This item is only available as the following downloads:
1 THE EFFICACY OF COMMUNITY BASED MONITORING IN NAMIBIA, THE EVENT BOOK SYSTEM By LUKE VICTOR ROSTANT A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2011
2 2011 Luke Victor Rostant
3 To my family, whose constant support keeps me going
4 ACKNOWLEDGMENTS This work would not have been possible without the support of numerous people and institutions. I would like to thank my advisors Dr. Michael Binford and Dr. Brian Child who kept pushing me and had faith in my abilities. I am also grateful for the invaluable advice provided by my other committee members Dr. Grenville Barnes, and D r. Todd Palmer. I am also indebted to my peers in the lab, especially Forrest Stevens and Andrea Gaughan who lent of their time to help me through many of the analyses contained herein. In Namibia I would like to thank the Ministry of Environment and Touri sm (MET), World Wildlife Fund for Nature (WWF) Integrated Rural Development and Nature Conservation ( IRDNC ) Namibia Nature Foundation ( NNF ) Polytechnic and University of Namibia (UNAM) for their help in making this research possible. I would especially like to thank the communal conservancy committees of Kwandu, Mayuni and Mashi for being patient with my questions, and for le n ding their support in the field. Busihu Bennety of IRDNC was invaluable, going above and beyond to help us with our research. Inde ed, without his help it is doubtful this research could have gone as smoothly as it did. In Namibia Gregory Stuart Hill, Richard Diggle, David Ward, Raymond Peters, Karen Nott, Chris Weaver, Bevan Munali Alfons Mosimane and Robin Beatty were always giving of their time to lend advise and often bent over backwards to help me with logistics. Ibo Zimmermann of the Polytechnic was instrumental in advising and helping me with logistics, as were the people of the National Botanical Research Institute (NBRI) At the MET, Toivo Uahengo worked hard behind the scenes to establish and renew our research permit every year. MET also accommodated us by providing logistical support in the field. The people of
5 Namushasha Lodge in the Caprivi were extremely kind and support ive of this research, as were the Chameleon Backpackers in Windhoek. I have received much of my support from people at the University of Florida, both in Gainesville and in the field. I thank all of the professors who gave of their time to meet with me to discuss my research, and all of the administrators who walked me through the many administrative requirements to help me fulfill this degree. This research would not have been possible without my many sources of funding Fulbright, the School of Natural Resources and Environment (SNRE), Danish International Development Agency (DANIDA), Graduate Student Council (GSC), Doris and Earl and Verna Lowe Scholarship, The Department of Geography, Institute of Food and Agricultural Studies (IFAS), Department of Bio logy, and African Studies. I would like to thank my friends in Gainesville, in Trinidad, and those I made in Namibia for helping me through this process, and putting up with my complaints. I would especially like to thank Katy Garland, who believed I could finish when I had lost faith and whose companionship was invaluable to me through this process Last but not least, this research could not have been finished without the undying support of my parents Frances R ostant and Geoffrey Rostant, my brothers Jer ome Rostant and Wayne Rostant and my sister in law Sarah Jane Rostant I thank them all for their love.
6 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 9 LIST OF FIGURES ................................ ................................ ................................ ........ 11 LIST OF ABBREVIATIONS ................................ ................................ ........................... 13 ABSTRACT ................................ ................................ ................................ ................... 15 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 17 Statement of the Problem ................................ ................................ ....................... 17 Research Questions ................................ ................................ ............................... 20 The Evolution of Community Based Natural Resource Management in Namibia ... 21 The Evolution of Community Based Monitor ing in Namibia ................................ .... 26 Anthropogenic Effects on Savanna Ecosystems ................................ .................... 29 Overview of Dissertation ................................ ................................ ......................... 31 2 AN EVALUATION OF THE EVENT BOOK SYSTEM IN THE CAPRIVI, NAMIBIA ................................ ................................ ................................ ................. 34 Research Objectives ................................ ................................ ............................... 37 The Even t Book System ................................ ................................ ......................... 38 Methods ................................ ................................ ................................ .................. 41 Study Site ................................ ................................ ................................ ......... 41 Data Collection ................................ ................................ ................................ 42 Event Book System interviews ................................ ................................ ... 42 Event Book System data ................................ ................................ ............ 42 Governanc e interviews ................................ ................................ .............. 43 Data Analyses ................................ ................................ ................................ .. 43 Perceptions of problem animals ................................ ................................ 43 Uses and sustainability of the Event Book System ................................ .... 44 Results ................................ ................................ ................................ .................... 44 Discussion ................................ ................................ ................................ .............. 47 Summary ................................ ................................ ................................ ................ 55 3 HUMAN WILDLIFE CONFLICT IN THE CAPRIVI, NAMIBIA: WHAT CAN WE LEARN FROM COMMUNITY BASED MONITORING? ................................ .......... 66 The Event Book System ................................ ................................ ......................... 70 Methods ................................ ................................ ................................ .................. 72
7 Study Site ................................ ................................ ................................ ......... 72 Data Colle ction ................................ ................................ ................................ 74 Elephant population data ................................ ................................ ........... 74 Problem animal incident data ................................ ................................ ..... 74 Temporal datasets ................................ ................................ ..................... 75 Spatial datasets ................................ ................................ ......................... 76 Data Analysis ................................ ................................ ................................ ... 77 E lephant population trends and comparison to past studies ...................... 77 Temporal patterns of crop and livestock raiding ................................ ......... 77 Spatial patterns of crop and livestock raiding ................................ ............. 78 Results ................................ ................................ ................................ .................... 79 Elephant Population Trends and Trends in Crop Raiding ................................ 79 Type of Conflict and Comparison to Past Survey ................................ ............. 79 Temporal Trends in Elephant Crop Raiding ................................ ..................... 80 Effect of moon phase ................................ ................................ ................. 80 Plant productivity and crop raiding ................................ ............................. 80 Temporal Trends in Livestock Raiding ................................ ............................. 81 Effect of moon phase ................................ ................................ ................. 81 Plant productivity and livestock raiding ................................ ...................... 81 Spatial Patterns o f Crop and Livestock Raiding ................................ ............... 81 Discussion ................................ ................................ ................................ .............. 82 Summary ................................ ................................ ................................ ................ 91 4 EFFEC TS OF FIRE FREQUENCY AND CATTLE DENSITY ON VEGETATION IN THE KWANDU, MAYUNI AND MASHI CONSERVANCIES, NAMIBIA ............ 115 Methods ................................ ................................ ................................ ................ 119 Stu dy Site ................................ ................................ ................................ ....... 119 Sampling Design ................................ ................................ ............................ 119 Data Collection ................................ ................................ ............................... 120 Data Analysi s ................................ ................................ ................................ 122 Results ................................ ................................ ................................ .................. 125 Herbaceous Guild ................................ ................................ ........................... 125 Shrub Guild ................................ ................................ ................................ .... 126 Tree Guild ................................ ................................ ................................ ....... 126 Multivariate Analyses ................................ ................................ ..................... 127 Discussion ................................ ................................ ................................ ............ 128 Summary ................................ ................................ ................................ .............. 135 5 CONCLUSION ................................ ................................ ................................ ...... 148 Overall Findings ................................ ................................ ................................ .... 148 Event Book System Shortcomings ................................ ................................ ........ 153 Event Book System Recommendations ................................ ................................ 156 Significance of the Re search ................................ ................................ ................ 158 APPENDIX : EVENT BOOK SYSTEM SEMI STRUCTURED QUESTIONNAIRE ....... 160
8 LIST OF REFERENCES ................................ ................................ ............................. 168 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 179
9 LIST OF TABLES Table page 2 1 Breakdown of the respondents for the Event Book System intervi ews. .............. 58 2 2 Spearman rank correlation values for the comparison of perceived rankings of problem animal species, and the actual rankings of these species. ............... 60 2 3 Uses of the EBS identified by communal conservancy sorted by user group.. ... 61 2 4 Uses of the EBS identified by NGOs sorted by user group. ............................... 62 2 5 Percentage of each respondent type saying that the pre defined user group used the EBS. ................................ ................................ ................................ ..... 63 2 6 ignificant difference observed between the perception of use by community members ................................ ..... 63 2 7 Data from Collomb et al. 2010 showing measures of horizontal accountability for each conservancy. ................................ ................................ ........................ 63 2 8 of sustainability of the EBS by different respondent types. ................................ 64 2 9 Characteristics of the Event Book System for community resource monitors verses wildlife monitors. ................................ ................................ ..................... 65 3 1 Imperfect model results of the zero inflated negative binomial model of elephant crop raiding in the communal conservancies ................................ ..... 109 3 2 Perfect model results of the zero inflated negative binomial model of elephant crop raiding in the communal conservancies ................................ ..... 109 3 3 Imperfect model results of the zero inflated negative binomial regression model of livestock raiding by hyena, leopard and lion ................................ ...... 110 3 4 Perfect model results of the zero inflated negative binomial regression model of livestock raiding by hyena, leopard and lion ................................ ................. 110 4 1 Descriptive statistics of the environmental vari ables included in the multivariate analyses ................................ ................................ ........................ 138 4 2 Correlations among the environmental variables used in the BIO ENV procedure. ................................ ................................ ................................ ........ 1 38 4 3 Shapiro Wilk test results for normality in the environmental datasets.. ............. 139 4 4 Dominant grass species found in 25 transects ................................ ................. 139
10 4 5 Spearman rank correlations between grazing value, veld condition, cattle density and fire frequency for the herbaceous guild.. ................................ ....... 140 4 6 Top 20 species of shrub counted in the stu dy area.. ................................ ........ 140 4 7 Top 20 species of tree counted in the study area. ................................ ............ 141 4 8 Results of the BIOENV procedures for shrubs, trees a nd environmental variables. ................................ ................................ ................................ .......... 144 4 9 Results of the categorical data analysis of fire frequency and cattle densities with NMDS ordinations of shrub density and tree biomass. ............................. 147
11 LIST OF FIGURES Figure page 2 1 Map of the study region showing the location of the focal communal conservancies, ................................ ................................ ................................ .... 57 2 2 Organogram of the different types of respondent interviewed in the Event Book System dataset.. ................................ ................................ ........................ 59 2 3 EBS ................................ ............. 64 3 1 An example of a problem animal incident yellowbook sheet from Mayuni Conservancy, 2005.. ................................ ................................ ........................... 92 3 2 Flow diagram of the temporal and organisational structure of the problem animal incident module of the EBS. ................................ ................................ .... 93 3 3 Areas under active crops mapped using SPOT 2010 imagery. .......................... 94 3 4 Mean human population density within each grid cell over the communal conservancy region. ................................ ................................ ........................... 95 3 5 Zero inflated nature of human wildlife conflict data.. ................................ ........... 96 3 6 Total number of elephant crop raiding incidents in the three focal conservancies over the course of the survey period.. ................................ ......... 97 3 7 Number of predation incidents on livestock by lion, leopard and hyena.. ........... 98 3 8 Elephant densities over time for the period 1985 to 2005. ................................ .. 99 3 9 Differences observed in the distribution of human wildlife conflict in the study area. ................................ ................................ ................................ ................. 100 3 10 Number of elephant crop raiding i ncidents during the lunar cycle in the communal conservancies. ................................ ................................ ................ 101 3 11 Trend in mean NDVI values and elephant crop raiding incidents in the study region. ................................ ................................ ................................ .............. 102 3 12 Lagged correlation correlogram showing a peak in correlation at about 48 days ................................ ................................ ................................ .................. 103 3 13 Prewhitened data showing no strong effect of NDVI magnitude on subseq uent levels of elephant crop raiding. ................................ ..................... 104
12 3 14 Number of livestock raiding incidents during the lunar cycle in the communal conservancies. ................................ ................................ ................................ .. 105 3 15 Trend in mean NDVI and livestock raiding over the 4 year period. ................... 106 3 16 Lagged correlation correlogram showing no significant relationship between mean NDVI and livestock raiding. ................................ ................................ ..... 107 3 17 Prewhitened data showing the effect of NDVI magnitude on subsequent levels of livestock raiding.. ................................ ................................ ................ 108 3 18 Ma p of the modeling results of elephant crop raiding. ................................ ...... 111 3 19 Results of the zero inflated negative binomial model for elephant crop raiding showing the residual values for the model. ................................ ....................... 112 3 20 Map of the modeling results of livestock raiding.. ................................ ............. 113 3 21 Results of the zero inflated negative binomial model for livestock r aiding by hyena, leopard and lion showing the residual values for the model. ................. 114 4 1 Location of transect sites in the study region in relation to the communal conservancies and the fire frequ ency from 1989 to 2005. ................................ 136 4 2 Location of transect sites in the study region in relation to the communal conservancies and the cattle density.. ................................ .............................. 137 4 3 Cumulative count of the number of species of shrub observed in 29 transects 142 4 4 Mean abundance of each shrub species verses the number of transects in which the s pecies was found. ................................ ................................ ........... 142 4 5 Cumulative count of the number of species of tree observed in 29 transects ... 143 4 6 Mean abundance of t ree species verses the number of transects in which the species was found. ................................ ................................ ........................... 143 4 7 Superimposition of categorical variables onto NMDS plots of shrub community density. ................................ ................................ ........................... 145 4 8 Superimposition of categorical variables onto NMDS plots of tree community biomass. ................................ ................................ ................................ ........... 146
13 LIST OF ABBREVIATION S AfESG African Elephant Specialist Group AGM Annual general meeti ng CAMPFIRE CBNRM Community based natural resource management CEC Cation Exchange Capacity CITES Convention on the International Trade in Endangered Species of wild flora and fauna CRMs Community resource monitors dbh diameter at breast height EBS Event Book System ETM+ Enhanced Thematic Mapper Plus GIS Geographical information system HACSIS Human Animal Conflict Self Insurance Scheme HEC Human elephant conflict HECWG Human Elephant Conflict Workin g Group HWC Human wildlife conflict IRDNC Integrated Rural Development and Nature Conservation IUCN International Union for the Conservation of Nature KAZA Kavango Zambezi Transfrontier Conservation Area MET Ministry of Environment and Tourism MNC Mudumu N orth Complex MODIS Moderate Resolution Imaging Spectroradiometer MOMS Management Oriented Monitoring System NACSO Namibian Asso ciation of CBNRM Support Organis ations NCA Northern Communal Areas
14 NDVI Normalized difference vegetation index NGOs Non governmen tal organizations NMDS Non metric multidimensional scaling RDC Rural District C ouncils ROMS Research Oriented Management System SADF South African Defence Force SSC Species Survival Commission TM Thematic Mapper WMs Wildlife monitors WWF World Wildlife Fun d WWF LIFE World Wildlife Fund Living in a Finite Environment ZINB Zero inflated negative binomial ZIP Zero inflated Poisson
15 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy THE EFFICACY OF COMMUNITY BASED MONITORING IN NAMIBIA, THE EVENT BOOK SYSTEM By Luke Victor Rostant August 2011 Chair: Michael W. Binford Cochair: Brian Child Major: Interdisciplinary Ecology Co mmunity based natural resource management strategies have increased the world over, and with this, the design and implementation of community based monitoring strategies have also increased. Research of these forms of monitoring is important in recognizing whether or not comm unities are capable of monitoring and managing their own natural resources. While there has been the proliferation of these approaches to monitoring, most research has focused on data quality issues, comparing them to concurrent more pr ofessional monitoring systems, with few exa mining the way that the data is incorporated into management. Thi s dissertation seeks to fill this knowledge gap by examining a community based monitoring system in Namibia called the Event Book System in this con text. I use interview, geospatial and temporal analyses and field data to examine how natural resource data is being institutionalized, what the uses and decisions of the monitoring system are, how this information is disseminated, what conditions within t he monitoring system make it more amenable to adaptive management, and determine whether or not the data can identify long term spatial and temporal trends. I also fill the knowledge gap of vegetation community
16 composition in the study region by examining species distribution of the herbaceous, shrub and tree guilds, and determine the effect of fire frequency and cattle density on these communities. The research was conducted in three communal conservancies in the Caprivi of Namibia. Semi structured intervi ews we re conducted with th ose collecting and managing the monitoring system, as well as those who help implement the system. Collection, conversion, and analysis of human wildlife conflict data was carried out to develop a model of both temporal and spatia l determinants of elephant crop raiding and livestock raiding. Results suggest that there is little horizontal accountability within the monitoring system, while much of its use comes from compliance reporting to government, non governmental organizations, and donors. There were few ex ample s of conservancies using the information in an adaptive management context, and best practices for adaptive management of natural resources were identified. Temporally, it was found that there was a relationship between e lephant crop raiding and plant productivity derived from satellite imagery, and that crop and livestock raiding were dependent upon the moon phase. Spatially, the models were able to identify variables which contributed to the patterns of both crop and liv estock raiding. Vegetation community analysis recognized lower grazing value with increasing cattle density and lower fire frequency, found the dominant shrub and tree species to have bush encroaching potential, and that patterns of shrub s and tree s were d ependent on cattle density and fire frequency. The research recommends incentivizing horizontal accountability, and building on those aspects of the monitoring system which contribute towards adaptive capacity. The research also shows that this community b ased monitoring data can be used t o model natural resource trends
17 CHAPTER 1 INTRODUCTION Statement of the Problem A gradual shift away from the strictly protectionist approach to conservation (the fortress conservation model) towards community conservat ion has occurred partly as a response to the recognition that much of the biodiversity throughout the world is not restricted to protected areas and there is thus the need to conserve regions external to them (Naughton Treves et al. 2005). At the same time the social injustices of past removal of people for protected area formation have led to resentment on the part of some who reside on the outskirts of these protected areas (Guha 1997). The social and economic costs of strictly protectionist protected ar ea establishment include population displacement of local peoples (Rangarajan and Shahabuddin 2006), high economic costs of maintaining these forms of protected areas (Leader Williams and Albon 1988), the opportunity costs of this form of land use as oppos ed to other forms of human land use (Norton Griffiths and Southey 1995) and the hazards of neighboring communities surrounding parks to wild animals through crop raiding (Sitati et al. 2005). Though tourism based economies are often cited as benefits to pr otected area establishment, the costs of such activities may be ignored (Adams 2004) and the benefits of protected areas economies are often not distributed equitably (Paudel 2006; Collomb et al. 2010). It has thus been recognized that the establishment of strictly protected areas often leads to benefits which are derived globally, often by remote and relatively wealthy people, while the costs of their establishment are borne locally (Balmford and Whitten 2003). The recognition of the social and economic im pact of strictly protected areas began to be widely recognized
18 social aspects of protected area establishment (Adams and Hulme 2001 a ). While highlighting the problems associated with the fo rtress conservation model I recognize that this form of protected area is perhaps the most effective in cases where the conservation of biological resources is of paramount concern, especially where the protected area contains critically endangered specie s (Bruner et al. 2001, Adams and Hulme 2001b). People argued that since most of the costs of protected area establishment were borne by local peoples, then conservation goals had to be realigned to provide for them if they were to be successful (Child 200 4). The participation of communities in natural resource management and conservation is cited as a critical element for success since this facilitates involvement of local residents who are important for achieving program goals (Robinson and Bodmer 1999; D anielsen et al. 2005 a ). There are several tenets upon which the community conservation narrative has been built (Adams and Hulme 2001 b ), including the participatory inclusion of local peoples as conservation partners, contribution to sustainable livelihood s, and that there should be some form of economic return for local people and the wider economy. The shift in focus from conservation being strictly based on ecosystem and biodiversity preservation to economic and social well being has been criticized by s ome as detrimental to conservation goals since they argue these approaches waste scarce conservation resources which would be better spent in strictly protected areas (Bruner et al. 2001) which are well resourced and properly managed (Balmford et al. 2002) Many of those who support the traditional establishment of parks with social exclusion cite research which has been conducted mostly by social scientists who demonstrate the shortcomings of community
19 conservation approaches (Adams and Hulme 2001 a ). While recognizing the problems with community conservation, advocates of this approach to conservation insist that many of the failures experienced thus far have more to do with poor implementation and understanding than a pervasive issue with the concept itsel f (Child and Barnes 2010). The growth of participatory management approaches to conservation required that monitoring systems be designed and implemented to give communities the information they needed to successfully sustainably manage their resources (G arcia and Lescuyer 2008). There is growing debate about the usefulness of the monitoring data being supplied by community based monitoring, some arguing that communities lack the capacity, sustained interest and resources needed for effective monitoring (G arcia and Lescuyer 2008) Conversely, monitoring systems implemented by outside organizations may be successful in the short term, but in the long term they are not sustained, mainly because the monitoring is extractive not being returned to the communiti es in a form useful to them (Danielsen et al. 2005 a ). Locally based monitoring programs seek to promote better management of natural resources in a variety of ways (Danielsen et al. 2005 a; Fraser et al. 2006 ; Garcia and Lescuy er 2008) including enhancing c ommunication amongst local people which can lead to the development of more sustainable practices, empowering local communities so that they are better able to communicate with other stakeholders, and reducing the transaction costs for decision making whic h enables a faster response to stochastic events. There is growing concern that many of these community based monitoring systems are falling short of their desired outcomes (Garcia and Lescuyer 2008), and, coupled with this, there is interest in testing wh ether or not these monitoring approaches can detect true local or larger
20 scale natural resource trends which address the shortfalls of conventional (professional) monitoring. Research into the mechanisms of community based monitoring programs is also vital to the adaptability of communities to change, since adaptive management of these systems is dependent upon the monitoring systems which inform them. Research Questions The overarching research question in this dissertation is how well does the community based monitoring system in Namibia contribute towards management in the communities of Namibia? While many different types of community based monitoring systems have been adopted (Danielsen et al. 2009), most of the se analyses focus upon their comparison t o concurrent more professional monitoring systems in terms of the quality of the information they produce, shying away from fundamental evaluations of these systems in terms of their ability to aid the management of natural resources by the communities con du cting them. In this research, I focus on the latter aspect of community based monitoring b y examining the way that data is incorporated into management, the way that natural resource information is disseminated, and the conditions which foster a greater adaptive response in t he face of stochastic events. I then use the community based monitoring data to examine the potent ial of this data to model human wildlife conflict. Finally, because the monitoring system did not focus on vegetation communities; becau se the communities themselves expressed a need for vegetation to be monitored; and because an understanding of vegetation responses to anthropogenic effects is important for conservancy management, I examined the vegetation communities in three communal co nservancies to identify the underlying patterns of these communities and determine to what degree they may be affected by anthropogenic effects. In so doing, I investigate an important aspect of the natural
21 resources in these communal conservancies which a re being largely ignored by the community based monitoring system. I examine a community based natural resource monitoring system in Namibia known as the Event Book System, answering the following research questions: Are those conducting the monitoring s ystem familiar with the patterns the data shows? Who gets the information being collected by the monitoring system? Who uses the information being collected by the monitoring system? What uses and decisions are being made based upon the data being collect ed by the monitoring system? What conditions within the monitoring system make it more (or less) amenable to adaptive management? Is there potential in understanding the spatial and temporal patterns of human wildlife conflict based upon the com munity base d monitoring system? I investigate the vegetation communities in three communal conservancies in Namibia by answering the following research question: How do fire frequency, cattle density, and edaphic factors interact to affect the vegetation community st ructure and function in several communal conservancies ? The E volution of Community Based Natural Resource Management in Namibia There are m any characteristics found in southern Africa which lend themselves to some form of community based natural resource m anagement (CBNRM) approach to conservation (Adams and Hulme 2001 b ; Brown 2004; Child 2004). For the most part, the southern African approach to conservation seeks to manage natural resources in ng the inherent biodiversity found in the region to satisfy social and economic needs. For community based conservation to succeed, Adams and Hulme (2001 b ) outline several conditions
