Diversity and

2056 Distributions, vol. 23, no. 6, 592-603.

2057 Aiello-Lammens ME, Boria RA, Radosavljevic A, Vilela B (2015). spThin: an R package

2058 for spatial thinning of species occurrence records for use in ecological niche models.

2059 Ecography, 38(5), 541–545.

2060 Angelieri, C.C.S., et al., (2016). Using species distribution models to predict potential

2061 landscape restoration effects on puma conservation. PLoS One, vol. 11, no. 1,

2062 e0145232.

2063 Athreya, V., et al., (2013). Big cats in our backyards: persistence of large carnivores in a

2064 human dominated landscape in India. PloS one, vol. 8, no. 3, e57872.

2065 Beier, P., et al., (2008). Forks in the road: choices in procedures for designing wildland

2066 linkages. Conservation Biology, vol. 22, no. 4, 836-851.

2067 Berger, J., et al., (2006). Connecting the dots: an invariant migration corridor links the

2068 Holocene to the present. Biology Letters, vol. 2, no. 4, 528-531.

2069 Cardillo, M., et al., (2005). Multiple causes of high extinction risk in large mammal

2070 species. Science (New York, N.Y.), vol. 309, no. 5738, 1239-1241.

2071 Carroll, C., et al., (2012). Use of linkage mapping and centrality analysis across habitat

2072 gradients to conserve connectivity of gray wolf populations in western North

2073 America. Conservation Biology, vol. 26, no. 1, 78-87.

2074 Castilho, C.S., et al., (2015). Evaluating landscape connectivity for Puma concolor and

2075 Panthera onca among Atlantic forest protected areas. Environmental

2076 Management, vol. 55, no. 6, 1377-1389.

2077 Coffin, A.W. (2007). From roadkill to road ecology: a review of the ecological effects of

2078 roads. Journal of Transport Geography, vol. 15, no. 5, 396-406.

2079 Congalton, R.G. (1991). A review of assessing the accuracy of classifications of remotely

2080 sensed data. Remote Sensing of Environment, vol. 37, no. 1, 35-46.

2081 Copernicus Sentinel data (2018). https://asf.alaska.edu/ Accessed 01/06/2018, processed by

2082 ESA.

2083 Crooks, K.R., et al., (2011). Global patterns of fragmentation and connectivity of

2084 mammalian carnivore habitat. Philosophical Transactions of the Royal Society B:

2085 Biological Sciences, vol. 366, no. 1578, 2642-2651.

2086 Cumming, D.H. (2008). Large scale conservation planning and priorities for the Kavango-

2087 Zambezi Transfrontier Conservation Area. Unpublished report commissioned by

2088 Conservation International.

2089 Cushman, S.A., et al., (2013). Biological corridors and connectivity, In: Macdonald, DW;

2090 Willis, KJ, eds. Key Topics in Conservation Biology 2.Hoboken, NJ: Wiley-Blackwell,

2091 384-404.

2092 Cushman, S.A., et al., (2016). A multi-scale assessment of population connectivity in

2093 African lions (Panthera leo) in response to landscape change. Landscape

2094 Ecology, 31(6), 1337-1353.

2095 Cushman, S.A., et al., (2018). Prioritizing core areas, corridors and conflict hotspots for

2096 lion conservation in southern Africa. PloS one, 13(7), p.e0196213.

2097 Di Marco, M., et al., (2014). A retrospective evaluation of the global decline of carnivores

2098 and ungulates. Conservation Biology, vol. 28, no. 4, 1109-1118.

2099 Di Minin, E., et al., (2013). Conservation businesses and conservation planning in a

2100 biological diversity hotspot. Conservation Biology, vol. 27, no. 4, 808-820.

2101 Drusch, M., et al., (2012). Sentinel-2: ESA's optical high-resolution mission for GMES

2102 operational services. Remote Sensing of Environment, vol. 120, 25-36.

2103 eCognition Developer, T. 2014. 9.0 User Guide. Trimble Germany GmbH: Munich,

2104 Germany.

2105 Elith, J., et al., (2011). A statistical explanation of MaxEnt for ecologists. Diversity and

2106 Distributions, vol. 17, no. 1, 43-57.

2107 Elliot, N.B., et al., (2014). Movements vary according to dispersal stage, group size, and

2108 rainfall: the case of the African lion. Ecology, vol. 95, no. 10, 2860-2869.

2109 ESRI (2012). Environmental Systems Research Institute. ArcGIS Release 10.1. Redlands,

2110 CA.

2111 Fahrig, L. (2003). Effects of habitat fragmentation on biodiversity. Annual Review of

2112 Ecology, Evolution, and Systematics, vol. 34, no. 1, 487-515.

2113 Fitzpatrick, M.C., et al., (2013). MaxEnt versus MaxLike: empirical comparisons with ant

2114 species distributions. Ecosphere, vol. 4, no. 5, 1-15.

2115 Forman, R.T., et al., (2003). Road ecology: science and solutions, Island press.

2116 Funston, P.J., et al., (2013). Insights into the management of large carnivores for profitable

2117 wildlife-based land uses in African savannas", PloS One, vol. 8, no. 3, pp. e59044.

2118 Hayward, M.W. & Kerley, G.I. (2008). Prey preferences and dietary overlap amongst

2119 Africa's large predators. African Journal of Wildlife Research, vol. 38, no. 2, 93-109.

2120 IUCN (2006b). Conservation strategy for the lion in eastern and southern Africa. IUCN

2121 SSC Cat Specialist Group, Gland, Switzerland and Cambridge, UK.

2122 Jackson, C.R., et al., (2016). Evaluating habitat connectivity methodologies: a case study

2123 with endangered African wild dogs in South Africa. Landscape Ecology, 1-15.

2124 Jacobson, A.P., et al., (2016). Leopard (Panthera pardus) status, distribution, and the

2125 research efforts across its range. PeerJ, 4, p.e1974.

2126 KAZA (2014). Master Integrated Development Plan. KAZA Secretariat. Victoria Falls.

2127 KAZA (2011a). Treaty Between the Governments of Angola, Botswana, Namibia, Zambia

2128 and Zimbabwe on the Establishment of the Kavango–Zambezi Transfrontier

2129 Conservation Area. Kavango–Zambezi TFCA Secretariat, Kasane, Botswana.

2130 Lindsey PA, et al., (2013). The bushmeat trade in African savannas: Impacts, drivers, and

2131 possible solutions. Biological Conservation. Apr 1; 160:80-96.

2132 Lines, R., et al., (2018). Status of terrestrial mammals at the Kafue–Zambezi interface:

2133 implications for transboundary connectivity. Oryx 1-10.

2134 Linnell, J.D., (2011). Wildlife-Human Interactions: From Conflict to Coexistence in

2135 Sustainable Landscapes, Norsk Institutt for Naturforskning (NINA).

2136 Loveridge, J., et al., (2017). The landscape of anthropogenic mortality: how African lions

2137 respond to spatial variation in risk. Journal of Applied Ecology, vol. 54, no. 3, 815-

2138 825.

2139 Machina, H. (2005). Land is Life. Land Policy and Administration in Zambia, 1-5.

2140 Maxwell, S.L., et al., (2016). Biodiversity: The ravages of guns, nets and

2141 bulldozers, Nature News, vol. 536, no. 7615, 143.

2142 McRae, B.H. & Kavanagh, D.M. (2011). Linkage mapper connectivity analysis

2143 software. The Nature Conservancy.

2144 Merow, C., et al., (2013). A practical guide to MaxEnt for modeling species’ distributions:

2145 what it does, and why inputs and settings matter. Ecography, vol. 36, no. 10, 1058-

2146 1069.

2147 Mueller, T., et al., (2011). How landscape dynamics link individual‐to population‐level

2148 movement patterns: a multispecies comparison of ungulate relocation data. Global

2149 Ecology and Biogeography, vol. 20, no. 5, 683-694.

2150 Munthali, S.M., et al., (2018). Integration of Ecological and Socioeconomic Factors in

2151 Securing Wildlife Dispersal Corridors in the Kavango-Zambezi Transfrontier

2152 Conservation Area, Southern Africa. Selected Studies in Biodiversity, 181.

2153 Musgrave, M. 2016. Scale, Governance and Change in Zambezi Teak Forests: Sustainable

2154 Development for Commodity and Community. Cambridge Scholars Publishing.

100

2155 Oriol‐Cotterill, A., et al., (2015). Landscapes of coexistence for terrestrial carnivores: the

2156 ecological consequences of being downgraded from ultimate to penultimate predator

2157 by humans. Oikos, vol. 124, no. 10, 1263-1273.

2158 Phillips, S.J., et al., (2006). Maximum entropy modeling of species geographic

2159 distributions. Ecological Modelling, vol. 190, no. 3-4, 231-259.

2160 PPF (2008). Integrated Development Plan for the Zambian Component of the Kavango-

2161 Zambezi Transfrontier Conservation Area. Peace Parks Foundation, Stellenbosch,

2162 South Africa.

2163 PPF (2016). In house dataset for the Zambian Component of the Kavango-Zambezi

2164 Transfrontier Conservation Area. Unpublished data. Peace Parks Foundation,

2165 Stellenbosch, South Africa.

2166 Purvis, A., et al., (2000). Predicting extinction risk in declining species. Proceedings

2167 Biological Sciences, vol. 267, no. 1456, 1947-1952.

2168 QGIS Development Team 2016. QGIS Geographic Information System. Open Source

2169 Geospatial Foundation Project. http://qgis.osgeo.org

2170 Ripple, W.J., et al., (2014). Status and ecological effects of the world’s largest

2171 carnivores. Science, vol. 343, no. 6167, 1241484.

2172 RWCP & IUCN/SSC (2015). Regional Conservation Strategy for the Cheetah and African

2173 Wild Dog in Southern Africa; Revised and Updated, August 2015. IUCN, Gland

2174 Switzerland.

101

2175 Schaller, G.B. (1976). The Serengeti Lion: A Study of Predator-Prey Relations. University

2176 of Chicago Press.

2177 Trombulak, S.C. & Frissell, C.A. (2000). Review of ecological effects of roads on

2178 terrestrial and aquatic communities. Conservation Biology, vol. 14, no. 1, 18-30.

2179 UNEP-WCMC (2015). United Nations Environment Programmes World Conservation

2180 Monitoring Centre. Protected Planet

2181 Database https://www.protectedplanet.net/ Accessed 01.05.2015.

2182 Watson, F.G., et al., (2015). Human encroachment into protected area networks in Zambia:

2183 implications for large carnivore conservation. Regional Environmental Change, vol.

2184 15, no. 2, 415-429.

2185 Wittemyer, G., et al., (2008). Accelerated human population growth at protected area

2186 edges. Science (New York, N.Y.), vol. 321, no. 5885, 123-126.

