Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Effects of linear barriers on African buffalo (Syncerus caffer) movement in a transfrontier conservation… Thompson, Allison 2015

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2015_september_thompson_allison.pdf [ 2.57MB ]
Metadata
JSON: 24-1.0166309.json
JSON-LD: 24-1.0166309-ld.json
RDF/XML (Pretty): 24-1.0166309-rdf.xml
RDF/JSON: 24-1.0166309-rdf.json
Turtle: 24-1.0166309-turtle.txt
N-Triples: 24-1.0166309-rdf-ntriples.txt
Original Record: 24-1.0166309-source.json
Full Text
24-1.0166309-fulltext.txt
Citation
24-1.0166309.ris

Full Text

EFFECTS OF LINEAR BARRIERS ON AFRICAN BUFFALO (SYNCERUS CAFFER) MOVEMENT IN A TRANSFRONTIER CONSERVATION AREA by  Allison Thompson  B.Sc., The University of British Columbia, 2015  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  MASTER OF SCIENCE in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Resource Management and Environmental Studies)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  June 2015  © Allison Thompson, 2015 ii  Abstract Understanding the movement of large wildlife species is crucial for biodiversity conservation efforts. Despite a recognition of the need for landscape scale conservation efforts addressing connectivity, our understanding of the effects of linear landscape features on wildlife movement remains incomplete, particularly in developing country contexts. In this study, we used a six-year GPS collar dataset of the locations of 36 buffalo, in Namibia and surrounding countries, to investigate permeability of linear features to buffalo movement, and the effect of proximity to these linear features and human settlements on buffalo movement, including behaviour when crossing potential barriers. We found that buffalo alter their behaviour in proximity to linear barriers, increase their speed when crossing roads, and are less likely to cross roads when they are near human settlements. These findings can aid conservation planners’ understanding of how large animals respond to landscape features to better manage for connectivity at the landscape scale. iii  Preface The buffalo GPS data were collected by WWF-US and the Ministry of Environment and Tourism Namibia, provided to me by Dr. Robin Naidoo. The rest of my dataset was assembled by me, under the direction of Dr. Robin Naidoo. The analysis was performed by me, under the direction of Dr. Robin Naidoo with relevant input from Dr. Kai Chan. The writing was also done by me, with editing and helpful comments from Dr. Robin Naidoo and Dr. Kai Chan.  iv  Table of Contents  Abstract .......................................................................................................................................... ii Preface ........................................................................................................................................... iii Table of Contents ......................................................................................................................... iv List of Tables ................................................................................................................................ vi List of Figures ............................................................................................................................. xiii List of Abbreviations ................................................................................................................. xiv Acknowledgements ......................................................................................................................xv Dedication ................................................................................................................................... xvi Chapter 1: Introduction ................................................................................................................1 1.1 Ecological and biodiversity decline, and the growth of applied conservation .................. 1 1.2 Movement ecology applied to conservation ...................................................................... 2 1.3 KAZA and CBNRM programs in Namibia ....................................................................... 4 1.4 The African buffalo............................................................................................................ 7 1.5 Linear barrier impacts on wildlife...................................................................................... 8 1.6 Research objectives and questions ..................................................................................... 9 Chapter 2: Effects of Linear Barriers on African buffalo (Syncerus caffer) Movement in a Transfrontier Conservation Area ...............................................................................................11 2.1 Introduction ...................................................................................................................... 11 2.2 Study area......................................................................................................................... 14 2.3 Methods............................................................................................................................ 16 2.3.1 The study dataset....................................................................................................... 16 v  2.3.2 Environmental variables ........................................................................................... 18 2.3.3 Statistical analysis ..................................................................................................... 21 2.4 Results .............................................................................................................................. 24 2.4.1 Movement models ..................................................................................................... 24 2.4.2 Crossing models ........................................................................................................ 26 2.5 Discussion ........................................................................................................................ 29 2.5.1 Which linear features on the landscape are barriers, and which of those are impermeable or semi permeable? ......................................................................................... 29 2.5.2 How do buffalo behave in proximity to linear features? .......................................... 30 2.5.3 How do buffalo behave when crossing linear features? ........................................... 32 2.5.4 General discussion and conclusions.......................................................................... 33 Chapter 3: Conclusion .................................................................................................................35 3.1 Overall significance and contribution .............................................................................. 35 3.2 Analysis and integration of the research .......................................................................... 36 3.3 Strengths and limitations.................................................................................................. 38 3.4 Potential applications ....................................................................................................... 40 3.5 Possible future research directions ................................................................................... 41 Bibliography .................................................................................................................................43 Appendices ....................................................................................................................................53 Appendix A Additional Statistical Results ............................................................................... 53 A.1 Additional results from movement model, split by collaring location ....................... 53 A.2 Additional results from crossing model ...................................................................... 67  vi  List of Tables Table 2.1 Buffalo collars, sites, migration types ……………………………….……………… 16 Table 2.1 Movement model, speed, dry season, all collars …………………….……………… 26 Table 2.2 Movement model, speed, wet season, all collars ……………………………………. 26 Table 2.3 Crossing model, Mahango road, 2000m …………………………………………….. 28 Table 2.4 Crossing model, Mahango road, 2000m …………………………………………….. 28 Table 2.5 Crossing model, Mahango road, 2000m …………………………………………….. 29 Table A.1 All collars, dry season, turn angle model …………………………………………… 53 Table A.2 All collars, wet season, turn angle model …………………………………………... 54 Table A.3 Buffalo core, dry season, turn angle model ……………………………………….... 54 Table A.4 Buffalo core, dry season, speed model ……………………………………….....….. 55 Table A.5 Buffalo core, wet season, turn angle model ………………………………………… 55 Table A.6 Buffalo core, wet season, speed model …………………………………………...… 56 Table A.7 Eastern floodplain, dry season, turn model …………………………………………. 56 Table A.8 Eastern floodplain, dry season, speed model ……………………………………….. 57 Table A.9 Eastern floodplain, wet season, turn angle model …………………………………... 57 Table A.10 Eastern floodplains, wet season, speed model …………………………………….. 57 Table A.11 Horseshoe, dry season, turn angle model ………………………………………….. 58 Table A.12 Horseshoe, dry season, speed model ………………………………………………. 58 Table A.13 Horseshoe, wet season, turn angle model …………………………………………. 59 Table A.14 Hoseshoe, wet season, speed model ………………………………………………. 59 Table A.15 Mahango, dry season, turn angle model …………………………………………... 60 Table A.16 Mahango, dry season, speed model ……………………………………………….. 60 vii  Table A.17 Mahango, wet season, turn angle model ………………………………………...… 61 Table A.18 Mahango, wet season, speed model ……………………………………………….. 61 Table A.19 Nkasa Rupara, dry season, turn angle model ……………………………………… 62 Table A20 Nkasa Rupara, dry season, speed model …………………………………………… 62 Table A.21 Nkasa Rupara, wet season, turn angle model ……………………………………... 62 Table A.22 Nkasa Rupara, wet season, speed model ………………………………………….. 63 Table A.23 Mudumu, dry season, turn angle model …………………………………………… 63 Table A.24 Mudumu, dry season, speed model ………………………………………………... 64 Table A.25 Mudumu, wet season, turn angle model …………………………………………... 64 Table A.26 Mudumu, wet season, speed model ……………………………………………….. 65 Table A.27 Susuwe, dry season, turn angle model …………………………………………….. 65 Table A.28 Susuwe, dry season, speed model ……………………………………………….… 66 Table A.29 Susuwe, wet season, turn angle model …………………………………………….. 66 Table A.30 Susuwe, wet season, speed model …………………………………………………. 67 Table A.31 77264, dry season, angle model ………………………………………………….... 67 Table A.32 77264, dry season, speed model …………………………………………………... 68 Table A.33 77264, wet season, angle model ……………………………………………...…… 68 Table A.34 77264, wet season, speed model ……………………………………………...…… 69 Table A.3577264new, dry season, angle model ……………………………………………..… 69 Table A.36 77264new, dry season, speed model ………………………………………………. 70 Table A.37 77264new, wet season, angle model …………………………………………….… 70 Table A.38 77264new, wet season, speed model …………………………………………….... 70 Table A.39 77266, dry season, angle model …………………………………………………… 71 viii  Table A.40 77266, dry season, speed model ……………………………………………...…… 71 Table A.41 77266, wet season, angle model …………………………………………………... 71 Table A.42 77266, wet season, speed model …………………………………………………... 72 Table A.43 77266new, dry season, angle model …………………………………………….… 72 Table A.44 77266new, dry season, speed model ………………………………………………. 73 Table A.45 77266new, wet season, angle model ………………………………………………. 73 Table A.46 77266new, wet season, speed model ……………………………………………… 74 Table A.47 AM290, dry season, angle model …………………………………………………. 74 Table A.48 AM290, dry season, speed model …………………………………………………. 75 Table A.49 AM290, wet season, angle model …………………………………………………. 75 Table A.50 AM290, wet season, speed model …………………………………………………. 75 Table A.51 94043, dry season, angle model …………………………………………………… 76 Table A.52 94043, dry season, speed model …………………………………………………... 76 Table A.53 94043, wet season, angle model ……………………………………………...…… 77 Table A.54 94043, wet season, speed model …………………………………………………... 77 Table A.55 AG275, dry season, angle model ………………………………………………….. 78 Table A.56 AG275, dry season, speed model ………………………………………………….. 78 Table A.57 AG275, wet season, angle model ………………………………………………….. 79 Table A.58 AG275, wet season, speed model …………………………………………………. 79 Table A.59 AG273, dry season, angle model ………………………………………………….. 80 Table A.60 AG273, dry season, speed model ………………………………………………….. 80 Table A.61 AG273, wet season, angle model ………………………………………………….. 80 Table A.62 AG273, wet season, speed model …………………………………………………. 81 ix  Table A.63 AG276, dry season, angle model  ………………………………………………….. 81 Table A.64 AG276, dry season, speed model ………………………………………………….. 81 Table A.65 AG276, wet season, angle model ………………………………………………….. 82 Table A.66 AG276, wet season, speed model …………………………………………………. 82 Table A.67 AG277, dry season, angle model  ………………………………………………….. 82 Table A.68 AG277, dry season, speed model ………………………………………………….. 83 Table A.69 AG277, wet season, angle model ………………………………………………….. 83 Table A.70 AG277, wet season, speed model …………………………………………………. 83 Table A.71 AM291, dry season, angle model …………………………………………………. 84 Table A.72 AM291, dry season, speed model …………………………………………………. 84 Table A.73 AM291, wet season, angle model …………………………………………………. 85 Table A.74 AM291, wet season, speed model …………………………………………………. 85 Table A.75 77259, dry season, angle model …………………………………………………… 86 Table A.76 77259, dry season, speed model …………………………………………………... 86 Table A.77 77259, wet season, angle model …………………………………………………... 87 Table A.78 77259, wet season, speed model …………………………………………………... 87 Table A.79 77261new, dry season, angle model ………………………………………………. 88 Table A.80 77261new, dry season, speed model ………………………………………………. 88 Table A.81 77261new, wet season, angle model ………………………………………………. 89 Table A.82 77261new, wet season, speed model ……………………………………………… 89 Table A.83 77262, dry season, angle model …………………………………………………… 90 Table A.84 77262, dry season, speed model …………………………………………………... 90 Table A.85 77262, wet season, angle model …………………………………………………... 91 x  Table A.86 77262, wet season, speed model …………………………………………………... 91 Table A.87 77265new, dry season, angle model …………………………………………….… 92 Table A.88 77265new, dry season, speed model ………………………………………………. 92 Table A.89 77265new, wet season, angle model ………………………………………………. 93 Table A.90 77265new, wet season, speed model ……………………………………………… 93 Table A.91 SAT508, dry season, angle model ………………………………………………… 94 Table A.92 SAT508, dry season, speed model ………………………………………………… 94 Table A.93 SAT508, wet season, angle model ………………………………………………… 95 Table A.94 SAT508, wet season, speed model ………………………………………………... 95 Table A.95 SAT509, dry season, angle model ………………………………………………… 96 Table A.96 SAT509, dry season, speed model ………………………………………………… 96 Table A.97 SAT509, wet season, angle model ………………………………………………… 97 Table A.98 SAT509, wet season, speed model ………………………………………………... 97 Table A.99 77263, dry season, angle model …………………………………………………… 98 Table A100 77263, dry season, speed model ………………………………………………….. 98 Table A.101 77263, wet season, angle model …………………………………………………. 99 Table A.102 77263, wet season, speed model …………………………………………………. 99 Table A.103 77263new, dry season, angle model ……………………………………………. 100 Table A.104 77263new, dry season, speed model ……………………………………………. 100 Table A.105 77263new, wet season, angle model ……………………………………………. 101 Table A.106 77263new, wet season, speed model …………………………………………… 101 Table A.107 94041, dry season, angle model  ………………………………………………… 102 Table A.108 94041, dry season, speed model ………………………………………………... 102 xi  Table A.109 94041, wet season, angle model ………………………………………………... 103 Table A.110 94041, wet season, speed model ………………………………………………... 103 Table A.111 94042, dry season, angle model ………………………………………………… 104 Table A.112 94042, dry season, speed model ………………………………………………... 104 Table A.113 94042, wet season, angle model ………………………………………………... 105 Table A.114 94042, wet season, speed model ………………………………………………... 105 Table A.115 AG29, dry season, angle model  ………………………………………………… 106 Table A.116 AG269, dry season, speed model ……………………………………………….. 106 Table A.117 AG269, wet season, angle model ……………………………………………….. 107 Table A.118 AG269, wet season, speed model ………………………………………………. 107 Table A.119 AG270, dry season, angle model ……………………………………………….. 108 Table A.120 AG270, dry season, speed model ……………………………………………….. 108 Table A.121 AG270, wet season, angle model ……………………………………………….. 109 Table A.122 AG270, wet season, speed model ………………………………………………. 109 Table A.123 77265, dry season, angle model  ………………………………………………… 110 Table A.124 77265, dry season, speed model ………………………………………………... 110 Table A.125 77265, wet season, angle model ………………………………………………... 111 Table A.126 77265, wet season, speed model ………………………………………………... 111 Table A.127 AG278, dry season, angle model ……………………………………………….. 112 Table A.128 AG278, dry season, speed model ……………………………………………….. 112 Table A.129 AG278, wet season, angle model ……………………………………………….. 113 Table A.130 AG278, wet season, speed model ………………………………………………. 113 Table A.131 AG279, dry season, angle model ……………………………………………….. 114 xii  Table A.132 AG279, dry season, speed model ……………………………………………….. 114 Table A.133 AG279, wet season, angle model ……………………………………………….. 115 Table A.134 AG279, wet season, speed model ………………………………………………. 115 Table A.135 AM291, dry season, angle model ………………………………………………. 116 Table A.136 AM291, dry season, speed model ………………………………………………. 116 Table A.137 AM291, wet season, angle model ………………………………………………. 117 Table A.138 AM291, wet season, speed model ………………………………………………. 117 Table A.139 Mahango crossing model, 1000m ………………………………………………. 118 Table A.140 Mahango crossing model, 5000m ………………………………………………. 118 Table A.141 Ring road crossing model, 1000m ……………………………………………… 118 Table A.142 Ring road crossing model, 5000m ……………………………………………… 119 Table A.143 Tar road crossing model, 1000m ……………………………………………….. 119 Table A.144 Tar road crossing model, 5000m ……………………………………………….. 119 Table A.145 Buffalo Core (Tar road) crossing model, 1000m ……………………………….. 120 Table A.146 Buffalo Core (Tar road) crossing model, 2000m ……………………………….. 120 Table A.147 Buffalo Core (Tar road) crossing model, 5000m ……………………………….. 120 Table A.148 Mudumu (Ring road) crossing model, 1000m ………………………………….. 121 Table A.149 Mudumu (Ring road) crossing model, 2000m ………………………………….. 121 Table A.150 Mudumu (Ring road) crossing model, 5000m ………..………………………… 121 Table A.151 Susuwe (Tar road) crossing model, 1000m …………………………………….. 122 Table A.152 Susuwe (Tar road) crossing model, 2000m …………………………………….. 122 Table A.153 Susuwe (Tar road) crossing model, 5000m …………………………………….. 122 xiii  List of Figures Figure 2.1 Map of the study area …………………………………………………………….… 15 Figure 2.2 Buffalo GPS fixes and collaring sites ………………………………………………. 18  xiv  List of Abbreviations AICc………………………………. Akaike’s Information Criterion, finite sample size correction CBNRM……………………………………… Community-based natural resource management EVI…………………………………………………………...……… Evaluated vegetation index GPS…………………………………………………………… Geographical Positioning Systems KAZA………………………………………………………………………….. Kavango-Zambezi TFCA………………………………………………...…...……… Transfrontier conservation area WWF…………………………………………………………..... World Wildlife Fund for Nature  xv  Acknowledgements Thank you to my supervisors, Dr. Kai Chan and Dr. Robin Naidoo for all of your guidance and support, and help with funding. Thank you to the NSERC Canadian Graduate Scholarship (master’s) program for funding this work. To my grandparents, Douglas and Eva Thompson for providing me with financial support for education. To my parents for financial and emotional support. To Ryan, Olive, and Walter for always being there for me. Thank you to the rest of my family and friends for reminding me that life is so much more than school. Thank you to my lab group for help with statistics, presentations, and insightful discussion. xvi  Dedication This thesis is dedicated to my family.1  Chapter 1: Introduction  1.1 Ecological and biodiversity decline, and the growth of applied conservation The field of conservation biology sprung from a growing recognition of rapid decline of biodiversity and a desire to stop or reverse this trend. It is well recognized that conservation biologists often must work with imperfect information to provide recommendations to managers and planners (Soule, 1985). Over time, both our understanding of the importance of ecosystems to human wellbeing and ecosystem function, and the breadth and depth of science supporting conservation efforts, have grown (e.g. Costanza & Folke, 1997; Armsworth et al., 2007; Naidoo et al., 2008; Visconti et al., 2011; Cardinale et al., 2012). At the same time, biodiversity loss has continued at historically unprecedented rates (Dirzo & Raven, 2003; Barnosky et al., 2011). Applied approaches to conservation have changed in response to our evolving understanding. There is still a recognized need for core conservation areas that exclude much human activity, but increasingly conservation initiatives involve whole landscapes, including human use areas, and often aim to pair biodiversity conservation with poverty alleviation (Rands et al., 2010; Turner et al., 2012). Programs such as payments for ecosystem services and community-based natural resource management are implemented at local scales and encourage local people to manage landscapes for the conservation of biodiversity, ecosystem services, and/or ecosystem function (Munthali, 2007; Naidoo et al., 2011). Transfrontier conservation areas build upon the traditional approach to conservation of national parks, by linking core conservation areas across borders in a network of protected areas, encouraging cooperation between managers to create landscape level conservation plans. All three of the aforementioned conservation approaches aim to link income generation with conservation. As these approaches involve (in part, or 2  completely) conservation in working landscapes, an understanding of the impacts of human development on ecological systems is necessary, as is a landscape-scale understanding of the requirements of target species. Ecological studies of target conservation species can provide information supporting the implementation and monitoring of such conservation initiatives.  1.2 Movement ecology applied to conservation The impacts of changes on species, such as human development, or conservation programs, can be difficult to evaluate, as population-level indicators can take years, or even decades, to manifest. The field of movement ecology contributes to conservation planning and monitoring by striving to understand the proximate and ultimate drivers of animal movement (Bucholz, 2007; Nathan, 2008). Alteration in movement patterns in response to a conservation intervention, human development, or changes in climate can provide clues for the long term effect of such changes on the wellbeing of wildlife species (Doerr et al., 2011).  Biotelemetry, the use of remote sensing tools to study movement ecology, is a fast evolving suite of technology and accompanying techniques, which has contributed to conservation efforts by allowing researchers to track wildlife to places where humans cannot easily access or sample (Cooke et al., 2004). The use of remote sensing technology in wildlife studies has grown from a focus on the use of aerial photography to the inclusion of radio collars and GPS satellite collars that can track wildlife at ever increasing temporal and spatial resolution (Cagnacci et al., 2010). The tools of movement ecology have also expanded to include the use of complementary technology to monitor wildlife, such as camera traps (McCarthy et al., 2010; Pettorelli et al., 2010) and biological sensors (e.g. collars on cattle that record head movements 3  to identify grazing behaviour, see Augustine & Derner, 2013), and the pairing of GPS technologies with additional sensors to add extra dimensions to the understanding of wildlife movement.  The relative proliferation of GPS and associated technology and its rapid improvement has led to the publication of numerous research papers documenting and analyzing the movement paths of wildlife species ranging from very small (insects, e.g. Kissling et al., 2014) to very large (elephants, whales, e.g. Mate et al., 2011; Mashintonio et al., 2014), including aquatic and avian species (e.g. Pérez-García et al., 2012; Villegas-Ríos et al., 2013). As a result, a wide variety of models have been built to accommodate such a range of study species and different types of movement, from random walks to movement models that incorporate multiple behavioural strategies, such memory-based movement, area restricted search, directed travel, and others (e.g. Dalziel et al., 2008; Fryxell et al., 2008; Johnson et al., 2008; Fagan et al., 2013). The increasing sophistication of these models has improved the scientific understanding of the patterns underlying movement. With a broad set of tools for modeling movement, a focus has developed on producing results that can be applied to conservation management.  