22 which should be met. Much of southern Africa possesses charismatic mammal s which is one of its competitive advantages when compared with other regions of the world. The indigenous biodiversity in southern Africa can be used to support wildlife tourism, safari hunting, and other ventures. If managed properly, these wildlife reso urces should yield a sustainable revenue flow. Brown (2004) and Child (2004) also suggest that the semi arid and dry sub humid regions of southern Africa lend themselves to the sustainable use of natural resources as a more viable land use alternative. Th is is in large part attributed to the unreliable nature of agriculture in regions where rainfall is lower than 600 mm (Brown 2004). Significantly greater returns can be achieved using the megafauna and other indigenous biodiversi ty found in southern Africa as a result of the unreliable returns attributed to cultivation as a land use option. Brown (2004) suggests that this is the situation throughout Namibia, where indigenous biodiversity generates potentially more economic returns than agriculture, and wher e rainfall varies from less than 25 mm in the Namibian desert to the west, and over 600 mm in the east (Jones and Murphree 2001). It would therefore seem prudent to manage for the natural habitat in Namibia, so that indigenous species can be sustainably ut ilized If land management is thought to be better suited to indigenous biodiversity under the aforementioned conditions, why are many communities still focusing on agriculture? Throughout southern Africa, pre independence conservation policy and practice based on centralized and racial lines led to the alienation of most rural people from their wildlife (Taylor 2001). Since they were not allowed to use wildlife resources, alternative land uses were pursued which conflicted with native species and often le d to land
23 degradation. Attempts to return certain access rights to natural resources through changes in existing laws, devolution of power, and economic empowerment have occurred recently in parts of southern Africa (Taylor 2001). In Zimbabwe, this came in was their g reatest source of revenue. This was seen as a major weakness of the CAMPFIRE program (Jones and Murphree 2001), notwithstanding the turmoil observed in contemporary Zimbabwe, since little benefit would reach down to the community level. In Namibia, the com munal conservancy system was adopted. Namibia has a long history of inequitable distribution of land based on its colonial history. G erman settlers from 1888 to 1917 appropriated much of the more fertile central regions of the country (called South West A frica) removing the black tribal groups to reserves in less fertile regions (Jones and Murphree 2001). The German colonists cut the rest of the country in social, economic, and commercial terms (Gottfried 2001). During World War I, South Africa joined the British cause and began occupying South West Africa in 1915. In 1920 the League of Nations placed South West Africa under Brit ish control, with its administration controlled by South Africa. The So uth African administration (1915 to 1990) continued the process of consolidating the reserves into homelands based on the apartheid policy (Jones and Murphree 2001), relegating black tr ibal groups to land which was predominantly unsuitable for crop growing and livestock rearing. The legacy of the colonial period
24 concerning the ownership of wildlife continued throughout this period, which allowed hunting by white settlers and visiting spo rtsmen, while removing rights to access wildli fe from tribal groups. White commercial farmers often saw wildlife as competing for livestock resources, and considered them pests that interfered with crop farming. T he conversion of communal and freehold wil d habitat ensued through agriculture, and the illegal use of wildlife through poaching, led to rapid declines in wildlife numbers (Jones and Murphree 2001). This changed in 1968, when white commercial farmers were given proprietorship over wildlife on thei r farms, allowing them to benefit from hunting for own use, trade in game species, cull for sale of meat and trophy hunting by foreigners. White farmers now had a vested interest in sustainably managing wildlife and today much of the wildlife is found on commercial farms, and a multi million dollar industry based on wildlife on commercial farms has developed. Laws allowing for the rights to wildlife resources to communities in communal regions were not enacted until after Namibian independence in 1990 so that wildlife continued to decline in most of these regions (Jones 1999; Jones and Murphree 2001). Community conservation in Namibia developed in the Kunene region of the northwest out of concern by conservationists and local leaders about a major decline in wildlife numbers caused by severe drought and heavy poaching in the 1970s (Jones 1999). In 1982, Chris Eyre, a government extension officer, and Garth Owen Smith, a former agricultural extension officer and government ranger, worked with the traditional authorities in the region to establish a community game guard system. This devolved some power over natural resources back to the communities. The number of game animals increased significantly over the next ten years with an end to the drought and
25 the a n ti poaching activities of the game guards (Durbin et al. 1997). Through integrated work with the communities of north western Namibia, the value for wildlife increased the communities deriving benefits through the growing tourism industry. Owen Smith work ed with Margaret Jacobson, an anthropologist studying the Himba semi nomadic pastoralists of Northern Kunene, to ensure that the negative effects of tourism did not develop in the region unchecked by the traditional authorities. Together, Owen Smith and Ja cobson formed an NGO called Integrated Rural Development and Nature Conservation (IRDNC) and spread the community game guard system to the north eastern Caprivi region. Post and To urism (later changed to the Ministry of Environment and Tourism, MET) collaborated with the IRDNC to develop new policies to help improve conservation on communal lands (Jones 1999). The government wished to replicate the success exhibited on private farms to the communal lands which made up a substantial portion of the country. Their primary goals were to enable rural communities to benefit from wildlife, and also gain rights over tourism concessions in these regions (MET 1995). This policy proposed that c onditional stewardship over some game species, the use of other species through a permit system, and the right to trade in game species be permitted if communal areas formed a common property resource ma nagement To form a conservancy, communities had to clearly define their boundaries, have defined membership, form a committee representative of the membership, establish a legal constitution recognized by government, and plan for the equitable distribution of benefits to m embers (MET 1995). The policy was put into
26 effect under the Nature Conservation Amendment Act of 1996, allowing the MET to declare communal area conservancies once they had satisfied the conditions cont ained within the act (NACSO 2007 ). By the end of 2009, 59 communal conservancies had and contained over 230 000 people within them (NACSO 2010). The rapid increase in the number of emerging conservancies has also coincided with increasing wildlife populations throughout Namibia (Jones and Weaver 2009). While the re have been increases in wildlife, this has also resulted in increasing levels of human wildlife conflict which, if left unresolved, may undermine these conservation effo rts (NACSO 2010). The rapid growth in the number of conservancies have also stretched the resources of facilitating NGOs and supporting government officials (NACSO 2010), and with conservancies gradually generating their own income, there have been problem s associated with misappropriation of money and benefits (Child and Barnes 2010), with some conservancy committees making all major decisions themselves without the inclusion of members (NACSO 2010); a practice of elite capture which successful CBNRM strat egies are meant to overcome. The Evolution of Community Based M onitoring in Namibia In order to maintain conservancy status, the communities had to show they were managing their resources sustainably, especially with regards to wildlife. The aims of the c onservancy movement were to empower local communities, and to enable them to adaptively manage their own natural resou rces. Thus, the government, NGO s and rural communities assisted in developing a CBNRM monitoring system in Namibia. The management of natu ral resources within communal conservancies is complicated by the wide range of variables involved. When one considers the huge land area under
27 conservancy status, the Namibian government simply does not have the resources to monitor all the natural resour ces (Stuart Hill et al. 2005). If monitoring o f the communal conservancies were to be implemented successfully, then it was decided it should be the responsibility of the communities to monitor their natural resources. The first attempt at community monito ring involved the employment of community game guards to collect data on the natural resource trends occurring within their conservancies. These forms were then handed back to the NGO, who was then supposed to analyze the data and feed the results back to the communities. The analysis of this information, when carried out, was often not fed back to the community, was complicated, and provided results (Stuart Hill et al. 2005; Danielsen et al. 2010 a ). This first a ttempt at monitoring may be described as functional in nature, where local labor and knowledge were used to collect information (Lawrence 2006) but ultimately the decisions were still made by facilitating NGOs and government. Gregory Stuart Hill and Bevan Munali conducted a strengths, weaknesses, opportunities and threats exercise of the existing monitoring system, the results of which were presented to Richard Diggle (IRDNC). Based on the limitations of this system, at a meeting at the IRDNC office in Kong ola, in the Caprivi region of Namibia, Gregory Stuart Hill, Bevan Munali and senior conservancy game guards from five surrounding conservancies discussed how to truly devolve monitoring to the conservancies, monitoring which would be community based and ow ned by the community. Based upon this meeting, the Event Book System (hereafter referred to as the EBS) was developed. In the EBS, the communities decide what to monitor, technicians only providing assistance when requested, and all data collection and ana lysis is undertaken
28 locally by conservancy members. Stuart Hill et al. (2005) regard the EBS as a management oriented monitoring system (MOMS), where priorities have been aligned more towards adaptive management through acquiring bits of information on man y management objec tives, as opposed to a research oriented monitoring system (ROMS), where the focus may be on publication, and where scientists get detailed information on fewer natural resources. The latter is often characterized by regular modification and increasing complication, which inevitably leads to the system breaking down when the scientists who developed it leave. They recommend that these two systems can comple ment one another if the ROMS focus can be realigned to investigate reasons for success or failure of MOMS management initiatives. Stuart Hill et al. (2005) have cited examples of the EBS being used by government, NGOs and donor agencies, in the form of co mpliance reporting, helping them to focus technical support in different conservancies, assisting in the design of a human wildlife conflict compensation scheme, setting regional harvest quotas for of their ivory stockpile at the Convention on the International Trade in Endangered Species of wild flora and fauna (CITES) 2002 Cong ress of the Parties meeting. This research focus es on what uses and decisions were being made by the conservancies themselv es since these were the ones who the monitoring system was p rimarily designed to benefit. I will also investigate whether those conducting the monitoring and analysis at the community level are familiar with the uses described at higher institution al and o rganizational levels. I conducted this research in the Kwandu, Mayuni and Mashi Conservancies of the
29 eastern Caprivi since this region is where the EBS was initiated, and so should have had the most experience in terms of incorporating the data into manage ment action. Anthropogenic Effects on Savanna Ecosystems Understanding how the vegetation communities in the communal conservancies of Caprivi Namibia are responding to anthropogenic effects is important for a number of reasons. At the local level, commu nities depend upon vegetation for building materials, craft resources, medicines, food, firewood and, when sold commercially, as a source of income (Mendelsohn and Roberts 1997) People also need reliable forage for their cattle, and it is important to mea sure range condition to determine how many cattle can be sustainably supported on the landscape. The contribution of wildlife to the economy in the Caprivi region has grown since conservancy system inception (NACSO 2010) and these wildlife resources are al so dependent upon vegetation resources in conservancies. In the regional context, the Caprivi region also forms part of the Kavango Zambezi Transfrontier Conservation Area (KAZA) and conservancies in the region form a vital connection between different ty pes of co nservation areas. Altering t he vegetation communities in these communal conservancies could therefore affect the connectivity of the KAZA region, and may undermine conservation efforts. Because the majority of the benefits delivered to the communi ty come from wildlife (NACSO 2010), there may be less emphasis on monitoring vegetation within the Event Book System. Vegetation monitoring is undertaken by community resource monitors, but this is restricted to the measurement of qualitative attributes of those species being used to make crafts. S emi structured interviews conducted with the communal conservancy management committees and community monitors also revealed that they thought they should be focusing more on monitoring vegetation within their
30 con servancies. To this end I measured vegetation communities within the communal conservancies to examine the effect anthropogenic activities were having on vegetation communities. While the distribution and type of savannas in Africa is determined by climat ic and edaphic factors, humans can modify these patterns through local land use practices (Scholes and Archer 1997; Higgins et al 1999 ; van Langevelde et al 2003). Holdo et al. (2009 ) cites soil moisture, fire, and mammalian herbivory as the three main fac tors affecting the biomass of woody and herbaceous vegetation in African savannas and humans can have profound effects on the latter two factors Humans can directly alter the vegetation in savanna habitats through conversion of the l andscape to support c rop fields or settlement and by harvesting wood and other resources for fuel and building materials ( Shackle ton et al 1994). These effects may be more localized than the indirect effects which can spread further, and include the setting of fires ( Dublin e t al 1990; Sheuyange et al 2005), the removal of wi ldlife which in turn affects tree grass ratios ( Leuthold 1994) and cattle rearing ( Moleele and Perkins 1998) All three of these indirect effects can interact on the savanna landscape. The effects of fir e on savanna vegetation depend upon the intensity and frequency of fires (Scholes and Archer 1997). More i ntense fires can lead to a loss of woody biomass ( Smit et al 2010), a nd more frequent fires allow grasses to persist in the system by opening up the c anopy and controlling the spread of shrubs (She u yange et al 2005). Browsing and grazing interact as well to shape tree grass biomass in Africa n savanna ecosystems (Holdo et al. 2009 ). In areas where browsers such as elephants decreased, there has been an i ncrease in woody biomass ( Leuthold 1994) Since woody biomass
31 and grasses compete for soil moisture in savanna ecosystems (Scholes and Archer 1997), overgrazing by cattle may give woody biomass a competitive advantage leading to bush encroachment (Moleele and Perkins 1998). An excellent example of how different disturbances interact to influence savanna vegetation is in the Serengeti Mara region of Kenya. Dublin et al (1990) suggest that t he rinderpest pand emic in East Africa in the 1890 s resulted in the de cimation of wildlife populations, which, together with a decline in anthropogenic fires and reduction in elephant numbers through poaching, resulted in i n creased woody vegetative cover and a conversion of the grassland state Subsequent interaction between increased intensity of fires in the 1960s, and herbivore effects on the recruitment of woody plant species in the 1980s led to the woodland being converted back into a grassland state. The interactions among anthropogenic effects are complex, and chapter 4 of this dissertation investigates how fire frequency and cattle density are interacting to produce the patterns of vegetation within the Kwandu, Mayuni and Mashi conservancies. Overview of Dissertation This d issertation is divided into five chapters whic h complement one another by first examining the community based monitoring program used in three focal communal conservancies in the Caprivi of Namibia in terms of the way it is being used by the conservancies, by examining the potential of using the spati al and temporal data of the monitoring system to model human wildlife conflict, and by examining an aspect of the natural resources in the conservancies which has been largely ignored by the monitoring system (its vegetation communities and the effects of fire frequency and cattle density on these communities)
32 Chapter 2 focuses on an evaluation of the EBS with regards to the way the information is being collected, integrated into uses and decisions, how the information is being disseminated, and the condi tions in the EBS which lead to improved management. I used three datasets in this analysis, the primary dataset coming from semi structured interviews of people directly using the EBS. These included the communal conservancy monitors, the management commit tee of the conservancies, and key informants from government and NGOs who assist in EBS implem entation and analysis. T he actual EBS data itself was used to gain some perspective of how well the communities conducting the monitoring are incorporating it int o knowled ge about their conservancies. I also compare the perceptions of use of the EBS by communal conservancy members given by the respondents in the primary dataset, to another study conducted by Collomb et al. 2010 to determine how well information is being disseminated within these conservancies. The results of these analyses are used in a discussion of the role the EBS plays in knowledge generation for the conservancies, the conditions in the EBS which lend themselves more to an adaptive management re sponse by the community, and the shortcomings of the EBS. Chapter 3 uses the human wildlife monitoring dataset from the EBS to determine whether the monitoring system can detect temporal and spatial patterns of elephant crop raiding and livestock raiding. It is expected that elephant crop raiding will show a distinct seasonality in accordance with the seasonal nature of crop cultivation in the study region, whil e livestock raiding will not. I also investigate the effect of moon phase on both crop an d livest ock raiding activity. I examine spatial determinants of crop and livestock raiding which are reported to affect the patterns of both forms of human wildlife
33 conflict in the scient ific literature. In so doing, I identify which factors affect the patterns of human wildlife conflict in these three communal conservancies, and identify improvements which can be made in the analysis to better understand how to mitigate conflict. Chapter 4 examines the effects of fire frequency and cattle density on the vegetatio n communities observed in the three conservancies. I hypothesize that the combined effects of high cattle density and reduced fire frequency result in a reduction of overall grazing value and veld condition, and that these anthropogenic effects also result in shrub and tree woody biomass in the conservancies which is dominated by species cited in the literature as potential woody encroachers. I measured the vegetation in three vegetation guilds herbaceous, shrub and tree, and determined the dominant speci es in each of these, followed by a discussion of their characteristics. I then performed multivariate analyses on the vegetation community data of shrubs and trees to determine whether or not fire frequency and cattle density a re included in models correla ting the environmental and vegetation community datasets. Chapter 5 integrates the findings from the previous chapters, and synthesizes some of the challenges of the EBS I make recommendations for best practices for the design and implementation of commun ity based monitoring systems, using the EBS as a model for improvement. I also summarize the findings of Chapter 4, making recommendations for the continued monitoring of vegetation communities in these communal conservancies. Finally, I revisit the ways t hat this research is relevant for not only informing researchers and practitioners, but most importantly to the communities whom we are increasingly relying upon to sustainably manage ecosystems.
34 CHAPTER 2 AN EVALUATION OF THE EVENT BOOK SYSTEM IN THE CAP RIVI, NAMIBIA A shift towards the involvement of local communities in monitoring of natural resources has taken place within recent times (Danielsen et al. 2005a). With this shift, it has become increasingly important to develop monitoring data collection methods which are relevant to local people and also facilitate communication with other stakeholders (Garcia and Lescuyer 2008). Several principles have been cited for the successful implementation and sustaining of locally based monitoring systems (Daniel sen et al. 2005a). These include monitoring of beneficial resources within the community, making the benefits of monitoring by locals exceed the costs, preventing politics and conflict between different stakeholders from interfering with the involvement of local participants, trying to build upon traditional institutions, allowing local people to store and analyze the data, and making the data accessible locally. Much of the focus in the literature on community based natural resource monitoring systems has been on their ability to depict true trends in relation to more professional monitoring systems (McLaren and Cadman 1999 (bird calls) Bray and Schramm 2001 (anglers) ; Genet and Sargent 2003 (amphibian calls) ), with few of these studies being concentrated in developed countries (Mumby et al. 1995 (reef fish identification) Darwall and Dulvy 1996 (reef fish identification) ; Hellier et al.1999 (rainforest biodiversity) ; Noss 1999 (rainforest census of game) ). The problem with evaluation of locally based moni toring against more professional monitoring techniques is that it is seldom practiced because it is logistically difficult, and may suggest suspicion of community data by scientists, NGOs, donors and government agencies (Danielsen et al. 2005a). While accu racy and precision are important aspects of community based monitoring systems, in the
35 absence of comparative professional datasets it is perhaps just as important to ensure that there are mechanisms in place in order to transform the data into management action. It makes little sense to have a monitoring system in place if the data it produces comes to no use. It has been suggested that while professional monitoring techniques are more likely to influence decisions high er up in the governance hierarchy (na tional and global policies), decisions based on local monitoring data can be taken faster and be locally relevant (Danielsen et al. 2010b). In a study of 104 environmental monitoring schemes, Danielsen et al. 2010b found that those involving local people w ere more effective regarding resource use at the village level, and decisions were typically implemented in less than a year. The speed that these decisions can be made makes them more amenable to adaptive management processes, and is dependent upon the st rength of the institution established to process the data; and the levels of bureaucracy tied to the institutions making these decisions. It therefore, stands to reason that weak institutional arrangements and high levels of bureaucracy would result in les s adaptable monitoring programs. But which institutional arrangements can be considered best to support locally based monitoring? Danielsen et al. (2005a) suggests that in cases where the institution for monitoring is nested within an existing local instit ution it is more likely to succeed since the decisions made are more respected by local actors. One visio n of such local institutions is referred to as a community based natural resource management (CBNRM) organizatio n In Namibia, for example, the common property institution is resources once they define their boundaries, define their members, have a
36 representational committee, establish a recognized constitution, and have equit able distribution of benefits to their members (MET 1995). Any examination of community based natural resource monitoring should thus also include an analysis of the CBNRM system to identify conditions which facilitate successful monitoring of natural reso urces. In the southern African context, Child and Barnes (2010) outline the strategies for the successful implementation and sustaining of CBNRM programs. In their discussion, they argue that successful CBRNM involves not only the devolution of rights, res ponsibilities and opportunities to ordinary people; but that the way these are transferred is just as important. While some may argue that it is more efficient to have representational governance in the form of a committee based management governance frame wo rk this form of natural resource management runs the risk of elite capture of resources, which can lead to alienation of the community it is purported to support; the very problem many CBNRM programs are trying to overcome (Overdevest 2000; Child and Ba rnes 2010). Participatory democracy, where significant decisions are made through face to face interaction with the whole community is more likely to result in successful CBNRM (Child and Barnes 2010). When people are allowed face to face communication to discuss social dilemmas, researchers have consistently found significant improvement i n levels of cooperation (Ostrom 1998).Village group discussion was found to be the most effective monitoring method for improving cooperation between stakeholders, for pr oviding incentives for participation in regulations governing resources, and in empowering local participants in an examination of a community based monitoring program in the Phillipines (Danielsen et al. 2005 b ). This is because more people are informed ab out what is taking place in their community, making the
37 managers in the community accountable and the system as a whole more democratic in nature. This form of accountability is termed horizontal accountability, where managers must answer to the people the y represent, as opposed to vertical accountability, where managers are accountable to more powerful actors (such as the government) creates a forum for addressing conflicts and allowi ng for the development of shared goals and compromise s in decision making (Rammel et al. 2007). In this paper, I examine a community based monitoring system from Namibia called the Event Book System (EBS). I shift focus away from the data quality investiga tions which have dominated community based natural resource monitoring literature, and instead focus on the products of this system in the form of its uses and management decisions, and the way that the system is institutionalized. I argue that within the system itself a dichotomy has developed between those that monitor fugitive resources (wildlife) within communal conservancies in Namibia (wildlife monitors), and those that monitor craft resources (community resource monitors), and discuss the advantages and disadvantages of each approach. I also compare my dataset to another dataset describing these three communal conservancies to assess horizontal accountability of natural resource information to the conservancy (Collomb et al. 2010). In so doing, I illu strate how differences in the way monitoring systems are designed can in turn affect the way people perceive their utility and sustainability, and make recommendations based upon my observations. Research O bjectives The overarching objective of this resear ch was to determine under which conditions the community based monitoring system, the Event Book System, may lead
38 to improved management. This research sought to understand the monitoring system through an examination of how the data was collected and the degree of knowledge about this data, what decisions were made based upon the data, and whether or not there was participation in the decision making process. Secondarily, I examined which conditions were most favorable for translating the EBS data into man agement action. I focus on the ways the EBS is (i) being incorporated into kno wledge about the conservancy, (ii) used in compl iance monitoring, (iii) de veloping confidence in CBNRM, (iv) empowering local communities, (v) decision making, and in adaptive ma nagement within the conservancies. The Event Book System A comprehensive discussion of the Event Book System may be found in Stuart Hill et al. (2005). Here I focus on the way it is introduced, the way that the data is collected, and the institutional st ructure of the system; items pertinent to my analysis. The EBS is jointly implemented by supporting scientists in NGOs and the communities which community monitoring, bu t a melding of both into the beginnings of an adaptive management system. The process often times begins before a community has been officially granted conservancy status. The community invites the facilitating NGO for the region to help them monitor their wildlife. Quite often the communities hear about the EBS from already established conservancies. The NGO meets with the communities and the participants indicate what they wish to monitor. Over time, NGO s have used a participatory development process to d evelop a number of monitoring modules such as the problem animal incident module, and the craft resources module. Since the community decides what to monitor, and decides which combination of monitoring
39 modules to adopt, this should better sustain interest in monitoring, and thus sustain longevity of the system as a whole. It should be noted, however, that there are some issues which the conservancy is obligated to report on in order to satisfy government requirements (and maintain conservancy status), such as the module for wildlife poster that visually describes the responsibilities of the wildlife monitors and community resource monitors. Training of these data collectors is t hen carried out with respect to how to manage information flows and enter these into the datasheets, first on a daily basis, and then on a monthly basis, leading up to the first annual training session. The system is implemented incrementally, building up from a few modules, and then sequentially developing the management information system. The conservancies gain experience using the system over time, finally developing the capacity to interpret and use the information derived from it. The EBS operates at three temporal and institutional levels: Events: Wildlife monitors and community resource monitors collect data in yellow daily incident reporting cards, with separate cards for different events (i.e. different modules), be it a problem animal incident, c raft sale, or daily precipitation. Each data collector has their own personalized set of modules which are often kept within a carrying bag, and are with them at all times. Monthly collation: Natural resource supervisors, who are employed by the conservanc ies, meet with these data collectors on a monthly basis, correct and collate the information from the daily yellow reporting cards and enter this into blue monthly cards. Initially this is facilitated by the NGO, who then reduce their input as the conserva ncy gains more experience with the system. Annual collation: An annual joint audit of the EBS in each conservancy is carried out where the facilitating NGOs and MET meet with the conservancies, check and correct errors, and then work with the conservancy m anager or elected chairman to collate the information from the blue monthly reporting charts into a long term red (or green in the case of the community resource monitors) charts.
