<%BANNER%>

Anthropogenic Burning in the Okavango Panhandle of Botswana: Livelihoods and Spatial Dimensions


PAGE 1

ANTHROPOGENIC BURNING IN TH E OKAVANGO PANHANDLE OF BOTSWANA: LIVELIHOODS AND SPATIAL DIMENSIONS By LIN CASSIDY A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2003

PAGE 2

Copyright 2003 by Lin Cassidy

PAGE 3

ACKNOWLEDGMENTS I would like to thank my committee members (Dr Christy Gladwin, Food and Resource Economics Department, Dr Michael Binford, Land Use and Environmental Change Institute, Department of Geography; and Dr Mark Brown, Center for Wetlands, Department of Environmental Engineering) for their support and guidance. Dean Stephen Humphrey, College of Natural Resources and Environment and Dr Pete Hildebrand, Food and Resources Economics Department gave me much encouragement. Thanks are also due to Dr Jane Southworth, Geography Department and Dr Mickie Swisher, Family, Youth and Community Science Department for assistance with data interpretation. I am extremely grateful to Dr Helen-Jane Armstrong and the Map and Imagery Library within the University of Floridas Marston Science Library for purchasing the satellite images used in this study. There are many people in Botswana whom I would like to acknowledge. Most important are the people of Mogotho, Sekondomboro, Samochima and Nxamasere who gave freely of their knowledge and time. I am grateful to Monty Montshiwa of Ecosurv (Pty) Ltd for translating the questionnaire, and to Hannelore Bendsen of the Harry Oppenheimer Okavango Research Centre for introductions, access to literature and general support. I am indebted to my research assistants, Charlie John, Manyima Manyima and Oteng Segakisi for their unflagging enthusiasm and meticulous data collection. iii

PAGE 4

I would like to express my gratitude to Steve Harpt and Bana ba Metsi School for giving us a base on the east side of the panhandle, which made life in the study area a lot richer and easier. I am also extremely grateful to Birgit and Reiner Khler and family, Robyn and Charles Sheldon, Eve and Hugh Murray-Hudson, and Jan and Eileen Drotsky for their hospitality. This thesis would not have been possible without funding from a Compton Fellowship, a Tropical Conservation and Development Fellowship and a Tropical Conservation and Development Field Research Grant. I greatly appreciate this support. iv

PAGE 5

TABLE OF CONTENTS page ACKNOWLEDGMENTS.................................................................................................iii LIST OF TABLES...........................................................................................................viii LIST OF FIGURES...........................................................................................................xi ABSTRACT.....................................................................................................................xiv CHAPTER 1 INTRODUCTION..........................................................................................................1 Statement of the Problem.............................................................................................1 Research Objectives.....................................................................................................1 Literature Review.........................................................................................................2 Indigenous Use of Fire in Wildlands............................................................................2 Fire in Wetlands...........................................................................................................4 Sustainable Livelihoods...............................................................................................5 Background and Research Setting................................................................................8 Botswana......................................................................................................................8 Study Area..................................................................................................................10 Focal Villages.............................................................................................................18 Plan of Study..............................................................................................................22 2 METHODS...................................................................................................................24 Spatial Assessment.....................................................................................................24 Definition of Study Area............................................................................................25 Location and Distribution of Key Vegetation Resources...........................................25 Location and Distribution of Burnt Areas..................................................................26 Identification of Village Resource Areas...................................................................32 Calculation of Geographic Extent of Resource and Burn Areas................................33 Socio-economic Assessment......................................................................................33 Extent of Reliance on Key Natural Resources...........................................................36 Variations in Reliance on Key Resources..................................................................40 Effect of Fire on Availability of Key Natural Resources...........................................42 Conflict Surrounding Burning....................................................................................24 v

PAGE 6

3 RESULTS.....................................................................................................................47 Extent and Distribution of Fire in the Panhandle and Key Resource Areas..............47 Key Resources Areas for Focal Villages....................................................................51 Burning in the Key Resource Areas...........................................................................55 Temporal Distribution of Fire....................................................................................60 Discussion Theme: Local Knowledge of Fire Behavior...........................................61 Reliance on the Wetlands Natural Resource Base.....................................................62 Spatial Variations in Reliance on Key Resources......................................................70 Wealth Variations in Reliance on Key Resources......................................................79 Gender Variations in Reliance on Key Resources.....................................................81 Ethnic Variations in Reliance on Key Resources.......................................................85 Temporal Distribution of Resource Collection..........................................................93 Discussion Theme: Reasons for Temporal Variation in Collecting Plant Resources....................................................................................................................96 Effect of Fire on Access to Key Natural Resources...................................................99 Effect of Fire on Grazing..........................................................................................100 Effect of Fire on Availability of Thatching Grass....................................................101 Effect of Fire on Availability of Reeds....................................................................102 Effect of Fire on Access to Palm Leaves, Water Lily Bulbs and Papyrus...............103 Effect of Fire on Access to Fish...............................................................................104 Effect of Fire on Access to Wildlife.........................................................................105 Effect of Fire on Livelihood Sustainability..............................................................106 Conflict Over Burning..............................................................................................124 Discussion Theme: Lack of Conflict Due to Common Purpose.............................124 Different Use of Panhandle Resources.....................................................................125 Timing of Fires.........................................................................................................127 Management and Consultation in the Use of Fire....................................................129 4 DISCUSSION.............................................................................................................133 Livelihoods Assessment...........................................................................................133 Vulnerability Context...............................................................................................133 Livelihood Assets.....................................................................................................134 Policies, Institutions and Processes..........................................................................138 Livelihood Strategies................................................................................................138 Livelihood Outcomes...............................................................................................139 Conclusions and Recommendations.........................................................................141 Extent and Distribution of Fire in the Panhandle and Key Resource Areas............141 Reliance on the Wetlands Natural Resource Base...................................................142 Affect of Fire on Access to Key Natural Resources................................................143 Conflict Over Burning..............................................................................................144 Recommendations for Further Study.......................................................................145 Recommendations for Improved Fire Management.................................................145 vi

PAGE 7

APPENDIX A HOUSEHOLD QUESTIONNAIRE..........................................................................148 B INFORMAL INTERVIEW TRANSCRIPTS............................................................171 Reed/Thatch Perspective...........................................................................................171 Livestock Perspective...............................................................................................184 Fishing Perspective...................................................................................................187 Water Lily Perspective..............................................................................................196 Palm Leaf Perspective...............................................................................................199 Non-User Perspective...............................................................................................202 Wealthy Man (owns Land Rover 110), Sekondomboro Village.................................202 Kgotla Meetings........................................................................................................203 Kgotla Meeting Sekondomboro Village......................................................................203 Kgotla Meeting Samochima Village............................................................................203 Kgotla Meeting Nxamasere Village.............................................................................204 LIST OF REFERENCES.................................................................................................206 BIOGRAPHICAL SKETCH...........................................................................................211 vii

PAGE 8

LIST OF TABLES Table page 1 Characteristics of focal villages.....................................................................................22 2 Contributions of spectral bands to principal components from subset of exposed ground image...............................................................................................................29 3 Pixel values for areas of low and high reflectance in exposed area subset....................30 4 Population and sample sizes of focal villages................................................................34 5 Average sizes and prices for plant resources and fish...................................................37 6 Values of goods and flows used to create wealth index variable..................................39 7 Wetland resources included in natural resources value.................................................40 8 Measures of association by data type.............................................................................41 9 Names of languages spoken by different ethnic groups................................................41 10 Size of village resource ranges....................................................................................51 11 Extent of fire within village resource areas.................................................................56 12 Comparison of reliance on different resources............................................................63 13 Household reliance on different resources...................................................................64 14 Proportional contribution of wetland resources to households by village...................70 15 Numbers of cattle owned by village............................................................................71 16 Thatching grass collected in 2001 by village...............................................................72 17 Reeds collected in 2001 by village..............................................................................73 18 Reed bundles sold in 2001 by village..........................................................................74 19 Households collecting palm leaves by village.............................................................74 viii

PAGE 9

20 Bunches of palm leaves collected in 2001 by village..................................................74 21 Households with basket-makers..................................................................................75 22 Source of palm leaves for basket-making....................................................................76 23 Collection of water lily bulbs by village......................................................................77 24 Households usually fishing by village.........................................................................78 25 Fish caught in 2001 by village.....................................................................................78 26 Fish sold in 2001 by village.........................................................................................78 27 Household size by village............................................................................................80 28 Gender distribution of household heads......................................................................82 29 Wealth index by gender of household head.................................................................82 30. Total household size by gender of household head....................................................83 31 No. of household members 15 years and over by gender of household head..............83 32 Bundles of thatching grass collected by household head.............................................84 33 Bundles of reeds collected by gender of household head............................................84 34 Proportional contribution of wetland resources to households in 2001 by ethnic group...........................................................................................................................86 35 Livestock ownership and use of floodplain for grazing in 2001 by ethnicity.............88 36 Bundles of thatching grass collected in 2001 by ethnicity..........................................88 37 Collection of palm leaves by ethnic group..................................................................89 38 Collection of water lily bulbs by ethnic group.............................................................90 39 Involvement in fishing by ethnic group.......................................................................91 40 Seasonal Calendar of Flood and Collection of Main Resources..................................95 41 Extent of resources burnt within village resource areas in 2001.................................99 42 Effect of social characteristics on no. of adults in the household..............................108 43 Effect of social characteristics on no. of wage jobs in household.............................108 44 Regression table for factors influencing wealth in 2001...........................................110 ix

PAGE 10

45 Relative contribution of fire to range of resources accessed in 2001........................112 46 Relative contribution of fire to amount of fish caught in 2001..................................114 47 Relative contribution of fire to amount of reeds collected in 2001...........................116 48 Relative contribution of fire to amount of thatching grass collected in 2001............118 49 Rrelative contribution of fire to numbers of baskets made in 2001...........................120 50 Relative effect of fire on proportional contribution of wetland resources to livelihoods in 2001...................................................................................................122 51 Reasons believed for why people set fires experienced in 2001...............................125 52 Resource affected by user held responsible for fire...................................................126 x

PAGE 11

LIST OF FIGURES Figure page 1 Sustainable livelihoods diagram......................................................................................7 2 Map of Botswana showing location of study area...........................................................9 3 Map of the study area showing location of focal villages.............................................11 4 Map of average annual rainfall isohyets relative to the Okavango and its major tributaries and basin......................................................................................................12 5 Representative views of vegetation in the Okavango panhandle...................................14 6 Use of wetland resources in the Okavango panhandle..................................................20 7 Savanna fire showing active fire front and burnt area...................................................27 8 Signature plots of the 15-class unsupervised classification of the October scene exposed area subset.......................................................................................................31 9 Survey variables expected to influence access to wetland resources............................45 10 Study area map.............................................................................................................48 11 Map showing main vegetation groups in vicinity of focal villages.............................49 12 Map showing extent of burning in 2001......................................................................50 13 Map of grazing and wetland plant resource areas for each focal village.....................52 14 Comparison of available wetland grazing for focal villages.......................................53 15 Comparison of available plant resources for focal villages.........................................54 16 Map showing graze resources burnt in the grazing areas of each focal village...........57 17 Proportion of grazing vegetation burnt in graze ranges...............................................58 18 Map showing papyrus /reed /thatch burnt in the plant collection areas of each focal village..........................................................................................................................59 xi

PAGE 12

19 Proportion of plant resources burnt in collection ranges.............................................60 20 Proportional contribution of wetland resources to household livelihoods...................64 21 Frequency distribution of thatching grass bundles collected in 2001..........................66 22 Distribution of fish caught and fish sold in 2001.........................................................69 23 Distribution of cattle ownership by village in 2001.....................................................71 24 Extent of water lily bulb collection by village in 2001................................................76 25 Percentage distribution of wealth over households......................................................80 26 Distribution of ethnic groups by village......................................................................85 27 Proportional contribution of wetland resources to household livelihood in 2001 by ethnic group..................................................................................................................87 28 Quantities of water lily bulbs collected in 2001 by ethnic group................................90 29 Numbers of fish caught and number sold in 2001 by ethnic group.............................92 30 Map showing change in open water between the 2001 wet and dry seasons..............94 31 Decision tree determining early or late collection of reeds and thatch........................98 32 Numbers of bundles of thatching grass collected by whether fire had been experienced or not in 2001.........................................................................................101 33 Similarities in reports of fire in thatching grass and reeds in 2001...........................103 34 Numbers of fish caught categorized by fire experienced or not in 2001...................105 35 Results of the Stage 1 multiple regression.................................................................109 36 Results of the Stage 2 multiple regression.................................................................111 37 Results of the first Stage 3 multiple regression.........................................................113 38 Results of the second Stage 3 multiple regression.....................................................115 39 Results of the third Stage 3 multiple regression........................................................117 40 Results of the fourth Stage 3 multiple regression......................................................119 41 Results of the fifth Stage 3 multiple regression.........................................................121 42 Results of the sixth Stage 3 multiple regression........................................................123 xii

PAGE 13

43. Proportion of 2001 fire events reported by month....................................................128 44. Differences in the role of fire in the livelihoods of rich and poor households.........140 xiii

PAGE 14

Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science ANTHROPOGENIC BURNING IN THE OKAVANGO PANHANDLE OF BOTSWANA: LIVELIHOODS AND SPATIAL DIMENSIONS By Lin Cassidy May 2003 Chair: Christina H. Gladwin Department: College of Natural Resources and Environment This thesis examines how illegal, anthropogenic wildland fire affects peoples access to and use of resources in the wetlands of the panhandle region of Botswanas Okavango Delta. The study focuses on two villages close to the main river channel, and two near the shallower floodplains of the panhandle area. It addresses four specific questions: What is the extent and distribution of fire in the Okavango panhandle in general, and in key resource areas in particular, in a given year? How much do households rely on wetland resources? How does fire affect the availability of these natural resources? Is there any conflict among people over burning? Two approaches to answering these questions are used. The first is a spatial analysis based on the interpretation of two Landsat ETM satellite images. One scene is from March 2001 at the end of the rainy season, which coincides with the peak of the annual wetland flood. This provides information on the availability of wetland resources. The other is from October 2001, at the end of the dry season, and provides the basis for xiv

PAGE 15

identifying burnt areas. The second part of the analysis is a livelihoods assessment based on a quantitative household survey and informal theme-based interviews. Information on the location, types, and quantities of resources used is examined in terms of household-level socio-economic attributes as well as whether fire occurred in the resource areas or not. Results show that in 2001, fire covered about 5% of the study area as a whole. More burning occurred in the resource areas of the villages near the floodplains than those near the main river channel. Wealthier households, those from the dominant ethnic groups, and those headed by men, tended to collect greater amounts of wetland resources. However, the proportional contribution of these resources to their total livelihood was significantly less than it was for poorer households with reduced socio-cultural status, who had few alternative livelihood strategies. A quarter of all households experienced fire in 2001; however fire appears overall to have a positive effect on the availability of wetland resources. In addition, socio-economic factors affect access to the resources more than fire does. There is little conflict due to burning because, although it is illegal, most people see it as a way of improving the resource base. The limited conflict that occurs is associated with poor timing, and the lack of warning that a fire is to be set. This in turn is due to the secretiveness that the illegality of the practice induces. xv

PAGE 16

CHAPTER 1 INTRODUCTION Statement of the Problem Every year, Africas subequatorial savannas experience huge fires that consume a total of approximately 1.3 million square kilometers (Valenti 2000). Fires in Botswanas savannas and wetlands are no exception. Uncontrollable because of their remoteness and intensity, fires often range for hundreds of kilometers, until they burn themselves out. Nearly all these fires are started by people, just as they have been for millennia (Pyne 1995). The vast size of the fires suggests extensive destruction, and certainly every year they consume resources on which many people depend. Aside from the potential environmental impacts, this apparent contradiction raises several questions about the social dimensions of anthropogenic fires. If fires really were a threat to the peoples use of natural resources, why would they set them? How do people view the role of fire in their livelihoods? Research Objectives Specifically, this thesis examines how fire affects peoples access to and use of resources in the wetlands of the panhandle region of Botswanas Okavango Delta. In this study the focus is narrowed to the Okavango panhandle because of the greater concentration of people and natural resources there. It provides a bounded study area, allowing fire management issues to stand out. Until now, reports on the extent of burning have been mainly anecdotal and speculative, with some estimates of around 75% of the total area of the Okavango Delta 1

PAGE 17

2 being burnt each year (Patterson 1976). While it seems clear that large areas of the Okavango panhandle, like the rest of the delta and other areas in Botswana, are burnt every year, little information exists on the frequency, extent, or distribution of burning. No data have been recorded about how fire affects peoples access to natural resources in the panhandle. In attempting to fill some of the information gaps, this thesis pursues the following questions: What is the extent and distribution of fire in the Okavango panhandle in general, and in key resource areas in particular, in a given year? How much do households rely on wetland resources? How does fire affect the availability of these natural resources? Is there any conflict among people over burning? Literature Review Indigenous Use of Fire in Wildlands Fires in wildlands are inherently neither good nor bad. The desirability of fire is a human evaluation of how it influences the environments ability to meet the varied needs of its inhabitants. People have been purposefully manipulating their environment with fire for as long as they have been able to control it (Pyne 1995). In many developing countries people continue to use fire to change their surroundings. This often leads to conflict between the local population concerned with their own immediate needs and the authorities who represent a constituency and who see the lack of control as a threat to the environment and peoples safety. And yet, as is discussed below, research shows that local people have a clear understanding of fire behavior and how to use it for specific purposes.

PAGE 18

3 A study of Aboriginal communities living in the Northern Territory of Australia documents the burning patterns caused by indigenous natural resource users in Kakadu National Park (Roberts 1995). This study showed the reasons why people burned, their methods for setting fire, and how the fires spread. Respondents showed clear preferences for smaller patchier fires over large intense burns, and identified times of year when burning would have less impact on the regeneration of plant species they used. Park rangers were able to learn from the Aboriginal knowledge of the effects of burning on the environment, and of the effect of season on burning. Similarly, in north-eastern Luzon, in the Philippines, fire is used by different local residents, such as farmers, ranchers and grass collectors, in ways that achieve specific objectives for given land use practices (Masipiquea et al. 2000). While they see fire as an important tool, they understand the implications of resource loss. Again, knowledge of the best season for certain types of burn was shown. Because optimal burning times vary for different land uses, conflicts arise because the fires spread beyond the specific resource area. An important addition to the many uses of fire that is noted here is that of a weapon of injustice, jealousy or revenge (ibid.). In sub-Saharan Africa, fire has been attributed to very specific aspects of pasture management (West 1965). In Botswana, too, oral traditions show that tribal chiefs used to regulate both the timing and location of veld fires. For example, fires could not be set in the same place in consecutive years. Burning could also not be done until late winter, after the thatching grass species had dropped their seeds, so ensuring the availability of grass the following year. However, under pressure from white settlers accustomed to European management practices, the colonial government convinced the chiefs that

PAGE 19

4 burning was destructive. By the 1930s it was considered an offense to start a fire, or to leave one burning (Schapera 1970). This restriction became further entrenched after independence, with the promulgation of the Herbage Preservation (Prevention of Fires) Act in 1978 (Government of Botswana 1978). This law, combined with the loss of power of the chiefs under the post-independence centralized government, has led to the setting of fires being done in secret and with no control. The Okavango panhandle has been inhabited for thousands of years, sparsely at first by the nomadic San peoples, with more intensive settlement occurring since the mid-18th century with the influx of different Bantu peoples, such as the Hambukushu, Bayei, Xereku, Batswana and Bakgalagadi (Afriye 1976, Tlou 1985). It is highly probable that people have been setting fires in the Okavango for as long as they have lived there. However, the short-lived nature of the Deltas wetland vegetation and the frequent reworking of the sediments mean that there is very little historic record. Fire in Wetlands Fires are an important ecological component of wetlands. Most wetlands show fluctuations in their moisture regimes, and where conditions support fire, marshy wetlands (into which category the Okavango would fit) can burn frequently (Lugo 1995). Experimental studies in wetlands in Florida show that wet periods tend to have a greater impact on plant composition and the availability of nutrients than does fire (Environmental Science and Engineering, Inc. 1994). The effects of fire on small mammal and bird populations, and on plant biomass and composition, in Florida wetlands tend to be short-term (less than 6 months), unless conditions are dry enough so that the ground burns (Wade et al. 1980).

PAGE 20

5 The number of small animals killed in Everglades sawgrass fires tends to be much higher in unusually dry years. However, fire appears to benefit some species through removing the dense accumulation of vegetation (ibid.). In the Kissimmee River marsh, research showed that the number of both species and individuals were greater on burned than on unburned plots. In addition, when fire was tied to changes in water levels, the production of fish and macroinvertebrates increased (ibid.). In the Okavango, research on the effects of fire has been limited to peat fires (Ellery et al. 1989, Gumbricht et al. 2001). This is partly because by burning the soil, peat fires bring about more permanent changes than do surface fires, and partly because these persist for longer periods, making their study is more feasible. Peat fires are associated with long-term reduction in flow of river channels, and resultant drying out of peat soils. In comparison to surface fires, peat fires develop slowly, may persist for several years, and are often not visible at the surface (Gumbricht et al., op. cit.). While the University of Munich has recently begun to study the effect of fire on vegetation in the seasonal areas of the Okavango Delta, no links have yet been made to how fire affects the livelihoods of the people who rely on the Okavango for many resources. Sustainable Livelihoods The concept of sustainable livelihoods arose as part of the sustainable development debate linking poverty and environmental degradation (World Commission on Environment and Development 1987). There was a growing awareness of the need to fully capture the perspective of the primary target of development objectives: the rural poor. It was felt that to be more effective, development approaches needed to become more people-centered. Definitions of poverty became broader and more multi-disciplinary. The concept of sustainable livelihoods encompasses and links the

PAGE 21

6 notions of capability, equity and sustainability. To clarify how these notions are linked, a working definition was proposed as follows: A livelihood comprises the capabilities, assets (stores, resources, claims and access) and activities required for a means of living; a livelihood is sustainable which can cope with and recover from stress and shocks, maintain or enhance its capabilities and assets, and provide sustainable livelihood opportunities for the next generation . . (Chambers and Conway 1992, pp 7-8). The implication of this is that livelihoods should be both environmentally and socio-economically sustainable. The concept of sustainable livelihoods has been taken up as a practical tool by various international organizations, which have refined the definition and prepared different versions of an analytical framework (Carney et al. 1999, Neefjes 2000). However, all are similar and contain the same three basic elements: An outcome, which is the sustainable livelihood, Strategies that rural people select to achieve this outcome, and The environmental and social context in which the livelihoods are located. For most of the approaches two additional elements are broken out from the context to be given specific emphasis: institutional processes and livelihood resources are highlighted (Carney et al. 1999). The UKs Department for International Developments (DfID) Sustainable Livelihoods Framework is shown diagrammatically in Figure 1. It shows the vulnerability context on the left that is, the macro-level environment within which people live, and over which they have little control. This may change suddenly, presenting stresses or shocks to household attempting to survive at the micro level. The diagram also shows the assets available to rural people. The pentagon should be seen as flexible in shape, stretching out to those points of capital that are in greater

PAGE 22

7 SUSTAINABLE LIVELIHOODS FRAMEWORK LIVELIHOOD ASSETS VULNERABILITY CONTEXTShocksTrendsSeasonality KEYH = Human Capital S = Social CapitalN = Natural Capital P = Physical CapitalF = Financial Capital LIVELIHOOD STRATEGIESHNFPS Influence & Access POLICIES, INSTITUTIONS & PROCESSESLevels of GovernmentPrivate LawsSector CulturePoliciesInstitutions LIVELIHOOD OUTCOMESMore incomeIncreased well-beingReduced vulnerabilityImproved food securityMore sustainable use of NR base SUSTAINABLE LIVELIHOODS FRAMEWORK LIVELIHOOD ASSETS VULNERABILITY CONTEXTShocksTrendsSeasonality KEYH = Human Capital S = Social CapitalN = Natural Capital P = Physical CapitalF = Financial Capital LIVELIHOOD STRATEGIESHNFPS Influence & Access POLICIES, INSTITUTIONS & PROCESSESLevels of GovernmentPrivate LawsSector CulturePoliciesInstitutions LIVELIHOOD OUTCOMESMore incomeIncreased well-beingReduced vulnerabilityImproved food securityMore sustainable use of NR base Figure 1. Sustainable livelihoods diagram. Redrawn from DfIDs diagram. (DfID 1999) http://www.livelihoods.org/info/guidance_sheets_rtfs. abundance, and shrinking inwards where a resource is in short supply (DfID 1999). The transforming structures and processes represent the administrative and legal context. These include informal and formal organizations, laws and policies. They can operate at all levels, from individuals to district level to the international arena (ibid.). For example, it is as important to understand who in the household controls the allocation of income as it is to understand the effect of the land use policies at individual, household and community levels. The strategies are the ways in which people meet or try to meet their goal of a sustainable livelihood. Strategies will vary according to social factors such as gender, class, and ethnicity, as well as according to spatial location. For example, the San have traditionally relied on hunting game while groups of Bantu origin have focused on keeping livestock. It is important to see how not only the types, but also the number, of

PAGE 23

8 livelihood strategies available to different people vary. People with more livelihood strategies may be more adaptable, and therefore less vulnerable to external shocks and stresses. Different strategies may also lead to competition and conflict. Background and Research Setting Botswana Botswana is a land-locked, semi-arid country in southern Africa, lying between 18 and 27 South and between 20 and 29 East. Much of the country comprises semi-arid savanna, with average annual rainfall varying from about 650mm in the northeast to 250 mm in the southwest. Climate characteristics specific to the study area are discussed in further detail below. It is sparsely settled, mainly because of its low rainfall and lack of surface water. The country has had a stable, multi-party democracy since its peaceful independence from the British Protectorate in 1966. Because of this it has attracted a lot of international NGO and donor support, particularly during the 1980s and 1990s. Diamonds provide about 80 percent of the countrys foreign earnings, and form the backbone of the economy, which as a result is highly centralized. Income from diamonds is used to subsidize rural livelihoods that to a large extent are based on subsistence agriculture. The traditional significance accorded to livestock ownership has supported a strong livestock industry, subsidized heavily by Government and marketed to the European Union through a preferential trade agreement. Most rural households grow some crops, but in much of the country in many years the yields are insufficient to meet annual household food requirements, and crop production is in almost all cases sustained only with central government assistance. Households grow maize, sorghum and millet,with melons and beans as intercrops.

PAGE 24

9 Figure 2. Map of Botswana showing location of study area.

PAGE 25

10 In striking contrast to most of Africa, there are no traditional markets in any of the settlements. This is primarily because there is very little specialization in livelihood strategies, with most households producing the same goods. Agricultural productivity is also low, so there is very little or no surplus to sell. All agricultural and natural produce is typically obtained for subsistence consumption. Some informal trade takes place between households, but this is typically on an ad hoc basis. Any other goods that a household needs to purchase are obtained from retail shops. Nearly all rural households rely directly on natural resources as part of their livelihoods. This is most noticeable in Ngamiland District, where the Okavango Delta is located. The Okavango system is one of the most important economic and ecological features of the country. As a Ramsar Convention Site wetland on tribal land, proper management that includes resident communities is vital to its long-term sustainability. As a permanent source of fresh water, it supports an abundance of plant and animal life that contrasts starkly with the rest of the semi-arid northwestern region of the country. Although the fresh water and concentration of resources of the Okavango Delta could accommodate far greater settlement, the presence of vector-borne diseases that affect both humans and livestock has until recently kept population levels low (Tlou, op. cit.). Study Area For the purposes of this research, the study area is defined as the wetland system of the Okavango panhandle. It comprises the area reaching from the Namibian border in the northwest to a point some 2.5 kilometers southeast of the village of Seronga a total straight-line distance of approximately 92 km. About 10 km south of this point the panhandle fans out to form the delta. In this study, the term panhandle will generally be used to refer to the study area.

PAGE 26

11 Figure 3. Map of the study area showing location of focal villages. The panhandle is the diagonal green feature, showing the Okavango river meandering from northwest to southeast. Bright green indicates dense, healthy vegetation. Note that since this is the dry season, vegetative growth is concentrated along the main channels of permanent water. Pink areas lining the edge of the panhandle are sandy areas associated with villages, heavy grazing and bare fields at the end of the dry season.

PAGE 27

12 Biogeophysical characteristics of the Okavango panhandle The Okavango panhandle is a broad trough formed by parallel geological fault lines between 12 and 15 km apart. These faults lie at right angles to the main fault lines that limit the distal end of the delta itself (Hutchins et al. 1976). Since this is the southern hemisphere, winters run from June to August. In Botswana they are dry and cool. The hot and moist summers are from October to April. Rain falls from November to March, peaking in January/February. The Okavango panhandle has a fairly arid climate, with an average annual rainfall of 560 mm. The rainy season coincides with the arrival of the Figure 4. Map of average annual rainfall isohyets relative to the Okavango and its major tributaries and basin. Figures are given in millimeters. The 500 mm isohyet passes through the Okavango delta just south of the panhandle. (Redrawn from Nicholson et al. 1997 and United Nations 2000.)

PAGE 28

13 annual flood wave fed by an average of 1500 mm/annum of rainfall in the Planalto Central highlands of Angola (Figure 4). The Okavango river carries an average of 11 billion m3 of water a year. The river is comprised of a main channel, which ranges from 75 to 250 m wide, and several side channels and lagoons. With the arrival of the floodwaters the river fills up the extensive floodplains that stretch between the fault lines. During the dry winter months the pulse of the flood moves down through the Delta, and the water level in the panhandle drops, drying out the outer reaches of the floodplains. Not only does the river loop tightly as it flows southwards, but it also meanders from one side of the panhandle to the other. This changes the water and vegetation characteristics close to the banks, with important consequences for peoples access to various resources, as is discussed below. There is a distinct change in vegetation moving outwards from the channel margins through the back swamp vegetation to the seasonal floodplains (Ellery et al. 2000). Dense stands of papyrus (Cyperus papyrus) flank the river, while reeds (Phragmites australis) and taller (thatching) grass (Miscanthus junceus) grow on slightly elevated, but flooded, land. Behind the papyrus, the floodplains are mainly covered in shorter aquatic grasses and sedges (Cyperus articulatus, C. denudata, Cladium mariscus, Panicum repens, inter alia). Small islands have formed in the panhandle where river channels have moved, leaving perched ridges of sand. These are covered primarily with phoenix palms (Phoenix reclinata). The floodplains are very important breeding grounds for native tilapia fish (mainly Oreochromis and Tilapia spp.). This is in part due to the slower flow of the water, greater cover, and decreased vulnerability to predation by tiger fish (Hydrocynus

PAGE 29

14 A B Figure 5. Representative views of vegetation in the Okavango panhandle. A) A permanent river channel, showing reeds in the foreground, and papyrus on the far bank. B) A seasonally wet floodplain. C) A seasonally flooded area where water lily bulbs are collected and where basket fishing is done. D) Overlooking an area of dense aquatic vegetation, taken from the sandveld bank. All pictures taken by L. Cassidy (June/July 2002).

PAGE 30

15 C D Figure 5. continued.

PAGE 31

16 vittatus). Mainly, however, the much higher productivity and ecological complexity provide more food for both adults and young (pers. comm., Roger Bills, J.L.B Smith Institute of Ichthyology, Rhodes University). Fish, reeds and thatching grass are key resources on which rural households rely both for subsistence use and small-scale trade (Mosepele 2001, Murray-Hudson et al. 1994). The panhandle floodplains are also important for the grazing of livestock and, to a lesser extent, wildlife such as sitatunga (Tragelaphus spekei), lechwe (Kobus leche) and hippopotamus (Hippopotamus amphibius). Socio-economic characteristics of the Okavango panhandle Land tenure in the study area, as with most of Botswana, is communal. With the centralization of government after independence, the authority of the chiefs has been weakened, and traditional rules of access and obligations have been eroded. Use of resources takes place under open access. Because population densities are low, the negative consequences of this are not obvious, and the environmental tragedy of the commons (Hardin 1968) is extremely localized around the larger settlements.1 Within the wetlands, the ability for individuals to maximize use of resources is further limited by the inaccessibility due to the deep channels and dense vegetation. There are approximately 19,000 people living in the study area (Central Statistics Office 2002). By far the majority are of Hambukushu origin. The Hambukushu are traditionally matrilineal (Larson 1980), and without the rigorous social structure of the Tswana that comprise 80% of the national population. Other ethnic groups present 1 Around larger villages in the study area, firewood and trees suitable for building poles may have been used up within a radius of 2 km. Each year, grazing within a similar area results in all grass cover being removed. For smaller villages, such as the four focused on in this study, the radius is probably less than 1 km.

PAGE 32

17 include the Bayei, Bugakhwe and Xanikhwe San, Bakgalagadi, Batawana and Ovaherero (Afriye 1976, van Hoof et al. 1991 and 1993). For all groups, the household is the primary economic unit. Households tend to comprise three generations, and are generally large. As with most of the other rural parts of Botswana, the proportion of women is relatively high due to out-migration of men in search of wage labor (van Hoof et al. 1991 and 1993). Class, as a definition of how a household meets its livelihood needs, is only beginning to emerge. That is, it is not possible to identify separate social classes based on a single economic activity. Most households pursue several livelihood strategies (Scott Wilson 2000). Wealth is usually determined by how many sources of income a household has, rather than which type. Household size to some extent defines labor availability, and the ability to diversify livelihood strategies (Cassidy 1997). There are very few wage employment opportunities in the study area, and most households receive cash in the form of remittances from absent household members. For this reason, reliance on the natural resource base is still high. The reason households need to maintain diverse livelihood strategies is to provide a buffer against the unpredictable conditions under which they live. However, the recent (1995/6) slaughter of all cattle in Ngamiland District to eradicate an outbreak of cattle lung disease has caused irreversible economic dislocations (Fidzani et al. 1999). Six years later, the district herd is still less than half its 1995 size (Scott Wilson, op. cit.), and it is only the larger herd-owners who have been able to restock, which has reinforced wealth disparities. Not only is cattle ownership a source of wealth in itself, but it is also critical to the ability of households to plough fields for crops. As a result, the proportional contribution of natural resources, particularly to poorer households, has

PAGE 33

18 increased. Much of the increased resource use is within the panhandle, where important foods and building materials are found. Key natural resources from the panhandle Fish are a large part of peoples diets. Some 65% of the population depend on fishing (Fidzani et al., op. cit.). House walls are constructed of reeds, then plastered with mud. Outdoor enclosures are also built of reeds, while roofs are made with thatching grass (Larson, op. cit.). Papyrus is used to make sleeping mats and baskets. Leaves from young mokola palms (Hyphaene ventricosa) are used for making baskets that have a range of uses. Water lily bulbs (Nymphaea capensis) and fruit from a variety of riparian trees are collected and eaten (Murray-Hudson et al. 1994). Peoples access to, and extent of use of, these resources are not determined by socio-economic factors alone. Distance and accessibility are limiting factors, and may determine whether people can collect for themselves, or may need to buy from others, or even do without the resource altogether. Anecdotal evidence suggests that collection of resources usually takes place within a regularly used collection area that reflects these limiting factors. Similarly, grazing of livestock within the panhandle occurs according to the availability of suitable fodder species and the maximum herding distance. Focal Villages Twelve villages are situated along the edge of the study area. Between these, cattle posts and field areas are scattered. None of the settlements are within the panhandle itself, they are all up on the sandveld, on the edge of the panhandle. Settlement is linear: the roads run parallel to the panhandle, and the villages are strung out along the roads. Only four villages were considered for this study: Mogotho, Sekondomboro, Samochima and Nxamasere. These were selected on the basis of their location, and to some extent on

PAGE 34

19 whether there was an indication that there had been fire in their collection and grazing areas in 2001. Two villages represent each side of the panhandle. Two are river villages that have the main channel pass close to their bank, while the other two are floodplain villages. Maps show that Nxamasere is a relatively drier area. Its floodplains contain several larger islands, and floods last a shorter period. The location of the settlements is shown in Figure 3. Basic demographic and development information is summarized in Table 1. Remoteness increases for the villages in the following order: Nxamasere, Samochima, Sekondomboro, Mogotho. The nearest major town and district headquarters is Maun, where alternative food and building supplies, as well as wage employment, can be obtained. By vehicle it is 3.5 hours (350 km) from Nxamasere, 4 hours (370 km) from Samochima, 6 hours (420km) from Sekondomboro and 6.5 hours (450 km) from Mogotho. However, by far the majority of people do not have vehicles, and traveling times on foot or by donkey cart are considerably longer. In addition, road access to the east bank villages of Mogotho and Sekondomboro is only possible after traveling north and crossing with the river ferry 5 km south of the Namibian border, and 4x4 vehicles are needed. This makes these two villages far more remote than distance alone suggests. Remoteness often affects settlement size, while settlement size is a criterion for levels of infrastructural development (Ngamiland District Council 1997). Remoteness is also believed to play a part in determining extent of reliance on natural resources, because of the lack of cash income opportunities and resource alternatives.