2187 Wolf, C. & Ripple, W.J. (2018). Rewilding the world's large carnivores. Royal Society

2188 Open Science, vol. 5, no. 3, 172235.

2189 Wolf, C. & Ripple, W.J. (2017). Range contractions of the world's large carnivores. Royal

2190 Society Open Science, vol. 4, no. 7, 170052.

2191 Woodroffe, R. (2000). Predators and people: using human densities to interpret declines of

2192 large carnivores. Animal Conservation forum. Cambridge University Press, 165.

2193 Worboys, G., et al., (2010). Connectivity conservation management: a global guide (with

2194 particular reference to mountain connectivity conservation), Earthscan.

102

2195 Yirga, G., et al., (2016). Spotted hyena (Crocuta crocuta) concentrate around urban waste

2196 dumps across Tigray, northern Ethiopia. Wildlife Research, vol. 42, no. 7, 563-569.

2197 ZAWA 2010. General Management Plan for Kafue National Park. ZAWA, Chilanga,

2198 Zambia.

2199

2200

2201

2202

2203

2204

2205

2206

2207

2208

2209

2210

2211

2212

103

2213 Chapter 4: Utility of Human footprint pressure mapping for large

2214 carnivore conservation: The Kafue-Zambezi Interface

2215 Authors: Lines, R.,1* Bormpoudakis, D1., Xofis, P2., MacMillan. D1., Pieterse, L3 and

2216 Tzanopoulos. J1

2217 1 Durrell Institute of Conservation and Ecology, School of Anthropology and Conservation,

2218 University of Kent, UK

2219 2 International Hellenic University, Department of Forestry and Natural Environment,

2220 Drama, Greece

2221 3 Green Trust Zambia Ltd, Plot 14126, Chipepo Road, Siavonga, Zambia.

2222 * corresponding author [email protected]

2223 Abstract

2224 Increasing human pressures are driving large carnivore declines within and beyond wildlife

2225 managed areas. Edge effects, attractive sinks and bottlenecks constrain connectivity,

2226 isolating populations, and hindering development of wildlife-based land uses at local to

2227 Transboundary scales.

2228 The development of proxies for monitoring changes to human disturbance against which

2229 species responses can be quantified provides a novel conservation tool to model and map

2230 species distribution and likelihood of distribution change. This can improve our

2231 understanding of threshold pressures at which species can persist, are extirpated or might

2232 recolonize human-dominated landscapes.

2233 We integrated high resolution remote sensing data and in situ mapping of human pressure

2234 variables with an existing landcover map to generate a site-specific, fine scale human

104

2235 footprint pressure map for 3.9 m ha of rangelands at the Kafue-Zambezi interface - a key

2236 linkage in the Kavango-Zambezi Transfrontier Conservation Area programme. We then

2237 modelled human footprint pressure against empirically derived occurrence data for Lion

2238 (Panthera leo), Leopard (Panthera pardus) and Spotted Hyena (Crocuta crocuta). Using

2239 these data, we then generated human footprint pressure threshold ranges at which each

2240 species were persisting or extirpated within ten wildlife managed areas linking Kafue

2241 National Park to the Zambezi River.

2242 Fine scale model results overcame many limitations inherent in existing large-scale human

2243 footprint pressure models, providing encouraging direction for this approach. Human

2244 footprint pressure thresholds at which species persisted or were extirpated from wildlife

2245 managed areas were broadly in line with expectations, indicating this approach is valid for

2246 site- and species-specific modelling.

2247 Model performance will improve as additional datasets come available and with improved

2248 understanding of how asymmetrical and nonlinear threshold responses to footprint pressure

2249 changes across spatial-temporal scales. This approach has broader utility for local and

2250 regionwide conservation planning where mapping and managing the ubiquitous impact of

2251 humanity will determine species persistence throughout protected area networks.

2252 Key words: Human footprint pressure, Thresholds, Large Carnivores, Kafue, Kavango-

2253 Zambezi Transfrontier Conservation Area, Connectivity

2254 Word count: 6110

2255

2256

105

2257 4.1 Introduction

2258 The ubiquitous impact of humanity on the planet stretches from the deep ocean to

2259 mountaintops, manifesting through our direct demands on natural resources and indirect

2260 effects of these demands on wider global systems (Vitousek et al.,1997; Wackernagel et al.,

2261 2002). The wide-ranging implications of increasing spatial-temporal resource demands

2262 include loss and fragmentation of key wildlife habitats (Riitters et al.,2000; Foley et al.,

2263 2005), constraining species movement (Tucker et al., 2018) and the reduction and

2264 extinction of wildlife populations at multiple scales (Maxwell et al., 2016; Ceballos &

2265 Ehrlich, 2002; Ceballos et al., 2015).

2266 While the decline in human pressures on natural systems is presenting new opportunities

2267 for rewilding and carnivore conservation throughout much of continental Europe (Navarro

2268 & Pereira, 2015), many of the world’s developing regions supporting large tracts of

2269 existing wildlife habitat and high levels of biodiversity (Brooks et al., 2006) are

2270 experiencing intensifying spatio-temporal human pressures in and around protected areas

2271 (Newmark, 2008; Geldmann et al., 2014). Increased human resource demands in these

2272 areas are also impacting conservation efforts and political support for the maintenance and

2273 expansion of wildlife-based land uses and wildlife economies at regional, national and

2274 transboundary scales (Duffy, 2006; Gren et al., 2018).

2275 These anthropogenic pressures have been characterized as the human footprint and

2276 manifest through inter alia population growth, the expansion of built areas and settlement,

2277 transport infrastructure and linkages, agropastoralism and extractive industries (Sanderson

2278 et al., 2002).

106

2279 Significant human footprint pressure habitually results in profound and complex effects

2280 impacting the structure and function of ecosystems, including changes to key resources

2281 driving socioecological system productivity and resilience (Walker & Salt, 2012) and

2282 livelihood opportunities for communities residing within them (Venter et al., 2016).

2283 Elevated human footprint pressure decrease structural and functional connectivity between

2284 wildlife managed areas for many species of conservation concern (Ayram et al.,2017).

2285 Existing human footprint pressure analyses have traditionally been generated at relatively

2286 low resolution to provide overviews and indicators of human footprint pressure impacts on

2287 states of interest, such as habitat change and protected area integrity at global scales

2288 (Venter et al., 2016; Jones et al., 2018; Watson et al., 2016).

2289 Increasingly the focus of human footprint pressure is shifting to consider its utility as a

2290 proxy or predictive indicator for measuring and understanding finer scale impacts on

2291 species and processes, including studies on species movement (Tucker et al.,2018),

2292 behaviour (Gaynor et al.,2018), extinction risk (Di Marco et al.,2018), range use (Di Marco

2293 & Santini, 2015) and more broadly as a conservation planning tool (Tombulak et al., 2010).

2294 These approaches seek to overcome many of the questions and limitations surrounding data

2295 availability, accuracy and resolution posed by conventional larger scale multivariate

2296 models.

2297 Generating site- and species-specific human footprint pressure models that can be used as a

2298 proxy or indicator of species-level habitat suitability and sensitivity to human pressure can

2299 aid our understanding of thresholds at which species persist, are extirpated or are likely to

2300 recolonize both protected and non-protected areas, leading to improved application of

2301 conservation science in management (Sutherland et al., 2004). But beyond large scale

107

2302 assessments (Di Marco et al., 2018), these tools are poorly understood and developed

2303 owing chiefly to an absence of integrated fine-scale remote sensing and in situ data,

2304 precluding appropriate accuracy and resolution (Woolmer et al., 2008).

2305 The Kavango-Zambezi Transfrontier Conservation Area (KAZA TFCA) in central southern

2306 Africa seeks to promote connectivity between clusters of wildlife managed areas at the

2307 interface of five neighboring countries (KAZA, 2011a). Connectivity at the species and

2308 scale of interest are poorly studied within and between many of the proposed landscape-

2309 scale linkages (Cumming, 2011), but with ubiquitous human pressure increasing

2310 throughout the region (Newmark, 2008; Wittemyer et al., 2008), there is a need to

2311 understand how human footprint pressure are impacting connectivity for key species of

2312 interest throughout core linkages.

2313 Large carnivore guilds exert significant top-down influence on ecosystems, imparting

2314 strong regulatory pressures driving ecosystem structure and function (Ripple et al., 2014,

2315 Estes et al., 2011). They are highly susceptible to direct and indirect human activities

2316 including (legal and illegal) hunting, reduction of wild prey and habitat fragmentation and

2317 loss (Maxwell et al., 2016; Riggio et al., 2018). Large carnivores are also a key asset for the

2318 development of wildlife-based land uses and wildlife economies (Funston et al., 2013), and

2319 have been identified by the KAZA TFCA Programme as target species for conservation

2320 action, including the stabilization and growth of populations in key habitats, and

2321 maintenance of secure and active connectivity pathways between core wildlife managed

2322 areas (KAZA, 2018). In concert these factors indicate large carnivores as appropriate target

2323 species against which to model human footprint pressure.

108

2324 This study aims to generate a site-specific, fine scale map of human footprint pressure to

2325 test 1) the validity of this appropriate for predicting species occurrence and 2) explore if

2326 this approach can determine discernible human footprint pressure thresholds at which target

2327 species persist or are extirpated at the wildlife managed area scale.

2328

2329 4.2 Methods

2330 4.2.1 Study Area

2331 The KAZA TFCA covers c.520,000 km2 of central Southern Africa, spanning the borders

2332 of Angola, Botswana, Namibia, Zambia and Zimbabwe, and centered around the Kavango

2333 and Zambezi River basins (KAZA, 2014). The KAZA TFCA landscape incorporates a

2334 network of ~70 protected areas in IUCN I-VI and Not Reported categories (UNEP-WCMC,

2335 2015). These protected areas are characterized by a wide spectrum of investment and

2336 management effectiveness (Lindsey et al., 2014).

2337 Spatially, these protected areas fall broadly into three major and five periphery clusters,

2338 with connectivity between protected area clusters identified as one of KAZA TFCA

2339 Programme’s central objectives (KAZA, 2014).

2340 Kafue National Park and surrounding protected areas, collectively known as the Greater

2341 Kafue System, represents the KAZA TFCA’s major northern cluster (Fig. 4.1) and

2342 Zambia’s majority contribution to the KAZA TFCA Programme (KAZA, 2014).

2343 Connectivity between Kafue National Park and adjacent protected areas, centered on Chobe

2344 National Park and East Zambezi Region in Namibia, is contingent on movement across

2345 eight partially and nominally protected areas plus an adjacent open Communal Areas

2346 identified by Lines et al. (2020 in review) as potentially important for corridor planning. In

109

2347 concert these areas span ~13,000 km2, extending140-170 km from the Kafue National Park

2348 border south-southwest towards the Zambezi River at the confluence of Zambia, Namibia

2349 and Botswana (Lines et al., 2018).