Movement ecology offers many contributions to applied conservation, including identifying space use patterns as well as understanding interactions between wildlife and their environment. The ability to identify species home ranges and their seasonal and inter-annual changes provides an overview of the spatial needs of wildlife from the daily scale up to the multi-year scale (Chadwick et al., 2010; Getz et al., 2007; Naidoo et al., 2012; Ryan et al., 2006, Williams et al., 2011). Pairing GPS datasets with indigenous or local knowledge has helped researchers to provide a more complete picture of species needs, adding layers of detail that may not be easily quantified (e.g. Service et al., 2014). Researchers have used complementary spatial 4  data to understand how animal behaviour is influenced by the environment (Augustine & Derner, 2013; Morales et al., 2004), which can lead to identifying the outcomes of conservation interventions (such as artificial water holes) (Loarie et al., 2009), and understanding the multiple scales at which resources influence movement (Frair et al., 2005; Fryxell et al., 2008). Spatially explicit data has also been paired with GPS tracking data to examine how a wide range of species respond to anthropogenic development including roads, fences, settlements, and other features (including underwater man-made structures) (e.g. Russel et al., 2014; Leblond et al., 2013, Frair et al., 2014), and the conditions under which linear features on the landscape can become barriers (Matawa et al., 2012; Cozzi et al., 2013; Beyer et al., 2014; Elliot et al., 2014). The results of these studies then can be further applied to build “resistance” surfaces (maps that are coded based on varying levels of resistance to wildlife movement across the landscape) to help planners identify potentially suitable movement corridors (Richard & Armstrong, 2010; Zeller et al., 2012; Squires et al., 2013; Zeller et al., 2014; Elliot et al., 2014). Researchers have gained insight into drivers of human-wildlife conflict (e.g. Gunn et al., 2013), enabling better conflict mitigation strategies. With the growth of movement ecology studies, an understanding of animal space use at multiple temporal and spatial scales, and how individuals respond to their environment and changes in it, managers can better understand species needs and apply that understanding to conserving habitat.  1.3 KAZA and CBNRM programs in Namibia In southern Africa, innovative programs have been developed to address conservation and human development needs simultaneously, recognizing that wildlife and human lives are directly intertwined. Namibia, a country in southern Africa, is taking part in two such programs – 5  a transfrontier conservation area (TFCA) initiative, and a community-based natural resource management program (CBNRM). TFCAs consist of a network of different land-use areas, from core conservation to human use, spanning international borders, and aim to balance the needs of local people living within the TFCA with the biodiversity conservation goals. The direct sources of economic benefits that people derive from TFCAs include nature-based tourism, hunting, and plant-based products, among others (Thomson et al., 2013). Namibia’s CBNRM program is also aimed at conserving biodiversity while promoting economic development, by transferring the rights to manage natural resources local groups (conservancies), allowing them to manage, and profit from, the resources within their conservancies (NACSO, 2014).  The Kavango-Zambezi (KAZA) TFCA is a new TFCA in southern Africa that includes parts of Angola, Botswana, Namibia, Zambia, and Zimbabwe, that will be approximately 500,000 km2 when complete, roughly the size of France (Russell Taylor Pers. Comm.; KAZA TFCA, 2013). Key issues for TFCAs include identifying areas for conservation zoning and those for economic development, fostering cooperation among land use managers across borders, and ensuring connectivity for wildlife throughout the TFCA to maintain healthy animal populations (Munthali, 2007). Connectivity is especially important in KAZA as in addition to natural barriers and human infrastructure, veterinary fences have been erected to prevent the spread of disease from wild buffalo (Syncerus caffer) to cattle. Buffalo are among many large wildlife species in the region, which require large tracts of habitat to maintain healthy populations (Naidoo et al., 2014; Naidoo et al., 2014a).  The CBNRM program in Namibia has contributed to the recovery of many of the country’s large wildlife species (NACSO, 2014). Wildlife are an important source of benefits for conservancies, as benefits from trophy hunting and photographic safaris currently provide the 6  greatest income to communities registered in the CBNRM program (Naidoo et al., 2011a). Additional research has found that biodiversity is correlated to conservancy income in Namibia, and the more wildlife species (especially ‘Big 5’ iconic species) a conservancy has, the more income the conservancy derives (Naidoo et al., 2011b). Communities also benefit from wildlife because they own the meat from animals that are killed in the trophy hunt, which provides an important source of protein. The meat from certain species is also used in traditional feasts, and has a cultural importance for local people.  The benefits from wildlife-based endeavors are clear, but a potential downside is human-wildlife conflict as habitats overlap and wildlife may raid crops and livestock for food, transmit diseases to wildlife, and cause direct human harm (Mulonga et al., 2003). Understanding the drivers of wildlife movement is important in the context of CBNRM as understanding movement patterns can help managers plan for conservation and the reduction of mitigating human-wildlife conflict (Gunn et al., 2013). Namibia’s Kavango and Zambezi regions (formerly the Caprivi Strip) lie at the heart of KAZA TFCA and include 17 conservancies in the CBNRM program. It is an ideal study area for wildlife movement as the benefits (and costs) of wildlife to local people are transparent, but there are clear threats to the health of wildlife populations. Connectivity is a major concern in the TFCA context as KAZA spans 5 partner countries with major migratory routes through the area and the challenge of coordinating management throughout the system. At the CBNRM level, connectivity is an important issue as conservancies continue to grow, attracting more people to the region and building infrastructure, both of which can be deterrents to wildlife movement.  7  1.4 The African buffalo The African buffalo (Syncerus caffer) is an important species in the KAZA TFCA and Namibia’s CBNRM program, both ecologically and for socio-economic reasons. Buffalo are large herbivores whose diet ranges from approximately 75-100% grass (Cerling et al., 2003; Codron et al., 2007; Tshabalala et al., 2009). As large grazers, buffalo can change plant communities directly through their grazing patterns, and indirectly by influencing the distribution of other smaller grazing species (McNaughton, 1985; Bar-David et al., 2009). Buffalo are also an important prey species for lions, top predators in the ecosystem, and along with lions, they are a member of the “Big 5” most dangerous species to hunt, and therefore a top target for trophy hunting and photographic safaris, which benefits local people (Naidoo et al., 2010). In addition playing a part in tourism benefits, buffalo are a regionally important species as they are a target of subsistence hunting and are used for meat in traditional feasts (Naidoo et al., 2011). At the opposite end of the spectrum, buffalo can also have a negative impact on human wellbeing. Human-wildlife conflict also arises because buffalo occasionally consume or degrade crops. Buffalo can also carry foot-and-mouth disease and bovine tuberculosis (Vosloo et al., 2001; Miguel et al., 2013). The transfer of Bovine tuberculosis and foot-and-mouth disease to cattle can have devastating effects, as the disease can cause lameness, and in severe cases, mortality. Accordingly, regions with foot-and-mouth disease in cattle have very limited access to export markets for beef (Thomson et al., 2013).  Buffalo are large animals (weighing 500-900kg on average) and require varying sizes of intact habitat to survive. Historically, buffalo range selection was constrained by the availability of surface water and grazing, and the only linear features restricting movement were large, impassable rivers. In the current day, many more potential barriers to movement exist, due to 8  human development, such as roads, fences, and settlements. Along with nutritional and water needs, anthropogenic and natural linear features on the landscape play a factor in determining buffalo range selection and movement, from daily movement decisions up to inter-annual migration and dispersal patterns (Naidoo et al., 2012). For buffalo in the Kavango-Zambezi region of Namibia, home range size has been shown to range from as small as 0.60 km2 up to 448 km2 (Naidoo et al., 2012). Buffalo in this region have exhibited a range of behaviour from long distance migration (up to 100km one-way) and dispersal (greater than 200km), to non-migrant behaviour where the home range stays constant year round (Naidoo et al., 2012; Naidoo et al., 2014). Although buffalo populations appear to be growing in Namibia under the CBNRM program, a recent study found that populations are less genetically related than historically, indicating restricted gene flow and possibly impediments to connectivity (Epps et al., 2013). Elsewhere in KAZA, buffalo have been found to have lower heterozygosity than in the past, which increases their susceptibility to population loss from stochastic events (Smitz et al., 2014). These changes have been attributed in part to fragmentation from human development and increasing distance between buffalo herds (Epps et al., 2013, Smitz et al., 2014)  1.5 Linear barrier impacts on wildlife In light of a growing understanding of the space needs of large herbivore species and the potential effects of human infrastructure on conservation efforts, researchers have begun to use movement ecology techniques to study the impacts of linear barriers on wildlife movement. The effect of roads on wildlife has been well studied on a both herbivores and carnivores, large and small (Forman & Alexander, 1998; Spellerberg, 2007; Forman, 2010). The impact of fences, 9  pipelines, and other potential barriers on wildlife movement has also been studied (Dyer et al., 2002; Sheldon, 2005; Gates et al., 2012; Semeniuk et al., 2012; Beyer et al., 2014). Many of these linear barrier studies have taken place in North America (Dyer et al., 2001; Sheldon, 2005; Waller & Servheen, 2005; Northrup et al., 2012; Roever et al., 2010; Lendrum, 2013; Seidler & Long, 2014).  The southern hemisphere has had less research on the impact of barriers, although this field is growing. Researchers in southern Africa have begun to study the impact of barriers on movement in the context of TFCAs as well as other conservation areas, many of which are completely fenced in in this part of the world (Loarie et al., 2009; Mbaiwa & Mbaiwa, 2006; Vanak et al., 2010a; Gates et al., 2012). In KAZA, the effect of roads has been studied on lions, as well as a predator guild (Cozzi et al., 2013; Elliot et al., 2014). It has also been established in KAZA that human land use, such as small agriculture, near natural linear barriers, rivers, can lead to an increase in human-wildlife conflict with large herbivore species (Matawa et al., 2012). Growing programs supporting conservation and livelihood goals are excellent case studies to understand the impact of current linear barriers on wildlife movement, and gain an understanding that can be used to predict the impact of future linear structures on movement.  1.6 Research objectives and questions The objective of this thesis was to gain an understanding of the current state of connectivity in the Kavango and Zambezi regions of Namibia, as perceived by a large herbivore species, the African buffalo. This is important in the context of both the KAZA TFCA and the CBNRM program. For the KAZA TFCA to function as intended, wildlife must be able to move between core conservation areas through the surrounding landscape. We aim contribute to an understanding of whether the TFCA is functioning as a connected network of protected areas, or 10  as a collection of isolated patches. As a part of the CBNRM program, buffalo are an important species and their long-term conservation is tied in to the success of the program. The Kavango and Zambezi regions of Namibia are undergoing human population growth, and this baseline assessment will establish the current state of connectivity for buffalo, and allow managers to anticipate future connectivity changes as human density in the region increases. To assess connectivity in the study area, we specifically studied the impact of linear barriers on the movement of wildlife. We asked the research questions: which features in the study area constitute linear barriers, and are these barriers impermeable or semi-permeable to buffalo movement? Do buffalo alter their behaviour in proximity to linear barriers on the landscape, such as rivers, roads, or fences? Do buffalo alter their movement when crossing semi-permeable linear barriers (e.g. roads), and do factors such as human settlement affect crossing behaviour? The thesis is comprised of one main chapter containing the study, and a closing chapter with a broader discussion and general conclusions.   11  Chapter 2: Effects of Linear Barriers on African buffalo (Syncerus caffer) Movement in a Transfrontier Conservation Area  2.1 Introduction Understanding large-scale wildlife movement patterns is critical for biodiversity conservation. Broad scale movement processes such as migration and dispersal enable species to optimize nutritional requirements, find mates, avoid predation, adapt to changing environmental conditions, and maintain genetic flow between populations (Zeller et al., 2012). These processes require connectivity between habitats and are threatened by human-induced landscape change (Trakhtenbrot et al., 2005; Kokko & López-Sepulchre, 2006). Loss of connectivity can lead to population collapse and local extinction (Saunders et al., 1992; Bolger et al., 2008). There is a recognized need for landscape-scale conservation approaches to manage for connectivity, and in cases where landscapes span multiple countries; one of the approaches taken is the creation of transfrontier conservation areas (Rands et al., 2010). Transfrontier conservation areas (TFCAs) are one critical tool in the effort to maintain large-scale wildlife movement. As networks of protected areas that span national borders and incorporate multiple land use types, the goal of TFCAs is to conserve biodiversity while encouraging sustainable economic development for the people living within the TFCA (Munthali, 2007). Theoretically, by fostering cooperation among park managers and also local people in their attitudes toward wildlife, TFCAs can manage wildlife at the landscape scale, conserving habitat connectivity (Munthali, 2007). Despite this vision, on the ground there are many factors that may compromise the ability of managers to achieve conservation success. 12  Anthropogenic and natural linear features on the landscape may hinder the ability of wildlife to migrate and disperse (Dyer et al., 2002; Dussault et al., 2007; Loarie et al., 2009; Sawyer et al., 2013a; Beyer et al., 2014). TFCA managers must work within the bounds of the previously existing human infrastructure as well as plan for anticipated future growth. A crucial component of TFCAs is managing the effects of linear features already present on the landscape, including roads, pipelines, rivers, and fences. Linear features on the landscape can have a multitude of effects of on wildlife movement. The physical structures themselves can vary in their permeability (Dyer et al., 2002; Sheldon, 2005; Seidler & Long, 2014). Different types of physical structures affect animal movement in different ways, such as fences of different heights and materials, underground versus above ground pipelines, and gravel versus paved roads (Spellerberg et al., 2007; Ferguson & Hanks, 2010). In addition to potentially impeding movement, linear features can create edge habitat surrounding them, which can influence the species assemblages that will use such habitat (Forman, 2010; Vanak et al., 2010; Beyer et al., 2014; Frair et al., 2014). Other indirect effects of linear features can also influence animal movement, such as the use of man-made linear corridors by wolves in northern Canada to hunt prey, causing learned avoidance of linear features by caribou (Dyer et al., 2001). There are important questions about the effect of linear features on animal movement that have implications for landscape scale connectivity and biodiversity conservation.  The field of movement ecology provides the tools necessary to answer the above questions. Movement ecology has been rapidly growing as GPS and other tracking technology decreases in size and prize, and increases in battery life (Cagnacci et al., 2010).  This technology has allowed researchers to study animal movement at the micro scale (such as tagging insects, or recording 13  animal movement every few seconds), in environments humans cannot easily access (underwater, in dense jungle), and up to the macro scale (recording full migrations over hundreds of kilometres) (Kissling et al., 2014; Naidoo et al., 2014a; Naidoo et al. 2014b). The types of studies using GPS or radio collar technology have grown from exploring spatiotemporal patterns within the movement trajectory (e.g. Morales et al., 2004; Horne et al., 2007; Smouse et al., 2010), to studying the connection between the external environment and movement patterns (e.g. Frair et al., 2005; Roever et al., 2010; Johnson et al., 2011). The growth of both movement ecology studies and accompanying analytic methods has also allowed researchers to examine how linear features (both natural and anthropogenic) affect the movement of species of conservation concern. There has been a concentration of studies in North America as the rapid development of the natural gas and oil industries have caused declining populations of large mammals (e.g. Dyer et al., 2001; Sheldon 2005; Morellet et al., 2011). In the southern hemisphere, researchers have examined the impact of roads, water sources, and fences on the movement of large wildlife species (e.g. Graham & Douglas-Hamilton, 2009; Loarie et al., 2009; Elliot et al., 2014). With the development of multiple TFCAs in southern Africa (Ramutsindela, 2007), and the focus on connectivity throughout these cross-border parks, there has been a more recent effort to quantify connectivity in the region for large predators (Cozzi et al., 2013; Elliot et al., 2014). In southern Africa, there is the opportunity to mitigate against the potential impacts of human population growth on connectivity by planning for transboundary movement. Understanding the impacts of linear features, both natural (e.g. rivers), and man-made (e.g. roads) on species of conservation concern is crucial for large scale planning for healthy wildlife populations. 14  Here we use the case of the African buffalo (Syncerus caffer) in the Kavango-Zambezi TFCA to pose and answer a set of research questions about the movement of wildlife in relation to linear barriers. Research questions include:  which features in the study area constitute linear barriers, and are these barriers impermeable or semi-permeable to buffalo movement? Do buffalo alter their behaviour in proximity to linear barriers on the landscape, such as settlements, roads, or fences? Do buffalo alter their movement when crossing semi-permeable linear barriers (e.g. roads)?  2.2 Study area The study area is comprised of the Kavango and Zambezi regions of Namibia (formerly known as the Caprivi Strip), with a buffer of approximately 50km to the north and south. The Kavango and Zambezi regions are in northeast Namibia, taking the form of a narrow arm jutting out of the northeast corner of the country (Figure 2.1). The study area lies at the heart of the Kavango Zambezi TFCA, a new TFCA in southern Africa that includes parts of Angola, Botswana, Namibia, Zambia, and Zimbabwe and will be approximately 500 000 km2 when complete, roughly the size of France (Russell Taylor Pers. Comm., KAZA TFCA, 2013). The Kavanzo and Zambezi regions combined are approximately 450km long (east-west) and is 30km at the narrowest point. The study area is relatively flat topographically (~930-1100 m above sea level) and experiences seasonal rainfall, as the majority of annual precipitation (~650 mm per year) falls between November and April (Naidoo et al., 2012). Major geographical features include the Chobe river flowing along the southeastern border section, the Zambezi river on the northeast border, and the Okavango and Kwando rivers which intersect the study area flowing north-south (Figure 2.1). The study area contains a variety of habitat types, including a 15  mixture of floodplains, grasslands, shrub savannas, tree savannas, and woodland. There are 5 core conservation areas in the study area surrounded by a matrix of human-use and natural landscape (Figure 2.1). Compared to the rest of Namibia, the Kavango and Zambezi regions have a relatively dense human population, at 6.1 persons per square kilometre (the national average is 2.1 persons per square kilometre) (Namibia Statistics Agency, 2011).   Figure 2.1 Map of the study area including the Kavango and Zambezi regions of Namibia and surrounding area. Potential barriers to buffalo movement and protected areas are also identified.  The study area contains linear features, both man-made and natural, posing as potential barriers to buffalo movement. The human-made barriers include a veterinary fence running east-west along most of the southern border between Namibia and Botswana, covering the western two-thirds of the study area, a north-south veterinary fence in northern Botswana connecting with the border fence, a fence surrounding Mahango Conservation area, and a small section of fence north of Mahango (Figure 1). There are also roads through this region, including a major highway (the Tar Road), and two secondary roads (the Mahango road and the Ring road) (Figure 2.1). The aforementioned rivers comprise potential natural barriers to buffalo movement.  16  2.3 Methods 2.3.1 The study dataset The study dataset comprised of 36 African buffalo (Syncerus caffer) that were fitted with GPS collars in the study area from 2007-2012 (Table 2.1, Figure 2.2). Animals were darted and anesthetized from helicopters, and then collars were attached. The GPS collars were either Sirtrack or Africa Wildlife Tracking brands, programmed to record the animal’s location every 5 hours. The time interval of 5 hours was chosen as a compromise between frequency of observation and collar life (due to battery life constraints), and also to obtain locations of buffalo covering every hour of the day over the collar's lifespan. The temporal span of the entire dataset of GPS locations was September 22, 2007 to May 7, 2013. Of the buffalo collared, 5 were male and 31 were female. Three male buffalo were excluded from the analysis due to insufficient data points. More female buffalo were collared than males due to relative ease on collars – males are harder on collars, often destroying them – and the fact that females aggregate in breeding herds of between 50-500 individuals, so the GPS recordings would represent herd movement patterns. The animals were collared in 7 sites across the study area, which are used to define their spatial groupings in the statistical analysis: Mahango, Buffalo Core, Horseshoe, Susuwe (all part of Bwabwata Park), Mudumu, Nkasa Rupara, and the Eastern Floodplains (Figure 2.2). Table 2.1 Buffalo collars, their collar sites, and migration types.  