PAGE 35

20 A B Figure 6. Use of wetland resources in the Okavango panhandle. A) Thatching grass stored outside a reed enclosure. B) Hambukushu couple floating thatching grass down the main channel on a papyrus raft. C) Cattle being swum across the main river. D) Girls going basket-fishing. Picture A) taken by L. Cassidy, B) by R. Forrester, and C) and D) by Love Botswana (www.lovebotswana.org).

PAGE 36

21 C D Figure 6. continued.

PAGE 37

22 Table 1. Characteristics of focal villages. Village Mogotho Sekondomboro Samochima Nxamasere Panhandle side East East West West Ecosystem type River Floodplain River Floodplain Population a 557 655 847 1328 Sex ratio a (proportion male) 0.44 0.45 0.46 0.43 Approx no. of households b 85 100 130 200 Road surface to village c Dirt Dirt Tar Tar Road surface within village c Dirt Dirt Dirt Dirt School c 1 primary 1 primary none 1 primary, day care facilities Health facilities c Health post, 1 nurse, 1 welfare Health post, 1 nurse, 1 welfare Health post, 1 nurse, 1 welfare Clinic, 1 matron, 1 nurse, 1 welfare Water supplies c Reticulated to communal standpipes Reticulated to communal standpipes none, collected from river 1 km away Reticulated to communal standpipes Postal facilities c via bag sent to neighboring village via bag sent to neighboring village via post office at neighboring village post office Communications c radio at health post radio at health post radio at health post public telephones a. Central Statistics Office 2002 b. based on an average of 6.5 people per household (Cassidy 1997) c. Ngamiland District Council 1997. Plan of Study In examining how fire affects resources that play a key role in peoples livelihoods in the panhandle, it was necessary, firstly, to identify how much of key resources in the focal villages collection areas were burnt. Secondly, the ways in which people evaluated

PAGE 38

23 the effect of this burning on their access to the resources needed to be determined. To accomplish these tasks, two approaches were used. The first was a geographic, or spatial, assessment of the location of fires, resources and collections areas that people in the focal villages used. This was done by analyzing satellite imagery, and by collecting geographic coordinates of resource types and interviewing people in the field. Once the areas of interest had been defined, it was possible to measure these using spatial modeling software. The second approach was to ask the people living in the area about their livelihoods, including how much they collected of each resource, and what their opinions about the effects of fire and other aspects of burning were. Information from a quantitative household survey was encoded and analyzed. Qualitative themes were developed from informal interviews. In order to be able to provide a quantified assessment, and to allow future comparisons, it was necessary to define spatial and temporal boundaries to the study. To this end, a defined area was clipped from the satellite imagery to reflect only the wetland areas of the panhandle. Analysis was restricted to a one-year timeframe throughout the research.

PAGE 39

CHAPTER 2 METHODS Primary data for this study were obtained through the analysis of satellite imagery and from formal and informal interviews conducted in the study area. Spatial Assessment The spatial assessment is based largely on the interpretation of two Landsat Enhanced Thematic Mapper (ETM) satellite scenes, from Path 175, Row 073. Landsat ETM has a spatial resolution, or ground cell size, of 28.5 x 28.5 meters. These cells are small enough to allow differentiation of features such as islands, river channels, clusters of dense vegetation, and burn scars. The first scene was taken at the end of the rainy season, and the peak of the flood, on 28 March 2001. The second was taken on 6 October 2001, at the end of the dry season. This date should capture most of the dry seasons burn scars, being late in the dry season, while still being early enough that regrowth would have been minimal. Both scenes have less than 1 % cloud cover. Image rectification, interpretation, analysis and map creation were done using ERDAS Imagine version 8.5 software. Initial subsets focusing on the area immediately surrounding the panhandle were created for ease of display. The subset of the October image was then rectified to the March image with a total RMS error of 0.024 using 40 ground control points. 24

PAGE 40

25 Definition of Study Area The study area was created by performing a 2-class ISODATA1 unsupervised classification of the March (wet season) initial subset. This was done by setting clustering to initialize from statistics along the diagonal axis, and with standard deviations of 1. Iterations were set to 20, with a convergence threshold of 0.95. This separated wet from dry areas. The dry area was recoded to 0, leaving a single value map layer of the wetland. The study area was consolidated by filling any islands that showed up as dry areas within the boundaries, and by trimming any wet areas that had been included from the dry land. This map layer was used to further subset both the March and October images. All further analysis was performed on these final subsets, hereafter referred to as the March scene or the October scene. This area is shown at the beginning of the results chapter. Location and Distribution of Key Vegetation Resources A 15-class ISODATA unsupervised classification (parameters as above) was carried out on the March image. In assigning vegetation categories to the classes, information from ground control points and photographs taken in the field were used. In the field it was clear that most of the vegetation types were mixed. There were few accessible places where one of the dominant species occupied a full ground cell, let alone the 100 m x 100 m that would allow one to be sure of having a true signal from the center cell. However, by examining spectral signatures, it was possible to separate areas of dense aquatic vegetation with high biomass, which comprised reeds, papyrus and thatching grass the three wetland plant resources of greatest interest. These three 1 ISODATA stands for Iterative Self-Organizing Data Analysis Technique. This algorithm makes multiple passes over the image, and after initial parameters are set by the user, organizes the clusters itself (Jensen 1996).

PAGE 41

26 resources were therefore treated as one category, justifiable because respondents typically obtained these resources from the same location. By comparing photographs and GPS readings to coordinates on the classified image, it became clear that six of the classes generated could be included in the dense aquatic vegetation category. This category is considered as plant resources. The six plant resources classes were recoded and combined in a new single value map layer. Areas of short grasses and sedges were also discernible, and verified by informants as being important for grazing. Based on their spectral signatures, four classes were combined to define vegetation areas suitable for grazing, even if these were in fact too inaccessible to be used as such. A recoded map layer containing only grazing areas was generated. Location and Distribution of Burnt Areas Burn scars in marshy wetlands are ephemeral, and are masked by the subsequent flood and plant regrowth. By the time of the visit to the study area eight months after the October scene was acquired, it was not possible to obtain any ground data to verify the interpretations of the satellite image. For this reason, more rigorous analysis of the burn areas as identified in the October scene was necessary. The first step taken was a visual analysis of the October scene. In a MODIS-based study of southern Africa, the 1.24 micrometer (m) band gave the best discrimination between burned and unburned areas, followed by 0.86 m, and then 1.64 m (Roy et al. 2002). This corresponds roughly to bands 5 (MIR1) and 4 (NIR). These two bands were displayed with MIR2 to show the best contrast of burned areas in the October scene. Fire

PAGE 42

27 scars, with active fire fronts or with distinctive shapes, on the savanna surrounding the panhandle gave verification to this interpretation. Figure 7. Savanna fire showing active fire front and burnt area. This fire lies on the edge of the wetland (bright green area), and, together with other similar burnt areas, was used to provide verification of burnt areas within the panhandle. Smoke can also be seen north of the flame pixels (red) in the upper right part of the image.

PAGE 43

28 Figure 7 shows an active fire (in red, where cells have saturated pixel values1 (255) in both MIR2 and thermal bands) and a fire scar behind it, about 20 km east of the study area. The next stage was to conduct a stage-by-stage hierarchical classification, essentially as a process of eliminating known features to allow wider separation among the remaining features. This was conducted as follows: 1. A 15-class ISODATA unsupervised classification (parameters as above) of the October scene was done to allow for the creation of masks to remove obvious or known features. 2. Open water (extremely low signatures in bands 2, 3, and particularly 4 and 5) covered 2 classes, and this was recoded as a map layer and applied as a mask to the October scene, creating a new image with no open water. 3. Next, the no open water image was classified into a new map layer using a 5-class ISODATA unsupervised classification. 4. The classes numbered 1and 2 were identified as corresponding to healthy vegetation, by three methods: Firstly, the swipe utility was used over a viewer display of the image in r,g,b = 5,4,3 format. Classes 1 and 2 consistently covered the green (indicating high reflecting in the NIR band) pixels. Nearly all uncovered pixels were shades of pink, red and magenta. Secondly, classes 1and 2 were swiped over a view display of a transformed NDVI image, and were shown to correspond to areas of healthy growth. Finally the spectral signatures generated for the 3-class map layer showed that both Class 1 and Class 2 were still fairly wet areas, with low values in bands 2 and 3, and rising in bands 4 and 5, with band 7 dropping extremely low. In contrast, Classes 3, 4 and 5 dropped sharply in the NIR band, and had extremely high values in band 5 and high values in band 7. 5. Classes 1 and 2 were then recoded as a map layer and applied as a mask to the no open water image. 1 In 8-bit data, pixel values range from 0-255, with 0 representing total absorbance, and 255 the maximum measurable by the sensor.

PAGE 44

29 6. Having now removed all open water and all healthy vegetation, by a process of elimination all remaining areas would have to be senescent vegetation, or exposed sand or soil due to burning or grazing. 7. A principal components analysis (PCA) was then carried out on this final masked no water, no healthy vegetation scene. The first component accounted for 99.2% of the variance in the scene. The major contribution to this band came from the thermal, MIR1, and blue bands. The second component accounted for 0.62% of the variance, while the third component accounted for 0.11%. The contributions of the various spectral bands to each of the first 3 principal components are shown in Table 2. When considering all these bands together, most of the variance of the non-vegetated or sparsely vegetated areas is explained by the thermal, MIR, and NIR bands, as suggested for the initial visual analysis. Table 2. Contributions of spectral bands to principal components from subset of exposed ground image. PC 1 PC 2 PC 3 Band 1 (blue) 0.3122 -0.2045 -0.1214 Band 2 (green) 0.2593 -0.0959 -0.2605 Band 3 (red) 0.2800 0.0797 -0.3247 Band 4 (NIR) 0.2450 -0.2335 -0.6616 Band 5 (MIR1) 0.4424 0.5127 -0.1413 Band 6 (thermal) 0.6414 -0.4610 0.5493 Band 7 (MIR2) 0.2996 0.6424 0.2292 While it is possible that all the non-vegetated or sparsely vegetated area could be burnt, the wetland conditions, with fire burning on vegetation on top of water, or across areas where plants can start resprouting within days, make it difficult for the sensor to define it as such (Stroppiana et al. 2002). However, charcoal and ash deposits would suggest that burnt areas reflect less in NIR and MIR bands, and emit more in the thermal band than sand with sparse vegetation. Conversely, sandy areas with sparse vegetation would have higher reflectance in red, NIR and MIR bands, and less so than burnt areas in the thermal band. Overall, therefore, burnt areas are likely to appear darker than grazed or otherwise sparsely vegetated areas in the satellite image. 8. Signature changes across the spectral bands for various sites on the exposed area no open water, no healthy vegetation subset were examined. Consistent distinctions could be made between two types of area, which corresponded roughly to whether

PAGE 45

30 they appeared to have low or high reflectance. The variations in pixel values for the two types of area are shown in Table 3. Table 3. Pixel values for areas of low and high reflectance in exposed area subset. Band Areas of Low Reflectance Areas of High Reflectance Red < 80 > 90 NIR < 70 > 70 MIR1 < 130 > 150 MIR2 < 100 > 100 Thermal > 200 < 190 Based on these signatures, low reflectance areas are assumed to relate to burnt areas, while areas of high reflectance are more likely to be areas of sparse or senescent vegetation, consistent with grazing. This information was used to guide interpretation of the final classification. 9. In the final stage of the hierarchical classification, a 15-class ISODATA unsupervised classification was done on the exposed area subset (the image from which all open water and healthy vegetation had been masked). The signatures of two of the classes matched the low reflectance area pixel values identified above (see Figure 8). A third class matched all the criteria except that the red pixel values were all over 90, while a fourth class matched all the criteria except that the thermal pixel values were all around 190. While these may also be burnt areas, it was decided that only the classes matching the ranges tabulated above for all the bands should be included. 10. The two classes that most closely matched the pixel values in Table 3 were then recoded as burned areas into a single value burn map layer, completing the hierarchical classification process. In spite of the limited ground data on burnt areas, a rough accuracy assessment of the burn map layer was carried out. Coordinates of places repeatedly identified by respondents as having burned in 2001, as well as those of 3 active burns in the October image were used to create a set of 13 points for use in conducting an accuracy assessment of the burnt areas. This was done on a composite of the burn map layer overlaid on the March classification. Of the 13 points, 9 were classified as such in the map layer, while 3

PAGE 46

31 Figure 8. Signature plots of the 15-class unsupervised classification of the October scene exposed area subset, showing the two classes selected for recoding into the burn map layer. Values on the Y-axis are mean pixel values, while the X-axis shows the 7 Landsat bands.

PAGE 47

32 were classified as dense aquatic vegetation and 1 as sedge/short grass. This suggests considerable under-classification of the burnt areas in the map layer, and supports the likelihood of errors of omission in the classification process. It is highly probable that the actual area burnt in 2001 was larger, because vegetation burnt earlier in the season would have had longer to grow, and charcoal and ash would have dispersed. Similar low misclassifications have been reported in northern Australia (Stroppiana et al., op. cit.). Only one ground control point identified as vegetation was identified as burnt in the map layer. Overall, an accuracy level of approximately 70% can be assessed, with the reservation that the number of points is limited, and that the accuracy is related to rapid changes in burnt areas. Although this assessment is statistically insufficient, it is considered to be adequate as an indication. Once the burn map layer had been created and verified, the area of burn in the papyrus/reed/thatch map layer and the grazing map layer was calculated by using the burn map layer as a mask to subset those. Identification of Village Resource Areas Collection and grazing areas were defined on the basis of information gathered from the household survey and informal interviews. In some instances, GPS co-ordinates were taken. In the field, informants helped develop hand-drawn maps of places used for various resources. These were then sketched onto 1:50 000 topographic sheets, and modified to accommodate other information on accessibility. Using ERDAS Imagine, areas of interest were developed to correspond to each villages grazing area and plant resource collection area, as defined on the maps. These areas of interest were used to subset the March image. The result was eight images, one for each of the grazing and collection areas of the villages. Again, these were recoded to provide single value map layers. It must be noted that these areas correspond to the range within which people

PAGE 48

33 (and livestock) move, and within which the resources fall. The actual resources cover only a proportion of these ranges. Burning within these areas was calculated by using these range map layers as masks to create subsets of the burn map layer. Burning of the resources within these areas was also calculated by using the range map layers as masks this time of the previously created burn-resources subsets referred to above. Calculation of Geographic Extent of Resource and Burn Areas ERDAS Imagine automatically measures the total area of a given map layer, and reports the area in hectares in the Raster Attributes table under the image viewer. These data were exported to a Microsoft Excel spreadsheet to calculate proportions. Socio-economic Assessment Most of the socio-economic data were collected in the study area between late May and mid July 2002. Prior to starting the research, meetings were held with the village chiefs to obtain their permission to work in the villages. With the exception of Mogotho, the chiefs all called public meetings to introduce the study to the community. Three research assistants helped administer a quantitative household survey, and to conduct informal interviews on certain discussion topics. Throughout this study the unit of analysis is the household, unless otherwise stated. This is considered the most appropriate scale for quantification of resources use, as well as for comparisons between types and extent of use. Due to time constraints, it was not possible to conduct a survey of all the households in each village. For statistical validity within each village, a minimum sample of 30 per village would be preferred. According to the Central Limit Theorem, the shape of the sampling distribution more closely resembles the shape of a normal distribution as sample size increases. Where the sample

PAGE 49

34 size is greater than or equal to 30, a normal distribution shape is reached if the population itself is normally distributed (Agresti and Finlay 1999). Within each village sampling was done using a systematic random approach (ibid.). The estimated total number of households in each village was divided by 30 to identify the skip number (k). The sample was then randomized by selecting the first household at random for any value between 1 and k. After that every kth household was selected by walking transects through the village that were perpendicular to the panhandle bank, thus avoiding any bias associated with the linear features of the settlements. The number of households was in fact not known, and these were estimated by dividing the population by 6.5, the 1997 estimated mean household size for the Okavango (see Chapter 1). Thirty-five questionnaires were administered in Mogotho and 36 in each of the other villages, giving a total of 143 cases. Table 4. Population and sample sizes of focal villages. Village Mogotho Sekondomboro Samochima Nxamasere Total Population a 557 655 847 1328 3387 Est. # of households 85 100 130 200 515 Sample 35 36 36 36 143 a. Central Statistics Office 2002. The results of the survey are considered to be representative of all households within the focus villages. For a 95% confidence interval and assuming maximum heterogeneity, the total sample provides a margin of error of 0.08 %, while the samples for each village have margins of error of 0.16%.

PAGE 50

35 The survey was tested on five households in Mogotho and a few minor modifications were made to the questionnaire as a result. The questionnaire (see Appendix A) was conducted in the national language Setswana, which is understood by the researcher and almost all respondents. Where necessary, the research assistants provided translations into Thimbukushu, Xanikhwe or Seyei or Sekgalagadi. The questionnaire gathered information on household characteristics, tools and assets owned, economic activities, quantities of resources used and the effects of fire on these resources. A strict one-year point of reference was adhered to for all questions. The data were all coded and entered into a database, and analyzed using SPSS version 10.0 software. Two types of variables were recorded: nominal (qualitative) data and interval (quantitative) data. Forty-three informal interviews were also conducted 8 in Mogotho, 11 in Sekondomboro, 15 in Samochima and 9 in Nxamasere (see Appendix B). Information was solicited in as neutral and open a manner as possible, and informants were asked to look at both positive and negative aspects of the topics raised. Respondents were encouraged to focus on one particular resource for the discussion, even if they used several. Discussions were held across several themes, to provide deeper contextual information about peoples views regarding access to resources and about burning. According to Spradley (1979), themes form part of the system of meaning that underpins a given cultural context. By focusing the informal interviews on specific topics, recurring themes were able to emerge. Ethnographic analysis of these themes was used to provide a systematic interpretation of the perspective of those reliant on the resource

PAGE 51

36 base (ibid.). This allowed the social constructs of the respondents to be presented without being filtered by external theories or value judgments. Extent of Reliance on Key Natural Resources Information on the extent of reliance on natural resources is taken mostly from the quantitative household survey. Information for the panhandle as a whole was first assessed. The proportions of households relying on each of the key resources, as well as the central tendencies of the amounts of these resources, were examined. Descriptive reports such as variability and distribution were generated. Quantification of resource use Some households had a clear idea of exactly how much of a particular resource they had collected in 2001. However, for many, particularly with regard to resources collected sporadically, it was necessary to calculate this based on number of collecting days, and amounts per day. In addition, an attempt was made to standardize units. For reeds and thatching grass, small bundles were taken as the unit, and larger bundles were counted as double. Wherever possible, actual bundles were examined. Water lily bulbs were difficult to assess. Typically people would go out with an enamel dish, but sometimes they would use plastic shopping bags. These quantities were estimated in terms of a 20 cm diameter dish. A recent fisheries survey (Mosepele 2001) that involved keeping diaries meant that many respondents were more aware of how much fish they caught. While reeds and thatching grass were collected within a very short time-frame, fish tended to be caught all year, or for several months at a time. For respondents who said they fished every day of the year, the number of days used was 300, as this was more probable, considering other social and economic factors. For those fishing for several months at a time, months were calculated at 25 days each.

PAGE 52

37 However, when assessing the number of fish, size was not taken into consideration. An exception to this was fish caught by traditional basket. These fish are typically less than 10 cm long, and are collected in bowls. Because the focus in this study is on the Table 5. Average sizes and prices for plant resources and fish. Item Unit Size a Prices b Thatching grass large bundle 20 cm to 30 cm P 20 Thatching grass small bundle 10 cm to 15 cm P 10 Reeds large bundle 30 cm to 60 cm P 20 Reeds small bundle 15 cm to 30 cm P 10 Mabinda (reed) mats each large, 2 m x 3 m P 20 Mabinda (reed) mats each small, 1.5 x 3 m P 10 Water lily bulbs c dish 20 cm P 2-P 3 Palm leaves wrist bundle P5 for 5 cm to7 cm P 5 Baskets tatana flat 25 cm P 30 Baskets big closed 50 cm tall P 100 Baskets small closed 25 cm tall P 50 Baskets seteko dish-shaped 75 cm P 50 Papyrus shoots for eating c stalk P0.25 Papyrus for mats c large bundle P 20 Papyrus sleeping mats each large P 100 Papyrus sleeping mats each small P 85 Tilapia Fish (fresh) each or kg 30cm, P4.50 or 1 kg P9-P10 Tilapia Fish (fresh) each or kg 15 cm or 500g P5 Tilapia Fish (dried) kg P2.50 Barbel Fish (fresh) each large fish P 6 a. = diameter b. Values are given in Botswana Pula. At the time of writing US$1 = Pula 6. Prices are based on average costs given by respondents in informal interviews. c. These items are very rarely sold.

PAGE 53

38 contribution to livelihoods, and not on the impact of extraction on the ecosystem, these bowls were quantified in terms of their equivalent size of net-caught fish. Each bowl collected was considered to equal 2 fish. The approximate sizes and local values of key resources are given in Table 5. Quantification of wealth In order to allow comparisons between households in terms of wealth, a wealth index (WI) variable was developed. This reflected both goods, such as the list of tools and assets owned, and flows, such as cash income opportunities from employment or small businesses. Values ascribed to the various assets are important more in terms of magnitude than in the actual value given. For each household, the value of each item or activity (as shown in Table 6) was summed into a new variable based on whether the item or activity was present (1) or absent (0) in the past year, independently of the number of items or extent of activity. An important exception to these are livestock owned and wage employment, where actual numbers were multiplied by the representative values. An example from a random household would be: WI for Household X = (beer selling: 1*200) + (street vendor: 0*500) + (carpentry: 0*500) + (thatching: 0*200) + (brick-making: 0*500) + (well-digging: 1*200) + (smithy: 0*200) + + (grows crops: 1*200) + (cattle: 3*2000) + (goats: 5*250) + (donkeys: 0*100) + (chickens: 7*5) + (wage jobs: 1*5000) + (drought relief jobs: 4*100). The inclusion of livestock is important because people base social status to livestock ownership (particularly cattle). The relationship between wealth and cattle ownership is interesting because people in the study area do not regularly trade their livestock, only selling or slaughtering for special events such as weddings or funerals, or

PAGE 54

39 if there is an urgent need for cash. The role of cattle in determining wealth is explored further in a causal path analysis discussed below. Table 6. Values of goods and flows used to create wealth index variable. Item Value a Item Value Beer selling 200 Hoe 20 Street vendor 500 Plough 100 Carpentry 500 Yokes 20 Thatching 200 Harrows 20 Brick-making 500 Spade 20 Well-digging 200 Axe 20 Smithy 200 Gun 2000 Bakery 500 Spear 20 Poultry project 500 Fish net 200 Knitting for sale 100 Fishing basket 20 Sewing for sale 100 Large pots for paid cooking 100 Bicycle 200 Jerry cans for water 100 Cart 1000 Drums for water 200 Vehicle 50000 Mosquito net 50 Tractor 75000 Grows crops 200 Dug-out canoe 1000 Cattle no.*2000 Sledge 500 Goats no.*250 Wheelbarrow 100 Donkey no.*100 Standpipe in yard 200 Chickens no.*5 Water pump 500 No. of full-time jobs no.*5000 Generator 1000 No. of drought relief jobs no.*100 Radio 100 Lamps 20 Stove 100 a. Values are given in Botswana Pula. Prices are based on current average costs of items, allowing some devaluation. Annual income from small businesses was estimated based on the purchase potential in the villages. The value of cattle is accurate at current prices.

PAGE 55

40 Proportional contribution of key wetland resources to livelihoods Another measure of assessing the importance of wetland resources to individual households is in terms of the proportion they contributed to the total household livelihood in 2001. To calculate this, the amounts of the resources listed in Table 7 that households collected in 2001 were multiplied by the values presented in Table 5. Table 7. Wetland resources included in natural resources value. reeds thatching grass palm leaves water lily bulbs papyrus for eating papyrus for mats fish These figures were then summed into to provide a natural resource value (NRV) and measured against the wealth index (WI) as follows: Proportional contribution = NRV/(NRV + WI), where NRV = natural resources value and WI = wealth index. This provided another variable for assessing variations in reliance on the key resources. It must be noted that there are also dryland resources such as poles, timber and firewood that are not considered in the total livelihoods. In addition, some wetland resources are not collected every year. However, it is felt that the likelihood of any household collecting a particular resource in a given year is the same, and so the proportional contribution for a year is indicative of most years. Variations in Reliance on Key Resources Differences in extent of reliance on certain resources were assessed in terms of spatial location, gender of household head, ethnicity and wealth. These were primarily

PAGE 56

41 tests of measures of association, with all tests set to a 0.05 significance level (95% confidence interval). The combination of data types determined the tests conducted, as shown in Table 8. Table 8. Measures of association by data type. Nominal Interval Nominal Crosstabs (Chi-square Phi and Cramers V; Fisher exact test for small samples) Comparisons of means (independent sample t-test for 2 samples; ANOVA for more than 2 samples, F-value as measure of association), Fishers LSD to follow up which samples vary significantly. Interval Correlations (Pearsons correlation co-efficient) Spatial variation is assessed on the basis of the 4 focal villages. Analysis of gender differences was limited to household head. Households in the study were categorized as male-headed, female (de jure) headed, and female (de facto) headed. This last category was defined where the male head was absent for longer than 6 months of each year. Table 9. Names of languages spoken by different ethnic groups. Ethnic Group Language Comment Hambukushu Thimbukushu Xereku Xereku Related to the Hambukushu Bakgalagadi Sekgalagadi Bugakhwe Bugakhwe A San group Xanikhwe Xanikhwe Sometimes known as River San Bayei Seyei Batswana Setswana

PAGE 57

42 Main language spoken was used to characterize the ethnicity of a household. This was done because the government has avoided any emphasis on ethnic or tribal identities. As a result, people are unused to talking about this. The groups and their languages are shown in Table 9. In the text of this thesis, the ethnic group name will be used in preference to the language. Effect of Fire on Availability of Key Natural Resources Figures were taken from the spatial analysis of the satellite imagery to determine what proportion of resources were burnt. This information was supplemented by the responses of people in households that had experienced fire where they normally collect resources, as to whether fire had improved or reduced their access to specific resources. Only households that usually collected a given resource were asked if there had been fire in the resource area in 2001. Although there is no direct evidence to support this, it is felt that the number of yes responses was lower than the number of households that actually knew there had been a fire there. This feeling is due to some hesitancy witnessed as some respondents answered the question, and may in part be because people were wary of talking about an illegal activity, or embarrassed to say they didnt know, even if this was because they themselves werent the collector, or the household hadnt collected that year. There was also a sense that, if the fire hadnt affected some one, for them it had not occurred. Qualitative responses on how fire affects resources were taken from the informal interviews. These were categorized and built into ethnographic themes as described above. They were also used to provide contextual information for the analysis of the household survey responses relating to this issue.

PAGE 58

43 Causal path linking sustainable livelihoods, resource use and the effects of fire During the initial stages of data analysis it became clear that socio-economic factors and not just fire played an important role in determining access to resources. In order to be able to bring the focus back to livelihoods and their sustainability, and to compare the relative effects of fire and livelihood sustainability on access to resources, a causal path analysis was developed using both original and derived variables from the household survey. An important assumption is that the wealth index (WI) is an adequate proxy for livelihood sustainability, at least in the short term and at household level. As is shown in Table 6, both assets (goods) and income opportunities (flows) were considered when creating the WI variable. While bivariate tests showed that wealth was central to resource use, it was clear that it was not the sole factor. Significant associations, as show in Chapter 3, revealed that other factors influence wealth, and that wealth is not the sole factor determining access to resources. Given the complex range of variables involved, three stages of multiple regressions were used to test the likely causal path, presented diagrammatically in Figure 9. This diagram is based on the survey variables that had strong associations in bivariate tests. Social characteristics, as represented by gender of household head and ethnicity, were assumed to be the primary factors influencing the number of adults in the household (people aged 15 years and over), as well as the number of jobs. Bivariate analysese did not show significant associations between gender of household head and number of cattle owned, nor with ethnicity or cattle owned. Nevertheless, these relationships are tested again here to confirm this finding. The predicted influence of social characteristics is shown in Stage 1 in the diagram in Figure 9. Number of wage jobs is thought to be the most important flow, while cattle and adult labor are thought to be the most important

PAGE 59

44 assets, in determining wealth. Indeed, cattle and wage employment have already been assumed to be part of wealth, hence their inclusion in the WI variable in the first place. However, in order to show their importance, the WI variable was altered for this exercise. Three key items (values for canoe ownership, cattle owned and wage jobs) that individually had strong associations with resource use were taken out to create a modified WI variable (referred to in the diagram as MVI) so that their relative input to wealth (as a proxy for livelihood sustainability) could be determined. The second stage of the multiple regressions tested the MVI for dependency on number of adults in the household, number of cattle owned and number of wage jobs in the household. The third and final stage of the analysis returned the focus to the role of wealth relative to other factors notably fire in determining various aspects of resource use. Several of the variables contained nominal data with more than 2 categories. This meant that they could not simple be recoded to numeric presence/absence values. In order to include them in the analysis, it was necessary to create dummy variable by moving each category to a separate variable. This was done for ethnicity and location (village). Bugakhwe and Xanikhwe were combined into a single group San because their individual sample sizes were small. Hambukushu was chosen as the referent variable for ethnicity, and Mogotho as the referent for location. These two variables were therefore not entered into the regression equations. Female de facto households were recoded as male-headed, and the variable was assigned new values to allow gender of household head to be considered as either male or not. Ownership of a canoe and experience of fire were recoded in a similar manner to indicate presence or absence.

PAGE 60

45 Figure 9. Survey variables expected to influence access to wetland resources. The variables were selected on the basis of significant bivariate associations. This causal path diagram shows the 3 stages for which multiple regressions will be run. In the first stage, no. of adults, no. of wage jobs and no. of cattle are separately tested as dependent on ethnicity and gender of household head. In the second stage the modified wealth index (MVI) is tested as dependent on no. of adults, no. of wage jobs, and no. of cattle. In the third and final stage, various variables that comprise access to wetland resources are separately tested as dependent on no. of adults, ethnicity, MVI, canoe ownership, location/village (where presence of resources may differ), and whether fire was experienced or not. This final stage allows the effect of fire on access to resources to be evaluated relative to other factors.

PAGE 61

46 Temporal variations in effects of fire Given that the timing of the fires (how early in the season) appeared to be a key factor in determining the negative effect on livelihoods, it seems likely that the timing of collection (how late in the season) by a given household would also make them more susceptible to the negative effects of fire. A decision tree model based on factors determining whether households could go early to collect resources or not was therefore developed from the relevant theme. Ethnographic decision tree modeling allows one to predict group behavior in situations where decisions are made by individuals or in this study, individual households (Gladwin 1989). A set of alternatives or if then rules determine the path of the tree, and reveal differences between households in their ability to access resources soon enough to avoid loss due to fire. Conflict Surrounding Burning Measuring conflict between people also proved difficult, because for many respondents, conflict meant confrontation. This did not happen because, as became clear during the course of the survey, the identity of those responsible was normally not known. Substitute questions such as whether people lost access to resources because of fire, or whether they believed that fire was bad, and why, were asked. Some post hoc statistical analysis was possible based on the timing of reported fires, and which resource-user types were typically held responsible for setting fires. Comparing cross-tabulations of month of fire versus origin of fire for each resource type allowed the overall number of fire events experienced to be identified. The informal interviews provided several themes linking conflict to the lack of control over burning.

PAGE 62

CHAPTER 3 RESULTS Extent and Distribution of Fire in the Panhandle and Key Resource Areas The wetlands of the panhandle, as defined in Chapter 1, measure some 88790 hectares (ha), including permanently dry islands within its boundaries. The spatial extent of the study area is shown in Figure 10. Open water (with no surface vegetation) accounts for about 10% of this area. 1 Dense aquatic vegetation, such as reeds, papyrus and tall grasses, covers more than 40% of the panhandles area. Short grasses and sedges (floodplain vegetation) account for some 30% of the study area, including those parts which may be covered in shallow floodwaters for part of the year, but which dry out seasonally. These vegetation groups are shown graphically in Figure 11. Figure 11 shows only that portion of the panhandle in the vicinity of the focal villages to provide greater resolution for the different classes. By early October, at least 3690 ha (approximately 4.1%) of the wetland area had burnt (Figure 12) during the 2001 dry season. Most of the burning took place on the floodplain areas. While 4.7% of the floodplains short grass and sedge vegetation burnt, only 2.2% of the dense aquatic vegetation did. 1 All the measurements and proportions in this section are derived from interpretation of the satellite imagery. 47

PAGE 63

48 Figure 10. Study area map. Green areas denote healthy vegetation, while areas that appear black are open water, and pink areas are exposed soil. All 12 villages along the panhandle are located on the dry land, but only the four villages that are the focus of this study are shown.

PAGE 64

49 Figure 11. Map showing main vegetation groups in vicinity of focal villages. Grazing areas correspond to the Open water seasonal and Sedge / short grass classes. Classes were derived from the March 2001 Landsat scene using an unsupervised 15-class classification. The sedge / short grass category was merged from four of these classes, while the papyrus/reeds/ thatch class was merged from 6 of them. This map shows why Mogotho and Samochima are considered to be river villages and Nxamasere and Sekondomboro floodplain villages.

PAGE 65

50 Figure 12. Map showing extent of burning in 2001. Burnt areas are concentrated near the floodplain villages, while the river villages experienced less fire.