2350 2351 Figure 4.1. The Kavango-Zambezi Transfrontier Conservation Area landscape, indicating 2352 study area, clusters of wildlife managed areas (WMAs) and their degrees of protection. 2353 Protected = National Parks; IUCN II; Partially Protected = IUCN III-VI; Nominally 2354 Protected = IUCN Not Reported (adapted from KAZA, 2011a).

2355

2356

2357

2358

110

2359 The landscape is historically, and remains, characterized by dynamic spatiotemporal human

2360 pressures, though few data on the areas’ wildlife and human population are available prior

2361 to the 1960’s (Lines et al., 2018; CSO, 2019). Much of the study area was sparsely settled

2362 until the development of a railway from Livingstone to Mulobezi from 1923-4 to exploit

2363 the region extensive tracts of Zambezi Teak forest (Baikiaea plurijuga). Access to formerly

2364 remote areas had profound impacts on its people and wildlife (Calvert, 2005). Southern

2365 areas around Simalaha, bordering the Zambezi River, were heavily depopulated during the

2366 1966-1990 Angolan Bush War, and thereafter increasing numbers of agro-pastoralists have

2367 settled this landscape (Yeta, pers comms), with significant increases in human disturbance

2368 (Lines et al., 2018). Systematic census from 2000 onward indicate Districts with

2369 boundaries intersecting the study area have experienced annualized population growth of

2370 ~2.8%, with an average population density of ~4.5/km2 (CSO, 2019). But these larger scale

2371 surveys hide significant finer scale variation.

2372 4.2.2 Generating human footprint pressure maps

2373 Early Geographical Information System-based versions of the human footprint pressure

2374 sought to build on the concept of Ecological Footprint mapping (Rees et al.,1996), utilizing

2375 availability of new Earth observational data sets and advances in satellite imagery

2376 capabilities covering human activities and the physical world, including land use and cover,

2377 transport linkages and human population density. This increase in resolution and grain has

2378 facilitated the development of geographical proxies for inferring variation in global human

2379 influences believed to have the most important direct pressure on wildlife and wildlands

2380 (Sanderson et al., 2002). With this baseline multivariate model Venter et al. (2016)

2381 duplicated the methodology of aggregating pressure scores at a global 1 km2 resolution,

111

2382 based on long-term datasets, to generate updated human footprint pressure and trends over

2383 time.

2384 While the Kafue-Zambezi landscape lacks long term datasets from which to derive trend

2385 data, our reworking of the Sanderson et al. (2002) and Venter et al. (2016) methodology

2386 sought to integrate the highest resolution data sets currently available for the landscape to

2387 generate comparable outputs at two orders of magnitude fine scale. There is no current

2388 available data on pastureland, which we omitted from our fine scale reworking, leaving 7

2389 pressure scores (Table 4.1).

2390 Individual variable pressure scores weights are estimates of relative human pressures

2391 following Sanderson et al. (2002), with many co-occurring at the same spatial scale. Using

2392 our calculation the maximum human footprint pressure score is 43.8

2393 1. Settlement data was derived from Bonafilia et al. (2019) at 30m resolution. All pixels

2394 overlapping settlement areas were given a pressure score of 10 representing the highest

2395 level of direct pressure (implying settled area were unsuitable for wildlife), with all

2396 other pixels given a score of 0. The raster layer was then rescaled to 10 m resolution

2397 and written to file.

2398 While recognizing the complexity and limitations surrounding downscaling, we argue

2399 here that scaling from 30 m to 10 m represents a small downscaling in context of

2400 settlement characteristics, insofar as the area immediately outside buildings might

2401 reasonably be considered part of the building/living area in local rural contexts, and

2402 thus downscaling should not pose real or theoretical problem to broader analyses.

2403

112

2404 2. Human Population Density data was unavailable at sufficiently fine scale for the

2405 landscape to include as a stand-alone data layer. Given the largely homogenous nature

2406 of settlement throughout the area – an absence of large multi-story buildings and dense

2407 conurbations versus ubiquitous single-story concrete block and tin buildings with

2408 scattered clay (adobe) and grass huts throughout rural area (Lines, pers obs), we

2409 generated average population density for the study area at the District scale using 2019

2410 population census projected data (CSO, 2019). We then overlaid this on the rescaled

2411 settlement layer and applied Venter el al’s. (2016) methodology to provide an addictive

2412 pressure score of 7.363 for every pixel overlaying settlements. This layer was written to

2413 file at 10 m resolution.

2414 3. Roads: The roads layer pressure scoring followed Venter el al. (2016) providing a

2415 pressure score of 8 for 0.5 km either side of roads, indicating high direct human

2416 pressures on and close to roads with immediate access, and a pressure score of 4

2417 decaying out to 15 km from 0.5 km either side of the roads, indicating lower indirect

2418 pressures moving away from the roads, considered the approximate distance a person

2419 might reasonably access on foot within a day.

2420 Firstly, we manually modified a vector roads layers supplied by Peace Parks

2421 Foundation (unpublished data) to define major tar and secondary dirt roads linking

2422 settlement nodes in QGIS (QGIS, 2018), omitting tertiary dirt tracks from analyses due

2423 to their dynamic nature and inconsistent mapping. We then created a base raster layer

2424 using R (R core team, 2018) with 10 m pixels to cover the study area for both road

2425 types giving a value of 1 for each pixel that intersects a road and 0 for all other pixels,

2426 then generated a distance raster showing the distance of each ‘0’ pixel to each ‘1’ pixel

2427 (the roads) using the GRASS add-on in QGIS with the command r.grow.distance and 113

2428 the ‘Euclidian’ parameter. This created two rasters with continuous values of distance

2429 in meters for each road pixel. We assigned all pixels 0m to 500 m from the road with a

2430 pressure score of 8, and all pixels more than 15 km from the road to N/A so they are

2431 ignored when the exponential decay model was applied to calculate human pressure

2432 score within 15 km of roads Starting with a pressure score of 4 at 500 m from roads we

2433 applied the exponential decay function to a pressure score of 0.25 out to 15 km (you

2434 cannot decay out to 0 or the calculation never ends). Both rasters were then aggregated

2435 and the final layer was written to file at 10 m resolution.

2436 4. Railways represent direct drivers of habitat conversion and conduits of access into

2437 wildlife areas, and indirect effects in adjacent areas, similar to roads, although as

2438 passengers cannot commonly disembark at will, indirect effects away from the railway

2439 line are considered minimal. Following Venter el al. (2016) we gave railways a direct

2440 pressure score of 8 for a distance of 500 m either side of the railway using the same

2441 method as for roads at 10 m resolution. The layer was written to file.

2442 5. Navigable Waterways, like roads, provide direct access to wildlife habitat along the

2443 waterway, and indirect access in periphery areas. The Zambezi River is the only

2444 permanent navigable waterway in our study area, and following Venter el al. (2016), we

2445 gave this waterway a direct pressure score of 4, exponentially decaying out to 15 km as

2446 with the roads layer methodology. The layers were written to file at 10 m resolution

2447 concurrent with the base landcover map generated from Lines et al. (2020, in review).

2448 6. Arable land throughout the Kafue-Zambezi interface is characterized by majority maize

2449 and pulses cultivated using the traditional Chitemene low input, rain fed, slash and burn

2450 farming method (Musgrave, 2016). Arable landcover classifications are considered by

114

2451 Venter el al. (2016) to provide intermediate disturbance to wildlife though direct

2452 reduction of wildlife habitat.

2453 For Arable land the Arable class was selected from the 10 m resolution landcover map

2454 classification generated from Lines et al. (2020, in review) and rasterized onto the 10m

2455 base layer, with a pressure score of 7 for pixels intersecting an area of Arable land, and

2456 0 for all other pixels. This layer was then written to file.

2457 7. Night-time light infrastructure, while sparse and low intensity throughout much of our

2458 study area, is considered a direct human pressure limiting wildlife through a range of

2459 negative impacts (Rich & Longcore, 2013).

2460 The "vcm-orm-ntl" (VIIRS Cloud Mask - Outlier Removed - Night-time Lights) annual

2461 average layer was used (NOAA, 2019) and rescaled to 10 m. Pixels with a value 0 (no

2462 light) were excluded from the layer, and following Venter et al. (2016) we applied the

2463 same log formula used for pop density resulting in pixels with scores ranging from 0 to

2464 8.971. This layer was then written to file.

2465 8. Aggregating the layers: The score layers were added together in R, to give the final

2466 modified human footprint pressure, which was then saved to file-at 10m resolution.

2467

2468

2469

2470

2471

2472

115

2473 Table 4.1. Modifications to Human Pressure Variables and Pressure Scoring.

Naïve Variable Pressure Score Source Details Resolution

Settlement 0,10 Bonafilia et al., 2019 30m All settled areas mapped given score of 10

Population Pressure score=3.333×log (population 0–10 Continuous Bonafilia et al., 2019 30m Density density+1)

0,8 Direct impacts Direct pressure score of 8 for 500m either Roads 0–4 Indirect PPF, unpublished data 10m side of road, exponentially decaying out to 4 impacts at 15 km

Direct pressure score of 8 for 500m either Railways 0-8 PPF, unpublished data 10m side

Navigable Pressure score of 4 exponentially decaying 0–4 PPF, unpublished data 10m Water out to 15 km

Arable 0,7 Lines et al., 2019 10m All areas mapped as crops given score of 7

Night 0-9 NOAA, 2019 100m Pressure score=3.333×log (NTL+1) Lights/NTL 2474

2475 4.2.3 MaxEnt Habitat Suitability Modelling

2476 Following the methods described in Lines et al. (2020, in review), habitat suitability maps

2477 for lion, leopard and spotted hyaena were generated using MaxEnt (Phillips et al., 2008)

2478 which performs well compared to other modelling techniques using presence only data

2479 (Elith et al., 2011), and has been repeatedly used to model large carnivores distribution

2480 (e.g. Jackson et al., 2016; Di Minin et al., 2013; Angelieri et al., 2016; Ahmadi et al.,

2481 2017).

2482 We incorporated empirically generated occurrence data from Lines et al. (2018). In total,

2483 102 x 4 km transects, optimized for site conditions, were surveyed on foot three times by

2484 the author and two experienced local trackers from the safari hunting industry, amounting

2485 to 1,224 km of spoor transects during the dry season of May–October 2015, based on a

2486 pilot study to determine optimal sampling effort to detect target species and cover the

116

2487 landscape in a single field season (Lines & MacKenzie, unpublished data, see Appendix 1

2488 for more detail and explanations).