Collar ID Sex Collar site Migratory behaviour* Number of data points 101108 F Horseshoe Migrant  1568 101109 F Susuwe n/a  855 77259 F Buffalo Core Expansionist  1332 77259new F Mudumu Expansionist  1631 77259new2 F Susuwe n/a  841 77260 F Susuwe Migrant  2132 17  Collar ID Sex Collar site Migratory behaviour* Number of data points 77260new F Mudumu Non-migrant  3350 77261 F Susuwe Migrant  303** 77261new F Buffalo Core Migrant  3584 77262 F Buffalo Core Expansionist  2240 77263 F Susuwe Migrant  1289 77263new F Susuwe Migrant  1901 77264 F Nkasa Rupara Non-migrant  1797 77264new F Nkasa Rupara Non-migrant  3719 77265 F Mudumu Expansionist  2793 77265new F Buffalo Core Expansionist  1148 77266 F Nkasa Rupara Non-migrant  3331 77266new F Nkasa Rupara Non-migrant  2615 94041 F Susuwe Migrant  3114 94042 F Susuwe Migrant  1817 94043 F Horseshoe Migrant  2502 AG269 F Susuwe Migrant  3081 AG270 M Susuwe Migrant  241** AG271 M Eastern Floodplains n/a  69** AG273 M Eastern Floodplains Disperser  756 AG274  M Horseshoe n/a  63** AG275 F Horseshoe Migrant  1235 AG276 F Eastern Floodplains Non-migrant  1038 AG277 F Eastern Floodplains Non-migrant  3881 AG278 F Mudumu Expansionist  4239 AG279 F Mudumu Non-migrant  702 AG280 M Nkasa Rupara n/a  22** AM290 F Nkasa Rupara Non-migrant  3431 AM291 F Mahango Migrant  2138 SAT508 F Buffalo Core n/a  948 SAT509 F Buffalo Core n/a  940 * migratory type identified by Naidoo et al. (2012), ** not enough data points to model this collar  18   Figure 2.2 Buffalo GPS fixes and their collaring sites. GPS fixes are the coloured circles; different coloured triangles represent different sites. Purple is Mahango, green is Buffalo Core, pink is Horseshoe, teal is Susuwe, orange is Mudumu, blue is Nkasa Rupara, and yellow is the Eastern Floodplains. GPS locations were recorded every 5 hours from September 2007 – May 2013.  2.3.2 Environmental variables We compiled a set of environmental variables that we expected to affect buffalo movement behaviour when near or crossing linear barriers. Environmental variables were selected by performing a literature review of buffalo ecology and behaviour, and a literature review of the environmental variables selected in similar movement studies of large ungulates (Sinclair 1977; Prins, 1996; Dyer et al., 2002; Winnie et al., 2008, Loarie et al., 2009; Leblond et al., 2012; Matawa et al.; 2012; Naidoo et al., 2012; Elliot et al., 2014; Naidoo et al., 2014a). Satellite images were obtained detailing vegetation, fire, and rainfall. These variables directly impact resource distribution, which can alter buffalo movement, as buffalo rely heavily on both water and fresh graze (Sinclair, 1977). Processed satellite data containing ‘enhanced vegetation index’ was obtained from the NASA Reverb server covering the course of the study period (EVI – MOD13Q1, NASA LP DAAC, 2001), 16 day composite). EVI is a proxy for vegetation quality; studies have shown that buffalo select for vegetation quality when not restricted by water (Sinclair, 1972; Traill & Bigalke, 2007; Tshabalala et al., 2009; Ryan et al., 19  2012). Daily estimates of fires on the landscape (fire index), were also obtained from NASA Reverb server for the course of the study period (MOD14A1 daily, NASA LP DAAC, 2001b). Fire was included in the study because vegetation growth rates are highest within 3 months of a fire, providing quality forage that may be an attractant for buffalo (Vijver et al., 1999). Daily estimated rainfall data, in the form of satellite images, was obtained from the Tropical Rainfall Monitoring Mission via a NASA server, over the course of the study period (TRMM 3B42 daily v7, TRMM 2013). In savannas, migratory animals follow rainfall gradients; during the growing season they are able to choose high quality vegetation, and during the dry season they select higher rainfall areas (Bauer et al., 2011). Satellite imagery was processed to extract EVI, fire (which was calculated as days since last fire), and rainfall measurements for individual observations throughout the study period. The rainfall data were highly skewed with many more zero than non-zero values at buffalo GPS locations, so this variable was transformed to a binary variable with 0 representing no rain, and anything greater than 0 representing rainfall on that day at the buffalo's location. Fire index was also transformed to a binary variable, fire, where 1 represented there having been a fire within the past 90 days, and 0 represented no fire within 90 days. Spatially explicit vector layers were obtained detailing rivers, channels (smaller waterways), and pans (ephemeral water sources), settlements, and omurambas (narrow linear depressions in the landscape) in the study area, from the Ministry of Environment and Tourism, Namibia and WWF in Namibia. Buffalo must remain close to permanent water sources such as rivers and channels during the dry season, which limits their movements (Stark, 1986; Traill & Bigalke, 2007). Pans are depressions that contain water holes that fill in the wet season, and buffalo use these resources to expand their access to fresh graze.  Our pans layer was derived 20  from a traditional knowledge study surveying land users in the western part of the study area about pans that they use (Taylor et al., 2006).  Vector-based maps for roads and the veterinary fence were also used. The settlement data were derived from a land use map of the study area (World Wildlife Fund, 2005). Omurambas are narrow linear depressions across the landscape occurring at lower elevation than the surrounding landscape. They are preferred habitat for buffalo as they often contain ephemeral water during the wet season as well as grass for foraging. The omuramba layer was derived from a map of vegetation type (Mendelsohn & Roberts, 1997).  The vector features were linked to buffalo location data in ArcMap 10 (ESRI, 2011). Distance to road, river, fence, and settlement was calculated for every buffalo location. For road, river, and fence, distance to feature was converted to a binary variable indicating whether a GPS location was within a certain distance to the feature or outside. For river, the threshold was the river’s floodplain; for roads and fences a threshold of 2km was chosen based on a study examining linear anthropogenic features affecting the movement of large ungulates, where it was an intermediate threshold level of potential disturbance (Leblond et al., 2012). For omurambas, the binary variable indicated whether each location was in an omuramba or not. Each GPS location was classified temporally by whether it was recorded during day or night, and wet or dry season. Information on daily sunrise and sunset times were used to differentiate day/night observations, and movement patterns were examined in conjunction with rainfall patterns to determine season change dates for each year of the study. The temporal initiation of zebra migrations in the study area has been found to coincide with large episodic rainfall events and buffalo net squared displacement (an indicator of migration) is correlated with rainfall (Naidoo et al., 2012; Naidoo et al., 2014a). 21   2.3.3 Statistical analysis All statistical analyses were performed using the R statistical software (R Core Team, 2014).The ltraj function from the “AdehabitatLT” package (Calenge, 2006) was used to generate step lengths (straight line distance between successive locations) and turn angles (the relative turn angle that the animal has taken compared to the direction of its previous step) for every observation (n= 54 729 observations, 36 individuals) in the dataset. Speed was derived as an additional variable by dividing step length by the time expired between successive locations. For each individual dataset, all observations > 5.1 hours apart were removed, and a time lag of one unit was created for speed turn angle. Each dataset was individually examined for GPS recording errors. Two statistical analyses were performed, both using an information-theoretic approach (Burnham & Anderson, 2001). This method compares models in a candidate set between each other and ranks them based on the relative fit of each model given the data. An all-subsets multimodel averaging approach was used for both analyses, in order to explore and determine which coefficients had the greatest impact on buffalo movement, including the relative magnitude and direction of these effects (Burnham & Anderson, 2001). This resulted in 2k candidate models for each analysis, where k is the number of parameters (including the intercept). The MuMIn package was used for this analysis (Barton, 2014). We used Akaike’s information criterion, with a finite sample size correction (AICc) to assess the relative fit of each candidate model to the data. AICc is recommended for use when the sample size is not many times larger than the square of the number of parameters. Use of AICc in this case avoids over fitting models (Burnham & Anderson, 2011). From AICc rankings, coefficients for the average 22  model were calculated as weighted averages of the coefficients from the full set of models, using Akaike weights derived from the AICc values. Model averaging was used to take into account uncertainty in model selection. Shrinkage coefficients were used in the model averaging process. Shrinkage coefficients are closer to zero as they include a zero for each model where the variable is not present when calculating of the average-model coefficient for each variable (Burnham & Anderson, 2001). The first statistical analysis was an exploratory analysis of the relative impact of our various environmental variables, including linear features, on buffalo movement. The proxies for movement used in this analysis were speed and relative turn angle. These were modeled separately as dependent variables in linear models. Models were run for each individual GPS collar, as well as aggregated by collaring site (hereafter ‘site’, Figure 2.2), and all collars aggregated. We used collaring sites to group collars because each site has a distinct set of linear features that we expected to have important localized effects on movement and permeability of linear barriers. Not all individual models contained all independent variables, since not all variables were relevant for the various study sites (for example, proximity to the veterinary fence is not relevant for animals in the eastern floodplains as the nearest such fence is over 200km away). Independent variables were chosen at the site level, so all models for each buffalo in a site (e.g., Susuwe, Horseshoe) contained the same set of independent variables (Figure 2.1). Dry and wet season movements were modeled separately. The variables of interest in the general movement models were: road buffer, river floodplain, and fence buffer. The models included additional variables that we hypothesized would influence buffalo movement: day, omuramba, rain, fire angle, speed, EVI and EVI2, and lag speed/angle. 23  The second statistical analysis was an analysis of buffalo crossing behaviour in relation to potential linear barriers in the study area. Linear features in the study area were identified, including the main tar road, two secondary roads (Mahango and the ring road), the veterinary fence, and two major rivers: the Kavango and the Kwando rivers (Figure 2.1). The movement paths for all 36 animals were examined and each time they crossed one of the linear features, this was recorded.  A buffalo was deemed to have crossed a linear feature when the shortest straight line path between one location and the next location crossed a linear feature (given the nature of our linear features including roads and rivers, there were very few opportunities for animals to travel around these features). We expected fences to be impermeable barriers to buffalo as the veterinary fence in Namibia is aimed at prohibiting buffalo crossings, and is a barrier to other large wildlife in the region (Cozzi et al., 2013), rivers to be semi-permeable (as buffalo can cross rivers, depending on river depth and width, e.g. Ryan et al., 2006), and roads to be semi-permeable barriers (based on studies of other large herbivores, e.g. Dyer et al., 2002; Sheldon, 2005). Buffers of 1km, 2km, and 5km were created surrounding each linear feature. These buffers were selected based on estimates of the influence of linear barriers on other large herbivores (Leblond et al., 2012). For each animal, the number of data points within each buffer was recorded. The percentage of crossing events compared to observations within the buffers varied from 0.5% to 11.3%. To avoid small sample bias, we used Firth’s bias reduced logistic regression with the logistf package in R (Firth, 2008; Heinze et al., 2013). This method is used on rare events logistic data to account for the tendency to under-predict events with rare event data (Firth, 2008). The dependent variable was crossing a linear barrier (1 = crossed, 0 = not 24  crossed). The independent variables in this set of models included day (1 for daytime and 0 for night), season (1 for dry, 0 for wet), speed, angle, fraction tree cover, distance to nearest settlement (when applicable), and in or out of an omuramba (when applicable). We expected buffalo to be more likely to cross linear features when farther from settlements, as traffic, noise, and human presence are all higher closer to settlements and similar effects have been found with other herbivores (Sheldon 2005; Coulon et al., 2008). Presence in or out of an omuramba was included as an independent variable because omurambas are often used as travel corridors for buffalo and also contain ephemeral water sources that fill in the wet season, so we expected that buffalo may be more likely to cross linear barriers when traveling in an omuramba (Naidoo et al., 2012). Additionally, we expected to observe buffalo to display directed movement patterns (smaller turn angles, higher speeds), when crossing roads, as roads are perceived negatively by many herbivores (e.g. Leblond et al., 2012). We expected buffalo to be more likely to cross roads at nighttime, as traffic is likely lower. We also expected buffalo to be more likely to cross roads in the wet season, when they expand their range to track high quality forage.   2.4 Results 2.4.1 Movement models Site-aggregated models and the ‘all collars’ aggregated models are reported in this section. For individual collar results, see Appendix A2. In the all collars models, buffalo moved significantly slower in river floodplains across the study area in both the wet and dry seasons (Table 2.1, 2.2), with a larger turn angle in the wet season (Appendix A1, A2). When split into site-level models, the results supported the all collars models, with buffalo moving faster in the river floodplain for all sites (in at least one season) except Mahango. Turn angle in the river 25  floodplain at the site level was variable; individuals in Buffalo Core had a larger turn angle (wet season) and in Susuwe, a smaller turn angle (dry season). For the road buffer variable, in the all collars models buffalo moved significantly slower within the road buffer (dry season), with smaller turn angles (Table 2.2). Additionally, the road buffer variable was included in three of the site models (Buffalo Core, Mahango, and Susuwe). For the Buffalo Core site model, individuals moved faster in the road buffer. Conversely, buffalo moved slower in the road buffer for the Mahango (dry season) and Mudumu (wet season) site models, with smaller turn angles (dry season). Distance to settlement was not included in the all collars models because two of the five sites either did not have settlements within access of the animals, or data was not available. At the site level, buffalo moved significantly slower when closer to human settlements in Horseshoe, Mudumu, and Nkasa Rupara. Conversely, in the wet season Susuwe model, buffalo moved faster when closer to human settlements. There were too few data points within the fence buffer to model the effects of proximity to the fence when all collars were grouped together; at the site level, fence buffer was not significant at the three sites it was relevant for (Buffalo Core, Horseshoe, Mahango). With respect to the additional variables in the movement analysis, buffalo moved faster with increased rainfall, faster in omurambas, fastest at intermediate EVI values in the dry season and at lower EVI in the wet season, faster at night, faster with a higher lag speed, and faster with a smaller turn angle (speed and angle had a significant negative relationship in the 6 of 16 speed models and 7 of 16 turn angle models) (Tables 2.1, 2.2 and Appendix A1). Fire was only included in one model, the Eastern Floodplains model, and these animals moved faster in areas where fire had occurred within 90 days.  26  Table 2.2 All collars, dry season, speed. Variable Estimate Std. Error z-value Pr(>|z|) Lower Confidence Interval (2.50%) Upper Confidence Interval (97.50%) River -0.146 0.016 9.254 < 0.001 -0.177 -0.115 Road -0.070 0.033 2.112 0.0347 -0.136 -0.005 Day -0.054 0.015 3.518 < 0.001 -0.085 -0.024 In omuramba 0.1630 0.034 4.72 < 0.001 0.095 0.231 Rain 0.0949 0.023 4.176 < 0.001 0.050 0.139 Turn angle -0.089 0.015 5.871 < 0.001 -0.118 -0.059 EVI 0.066 0.016 4.109 < 0.001 0.034 0.097 EVI^2 -0.075 0.018 4.084 < 0.001 -0.112 -0.039 Lag speed 0.058 0.015 3.812 < 0.001 0.028 0.089  Table 2.3 All collars, wet season, speed Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River -0.114 0.017 6.851 < 0.001 -0.147 -0.081 Road 0.019 0.025 0.788 0.431 -0.029 0.068 Day -0.118 0.013 9.284 < 0.001 -0.143 -0.093 In omuramba 0.083 0.021 3.869 < 0.001 0.041 0.125 Rain 0.062 0.013 4.834 < 0.001 0.037 0.087 Turn angle -0.096 0.013 7.653 < 0.001 -0.121 -0.072 EVI -0.045 0.013 3.317 < 0.001 -0.071 -0.018 EVI^2 -0.065 0.013 5.106 < 0.001 -0.090 -0.040 Lag speed 0.192 0.013 15.109 < 0.001 0.167 0.217   2.4.2 Crossing models Buffers around of 1, 2, and 5km, around roads in the study area, were all used for modeling the influence of different variables on crossing behaviour; the results reported in this section are only those for 2km buffers since results were similar across all buffer distances (see Appendix A3 for 1 and 5km buffer results). Our results indicate both impermeable and semi-permeable barriers across the study area. Major rivers acted as an impermeable barrier to buffalo movement, except one occasion where a 27  buffalo in the Eastern Floodplains crossed the Chobe River. Out of 54 729 total data points collected, just over a quarter of those (15 488) were within 2km of a river. The Mahango park fence similarly acted as an impermeable barrier to buffalo movement; there were no crossings out of 505 observations within 2km of the Mahango fence (out of 1 973 data points in Mahango). The veterinary fence was a semi-permeable barrier to movement. There were two instances of buffalo crossing the veterinary fence out of 1 655 data points that were within 2km of it. The roads in the study were semi-permeable barriers to movement. In Mahango, buffalo were observed to cross the road 48 times out of 505 instances of the animal being within 2km of the road. The Ring Road was crossed 115 times in 1 684 instances of an animal within 2km of the road). All 115 crossings were from Mudumu animals (1 523 observations); the Nkasa Rupara animals never crossed the road (161 instances)  The tar road in the study area was crossed by Buffalo Core and Susuwe animals, but not by Horseshoe animals. There were a total of 2 367 observations within 2km of the tar road and 128 crossings. Horseshoe buffalo were observed within 2km of the tar road 142 times while never crossing the road. There were 587 Buffalo Core animal observations within 2km of the tar road, and the road was crossed 12 times. Susuwe animals were recorded within 2km of the tar road 1 638 times, and crossed the road 116 times.  Crossing behaviour appears to be affected by the three roads in the study area. For all three roads, buffalo moved significantly faster when crossing the road than when not crossing (Tables 2.3-2.5). Additionally, for the Tar and Mahango roads, relative turn angle was significantly smaller when crossing versus not crossing (Tables 2.3, 2.4). In the Tar and Ring road crossing models, distance to settlement was positive and significant, indicating that buffalo were more likely to cross the road further from human settlements (Table 2.4). When models 28  were further split by site and individual collar, distance to settlement was significant and positive for the Mudumu crossing model and collar AG278 model (within Mudumu). For the rest of the collars that had enough data points to be modeled, speed was significant but not distance to settlement. In Mahango, buffalo were more likely to cross the road in the wet season than the dry season.  Table 2.4 Tar road crossing model, 2000m. Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.50%) Upper Confidence Interval (97.50%) Angle -0.565 0.216 2.615 0.009 -0.988 -0.141 Settlement 0.793 0.285 2.785 0.005 0.235 1.351 Day -0.380 0.234 1.623 0.104 -0.838 0.079 Dry season 0.194 0.266 0.729 0.466 -0.328 0.716 EVI -0.172 0.251 0.685 0.493 -0.664 0.320 In omuramba 0.222 0.220 1.011 0.312 -0.209 0.653 Speed 2.13 0.263 8.089 <0.001 1.614 2.646  Table 2.5 Ring road crossing model, 2000m.  Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Angle 0.063 0.212 0.299 0.765 -0.352 0.479 Settlement 2.041 0.462 4.416 <0.001 1.135 2.947 Day -0.062 0.210 0.294 0.769 -0.472 0.349 Dry season -0.164 0.387 0.425 0.671 -0.922 0.593 EVI -0.452 0.479 0.943 0.346 -1.392 0.488 Speed 1.893 0.204 9.258 <0.001 1.492 2.293      29  Table 2.6 Mahango crossing model, 2000m. Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Angle -1.048 0.37822 2.771 0.006 -1.78945 -0.30686 Day -0.090 0.28206 0.318 0.751 -0.64239 0.463277 Dry -4.190 0.95953 4.367 <0.001 -6.07095 -2.30968 Speed 1.792 0.30636 5.851 <0.001 1.191974 2.392878 EVI 0.096 0.32347 0.298 0.765 -0.53746 0.730497 In omuramba -0.382 0.6839 0.559 0.576 -1.7225 0.958341  2.5 Discussion In this study, we used a dataset of GPS collar locations and a multimodel averaging approach to demonstrate that connectivity in the KAZA TFCA is impacted in some places by man-made linear features, and that buffalo alter their behaviour in proximity to both man-made and natural linear barriers.  2.5.1 Which linear features on the landscape are barriers, and which of those are impermeable or semi permeable? Our analysis revealed that permeability of linear barriers varied both among and within types of features on the landscape. The impermeability of the rivers in our study area restricts east-west movement of buffalo, and in the far east of the study area, north-south movement. Fences were only crossed when external circumstances (such as low water level, or a likely breach by elephants) allowed buffalo to cross, further restricting southward movement into Botswana in the eastern half of the study area. The roads were semi-permeable barriers, crossed in some places but not others, making northward movement partially (but not completely) compromised in some areas. There is a wealth of previous work examining the barrier effect of roads, and our results 30  are consistent with other work on large ungulates finding that the permeability of roads is variable depending on other factors such as traffic and fencing (Forman & Alexander, 1998; Shepard et al., 2008; Trombulak & Frissell, 2012). Rivers are impermeable barriers to cheetahs, hyenas, and wild dogs elsewhere in KAZA, but are permeable to lions (Cozzi et al., 2013). By gaining a more complete understanding of how linear barriers affect wildlife movement when in proximity to, and crossing, these features, managers can begin to understand how certain interventions may help restore connectivity, or mitigate further losses in connectivity.   2.5.2 How do buffalo behave in proximity to linear features? We found that buffalo moved more slowly in the river floodplains, indicating that floodplains are used for encamped behaviour, such as grazing, drinking, and ruminating (Morales et al., 2004; Fryxell et al., 2008). This is consistent with previous work showing that when buffalo do not have easy access to many smaller permanent water sources (such as permanent water holes), they select habitat close to rivers and lakes (Sinclair, 1977; Prins, 1996; Redfern et al., 2003). River floodplains are also preferred land for agriculture, so there is the potential for human-wildlife conflict when buffalo range outside of protected areas. In a study of the Zambezi valley, researchers found buffalo use of river-adjacent habitats led to increased human-wildlife conflict in the area, as human development was also concentrated along the river floodplain, although buffalo did avoid agricultural land uses in the river floodplains (Matawa et al., 2012). Currently, the majority of buffalo use of river floodplains in the study area does not overlap with human use, but as the region develops, agriculture may expand. Potential solutions exist, such as the expansion of protected area for buffalo, or use of artificial water sources for wildlife. A conservation intervention in South Africa allowed elephants to rely much less on river floodplain 31  habitat (Loarie et al., 2009). As a consequence, though, the animals did not vary their space use seasonally. This could lead to overexploitation of the local flora with no time for recovery, which could lead to significant habitat changes. Habitat alterations by large herbivores can cause cascading effects to other herbivores, ultimately altering ecosystem structure and function and potentially rendering the habitat unsuitable for certain species (Ripple, 2015).  