40 At the end of each year, the data sheets used by the game guards and resour ce monitors are collected and stored together with the blue monthly reporting charts and are filed according to year within the conservancy office. The long term charts are filed hin the system allows the community to avoid confusion in terms of information flows. Most wildlife monitors are men while most community resource monitors are women, but there are also differences in the natural resources they monitor, and in the levels of interaction with different stakeholders. Wildlife monitors mainly collect information on fugitive resources in the conservancy. As such, most of the modules they use in the yellow book correspond to some aspect of wildlife, be it problem animal incident s, predator observation, poaching of wildlife, or rare and endangered species observation. The data that they collect is mainly fed upwards towards the conservancy management committee during monthly collation of the yellow book data. While data is collect ed from the conservancy membership (for example in the form of problem animal reports), feedback on wild animal resources to members usually takes place during the annual general meetings. Community resource monitors, by contrast, tend to work directly wit h people rather than with wildlife, mainly collecting information on resources which are used by the community for making crafts. Many of the modules used in the yellow book correspond to some aspect of plant resource status (as in the case of dye tree mon itoring) or on the collection and sale of crafts. Community resource monitors are also charged with having awareness meetings on issues affecting the conservancy membership, such as information exchange on HIV and AIDS. The fact that the CRMs interact dire ctly with those community members making crafts means that they are
41 feeding information from monitoring back to the community. Data collected is also fed upwards to the conservancy management committee during monthly collation of the yellow book data. Info rmation on the community resource monitoring activity is disseminated to the rest of the conservancy membership during annual general meetings. Methods Study Site Since the EBS was initiated in the Caprivi region of Namibia, the evaluation of how this mon itoring system is performing was conducted here as these communities have the most experience in terms of implementation, and should thus have the most experience in terms of interpreting and utilizing this information. Three focal conservancies were resea rched: the Kwando, Mayuni and Mashi conservancies, three of the four conservancies where the EBS was initiated in 2000. At the time of interviewing, these conservancies had at least 7 years of experience with the EBS. The Caprivi region is surrounded by An gola and Zambia to the north, Zimbabwe in the southeast, and Botswana to the south, and lies between the Okavango River in the west, and the Zambezi and Chobe Rivers in the east (see Figure 2 1) The Kwando River divides the region into the East and West C aprivi. The area is roughly 20000 km 2 and has the highes t average rainfall in Namibia between 600 and 700 mm Caprivi, and thus this region has a high potential for increasing w ildlife densities and diversity. It is part of a wider region thought to possess the largest free ranging population of elephants left in Africa (Craig 1996). Figure 2 1 illustrates the location of
42 the study region, showing where the communal conservancies are located along the Kwando River, and the location of the nearby Bwabwata and Mudumu National Parks. Data Collection Event Book System interviews The primary dataset involved semi structured interviews conducted during 2007 and 2008 with people direct ly involved in the EBS. These interviews were used to investigate different stakeholder opinions on the EBS. Stakeholders were divided into two main categories: those in the communal conservancy who use the Event Book System: communal conservancy managem ent including managers, treasurers, enterprise officers, field officers and secretaries. The data collectors key informants who facilitat e t he implementation of the system The breakdown of the respondents is given in Table 2 1, while an organogram of those interviewed is provided in Figure 2 2, showing the EBS data flow amongst respondents. Some CRMs from conservancies outside of the focal area were interviewed to increase the sample size so this group could be more comparable to the Event Book System data I decided to compare the perceptions of the most problematic species given from the EBS conservancy interview respondents, with t he actual data provided by the EBS since this would give some indication of whether or not those carrying out monitoring were knowledgeable about the trends in the data. To this end, all of the problem animal incident data from 2004 to 2007 for each conser vancy was photographed, and was then translated into an electronic database where it could then be easily manipulated and
43 ranked by the most problematic species in each conservancy 1 This ranking could then be compared with the perceived rankings of probl em animal species given by respondents in the EBS interviews. Governance interviews The governance dataset comes from structured questionnaire interviews of people within these three communal conservancies in 2007 who were not directly involved with the m onitoring system, the details of which may be found in Collomb et al. 2010. 115 respondents were used in this analysis (Kwando = 37, Mayuni = 46, Mashi = 32), with aggregate indices (between 1 100) being calculated for each conservancy with respect to info rmation transfer on natural resources and financial resources. This dataset was used as a proxy for the degree of horizontal accountability in the three communal conservancies, and a comparison with perceptions expressed by EBS conservancy respondents. Da ta Analyses Perceptions of problem animals C ommunal conservancy respondent rankings for the species causing the most problems in the conservancy were scored for 16 species, and the overall rank for each species was calculated for each conservancy. This r anking was then compared to the actual rankings of these species for each year, and the mean actual ranking over the 4 year period. Spearman rank correlations were then used to compare the perceived ranking to the actual rankings. 1 This time period was chosen to allow the communal conservancies at least 3 years of experience with the monitoring system, and also allow for independent auditing of the datasets by NGOs and the Ministry of Environment and Tourism
44 Uses and sustaina bility of the Event Book System The communal conservancy respondents in the EBS interviews were asked which user groups used the EBS, and were also asked what these groups used the EBS for. As a follow up, respondents were also asked what decisions were made bas ed on the EBS data. This information was used t o develop a narrative describing the main uses of the EBS. The EBS interview respondents were separate d into four groups (CRMs, WMs, M anagement and key informants) and the perceptions of use of the EBS by diff erent stakeholders (user groups) were tallied. The same was done for the observe d between the four respondent groups in their perception of use, as well as in the perception of sustainability. Results Table 2 2 below shows that there is significant congruence between the perceived and actual species causing the most conflict within th e three different conservancies. There is only one case where there is a non significant Spearman rank correlation between the perceived rankings of problem animals and the actual rankings of problem animals and that occurs in Kwandu Conservancy in 2006. I t thus appears that the conservancies are knowledgeable about this aspect of their natural resource since their perceptions of the most problematic species in their communities match the reality observed in the EBS. While there was general agreement betwee n perceived and actual problem animals, it should be noted that lions in particular were perceived to be more of a problem than the data actually suggested. Lions ranked 4.5 of 16 and 5 of 16 in Kwandu and Mayuni Conservancies when the actual rankings from the EBS are 13 of
45 16 and 12 of 16 respectively. In fact, the EBS data recorded 7 of 2251 and 4 of 577 problem animal incidents perpetrated by lions in Kwandu and Mayuni conservancies for the years 2004 to 2007 respectively, none of which involved humans. The fact that lions were perceived as being one of the most problematic in these conservancies is not supported by the EBS data in terms of the number of incidents, especially since there are other predators not cited which rank higher (leopard incidents w ere recorded 35 times in Kwandu while hyena incidents were recorded 22 times in Mayuni during the same period). I will expand upon this further in the discussion section. I then compared perceived uses of the EBS by community people who implemented it to t hat of NGO support agencies. The different responses from conservancy respondents as to whether or not particular user groups used the EBS are given in Table 2 cate that there are several perceived uses, with communal conservancy management and NGOs using it the most. Table 2 4 shows the responses given by key informants about whether or not different user groups used the EBS. These responses were less vague than those given by the conservancy respondents, but were similar to the uses identified by the conservancy respondents. Of note in both Tables 2 3 and 2 4 are those uses which focus on performance monitoring of the community. These uses may be considered indi cators of accountability within the monitoring system. Wildlife data was focused within the conservancy management, and was transferred upward to NGOs and MET, reflecting the more nationalized configuration of wildlife management, compared to data on plan ts and crafts which emphasized local
46 user groups more, and outsiders less. Table 2 5 shows variability between respondent types in their perceptions of how EBS data is used by a particular user group. Of note are the majority of the CRMs who believe regula r members were using the EBS, indicated that CRM responses compared to other respondents were significantly different (Table 2 6). The perceptions of the CRMs with respect t o the membership use may be more the exception than the rule since Collomb et al. 2010 found that the information transfer indices calculated for natural resources were low for all three conservancies, in keeping with the perception of WMs, management and NGOs in the EBS interviews (Table 2 7). Respondents varied in their perception of sustainability of the EBS (Figure 2 3). While most of the CRMs believed that the monitoring program could continue without NGO support, the opposite trend was found with the WMs, with Management divided equally and key informants also believing that the conservancies could carry on the monitoring program themselves. Table 2 comparing the perceived responses of each respondent type There was a significant difference between the CRMs and WMs in their perception of the EBS surviving in the absence of NGO support, but no significant difference was found when comparing the perceptions of CRMs to Management and CRMs to key informants. C onversely, there was a significant difference observed between the WMs and key informants. The differences observed between the CRMs and the WMs may be tied to the differences in the type of natural resources they are charged with monitoring, scale, and th e length and immediacy of decision chains and economic impact.
47 Discussion This paper analyzes management within the communal conservancies of Namibia. I have evaluated how well perceptions of those conduc ting the system match the actual data, examined the perceived usage of the EB S by respondent type, and analyz ed perceptions of sustainability of the EBS by respondent type. In so doing, I relate these perceptions to the roles and responsibilities of the r espondents as described in the introduction. The perceptions of those in the community who are intimately involved in data collection and management match with respect to problem animals in their conservancies overall. However, teasing apart the degree to which these perceptions have been influenced by the EBS is difficult. The inflated perception of lions as problem animals indicates that some species are perceived to be more problematic than they actually are, and that perceptions of problem species may still be considerably weighted by cultural narratives within their communities. This should also be considered in the experiences, while the actual rankings are restricted to 4 years of data. Attitudes and perceptions towards felid conflict often distort the degree of conflict actually taking place (Inskip and Zimmermann 2009), and this may be the situation observed here. Other than skewed perceptions for some of the problem spe cies, the three conservancies appear knowledgeable about the data they are collecting. Knowing the overall trend in problem animals is an important aspect of the monitoring system since it means that data collectors and managers are paying attention to the trends in the conservancies, which is a first step in partitioning limited resources to effectively manage them.
48 Knowledge generation and incorporation into management institutions are important, but it is also important to find out what value this infor mation is producing. The uses derived and decisions made through monitoring data may be used as a proxy for value (Danielsen et al. 2005b). Several uses of the EBS were identified by the different respondent types. Most of these uses fall under the categor y of vertical accountability, where the EBS data is used to create reports so that the conservancy managers are accountable to the government (MET), or the NGOs are accountable to donors, for example. Compliance reporting contributes towards building confi dence in the CBNRM system by government and other stakeholders higher in the hierarchy (vertically upwards), while also enhancing pride in communities, and the confidence generated therein may be a useful mechanism for empowering local communities. Complia nce reporting to the government by the conservancy is important since non compliance may result in a loss of conservancy status, and the benefits that this confers upon the people within the conservancy. Accountability to the donors is important to the NGO s who depend upon them for funding, and who need to show the donors that their investment in monitoring is producing valuable information from the conservancies. These examples illustrate how the vertical transfer of information upwards is incentivized by the threat of a loss of conservancy status, or loss of funding. Some of the uses identified by the respondents relate to the management of natural resources, but for the most part the answers given were vague. Specific reference was made to the Human Anim al Conflict Self Insurance Scheme (HACSIS). The HACSIS is used to assess the degree of severity of a problem animal incident, and is linked to a compensation program under a specific set of conditions. While only
49 partial compensation is available, the HACS IS program mitigates damage done to crops and livestock. There is debate, however, on how much the EBS actually contributes towards HACSIS, some key informants maintaining that the claim procedure in HACSIS is entirely independent, while others claim that the EBS allows them to identify where an incident has occurred so that they can investigate the claim. Coupled with this, there are already concerns over the limited compensation being provided, and the ability of conservancies, NGOs and the government to sustain compensation, especially in light of expanding wildlife populations and a concurrent anticipated rise in human wildlife conflict (NACSO, 2007) WMs also use the EBS problem animal incident data to focus their mitigation efforts. For example, if elephants are raiding the crops in a field the WMs will focus efforts to guard these fields. CRMs would often cite the use of the EBS to decide which areas within the conservancy were in danger of being overexploited and in need of a respite from harvesting by the community. The latter two examples illustrate some of the more adaptive responses to problems within the conservancy. The las t broad category of uses identified may be those which help market the conservancies. The EBS is used to market the conservancy by showing trophy hunters the location of target species, allowing them to satisfy customer demands more efficiently. Conservanc y management uses the EBS to lobby for hunting quotas via the facilitating NGO. They use the trend data from natural resource monitoring to show the MET which species are increasing in population size and also in number of problem animal incidents, and MET triangulates this data to determine whether or not quotas can stringent methodology in place to identify how to incorporate the EBS data, and quota
50 setting still relie s mainly on population estimates modeled by NGOs and the MET. It remains a challenge of NGOs and government to find complementary ways to employ the data from the EBS into national and international policy decisions. In spite of this challenge, the EBS has already been used to influence national and international policy regarding the use of ivory stockpiles in Namibia. At the 2002 Convention on the International Trade in Endangered Species of wild flora and fauna (CITES) Congress of the Parties meeting, the Namibian delegation wanted to lobby for a onetime 10000 kg sale of ivory, but had not satisfied the requirements of the Monitoring of the Illegal Killing of Elephants (MIKE) program. Using EBS datasets, they were able to fill in some of the gaps in data, thus convincing delegations to allow this one time sale. Based on my analysis of the EBS, it appears that the monitoring system is being utilized, but the utility derived from the system is mainly for compliance monitoring for wildlife, and for monitoring the overall progress of the CBNRM program. The incentivized nature of this compliance monitoring to NGOs, government and donors means that there is value in the monitoring system in so far as it provides information for these stakeholders to legitimize th eir efforts, which in turn leads to a persistence of the conservancy system as a whole. In terms of conservation management interventions, the WMs use the monitoring system to mitigate problem animal incidents, and the CRMs use the monitoring system to con trol the use of craft resources, but many of the key informants lamented the fact that the use of the EBS by conservancies had not yet reached its full potential, with some of the opinion that much of the decisions made in the conservancy regarding natural resource management were still being based on
51 existing traditional knowledge. For example, coupling of natural resource, economic and institutional modules would allow for more holistic adaptive management. The results show that the perception of communit y membership use by the CRMs is more an exception than the rule, especially in light of the comparison made to the Collomb et al. ( 2010 ) very least the opinion of the WMs might match those of the CRMs since they both monitor natural resources. The design of the CRM portion of the EBS, lends itself more to horizontal accountability for several reasons which we summarize in Table 2 9. First and foremost is the nature of the resource being monito red. CRMs monitor the raw materials for the manufacture of crafts to detect changes in abundance or status. WMs on the other hand monitor mobile wild animals which are more challenging to monitor. The WMs also monitor more modules compared with the CRMs, a nd must work with the entire conservancy, while the CRMs focus their efforts on a specific user group (craft makers) in the conservancy. CRMs work with fewer community members which make it easier to interact with them. Most of the CRM duties put them in c onstant communication with the craft makers within the conservancies, where they are either taking their crafts to market, giving them money back from the market, or educating the craft makers on how to improve their techniques. Working in manageable group s has been cited as an important facet of successful collective action for the sustainable management of natural resources (Poteete and Ostrom 2004; Child and Barnes 2010). The transaction costs of CRM recommendations are lower with respect to managing nat ural resources since they can instruct the few craft makers under their purview to stop using a particular area or resource when it is in danger of irreparable harm. WMs
52 on the other hand must collect information from all the conservancy, and this informat ion then goes to the conservancy managers, who are then responsible for assimilating the information, using it in management, and feeding it back to the conservancy membership. Managers are responsible for disseminating information to the communities throu gh awareness, Khuta and annual general meetings, but the frequency and detail of this varies depending upon the conservancy. Mayuni conservancy, for example, had put off having its annual general meeting for four years before having it finally in June of 2 007. Even when the meetings are held, conservancy members may consider the opportunity costs of attendance to be too high, since meetings can be long (often lasting more than a day), meaning that many stay for the duration of the meeting (Collomb et al. 2010). Finally, the link between the work of the CRM and the benefits derived thereof is tight compared to that of the WMs. CRMs are able to provide benefits in the form of money and training directly to the craft makers, while the benefits returned to the conservancy members through the WMs is more convoluted in nature. For example, a problem animal may be reported to a WM, and then this report is taken to the management, who then contact the MET or trophy hunter to take action against the offending sp ecies. Any benefits derived through this process are shared within the conservancy, with some being shared amongst joint management institutions, or co management entities (the Mudumu North Complex (MNC), for example, was formed which is a joint management system consisting of conservancies (Kwandu, Mayuni, Mashi and Sobbe), community forests (Kwandu, Lubuta and Masida) and protected areas (Bwabata East and Mudumu National Parks)). Thus the incentive and institutional configurations of the two systems are r emarkably
53 different, although the mechanisms used are essentially the same in the form of the EBS. The design of the CRM event book modules lends itself to community feedback, if even to only those participating in craft making. Feedback is thus incentiviz ed, since this is a part of their job. By contrast, conservancy committees are accountable to those groups giving them training and funding, but there is little incentive in place for them to devolve information about wildlife from the EBS to conservancy m embership. The dearth of information feedback to conservancy membership in Namibia has been cited by others (Jones and Mosimane 2000; Jones and Weaver 2009; Collomb et al. 2010). Attempts have been made to get more regular feedback of WM data to the commun ity through village representatives, but these village representatives appear to focus their reporting to the traditional authorities in the conservancy (area and village headmen). There is clearly a design flaw in the conservancy system as a whole with re gards to information dissemination, and the translation of data into local decision making processes such as quota setting. One way this may be achieved is by adding in a compliance module which focuses on feedback meetings to the community. Such a module should include the area where a meeting took place, the number of people who attended, and the information shared. Another way is by truly devolving the quota setting process to the community, with the support of NGOs and government. This was done with som e success in the CAMPFIRE p roject in Zimbabwe (Taylor, 2001 ). The results also show that the CRMs anticipate sustaining the monitoring system with the communal conservancy me mbership, their ability to make decisions about
54 natural resource use, and the lower cost of this form of monitoring leaves them more optimistic about their abilities to conduct the monitoring system without outside support. The key informants from NGOs and MET varied in their opinion as to the sustainability of the EBS in the absence of a facilitating NGO. Most agreed that the conservancies that have been participating in the monitoring system for some time have the experience to carry on the system themsel ves, but at the same time said that the system would probably fail unless the conservancies were provided with support in the form of the physical materials used. In 2007 2008, an attempt was made to evaluate the long term sustainability of the EBS in some of the conservancies of Caprivi who appeared to be managing the monitoring system well (Caprivi Senior Management Forum 2007). Examination of the monitoring system showed that they scored poorly in this exercise. The independent auditing of the EBS by NGO s or the MET was cited as an important aspect for the successful persistence of monitoring the natural resources within these communities. Without independent auditing, the quality of the data being collected would be called into question. In order to gain legitimacy with higher level decision makers, any information collected by communities should be shown to possess few errors, and the communities collecting the data need to show that they are competent and serious about data collection (Garcia and Lescuy er 2008). As such, in the absence of independent audits of the EBS, it is unlikely that any serious effort may be made to incorporate information being collected at the local level into policy making at higher levels of governance. I recommend that NGOs an d government work with the communal conservancies to find ways to better incorporate EBS data into management action. If this happens, then the conservancy committee will come to value the
55 monitoring system more, and this may in turn lead to more conscient ious record keeping. Summary This study has shown that the conservancy committees of Kwandu, Mayuni and monitoring system, at least with regards to species causing the most con flict in their conservancy. This knowledge may still be influenced by traditional beliefs for some species, but for the most part is being incorporated into knowledge on natural resource trends in the conservancies. The way this added information is being utilized may be divided into three main categories, compliance monitoring, natural resource management, and marketing. Most of the utility for the data has come in the form of compliance monitoring in the form of reporting upwards to the NGOs, government, and donors. This is because the vertical transfer of information is incentivized by the threat of loss of conservancy status and funding. Many key informants believe that the use of the EBS has not realized its full potential, and there is the need to find ways of using the data more in a natural resource management context, as is the case with the mitigation of human wildlife conflict, and sustainable harvesting of craft resources. As with other studies, I found that horizontal accountability to the commun al membership was low, and this can lead to elite capture of information and resources by the conservancy committee, and ultimately a failure in successful CBNRM implementation. I suggest that a module be designed specifically for recording of meetings tha t the conservancy committee has with the community, and that this be included in compliance reporting so that there is added incentive to devolve information. I also suggest that efforts be made to devolve more decision making to the communities, such as q uota setting for wildlife
56 harvests. Finally, while most CRMs and key informants think the monitoring system can be sustained, I believe that unless independent auditing of the EBS is continued then the data it produces will not gain legitimacy. All parties should work together to find ways of incorporating the data into management action, which will make data collectors more diligent in their efforts when they realize the added value of this activity.
57 Figure 2 1 Map of the study region showing the location of the focal communal conservancies, and the locations of the Bwabwata East and Mudumu National P arks.
58 Table 2 1 Breakdown of the respondents for the Event Book System interviews. Category Conservancy Number of Respondents Conservancy management (9) Kwandu 3 Mayuni 3 Mashi 3 Community resource monitors (15) Kwandu 2 Mayuni 2 Mashi 3 Other 8 Wildlife monitors (16) Kwandu 8 Mayuni 3 Mashi 5 Key Informants (8) MET 1 IRDNC 4 WWF LIFE 3 Total 48
59 Figure 2 2. Organogram of the different types of respondent interviewed in the Event Book System dataset. Arrows show the flow of information from the Event Book.