2489 To account for sampling bias problems relevant to prediction accuracy and model fitting,

2490 we spatially filtered occurrence records for all species by thinning records within 500m of

2491 each other (incorporating a 500 x 500 m pixel-size grid of the area) using spThin package

2492 in R (Aiello-Lammens et al., 2015). In total, 43 occurrence records were used for lions, 84

2493 for leopards and 78 for spotted hyenas. While we sought to incorporate occurrence data for

2494 the entire extant large carnivore guild known from the Greater Kafue System, sample sizes

2495 were too small for Cheetah (Acinonyx jubatus) and African wild dog (Lycaon pictus) to

2496 include in final analyses. The predictor variable modelled against single occurrence was the

2497 aggregated human footprint.

2498 The relationship of large carnivores to changing human footprint pressure is well

2499 established at the global scale (Di Marco et al., 2013, 2015). However, application of this

2500 relationship towards an understanding of thresholds at which species occur, are locally

2501 extirpated, or might recolonize areas is poorly developed, irrespective of its clear utility as a

2502 conservation tool (Tombulak et al., 2014). A supplemental exploratory analysis was

2503 undertaken to investigate species sensitivity to human footprint pressure at the wildlife

2504 managed area scale.

2505 To undertake this analysis we extracted mean human footprint pressure at the wildlife

2506 managed area scale against empirically derived occurrence data for lion, leopard and

2507 spotted hyena following Lines et al. (2018). Outputs were compared against species-level

2508 sensitivities to extinction from Di Marco et al. (2018), and the ranking of sensitivities to

2509 localized extirpation following Riggio et al. (2018).

117

2510 4.3 Results

2511 4.3.1 Human footprint pressure

2512 The 10 m resolution map (Fig. 4.2) indicates areas of high to low human pressure, with

2513 notable areas of highest human footprint pressure around Sesheke/Katima Mulilo in the

2514 southwest, along the Zambia/Namibia border following the east-west tar road, along much

2515 of the Zambezi River and in the central/eastern areas dominated by access roads, settlement

2516 and agricultural development. Broadly, settlement and agricultural development is

2517 widespread throughout the landscape, concurrent with the formal and informal road

2518 network.

2519 Areas of low apparent human footprint pressure include Kafue National Park (where

2520 settlement and agriculture are illegal and non-existent), and adjacent areas of northern

2521 Mulobezi, Sichifulo Game Management Areas, Nachitwe and Martin Forests.

118

2522

2523

2524 2525 Figure 4.2. Dissagregated Human Footprint layers overlaid on wildlife managed areas. 2526 Darker areas represent higher human footprint pressure.

119

2527 2528 Figure 4.3. Human footprint pressure, Kafue-Zambezi Interface, 10m resolution.

2529

2530 4.3.2 MaxEnt Habitat Suitability Modelling Outputs

2531 The best performing model was for lion with an AUC of 0.76 followed by leopard (AUC

2532 0.70) and finally Spotted hyena (AUC 0.61), indicating strong to moderate model

2533 performance, and a negative correlation between aggregated human pressure and species

2534 occurrence (Fig 4.4).

2535 Human footprint pressure had the greatest impact on lion, then leopard and hyena, with

2536 extensive unsuitable areas for all species in central-southern areas, and especially along the

2537 river parallel to the main tar road where much settlement and agriculture also occurs.

2538 Another notable linear feature of human pressure affecting all species followed the railway

2539 line and parallel roads, interspersed with settlement and arable lands, where transport

120

2540 infrastructure, settlement and agricultural development are driving the human footprint

2541 pressure throughout this landscape.

2542 At the wildlife managed area scale human footprint pressure was lowest in Kafue National

2543 Park, Mulobezi and western parts of Sichifulo Game Management Areas. Both Nachitwe

2544 and Martin Forest Reserves appear relatively intact, though exhibiting elevated pressure

2545 towards their western boundaries. Machili Forest Reserve is heavily impacted throughout

2546 by human pressure. There are still areas within Nyawa communal lands with relatively low

2547 human pressure and again to the northeast and eastern sections of Simalaha, extending into

2548 the adjacent Open Area. Extensive pressure exists around the settlement of Bombwe,

2549 formerly a registered Forest Reserve, although now degazetted in recognition of

2550 unmanaged human pressures degrading the forest and natural resources (Inyambo-Yeta,

2551 pers comms). Simalaha Wildlife Recovery Sanctuary, sandwiched between the Zambezi

2552 River and main Tar road running between the borders of Namibia and Zimbabwe, is subject

2553 to elevated human pressure, including settlement and agriculture both within the Sanctuary

2554 and on its borders.

2555

2556

2557

2558

2559

2560

121

2561 Lion: HFP / Mean AUC: 0.764

2562 2563 Leopard: HFP / Mean AUC: 0.699

2564 2565 Spotted hyena: HFP / Mean AUC: 0.606

2566 2567 Figure 4.4. Human footprint pressure-based Habitat Suitability Models 122

2568 Response curves in Fig 4.4 are an indication of species sensitivity to human footprint 2569 pressure, showing highest sensitivity for lion followed by leopard then spotted hyena, as 2570 expected by availability of suitable habitat.

2571 2572 4.3.3 Human footprint pressure thresholds and Species Persistence

2573 The gradient of mean human footprint pressure at the wildlife managed area scale varied

2574 from 1.1 in Kafue National Park to 6.8 in the western section of the Simalaha Sanctuary,

2575 with IUCN categorized protected areas experiencing lowest mean Human footprint pressure

2576 (Table 4.2). With the exception of Machili Forest (human footprint pressure 5.6) there was

2577 a steady increase in human footprint pressure moving south away from Kafue National

2578 Park towards the Zambezi River.

2579 The human footprint pressure threshold (Table 2) at which each species occurs in each

2580 wildlife managed area revealed broadly similar high sensitivities. Lion exhibited the highest

2581 sensitivity to human footprint pressure with an occurrence threshold between 2.4-3.7,

2582 followed by hyena and leopard, with threshold values between 3.7-4.6, mirroring species

2583 sensitivity presented by Riggio et al. (2018).

2584 An apparent anomaly is Machili forest reserve, with extensive settlement, agriculture and

2585 transport infrastructure, having a mean human footprint pressure score of 5.6, and with both

2586 leopard and spotted hyena occurring.

2587 The proposed supplemental addition to the protected area network identified in Lines et al

2588 (2020 in review), Open Area extension 1 (& 2), has a mean human footprint pressure of 3.9,

2589 within threshold limits presented here sufficient to imply likelihood of suitable habitat for

2590 leopard and Spotted hyena.

123

2591 Table 4.2: Mean human footprint pressures by Wildlife Managed Areas 2592 against Species Occurrence Area Mean Occurrence Area IUCN Ha HFP Leopard Lion Hyena Kafue National Park II 206,314 1.1 Yes Yes Yes Mulobezi GMA VI 347,481 1.2 Yes Yes Yes Sichifulo GMA VI 133,734 2.2 Yes Yes Yes Nachitwe Forest Reserve unreported 71,075 2.3 Yes Yes Yes Martin Forest Reserve unreported 62,948 2.4 Yes Yes Yes Nyawa West unreported 57,439 3.7 Yes No Yes Simalaha Conservancy unreported 181,936 4.6 No No No Open Area extension 1* unreported 32,712 3.9 Yes? No? Yes? Open Area extension 1+2* unreported 73,660 3.9 Yes? No? Yes? Nyawa East unreported 39,565 4.6 No No No Machili Forest Reserve unreported 49,269 5.6 Yes No Yes Simalaha Sanctuary E unreported 11,020 5.8 No No No Simalaha Sanctuary W unreported 11,282 6.8 No No No 2593 *proposed extensions to protected area network within HFP thresholds limits for selected species 2594

2595 4.4 Discussion

2596 Human footprint pressure modelling has traditionally been undertaken at global scales, and

2597 typically at low spatial resolution, characteristic of available datasets (Sanderson et al.,

2598 2002; Leu et al., 2008). Where analyses have sought to overcome resolution constraints

2599 through availability of finer scale data, rescaling has facilitated improved accuracy and

2600 applications for conservation planning (Woolmer et al., 2008), including deriving impacts

2601 of human pressure at more appropriate site-and species-specific scales where proxy utility

2602 as a conservation tool is most valuable (Trombulak et al., 2010). Our study takes this

2603 approach a step further, successfully overcoming limitations to Venters et al’s. (2016)

2604 existing model by generating and integrating site-specific, multiple high-resolution data

2605 sets at two orders of magnitude finer scale, then applying it directly to key questions

2606 surrounding the impacts of human footprint pressure on large carnivores throughout a

124

2607 network of wildlife managed areas under varying degrees of human footprint pressure

2608 representative of a key proposed corridor in the KAZA TFCA.

2609 Model output performed best for lion with lower predictive power for leopard and hyena -

2610 species known for intrinsic ecological traits and behavioural plasticity enabling greater

2611 coexistence in human dominated landscapes than lion, which exhibits very high sensitivity

2612 to human disturbance (Riggio et al., 2018). Site-and species-specific sensitivity to complex,

2613 interrelated human disturbance variables is hypothesised to explain why the best

2614 performing model output was for the species most sensitive to human pressure.

2615 While there are limits to the predictive power of single variable models, predictive power

2616 of this model would likely benefit from supplemental data layers, when available, notably

2617 interference and exploitative pressures of pastoralism and (legal and illegal) wildlife

2618 consumption (Everatt et al., 2019), and synergistic effects between human behaviour and

2619 climate change (Brodie, 2016). There is also debate over the appropriate scale or extent at

2620 which to measure human footprint pressure, whether that be at the population level,

2621 proportion of species total range, home range or other scales (Di Marco et al., 2013).

2622 Additionally, there is scope for a greater understanding of site-and species-specific pressure

2623 score calibrations, including impacts of formal and informal road linkages (Woolmer et al.,

2624 2008).

2625 Model response curves and secondary explorative analyses of thresholds at which species

2626 are extirpated at the wildlife managed area scale closely match global mammalian human

2627 footprint pressure extinction thresholds (Di Marco et al., 2018) along with Riggio et al’s.

2628 (2018) sensitivity analysis of African large mammals with high susceptibility to human

2629 disturbance. Collectively these data provide compelling evidence that human footprint

125

2630 pressure scores ranging between 2.4-4.6 represent a threshold limit for these three species

2631 of large carnivores beyond which they are unlikely to persist in human-dominated

2632 landscapes using Venter et al’s. (2016) existing pressure score methodology. The

2633 persistence of leopard and hyena in Machili Forest, with a mean human footprint pressure

2634 score of 5.6, is likely explained by proximity of this area to extensive lower human

2635 footprint pressure areas closer to Kafue National Park, considered the landscapes core

2636 wildlife managed area. In this regard Machili Forest could be characterised as a threshold

2637 area or an attractive sink, limiting range expansion of these species to broader areas with

2638 lower human footprint pressure.