The effect of road proximity on behaviour has been studied on other large mammals, including ungulates, but not extensively on buffalo (Dyer et al., 2001; Dyer et al., 2002; Sheldon, 2005; Waller & Servheen, 2005; Holdo et al., 2011; Leblond et al., 2012; Rytwinski & Fahrig, 2013; Sawyer et al., 2013b; Beyer et al., 2014; Elliot et al., 2014; Frair et al., 2014; Seidler & Long, 2014). Studies have found that large ungulates avoid roads up to a distance of 1.25km, and behavioural effects of roads on ungulate movement extend up to 5km, including the effect of moving faster when closer to roads for both mule deer and caribou (Dyer et al., 2001; Sawyer et al., 2009; Leblond et al., 2012; Sawyer et al., 2013b; Beyer et al., 2014). In this study, buffalo moved faster within the Tar Road buffer, suggesting that the habitat surrounding the tar road is sub-optimal (Sawyer et al., 2013b). The Tar road is the road in the study area with the highest traffic and has many settlements along it (Figure 1), so this could contribute to creating adjacent undesirable habitat, due to increased human presence and the noise and vehicle disturbance associated with human settlements. Conversely, buffalo moved slower in the Mahango and Ring Road buffers than outside. There are no settlements along the Mahango road, and none along the section of the Ring Road where the Mudumu buffalo reside, so there might be less of an indirect undesirable habitat effect. The Ring and Mahango roads are also less than 5km from major rivers in both Mudumu and Mahango sites, so the habitat within 2km of the road could be ideal graze.  32  The differing effects of the roads in the study area on buffalo movement show that the effect on behaviour is more nuanced than simply the physical structure of a road, and there must be other contributing factors, such as traffic, or proximity to necessary resources such as water. All three roads in the study area cross through protected areas, and therefore their effects on animal behaviour are important for conservation planning. Planners must be able to anticipate when a road will cause a buffer of undesirable habitat, as this can fragment habitat in protected areas. The results of the crossing analysis add additional insight into the impact of roads in the study area on buffalo movement.  2.5.3 How do buffalo behave when crossing linear features? Our finding that buffalo moved faster when crossing roads suggests that roads have a negative effect on buffalo in the study area and that buffalo perceive roads as undesirable habitat. We also found that buffalo were more likely to cross the tar road when further from settlements, suggesting an indirect effect of settlements that causes the road to become an impermeable barrier. This is likely the extra traffic, physical infrastructure, and human presence located in and around settlements. Lions in KAZA TFCA have also been found to avoid settlements (Elliot et al., 2014), and studies in the northern hemisphere have revealed traffic to be an important factor in determining the permeability of roads to large mammal movement, along with road network density and road fencing (Dyer et al., 2002; Sheldon, 2005; Waller & Servheen, 2005; Leblond et al., 2012; Sawyer et al., 2013a; Frair et al., 2014; Siedler & Long, 2014). With the development of KAZA TFCA, there is an expected increase in human presence that will take the form of increased traffic from tourists and additional human infrastructure to accommodate this tourism. This anticipated development could reduce the permeability of roads in the study area to 33  buffalo movement, decreasing connectivity and possibly barring wildlife from dispersing and migrating between seasonal ranges. A limitation of our study is that the buffalo were almost all adult females so we were unable to test for differences in behaviour related to linear barriers by age and sex. Of the two male buffalo included in the analysis, one was in the eastern floodplains, where there were no anthropogenic linear barriers to movement, and one was in Susuwe with insufficient observations to model this animal’s road-crossing behaviour individually. A study of lion movement in KAZA revealed that behaviour around barriers (avoidance, willingness to cross) was related to sex and demographic status, which predicted different levels of risk adverse behaviour in individuals (Elliot et al., 2014). For a complete understanding of buffalo behaviour in relation to linear barriers, it’s important to consider the behaviour of not just herding females, but also reproductive-aged males that do not travel in herds. Young, dispersing juvenile lions were more risk averse, which could also be the case for young male buffalo (Elliot et al., 2014). Failure to take these differences into account could limit the success of a conservation plan.  2.5.4 General discussion and conclusions In planning for the conservation side of the KAZA initiative, priority corridors for wildlife movement have been identified. The top priority corridor passes from Botswana through Namibia to Angola and Zambia, crossing through Nkasa Rupara, Mudumu, Susuwe, and Horseshoe (Ferguson & Hanks, 2010). The potential barriers to the connectivity of this corridor include settlements, roads, fences, and land use. This study contributes to our understanding of how those potential barriers buffalo, a large, wide-ranging mammal. The major north-south barrier to this migratory route is the Tar Road, which is currently crossed by Susuwe animals but 34  not Horseshoe animals.  We also found that proximity to human settlements influences the ability of buffalo to cross roads in the study area, and that fences are impermeable barriers. These empirical results can guide management of this corridor in KAZA to ensure that it links wildlife populations on the ground, and continues to do so as the region develops.  Overall, we found that connectivity in the study area is compromised in some places by human development. This supports the findings of a recent genetic analysis that revealed that gene flow between buffalo in the study area are is more restricted than it historically (Epps et al., 2013). As heterozygosity in buffalo populations elsewhere in the KAZA TFCA have also decreased, it is possible that the loss in connectivity has had a negative effect on the genetic diversity of buffalo in the Namibia part of KAZA as well.  We also found areas of high buffalo preference, river floodplains, which are also areas where human development tends to focus. These findings highlight the tensions between human development and biodiversity conservation that are on the ground realities in conservation landscapes where wildlife and humans must coexist, and specifically in the development of KAZA TFCA. Through coordinated management efforts, there is the possibility of mitigating further losses in connectivity by carefully managing human development along linear features, both the rivers and roads. This study contributes to a quantitative understanding of the impact of linear structures on connectivity, which could add an extra layer of insight to connectivity planning in KAZA TFCA and other multi-use landscapes.   35  Chapter 3: Conclusion 3.1 Overall significance and contribution The results of this thesis demonstrate that there are a number of linear features in the study area acting as barriers to movement of varying permeability. These results indicate that connectivity in the study area is compromised by a combination of man-made and natural features. The movement and crossing models reveal that buffalo alter their behaviour when in proximity to, and crossing, linear barriers. Natural linear features such as rivers contain adjacent habitat for buffalo, which slow their movement in river floodplains. Humans also prefer this high quality habitat, and so the potential for human-wildlife conflict may arise as the human population in this region continues to grow (this has been observed in adjacent Zambia; Matawa et al., 2012). Buffalo increase their speed when crossing roads, and are more likely to cross roads further from human settlements. This suggests that roads are undesirable habitat for buffalo. The avoidance of crossing near human settlements is likely due to the associated human presence, infrastructure, and traffic. Roads bisect the study area from east to west and are a potential source of further connectivity loss in the region. Fences were opportunistically crossed, indicating that, were there no fence, buffalo would likely roam freely between Botswana and Namibia. These results are significant for the planning of the KAZA TFCA and Namibia’s CBNRM program, both of which rely on healthy wildlife populations for generating revenue. As the human population in Namibia (and sub-Saharan Africa in general) is increasing, it’s likely that these connectivity issues will become exacerbated in the future, as the landscape matrix outside of protected areas becomes more crowded, and roads become busier. Conservation managers will likely need to think creatively about how to maintain functional 36  connectivity within KAZA to ensure that it works as a connected network of protected areas, and not a set of isolated habitat patches.  3.2 Analysis and integration of the research The research performed in this thesis contributes a novel application to two niches of movement ecology studies: those that aim to provide tangible results for applied conservation problems, and those that aim to further the scientific understanding of linear barriers on wildlife movement. There is some overlap between these two aims in movement ecology – all studies of impacts of linear barriers can contribute to conservation initiatives, but not all are directly relevant to a specific initiative. Many studies of the impacts of linear barriers on movement occur in industrial landscapes, due to the prevalence of roads, fences, seismic lines, and other human developments in these areas that pose as threats to wildlife movement. In the KAZA TFCA, connectivity has been studied in lions, and a predator guild including lions, cheetahs, hyena, and wild dog (Cozzi et al., 2013; Elliot et al., 2014). This study adds an understanding of impacts of linear barriers on the movement of a large herbivore to the body of connectivity knowledge for KAZA. The aforementioned studies examined crossing behaviour in the first case, and full movement paths in the second case; our study examined both crossing behaviour and behaviour in proximity to impermeable barriers, demonstrating that understanding both crossing and non-crossing proximate behaviour are important to understanding the influence of linear features and barriers on wildlife movement. There are other similar initiatives to KAZA and the CBNRM program in southern Africa, and the findings of this study can apply broadly to the understanding of barriers on wildlife movement throughout multi-use regions aiming to balance biodiversity conservation with human wellbeing. 37  With respect to methods in the analysis of the impacts of linear barriers on movement, we are unaware of another study that has employed multi-model averaging. This method is relatively simple compared to those employed in other studies, but we have shown that it can provide meaningful results contributing to conservation planning and the general understanding of wildlife movement, especially in response to potential barriers to movement. Our general movement model method is similar to the step selection function (e.g. Coulon et al., 2008; Roever et al., 2010), but rather than comparing used steps to “available” steps, we modeled observed movement metrics. Unlike the more sophisticated state space models, our model does not incorporate multiple behavioural modes into the model itself, but from our results we were able to infer different behaviours in relation to environmental variables (such as buffalo moving more slowly, with wider turn angles, in the river floodplain, indicating encamped behaviour). Additionally, in the study of linear barrier effects on wildlife movement, researchers in the northern hemisphere have begun to quantify not just the effects of roads on movement, but the additional effects of traffic on wildlife behaviour in proximity to, and when crossing roads. Findings have demonstrated that traffic volume can create a threshold to the willingness of wildlife species to cross a road (e.g. Sheldon, 2005; Northrup et al., 2012). This has not been examined in the southern hemisphere, as far as we know. While we did not have traffic data, we were able to use human settlement data to find a similar effect – that buffalo were less likely to cross roads when closer to human settlements. This finding supports what has been found in the northern hemisphere, and demonstrates that the effect of traffic can be approximated with data that is often more freely available and does not require field work (such as recording traffic levels).   38  3.3 Strengths and limitations The scope of the thesis research was limited firstly by the study dataset. The data were collected prior to my entry into the master’s program. The selection of (approximately) five-hour intervals between GPS recordings fixes limited the types of behaviour could be studied. There has been a focus in the movement ecology literature on scales of movement (both temporal and spatial), and what GPS fix intervals are appropriate for scales of behavioural study  (e.g. Levin, 2000; Leblond et al., 2011; Kawai & Petrovskii, 2012; Thiebault & Tremblay, 2013; Benhamou, 2014). For example, a collar set with a fix rate of <60 seconds is useful for studying fine scale movements such as area-restricted searches or similar foraging search patterns, but is not useful for the study of home range size, or migration (due to battery life restrictions of GPS units) (Fryxell et al., 2008; Postlethwaite & Dennis, 2013). Daily or less frequent GPS are useful for quantifying home ranges or tracking the larger scale movements of introduced species as they establish new home ranges (John et al., 2008). Additionally, as temporal resolution of GPS fixes becomes coarser, it becomes more difficult to use turn-angle data to distinguish between foraging and travelling states (Postlethwaite & Dennis, 2013). This can be seen in our study results, as the speed models had many more significant variables than the turn-angle models. Another limitation to the study was the resolution of freely available satellite data. The highest resolution for freely available satellite data (EVI, rainfall, and fire data) was 250x250m. The dataset and resolution of available spatial guided our selection of analysis. Initially, we hoped to build a predictive model of buffalo movement, but buffalo select forage based on grass species, or the protein content in grasses, and a 250x250m resolution is too coarse to distinguish grass species composition. Additionally, migrating or dispersing buffalo track pans in the wet season for their water sources, but pans are also smaller than 250x250m (in most cases) and 39  could not be detected by the EVI data. Buffalo are driven by bottom up process more than top down (although predation does have an influence on movement) (Grange & Duncan, 2006; Winnie et al., 2008), and it would be difficult to model their movement accurately without taking into account these two major drivers of movement. Just as the GPS fix resolution limits the type of analysis, it also suits certain analyses quite well. The GPS fix rate of five hours allowed the collars to run for multiple years. Data were collected over the course of six years on 36 different buffalo, and overall there are data from six wet seasons and five dry seasons. Having multi-year data dampens the effect of seasonal variability on buffalo movement, and helps to draw out the effect of the variables of interest, linear barriers. Additionally, the 36 buffalo in the dataset were collared in seven different sites, allowing us to explore the different movement patterns between sites with different features (e.g. some sites were partially or fully fenced, and some had no fences nearby).  The analysis would have benefitted from additional data describing factors relevant to buffalo movement, such as hunting pressure and lion predation. As the conservancies in Namibia’s CBNRM program rely on trophy hunting for a significant portion of their income, it would be important to understand if hunting pressure affects buffalo movement (NACSO, 2014). More generally, including hunting and predation in the analysis would create a fuller understanding of the drivers of buffalo movement, allowing us to more accurately predict the effect of linear barriers on movement. The simplicity of the analysis is one of the strengths of the study. The analysis performed was transparent and replicable, requiring knowledge of the software programs R and ArcGIS, and much of the data, particularly relating to environmental and infrastructure development is freely available. More sophisticated methods are often employed in movement modeling, with 40  state-space models (requiring the ability to understand and build state-space equations) being at the forefront of movement ecology methods. Although we sacrificed sophisticated computation methods, this study shows that relevant results can be obtained from using much simpler methods. The interpretation of movement model results is limited by the way we constructed this set of models. We chose to model all collars individually, as well as grouped by site, and finally all animals grouped together. For the model with all buffalo together, the individuals with a larger number of data points would have a stronger influence on parameter estimation than the individuals with less data points. This may have resulted in some bias in the results toward over-representing the sites that had more data points. We compensated for this by modeling all collars individually; this allowed us to compare what variables were significant between the overall models and the site/individual models. This approach allowed us to discern patterns that were common across the entire study area, such as buffalo moving more slowly in the river floodplains, versus site level or individual differences in what variables were important to movement. An alternate approach would have been to construct a hierarchical model with individual as a random effect, which would have been a compromise between modeling all buffalo individually and the fully pooled model (Jackman, 2008). As we have individual and pooled results, our analysis spans what we would expect a hierarchical analysis to produce.  3.4 Potential applications The findings from this study are relevant for the planning of Namibia’s CBNRM program and the KAZA TFCA. Both programs are focused on the conservation of wildlife species while also encouraging economic growth. The population in Namibia is growing, and with this growth 41  will come further human development (Namibia Statistics Agency, 2011). This development has the potential to impact buffalo movement by increasing traffic along roads, increasing the size of urban areas, adding infrastructure such as fences and roads, and crowding high quality habitat areas (such as river floodplains) preferred by both humans and buffalo. This study has provided the beginnings of an understanding of how linear features in the study area affect buffalo movement, and what factors contribute to the permeability of roads. As noted in Chapter 2, researchers have identified a set of important corridors throughout the KAZA TFCA, and the top priority corridor passes through the Kavango and Zambezi regions of Namibia. The researchers outlined potential impediments to movement through these corridors, including human use. This study confirms that anthropogenic presence and structures can potentially impact the functioning of this corridor, especially with further human growth expected in the region.  3.5 Possible future research directions Building upon this analysis, it is possible to delve further into the impacts of barriers on buffalo movement. Similar studies have built least-cost path networks based on resistance surfaces to demonstrate the best possible dispersal and migratory corridors for wildlife species (Coulon et al., 2008; Cushman & Lewis, 2010; Squires et al., 2013; Elliot et al., 2014). These studies can help planners identify current best options for corridors, and view visually on a map the least amenable regions to wildlife travel. With the addition of traffic data, it would be possible to more completely understand the role of the roads in the study area as barriers to movement. Studies have found traffic thresholds where a road switches from semi-permeable to impermeable; such a threshold would be very 42  useful to planners in the region and simple monitoring techniques could mitigate the loss of road permeability to wildlife. Hunting pressure would be another relevant layer of data to add into the study in the future. Anthropogenic hunting pressure data could be collected through the CBNRM program in Namibia; currently kills are recorded at the resolution of the conservancy, which is too low resolution to be of use in a statistical analysis (R. Naidoo pers. comm.). Lion hunting pressure could be accounted for by GPS collaring lions and tracking their movement, and/or by having local people and hunters collect data on buffalo kills, as other than humans, lions are the main predator for buffalo.43  Bibliography Armsworth, P. R., Chan, K. M. a, Daily, G. C., Ehrlich, P. R., Kremen, C., Ricketts, T. H., & Sanjayan, M. A. (2007). Ecosystem-service science and the way forward for conservation. Conservation Biology : The Journal of the Society for Conservation Biology, 21(6), 1383–4.  Augustine, D. J., & Derner, J. D. (2013). Assessing herbivore foraging behavior with GPS collars in a semiarid grassland. Sensors, 13(3), 3711–23.  Bar-David, S., Bar-David, I., Cross, P. C., Ryan, S. J., Knechtel, C. U., & Getz, W. M. (2009). Methods for assessing movement path recursion with application to African buffalo in South Africa. Ecology, 90(9), 2467–79.  Barnosky, A. D., Matzke, N., Tomiya, S., Wogan, G. O. U., Swartz, B., Quental, T. B., Marshall, C., McGuire, J. L., Lindsey, E. L., Maguire, K. C., Mersey, B., & Ferrer, E. A. (2011). Has the Earth’s sixth mass extinction already arrived? Nature, 471(7336), 51–7.  Barton, K. (2014). MuMIn: Multi-model inference. R package version 1.10.0. Bauer, S., Nolet, B. A., Giske, J., Chapman, J. W., Åkesson, S., Hedenström, A., & Fryxell, J. M. (2011). Cues and decision rules in animal migration. In Animal Migration: A Synthesis (eds J. Milner-Gulland, J. M. Fryxell, & A. R. E. Sinclair), pp. 69–77. Oxford University Press, Oxford, UK. Benhamou, S. (2014). Of scales and stationarity in animal movements. Ecology Letters, 17(3), 261–72.  Beyer, H. L., Gurarie, E., Borger, L., Panzacchi, M., Basille, M., Herfindal, I., Van Moorter, B., Lele, S., & Matthiopoulos, J. (2014). “You shall not pass!”: quantifying barrier permeability and proximity avoidance by animals. Journal of Animal Ecology, 1-11.  Bolger, D. T., Newmark, W. D., Morrison, T. A, & Doak, D. F. (2008). The need for integrative approaches to understand and conserve migratory ungulates. Ecology Letters, 11(1), 63–77.  Buchholz, R. (2007). Behavioural biology: an effective and relevant conservation tool. Trends in Ecology and Evolution, 22(8), 401–407.  Burnham, K., & Anderson, D. (2001). Model Selection and Multimodel Inference: A Practical Approach (2nd ed). Springe-Verlag New York, New York, USA. Cagnacci, F., Boitani, L., Powell, R. A, & Boyce, M. S. (2010). Animal ecology meets GPS-based radiotelemetry: a perfect storm of opportunities and challenges. Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences, 365(1550), 2157–62. Calenge, C. (2006). The package “adehabitat” for the R software: A tool for the analysis of space and habitat use by animals. Ecological Modelling, 197(3-4), 516–519.  Cardinale, B., Duffy, J., Gonzalez, A., Hooper, D., Perrings, C., Venail, P., Narwani, A., Mace, G. M., Tilman, D., Wardle D. M., Kinzig, A. P., Daily G. C., Loreau, M., Grace, J. B., Larigauderie, A., Srivastava, D. S., & Naeem, S. (2012). Biodiversity loss and its impacts on humanity. Nature, 486, 59–67. 44  Cerling, T., Harris, J., & Passey, B. (2003). Diets of East African Bovidae based on stable isotope analysis. Journal of Mammalogy, 84(2), 456–470.  Chadwick, J., Fazio, B., & Karlin, M. (2010). Effectiveness of GPS-Based Telemetry to Determine Temporal Changes in Habitat use and Home-Range Sizes of Red Wolves. Southeastern Naturalist, 9(2), 303–316.  