PAGE 66

51 Key Resources Areas for Focal Villages The spatial extent of the plant collection ranges and the livestock grazing ranges, as developed in the methods section, for the four villages is shown in Figure 13, while Table 10 summarizes the values. It is striking to note that for both range types, the two floodplain villages have roughly double the area of the other two. The larger size is primarily accounted for by the greater distance from the panhandle banks to the main river channels, as the length along the banks is similar. There is considerable overlap between the grazing and plant collection ranges of each village, but generally within the ranges these two main resource types are discrete. Table 10. Size of village resource ranges River villages Floodplain villages Village Mogotho Samochima Nxamasere Sekondomboro Plant collection range (ha) 678.6 764.8 1463.5 1306.75 Proportion of total study area 0.76% 0.86% 1.65% 1.47% Length along banks (km) 12.3 11.7 12.4 8.1 Width (km) 0.58 0.98 1.10 (2.72 at extreme) 6.87 Grazing range (ha) 1243.1 1157.3 3407.39 2858.63 Proportion of total study area 1.40% 1.30% 3.84% 3.22% Length along banks (km) 14.2 14.6 14.6 11.7 Width (km) 1.05 0.85 (2.02 at extreme) 2.55 2.40 (5.34 at extreme)

PAGE 67

52 Figure 13. Map of grazing and wetland plant resource areas for each focal village. Resource areas were based on information provided by informants during the survey, and from GPS coordinates collected in the field. There is considerable overlap between graze areas and plant collection areas for each village. Both types of area extend much further into the panhandle for the floodplain villages, while those of the river villages are constrained close to the panhandle edge. This is because dense aquatic vegetation limits access.

PAGE 68

53 Grazing areas The importance of floodplains for grazing is shown by both the size and the proportion of short grass/sedge vegetation available to Sekondomboro and Nxamasere villages. These data are presented in Figure 14. 0.010.020.030.040.050.060.070.0MogothoNxamasereSekondomboroSamochima Size of grazing area(sq km) Percent of grazingarea available as graze Figure 14. Comparison of available wetland grazing for focal villages. The Y-axis gives square kilometer values for size of grazing area and percent of grazing area for available graze. Only in the case of Sekondomboro do the actual graze resources cover more than half of the grazing area. Plant collection areas The availability of resources within these ranges also varies considerably. The river villages have the greatest concentration of reeds, thatching grass and papyrus. In terms of spatial extent, these three resources are considered together (see for example, the

PAGE 69

54 papyrus/reed/thatch class shown in Figure 11). Together these three wetland plant resources cover 59.7% of Samochimas, and 45.2% of Mogothos plant collection area. This is in contrast to the floodplain villages, where these resources cover 44.5% of Sekondomboros and only 23.4% of Nxamaseres areas, due to its drier conditions, as is shown in Figure 15. 0.010.020.030.040.050.060.070.0MogothoNxamasereSekondomboroSamochima Size of plant collection area(sq km) Percent of collection areacovered in papyrus /reeds/thatch Figure 15. Comparison of available plant resources for focal villages. The Y-axis gives square kilometer values for size of plant collection area and percent of the collection area covered in papyrus/reeds/thatch. Water lily bulb collection was not mapped. These areas are related to shallow open water, of which the largest patches are found at Sekondomboro. Hyphaene palms grow on larger islands where floodwaters do not reach. These islands are scarce except around Nxamasere. Other resources we use are the riparian trees, such as the fruit of mochaba

PAGE 70

55 [Ficus sycomorus], mokochum [Diospyros mespiliformis], and motsaodi [Garcinia livingstonei]. There are also many leafy plants that are important as food in our lives. If these things are found close to the village, we sometimes make specific trips to collect them, but if they are far away we will just look for them when we are harvesting reeds. (Water lily bulb collector, Sekondomboro, June 2001) Fishing Fishing takes place in several different types of locations, depending on the techniques used. Traditional Hambukushu basket fishing is done by women in the shallow vegetated floodplains. According to responses from the survey, the vegetation is important as it harder for fish to escape. Because access is on foot, basket fishing usually takes place within 500 m of the edge of the panhandle. Some women walk several kilometers along the bank to find suitable places. This was particularly evident at Samochima. Net fishing is mainly done in areas of unvegetated water on the floodplains, although nets are also strung along quieter lees of channels and lagoons. Hook-and-line fishing is done on the big areas of permanent water, and, like net fishing, is dependent on access by dugout canoe or motorboat. Motor boats are used by small artesanal fishermen, who have a co-operative based at Samochima. This co-operative is supported by subsidies and technical advice from the fishing division of the governments agricultural extension agency. The boats give them the ability to range up to 50 km away along the main rivers. Burning in the Key Resource Areas The percentage burnt within all the total villages plant collection areas (8.65%) and grazing areas (10.8%) is much higher than the 4.16% for the study area as a whole.

PAGE 71

56 Nevertheless, this is still a small proportion of the resources areas. There were differences between the villages, as Table 11 indicates. Notice how greater areas and proportions were burnt in the plant collection areas of the floodplain villages. Table 11. Extent of fire within village resource areas River villages Floodplain villages Village Mogotho Samochima Nxamasere Sekondomboro Area burnt in plant collection range (ha) 46.57 11.05 252.84 53.86 Proportion of collection range burnt 6.86% 1.45% 17.28% 4.12% Area burnt in grazing range (ha) 76.83 34.26 613.48 209.80 Proportion of grazing range burnt 6.18% 2.96% 18.0% 7.34% Fire in grazing areas Nxamasere had the highest proportion of its graze area burnt (see Figure 16 and Figure 17). In terms of actual graze resources, the two river villages both had less than 2% burnt, while the floodplain villages had more than double this proportion.

PAGE 72

57 Figure 16. Map showing graze resources burnt in the grazing areas of each focal village. Maroon indicates actual graze resources burnt, while tan indicates burning of other vegetation types.

PAGE 73

58 0.02.04.06.08.010.012.014.016.018.020.0MogothoNxamasereSekondomboroSamochima Percent of graze area burnt Percent of graze resources(within graze area) burnt Figure 17. Proportion of grazing vegetation burnt in graze ranges. Nxamasere and Sekondomboro, the floodplain villages, had about twice the area of graze resources burnt compared to the river villages. Fire in plant collection areas The proportion of the actual resources that burnt within the collection areas appears to be much less than for the collection areas themselves. All the villages had less than 2% of their reeds/thatch/papyrus burnt, with the exception of Nxamasere. This village had 5.6% of these plant resources burnt in their collection area. Samochima appears to have experienced very little fire at all (see Figure 19 and Figure 18).

PAGE 74

59 Figure 18. Map showing papyrus /reed /thatch burnt in the plant collection areas of each focal village. Red indicates burning of actual resources, while beige indicates burning of other types of vegetation.

PAGE 75

60 0.02.04.06.08.010.012.014.016.018.020.0MogothoNxamasereSekondomboroSamochima Percent of plant collectionarea burnt Percent of papyrus /reed/thatch (in collection area)burnt Figure 19. Proportion of plant resources burnt in collection ranges. Temporal Distribution of Fire Because only one dry season satellite image was analyzed, it is not possible to know exactly when fires occurred. However, responses from the household survey indicate that more than half the fires people observed in their resource areas were in September. No fires were reported for the months of February to July. According to informal interview respondents, this is because the flood waters are still too high at this time of year, and so it is too wet to burn. Some people noted that fishermen did in fact set many small fires in the vegetation immediately around the openings where they set their nets, and that these fires did not spread at all because of the water.

PAGE 76

61 Discussion Theme Local Knowledge of Fire Behavior People are very aware of fire and how it moves within the panhandle, as the following excerpts from the informal interviews show. However, their expressions are a reflection not only of their experiences, but also of whether they see fire as good or bad. There is no way that fire can be controlled, especially in the delta because it is very difficult to stop, it burns from the first place to the last one, and results in the whole ecosystem burning. Whether fire can be controlled depends on how dry it is. If it is very dry, no-one can stop it. Generally fires in the delta cannot be controlled. Delta fires are usually stopped by water or hippo paths. In the old days we used to burn regularly, but now it is less frequent and this is why the fires are so big these days. If you burn every other year, the fires dont go so far. Even if you burn every year, you can still get big fires. There is no difference in the size or intensity of fire whether you burn every year or once in seven years. Fires go according to the wind. Usually fires do burn everything because of the wind. It is the wind which always brings the fires here. Wind and the amount of grass are the cause of big dangerous fires that go long distances. The direction of the fire usually depends on the direction of the wind. The size of the fire depends on the amount of old vegetation or twigs burning. The smaller the amount, the smaller the fire. Fires are bigger the year after a big flood, because there is more to burn. Even early in the winter [June July], there is so much wind and the fuels are so light and hot, they spread easily. Burning before winter isnt possible because of the floodwater, which stops burning. Flood doesnt affect fire, the vegetation always gets dried out.

PAGE 77

62 Fire never kills everything, wherever there is water patches of vegetation do remain. These things all re-grow from their base or roots so they are not damaged by fire. The interaction of fire and water is essential. When there is drought, even the roots burn because there is no water in the soil. We expect rain in October and November, so September is the best time for fire. Reliance on the Wetlands Natural Resource Base Only three households out of the 143 surveyed (2.1%) did not use any of the wetland resources asked about in the survey. An unexpected finding was that very few people (8.39%) grew crops in the floodplains, in spite of the greater level of nutrients compared to the conditions on the sandveld. Respondents explained that the land authority had removed all rights to floodplain fields because of increased conflicts with hippos. Where floodplain fields did exist, these were typically on the banks beyond the reach of floodwaters. The association between use of one resource and that of another is measured in two ways whether a household usually uses each of the resources (nominal data), and how much of each of the resources they use (interval data). The nominal data relating to whether a household usually used one resource compared to its usual use of others is shown in Table 12. Significant values are shown in bold. In 2001, wetland resources contributed on average 14.7% of household livelihoods, as measured as a proportion of the households total wealth (which will be discussed in greater detail below), with a median value of 5.2%. For 10% of households, however, wetland resources contribute at least 50% of the total livelihood. The range of values is shown graphically in Figure 20. The proportion of households relying on various resources are shown in Table 13.

PAGE 78

63 Table 12. Comparison of reliance on different resources. The table presents the results of cross-tabular analyses where usual use of one resource is compared against usual use of another. Significant values are shown in bold. Use of reeds, thatching grass and fish have strong associations with each other. Use of palm leaves is significantly related to use of water lily bulbs and papyrus for eating. Grazing Reeds a Palm leaf Thatching grass Water lily bulbs Papyrus for eating Papyrus for mats Fishing Grazing .094 .354 .161 .110 .157 .121 .140 .165 .010 .919 .005 .960 .068 .501 Reeds .139 .097 .437 .000 .098 .239 .066 .429 .124 .137 .177 .035 Palm leaf .050 .551 .275 .001 .199 .017 .125 .136 .056 .503 Thatching grass .080 .337 .128 .125 .175 .036 .163 .051 Water lily bulbs .177 .035 .190 .023 .246 .003 Papyrus for eating .188 .024 .268 .001 Papyrus for mats .138 .098 Fishing a. The first figure is the value of Phi and Cramers V (see Methods). The second is the significance level.

PAGE 79

64 Table 13. Household reliance on different resources. Proportion of HH generally collecting Proportion of HH collecting 2001 Mean amount collected 2001 Median amount collected 2001 Std. Dev. Grazing 37.1% n/a n/a n/a n/a Reeds 95.1% 69.2% 31.0 bundles 20 bundles 37.51 Palm leaf 27.3% 16.8% 3.9 bunches 3 bunches 2.9 Thatching grass 95.8% 74.1% 30.2 bundles 20 bundles 39.4 Water lily bulbs 35.0% 22.4% 21.1 dishes 14.3 dishes 17.9 Papyrus for eating 27.3% 18.9% 18.1 bundles 10 bundles 28.2 Papyrus for mats 41.3% 21.7% 4.3 bundles 2 bundles 5.7 Fishing a 37.8% 29.4% 1168 fish 550 fish 1976.7 a. It is believed that the proportion of households consuming fish is much higher, and that the relatively low value reflects a growing commercialization of the resource. .875.750.625.500.375.250.1250.000No. of households6050403020100 Figure 20. Proportional contribution of wetland resources to household livelihoods. The proportion of total household livelihood that was derived from wetland resources in 2001 is shown for the survey sample (N = 143). The mean proportion was 0.147, with a standard deviation of 0.21. The proportional contribution is calculated by NRV/(NRV + WI), where NRV = natural resources value and WI = wealth index.

PAGE 80

65 Grazing of livestock in the floodplain is determined by several factors. Not all households own livestock, and not all types of livestock graze. Trying to assess the extent of use is more difficult. While 51.7% of households owned goats, these are browsing animals and use the sandveld instead of the wetlands. It is for cattle that the grazing resources of the panhandle are mainly used. Only 39.8% of households own cattle, and the distribution of the numbers across the households is extremely skewed, as will be discussed below. Nevertheless, 37.1% of all households, and 61.4% of all cattle owners, graze their livestock in the panhandle when conditions are dry enough. Donkeys and horses account for the difference. Thatching grass appears to be the most important key resource collected 95.8% of households normally collect thatching grass. This resource is used exclusively for making roofs, which are typically replaced every three years. For this reason, the proportion of households collecting in 2001 was slightly lower 74.1%. The mean amount collected for all households who normally collect grass, including those that did not collect in 2001, was 23.4 bundles. If the number of traditional (as opposed to government employee) households is taken at 2500 for the study area as a whole, this suggests a level of extraction of between 55 000 and 60 000 bundles. While the mean number of thatching grass bundles collected may provide a broad scale idea of levels of extraction, it does not accurately reflect the variance or skewedness of use. Seven outliers have skewed the data to the right. Five of these cases have extreme values, of which two households collected 270 and 300 bundles last year. The mode (25.5% of 2001 collectors) was 20 bundles per household. Even without the two extreme outliers there is still considerable skewedness and kurtosis (greater clustering), as

PAGE 81

66 is shown in relation to the normal curve in Figure 21. Only seven households (6.6%) sold thatching grass last year, all selling 20 bundles or less. 240 250160 17080 900 10Frequency50403020100 Figure 21. Frequency distribution of thatching grass bundles collected in 2001. For the households who collected grass that year (N = 106), the mean number of bundles was 30.2, with a standard deviation of 39.4. Reed collection follows a very similar pattern to thatching grass. Ninety-five % of households regularly collect reeds, which are used to build the walls of both huts and yard enclosures. Less commonly nowadays, reeds are also flattened and woven into rigid mabinda mats that are used for sitting on or for wall or floor coverings. In 2001 69.2% of households collected reeds, with the mean number of bundles 22.59. Assuming the 2500 households referred to above, this once again suggests approximately 55 000 to 57 000 bundles of reeds for the study area as a whole.

PAGE 82

67 As with thatching grass, there were two extreme outliers one of which was the same household fro both resource types. The outlying households in reed collection harvested 200 and 275 bundles last year, while both the median and mode for bundles of reeds collected were 20, with a standard deviation of 37.5. This shows how skewed to the right distribution is, as two standard deviations to the left would give negative values. Twenty-three percent of those who collected reeds in 2001 sold bundles, compared to the 6.6% selling thatching grass. Nevertheless, the quantities sold remained very small (mode = 10 bundles, mean = 10.7 bundles). There is no significant correlation between amount of reeds collected and amount sold (Pearsons r =0.066, p = 0.514). Given this, and the lack of markets and the small quantities sold, it is likely that the sale of reeds reflects a once-off need for cash rather than increasing commercialization of the resource. Only 15.4% of households had members who still made mabinda mats, typically done by men. During the interviews several respondents said that few people knew how to do this any more. In the past year, only 7% actually made mats, while 2.8% of households sold mats. One man made and sold 100 under commission from a nearby safari lodge, but the other households sold between 6 and 12 mats each. Palm leaf collection and basket-making will be discussed in the section on spatial variation in reliance, as this resource is mainly found in one area. Note that papyrus for eating is in most cases not collected in bundles. Instead people eat it while out fishing or collecting other resources. When brought home it is usually for children, and must be eaten that day or else it dries out. Ten households in the sample (7%) sold papyrus sleeping mats in 2001. All sold 6 or less. Some respondents showed how empty grain sacks were now being used for sleeping mats, and explained they were easier to make. It

PAGE 83

68 is not sure if sufficient quantities of these sacks are permanently available, or if their presence and use is limited to how long welfare relief is provided after the cattle slaughter. Reliance on fish needs to be examined both in terms of the proportion of households collecting, as well as the amount of fish being sold. The proportion of households who said they usually catch fish in the delta was 37.8%. Those actually fishing in 2001 comprised 29.4%. Most households caught enough fish (mode = 600 fish, median = 550 fish) for their households. However, the mean value (1168.1) reflects considerable right-skewedness, which is explained by 35.7% of those fishermen catching in 2001 who caught between 800 and 10000 fish each. Fish sale quantities are typically low, implying local reliance on a resource collected by others within the study area. These data, together with the low proportion of households catching for themselves, suggest that a large number of households buy fish caught by others. Increasingly, though, the fishermen of the Samochima co-operative are selling outside the study area as they have both the quantities and the equipment to make transportation viable. .There is a significant correlation between size of catch and numbers of fish sold (Pearsons = 0.652, one-tailed significance < 0.001). Of the households selling fish, 3 sold less than 250. The mean number of fish sold by the 42.9% of fishermen who caught in 2001 was 1274. The comparative distribution of fish caught and fish sold is shown in Figure 22.

PAGE 84

69 10250775052502750250Frequency2520151050 A 625052504250325022501250250Frequency6543210 B Figure 22. Distribution of fish caught and fish sold in 2001. A) For those households catching fish that year (N = 42), the mean number caught was 1168, with a standard deviation of 1976.7. B) For those selling fish (N = 18) the mean number sold was 1274, with a standard deviation of 1427.6.

PAGE 85

70 Spatial Variations in Reliance on Key Resources There are strong differences between villages in the proportional contribution of wetland resources to household livelihoods (F = 2.657, p = 0.051). In both Mogotho and Sekondomboro wetland resources comprise almost 20% of the total household livelihood, compared to 8.8% in Nxamasere (see Table 14). To some extent, this could reflect Nxamaseres distance away from the edge of the panhandle, as well as the fact that it is closest to Maun where other types of building materials can be obtained. Given that ethnicity is also associated with village location, cultural differences may also be contributing to this association. This is discussed in greater detail below. In all villages the proportion of households owning livestock of any kind (including chickens) is similar ranging from 77.8% in Sekondomboro to 94.3% in Mogotho. Those owning cattle were less, as is shown in the box-plots in Figure 23 (where each box shows the central 50% for each group, while whiskers indicate the bottom and top percentiles. Larger boxes reflect greater spread of response values, with the sample size given in the X-axis above the name of the group), and in Table 15. Table 14. Proportional contribution of wetland resources to households by village. Village Mean proportion of household livelihood in 2001 Median proportion of household livelihood in 2001 Mogotho 19.3% 10.5% Sekondomboro 19.6% 11.3% Samochima 11.2% 3.8% Nxamasere 8.8% 2.6%

PAGE 86

71 NxamasereSamochimaSekondomboroMogothoNo. of cattle200180160140120100806040200 Figure 23. Distribution of cattle ownership by village in 2001. The boxes represent the interquartile ranges that contain 50% of the values for each village. The line across the box is the median value. The whiskers extend to the lowest and highest values, excluding outliers. Open circles indicate outliers, while extreme outliers are shown by stars Table 15. Numbers of cattle owned by village. Village Proportion of households owning cattle Mean # owned by these households Median # owned by these households Mogotho 37.1% 24.6 6 Sekondomboro 33.3% 12.8 5.5 Samochima 30.5% 17.7 5 Nxamasere 41.7% 16.6 12 Total numbers of cattle are very low. It is clear that people are taking time to replace their herds, and more than half the people in all villages own no cattle at all.

PAGE 87

72 Ninety-three percent of all cattle owners in Nxamasere, and 63.6% in Samochima grazed their livestock in the panhandle at some point in 2001. This compares with 47.4% for both Mogotho, and 41.7% for Sekondomboro, the east bank villages. The differences between the villages are significant (Phi and Cramers V = 0.419, P = 0.018), and the figure for Sekondomboro is surprisingly low, given the availability of wetland grazing. It may be that given the current low stocking rates, people prefer to graze their cattle on the sandveld where grasses are more palatable. Direct access to thatching grass shows some spatial variation. As Table 16 shows, almost all households in Mogotho collected thatching grass in 2001, compared to the west bank villages of Samochima and Nxamasere, where less than two-thirds did. While the villages of Mogotho and Sekondomboro are perched right on the edge of the panhandle, Nxamasere and Samochima are 6 km and 1 km away respectively. This increased distance may make collection difficult for some households. In Samochima, the mean number of bundles collected is highest. This fact, combined with fewer households collecting, suggests the possibility of greater trade in the resource in this village. Table 16. Thatching grass collected in 2001 by village. Village Proportion of households collecting Mean no. of bundles collected Median no. of bundles collected Sample sum amount Mogotho 94.3% 31.4 20 1035 Sekondomboro 83.3% 25.6 20 767 Samochima 58.3% 40.4 20 848 Nxamasere 61.1% 25.2 26.5 555 Total 74.1% 30.2 20 3205

PAGE 88

73 The concentration of grass collection into fewer households in Samochima would appear to suggest greater trade in this resource. However, this notion is not borne out by data on quantities of thatching grass sold. Only one household in the sample sold thatching grass in 2001, and that was only 10 bundles. Figures for the other villages were similarly low: Mogotho two households each selling 7 bundles; Sekondomboro one household selling 20 bundles; and Nxamasere two households selling 10, and one selling 5 bundles. Only 52.8% of households in Nxamasere collected reeds last year, compared to 82.9%, 77.8% and 63.9% for Mogotho, Sekondomboro and Samochima respectively. There was a significant difference in the mean number of bundles collected in Mogotho compared to Nxamasere (t = 2.287, p = 0.028). This reflects the proportion of resources available in the collection areas for the villages as identified above. Compared to thatching grass, there was greater trade in reeds in 2001, particularly in Mogotho, where 41.4% of households that collected reeds sold some of their bundles. Quantities again are low, with the median values at 10 bundles for Nxamasere and Mogotho, and 5 bundles for Sekondomboro and Samochima. Table 17. Reeds collected in 2001 by village. Village Proportion of households collecting Mean no. of bundles collected Median no. of bundles collected Sample sum amount Mogotho 82.9% 34.1 26 988 Sekondomboro 77.8% 33.4 20 936 Samochima 63.9% 32.3 20 744 Nxamasere 52.8% 21.3 20 404 Total 69.25 31.0 20 3072

PAGE 89

74 Table 18. Reed bundles sold in 2001 by village. Village Proportion of households selling Mean no. of bundles sold Median no. of bundles sold Sample sum amount Mogotho 34.2% 13.25 10 159 Sekondomboro 2.8% 5.0 5 5 Samochima 8.3% 5.77 5 17 Nxamasere 19.4% 9.3 10 65 Total 16.1% 10.7 10 246 Table 19. Households collecting palm leaves by village. Village Usually collecting Collecting in 2001 Mogotho 0.0% 0% Sekondomboro 5.6% 2.8% Samochima 25.0% 8.3% Nxamasere 77.8% 55.6% Table 20. Bunches of palm leaves collected in 2001 by village. Village Proportion of households collecting Mean no. of bunches collected Median no. of bunches collected Sample sum amount Mogotho 0% 0 0 0 Sekondomboro 2.9% 2.0 2 2 Samochima 5.56% 5.3 4 16 Nxamasere 55.56% 3.7 3 75 Total 66.67% 3.9 3 93 As suggested earlier, there are significant variations in the participation by different villages in palm leaf collection (Pearson chi-square = 68.07, p < 0.001). The proportion of households that regularly collect palm leaves is shown in Table 19. All of the Samochima households that regularly collect palm leaves travel to Nxamasere to do so, supporting information from the informal interviews that this is where the resource is concentrated. In 2001, only 33.3% of Samochima palm leaf collectors actually collected,

PAGE 90

75 compared to 72.4% of Nxamasere collectors. Only 2 households from Nxamasere sold palm leaves in 2001. Amounts collected are shown in Table 20. In spite of the low levels of palm leaf collection, a high proportion of households in all villages have members who make baskets (Table 21). The significant difference shown for Nxamasere compared to the other villages (Pearson chi-square = 19.91, p < 0.001) is probably related to the availability of the resource. T he ways in which households making baskets in 2001 obtained palm leaves is shown in Table 22. Mohembo, the northern-most village in the panhandle was regularly mentioned as a place for buying palm leaves. The number of baskets made by these households is not significantly different, but the number sold is (F = 2.86, p = 0.045). No households sold baskets in Sekondomboro last year, while 58.3 % of Nxamasere basket-makers sold a mean of 4.7 baskets each. The market for baskets is through a grassroots co-operative, Thokadi Women, who sell to a wholesaler in Botswanas capital city as well as informally to tourists from the nearby safari lodge. Table 21. Households with basket-makers. Village name Makes baskets Mean no. of baskets made per household Median no. of baskets made per household Mogotho 45.7% 3.2 2 Sekondomboro 38.9% 2.7 2 Samochima 47.2% 3.1 2 Nxamasere 86.1% 4.8 3 Total 54.5% 3.6 2

PAGE 91

76 Table 22. Source of palm leaves for basket-making. Collected by HH member Bought Both collected and bought Gift Total Sample number 25 30 1 4 60 Proportion 41.7% 50.0% 1.7% 6.% 100% The number of households relying on water lily bulbs for food varies only slightly, as is shown in Table 23. However, there are significant differences (F = 2.89, p =0.05) in the extent to which households in the different villages use the resource (see boxplot in Figure 24). Of those that collected water lily bulbs, households in Samochima collected a mean of 52.5 dishes, compared to 18.6, 22.5 and 15.7 for Mogotho, Sekondomboro and Nxamasere respectively. NxamasereSamochimaSekondomboroMogothoNo. of dishes of water lily bulbs706050403020100-10 Figure 24. Extent of water lily bulb collection by village in 2001. The boxes represent the interquartile ranges that contain 50% of the values for each village. The line across the box is the median value. The whiskers extend to the lowest and highest values, excluding outliers. Open circles indicate outliers, while extreme outliers are shown by stars.

PAGE 92

77 Table 23. Collection of water lily bulbs by village. Village name Usually collect Collecting in 2001 Mogotho 25.7% 25.7% Sekondomboro 38.9% 30.6% Samochima 30.6% 5.6% Nxamasere 44.4% 27.8% The proportion of households that usually collects papyrus for mat-making is remarkably similar for all villages, ranging from 38.9 % to 42.9%. However, there were marked differences between the proportions that actually collected in 2001. While 42.9% of Mogotho households collected papyrus for mats in 2001, only 16.7% in Sekondomboro, 5.5% in Samochima, and 22.2% in Nxamasere did. Quantities in Sekondomboro were lowest, with a mean of 2.3 bundles. It was in this village that people mentioned that they were making sleeping mats out of grain sacks nowadays. The highest mean number of bundles collected was in Nxamasere (5.1 bundles). With regards to papyrus collected for eating there was a significant difference for the different villages (Phi and Cramers V = 0.371, p < 0.001). While 48.6 % of Mogotho households usually collect papyrus shoots for eating, only 2.8% in Nxamasere do. Many people that collected papyrus for eating said they did so when out fishing or collecting other resources. In 2001, no households in Nxamasere collected, and only 11.1 % (N = 1) of those that rely on the resource in Samochima did. The median bundle equivalent collected in Sekondomboro was 13, compared to 5 in Mogotho. As expected, fewer households in Nxamasere were involved in fishing than in the other villages (see Table 24). In Sekondomboro and Samochima, the proportion of households actually catching fish in 2001 was much lower. The mean number of fish

PAGE 93

78 caught by households in Nxamasere in 2001 was lower, but not significantly so at the 0.05 level. Mean and median values for 2001 fishing households are shown in Table 25. There were no significant differences between villages. Table 24. Households usually fishing by village. Village Households involved in fishing Households fishing in 2001 Households selling fish in 2001 Mogotho 31.4% 31.4% 20.0% Sekondomboro 50.0% 38.9% 13.8% Samochima 41.7% 19.4% 11.1% Nxamasere 27.8% 27.8% 5.6% Table 25. Fish caught in 2001 by village. Village Proportion of households fishing Mean no. of fish caught Median no. of fish caught Sample sum amount Mogotho 31.4% 1209.6 1200 13305 Sekondomboro 38.9% 1498.5 280 20979 Samochima 19.4% 1294.3 600 9060 Nxamasere 27.8% 571.4 75 5714 Total 29.4% 1168.0 550 49058 Likewise, there were no significant differences between villages in amounts of fish sold, although Samochimas higher mean figures may reflect that some of the households are members of the fishing co-operative.

PAGE 94

79 Table 26. Fish sold in 2001 by village. Village Proportion of households selling Mean no. of fish sold Median no. of fish sold Sample sum amount Mogotho 20% 1082.9 1200 7580 Sekondomboro 13.8% 940.0 850 4700 Samochima 11.1% 1912.5 800 7650 Nxamasere 5.6% 1505.0 1505 3010 Total 12.6% 1274.4 925 22940 Wealth Variations in Reliance on Key Resources Wealth, as defined using the wealth index described in Chapter 2, is not equally distributed across households. Most households have a small proportion of the total wealth within the study area, as is shown in Figure 25. Because other studies have shown that wealth is related to household size, the first thing to test is whether there is a positive correlation between the wealth index and the numbers of family members. While there is no significant correlation for overall household size (Pearson = 0.068, one-tailed p = 0.200)., the correlation figures for wealth index vs. number of adults (15 years and over) are stronger (Pearson = 0.120, one-tailed p = 0.076), although not significant at the 95% level. Household sizes in all villages were in fact slightly larger than recorded in earlier studies (see Chapter 1), and are shown in Table 27. If this were due to household consolidation as a result of AIDS, that would explain the limited correlations with wealth. For any particular resource, there are no significant differences in the mean wealth of households that are involved in the collection of it, compared to the mean wealth of households that are not. Only the sale of thatching grass was significantly related to wealth (Pearson = 170, p = 0.041).

PAGE 95

80 01020304050607080901001112131415161718191101111121131141Number of householdsPercent of total wealth Figure 25: Percentage distribution of wealth over households. The blue line represents how equal distribution of the percentage of total wealth (the summation of the WI for all cases in the survey) would appear, while the pink line shows the actual distribution. The distribution has a Gini co-efficient of 0.7303, where equal distribution = 0 and completely unequal distribution = 1. Fifty percent of the total wealth is owned by 7 of the 143 surveyed households. Table 27. Household size by village. Village Mean Median Mogotho 7.54 7.00 Sekondomboro 8.50 8.00 Samochima 8.08 7.00 Nxamasere 7.06 7.00 Total 7.80 7.00

PAGE 96

81 There are no significant correlations between wealth and the amounts of each of the resources that households collected in 2001. However, there is a significant negative correlation between wealth and the proportional contribution of wetland resources to the household livelihood (t = -2.361, p = 0.021). Given the importance of cattle to wealth, it is not surprising that there was also a strong inverse association between number of cattle owned and proportional contribution to household livelihood from wetland resources (t = -0.172, p = 0.058). Household size was significantly linked to households that usually collected reeds (F = 4.583, p = 0.034) and that usually made baskets (F = 6.536, p = 0.012). While there was a significant correlation between fish collection and household size (Pearson = 0.403, p = 0.019), there were no strong relationships between household size and plant resources collected or sold. Gender Variations in Reliance on Key Resources The high proportion of households headed by women in their own right (de jure) was unexpected (see Table 28) given patterns elsewhere in Botswana. For example, the dominant Tswana tribes are patrilineal and traditionally women have not been recognized as decision-makers. This means that until recently, there have been extremely few households with de jure female heads, although females heading households in their husbands absence (de facto) have been more common. In the study area, respondents revealed that in many instances, male partners to the female heads were present, but since the couple was not married, and lived in the womans house, she was considered both the head and the decision-maker that is, the de jure head. While this may in part be due to the breakdown in marriage traditions, it has an original basis in the groom living with the bride at her parents homestead for the period (often years) during which bridewealth was

PAGE 97

82 paid (Larson 1980). The average wealth (see Table 29) of male headed households was greatest, followed by female de facto and then female de jure, but the differences were not significant. However, one aspect of wealth that is important is ownership of a dugout canoe, because of the role it plays in access within the panhandle. A higher proportion of male-headed households than female headed households own canoes (Phi and Cramers V = 0.203, p = 0.053). Table 28. Gender distribution of household heads. Head of Household Proportion Male 46.2% Female de jure 51.0% Female de facto 2.8% Table 29. Wealth index by gender of household head. Mean Median Min Max Std Dev Male 26252.5 9272.5 360.0 441905.0 61195.5 Female de facto 24117.5 8457.5 6305.0 73250.0 32816.9 Female de jure 22055.34 6270 0.0 302875.0 51666.6 Total 24050.17 7650 0.0 441905.0 55630.7 Looking at labor-availability as determined by household size, on average male-headed households are largest, and female de jure households are smallest. In this case, while the difference may not seem much, it is significant (t = 2.598, p = 0.010). Again, household size is considered important as it plays a role in maximizing access to resources. The significance of the difference increases when looking at adult members (t

PAGE 98

83 = 3.147, p = 0.002). Gender of household head does not vary significantly across the different villages or language groups. Table 30. Total household size by gender of household head. Mean Median Min Max Std Dev Male 8.8 7.0 1.0 32.0 5.2 Female de facto 7.3 7.0 3.0 12.0 3.8 Female de jure 7.0 6.0 1.0 12.0 2.7 Total 7.8 7.0 1.0 32.0 4.1 Table 31. No. of household members 15 years and over by gender of household head. Mean Median Min Max Std Dev Male 5.26 5.0 1 14 2.5 Female de facto 3.75 3.5 2 6 1.7 Female de jure 4.10 4.0 1 7 1.8 Total 4.62 4.0 1 14 2.2 Neither does gender of household head appear to play a role in determining the proportional contribution of wetland resources to the household. However, there are some differences in the extent of use of some resources. For example, there is very little difference in the mean number of cattle owned by the different household types, although there is a significant difference in whether livestock are grazed on the floodplains or not (Phi and Cramers V = 0.298, p = 0.013). Some respondents stated that floodplain grazing is for poorer people who dont have cattleposts located away from the village. While 70.5% of female de jure headed households graze their livestock on the floodplain, only 41.2% of male-headed, and 33.3% of female de facto headed do.

PAGE 99

84 Since nearly all households usually collect thatching grass, it is not surprising that there is no significant gender difference regarding participation in this activity. Amongst those who collected in 2001, however, there were strong differences in the mean amounts of thatching grass collected (F = 3.066, p = 0.051), with male-headed households collecting almost twice as much on average as female headed households. Reed collection is similar to that of thatching grass. However, although the mean number of reed bundles collected was greater for men, this was not significantly so. Table 32. Bundles of thatching grass collected by household head. Household head % hholds collecting Mean no. collected Median no. collected Sample sum amount (bundles) Male 75.8% 40.1 25.0 2004 Female de facto 100% 20.5 20.0 82 Female de jure 71.2% 21.5 20.0 1119 Total 74.1% 30.2 20.0 3205 Table 33. Bundles of reeds collected by gender of household head. Household head % hholds collecting Mean no. collected Median no. collected Sample sum amount (bundles) Male 77.3% 37.2 30.0 1898 Female de facto 100% 23.5 24.0 94 Female de jure 60.3% 24.6 20.0 1080 Total 69.3% 31.0 20.0 3072 There was no discernable association with selling either thatching grass or reeds and gender of household head. While a higher proportion of male-headed households than female-headed households usually collected palm leaves, papyrus, and fish, this was

PAGE 100

85 not significantly so. More female de jure headed households usually collected water lily bulbs, but the difference was not significant. In terms of quantities both collected and sold, there were no s associations between gender of household head and mean amounts. Ethnic Variations in Reliance on Key Resources There is a very strong relationship between ethnicity and spatial distribution (Phi = 0.624, Cramers V = 0.361, p < 0.001). This is important because it is often the location that explains use of a resource more than do different cultural practices. The Hambukushu are most numerous in all settlements (see Figure 26). NxamasereSamochimaSekondomboroMogotho403020100 BakgalagadiXerekuBugakhweXanikhweBatswanaBayeiHambukushu Figure 26. Distribution of ethnic groups by village. The Xereku live only in the northernmost villages, while the Bayei and Batswana live in the villages on the southwest side of the panhandle. The distribution of the ethnic groups reflects the direction from which they migrated. Ethnicity of the household was determined on the basis of main language spoken in the household.