60 Table 2 2 Spearman rank correlation values for the comparison of perc eived rankings of problem animal species, and the actual rankings of these species in Kwandu, Mayuni and Mashi conservancies. Conservancy Statistic Perceived Actual 04 Actual 05 Actual 06 Actual 07 Mean actual rank Kwandu Correlation Coefficient 1 0 .742 ** 0 .626 ** 0.478 0 .683 ** 0 .642 ** Sig. (2 tailed) 0.001 0.01 0.061 0.004 0.007 N 16 16 16 16 16 16 Mayuni Correlation Coefficient 1 0 .735** 0 .614* 0 .693** 0 .719 ** 0 .737 ** Sig. (2 tailed) 0.001 0.011 0.003 0.002 0.001 N 16 16 16 16 16 16 Mashi Co rrelation Coefficient 1 0 .876 ** 0 .917 ** 0 .718 ** 0 .751 ** 0 .861 ** Sig. (2 tailed) 0 0 0.002 0.001 0 N 16 16 16 16 16 16 *Significant at the 0.05 level ** Significant at the 0.01 level
61 Table 2 3 Uses of the EBS identified by communal conservancy sort ed by user group. The most frequent uses identified are ranked first. Community members Community management NGO MET Donors Researcher As a guide to instruct craft makers 1 To keep a record/make reports 1,2,3 Train/support conservancy 1,2,3 To understand natural resource trends 1,2,3 To check on conservancy performance 1,2,3 To examine wildlife conflict 2,3 To understand natural resource trends 1, 2, 3 To manage natural resources 1,2,3 To make reports 1,3 To make reports 1,2,3 To get information o n the conservancy 1,2,3 For problem animal information 2,3 To report to the community members 1,2 Ensure conservancy is working 1,2,3 Ensure conservancy is working 1,3 Ensure monitors are working 1,2 Ensure monitors are working 1,2 To at tract donors, hunters 1,2,3 To manage wildlife 2 To promote craft making 1 To understand natural resource trends 1,2,3 Quota setting of wildlife 2 Quota setting of wildlife 3 1 = CRMs; 2 = WMs; 3 = conservancy management
62 Table 2 4 Uses of the EBS identified by NGOs sorted by user group. The most frequent uses identified are ranked first. Community members Community management NGO MET Donors Researchers Traditional authority/Area headmen use for information For administering HACSIS scheme To report to donors Perfor mance monit oring Performance monitoring Wildlife conflict research (especially elephants) To assess finances in the conservancy Used for communication of natural resource trends Evaluate feasibility of HACS IS Quota setting Trends in wildlife and other natural resources Wildlife research (lions, wild dogs, hyena) To lobby for hunting quotas Quota setting and wildlife trends Ground truth human wildlife conflict To inform trophy hunters
63 Table 2 5 Percentage of each respondent type saying that the pre defined user group used the EBS. User group (defined a priori) Question: Who uses the data you are collecting in the EBS? Respondent type CRM % N=15 WM % N=16 Management % N=9 Key i nformants % N=8 Overall % N=48 Community membership 93 44 44 38 54 Community management 93 94 78 88 90 NGO 100 94 89 88 94 MET 60 88 78 50 71 Donors 60 75 89 88 60 Researchers 53 81 100 88 63 Table 2 6 exact test results illustrating the significant difference observed between the perception of use by community members for CRMs and WMs, CRMs and Management, and CRMs and Key Informants WM Management Key Informant df 1 1 1 n 31 24 23 p 0.006 0.015 0.009 Table 2 7 Data from Collomb e t al. 2010 showing measures of horizontal accountability for each conservancy. All values are on a scale from 0 100, and the composite value is the mean of the natural resource information and financial information indices. Kwandu Mayuni Mashi Index val ue n Index value n Index value n Natural resource information 34.36 37 36.41 46 32.03 32 Financial information 30.59 37 53.08 46 19.79 32 Composite value (information transfer index) 32.2 37 45.03 46 26.79 32
64 Figure 2 3 Response management; 8 key informant). Table 2 8 Exact Test p values looking at the relationship between perceptions of s ustainability of the EBS by different respondent types. The number of observations compared is in parentheses. CRM WM Management Key Informant CRM 0.008**(31) 0.657(24) 1(23) WM 0.137(25) 0.008**(24) Management 0.608(17) **Significant at the 0.01 level
65 Table 2 9. Characteristics of the Event Book System for community resource monitors verses wildlife monitors. Details of how some of these characteristics make CRMs more horizontally accountable are provided in the di scussion CRM WM Sex Female Male No of employees/conservancy 2 5 Characteristics of resource static (crafts, craft resources) mobile (wildlife) Modules used meetings, dye tree monitoring, craft sales, crafts collected, rainfall meetings, pr oblem animals, predators, poaching, wildlife mortalities, rare and endangered species, rainfall Feedback to the community meet with small groups of craft makers information disseminated at annual general meetings through conservancy committee Ec onomic impact small, but direct (potentially) large, but indirect
66 C HAPTER 3 HUMAN WILDLIFE CONFLICT IN THE CAPRIVI, NAMIBIA: WHAT CA N WE LEARN FROM COMMUNITY BASED MONITORING? While many people perceive the African elephant, Loxodonta africana as a m ajestic and magnificent animal, to those who share their range elephants can be dangerous pests, adversely affecting human security (Dublin and Hoare 2004; Warner 2008). There are perhaps few animals which stir such a wide variety of differing opinions and approach. Elephants are ecosystem engineers (Pringle 2008), flagship species for conservation agendas (Bowen Jones and Entwistle 2002), species of economic importance (van Kooten 2008), and, the focus of this paper, as causes of crop failure (Hoare 1999b) Many believe that the future of the African elephant is dependent upon the way we manage conflict arising from their interaction with humans (Fernando et al. 2005; Jackson et al. 2008), especially given the fact that much of the exists outside of protected areas (Blanc et al. 2007; Junker 2008), and that there have been increasing reports of human elephant conflict (HEC) across many parts of Africa (Hoare and Du Toit 1999; Sitati et al. 2003; Jackson et al. 2008). To help mitigate HEC, there is an increasing body of research aimed at determining those factors responsible for observed levels of HEC, as well as ways to help lessen the impacts of this problem (Naugh ton Connell Rodwell et al. 2000; Sitati et al. 2003; Os born 2004; Fernando et al. 2005; Sitati et al. 2005; Jackson et al. 2008; Hedges and Giunaryadi 2009). Human wildlife conflict (HWC) is not restricted to elephants, but many species of wildlife interact negatively with humans who share their environments. Conflict with humans includes damage to livestock or crops, injury to people, and, in some cases, death. Those that survive an attack by a wild animal may also suffer psychologically,
67 and there are also opportunity costs associated with conflict when, for example, farmers must spend nights in their fields guarding crops, or when people avoid walking at night to prevent attack. The costs of conflict with wild predators are also seldom recognized by westerners who may only consider how magnificent lions, leo pards and other large predators are to them, while, to local herders, these predators are often considered pests. Human carnivore conflict is increasing in many parts of the world (Treves and Karanth 2003), and carnivores may be predisposed to conflict bec ause of their large home ranges, dietary habits and replacement of prey species by domestic livestock (Linnell et al. 2001; Inskip and Zimmermann 2009). People, in turn, contribute to conflict with wildlife through agricultural expansion, habitat loss and fragmentation, poaching, and persecution as problem species (Parker and Graham 1989; Fernando et al. 2005; Hazzah et al. 2009; Inskip and Zimmermann 2009). Successful conservation initiatives therefore require that we balance the needs of wildlife with the needs of people that occupy the same habitats as those species. Understanding the spatial and temporal patterns of HWC will help mitigate these problems, but is complicated by variability over the landscape in the distribution and phenology of crops, th e density of people and livestock, ethology of problem species, and the proximity of refuge for wild animals. In this paper, I will examine the spatial and temporal patterns of HWC in three communal conservancies in Namibia, and examine how one can monitor these patterns in widespread and numerous communities using community based monitoring data. One important aspect of the research involving HWC and mitigation strategies is determining what information is needed to understand and manage the problem adequa tely (Dublin and Hoare 2004). For example, the Human
68 Elephant Conflict Working Group (HECWG) of the African Elephant Specialist Group (IUCN) species survival commission (SS C), has produced a manual illustrating a stepwise method for quantifying human elephant conflict (Hoare 1999 b ). While such manuals appear to be comprehensive in nature and are often touted as workable methods for communities to conduct HEC analysis themsel ves with minimal training, there are other community based monitoring approaches that may produce valuable information and which have yet to be examined regarding their potential to help describe this phenomenon. One such comm unity based monitoring program is the Event Book System (EBS) in Namibia (Stuart Hill et al. 2005). In 1996 the government of Namibia passed landmark legislation which allowed the Ministry of Environment and Tourism to declare once communities wishing to do so had defined their boundaries, defined their membership, defined a constitution, formed a committee representative of the membership, and had developed a plan for the equitable distribution of benefits to the membership (ME T 1995). The main benefit of conservancy formation in Namibia is that the communities are given access rights to natural resources they were previously prohibited fro m accessing The development of communal conservancies in Namibia has grown tremendously o ver time from just 4 registered in 1998, to 59 conservancies at the end of 2009 (NACSO 2010 ). These 5 9 conservanci es contain over 230 000 people, cover an area of over 13 million hectares, and comprise about 16.5 h a large area is prohibitively expensive for the Namibian government to conduct themselves, and they
69 decided to design and implement a monitoring system which the conservancies could conduct. While the EBS has a number of different components called modul es relating to business and natural resource management, I will focus upon its use in recording HWC. In this paper, I analyze EBS data from 3 conservancies in the Caprivi of Namibia to determine what temporal and spatial factors affect the distribution of h uman wildlife conflict. Several factors have been cited as influencing levels of HWC. Moon phase affects the number of crop raiding incidents by elephants, with less activity taking place during the full moon (Barnes et al. 2007). Predation is also thoug ht to be affected by the moon phase, with prey animals often avoiding the brightest nights (Sabato et al. 2006; Brown and Peinke 2007; Ciechanowski et al. 2007). Variation in the frequency of crop raiding by elephants has been found to be dependent on the quality and availability of wild food resources (Osborn 2004; Rode et al. 2006), so I expect there to be a distinct pattern to these incidents, while anticipating no pattern with respect to plant productivity on predation of livestock. Spatially, I examine several factors which have been shown to affect the levels of crop and livestock raiding. These include crop area (Sitati et al. 2003), distance to protected area (Naughton Treves 1998; Smith and Kasiki 1999), distance to households or settlements (Sitati et al. 2003), mean human population density ( Naughton Treves 1998, Hoare 1999 a ; Smith and Kasiki 1999 Kagoro Rugunda 2004 ), and livestock density (Stahl et al. 2002; Inskip and Zimmermann 2009; Kaartinen et al. 2009). I also examine (i) the use of EBS data to compare the levels of HWC with recorded trends in elephant populations in the region (Martin 2005; Ch ase and Griffin onnell Rodwell et al. 2000; Mulonga et al. 2003),
70 (iii) assess the limitations of using HWC data from the EBS, and (iv) recommend ancillary datasets which can be used to help model and mitigate these problems in the region. This study will contribute to the body of knowledge regarding HWC, and also to our understanding of the use of natural resource m onitoring implemented by communities. I argue that the use of community based monitoring data may help legitimize community efforts, which can help garnish support and investment in such monitoring programs especially when professionals are able to model phenomena using this data. The Event Book System The EBS was developed as a collaborative effort between the Ministry of Environment and Tourism (MET), community based natural resource management (CBNRM) support agencies (including the World Wildlife Fund (WWF) in Namibia and Integrated Rural Deve lopment and Nature Conservation (IRDNC)) and through a participatory technology development approach with communities (Stuart Hill et al. 2005). The EBS is a collaborative monitoring system, with local data interp retation (Danielsen et al. 2009), where local people decide what data needs to be collected, collect the data and are involved in interpreting and analyzing the data, using this for management decision making with advice, training and support supplied by s cientists, non governmental organizations (NGOs), and government when requested. The system has been initiated often in conservancies before they have been legally recognized The community asks for NGO support, and the facilitating NGO meets with the comm unities to ask what they wish to monitor. Following the meeting, NGOs and MET personnel help implement the system by training designated community members how to fill out
71 ex time, a series of monitoring have been developed and tested. These were initially tailor made to each community, but experien ce suggested that they could be standardized, reducing development costs and increasing comparability and cross community learning. NGOs play an integral part in supporting the monitoring syst em since they are often the ones training the communities on its use, ensuring that the data entered is accurate, and transferring the data to other stakeholders such as donors to the conservancy system. My the date of a problem animal incident, the village the complainant belongs to, spatial data on the location of the incident, the species responsible, and the type of damage done to crops, livestock, people or infrastructure (examp le in Figure 3 1). A flow diagram of the entire process is illustrated in Figure 3 2. Each WM in the community is which they keep on their person at all tim then collated data is carried out by the MET and/or facilitating NGO on a semi annual or annual basis, ensuring that no one problem animal incident is duplicated and that everything has been transferred to the monthly (blue) reporting char t accurately. Once the data has been checked, the chairman of the conservancy then transfers the blue monthly trends. NGOs or government only record the blue (monthly) and red (long term) charts, which are used in national
72 analyses including the Namibian Association of CBNRM Support Organizations ( NACSO ) review of books are thus only aggregated and much info rmation is lost in the process Methods Study Site The study site is located in the Caprivi region of Namibia (Figure 2 1 ). The region is surrounded by Angola and Zambia to the North, Zimbabwe to the East, and Botswana to the South, and is bounded by the Okavango Rivers in the West and the Zambezi and Chobe Rivers in the East. The Caprivi is roughly 20000 km 2 and has the highest annual average rainfall in Namibia at between 600 and 700 mm ( Mendelsohn and Roberts onnell the Caprivi, making it a region of relatively high potential for increasing w ildlife densities onnell Rodwell et al. 2000; Martin 2006). The apartheid style government enforced by South Africa during the Angolan conflict (1969 1989) extended to the use of wildlife resources, restricting local people from deriving any benefit from them, while at the same time the military in the region killed wildlife to serve their needs. This led to resentment on the part of local peoples who saw the wildlife in the region as pests that ought to be eradicated. Only after Namibian independence in 1990 did this change. Amendments to the Nature Conservation Act in 1996 enabled the involvement of local people in conservation, and the support of a local game guard approach by NGOs, con tributed to wildlife recovery The EBS was initiated in Caprivi conservancies, and has been in place for nearly a decade in many of these conservancies. I focused on analysis of elephant crop raiding
73 and livestock raiding by hyena, leopard and lion in the Kwando, Mayuni and Mashi conservancies, three of the four conservancies which initiated the monitoring system. These conservancies are located on the Kwandu River in eastern Caprivi, and area adjacent to one another (Figure 2 1 ). The Caprivi is located in the center of a large semi arid and dry sub humid K alahari woodlands landscape which stretches into a ll of the surrounding countries (Mendelsohn and Roberts 1997). Burkea africana (Hook.) and Baikiaea plurijuga (Harms) forest dominant regions in the north (Kwandu, Mayuni, Bwabwata East and the State Forest ), and Colophospermum mopane (Kirk ex Benth.) forest dominates the old floodplain regions in the south (most of Mashi conservancy, and Mudumu) with Okavango Kwando valley woodland and grassland persisting around the Kwando River (Mendelsohn and Roberts 199 7). This mosaic of woodland, shrubland, and grassland is influenced by many different factors including soil type, fire history, rainfall, wildlife, and cattle. My study region is bisected by the Kwando River, which serves as an important source of water for both people and wildlife, especially during the driest portion of the year (May to September). The entire Caprivi region is part of the range of the largest remaining elephant populations in Africa (Junker 2008) which are not restricted within protecte d areas, but utilize areas outside of these, competing with people for natural resources such as water, food and space. While Namibia does not have a high population density overall, high population densities are found in Caprivi, especially in Katima Muli lo in Eastern Caprivi (13 people/km 2 ). People in the communal
74 conservancies typically engage in subsistence farming and cattle rearing and are therefore vulnerable to wildlife conflict. Data Collection Elephant population data Elephant population data f or the Bwabwata Eas t and Mudumu National Parks was taken from Martin (2005) and Chase and Griffin (2009), adding supplementary information on data points and quality assessment by correspondence with investigators who either conducted the surveys in these areas, or who wrote the reports based on these surveys. The limited data for the State Forest region was judged to be of poor quality, and this area was not covered in the most recent aerial surveys so I did not use this data. Problem animal incident data Wildlife conflict data is recorded in a 2km by 2km grid reference system that was adopted from the South African Defense Force (SADF) grid system. Every time a problem animal incident is reported, the wildlife monitors investigate the incident and describ e it including a grid reference. I collated data on human wildlife conflict for the Kwandu, Mayuni and Mashi conservancies from 2004 to 2007 by photographing the yellow book pages. This time period was chosen to allow at least 3 years of practice with the EBS in each conservancy, while also being independently audited and corrected by NGOs and the MET. The photographs were then translated into an electronic geographical information system (GIS) database allowing for examination of the data in both temporal (daily) and spatial (4 km 2 resolution over an area of approximately 640 km 2 ) scales. Crop raiding specifically refers to incidents perpetrated
75 by elephants. The data for hyena, leopard and lion were combined into one dataset for livestock raiding. Temporal d atasets The timing of crop raiding and livestock raiding was compared to moon phase and a remotely sensed vegetation index. The months of crop and livestock raiding in the study region were matched with the moon phase over the study period (waxing, full waning, new). Each moon phase was defined as the week in the middle of which the phase occurred. I also modeled crop raiding over the entire study period with the variation in the mean value for the normalized difference vegetation index (NDVI) within t he conservancy are a over the entire study period. I used NDVI as a proxy for vegetation productivity with higher values indicating higher vegetation productivity. The index is calculated as follows: NDVI = (NIR RED)/(NIR+RED) where NIR represents the near infrared band of the satellite, and RED represents the red band of the satellite. I used the Moderate Resolution Imaging Spectroradiometer (MODIS) MOD13Q1 product from October 2003 to January 2008 which was downloaded, stacked by NDVI, reprojected, and su bset for the area of the communal conservancies using ENVI v.4.7 (ITT Visual Information Systems). This product is provided every 16 days at 250 meter spatial resolution. I found the mean NDVI value over the entire conservancy area (i.e. Kwandu, Mayuni and Mashi combined) giving us 99 data points for the 2004 to 2007 period which I then matched to incidents of elephant crop raiding. While some studies have used remote sensing in the past to help describe the phenomenon of elephant movements (Ngene 2010), th eir ability to persist in
76 agricultural landscapes (Murwira and Skidmore 2005), the effect they have on the vegetation (Young et al. 2009) and the effect vegetation productivity has on their reproduction (Wittemyer et al. 2007), few studies have examined th e relationship between elephant crop raiding and plant phenology on a broad scale using satellite indices (Graham 2006 ; this paper ). Spatial d atasets The modified SADF 4km 2 grid system provided 202 spatial samples for the study area (see Figure 3 4 fo r th e distribution of this grid ). Six independent variables were used to examine the spatial occurrence of crop and livestock raiding on this grid system using ArcGIS v.9.3 (ESRI Inc. Redlands, CA): Latitude derived from the centroid of each grid cell Longit ude derived from the centroid of each grid cell Percentage land area under crops in each cell of the communal conservancy region was derived from Google Earth Imagery. Areas under ac tive crops were digitized from two May 4th 2010 SPOT image s at 2.5 m res olution (lat/long center: Figure 3 3 )). The spatial accuracy of the digitizing was evaluated using training sam ples collected in 2007 and 2008 The distance to refuge was derived using the cent roid of each grid cell, with Bwabwata East National Park, the State Forest, and Mudumu National Park being considered as refuge areas for elephant. The centroid of each grid cell was used to avoid bias in estimating this distance, especially since other va riables such as crop area may have different distribut ions depending on the grid cell The distance to households was derived in relation to the centroid of each grid cell, with the household layer derived from the most recently available aerial photograph y (1:20000) taken in 1996 (Mendelsohn and Roberts 1997). Mean human population density in persons per square kilometer was derived using zonal statistics within each grid cell (Figure 3 4 ). The human population density was derived from a model incorporatin g satellite information, land cover information and census data dev e loped by The Afripop Project ( 2010 ).This data has been shown to be a reliable data source compared to other models of human population density (Tatem et al. 2007; Linard et al. 2010 ).
77 Th e livestock raiding model used mean livestock density which was derived from data collected by the Department of Veterinary Services of the Ministry of Agriculture, Water and Forestry in Katima Mulilo. The data corresponds to animal crush pens which are us ed as vaccination points throughout the study area. I used the Mendelsohn and Roberts (1997) method to create a density map for livestock, assuming that cattle came from within a 10 km radius of a crush pen. Zonal statistics were then used to find the mean number of livestock within each grid cell. Data Analysis Elephant population trends and comparison to past studies Based on the quality assessment of the available elephant population data (Martin 2005; Chase and Griffin 2009), I found four data points for Bwabwata East National Park, and six data points for Mudumu National Park. All of these data points were recorded during dry season aerial surveys. I combined the data for Bwabwata East and Mudumu National Parks to examine the trend in elephant density in the study region. The trend in elephant populations was then compared to the overall trend in elephant crop raiding incidents for the three communal conservancies. Rodwell et al. (2000) and Mulonga et al. (2003) was compared to the geographic location of wildlife conflict report ed in the EBS from 2004 to 2007 to determine whether any similar patterns were observed. Temporal patterns of crop and livestock raiding A Wilcoxon matched pairs signed ranks te st was used to test for differences between the new and the full moon phase in terms of the number of crop and livestock raiding incidents during each month where any raiding incident was recorded in the 4 year study period (2004 2007). I also tested wheth er there were significant differences between the new, waxing and waning moon phases using a Kruskal Wallis test.
78 For the analysis of NDVI and crop and livestock raiding incidents, a cross correlation function was applied to identify if there was any ident ifiable lag between the peak in NDVI, and the peak in crop raiding using R statistical software (v.2.10.1). Removal of seasonality in the dataset (prewhitening) was then carried out to identify cross correlations on the residual values, which helped identi fy whether or not higher or lower levels of HWC were associated with higher or lowe r NDVI values (Shumway 2006). Spatial patterns of crop and livestock raiding The data for both elephant crop raiding and livestock raiding were both zero inflated, with man y grid cells not having any crop or livestock raiding incident over the entire study period (Figure 3 5 A) and B)). To account for this, I used a zero inflated model, comparing two models of both crop and livestock raiding. Six independent variables were s tandardized and then included in models of crop or livestock raiding (described in the data collection section ). I compared a zero inflated Poisson (ZIP) regression to a zero inflated negative binomial regression (ZINB) to determine which was better able t o explain the dependent variable (crop raiding or livestock raiding). Zero inflated models are modeled in two parts, the first being described as the perfect state, in which there are no counts (zero), and the second is the imperfect state, where positive events can occur, but these are not certain (i.e. you can get both zero and non zero values) (Agarwal et al. 2002; Minami et al. 2007). In the case of ZINB regression, the perfect state is modeled with a logistic regression, while the imperfect model is mo deled with a complete negative binomial distribution. For a ZIP regression, the perfect model is modeled the same as in a ZINB regression, but the imperfect model is modeled with a Poisson distribution. When I describe the results, I refer to the perfect
79 a nd imperfect models separately, since the perfect model is es sentially only modeling Results Over the four year period, there were 1193 elephant crop raiding incidents. in the 3 conservancies, with the majority of these occurring in Kwandu Conservancy (Figure 3 6 ). Livestock depredation by hyena, leopard and lion showed the opposite trend, with the majority of incidents taking place in Mashi Conservancy (Figure 3 7 ) Elephant Population Trends and Trends in Crop R aiding The change in the density of el ephants in the study region showed a general increase from 1985 to 2005 (Figure 3 8 ), but appears to fluctuate later on between1994 to 2005. There are not sufficient data points to warrant any useful comparison between the elephant densities over time, and the frequency of crop raiding in the communal conservancies which are surrounded by the Bwabwata East National Park and Mudumu National Park since only one data point (2005) falls within th e 2004 to 2007 EBS study period Type of Conflict and Comparison t o Past S urvey Rodwell et al. (2000) compared the proportion of wildlife conflict being conducted by elephant, lion, and hyena in 14 villages in this region, and while their study extended further southwards beyond the Mudumu National Park, Figure 3 9 shows results similar to the study they conducted from 1991 to 2000 where a higher proportion of predation by lions and hyena occurs on livestock in the southern region of Ro dwell et al. (2000) also found that most of the elephant crop raiding in the study region occurred in the northern parts of their study region, a region which corresponds with the Kwandu
80 Conservancy which has consistently had the highest levels of elephant crop raiding over the study period (Figure 3 6 ). Tempor al Trends in Elephant Crop R aiding Effect of moon phase At least one elephant crop raiding incident was evident in 35 of the 48 months in the study period. Figure 3 10 shows the trend in the number of crop raiding incidents matched with their respective day of the lunar cycle, with the new moon at day 1 and the full moon at day 14, 15 or 16. There were significantly more crop raiding incidents during the new moon phase compared with the full moon pha se (Wilcoxon matched pairs signed ranks test Z = 2.028, P = 0.043). There was no significant difference in the new, waxing, and waning phases (Kruskal Wallis test ). Plant productivity and crop raiding Figure 3 11 shows that there is close correspondence between the peaks in NDVI and the peaks in elephant crop raiding in the study region. The lagged correlation correlogram between N DVI and elephant crop raiding in the study region reveals that the largest correlation occurs at a lag of about 48 days, meaning that the peak in NDVI is followed by a peak in elephant crop raiding 50 days later (Figure 3 12 ). Pre whitening of the datasets revealed that there was no significant effect in terms of the magnitude of mean NDVI and the magnitude of elephant crop raiding (Figure 3 1 3 ). One interesting aspect of the pre whitening process did reveal that there may be more crop raiding incidents if the productivity on the landscape is lower in the previous dry season (at about 192 days or about 6 months pr ior to the peak in crop raiding ).
81 Temporal Trends in Livestock R aiding Effect of moon phase Of the 48 months examined, 44 had at least one livest ock raiding incident from hyena, leopard or lion. Figure 3 14 shows the trend in the number of crop raiding incidents matched with their respective day of the lunar cycle, with the new moon at day 1 and the full moon at day 14, 15 or 16. There were signifi cantly more crop raiding incidents during the new moon phase compared with the full moon phase (Wilcoxon matched pairs signed ranks test Z = 1.992, P = 0.046). There was no significant difference in the new, waxing, and waning phases (Kruskal Wallis test 2 = 2.146, df = 2, P = 0.342 ). Plant prod uctivity and livestock raiding Figure 3 15 shows that there is no definitive relationship between the peaks in NDVI and livestock raiding in the study region. Likewise, the lagged correlation correlogram between ND VI and livestock raiding shows no significant lagged effect, as does the correlogram of the pre whitened datasets (Figures 3 16 and 3 17 respectively) Spatial Patterns of Crop and Livestock R aiding The zero inflated negative binomial regression model resu lts for the prediction of crop raiding and livestock raiding are shown in Tables 3 1 to 3 4 Both models were statistically significant (Wald test, p < 0.0001 for crop raiding; p = 0.0037 for livestock raidin g). For elephant crop raiding, percentage of cro ps in the grid and latitude were the only statistically significant predictors in the imperfect model. In the perfect model, longitude and mean number of persons /km 2 were statistically significant predictors of no elephant crop raiding. For livestock raidi ng, latitude was the only statistically significant predictor in the imperfect model at p< 0.05, while mean human population
82 density and distance to households were significant at the p<0.1. In the perfect model, longitude was the only statistically signi ficant variable at p < 0.1.The dispersion parameter in both models (Log(theta)) was significantly different from zer o, indicating that the data is over dispersed and a negative binomial model was more appropriate than the Poisson model. I used a Vuong test to confirm that the zero inflated negative binomial model was a significant improvement over the standard negative bino mial regression for both models. Figures 3 18 to 3 21 show the observed, predicted and residual values for each model. For both crop rai ding and livestock raiding, the models are able to identify hotspots for conflict, with crop raiding incidents being concentrated in the North of the study area, and livestock raiding in the Southeast. In terms of modeling the magnitude of crop raiding, th e model is better able to detect areas of no crop raiding, underestimating some grids that have crop raiding (for example grid 96, 16 Figure 3 19 ) while there is great overestimation on others (for example grid 96, 16 Figure 3 19). Still, most of the resid ual values lie in the 9.9 to +10 range. The magnitude of livestock raiding events fare s better, also able to detect areas of no livestock raiding, the major problem being underestimation of these incidents (for example grid 102, 33 Figure 3 21 ). Discussio n This research shows how one can examine human wildlife conflict both temporally and spatially using community based monitoring data over a widespread area. I examined the community based monitoring data from three communal conservancies in Namibia to se e whether or not trends in elephant crop raiding incidents matched elephant population trends in the region. I also compared the EBS problem animal
83 proportions to those of a past study to see what patterns emerged, and modeled temporal and spatial aspects of elephant crop raiding in the region. The dry season data on elephant population density in the Bwabwata Eastern Core region and the Mudumu National Park (combined) showed that elephant densities in the region have been increasing overall, but there app ears to be some stabilization in the latter years. Elephant crop raiding in the region has also increased over time in the study region, with 156 incidents in 2004, and 442 incidents in 2007. I cannot infer whether this has anything to do with the populati on trend since I only have one elephant density data point to compare, and the majority of crop raiding incidents recorded also take place during the rainy season, while density measurements were recorded in the dry season. In the context of the regional p opulation, Junker et al. (2008) found that the population of elephants in Northern Botswana was stabiliz ing over time from 1996 to 2004 after an increase from 1973 to 1993. The increase in levels of elephant crop raiding in my study region may thus not be the result of an increase in elephant populations per se, but due to a shift in elephant range The expansion of areas under some form of conservation status in Namibia, and the cessation of conflict in Angola may have allowed for the expansion of elephant populations back into their historic range (Chase and Griffin 2009), which may have inevitably led to increasing levels of human elephant conflict in more recent times. With the elephant populations appearing to stabilize (Junker et al. 2008), the number of elephants participating in crop raiding might also stabilize. That said; elephant numbers are not the only variable contributing to the level of crop raiding experienced.