Codron, D., Codron, J., Lee-Thorp, J. a., Sponheimer, M., de Ruiter, D., Sealy, J., Grant, R., Fourie, N. (2007). Diets of savanna ungulates from stable carbon isotope composition of faeces. Journal of Zoology, 273(1), 21–29. Cooke, S. J., Hinch, S. G., Wikelski, M., Andrews, R. D., Kuchel, L. J., Wolcott, T. G., & Butler, P. J. (2004). Biotelemetry: a mechanistic approach to ecology. Trends in Ecology & Evolution, 19(6), 334–43. Costanza, R., & Folke, C. (1997). Valuing ecosystem services with efficiency, fairness, and sustainability as goals. In Nature’s Services: Societal Dependence on Natural Ecosystems (ed G. C. Daily), pp. 49–68. Island Press, Washington, DC, USA Coulon, A., Morellet, N., Goulard, M., Cargnelutti, B., Angibault, J. M., & Hewison, A. J. M. (2008). Inferring the effects of landscape structure on roe deer (Capreolus capreolus) movements using a step selection function. Landscape Ecology, 23(5), 603–614.  Cozzi, G., Broekhuis, F., Mcnutt, J. W., & Schmid, B. (2013). Comparison of the effects of artificial and natural barriers on large African carnivores: Implications for interspecific relationships and connectivity. Journal of Animal Ecology, 82(3), 707–715.  Cushman, S. A., & Lewis, J. S. (2010). Movement behavior explains genetic differentiation in American black bears. Landscape Ecology, 25(10), 1613–1625. Dalziel, B. D., Morales, J. M., & Fryxell, J. M. (2008). Fitting probability distributions to animal movement trajectories: Using artificial neural networks to link distance, resources, and memory. The American Naturalist, 172(2), 248–58. Dirzo, R., & Raven, P. H. (2003). Global sate of biodiversity and loss. Annual Review of Environment and Resources, 28(1), 137–167.  Doerr, V. a J., Barrett, T., & Doerr, E. D. (2011). Connectivity, dispersal behaviour and conservation under climate change: A response to Hodgson et al. Journal of Applied Ecology, 48, 143–147. Dussault, C., Ouellet, J. P., Laurian, C., Courtois, R., Poulin, M., & Breton, L. (2007). Moose movement rates along highways and crossing probability models. The Journal of Wildlife Management, 71(7), 2338-2345. Dyer, S. J., O’Neill, J. P., Wasel, S. M., & Boutin, S. (2001). Avoidance of industrial development by woodland caribou. The Journal of Wildlife Management, 65(3), 531–542.  Dyer, S. J., O’Neill, J. P., Wasel, S. M., & Boutin, S. (2002). Quantifying barrier effects of roads and seismic lines on movements of female woodland caribou in northeastern Alberta. Canadian Journal of Zoology, 80(5), 839–845. Elliot, N. B., Cushman, S. a., Macdonald, D. W., & Loveridge, A. J. (2014). The devil is in the dispersers: Predictions of landscape connectivity change with demography. Journal of 45  Applied Ecology, 51(5), 1169–1178. Epps, C. W., Castillo, J. A., Schmidt-Küntzel, A., Du Preez, P., Stuart-Hill, G., Jago, M., & Naidoo, R. (2013). Contrasting historical and recent gene flow among African buffalo herds in the Caprivi Strip of Namibia. Journal of Heredity, 104(2), 172–181.  ESRI. (2011). ArcGIS Desktop. Environmental Systems Research Institute, Redlands, CA, USA. Fagan, W. F., Lewis, M. A., Auger-Méthé, M., Avgar, T., Benhamou, S., Breed, G., LaDage, L., Schlagel, U. E., Tang, W., Papastamatiou, Y. P., Forester, J. & Mueller, T. (2013). Spatial memory and animal movement. Ecology Letters, 1-14. Ferguson, K., & Hanks, J. (2010). Fencing Impacts: A review of the environmental, social and economic impacts of game and veterinary fencing in Africa with particular reference to the Great Limpopo and Kavango-Zambezi Transfrontier Conservation Areas. University of Pretoria, Pretoria, South Africa.  Firth, D. (2008). Bias Reduction of Maximum Likelihood Estimates. Biometrika, 80(1), 27–38. Forman, R. T. T. (2010). Estimate of the area affected ecologically by the road system in the United States. Conservation Biology, 14(1), 31–35. Forman, R. T. T., & Alexander, L. E. (1998). Roads and Their Major Ecological Effects. Annual Review of Ecology and Systematics, 29(1998), 207–231.  Frair, J. L., Merrill, E. H., Beyer, H. L., Manual, J., & Frair, L. (2014). Thresholds in landscape connectivity and mortality to growing road networks risks in response. Journal of Applied Ecology, 45(5), 1504–1513. Frair, J. L., Merrill, E. H., Visscher, D. R., Fortin, D., Beyer, H. L., & Morales, J. M. (2005). Scales of movement by elk (Cervus elaphus) in response to heterogeneity in forage resources and predation risk. Landscape Ecology, 20(3), 273–287. Fryxell, J. M., Hazell, M., Börger, L., Dalziel, B. D., Haydon, D. T., Morales, J. M., McIntosh, Therese, & Rosatte, R. C. (2008). Multiple movement modes by large herbivores at multiple spatiotemporal scales. Proceedings of the National Academy of Sciences, 105(49), 19114-19119. Gates, C. C., Jones, P., Suitor, M., Jakes, A., Boyce, M. S., Kunkel, K., & Wilson, K. (2012). The Influence of Land Use and Fences on Habitat Effectiveness, Movements and Distribution of Pronghorn in the Grasslands of North America. In Fencing for Conservation: Restriction of Evolutionary Potential or a Riposte to Threatening Processes? (eds M. J. Somers & M. Hayward), pp. 277–294. Springer-Verlag New York, New York, USA. Getz, W. M., Fortmann-Roe, S., Cross, P. C., Lyons, A. J., Ryan, S. J., & Wilmers, C. C. (2007). LoCoH: nonparameteric kernel methods for constructing home ranges and utilization distributions. PLoS One, 2(2), 1-11. Graham, M., & Douglas‐Hamilton, I. (2009). The movement of African elephants in a human‐dominated land‐use mosaic. Animal Conservation, 12, 445–455. Grange, S., & Duncan, P. (2006). Bottom-up and top-down processes in African ungulate communities : resources and predation acting on the relative abundance of zebra and 46  grazing bovids. Ecography, 29(6), 899-907. Gunn, J., Hawkins, D., Barnes, R. F. W., Mofulu, F., Grant, R. A., & Norton, G. W. (2013). The influence of lunar cycles on crop-raiding elephants; evidence for risk avoidance. African Journal of Ecology, 52(2), 129–137. Heinze, G., Ploner, M., Dunkler, D., & Southworth, H. (2013). logistf: Firth’s bias reduced logistic regression. R package version 1.21. Holdo, R. M., Fryxell, J. M., Sinclair, A. R. E., Dobson, A., & Holt, R. D. (2011). Predicted impact of barriers to migration on the Serengeti wildebeest population. PloS One, 6(1), e16370. Horne, J. S., Garton, E. O., Krone, S. M., & Lewis, J. S. (2007). Analyzing animal movements using Brownian bridges. Ecology, 88(9), 2354–63. Ito, T. Y., Lhagvasuren, B., Tsunekawa, A., Shinoda, M., Takatsuki, S., Buuveibaatar, B., & Chimeddorj, B. (2013). Fragmentation of the habitat of wild ungulates by anthropogenic barriers in Mongolia. PloS One, 8(2), e56995, Jackson, S. (2009). Hierarchical Statistical Models. In Bayesian Analysis for the Social Sciences pp. 301–378. Wiley, West Sussex, UK.  Jaeger, J. A. G., & Fahrig, L. (2004). Effects of Road Fencing on Population Persistence. Conservation Biology, 18(6), 1651–1657. Johnson, C. J., Parker, K. L., & Heard, D. (2002). Movement parameters of ungulates and scale-specific responses to the environment. Journal of Animal Ecology, 71(2), 225–235. Johnson, D., London, J., & Lea, M. (2008). Continuous-time correlated random walk model for animal telemetry data. Ecology, 89(5), 1208–1215. Kawai, R., & Petrovskii, S. (2012). Multi-scale properties of random walk models of animal movement: lessons from statistical inference. Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences, 468(2141), 1428–1451.  KAZA TFCA. (2013). Kavango-Zambezi transfrontier conservation area: About us. http://www.kavangozambezi.org/about-us [accessed March 8 2015]. Kissling, D. W., Pattemore, D. E., & Hagen, M. (2014). Challenges and prospects in the telemetry of insects. Biological Reviews, 89(3), 511–530. Kokko, H., & López-Sepulcre, A. (2006). From individual dispersal to species ranges: perspectives for a changing world. Science, 313(5788), 789–91.  Leblond, M., Dussault, C., & Ouellet, J.-P. (2012). Avoidance of roads by large herbivores and its relation to disturbance intensity. Journal of Zoology, 289(1), 32–40. Leblond, M., Frair, J., Fortin, D., Dussault, C., Ouellet, J. P., & Courtois, R. (2011). Assessing the influence of resource covariates at multiple spatial scales: an application to forest-dwelling caribou faced with intensive human activity. Landscape Ecology, 1433–1446.  Lendrum, P. E., Anderson, C. R., Monteith, K. L., Jenks, J. A., & Bowyer, R. T. (2013). Migrating mule deer: Effects of anthropogenically altered landscapes. PloS One, 8(5), e64548. 47  Levin, S. A. (2000). Multiple scales and the maintenance of biodiversity. Ecosystems, 3(6), 498–506. Loarie, S. R., Aarde, R. J. Van, & Pimm, S. L. (2009). Fences and artificial water affect African savannah elephant movement patterns. Biological Conservation, 142(12), 3086–3098. Mashintonio, A. F., Pimm, S. L., Harris, G. M., Aarde, R. J. Van, & Russell, G. J. (2014). Data-driven discovery of the spatial scales of habitat choice by elephants. PeerJ, 19(2), e504. Matawa, F., Murwira, A., & Schmidt, K. S. (2012). Explaining elephant (Loxodonta africana) and buffalo (Syncerus caffer) spatial distribution in the Zambezi Valley using maximum entropy modelling. Ecological Modelling, 242, 189–197.  Mate, B. R., Best, P. B., Lagerquist, B. A., & Winsor, M. H. (2011). Coastal, offshore, and migratory movements of South African right whales revealed by satellite telemetry. Marine Mammal Science, 27(3), 455–476. Mbaiwa, J. E., & Mbaiwa, O. I. (2006). The effects of veterinary fences on wildlife populations in Okavango delta, Botswana. International Journal of Wilderness, 12(3), 17-41. McCarthy, J. L., McCarthy, K. P., Fuller, T. K., & McCarthy, T. M. (2010). Assessing Variation in Wildlife Biodiversity in the Tien Shan Mountains of Kyrgyzstan Using Ancillary Camera-trap Photos. Mountain Research and Development, 30(3), 295–301. McNaughton, S. (1985). Ecology of a grazing ecosystem: the Serengeti. Ecological Monographs, 55(3), 259–294. Mendelsohn, J., & Roberts, C. (1997). An environmental profile and atlas of Caprivi. Gamsberg Macmillan, Windhoek, Namibia. Miguel, E., Grosbois, V., Caron, A., & Boulinier, T. (2013). Contacts and foot and mouth disease transmission from wild to domestic bovines in Africa. Ecosphere, 4(4), 1–32. Morales, J. M., Haydon, D. T., Frair, J., Holsinger, K. E., & Fryxell, J. M. (2004). Extracting more out of relocation data: building movement models as mixtures of random walks. Ecology, 85(9), 2436–2445. Morellet, N., Moorter, B., Cargnelutti, B., Angibault, J.-M., Lourtet, B., Merlet, J., Ladet, S., & Hewison, a. J. M. (2011). Landscape composition influences roe deer habitat selection at both home range and landscape scales. Landscape Ecology, 26(7), 999–1010. Munthali, S. M. (2007). Transfrontier conservation areas: Integrating biodiversity and poverty alleviation in Southern Africa. Natural Resources Forum, 31(1), 51–60. Naidoo, R., Balmford, A., Costanza, R., Fisher, B., Green, R. E., Lehner, B., Malcom, T. R., & Ricketts, T. H. (2008). Global mapping of ecosystem services and conservation priorities. Proceedings of the National Academy of Sciences, 105(28), 9495–9500. Naidoo, R., Chase, M. J., Beytell, P., Du Preez, P., Landen, K., Stuart-Hill, G., & Taylor, R. (2014a). A newly discovered wildlife migration in Namibia and Botswana is the longest in Africa. Oryx.  doi:10.1017/S0030605314000222. Naidoo, R., du Preez, P., Stuart-Hill, G., Bytell, P., & Taylor, R. (2014b). Long range migrations and dispersals of African buffalo (Syncerus caffer) in the Kavango–Zambezi Transfrontier Conservation area. African Journal of Ecology, 52(4), 581–584. 48  Naidoo, R., du Preez, P., Stuart-Hill, G., Chris Weaver, L., Jago, M., & Wegmann, M. (2012). Factors affecting intraspecific variation in home range size of a large African herbivore. Landscape Ecology, 27(10), 1523–1534.  Naidoo, R., Du Preez, P., Stuart-Hill, G., Jago, M., & Wegmann, M. (2012). Home on the range: Factors explaining partial migration of African buffalo in a tropical environment. PLoS ONE, 7(5), e36527. Naidoo, R., Stuart-Hill, G., Weaver, L. C., Tagg, J., Davis, A., & Davidson, A. (2010). Effect of diversity of large wildlife species on financial benefits to local communities in northwest Namibia. Environmental and Resource Economics, 48(2), 321–335.  Naidoo, R., Weaver, L. C., De Longcamp, M., & Du Plessis, P. (2011a). Namibia’s community-based natural resource management programme: an unrecognized payments for ecosystem services scheme. Environmental Conservation, 38(4), 445–453.  Naidoo, R., Weaver, L. C., Stuart-Hill, G., & Tagg, J. (2011b). Effect of biodiversity on economic benefits from communal lands in Namibia. Journal of Applied Ecology, 48(2), 310–316. Namibia Statistics Agency. (2011). Namibia 2011 Population & Housing Census Main Report. Namibia Statistics Agency, Windhoek, Namibia. NASA Land Processes Distributed Active Archive Center (LP DAAC). (2001a). MOD13Q1. USGS/Earth Resources Observation and Science (EROS) Center, Sioux Falls, South Dakota, USA. NASA Land Processes Distributed Active Archive Center (LP DAAC). (2001b). MOD14A1. USGS/Earth Resources Observation and Science (EROS) Center, Sioux Falls, South Dakota, USA. NACSO. (2014). The state of community conservation in Namibia: A review of communal conservancies, community forests, and other CBNRM initiatives. National Association of CBNRM Support Organizations. Windhoek, Namibia. Nathan, R. (2008). An emerging movement ecology paradigm. Proceedings of the National Academy of Sciences of the United States of America, 105(49), 19050–19051. Northrup, J. M., Pitt, J., Muhly, T. B., Stenhouse, G. B., Musiani, M., & Boyce, M. S. (2012). Vehicle traffic shapes grizzly bear behaviour on a multiple-use landscape. Journal of Applied Ecology, 49(5), 1159–1167. Pérez-García, J. M., Margalida, A., Afonso, I., Ferreiro, E., Gardiazábal, A., Botella, F., & Sánchez-Zapata, J. A. (2012). Interannual home range variation, territoriality and overlap in breeding Bonelli’s Eagles (Aquila fasciata) tracked by GPS satellite telemetry. Journal of Ornithology, 63–71.  Pettorelli, N., Lobora, a. L., Msuha, M. J., Foley, C., & Durant, S. M. (2010). Carnivore biodiversity in Tanzania: Revealing the distribution patterns of secretive mammals using camera traps. Animal Conservation, 13(2), 131–139. Postlethwaite, C. M., & Dennis, T. E. (2013). Effects of temporal resolution on an inferential model of animal movement. PloS One, 8(5), e57640. 49  Prins, H. H. T. (1996). Ecology and behaviour of the African buffalo: Social inequality and decision making. Chapman & Hall, London, UK. R Core Team. (2014). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria Ramutsindela, M. (2007). Transfrontier Conservation in Africa: At the Confluence of Capital, Politics and Nature. CABI Publishing, Cambridge, MA, USA. Rands, M., Adams, W., & Bennun, L. (2010). Biodiversity conservation: challenges beyond 2010. Science, 329(5997), 1298–1303. Redfern, J. V, Grant, C. C., Biggs, H. C., & Getz, W. M. (2003). Surface water constraints on herbivore foraging in the Kruger National Park, South Africa. Ecology, 84(8), 1–52. Richard, Y., & Armstrong, D. P. (2010). Cost distance modelling of landscape connectivity and gap-crossing ability using radio-tracking data. Journal of Applied Ecology, 47(3), 603–610.  Ripple, W. J., Newsome, T. M., Wolf, C., Dirzo, R., Everatt, K. T., Galetti, M., Hayward, M. W., Kerley, G. I. H., Levi, T., Lindsey, P. A., Macdonald, D. W., Malhi, Y., Painter, L. E., Sandom, C. J., Terborgh, J., & Van Valkenburgh, B. (2015). Collapse of the world’s largest herbivores. Science Advances, 1(4), e1400103. Roever, C. L., Boyce, M. S., & Stenhouse, G. B. (2010). Grizzly bear movements relative to roads: application of step selection functions. Ecography, 33(6), 1113–1122.  Russell, D. J. F., Brasseur, S. M. J. M., Thompson, D., Hastie, G. D., Janik, V. M., Aarts, G., McClintock, B. T., Matthiopoulos, J., Moss, S. E., W., & McConnell, B. (2014). Marine mammals trace anthropogenic structures at sea. Current Biology, 24(14), R638–R639. Ryan, S. J., Cross, P. C., Winnie, J., Hay, C., Bowers, J., & Getz, W. M. (2012). The utility of normalized difference vegetation index for predicting African buffalo forage quality. The Journal of Wildlife Management, 76(7), 1499–1508. Ryan, S. J., Knechtel, C. U., & Getz, W. M. (2006). Range and Habitat Selection of African Buffalo in South Africa. The Journal of Wildlife Management, 70(3), 764–776.  Rytwinski, T., & Fahrig, L. (2013). Why are some animal populations unaffected or positively affected by roads? Oecologia, 173(3), 1143-56. Saunders, D. A., Hobbs, R. J., Margules, C. R., Saunders, D. A., & Hobbs, R. J. (1992). Biological consequences of ecosystem fragmentation: A review. Biological Conservation, 59(1), 18-32. Sawyer, H., Kauffman, M. J., Middleton, A. D., Morrison, T. A., Nielson, R. M., & Wyckoff, T. B. (2013). A framework for understanding semi-permeable barrier effects on migratory ungulates. Journal of Applied Ecology, 50(1), 68–78. Sawyer, H., Kauffman, M. J., Nielson, R. M., & Horne, J. S. (2009). Identifying and prioritizing ungulate migration routes for landscape-level conservation. Ecological Applications, 19(8), 2016–25. Seidler, R. G., & Long, R. (2014). Identifying Impediments to Long Distance Mammal Migrations. Conservation Biology, 29(1), 99-109. 50  Semeniuk, C. A., Musiani, M., Hebblewhite, M., Grindal, S., & Marceau, D. J. (2012). Evaluating risk effects of industrial features on woodland caribou habitat selection in west central Alberta using agent-based modelling. Procedia Environmental Sciences, 13(2011), 698–714. Service, C. N., Adams, M. S., Artelle, K. A., Paquet, P., Grant, L. V., & Darimont, C. T. (2014). Indigenous knowledge and science unite to reveal spatial and temporal dimensions of distributional shift in wildlife of conservation concern. PLoS ONE, 9(7), e101595. Sheldon, D. & Lindzey, F. (2005). Movement and distribution patterns of Pronghorn in relation to roads and fences in southwestern Wyoming. Wyoming Cooperative and Wildlife Research Unit, Wyoming, US. Shepard, D. B., Kuhns, A. R., Dreslik, M. J., & Phillips, C. A. (2008). Roads as barriers to animal movement in fragmented landscapes. Animal Conservation, 11(4), 288–296. Sinclair, A. R. E. (1972). Food selection and competition in the East African buffalo (Syncerus caffer Sparrman). African Journal of Ecology, 10, 77–89.  Sinclair, A. R. E. (1977). The African Buffalo: a study of resource limitation of populations. University of Chicago Press, Chicago, US. Smitz, N., Cornélis, D., Chardonnet, P., Caron, A., de Garine-Wichatitsky, M., Jori, F., Mouton, A., Latinne, A., Pigneur, L-M., Melletti, M., Kanapeckas, K., Marescaux, J., Pereira, C., & Michaux, J. (2014). Genetic structure of fragmented southern African populations of buffalo (Syncerus caffer caffer). BMC Evolutionary Biology, 14(203), 1–38. Smouse, P. E., Focardi, S., Moorcroft, P. R., Kie, J. G., Forester, J. D., & Morales, J. M. (2010). Stochastic modelling of animal movement. Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences, 365(1550), 2201–11. Soule, M. E. (1985). What is Conservation Biology? A new synthetic discipline addresses the dynamics and problems of perturbed and ecosystems. BioScience, 35(11), 727–734. Spellerberg, I. F. (2007). Ecological effects of roads and traffic : A literature review. Global Ecology and Biogeography, 7(5), 317–333.  Squires, J. R., DeCesare, N. J., Olson, L. E., Kolbe, J. A., Hebblewhite, M., & Parks, S. A. (2013). Combining resource selection and movement behavior to predict corridors for Canada lynx at their southern range periphery. Biological Conservation, 157, 187–195.  Stark, M. A. (1986). Daily movement, grazing activity and diet of savanna buffalo, Syncerus caffer brachyceros, in Benoue National Park, Cameroon. African Journal of Ecology, 24, 255–262. Taylor, J., Murphy, C., & Mayes, S. (2006). Land and natural resource mapping by San communities and NGOs: experiences from Namibia. Participatory Learning and Action 54, 79–84.  Thiebault, A., & Tremblay, Y. (2013). Splitting animal trajectories into fine-scale behaviorally consistent movement units: breaking points relate to external stimuli in a foraging seabird. Behavioral Ecology and Sociobiology, 67(6), 1013–1026. Thomson, G. R., Penrith, M.-L., Atkinson, M. W., Atkinson, S. J., Cassidy, D., & Osofsky, S. A. 51  (2013). Balancing livestock production and wildlife conservation in and around southern Africa’s transfrontier conservation areas. Transboundary and Emerging Diseases, 60(6), 492–506. Traill, L., & Bigalke, R. (2007). A presence‐only habitat suitability model for large grazing African ungulates and its utility for wildlife management. African Journal of Ecology, 45(3) 347–354. Trakhtenbrot, A., Nathan, R., Perry, G., & Richardson, D. M. (2005). The importance of long-distance dispersal in biodiversity conservation. Diversity and Distributions, 11(2), 173–181. Trombulak, S. C., & Frissell, C. A. (2012). Review of ecological effects of roads on terrestrial and aquatic communities and of aquatic ecological effects of roads on terrestrial communities, Conservation Biology 14(1), 18–30. Tropical Rainfall Measurement Mission Project (TRMM). (2013). TRMM 3B42 V7. Greenbelt, MD, USA: NASA Goddard Earth Sciences Data and Information Services Center (GES DISC). Tshabalala, T., Dube, S., & Lent, P. C. (2009). Seasonal variation in forages utilized by the African buffalo (Syncerus caffer) in the succulent thicket of South Africa. African Journal of Ecology, 48(2), 438–445. Turner, W. R., Brandon, K., Brooks, T. M., Gascon, C., Gibbs, H. K., Lawrence, K. S., Mittermeier, R. A., & Selig, E. R. (2012). Global biodiversity conservation and the alleviation of poverty. BioScience, 62(1), 85–92.  Vanak, A. T., Thaker, M., & Slotow, R. (2010). Do fences create an edge-effect on the movement patterns of a highly mobile mega-herbivore? Biological Conservation, 143(11), 2631–2637. Vijver, C. A. D. M. Van De, Poot, P., & Prins, H. H. T. (1999). Causes of increased nutrient concentrations in post-fire regrowth in an East African savanna. Plant and Soil 214, 173–185. Villegas-Ríos, D., Alós, J., March, D., Palmer, M., Mucientes, G., & Saborido-Rey, F. (2013). Home range and diel behavior of the ballan wrasse, Labrus bergylta, determined by acoustic telemetry. Journal of Sea Research, 80, 61–71.  Visconti, P., Pressey, R. L., Giorgini, D., Maiorano, L., Bakkenes, M., Boitani, L., Alkemade, R., Falcucci, A., Chiozza, F., & Rondinini, C. (2011). Future hotspots of terrestrial mammal loss. Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences, 366(1578), 2693–702.  Vosloo, W., Bastos, A. D., Michel, A, & Thomson, G. R. (2001). Tracing movement of African buffalo in southern Africa. Revue Scientifique et Technique (International Office of Epizootics), 20(2), 630–9.  Waller, J. S., & Servheen, C. (2005). Effects of transportation infrastructure on grizzly bears in northwestern Montana. The Journal of Wildlife Management, 69(3), 985–1000.  Williams, P. J., Gutiérrez, R. J., & Whitmore, S. A. (2011). Home range and habitat selection of spotted owls in the central Sierra Nevada. The Journal of Wildlife Management, 75(2), 333–52  343.  Winnie, J. J., Cross, P., & Getz, W. (2008). Habitat quality and heterogeneity influence distribution and behavior in African buffalo (Syncerus caffer). Ecology, 89(5), 1457–1468. World Wildlife Fund. (2005). KAZA LandCover 2005. Zeller, K. A., McGarigal, K., Beier, P., Cushman, S. A., Vickers, T. W., & Boyce, W. M. (2014). Sensitivity of landscape resistance estimates based on point selection functions to scale and behavioral state: pumas as a case study. Landscape Ecology, 29(3), 541–557.  Zeller, K. A., McGarigal, K., & Whiteley, A. R. (2012). Estimating landscape resistance to movement: A review. Landscape Ecology, 27(6), 777–797. 