2639 The identification of potential additions to the protected area network in Open Areas east of

2640 the Simalaha, first suggested in Lines et al. (2018), and further posited here, serves two-

2641 fold purposes: 1. The increase in wildlife habitat for a range of species and 2. The likely

2642 increase in scope and scale for connectivity between Kafue National Park and the Zambezi

2643 River for both leopard and spotted hyena in areas of low human habitation and agricultural

2644 development, limiting scope for negative interactions (human-wildlife conflict) in the

2645 otherwise increasingly human dominated landscapes that characterises the central-southern

2646 extents of the Kafue-Zambezi interface.

2647 These site-specific, high resolution maps have broad utility as a baseline against which

2648 subsequent changes to human footprint pressure can be mapped and modelled over time as

2649 more data sets come available to refine this iterative process. Pressure score calibration

2650 merits more explicit treatment to improve model response given species and/or process of

2651 interest.

126

2652 The value and applicability of generating standardised approaches to mapping human

2653 footprint pressure surrounds their use as a proxy or indicator of broader drivers impacting

2654 habitat degradation, ecosystem function, species loss, or potential for species recovery

2655 which presents new conservation challenges surrounding e.g. human-wildlife conflict.

2656 Progress with human footprint pressure modelling depends in part on understanding and

2657 addressing limitations and assumptions of model development (Halpern & Fujita, 2013),

2658 and recognition of the dynamic nature of human pressure in terms of asymmetrical and

2659 nonlinear threshold responses to total footprint pressure changes across spatial-temporal

2660 scales (Toews, 2016; Venter et al., 2016).

2661

2662 4.5 Conclusion

2663 Our human footprint pressure analyses demonstrate utility for determining habitat

2664 suitability for a suite of large carnivores of conservation value, including predicting species

2665 persistence, extirpation and potential for recolonization, even at this preliminary, proof-of-

2666 concept stage of model development.

2667 Model output broadly follows existing data in support of understanding human pressure

2668 impacts on landscape-level connectivity at the Kafue-Zambezi interface, providing a

2669 valuable additional tool in conservation planning for this landscape, the broader KAZA

2670 region and beyond.

2671 As additional data layers become available and the pressure score calibration process

2672 evolves, site-specific human pressure maps can be expanded to the broader Zambian and

127

2673 KAZA landscape to model how spatiotemporal human pressure impacts species and

2674 processes of interest to key Programme objectives.

2675

2676

2677

2678

2679

2680

2681

2682

2683

2684

2685

2686

2687

2688

2689

2690

2691 128

2692 References

2693 Ahmadi, M., et al. (2017). Combining landscape suitability and habitat connectivity to

2694 conserve the last surviving population of cheetah in Asia. Diversity and Distributions,

2695 23(6), 592-603.

2696 Aiello-Lammens ME, Boria RA, Radosavljevic A, Vilela B (2015). spThin: an R package

2697 for spatial thinning of species occurrence records for use in ecological niche models.

2698 Ecography, 38(5), 541–545.

2699 Angelieri, C. C. S., et al. (2016). Using species distribution models to predict potential

2700 landscape restoration effects on puma conservation. PLoS One, 11(1), e0145232.

2701 Ayram, C. A. C., et al. (2017). Anthropogenic impact on habitat connectivity: A

2702 multidimensional human footprint index evaluated in a highly biodiverse landscape of

2703 Mexico. Ecological Indicators, 72, 895-909.

2704 Bonafilia, D., et al. (2019). Building High Resolution Maps for Humanitarian Aid and

2705 Development with Weakly-and Semi-Supervised Learning. Proceedings of the IEEE

2706 Conference on Computer Vision and Pattern Recognition Workshops, 1-9.

2707 Brodie, J. F. (2016). Synergistic effects of climate change and agricultural land use on

2708 mammals. Frontiers in Ecology and the Environment, 14(1), 20-26.

2709 Brooks, T. M., et al. (2006). Global biodiversity conservation priorities. Science (New

2710 York, N.Y.), 313(5783), 58-61.

2711 Calvert, G. M. (2005). Sitimela: A History of the Zambesi Sawmills Logging Railway,

2712 1911-1972. Barotse Development Trust.

129

2713 Ceballos, G., et al. (2015). Accelerated modern human–induced species losses: Entering the

2714 sixth mass extinction. Science Advances, 1(5), e1400253.

2715 Ceballos, G. and Ehrlich, P. R. (2002). Mammal population losses and the extinction crisis.

2716 Science (New York, N.Y.), 296(5569), 904-907.

2717 CSO (2019). Central Statistics Office of Zambia. Available at:

2718 http://www.zamstats.gov.zm/ Accessed 23.09.2019.

2719 Cumming, D. (2011). Constraints to conservation and development success at the wildlife-

2720 livestock-human interface in southern African transfrontier conservation areas: a

2721 preliminary review. Wildlife Conservation Society.

2722 Di Marco, M., et al. (2013). Comparing multiple species distribution proxies and different

2723 quantifications of the human footprint map, implications for conservation. Biological

2724 Conservation, 165, 203-211.

2725 Di Marco, M. and Santini, L. (2015). Human pressures predict species’ geographic range

2726 size better than biological traits. Global Change Biology, 21(6), 2169-2178.

2727 Di Marco, M., et al. (2018). Changes in human footprint drive changes in species extinction

2728 risk. Nature Communications, 9(1), 4621.

2729 Di Minin, E., et al. (2013). Conservation businesses and conservation planning in a

2730 biological diversity hotspot. Conservation Biology, 27(4), 808-820.

2731 Duffy, R. (2006). Global governance and environmental management: the politics of

2732 transfrontier conservation areas in Southern Africa. Political Geography, 25, 89-112.

130

2733 Elith, J., et al. (2011). A statistical explanation of MaxEnt for ecologists. Diversity and

2734 Distributions, 17(1), 43-57.

2735 Estes, J. A., et al. (2011). Trophic downgrading of planet Earth. Science (New York, N.Y.),

2736 333(6040), 301-306.

2737 Everatt, K. T., et al. (2019). Africa's apex predator, the lion, is limited by interference and

2738 exploitive competition with humans. Global Ecology and Conservation, e00758.

2739 Foley, J. A., et al. (2005). Global consequences of land use. Science (New York, N.Y.),

2740 309(5734), 570-574.

2741 Funston, P. J., et al. (2013). Insights into the management of large carnivores for profitable

2742 wildlife-based land uses in African savannas. PloS One, 8(3), e59044.

2743 Gaynor, K. M., et al. (2018). The influence of human disturbance on wildlife nocturnality.

2744 Science (New York, N.Y.), 360(6394), 1232-1235.

2745 Geldmann, J., et al. (2014). Mapping change in human pressure globally on land and within

2746 protected areas. Conservation Biology, 28(6), 1604-1616.

2747 Gren, M., et al. (2018). Economics of wildlife management—an overview. European

2748 Journal of Wildlife Research, 64(2), 22.

2749 Halpern, B. S. and Fujita, R. (2013). Assumptions, challenges, and future directions in

2750 cumulative impact analysis. Ecosphere, 4(10), 1-11.

2751 Jackson, C. R., et al. (2016). Evaluating habitat connectivity methodologies: a case study

2752 with endangered African wild dogs in South Africa. Landscape Ecology, 1-15.

131

2753 Jones, K. R., et al. (2018). One-third of global protected land is under intense human

2754 pressure. Science (New York, N.Y.), 360(6390), 788-791.

2755 KAZA (2018). The Kavango Zambezi Transfrontier Conservation Area Carnivore

2756 Conservation Strategy. KAZA TFCA Secretariat, Hwange, Zimbabwe.

2757 KAZA (2014). Master Integrated Development Plan. KAZA Secretariat. Victoria Falls.

2758 KAZA (2011a). Treaty Between the Governments of Angola, Botswana, Namibia, Zambia

2759 and Zimbabwe on the Establishment of the Kavango–Zambezi Transfrontier

2760 Conservation Area. Kavango–Zambezi TFCA Secretariat, Kasane, Botswana.

2761 Leu, M., et al. (2008). The human footprint in the west: a large‐scale analysis of

2762 anthropogenic impacts. Ecological Applications, 18(5), 1119-1139.

2763 Lindsey, P., et al. (2014). Underperformance of the Zambian protected area network: steps

2764 needed to improve functionality and effectiveness. PloS One, 9(5), e94109.

2765 Lines, R., et al. (2018). Status of terrestrial mammals at the Kafue–Zambezi interface:

2766 implications for transboundary connectivity. Oryx, 1-10.

2767 Maxwell, S. L., et al. (2016). Biodiversity: The ravages of guns, nets and bulldozers.

2768 Nature News, 536(7615), 143.

2769 Musgrave, M. (2016). Scale, Governance and Change in Zambezi Teak Forests:

2770 Sustainable Development for Commodity and Community. Cambridge Scholars

2771 Publishing.

2772 Navarro, L. M. and Pereira, H. M. (2015). Rewilding abandoned landscapes in Europe. In:

2773 Rewilding European Landscapes. Springer, Cham, 3-23.

132

2774 Newmark, W. D. (2008). Isolation of African protected areas. Frontiers in Ecology and the

2775 Environment, 6(6), 321-328.

2776 NOAA (2019). National Oceanic and Atmospheric Administration. National Geophysical

2777 Data Center. DMSP data collected by the US Air Force Weather Agency. Accessed

2778 15/09/2019.

2779 Phillips, S. J. and Dudík, M. (2008). Modelling of species distributions with Maxent: new

2780 extensions and a comprehensive evaluation. Ecography, 31(2), 161-175.

2781 QGI (2018). QGIS Geographic Information System. Open Source Geospatial Foundation

2782 Project. http://qgis.osgeo.org.

2783 R Core Team (2018). A language and environment for statistical computing. R Foundation

2784 for Statistical Computing, Vienna, Austria. https://www.R-project.org/

2785 Rees, W., et al. (1996). Our Ecological Footprint: Reducing Human Impact on the Earth.

2786 New Society Publishers Gabriola Island, BC.

2787 Rich, C. and Longcore, T. (2013). Ecological Consequences of Artificial Night Lighting.

2788 Island Press.

2789 Riggio, J., et al. (2018). Sensitivity of Africa’s larger mammals to humans. Journal for

2790 Nature Conservation, 43, 136-145.

2791 Riitters, K., et al. (2000). Global-scale patterns of forest fragmentation. Conservation

2792 Ecology, 4(2).

2793 Ripple, W. J., et al. (2014). Status and ecological effects of the world’s largest carnivores.

2794 Science, 343(6167), 1241484.

133

2795 Sanderson, E. W., et al. (2002). The human footprint and the last of the wild: the human

2796 footprint is a global map of human influence on the land surface, which suggests that

2797 human beings are stewards of nature, whether we like it or not. Bioscience, 52(10),

2798 891-904.