84 The distribution of elephant crop raiding incidents and livestock incidents by hye na though their study stretched further south along the Kwando River. The similarities to this past survey lend credence to the data that the community is collecting. T he ability of the EBS to describe accurately the trends in human wildlife conflict was also noted by Mulonga et al. (2003) where they compared the number of human wildlife conflict incidents reported to the MET to that of the EBS data recorded over the sam e time period (1996 2001). This study illustrated similarity in the proportions of incidents perpetrated by different species, and also suggested that the EBS might be more accurate since it is easier to contact WMs to report incidents rather than MET em ployees (Mulonga et al. 2003). Many studies show that elephant crop raiding typically occurs at night (Naughton Treves 1998; Sitati et al. 2003; Fernando et al. 2005; Sitati et al. 2005; He dges and Giunaryadi 2009), and the effect of the moon p hase on cro p raiding is clear. Barnes et al. (2006) also found that there were significantly more crop raiding incidents during the new moon phase compared with the full moon phase, and postulated that the elephants perhaps feel safer on moonless nights, and are thus more likely to venture into crop fields. People guarding fields may also be better able to detect invading elephants when there is more light as well as dogs which might accompany them and alert people to elephant presence. In this case we might consider people as potential predators of elephants, and several nocturnal animals have been shown to be less active during the full moon to avoid predation (Lima and Dill 1990; Lima 1998; Griffin et al. 2005). The same can be said of the predators who perpetrated less livestock depredation during
85 the full moon phase. The EBS showing this trend may be considered further validation of its data quality. The fact that there is seasonality to the crop raiding incidents in the Caprivi is not surprising, with elephants r aiding crops when they are at their optimal level, and are ready to be harvested. This has been shown in other studies examining elephant crop raiding (Osborn 2004; Rode et al. 2006). Fewer studies, however, have looked at the relationship between wild pla nt productivity in relation to the levels of crop raiding (Hoare and Du Toit 1999; Smith and Kasiki 1999, Sitati et al. 2003 ). My study took a broad approach to estimating the productivity on the landscape by looking at mean NDVI values across the study re gion at a 250m spatial resolution, and I found that there was a 48 day lag in peak crop raiding behind the peak in productivi ty over the conservancy area, with no strong effect in terms of the magnitude of NDVI. Osborn (2004) observed a similar lag in the types of food consumed by elephants, where the elephants were more likely to raid fields when the quality of wild grasses were on the decline. In this case, the elephants consumed less grass as the moisture content of the grasses declined, which may have s erved as a trigger for the increased levels of crop raiding. Elephants in this study were observed to select the food with highest nutrient availability at a given time, thus selecting crops which maintained their nutrient quality after that of grass quali ty declined, and just before the quality of wild browse increased. The NDVI crop raiding pattern suggests that as the quality of wild browse declines, elephants increasingly engage in risky forays into crop fields, and also identifies the need for further research on this phenomenon. It should also be noted that while the prewhitening of the datasets to remove seasonality did not identify strong correlations
86 between the magnitude of NDVI and the magnitude of crop raiding, the highest correlation ( 0.415) wa s found at a lag of 192 days (Figure 3 13 ). This indicates that if the previous dry season has lower productivity (NDVI), then this may result in a greater frequency of crop raiding incidents 6 7 months later. Perhaps this is related to the lower availabil ity of wild browse after periods of low productivity, caused by either lower rainfall, or perhaps a higher frequency of fires on the landscape. If this is the case, then local people may be able to identify whether or not they are likely to have increased crop raiding incidents if the quality of the wild browse prior to harvesting is in poor condition, and assign additional resources based on this prediction to help mitigate these impacts. Further investigation into this phenomenon through expansion of the time period and comparison of vegetation productivity on a smaller scale would be useful, as well as an examination of crop phenology. Unlike elephant crop raiding, livestock raiding did not appear to be seasonally linked to NDVI values over the study peri od. This makes intuitive sense, since crops have a distinct season, while livestock roam across the landscape throughout the year and may be attacked w henever the opportunity arises. That said, extension of this temporal dataset may reveal changes in lives tock raiding with climatic changes that adversely affect wildlife, an issue that warrants further investigation. The use of a gridded reference system to identify spatial aspects of human wildlife conflict was used by Sitati et al. (2003) in their examinat ion of spatial correlates of human elephant conflict. The ZINB regression model of elephant crop raiding modeled both the distribution of incidents across the landscape and approximated the magnitude of these events. The only variables statistically signif icant in the model were the
87 percentage of crops and latitude in the imperfect model, while longitude and mean human population density were statistically significant in predicting areas of no elephant crop raiding, the perfect model. It makes sense that ar eas with a higher proportion of crops would more likely experience elephant crop raiding. Since Kwandu Conservancy in the northern portion of my study region has consistently had the most elephant crop raiding incidents, latitude is an important variable c ontributing to the overall model. Kwandu has both the highest area under active fields (13.1 km 2 or 7% of the total area), as well as the highest population density ( 18 persons km 2 ) compared to both Mayuni (6.03 km 2 (4%); 13 persons km 2 ) and Mashi ( 13 km 2 (3%); 6 persons km 2 ). These two factors combined are the main reasons for the higher levels of elephant crop raiding observed in Kwandu; there are more people to make reports, and more crop fields for the elephants to raid. This analysis showed that human population density was positively correlated with elephant crop raiding. Hoare and Du Toit (1999) found a threshold phenomenon where elephant density varied widely at low human population densities, while at high human population densities there was a sig nificant human effect on elephant habitat, which only allowed low elephant densities to exist. They hypothesized that there was a threshold human density above which the conversion of the landscape to agricultural cover would result in a sharp decline in t he number of elephants (and by extension the number of elephants able to propagate crop raiding). In the study region, the low human population density means that the landscape has not been converted above this threshold so that elephants have nocturnal re fuge within the conservancies, and the presence of elephant refuges in the form of the national parks and the fores t reserve
88 afford the elephant sufficient daytime refuge nearby. Longitude was also a significant determinant of no elephant crop raiding, wit h less crop raiding close to the Kwando River in the West, and more incidents occurring eastwards. The distance of a sample grid to the refuge boundary was not a significant factor accounting for the presence of elephant crop raiding. Most of the crops in the study region were not grown near the national park or forest reserve boundary, but were often placed along the main roads within the conservancies, with fewer fields observed in grids nearer to the margins of the conservancies (Figure 3 3 ). The average daily movement of elephants in dry regions averages 6 km (Loarie et al. 2009), and the average distance of each grid cell to a refuge was about 3.5 km. Thus, elephants raid crops in most grid cells within a day, contributing to the non significance of dis tance to refuge in my model. The fact that the distance to households was also found to be non significant in the model lends further evidence to elephants wandering over this landscape unimpeded, (or raiding from nearby refuges) entering crop fields at ni ght, and then returning to the protected refuges during the day. The ZINB regression examining livestock depredation by hyena, leopard and lion was also statistically significant and was able to identify hotspots of human predator conflict in the study reg ion. In the imperfect model, latitude was the most significant explanatory variable for the magnitude of livestock raiding since most of the livestock raiding occurs in the southern part of the study region. The distance to households and mean human popul ation density were also somewhat significant determinants, with both having a positive correlation with livestock raiding. It makes sense that the higher human population would result in higher livestock raiding as there would be more
89 people reporting prob lem animals. The predators may also avoid village areas and stalk livestock when they are further away from households to avoid the risk of persecution since they may associate these structures with people. For the perfect model, longitude contributed the most to explaining the areas of no livestock raiding, with grid cells closer to the river not having as many incidents as those further inland. This is in agreement with the overall pattern observed with livestock raiding (Figure 3 20 A ). Distance to refug e was not a statistically significant predictor of livestock depredation, probably because these predators are able to move over the landscape unimpeded, taking refuge within the conservancies themselves, and probably returning to the safety of the nationa l parks and forest reserve during the day. Surprisingly, mean livestock density was not a large contributor to the model. This may be an artifact of the quality of this dataset not mirroring reality, or perhaps those areas of high livestock density also co rrespond with areas of little vegetation cover, a high density of people, or a combination of these which may deter predators from engaging in risky raiding behavior Livestock depredation behavior may simply be an opportunistic trait, where predators atta ck livestock if the opportunity arises, and is thus not dependent upon the density of the prey animals. While these analyses illustrate that the EBS can be used for more detailed analysis of elephant crop raiding and livestock raiding, some additional deta ils would make the monitoring system better for modeling of human wildlife conflict. The errors in estimation of the magnitude of crop raiding and livestock raiding (Figure 3 19 and 3 21 ) may b e minimized by including these in future analyses. The wildlife monitors record the type of crop is damaged during each event, but not the amount of damage incurred so
90 we cannot ascertain if the wildlife is actively feeding on crops, or if the damage is incidental as they pass through fields. It would be useful to add an estimate of the area of crops damaged as a subjective evaluation by the wildlife monitors, or perhaps a simple count of the number of steps it takes to cover the perimeter of the damaged area. Alternatively, a parallel sampling system could be develope d to compliment the EBS by getting more details on each of the incidents. Information on the sex and number of animals committing such incidents could also be included in monitoring. Ancillary datasets would help improve model accuracy. Information on the proportion of different crops grown in the study region would help us assess if species are selective. Radio tracking elephant individuals would help with the modeling of spatial determinants of elephant crop raiding; for example, are there elephants movin g between the forest reserve and Bwabwata National Park during the height of the crop raiding season? For predators, data on the populations of hyena, leopard and lion in the study region; their distribution across the landscape; feeding ecology and behavi or ; and whether or not kraals and other preventative methods help mitigate livestock depredation would go a long way in helping to make the model more accurate. Working with other scientists in the region may help fill in some of the knowledge gaps regardi ng predators in the region. Lise Hannsen of the Predator Conservation Trust has already used the hyena livestock raiding maps developed through this project to decide where to bait hyena so she might study them (pers comm. 2010). Such collaborative efforts will help develop a greater understanding of human wildlife conflict and the most effective means of mitigating this problem.
91 Summary This study highlighted the use of community based monitoring data to model human wildlife conflict. I showed that the pat terns of HWC match what was observed in the past, and suggest that the increases observed in elephant crop raiding may be more a result of elephant expansion into their historical range than an increase in elephant population size over the study period. Te mporal data showed less crop raiding by elephants and livestock raiding during the full moon phase, and a 48 day lag in the peak of crop raiding compared to plant productivity (NDVI). Spatially, I illustrate the use of ZINB regression models for HWC resear ch, and found several variables which appear responsible for the spatial patterns observed. Future analysis of crop raiding would be improved by including data on the area of crops damaged, the proportion of different species of crops in a given field, the sex and number of animals committing crop raiding incidents, and, specific to elephants, the elephant pathways frequented. For predators, I suggest the inclusion of regional information on their ecology and behavior as well as the recording of preventati ve methods used in livestock raiding mitigation would help support the information needs of the conservancy, this study shows that this management oriented system also sup ports more sophisticated analyses. The fact that the community based data in this case study was able to explain trends in elephant crop raiding suggests that when community based monitoring systems are carefully designed and implemented, they can be robus t and useful. Stakeholders should not shy away from using community based data, but should encourage efforts to incorporate this data into management, working with communities to find flaws in the system and adjust them to meet new challenges.
92 Figure 3 1. An example of a problem animal incident yellowbook shee t from Mayuni Conservancy, 2005 ( The name of the CGG and names and signatures of complainants have been obscured to preserve their anonymity )
93 Figure 3 2 Flow diagram of the temporal and organ isational structure of the problem animal incident module of the EBS.
94 Figure 3 3 Areas under active crops mapped using SPOT 2010 imagery.
95 Figure 3 4 Mean human population density within each grid cell over the communal conservancy region.
96 Figure 3 5 Zero inflated nature of human wildlife conflict data. Frequency distribution of A) elephant crop raiding and B) livestock raiding taken from 4km 2 sample grids from 2004 to 2007. Both datasets warrant examination using a zero inflated Poisson o r negative binomial regression.
97 Figure 3 6 Total number of elephant crop raiding incidents in the three focal conservancies over the course of the survey period. Each conservancy shows an overall increase over time, consistent with the increases in ele phant densities observed in Figure 3 8
98 Figure 3 7 Number of predation incidents on livestock by lion, leopard and hyena. Mashi Conservancy had the most number of incidents, with most being perpetrated by hyena. An exception is the leopard incidents w hich were predominantly reported in Kwandu Conservancy (35) (Mayuni (9); Mashi (15)).
99 Figure 3 8 Elephant densities over time for the period 1985 to 2005. There is an overall increasing trend, with elephant density seeming to flu ctuate between 1 and 2 elephant/ km 2
100 Figure 3 9 Differences observed in the distribution of human wildlife conflict in the study area. A) elephant crop raiding, B) hyena and lion attacks on livestock in the study region from 2004 to 2007. This illustr ates the higher proportion of livestock are affected in the southern part of the study area, similar to what was found By Rodwell (2000)
101 Figure 3 10 Number of elephant crop raiding incidents during the lunar cycle in the communal conservanci es. Data is based on 35 lunar months, with the new moon beginning at day 1 or 29 (indicated by filled circles), and the full moon on night 14, 15 or 16 (indicated by stippled bars and open circle). The 3 period moving average trendline shows the lowest val ue occurring during the full moon.
102 Figure 3 11 Trend in mean NDVI values and elephant crop raiding incidents in the study region.
103 Figure 3 12 Lagged correlation correlogram showing a peak in correlation at about 48 days i.e. there is a peak in elep hant crop raiding 48 days after the peak in NDVI.
104 Figure 3 13 Prewhitened data showing no strong effect of NDVI magnitude on subsequent levels of elephant crop raiding. The most significant value occurs at a lag of 192 days with a cross correlation factor of 0.4, suggesting that if there is a drier than usual period before the next rainy season, then there may be a higher incidence of crop raiding.
105 Figure 3 14 Number of livestock raiding incidents during the lunar cycle in the communal conserv ancies. Data is based on 44 lunar months, with the new moon beginning at day 1 or 29 (indicated by filled circles), and the full moon on night 14, 15 or 16 (indicated by stipled bars and open circle). The 3 period moving average trendline shows the lowest value occurring just after the full moon.
106 Figure 3 15 Trend in mean NDVI and livestock raiding over the 4 year period.
107 Figure 3 16 Lagged correlation correlogram showing no significant relationship between mean NDVI and livestock raiding.
108 Figure 3 17 Prewhitened data showing the effect of NDVI magnitude on subsequent levels of livestock raiding. While there were four statistically significant correlations, none of these was strong enough to be considered a discernable pattern.
109 Table 3 1 Im perfect model results of the zero inflated negative binomial model of elephant crop raiding in the communal conservancies Imperfect model Estimate Std. Error z value Pr(>|z|) Significance level Intercept 1.67171 0.15381 10.869 <0.0001 *** Percentage crop s 0.38678 0.17133 2.258 0.02397 ** Distance to refuge 0.11243 0.23912 0.47 0.63821 Longitude 0.0581 0.21052 0.276 0.78258 Latitude 0.51493 0.1795 2.869 0.00412 *** Mean population density 0.06717 0.15838 0.424 0.67147 Distance to households 0.16 81 0.20933 0.803 0.42195 Log(theta) 0.4264 0.17307 2.464 0.01375 ** Significance codes: 0.01***, 0.05**, 0.1* Table 3 2 Perfect model results of the zero inflated negative binomial model of elephant crop raiding in th e communal conservancies Per fect model Estimate Std. Error z value Pr(>|z|) Significance level Intercept 5.28913 2.3422 2.258 0.0239 ** Percentage crops 0.06353 1.04684 0.061 0.9516 Distance to refuge 0.4026 0.75762 0.531 0.5951 Longitude 2.25183 1.06734 2.11 0.0349 ** Latitude 0.20045 0.82946 0.242 0.809 Mean population density 10.227 4.51911 2.263 0.0236 ** Distance to households 0.55412 0.66871 0.829 0.4073 Significance codes: 0.01***, 0.05**, 0.1*
110 Table 3 3 Imperfect model results of the zero in flated negative binomial regression model of livestock raiding by hyena, leopard and lion in the communal conservancies Imperfect model Estimate Std. Error z value Pr(>|z|) Significance level Intercept 0.84143 0.25523 3.297 0.000978 *** Distance to refug e 0.03139 0.32287 0.097 0.922552 Longitude 0.05483 0.22331 0.246 0.806035 Latitude 0.65304 0.29545 2.21 0.027081 ** Mean population density 0.35659 0.19499 1.829 0.06744 Distance to households 0.56605 0.29162 1.941 0.05225 Mean livestock den sity 0.2104 0.25095 0.838 0.401805 Log(theta) 0.53021 0.2692 1.97 0.048889 ** Significance codes: 0.01***, 0.05**, 0.1* Table 3 4 Perfect model r esults of the zero inflated negative binomial regression model of livestock raiding by hyena, leopard and lion in the communal conservancies Perfect model Estimate Std. Error z value Pr(>|z|) Significance level Intercept 3.4856 2.7014 1.29 0.197 Distance to refuge 0.3281 0.8153 0.402 0.6874 Longitude 2.3349 1.2336 1.893 0.0584 Latitude 0.625 8 0.8219 0.761 0.4465 Mean population density 6.6418 4.765 1.394 0.1634 Distance to households 1.7592 1.2449 1.413 0.1576 Mean livestock density 0.6565 0.9761 0.673 0.5012 Significance codes: 0.01***, 0.05**, 0.1*
111 Figure 3 18 Map of the modeling results of elephant crop raiding. A) The observed e lephant crop raiding incidents B) the predicted values. The figure illustrates that the model was able to identify hotspots of raiding, but the magnitude of the problem is overest imated in some of the cells.
112 Figure 3 19 Results of the zero inflated negative binomial model for elephant crop raiding showing the residual values for the model.
113 Figure 3 20 Map of the modeling results of livestock raiding. A) The observed livestock raiding incidents B) the predicted values. The figure illustrates that the model was able to identify hotspots of raiding, but the magnitude of the problem is mostly underestimated estimated in some of the cells.
114 Figure 3 21 Resul ts of the zero inflated negative binomial model for livestock rai ding by hyena, leopard and lion showing the residual values for the model
115 CHAPTER4 EFFECTS OF FIRE FREQ UENCY AND CATTLE DEN SITY ON VEGETATION IN THE KWANDU, MAYUNI AND M ASHI CON SERVANCIES, NAMIBIA Tropical savanna many abiotic and biotic factors which interact on the landscape in co mplex ways to produce different woody and herbaceous vegetation patterns in savanna ecosystems (Scholes and Archer 1997; Sankaran et al. 2 005). The complexity of savanna ecosystems is also affected by the presence of people on the landscape, who have the p otential to alter natural processes which change the overall structure and function of savanna ecosystems, and in turn affect the type and quantity of ecosystem services they provide ( Higgins et al. 1999 ) Current land use practices such as agriculture and livestock husbandry in southern Africa may not be sust ainable in the context of devolution of the management of common pool resources such as wildlife to communities since these land uses compete with the very resources upon which wildlife are dependent ( Adams and Hulme 2001a; Brown 2004). In Namibia, landmark legislation in 1996 led to the formation of common property institutions called conservancies where communities could derive benefits from wildlife (NACSO 2010) Since its inception, the communal cons ervancy system has grown throughout the country, and has resulted in a steady increase in benefits to communities derived from wildlife over time (NACSO 2010). At the same time, conservan cies continue to engage in land use practises that have the potential to undermine conservation efforts. Subsistence crop farming in the communal conservancies contributes to habitat loss and fragment ation of the landscape, creating barriers for the movement of wildlife (Martin 2006), while contributing to human wildlife
116 co nflict since wild animals raid crop fields, resulting in their persecution by people Rodwell et al. 2000) C onversion of the landscape for crop farming is not the only land use practise that can undermine conservation efforts O vergrazing by liv estock and manipulation of the fire regime in these conservancies can also alter the vegetation structure and function in African savanna ecosystems affecting the ability of wildlife to persist within them ( Dublin et al. 1990 ; du Toit and Cumming 1999 ) Th e Caprivi of Namibia has had a history of fire suppression and livestock husbandry (Mendelsohn and Roberts 199 7) which may affect the patterns of vegetation communities observed today. To this end, the research question of this paper is how do fire frequen cy, cattle grazing density, and edaphic factors interact and affect the vegetation community structure and function in several communal conservancies ? Edaphic fa c tors are important determinants of ve getation composition in savanna ecosystems (Scholes and A rcher 1997; van Langevelde et al 2003). Woody (shrub and tree) and herbaceous (grass) vegetation compete with one another for soil moisture, for example, and this is in turn affected by soil characteristics (Scholes and Archer 1997). As such, it is importa nt to consider edaphic factors in analysing vegetation community composition. African savanna ecosystems are also often characterized by the presence of two herbivore feeding guilds; the grazers such as buffalo ( Syncerus caffer ) which feed on the herbaceou s layer, and the browsers such as elephants ( Loxodonta africana ) which feed on woody vegetation (Holdo 2009). Grazing and browsing animals interact with abiotic conditions in the environment and with one another to affect the tree grass ratio in African sa vanna s ( Scholes and Archer 1997 ; Holdo 2009). Grazers feeding on the herbaceous layer, for example, may regulate th e effect of fire in the savanna
117 ecosystem by reducing the fuel available, thus allowing woody vegetation to grow (van Langevelde et al 2 003). At the same time, browsers fee ding on woody vegetation open the canopy allowing grasses to persist. van L angevelde et al. (2003) identified two ways that woody biomass increases in African savannas The first is through a reduction in fire intensity thro ugh increased grazing which reduces fuel for fires, and the second is decreased browsing that leads to woody biomass increases which eventually outcompete grasses, thus reducing the fuel available for fires to affect woody biomass. Both of these processes are occu rring simultaneously in the communal conservancies of this study region The grazing guild in the communal conservancies is now dominated by livestock, and the browsing guild has been reduce d through heavy poaching in the 1970s and 1980s (Chase and Griffin 2009), and through persecution Rodwell et al. 2000). W ith reduced browsing and increased grazing in communal conservancies I expect the re to be a reduction in the quantity and quality of herbace ous vegetation, while wo ody vegetation, given a competitive advantage by the alteration, results in the savanna being gradual ly transformed into woody scrub 1 At higher densities, cattle can reduce the root reserve nutrients of grasses through repeated consumption, leading to a l oss of more palatable (and often perennial) species (van Oudts hoorn 2004) which then a ffords woody biomass the opportunity to increase. I hypothesize that higher cattle density and lower fire frequency interact to lower the overall quality of grazing for 1 Grass species can be defined in terms of a qualitative assessment of their grazing value, which is determined based upon the amount of production of leaf material, the palatability of the grass, the nutritional value of the gr ass, and its growth vigour (how quickly the species can recover from being grazed), and the quality of the new plant material ( van Oudtshoorn 2004)
118 l ivestock, and the dominant woody shrub and tree species present would be those indicative of a pert urbed, woody encroached savanna ecosystem. To test this hypothesis, I studi ed vegetation communities in 29 transects in three Namibian communal conservancie s with different fire frequencies and cattle densities. I measured the characteristics of dominant grass species, and then compared this to fire frequency and cattle density at each of these sites. I measured s hrub and tree commun i ty diversity and abundanc e. T he characteristics of the more common and abundant species were examined in terms of the risk they pose to increasing woody bio mass through bush encroachment. Finally, the BIO ENV procedure (Clarke and Ainesworth 1993) was employed to determine whether fire frequency and cattle density were included as environmental variables which best described the patterns of shrub and tree communities in the vegetation samples. T he persistence of wildlife in these communal conservancies may be contingent upon the p ersistence of the floristic diversity in the savanna ecosystem (du Toit and Cumming 1999) and there is a need to measure the vegetation communit y composition in these conservancies to not only provide baseline data, but also to determine whether anthropog enic disturbances are having a deleteri ous effect on plant communities. There has been little vegetation monitoring within the communal conservancies to date, and this research helps to fill this gap in knowledge about vegetation community composition at a fine scale, allowing for the detection of changes in plant communities over time by possibly re measuring these transects in the future. This research will also add to our understanding of the effects of fire frequency and cattle density in semi arid sout hern African ec osystems, which is critical to our understanding of how changes in
119 these and other anthropogenic effects may alter vegetation communities which in turn affect the types of wildlife accommodated on the landscape. Methods Study Site The study site was in three communal conservancies along the Kwando River Kwandu, Mayuni and Mashi C onservancies in the Caprivi Strip of Namibia (Figure 2 1) with a few transects established just outside this area. The vegetation of the area has two broad categor ies : Mopane woodlands and Kalahari woodlands (Mendelsohn and Roberts 1997) Both of these are defined by the dominance of Mopane trees ( Colophospermum mopane (Kirk ex Benth.) J Lonard ) in the former, and varying dominant trees on Kalahari sands in the lat ter. Most of the transects I established are in the Kalahari woodlands and are dominated by Zambezi Teak ( Baikiaea plurijuga (Harms) ) with different shrubs and grasses dominating the understory (Mendelsohn and Roberts 1997) The differences observed in th e dominant tree type are related to edaphic factors, with well drained Kalahari sandy soil being dominated by Teak, while Mopane dominates on soil with a higher proportion of clay. The mean annual rainfall in the study site varies fro m 600 to 700 mm annual ly (Mende lsohn and Roberts, 1997). Sankaran et al (2005) as those savannas where disturbances are more important than mean annual rainfall in determining the tree grass rati o. Sampling D esign Twenty nine transect sites were established throughout the study region based on a system atic random sampling method where I would travel 1 5 km along tracks within the communal conservancies, flip a coin to determine which side of the track I would go,