PAGE 101

86 The Xereku, who were later arrivals from the north, are only found in the northern villages, while the Bakgalagadi are only found in the western villages.2 The bars show the sample distribution of household ethnicity across the four villages. The San households (Bugakhwe and Xanikhwe) had larger households, but there were no significant differences in household size between the 6 ethnic groups. The average wealth of the San was lowest, with Batswana and Bayei having the highest average wealth, but again, not significantly so. There was considerable variation according to ethnic group in the mean proportional contribution of wetland resources to households, but the differences were not significant at the 95% level (F = 1.556, p = 0.165). It is interesting to note that the poorer San households (Bugakhwe and Xanikhwe had the highest proportion of their livelihoods coming from wetland resources). The variations in values are shown in Table 34 and Figure 27. Table 34. Proportional contribution of wetland resources to households in 2001 by ethnic group. Ethnic Group Mean contribution to household livelihood in 2001 Median contribution household livelihood in 2001 Hambukushu 13.6% 4.6% Bayei 1.6% 1.3% Batswana 2.2% 5.3% Xanikhwe 30.1% 14.4% Bugakhwe a 31.7% 31.7% Xereku 19.1% 14.6% Bakgalagadi 11.4% 3.2% a. The similarity between the mean and median values for the proportional contribution of wetland resources to Bugakhwe livelihoods is probably a factor of the small number of Bugakhwe households in the sample. 2 It must be remembered that the 4 focal villages are not the only villages in the panhandle. However, the findings follow the pattern of settlement that has been previously documented (see literature in Chapter 1).

PAGE 102

87 BakgalagadiXerekuBugakhweXanikhweBatswanaBayeiHambukushuProportion of livellihood1.00.90.80.70.60.50.40.30.20.100.00-.10 Figure 27. Proportional contribution of wetland resources to household livelihood in 2001 by ethnic group. The boxes represent the interquartile ranges that contain 50% of the values for each ethnic group. The line across the box is the median value. The whiskers extend to the lowest and highest values, excluding outliers. Open circles indicate outliers, while extreme outliers are shown by stars. There is a very strong association between ethnicity and the ownership of livestock of any kind (Phi and Cramers V = 0.356, p = 0.006), with Bakgalagadi and Hambukushu having the highest proportion owning, and no Bugakhwe households (N = 2) owning any livestock at all. However, once livestock ownership is established, there is no association between ethnicity and the number of cattle owned, nor in use of the floodplains for grazing of livestock.

PAGE 103

88 Table 35. Livestock ownership and use of floodplain for grazing in 2001 by ethnicity. Ethnic group Proportion of households keeping livestock Proportion of livestock owners using floodplains for grazing Hambukushu 89.3% 50% Bayei 66.7% 100% Batswana 66.7% 50% Xanikhwe 62.5% 100% Bugakhwe 0.00% n/a Xereku 85.7% 40% Bakgalagadi 90.0% 75% Total 85.3% 54.1% As is shown in Table 36, mean amounts of thatching grass collected by different ethnic groups were significantly different (F = 19.014, p < 0.001), even when filtering out the two categories where N = 1 (F = 3.025, p = 0.021). Bakgalagadi collected the least, while Xanikhwe collected the most. There is however no relation between the sale of thatching grass and ethnic group. Table 36. Bundles of thatching grass collected in 2001 by ethnicity. Ethnic group Proportion of households collecting Mean no. of bundles collecting Median no. of bundles collecting Sample sum amount Hambukushu 78 24.78 20.00 1933 Batswana 1 300.00 300.00 300 Xanikhwe 8 49.50 20.00 396 Bugakhwe 1 20.00 20.00 20 Xereku 12 37.17 22.50 446 Bakgalagadi 6 18.33 17.50 110 Total 106 30.24 20.00 3205

PAGE 104

89 The strong association between ethnicity and the harvesting of palm leaves appears to be an artifact of village location: when controlling for location the relationship is not statistically significant. The mean number of bundles for the 4 groups that collected in 2001 ranged from 2.2 bundles to 4.71. Table 37. Collection of palm leaves by ethnic group. Ethnic group Usually collect Collecting in 2001 Hambukushu 22.3% 13.6% Bayei 66.7% 33.3% Batswana 66.7% 0% Xanikhwe 50.0% 50.0% Bugakhwe 0% 0% Xereku 7.1% 0% Bakgalagadi 70.0% 50.0% Baskets made and sold did not differ much between households of different ethnicity. Although basket-making has come to be known as a Hambukushu skill, all ethnic groups make and use different kinds of baskets. Collection of water lily bulbs was also associated with ethnicity (Phi and Cramers V = 0.294, p = 0.051), but in this case does not seem to be due solely to location. There were also significant variations in the mean amounts collected by the different ethnic groups (F = 5.685, p = 0.002) as is shown in Figure 28. The Xanikhwe, who are also known as the River San, consider water lily bulbs a delicacy, and respondents explained how they used a hook to pull up the bulbs when water was too deep to use feet or hands.

PAGE 105

90 BakgalagadiXerekuBugakhweXanikhweHambukushuNo. of dishes of water lily bulbs706050403020100-10 Figure 28. Quantities of water lily bulbs collected in 2001 by ethnic group. The boxes represent the interquartile ranges that contain 50% of the values for each ethnic group. The line across the box is the median value. The whiskers extend to the lowest and highest values, excluding outliers. Open circles indicate outliers. Groups not usually collecting water lily bulbs (Bayei and Batswana) are not shown. Table 38. Collection of water lily bulbs by ethnic group. Ethnic group Usually collect Collecting in 2001 Hambukushu 28.2% 18.4% Bayei 0% 0% Batswana 33.3% 0% Xanikhwe 62.5% 37.5% Bugakhwe 50.0% 50.0% Xereku 57.1% 42.9% Bakgalagadi 60.0% 30.0%

PAGE 106

91 Table 39 shows that the variation in involvement in fishing for the different ethnic groups. Xereku are far more likely to fish than any other ethnic group (Phi and Cramers v = 0.402, p < 0.001). Even though the Bugakhwe consider themselves dryland San, it is likely that they do rely to some extent on fishing, and that the zero value in Table 39 is due to the small number of Bugakhwe households in the sample. On the other hand, the Batswana in the study area do not have a history of fishing. In terms of quantity, Xanikhwe rely more heavily on fish, both for own consumption and for sale, although not significantly so (see Figure 29). Table 39. Involvement in fishing by ethnic group. Ethnic group Usually catch Catching in 2001 Hambukushu 34.0% 24.3% Bayei 33.3% 33.3% Batswana 0% 0% Xanikhwe 37.5% 37.5% Bugakhwe 0.0% 0.0% Xereku 92.9% 71.4% Bakgalagadi 20.0% 20.0%

PAGE 107

92 BakgalagadiXerekuXanikhweBayeiHambukushuNo. of fish caught11000100009000800070006000500040003000200010000-1000 A BakgalagadiXerekuXanikhweBayeiHambukushuNo. of fish sold70006000500040003000200010000-1000 B Figure 29. Numbers of fish caught and number sold in 2001 by ethnic group. The boxes represent the interquartile ranges that contain 50% of the values for each ethnic group. The line across the box is the median value. The whiskers extend to the lowest and highest values, excluding outliers. Open circles indicate outliers, while extreme outliers are shown by stars. Bugakhwe and Batswana are not shown in the graph because in the survey sample households from these ethnic groups were not involved in fishing.

PAGE 108

93 Temporal Distribution of Resource Collection To a large extent, water determines accessibility of resources. Collection of most of the plant resources can only be done once the floods have receded enough. Only 25.9% of households own a dugout canoe, allowing them earlier and further access into the panhandle. Many respondents in the informal interviews cited this as a reason for why some households are able to collect resources sooner than others. Grazing on the floodplains usually takes place from July to December. Not only is this related to the drying out of the shallow flooded areas, but it also corresponds to the end-of-season reduction of available grazing on the sandveld. Collecting reeds and thatching grass is strictly seasonal. While these plants are ready for harvesting by late June, most people harvest in August and September, as many of the reed and grass beds are inaccessible until then. Depth of water also determines when water lily bulbs are collected. Bulbs are usually collected in waist-deep, open water by women. Typically bulb collection is concentrated around July and August, but some households go every week of the year, moving to where conditions are suitable. The relation between the flood, time of year and harvesting of resources is summarized in Table 40. Palm leaves are collected all year round. The Hyphaene palms grow on the dryland edges of the panhandle as well as on islands. Where conditions are suitable, such as around Nxamasere, it is usually possible to access some of these plants by foot most months of the year. Papyrus for eating is typically collected opportunistically by fishermen and children. While young shoots can be found throughout the year, most of the sprouting is in June and July. In Nxamasere, people say they collect these papyrus shoots when the flood is full. In the river villages of Mogotho and Samochima, collection of papyrus for mats is concentrated in September, October and November,

PAGE 109

94 although some harvesting takes place in all months. In Sekondomboro most people collect in December, while in Nxamasere papyrus for mats is collected in August and September, when the water level starts to drop. Figure 30. Map showing change in open water between the 2001 wet and dry seasons. Note that this map shows only open water. Much of the rest of the study area stays flooded year round, but is covered in dense aquatic vegetation that is largely impenetrable.

PAGE 110

95 Table 40. Seasonal Calendar of Flood and Collection of Main Resources. Season Month State of Annual Flood Main resources women responsible for Main resources men responsible for November First pulse of flood arrives Papyrus for mats Grazing livestock on floodplain Fishing December Water levels rise in main channels, and start to spill out into floodplains Grazing livestock on floodplain Fishing January Second pulse of floods arrives Fishing Summer February Fill up floodplains Fishing March Fill up floodplains Fishing April Floods at peak in northern villages Fall May Floods at peak in southern villages June Water flow slows on floodplains July Water levels begin dropping on floodplains Water lily bulb collection Grazing livestock on floodplain Winter August Water continues to recede Water lily bulb collection Collecting thatch and reeds Grazing livestock on floodplain Fishing September Water drops rapidly Collecting thatch and reeds Papyrus for mats Grazing livestock on floodplain Fishing Spring October Water confined to main channels and deeper lagoons on floodplains Papyrus for mats Grazing livestock on floodplain Fishing

PAGE 111

96 Fish are caught all year round by some, but there are distinct seasons. December/January is when the fish are most plentiful. Line fishermen dont fish in April/May when the flood is at its highest and the fish are most dispersed, because returns for effort are lower. Some fishermen say that even in June the numbers of fish are low because of the cold. Most fishermen using nets or lines fish for 6 months of the year, from August to March. Discussion Theme Reasons for Temporal Variation in Collecting Plant Resources During the informal interviews people were asked what determined the timing of collecting plant resources. There was a range of responses, as is shown in the quotations below. Some people go earlier than others to collect reeds from the delta because they want good reeds, and rush when there is a chance to collect the nearest reeds. Others do not have a [dugout canoe] to use and others are sick or have funeral to attend, that may be why they go later to collect. People usually look at the condition of their houses and compound walls to decide if they need to go earlier to collect. If their walls and roofs are old, they rush so that they can repair their houses before the rainy season. Some also go earlier in order to get the nearest grasses. But the water level also controls the time they go to collect. Floodwater controls the time of the year that people go to collect building materials. We have to wait till the level of water is low so that we can reach the beds easily without fear of crocodiles and pythons. We usually start going around August until December. Those who go earlier start in July they are usually young people! Older men and women usually go late, around September. Those who cut late do so because they are afraid of the water its too deep. Those who go early want to get quality reeds its a competition. Those who are brave

PAGE 112

97 enough to deal with deep water get the best reeds. People do compete for the best reeds, this may be why some people go earlier to get reeds of better quality. High floodwaters and crocodiles may slow others from going out, and they must wait until the water levels have become very low (e.g. around August or September). Those with thatching grass and reeds which are old usually go there earlier so that they can fix their homes before the rainy season starts. Some people go early to harvest depending on if they need new stuff for their houses. Some want to collect palm leaves, so they go later. Some go earlier to get the closer ones, and others are afraid of the water. One reason why some people may go later to collect stuff is if they have some left over from the previous year. If not, then you go early. Some people may just be lazy. Those without canoes have to wait either to borrow one when the owner is finished, or until the water is low. Those who have canoes can go earlier. People have to collect from close by because they have no transport and carry things on their heads. Sometimes we have to wait for weekends so that the children can help us carry. The main school holiday is in August. When your family is big, you can collect more, because they can help carry. Small families get little. Some people go early so that they can collect enough to sell, so that they can buy soap and school uniforms. Some go there earlier so that they can come back and attend their fields earlier. Tswii (water lily bulbs) is usually collected from August to November when the water level has gone down. People collect at this time because of the crocodiles. When the floodwater is low, the crocodiles stay in the main river, thereby giving people a chance to collect tswii in the lagoons.

PAGE 113

98 Decision tree model collecting reeds or grass earlier or later Figure 31. Decision tree determining early or late collection of reeds and thatch.

PAGE 114

99 As the quotations presented above show, there is a range of factors that influence the decisions household members make about when they go and collect reeds and grass. These factors, summarized from the informal interviews, can be presented as a decision tree model. It is interesting to note how some of these factors, such as ownership of a canoe, and household size, correspond to variables from the household survey that are strongly correlated to the collection of these resources. Effect of Fire on Access to Key Natural Resources As noted earlier, in 2001, fire was mapped as occurring over approximately 3690 ha, or just over 4% of the study area. The proportions of actual resources burnt, which were shown graphically earlier in this chapter, are presented in Table 41. Table 41. Extent of resources burnt within village resource areas in 2001. River villages Floodplain villages Village Mogotho Samochima Nxamasere Sekondomboro Total Area of papyrus/ reeds/ thatch burnt in range (ha) 11.84 3.84 81.69 13.72 111.09 Proportion of papyrus/ reeds/ thatch burnt in range 1.74% 0.50% 5.58% 1.05% 6.58% Area of grazing resources burnt within range (ha) 17.01 10.90 184.55 111.72 324.18 Proportion of grazing resources burnt within range 1.37% 0.94% 5.42% 3.91% 8.87% There was less burning in the resource areas of the river villages compared to the floodplain villages. While the proportions of resources in all the village resources areas

PAGE 115

100 affected by fire appear mathematically small, 25.9% of all households experienced fire in at least one resource. There was a strong association between experiencing fire in one resource and in another. How people perceived the effects of fire does not appear to be related to whether they experienced fire or not: In both cases the majority overall favored burning: 61.5% of all those that had experienced fire in 2001, and 56.4% of all those who had not, said that fire was generally good for resources. Effect of Fire on Grazing Of the 37.1% of households that uses the floodplains for grazing, 24.5% said there had been a fire in their grazing area in 2001, and 15.1% said it had been set intentionally. Of those who said this, 72.7% felt that the reason people had set the fire was to improve grazing. No-one lost any livestock due to the fire, and 69.2% believed that their livestock had benefited from improved grazing conditions as a result of the fire. There was a significant association between village and whether fire had been experienced in grazing areas or not (Phi and Cramers V = 0.416, p = 0.027). Higher proportions of households in the two northern-most villages, Sekondomboro and Samochima, experienced fire than in Mogotho and Nxamasere. This is in contradiction of the finding from the satellite imagery interpretation, which shows the greatest area burnt was in Nxamasere. One possible explanation for this may be data quality. Survey respondents in Nxamasere may have been less aware of fire than those in the other villages. This is because the floodplains at Nxamasere lie about 6 km from the village where the interviews were conducted. In addition, because livestock are grazed some distance from the village, those household members who normally herd cattle and who might have had greater awareness of burning, were less likely to have been in the village responding to the survey.

PAGE 116

101 Effect of Fire on Availability of Thatching Grass Of the households usually collecting thatching grass, 25.5% said there had been fire in their collection area in 2001. Most of this subset of respondents (60%) said they experienced fire in September, but some fires were reported as occurring in August. This is important, because 40% of the 25% experiencing fire said that it destroyed grass-beds before they had finished collecting. However, 67.6% said that the thatching grass was growing better as a result of the fire, and 32.3% said both that it destroyed before they collected, and that fire improved its growth. While this seems contradictory, it reflects temporal changes. Immediately after the fire experienced, there was a reduction in availability of thatching grass, but after a season of growth, respondents felt that the fire had stimulated improved quality and quantity of grass for the coming years harvest. noyesNo. of grass bundles collected1501401301201101009080706050403020100 Figure 32. Numbers of bundles of thatching grass collected by whether fire had been experienced or not in 2001. The boxes represent the interquartile ranges that contain 50% of the values for each group. The line across the box is the median value. The whiskers extend to the lowest and highest values, excluding outliers. Open circles indicate outliers, while extreme outliers are shown by stars. For those households that had experienced fire in 2001, the number of bundles collected was higher than for those that had not.

PAGE 117

102 The mean number of bundles of thatching grass was higher, and less variable for those households who said there had been a fire in the area, than for those who said there wasnt (t= 1.674, p = 0.105, unequal variances). There were no significant differences between mean amounts collected and whether people said that fire had destroyed grass-beds before they had finished collecting, nor with whether they said that fire had improved the thatching grass. Effect of Fire on Availability of Reeds Those experiencing fire in reed beds in 2001 comprised 26.5% of households usually collecting reeds. Because reeds and thatching grass collection is highly correlated, and because collection usually takes place in the same locations, it is not surprising that some of the responses on experience of fire are similar to those described above, especially in terms of source and timing of fires. In almost all instances, respondents who experienced fire in both reeds and thatching grass described a single event affecting both resources. However, fewer people (36.1%) felt that fire had destroyed reed-beds before they had finished collecting, and more (74.3%) believed that fire had improved the reeds. There was also a positive association between people feeling both that reeds had been destroyed, and that they had been improved by the fire event (Pearsons chi-square = 3.52, p = 0.61, but Fishers exact test p = 0.066 one-sided). The mean number of reed bundles collected by those who experienced fire compared to those that did not was very similar (29.6 to 31.6 bundles). However, unlike thatching grass, the mean number of bundles of reeds was smaller for those who said that fire had destroyed reeds before they finished collecting, and bigger for those who said that fire had improved reeds.

PAGE 118

103 051015202530354045MogothoNxamasereSamochimaSekondomboroPercent of households Fire in thatching grass Fire in reeds Figure 33. Similarities in reports of fire in thatching grass and reeds in 2001. In most cases, households reported that the fire experienced in the reed bed where they normally collect was the same one experienced in the usual thatching grass collection area. Effect of Fire on Access to Palm Leaves, Water Lily Bulbs and Papyrus Only three of the 39 harvesters in the sample reported fire in the young palm trees where they collected in 2001. Respondents explained that the palms were found at the inside of larger islands, and that wetland fires rarely penetrated through the riparian trees to reach the palms. The respondents did not know how the fires had started. Although the size of the sample subset is small for meaningful statistics to be computed, the mean size of bundles collected by those experiencing fire was smaller than that of households that did not. Water lily bulb collectors tend to collect from shallower pools of open water within the sedge and short grass areas suitable for grazing, or at the edges of deeper lagoons or

PAGE 119

104 channels. 24% of collectors said there was fire in 2001 in places where they usually collect. There were no significant differences between the mean amounts collected of those who reported fire versus those who did not. None of the respondents felt that the fire had any bad effects, and 91.7% believed that the fire had improved access to the water lily bulbs. In the informal interviews it was explained that the fire does not affect the plants directly, since they are submerged in the water. Instead, it removes the dense hippo grass that floats on top of the water and closes over open pools. Once the hippo grass is removed, the water lilies not only get sunlight, but also have room to grow. In the households that usually collect papyrus for mats, 16.9% were aware of fire in the area where they normally collect. In terms of the perceived effect of fire, 77.8% said that the fire had not prevented them from harvesting enough, and 60% believed that fire had improved the papyrus. Looking at papyrus collected for eating, 23.1% of collectors experienced fire where they take shoots. Of this group, 88.9% said that it had not been destroyed by the burning, and 77.8% said that the fire had improved the papyrus shoots. Effect of Fire on Access to Fish For some respondents, it was difficult to accept that fire could affect an organism that was completely inside, and therefore protected by, water. Nevertheless, 17.6% of households with fishers felt that in 2001 fire had affected the availability of fish. Within this subset, 66.7% believed the numbers had increased as a result of burning, while 22.2% believed their numbers had decreased. The main reason cited both during the household survey and the informal interview was that fire stimulated growth of new shoots that were palatable for fish. In addition, fire removed vegetation leaving the sand clear for the tilapia to make their nests once the flood filled the floodplains.

PAGE 120

105 noyesNo. of fish caught120001000080006000400020000-2000 Figure 34. Numbers of fish caught categorized by whether fire had been experienced or not in 2001. The boxes represent the interquartile ranges that contain 50% of the values for each group. The line across the box is the median value. The whiskers extend to the lowest and highest values, excluding outliers. Open circles indicate outliers, while extreme outliers are shown by stars. There was no significant difference in the number of fish caught by those households that thought fire had affected fish numbers in 2001 compared to those who thought it had not. A further 11.1% felt that while the actual numbers had not increased, fires made it easier to catch fish. Nevertheless, there was no significant difference in the numbers caught in the past year between those who thought there had, and those who thought there hadnt, been an affect on the number of fish, as is shown in Figure 34. Effect of Fire on Access to Wildlife People were very reluctant to talk about hunting within the panhandle. This is because hunting is illegal without a license, and licenses are not given for hunting within

PAGE 121

106 this section of the wetland area. Only 1.4% of households claimed to do this. It is therefore difficult to assess how fire affects access to this resource. Hunters are cited as one of four resource-user types believed to be responsible for fires set in 2001, but this information cannot be substantiated. How much wetland wildlife both animals and birds is still considered a resource by most people is unknown. From informal interview responses, it appears as if wildlife is seen less as part of the livelihood support system, and more as either a competitor or minor casualty in the use of fish and plant resources, as is shown in the quotations below. Crocodiles are the most destructive creatures in the river because they get trapped in our fishing nets and cut them, making large holes which allow fish to just pass through, therefore our production is brought down. They even eat the fish that they find already trapped there. In addition, otter also cut the nets and feed on the trapped fish. Hippos too break our nets by displacing them. (In the old days hippos were few in number compared to these days.) Turtles also eat the trapped fish in the net and bring our catch down. Most of the people like the use of fire because its impacts are profitable to them, and are only bad for animals and birds. Fires are bad for animals and birds because it burns when they are breeding. Effect of Fire on Livelihood Sustainability The preceding sections have revealed that spatial location, ethnicity, gender of household head, ownership of key assets, large households, and other income opportunities are all associated with resource collection. In addition, use of one wetland resource is often strongly linked to use of others. Reeds and thatching grass are central to most peoples lives. Multiple regressions run to examine the three-stage causal path described in Chapter 2 show the relative importance of these factors in determining the

PAGE 122

107 access to resources, as well as the relative importance of fire with regard to livelihood sustainability. 3 The first stage of the analysis examined number of adults as a dependent of gender of household head and ethnicity was carried out. This regression shows that the model is significant (F = 2.015, p = 0.068), but does not fit the data too well (R2 = 0.082). Values showing the contribution of each independent variable are given in Table 42. Likewise, gender of household head and ethnicity determine the number of wage jobs in the household (F = 2.279, p = 0.040) but much of the variance is not explained by the model (R2 = 0.091). Batswana are likely to have significantly more jobs than Hambukushu, the referent ethnic group, while San have fewer as is shown in Table 43. The value for San is not surprising, because they are known to be marginalized. Gender of household head is also significant, but only at the 0.1 level. Results for Stage 1 of the causal path analysis are presented in Figure 35. As the bivariate results have shown, the number of cattle owned are not dependent on ethnicity or gender of household head (F = 0.3156, p = 0.928, and R2 = 0.014). This is probably because any association is more to do with whether the household keeps livestock at all than with the actual numbers kept. 3 The village of Mogotho was excluded from these analyses because it was the reference constant for the villages. With regard to ethnicity, the Hambukushu group was excluded for the same reason.

PAGE 123

108 Table 42. Regression table for effect of social characteristics on no. of adults in the household. Coefficientsa 4.129 .282 14.648 .000 1.054 .368 .237 2.864 .005 -1.813 1.280 -.117 -1.417 .159 .344 .723 .040 .476 .635 -7.7E-02 .624 -.010 -.123 .903 .339 .724 .039 .468 .640 -1.165 1.280 -.075 -.910 .365 (Constant) Gender of hh head Bayei San Xereku Bakgalagadi Batswana Model 1 B Std. Error UnstandardizedCoefficients Beta StandardizedCoefficients t Sig. Dependent Variable: No. of adults in householda. Table 43. Regression table for effect of social characteristics on no. of wage jobs in household. Coefficientsa .436 .095 4.578 .000 .216 .124 .143 1.739 .084 -.175 .433 -.033 -.405 .686 -.345 .245 -.116 -1.410 .161 -.157 .211 -.062 -.742 .459 .134 .245 .045 .546 .586 1.086 .433 .206 2.508 .013 (Constant) Gender of hh head Bayei San Xereku Bakgalagadi Batswana Model 1 B Std. Error UnstandardizedCoefficients Beta StandardizedCoefficients t Sig. Dependent Variable: No. of wage jobs in householda.

PAGE 124

109 Figure 35. Results of the Stage 1 multiple regression. Gender and ethnicity explain two important sources of wealth for households: number of adults and number of wage jobs.

PAGE 125

110 Stage 2 of the analysis looks at the factors influencing wealth, as represented by MWI. While offering only partial explanation of the variance (R2 = 0.375), the regression is highly significant (F = 27.814, p < 0.001). Number of cattle owned is the most significant factor related to wealth (t= 6.586, p < 0.001), followed by number of wage jobs (t = 3.415, p = 0.001). These results are shown in Table 44 and Figure 36. Table 44. Regression table for factors influencing wealth in 2001. Coefficientsa 573.738 2085.550 .275 .784 264.320 416.920 .044 .634 .527 4441.612 1300.676 .250 3.415 .001 279.674 42.466 .470 6.586 .000 (Constant) No. of adults in hh No. of wage jobs in hh No. of cattle owned by hh Model 1 B Std. Error UnstandardizedCoefficients Beta StandardizedCoefficients t Sig. Dependent Variable: Modified Wealth Index (MVI)a. The causal model outlined in Chapter 2 was designed to test whether fire is an important determinant of a households access to resources, and whether it affects not only the amount of resources accessed by the household, but also the diversity or range of types of resources available to it. The effect of fire on the proportional contribution to the households livelihood (expressed in the causal path analysis as the MWI) also needed to be tested. By including other factors that are known to affect access to resources, such as location, ownership of canoe, ethnicity (through cultural preferences, number of adults and wealth, the relative contribution of fire to the models explaining various aspects of resources access can be shown. Multiple regressions based on these variables were run as Stage 3 of the causal path analysis.

PAGE 126

111 Figure 36. Results of the Stage 2 multiple regression. The importance of cattle and wage employment is clear.

PAGE 127

112 The first Stage 3 regression looked at the range of resources used. While the overall model is not significant (F = 1.252, p = 0.255, R2 = 0.104) there are some interesting individual coefficients. Results in Table 45 show that the effect of fire is significant at the 0.2 level (t = 1.424, p = 0.157). Also, households collected a greater range of resources at Nxamasere than at Mogotho, the reference constant (t = 1.332, p = 0.185). Table 45. Regression showing relative contribution of fire to range of resources accessed in 2001. Coefficientsa 3.067 .399 7.692 .000 6.858E-02 .063 .094 1.085 .280 1.102E-05 .000 .091 .954 .342 .563 .423 .151 1.332 .185 .148 .415 .040 .357 .722 8.899E-03 .421 .002 .021 .983 -5.63E-03 .969 .000 -.006 .995 .544 .550 .086 .990 .324 -.557 1.061 -.049 -.525 .601 .321 .510 .059 .630 .530 .624 .568 .098 1.100 .274 .412 .325 .111 1.267 .207 .476 .334 .129 1.424 .157 (Constant) No. of adults in hh MVI Nxamasere Samochima Sekondomboro Bayei San Batswana Xereku Bakgalagadi Canoe Fire experienced Model 1 B Std. Error UnstandardizedCoefficients Beta StandardizedCoefficients t Sig. Dependent Variable: No. of wetland resources useda.

PAGE 128

113 Figure 37. Results of the first of the Stage 3 multiple regressions: range of resources collected.

PAGE 129

114 After looking at the range of resources, the actual amounts of the main wetland resources were also examined, resulting in several regressions being run. The first one uses the amount of fish caught by the household as the dependent variable. This regression has a good overall fit (F = 2.005, p = 0.052) but with some variance unaccounted for (R2 = 0.344). Fire experience has an insignificant (but positive) effect on the amount of fish caught (t = 0.359, p = 0.721), as is shown in Table 46. Explanatory variables of far greater importance are the number of adults in the household (t = 3.174, p = 0.003) and wealth (t = 1.576, p = 0.123), although the latter is only significant at the 0.2 level. Table 46. Regression showing relative contribution of fire to amount of fish caught in 2001. Coefficientsa -1190.596 788.479 -1.510 .139 367.203 115.708 .414 3.174 .003 3.084E-02 .020 .215 1.576 .123 -561.080 807.641 -.122 -.695 .491 -245.693 746.232 -.061 -.329 .744 4.367 761.796 .001 .006 .995 400.966 1783.608 .030 .225 .823 2049.954 1035.084 .262 1.980 .054 171.907 642.411 .041 .268 .790 99.288 1298.169 .010 .076 .939 430.834 520.751 .118 .827 .413 203.614 566.522 .055 .359 .721 (Constant) No. of adults in hh MVI Nxamasere Samochima Sekondomboro Bayei San Xereku Bakgalagadi Canoe Fire experienced Model 1 B Std. Error UnstandardizedCoefficients Beta StandardizedCoefficients t Sig. Dependent Variable: No. of fish caughta.

PAGE 130

115 Figure 38. Results of the second Stage 3 multiple regression: number of fish caught.

PAGE 131

116 The third regression in Stage 3 uses the amount of reeds collected by households in 2001 as the dependent variable. The factors that contribute most to this model are ethnicity and location (see Table 47). While it makes sense that fewer reeds were collected at Nxamasere because it is a drier place, it was unexpected to find that Samochima households also collected fewer reed bundles. The fit of the data to this regression model values is however low. Although the F-statistic is significant at the 0.2 level (F = 1.520 and p = 0.126) the R2 value at 0.129 suggests a lot of unexplained variance. As with the regression for numbers of fish caught, fire is insignificant as an explanatory variable in the model. Table 47. Regression showing relative contribution of fire to amount of reeds collected in 2001. Coefficientsa 24.545 8.852 2.773 .006 1.176 1.380 .075 .852 .396 -5.82E-05 .000 -.023 -.237 .813 -22.211 9.256 -.280 -2.400 .018 -14.356 9.009 -.182 -1.594 .114 -1.744 9.359 -.022 -.186 .852 -3.977 20.582 -.017 -.193 .847 -13.014 11.689 -.098 -1.113 .268 59.783 22.531 .253 2.653 .009 -3.927 11.220 -.033 -.350 .727 10.343 12.065 .078 .857 .393 5.924 7.108 .075 .833 .406 2.301 7.220 .030 .319 .750 (Constant) No. of adults in hh MVI Nxamasere Samochima Sekondomboro Bayei San Batswana Xereku Bakgalagadi Canoe Fire experienced Model 1 B Std. Error UnstandardizedCoefficients Beta StandardizedCoefficients t Sig. Dependent Variable: No. of reed bundles collecteda.

PAGE 132

117 Figure 39. Results of the third Stage 3 multiple regression: amount of reeds collected.

PAGE 133

118 The next regression examines the amount of thatching grass collected in 2001 as the dependent variable. The overall fit of the data to this model is good (F = 3.187, p = 0.001, R2 = 0.236). The most significant explanatory variable here is fire, however, once again it has a positive effect. That is, the number of bundles of thatching grass collected by a household was higher if fire was experienced (t = 2.950, p = 0.04). As with reeds, ethnicity and location also account significantly for some of the variation in amounts of grass collected, as is shown in Table 48. Table 48. Regression showing relative contribution of fire to amount of thatching grass collected in 2001 Coefficientsa 19.730 8.514 2.317 .022 1.529 1.366 .092 1.119 .265 -1.50E-05 .000 -.006 -.061 .951 -23.344 9.213 -.274 -2.534 .013 -16.600 8.859 -.199 -1.874 .063 -18.726 9.184 -.218 -2.039 .044 -11.358 20.446 -.045 -.556 .580 26.826 11.609 .190 2.311 .022 87.086 22.344 .347 3.897 .000 9.185 11.131 .073 .825 .411 -2.409 12.005 -.017 -.201 .841 4.931 7.050 .058 .699 .486 21.156 7.171 .255 2.950 .004 (Constant) No. of adults in hh MVI Nxamasere Samochima Sekondomboro Bayei San Batswana Xereku Bakgalagadi Canoe Fire experienced Model 1 B Std. Error UnstandardizedCoefficients Beta StandardizedCoefficients t Sig. Dependent Variable: No. of grass bundles collecteda.

PAGE 134

119 Figure 40. Results of the fourth Stage 3 multiple regression: amount of thatching grass collected.

PAGE 135

120 Because baskets are made from palm leaves collected in the panhandle, and they also contribute towards wealth, they too are susceptible to being affected by fire and other factors. For this reason the same regression was run for this variable, with results as follows: R2 = 0.297, F = 2.289 and p = 0.017. Once again, fire (t = 2.466, p = 0.016) is significant factor, and again it has a positive effect. The significance of ethnicity in this regression probably reflects the lack of involvement of some ethnic groups in the tradition of basket-making (see Table 49). The ownership of a canoe is important in this model (t = 2.466, p = 0.016), with canoe owners making more baskets. Table 49. Regression showing relative contribution of fire to numbers of baskets made in 2001. Coefficientsa 3.034 1.306 2.323 .023 -.246 .197 -.142 -1.247 .217 2.411E-05 .000 .076 .667 .507 -.970 1.386 -.123 -.699 .487 -1.663 1.363 -.178 -1.220 .227 -1.586 1.557 -.157 -1.019 .312 .155 2.749 .006 .056 .955 4.372 1.881 .277 2.325 .023 -1.883 2.861 -.077 -.658 .513 -2.152 1.713 -.148 -1.257 .213 3.870 1.512 .304 2.560 .013 2.074 .963 .234 2.154 .035 2.697 1.093 .318 2.466 .016 (Constant) No. of adults in hh MVI Nxamasere Samochima Sekondomboro Bayei San Batswana Xereku Bakgalagadi Canoe Fire experienced Model 1 B Std. Error UnstandardizedCoefficients Beta StandardizedCoefficients t Sig. Dependent Variable: No. of baskets madea.