53  Appendices Appendix A  Additional Statistical Results  This appendix contains additional statistical results, referred to in the results section of Chapter 2. A.1 Additional results from movement model, split by site Table A.1 All collars, dry season, turn angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer 0.013 0.015 0.875 0.381 -0.016 0.043 Road buffer -0.079 0.026 3.039 0.002 -0.130 -0.028 Rain 0.022 0.023 0.955 0.340 -0.023 0.067 Lag angle 0.080 0.013 6.076 < 0.001 0.054 0.106 Speed -0.082 0.013 6.140 < 0.001 -0.108 -0.056 EVI2 -0.001 0.009 0.159 0.874 -0.018 0.015 EVI -0.001 0.007 0.107 0.915 -0.015 0.014 In omuramba -0.002 0.016 0.144 0.885 -0.035 0.030 Day 0.000 0.007 0.010 0.992 -0.014 0.013            54  Table A.2 All collars, wet season, turn angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer 0.045 0.018 2.506 0.012 0.010 0.081 Day -0.014 0.014 0.998 0.318 -0.043 0.014 Rain 0.013 0.014 0.943 0.346 -0.015 0.041 EVI -0.025 0.016 1.634 0.102 -0.056 0.005 Lag angle 0.055 0.012 4.474 <0.001 0.031 0.079 Speed -0.098 0.012 7.990 <0.001 -0.122 -0.074 In omuramba -0.002 0.011 0.197 0.844 -0.025 0.020 Road buffer -0.002 0.012 0.138 0.890 -0.025 0.022 EVI2 -0.001 0.007 0.099 0.921 -0.013 0.012  Table A.3 Buffalo Core, dry season, turn angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) In omuramba -0.117 0.090 1.297 0.194 -0.293 0.060 EVI -0.076 0.047 1.626 0.104 -0.167 0.016 Speed -0.036 0.041 0.882 0.378 -0.116 0.044 Settlement -0.021 0.035 0.603 0.547 -0.090 0.048 River buffer 0.017 0.033 0.518 0.604 -0.048 0.082 Pan 0.006 0.022 0.257 0.797 -0.038 0.050 Road buffer 0.010 0.046 0.214 0.831 -0.080 0.100 Day -0.003 0.020 0.154 0.878 -0.042 0.036 Rain 0.003 0.025 0.120 0.904 -0.046 0.053 EVI2 -0.006 0.032 0.178 0.858 -0.068 0.057 Lag angle 0.002 0.019 0.119 0.905 -0.036 0.040       55  Table A.4 Buffalo core, dry season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Intercept -1.815 0.027 67.938 < 0.001 -1.867 -1.762 River buffer -0.251 0.057 4.403 < 0.001 -0.362 -0.139 Road buffer -0.216 0.133 1.624 0.104 -0.476 0.045 Day -0.226 0.048 4.704 < 0.001 -0.320 -0.132 Rain 0.088 0.078 1.137 0.255 -0.064 0.240 Angle -0.040 0.051 0.785 0.432 -0.139 0.060 Pan 0.114 0.062 1.828 0.067 -0.008 0.236 Settlement -0.117 0.067 1.744 0.081 -0.248 0.014 EVI 0.051 0.057 0.888 0.374 -0.061 0.163 Lag speed 0.174 0.049 3.514 < 0.001 0.077 0.270 EVI2 0.016 0.048 0.340 0.734 -0.078 0.111 In omuramba -0.003 0.050 0.068 0.946 -0.102 0.095  Table A.5 Buffalo core, wet season, turn angle model Variable Estimate Std. Error Ad z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer 0.140 0.052 2.714 0.007 0.039 0.241 Day 0.031 0.034 0.892 0.373 -0.037 0.098 Lag angle 0.050 0.038 1.320 0.187 -0.024 0.124 Speed -0.077 0.036 2.139 0.032 -0.147 -0.006 Fence buffer -0.029 0.054 0.534 0.594 -0.134 0.076 EVI -0.013 0.025 0.506 0.613 -0.062 0.037 In omuramba 0.018 0.038 0.488 0.626 -0.056 0.092 Road buffer 0.016 0.040 0.395 0.693 -0.063 0.094 Settlement -0.004 0.022 0.200 0.841 -0.048 0.039 Pan -0.002 0.017 0.140 0.889 -0.036 0.031 Rain 0.001 0.016 0.052 0.959 -0.030 0.032 EVI2 0.001 0.018 0.078 0.938 -0.033 0.036     56  Table A.6 Buffalo core, wet season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer 0.123 0.074 1.664 0.096 -0.022 0.267 Road buffer -0.219 0.075 2.915 0.004 -0.366 -0.072 Day -0.129 0.037 3.505 0.000 -0.202 -0.057 Angle -0.082 0.043 1.892 0.058 -0.167 0.003 Pan -0.102 0.045 2.269 0.023 -0.189 -0.014 EVI -0.229 0.038 6.038 < 0.001 -0.303 -0.154 EVI2 0.148 0.042 3.554 < 0.001 0.066 0.230 Lag speed 0.114 0.039 2.938 0.003 0.038 0.190 Fence buffer -0.025 0.057 0.431 0.666 -0.137 0.087 Settlement -0.021 0.041 0.524 0.600 -0.102 0.059 Rain 0.009 0.025 0.378 0.705 -0.039 0.057 In omuramba 0.012 0.037 0.333 0.739 -0.060 0.085  Table A.7 Eastern floodplain, dry season, turn angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Day 0.017 0.035 0.505 0.614 -0.050 0.085 River buffer 0.016 0.034 0.480 0.631 -0.050 0.083 Fire -0.033 0.069 0.482 0.630 -0.169 0.102 Lag angle 0.015 0.032 0.453 0.650 -0.048 0.078 Speed -0.011 0.029 0.372 0.710 -0.068 0.046 EVI2 0.010 0.037 0.271 0.787 -0.062 0.082 EVI 0.007 0.026 0.256 0.798 -0.043 0.057 Rain 0.000 0.034 0.011 0.991 -0.067 0.067        57  Table A.8 Eastern floodplain, dry season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer -0.114 0.061 1.884 0.060 -0.233 0.005 Day -0.260 0.050 5.220 0.000 -0.357 -0.162 Fire 0.274 0.117 2.335 0.020 0.044 0.504 EVI 0.319 0.050 6.385 <2e-16 0.221 0.417 EVI2 -0.391 0.071 5.544 <2e-16 -0.529 -0.253 Rain 0.037 0.066 0.564 0.573 -0.092 0.167 Angle -0.010 0.031 0.313 0.755 -0.070 0.051 Lag Speed 0.004 0.028 0.155 0.877 -0.050 0.059  Table A.9 Eastern floodplain, wet season, turn angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer 0.050 0.045 1.111 0.267 -0.038 0.137 Day -0.233 0.035 6.701 <0.001 -0.301 -0.165 Lag angle 0.027 0.035 0.760 0.447 -0.043 0.097 Speed -0.053 0.043 1.247 0.213 -0.136 0.030 EVI 0.013 0.028 0.449 0.654 -0.043 0.068 EVI2 0.008 0.024 0.349 0.727 -0.039 0.056 Rain 0.006 0.021 0.268 0.789 -0.035 0.047  Table A.10 Eastern floodplain, wet season, speed model Variable Estimate Std. Error z-value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer -0.126 0.040 3.122 0.002 -0.205 -0.047 Day -0.199 0.035 5.654 <0.001 -0.269 -0.130 Angle -0.052 0.042 1.235 0.217 -0.136 0.031 EVI 0.246 0.037 6.643 <0.001 0.174 0.319 EVI2 -0.217 0.036 5.943 <0.001 -0.288 -0.145 Rain -0.008 0.023 0.357 0.721 -0.054 0.037 Lag speed -0.003 0.020 0.173 0.863 -0.042 0.035  58  Table A.11 Horseshoe, dry season, turn angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Settlement -0.047 0.061 0.769 0.442 -0.166 0.073 Lag angle 0.049 0.053 0.920 0.358 -0.056 0.154 Speed -0.042 0.051 0.821 0.412 -0.141 0.058 EVI 0.054 0.063 0.845 0.398 -0.071 0.178 EVI2 -0.019 0.053 0.355 0.723 -0.124 0.086 Rain -0.016 0.050 0.328 0.743 -0.115 0.082 Day -0.012 0.032 0.373 0.709 -0.075 0.051 In omuramba -0.002 0.043 0.057 0.955 -0.088 0.083 River buffer 0.004 0.036 0.098 0.922 -0.068 0.075  Table A.12 Horseshoe, dry season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Day 0.100 0.066 1.517 0.129 -0.029 0.229 Rain 0.247 0.097 2.542 0.011 0.057 0.438 Lag speed 0.224 0.053 4.237 0.000 0.120 0.327 Angle -0.036 0.052 0.691 0.490 -0.137 0.066 Settlement -0.034 0.061 0.565 0.572 -0.153 0.085 EVI -0.024 0.055 0.436 0.663 -0.131 0.084 In omuramba 0.005 0.048 0.111 0.912 -0.089 0.100 River buffer 0.009 0.044 0.201 0.840 -0.077 0.095 EVI2 0.006 0.047 0.137 0.891 -0.086 0.099         59  Table A.13 Horseshoe, wet season, turn angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Speed -0.082 0.051 1.613 0.107 -0.181 0.018 EVI -0.030 0.043 0.691 0.489 -0.114 0.055 EVI2 0.018 0.034 0.520 0.603 -0.049 0.085 In omuramba -0.014 0.033 0.433 0.665 -0.078 0.050 Rain -0.006 0.024 0.249 0.803 -0.054 0.041 Day -0.005 0.023 0.208 0.835 -0.049 0.040 Settlement -0.010 0.028 0.347 0.729 -0.066 0.046 Fence buffer 0.007 0.037 0.198 0.843 -0.065 0.079 Lag angle 0.001 0.021 0.055 0.956 -0.040 0.042  Table A.14 Horseshoe, wet season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Day -0.047 0.052 0.903 0.367 -0.150 0.055 Angle -0.098 0.055 1.782 0.075 -0.206 0.010 Settlement 0.124 0.057 2.177 0.030 0.012 0.236 EVI -0.347 0.050 6.984 <0.001 -0.445 -0.250 Lag speed 0.567 0.047 12.135 <0.001 0.475 0.658 Fence buffer -0.035 0.063 0.546 0.585 -0.158 0.089 In omuramba 0.016 0.037 0.425 0.671 -0.056 0.087 Rain 0.013 0.033 0.388 0.698 -0.052 0.078 EVI2 0.005 0.026 0.178 0.858 -0.047 0.056         60  Table A.15 Mahango, dry season, turn angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Road buffer -0.177 0.085 2.084 0.037 -0.343 -0.010 Day -0.065 0.073 0.888 0.375 -0.207 0.078 In omuramba 0.174 0.163 1.068 0.285 -0.145 0.493 EVI2 -0.042 0.059 0.719 0.472 -0.157 0.073 Lag angle 0.033 0.054 0.604 0.546 -0.073 0.138 Rain -0.026 0.055 0.469 0.639 -0.133 0.081 River buffer -0.018 0.053 0.343 0.732 -0.122 0.085 Speed -0.008 0.035 0.217 0.828 -0.075 0.060 EVI 0.008 0.039 0.213 0.831 -0.067 0.084  Table A.16 Mahango, dry season, speed model Variable Estimate Std. Error value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Road buffer 0.339 0.115 2.955 0.003 0.114 0.564 Day -0.381 0.099 3.837 0.000 -0.576 -0.186 In omuramba 0.142 0.204 0.694 0.488 -0.258 0.542 Rain 0.096 0.118 0.816 0.414 -0.135 0.327 EVI2 0.236 0.103 2.287 0.022 0.034 0.438 Lag speed -0.152 0.116 1.311 0.190 -0.379 0.075 EVI 0.020 0.065 0.304 0.761 -0.107 0.147 Angle -0.011 0.052 0.216 0.829 -0.113 0.091 River buffer -0.003 0.062 0.050 0.960 -0.125 0.119         61  Table A.17 Mahango, wet season, turn angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) In omuramba 0.226 0.164 1.378 0.168 -0.095 0.546 Lag angle 0.255 0.072 3.543 <0.001 0.114 0.397 Speed -0.119 0.089 1.338 0.181 -0.292 0.055 Fence buffer 0.060 0.084 0.707 0.479 -0.106 0.225 River buffer 0.054 0.114 0.475 0.635 -0.170 0.279 EVI 0.022 0.054 0.414 0.679 -0.084 0.128 Road buffer -0.015 0.047 0.317 0.751 -0.106 0.077 EVI2 0.004 0.026 0.168 0.867 -0.047 0.055 Day 0.011 0.044 0.245 0.806 -0.076 0.098 Rain -0.003 0.037 0.070 0.944 -0.075 0.070  Table A.18 Mahango, wet season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Day -0.485 0.078 6.193 <0.001 -0.638 -0.331 In omuramba 0.211 0.173 1.216 0.224 -0.129 0.551 Angle -0.156 0.092 1.692 0.091 -0.336 0.025 Lag speed -0.577 0.076 7.605 <0.001 -0.725 -0.428 Rain 0.041 0.068 0.608 0.543 -0.092 0.175 Road buffer -0.027 0.058 0.454 0.650 -0.141 0.088 Fence buffer 0.027 0.064 0.419 0.675 -0.099 0.152 River buffer 0.025 0.093 0.266 0.790 -0.158 0.208 EVI 0.008 0.045 0.185 0.854 -0.080 0.097 EVI2 -0.006 0.028 0.217 0.828 -0.061 0.048        62  Table A.19 Nkasa Rupara, dry season, turn angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Rain 0.056 0.053 1.069 0.285 -0.047 0.159 Lag angle 0.139 0.028 4.954 <0.001 0.084 0.193 Day -0.017 0.026 0.631 0.528 -0.069 0.035 Settlement -0.015 0.025 0.599 0.549 -0.065 0.035 EVI 0.012 0.023 0.500 0.617 -0.034 0.057 EVI2 -0.009 0.025 0.374 0.708 -0.058 0.040 Speed -0.002 0.015 0.104 0.917 -0.031 0.028 River buffer 0.002 0.017 0.121 0.903 -0.031 0.036  Table A.20 Nkasa Rupara, dry season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer -0.048 0.037 1.300 0.194 -0.121 0.025 Day 0.270 0.027 10.120 <0.001 0.218 0.322 Rain 0.077 0.051 1.514 0.130 -0.023 0.176 Settlement 0.206 0.027 7.703 <0.001 0.154 0.258 Lag speed 0.052 0.033 1.581 0.114 -0.012 0.116 EVI2 0.018 0.031 0.600 0.549 -0.042 0.079 EVI -0.002 0.015 0.145 0.885 -0.031 0.027 Angle 0.000 0.014 0.000 1.000 -0.027 0.027  Table A.21 Nkasa Rupara, wet season, turn angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer -0.030 0.037 0.808 0.419 -0.101 0.042 Day -0.036 0.034 1.063 0.288 -0.103 0.031 Settlement -0.063 0.035 1.787 0.074 -0.132 0.006 EVI 0.032 0.036 0.885 0.376 -0.039 0.104 EVI2 -0.129 0.034 3.784 < 0.001 -0.196 -0.062 Lag angle 0.114 0.028 4.021 < 0.001 0.058 0.169 Speed -0.030 0.033 0.921 0.357 -0.094 0.034 Rain 0.003 0.016 0.200 0.842 -0.029 0.036 63   Table A.22 Nkasa Rupara, wet season, speed model Variable Estimate Std. Error z-value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer 0.027 0.032 0.838 0.402 -0.036 0.091 Day -0.055 0.030 1.828 0.067 -0.115 0.004 Rain 0.120 0.025 4.712 <0.001 0.070 0.169 Angle -0.026 0.028 0.922 0.356 -0.081 0.029 Settlement 0.212 0.025 8.463 <0.001 0.163 0.261 EVI -0.100 0.030 3.374 0.001 -0.158 -0.042 EVI2 0.104 0.030 3.509 <0.001 0.046 0.162 Lag speed -0.053 0.031 1.751 0.080 -0.113 0.006  Table A.23 Mudumu, dry season, turn angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Road buffer -0.159 0.043 3.718 <0.001 -0.243 -0.075 Day 0.058 0.037 1.554 0.120 -0.015 0.130 Pan 0.089 0.037 2.414 0.016 0.017 0.162 Lag angle 0.098 0.031 3.222 0.001 0.039 0.158 Speed -0.096 0.031 3.088 0.002 -0.157 -0.035 River buffer 0.028 0.041 0.699 0.484 -0.051 0.108 Settlement 0.026 0.038 0.681 0.496 -0.048 0.100 EVI 0.020 0.034 0.582 0.561 -0.046 0.086 EVI2 -0.017 0.034 0.503 0.615 -0.085 0.050 Rain 0.003 0.028 0.118 0.906 -0.052 0.059        64  Table A.24 Mudumu, dry season, speed model Variable Estimate Std. Error value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer -0.284 0.046 6.222 <0.001 -0.374 -0.195 Day -0.104 0.036 2.913 0.004 -0.174 -0.034 Rain 0.255 0.060 4.271 <0.001 0.138 0.372 Angle -0.108 0.034 3.150 0.002 -0.175 -0.041 Settlement 0.261 0.039 6.677 <0.001 0.184 0.337 EVI -0.139 0.040 3.435 0.001 -0.218 -0.060 Lag speed -0.245 0.033 7.417 <0.001 -0.309 -0.180 Road buffer 0.011 0.031 0.374 0.708 -0.049 0.072 EVI2 0.009 0.030 0.311 0.756 -0.049 0.068 Pan -0.003 0.020 0.159 0.873 -0.042 0.035  Table A.25 Mudumu, wet season, turn angle model Variables Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Rain 0.067 0.036 1.876 0.061 -0.003 0.137 Speed -0.133 0.029 4.552 <0.001 -0.191 -0.076 River buffer -0.029 0.042 0.682 0.496 -0.111 0.054 Pan -0.017 0.027 0.637 0.524 -0.071 0.036 Settlement -0.006 0.021 0.274 0.784 -0.047 0.035 Road buffer -0.003 0.032 0.098 0.922 -0.066 0.060 EVI2 -0.001 0.018 0.070 0.944 -0.036 0.034 EVI 0.001 0.016 0.059 0.953 -0.030 0.032 Lag angle -0.001 0.015 0.074 0.941 -0.031 0.029 Day -0.001 0.015 0.059 0.953 -0.031 0.029        65  Table A.26 Mudumu, wet season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Road buffer 0.024 0.033 0.725 0.468 -0.040 0.088 Day -0.127 0.023 5.584 <0.001 -0.172 -0.083 Rain 0.086 0.023 3.720 <0.001 0.041 0.132 Angle -0.081 0.023 3.524 <0.001 -0.126 -0.036 EVI -0.330 0.024 13.872 <0.001 -0.377 -0.283 EVI2 0.102 0.025 4.149 <0.001 0.054 0.150 Lag speed 0.288 0.023 12.517 <0.001 0.243 0.333 In omuramba 0.019 0.027 0.733 0.463 -0.033 0.072 Fence butter -0.020 0.043 0.465 0.642 -0.104 0.064 Settlement 0.001 0.012 0.101 0.919 -0.023 0.025  Table A.27 Susuwe, dry season, turn angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer -0.083 0.042 1.976 0.048 -0.166 -0.001 Road buffer 0.100 0.079 1.271 0.204 -0.054 0.255 Rain 0.050 0.048 1.043 0.297 -0.044 0.145 Pan 0.110 0.035 3.181 0.001 0.042 0.177 EVI -0.096 0.034 2.856 0.004 -0.162 -0.030 Speed -0.128 0.029 4.457 <0.001 -0.184 -0.071 Day -0.006 0.018 0.312 0.755 -0.042 0.030 In omuramba 0.012 0.033 0.371 0.710 -0.052 0.076 EVI2 -0.007 0.025 0.265 0.791 -0.055 0.042 Lag angle 0.002 0.015 0.108 0.914 -0.028 0.031 Settlement -0.005 0.027 0.187 0.852 -0.059 0.049       66  Table A.28 Susuwe, dry season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer -0.170 0.041 4.137 <0.001 -0.250 -0.089 Day -0.262 0.032 8.303 <0.001 -0.324 -0.200 Angle -0.143 0.031 4.605 <0.001 -0.205 -0.082 Pan -0.089 0.043 2.089 0.037 -0.173 -0.006 EVI -0.046 0.042 1.090 0.276 -0.130 0.037 EVI2 0.127 0.051 2.510 0.012 0.028 0.227 Lag speed -0.114 0.032 3.572 <0.001 -0.176 -0.051 Road buffer 0.029 0.057 0.515 0.607 -0.082 0.140 Rain 0.008 0.027 0.295 0.768 -0.046 0.062 Settlement -0.014 0.035 0.403 0.687 -0.083 0.055 In omuramba 0.003 0.028 0.115 0.909 -0.052 0.059  Table A.29 Susuwe, wet season, turn angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Day 0.022 0.027 0.802 0.423 -0.032 0.076 Speed -0.083 0.027 3.040 0.002 -0.136 -0.029 River buffer 0.031 0.051 0.598 0.550 -0.070 0.131 Rain 0.010 0.020 0.486 0.627 -0.030 0.050 Pan 0.009 0.020 0.427 0.669 -0.031 0.049 Road buffer 0.007 0.022 0.302 0.762 -0.036 0.049 EVI2 0.003 0.017 0.172 0.863 -0.031 0.037 Lag angle 0.002 0.014 0.166 0.868 -0.025 0.030 Settlement -0.007 0.019 0.359 0.720 -0.045 0.031 EVI 0.000 0.014 0.025 0.980 -0.028 0.028 In omuramba 0.002 0.017 0.134 0.893 -0.030 0.035      67  Table A.30 Susuwe, wet season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Road buffer 0.020 0.032 0.628 0.530 -0.043 0.084 Day -0.171 0.026 6.602 <0.001 -0.221 -0.120 Rain 0.148 0.026 5.668 <0.001 0.097 0.199 Angle -0.080 0.027 2.955 0.003 -0.134 -0.027 Settlement -0.101 0.030 3.325 0.001 -0.161 -0.041 EVI -0.304 0.028 11.030 <0.001 -0.358 -0.250 EVI2 0.162 0.032 5.079 <0.001 0.099 0.224 Lag speed 0.159 0.026 6.073 <0.001 0.107 0.210 Pan -0.012 0.024 0.491 0.623 -0.059 0.035 In omuramba 0.011 0.024 0.467 0.641 -0.036 0.059 River buffer -0.020 0.045 0.440 0.660 -0.108 0.068  A.2 Additional results from movement model, split by collar Table A.31 77264, dry season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Settlement -0.062 0.088 0.705 0.481 -0.236 0.111 Speed 0.094 0.099 0.951 0.342 -0.100 0.287 EVI2 -0.028 0.076 0.361 0.718 -0.177 0.122 Day 0.022 0.060 0.374 0.708 -0.095 0.140 Rain 0.030 0.081 0.372 0.710 -0.129 0.189 River buffer 0.006 0.050 0.122 0.903 -0.092 0.104 Lag angle -0.012 0.049 0.240 0.810 -0.108 0.085 EVI 0.006 0.046 0.138 0.890 -0.083 0.096      68  Table A.32 77264, dry season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Day 0.255 0.100 2.542 0.011 0.058 0.451 Rain 0.121 0.137 0.884 0.377 -0.148 0.390 Angle 0.090 0.098 0.920 0.357 -0.102 0.282 Settlement 0.174 0.113 1.541 0.123 -0.047 0.396 Lag speed 0.048 0.079 0.604 0.546 -0.107 0.203 River buffer -0.024 0.065 0.360 0.719 -0.151 0.104 EVI 0.016 0.053 0.293 0.769 -0.089 0.119 EVI2 -0.01 0.062 0.104 0.917 -0.128 0.115  Table A.33 77264, wet season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer 0.139 0.160 0.867 0.386 -0.175 0.453 Settlement -0.427 0.159 2.675 0.007 -0.739 -0.114 EVI2 -0.422 0.150 2.816 0.005 -0.716 -0.128 Speed 0.035 0.083 0.423 0.672 -0.128 0.198 Lag angle -0.006 0.060 0.093 0.926 -0.123 0.112 EVI 0.001 0.070 0.018 0.986 -0.137 0.140 Rain 0.000 0.058 0.008 0.994 -0.114 0.115 Day 0.000 0.058 0.002 0.998 -0.115 0.115          69  Table A.34 77264, wet season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Settlement 0.323 0.106 3.028 0.002 0.114 0.532 EVI -0.095 0.105 0.899 0.369 -0.301 0.112 Lag speed 0.078 0.094 0.831 0.406 -0.106 0.262 Day 0.038 0.072 0.532 0.595 -0.102 0.179 Rain 0.030 0.067 0.454 0.650 -0.101 0.161 Angle 0.032 0.067 0.472 0.637 -0.100 0.163 River buffer 0.020 0.068 0.285 0.776 -0.115 0.154 EVI2 -0.005 0.061 0.080 0.936 -0.126 0.116  Table A.35 77264new, dry season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Lag angle 0.054 0.066 0.816 0.415 -0.075 0.183 Rain 0.053 0.094 0.563 0.573 -0.131 0.236 River buffer -0.027 0.057 0.479 0.632 -0.139 0.084 Day -0.007 0.034 0.195 0.846 -0.073 0.060 Settlement -0.002 0.032 0.072 0.943 -0.065 0.060 EVI -0.002 0.033 0.075 0.940 -0.067 0.062 Speed -0.001 0.032 0.040 0.968 -0.063 0.061 EVI2 -0.001 0.044 0.026 0.979 -0.087 0.085          70  Table A.36 77264new, dry season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer -0.067 0.070 0.951 0.342 -0.204 0.071 Day 0.195 0.051 3.814 <0.001 0.095 0.296 Settlement 0.159 0.055 2.898 0.004 0.051 0.266 Lag speed 0.045 0.055 0.817 0.414 -0.063 0.152 EVI2 0.040 0.065 0.622 0.534 -0.086 0.167 Rain 0.006 0.048 0.131 0.896 -0.088 0.101 EVI -0.004 0.028 0.152 0.879 -0.060 0.051 Angle 0.000 0.026 0.001 1.000 -0.051 0.051   Table A.37 77264new, wet season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Rain -0.089 0.072 1.222 0.222 -0.231 0.054 EVI 0.115 0.073 1.576 0.115 -0.028 0.258 Speed -0.078 0.070 1.126 0.260 -0.215 0.058 Settlement -0.045 0.060 0.754 0.451 -0.163 0.072 Lag angle 0.021 0.044 0.481 0.630 -0.065 0.108 River buffer 0.018 0.047 0.381 0.703 -0.075 0.111 EVI2 -0.015 0.051 0.288 0.773 -0.114 0.085 Day -0.006 0.032 0.182 0.856 -0.068 0.056  Table A.38 77264new, wet season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Day -0.039 0.053 0.737 0.461 -0.143 0.065 Rain 0.079 0.066 1.193 0.233 -0.051 0.208 Angle -0.062 0.061 1.011 0.312 -0.181 0.058 Settlement 0.220 0.052 4.179 <0.001 0.117 0.322 EVI2 0.173 0.082 2.114 0.035 0.013 0.334 Lag speed 0.014 0.036 0.382 0.703 -0.056 0.084 River buffer 0.006 0.035 0.178 0.859 -0.062 0.074 EVI 0.006 0.032 0.198 0.843 -0.057 0.069 71  Table A.39 77266, dry season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Rain 0.066 0.083 0.789 0.430 -0.097 0.229 Settlement 0.158 0.060 2.653 0.008 -0.275 -0.041 Lag angle 0.167 0.057 2.948 0.003 0.056 0.278 River buffer 0.034 0.054 0.631 0.528 -0.072 0.141 Day -0.022 0.044 0.504 0.614 -0.108 0.064 EVI2 -0.022 0.052 0.424 0.672 -0.125 0.080 Speed -0.009 0.032 0.271 0.787 -0.072 0.054 EVI 0.005 0.030 0.177 0.859 -0.053 0.064  Table A.40 77266, dry season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Day 0.284 0.048 5.882 <0.001 0.189 0.378 Settlement 0.115 0.058 1.971 0.049 0.001 0.229 EVI -0.054 0.055 0.971 0.332 -0.163 0.055 EVI2 0.073 0.075 0.972 0.331 -0.074 0.219 Rain 0.023 0.052 0.438 0.661 -0.080 0.125 Angle -0.009 0.029 0.309 0.757 -0.067 0.049 River buffer -0.004 0.028 0.159 0.874 -0.060 0.051 Lag speed 0.003 0.025 0.126 0.899 -0.046 0.053  Table A.41 77266, wet season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer -0.053 0.070 0.755 0.450 -0.191 0.085 EVI2 -0.199 0.064 3.109 0.002 -0.325 -0.074 Lag angle 0.261 0.057 4.582 <0.001 0.149 0.372 EVI 0.013 0.043 0.311 0.755 -0.070 0.097 Speed -0.011 0.035 0.302 0.763 -0.080 0.059 Settlement -0.013 0.038 0.338 0.735 -0.087 0.061 Day -0.011 0.036 0.301 0.763 -0.081 0.059 Rain 0.009 0.035 0.255 0.799 -0.059 0.077 72   Table A.42 77266, wet season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Rain 0.163 0.048 3.407 0.001 0.069 0.256 EVI -0.046 0.058 0.785 0.433 -0.161 0.069 EVI2 0.045 0.056 0.795 0.426 -0.066 0.155 Lag speed -0.074 0.055 1.345 0.179 -0.182 0.034 Settlement -0.034 0.046 0.741 0.459 -0.125 0.056 Angle -0.011 0.029 0.362 0.717 -0.068 0.047 Day 0.001 0.024 0.022 0.982 -0.046 0.047 River buffer -0.001 0.027 0.048 0.962 -0.055 0.052  Table A.43 77266new, dry season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Fire 0.312 0.207 1.503 0.133 -0.095 0.719 Lag angle 0.195 0.103 1.898 0.0577 -0.006 0.397 EVI -0.042 0.075 0.556 0.578 -0.189 0.106 Rain 0.046 0.118 0.387 0.699 -0.185 0.276 Speed -0.014 0.052 0.273 0.785 -0.115 0.087 Day -0.014 0.051 0.269 0.788 -0.114 0.087 Settlement -0.007 0.046 0.155 0.877 -0.098 0.084 EVI2 -0.007 0.061 0.110 0.912 -0.127 0.113 River buffer 0.000 0.054 0.003 0.997 -0.106 0.107         73  Table A.44 77266new, dry season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer -0.120 0.110 1.088 0.276 -0.337 0.096 Day 0.317 0.074 4.301 <0.001 0.173 0.462 Settlement 0.253 0.078 3.253 0.001 0.101 0.406 EVI 0.073 0.085 0.855 0.393 -0.094 0.240 Lag speed 0.196 0.085 2.302 0.021 0.029 0.363 EVI2 -0.044 0.084 0.529 0.597 -0.208 0.120 Rain 0.060 0.119 0.507 0.612 -0.173 0.294 Angle -0.004 0.038 0.100 0.920 -0.079 0.071 Fire 0.002 0.074 0.030 0.976 -0.142 0.147  Table A.45 77266new, wet season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Lag angle 0.035 0.059 0.588 0.557 -0.081 0.151 Speed -0.034 0.059 0.577 0.564 -0.150 0.082 Day -0.030 0.056 0.532 0.595 -0.139 0.080 Road buffer 0.024 0.082 0.293 0.770 -0.138 0.186 EVI2 0.008 0.042 0.195 0.845 -0.074 0.090 River buffer -0.009 0.048 0.185 0.853 -0.104 0.086 Rain 0.007 0.038 0.187 0.851 -0.067 0.081 Settlement 0.001 0.039 0.028 0.978 -0.075 0.077 EVI -0.005 0.037 0.131 0.895 -0.078 0.068         74  Table A.46 77266new, wet season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Road buffer 0.426 0.158 2.688 0.007 0.115 0.737 Day -0.097 0.077 1.258 0.209 -0.248 0.054 Settlement 0.375 0.079 4.743 <0.001 0.220 0.529 EVI -0.216 0.074 2.894 0.004 -0.362 -0.070 EVI2 0.125 0.092 1.358 0.175 -0.056 0.306 Angle -0.035 0.057 0.621 0.535 -0.146 0.076 Rain 0.039 0.060 0.657 0.511 -0.078 0.157 Lag speed -0.012 0.040 0.310 0.757 -0.090 0.065 River buffer 0.013 0.048 0.263 0.793 -0.081 0.106  Table A.47 AM290, dry season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Settlement -0.124 0.061 2.030 0.042 -0.243 -0.004 EVI 0.113 0.063 1.776 0.076 -0.012 0.237 Lag angle 0.100 0.063 1.581 0.114 -0.024 0.224 River buffer -0.025 0.049 0.516 0.606 -0.121 0.070 EVI2 0.012 0.040 0.296 0.767 -0.066 0.090 Speed -0.008 0.031 0.271 0.786 -0.068 0.052 Rain -0.009 0.042 0.207 0.836 -0.091 0.073 Day 0.002 0.029 0.076 0.940 -0.054 0.058          75  Table A.48 AM290, dry season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer -0.077 0.076 1.004 0.315 -0.226 0.073 Day 0.362 0.059 6.138 <0.001 0.2462 0.477 Settlement 0.084 0.068 1.229 0.219 -0.050 0.218 Rain 0.044 0.072 0.617 0.538 -0.096 0.185 EVI2 0.024 0.051 0.476 0.634 -0.076 0.125 Angle -0.010 0.034 0.299 0.765 -0.076 0.056 EVI -0.052 0.029 0.002 0.999 -0.058 0.057 Lag speed -0.019 0.027 0.068 0.946 -0.058 0.054  Table A.49 AM290, wet season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer -0.115 0.082 1.401 0.161 -0.276 0.046 Rain 0.056 0.063 0.896 0.370 -0.067 0.179 Settlement -0.152 0.062 2.448 0.014 -0.274 -0.030 EVI2 -0.034 0.057 0.593 0.553 -0.146 0.078 Day -0.041 0.057 0.723 0.470 -0.154 0.071 Lag angle 0.028 0.048 0.581 0.561 -0.066 0.122 EVI -0.005 0.035 0.144 0.885 -0.074 0.064 Speed -0.003 0.029 0.105 0.917 -0.059 0.053  Table A.50 AM290, wet season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Day -0.043 0.053 0.803 0.422 -0.147 0.062 Rain 0.045 0.055 0.809 0.419 -0.064 0.153 Settlement 0.152 0.052 2.905 0.004 0.050 0.255 EVI -0.085 0.074 1.142 0.254 -0.231 0.061 EVI2 0.098 0.080 1.220 0.222 -0.059 0.255 Lag speed -0.243 0.048 5.061 <0.001 -0.338 -0.149 River buffer 0.006 0.032 0.180 0.857 -0.058 0.069 Angle -0.004 0.026 0.153 0.879 -0.055 0.047 76    Table A.51 94043, dry season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Omuramba -0.208 0.211 0.986 0.324 -0.621 0.205 Settlement -0.171 0.129 1.322 0.186 -0.425 0.083 Lag angle 0.076 0.086 0.885 0.376 -0.093 0.246 Rain -0.079 0.123 0.640 0.522 -0.320 0.163 EVI 0.092 0.110 0.831 0.406 -0.125 0.308 Speed -0.016 0.049 0.316 0.752 -0.112 0.081 EVI2 -0.037 0.097 0.379 0.705 -0.227 0.154 River buffer -0.020 0.070 0.286 0.775 -0.157 0.117 Day -0.005 0.043 0.120 0.904 -0.090 0.080  Table A.52 94043, dry season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Variable Rain 0.584 0.155 0.155 3.757 <0.001 0.279 0.889 Settlement -0.313 0.129 0.129 2.419 0.016 -0.567 -0.059 EVI -0.272 0.136 0.136 1.992 0.046 -0.539 -0.004 Lag speed 0.271 0.097 0.097 2.798 0.005 0.081 0.461 River buffer -0.048 0.102 0.102 0.468 0.640 -0.247 0.152 EVI2 -0.024 0.096 0.096 0.247 0.805 -0.213 0.165 Angle -0.009 0.049 0.049 0.175 0.861 -0.104 0.087 Omuramba -0.017 0.114 0.114 0.149 0.882 -0.240 0.206 Day 0.000 0.048 0.048 0.003 0.998 -0.095 0.095      77  Table A.53 94043, wet season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Day -0.060 0.068 0.884 0.377 -0.194 0.073 Settlement -0.056 0.072 0.784 0.433 -0.197 0.084 EVI -0.060 0.073 0.817 0.414 -0.203 0.084 EVI2 0.053 0.074 0.713 0.476 -0.092 0.197 Speed -0.073 0.071 1.034 0.301 -0.212 0.066 Lag angle 0.024 0.047 0.512 0.609 -0.068 0.116 Omuramba -0.022 0.050 0.437 0.662 -0.121 0.077 Rain -0.001 0.033 0.036 0.971 -0.067 0.064 Road buffer 0.007 0.064 