120 and then start the transect 150 250 m away from the track. Community members who accompanied me provided environmental history information about the site, and if the previously cleared for agriculture a transect site wa s established. The site was marked using a GPS, and a 50 m length of rope was stretched away from the track The location of transect sites in relation to the conservancies as well as t he fire frequency and cattle d ensity are shown in Figures 4 1 and 4 2 r espectively Data C ollection Herbaceous 1 m 2 quadrats were set 3 metres away from the transect line, and spread every 3 metres along it. I recorded data for a total of 15 quadrats, 7 on one side of the transect line, while 8 were recorded on the other. Th e percentage cover of the most dominant grass was recorded in each quadrat Reference samples for g rasses that could not be identified on site were collected and identified later. W oody vegetation between 0.5 m and 3m height were classified as shrubs Shr ubs were recorded along the first 25 m of the transect line, and 2.5 m on either side of this line, giving an area of 125 m 2 Trees (any woody plant over 3 m high) were recorded the entire length of the transect line, and 5 m on either side of it (an area of 500 m 2 ). The species, height, number of stems, and type of coppicing were recorded. D iameter at breast height (dbh) of the thickest stem was recorded for trees At each of the transect sites, a hole was dug 10 20 cm in the ground at the 25 m mark along the transect line and a soil sample taken after mixing and stored in sealed Ziploc bags Each of these samples was then analysed by the Agricultural Laboratory of the Ministry of Agriculture in Windhoek, Namibia The percentage moisture in the sample, eff ective cation exchange capacity (CEC me/100g), exchangeable calcium, magnesium, potassium and sodium (me/100g), phosphorous, calcium, magnesium and
121 percentage, and per centage of sand, clay and silt in each of the samples were measured. Fire frequency for 1989 to 2005 was derived from GIS data of fire maps of the Caprivi region provided by Integrated Rural Development and Nature Conservation (IRDNC), one of the facilita ting non governmental organizations ( NGO s) of the Kwando, Mayuni and Mashi Conservancies. Fires were mapped each year from a variety of satellite instruments (Landsat Thematic Mapper ( TM ) Landsat Enhanced Thematic Mapper Plus (E TM +) Moderate Resolution I maging Spectroradiometer ( MODIS) ) and the map product (Figure 4 1) is a result of these mapping efforts over this period by a number of people, including Verlinden and Laamanen (2006) who used Landsat TM Quicklooks to map fire scars in the entire Caprivi region from 1989 to 2001. Cattle density was modelled using data derived from the Department of Veterinary Services of the Ministry of Agriculture in Katima Mulilo. The data w as taken from 12 animal crush pens used as vaccination points throughout the com munal conservancies. I employed the same methodology used by Mendelsohn and Roberts (1997) to model cattle density, where it was assumed that any cattle being recorded in these crush pens came from within a radius of 10 km (Figure 4 2) ArcGIS v.9.3 (ESRI Inc. Redlands, CA) was used for all spatial analysis employing the Spatial Analyst Tools extension with a search radius if 10 km and output cell size of 1 km 2 for modelling cattle density The altitude at each transect site was extracted from a digital el evation model (DEM) provided by the Ministry of Agriculture, Water and Forestry in Windhoek, Namibia. This DEM used the Shuttle Radar Topography Mission (SRTM) data which
122 has a spatial resolution of 3 arc seconds (about 90 m resolution) for the entire glob e. The SRTM dataset has been validated with independent data and has more accurate results over flatter terrain (Rodriguez et al. 2006, Berry et al. 2007), which characterizes the study site. Data A nalysis The grazing value and indication of veld conditi ons characteristic of the do minant grass species were evalu ated for each transect using the categories established by Mller and van Eck (2007), and supplemented by van Oudtshoorn (2004) and Quattrocchi (2006). These grazing values and veld condition value s for each species range from 1 to 3 (1 being the lowest grazing value or veld condition score) and are categorical values assigned to each species of grass based on several criteria, including the leaf production of the species, the palatability of the sp ecies, its nutritional value, growth vigour, and whether or not the species is indicative of overgrazed veld. The average grazing value of the transect site was then calculated by weighting the grazing value of each dominant species in each of the 15 quadr ats as well as the number of quadrats where grasses were absent, to come up with an overall value between 0 and 3 For example, if Eragrostris rigidior (with a grazing value of 2) was the dominant grass in 8 of 15 transects, and the remaining 7 transects had no grass, the grazing value overall would be ((2 x 8) + (0 x 7))/15 = 1.1 The same was done for calculating veld condition at each transect. I used Spearman rank correlation s to determine whether or not grazing value and veld condition were correlated with cattle density and fire frequency. The dominant shrub and tree species were determined for both community datasets and an examination of the characteristics of the most dominant species was
123 conducted in terms of their potential as bush encroaching species I also examined whether or not both communities were adequately sampled using species a ccumulation curves, determined the prevalence of each species by comparing their mean abundance to the number of plots they occurred in, and determined whether or not shrub and tree density was correlated with the number of species at a transect site. Ordination techniques were used to represent the numerous species and environmental variables in low dimensional space (Legendre and Legendre 1998; Zuur et al. 200 7). I first used a modified Mantel test called the BIO ENV procedure ( Clarke and Ainsworth 1993) a function in the vegan package (version 1.17 8) of R statistical software, which searches for the best combination of environmental variables to match this d issimilarity matrix to the dissimilarity matrix of the transect shrub data and tree data. Those variables chosen by the BIO ENV procedure are the ones which have the most influence over the patterns observed in the biotic variables. If fire frequency and c attle density are important determinants of shrub and tree composition, then I expect these two variables to be included in the result of the BIO ENV procedure. This procedure uses a Euclidean distance dissimilarity matrix for the environmental dataset, an d the Bray Curtis dissimilarity matrix for the biotic (community) dataset, and then calculates a correlation between the two matrices using the Spearman rank correlation ( w ) I then subjectively compared the ordination plots of the biotic datasets to the subset of environmental variables chosen through the BIO ENV procedure as recommended by Clarke and Ainsworth (1993). If the subset of environmental variables chosen throu gh the BIO ENV procedure and the biotic dataset is high, then I would expect there to be similarity in the distribution of transect sites in these two ordination
124 plots. The environmental dataset and shrub and tree community datasets were converted as follo ws to conduct the BIO ENV procedure: Highly correlated environmental variables were removed and variables included were transformed to make their distributions normal and unskewed (Clarke and Ainsworth 1993 Legendre and Legendre 1998 ). I used collinearity diagnostics to identify which environmental variables were high ly correlated with one another, using a standard conservative cut off value of 0.7 (which is somewhat subjective, but which has been used in other studies (Michalski and Peres 2007) and is als o recommended by Pallant (2007)) coupled with a priori knowledge of the study site to determine which environmental variables to include in the model This resulted in a reduction of the number of environmental variables included in the BIO ENV and ordinat ion analyses to 10: soil moisture (Moist), cation exchange capacity (CEC), exchangeable sodium (ExNa), sodium concentration (Nappm), pH, Easting, Northing, Altitude (Alt), fire frequency (Ffreq) and cattle density (Cadens). A summary of these 10 variables in terms of their descriptive statistics and collinearity diagnostics is shown in Table 4 1 and 4 2 respectively The transformations performed on these variables are shown in Table 4 3 While abundance data is often used in ordination analyses of communi ty data, Warwick (198 8) argues that the use of biomass is more meaningful. As such, the biomass of each tree in the datasets was estimated using the allometric equation developed by Netshiluvhi and Scholes (2001) T o balance the contribution of uncommon an d common species a 4th root transformation was performed on both the shrub densities and the tree biomass since there were at least 3 orders of magnitude difference between the most common and rare species in both the shrub and tree datasets whi ch is recom mended by Warwick 198 8 and Clarke and Warwick 2001 Since interpretation of the ordination plots for the community datasets and environmental datasets proved futile, t o make the ordination plots of both shrub density and tree biomass more interpretable the fire frequency and cattle datasets were converted into ordinal data divided into three categories Fire frequency was divided into low (1 to 6 of 16 years), medium (7 to 11 of 16 years), and high (12 to 16 of 16 years) categories, similar to those designa ted by Sheuyange et al. (2005); while cattle density was divided into low ( < 5 cattle/km 2 ), medium ( 5 to 10 cattle/km 2 ) and high ( >10 cattle/km 2 ) categories based on the distribution of cattle in the study region (the range of cattle densities for the tran sects was 0.44 to 14.2 ca t tle/km 2 ) These three categories
125 were then superimposed onto the NMDS ordination plots of shrub community data and tree community data, and the significance between the class centroids of these factor variables and the NMDS ordina tions was tested using the envfit function in the vegan package of R statistical software. This function returns a significance valu e of the centroids based on an R 2 value as the goodness of fit statistic. Results Herbaceous G uild 23 species of grasses we re identified as the dominant species in 25 of the vegetation transects (Table 4 4 ). The remaining 4 transects had a significant proportion of the quadrats (6 or more) with grasses which could not be identified. The most a bundant grass in the region was Di gitaria milanjiana (Rendle ex Stapf) ( 13 of 25 transects ), followed by Aristida stipitata (Hack. ex Schinz) ( 7 of 25 transects ) The mean grazing value calculated for the 23 species was 1.78 (out of 3), while the mean veld condition score was 2 (out of 3). Thus, overall, the species exhibited average grazing quality, with veld condition in a transitional phase. Overall, transects had an average grazing quality 1.6 (out of 3) and transitional veld condition (1.7 out of 3) Table 4 5 shows the results of Spe arman rank correlation between grazing value, veld condition, fire frequency and cattle density. Unsurprisingly, there was a strong positive correlation between grazing value and veld condition since most of the species with a particular grazing value ofte n had the same veld condition score (Table 4 4 ). Both grazing value and veld condition were negatively correlated with cattle density, and positively correlated with fire frequency.
126 Shrub G uild Forty five species of shrub from 15 families were found in t he transects. The 20 m ost abundant species are given in Table 4 6 Shrub species richness in the transects ranged from 1 to 13. A cumulative count of the number of species of shrub found verses the number of plots is shown in Figure 4 3 This species accum ulation curve for the shrub guild shows that the communal conservancies were sufficiently sampled as the curve approaches a vertical asymptote. The most common (found in most transects) shrub found was Baphia massaiensis (Taub) (20 transects), followed by Terminalia sericea (Burch. ex. DC.) (19) and Combretum collinum (Fresen.) (19). Few species over the entire study region had a mean abundance in excess of 0.1 individuals/m 2 A graph of shrub prevalence (Figure 4 4 ) shows that only one species occurring in m any transects has a high abundance, B. massaiensis while there are some specie s with high abundance present in few transects ( Gymnosporia senegalensis (Lam. ex. Loes.) Combretum elaeagnoides (Klotzsch) Markhamia zanzibarica (Bojer ex. DC.) K. Schum. ) Th e number of shrub species present in a transect was positively correlated with the total shrub density in th =0.74, n =29, p<0.001). B. massaiensis was significantly negatively correlated with cattle density (Spearm = 0. 48. n=20, p<0.05). T here was no other significant correlation between the more common species of shrub and fire frequency or cattle density. Tree G uild 36 species of tree from 16 families were found in the transects. The 20 most abundant species are given in Table 4 7 Tree species richness ranged from 1 to 13. A cumulative count of the number of species of tree found in each plot is shown in Figure 4 5 also showing a typical shape, with fewer new species being found as the number of
127 plots increased. The most common tree found was C. collinum (19 transects), followed by T. sericea (18) and B. plurijuga (13). Few species over the entire study region had a mean abundance in excess of 0.02 individuals/m 2 (200/hectare). A graph of tree prevalence shows that n one of these more common species has a high density, with the highest density found in one transect with Albizia anthelmintica (A.Rich) Brogn. (Figure 4 6 ). The number of tree species present in transects was positively correlated with the total tree densit =0.78, n =29, p<0.001). There was no significant correlation found between the more common species of tree and fire frequency or cattle density. Multivariate A nalyses Summary results of the BIO ENV models are shown in Table 4 8 The correlation between the dissimilarity matrices and the environmental variables which contribute to this model are also shown. The highest correlation between matrices was found for the model of tree community biomass with 7 variables contrib uting to the model (0.61). The correlation for the shrub density model is simila r at 0.59. F ire frequency and the transformed cattle density variable were included in the best model approximating the matrix of the shrub and tree communities meaning that th ey significantly contribute to the vegetation composition of both these guilds Given the correlations between the biotic matrices and environmental matrices, I expected that ordination plots would show similar distributions of the transect sites. The ordi nation plots of the best models however, wer e considerably different when compare d with one another which is attributed to the somewhat high stress value s of the community datasets (between 0.1 and 0.2, Table 4 8 ) coupled with the unexplained distributi on of the communities by the subset of environmental variables Stress is an indicator of the appropriate number of
128 dimensions to include in a NMDS model (Zuur et al. 2007), and values lower than 0.05 are ideal, and are indicative of an appropriate configu ration. While a 3 dimensional NMDS configuration would produce a smaller stress value, it would make interpretation of the ordination plot more difficult when compared with the plot of the environmental variables (Clarke and Ainsworth 1993) The environmen tal ordination stress values were below 0.1 for all models, which indicate s a good 2 dimensional configuration (Table 4 8 ) The superimposed plots of fire frequency and cattle density onto the shrub and tree NMDS ord inations are shown in Figures 4 7 and 4 8 Table 4 9 gives the R 2 values and their significance for both shrub and tree datasets. Shrub community density was statistically significantly distinguished by both fir e frequency (p < 0.1) and cattle density (p < 0.05), while tree community biomass wa s statistically significantly distinguished by fire frequency (p<0.01). Discussion The interaction of fire frequency and cattle density in the communal conservancy region affects the herbaceous vegetation layer in this study, with the grazing value and vel d condition negatively correlated with cattle density, and positively correlated with fire frequency The significant negative correlation of grass value and veld condition with cattle density, indicate that as the cattle density increases, grass species s witch from more vigorous types, to less vigorous ones In general, the grasses found at the transect sites were of average grazing value, and the veld was in a transitional condition (meaning they were on the decline) The most dominant grass species found ( D. milanjiana ) can be classified as either an annual or perennial grass (depending on soil quality), is of high grazing value, highly palatable, drought resistant, and can withstand
129 heavy grazing (Quattrocci 2006). Its ability to withstand high levels of grazing may be the reason it has been able to persist in the communal conservancies. Though listed as highly palatable, it should be noted that the quality of browse produced depends upon fertility of the soil and the age of the plant (Tropical Forages 20 11). A. stipitata in contrast to D. milanjiana is an annual tufted grass which occurs on deep sandy soil, is associated with disturbed areas and indicative of overgrazed veld (van Oudtshoorn 2004; Mller and van Eck 2007). While fire frequency was positi vely correlated with grazing value and veld condition, fire frequenc y generally declines in savanna s where grazers consume the herbaceous layer since this reduces the available fuel load (Scholes and Archer 1997; Hoffman n 1999; Holdo 2007) and so it is dif ficult to determine in this study whether or not fire frequency contributes to enhanced grazing value and veld condition, or is more an artefact of the relatively high densities of livestock observed in portions of the communal conservancies. Of all of the species of shrub found in the transects, B. massaiensis was both common and abundant in many of the transects. This species is typical of deep sandy soils in in regions receiving above 500 mm rainfall a nnually (Curtis and Mannheimer 2005), and has been re ported to have lower mortality in response to fire beyond the 2cm stem diameter size class where the bark becomes thicker and more fire resistant (Holdo 2005). Below t his size class, while the above ground biomass may be killed in a fire, the plant may per sist and coppice at ground level. B. massaiensis in this study did display ground coppicing, with a mean of 9 stems per plant (Table 4 6 ), but further study is needed to determine to what degree fire frequency in the region affects this If fire is exclude d from the environment, then B. massaiensis has the potential to grow more
130 resistant to fires, and contribute to the problem of bush encroachment given the increasing fire resistance of its bark as it grows (Holdo 2005) While B. massaiensis is classified as a bush encroaching species, t his shrub has also been cited as an important browse species for cattle during dry periods since it often occurs at browse height for cat tle (Verlinden and Dayot 2005). T. sericea was also found in many of the transect sites but was on average less abundant compared with B. massaiensis (Figure 4 4 ). It is widespread in Namibia, and is found in sandy soils where mean annual precipitation exceeds 150 mm (Curtis and Mannheimer 2005). This species has been cited as a woody encro aching species in overgrazed areas of southern Africa (Moleele et al. 2002; H ipondoka and Versfeld 2006), was found to have high allelopathic potential (Nakafeero et al. 2007), and may inh ibit the growth of other plants In this region, T. sericea trees fo rm the bulk of the biomass in places near to the Kwando River sometimes giving the appearance of a monodominant forest (pers obs) but the effect it has on the growth of other species in this study region warrants further investigation The third most com mon species of shrub, C. collinum is a common species found in the northeast of Namibia on sandy soils (Curtis and Mannheimer 2005). Most pastoralists in the Omaheke region of eastern Namibia consider C. collinum an encroaching species although it is also an important browsing species for cattle, especially in the driest periods of the year (Katjiua and Ward 2007). While C. collinum was the most widespread tree recorded in 19 transects, T. sericea was perhaps the most common and abundant species (Figure 4 6 ) T. sericea trees were found in 18 transects, and comprised over 40% of the trees in 12 of these transects. As mentioned, this species is widely regarded as a bush encroaching
131 species, and, because it is more fire tolerant than other species such as B. plurijuga (Zambezi Teak) (Holdo 2005; Gambiza et al. 2008) it may have the potential to spread in those parts of the conservancies exposed to more regular and high intensity fires T he classification provided by Mendelsohn and Roberts (1997) which categor izes much of the communal area into some type of B. plurijuga tree dominated vegetation may be a misnomer since of the 13 transects in which this species was present, only 3 of these had proportions over 40 %. B. plurijuga is found in north eastern Namibia i n regions over 500 mm annual rainfall and is of conservation concern since it has been overexploited for timber and other uses (Curtis and Mannheimer 2005). Frequent intense fire in Zambezi Teak woodlands may result in increased mortality of seedlings ( Gambiza et al. 2008) which can lead to a change in the dominant species present over time This may be reflected in the shrub dataset for this study, where on ly 28 teak shrubs were found in 8 transects Given the low number of teak plants in the shrub comm unity, and the low number of mature teak trees in our datasets, there is a need to investigate how this species can persist in the communal conservancies. Chidumayo et al. (1996) suggest that livestock grazing or early burning may help sustain savanna wood lands (since these have relatively little impact on woody plants), but conversely Gambiza et al. (2008) note that heavy grazing can increase seedling predation. Goheen et al. (2010) also noted that large herbivores can facilitate seedling and tree establis hment by reducing the grass cover for rodents which are the main seed and seedling predators at the Mpala R esearch Centre in central Kenya. The BIO ENV procedure conducted on both shrubs and trees show s moderate correlations between the three models (Tabl e 4 8 ). Both fire frequency and cattle
132 density were included in the shrub density model, where six of the ten variables produced the highest correlation between matrices. This means that both of these disturbance effects contribute to the distribution of s pecies of shrubs in these models Other variables included in the model were soil moisture, easting, altitude, and sodium concentration. These variables are associated with edaphic factors affecting vegetation communities, and interact with one another on the landscape (as well as disturbance regimes) to produce different vegetation communities. Soil moisture and sodium concentration, for example, is lower in soils with higher % sand content, since water percolates through sandy soils more taking sodium and other minerals with it, while altitude may be serving as a proxy for soil depth; higher elevations having a deeper Kalahari sand layer. Easting was included in the model as well, indicating that there were significantly different communities in the deeper sandy soil to the East, compared to those closer to the Kwando River. Since the tree biomass model had a higher correlation with the environmental dataset compared with the tree density model, I restrict my discussion to the results of the former. In add ition to those variables described already, CE C was also included in the tree biomass model, another edaphic factor contributing towards the correlation between the tree biomass community matrix and the environmental matrix. CEC was also lower in soils wit h a higher % sand, which is why it is included in the tree biomass model. I expected that the NMDS plots comparing the biotic datasets to their corresponding environmental datasets would show similarities in the distribution of the transect sites given the reasonably high correlations ( 0.6) but this was not the case Clarke and Ainsworth (1993) note that the comparison of community and environmental
133 plots is separate from the BIO ENV procedure itself, and in cases where the stress values are over 0.15 (as is the case with my biotic NM DS models Table 4 8 ) there is no guarantee that the 2 dimensional visualization of the community and environ mental datasets will match one another Still, much of the disagreement between the ordination plots may stem from the fact that while the environm ental variables did explain much of the distribution in the shrub and tree vegetation communities (with correlations there is still much of the composition of these communities left unexplained by the environmental variables in the models, which leads to complex and difficult interpretation. In lieu of matching biotic and environmental NMDS ordination plots, Fig ures 4 7 and 4 8 allow us to interpret the connectivity of different transects based upon fire frequency and cattle density categories While side by side NMDS ordination plots of the community and environmental datasets were difficult to i nterpret, the ce ntroids of Figures 4 7 and 4 8 show that many of the transects with similar fire and cattle categories were grouped together, with some overlapping more than others (for example the low cattle density category su perimposed on the NMDS for tree biomass ) C ollapsing discrete and continuous data into ordinal data may not be ideal; but in this case it serves as a heuristic tool demonstrating the connections between different fire frequency and cattle density regime s. This is corroborated by the R 2 values in Ta ble 4 9 which indicate that these different fire frequency and cattle density categories do contribute to the formation of the vegetation communities of shrub, and fire frequency to the vegetation communities of tree. The lack of a statistically significan t relationship between the centroids for cattle density categories with the tree biomass community
134 ordination may be attributed to the differential exposure of these shrub and tree guilds to herbivory by cattle, which are only able to feed on trees up till a certain height. This study shows that all three vegetation guilds in these communal conservancies were affected by fire frequency and cattle density While this serves as a useful evaluation of the effects of fire frequency and cattle density on differe nt v egetation guilds, there are several other factors which may be contributing to the vegetation patterns observed that are not accounted for in this research. While the frequency of fire in a savanna landscape is important, fire intensity has also been r ecognised as an important factor in the vegetation communit y dynamics of savannas (van Lang evelde et al. 2003). Increased fire intensity affect s shrub stem mortality, and more intense fires are related t o higher grass biomass (van Lang evelde et al.2003). T hus, inclusion of either fire intensity data or grass biomass data would perhaps lead to a better understanding of the changes taking place in these vegetation communities. Communities can be trained to estimate grass biomass by comparing photos of already calibrated grass biomass estimates, to the grass present at the transect sites. While I used altitude as a proxy for soil depth, the depth of the sandy soil layer at each transect site would be a better representation of this variable since this may affec t the ability of grasses to compete with woody vegetation for soil moisture. Cattle density could also be modelled more accurately through the use of dung counts, and in order to more accurately determine fire frequency, community members can visit transec ts regularly for on the ground verification that a particular site has burnt as identified via satellite imagery. I anticipate that the baseline dataset and analyses developed through this research will allow the communities to assess the changes in their conservancies to
135 ensure that they are better able to manage the effects of fires, cattle, and other disturbances in their communities Summary The interaction of fire frequency and cattle density in the communal conservancy region affects the herbaceous ve getation layer in this study, with the grazing value and veld condition negatively correlated with cattle density, and positively correlated with fire frequency. The most dominant species of shrub and tree in the communal conservancies have all been recogn ized as potential bush encroaching species, with B. massaiensis being the most dominant species of shrub, and T sericea the most dominant species of tree Multivariate analyses included both fire frequency and cattle density in models correlating shrub com munity densities and tree community biomass, but adjacent ordination plots of these NMDS ordinations were difficult to interpret. Converting the fire frequency and cattle density datasets into an ordinal scale helped with interpretation, and the centroids of these categories were significantly separated from one another for fire frequency and cattle density in the shrub NMDS ordination, while the centroids of fire frequency categories were significantly separated in the tree NMDS ordination. This means that differential levels of fire frequency and cattle density do affect the shrub community composition, while fire frequency affects tree community composition. Inclusion of additional variables such as fire intensity, grass biomass and soil depth, and improv ed accuracy of the variables already included, may help improve the accuracy of models of community composition in the future. Regular measurement of these vegetation transects in the future will help communal conservancies track changes occurring in their vegetation communities by comparing them with the baselines established through this study.
136 Figure 4 1 Location of transect sites in the study region in relation to the communal conservancies and the fire frequency from 1989 to 2005. Data is contribut ed by Integrated Rural Development and Nature Conservation. High = area burnt 12 to 16 of 16 years; Medium = area burnt 7 to 11 of 16 years; Low = area burnt 1 to 6 of 16 years. Inset: study area in relation to surrounding countries.
137 Figure 4 2 Locatio n of transect sites in the study region in relation to the communal conservancies and the cattle density. Data is modelled from cattle crush pens and were collected by the Department of Veterinary Services of the Ministry of Agriculture, Water and Forestry in Katima Mulilo.