PAGE 136

121 Figure 41. Results of the fifth Stage 3 multiple regression: number of baskets made.

PAGE 137

122 The final regression in Stage 3 (and in the causal path analysis) looks at the relative contribution of wetland resources to household livelihood. The results are significant (F = 2.305, p = 0.11), although the fit of the data to this model left some variance unexplained (R2 = 0.175). The most highly significant variable was ethnicity, with the proportional contribution of wetland resources to San households higher than for most others (t = 2.924, p = 0.004). These data are not surprising given the hunter-gatherer origins of the San, and the difficulties they face in adopting other livelihood strategies. Wealthier households depended significantly less on wetland resources for their livelihoods (t = -2.213, p = 0.029). Table 50. Regression showing relative effect of fire on proportional contribution of wetland resources to livelihoods in 2001. Coefficientsa 3.034 1.306 2.323 .023 -.246 .197 -.142 -1.247 .217 2.411E-05 .000 .076 .667 .507 -.970 1.386 -.123 -.699 .487 -1.663 1.363 -.178 -1.220 .227 -1.586 1.557 -.157 -1.019 .312 .155 2.749 .006 .056 .955 4.372 1.881 .277 2.325 .023 -1.883 2.861 -.077 -.658 .513 -2.152 1.713 -.148 -1.257 .213 3.870 1.512 .304 2.560 .013 2.074 .963 .234 2.154 .035 2.697 1.093 .318 2.466 .016 (Constant) No. of adults in hh MVI Nxamasere Samochima Sekondomboro Bayei San Batswana Xereku Bakgalagadi Canoe Fire experienced Model 1 B Std. Error UnstandardizedCoefficients Beta StandardizedCoefficients t Sig. Dependent Variable: No. of baskets madea.

PAGE 138

123 Figure 42. Results of the sixth Stage 3 multiple regression: proportional contribution of wetland resources to household livelihoods

PAGE 139

124 Together these regressions show that ethnicity and location are the most common factors determining access to wetland resources. Fire does not always play a role, but when it does, it has a positive effect. Notably, the extent of reliance on wetland resources, as a proportion of the households total wealth is lowest for wealthy households, and high for the marginalized San. This causal path analysis confirms that fire is only one factor determining use of resources. It shows that burning improves peoples access to wetland resources. Conflict Over Burning In a yes/no response in the household survey, 62.4% of respondents felt that generally, fires were good. During the informal interviews, where people were able to discuss fire more freely, 85.7% (N = 42) believed that, overall, fire in the panhandle was good. Informal interview respondents were further asked about the characteristics of a good fire, and then about those of a bad fire. Good and bad fires were explained in terms of timing, frequency, and size, with many respondents stating that less frequent burning led to bigger fires. Assessing levels of conflict with only a single snapshot in time can be best achieved by expressing the views of respondents through ethnographic themes. Several themes and supporting data for them are presented below, before the focus is returned to the effect of fire on sustainable livelihoods. Discussion Theme Lack of Conflict Due to Common Purpose Many respondents felt that there was no conflict at all, because burning, when managed, was good for all resource types. In general there are no conflicts when there is fire because it is beneficial to everyone. There is no conflict within the village because of burning on the floodplains. There are fires every year but we dont always know from which village

PAGE 140

125 the fires start. Nevertheless, there are no conflicts between villages. There are no conflicts about burning because usually people burn at the right time. There is no conflict between families because the fire improves the quality of reeds and grass. We have never had conflicts over fire in this village, although in last years fire some have lost their reeds that they had already cut. But there are no conflicts over burning. I have never lost reeds or grass bundles to fire. Fires from other villages are not problem sometimes they even help fishermen. Different Use of Panhandle Resources In the household survey, for each of the resources, about 75% of respondents experiencing fire in 2001 believed that it had been set intentionally by adults. They also gave the motives they believed were behind the fires. These are summarized in Table 51. Table 51. Reasons believed for why people set fires experienced in 2001. Percent Sample size Improve grazing 33.3 6 Create open spaces for fishing 27.8 5 Hunters to improve access 22.2 4 Improve quality of reeds, grass and/or papyrus 16.7 3 Total 100.0 18 From these responses it was possible to identify the resource-user type that people believed set the fire(s). Livestock grazers are considered the main source of fires, followed by fishermen, hunters, and then reed and thatch collectors. For each resource affected, the proportion of responses identifying a certain user type is shown in Table 52. For example, 62.5% of households using the floodplains for grazing livestock believe that

PAGE 141

126 (other) cattle owners are responsible for setting the fires, while 12.5% each blamed fishermen, hunters and reed/thatch collectors. Table 52. Resource affected by user held responsible for fire. Cattle owners Fishermen Hunters Reed/thatch collectors Total Sample size Livestock 62.50% 12.50% 12.50% 12.50% 100% 8 Thatching grass 33.33% 26.67% 20.00% 20.00% 100% 15 Reeds 31.25% 31.25% 18.75% 18.75% 100% 16 Palm leaves n/a n/a n/a n/a n/a 0 Water lily bulbs 28.57% 28.57% 14.29% 28.57% 100% 7 Papyrus for eating 50.00% 16.67% 33.335 0.00% 100% 6 Papyrus for mats 20.00% 20.00% 0.00% 60.00% 100% 5 Fish 42.86% 14.29% 28.57% 14.29% 100% 7 All resources 37.50% 23.44% 18.75% 20.31% 100% 64 In the informal interviews, livestock grazers were also in favor of burning, and thought to be behind the setting of fires. Fishermen, too, were often identified as a source of fires, and who themselves seemed most outspoken about the benefits of fire and its importance in improving access. Although all households use more than one resource, the extent of reliance on each varies. Several respondents expressed the need for trade-offs of losing access to one resource due to fire in order to improve another. Discussion Theme Conflict as a result of differing relations to the resource base Several people felt that users of certain resource types were more likely to want to set fires, and that in some cases, to have less consideration for the resource needs of others.

PAGE 142

127 There are conflicts between people those who burn are the ones who want to graze their cattle. Conflicts come from when those who cut reeds and grass early set fires so that their cattle can graze. I wish we could burn every year in order to have better pastures. Mainly the problem is that they burn too early and that leaves no food for livestock. Hunters like fire because it creates space for them to reach hidden places in the delta where there are more animals, and those who collect grass and reeds also like it because after burning new and attractive thatch and reeds grow. Fishermen also like fire because it is easy for them to spread their nets and even reach channels where there are more fish. There are many fires that side, because there are the small ponds that fishermen like to keep clear of grasses and sedges that is why they burn all the time. It is always the fishermen who are burning. When one burns, fish are attracted from the main river into the floodplains to eat the emerging new shoots. Fire makes the number of fish increase the next year. More have bred, because the ash has more nutrients which makes the plants grow. More sun reaches the fish nests and eggs, and these eggs need sun. It is always the fishermen who are burning. All they care about is more fish. In the old days they could be patient, now it is a business and they are running after money. The fires are started by people not fishermen just people, in order to improve reeds and thatching grass. Timing of Fires Half of the respondents felt that fires were bad if set at the wrong time that is, too early, or before everyone has finished taking what they need from the panhandle. Typically this was said in reference to reeds and thatching grass. Most respondents felt

PAGE 143

128 that fires should be set at selemo, the ploughing season, which roughly translates to October or November. By this time, not only would people have finished collecting, but also it was close enough to the start of the rains for new shoots of both reeds and thatching grass to sprout properly. Looking at the timing of the 17 fire events reported in the household survey, it is clear that most occurred in August and September, too early for people to have finished collecting (see Figure 43). The timing of these fires relate to both the start of floodplain grazing, and to the increase in fishing, which in turn relates to the reasons people gave for why fires were set. Most respondents said that fires set in winter (June or July) were not possible because the floodwaters were too high to allow the water to spread, and plants were simply damaged instead of being replaced by newly emerging shoots. However, some fishermen like these conditions as it allows them to set small fires around the lagoon edge where they fish. 0.0%5.0%10.0%15.0%20.0%25.0%30.0%35.0%40.0%45.0%AugustSeptemberOctoberNovemberDecemberJanuary Percent of fire eventsexperienced Figure 43. Proportion of 2001 fire events reported by month. This graph is based on the 17 separate fire incidents identified from information collected in both the household survey and informal discussions.

PAGE 144

129 Discussion Theme Conflict as a result of bad timing Even the discussion on conflict due to differing resource use contains an element of timing, so it is not surprising to see that the majority of comments centered on this issue. People felt that the issue was not whether the wetlands should be burned, but when. In the olden days people burnt at the right time, just before the rains, and there was no conflict between families but nowadays conflicts are many between villages because others burn at the wrong time or early. Conflicts only arise when the grass and reed beds are burned before everyone can collect. There is increased conflict about burning, because without the chiefs control people burn at the wrong time, while reeds are still there waiting to be collected. Conflicts usually arise when there was fire before or during the period of collection. Conflicts arise when people burn when others are still collecting. Conflicts between families due to fire do arise, but only when the fire was set before they can collect, and when the culprit is known. I dont really know about conflicts. The year before last, 2000, there were conflicts. Some people had not had a chance to cut, and others who had not moved them, lost their bundles in the fire. But no-one was found responsible. People live far from the river, and dont see what is happening. In the olden days people burnt at the right time and there was no conflict between families but nowadays conflicts are many between villages because others burn at the wrong time or early. Management and Consultation in the Use of Fire As discussed in Chapter 1, current fire laws reflect the central government standpoint that the destructive effects of fire outweigh any benefits, and that as a result, all fires should be suppressed. However, instead of stopping burning, the laws have simply made the practice secret. While this means there is usually no overt conflict, it

PAGE 145

130 also has led to greater frustration among community members. It is clear from the comments below that respondents feel that it is precisely because circumstances force people to set fires in secret that these now have negative impacts on their lives. Discussion Theme Lack of confrontational conflict due to unknown source In terms of expressing conflict as a result of badly timed fires, people tended to be more resigned than angry. This is mainly because the perpetrators are usually not known, as is shown through the comments below. There is no conflict between families because culprits do it secretly, and there is no way to know who to have the conflict with. There are no conflicts after fire because the culprits are not known. There is no-one we can directly accuse because the specific culprits always do the burning secretly. In the old days there were no conflicts because we used to discuss it in the village beforehand. People now have conflicts because burning is done secretly and no-one is warned. As to conflicts, there are no conflicts about fire because we never know who burnt the floodplains. There are no conflicts after fire because the culprits are not known. Conflicts are not common because the culprits are not known, so it is difficult to argue for something that you are not sure of. I have never heard of any conflicts. There cant be any, because we dont know who sets the fire. Some do quarrel and there is a lot of talking that takes place, but there is never any proof. It has been a long time since we have been affected by a big fire. There are conflicts over burning, but they end nowhere, the people are unknown so there is nothing you can do. But normally those who burn wait until people have taken their bundles from the river and then burn. There is no way you can hear that they are going to do it, it is done secretly. There are always conflicts between villages when a

PAGE 146

131 fire has been set, especially ones that destroy their property. Some just go and burn out of jealousy. It is difficult to confront those who set wrong fires as there is no way of telling who it is but there are conflicts. Discussion Theme How burning should be managed Many people believed that conflicts could be reduced if burning was not done secretly, so that people could plan for and agree on the timing of fires. We would like to be able to decide as a community when fires should be set. Before people used to come together to discuss but now, because it is illegal, people just do it as a secret without discussing with everyone. If Government could give permission to burn that would be good. They could call a kgotla meeting to choose a person to decide on when and where the burns would be OK. It should not be done individually, it must be done in the kgotla, especially in case the fire gets too strong and needs many people to help put it out. We would like it if Government could allow us to burn, and I hope the authority is given to the chief, it should be the chiefs who decide. If they called a kgotla meeting, they could give people a chance to finish collecting their reeds and grass. Maybe it would be easier to enforce the laws and control burning if things were controlled from within the village. We can just call a kgotla meeting every year, and choose the right time and the people who will do the burning. If fire were to be allowed and controlled, only the government would do it. Also if the tribe agrees with the statement, maybe the chief might be the one to control the burning, because they are his people. If fires are set in time not too early then it would be good to legalize fire. The chief should decide about that. The chief should be the one to decide about burning. He should be the one to call people to go together to manage fires.

PAGE 147

132 There is no ARB committee in Mogotho. We have been told to form one, but this hasnt happened yet. But this could be the organization to deal with burning. The Government should be the one to decide when there could be fires. They can have someone who comes from Gaborone to set the fires. They should have a kgotla meeting where they decide when and where they want that Government person to set the fires. It would be no problem to coordinate with downwind villages, because even they will choose the same time of year.

PAGE 148

CHAPTER 4 DISCUSSION Livelihoods Assessment It is clear that livelihoods in the panhandle vary considerably, both in general and in relation to wetlands resources. However, all households rely on these resources, and for most this reliance is high both in absolute terms and in the proportional contribution they make to household livelihoods. There is a range of factors that affect access to and use of the panhandles resources. Fire is only one of these. This section examines the results of this study using the sustainable livelihoods framework. Vulnerability Context As a constant and reliable source of food and building materials, the resources of the Okavango panhandle act as a buffer against many of the shocks and stresses faced by households. Fire has the potential to be a severe stressor, precisely because it affects the buffer, or safety net, which the wetland resources constitute. Because fires are set in secret, they are unpredictable and no one can plan for them. There is no consultation between or within communities, and fires are set by some resource users without accommodating the needs of others. Since the fires are left untended, they spread out of control and reach far beyond the area targeted for burning. For people interviewed in the Okavango panhandle, the threat to their access to reeds and thatching grass comes more from the timing of the fire rather than simply whether the plant beds burn or not. Fires set too early remove these key resources before people have had a chance to collect them. Most importantly, it is the households for 133

PAGE 149

134 whom natural resources comprise the largest part of their livelihood that are most vulnerable to early fires. These are the households that are most constrained in terms of their ability to harvest resources earlier in the collecting season. From a long-term perspective, the combination of fire and drought, or fire and drier flood cycles, leads to more serious consequences. As was reported in one village, fires that are set when the soil is dry can destroy the roots and corms of important resources such as reeds and thatching grass, permanently removing their stock. Burning in these conditions can also lead to peat fires that can burn for several years until shifts in the system bring floods back. Peat fires not only remove nutrients, but also change the gradient in this very flat system, permanently altering the flow dynamics of some river channels. Livelihood Assets In the Okavango panhandle, as with much of rural Botswana, increased access to one form of capital is usually linked to increased access to all the others. Male-headed households and those from the Hambukushu, Bayei and Xereku tribes tend to have comparatively more of all types of capital than female-headed households and those from the San tribes. There are also disparities between the relative contributions from natural capital versus financial capital along the same social divisions. Human capital In unpredictable environments with subsistence economies, the ability to diversify livelihood activities is crucial to survival. In the Okavango panhandle, as with the rest of rural Botswana, large households mean increased access to labor. The number of adults in the household is strongly correlated with wealth, with cattle ownership, with

PAGE 150

135 commercial fishing and with wage employment outside the study area. Not surprisingly, male-headed households tend to have more adult members. Because larger households can diversify more, wetland resources contribute a smaller proportion of their livelihood compared to smaller households. This means that any loss of resources to fire will affect the bigger households less. Bigger households, with more labor, can also finish collecting their resources earlier, before most of the fires are set. The link between household size on the one hand and ownership of cattle and commercial fishing on the other is important for another reason. Most respondents, including cattle owners and fishermen, stated that the main reasons people burnt were for grazing and fishing. That is, larger households have more reason, and may be more likely, to set fires. Social capital Generally, social capital within the panhandle villages is strong. There are several formal community organizations, as well as informal groups and networks. Links extend beyond the individual settlements, as family members are dispersed across the settlements. Sharing, borrowing and labor exchange among households are common practices. In spite of the limited power of the chief, communities still gather to make decisions on development issues within the spheres permitted by central government. Yet communities have no legal rights to use fire in the wetland or on other communal areas, or to decide on how to control it. Because it is illegal, fire-setting takes place outside of social networks. The lack of communication surrounding burning, particularly with fires set early in the season, undermines trust and creates suspicion between community members. It also leaves non-fire setters, such as smaller, poorer, female

PAGE 151

136 headed or San households, vulnerable. Even if decision-making on fire use were to be allowed at village level, there are differences in social capital that would limit the ability of some sectors of the community to influence where and when fires are set. While the majority of households are from matrilineal groups, these are still patriarchal, and female-headed households do not have the same status in the community, nor do women have a voice at community meetings. Likewise, the San tend to be stigmatized and marginalized. Given the link between these social categories and dependence on natural resources and with greater poverty, the people who most stand to be affected by fire would be less likely to have a voice concerning its use. Physical capital Access to physical capital at the household level, both of producer goods and infrastructure, is one of the key indicators of wealth. Looking at infrastructure, all households rich and poor alike rely on thatching grass and reeds to build their houses. The effect of fire on these resources (whether there is loss one year, or increased amounts and quality the next) affects everybody. But access to these resources differs especially when considering transport, another component of physical capital. Households with the most producer goods, such as tools and equipment, can pursue various strategies and are clearly better off. This is noticeable with regard to assets such as fishing equipment and dugout canoes. Again, wealthier, male-headed households, particularly of the Hambukushu and Xereku ethnic groups, have greater access to this form of capital and can therefore use their time more efficiently. Specifically, ownership of a dugout canoe confers the advantage of reaching resources further away, earlier in the season, and transporting them back more quickly and easily. This not only provides them with some

PAGE 152

137 protection against the negative effects of fire on their livelihoods, but also the opportunity to set fires themselves. Financial capital Again, it is the larger, male-headed households that have more members in wage employment, and so with greater access to cash. Beyond this, given that the wetlands are used primarily for subsistence use, the connection between these resources and cash is small. The main link between financial capital and fire is through the increasing commercialization of fishing. Not only did fishermen themselves state that fire was important for stimulating new shoots to attract fish closer, and to create space for their nets, but government fisheries officers confirmed that fire was increasingly being used. Natural capital All households in the Okavango panhandle use natural resources from the wetland. Wealthier households use greater amounts of all the natural resources than poorer households, but as has been shown, poorer households rely on natural resources for a greater proportion of their livelihoods. Most survey respondents believe that even with some loss one year, overall fire improves both the quality and quantity of the wetland resources they use, and that in some instances, the absence of fire decreases the availability of some resources. All felt that there were more than enough of the key resources to go around. Fire is seen as an agent of renewal, and of reviving the ecosystem and ensuring continued natural capital. Because the wetland is an annual system, fire is seen as a small but critical part of the seasonal cycle. Even with fires set early, most people felt that this was less of an environmental issue and more one of access.

PAGE 153

138 Policies, Institutions and Processes Several political factors affect the use of fire. The first is the centralization of governance, with decisions made far from the study area by people who may never have seen the Okavango panhandle, let alone understand the importance local people accord fire in the functioning of the ecosystem. In addition, although the Herbage Preservation Act is currently under revision, some departments in central government do not understand the important role that burning can play when managed, overt and planned. They still see fire as a threat to, rather than part of the environment. Some policy-makers further believe that the growing ecotourism industry does not leave any room for wildland fire. The second political factor is the law itself. It is hard for people who have a tradition of using fire as a management tool to see it as illegal. Another factor is the large gap between the creation of the law and its enforcement. The extreme remoteness and limited accessibility of the Okavango panhandle makes it impossible for central government authorities to police the setting fires. The diminished role of chiefs and local authorities means that their local knowledge does not get considered when policies and laws are formulated. Neither do their mandates extend to enforcement of common law such as the Herbage Preservation Act. Even if local institutions were to take on fire management responsibilities, the higher status of those with greater social capital such as wealthier, male-headed households would lead to their views dominating the decision making process. Livelihood Strategies Fire is clearly more than just an environmental stressor it can also be seen as an activity supporting livelihood strategies. The issue is who gets to use fire, and who does and doesnt benefit from it. From the preceding discussion, it is clear that variations in

PAGE 154

139 the different types of capital determine the relationship different households have with fire. Currently, the use of fire appears to be controlled by larger, wealthier male-headed households. These households are able to escape the negative effects of early burning by collecting resources earlier. They also have other livelihood sources to turn to in the absence of natural resources. Smaller, poorer households do not control when and where fires are set. They cannot predict their occurrence, and even if they could, they lack the necessary capital to avoid their negative effects. Although fire is a strategy useful to the collection of several resources, it is perhaps most important as pursued by fishermen. Given the increasing commercialization of fishing and the emergence of fishermen as a separate resource user group, burning in the wetland has the potential to create divisions among the community. Old traditional livelihood strategies favor the use of fire only late in the season, while commercial fishing encourages the setting of fires earlier. On an annual basis, therefore, fire affects peoples livelihood strategies with regard to the timing and seasonality of burning. However, in the long term fire affects strategies through the frequency or return intervals of burning. Fires that burn the same place every year could lead to irrevocable changes in vegetation composition, while fire suppression leads to the accumulation of large quantities of biomass, with the result that fires are much more intense, affecting larger portions of the panhandle. Fires potential to improve resources (and therefore livelihoods) over sustained periods can only be made dependable with proper management. Livelihood Outcomes The key factors determining livelihood sustainability appear to be gender of household head, ethnicity, ownership of cattle, and the diversity of livelihood strategies

PAGE 155

140 pursued. Fire appears to play a small role. Its importance as both a positive and negative influence increases for those households whose livelihoods are mostly based on wetland resources, because these households do not have alternative strategies to turn to. The survey shows that these households tend to be poorer and to have fewer alternative strategies. They are less likely to own dugout canoes. They are also smaller, with less labor, and older with frailer health. These households must wait until the floods have receded more fully to reach the collection areas, and often need longer to collect all they need. They rely on immediate access to each years produce, as well as on the functioning of the panhandle ecosystem for their long term sustainability. Figure 44. Differences in the role of fire in the livelihoods of rich and poor households. When fire is set secretively under unmanaged conditions, poorer households are vulnerable to temporary loss. Wealthier households are able to use fire to promote their livelihood sustainability.

PAGE 156

141 While wealthier households may be less dependent on wetland resources, they still need to pursue this as a livelihood strategy in order to contribute to their long term sustainability. For all households, fires need to occur, but they need to occur at the right time and at the right frequency. Fires at the right time can increase access to resources, and thereby reduce vulnerability. However, fires at the wrong time can have the opposite effect. Because of the lack of control and secretiveness of fire setting, it is very difficult to assess how burning affects the sustainable use of the natural resource base. As long as fire remains uncontrolled, it remains part of the vulnerability context for poorer households while wealthier households use it as a livelihood strategy for improving access to natural capital, as is shown schematically in Figure 44. Conclusions and Recommendations Overall, in 2001 fire did not appear to be a major factor affecting either the Okavango panhandle ecosystem itself or the livelihoods dependent on its wetland resources. Seasonal changes due to the annual flood regime appear to be far more important in determining the functioning of the system and the availability of key resources. Fires are always affected by the previous years flood. This determines how much growth there was to provide biomass for fuel, as well as limiting the spread where open water remains. Extent and Distribution of Fire in the Panhandle and Key Resource Areas The limited area of the Okavango Panhandle burnt in 2001 shows that while large fires do occur, they do not occur every year. The massive fire reported in 2000 was not repeated the following year and, having removed a lot of senescent vegetation, probably contributed to the limited spread of the 2001 fires. Most of the area burnt appears to have

PAGE 157

142 been on the floodplains rather than in the dense aquatic vegetation. Although the total area burnt in 2001 may appear to be mathematically small (at approximately 4% of the study area), its ecological significance cannot be determined on the basis on a single year. Burning was more concentrated within the areas used by the four focal villages. Nevertheless, it does not appear to have caused any major reduction in the presence of any of the key resources. Fires occurred relatively early in the season, indicating that these were set as soon as conditions allowed, instead of waiting until immediately before the onset of the rains or the arrival of the flood. Reliance on the Wetlands Natural Resource Base It is important to remember that the wetland represents only a part of the natural resources base available to households in the study area. The dryland is also used extensively, particularly for grazing. The shape and size of each villages resource areas are determined by accessibility as well as availability of the various key resources. Spatial variations in the extent of reliance on wetland resources suggests that where these are harder to get, households find alternatives, or obtain the resources through trade. It is not only the presence or absence of certain resources that affects the extent of reliance on them. Another factor is access. There is less floodplain grazing available for the river villages, for example. While the river channels themselves may act as entry points into the panhandle, the thick aquatic vegetation flanking them is a barrier. The floodplain villages both have to go further to get to the reeds and papyrus beds, but they are also able to do so more easily once the floods recede. For all villages the availability of resources varies according to season, with water strongly influencing accessibility. Use of the wetland resource base intensifies from July to November, during the dry season. A given household does not collect all resources

PAGE 158

143 every year. Building materials do not need to be replaced annually, and social and labor constraints may also prevent some people from harvesting. There is considerable social variation in the extent of dependence on wetland resources. While larger, wealthier households extract greater quantities of each resource, the relative contribution that these quantities make to the households livelihood is much smaller than for smaller, poorer households. The matrilineal nature of the dominant culture does not mean that women have equal access to resources. Female-headed households, even though they may have male family members, are less likely to have the assets needed to secure the contribution of wetland resources to their livelihoods. Ethnicity also affects access to resources. The San are most reliant on wetland resources, although the absolute quantities they collect are less than those of the majority Hambukushu and Xereku groups. Generally, wealthier, male-headed households from the majority ethnic groups tend to collect the most resources while they also have the greatest range of economic activities. Affect of Fire on Access to Key Natural Resources The actual impact of fire on resources is very small. Even where fire was experienced, households were able to collect resources before they were burnt. Some households however, felt they could have collected more thatching grass if it had not been for burning in their collection area. With regard to this and other plant resources, it is important to reiterate that households do not collect every year, and most households felt that a fire one year would improve the availability of quality reeds and thatching grass the following year. Generally, fire was perceived as being beneficial, particularly in terms of renewal of vegetation.

PAGE 159

144 Concerning access to resources, the impact of burning was smaller still. Fires are a relatively small factor in determining both the range of resources, and the amounts of each, collected. Instead, social factors such as wealth, ethnicity and gender of household head are far more important. Fires effects on livelihood sustainability depend on the relative contribution that the resources it affects make to a given household overall income. Even for those households that rely heavily on the wetland resources, fire is more likely to be seen as having a positive effect than a negative one, particularly when looking beyond a single season or resource. Conflict Over Burning Defining conflict is key to interpreting the results of this study. From the study, it is clear that conflict in the sense of confrontation was minimal. However, this should not automatically lead to an interpretation that everyone supports all fire. Often there was no conflict because it was not possible to identify an agent responsible for the fire there was no one with whom to have conflict. Instead, there were feelings of frustration, sometimes expressed in terms of different use of the panhandles resources, with livestock grazers and fishermen believed to be the main source of fires. Nevertheless, most people believed that burning served a common purpose that was beneficial to all. As a result, feelings of resentment were rare. The sense of frustration is not related to whether fires are set or not, but to when they are set. Conflict is more an issue of the timing of burning. Most people felt that fires were being set too early, with implications both for those who could not finish collecting in time, as well as for the regeneration of the plants themselves. It is the secretiveness and lack of consultation surrounding the setting of fires that people find the most exasperating.

PAGE 160

145 Recommendations for Further Study Local residents know that fire has been beneficial to their access to wetland resources for generations without any visible changes in the aspects that are important to them. However, because the Okavango is a feature of national interest, policy-makers usually consider its value beyond the level of local livelihoods. This requires understanding the positive and negative impacts of fire on the system as a whole and not just on some of its components. This would require a long term study and time series analysis, both of the extent and distribution of fire, and of any changes in species composition. In terms of spatial analysis, areas of repeated overlap should be measured, and a long term average for the area burnt to be calculated. These data should be analyzed in conjunction with figures for annual flood inflows into the panhandle. Ideally, bi-monthly satellite images should be obtained in order to record fires in wetter areas where fire scars may be ephemeral due to rapid vegetation regrowth. Frequent and in-depth groundtruthing (by motorboat and dugout canoe) would provide further validation of burn scars identified in the satellite imagery. Recommendations for Improved Fire Management A single snapshot in time, such as this study, cannot lead to conclusions about the role fire plays in the functioning of the panhandle ecosystem. Studies elsewhere suggest that fire in wetlands is at best, extremely beneficial to productivity. At worst, in dry conditions, it can lead to permanent changes in the soil and channel structures. Typically though, it is insignificant in comparison to changes due to the dynamics of the water regime, and its effects are temporary.

PAGE 161

146 Nevertheless, both environment and people would benefit from a more predictable fire regime. It is important to acknowledge that fire has always been part of the system, and that people have been using fire in the panhandle for thousands of years. Their traditional practices have much validity. This lies both in the sound reasons they have for burning, and in the continuing functioning of the wetland ecosystem in spite of or because of thousands of years of fire. Fire suppression leads to a large build-up of biomass and consequently large and uncontrollable fires. It seems clear that people will continue to use fire to further their livelihoods whether the practice is illegal or not. This is because, while they know that setting fire is against the law, they feel that it is right to use it to improve the resource base. This knowledge is supported by the fact that the negative impacts of fire are largely associated with the furtiveness in its use, and not the use itself. It is because burning is illegal that it is done in secret and outside of any form of control or management. By legalizing wildland fires, they can be planned and set in a manner better suited to environmental conditions and peoples needs. Because the system is dynamic and changes from year to year, so too must burning practices. Whether and when fires are set in any year must be based on local conditions, amongst which need to be counted local populations. For people to support proper managed use, they need to be involved in decision-making, and to understand the process. Care needs to be taken to ensure that the opinions of those who most depend on the wetland resources, who are also those less likely to be heard in communal meetings, are considered. This can be done by working at the ward level within villages, or by encouraging interest groups to meet and define their position prior to larger village-level meetings.

PAGE 162

147 A single strategy for fire management limits the ability to accommodate local variations in both environmental and socio-economic conditions. It also reduces the ability for decisions to be adapted to different fire regimes. The hydrology, vegetation biomass, and biodiversity of the Okavango panhandle is completely different to that of the surrounding savanna, and indeed to the rest of the Okavango delta. In addition, the linear settlements flanking the panhandle act as a useful buffer between the panhandle and the dryland. It is unlikely that fire could spread from the panhandle onto the savanna or vice versa. This provides an opportunity to incorporate legal, prescribed burning in the management of the panhandle. Decentralizing fire management is justified because it means that local variations such as these can be included.

PAGE 163

APPENDIX A HOUSEHOLD QUESTIONNAIRE

PAGE 164

CONFIDENTIAL1 Effects of Fire on Livelihoods and the Environment Okavango Panhandle Data Checks Date Initials Field check Office check Data entered Data verified Questionnaire no.: Enumerator: Kgotla GPS Coordinates: UTM S: UTM E: Date:__________ Village:__________ Respondents age:__________ Gender of respondent:__________ Refusal Index (no. of households refusing to answer before this one):__________ Interview start time:__________ HOUSEHOLD STRUCTURE Ke tla a simolola ka go botsa dipotso tse di mabapi le ba lelwapa. La ntlha ke tla a botsa dipotso tse di mabapi le tlhogo ya lelwapa mme morago go latela tsa ba bangwe ba lelwapa. I would like to begin by asking questions about the people from this household. First I will ask some questions about the head of the household, and then about the other people living here. 1. Tlhogo ya lelwapa ke mang? Who is the head of your household? Gender and marital status: GENDER MARITAL STATUS EN: TICK ONE ONLY Rre (Male) Single Married/Living together Widowed/divorced/separated Mme (Female) Single Married/Living together Widowed/divorced/separated 1 Apart from the page numbering, this questionnaire is presented in the same layout and style as used in the field. 149

PAGE 165

150 1.1 FA O NYETSWE, o atle o nne kgwedi di feta borataro mo ngwageng le rraabo / mmabo? IF MARRIED MALE OR FEMALE, does the husband live here more than 6 months of the year? no yes 2. Go buiwa le leme lefe mo lwapeng la gago? What is the language most often spoken in your household? CODE LANGUAGE EN: TICK ONE ONLY 1 Sembukushu 2 Seyei 3 Setswana 4 Xanikhwe (Sesarwa) 5 Bugakhwe (also Sesarwa) Xereku Other:___________ 3. Ene e le ga kae ka mariga e e fetileng fa mongwe wa mo lelwapeng a tsenelela phutego ya Kgotla? How many times since this time last year has anyone in the household attended a kgotla meeting? 4. Re tla kopa goitse ka tsa itshetso le thutego mo lwapeng la gago. Mo go tsa ditiro, re tlaa go bua ka batho ba ba neng ba ntse ba bereka ngwaga e e fetileng. Mo di potso tsa ditiro, re batla go bua ka batho ba hirilweng fela : We would like to get information on the members of your household, their education and their jobs. For the jobs we want to know about people who have been employed by anyone for wages since this time last year. Gender (M/F) Age Resident in Panhandle (Y/N) Education (HIGHEST LEVEL ATTAINED) Job Title (IF NOT EMPLOYED, WRITE N/A) 1st person 2nd person 3rd person 4th person 5th person 6th person 7th person 8th person 9th person 10th person 11th person 12th person

PAGE 166

151 EN: USE THE FOLLOWING FOR COMPLETING Q.4: EDUCATION: Tertiary A levels Forms 4-5 Forms 1-3 Standard 7 Other None ECONOMY Ke tsile go botsa dipotso dingwe mabapi le ditiro tse batho ba di dirang, le didiriswia tse balelwapa ba nang natso, dijwalo, diruiwa le meamuso ya tlholego. Tswee-tswee itse gore tsotlhe tse ke khupa marama. I would now like to ask some questions about the jobs people do, and items that your household owns. Please remember that this information will be kept confidential. 5. A bangwe mo lelwapeng ba kile ba bereka mo namola leubeng ka tshimologo ya mariga e e fetileng? Have any members of the household worked under the Drought Relief/CBPP scheme since this time last year? no yes IF YES, tswee tswee bolela gore ke bomang: please list below for each person: MALE/FEMALE TOTAL # MONTHS POST CODE 7. Mo ditirong tsa kgwebo-potlana tse di latelang, a mongwe wa lelwapa o kile a bona madinyana mo go tsone ka tshimologo ya mariga e e fetileng? From which of the following small-scale business activities has any member of the household earned any income since this time last year:

PAGE 167

152 EN: READ LIST TO RESPONDENT. TYPE YES/NO Shebeen (Beer/kgadi) Semausu (Street vendor) Go betla (Carpentry) Go rulela (Thatching) Go itaya ditena (Brick-making) Go epa didiba (well-digging) Go thula (Blacksmith) Kapei ya borotho (Bakery) Go rua dikoko (Poultry project) Go loga (Knitting) Go roka (Sewing) Tse dingwe (Other):. 7. Tswee-tswee bolela gore ke dife didirisiwa tse di latelang tse ba lelwapa ba nang le tsone. Please tell me which of the following items members of your household own: EN: READ LIST TO RESPONDENT. ITEM YES/NO Baesekele (Bicycle) Kara ya ditonki (Animal drawn cart) Koloi (Motor vehicle) Terekere (Tractor) Mokoro (Dug-out canoe) Selei (Sledge) Kiribane (Wheelbarrow) Pompo mo lelwapeng (Standpipe in yard) Mashini go pompa metsi (Water pump) Jeneraitara (Generator) Wailese (Radio) Lebone la parahini (Paraffin lamps) Setofo (Paraffin or gas stove) Mogoma wa diatla (Hoe) Mogoma wa dikgomo (Plough) Dijokwe (Yokes) Reike/Harawe (Harrows) Garawe (Spade) Selepe (Axe) Tlhobolo (Gun) Segai sa ditlhapi (Fishing spear) Letlowa (Fishing net) Sekoko (Hambukushu fishing basket) Dipitsa tse ditona (Large cooking pots) Dikupu tsa 25 kana 50 (25 or 50 plastic drums) Didramo tsa metsi (Large metal drums) Tsa itshereletso monang (Mosquito nets)

PAGE 168

153 Crop Production Dipotso tse di latelang di remeletse mo temong. The following questions are about crop production. 8. A balelwapa ba a tle ba leme? Does the household grow crops? no (skip to 9) yes 8.1 A balelwapa ba na le masimo mo melapong? Does the household have fields in the floodplains of the river? no (skip to 9) yes 8.2 Lelwapa le na le tema di le kae mo melapong? How many acres/hectares does the household have in the floodplains? unit 8.3 Tshimo tsa balelwapa mo melapong di fa kae? Where are fields located? ________________________________________ (EN: MAKE SURE YOU CAN RELATE THIS NAME TO A PLACE ON THE MAP.) 8.4 Le kile la lema mo ngwageng e e fetileng? Did the household grow crops in their floodplain fields last season? ________ no (skip to 9) ________ yes 8.5 Mo temong e e fitileng, lelwapa lene le jwadile tema di le kae mo melapong? How many acres/hectares did the household plant in the floodplains this last season? __________unit 8.6 Thobo e ne e le e e kae mo temong mo melapong e e fetileng? What was the yield from the floodplain fields this last season of the crop(s)?