0.106 0.916 -0.118 0.131  Table A.54 94043, wet season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Day -0.261 0.065 4.019 <0.001 -0.388 -0.134 Omuramba 0.114 0.089 1.278 0.201 -0.061 0.289 Rain -0.060 0.073 0.825 0.410 -0.203 0.083 Angle -0.096 0.075 1.273 0.203 -0.244 0.052 Settlement 0.225 0.077 2.922 0.003 0.074 0.376 EVI -0.208 0.075 2.794 0.005 -0.355 -0.062 Lag speed 0.435 0.063 6.880 <0.001 0.311 0.559 Road buffer 0.020 0.075 0.264 0.791 -0.127 0.166 EVI2 -0.007 0.042 0.156 0.876 -0.088 0.075         78  Table A.55 AG275, dry season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Road buffer -0.058 0.140 0.412 0.680 -0.333 0.217 Settlement 0.037 0.075 0.489 0.625 -0.110 0.184 Omuramba 0.035 0.078 0.447 0.655 -0.117 0.187 Day -0.026 0.061 0.421 0.674 -0.146 0.094 Speed -0.015 0.052 0.290 0.772 -0.117 0.087 EVI2 -0.012 0.064 0.187 0.852 -0.138 0.114 Rain 0.008 0.069 0.116 0.907 -0.128 0.144 EVI -0.002 0.048 0.045 0.964 -0.097 0.093 River buffer -0.015 0.093 0.159 0.873 -0.198 0.168 Lag angle 0.003 0.044 0.079 0.937 -0.083 0.090  Table A.56 AG275, dry season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Settlement 0.113 0.124 0.911 0.362 -0.130 0.355 EVI2 0.167 0.154 1.087 0.277 -0.134 0.469 River buffer -0.094 0.169 0.554 0.579 -0.427 0.238 Lag speed 0.044 0.079 0.549 0.583 -0.112 0.199 Day 0.032 0.070 0.461 0.645 -0.106 0.171 Omuramba 0.022 0.071 0.304 0.761 -0.118 0.162 Rain 0.039 0.100 0.388 0.698 -0.157 0.235 Angle -0.016 0.056 0.282 0.778 -0.125 0.094 EVI -0.002 0.058 0.042 0.966 -0.117 0.112 Road buffer 0.000 0.118 0.002 0.999 -0.232 0.231        79  Table A.57 AG275, wet season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Speed -0.068 0.080 0.849 0.396 -0.224 0.089 EVI2 0.023 0.043 0.529 0.597 -0.061 0.106 EVI 0.031 0.060 0.516 0.606 -0.087 0.149 Lag angle -0.027 0.056 0.480 0.631 -0.137 0.083 Rain -0.026 0.057 0.462 0.644 -0.138 0.085 Day 0.025 0.055 0.458 0.647 -0.083 0.133 Settlement -0.006 0.040 0.156 0.876 -0.084 0.072 Omuramba 0.005 0.041 0.126 0.900 -0.075 0.085 Fence buffer 0.001 0.058 0.014 0.989 -0.113 0.114  Table A.58 AG275, wet season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Omuramba 0.110 0.095 1.159 0.246 -0.076 0.296 Angle -0.063 0.078 0.808 0.419 -0.217 0.090 EVI -0.207 0.086 2.416 0.016 -0.375 -0.039 EVI2 0.051 0.059 0.865 0.387 -0.064 0.166 Day -0.026 0.057 0.465 0.642 -0.137 0.085 Rain 0.023 0.055 0.424 0.672 -0.084 0.131 Fence buffer -0.035 0.082 0.420 0.674 -0.195 0.126 Lag speed -0.011 0.044 0.241 0.810 -0.096 0.075 Settlement -0.006 0.041 0.154 0.877 -0.087 0.074        80  Table A.59 AG273, dry season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Rain 0.193 0.209 0.916 0.359 -0.219 0.605 Day 0.108 0.162 0.666 0.505 -0.210 0.427 River buffer 0.142 0.211 0.670 0.503 -0.273 0.557 Speed 0.073 0.137 0.534 0.593 -0.196 0.343 Lag angle 0.025 0.098 0.249 0.803 -0.168 0.217 EVI2 -0.007 0.059 0.125 0.901 -0.125 0.110 EVI -0.013 0.092 0.141 0.888 -0.195 0.169  Table A.60 AG273, dry season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Lag speed 0.657 0.222 2.926 0.003 0.2167 1.097 Angle 0.087 0.167 0.516 0.606 -0.243 0.417 EVI2 0.055 0.113 0.485 0.627 -0.167 0.277 River buffer -0.038 0.155 0.244 0.807 -0.345 0.269 EVI -0.026 0.121 0.214 0.830 -0.266 0.213 Rain -0.004 0.115 0.035 0.972 -0.231 0.223 Day -0.002 0.108 0.023 0.982 -0.215 0.211  Table A.61 AG273, wet season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Day -0.293 0.078 3.744 <0.001 -0.447 -0.140 Lag angle -0.073 0.085 0.869 0.385 -0.240 0.093 Speed -0.029 0.060 0.492 0.623 -0.147 0.088 EVI2 0.015 0.034 0.461 0.645 -0.051 0.082 EVI 0.013 0.048 0.278 0.781 -0.081 0.107 Rain 0.011 0.045 0.252 0.801 -0.076 0.098 River buffer 0.001 0.048 0.024 0.981 -0.094 0.096   81  Table A.62 AG273, wet season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer -0.219 0.123 1.782 0.075 -0.460 0.022 Rain -0.050 0.078 0.647 0.517 -0.202 0.102 Day -0.047 0.076 0.614 0.539 -0.195 0.102 Angle -0.025 0.058 0.426 0.670 -0.140 0.090 EVI2 -0.005 0.026 0.195 0.845 -0.057 0.046 EVI -0.005 0.044 0.108 0.914 -0.091 0.081 Lag speed -0.001 0.043 0.02 0.984 -0.086 0.084  Table A.63 AG276, dry season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Speed -0.265 0.151 1.753 0.080 -0.560 0.031 River buffer 0.074 0.115 0.638 0.523 -0.153 0.300 Day 0.080 0.119 0.671 0.502 -0.153 0.313 Rain -0.075 0.131 0.572 0.568 -0.332 0.182 EVI 0.002 0.071 0.024 0.981 -0.138 0.142 EVI2 -0.035 0.201 0.175 0.861 -0.430 0.360 Lag angle -0.005 0.063 0.072 0.943 -0.128 0.118  Table A.64 AG276, dry season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Day -0.167 0.174 0.962 0.336 -0.509 0.174 Angle -0.251 0.180 1.391 0.164 -0.604 0.103 EVI 0.683 0.157 4.319 <0.001 0.373 0.993 Lag speed 0.298 0.189 1.571 0.116 -0.074 0.669 River buffer -0.035 0.098 0.358 0.720 -0.228 0.158 EVI2 -0.056 0.254 0.219 0.827 -0.556 0.444 Rain 0.016 0.098 0.164 0.869 -0.176 0.208   82  Table A.65 AG276, wet season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Day -0.294 0.067 4.349 <0.001 -0.426 -0.161 River buffer -0.029 0.059 0.485 0.628 -0.144 0.087 Speed -0.021 0.049 0.432 0.666 -0.118 0.076 EVI2 0.025 0.059 0.415 0.678 -0.091 0.140 Rain 0.005 0.036 0.130 0.896 -0.066 0.075 Lag angle -0.003 0.035 0.096 0.924 -0.072 0.066 EVI 0.010 0.042 0.235 0.814 -0.073 0.093  Table A.66 AG276, wet season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Day -0.097 0.089 1.082 0.279 -0.272 0.079 EVI 0.183 0.093 1.964 0.050 <0.001 0.366 EVI2 -0.389 0.092 4.208 <0.001 -0.569 -0.208 Angle -0.017 0.049 0.359 0.720 -0.113 0.078 Lag speed 0.024 0.056 0.424 0.671 -0.086 0.133 Rain -0.006 0.039 0.156 0.876 -0.083 0.070 River buffer <0.001 0.045 0.004 0.997 -0.088 0.088  Table A.67 AG277, dry season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Fire -0.032 0.069 0.465 0.642 -0.166 0.103 Lag angle 0.012 0.033 0.382 0.702 -0.051 0.076 EVI 0.011 0.031 0.341 0.733 -0.050 0.072 EVI2 0.012 0.040 0.293 0.770 -0.066 0.089 Rain -0.005 0.044 0.121 0.904 -0.091 0.080 Day 0.003 0.025 0.104 0.917 -0.047 0.052 River buffer <0.001 0.025 0.013 0.989 -0.049 0.050 Speed <0.001 0.025 0.017 0.987 -0.048 0.049  83  Table A.68 AG277, dry season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Day -0.298 0.056 5.340 <0.001 -0.408 -0.189 Fire 0.182 0.127 1.434 0.152 -0.067 0.430 Rain 0.135 0.114 1.179 0.238 -0.089 0.358 EVI 0.318 0.055 5.806 <0.001 0.211 0.426 EVI2 -0.417 0.073 5.679 <0.001 -0.561 -0.273 Lag speed -0.133 0.067 1.983 0.047 -0.265 -0.002 River buffer -0.025 0.046 0.536 0.592 -0.115 0.066 Angle 0.001 0.028 0.019 0.985 -0.054 0.055  Table A.69 AG277, wet season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer 0.101 0.059 1.702 0.089 -0.015 0.216 Day -0.192 0.047 4.077 <0.001 -0.284 -0.100 Lag angle 0.098 0.057 1.731 0.084 -0.013 0.209 Speed -0.032 0.046 0.698 0.485 -0.122 0.058 EVI 0.002 0.028 0.065 0.948 -0.053 0.056 EVI2 0.000 0.028 0.003 0.998 -0.055 0.055 Rain 0.001 0.025 0.058 0.954 -0.047 0.050  Table A.70 AG277, wet season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer -0.123 0.057 2.138 0.033 -0.235 -0.010 Day -0.238 0.046 5.161 0.001 -0.328 -0.148 EVI 0.290 0.050 5.808 <0.001 0.192 0.387 EVI2 -0.208 0.053 3.959 0.001 -0.311 -0.105 Angle -0.032 0.045 0.708 0.479 -0.119 0.056 Lag speed -0.017 0.036 0.464 0.642 -0.087 0.054 Rain -0.001 0.024 0.023 0.982 -0.047 0.046  84  Table A.71 AM291, dry season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Road buffer -0.177 0.085 2.084 0.037 -0.343 -0.010 Day -0.065 0.073 0.888 0.375 -0.207 0.078 In omuramba 0.174 0.163 1.068 0.285 -0.145 0.493 EVI2 -0.042 0.059 0.719 0.472 -0.157 0.073 Lag angle 0.033 0.054 0.604 0.546 -0.073 0.138 Rain -0.026 0.055 0.469 0.639 -0.133 0.081 River buffer -0.018 0.053 0.343 0.732 -0.122 0.085 Speed -0.008 0.035 0.217 0.828 -0.075 0.060 EVI 0.008 0.039 0.213 0.831 -0.067 0.084  Table A.72 AM291, dry season, speed model Variable Estimate Std. Error value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Road buffer 0.339 0.115 2.955 0.003 0.114 0.564 Day -0.381 0.099 3.837 <0.001 -0.576 -0.186 In omuramba 0.142 0.204 0.694 0.488 -0.258 0.542 Rain 0.096 0.118 0.816 0.414 -0.135 0.327 EVI2 0.236 0.103 2.287 0.022 0.034 0.438 Lag speed -0.152 0.116 1.311 0.190 -0.379 0.075 EVI 0.020 0.065 0.304 0.761 -0.107 0.147 Angle -0.011 0.052 0.216 0.829 -0.113 0.091 River buffer -0.003 0.062 0.050 0.960 -0.125 0.119         85  Table A.73 AM291, wet season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) In omuramba 0.226 0.164 1.378 0.168 -0.095 0.546 Lag angle 0.255 0.072 3.543 <0.001 0.114 0.397 Speed -0.119 0.089 1.338 0.181 -0.292 0.055 Fence buffer 0.060 0.084 0.707 0.479 -0.106 0.225 River buffer 0.054 0.114 0.475 0.635 -0.170 0.279 EVI 0.022 0.054 0.414 0.679 -0.084 0.128 Road buffer -0.015 0.047 0.317 0.751 -0.106 0.077 EVI2 0.004 0.026 0.168 0.867 -0.047 0.055 Day 0.011 0.044 0.245 0.806 -0.076 0.098 Rain -0.003 0.037 0.070 0.944 -0.075 0.070  Table A.74 AM291, wet season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Day -0.485 0.078 6.193 <0.001 -0.638 -0.331 In omuramba 0.211 0.173 1.216 0.224 -0.129 0.551 Angle -0.156 0.092 1.692 0.091 -0.336 0.025 Lag speed -0.577 0.076 7.605 <0.001 -0.725 -0.428 Rain 0.041 0.068 0.608 0.543 -0.092 0.175 Road buffer -0.027 0.058 0.454 0.650 -0.141 0.088 Fence buffer 0.027 0.064 0.419 0.675 -0.099 0.152 River buffer 0.025 0.093 0.266 0.790 -0.158 0.208 EVI 0.008 0.045 0.185 0.854 -0.080 0.097 EVI2 -0.006 0.028 0.217 0.828 -0.061 0.048        86  Table A.75 77259, dry season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Pan 0.105 0.111 0.948 0.343 -0.112 0.322 Lag angle -0.071 0.093 0.761 0.447 -0.254 0.112 River buffer -0.097 0.142 0.680 0.497 -0.375 0.182 EVI -0.052 0.089 0.582 0.561 -0.226 0.122 EVI2 0.047 0.093 0.504 0.614 -0.135 0.229 Settlement 0.013 0.067 0.196 0.844 -0.118 0.145 Speed -0.020 0.059 0.339 0.734 -0.136 0.096 Day 0.006 0.048 0.130 0.896 -0.088 0.101 Fence buffer -0.011 0.059 0.183 0.855 -0.127 0.106 Rain 0.005 0.053 0.087 0.931 -0.100 0.109  Table A.76 77259, dry season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Fence buffer -0.681 0.123 5.517 <0.001 -0.923 -0.439 River buffer -0.501 0.169 2.961 0.003 -0.832 -0.169 EVI2 0.148 0.155 0.956 0.339 -0.155 0.451 Lag speed -0.125 0.124 1.004 0.315 -0.368 0.119 Angle -0.030 0.072 0.421 0.674 -0.172 0.111 Day 0.025 0.069 0.362 0.717 -0.111 0.162 Pan -0.009 0.057 0.154 0.877 -0.121 0.103 Rain 0.014 0.065 0.216 0.829 -0.113 0.141 EVI -0.008 0.061 0.122 0.903 -0.128 0.113 Settlement 0.005 0.062 0.086 0.932 -0.116 0.126        87  Table A.77 77259, wet season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Day 0.103 0.123 0.831 0.406 -0.139 0.345 Settlement -0.088 0.123 0.714 0.475 -0.329 0.153 Speed -0.051 0.093 0.548 0.584 -0.234 0.132 Fence buffer 0.049 0.102 0.477 0.633 -0.151 0.249 EVI2 -0.057 0.109 0.521 0.603 -0.272 0.158 Omuramba -0.042 0.124 0.333 0.739 -0.286 0.203 Lag angle -0.024 0.070 0.341 0.733 -0.162 0.114 EVI 0.046 0.100 0.459 0.647 -0.151 0.243 River buffer -0.051 0.143 0.354 0.723 -0.331 0.230 Pan -0.015 0.065 0.237 0.813 -0.143 0.112 Rain 0.006 0.057 0.106 0.915 -0.106 0.118  Table A.78 77259, wet season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer 0.330 0.288 1.145 0.252 -0.235 0.894 Day -0.451 0.129 3.495 <0.001 -0.703 -0.198 Settlement -0.099 0.138 0.715 0.475 -0.371 0.173 EVI 0.143 0.155 0.925 0.355 -0.160 0.447 Pan 0.060 0.109 0.552 0.581 -0.154 0.275 Angle -0.048 0.096 0.494 0.621 -0.236 0.141 Lag speed 0.044 0.094 0.470 0.638 -0.140 0.229 Rain 0.038 0.088 0.431 0.667 -0.135 0.211 EVI2 -0.015 0.089 0.171 0.865 -0.190 0.159 Fence buffer -0.024 0.091 0.268 0.789 -0.202 0.154 Omuramba -0.003 0.115 0.027 0.978 -0.228 0.222      88  Table A.79 77261new, wet season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer 0.131 0.090 1.454 0.146 -0.046 0.308 Road buffer 0.073 0.103 0.711 0.477 -0.129 0.275 Omuramba -0.198 0.096 2.059 0.040 -0.386 -0.010 EVI -0.057 0.067 0.845 0.398 -0.188 0.075 Day -0.039 0.053 0.731 0.465 -0.143 0.065 Speed -0.028 0.047 0.598 0.550 -0.120 0.064 Pan 0.029 0.050 0.592 0.554 -0.068 0.127 Settlement -0.030 0.053 0.562 0.574 -0.134 0.074 EVI2 -0.020 0.056 0.359 0.720 -0.130 0.089 Lag angle -0.007 0.030 0.244 0.808 -0.066 0.051 Rain 0.021 0.061 0.342 0.732 -0.099 0.141 Fence buffer 0.010 0.056 0.170 0.865 -0.100 0.119  Table A.80 77261new, wet season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer -0.171 0.140 1.214 0.225 -0.446 0.105 Day -0.532 0.075 7.131 <0.001 -0.678 -0.386 Rain 0.376 0.157 2.388 0.017 0.067 0.684 Settlement 0.123 0.101 1.217 0.224 -0.075 0.322 Angle -0.048 0.072 0.673 0.501 -0.190 0.093 EVI2 0.062 0.103 0.597 0.550 -0.141 0.264 Fence buffer -0.104 0.139 0.751 0.453 -0.377 0.168 Road buffer -0.066 0.123 0.534 0.593 -0.306 0.175 Pan -0.015 0.052 0.294 0.769 -0.117 0.086 Lag speed 0.013 0.046 0.280 0.779 -0.077 0.102 EVI -0.012 0.055 0.211 0.833 -0.120 0.097 Omuramba 0.000 0.062 0.002 0.999 -0.121 0.121     89  Table A.81 77261new, wet season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) EVI -0.076 0.064 1.175 0.240 -0.202 0.051 Speed -0.097 0.065 1.494 0.135 -0.224 0.030 Lag angle 0.033 0.049 0.664 0.507 -0.064 0.129 EVI2 0.037 0.060 0.621 0.534 -0.080 0.154 Settlement 0.021 0.043 0.485 0.628 -0.063 0.104 Omuramba 0.010 0.035 0.280 0.780 -0.060 0.080 Rain 0.005 0.029 0.187 0.851 -0.051 0.061 Day 0.005 0.028 0.180 0.857 -0.051 0.061 Fence buffer -0.015 0.060 0.259 0.796 -0.133 0.102 Road buffer 0.001 0.040 0.024 0.981 -0.077 0.078 Pan -0.002 0.028 0.080 0.936 -0.058 0.053  Table A.82 77261new, wet season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Day -0.106 0.076 1.389 0.165 -0.255 0.043 Rain -0.087 0.074 1.176 0.239 -0.233 0.058 Angle -0.111 0.075 1.475 0.140 -0.259 0.037 Settlement 0.045 0.064 0.705 0.481 -0.081 0.171 EVI -0.335 0.061 5.444 <0.001 -0.455 -0.214 Lag speed 0.035 0.056 0.614 0.539 -0.076 0.145 Road buffer -0.027 0.064 0.418 0.676 -0.153 0.100 EVI2 0.008 0.043 0.189 0.850 -0.076 0.092 Pan 0.009 0.038 0.230 0.818 -0.065 0.082 Omuramba 0.005 0.037 0.131 0.895 -0.068 0.077 Fence buffer 0.001 0.061 0.019 0.985 -0.119 0.122      90  Table A.83 77262, dry season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Fence buffer -0.520 0.161 3.235 0.001 -0.836 -0.205 Pan 0.102 0.103 0.985 0.325 -0.101 0.304 Lag angle 0.025 0.059 0.430 0.667 -0.090 0.141 EVI -0.026 0.060 0.428 0.669 -0.143 0.092 Rain 0.027 0.070 0.387 0.698 -0.110 0.164 Settlement -0.055 0.092 0.598 0.550 -0.235 0.125 EVI2 -0.018 0.063 0.286 0.775 -0.141 0.105 Day 0.008 0.044 0.171 0.864 -0.079 0.095 River buffer -0.015 0.058 0.261 0.794 -0.129 0.098 Speed 0.005 0.043 0.122 0.903 -0.079 0.090 Omuramba 0.000 0.106 0.002 0.999 -0.209 0.209  Table A.84 77262, dry season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Fence buffer -0.332 0.228 1.455 0.146 -0.780 0.115 Omuramba 0.190 0.255 0.742 0.458 -0.311 0.690 Rain -0.181 0.150 1.209 0.227 -0.474 0.112 EVI 0.034 0.075 0.449 0.653 -0.114 0.181 River buffer -0.060 0.100 0.601 0.548 -0.257 0.136 Lag speed -0.031 0.072 0.436 0.663 -0.172 0.110 EVI2 0.041 0.093 0.444 0.657 -0.140 0.223 Pan 0.013 0.060 0.224 0.823 -0.104 0.131 Angle 0.008 0.053 0.157 0.875 -0.097 0.113 Settlement 0.011 0.067 0.162 0.871 -0.121 0.142 Day 0.001 0.050 0.015 0.988 -0.098 0.100      91  Table A.85 77262, wet season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Rain -0.058 0.087 0.670 0.503 -0.229 0.112 Speed 0.068 0.091 0.741 0.459 -0.111 0.246 Settlement -0.064 0.096 0.669 0.504 -0.252 0.124 EVI -0.032 0.072 0.448 0.654 -0.175 0.110 Day 0.017 0.055 0.311 0.756 -0.092 0.126 EVI2 -0.016 0.067 0.235 0.814 -0.148 0.116 Pan 0.000 0.055 0.003 0.998 -0.108 0.108 Lag angle 0.005 0.047 0.109 0.913 -0.087 0.097 River buffer -0.013 0.062 0.213 0.832 -0.134 0.108  Table A.86 77262, wet season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Angle 0.105 0.133 0.790 0.430 -0.155 0.365 Pan -0.268 0.172 1.557 0.119 -0.605 0.069 Lag speed 0.087 0.125 0.695 0.487 -0.158 0.331 EVI 0.096 0.134 0.717 0.473 -0.167 0.359 Day 0.082 0.121 0.678 0.498 -0.155 0.320 Settlement 0.057 0.122 0.463 0.644 -0.183 0.297 River buffer -0.016 0.083 0.195 0.845 -0.178 0.146 EVI2 -0.022 0.101 0.219 0.827 -0.221 0.177 Rain -0.025 0.081 0.309 0.758 -0.184 0.134        92  Table A.87 77265new, dry season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Settlement 0.095 0.135 0.703 0.482 -0.170 0.359 EVI2 -0.149 0.197 0.756 0.450 -0.537 0.238 Speed -0.128 0.139 0.919 0.358 -0.402 0.145 Lag angle -0.061 0.106 0.575 0.566 -0.270 0.147 Rain 0.063 0.147 0.428 0.669 -0.226 0.352 Pan -0.018 0.087 0.206 0.837 -0.189 0.153 EVI -0.020 0.076 0.267 0.790 -0.170 0.129 River buffer -0.034 0.121 0.277 0.782 -0.273 0.205 Fire 0.020 0.143 0.143 0.887 -0.261 0.301 Day 0.005 0.065 0.073 0.942 -0.123 0.132 Fence buffer 0.019 0.110 0.171 0.864 -0.198 0.236  Table A.88 77265new, dry season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer -0.204 0.258 0.789 0.430 -0.709 0.302 Day 0.206 0.180 1.143 0.253 -0.147 0.558 Fire -0.240 0.322 0.744 0.457 -0.873 0.393 Angle -0.154 0.163 0.940 0.347 -0.474 0.167 Pan -0.136 0.172 0.789 0.430 -0.474 0.202 Settlement 0.048 0.121 0.392 0.695 -0.190 0.285 EVI 0.046 0.110 0.418 0.676 -0.170 0.263 Rain 0.037 0.141 0.263 0.792 -0.241 0.316 Fence 0.020 0.136 0.144 0.885 -0.247 0.287 EVI2 0.033 0.132 0.248 0.804 -0.227 0.293 Lag speed -0.016 0.080 0.195 0.845 -0.172 0.141      93  Table A.89 77265new, wet season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer 0.361 0.127 2.833 0.005 0.111 0.611 Speed -0.178 0.101 1.765 0.078 -0.375 0.020 Omuramba 0.081 0.131 0.619 0.536 -0.175 0.337 Road buffer 0.035 0.115 0.305 0.760 -0.190 0.261 Day -0.014 0.050 0.279 0.781 -0.112 0.084 EVI 0.011 0.049 0.229 0.819 -0.084 0.107 Rain -0.007 0.045 0.162 0.871 -0.095 0.080 Settlement -0.009 0.053 0.177 0.860 -0.113 0.094 Lag angle 0.002 0.042 0.050 0.960 -0.081 0.085 EVI2 -0.004 0.063 0.057 0.955 -0.128 0.121 Pan -0.001 0.044 0.013 0.989 -0.087 0.085  Table A.90 77265new, wet season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer -0.109 0.132 0.828 0.408 -0.367 0.149 Road buffer -0.481 0.181 2.659 0.008 -0.836 -0.127 Day -0.080 0.085 0.938 0.348 -0.246 0.087 Omuramba 0.136 0.146 0.928 0.353 -0.151 0.423 Angle -0.151 0.089 1.693 0.090 -0.327 0.024 Settlement -0.171 0.106 1.608 0.108 -0.379 0.037 EVI -0.302 0.076 3.940 <0.001 -0.452 -0.152 EVI2 0.176 0.134 1.314 0.189 -0.087 0.439 Rain 0.037 0.065 0.570 0.569 -0.090 0.165 Lag speed -0.029 0.059 0.491 0.623 -0.145 0.087 Pan 0.017 0.054 0.319 0.749 -0.089 0.124      94  Table A.91 SAT508, dry season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Rain 0.270 0.218 1.233 0.218 -0.159 0.699 Speed -0.155 0.193 0.801 0.423 -0.534 0.224 Lag angle 0.092 0.154 0.595 0.552 -0.211 0.396 Day 0.122 0.176 0.689 0.491 -0.224 0.468 Settlement -0.120 0.274 0.436 0.663 -0.659 0.419 EVI 0.025 0.107 0.231 0.817 -0.187 0.237 Fire -0.069 0.193 0.355 0.723 -0.449 0.312 Pan -0.006 0.134 0.046 0.963 -0.271 0.259 EVI2 0.044 0.200 0.218 0.828 -0.350 0.438 Road buffer -0.076 0.290 0.261 0.794 -0.648 0.495  Table A.92 SAT508, dry season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Road buffer -0.531 0.520 1.017 0.309 -1.555 0.492 Day -0.875 0.232 3.733 <0.001 -1.334 -0.416 Fire 0.740 0.475 1.549 0.121 -0.196 1.676 Angle -0.105 0.192 0.541 0.588 -0.484 0.274 EVI 0.084 0.183 0.459 0.646 -0.276 0.445 EVI2 0.074 0.274 0.267 0.790 -0.468 0.615 Settlement 0.002 0.332 0.005 0.996 -0.653 0.656 Rain 0.042 0.146 0.288 0.773 -0.246 0.330 Pan 0.001 0.173 0.003 0.997 -0.341 0.343 Lag speed 0.005 0.119 0.046 0.964 -0.230 0.241        95  Table A.93 SAT508, wet season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Rain 0.074 0.085 0.868 0.385 -0.093 0.241 Speed -0.185 0.091 2.028 0.043 -0.363 -0.006 Lag angle 0.048 0.073 0.663 0.507 -0.094 0.191 Day 0.035 0.064 0.545 0.586 -0.091 0.160 Settlement 0.024 0.065 0.370 0.711 -0.104 0.152 EVI 0.013 0.050 0.263 0.793 -0.085 0.111 Pan -0.010 0.059 0.169 0.866 -0.127 0.107 EVI2 -0.002 0.058 0.028 0.977 -0.115 0.112  Table A.94 SAT508, wet season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Angle -0.215 0.108 1.988 0.047 -0.428 -0.003 Pan -0.443 0.158 2.796 0.005 -0.754 -0.133 Settlement 0.134 0.152 0.881 0.378 -0.165 0.433 EVI -0.267 0.130 2.054 0.040 -0.522 -0.012 EVI2 0.046 0.108 0.422 0.673 -0.166 0.258 Day -0.016 0.056 0.291 0.771 -0.125 0.093 Lag speed -0.003 0.049 0.072 0.943 -0.099 0.092 Rain 0.000 0.049 0.004 0.997 -0.097 0.097         96  Table A.95 SAT509, dry season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer 2.456 0.965 2.532 0.011 0.555 4.357 Rain -0.196 0.271 0.718 0.473 -0.730 0.338 Settlement 0.919 0.492 1.859 0.063 -0.050 1.888 EVI2 -0.394 0.432 0.909 0.363 -1.244 0.456 EVI 0.128 0.222 0.576 0.565 -0.308 0.565 Omuramba 0.090 0.225 0.399 0.690 -0.353 0.534 Speed 0.065 0.165 0.389 0.697 -0.261 0.390 Road buffer -0.015 0.135 0.112 0.911 -0.282 0.251 Day 0.019 0.122 0.150 0.880 -0.223 0.260 Pan -0.020 0.133 0.148 0.882 -0.283 0.244 Fire 0.038 0.247 0.154 0.878 -0.449 0.526 Lag angle 0.014 0.121 0.117 0.907 -0.224 0.253  Table A.96 SAT509, dry season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Road buffer -0.331 0.301 1.096 0.273 -0.923 0.261 Day -0.558 0.283 1.961 0.050 -1.115 0.000 Settlement -0.573 0.361 1.580 0.114 -1.283 0.138 Rain 0.097 0.208 0.465 0.642 -0.312 0.506 Fire 0.165 0.349 0.469 0.639 -0.523 0.852 Angle 0.065 0.159 0.408 0.683 -0.249 0.379 EVI2 -0.112 0.262 0.425 0.671 -0.627 0.404 Lag speed -0.047 0.161 0.291 0.771 -0.365 0.271 EVI 0.034 0.141 0.237 0.813 -0.245 0.313 Omuramba -0.048 0.181 0.260 0.795 -0.405 0.310 River buffer 0.131 0.510 0.255 0.799 -0.874 1.136 Pan 0.006 0.122 0.050 0.960 -0.235 0.247     97  Table A.97 SAT509, wet season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer 0.174 0.132 1.314 0.189 -0.086 0.434 EVI -0.067 0.085 0.789 0.430 -0.234 0.100 Lag angle 0.066 0.082 0.806 0.420 -0.095 0.228 Speed -0.036 0.066 0.546 0.585 -0.165 0.093 Rain -0.031 0.063 0.484 0.629 -0.154 0.093 Road buffer -0.030 0.088 0.337 0.736 -0.203 0.143 Pan 0.021 0.054 0.382 0.702 -0.085 0.126 Day 0.012 0.046 0.253 0.801 -0.079 0.102 EVI2 -0.019 0.078 0.240 0.810 -0.172 0.135 Omuramba -0.008 0.085 0.089 0.929 -0.174 0.159 Settlement -0.006 0.062 0.099 0.921 -0.129 0.116   Table A.98 SAT509, wet season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Road buffer -0.451 0.158 2.848 0.004 -0.762 -0.141 Day -0.481 0.085 5.673 <0.001 -0.647 -0.315 Rain 0.135 0.106 1.263 0.207 -0.074 0.343 EVI -0.323 0.089 3.639 <0.001 -0.497 -0.149 Lag speed -0.153 0.106 1.448 0.148 -0.360 0.054 Angle -0.046 0.075 0.611 0.541 -0.192 0.101 EVI2 0.059 0.118 0.501 0.616 -0.172 0.291 Pan 0.027 0.064 0.427 0.669 -0.098 0.152 River buffer 0.043 0.096 0.442 0.658 -0.146 0.231 Settlement 0.028 0.074 0.375 0.707 -0.118 0.174 Omuramba -0.005 0.091 0.059 0.953 -0.183 0.173     98  Table A.99 77263, dry season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer -0.279 0.114 2.452 0.014 -0.502 -0.056 Pan 0.221 0.109 2.023 0.043 0.007 0.434 Settlement 0.064 0.092 0.694 0.488 -0.116 0.243 Speed -0.045 0.078 0.579 0.563 -0.198 0.107 Day -0.045 0.081 0.557 0.577 -0.204 0.114 Lag angle -0.018 0.056 0.317 0.751 -0.128 0.092 EVI -0.022 0.068 0.326 0.745 -0.156 0.111 EVI2 -0.014 0.063 0.229 0.819 -0.138 0.109 Rain -0.002 0.054 0.038 0.970 -0.108 0.104  Table A.100 77263, dry season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Day -0.337 0.104 3.242 0.001 -0.540 -0.133 Rain -0.173 0.139 1.241 0.215 -0.446 0.100 Settlement -0.114 0.115 0.988 0.323 -0.340 0.112 EVI2 0.072 0.111 0.642 0.521 -0.147 0.290 Lag speed -0.110 0.111 0.987 0.324 -0.328 0.108 EVI 0.046 0.085 0.535 0.593 -0.122 0.213 Angle -0.055 0.088 0.630 0.528 -0.228 0.117 River buffer -0.044 0.089 0.496 0.620 -0.219 0.131 Pan -0.020 0.063 0.320 0.749 -0.143 0.103         99  Table A.101 77263, wet season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer -0.485 0.222 2.180 0.029 -0.920 -0.049 Road buffer 0.116 0.160 0.723 0.470 -0.198 0.430 Omuramba 0.067 0.117 0.569 0.569 -0.163 0.296 EVI2 0.062 0.120 0.514 0.607 -0.173 0.297 Day 0.027 0.074 0.368 0.713 -0.117 0.172 Pan -0.036 0.081 0.439 0.661 -0.194 0.123 Lag angle -0.020 0.066 0.304 0.761 -0.149 0.109 Rain -0.017 0.067 0.258 0.797 -0.149 0.114 Settlement 0.014 0.077 0.181 0.856 -0.137 0.165 Speed -0.007 0.056 0.128 0.899 -0.117 0.103 EVI 0.010 0.083 0.125 0.901 -0.152 0.173  Table A.102 77263, wet season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Omuramba -0.187 0.201 0.928 0.354 -0.582 0.208 Lag speed 0.221 0.175 1.261 0.207 -0.123 0.565 Rain -0.064 0.122 0.519 0.604 -0.304 0.177 EVI2 -0.068 0.147 0.461 0.645 -0.357 0.221 Road buffer -0.083 0.169 0.489 0.625 -0.415 0.249 EVI -0.026 0.104 0.247 0.805 -0.230 0.179 River buffer 0.061 0.169 0.360 0.719 -0.271 0.393 Angle -0.022 0.084 0.258 0.796 -0.188 0.144 Pan -0.024 0.087 0.280 0.780 -0.196 0.147 Settlement 0.007 0.087 0.079 0.937 -0.164 0.178 Day 0.008 0.076 0.102 0.919 -0.142 0.157      100  Table A.103 77263new, dry season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Rain 0.170 0.148 1.151 0.250 -0.120 0.460 Speed -0.176 0.110 1.596 