2799 Sutherland, W. J., et al. (2004). The need for evidence-based conservation. Trends in

2800 Ecology & Evolution, 19(6), 305-308.

2801 Toews, M. (2016). Managing Human Footprint with Respect to its Effects on Large

2802 Mammals: Implications of Spatial Scale, Divergent Responses and Ecological

2803 Thresholds. Doctoral Dissertation.

2804 Trombulak, S. C., et al. (2010). The human footprint as a conservation planning tool. In:

2805 Landscape-Scale Conservation Planning. Springer, 281-301.

2806 Tucker, M. A., et al. (2018). Moving in the Anthropocene: Global reductions in terrestrial

2807 mammalian movements. Science (New York, N.Y.), 359(6374), 466-469.

2808 UNEP-WCMC (2015). United Nations Environment Programme's World Conservation

2809 Monitoring Centre. Protected Planet Database https://www.protectedplanet.net/

2810 Accessed 01.05.2015.

2811 Venter, O., et al. (2016). Global terrestrial Human Footprint maps for 1993 and 2009.

2812 Scientific Data, 3, 160067.

2813 Vitousek, P. M., et al. (1997). Human domination of Earth's ecosystems. Science,

2814 277(5325), 494-499.

134

2815 Wackernagel, M., et al. (2002). Tracking the ecological overshoot of the human economy.

2816 Proceedings of the National Academy of Sciences of the United States of America,

2817 99(14), 9266-9271.

2818 Walker, B. and Salt, D. (2012). Resilience Thinking: Sustaining Ecosystems and People in

2819 a Changing World. Island Press.

2820 Watson, J. E., et al. (2016). Catastrophic declines in wilderness areas undermine global

2821 environment targets. Current Biology, 26(21), 2929-2934.

2822 Wittemyer, G., et al. (2008). Accelerated human population growth at protected area edges.

2823 Science (New York, N.Y.), 321(5885), 123-126.

2824 Woolmer, G., et al. (2008). Rescaling the human footprint: a tool for conservation planning

2825 at an ecoregional scale. Landscape and Urban Planning, 87(1), 42-53.

2826

2827

2828

2829

2830

2831

2832

2833

2834

135

2835 Chapter 5: Final Discussion

2836 5.1 Recapitulation of Study Scope & Key Findings

2837 The KAZA TFCA promotes landscape connectivity between core and periphery wildlife

2838 managed area clusters, both within and between five bordering countries. The Greater

2839 Kafue System represents Zambia’s majority component in this TFCA but narratives

2840 surrounding connectivity have been promoted largely in the absence of any systematic

2841 research or empirical evidence of species-level connectivity prior to this study.

2842 Given the dearth of quantifiable evidence in support of connectivity narratives this study

2843 sought to delve into key questions surrounding the status and argument for landscape

2844 connectivity at the large carnivore scale between Kafue National Park and the Zambezi

2845 River, bordering adjacent wildlife managed areas in Namibia and Botswana.

2846 I approached the study from a holistic perspective, building an argument for connectivity

2847 based on accumulating contemporary evidence while considering historical data and

2848 information compiled from a range of published and grey sources.

2849 Historical data suggests widespread occurrence of the intact large carnivore guild and

2850 majority common prey species throughout the Kafue-Zambezi landscape until emergence

2851 of widespread commercial hunting in late Nineteenth century, followed by further wildlife

2852 declines caused by Rinderpest epizootic, expansion of commerce, transport infrastructure,

2853 human settlement and associated human disturbance. Spillover of impacts from Regional

2854 border wars further increased pressure on wildlife following Zambia’s Independence in

2855 1964 at a time when wildlife conservation resources declined.

136

2856 In recent decades wildlife conservation budgets have remained significantly below

2857 minimum requirements throughout this landscape, and near zero in many wildlife managed

2858 areas. In concert with widespread increases in human settlement, infrastructure and agro-

2859 pastoralism activities high quality wildlife habitat has reduced and fragmented.

2860 Wildlife occurrence and distribution has declined further, especially in southern wildlife

2861 managed areas. There were no quantifiable data on any large carnivores occurrence within

2862 60km of the Zambezi river, and large scale reductions in common prey assemblages and

2863 distribution.

2864 While large carnivore mobility implies scope for movements between Kafue National Park

2865 and the Zambezi river, and pockets of suitable habitat exist, we conclude that there are no

2866 resident large carnivores throughout large areas of the Kafue-Zambezi interface.

2867 Although somewhat beyond the scope of this study, I was unable to establish quantifiable

2868 or anecdotal records for any large carnivore movements across the Zambezi river.

2869

2870 5.2 Contributions to conservation science

2871 5.2.1 Chapter 2: Species Assemblages and Distribution

2872 The absence of accurate baseline data and metrics limits capacity to effectively monitor and

2873 evaluate conservation management performance (Sutherland et al., 2004; Hockings et al.,

2874 2000). Existing area-specific literature (PPF, 2008) failed to detect or accurately quantify

2875 the large mammal assemblage from Kafue National Park through the Zambezi River.

2876 Resulting outputs effectively provided presence only data for limited species, from limited

137

2877 areas, and misclassified the extent of the range for many species of conservation concern,

2878 including the extant large carnivore guild.

2879 While no systematic research apart from this study has been undertaken on carnivores and

2880 their prey throughout this landscape, larger scale research on Cheetah (Durant et al., 2017)

2881 and African wild dog (RWCP & IUCN/SSC, 2015) failed to incorporate known range

2882 beyond Kafue National Park. Findings from Stein et al (2016) and ZAWA (2009)

2883 approximates findings from this study for Leopard and Lion distribution. Current

2884 reassessment of Spotted Hyena status (Weise et al., in prep) utilises data from this study.

2885 Our published research compiled the most comprehensive data available on five large

2886 carnivores and 26 principle prey species against which performance of existing and

2887 subsequent conservation interventions throughout ten wildlife managed areas can be

2888 monitored and evaluated over time to ascertain changes to species assemblage and

2889 distribution by wildlife managed area. Management implications concern species-level

2890 connectivity throughout the Kafue-Simalaha Wildlife Dispersal Area and longer-term

2891 persistence of large carnivores throughout the Greater Kafue System as a driver of balanced

2892 and resilience food webs and wildlife-based land uses.

2893 5.2.2 Chapter 3: Landcover Mapping, Habitat Suitability and Connectivity Modelling

2894 The life history traits of large carnivores make them highly susceptible to human

2895 disturbance (Crooks, 2002). Increasing human disturbance within and surrounding wildlife

2896 habitat is having devastating effects on these species (Maxwell et al., 2016; Venter et al.,

2897 2016), with significant secondary effects throughout food webs (Ripple et al., 2014; Estes

2898 et al., 2011) hindering the development of wildlife-based land uses (Funston et al., 2013).

138

2899 With the advent and availability of high-resolution satellite imagery and associated

2900 analytical software there has been a revolution in the practical value and application of

2901 remote sensing for conservation science and planning (de Klerk & Buchanan, 2017). These

2902 developments are expanding our knowledge and understanding of drivers impacting species

2903 distribution and persistence throughout increasingly human-impacted environments (Kerr

2904 & Ostrovsky, 2003; Pettorelli et al., 2005; Watson et al., 2015).

2905 By incorporating extensive ground-truthing and the latest Sentinel imagery we were able to

2906 develop high resolution landcover maps for 32,000km2, covering the majority of the Kafue-

2907 Zambezi interface and all ten wildlife managed areas within the proposed Kafue-Simalaha

2908 Wildlife Dispersal Corridor at over 91% accuracy. This output clearly shows the extent and

2909 spatial characteristics of human activities on the landscape, both within formal wildlife

2910 managed areas and peripheral Open Communal Areas. Agriculture, settlement, transport

2911 links and associated human disturbance pressures are pervasive wherever permanent and

2912 seasonal water access permits settlement and grazing camps in all but the most remote

2913 Game Management Areas and Kafue National Park.

2914 This landcover map represents the highest current resolution against which all subsequent

2915 landcover analyses will be measured and provides a valuable tool for land use planning

2916 across for a range of stakeholder activities including zoning areas for wildlife and tourism

2917 development, agricultural expansion, forestry and species-level conservation planning.

2918 Subsequent habitat suitability and current flow models indicate priority areas required to

2919 maintain species persistence and connectivity throughout this landscape and highlights the

2920 value of unprotected (Open) areas for large carnivore conservation (RWCP & IUCN/SSC,

2921 2015). Linkage Mapper corridor routing performed largely to expectations, and broadly

139

2922 complimented existing analyses (Cushman et al., 2018 & 2016), indicating narrowing and

2923 sub-optimal current flow in southern areas of the proposed Kafue-Zambezi Wildlife

2924 Dispersal Corridor, both at single and multi-species scales. These findings provide

2925 management with targeted areas for conservation intervention to increase species-level

2926 persistence and movement throughout the landscape.

2927 5.2.3 Chapter 4: Human footprint pressure

2928 Acknowledging the near-ubiquitous human impacts on natural resources within and

2929 bordering protected areas, the development and refinement of functional proxies predicting

2930 species response to human disturbance represents a valuable development in conservation

2931 biology (Zipkin et al., 2010; Di Marco et al., 2014). Importantly, novel methods to monitor

2932 and evaluate human impacts on habitats and species of interest over time provide planners

2933 and managers with valuable tools to assess and the effectiveness of conservation

2934 interventions and direct resources

2935 I tested the utility of this approach through analysis of large carnivore occurrence to a

2936 composite measure of human footprint pressure based on existing methodologies

2937 (Sanderson et al., 2002; Venter et al., 2016). My site-calibrated datasets utilised the highest

2938 available data resolution, overcoming much limitation in existing analyses.

2939 Outputs provided a striking visual representation of human pressure throughout 33,000km2

2940 of wildlife managed areas and rangelands comprising the majority Zambian component of

2941 the KAZA TFCA, highlighting both the extent and intensity of pressures, and facilitating

2942 finer scale analyses against which to model species response.

140

2943 Response of species to human footprint pressure at the wildlife managed area scale expands

2944 our understanding of thresholds at which species can exist in increasingly human-

2945 dominated landscapes at scales more appropriate to understanding species persistence.

2946 Encouragingly my human footprint pressure mapping analyses produced outputs broadly

2947 similar to my landcover, habitat suitability and current flow analyses. Notably I identified

2948 important areas for addition to the protected area network in the Kafue-Zambezi Wildlife

2949 Dispersal Corridor that incorporated human footprint pressure thresholds below which large

2950 carnivores might reasonably occur, and concurrently are not utilized extensively by local

2951 communities.

2952 Our proof-of-concept analysis is pertinent in context of trends in human pressure and

2953 resource demands within and surrounding wildlife managed areas (Wittemyer et al., 2008),

2954 and given challenges surrounding the maintenance, integrity and connectedness of African

2955 protected areas (Newmark, 2008).