PAGE 169

154 EN: ASK FOR # OF BAGS FOR GRAIN CROPS AND # OF FRUIT FOR MELONS/PUMPKIN. ASK HOW MUCH OF EACH CROP GROWN WAS SOLD. POST CODE CROP AMOUNT HARVESTED AMOUNT SOLD 8.7 A go ile go na le molelo mo kgaolong eno mo nwgageng e e fetileng? Was there a fire in this area since this time last year? _____ no (skip to 9) _____ yes 8.8 Molelo one o tlhaga ntlheng kae? Where (geographically) did the fire originate from? ____________________________ 8.9 Molelo one wa tlhaga kgeding efe? In which month was this fire? ____________________ 8.10 Molelo one o simolotse jang? How was fire started? (EN: DO NOT PROMPT) CODE HOW FIRE STARTED EN: TICK ONE ONLY 1 Legadima (Lightning) 2 Bana ba tshameka (Children playing) 3 Bagolo ka maikaelelo (Set intentionally by adults) 4 Ka phosego (Fire escaped from yard/campsite/cigarette) Gakeitse (Dont know) Other:___________________________________ IF SET INTENTIONALLY BY ADULTS: 8.11 Ga o bona ke eng gone go tshubiwa? What do you think is the reason that people started the fire? ___________________________________________________________________________________ ___________________________________________________________________________________ 8.12 A gone go na le dijwalo tsa lelwapa tse dineng tsa senngwa ke go tshuba mo melapong mo temong e e fetileng? Were any of the households floodplain crops damaged by fire this last season? _____ no _____ yes

PAGE 170

155 8.13 A bangwe ba lelwapa ba dirisa molelo go thokafatsa masimo mo melapong mo ngwageng e e fetileng? Did household use fire to clear or improve floodplains fields since this time last year? no yes Livestock Jaanong ke tlile go botsa dipotso tse di mabapi le leruo. I would now like to ask some questions about livestock. 9. A bangwe ba lelwapa ba na le leruo lengwe? Do any members of the household keep any livestock? no (skip to 10) yes 9.1 Tswee-tswee bolela mofuta wa leruo le palo: Please give livestock type and numbers kept: EN: UNDER ARRANGEMENT FOR CATTLE BELONGING OUTSIDE THE HOUSEHOLD, NOTE WHETHER FAMILY, MAFISA, OR PAID HERDING TYPE TOTAL # KEPT # OWNED BY HOUSEHOLD MEMBERS # OWNED BY NON-HOUSEHOLD MEMBERS ARRANGEMENT Dikgomo (Cattle) Dipudi (Goats) Didonki (Donkeys) Dipitse (Horses) Dinku (Sheep) Dikoko (Chickens) Tse dingwe (Other):.. 9.2 A ba lelwapa ba kile ba rekisa leruo lengwe la bone le le tshelang go simolola mo nakong e no mo ngwageng e e fetileng? Did the household sell any of its livestock AS LIVE ANIMALS since this time last year? no yes IF YES, Tswee tswee, bolela palo (please give numbers):

PAGE 171

156 TYPE NUMBER Dikgomo (cattle) Dipudi (goats) Didonki (Donkeys) Dipitse (Horses) Dinku (sheep) Dikoko (chickens) Tse dingwe (Other):__________________ 9.3 A ba lelwapa ba kile ba bolaya leruo lengwe la bone go simolola mo nakong e no mo ngwageng e e fetileng? Did the household slaughter any of its livestock since this time last year? no yes IF YES, Tswee tswee, bolela palo (please give numbers): TYPE NUMBER Dikgomo (cattle) Dipudi (goats) Didonki (Donkeys) Dipitse (Horses) Dinku (sheep) Dikoko (chickens) Tse dingwe (Other):__________________ 9.4 A leruo la lona le lengwe le fulela mo melapong? Are any livestock grazed on the floodplains? no (skip to 10) yes 9.5 Le fudisa fa kae mo melapong? Where on the floodplains are livestock normally grazed? _______________________________ (EN: MAKE SURE YOU CAN RELATE THIS NAME TO A PLACE ON THE MAP.) 9.6 A go ile go na le molelo mo kgaolong eno mo nwgageng e e fetileng? Was there a fire in this area since this time last year? _____ no (skip to 10) _____ yes

PAGE 172

157 9.7 Molelo one o simolotse kae? Where (geographically) did the fire originate from? ____________________________ 9.8 Ene ele ka kgwedi efe? In which month was this fire? ____________________ 9.9 Molelo one o simolotse jang? How was fire started? (EN: DO NOT PROMPT) CODE HOW FIRE STARTED EN: TICK ONE ONLY 1 Legadima (Lightning) 2 Bana ba tshameka (Children playing) 3 Bagolo ka maikaelelo (Set intentionally by adults) 4 Ka phosego (Fire escaped from yard/campsite/cigarette) Gakeitse (Dont know) Other:___________________________________ IF SET INTENTIONALLY BY ADULTS: 9.10 Ga o bona ke eng batho ba ne ba thoseditse molelo oo? What do you think is the reason that people started the fire? ___________________________________________________________________________________ 9.11 A balelwapa ba kile ba latlhegelwa ke leruo lengwe ka ntlha ya melelo ya naga mo melapong go simolola mo nakong eno e e fetileng go fitlhelela gompieno? Did the household lose any livestock to fires on the floodplain since this time last year? no yes 9.12 A leruo la gago le ne la sologelwa molemo ka go tshubiwa moo ga naga mo ngwageng e e fetileng? Did your livestock benefit from improved grazing conditions on the floodplain after fires since this time last year? no yes

PAGE 173

158 Natural Resources Use Dipotso tse di latelang ke tse di mabapi le meamuso a naga. The next few questions are about the use of natural resources. 10. Go na le ba ba kgetlhang matlhaka mo lwapeng la lona? Does anyone in the household collect reeds? no (skip to 11) yes 10.1 Ba kgethile dingata tse kae mo ngwageng e e fetileng? _______ How many bundles did they cut in the past year? 10.2 Ba rekisitse dingata tse kae mo ngwageng e e fitileng? How many bundles did the household sell in the past year? _______ 10.3 Ba kgetlha letlhaka kae? Where do they usually collect them? __________________________________ (EN: MAKE SURE YOU CAN RELATE THIS NAME TO A PLACE ON THE MAP.) 10.4 A go ile go na le molelo mo kgaolong eno mo nwgageng e e fetileng? Was there a fire in this area since this time last year? _____ no (skip to 11) _____ yes 10.5 Molelo one o simolotse kae? Where (geographically) did the fire originate from? ____________________________ 10.6 E ne e le ka kgwedi efe? In which month was this fire? ____________________ 10.7 Molelo one o simolotse jang? How was fire started? (EN: DO NOT PROMPT) CODE HOW FIRE STARTED EN: TICK ONE ONLY 1 Legadima (Lightning) 2 Bana ba tshameka (Children playing) 3 Bagolo ka maikaelelo (Set intentionally by adults) 4 Ka phosego (Fire escaped from yard/campsite/cigarette) Ga keitse (Dont know) Other:___________________________________

PAGE 174

159 IF SET INTENTIONALLY BY ADULTS: 10.8 Fa o bona maikaelelo a go tshuba e ne ele eng? What do you think is the reason that people started the fire? ___________________________________________________________________________________ 10.9 Go na le mafelo a go na le letlhaka aa neng a tshubiwa molelo pele le ka kgona go kgetlha mo ngwageng e e fetileng? Were any reedbeds in this area destroyed by fire before you could collect this past year? no yes 10.10 A lethaka le ne la sologelwa molemo ka go tshubiwa moo ga naga mo ngwageng e e fetileng? Were the reedbeds where you collect improved by any fire since this time last year? no ______ yes 11. Go na le bangwe mo lwapeng ba ba logang diphatle tsa matlhaka? Does anyone in the household make mabinda mats? _____ no (skip to 12) _____ yes 11.1 Fa gole jalo, ba logile phatle tse kae mo ngwageng ee fetileng? If yes, how many did they make in the past year? ____________ 11.2 Ba rekisitse tse kae mo ngwageng ee fetileng? How many did they sell in the past year? ____________ 11.3 Ba dirisitse ngata tse kae tsa letlhaka? How many bundles of reeds were used for mabinda mats in the past year?_________ 11.4 Ba ne ba bona letlhaka jang? What was the main way that the reeds used for the mats were obtained? CODE HOW REEDS OBTAINED EN: TICK ONLY ONE 1 Re ikgetletse mo lwapeng (Collected by someone in the household) 2 Re rekile (Bought) 3 Re ikgetletse le re rekile (Both collected and bought) Ga keitse (Dont know) Other:___________________________________

PAGE 175

160 12. A bangwe mo lelwapeng ba kgetla mokola mo melapong? Does anyone in the household collect palm leaves within the panhandle? no (skip to 13) yes 12.1 Fa gole jalo ba kgetlele ngata tse kae mo ngwageng e e fetileng? _______ How many bunches did they cut there in the past year? 12.2 Le rekesitse ngata tse kae mo ngwageng ee fetileng? How many bunches did the household sell in the past year? _______ 12.3 Le kgetla fa kae mo melapong? Where in the panhandle do they usually collect them? ________________________ (EN: MAKE SURE YOU CAN RELATE THIS NAME TO A PLACE ON THE MAP.) 12.4 A go ile go na le molelo mo kgaolong eno mo nwgageng e e fetileng? Was there a fire in this area since this time last year? _____ no (skip to 13) _____ yes 12.5 Molelo one o simolotse fa kae? Where (geographically) did the fire originate from? ____________________________ 12.6 E ne e le ka kgwedi efe? In which month was this fire? ____________________ 12.7 Molelo one o simolotse jang? (EN: DO NOT PROMPT) CODE HOW FIRE STARTED EN: TICK ONE ONLY 1 Legadima (Lightning) 2 Bana ba tshameka (Children playing) 3 Bagolo ka maikaelelo (Set intentionally by adults) 4 Ka phosego (Fire escaped from yard/campsite/cigarette) Ga keitse (Dont know) Other:___________________________________ IF SET INTENTIONALLY BY ADULTS: 12.8 Fa o bona maikaelelo a go tshuba e ne ele eng? What do you think is the reason that people started the fire? ___________________________________________________________________________________

PAGE 176

161 12.9 Go na le mekolwane mengwe e eneng ya tshubiwa ke molelo go tsweng ngwaga ee fetileng? Were any small palm trees in this area destroyed by fire since this time last year? no yes 12.10 A mekolwane le ne la sologelwa melemo ka go tshubiwa moo ga naga mo ngwageng ee fetileng? Were the palm trees where you collect improved by any fire since this time last year? no yes 13. A bangwe mo lwapeng ba loga manki? Does anyone in the household make baskets? _____ no (skip to 14) _____ yes 13.1 Ba logile dimanki dile kae mo ngwageng ee fetileng? If yes, how many did they make in the past year? ____________ 13.2 Ba rekisitse dile kae? How many did they sell in the past year? ____________ 13.3 Ba dirisitse ngata tse kae tsa mokolwane? How many bunches of palm leaves were used for making baskets in the past year?_________ 13.4 Ba ne ba bona mekolwane jang? What was the main way that the palm leaves for the baskets were obtained? CODE HOW PALM LEAVES OBTAINED EN: TICK ONE ONLY 1 Re ikgetletse mo lwapeng (Collected by someone in the household) 2 Re rekile (Bought) 3 Re ikgetletse le re rekile (Both collected and bought) Ga keitse (Dont know) Other:___________________________________ 14. Go bangwe mo lwapeng ba ba kgetlang bojang mo melapong? Does anyone in the household collect thatching grass from within the panhandle? no (skip to 15) yes

PAGE 177

162 14.1 Fa gole jalo ba kgetlile ngata tse kae mo ngwageng e e fetileng? How many bundles did they cut in the past year? _______ 14.2 Ba rekisitse tse kae mo ngwageng e e fetileng? How many bundles did the household sell in the past year? _______ 14.3 Ba kgetla kae? Where do they usually collect them? __________________________________ (EN: MAKE SURE YOU CAN RELATE THIS NAME TO A PLACE ON THE MAP.) 14.4 A go ile go na le molelo mo kgaolong eno mo nwgageng e e fetileng? Was there a fire in this area since this time last year? _____ no (skip to 15) _____ yes 14.5 Molelo one o simolotse kae? Where (geographically) did the fire originate from? ____________________________ 14.6 E ne ele ka kgwedi efe? In which month was this fire? ____________________ 14.7 One o simolotse jang? How was fire started? (EN: DO NOT PROMPT) CODE HOW FIRE STARTED EN: TICK ONE ONLY 1 Legadima (Lightning) 2 Bana ba tshameka (Children playing) 3 Bagolo ka maikaelelo (Set intentionally by adults) 4 Ka phosego (Fire escaped from yard/campsite/cigarette) Gakeitse (Dont know) Other:___________________________________ IF SET INTENTIONALLY BY ADULTS: 14.8 Fa o bona maikaelelo a go tshuba e ne ele eng? What do you think is the reason that people started the fire? ___________________________________________________________________________________

PAGE 178

163 14.9 A kgaolo dingwe tsa bojang di kile tsa tshubiwa ka molelo pele le ka kona go kgetla mo ngwageng ee fetileng? Were any thatching grass beds in this area destroyed by fire before you could collect this past year? no yes 14.10 A bojang le ne la sologelwa melemo ka go tshubiwa moo ga naga? Were the thatching grass beds where you collect improved by any fire since this time last year? no yes 15. Go bangwe mo lwapeng ba ba bapala tswii? Does anyone in the household collect water lily bulbs? no (skip to 16) yes 15.1 Ba bapetse dingata di le kae mo ngwageng ee fetileng? _______ How much did they collect in the past year? 15.2 Ba rekisitse dingata di le kae? How much did the household sell in the past year? _______ 15.3 Tswii ba i tsaya kae gantsi? Where do they usually collect them? __________________________________ (EN: MAKE SURE YOU CAN RELATE THIS NAME TO A PLACE ON THE MAP.) 15.4 A go ile go na le molelo mo kgaolong eno mo nwgageng e e fetileng? Was there a fire in this area since this time last year? _____ no (skip to 16) _____ yes 15.5 Molelo one o simolotse kae? Where (geographically) did the fire originate from? ____________________________ 15.6 E ne ele ka kgwedi efe? In which month was this fire? ____________________

PAGE 179

164 15.7 One o simolotse jang? How was fire started? (EN: DO NOT PROMPT) CODE HOW FIRE STARTED EN: TICK ONE ONLY 1 Legadima (Lightning) 2 Bana ba tshameka (Children playing) 3 Bagolo ka maikaelelo (Set intentionally by adults) 4 Ka phosego (Fire escaped from yard/campsite/cigarette) Gakeitse (Dont know) Other:___________________________________ IF SET INTENTIONALLY BY ADULTS: 15.8 Fa o bona maikaelelo a go tshuba e ne ele eng? What do you think is the reason that people started the fire? ___________________________________________________________________________________ 15.9 A tswii mo kgaolong e e kile ya sengwa ke molelo pele le ka kgona go bapala mo ngwageng ee fetileng? Was any tswii in this area destroyed by fire before you could collect this past year? no yes 15.10 A tswii le ne la sologelwa melemo ka go tshubiwa moo ga naga? Was the tswii where you collect improved by any fire since this time last year? no yes 16. Go bangwe mo lwapeng ba ba kgetlang koma go ja? Does anyone in the household collect papyrus to eat? no (skip to 17) yes 16.1 Ba kgetlile ngata di le kae mo ngwageng ee fetileng? _______ How many bundles did they collect in the past year? 16.2 Ba rekisitse ngata di le kae? How many bundles did the household sell in the past year? _______ 16.3 Koma e ba e jang ba i tsaya kae gantsi? Where do they usually collect papyrus for eating? __________________________________ (EN: MAKE SURE YOU CAN RELATE THIS NAME TO A PLACE ON THE MAP.)

PAGE 180

165 16.4 A go ile go na le molelo mo kgaolong eno mo nwgageng e e fetileng? Was there a fire in this area since this time last year? _____ no (skip to 17) _____ yes 16.5 Molelo one o simolotse kae? Where (geographically) did the fire originate from? ____________________________ 16.6 E ne ele ka kgwedi efe? In which month was this fire? ____________________ 16.7 One o simolotse jang? How was fire started? (EN: DO NOT PROMPT) CODE HOW FIRE STARTED EN: TICK ONE ONLY 1 Legadima (Lightning) 2 Bana ba tshameka (Children playing) 3 Bagolo ka maikaelelo (Set intentionally by adults) 4 Ka phosego (Fire escaped from yard/campsite/cigarette) Gakeitse (Dont know) Other:___________________________________ IF SET INTENTIONALLY BY ADULTS: 16.8 Fa o bona maikaelelo a go tshuba e ne ele eng? What do you think is the reason that people started the fire? ___________________________________________________________________________________ 16.9 A koma e ba jang mo kgaolong e e kile ya sengwa ke molelo pele le ka kgona go kgetla mo ngwageng ee fetileng? Was any papyrus (for eating) in this area destroyed by fire before you could collect this past year? no yes 16.10 A koma e ba e jang le ne la sologelwa melemo ka go tshubiwa moo ga naga? Was the papyrus (for eating) where you collect improved by any fire since this time last year? no yes

PAGE 181

166 17. Go bangwe mo lwapeng ba ba kgetlang koma go dira moseme? Does anyone in the household collect papyrus for mat-making? no (skip to 18) yes 17.1 Ba kgetlile ngata se se kae mo ngwageng ee fetileng? _______ How many bundles did they collect in the past year? 17.2 Ba rekisitse ngata di le kae? How many bundles did the household sell in the past year? _______ 17.3 Ba rekisitse meseme di le kae? How many papyrus mats did the household sell in the past year? _______ 17.4 Koma e ba e dirang moseme ba i tsaya kae gantsi? Where do they usually collect papyrus for mats? -__________________________________ (EN: MAKE SURE YOU CAN RELATE THIS NAME TO A PLACE ON THE MAP.) 17.5 A go ile go na le molelo mo kgaolong eno mo nwgageng e e fetileng? Was there a fire in this area since this time last year? _____ no (skip to 18) _____ yes 17.6 Molelo one o simolotse kae? Where (geographically) did the fire originate from? ____________________________ 17.7 E ne ele ka kgwedi efe? In which month was this fire? ____________________ 17.8 One o simolotse jang? How was fire started? (EN: DO NOT PROMPT) CODE HOW FIRE STARTED EN: TICK ONE ONLY 1 Legadima (Lightning) 2 Bana ba tshameka (Children playing) 3 Bagolo ka maikaelelo (Set intentionally by adults) 4 Ka phosego (Fire escaped from yard/campsite/cigarette) Gakeitse (Dont know) Other:___________________________________

PAGE 182

167 IF SET INTENTIONALLY BY ADULTS: 17.9 Fa o bona maikaelelo a go tshuba e ne ele eng? What do you think is the reason that people started the fire? ___________________________________________________________________________________ 17.10 A koma e ba e dirang moseme mo kgaolong e e kile ya sengwa ke molelo pele le ka kgona go kgetla mo ngwageng ee fetileng? Was any papyrus (for mats) in this area destroyed by fire before you could collect this past year? no yes 17.11 A koma e ba e dirang le ne la sologelwa melemo ka go tshubiwa moo ga naga? Was the papyrus (for mats) where you collect improved by any fire since this time last year? no yes 18. A ba bangwe mo lelwapeng ba tshwara ditlhapi? Does anyone in the household catch fish from within the panhandle? no (skip to 19) yes 18.1 Fa gole jalo ba tshwere dile dikae mo ngwageng e e fetileng? _______ How many fish did they catch in the past year? 18.2 Ba rekisitse dile kae? How many fish did the household sell in the past year? _______ 18.3 Ba di tshwara kae? Where do they usually catch them? __________________________________ (EN: MAKE SURE YOU CAN RELATE THIS NAME TO A PLACE ON THE MAP.) 18.4 A melelo ya naga e amile selekanyo sa ditlhapi mo ngwageng ee fetileng? Did fire affect the number of fish they caught in the past year? no (skip to 19) dont know (skip to 19) yes

PAGE 183

168 18.5 IF YES, Jaang? How? CODE HOW FISH AFFECTED EN: TICK ONE ONLY 1 Palo ya koketsego (Numbers increased) 2 Palo ya phokotsego (Numbers decreased) Other:___________________________________ 18.6 Molelo one o simologile kae? Where (geographically) did the fire originate from? ____________________________ 18.7 Ene ele ka kgwedi efe? In which month was this fire? ____________________ 18.8 One o simolotse jang? How was fire started? (EN: DO NOT PROMPT) CODE HOW FIRE STARTED EN: TICK ONE ONLY 1 Legadima (Lightning) 2 Bana ba tshameka (Children playing) 3 Bagolo ka maikaelelo (Set intentionally by adults) 4 Ka phosego (Fire escaped from yard/campsite/cigarette) Gakeitse (Dont know) Other:___________________________________ IF SET INTENTIONALLY BY ADULTS: 18.9 Fa o bona maikaelelo a go tshuba e ne ele eng? What do you think is the reason that people started the fire? ___________________________________________________________________________________ 19. A ba bangwe mo lelwapeng a ba tshoma diphologolo kana dinonyane? (REMIND AND REASSURE THEM THIS IS CONFIDENTIAL. TELL THEM WE WONT BE ASKING QUANTITIES OR KINDS OF ANIMALS) Does anyone in the household hunt wildlife or birds? no (skip to end) yes 19.1 Go na le bangwe mo lwapeng ba ba tsomang mo melapong? Is any hunting done on the floodplains? no (skip to end) yes

PAGE 184

169 19.2 Ba tsoma kae mo melapong? Where on the floodplains is hunting normally done? ___________________________________ (EN: MAKE SURE YOU CAN RELATE THIS NAME TO A PLACE ON THE MAP.) 19.3 A melelo ya naga e amile selekanyo sa dipologolo mo ngwageng ee fetileng? Did fire affect the number of animals they caught in the past year? no (skip to end) dont know yes 19.4 IF YES, Jaang? How? CODE HOW WILDLIFE AFFECTED EN: TICK ONE ONLY 1 Palo ya koketsego (Numbers increased) 2 Palo ya phokotsego (Numbers decreased) Other:___________________________________ 19.5 Molelo one o simologile kae? Where (geographically) did the fire originate from? ____________________________ 19.6 Ene ele ka kgwedi efe? In which month was this fire? ____________________ 19.7 One o simolotse jang? How was fire started? (EN: DO NOT PROMPT) CODE HOW FIRE STARTED EN: TICK ONE ONLY 1 Legadima (Lightning) 2 Bana ba tshameka (Children playing) 3 Bagolo ka maikaelelo (Set intentionally by adults) 4 Ka phosego (Fire escaped from yard/campsite/cigarette) Gakeitse (Dont know) Other:___________________________________ IF SET INTENTIONALLY BY ADULTS: 19.8 Fa o bona maikaelelo a go tshuba e ne ele eng? What do you think is the reason that people started the fire? ___________________________________________________________________________________

PAGE 185

170 EN: THANK THE RESPONDENT FOR THEIR TIME AND CONTRIBUTIONS. TELL THEM THAT THE RESULTS OF THE SURVEY WILL BE SENT TO THE KGOTLA BY JUNE 2003. (Ke lebogela nako eo nneleng nayo le bopelo telele jwa gago gore thusa mo tshekatshokong ya rona., le ka moso. Maduo atla rumelwa go lona mme a ka bonwa fa kgotleng mo kgweding ya Seetebosigo ngwaga ee tlang, 2003.) ASK THEM FOR ANY COMMENTS ON THE QUESTIONNAIRE OR ON THE USE OF THE RESOURCES IN THEIR AREA. (A o na le sengwe se oneng o batla go se tlatsa?) ________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ ________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ ______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ RECORD THE FINISHING TIME. Finishing time:_________ Total interview time:_________

PAGE 186

APPENDIX B INFORMAL INTERVIEW TRANSCRIPTS

PAGE 187

Reed/Thatch Perspective Woman, Mogotho Village We get reeds and thatch in the Okavango River and use some of the floodplains as a grazing area. Burning affects the growth rates of thatching grass and reeds because when it burns new thatching grass and reeds emerge. Burning is important because it removes the older grass so that the new grass and reeds can emerge easily without disturbance from the older ones. Sometimes burning is bad because it destroys the emerging new grasses and reeds when set not in time, like during the flooding season. The fire which is good is the one which burns at the right season, like November to December because the grass and reeds will be dry and can burn easily, and most of the people will be through with collecting reeds and grass by those months. In the olden days people burnt at the right time and there was no conflict between families but nowadays conflicts are many between villages because others burn at the wrong time or early. Old Woman, Mogotho Village We get water lily (tswii) and reeds from the delta. Some grass types crowd out reeds, and this affects the growth rate of reeds. Burning is the only solution to make space for more reeds. When we burn, thatch and reeds grow well. Floodwater improves the quality of reed, because reeds are water loving plants. The fire which is useful is the one which burns everything so that new reed and grass beds can be large enough fore everyone to collect, thereby reducing competition. Fire also makes more space for water lily to grow. Small fires are bad because they burn every year, and they dont create enough space for grass, reeds and lilies to grow freely, which means some plants may get used up. There is no conflict between families because the fire improves the quality of reeds and grass, and because culprits do it secretly, and there is no way to know who to have the conflict with. In the old days fire was always there because the law enforcement was not strong as it is nowadays. In the old days people used to burn after collecting reeds, such as in September so that at the rainy season new reeds could emerge. In the old days they burnt every 2 years they did this so that there was no accumulation of old grass and reeds. People get injured when these old reeds are not burned for example, they can get stabbed by the old hardened stalks. The kgosi should be given the power to permit people to burn, but it depends on the peoples wishes. Some people go earlier than others to collect reeds from the delta because they want good reeds, and rush when there is a chance to collect the nearest reeds. Others do not have a mokoro to use and others are sick or have funeral to attend, that may be why they go later to collect. It is not good to burn at the start of the winter because of the floodwater the reeds and grass are still wet. Nowadays you dont see fire often because of the law. Old Man, Mogotho Village Mostly we collect thatching grass and reeds. The accumulation of grass, reeds and papyrus really disturbs the movement of fish. Fire makes food for fish so that they increase in number because it gets rid of the old dry ones and the new and nutritious 171

PAGE 188

172 shoots that they feed on grow. Floodwater encourages rapid growth of reeds and grass. Fire also gets rid of the old reeds and thatch, so that new ones grow. When grazing land is flooded, there is less grazing area. Fire improves the grazing land by removing the old grass and bringing new palatable grass for livestock even fish. I wish we could burn every year in order to have better pastures. A good fire is one that is set after everyone has collected reeds and grass. Bad fires are the ones that are set before everyone can collect, especially those set around August. There are no conflicts after fire because the culprits are not known. As for when and where people collect, everyone is free to collect reed and thatch wherever they wish. People usually look at the condition of their houses and compound walls to decide if they need to go earlier to collect. If their walls and roofs are old, they rush so that they can repair their houses before the rainy season. Some also go earlier in order to get the nearest grasses. But the water level also controls the time they go to collect. It is not common to set fires at the start of the winter because of the floodwater. The suitable time for burning is after August. In the past years fires have not been frequent only in the old days were fires common. Grazing pastures were cleared with fire just before the rainy seasons. Even in the past the kgosi did not allow fires, so they were set secretly. But I think the kgosi should be give the power to allow people to burn. Nowadays fires are not often and they are small compared to the old days. This is because fires are put out immediately so they cant spread far. People mostly burn in spring from August to December. Focus Group Meeting Thatch and Reed Collectors, Sekondomboro Village Meeting with 9 women, the chief and 2 other men. The most important places for these resources are Kgaolatlhogo and Hogu. August is the start of the collecting season, earlier the reeds are still too wet. September is the best time. By November the reeds are too dry to cut. Fires happen every 2 years at both these places, and they are always big fires. Some would prefer it if we could burn every year to clean the place up, but others feel that every 2 years is good for both reeds and grass. Things have changed over the years. Fires used to be burnt anywhere, anytime, but now because of the law there are only secret and accidental fires. In the old days it used to burn every year, but this was bad. Reeds take time to grow, more than one year, that is why it should not be burnt every year. We would like to be able to decide as a community when fires should be set. If fires are started by villages in the south, this could be a problem. Even now villages dont come to work together. We dont know if other village could come and work with us in making these decisions. No-one had heard of the ARB there is no committee here. There is no-one in the village who can be responsible for fires they are too dangerous. Someone must be hired. At the moment no-one in the village can decide, maybe the government could give the kgosi these powers. We have never had conflicts over fire in this village. Although

PAGE 189

173 some have lost their reeds that they had already cut in last years fire. This fire was in September. Fires should be set in December, that way everyone will have had a chance to take their bundles from the river. We never see lightning fires they are always set by people. Yes, they are always set deliberately so that new grass can emerge. Some people set fires out of jealousy. There is never a good reason to burn at the start of winter, it kills the grass too soon. Fires are bigger than they used to be in the old days. Big fires are worse because they destroy everything. What decides whether someone goes early or late to collect reeds and thatch is water, and the amount of reeds that some-one wants. We also want it to be burnt so that we can get tswii more easily. [Vociferous about the] need to remove old reed stumps which stab their feet. This is one of the main reasons people burn for. Old Man, Sekondomboro Village We use various things from the river, like thatching grass and reeds which are our building materials. Fishing: Crocodiles are the most destructive creatures in the river because they get trapped in our fishing nets and cut them, making large holes which allow fish to just pass through, therefore our production is brought down. They even eat the fish that they find already trapped there. In addition, otter also cut the nets and feed on the trapped fish. Hippos too break our nets by displacing them. (In the old days hippos were few in number compared to these days.) Turtles also eat the trapped fish in the net and bring our catch down. Reeds: Fires are the most dangerous things that affect the reed beds sometimes burning everything, leaving people with nothing to collect. The same for thatching grass. Floodwater brings along with it a larger number of fish compared to the dry season, when only a few number are found in the main streams (river). It is even easy to catch fish in the floodwater because even canoes can make their way through, while in the dry season, only those with speed boats can catch fish in the main streams. Fire creates more space for fish to travel from the main streams to the nearest lagoons. It makes space for fish nets to be spread easily without any disturbance from hippo grass or reeds. Fire also encourages the growth of tswii which is a favourite food of fish, and even people. Fires also encourages the growth of better reeds and grass and gets rid of dry and old ones. Generally fires are good because they bring life to the people, but at the same time they cause destruction to the system in the delta. Animals and birds die and some find no way to stay because their habitats are destroyed. Every kind of fire can be bad because it brings destruction to many animal which are living in the delta. Most of the people like the use of fire because its impacts are profitable to them, and are only bad for animals and birds. Hunters like fire because it creates space for them to reach hidden places in the delta where there are more animals, and those who collect grass and reeds also like it because after burning new and attractive thatch and reeds grow. Fishermen also like fire because it is easy for them to spread their nets and even reach channels where there are more fish. Tswii also grows well after fire, so providing food

PAGE 190

174 for people and fish. In general there are no conflicts when there is fire because it is beneficial to everyone. Conflicts only arise when the grass and reed beds are burned before everyone can collect, but there is no-one we can directly accuse because the specific culprits always do the burning secretly. Even in the old days, fires were strictly prohibited, even by the kgosi himself, so they were still done secretly. In the old days when the law enforcement wasnt strong, fires were frequent, but nowadays a year or two can pass with no fires due to strong law enforcement. This law enforcement really made the life of the Okavango residents tough because they cant get enough food from the delta without fire. There is no way that fire can be controlled, especially in the delta because it is very difficult to stop, it burns from one place to the last one, and results in the whole ecosystem burning. NThere are no specific places for different families to use in the delta, everyone decides for him or herself where to collect. For example, people from Xakao can come to the Kgaolatlhogo to collect reeds and thatching grass without any problem. We regard the delta as a gift for everyone from God. Floodwater does control the time of the year that people go to collect building materials. We have to wait till the level of water is low so that we can reach the beds easily without fear of crocodiles and pythons. We usually start going around August until December. Those who go earlier start in July they are usually young people! Older men and women usually go late, around September. It is not easy to burn at the start of winter because this is the season when the river is flooded and grass and reeds are still waterlogged, and therefore it is not easy to burn them. Even the water level is still high so fire cant drift as easily as when the grass and reeds are dry. In the old days fires were very frequent due to the weak law enforcement in those days, but nowadays they are not as frequent because the law enforcement is strong. In the old days fire could burn from as far as Seronga for hundreds of kilometres till it reached places like Mohembo. Fires were big and dangerous because they burned huge acres of land. Nowadays big fires are not as frequent. In the old days people were allowed to grow crops in the floodplains where they found very good yields because they grew crops all year round without waiting for rain. They were favoured by the ever wet soils in the floodplains, and the soil is fertile compared to that in the dryland farms. Nowadays people are strictly prohibited from farming in those areas, thats why their standards of living are poor. Old Man, Sekondomboro (at the outskirts of the village) I depend heavily on reeds from the river. Fire improves the quality of the reeds by getting rid of the old ones. Floodwater encourages the rapid growth of the reeds since reeds are water loving plants.

PAGE 191

175 Good fire is the one which pushes away snakes and creates space for easy collecting of reeds. We often meet pythons when we are collecting reeds. Bad fire is the one set before others can collect reeds and even before they can transport their reeds from the river. Some grassy plants hamper my access on the way to the reed bed that is why fire helps to create space. For mabinda mats, we go earlier, using mekoro, so that we can make and sell earlier, and even have some for the children to sleep on. Conflicts arise when other people steal others reeds. This is discovered by tracing their footprints until the culprits are found. In my opinion I would report these culprits to the police. In the old days, they used to burn each and every year, but now there is completely no fire. In the old days we believed that burning renews the vegetation, and anyone use to burn if they wished because everyone thought it was valuable. I dont really have any idea about who should be in charge of giving permission to burn. The law that states no burning is only good for the reeds and grasses, but not for people. As to who may use different areas, in the old days, the Xanikhwe used to harass other tribes telling them they owned the area and that no-one should collect there. They nearly killed my brother. But nowadays these things have come to an end due to the equality brought by the law. If people dont want to walk so far, or if they want quality reeds, they will go earlier to collect. But those who go earlier are usually those who are brave enough to collect at higher water levels, while those who go late are the ones who are scared of crocodiles and snakes in the water. In addition, funerals and sickness usually contribute to the late collection of reeds by other people. It is not easy to burn at the start of the winter season due to floodwater which prevents burning. The right time for burning is spring. Fires used to occur often in the past during this season. They would burn over a large distance, and in this way they were much bigger compared to how fires are nowadays. Village Leader, Sekondomboro Village Our household collects reeds and thatching grass. Crocodiles and otters affect fish. Floodwater affects fish because it kills them when the floodwater combines with the ash from fire. Fires dont affect fish directly because they stay in the water. Floodwater increases the growth rate of reeds since it comes with good nutrients. And fires only destroy the reeds, which decreases the number of reeds. But it also increases the rate of new shoots because it provides space for them to grow. Fires both destroys and helps in the growth of thatching grass, similar to the life of reeds. As for grazing, floodwater helps the re-growth of grazing because of the increase in nutrients. Also when it floods the grazing pastures are filled with water. When fires burn in these areas, the old grass is removed and when it floods the re-growth of new shoots starts.