0.111 -0.391 0.040 Omuramba -0.051 0.085 0.598 0.550 -0.219 0.117 EVI -0.065 0.094 0.684 0.494 -0.250 0.121 Day 0.054 0.084 0.641 0.522 -0.111 0.219 Road buffer 0.047 0.087 0.535 0.593 -0.124 0.218 EVI2 -0.049 0.103 0.473 0.636 -0.251 0.153 Settlement 0.023 0.060 0.383 0.702 -0.095 0.142 Pan 0.012 0.052 0.227 0.820 -0.091 0.115 Lag angle -0.011 0.050 0.225 0.822 -0.109 0.086  Table A.104 77263new, dry season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Day -0.513 0.107 4.783 <0.001 -0.723 -0.303 Angle -0.192 0.132 1.456 0.145 -0.450 0.066 Pan 0.113 0.123 0.915 0.360 -0.129 0.354 EVI -0.311 0.129 2.400 0.016 -0.564 -0.057 EVI2 0.354 0.200 1.773 0.076 -0.037 0.746 Lag speed -0.226 0.132 1.704 0.088 -0.485 0.034 Rain 0.067 0.124 0.538 0.591 -0.176 0.310 Omuramba -0.039 0.087 0.441 0.660 -0.210 0.133 Road buffer -0.015 0.072 0.208 0.836 -0.157 0.127 Settlement -0.006 0.056 0.101 0.920 -0.116 0.105        101  Table A.105 77263new, wet season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Rain 0.065 0.083 0.779 0.436 -0.098 0.227 EVI2 -0.096 0.123 0.780 0.435 -0.337 0.145 Lag angle -0.037 0.065 0.561 0.575 -0.165 0.091 Day 0.034 0.063 0.533 0.594 -0.090 0.157 Omuramba 0.019 0.057 0.325 0.745 -0.094 0.131 Speed 0.016 0.049 0.332 0.740 -0.079 0.112 EVI 0.014 0.049 0.280 0.780 -0.082 0.110 Settlement 0.001 0.043 0.021 0.984 -0.083 0.085 Road buffer -0.003 0.060 0.056 0.955 -0.122 0.115 Pan 0.001 0.040 0.013 0.989 -0.079 0.080  Table A.106 77263new, wet season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Road buffer 0.149 0.131 1.131 0.258 -0.109 0.406 Day -0.061 0.075 0.810 0.418 -0.209 0.087 Rain 0.119 0.092 1.295 0.195 -0.061 0.298 EVI2 0.041 0.085 0.484 0.628 -0.125 0.208 Settlement 0.015 0.050 0.306 0.760 -0.083 0.114 Omuramba 0.028 0.064 0.440 0.660 -0.098 0.155 Angle 0.014 0.044 0.318 0.750 -0.073 0.101 EVI -0.003 0.039 0.069 0.945 -0.080 0.075 Lag speed -0.002 0.037 0.054 0.957 -0.074 0.070 Pan -0.003 0.039 0.088 0.930 -0.079 0.073       102  Table A.107 94041, dry season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Pan 0.056 0.073 0.776 0.438 -0.086 0.199 Speed -0.095 0.077 1.225 0.220 -0.247 0.057 Rain 0.045 0.085 0.529 0.597 -0.122 0.212 EVI2 0.046 0.088 0.521 0.603 -0.127 0.219 Day 0.023 0.049 0.467 0.641 -0.073 0.118 River buffer -0.027 0.058 0.455 0.649 -0.141 0.088 EVI -0.018 0.049 0.373 0.709 -0.114 0.078 Lag angle 0.004 0.033 0.134 0.893 -0.061 0.070 Omuramba 0.002 0.082 0.021 0.984 -0.158 0.162 Settlement 0.016 0.051 0.303 0.762 -0.085 0.116  Table A.108 94041, dry season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Day -0.244 0.066 3.672 <0.001 -0.374 -0.114 Rain 0.141 0.128 1.100 0.271 -0.110 0.392 Angle -0.094 0.080 1.177 0.239 -0.250 0.062 Pan -0.074 0.082 0.899 0.369 -0.236 0.088 Lag speed -0.091 0.079 1.149 0.250 -0.247 0.064 River buffer -0.025 0.060 0.412 0.680 -0.142 0.093 EVI -0.005 0.041 0.114 0.909 -0.085 0.076 Settlement 0.015 0.054 0.276 0.783 -0.090 0.120 EVI2 0.005 0.059 0.087 0.931 -0.110 0.120 Omuramba 0.014 0.088 0.163 0.870 -0.159 0.188        103  Table A.109 94041, wet season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Speed -0.111 0.081 1.370 0.171 -0.271 0.048 Omuramba -0.029 0.058 0.499 0.618 -0.143 0.085 Day -0.026 0.051 0.497 0.619 -0.126 0.075 Settlement 0.010 0.042 0.242 0.808 -0.073 0.093 River buffer 0.030 0.091 0.323 0.747 -0.150 0.209 Lag angle 0.012 0.040 0.311 0.756 -0.065 0.090 Road buffer 0.019 0.054 0.353 0.724 -0.086 0.124 EVI2 0.009 0.057 0.156 0.876 -0.103 0.121 Pan 0.003 0.037 0.085 0.933 -0.069 0.075 Rain -0.002 0.035 0.070 0.944 -0.071 0.067 EVI 0.005 0.038 0.122 0.903 -0.069 0.078  Table A.110 94041, wet season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Road buffer 0.120 0.103 1.160 0.246 -0.082 0.322 Day -0.146 0.077 1.902 0.057 -0.297 0.004 Rain 0.229 0.070 3.268 0.001 0.092 0.366 Angle -0.125 0.076 1.646 0.100 -0.275 0.024 Settlement 0.070 0.081 0.865 0.387 -0.089 0.228 EVI -0.466 0.072 6.466 <0.001 -0.607 -0.325 EVI2 0.586 0.105 5.569 <0.001 0.380 0.792 Lag speed 0.287 0.068 4.223 <0.001 0.154 0.420 Pan 0.028 0.058 0.489 0.625 -0.086 0.143 River buffer -0.023 0.088 0.258 0.797 -0.195 0.150 Omuramba 0.006 0.045 0.129 0.897 -0.082 0.094      104  Table A.111 94042, dry season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Day 0.056 0.069 0.808 0.419 -0.080 0.192 EVI -0.058 0.071 0.823 0.410 -0.198 0.081 Settlement  0.0408 0.062 0.663 0.508 -0.080 0.162 EVI2 -0.063 0.101 0.626 0.531 -0.261 0.135 Lag angle -0.019 0.046 0.415 0.678 -0.109 0.071 Speed -0.012 0.040 0.298 0.766 -0.091 0.067 Road buffer 0.008 0.043 0.187 0.852 -0.077 0.093 Rain 0.002 0.049 0.038 0.969 -0.095 0.099 Pan -0.001 0.033 0.038 0.969 -0.066 0.064 Omuramba 0.004 0.036 0.110 0.912 -0.066 0.074  Table A.112 94042, dry season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Day -0.562 0.098 5.743 <0.001 -0.754 -0.370 Rain 0.164 0.169 0.970 0.332 -0.167 0.496 EVI -0.125 0.123 1.016 0.310 -0.366 0.116 EVI2 0.401 0.197 2.035 0.042 0.015 0.786 Lag speed -0.080 0.103 0.773 0.440 -0.282 0.123 Road buffer 0.087 0.121 0.721 0.471 -0.150 0.325 Pan 0.060 0.093 0.643 0.520 -0.122 0.241 Omuramba 0.037 0.078 0.468 0.640 -0.117 0.190 Angle -0.016 0.058 0.267 0.790 -0.130 0.099 Settlement -0.007 0.052 0.138 0.890 -0.109 0.095        105  Table A.113 94042, wet season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) EVI2 -0.084 0.115 0.726 0.468 -0.309 0.142 Speed -0.171 0.082 2.065 0.039 -0.332 -0.009 Rain 0.034 0.060 0.560 0.576 -0.085 0.152 Day 0.015 0.045 0.333 0.739 -0.073 0.103 EVI 0.014 0.045 0.300 0.764 -0.075 0.102 Settlement 0.002 0.037 0.040 0.968 -0.070 0.073 Lag angle -0.003 0.036 0.085 0.932 -0.074 0.068 Pan -0.004 0.037 0.098 0.922 -0.076 0.069 Road buffer 0.001 0.051 0.022 0.982 -0.099 0.102 Omuramba 0.003 0.039 0.071 0.943 -0.075 0.080  Table A.114 94042, wet season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Day -0.212 0.078 2.700 0.007 -0.365 -0.058 Omuramba 0.084 0.091 0.926 0.354 -0.094 0.263 Rain 0.196 0.083 2.366 0.018 0.034 0.358 Angle -0.173 0.084 2.056 0.040 -0.337 -0.008 Settlement 0.187 0.085 2.195 0.028 0.020 0.355 Pan 0.049 0.072 0.674 0.500 -0.093 0.190 Road buffer 0.073 0.104 0.698 0.485 -0.131 0.276 EVI2 -0.014 0.066 0.209 0.835 -0.143 0.116 Lag speed -0.003 0.038 0.091 0.927 -0.078 0.071 EVI -0.006 0.042 0.153 0.878 -0.088 0.075       106  Table A.115 AG269, dry season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Rain 0.194 0.133 1.458 0.145 -0.067 0.454 Pan 0.150 0.070 2.126 0.034 0.012 0.288 EVI -0.158 0.072 2.187 0.029 -0.300 -0.016 Speed -0.173 0.063 2.762 0.006 -0.296 -0.050 Day -0.040 0.057 0.702 0.483 -0.153 0.072 Settlement 0.012 0.040 0.305 0.760 -0.067 0.091 Lag angle 0.010 0.035 0.278 0.781 -0.059 0.078 River buffer 0.003 0.041 0.083 0.933 -0.076 0.083 EVI2 -0.003 0.054 0.058 0.953 -0.108 0.102  Table A.116 AG269, dry season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer -0.065 0.072 0.890 0.373 -0.207 0.078 Angle -0.160 0.056 2.866 0.004 -0.269 -0.051 Pan -0.035 0.053 0.662 0.508 -0.139 0.069 Lag speed -0.070 0.063 1.111 0.267 -0.192 0.053 EVI -0.022 0.044 0.489 0.625 -0.108 0.065 Settlement -0.015 0.043 0.350 0.726 -0.100 0.070 Day -0.011 0.034 0.339 0.734 -0.077 0.055 EVI2 0.014 0.052 0.261 0.794 -0.089 0.116 Rain -0.006 0.050 0.116 0.908 -0.104 0.092         107  Table A.117 AG269, wet season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer 0.099 0.100 0.990 0.322 -0.097 0.296 Road buffer -0.086 0.083 1.035 0.301 -0.248 0.077 Day 0.040 0.053 0.761 0.447 -0.064 0.145 Speed -0.015 0.036 0.407 0.684 -0.086 0.056 Omuramba 0.017 0.051 0.325 0.745 -0.083 0.117 Pan 0.017 0.042 0.405 0.685 -0.065 0.099 Rain -0.007 0.030 0.229 0.819 -0.066 0.052 Settlement 0.005 0.032 0.142 0.887 -0.058 0.067 EVI -0.004 0.034 0.114 0.910 -0.071 0.063 Lag angle 0.001 0.026 0.024 0.981 -0.051 0.052 EVI2 0.002 0.030 0.075 0.940 -0.057 0.061  Table A.118 AG269, wet season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer -0.128 0.107 1.189 0.235 -0.338 0.083 Day -0.172 0.051 3.368 0.001 -0.271 -0.072 Omuramba 0.273 0.076 3.591 <0.001 0.124 0.423 EVI -0.288 0.059 4.858 <0.001 -0.405 -0.172 Lag speed 0.160 0.052 3.046 0.002 0.057 0.263 Settlement -0.022 0.044 0.491 0.623 -0.107 0.064 Road buffer 0.018 0.046 0.389 0.697 -0.072 0.108 Rain 0.011 0.033 0.349 0.727 -0.052 0.075 Angle -0.009 0.030 0.303 0.762 -0.067 0.049 EVI2 -0.012 0.036 0.338 0.735 -0.084 0.059 Pan 0.007 0.034 0.201 0.841 -0.060 0.074      108  Table A.119 AG270, dry season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer 0.240 0.137 1.751 0.080 -0.029 0.508 Road buffer -0.253 0.092 2.753 0.006 -0.433 -0.073 Settlement 0.084 0.085 0.991 0.322 -0.083 0.252 EVI2 0.082 0.091 0.901 0.368 -0.097 0.261 Lag angle 0.103 0.071 1.449 0.147 -0.036 0.243 Pan 0.039 0.064 0.609 0.543 -0.087 0.165 Speed -0.033 0.053 0.628 0.530 -0.137 0.071 Day 0.040 0.058 0.698 0.485 -0.073 0.154 EVI 0.024 0.053 0.462 0.644 -0.079 0.127 Rain 0.011 0.061 0.188 0.851 -0.108 0.131  Table A.120 AG270, dry season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer -0.346 0.137 2.531 0.011 -0.614 -0.078 Angle -0.049 0.066 0.735 0.462 -0.179 0.081 Settlement 0.308 0.073 4.215 <0.001 0.165 0.451 EVI -0.055 0.074 0.743 0.458 -0.201 0.091 Lag speed -0.305 0.065 4.703 <0.001 -0.432 -0.178 Road buffer 0.040 0.073 0.544 0.586 -0.104 0.183 Rain 0.037 0.089 0.416 0.678 -0.137 0.211 EVI2 0.007 0.054 0.139 0.889 -0.098 0.113 Pan -0.007 0.041 0.181 0.857 -0.088 0.073 Day -0.003 0.034 0.075 0.940 -0.070 0.065        109  Table A.121 AG270, wet season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Lag angle -0.065 0.071 0.918 0.359 -0.205 0.074 Road buffer -0.046 0.093 0.500 0.617 -0.228 0.135 River buffer -0.033 0.067 0.492 0.622 -0.163 0.098 Speed -0.025 0.050 0.501 0.616 -0.123 0.073 Settlement 0.013 0.043 0.303 0.762 -0.071 0.097 Day -0.017 0.043 0.387 0.699 -0.101 0.068 Rain 0.009 0.037 0.233 0.816 -0.064 0.081 EVI2 -0.010 0.056 0.185 0.853 -0.119 0.099 Pan 0.003 0.033 0.079 0.937 -0.062 0.067 EVI -0.007 0.037 0.192 0.848 -0.079 0.065  Table A.122 AG270, wet season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer -0.167 0.114 1.469 0.142 -0.390 0.056 Pan -0.105 0.078 1.341 0.180 -0.259 0.049 EVI -0.167 0.078 2.140 0.032 -0.320 -0.014 EVI2 0.208 0.125 1.666 0.096 -0.037 0.452 Lag speed 0.034 0.056 0.597 0.550 -0.077 0.144 Angle -0.028 0.051 0.536 0.592 -0.128 0.073 Settlement 0.050 0.073 0.682 0.495 -0.094 0.194 Road buffer 0.040 0.088 0.450 0.653 -0.133 0.213 Day -0.017 0.043 0.395 0.693 -0.102 0.067 Rain -0.013 0.042 0.321 0.749 -0.096 0.069       110  Table A.123 77265, dry season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Speed -0.040 0.061 0.658 0.511 -0.160 0.079 Rain -0.040 0.081 0.486 0.627 -0.199 0.120 Road buffer -0.023 0.059 0.386 0.700 -0.138 0.093 River buffer -0.013 0.044 0.301 0.764 -0.099 0.073 EVI 0.009 0.037 0.232 0.816 -0.065 0.082 Settlement 0.009 0.040 0.230 0.818 -0.069 0.087 Lag angle 0.005 0.034 0.135 0.893 -0.062 0.071 Day -0.004 0.034 0.131 0.896 -0.071 0.063 EVI2 0.003 0.040 0.067 0.947 -0.075 0.081 Pan 0.001 0.034 0.042 0.966 -0.065 0.068  Table A.124 77265, dry season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer -0.525 0.076 6.852 <0.001 -0.675 -0.375 Day -0.140 0.082 1.716 0.086 -0.301 0.020 Rain 0.097 0.117 0.827 0.408 -0.133 0.326 Angle -0.047 0.065 0.723 0.470 -0.175 0.081 Pan -0.156 0.088 1.767 0.077 -0.329 0.017 Settlement 0.144 0.099 1.458 0.145 -0.050 0.338 EVI -0.148 0.084 1.762 0.078 -0.312 0.017 Lag speed -0.207 0.072 2.872 0.004 -0.348 -0.066 Road buffer 0.035 0.080 0.435 0.664 -0.122 0.191 EVI2 -0.010 0.046 0.205 0.837 -0.101 0.082        111  Table A.125 77265, wet season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Rain 0.070 0.088 0.795 0.427 -0.103 0.242 Speed -0.046 0.074 0.620 0.536 -0.190 0.099 Lag angle 0.027 0.060 0.454 0.650 -0.091 0.146 EVI 0.033 0.068 0.488 0.626 -0.100 0.166 Road buffer 0.021 0.068 0.302 0.763 -0.113 0.154 Day 0.014 0.049 0.290 0.772 -0.082 0.110 Settlement 0.004 0.046 0.095 0.925 -0.094 0.085 Pan 0.007 0.044 0.157 0.875 -0.079 0.093 EVI2 -0.005 0.042 0.128 0.898 -0.088 0.077 River buffer -0.001 0.044 0.033 0.974 -0.089 0.086  Table A.126 77265, wet season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer -0.146 0.106 1.379 0.168 -0.354 0.062 EVI2 0.119 0.093 1.273 0.203 -0.064 0.302 Angle -0.042 0.069 0.605 0.545 -0.178 0.094 Settlement 0.034 0.071 0.486 0.627 -0.104 0.173 Lag speed 0.017 0.051 0.345 0.730 -0.082 0.117 Rain 0.022 0.056 0.397 0.691 -0.087 0.131 Road buffer 0.027 0.072 0.367 0.713 -0.115 0.169 EVI -0.015 0.054 0.284 0.777 -0.122 0.091 Pan -0.012 0.046 0.262 0.793 -0.103 0.078 Day -0.002 0.040 0.056 0.955 -0.081 0.076       112  Table A.127 AG278, dry season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Road buffer -0.090 0.081 1.106 0.269 -0.249 0.069 Day 0.068 0.060 1.134 0.257 -0.050 0.187 Pan 0.110 0.066 1.672 0.094 -0.019 0.240 Lag angle 0.123 0.058 2.095 0.036 0.008 0.237 Speed -0.080 0.061 1.299 0.194 -0.200 0.041 Fire 0.044 0.087 0.505 0.613 -0.126 0.214 EVI2 -0.024 0.060 0.406 0.685 -0.142 0.093 Settlement -0.010 0.043 0.245 0.807 -0.094 0.073 EVI 0.022 0.049 0.444 0.657 -0.074 0.118 River buffer 0.011 0.042 0.254 0.799 -0.071 0.093 Rain -0.005 0.050 0.099 0.921 -0.104 0.094  Table A.128 AG278, dry season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Day -0.092 0.068 1.365 0.172 -0.225 0.040 Fire 0.121 0.133 0.908 0.364 -0.140 0.381 Rain 0.321 0.111 2.885 0.004 0.103 0.540 Angle -0.084 0.066 1.273 0.203 -0.212 0.045 Settlement 0.183 0.078 2.346 0.019 0.030 0.335 EVI -0.176 0.099 1.779 0.075 -0.369 0.018 EVI2 0.121 0.116 1.044 0.296 -0.106 0.348 Lag speed -0.235 0.053 4.413 0.000 -0.339 -0.130 Road buffer 0.035 0.064 0.552 0.581 -0.090 0.161 River buffer -0.016 0.051 0.323 0.747 -0.116 0.083 Pan -0.002 0.032 0.071 0.943 -0.065 0.061      113  Table A.129 AG278, wet season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Rain 0.046 0.053 0.879 0.379 -0.057 0.150 Pan -0.035 0.048 0.726 0.468 -0.128 0.059 Speed -0.164 0.048 3.398 0.001 -0.258 -0.069 EVI -0.021 0.040 0.539 0.590 -0.100 0.057 Settlement -0.028 0.046 0.592 0.554 -0.119 0.064 River buffer -0.047 0.086 0.548 0.584 -0.216 0.122 EVI2 -0.010 0.033 0.314 0.754 -0.075 0.054 Day 0.009 0.030 0.309 0.757 -0.049 0.067 Lag angle 0.006 0.027 0.239 0.811 -0.047 0.060  Table A.130 AG278, wet season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer 0.073 0.092 0.797 0.425 -0.106 0.252 Angle -0.140 0.040 3.488 <0.001 -0.218 -0.061 Settlement 0.103 0.052 1.967 0.049 0.000 0.205 EVI -0.164 0.040 4.123 <0.001 -0.242 -0.086 Lag speed -0.034 0.042 0.798 0.425 -0.116 0.049 EVI2 0.016 0.034 0.479 0.632 -0.050 0.082 Day -0.005 0.023 0.222 0.824 -0.050 0.039 Rain 0.004 0.022 0.203 0.839 -0.039 0.048 Pan 0.002 0.021 0.075 0.940 -0.039 0.042        114  Table A.131 AG279, dry season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Road buffer -0.590 0.348 1.687 0.092 -1.274 0.095 Settlement 0.190 0.241 0.787 0.432 -0.283 0.663 EVI -0.271 0.242 1.116 0.265 -0.747 0.205 EVI2 -0.839 0.345 2.415 0.016 -1.519 -0.158 Lag angle -0.141 0.182 0.768 0.442 -0.500 0.218 Day 0.067 0.140 0.479 0.632 -0.208 0.343 Fire 0.044 0.200 0.216 0.829 -0.352 0.439 Speed 0.025 0.096 0.254 0.799 -0.165 0.214 River buffer 0.013 0.152 0.086 0.932 -0.286 0.313 Pan -0.016 0.107 0.145 0.885 -0.227 0.196 Rain 0.014 0.103 0.138 0.890 -0.190 0.219  Table A.132 AG279, dry season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer -0.140 0.213 0.656 0.512 -0.560 0.279 Rain 0.137 0.196 0.694 0.487 -0.250 0.524 EVI2 0.183 0.290 0.629 0.530 -0.388 0.754 Pan -0.075 0.157 0.473 0.636 -0.384 0.235 Day -0.078 0.152 0.512 0.609 -0.377 0.221 Lag speed -0.039 0.117 0.333 0.739 -0.270 0.192 EVI -0.018 0.113 0.158 0.874 -0.241 0.205 Road buffer -0.011 0.162 0.065 0.948 -0.332 0.310 Settlement 0.006 0.128 0.049 0.961 -0.246 0.258 Angle 0.014 0.097 0.148 0.883 -0.177 0.206 Fire 0.044 0.209 0.209 0.834 -0.368 0.456      115  Table A.133 AG279, wet season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Road buffer 0.277 0.284 0.973 0.330 -0.281 0.835 Rain 0.125 0.104 1.195 0.232 -0.080 0.329 EVI2 0.036 0.065 0.557 0.578 -0.091 0.163 Day -0.035 0.068 0.507 0.612 -0.168 0.099 Lag angle -0.021 0.056 0.365 0.715 -0.131 0.090 Speed -0.016 0.054 0.294 0.769 -0.123 0.091 Settlement -0.016 0.057 0.286 0.775 -0.128 0.096 River buffer 0.013 0.066 0.204 0.838 -0.116 0.143 EVI 0.004 0.049 0.089 0.929 -0.091 0.100 Pan -0.007 0.047 0.147 0.883 -0.100 0.086  Table A.134 AG279, wet season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer -0.709 0.160 4.418 <0.001 -1.023 -0.394 Road buffer 0.351 0.349 1.003 0.316 -0.335 1.036 Rain -0.080 0.104 0.770 0.441 -0.284 0.124 Pan -0.194 0.126 1.536 0.124 -0.442 0.054 Settlement 0.144 0.144 0.997 0.319 -0.139 0.427 EVI 0.100 0.122 0.818 0.413 -0.139 0.338 EVI2 -0.222 0.110 2.014 0.044 -0.438 -0.006 Lag speed 0.768 0.109 7.034 <0.001 0.554 0.982 Day -0.017 0.059 0.290 0.772 -0.134 0.099 Angle 0.012 0.055 0.219 0.827 -0.096 0.120      116  Table A.135 AM291, dry season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Day -0.072 0.074 0.970 0.332 -0.216 0.073 Settlement 0.059 0.071 0.823 0.410 -0.081 0.199 EVI2 -0.035 0.052 0.680 0.496 -0.137 0.066 River buffer -0.188 0.221 0.849 0.396 -0.621 0.246 Rain -0.034 0.062 0.549 0.583 -0.157 0.088 Lag angle 0.025 0.049 0.511 0.609 -0.070 0.120 Pan -0.010 0.043 0.240 0.811 -0.095 0.074 Speed -0.014 0.040 0.341 0.733 -0.091 0.064 EVI 0.002 0.035 0.065 0.948 -0.067 0.071  Table A.136 AM291, dry season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) River buffer -0.858 0.352 2.435 0.015 -1.549 -0.167 Day -0.369 0.098 3.745 <0.001 -0.562 -0.176 Pan 0.353 0.116 3.057 0.002 0.127 0.580 EVI2 0.258 0.090 2.868 0.004 0.082 0.434 Lag speed -0.215 0.113 1.906 0.057 -0.435 0.006 Settlement 0.067 0.107 0.625 0.532 -0.142 0.276 EVI 0.047 0.089 0.531 0.595 -0.127 0.221 Rain 0.034 0.081 0.419 0.675 -0.125 0.192 Angle -0.022 0.061 0.367 0.713 -0.142 0.097         117  Table A.137 AM291, wet season, angle model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Settlement 0.103 0.095 1.087 0.277 -0.083 0.289 Lag angle 0.256 0.072 3.554 <0.001 0.115 0.398 Speed -0.117 0.089 1.315 0.189 -0.291 0.057 Pan -0.037 0.065 0.565 0.572 -0.164 0.091 EVI 0.014 0.048 0.293 0.769 -0.080 0.108 EVI2 0.005 0.029 0.168 0.867 -0.053 0.062 River buffer -0.026 0.065 0.403 0.687 -0.155 0.102 Rain -0.005 0.039 0.133 0.894 -0.081 0.071 Day 0.003 0.039 0.087 0.931 -0.072 0.079  Table A.138 AM291, wet season, speed model Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Day -0.497 0.076 6.534 <0.001 -0.646 -0.348 Angle -0.160 0.091 1.754 0.079 -0.338 0.019 Settlement 0.255 0.097 2.610 0.009 0.063 0.446 Lag speed -0.574 0.075 7.657 <0.001 -0.720 -0.427 River buffer 0.056 0.088 0.630 0.529 -0.117 0.229 Rain 0.032 0.062 0.521 0.603 -0.089 0.154 EVI2 -0.007 0.031 0.218 0.827 -0.069 0.055 EVI -0.006 0.043 0.128 0.898 -0.091 0.080 Pan 0.001 0.041 0.021 0.983 -0.080 0.082       118  A.3 Additional results from crossing model Table A.139 Mahango crossing model, 1000m. Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Angle -1.455 0.479 3.040 0.002 -2.393 -0.517 Dry season -2.837 1.015 2.793 0.005 -4.828 -0.846 Speed 1.928 0.442 4.358 <0.001 1.061 2.795 Day -0.140 0.3945 0.356 0.722 -0.914 0.633 EVI -0.354 0.672 0.527 0.598 -1.670 0.963 In omuramba -0.211 0.644 0.327 0.743 -1.472 1.051  Table A.140 Mahango crossing model, 5000m. Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.50%) Angle -1.184 0.307 3.862 <0.001 -1.785 -0.583 Day -0.253 0.300 0.843 0.399 -0.842 0.336 Dry season -3.372 0.673 5.011 <0.001 -4.689 -2.052 EVI 0.399 0.274 1.454 0.146 -0.139 0.936 Speed 0.686 0.172 3.997 <0.001 0.350 1.023  Table A.141 Ring road crossing model, 1000m. Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Settlement 2.234 0.718 3.11 0.002 0.826 3.641 Day 0.050 0.256 0.155 0.878 -0.462 0.541 EVI -0.469 0.562 0.835 0.404 -1.570 0.632 Speed 1.774 0.323 5.483 < 0.001 1.140 2.408 Angle 0.021 0.257 0.081 0.935 -0.483 0.5245 Dry season 0.048 0.415 0.116 0.908 -0.765 0.861     119  Table A.142 Ring road crossing model, 5000m. Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Angle -0.040 0.169 0.236 0.8132 -0.371 0.291 Settlement 0.989 0.281 3.517 <0.001 0.438 1.541 Day -0.125 0.193 0.644 0.520 -0.504 0.255 Dry season -0.494 0.388 1.273 0.203 -1.254 0.267 EVI -0.636 0.412 1.543 0.123 -1.443 0.172 Speed 1.901 0.122 15.547 <0.001 1.662 2.141  Table A.143 Tar road crossing model, 1000m. Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Angle -0.722 0.255 2.826 0.005 -1.222 -0.221 Settlement 0.333 0.323 1.031 0.303 -0.301 0.967 Day -0.279 0.285 0.979 0.327 -0.838 0.280 Dry season 0.115 0.282 0.406 0.685 -0.439 0.668 EVI -0.367 0.335 1.097 0.273 -1.024 0.289 In omuramba 0.260 0.265 0.981 0.327 -0.259 0.779 Speed 1.783 0.294 6.057 <0.001 1.206 2.360  Table A.144 Tar road crossing model, 5000m Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.50%) Angle -0.412 0.199 2.064 0.039 -0.802 -0.021 Settlement 1.128 0.241 4.676 <0.001 0.655 1.600 Day -0.466 0.201 2.313 0.021 -0.861 -0.071 Dry season 0.508 0.263 1.93 0.054 -0.008 1.025 EVI 0.026 0.203 0.129 0.898 -0.372 0.424 In omuramba -0.060 0.162 0.369 0.712 -0.377 0.258 Speed 2.321 0.225 10.314 <0.001 1.880 2.762    120  Table. A.145 Buffalo Core (Tar road) crossing model, 1000m Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Settlement -2.243 1.224 1.832 0.067 -4.643 0.157 EVI -2.037 1.461 1.395 0.163 -4.899 0.826 Angle -0.564 0.954 0.591 0.555 -2.435 1.307 Speed 0.516 0.956 0.539 0.590 -1.359 2.390 In omuramba 0.553 1.172 0.472 0.637 -1.744 2.850 Dry season 0.161 0.713 0.226 0.821 -1.236 1.558 Day -0.087 0.745 0.117 0.907 -1.548 1.374  Table A.146 Buffalo Core (Tar road) crossing model, 2000m Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Settlement -2.518 0.648 3.888 <0.001 -3.787 -1.248 Dry season 1.523 1.371 1.112 0.266 -1.163 4.2096 EVI -1.686 1.411 1.195 0.232 -4.452 1.0788 Speed 0.727 0.997 0.729 0.466 -1.227 2.6809 Angle -0.324 0.642 0.505 0.614 -1.582 0.9341 Day -0.164 0.582 0.282 0.778 -1.306 0.9771 In omuramba 0.264 0.893 0.296 0.767 -1.487 2.0154  Table A.147 Buffalo Core (Tar road) crossing model, 5000m Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Angle -0.880 0.753 1.169 0.243 -2.355 0.596 Settlement -2.185 0.450 4.856 <0.001 -3.067 -1.303 Dry season 1.784 1.167 1.529 0.126 -0.503 4.071 EVI -1.133 1.250 0.907 0.365 -3.584 1.317 Speed 1.514 0.967 1.565 0.118 -0.382 3.410 Day -0.295 0.578 0.511 0.609 -1.428 0.838 In omuramba -0.088 0.485 0.182 0.856 -1.040 0.863   121  Table A.148 Mudumu (Ring road) crossing model, 1000m Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Settlement 0.882 0.473 1.863 0.062 -0.046 1.810 Day 0.030 0.253 0.120 0.904 -0.466 0.527 EVI -0.320 0.472 0.679 0.497 -1.245 0.604 Speed 1.823 0.333 5.466 <0.001 1.169 2.476 Angle 0.020 0.257 0.079 0.937 -0.483 0.523 Dry season -0.013 0.408 0.033 0.974 -0.813 0.786  Table A.149 Mudumu (Ring road) crossing model, 2000m Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Angle 0.067 0.213 0.316 0.7521 -0.350 0.484 Settlement 1.116 0.338 3.302 0.001 0.453 1.778 Day -0.061 0.210 0.293 0.770 -0.473 0.350 Dry season -0.221 0.405 0.545 0.585 -1.014 0.573 EVI -0.330 0.425 0.776 0.438 -1.163 0.503 Speed 1.947 0.212 9.179 <0.001 1.531 2.362  Table A.150 Mudumu (Ring road) crossing model, 5000m Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Angle 0.0152 0.169 0.09 0.928 -0.317 0.347 Settlement 0.176 0.239 0.739 0.46 -0.291 0.644 Day -0.060 0.178 0.335 0.738 -0.410 0.290 Dry season -0.662 0.403 1.642 0.100 -1.451 0.128 EVI -0.448 0.393 1.14 0.254 -1.218 0.322 Speed 2.063 0.136 15.167 <0.001 1.797 2.330    122  Table A.151 Susuwe (Tar road) crossing model, 1000m Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Angle -0.685 0.274 2.500 0.012 -1.222 -0.148 Settlement 0.071 0.233 0.304 0.761 -0.386 0.528 Day -0.212 0.273 0.779 0.436 -0.747 0.322 Dry season 0.148 0.314 0.471 0.638 -0.467 0.763 EVI -0.399 0.355 1.125 0.261 -1.095 0.296 In omuramba 0.421 0.294 1.432 0.152 -0.155 0.996 Speed 1.901 0.317 5.993 <0.001 1.279 2.523  Table A.152 Susuwe (Tar road) crossing model, 2000m Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Angle -0.505 0.241 2.098 0.036 -0.977 -0.033 Settlement 0.361 0.298 1.212 0.225 -0.223 0.946 Day -0.285 0.240 1.187 0.235 -0.757 0.186 Dry season 0.0651 0.261 0.249 0.803 -0.446 0.577 EVI -0.335 0.306 1.095 0.274 -0.934 0.264 In omuramba 0.503 0.245 2.056 0.040 0.024 0.983 Speed 2.248 0.286 7.852 <0.001 1.687 2.809  Table A.153 Susuwe (tar road) crossing model, 5000m Variable Estimate Std. Error z value Pr(>|z|) Lower Confidence Interval (2.5%) Upper Confidence Interval (97.5%) Angle -0.294 0.218 1.351 0.177 -0.721 0.132 Settlement 0.7662 0.246 3.114 0.002 0.284 1.248 Day -0.463 0.216 2.149 0.032 -0.886 -0.041 Dry season 0.242 0.276 0.876 0.381 -0.299 0.783 EVI -0.166 0.252 0.657 0.511 -0.661 0.329 In omuramba 0.126 0.193 0.653 0.5135 -0.253 0.506 Speed 2.662 0.265 10.059 <0.001 2.143 3.181  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0166309/manifest

Comment

Related Items