2956

2957 5.3 Methodological Insights / Limitations

2958 Integration of qualitative and quantitative data sources provides a robust approach to

2959 generating species-level occurrence data given sufficient understanding of in situ field

2960 conditions and survey approaches required to cover a diverse mosaic of land uses and

2961 species at a large scale in a single sampling season.

2962 Initial attempts to operationalise the multi-species, multi-scale, Bayesian Hierarchical

2963 models of Whittington et al. (2015) were frustrated by imprecise parameter estimates from

2964 the model, likely derived from coding problems associated with modifying the existing

141

2965 framework from a single species to multi-species approach. Recent statistical developments

2966 (Petracca et al., 2019) appear to have extended the framework to account for multiple

2967 species, and, in doing so, mitigate estimation imprecision through shrinkage (the borrowing

2968 of statistical strength from a community level distribution).

2969 Notwithstanding initial modelling challenges, data was sufficiently robust to facilitate

2970 application of the alternative MaxEnt and Linkage Mapper approaches which produced

2971 results largely in line with expectations and related studies (Cushman et al., 2016 & 2018).

2972 Human footprint pressure mapping will benefit from supplemental layers as they become

2973 available, and more in-depth treatment of site-specific pressure variable calibrations and the

2974 appropriate scale at which to model species occurrence. There is also scope for approaching

2975 subsequent model analyses from an occupancy perspective (e.g. Petracca et al., 2019),

2976 incorporating detect/non-detection data sets versus simple occurrence.

2977 Additionally, pressures for which we have no local data, but are known to exert significant

2978 negative effects on habitat suitability (e.g. poaching and direct/indirect impacts of

2979 pastoralism and its spatial characteristics) will likely generate addictive pressures limiting

2980 habitat suitability. This implies model outputs are conservative and habitat suitability is

2981 over-estimated.

2982 Notwithstanding model performance, the responses of large carnivores to individual and

2983 cumulative human pressures is likely dynamic and non-linear, so we can expect a

2984 continuum of species-level responses to modelled pressures across the landscape over time

2985 as interacting human pressures change.

2986

142

2987 5.4 Broader Contribution of the Study to Literature

2988 5.4.1 Transboundary conservation

2989 While generalities surrounding the impacts of dynamic human pressures on wildlife are

2990 increasingly understood, the identification and calibration of human pressures at site-and

2991 species specific scales is a novel direction in applied research. My approach reflects real-

2992 world challenges in natural resource management throughout protected areas mosaics under

2993 varying degrees of conservation effectiveness. In the case of the KAZA-TFCA existing

2994 narratives surrounding landscape connectivity have failed to identify site-and species

2995 specific challenges within proposed landscape-level corridors and wildlife dispersal areas.

2996 These constraints limit practical application of conservation science to promote biodiversity

2997 conservation and the development of wildlife-based land uses. The strategic approach

2998 offered in this thesis and accompanying publications provides a repeatable and robust

2999 technique for understanding localised questions surrounding species persistence, extirpation

3000 and potential recolonization in key linkages throughout transboundary networks.

3001 5.4.2 Protected areas

3002 Mosaics of protected areas under varying conservation management intensity and

3003 effectiveness present a range of challenges for species persistence and movement, with few

3004 existing areas in Southern Africa designated with these specific considerations in mind.

3005 Given our approach and findings, clear utility exists for expanding protected areas to

3006 incorporate contemporary concerns, and scalable to other protected areas mosaics in the

3007 KAZA-TFCA and beyond where varying human pressures present site-specific challenges

3008 to wider conservation objectives.

3009

143

3010 5.4.3 Understudied landscapes

3011 The value of many understudied landscapes lies in their potential roles for supporting key

3012 aspects of biodiversity conservation and natural economies, including the persistence,

3013 recolonization and movement of key wildlife species between and within protected areas

3014 mosaics. The conclusions presented in this work capitalises on my novel integration of

3015 multi-methods throughout formerly unstudied landscapes, raising issues related to existing

3016 connectivity narratives whilst presenting new options to fulfil said objectives.

3017 5.4.4 Local issues

3018 Complexities surrounding inter alia land use ownership, livelihood practices and

3019 sensitivities of the local communities and their leaders to biodiversity conservation cannot

3020 be addressed by board brush approaches that fail to understand and integrate multi-faceted

3021 local issues into conservation and development planning. Typically, the cumulative

3022 pressures exerted by human populations at local scales vary over space and time, and this

3023 research has emphasised the value of understanding and capturing these nuances at a scale

3024 appropriate to planning for large carnivore conservation.

3025

3026

3027

3028

3029

3030 144

3031 References

3032 Ansell, W. F. H. (1978). The Mammals of Zambia. National Parks & Wildlife Service.

3033 Crooks, K. R. (2002). Relative sensitivities of mammalian carnivores to habitat

3034 fragmentation. Conservation Biology, 16(2), 488-502.

3035 Cushman, S.A., et al. (2018). Prioritizing core areas, corridors and conflict hotspots for lion

3036 conservation in southern Africa. PloS one, 13(7), p.e0196213.

3037 Cushman, S. A., et al. (2016). A multi-scale assessment of population connectivity in

3038 African lions (Panthera leo) in response to landscape change. Landscape Ecology,

3039 31(6), 1337-1353.

3040 de Klerk, H. M. and Buchanan, G. (2017). Remote sensing training in African conservation.

3041 Remote Sensing in Ecology and Conservation, 3(1), 7-20.

3042 Di Marco, M., et al. (2014). A retrospective evaluation of the global decline of carnivores

3043 and ungulates. Conservation Biology, 28(4), 1109-1118.

3044 Durant, S. M., et al. (2017). The global decline of cheetah Acinonyx jubatus and what it

3045 means for conservation. Proceedings of the National Academy of Sciences, 114(3),

3046 528-533.

3047 Estes, J. A., et al. (2011). Trophic downgrading of planet Earth. Science (New York, N.Y.),

3048 333(6040), 301-306.

145

3049 Funston, P. J., et al. (2013). Insights into the management of large carnivores for profitable

3050 wildlife-based land uses in African savannas. PloS One, 8(3), e59044.

3051 Gann, L. H. (1969). A History of Northern Rhodesia: Early Days to 1953. Humanities

3052 Press.

3053 Gibson, C. C. (1999). Politicians and Poachers: The Political Economy of Wildlife Policy

3054 in Africa. Cambridge University Press.

3055 Hockings, M. (2006). Evaluating Effectiveness: A Framework for Assessing Management

3056 Effectiveness of Protected Areas. IUCN, Gland, Switzerland.

3057 Kerr, J. T. and Ostrovsky, M. (2003). From space to species: ecological applications for

3058 remote sensing. Trends in Ecology & Evolution, 18(6), 299-305.

3059 MacKenzie, J. M. (1997). The Empire of Nature: Hunting, Conservation and British

3060 Imperialism. Manchester University Press.

3061 Macmillan, H. and Hugh, M. (2005). An African Trading Empire: The Story of Susman

3062 Brothers & Wulfsohn, 1901-2005. Vol. 16. IB Tauris.

3063 Martelli, G. (1970). Livingstone's River: A History of the Zambezi Expedition, 1858-1864.

3064 Simon & Schuster.

3065 Maxwell, S. L., et al. (2016). Biodiversity: The ravages of guns, nets and bulldozers.

3066 Nature News, 536(7615), 143.

146

3067 Newmark, W. D. (2008). Isolation of African protected areas. Frontiers in Ecology and the

3068 Environment, 6(6), 321-328.

3069 Petracca, L., et al. (2019). Modeling community occupancy from line transect data: a case

3070 study with large mammals in post‐war Angola. Animal Conservation, 23(4), 420-433.

3071 Pettorelli, N., et al. (2005). Using the satellite derived NDVI to assess ecological responses

3072 to environmental change. Trends in Ecology & Evolution, 20(9), 503-510.

3073 PPF (2008). Peace Parks Foundation. Integrated Development Plan: Zambian Component

3074 of the KAZA TFCA, Stellenbosch, South Africa, xiii + 153.

3075 Ripple, W. J., et al. (2014). Status and ecological effects of the world’s largest carnivores.

3076 Science, 343(6167), 1241484.

3077 Roeder, P. L., et al. (2006). Rinderpest in the twentieth and twenty-first centuries. In:

3078 Rinderpest and Peste Des Petits Ruminants. Elsevier, 105-VII.

3079 RWCP & IUCN/SSC (2015). Regional Conservation Strategy for the Cheetah and African

3080 Wild Dog in Southern Africa; Revised and Updated, August 2015.

3081 Sampson, R. (1972). The Man with a Toothbrush in His Hat: The Story and Times of

3082 George Copp Westbeech in Central Africa. Multimedia Publications.

3083 Sanderson, E. W., et al. (2002). The human footprint and the last of the wild: the human

3084 footprint is a global map of human influence on the land surface, which suggests that

3085 human beings are stewards of nature, whether we like it or not. Bioscience, 52(10),

3086 891-904. 147

3087 Stein, A., et al. (2017). Panthera pardus (errata version published in 2016). The IUCN Red

3088 List of Threatened Species 2016: e. T15954A102421779. Accessed 14/11/2019.

3089 Sutherland, W. J., et al. (2004). The need for evidence-based conservation. Trends in

3090 Ecology & Evolution, 19(6), 305-308.

3091 Suzman, J. (2001). An Introduction to the Regional Assessment of the Status of the San in

3092 Southern Africa. Legal Assistance Centre Windhoek.

3093 Venter, O., et al. (2016). Global terrestrial Human Footprint maps for 1993 and 2009.

3094 Scientific Data, 3, 160067.

3095 Watson, F. G., et al. (2015). Human encroachment into protected area networks in Zambia:

3096 implications for large carnivore conservation. Regional Environmental Change, 15(2),

3097 415-429.

3098 Whittington, J., et al. (2015). Estimating occupancy using spatially and temporally

3099 replicated snow surveys. Animal Conservation, 18(1), 92-101.

3100 Wittemyer, G., et al. (2008). Accelerated human population growth at protected area edges.

3101 Science (New York, N.Y.), 321(5885), 123-126.

3102 ZAWA (2009). Conservation Strategy and Action Plan for the African Lion Panthera leo.

3103 Zambia Wildlife Authority. Chilanga.

3104 Zipkin, E. F., et al. (2010). Multi-species occurrence models to evaluate the effects of

3105 conservation and management actions. Biological Conservation, 143(2), 479-484.

148

3106 Appendix: Methodological Explanations / Exploratory analysis / Data manipulation

3107 1: Spoor Tracking Surveys

3108 A number of key constraints were identified in developing an appropriate survey method:

3109 • Cryptic nature of target species;

3110 • Unknown status and ranging behaviour of target species throughout study area;

3111 • Large spatial scale of the survey area;

3112 • Requisite requirement to sample the entire area in a single season;

3113 • Limited and inconsistent network of access roads and tracks;

3114 • Variable substrate conditions;

3115 • Security and other logistics issues.