PAGE 192

176 Fire cleans the dirty areas so that re-growth takes place nicely, e.g. for tswii when it is burnt it grows well with no obstacles, which means it increases the number of water lilies. A fire that is said to be good is one that satisfies their needs. Accidental fires are not good, because people were not intending to burn. Fires that are set intentionally are also not good, because that person would have gone against the law. Fires which are set during the wrong times, such as people burning before others have cut reeds and grasses, or burning after they have cut but before they remove their bundles, are bad. There are always conflicts between villages when a fire has been set, especially ones that destroy their property. In 1999 cattle of Sekondomboro were burnt in the delta because of the fire. Some were killed and some were injured e.g. burning ears and tails, etc. The fire was said to have come from near Ngarange village. In 2001 a young girl was almost killed by fire in the delta, when they were cutting reeds. The fire was from Kgaolatlhogo, and it blocked their way of escape which means that fire can kill a person. The fire which happened in 1999 was dangerous to the point that even wild animals were killed, for example sitatungas were found burned. The fires were so big. Small fires were seen in 2001 in some areas, it is only that there was water around, without which it could have burnt the whole area. In most cases the fires that burn in Sekondomboro are from around here, not from Xakao or Ngarange sides. If fire were to be allowed and controlled, only the government would do it. Also if the tribe agrees with the statement, maybe the chief might be the one to control the burning, because they are his people. In the past no one controlled burning, but people only burned secretly because of the law. People burnt so that they could hunt animals properly because the obstacles would be few, leading to a clean area. People go and collect reed and grasses where they like. It depends on their feelings, on where to cut. Even between villages no one controls where each village will cut. If chiefs would be given the authority, they could control other people from other villages from coming to get reeds or thatch in their area, because they might be the ones doing the burning. It depends on how people feel on what time they go to collect. But another factor is water, if there is more water, people are afraid to go into the delta because of dangerous animals like crocodiles. In 2001 a crocodile bit a woman on the arm during the cutting season at Ngarange. In the past people were controlled by the chiefs on whether it was time for them to cut reeds or grass in the delta. People should not burn before winter, so that the grasses should grow strong and produce seeds to avoid extinction. People should rather burn during November or December. This is so that people will have finished cutting, and in winter no one will have started cutting yet. People should at least skip one year before they could burn in the same place. In the past every year you could see a fire in the delta and the dry bush. In winter you could see fire more often in the bush than in the delta. Nowadays fires dont occur as often, which means that when it burns the whole place gets burnt.

PAGE 193

177 Middle-aged Woman and Husband, Sekondomboro Village Fires should be set where and when it is completely dry then it is OK. If people burn when it is still wet, new shoots get killed and cant grow. During the rainy season and the start of the flood people should not burn, it destroys new growth. But some people do this out of jealousy. There is increased conflict about burning, because without the chiefs control people burn at the wrong time, while reeds are still there waiting to be collected. Conflicts come from when those who cut early set fires so that their cattle can graze. I cut early because I am afraid that people will burn. Those who cut late do so because they are afraid of the water its too deep. Those who go early want to get quality reeds its a competition. Those who are brave enough to deal with deep water get the best reeds. Everyone cuts anywhere there are no quarrels. Even before people would decide for themselves where and when to burn. Government should put a law that people can burn, and local officials should decide when and where. The Government should hire someone from the village to decide and oversee burning each year. This person must work hand in hand with the chief. It could be done on a volunteer basis. The problem is that fires cannot be stopped, they always just burn as far as the resources go, they can never be controlled. There is never a good reason to burn. These days fires are much more frequent, even every year the same place burns. Before people used to come together to discuss but now, because it is illegal, people just do it as a secret without discussing with everyone. Nowadays fires are much bigger, because some places are not burnt regularly. Some places never burn and then it is a disaster because the fire is very big. Some places burn very year, and this is no problem from a safety point of view. They should rotate burning, this will limit where fire will spread to and the intensity of the fire. In 2001, last year, a fire spread all the way from around Ngarange up to Shakawe. It destroyed reeds, even those already cut. It was after the rains had started. But there are more fires on the dry land than in the Okavango. But I dont see any adverse affects from fire except when it is big like in 2001. Annual fires would be better because we would get more reeds. On the dry land maybe the Basarwa burn to get fresh grazing to attract game for hunting. Old Woman, Samochima Village We get reeds and thatching grass in the Okavango. The bigger the flood, the better the growth of reeds and grass. Even fire, when set at the right time, improves the growth of reeds. We collect reeds opposite the fishing camp, across the main river, at Samochima, Hamonga, and Bodomachao. For fire to be useful to people who collect reeds and grass, it should burn everything in the area where we get those things. This will allow new things to emerge when the flood comes. Small fires are bad because it leaves other areas unburned, and they dont create enough space for grass and reeds to grow freely. The year before last there was very big

PAGE 194

178 fire in the floodplain, which came from Nxamasere side, going as far as Shakawe, and burning everything in its path. We benefited from it last year, as things had grown well, and this year there have been no fires in the floodplain. There is no conflict within the village because of burning on the floodplains. The VDC and the kgosi should decide or give the people permission to burn if the Government says this is OK. Some people go to collect earlier than others because they want more reeds and grass, and those who have canoes can go earlier. There is no reason to burn at the start of winter. Nowadays fires are bigger than in the old days because of more accumulation of vegetation (less frequent fires). In the old days fire were regular because anyone had the right to burn at the end of the collection period, but nowadays because of the law fire is not common in the floodplains. Reed Collector, an Older Woman, Samochima Village I mainly collect reeds from Molao. These are mostly for my home, not for sale. I also collect thatching grass from Molao. The way it is, is that the reeds line the river, and the grass is behind the reeds. For reeds to be good, there should be a big fires so that ALL the old reeds can be removed. September is when there should be fire, because then we are expecting rain. Even November is still OK. We only collect in July or August, so anytime after that is OK to burn. It is too late to cut reeds after September, they are too hard and stiff to cut. Even thatch is difficult to cut after September. Reeds and grass grow back really quickly. After one month it is already half a metre tall. I dont skip a year after burning before I can collect there again by the cutting season after a fire there is already a lot more reeds and grass. If you burn too early it is not good because some patches remain unburnt. Then there is less for us to cut, and we have to go further to find enough. In 2000 was the last fire, at Molao it gave us good reeds. There was no fire there in 2001, and now we have to search among the old stalks to find good reeds. But there are no conflicts over burning. I have never lost reeds or grass bundles to fire because usually people tell each other before they go and burn (we do know who set the fires). If fires are set in time not too early then it would be good to legalise fire. The chief should decide about that. Whether people go to cut reeds early or late depends on the state of their compound. If it is OK you go late, if not, you go early. Fires go according to the wind. Things have not changed much since the old days. Fires are generally the same, especially in terms of timing. They are also the same in terms of size and intensity. After we have collected this year we will burn. In the old days we used to burn every year, it is the same now.

PAGE 195

179 Young Woman (from big multi-compound household), Samochima Village We get reeds, fish, makhungura (lily flower buds), thatching grass and water from the river. Most of these things we get from Bodumachao, or from Sakunyima lagoon. We use traditional baskets to fish at Bodumachao. I like big fires, we want things to become ash, i.e. burn totally so that no stumps remain. Even if it looks totally burnt new shoots emerge from the roots. There is no such thing as a bad fire. When it is dry, just as the new flood is approaching, this is the best time to burn for tswii. When the flood comes, it first makes the soil wet and the bulbs spout, then when the flood is full they can grow well. November is the only time to burn, because then people here have finished cutting reeds. Even tswii should be burnt then. There are conflicts over burning, but they end nowhere, the people are unknown so there is nothing you can do. But normally those who burn wait until people have taken their bundles from the river and then burn. There is no way you can hear that they are going to do it, it is done secretly. I saw 3 fires last year they were in the same place as the previous year, where we collect reeds. But I think it would be better to skip a year. If you burn every year the reeds are very thin, like grasses. This is just because of too much fire, because it hasnt finished growing. Tswii doesnt suffer from annual burning. It is not OK to burn in winter, because people have not cut yet. We dont go too far to collect things, there is no need to the stuff is nearby. We cross the river to the dry floodplains. If Government could give permission to burn that would be good. They could call a kgotla meeting to choose a person to decide on when and where the burns would be OK. It should not be done individually, it must be done in the kgotla, especially in case the fire gets too strong and needs many people to help put it out. One reason why some people may go later to collect stuff is if they have some left over from the previous year. If not, then you go early. Some people may just be lazy. Those without mekoro have to wait either to borrow one when the owner is finished, or until the water is low. [when asked about the smoke visible that day] I dont know what fire is out there now, it is possibly from the other side. Normally we dont get small fires here. It is a different environment compared to the other side of the panhandle where there are many small islands. Nowadays fires are not as regular as in the old days. Then they burnt every year, now the fires are not as frequent, but they are as big as before.

PAGE 196

180 Sick Woman, at cattle post on Outskirts of Village, Samochima Village We collect thatching grass and reeds from the islands and floodplains, and fish from the river channels, using traditional baskets. The main place we get thatch and reeds from is the bank opposite Samochima village. Without fire, we cant get quality reeds. They are too thin. If there is a fire, they grow stronger. They must burn where we collect reeds, burn until you see the soil, then things will grown well. Fire doesnt affect fish directly, but the new emerging shoots are good for them. Burning is really a thing for reeds, not for basket-fishing, although the number of fish do increase. A bad fire is one that is set too early, whether on purpose or by accident. Fishermen are the ones who set the fires, and they dont talk to the reed collectors and let them know they are going to burn. In Samochima many lost reeds and thatch. Some had already bundled it, and some had not even cut yet when the fire came. But it was not a big problem because we were able to cut [opposite Prices] at Sasihara. The 2000 fire was not very big where we collect. Every year there is a fire on the other side of the river. There are many fires that side, because there are the small ponds that fishermen like to keep clear of grasses and sedges that is why they burn all the time. It would be good if we had permission to burn as it gives us better access to the islands and floodplains, and more space for reeds to grow. It must be the chief who decides when and where the fires can be set. People have to collect from close by because they have no transport and carry things on their heads. Sometimes we have to wait for weekends so that the children can help us carry. The main school holiday is in August. When your family is big, you can collect more, because they can help carry. Small families get little. Some people go early so that they can collect enough to sell, so that they can buy soap and school uniforms. Mabinda mat makers also go early so that they can get the soft reeds, not the dry ones, because soft ones are easier to weave. It is not good to burn in winter, there would be no reeds that year. Maybe fishermen have a reason to burn then, to get fish out of the floodplains into the ponds. In the old days burning was different. Nowadays, it is not alright, because people set fire for no reason. Before they would really think about where and when they were burning. Wind and the amount of grass are the cause of big dangerous fires that go long distances, and now there is more vegetation, because there is not regular fire and the vegetation builds up. Long ago they would always burn after people had collected, nowadays some burn early, which is bad. It is always the fishermen who are burning. All they care about is more fish. In the old days they could be patient, now it is a business and they are running after money. It is not just that they do it secretly, they do it because they care

PAGE 197

181 about catching lots of fish. If they were to allow people to burn, maybe the fishermen would listen to the chief, but maybe not. Young Man, Samochima Village There are 15 boat owners based at the fisheries co-op. Fire messes things up for the reed collectors, it even burns the piles they have cut. In 2000 there was one big fire, and we lost reeds there. In 2001 there were small fires, on islands and floodplains with no names, in places where people dont collect reeds and thatch. Sometimes if there is no effect or benefit, it is like these fires did not happen for us. Fire is good for thatching grass, if it doesnt burn, it grows up useless. 1999 was the biggest fire. It came from Mogotho all the way to Shakawe [according to most reports this was actually the 2000 fire]. But it stayed that side of the main channel. The channel controls where they fire goes. All fires are good for reeds. August is a good time for fire because there is a lot of wind to transport it, but most people havent finished collecting by then. What delays people from finishing is that they first cut for their houses, and then they go back to cut for sale. They cut these later and these are the ones that risk being burnt. Every fire slow, fast, intense, cool, is good, but they are best when it is windy. This is because we dont want patches to be left, we want everything to be burnt. Usually fires do burn everything because of the wind. People cut where they like, three is more than enough for everyone, they even leave thatch and reeds behind. Bad fires are those which dont burn everything. If the old stuff is left, it creates a layer through which the new shoots cant emerge. I have never heard of any conflicts. There cant be any, because we dont know who sets the fire. Some do quarrel and there is a lot of talking that takes place, but there is never any proof. It has been a long time since we have been affected by a big fire. We would like it if Government could allow us to burn, and I hope the authority is given to the chief, it should be the chiefs who decide. If they called a kgotla meeting, they could give people a chance to finish collecting their reeds and grass. Burning in August is OK for fishermen that is when the fish spread out to the floodplains. But people never burn in winter, there is too much water. Some people burn on the sandveld winter is the best time for this, because the trees have lost their leaves and the grass is dry and the place is ready for fire. By spring the new leaves etc. make it too wet to burn. Every year there are fires in the Delta there is never a year without fire. They are usually big fires, mostly in August and September. Some people may even be killed by fire, e.g. in 1986 or 1987 someone died in a fire on the sandveld. Even in the old days it used to burn every year. Nowadays there are fires every year, but not in the same place every year, before the same place could burn again and again, which is bad.

PAGE 198

182 Old Woman, Nxamasere Village I usually collect thatching grass and reeds from the Delta. Thatching grass: Flood water has no benefits, it is just a disturbance to us because it makes people wait a long time before it is low enough for us to go and cut the grass. For thatch, flood water is not good, because when there is only a little water it makes the grass grow fat. Fire is very important for the better growth of thatching grass. It clears unnecessary old and dry grass so that new ones can grow without competition from the old stalks. Reeds: Floodwater is good for growth of reeds. Poor floods result in slow growth of reeds. Floodwater also protects almost every plant growing in it from fire because the wetness stops the burning. Nevertheless, fire is essential for the better growth of reeds. It also clears space for easy collection of reeds. Good fires are those set for the purpose of clearing old plants. It should be set only in those places where people would like to get something afterwards. It shouldnt be sent when people are still collecting things from the river. The right time for good fires is around October to November when everyone has collected and the land will be ready for the coming rains so that it can recover. Bad fires are usually those that burn in the dry veld because they can even reach houses and cause a lot of destruction. Animals such as goats that rely heavily on browsing end up starving because, unlike cattle, they cant eat grass in the floodplains when the dry land has burned. Bad fires are the ones which burn before everyone can collect. Conflicts usually arise when there was fire before or during the period of collection but they end up unresolved because the culprits remain unknown. In the past, fires were frequent because they were not stopped as they are nowadays. Delta fires are very dangerous compared to dry veld fires because they cant be stopped due to the river which prevents people getting there to put them out. They are usually stopped by water or hippo paths. Fires were always starting from as far away as Sepopa. People living in the Okavango should join hands to decide on the control these fires so that they can be profitable instead of just destroying the vegetation. The time of year that people go to collect certain plants depends on the individuals will and the level of the floodwater. When the flood water level is high they cant go there, only when it is low. Age and other private things like sickness usually contribute to the time of collection. Young people and those with mekoro usually go earlier than elders. It is not easy for someone to set fire at the start of winter because the floodwater cant permit burning. the plants are also still green and have a high moisture content, therefore they cant burn easily.

PAGE 199

183 Woman Owner of Vegetable Garden Project, Samochima Village She mostly users water from the river, but also collects reeds and grass. Has had farm since 1996. Came from Maun, came because the permanent water presented an opportunity. In 2000 there was a really big fires all the way from the other side to the edge of the river here that is the only big fire Ive seen, although there were smaller ones nearby last year. From the big fire there were some deposits of ash, but Im not sure if they were of benefit to the garden or not. Fires can be dangerous and trap people out there, especially those without mekoro. I collect where it is shallower. The 2000 fire was stopped by the main Okavango channel. We saw it coming from far away. It took several days, almost a week to get here. People were not afraid, they new the river would stop it. That fire was in July, some people had not collected yet, but they were resigned rather than angry. The reeds were there again the next year if you burn it, it makes it come up faster, its good quality, its beautiful, and its straight. About 2 weeks after the fires the reeds were so thick you couldnt get a mokoro through. Some people do burn in winter, but not a lot. It is bad to do so because people dont get a chance to cut their reeds. The fires are started by people not fishermen just people, in order to improve reeds and thatching grass. Fires improve access, improve the flow of water, and get ride of the old vegetation that choke out new growth. People havent started collecting yet this year. They will start in August when the water has dropped enough. Opposite from here is the main place where people come to collect. It is their departure point. People dont use vehicles, they just carry their bundles on their heads. It is mostly women who collect reeds. Some collect over on the other side of the main channel. The launching place here is used by both boats and mekoro. Wife of Village Leader, Nxamasere Village Daughter also participated. Mainly we cut reeds in the delta. Fire makes reeds sprout well, also thatching grass, and papyrus for mats, and bulrush roots to eat. Bulrushes are found close to the main Nxamasere channel. We cut reeds and thatching grass at Tsetalago, even the papyrus we collect is at Tsetalago. People do burn, but we dont know who it is. They set fire while we are still trying to cut. Even if we dont know who sets the fire, it doesnt matter. As long as they set the fire at the right time, its OK. Bad fires are those that destroy other peoples property. There are conflicts but this is difficult because no-one knows who set the fires. Mainly the problem is that they burn too early and that leaves no food for livestock. I dont know who should control the fires. Some people have different opinions of when the place is ready for burning. Some think early is better, i.e., those who hunt want fires early. Those who have cattle want to wait for the rains or flood. Those who burn early may also be doing it out of spite. The law we have now is better: it means people dont burn every year as they used to in the old days, which was damaging to the land and its resources. In the old days it was good, they burnt every year when it was totally dry, now

PAGE 200

184 they burn at the wrong time when it is still wet. The problem is the law. [note switch of position] They need to employ people to patrol for the fires and people burning at the wrong time, and to stop these people. August is the best time for burning the sign should be when the trees put out new leaves and flowers. August fire is good for grazing, for thatch and reeds. Winter fires are bad as there is still too much water. It should be in spring when things sprout. I dont know about fires and fishermen no-one in our household fishes. I have only heard about the fire at Nxamasere Lodge last year. It brought smoke and ash into the village. We saw smoke and particles for about 3 weeks. After a fire like that it should skip a year before the grass and reeds are ready to cut. People still went to cut last year, and just made do with the smaller plants. Others found places that hadnt burnt. Livestock Perspective Women, Mogotho Village Household usually collects tswii. Excessive floodwater sometimes kills small fish, but fire doesnt affect fish. When floods are poor, reeds and thatching grass dont grow well, because they need a lot of water to grow well. When they dont burn those old grass and reed stalks, the quality wont be good. Sometimes when the water level is very low, fish die because of too much mud. When grazing areas are not burnt to get rid of old vegetation, the dry and old grass cant easily be eaten by livestock, so fire makes sure that new fresh grazing is there. The useful fire is that which is usually set when the water level is low and the vegetation is dry, so that afterwards new fresh ones grow. It should be set after everyone has collected reeds and thatching grass. Bad fires are those which are set before people can collect their grass and reeds. It is also dangerous because sometimes people can burn or get trapped because it is usually big. Family conflicts are not common because the culprits are not known, so it is difficult to argue for something that you are not sure of. Fires were frequent in the past few years, but now they are not as frequent as in the old days. (Most of the fires come from Seronga all the way through Mogotho down to Mohembo.) Some people should be given power to burn the veld and control fires. It should be the kgosi who selects these people. No-one decides which families use which areas. Every family has the right to use any area without any problem. Some people go and collect earlier than others because they expect there to be fire, and they want to be sure of collecting. But high floodwaters and crocodiles may slow others from going out, and they must wait until the water levels have become very low (e.g. around August or September). Those with thatching grass and reeds which are old usually go there earlier so that they can fix their homes before the

PAGE 201

185 rainy season starts. Some rush earlier so that they can collect from the nearest beds, resulting in those going late having to walk long distances to collect good ones. It is not easy to set fire at the start of the winter season because of the floodwater and the grass and reeds are still wet so they cant burn easily. It is therefore not common to see fires at the start of the winter season. In the past few years fires were not as frequent as in the old days when people were not exposed to the law as nowadays. Nowadays the fires are small because they are stopped before they can spread too far. Fires were frequent only in the spring, so that in the coming season, which is the rainy season, the new grass and reeds can grow well. Also in the spring, grass and reeds will be dry so they burn easily. Old Man, Sechenje Cattlepost, near Sekondomboro Village Although he only has 2 head of cattle, he uses the floodplains to graze his livestock. Crocodiles were the biggest problem in the river, having killed some of his cattle. High floods can result in cattle being forced to graze in the veld instead of on the floodplains. Fire improves the pastures by encouraging growth of new shoots which are palatable to cattle. All fires are good for improving grazing. After a fire in the floodplains, it only takes a weed for the new grass to grow. Sometimes these fires in the pasture areas are set by fishermen and this ends up being useful to livestock. In the past years, some of my neighbours lost some cattle in the big fire that burnt in the floodplains. There is no conflict over grazing areas, farmers dont argue because the grazing land is large enough for everyone. In the past, grazing areas were burned more frequently than nowadays. Big fires can be dangerous even to people, especially fires in dry papyrus beds. If Government were to give us permission to burn, the kgosi should be the overseer. We would really appreciate it if there was a new law that allowed us to burn. Even in the old days, no-one had the power to burn the grazing pastures. Without fire, all the reeds and the grazing would be of poor quality. The most common bad things that people do during the reed collecting season is that some steal the reeds of other people. In the past, people from as far as Shakawe and Samochima used to graze their livestock on the other side of the river at Xaedomo. This made people steal the livestock of others and get away by taking them across the river. Some livestock just start grazing in the floodplains, even when they are still flooded, before the water level goes down because they were trained by their owners to do so. In the old days, grazing pastures were burned around August, so that they could recover during the rainy season. Old Man, Sekondomboro Village We depend on many things from the river, including reeds, thatching grass, tswii, fish, makhungara, etc. Fires does not affect fish negatively, instead it brings food for fish. Fish only die in the spring when all the floodwater has dried completely except for in the lagoons where the water turns red and hot due to high temperatures. Fire improves the

PAGE 202

186 quality of reeds by getting rid of the old dry ones and when the floodwater comes, they grow well. This also applies to thatching grass. Grazing pastures after the floodwater has gone down, thats when the livestock graze in the floodplains. Fire also improves grazing by getting rid of the old dry grass which is not good for livestock, and the new palatable shoots grow well. Fires are usually started accidentally by children. I think all fires are bad nowadays because of most of the people, including government, are against the notion of burning. The good part of burning is that it brings good grazing pastures and new reeds afterwards. Conflicts arise when people burn when others are still collecting. They never say Hambukushu are the ones who caused the fire. In the past fires were frequent but now the law has taken serious action on those who burn the veld. I think government should keep on discouraging the burning of the veld. I think this has helped reduce the number of poachers and now the wildlife is improving. The time of year that people go to collect depends on the individual. There are other who there late because of laziness. Some go there earlier so that they can come back and attend their fields earlier. People usually go when the flood level is lower. People usually set fire from September to December when the water level is very low and the vegetation is dry, but during winter it is not easy due to the high waters. Nowadays fires are not there at all due to law enforcement while in the past some fires could burn for a month. Three Older Men, Nxamasere Village In this household we collect mainly thatch, reeds and palm leaves. We dont really fish, and we dont collect tswii because there is none around here. We usually collect around Tsetalago all of these things are found in one place. When there is a fire it burns and destroys things and there is nothing left, not even grazing. Those who cut reeds dont want fire because they burn year after year which kills the roots. They should only burn every 4 years that is a good fire. If they burn every 4 years there is a good build up of vegetation and things burn well and can clear the old stalks and things out of the way so that new shoots can emerge well. When it is dry like these days we dont need fire. It is too destructive. We havent noticed any effect on the accumulation of grass due to the CBPP slaughter campaign. This is because there is so little rain, that there is no grass. Good fires are those set on time and on purpose, and when we can expect rain. If no rain is expected, they shouldnt burn. These days fires are set anywhere and anytime. Some just go and burn out of jealousy. It is difficult to confront those who set wrong fires as there is no way of telling who it is but there are conflicts. Maybe it would be easier to enforce the laws and control burning if things were controlled from within the village. We can just call a kgotla meeting every year, and choose the right time and the people who will do the burning.

PAGE 203

187 In 2001 there were fires, mainly the one around Nxamasere Lodge. In 2000 there was one that started at Sepopa and it went all the way up. In the old days, fires were better. People would never burn the same place twice. They would never burn in drought years, they would only burn just before the rain. It is difficult to compare the size and intensity of fires to the old days because of the drought. On the one hand, fires tend to be small. But so much was destroyed in the 2000 fire that only the small patches that didnt burn have reeds growing. There is never a good reason to burn in winter. It must be spring. Things cant burn properly in winter. Some people go early to harvest depending on if they need new stuff for their houses. Some want to collect palm leaves, so they go later. Some go earlier to get the closer ones, and others are afraid of the water. Old Man (at cattlepost south of Nxamasere) We get reeds, thatch and fish from the Nxamasere floodplain. We get palm leaves from Shangandu. Fire is good, it makes papyrus, reeds and bulrushes grow well, even the grazing at Shangandu is better. The fire of 2000 has made things better. Fires need to clear everything to be useful. In the reed beds and on the islands, even palm leaves are improved. Bad fires are those that are set too soon, leaving some areas unburnt. There are conflicts between people those who burn are the ones who want to graze their cattle. But you cant really know who does it. We are here on the sand [i.e. some distance away] and the fires are there in the river. For example, I havent seen any fires since the 2000 one, but there may have been. Fires are not the same now as in the old days. Now there is very little burning. Before, they used to burn every year. The chief should be the one to decide when and where people could burn, but even if they give him that task, the fires take place far away from the village. Winters are a bad time to burn because it is too wet. The proper time is spring. Even in times of drought it is fine to set fire, but you need to burn near the river so that the flooding can stimulate new grass shoots. Some people are afraid of the deep water that is the main reason they wait till later for cutting thatch and reeds. When one burns, fish are attracted from the main river into the floodplains to eat the emerging new shoots. Whether ash affects fish or not depends on the size and speed of the flood. Sometimes it flushes it away, sometimes not. Fishing Perspective Village Leader, Mogotho Village The things we get from the river are fish, mud, reeds, papyrus for eating, bulrushes, and tswii, but fish are the main resource. Floodwater is important higher water means more

PAGE 204

188 fish, tswii, bulrushes. When there is a low flood, tswii rots quickly and then disappears. I use a mokoro, so high floods mean more of the areas are accessible to us. Fire is good. It helps a lot with fishing because it opens areas for putting nets, and because new shoots are good for fish to eat. When the floodwater first comes, it mixes with the ashes from fires, which makes a poison, an acid, which kills fish of all ages. Some people who dont have nets take advantage of this to get fish. People in Mogotho dont use traditional fishing baskets because Government has stopped them, because this method catches fish that are too small. The best time for burning for fish is September and October, when it is drier, and before the flood. Fires should be small, so that there is less ash to mix with the water. The best fire is the one that burns with a strong wind so that ashes can be blown away, and even in September there are strong winds. Ideally a big fire should be set every second year. If you set it every year some plants will be destroyed. Some fishermen do set small fires just in the area that they use. And some burn on specific islands, so that it burns out into the floodplains where it get stopped by the water. This acts as some kind of control. Hunting and grazing are also reasons for burning. Island fires are typically set in August so that they dont spread so fire. Those who collect reeds complain a lot they go to the police because they dont have a chance to finish collecting. The process is that complaints are registered through the chief at the kgotla. But complaints are not common, because people are afraid of jail the sentence can be more than 10 years. But some people dont go to collect in time because they are lazy, or because they dont have mekoro. There are fires every year but we dont always know from which village the fires start. Nevertheless, there are no conflicts between villages. The Government should tell the Agricultural Resources Board about burning. There is no ARB committee in Mogotho. We have been told to form one, but this hasnt happened yet. But this could be the organisation to deal with burning. Whether fire can be controlled depends on how dry it is. If it is very dry, no-one can stop it. Generally fires in the delta cannot be controlled. Even early in winter, there is so much wind and the fuels are so light and hot, they spread easily. There are no restrictions on who can collect where. Peopled who want lots (for selling or for a new house) of reeds and thatch are the ones who go early. People do compete for reeds, this may be why some people go earlier to get reeds of better quality. If you burn too early, the grass will be stunted, as it will start sprouting before the floodwater comes. In the old days we used to burn regularly, but now it is less frequent and this is why the fires are so big these days. If you burn every other year, the fires dont go so far. One big fire infrequently is better as it stops the same place being burnt year after year.

PAGE 205

189 Large Hambukushu Household, Mogotho Village We use thatching grass for our roofs, and reeds are also important. Other things we get are tswii and fish. More water means more fish because there is more food for them with the new shoots. Less water means more thatching grass. Fire doesnt really affect fish. Floodwater is more important than fire for fish. If there are low floods, there are fewer fish. Even reeds need more water. If there is a lot of water, there wont be a big fire, and reeds are affected less. They same is true for thatching grass. Fires destroy because, with thatch, it even destroys the roods and stems, and unless the seeds have been dispersed it wont renew. When it is flooded, we cannot graze our livestock, and there is also danger from crocodiles. All the resources of the delta are equally useful to us. Fires is a bit useful for grazing, it stimulates shoots, but otherwise there is no benefit from fire it kills everything, grass, reeds and fish. Even if fire creates more open space for tswii to grow, this is actually destruction. Floodwater is more effective at opening spaces. When it is dry fires are worse than when it is wet nothing can stop them and they can even jump to the dry land. Many fires are caused by lightning, not people. The most dangerous fires are when it is dry, and they can even turn entire reed beds to ash. Lightning fires come when it is getting dry and hot, but they are as bad as those set by people. Every year there are many fires in the delta. Fires dont affect fish directly because it burns when the place is dry and the fish are not there. Ash comes and mixes with water and kills fish, but afterwards the number of fish increases the next year. More have bred, because the ash has more nutrients which makes plants grow. More sun reaches the fish nests and eggs, and these eggs need sun. It is very difficult to know who has burnt if we heard rumours, it would make the whole village go against those people. In my opinion that very few people these days support burning. Those who like to burn are very few. Fire should always be forbidden it is our lives that are being destroyed. It is not possible to control fires as this would need patrols, and poor access makes these impossible. Before Independence there was some separation between tribes about where people could collect, but now we are all one tribe, and so people are all the same. The level of water decides when people go to collect reeds and grass. There is no competition for reeds, there are plenty. Generally we dont see winter fires, they start in September. In the old days there was more rain and big floods and therefore less fires than now. Now they occur every second year. In the old days fire were slight less often, such as every 3 years. Nowadays fires are bigger than before because there is less water and the place is dryer. The fire that has changed the most is on the dry land, where fields and livestock can be destroyed. Some people do use fire to clear new fields, but no-one uses fire in between harvests we dont burn the residue because it is fertiliser and you cant waste fertiliser.

PAGE 206

190 Fisherman, Mogotho Village The most important lagoons for fish are Guruhogo, Gxdwe, and Nende: there are lots of fish there. These are shallow lagoons when the water is low you can stand there. December/January are when the fish are most plentiful. In June the numbers go down due to the cold. During the flood, there are more fish in the lagoons. When the water drops fish go to the main stream. For thatch and reeds Kashika is the most important place, but otherwise on the edge of most islands. Grasses and sedges do close up the channels, but with the new floods the channels open up again. Fire would be useful to clear channels, but since it is illegal we want the Department of Water Affairs to come and keep the channels clear. If we were to use fire, it should be in November and December, so that the grass doesnt start to grow before the flood. Also, in September cattle are still grazing on the floodplains. Fires shouldnt be set every year, instead it should be every second year. If you burn every year it will kill things totally. Firs should burn the whole area, not just small places. Such big fires can be controlled, we can go with branches to where the grass is short and beat the fire down. The community works together, so this can be organised. Burning for fish is good it brings new fresh grass for young fish and increases the numbers. When there is a big fire combined with the new flood, some fish die but the net effect is an increase in numbers. Fires are much less now, there were more before as part of our traditions. In August and September there are no fires, as people are still getting grass and reeds. In November and December that is when we would burn. But fires should not be set every year, they should only be set when the place is full of (hippo) grass and needs clearing. Early fires are bad because the rain is still some time away, and it removes food for animals. The chief should be the one to decide about burning. He should be the one to call people to go together to manage fires. Focus Group Net Fishermen, Sekondomboro Village (4 men) We fish at Hogu or Kgaolathlogo, whichever appears to have more fish. Typically we dont fish there at high flood (May to June). In the high flood we fish around Sekondomboro. The best time for net fishing is when the new floods appear, when there is new grass around December and January. The place burns every year, these places are next to the main river, and the fish come. Fishing needs fire every year. Fish prefer new green grass, thats why they burn and should burn every year. Fishermen burn just around the ponds, the fire doesnt spread which is good, because it means the new shoots are only there, attracting the fish in to where they put their nets. Big fires are bad, and are not worth it as too many resources are destroyed. By looking where the water is, you can limit the spread of fire. August to

PAGE 207

191 October is when fires should be set, just before the new flood. It is hard to burn earlier, water is still high, and also it is bad for the environment. In the old days it was better, the law was less harsh, and if the grass was old you could just go and do it, no permission required. There were no conflicts because we used to discuss it in the village beforehand. People now have conflicts because burning is done secretly and no-one is warned. Fires from other villages are not problem sometimes they even help fishermen. It should be the chief who decides. Or Government could hire someone who could be the one to decide when to burn (although not actually set the fires). The one who sets it should be the one to control it we could work together. Fish are a gift from God, sometimes one gets a lot, sometimes one gets not so many. It changes for a person from day to day. There are still plenty of fish. To do well means knowing which river channels the fish are in that year. We have noticed that ash kills fish. When the flood comes in November/December with no rain, the fish dies, if it rains the ash is dissolved first so there is no problem. Even if some fish die, they are few compared to the new generation supported by the new grass. Death due to ash in a given lagoon depends on the speed of the new flood. If it is fast, you see dead fish for one day, if slow, for 3 days. Fishermen want fires every year, but reed collectors want it every second year. This is no problem because fishermen make small fires that they can control, and the reed beds are further away. Focus Group Women basket fishers Sekondomboro Village Three women (who were also part of reed/grass group, but stayed behind to discuss fishing). Fishing with baskets is becoming rare. Mostly we fish at Hogu. The best time for fishing with baskets is October to December. Fire is not good for basket fishing. We need grass to slow the fish down and to give us cover, and fire removes this, and then we cant catch them. We have suffered because the fire last year burnt everything. When it has burnt there are no fish, and even after the flood after the fire the fish are less [some argument over this] and we cant find any. It was agreed that after the second year the numbers of fish caught have returned to normal levels. With fire, fish dont come because they need the grass for breeding, and make their nests in between the plants. Village leader, Samochima Village We make use of many things from the river, including grass, reeds, tswii, etc., but most importantly fish. Sometimes when there is fire in the river, the shallow water gets hot and this disturbs the fish living there, resulting in them moving far away from that channel. This creates difficulties for the fishermen. Even the ash from the burning grass poisons the fish. The floodwater is not good for fishing because it makes fish migrate from the main river into the floodplains. When the floodwater dries up, those fish that dont go back to the river in time end up dying or even get exposed to predators.