138 Table 4 1 Descriptive statistics of the environmental variables included in the multivariate analyses Variable N Minimum Maximum Mean Std. Deviation Moist ( %) 29 0 0.53 0.20 0.13 CEC (me/100g) 29 0.40 7.02 2.42 1.70 ExNa (me/100g) 2 9 0 0.04 0.02 0.01 Nappm (ppm) 29 11 30 16.17 4.15 pHw 29 5.12 7.24 6.09 0.52 UTM Easting (m) 2 9 746991 769319 NA NA UTM Northing (m) 29 7999718 8047663 NA NA Alt (m) 29 962 997 977.76 10.74 Ffreq (yr) 29 1 15 8.14 3.68 Cadens (cattle/km 2 ) 29 0.44 1 4.20 5.32 3.58 soil moisture (Moist), cation exchange capacity (CE C), exchangeable sodium (ExNa), sodium concentration (Nappm ), pH (pHw) Easting, Northing, Altitude (Alt), fire frequency (Ffreq) and cattle density (Cadens) Table 4 2 Correlations among the environmental variables used in the BIO ENV procedure. None of the correlations is above 0.7, thus warranting their use in this procedure. N = 29 for all bivariate correlations. See Table 4 1 footnote for variable details Moist CEC ExNa Nappm pHw East ing Northing Alt Ffreq Cadens Moist 1.00 0.69 0.41 0.26 0.31 0.32 0.01 0.49 0.18 0.08 CEC 1.00 0.32 0.23 0.48 0.28 0.01 0.38 0.15 0.02 ExNa 1.00 0.31 0.11 0.14 0.01 0.24 0.10 0.03 Nappm 1.00 0.05 0.35 0.18 0.14 0.30 0.27 pHw 1. 00 0.25 0.24 0.25 0.22 0.34 Easting 1.00 0.20 0.24 0.37 0.23 Northing 1.00 0.50 0.26 0.45 Alt 1.00 0.20 0.03 Ffreq 1.00 0.55 Cadens 1.00
139 Table 4 3 Shapiro Wilk test results for normality in the enviro nmental datasets. Significance values above 0.05 required no transformation. The normalization transformation is also shown See Table 4 1 footnote for variable details Variable Statistic df Significance Transformation Alt 0.929 29 0.052 None required Ca dens 0.889 29 0.005 S qrt CEC 0.862 29 0.001 S qrt Easting 0.963 29 0.385 None required ExNa 0.93 29 0.054 None required Ffreq 0.964 29 0.415 None required Moist 0.946 29 0.142 None required Nappm 0.883 29 0.004 Log10 Northing 0.961 29 0.355 None requ ired pHw 0.962 29 0.375 None required Table 4 4 Dominant gras s species found in 25 transects, with the number of transects and quadrats in which they were identified, and the species grazing value and veld condition Species # Transects # Quadrats Gra zing value Veld condition Aristida adscensionis 2 14 1 1 Aristida congesta subs congesta 2 16 1 1 Aristida stipitata 7 40 1 1 Brachiaria nigropedata 1 1 3 3 Digitaria eriantha 3 3 3 3 Dactyloctenium giganteum 3 13 1 1 Digitaria m ilanjiana 13 78 3 3 Digitaria sanguinalis 3 12 1 1 Eragrostis cimicina 1 1 1 2 Eragrostis jeffreysii 1 1 1 2 Eragrostis pallens 3 6 1 2 Eragrostis rigidior 7 7 2 2 Melinis repens subs repens 1 3 2 2 Panicum fluviicola 2 6 2 2 Panicum kalaharense 2 2 2 2 Pogonarthria squarrosa 6 22 1 2 Schizachyrium exile 1 1 1 1 Setaria pumila 1 1 2 2 Sporobolus fimbriatus 3 4 3 3 Schmidtia pappophoroides 3 12 3 3 Setaria sphacelata var sphacelata 1 4 3 3 Tricholae na monachne 5 13 1 2 Tristachya superba 3 9 2 2
140 Table 4 5 Spearman rank correlations between grazing value, veld condition, cattle density and fire frequency for the herbaceous guild N = 25 for grazing value and veld condition; N = 29 for cattle density and fire frequency. Variable Grazing value Veld Condition Cattle Density Fire frequency Grazing value 1 0.982** 0.46* 0.437* Veld condition 1 0.43 4 0.467* Cattle density 1 0. 401 Fire frequency 1 ** p < 0.01 (2 tailed); p < 0.05 ( 2 tailed ) Table 4 6 Top 20 species of shrub counted in the study area. The proportion of the total number of shrubs (1682), mean height and mean number of stems is given. These 20 species accounted for 95% of all the individual shrubs enumerated. S pecies code Species name Number of plants Proportion of all plants Mean height (m) Mean number of stems Bm Baphia massaiensis 563 33.47 1.10 9.04 Bap Bauhinia petersiana 168 9.99 1.08 13.27 Ce Combretum elaeagnoides 149 8.86 1.65 8.12 Cc Combretum col linum 101 6.00 1.15 6.81 Mz Markhamia zanzibarica 97 5.77 0.84 2.96 Ts Terminalia sericea 93 5.53 1.28 6.39 Dc Dichrostachys cinerea 77 4.58 1.27 6.69 Cg Croton gratissimus 62 3.69 0.88 10.73 Gs Gymnosporia senegalensis 51 3.03 1.23 3.39 Gf Grewia fl avescens 47 2.79 1.20 13.26 Cp Colophospermum mopane 44 2.62 1.48 7.39 Bp Baikiaea plurijuga 28 1.66 1.55 8.64 Af Acacia fleckii 26 1.55 1.29 8.35 Baf Burkea africana 26 1.55 1.05 5.46 Pn Philenoptera nelsii 17 1.01 1.64 9.12 Ci Combretum imberbe 13 0.77 1.95 9.85 Ae Acacia erioloba 11 0.65 1.07 7.09 Cen Combretum engleri 9 0.54 1.44 6.56 An Acacia nigrescens 8 0.48 0.90 8.13 Co Combretum odeale 8 0.48 1.81 3.75
141 Table 4 7 Top 20 species of tree counted in the study area. The proportion of th e total number of tree (1158), mean height mean DBH, and mean number of stems is given. These 20 species accounted for 96% of all the individual trees enumerated Species code Species name Number of plants Proportion of all plants Mean height (m) Mean DB H (cm) Mean number of stems Ts Terminalia sericea 342 29.53 4.93 7.43 3.02 Cc Combretum collinum 117 10.10 5.36 10.81 2.24 Dc Dichrostachys cinerea 83 7.17 4.79 4.32 4.19 Bp Baikiaea plurijuga 67 5.79 8.12 22.63 3.19 Pn Philenoptera nelsii 66 5.70 4.7 1 7.57 2.38 Aan Albizia anthelmintica 61 5.27 5.57 5.61 5.10 Cm Combretum mossambicense 55 4.75 5.33 10.02 5.89 Cp Combretum psidioides 51 4.40 3.92 4.56 3.10 Ae Acacia erioloba 43 3.71 6.04 11.49 1.93 Cg Croton gratissimus 35 3.02 4.77 6.09 2.23 Af Acacia fleckii 28 2.42 4.18 5.32 6.86 Bm Baphia massaiensis 27 2.33 3.55 4.00 9.74 Baf Burkea africana 22 1.90 5.79 12.06 3.23 Pan Pterocarpus angolensis 22 1.90 5.57 12.86 1.32 An Acacia nigrescens 21 1.81 5.55 5.62 1.05 Co Combretum odeale 16 1.38 5 .31 12.75 1.31 Ci Combretum imberbe 15 1.30 4.14 6.30 9.67 Ce Combretum elaeagnoides 13 1.12 3.18 2.08 8.69 Ch Combretum hereroense 12 1.04 5.71 13.25 2.58 Pa Peltophorum africanum 11 0.95 4.44 9.86 2.09
142 Figure 4 3 Cumulative count of the number of species of shrub observed in 29 transects Figure 4 4 Mean abundance of each shrub species verses the number of transects in which the species was found. References to species codes may be found in Table 4 6 Only the 15 most abundant species are lab elled
143 Figure 4 5 Cumulative count of the number of species of tree observed in 29 transects Figure 4 6 Mean abundance of tree species verses the number of transects in which the species was found. References to species codes may be found in Table 4 7 Only the 15 most abundant species are labelled
144 Table 4 8 Results of the BIOENV procedures for shrubs, trees and environmental variables. The highest correlations are shown in the shrub density verses environmental variables and tree species biomass and environmental datasets. The stress values for both the community and environmental ordinations are also shown. Model Variables included Co rrelation w Stress value for community NMDS Stress value for Environmental NMDS Shrub species density vs Environmental Moist Easting Alt Ffreq LogNappm Sqrtcatdens 0.59 0.16 0.059 Tree species density vs Environmental Alt Ffreq SqrtCEC LogNappm 0. 51 0.19 0.008 Tree species biomass vs Environmental Moist Easting Alt Ffreq SqrtCEC LogNappm Sqrtcatdens 0.61 0.17 0.073
145 Figure 4 7 Superimposition of categorical variables onto NMDS plots of shrub community density. A) fire frequency, B) cat tle density. Labels are transect sites, while g reen, orange and red correspond to low, medium, and high categories of both fire frequency and cattle density The axes represent the two axes extracted in this NMDS solution, with a stress value of 0.16. The centroids of both categorical variables were significantly separated from one another (Table 4 9 )
146 Figure 4 8 Superimposition of categorical variables onto NMDS plots of tree community biomass. A) fire frequency, B) cattle density. Labels are transec t sites, while g reen, orange and red correspond to low, medium, and high categories of both fire frequency and cattle density The axes represent the two axes extracted in this NMDS solut ion, with a stress value of 0.17. The centroids of the fire frequency categorical variable were significantly separated from one another but not cattle density (Table 4 9 )
147 Table 4 9 Results of the categorical data analysis of fire frequency and cattle densities with NMDS ordinations of shrub density and tree biomass. Th is indicates that cattle density had the most influence on shrub density, while fire frequency had the most influence on tree biomass Categorical variable R 2 Significance Shrub with fire frequency 0.1611 0.056* Shrub with cattle density 0.1939 0.019** Tree with fire frequency 0.2642 0.005** Tree with cattle density 0.0992 0.217 Significance codes: 0.01***, 0.05**, 0.1*
148 CHAPTER 5 CONCLUSION Overall Findings The expansion of CBNRM institutions has been accompanied by an increase in the developm ent of monitoring systems which are designed to give communities the informa tion they require to manage their resources sustainably As such, researchers and practitioners have been examining these monitoring systems to identify shortfalls, and to build up on these to make them better. My main focus was an examination of the efficacy of community based monitoring in Namibia in the form of the Event Book System (EBS) In the first chapter, I outline the rationale for this research; review the historical conte xt of CBNRM in Namibia and of community based monitoring in Namibia and discuss the importance of monitoring vegetation communities in these communal conservancies After a major decline in wildlife populations in Namibia during the 1970s, community based natural resource management was given n ew life in the Kunene region ( Jones and Murphree 2001). Through NGO and government support, communities were able to gain rights to some of the wildli fe resources, and this example eventually contributed to the development of landmark legislation in 1996 for the development of communal conservancies (MET 1995). The increasing area and people under this new common property institution meant that governme nt and NGOs did not have the means to carry out monitoring of the natural resources themselves, which led to the development of a comm unity based monitoring program known as the Event Book System (EBS) (Stuart Hill et al. 2005).
149 While many studies focus u pon data quality issues with community based monitoring systems by comparing them to more professional monitoring systems (Noss 1999; Bray and Sch ramm 2001; Genet and Sargent 2003 ), few focus on the uses and decisions made by the communities conducting the se monitoring systems (Danielsen et al. 2005 b ). In Chapter 2 I focused on the ways the EBS was being used by different stakeholders, particularly the conservancy committees whom the monitoring system was designed to benefit. I used a combination of semi st ructured interviews with the conservancy committees and data collectors in Kwandu, Mayuni, and Mashi Conservancies, as well as interviews with key informants who help implement the EBS in the communal conservancies throughout Namibia. To identify whether p erceptions of wildlife patterns matched those exhibited by the EBS, I compared perceived rankings of the most problematic species with tho se of the actual data collected. This revealed that these three conservancies are familiar with this aspect of the EBS but that these may be still influenced by traditional beliefs. The perceived uses of the EBS by respondent type revealed that the EBS is being used mainly as a compliance monitoring tool (vertical accountability), is used in some respects for natural res ource management (where wildlife monitors decide where to focus mitigation of human wildlife conflict and where communi ty resource monitors sustainably manage craft resources), and is used in marketing the conservancy to trophy hunters and donors. There is little incentive for horizontal accountability within the monitoring system, which corresponds to other research findings which showed that most of the community did not have sufficient information on the natural re source trends (Collomb et al. 2010). Tha t being said, the community resource monitors (CRMs) appear better at being accountable to the
150 community, since this is built into the design of their monitoring protocol. CRMs also ch require more resources), and the link between the work they do and the benefits derived to the community is more tightly organized compared to the work of the wildlife monitors ( WMs ) In Chapter 3, I examine whether or not underlying spatial and tempor al patterns of human wildlife conflict can be identified using the EBS. Wild animals may be regarded as pests to those who share their range by destroying crops, livestock, and infrastructure (Naughton Treves 1998; Dublin and Hoare 2004; Warner 2008). At t he same time people come into conflict with wildlife by destroying or fragmenting their habitat, through poaching, and by persecuting them as pests (Parker and Graham 1989; Fernando et al. 2005; Hazzah et al. 2009; Inskip and Z immermann 2009). Managing hum an wildlife conflict requires an understanding of the factors which contribute to this problem, and valuable information may come from community based monitoring systems. There is thus the need to evaluate these moni toring systems in this respect. This cha pter examines the use of the Event Book System data from Namibia to determine which factors contribute to conflict with wildlife in three communal conservancies along the Kwando River in Caprivi. I combine the community monitoring data with satellite, moon phase, and GIS data to examine temporal and spatial patterns of crop and livestock raiding. I found significantly less crop and livestock raiding incidents during the full moon compared with the new moon and incidents of crop raiding were related to plant productivity with a lag of 48 days. Crop raiding was significantly related to the percentage of crops, location, and human population density,
151 while livestock raiding was significantly related to location, human population density and distance to househol ds. The zero inflated negative binomial models of crop and livestock raiding were able to detect hotspots of human wildlife conflict, but were less robust at modeling the magnitude of the problem. The EBS provides sufficient data to explain the temporal an d spatial trends in both elephant crop raiding and livestock raiding by hyena, leopards and lion, and stakeholders should make use of this valuable information to help mitigate these problems. In Chapter 4, I investigated in the Kwandu, Mayuni and Mashi co nservancies to determine how fire frequency, cattle density and edaphic factors interact to affect vegetation community structure. Fire frequency and cattle density are important anthropogenic effects which shape the community composition in African savann a ecosystems (Dublin et al. 1990; Moleele and Perkins 1998; Sheuyange et al. 2005). Since most of the benefits of communal conservancy establishment are derived from wildlife (NACSO 2010), understanding how anthropogenic factors act upon the vegetation in these communal conservancies is important as wildlife depend upon the vegetation to persist. The EBS focuses mainly upon wildlife monitoring, and little attention is paid to changes which may be taking place with the vegetation, so this chapter fill s this gap in knowledge, and provide s the communities with valuable baseline data to determine changes in vegetation composition in the future I used edaphic variables taken from soil samples, fire frequency data from 1989 to 2005, and cattle density data modele d from crush pens in the communal conservancies to examine the variables which contributed to the patterns of vegetation observed from sampling the herbaceous, shrub and tree vegetation guilds. I found that the vegetation communities
152 were being affected by both fire frequency and cattle density. Grazing value and veld quality was negatively correlated with cattle density, and positively correlated with fire frequency. The dominant shrub in the communal conservancies was B aphia massaiensis which is classifie d as a bush encroaching species (Verlinden and Dayot 2005). Terminalia seric ea was also found to be both abundant and widespread as a shrub, and was the most dominant tree species in all the transects. T. sericea has also been classified as a woody encroac hing species in overgrazed regions of southern Africa (Moleele et al. 2002, Hipondoka and Versfeld 2006). The patterns of shrub and tree communities were affected by cattle density and fire frequency when examin ed using ordination techniques, with both fir e frequency and cattle density being included in models predicting the most influential variables contributing to the patterns of shrubs and trees observed in the communal conservancies. Chapters 2, 3 and 4 in this dissertation all tie together to tell us something about the efficacy of community based monitoring in Namibia. Chapter 2 measures how well the information is being used in natural resource management, and how the information is being transferred to different stakeholders. Chapter 3 then examine s the information itself to determine whether or not underlying spatial and temporal patterns of human wildlife conflict are revealed. Chapter 4 then fills a gap in monitoring of the natural resources in the study region by the EBS by examining the distrib ution of different vegetation communities in the communal conservancies and determining the effects of fire frequency and cattle density on these conservancies These three chapters uncover some of the shortcomings of this monitoring system, while also ill ustrating the value of the information being collected in this community based monitoring system.
153 Event Book System Shortcomings Over the course of my research I was able to find many of the shortcomings of the EBS itself. Here I review these, and suggest ways that they may be resolved. These issues relate to matters of information transfer, sustainability, accuracy, and adaptability. It is clear from this and other studies (Child and Barnes 2010; Collomb et al. 2010) that the communal conservancies I exami ned are not adequately communicating with their constituents with respect to natural resource trends and are not sufficiently financially accountable to them. While Child and Barnes (2010) contend that representational forms of governance can be successful at managing natural resources, limited participation and access to the benefits of conservancy formation will eventually compromise sustainability. NACSO (2010) have noted that the lack of horizontal accountability may be resolved through increased govern ment regulation and supervision of conservancies by facilitating NGOs. Many conservancies in Caprivi have more recently taken action against their elected committees, firing committee members who are not performing, and in some cases removing them complete ly. The problem with firing of the entire committee is that it leaves a power vacuum, which may be more detrimental to the sustainability of the conservancy in the long run (NACSO 2010). Some efforts have been made in the Caprivi to make the conservancies more financially accountable by having professional accountants come in to develop some measure of control over their finances. In terms of dissemination of financial and other information, Anabeb and Orupembe Conservancies in the northwest of Namibia have experimented with sub dividing their conservancy into blocks where they hold quarterly meetings with members inside the blocks, and have a pre annual general meeting ( AGM ) meeting to set the agenda in the AGM and come to consensus on the most pertinent is sues they
154 face (NACSO 2010). This has led to greater communication with members, and the efficient running of the AGM, a model which is being experimented with in other conservancies. Just as the conservancy is mandated to show government they are monitori ng and managing their natural resources sustainably, I believe that incentives must be instituted to ensure the membership of the conservancy are getting information on a more regular basis. Since many community based monitoring systems become less intense or cease entirely when the initiating support entity leaves (Topp Jorgensen et al. 2005; Garcia and Lescuyer 2008), it is important to understand what factors contribute to their successful implementation and sustainability. Though 20 of the 59 conservanc ies are able to cover their own operating costs (NACSO 2010), and some believe that they have the experience to carry on monitoring on their own; it appears the EBS would ultimately fail without some form of external support. One key informant identified a trial conducted in Kwandu, Mayuni and Salambala conservancies from January to March of 2008 where the conservancies independently managed their event books without NGO assistance and, when audited; more errors were found within the system. The independent auditing of the EBS by NGOs or the MET was cited as an important aspect for the successful persistence of monitoring the natural resources within these communities. Without independent auditing, the quality of the data being collected would be called into question. In order to gain legitimacy with higher level decision makers, any information collected by communities should be shown to possess few errors, and the communities collecting the data need to show that they are competent and serious about data co llection (Garcia and Lescuyer 2008). As such, in the absence
155 of independent audits of the EBS, it is unlikely that any serious effort may be made to incorporate information being collected at the local level into policy making at higher levels of governanc e. Although independent auditing of the EBS does correct data the community has entered, how accurate is the actual data being collected? To date, there has been no comprehensive study which examines the accuracy of the data being collected within the EBS. For example, is the human elephant conflict (HEC) reported in Kwandu Conservancy an accurate depiction of what is taking place, or are these incidents inflated in order to gain NGO and government sympathy and perhaps attract more donor money to the conser vancy? If the levels of HEC reported are accurate, are most of the incidents of crop damage actually involving consumption by elephants, or are they more incidents where an elephant walking through crop fields inadvertently damages crops? The only way to a nswer these questions would be to set up an independent study examining elephant behavior in the conservancies comparing this to the EBS data. This approach, however, adds the cost of community based monitoring, and may foster resentment on the part of th ose carrying out the EBS. Another cause for concern for the EBS is that although its materials have been standardized throughout Namibia, the variability in the regional accuracy of the data may be called into question, making countrywide comparisons diffi cult. The EBS, for example, may be more accurate in the Caprivi region where there are higher levels of literacy, and where conservancies are on average smaller and thus easier to monitor compared with the lower literacy levels of the Kunene region of nort h western Namibia, with larger conservancies. Conservancies in the Kunene range between 348 to 4073 km 2 with
156 population densities ranging from 0.01 to 2.48 people/km 2 while those in Caprivi range from 73 to 930 km 2 with population densities ranging from 1 to 20.5 people/km 2 (NACSO 2010). It should be noted that while some of the key informant respondents question ed the accuracy of the EBS data, analyses such as those conducted in Chapter 3 illustrate that the use of an independent measure of plant productiv ity (NDVI) did show a distinct pattern with levels of elephant crop raiding, as did the frequency of crop and livestock raiding with respect to moon phase. I believe it is unlikely that the community would possess the capacity to fabricate data to match th ese trends, which does lend some credence to the quality of the data. At the community monitoring level, it may be more important to gain some idea of the trends exhibited, than to place excessive focus on data accuracy which might delay management action. Event Book System Recommendations As mentioned in the previous section, the EBS system continues to evolve as stakeholders identify shortcomings. Community based monitoring systems which are valued beyond the communities conducting monitoring contribute towards their persistence (Garcia and Lescuyer 2008) and their ability to influence larger scale policy contributes towards community empowerment (Fraser et al. 2006). While I acknowledge that the focus of the EBS is not on making it more amenable to use b y outside organizations, allowing for adaptation and interpretation by others may be beneficial. The analyses I conducted in Chapter 3 illustrate the value in the spatial and fine temporal data the conservancies are collecting, but, given the level of educ ation and access to resources needed to conduct such analyses, it is unlikely that communities will have the capacity to do these themselves. The onus is on scientists to analyze this data and then transform it into a digestible form for use by the communa l
157 conservancies. This can be difficult, since academic research objectives do not necessarily focus upon the adaptation of research findings towards benefiting communities, but may focus upon publication in peer reviewed journals Researchers may depend up on practitioners in non governmental organizations to translate their findings into management action. The research in chapter 3 illustrates an example of how such a partnership might work, where I was able to work with the NGOs in my study region to devel op a method of converting the spatial and temporal EBS data into maps which the communities can use to better manage human wildlife conflict. These efforts have extended beyond the Kwandu, Mayuni and Mashi Conservancies, and have now included mapping of pr oblem animal incidents in the Balyerwa, Impalila, Salambala, Sikunga and Wuparo Conservancies. The challenge of analyzing the spatial data in the EBS is that it would require a great deal of effort on the part of NGOs to record it, transcribe it, and then utilize the data and feed the results back to the communities in a timely fashion. This can be a cumbersome task, and by the time the results get back to the community, their utility would have diminished. Still, such analyses would help communities deter mine whether or not any management interventions they conduct are having the desired outcomes. In an ideal situation, stochastic events recorded in the EBS would be translated into a GIS database as they occurred, creating instant feedback for NGOs and com munities allowing them to respond quickly to these events. Another challenge to the EBS and other community based monitoring systems is that for the most part it appears that they are used for compliance monitoring, and few management decisions are being m ade using the data. This may improve given time, but NGOs, researchers and government should work with
158 the monitoring data to find novel ways of exploring problems within the conservancy so that the communities can experiment with different management stra tegies. Significance of the Research This research provides important insights into the operation of a community based monitoring system that has spread throughout Namibia. Through this research I was able to evaluate how well the monitoring system is gen erating knowledge which is being incorporated into management of three communal conservancies in Namibia. I have illustrated the uses and decisions being made by the communal conservancies based in part on the data collected by the EBS, and have demonstrat ed the conditions that foster more horizontal accountability of the data to the community membership. The design and implementation of community based monitoring systems can build upon this research so that they can be successful at helping communities man age their own natural resources. I have also demonstrated the value of using community based datasets to model human wildlife conflict, and have identified variables which are significant in the temporal and spatial distribution of elephant crop raiding an d livestock raiding, adding to the body of knowledge on these phenomena. The success of conservation initiatives will undoubtedly lead to increased interaction of humans and wildlife as the wildlife and human populations increase, and it is important for u s to find ways to mitigate conflict for there to be sustained success. Through collection and analysis of vegetation community and environmental data, I also added to the body of knowledge regarding the patterns of vegetation community which may be influen ced by fire frequency and ca ttle density in African savanna ecosystems. This research also has several broader impacts. Much of the information derived through this project has already been fed back to the facilitating NGOs in the region,
159 some of which ha s been disseminated to the communal conservancy management, and some of which has already been used in research looking at livestock raiding in these conservanc ies. All products of the analyse s will be made available to the communal conservancies in the re gion, which may make them better able to manage their natural from the information being generated throughout this project, which will ultimately aid them in natural resou rce management in the future.
160 APPENDIX EVENT BOOK SYSTEM SE MI STRUCTURED Q UESTIONNAIRE Questions to ask conservanc y committee members in Caprivi. The School of Natural Resources and the Environment at the University of Florida is conducting an independe nt research study about the communal conservancy system in Namibia. In order to understand the effectiveness of the communal monitoring system in the communal conservancies, research assistants are conducting face to face interviews with communal conservan cy members in Namibia. As a member of the conservancy committee, it is important that we gather information from you regarding the conservancy. Privacy is a key principle of this survey, so names and other identifying information will not be asked of you. There are no right or wrong answers, and most importantly candid and honest answers are most useful. If you have any questions about this survey please feel free to contact Luke Rostant, 081 331 8626 Lozi: Sikolo se si pahami sa Florida (USA) si ka be si e zize lipatisiso mwahala lipukelezo za Namibia kuamana ni sebelezo ya zona.ka ku kona ku utwisissa litaba/ misebelezo mo lizamaela kaufela ku ka ketiwa batusi ba ba ka toloka kapa kunga litaba kwa batu mwa lipukelezo za ku ba buzaka lipuzo ka mifuta ya ona inge ba italimezi. Sina ha luli limembala za katengo ka lipukelezo ki kwa butokwa hahulu kuli lu fumane bupaki/ misebelezo kuamana ni lipukelezo za luna. Ha luna ku zibahaza mabizo a batu mwa lipatisiso ze ni ka muta.mi lu ka lumelela maikuto kaufela a bat u ku sina kunga ya kiya butokwa kapa ye maswe. mi haiba muna ni maikuto kuamana ni patisiso ye mu lizeze muhala kwa nombol o ye (081 3318626) Luke Rostant
161 Conservancy: Village: Date and Time: Age: Sex: Position: Years in committee position: Other position s held prior to this on the conservancy committee: Period spent in this position/positions: What is your role in the conservancy committee? How long has your area been gazetted as a conservancy? What is the event book system?
162 When did yo u begin using the event book system? Which modules do you now monitor using this system (please tick all those that apply)? Others (please specify):___________________________ __________________________
163 When or how often do you monitor each of these modules? Is there anything else you think you should be monitoring? What do you do with the data? Who uses the data you collect using the event book system (please tick al l that apply)? Non Other (please specify):___________________________________________________ What do these people use the data for? Ordinary people ____ ________________________________________________ Communal conservancy managers__________________________________________ MET_________________________________ _____________________________
164 Donors________________________________________________________ ____ Re searchers____________________________ ___________________________ Non governmental organizations (N Other:_________________________________ ___________________________ Which of these animals cause the most problems i n your conservancy? Elephants Duikers Bushpigs Warthogs Springhare Hippos Crocodiles Hyaena Lions Wildogs Cheetahs Monkey Porcupine Baboon Jackals Other_________________________________ ____________________________
165 How often do you have meetings where you d iscuss the event book data? Once a month Once every two months Other (please specify):_____________________________________________________ Who attends these meetings? Where are the meetings held? Who owns the event book data? Do you feel proud about the d ata you are collecting using the event book? Do you think it is important to monitor natural resources? If yes to the above question, why do you feel it important to monitor natural resources.