3116

3117 To overcome constraints a planning/pilot study was developed to map all navigable tracks

3118 and assess substrate condition, to determine if vehicle or foot based surveys would be

3119 appropriate, and also sampling intensity required to detect target species and cover the

3120 entire study area in a single season. Necessary trade-offs were acknowledged.

3121 With the assistance of State wildlife officers and hunting safari operators we mapped the

3122 available road and track network by 4x4 vehicle and on foot and entered into a GIS. The

3123 network was then categorized as roads or tracks and further sub-categorized as graded

3124 (ideal substrate for spoor tracking), ungraded (functional with/without additional

3125 preparation) or overgrown (unusable for tracking). Using expert opinion it was clear

3126 insufficient graded roads precluded vehicle-based surveys.

149

3127 Taking 400 km2 as the approximate survey block area considered sufficient to cover home

3128 range of target species from IUCN Redlist data, I overlaid a 20x20 km grid on our

3129 road/track map, providing 31 blocks overlapping the study area. I then selected 2 sample

3130 blocks with >25 km of existing graded roads to determine survey effort required to detect

3131 target species – the minimum distance we could reasonably cover on foot during 3-4 hours

3132 of spoor tracking after dawn and again before dusk – optimum times for detecting tracks

3133 when the sun was not overhead and shadows highlighted fresh tracks.

3134 In each sample block 8 x 4 km tracks were spread throughout the road/track network on

3135 optimal substrate and walked by two experienced trackers and the project manager at 4

3136 km/hr, noting every large carnivore and common prey track within each 0.5 km sub-

3137 section. Tracks were marked so as not to be recounted on subsequent counts.

3138 While cheetah were known from the landscape from other sources they were not detected in

3139 the pilot study and omitted from formal analyses.

3140 Detection probability was then calculated for the remaining 4 large carnivore species at 0.5

3141 km sub-sections to determine optimal survey length and no of transect repeats required to

3142 optimize detection of target species. Data indicated 4 km transects surveyed three times

3143 would be sufficient to reach a P(det) of >0.61 for lion, leopard and spotted hyena. African

3144 wild dog were infrequently detected and omitted from further spoor tracking analyses.

3145 At this survey effort ten sample blocks could be surveyed in a single season, equating to

3146 102x4 km transects sampled three times.

150

3147 With these data we randomly selected 10 sample blocks from 22 that contained >25 km of

3148 useable roads/tracks, and overlaid transects over optimal substrate within these sample

3149 blocks (mean 40.8 km/sample block, range 28-64 kms, 7-16 transects).

3150 I optimized transect substrate with controlled burns and grading, dragging a heavy log over

3151 the track with a 4x4 until a consistent surface was achieved. Transects were then left a

3152 week before any surveys commenced. In that way I overcame limits to inconsistent surface

3153 tracking conditions and limited bias in transect selection and detection probability.

3154 Surveys commenced in the northern-most block adjacent to Kafue National Park and I

3155 worked south to the Zambezi River. 1-2 teams covered between 6-16 transects a day in

3156 each block, repeating the transects three times in each survey area over a period of 7-10

3157 days, with 2-3 days between repeat surveys.

3158

3159

3160

3161

3162

3163

3164

3165

3166

151

3167 Table 5: Summary of exploratory analyses to determine sampling strategy

Hyena Estimated coefficients

Naive occupancy estimate = 0.6471 hyaena leopard lion wild dog Individual Site estimates of psi_int 26.72 2.97 0.54 1.30 Site estimate Std.err 95% conf. interval p_int -0.35 -0.58 -2.66 -1.71 psi 1 site1 : 0.7129 0.0560 0.5922 - 0.8094 p_dist 0.29 0.42 0.78 0.31

Wild dog Parameter estimates Naive occupancy estimate = 0.1961 transect parameter hyaena leopard lion wild dog (km) Individual Site estimates of psi 1.00 0.95 0.63 0.79 Site estimate Std.err 95% conf. interval p 0.5 0.45 0.41 0.09 0.17 psi 1 site1 : 0.4452 0.1789 0.1625 - 0.7684 1 0.49 0.46 0.13 0.20

1.5 0.52 0.51 0.18 0.22 Lion 2 0.56 0.57 0.25 0.25 Naive occupancy estimate = 0.3529 2.5 0.59 0.62 0.33 0.28 Individual Site estimates of 3 0.63 0.67 0.42 0.31 Site estimate Std.err 95% conf. interval 3.5 0.66 0.71 0.52 0.35 psi 1 site1 : 0.5252 0.0988 0.3372 - 0.7062 P (det) 4 0.69 0.75 0.61 0.38

Leopard transect Number of surveys Naive occupancy estimate = 0.5980 (km) p 0.5 7 7 20 15 Individual Site estimates of 1 6 6 15 13 Site estimate Std.err 95% conf. interval 1.5 6 5 12 13 psi 1 site1 : 0.6053 0.0492 0.5059 - 0.6967 2 5 5 8 10

2.5 5 4 6 8

3 5 4 5 8

3.5 4 4 3 7

4 3 3 3 6 3168

3169

3170

3171

3172

3173

3174

3175

3176

152

3177 Appendix 2: Landcover Mapping

3178 Mapping dynamic and diverse landscapes, such as this study area, where a high degree of

3179 heterogeneity and variation is exhibited between dry and wet seasons, is challenging and the

3180 use of time series satellite data is indispensable (Lucas et al., 2007). The images were

3181 processed at Level 1C which includes a geometric correction and acquisition of Top of

3182 Atmosphere (ToA) Reflectance values. For the analysis only the 10 bands provided at spatial

3183 resolutions of 10 m (Bands 2,3,4 and 8) and 20 m (Bands 5,6,7,8a,11 and 12) were employed

3184 with the later resampled at 10 m. Prior to analysis an absolute atmospheric correction was

3185 applied using the Dark Object Subtraction algorithm (DOS). During the classification process

3186 various spectral ratios were calculated including Normalized Difference Vegetation Index

3187 (NDVI; equation 1), Normalized Difference Wetness Index (NDWI; Equation 2),

3188 Normalized Difference Moisture Index (NDMI; Equation 3), Near InfraRed/Red (NIR/Red;

3189 Equation 4).

(Band8−Band4) 3190 Equation 1: 푁퐷푉퐼 = (Band8+Band4)

(Band3−Band8) 3191 Equation 2: 푁퐷푊퐼 = (Band3+Band8)

(Band8−Band12) 3192 Equation 3: 푁퐷푀퐼 = (Band8+Band12)

Band8 3193 Equation 4: 푁퐼푅/푅퐸퐷 = Band4

3194

3195

3196

153

3197 Table 6: Sentinel 2 Spectral Bands and central wavelengths

Band Number Central Wavelength (nm) Spatial Resolution (m) Band 1 443 60 Band 2 490 10 Band 3 560 10 Band 4 665 10 Band 5 705 20 Band 6 740 20 Band 7 783 20 Band 8 842 10 Band 8a 865 20 Band 9 940 60 Band 10 1375 60 Band 11 1610 20 Band 12 2190 20 3198

3199 Various post-classification refinements were employed in order to avoid a noise in the final

3200 product. Isolated objects surrounded by one particular class and with a size smaller than the

3201 minimum mapping unit were reclassified at the enclosing class. Furthermore, isolated objects

3202 with a size smaller than the minimum mapping unit neighboring more than one class were

3203 classified to the one where it had the longest common border.

3204

3205

3206

3207

3208

3209

3210

3211

154

3212 Table 7: Accuracy assessments by landcover classification

Observed Arable Kalahari Mopane S/F Teak Water Total Accuracy Land Woodland Woodland Grasslands Forest Arable Land 2114420 522 125 95899 22 12 2211000 0.956 Kalahari 158 8923489 152422 489 4522 52 9081132 0.983 Woodland Mopane 15 898522 1452600 106 1256 589 2353088 0.617 Woodland S/F Grasslands 990 1230 458 2525589 58 15369 2543694 0.993 Teak Forest 12 11256 1235 58 1332689 0 1345250 0.991 Water 0 0 0 185699 146389 217325 549413 0.396

Column Total 2115595 9835019 1606840 2807840 1484936 233347 18083577 Producers 0.999 0.907 0.904 0.899 0.897 0.931 Overall 0.916 accuracy

Predicted Arable Kalahari Mopane S/F Teak Water Land Woodland Woodland Grasslands Forest Arable Land 0.117 0.000 0.000 0.005 0.000 0.000 0.122 Kalahari 0.000 0.493 0.008 0.000 0.000 0.000 0.502 Woodland Mopane 0.000 0.050 0.080 0.000 0.000 0.000 0.130 Woodland S/F Grasslands 0.000 0.000 0.000 0.140 0.000 0.001 0.141 Teak Forest 0.000 0.001 0.000 0.000 0.074 0.000 0.074 Water 0.000 0.000 0.000 0.010 0.008 0.012 0.030 Column Total 0.117 0.544 0.089 0.155 0.082 0.013 1.000

Θ1 0.916

Θ2 0.327 Κ 0.875 3213

3214

3215

3216

3217

3218

3219

155

3220 Appendix 3. Key Interventions and Outcomes in Promotion of Landscape Connectivity.

3221 3.1 Diagnosing challenges to maintaining biodiversity at the landscape scale.

3222 Majority challenges to the long-term maintenance of large carnivores and biodiversity

3223 throughout this landscape differ little in essence from other increasingly human-dominated

3224 wildlife areas throughout southern Africa and beyond. Encouraging our understanding of

3225 the science underpinning endangered species and biodiversity management has never been

3226 clearer, and with political support for the KAZA TFCA programme from a wide range of

3227 stakeholders, appropriate interventions to halt and reverse effects of deleterious human

3228 pressures are realistic and feasible.

3229 Human populations are increasing, education and poverty levels, especially surrounding

3230 protected areas and concerning youth, require financial and technical support. The current

3231 cost-benefit ratio to rural communities of co-existing with wildlife versus conversion of

3232 wildlife habitat into agro-pastoralism land uses disincentives development of green

3233 economies and capitalising from more sustainable wildlife-based land uses. Existing agro-

3234 pastoralist approaches are inefficient and typically degrade natural resources, jeopardise

3235 livelihood security and threaten wildlife populations. Political and financial support for law

3236 enforcement, combatting corruption and illegal wildlife trade remain key challenges.

3237 While site-specific variation in these broad challenges, including social, economic, and

3238 political factors, determines the tailored interventions required to improve biodiversity

3239 conservation and livelihoods, a number of broad approaches are applicable in the Kafue-

3240 Zambezi landscape.

156