166 List any decisions that you think have resulted directly or in part from the event book system: Please choose the most appropriate response according to your opinion: Strongly Agree Neutral Disagree Strongly agree disagree 1) The conservancy system has been beneficial 2) The conservancy system is more for foreigners 3) The conservancy system should be abolished 4) Restricting access to some resources in the conservancy is a good idea 5) The relationship between the community and the conservancy management is good 6) The conservancy management is doing a good job managing
167 the conservancy 7) The relationship between neighbouring conservancies is good 8) I feel proud that I am a part of this conservancy 9) It is better for the community to manage natural resources rather than outside people 10) We cannot manage natural resources properly unless we monitor them 11) Natural resources have improved since we became a conservancy 12) wild animals have increased since we became a conservancy 13) wild animals have caused more problems since we became a conservancy
168 LIST OF REFERENCES Adams WM (2004). Against extinction: the story of conservation. Earthscan, London Adams WM, Hulme D (2001a) Conservation and communities: changing narratives, policies and practices in A frican conservation. In: Hulme D and Murphree MW (eds) African wildlife and livelihoods: the promise and performance of community conservation. James Currey, London Adams WM, Hulme D (2001b) If community conservation is the answer in Africa, what is the qu estion? ORYX 35(3):193 200 Agarwal D, Gelfand A, Citron Poust, S (2002) Zero inflated models with application to spatial count data. Environ Ecol Stat 9(4):341 355 Balmford A, Bruner A, Cooper P et al (2002) Economic reasons for conserving wild nature. Sc ience 297(5583):950 953 Balmford A, Whitten T (2003) Who should pay for tropical conservation, and how could the costs be met? ORYX 37(2):238 250 Barnes RFW, Dubiure UF, Danquah E et al (2007) Crop raiding elephants and the moon. Afr J Ecol 45(1):112 115 Berry PAM, Garlick JD, Smith RG (2007) Near global validation of the SRTM DEM using satellite radar altimetry. Remote Sens Environ 106(1): 17 27 Blanc JJ, Barnes RFW, Craig CG et al (2007) African elephant status report 2007: an update from the African ele phant database. Occasional paper of the IUCN Species Survival Commission, no. 33. Gland, Switzerland Bowen Jones E, Entwistle A (2002) Identifying appropriate flagship species: the importance of culture and local contexts. ORYX 36(2):189 195 Bray G, Schra mm H (2001) Evaluation of a statewide volunteer angler diary program for use as a fishery assessment tool. N Am J Fish Manage 21(3):606 615 Brown C, and Peinke D (2007) Activity patterns of springhares from the Eastern Cape Province, South Africa. J Zool 2 72(2):148 155 conserve. In Conservation and the environment in Namibia 2004/2005. Ministry of Environment and Tourism, Windhoek Bruner AG, Gullison RE, Rice RE et al (2001) Effectiveness of parks in protecting tropical biodiversity. Science 291( 5501):125 128
169 Caprivi Senior Management Forum (2007) WWF/SDC Project Technical Progress Report Caprivi. 1 July 2007 to 3 June 2008. Integrated Rural Development and Nature Conserv ation (IRDNC), Windhoek Chase MJ, Griffin, CR (2009) Elephants caught in the middle: impacts of war, fences and people on elephant distribution and abundance in the Caprivi Strip, Namibia. Afr J Ecol 47(2):223 233. Chidumayo EN, Gambiza J, Grundy I (1996) Managing Miombo Woodland. In: Campbell BM (ed) The Miombo in transition: woodlands and welfare in Africa. CIFOR, Bogor Child B (2004) Parks in Transition. Biodiversity, rural development, and the bottom line. Earthscan, London Child B, Barnes G (2010) The conceptual evolution and practice of community based natural resource management in southern Africa: past, present and future. Environ Conserv 37(3):283 295 Child G (1995) Wildlife and people: the Zimbabwean success. Wisdom Foundation, New York Ciechanowsk i M, Zajac T, Bitas A et al (2007) Spatiotemporal variation in activity of bat species differing in hunting tactics: effects of weather, moonlight, food abundance, and structural clutter. Can J Zool/Rev Can Zool 85(12):1249 1263 Clarke KR, Ainsworth M (19 93) A method of linking multivariate community structure to environmental variables. Mar Ecol Prog Ser 92(3):205 219 Clarke KR, Warwick RM (2001) Change in marine communities: an approach to statistical analysis and interpretation. 2nd edition. PRIMER E, P lymouth Collomb J, Mupeta P, Barnes G, et al (2010) Integrating governance and socioeconomic indicators to assess the performance of community based natural resources management in Caprivi (Namibia). Environ Conserv 37(3):303 309 Craig C (1996) ELESMAP pro ject: final technical report (Unpublished report to USFWS). Namibian Nature Foundation, Windhoek Curtis B, Mannheimer C (2005) Tree atlas of Namibia. National Botanical Research Institute, Windoek Danielsen F, Burgess M, Funder M et al (2010a) Taking stock of nature in species rich but economically poor areas: an emerging discipline of locally based monitoring. In: Lawrence A (ed) Taking stock of nature; participatory biodiversity assessment for policy, planning, and practice. Cambridge University Press, Ca mbridge
170 Danielsen F, Burgess N, Jensen P et al (2010b) Environmental monitoring: the scale involvement. J Appl Ecol 47(6):1166 1168 Danielsen F, Burgess N, Balmford A (2005a) Monitoring matters : examining the potential of locally based approaches. Biodivers Conserv 14(11):2507 2542 Danielsen F, Jensen AE, Alviola PA et al (2005b) Does monitoring matter? A quantitative assessment of management decisions from locally based monitoring of protected areas. Biodivers Conserv 14(11):2633 2652 Danielsen F, Burgess ND, Balmford A et al (2009) Local participation in natural resource monitoring: a characterization of approaches. Conserv Biol 23(1):1 42 Darwall WRT, Dulvy NK (1996) An evaluation of the suita bility of non specialist volunteer researchers for coral reef fish surveys. Mafia Island, Tanzania A case study. Biol Conserv 78(3):223 231 Dublin HT, Hoare RE (2004) Searching for solutions: the evolution of an integrated approach to understanding and m itigating human elephant conflict in Africa. Hum Dim Wildlife 9(4):271 Dublin HT, Sinclair ARE, McGlade J (1990) Elephants and fire as causes of multiple stable states in the Serengeti Mara woodlands. J Anim Ecol 59(3):1147 1164 Durbin J, Jones BTB, Murphr ee MW (1997) Namibian community based natural resource management programme. World Wide Fund for Nature, Gland du Toit, JT and Cumming DHM (1999) Functional significance of ungulate diversity in African savannas and the ecological implications of the sprea d of pastoralism. Biodivers Conserv 8(12):1643 1661 Fernando P, Wikramanayake E, Weerakoon D et al (2005) Perceptions and patterns of human elephant conflict in old and new settlements in Sri Lanka: insights for mitigation and management. Biodivers Conserv 14(10):2465 2481 Fraser E, Dougill A, Mabee W et al (2006) Bottom up and top down: analysis of participatory processes for sustainability indicator identification as a pathway to community empowerment and sustainable environmental management. J Environ Ma nage 78(2):114 127 Gambiza J, Campbell BM, Moe SR et al (2008) Season of grazing and stocking rate interactively affect fuel loads in Baikiaea plurijuga harms woodland in northwestern Zimbabwe. Afr J Ecol 46(4):637 645 Garcia CA, Lescuyer G (2008) Monitori ng, indicators and community based forest management in the tropics: pretexts or red herrings? Biodivers Conserv 17(6):1303 1317
171 Genet K, Sargent L (2003). Evaluation of methods and data quality from a volunteer based amphibian call survey. Wildl Soc Bull 31(3):703 714 Goheen JR, Palmer TM, Keesing F et al (2010) Large herbivores facilitate savanna tree establishment via diverse and indirect pathways. J Anim Ecol 79 (2) : 372 382 discussion paper based on the Civil Society Conference on Land Reform in Namibia. Windhoek Graham, M (2006) Coexistence in a land use mosaic? Land use, risk and elephant ecology in Laikipia District, Kenya. Dissertation, University of Cambridge Griffin P, Griffin S, Waroquiers C et al (2005) Mortality by moonlight: predation risk and the snowshoe hare. Behav Ecol 16(5):938 944 Guha R (1997) The authoritarian biologist and the arrogance of anti humanism: wildlife conservation in the third world. The Ecologis t 27(1):14 20 Hazzah L, Mulder M, Frank L (2009) Lions and Warriors: social factors underlying declining African lion populations and the effect of incentive based management in Kenya. Biol Conserv 142(11):2428 2437 Hedges S, Giunaryadi D (2009) Reducing human elephant conflict: do chillies help deter elephants from entering crop fields? ORYX 44(1):139 146 Hellier A, Newton AC, Gaona SO (1999) Use of indigenous knowledge for rapidly assessing trends in biodiversity: a case study from Chiapas, Mexico. Biodi vers Conserv 8(7):869 889 Higgins SI, Shackleton CM, Robinson ER (1999) Changes in woody community structure and composition under contrasting landuse systems in a semi arid savanna, South Africa. J Biogeogr 26(3):619 627 Hipondoka MHT, Versfeld WD (2006) Root system of Terminalia sericea shrubs across rainfall gradient in a semi arid environment of Etosha National Park, Namibia. Ecol Indicators 6(3): 516 524 Hoare R (1999a) Determinants of human elephant conflict in a land use mosaic. J Appl Ecol 36(5):68 9 700 Hoare RE (1999b) A standardized data collection and analysis protocol for the human elephant conflict situation in Africa. IUCN African Elephant Specialist Group, Nairobi Hoare RE, Du Toit JT (1999) Coexistence between people and elephants in African savannas. Conserv Biol 13(3):633 639
172 Hoffmann WA (1999) Fire and population dynamics of woody plants in a neotropical savanna: Matrix model projections. Ecology 80(4): 1354 1369 Holdo RM (2005) Stem mortality following fire in Kalahari sand vegetation: e ffects of frost, prior damage, and tree neighbourhoods. Plant Ecol 180(1):77 86 Holdo RM (2007) Elephants, fire, and frost can determine community structure and composition in Kalahari woodlands. Ecol Appl 17(2):558 568 Holdo RM, Holt RD, Fryxell JM (2009) Grazers, browsers, and fire influence the extent and spatial pattern of tree cover in the Serengeti. Ecol Appl 19(1):95 109 Inskip C, Zimmermann A (2009) Review Human felid conflict: a review of patterns and priorities worldwide. ORYX 43(1):18 34 Jackson TR, Mosojane S, Ferreira SM et al (2008) Solutions for elephant Loxodonta africana crop raiding in northern Botswana: moving away from symptomatic approaches. ORYX 42(1):83 91 Jones B (1999) Policy lessons from the evolution of a community based approach t o wildlife management, Kunene Region, Namibia. J Int Dev 11(2):295 304 Jones B, Murphree MW (2001) The evolution of policy on community conservation in Namibia and Zimbabwe. In: Hulme D, Murphree MW (eds) African wildlife and livlihoods: the promise and pe rformance of community conservation. James Currey Ltd, Oxford Jones B, Weaver LC (2009) CBNRM in Namibia: growth, trends, lessons and constraints. In: Suich H, ChildB, Spenceley A (eds) Evolution and innovation in wildlife conservation: parks and game ranc hes to transfrontier conservation areas. Earthscan, London Jones BTB, Mosimane AW (2000) Empowering communities to manage natural resources: where does the new power lie? Case Studies from Namibia. In: Shackleton S, Campbell B (eds) Empowering communities to manage natural resources: case studies from Southern Africa. Division of Water, Environment and Forestry technology, CSIR, Pretoria Junker J (2008) An analysis of numerical trends in African elephant populations. Dissertation, University of Pretoria Jun ker J, van Aarde RJ, Ferreira, SM (2008) Temporal trends in elephant Loxodonta africana numbers and densities in northern Botswana: is the population really increasing? ORYX 42(1):58 65 Kaartinen S, Luoto M, Kojola I (2009) Carnivore livestock conflicts: d eterminants of wolf ( Canis lupus ) depredation on sheep farms in Finland. Biodivers Conserv 18(13):3503 3517
173 Kagoro Rugunda G (2004) Crop raiding around Lake Mburo National Park, Uganda. Afr J Ecol 42(1):32 41 ons and realities of vegetation change and browse consumption in the northern Kalahari, Namibia. J Arid Environ 69(4): 716 730 dichotomies of participation. Ethics Place Enviro n 9(3):279 298 Leader Williams N, Albon SD (1988) Allocation of Resources for Conservation. Nature 336(6199):533 535 Legendre P, Legendre, L (1998) Numerical Ecology. 2nd ed. Elsevier, Amsterdam Leuthold W (1996) Recovery of woody vegetation in Tsavo Natio nal Park, Kenya, 1970 94. Afr J Ecol 34(2):101 112 Lima S (1998). Nonlethal effects in the ecology of predator prey interactions what are the ecological effects of anti predator decision making? Bioscience 48(1):25 34 Lima SL, Dill LM (1990) Behavioural decisions made under the risk of predation a review and prospectus. Can J Zool/Rev Can Zool 68(4):619 640 Linard C, Gilbert M, Tatem AJ (2010) Assessing the use of global land cover data for guiding large area population distribution modelling. GeoJourna l Linnell J, Swenson J, Andersen R (2001) Predators and people: conservation of large carnivores is possible at high human densities if management policy is favourable. Anim Conserv 4:345 349 Loarie S, Van Aarde R, Pimm, S (2009) Fences and artificial wa ter affect African savannah elephant movement patterns. Biol Conserv 142(12):3086 3098 Martin R (2005) Transboundary species project, background study, elephants. Ministry of Environment and Tourism and the Namibian Nature Foundation). Ministry of Environm ent and Tourism and the Namibian Nature Foundation, Windhoek Martin R (2006) The Mudumu North Complex, Wildlife Co management in the Kwando Area of Caprivi. Ministry of Environment and Tourism, Windhoek McLaren A, Cadman M (1999). Can novice volunteers pro vide credible data for bird surveys requiring song identification? J Field Ornithol 70(4):481 490 Mendelsohn J, Roberts C (1997) An Environmental Profile and Atlas of Caprivi. Directorate of Environmental Affairs, Ministry of Environment and Tourism, Windh oek
174 MET (1995) Wildlife management, utilisation and tourism in communal areas. Ministry of environment and Tourism, Windhoek Michalski F and Peres CA (2007) Disturbance mediated mammal persistence and abundance area relationships in Amazonian rainforest fr agments. Conserv Biol 21(6):1626 1640 Minami M, Lennert Cody CE, Gao W et al (2007) Modeling shark bycatch: the zero inflated negative binomial regression model with smoothing. Fish Res 84(2):210 221 Moleele NM, Perkins JS (1998) Encroaching woody plant s pecies and boreholes: is cattle density the main driving factor in the Olifants Drift communal grazing lands, south eastern Botswana? J Arid Environ 40(3):245 253 Moleele NM, Ringrose S, Matheson W et al (2002) More woody plants? The status of bush encroac 11 Mller MAN, van Eck J (2007) Grasses of Namibia. Ministry of Agriculture, Water and Forestry, Winhoek Mulonga S, Suich H, Murphy C (2003) The conflict continues: human wildlife conflict and liv elihoods in Caprivi. Directorate of Environmental Affairs, Ministry of Environment and Tourism, Windhoek Mumby P, Harborne A, Raines P, et al (1995) A critical assessment of data derived from Coral Cay conservation volunteers. Bull Mar Sci 56(3):737 751 Mu rwira A, Skidmore A (2005) The response of elephants to the spatial heterogeneity of vegetation in a Southern African agricultural landscape. Landscape Ecol 20(2):217 234 challenges. N amibian Association of CBNRM Support Organisations, Windhoek challenges in 2009. Namibian Association of CBNRM Support Organisations Windhoek. Nakafeero AL, Reed MS, Moleele N (2007) Allelopathic potential of five agroforestry trees, Botswana. Afr J Ecol 45(4):590 593 Naughton Treves L (1998) Predicting patterns of crop damage by wildlife around Kibale National Park, Uganda. Conserv Biol 12(1):156 168 Naughton Treves L, Holland MB, Bra ndon K (2005) The role of protected areas in conserving biodiversity and sustaining local livelihoods Annu Rev Environ Resour 30(1):219 252
175 Netshiluvhi TR, Scholes R (2001) Allometry of South African woodland trees. CSIR, Division of Water, Environment and Technology, Pretoria Ngene SM (2010) Why Elephant Roam. Dissertation, University of Twente Norton Griffiths M, Southey C (1995) The opportunity costs of biodiversity conservation in Kenya. Ecol Econ 12(2):125 139 Noss AJ (1999) Censusing rainforest game s pecies with communal net hunts. Afr J Ecol 37(1):1 11 Rodwell CE, Rodwell T, Rice M et al (2000) Living with the modern conservation paradigm: can agricultural communities co exist with elephants? A five year case study in East Caprivi, Namibia. Biol Conserv 93(3):381 391 tal Accountability in New Democracies. J Democr 9(3):112 126 Osborn FV (2004) Seasonal variation of feeding patterns and food selection by crop raiding elephants in Zimbabwe. Afr J Ecol 42(4):322 327 Ostrom, E (1998 ) A behavioural approach to the rational choice theory of collective action: Presidential address, American Political Science Association, 1997 Amer Polit Sci Rev 92(1):1 22 Overdevest C (2000) Participatory democracy, representative democracy, and the nature of diffuse and concentrated interes ts: a case study of public involvement on a national forest district. Soc Nat Resourc 13(7):685 696 Pallant J (2007) SPSS survival manual. 3rd ed. Mc Graw Hill, New York Parker ISC, Graham AD (1989) Elephant decline (part I) downward trends in African elep hant distribution and numbers. Int J Environ Stud 34(4):287 Paudel NS (2006) Protected areas and the reproduction of social inequality. Policy Matters 14:155 169 Poteete A, Ostrom E (2004) Heterogeneity, group size and collective action: the role of insti tutions in forest management. Dev Change 35(3):435 461 Pringle R (2008) Elephants as agents of habitat creation for small vertebrates at the patch scale. Ecology 89(1):26 33 Quattrocchi U (2006) CRC world dictionary of grasses: common names, scientific nam es, eponyms, synonyms, and etymology. CRC/Taylor and Francis, Boca Raton
176 Rammel C, Stagl S, Wilfing H (2007) Managing complex adaptive systems a co evolutionary perspective on natural resource management. Ecol Econ 63(1):9 21 Rangarajan M, Shahabuddin G (2006) Displacement and relocation from protected areas: towards a biological and historical synthesis. Conservat Soc 4:359 378 Robinson JG, Bodmer RE (1999) Towards wildlife management in tropical forests. J Wildl Manage 63(1): 1 13 Rode KD, Chiyo PI, Cha pman CA et al (2006) Nutritional ecology of elephants in Kibale National Park, Uganda, and its relationship with crop raiding behaviour. J Trop Ecol 22:441 449 Rodriguez E, Morris CS, Belz JE (2006) A global assessment of the SRTM performance. Photogramm E ng Remote Sensing 72(3):249 260 Sabato M, de Melo L, Magni E et al (2006) A note on the effect of the full moon on the activity of wild maned wolves, Chrysocyon brachyurus Behav Processes 73(2):228 230 Sankaran M, Hanan NP, Scholes RJ et al (2005) Determi nants of woody cover in African savannas. Nature 438(7069):846 849 Scholes RJ, Archer SR (1997) Tree grass interactions in savannas. Annu Rev Ecol Syst 28(1):517 544 Shackleto n C, Griffin NJ, Banks DI et al ( 1994 ) Community structure and species compositi on along a disturbance gradient in a communally managed South African savannah. Vegetatio 115 (2): 157 167 Sheuyange A, Oba G, Weladji RB (2005) Effects of anthropogenic fire history on savanna vegetation in northeastern Namibia. J Environ Manage 75(3):189 198 Shumway R (2006) Time series analysis and its applications: with R examples. Springer, New York Sitati NW, Walpole MJ, Leader Williams N (2005) Factors affecting susceptibility of farms to crop raiding by African elephants: using a predictive model to mitigate conflict. J Appl Ecol 42(6):1175 1182 Sitati NW, Walpole MJ, Smith RJ et al (2003). Predicting spatial aspects of human elephant conflict. J Appl Ecol 40(4):667 677 Smit IPJ, Asner GP, Govender N et al (2010) Effects of fire on woody vegetation s tructure in African savanna. Ecol Appl 20(7):1865 1875 Smith N, Kasiki S (1999) A spatial analysis of human elephant conflict in the Tsavo ecosystem, Kenya. IUCN African Elephant Specialist Group, Kenya
177 Stahl P, Vandel J, Ruette S et al (2002) Factors affe cting lynx predation on sheep in the French Jura. J Appl Ecol 39(2):204 216 Stuart Hill G, Diggle R, Munali B et al (2005) The event book system: A community based natural resource monitoring system from Namibia. Biodivers Conserv 14(11):2611 2631 Tatem A, Noor A, von Hagen, C et al (2007) High resolution population maps for low income nations: combining land cover and census in East Africa. PLoS ONE 2(12) Taylor R (2001) Participatory natural resource monitoring and management: implications for conservati on. In: Hulme D, Murphree MW (eds) African wildlife and livelihoods: the promise and performance of community conservation. James Currey Ltd, Oxford The Afripop Project (2010) afripop. http://www.clas.ufl.edu/users/atatem/index_files/AfriPop.htm Cited 15 July 2010 Topp Jorgensen E, Poulsen MK, Lund J et al (2005) Community based monitoring of natural resource use and forest quality in montane forests and miombo woodlands of Tanzan ia. Biodivers Conserv 14(11):2653 2677 Treves A, Karanth K (2003) Human carnivore conflict and perspectives on carnivore management worldwide. Conserv Biol 17(6):1491 1499 Tropical Forages (2011) Factsheet: Digitaria milanjiana Tropical forages, an intera ctive selection tool. http://www.tropicalforages.info/key/Forages/Media/Html/Digitaria_milanjiana.htm Cited 9 Mar 2011 van Kooten G (2008) Protecting the Afric an elephant: a dynamic bioeconomic model of ivory trade. Biol Conserv 141(8):2012 2022 van Langevelde F, van de Vijver C, Kumar L et al (2003) Effects of fire and herbivory on the stability of savanna ecosystems. Ecology 84(2): 337 350 van Oudtshoorn F (20 04) Guide to grasses of Southern Africa. 2nd ed. Briza, Pretoria Verlinden A, Dayot B (2005) A comparison between indigenous environmental knowledge and a conventional vegetation analysis in north central Namibia. J Arid Environ 62(1): 143 175 Verlinden A Laamanen R (2006) Long term fire scar monitoring with remote sensing in Northern Namibia: relations between fire frequency, rainfall, land cover, fire management and trees. Environ Monit Assess 112(1 3):231 253
178 Warner, MZ (2008) Examining human elephant conflict in southern Africa: causes and options for coexistence. Department of Earth and Environmental Studies, University of Pennsylvania Warwick RM (1988) Analysis of community attributes of the macrobenthos of Frierfjord Langesundfjord at taxanomic leve ls higher than species. Mar Ecol Prog Ser 46(1 3):167 170 Wittemyer G, Rasmussen H, Douglas Hamilton I (2007) Breeding phenology in relation to NDVI variability in free ranging African elephant. Ecography 30(1):42 50 Young K, Ferreira S, van Aarde R (2009) Elephant spatial use in wet and dry savannas of southern Africa. J Zool 278(3):189 205 Zuur A, Leno EN, Smith G (2007) Analysing ecological data. Springer, New York
179 BIOGRAPHICAL SKETCH Luke Rostant was born in the twin island Repu blic of Trinidad and Tob ago. He graduated from the University o f the West Indies with a joint Bachelor of S cience g eneral degree in May of 2000, where he received first class hon ors in his joint major of botany and zoolog y. During this time he was employed as a research assistant on a number of projects, in cluding an Earthwatch projec t examining the ecology of the Manicou C rab, Eudaniella garmani in Tobago; a project working with the rehabilitation of offshore islands of the North Sound of Antigua for repopulation with the Antigu an Racer, Alsophis antiguae and an examination of bat community ecology in the forests of the Victoria Mayaro Forest Reserve in Trinidad. He received a postgraduate scholarship from the University of the West Indies in 2001, where he did his Master of P hi losophy research on the Manicou C rab, examining its growth, maturity and activity patterns. He received a Fulbright Faculty Improvement Scholarship in January 2004, and began his doctorate in in August of that year, pursuing a degree in interdisciplinary e cology through the School of Natural Resources and Environment, in the Department of Geography at the University of Florida. His dissertation focused on an examination of a community based monitoring system in Namibia, where he examined how well the monito ring system was being utilized by the communities, examined the ability of the monitoring system to detect spatial and temporal patterns in crop and livestock raiding and examined the effects of fire frequency and cattle density on the vegetation in these communities He received his PhD in August 2011.