PAGE 208

192 Floodwater doesnt disturb grazing, because during the high water season, livestock are grazing on the dryland, where there is still grass. According to the law fire is not good, but in the past fire renewed the veld. Fire is important to all the natural resources in the delta. Fire is generally good for many people, especially fishermen, grass and reed collectors and even hunters. Those who say fire is bad are the ones who dont use any of the resources from the river. The right time for burning is after everyone has collected reeds and thatching grass. This is around November/December. Firs set around August to October are bad because at this time most people are still collecting. Whether people go early or late to collect depends on their individual capability. Dryland fires are suitable around June and July, even August, because this is the time that the livestock will move to graze in the floodplains as the water level drops. There are no conflicts arising due to access to the grazing areas. Every one is free to collect grass or reeds wherever he or she wishes to. For example, peoples from as far as Xhauga and Xakao come to collect reeds in Samochima. Conflicts between families due to fire do arise, but only when the fire was set before they can collect, and when the culprit is known. The size of the fire depends on the amount of old vegetation or twigs burning. The small the amount, the smaller the fire. The direction of the fire usually depends on the direction of the wind. Burning before winter isnt possible because of the floodwater, which stops burning. (In dryland fires, there was a case in 1986 when one man from Okusi between Samochima and Shakawe was affected by the fire when he was taken there with others to try and stop it. He got so sick that he ended up dying.) In the past fires were more frequent than nowadays. If the Government allows fire, it should make some arrangements with the chief on the salary of the person who will be appointed to set fire. Last year the fire came from three directions. One was coming from across the river (opposite Samochima) towards Samochima, another came from Shakawe, and the last one was from Xhauga. The all met in Samochima. This was in August. Fisherman, Samochima Village I depend on the Okavango almost solely for fish. When floodwater comes, it brings more fish with it because it encourages growth of shoots which are good for fish as this is their food. When the floodwater drops, the fish come to the main rivers. Fire is very good for fish because fish feed on the charcoal. Without fire fish dont move freely due to the network formed by the overgrown old grass and reeds. Generally fire is good, when considering the positive impacts it has for both people and animals.

PAGE 209

193 Usually, fishermen dont set fire purposefully, it is children who are the ones that light fires for roasting their fish and then dont extinguish them, resulting in them catching nearby twigs etc. after they have left. Conflicts over using areas usually arise between the residents of the village and the white camp owners, who dont allow blacks to fish near their camps, saying the place is strictly reserved for their guests to fish in. But aside from this, within the community, no-one has the power to decide on the area each family can use for fishing. What we do is to try by all means to improve the living standards of others. The lagoons I use are Bodomachao (near Lloyds), Sepete (between Drotskys and Lloyds), Metsi Matlala (near Xaro), Kakoro (across from Xaro), Okaxha (beyond Xaro) and Bodomo (beyond Okaxha). Our fish nets often get torn by crocodiles and otters they even eat our catch. Another Fisherman, Samochima Village I depend heavily on reeds, thatching grass and fish. Looking at fish, floodwater is essential for the healthy growth of fish. It also provides new food for them. When the floodwater dries up or even just gets low, fish dont grow or even breed well. Floodwater creates more room for the breeding of fish because they will make use of the floodplains instead of competing for breeding space in the main river. The bad thing about floodwater is that it also brings more fish predators to our fish nets, bringing down the catch. There are more fish during the floods than when it is dry. This is because of more competition between the fish predators, e.g. crocodiles, otter and fishermen in a very limited area during the dry season. I am totally against the use of fire as a way of improving conditions for fish. In the old days they used to fish using traditional nets made of reeds tied together with sisal strings that didnt require any use of fire to create space. These would usually be used during the first appearance of the floodwater to catch fish that were migrating with the floodwater into new places. 30-year old Fisherman (from the fishing co-op), Samochima Village Floodwater spreads fish all over the floodplains. As a fishermen, I think fire is good. When it has been burnt we get a lot of fish because there is a lot of good space. Fire is good when it only burns certain places; it is bad when it burns the whole delta. In 1999 there was a fire which was said to be good because it burnt areas which people wanted to be burned. We fish all year round every month. During the rainy season, starting in November, this is when the fish are plentiful in the floodplains. Even hook fishing is done. During spring when the floodplains are dry some places are cut off with the fish in them which kills them.

PAGE 210

194 There are no conflicts among fishermen. Whites, such as PJ at Nxamasere Lodge at times try to stop people fishing and try to control the fishing network around the places neighbouring them. The same is at Drotskys and Xaro. There has never been a fire during the start of winter; they are only in the spring. The people who burn are the ones who use mekoro, because they can move around secretly because they dont make any noise. Maybe the VDC members should be given the authority to burn and control fire. The lagoons where I fish are: Samochima, Metsi Matlala (near Xaro Lodge), Kakuru (near Xaro), Okaxha (near Ngarange), Sekandoko (near Bana ba Metsi), Thungunu (near Matswabi), Seulu, Hogu. I fish using both gill net (January to June when the water is high) and hook and line (July to December). I use a mokoro when the water is low (July to December) and a boat with the gill nets when the water is high. Another Old Woman, Samochima Village We rely on the delta for reeds and fish. Fire is very important for the quality of both reeds and thatching grass. After fire, the new fresh ones grow and fire gets rid of the old dry stalks. Floodwater is important for the rapid growth of reeds. The floods are only regarded as bad when they come early before people have removed their reeds. Bad fire is the one that is set before everyone can collect. There are many cases where fire is set before others havent collected or removed their bundles from the river. Last year there were some bundles burnt to ashes. But fire is only profitable to people, because wild animals end up being killed, and the lucky ones end up with nowhere to stay. This ends up affecting the deltas ecosystem resulting in the extinction of other species which are few in number. The decision-maker about when to burn should be a man who can listen attentively to the peoples views/ideas. He should burn only when the residents allow. Even though conflicts do arise on who is to collect reeds where, What I do is just carry on with my job as long as I know that there is no one who can totally stop me from collecting. Early collection of reeds is determined by the level of the floodwater and the intention of the collector. That is, people who need to repair the houses or compounds go there earlier so that they can fix them before it rains. As for fishing, I use a basket. Leeches are the most disturbing things that we come across every day. There are a few cases where someone has been attacked by crocodiles while fishing with their baskets. The bad thing about the use of the fishing basket is that it can even catch the smallest fish which cant be eaten. This brings down the reproduction cycle because if we carry on fishing like this, it will end up with no new generation. That is why the government set the law which states that when you catch a small fish, throw it back into the river. This is to try and keep the cycle going to avoid their extinction.

PAGE 211

195 Middle-aged Fisherman, Samochima Village Besides fish we also get reeds and thatching grass from the river. We used to get tswii, but now there is none left because there is no fire. The main place where I catch is Xtinxo, near Xaro Lodge. This is a small lagoon in front of Xhauga village. Also at Auxwi and Nyamoka, which are far away, near Nxamasere, but on the main channel, accessible by boat. For a fire to be of benefit to fishermen it must be a big one, and go a long distance, to take away all the old vegetation this is the best. It never kills everything, wherever there is water patches of vegetation do remain. These things all re-grow from their base or roots so they are not damaged by fire. Even if you burn every year, you can still get big fires. There is no difference in the size or intensity of fire whether you burn every year or once in seven years. If you burn one side of the river only, it is bad for fishing because the fish hide in the old reeds on the unburnt side. When the rain starts is the best time to burn so that the ash can be washed away the diluted ash is good for fish to eat. There are no conflicts about burning because usually people burn at the right time. The last fire I remember was in 1999, it came up from Nxamasere in the spring. In 1998 there was one that came from Nyamoka. It is the wind which always brings the fires here. It is possible to burn small places, it is easy to put these fires out, but the small fires are useless. In the old days everyone was free to burn, even children, but maybe children would burn before people could collect their reeds and grass. The Government should be the one to decide when there could be fires. They can have someone who comes from Gaborone to set the fires. They should have a kgotla meeting where they decide when and where they want that Government person to set the fires. It would be no problem to coordinate with downwind villages, because even they will choose the same time of year. For access to resources there is no competition, no division of areas people just go where they want to. It depends on the individual whether she or he goes early or late. Some are afraid of deep water. People just tell each other informally that now is the right time to fish, collect reeds, etc., there are no formal meetings to start these activities. Even in the old days it was up to the people, not the chief to decide that now was the time. Even for ploughing. Having the chief announce it is Tswana, not Hambukushu culture. It is not possible to burn in winter. Nowadays if you set a fire it is going to be very very big, there will be terrible fires because there is a lot of grass because they dont burn every year anymore. There are no fires at all these days, not even secret ones.

PAGE 212

196 Bad fires are those set before everyone can collect what he or she needs from the river. These are the fires that are usually set at the start of winter. Truly speaking, fires are very Water Lily Perspective Focus Group Tswii Collectors Sekondomboro Village 5 women. They really want tswii, fire really helps this. But the time of the interview (June 2002) so far havent benefited from the September 2001 fire because the time for collecting tswii is July/August, when the water is lower, so maybe when they go out there they will find that there is a lot. Ndaminarundu is the place where they get tswii. Other info, papyrus for mats is always collected, never bought. It is collected in December. We go on foot to cut it. People make these mats every year. They cut it when it is still green. It is left to dry for a week before making into mats. Fire is good for papyrus, same as for thatching grass. Papyrus responds very quickly to burning. Other resources we use are the riparian trees, such as mochaba, mokochum, motsaodi. There are also many leafy plants (merogo) that are important in our lives. If these things are found close by to the village, we sometimes make specific trips to collect them, but if they are far away we will just look for them when we are harvesting reeds. Fire is good for all these things. Floods: 2000/2001 was a low flood, while the 2001/2002 flood was big. Big floods are better, things grow better. Flood doesnt affect fire, the vegetation always gets dried out. Fires are bigger the year after a big flood, because there is more to burn. The size of flood depends on how much water the sea releases. In the old days floods were very big maybe it is the developments that are taking the water. It never was as dry as this before, but now there are pipelines sucking water the floods will never be as big again. Old Man, Samochima Village I collect fish, reeds, thatching grass and lily bulbs from the river. Floodwater is important for fish because it reduces the competition for food by providing more food. Floodwater also provides a breeding area for fish by filling the floodplains, thereby reducing the competition for breeding areas. Fire creates more space for the free movement of fish from the main river to the channels where fishermen can easily catch them. As for reeds, floodwater is essential for the rapid growth of reeds since they are aquatic plants. Floodwater helps reeds recover very fast after fire. Fire gets rid of old useless reeds therefore promoting new and fresh reeds. Lily bulbs also grow better when the flood is high than in the dry season. This is because floodwater brings nutrients along with it, and deposits fertile soil which is important for the healthy growth of tswii. Without floodwater tswii grows slowly and is small in size. Fire creates more space for the easy collection of tswii. It gets rid of weeds (e.g. hippo grass) which compete with tswii for water and food.

PAGE 213

197 go to collect, Dihomeno, Botahi and Chorohane lagoons. In the low water season we go every day and get a full bowl. We collect for about 3 months of the year. dangerous towards the animals and birds that use the delta as their habitat. This is because it may bring some species to the point of extinction. Old Woman Tswii Collector, Samochima Village The old lady depends heavily on tswii and fish. Floodwater contributes a lot towards the health and rapid growth of tswii. Fire also encourages rapid growth of tswii. It also creates more space fore easy collection of tswii. Without fire I would not be able to collect it due to the overgrowth of hippo grass which is an obstacle. Fish are also more easily found after fire because they can easily move from the main river to the shallow lagoons where I can catch them with my basket. Fish and tswii make up my daily diet. For fire to be useful to people, it should be set only at place where they would like to collect something afterwards. They shouldnt burn unnecessarily. Bad fires are those which are set before everyone can collect reeds, but for the case of tswii, there are no bad fires because the fire doesnt affect it directly since it is in the water. Tswii only enjoys the positive impacts brought by fire. The other bad fire is the one which burns until reaching dry land, as it might also burn the fields. There are usually no conflicts because of burning on the floodplains. In the past 10 years fires were really frequent, each and every year there were fires in the delta. Now they are not as frequent as in the past due to the law enforcement. All people should be responsible for the control of fire everyone should be involved. No one decides which family uses which area because we all believe it is a gift from God for everyone. The level of the floodwater controls the time of year that people go to collect reeds and thatch. Owing a mokoro is also something that allows some people to go there earlier. Late collection is often a result of old age, and other things like sickness and funerals. Burning before the start of the winter season is not possible due to the floodwater. The plants are still waterlogged, so they cant burn easily. The right time is around August when all the plants are dry. In the past fires were big and frequent, and burnt after each year. Tswii Collectors, 2 Old Women, Samochima Village Hippo grass obstructs tswii the most, because it crowds it out. This grass is of no benefit to us, it is only food for hippos. There is nothing you can do to remove this grass, nothing but fire. We get tswii when the water is low, we collect it along the main channel. We just walk into the shallow areas. There are some lagoons near Xhauga and Sekwata where we also

PAGE 214

198 Fire cannot kill or disturb tswii, it only supports its growth. The fire needs to burn fast, or even slow, as long as it removes the hippo grass so that the tswii can emerge. The fire should be set in spring, this is the best time. The bulbs are always there in the mud. Once the grass has been removed, they can grow. We get tswii by feeling for it with our feet. After burning, we wait for the next spring to collect. The year of burning the fire comes too late for them to emerge. All fires are OK for tswii, because the process is just to remove the hippo grass, it does not affect the plants directly. People can collect wherever they want, there are not specific areas for different families. Even in the past this didnt happen. Hunger is what pushes some people to start collecting tswii earlier than others. Even when it is deep they will risk dying if they are hungry. So for tswii, also, the level of water determines when we can collect. Here it doesnt burn every year, so there is less tswii that there is at Sekondomboro across the river, where they burn every year. We dont collect to sell, only for eating. No-one sells tswii in the village. We havent seen a fire that has helped us in a long time, for about 20 years. In Sekondomboro there are no police that stop them burning. That is why they do it and get good tswii. The big fires dont cross the river and are therefore no benefit to us. Even at Sakunyima it is crowded out with hippo grass and no-one can collect tswii there. In the old days, it used to burn more frequently than now because there were no laws against it. Fires were bigger then. If people burnt, and the fire was put out by water in the river, they would relight it to make sure that everything burnt. The timing of the year for burning is just the same now as then spring. If it was legal, we would like to see a fire every year to get life going. Old Woman, Nxamasere I used to collect tswii in the past but nowadays I cant due to the accumulation of hippo grass which totally covers the lagoons where we used to collect from. We can only get very thin tswii that is unhealthy because it lacked enough food due to competition with the overgrown grasses. Without fire everything is useless. Hippo grass acts as a weed to tswii. Monkeys are the only animals that compete with people for tswii [!!!!!]. Fire clears off the hippo grass thereby giving tswii more chance for better growth. Without fire, tswii roots cant anchor themselves in the soil, they will just be suspended on top of the grass which cant provide it with food. Fish also benefit from tswii by eating its shoots which only grow after fire. It also creates space for easy collection and transportation of tswii. The overgrowth blocks the ways of mekoro, hindering people from reaching the lagoons deep in the delta. Tswii is usually collected from August to November when the water level has gone down. People collect at this time because of the crocodiles. When the floodwater is low, the crocodiles stay in the main river, thereby giving people a chance to collect tswii in the lagoons. Knives are used in the collection of tswii.

PAGE 215

199 leaves. Also, anyone is free to collect anywhere they wish to. The Wildlife Department In the past when there were many lagoons, there were no conflicts over which family or tribe should collect tswii where. Nowadays conflicts are there because there are few places. Other lagoons have been totally covered by hippo grass, bringing a lot of competition among the residents. Tswii was the easiest food we could collect from the river, comparing it to fishing which needs special techniques. Tswii also has some medicinal value. Those who suffer from coughs boil the skins and drink this liquid. Moyei Woman, Nxamasere Village Mostly I collect tswii from the river. Floodwater improves the quality of tswii. It grows faster and healthier. Fire doesnt disturb tswii because it is in the water where the fire cannot reach. Tswii is usually collected during the dry season (August to November). There are no conflicts arsing from the collection of tswii. We use knives to collect it. When the water is still high, tswii cant be collected because it is still young. After the water has dried, thats when it will be ready for eating. Fires clear large spaces for the free growth of tswii. It grows well in an open area. Palm Leaf Perspective Thokadi (basket-making group) Women, Nxamasere Village We collect palm leaves at Tsetalago. We use knives and hoes. These are the only implements the government allows us to use because they dont destroy the palm seeds. We dont collect palm leaves during winter time because they are brittle. Leaves are collected in spring and summer. Fires are not often seen in the palm beds, but if they are burnt, good mokola emerges the following year. Palm trees are usually not affected by fire because they are found on the islands, and fire doesnt easily spread in there. Floodplain fires cannot be controlled because of the accumulation of grass. Fires dont affect fish directly because they are in the water, and so are protected. But the eat new shoots after it has been burned. Fires are bad for wild animals and birds because they burn where they are breeding. In the past people managed to control fire because they knew the right time of burning, e.g. the dry season. They burnt knowing that the vegetation would recover when the rains started. In the past fires were not common because people knew the right times to burn and usually skipped some periods before burning. The chiefs should choose those who should burn and the law should be enforced tightly to make sure they are well controlled. As to conflicts, there are no conflicts about fire because we never know who burnt the floodplains which is pointless. There are also no conflicts between those collect palm

PAGE 216

200 papyrus for mats, when the water starts to drop. Papyrus for eating when the flood is full. is the one that hinders the collection of any natural resource, e.g. they restrict the use of hoes and axes because these can kill the palm trees. To make a basket we use one bundle of mokola leaves and one bundle of lechulo (lather) grass. We colour the palm leaves with motsentsila, mothlakola, motsiya roots which are boiled in water. The price of a basket depends on the quality of the baskets we make. They range from P10P20.00 for each basket. Our organisation was started with help form Kuru Development Trust which deals only with San groups, but our group deals with any tribe willing to join. Anyone can come and learn how to make baskets. As for other resources, for tswii to regenerate, after you collect it you peel back the skin in the water. These skins are the ones that grow up again into tswii. Floodwater increases the growth of tswii, reeds and thatching grass. When there is fire every year the thatch and reeds dont recover properly because they dont have the chance to grow as tall as they can. One complaint is that we should be allowed to burn because when it is burnt all sorts of plants that have been burnt grow nicely the following year. Old Woman and 35 year old Man, Nxamasere Village It is mainly reeds and thatch that we use, but even palm leaves are important. We go to Tsetalogo to collect grass and palm leaves, and there are 2 places for reeds: one for each tribe. The Hambukushu go to the one side of the Nxamasere valley, at Nabe, while the Bakgalagadi collect on the other side, north of Tsetalago. Since we are Hambukushu, we collect reeds at Nabe. This is an artefact from the time when there were two different villages, but after the 1960s the villages moved and merged into one. The palms are found on islands, not on the floodplains. Big fires are good to clear out the old stuff so that news resources can emerge. In the old days fires were set before the rain but the problem is that these days there is not much rain. If there is much rain, the burning of fires is not a problem. When palms are burnt, they dont produce good leaves. Fire is not that good for palm leaves. Because of the drought we should not burn as it will kill the thatching grass totally. So bad fires are those set in the time of drought. I dont really know about conflicts. The year before last, 2000, there were conflicts. Some people had not had a chance to cut, and others lost their bundles in the fire. But no-one was found responsible. People live far from the river, and dont see what is happening. The flood arrives here around February, and is at its fullest by March/April. In July it starts to go down, and by September it is dry, and stays dry until January. We expect rain in October and November, so September would be the best time for fire. In Nxamasere the seasons for collecting are as follows: Aug/Sep for cutting reeds and grass, and

PAGE 217

201 Palm leaves we get any time of the year. Fish, we get in spring, and follow the lagoons as the dry out. Tswii also we get in spring as the lagoons dry out. People could come together to decide when to burn. People could be happy to have localised permission, but I am not sure if it would make a difference, because some people (like fishermen) will always want to burn early, like in June, because of collecting honey, hunting, or small boys wanting to cook their fish and letting the fire escape. I dont know the history of fire in this place, but I do know that every year there are small fires. People go early to collect because they want to sell. They can be the ones setting fires so that others are forced to buy from them. Those with a lot of things to do go late. Some may be afraid of the cold. Here we dont need mekoro, but we just go on foot. It is not good that people burn in winter because it is too long until the rains come. In the old days they used to burn properly because they would set one big fire. Nowadays they only set small localised fires which benefit only the fishermen and hunters. These days there are only small fires because people are afraid of the Government. We have heard about the channel blockages, which is why the flood is low. That is another example that they dont burn much. The fires are more often than before, they set it year after year which is bad because it burns the grasses while they are still small. Basket maker (woman), Nxamasere Village I sell my baskets to the white people who come from the Lodge to visit the village. I collect palm leaves at Tsetalago. There are many small palm trees there, very many, enough for all the people. But people go less these days because of elephants. Also, elephants destroy the small palms. Everyone in Nxamasere knows how to cut well so that the palm does not die when we harvest from it. Fire is bad for palms, they cannot re-sprout. Phoenix palms can, but not the ones we use for baskets. I havent seen a fire recently in the palm areas on the islands. This is partly because the grazing of livestock reduces the fuel load, but also people are afraid of the law and dont burn, so fire tends not to be an issue. The Bakgalagadi cut at Tsetalago, while the Hambukushu cut at Shangandu. This division is there for all resources. I am part of a group that sells baskets, all organised from within Nxamasere village. We even sell to Botswanacraft. We have someone in the group who controls the quality. But definitely we should not burn in the palms, it really kills them.

PAGE 218

202 burn between harvests, because the termites eat the leftovers and return nutrients to the Non-User Perspective Wealthy Man (owns Land Rover 110), Sekondomboro Village People burn a lot its a problem. They burn reeds and burn grass. Its not good that they burn. Things get killed, every small animal and bird dies in these fires. Fish die the ash from the fire kills the baby fish. Small fish cant go into the deep water to escape the ashes. People burn themselves by standing on hidden coals such as when there are peat fires like the one in 1987. I fell into one. The peat fires happen after small floods, when the floods are too small to put out the peat fires. And when these fires burn, they change things that place used to be an old reed bed, and now it has been replaced by grass. Floodwater brings more fish. They spread out into the floodplains. Fish use these vegetated places for breeding. There is no benefit to people from fire. I disagree that it is need to clear out older vegetation. The floodwater does that and pushes the old vegetation to the banks. There are never any good fires in the river because it kills the animals living there. It doesnt help people in their livelihoods, in the long term they get less. In 1998, about September 15, at Sechenje, people set fire for hunting and they were caught by the local police and put in jail. Local residents were angry about the burning and told the kgotla to call the police. These are hunters who are burning. One cow was killed by the hunters, then they set fire to pretend it was killed by fire. Dryland fires set by hunters chase tortoises into the river. Thatching grass collectors dont like hunters because they burn. It is not those who graze who set fires, it is just the hunters. In winter people start to collect reeds, but this is the time people burn for hunting. Burning isnt necessary for livestock grazing, they can eat the dry grass, it is still palatable. In the past fires were more frequent than these days. Now they are less frequent because of the law. Fires are put out even before they spread further, whereas in the old days people liked fire and would let it burn. They liked fire because they all like hunting. Now they are educated. I didnt hear about the big fire last year, as I was not here. In 1999 there were just small fires, mainly on the bank. I havent heard that the Kgaolatlhogo burnt, apart from that fire. Someone should be employed to catch the people who sent fires it should be that the chief calls a meeting and selects someone to oversee this. Some people used to burn the small branches when first clearing a field. But we dont

PAGE 219

203 because it is too dry. soil burning prevents this happening. In the old days, fires was always prohibited and considered bad even the chief wouldnt allow people to burn. As for areas of use people can just go and collect where they want. There is plenty for everyone. We even leave lots unused each year. Floodwater (depth and crocodiles) determines when people can go. All the people go together and cut in the same month, its not that some cut earlier than others. There is never such a thing as a good fire, early or late they are bad. People set fires early for hunting. I wouldnt say that fires are bigger nowadays they are smaller because they get put out, because of the law. Small fires do start, but are put out before they were always left to burn. Now with the law, there are fewer early fires. At meetings at the kgotla, MPs are always always warning about fires, especially with reference to tourists who dont like to see burning. Kgotla Meetings Kgotla Meeting Sekondomboro Village Present: 13 men, including chief and 16 women. After introduction, the following comments received: The practice of burning is outdated. It affects animals, and even the resources that we use. It is not good for tourism. Burning is important: it helps us get fresh new resources. Now there are no fish caught, because there is too much grass. There is less thatching grass. When fires have been removed, there are no more reeds. I want people to be allowed to burn. We should be allowed to burn between August and September, and to meet conservation issues, fires should only be set every 2 years. Fire is a way of cleaning the government wont allow us to clean. I would like fires to be allowed, but if it escapes it has a bad side. Fire can bring new stuff and remove the old reeds that are in the way. Kgotla Meeting Samochima Village Present: 28 men (mostly elderly) and 11 women. Comments after the introduction are as follows: The place is dying because there is no fire. Even in the river in the old days it used to burn, now there is no new food. When there is drought, burning should be done in October so that our livestock benefit from the new grass shoots. Fires were important in the old days as there were good rains, but now fire is bad

PAGE 220

204 people even now are burning secretly. Maybe we can improve the situation by It would be better for decisions to be make locally, as we know what is needed and when it is good, and we can control the people. Fire is definitely of benefit new grass shoots that from new from the soil cant get through the old growth. Reeds especially need to be cleared out. Water lilies, people eat a lot, and this needs the floodplains to be cleared so they can grow. Even fish need new shoots. Without fire, the river is choked to death by old grasses. Fire cleans things. It removes the old stuff and lest new things grow. Fire is renewal. For livestock, fire is very important. All the men were vociferous about how much the environment is worse off for lack of burning, and that it should be reintroduced. The women did not speak, nor attempt to. Kgotla Meeting Nxamasere Village Present: 8 ward headmen and tribal clerk Comments after the introduction: No-one likes fire sometimes it destroys peoples things, sometimes it destroys Gods things. In the old days fire was OK because people burnt in time. These days Government has a law that prevents this and fire has become bad. I would like it if there could be a committee that could decide where and when we can burn, i.e. when the rain has started. Right now, I dont want there to be fires any more, because these days we have less rains and smaller floods. There was a fire at the lagoon which destroyed the reed bed, and now there are no reeds there anymore. There is no benefit from fire even the mogonono trees on the dry land are dying out because of fire, that is why there are none in Samochima. Even in the grazing areas, because there is no rain it is bad to burn. If ever you burn it should be November or December, when the rain comes. But in drought times, it kills everything, it destroys our nation. In the old days it was good because they burnt in time and switched areas so that the fire stopped when it hit a previously burnt area. Now with no regular fires, it just burns everything. There is no rain now, so the fires burn other plants which are dying out. Even if we were given control, others would still just go and burn, so things would not change burning would still be a problem. Others go out and steal others reeds, and then set fire to hide the fact they claim the reeds were burned in the fire. After that fire, you look, there is now no grass, and everything is just small and sparse. The interaction of fire and water is essential. When there is drought, even the roots burn because there is no water in the soil. That is why fire kills things completely. We cannot compare with the old days because that was when there was lots of water. Now Nxamasere is just too dry. Even now Government has the law but people still burn. There is no control,

PAGE 221

205 choosing committees to control, to and put the fires out and work like the game scouts and do patrols. Fires are usually set by poachers or those stealing reeds or thatching grass. If there is no rain or floodwater, fire is destruction. There is nothing that can grow after that if it burns. In the rivers, on the sand, the fires result in nothing. Really water is the most important thing. Right now, the big accumulation of old grasses upstream is what stops there being water here, or even reaching Maun. There needs to be fire, but in the right places, to remove the accumulated vegetation that is stopping the water from flowing. Year after year there is no rain, the sun itself is already burning things there is no space for fire. Nothing is growing because there is no rain and no flood. Those ARB committees are just volunteers, so they have no power and no reward to catch people and stop fires. It must be someone with authority and power who is employed to stop them. We are well aware that fire could benefit people, but not as things are now. If there was rain, a lot of rain, they could burn. The way to work it would be to have a small controlled burn and then wait for rain. If it comes, then they could burn further. We have police, we have the wildlife department, we have the ARB. If there was someone employed, it could get people to help go and put the fires in the delta out, as happens on the dryland. In 2000 there were peat fires. You cannot stop these or even see them. Even employed people couldnt stop it because it is physically impossible to stop. And these fires are the ones that kill the roots. River fires are more difficult than sandveld fires, because the fires are too big. We have never in our lives been able to stop one, only the water can stop them. If we had a committee that could be empowered to control that might be the only solution to manage it. August is the month for collecting reeds and thatching grass they need to set seeds before they are cut. That committee must enforce that people collect on time too.

PAGE 222

206 Gopal, B., Junk, W.J.& Davis, J.A. (eds.) Biodiversity in Wetlands: Assessment, LIST OF REFERENCES Afriye, E.K. 1976 Human Resources of the Okavango Area: Some Implications for Development Projects, in Symposium on the Okavango Delta, Botswana Society, Gaborone, Botswana. Agresti, A. and Finlay, B. 1999 Statistical Methods for the Social Sciences. Prentice Hall, Upper Saddle River, New Jersey, USA. Andersson, L-G. and Janson, T. 1997 Languages in Botswana: Language Ecology in Southern Africa. Longman Botswana, Gaborone, Botswana. Carney, D., Drinkwater, M., Rusinow, T., Neefjes, K., Wanmali, S., and Singh, N. 1999 Livelihoods Approaches Compared. Department for International Development. London, UK. Cassidy, L. 1997 OKACOM Diagnostic Assessment: Human Environment Botswana Sector. UNDP/GEF, Gaborone, Botswana. Central Statistics Office 2002 Population of Towns, Villages and Associated Localities in August 2001. Government Printer, Gaborone, Botswana. Cerulean S.I. and Engstrom R.T. (eds.) 1993 Fire in Wetlands, a Management Perspective. Proceedings of the Tall Timbers Fire Ecology Conference, No. 19. Tall Timbers Research Station, Tallahassee, Florida, USA. Chambers, R. and Conway, G.R. 1992 Sustainable Rural Livelihoods: Practical Concepts for the 21st Century. Institute of Development Studies, Brighton, UK. Department for International Development. 1999 Sustainable Livelihoods Guidance Sheets.. DfID. London, UK. http://www.livelihoods.org/info/guidance_sheets_rtfs. Site last visited May 2001. Ellery, W.N., Ellery, K., McCarty, T.S., Cairncross, B. and Oelofse, R. 1989 A Peat Fire in the Okavango Delta, Botswana, and its Importance as and Ecosystem Process in African Journal of Ecology Vol. 33 pp 7-21. Ellery, W.N., McCarty, T.S. and Dangerfield., J.M. 2000 Floristic Diversity in the Okavango Delta, Botswana as an Endogenous Product of Biological Activity in

PAGE 223

207 Function and Conservation. Vol.1. Backhuys Publishers, Leiden, The Netherlands. Environmental Science and Engineering, Inc. 1994 Upper St Johns River Marsh Controlled Burn Study: Response of Vegetation to a Controlled Burn in Sawgrass and Maidencane Plant Communities in the Upper St Johns River Basin Final Report. St Johns River Water Management District, Palatka, Florida, USA. Fidzani, B., Mlenga, W.S., Atlhopeng, M. and Shatera, M.M. 1999 Socio-economic Effects of CBPP in Ngamiland. Ministry of Agriculture, Gaborone, Botswana. Government of Botswana official webpages. Government of Botswana, Gaborone, Botswana. http://www.gov.bw/economy/index.html. Site last visited October 2002. Government of Botswana 1978 Herbage Preservation (Prevention of Fires) Act (Cap 38:02). Government Printers, Gaborone, Botswana. Gumbricht, T., McCarty, T.S., McCarty, J., Roy, D., Frost, P.E. and Wessels, K. 2001 Remote Sensing for the Detection of Sub Surface Peat Fires and Peat Fires Scars in the Okavango Delta, Botswana. Submitted to the South African Journal of Science. Reproduced with permission in McCarty, J. 2002 Remote Sensing for the Detection of Landscape Form and Function of the Okavango Delta, Botswana. Dissertation. Department of Land and Water Resources Engineering, KTH, Stockholm, Sweden. Hardin, G. 1968 The Tragedy of the Commons in Science, Vol. 162, pp 1243-1248. Hutchins, D., Hutton, S.M. and Jones, C.R. 1976 The Geology of the Okavango Delta in Symposium on the Okavango Delta, Botswana Society, Gaborone, Botswana. Jensen, J.R. 1996 Introductory Digital Image Processing: A Remote Sensing Perspective. Prentice Hall, Upper Saddle River, New Jersey, USA. Larson, T.J. 1980 The Hambukushu of Ngamiland, Ecology and Settlement Pattern of a Riverine People. University of Virginia, USA. Lugo, A.E. 1995 Fire and Wetland Management pp 1-9, in Cerulean, S.I. and Engstrom, R.T. (eds.) Fire in Weltands: A Management Perspective. Proceedings of the Tall Timbers Fire Ecology Conference, No. 19. Tall Timbers Research Station, Tallahassee, Florida, USA. Masipiquea, A.B., Persoon, G.A. and Snelder, D.J. 2000 The Use of fire in Northeastern Luzon (Philippines): Conflicting View of Local People, Scientists and Government Officials in Ellen, R., Parkes, P. and Bicker, A. (eds.),

PAGE 224

208 Indigenous Environmental Knowledge and its Transformations. Harwood Academic Publishers, The Netherlands. Mitsch, W.J. and Gosselink, J.G. 2000 Wetlands. John Wiley Inc, New York, USA. Mosepele, K. 2001 Description of the Okavango Delta Fishery. Fisheries Section, Ministry of Agriculture, Gaborone, Botswana. Murray-Hudson, M., Forrester, B. and Cherry, L. 1989 The Swamp Book: A View of the Okavango. Southern Book Publishers, Johannesburg, South Africa. Murray-Hudson, M., Parry, D., Moeletsi, B., Cassidy, L., and Murray, M. 1994 Natural Resource Utilisation: A Compilation of Documented Natural Resource Use in the Multiple Use Controlled Hunting Areas of the Okavango and Kwando Wildlife Management Areas. Tawana Land Board, Maun, and Department of Wildlife and National Parks, Gaborone, Botswana. Neefjes, K. 2000 Environments and Livelihoods: Strategies for Sustainability. Oxfam GB, Oxford, UK. Ngamiland District Council 1997 Ngamiland District Development Plan 5: 1997-2003. Ministry of Local Government, Lands and Housing. Government Printer, Gaborone, Botswana. Nicholson, S.E., Kim, J. and Ba, M.,B. 1997 The Mean Surface Water Balance over Africa and its Interannual Variability in Journal of Climate, Vol. 10, pp 2981-3002. Patterson, L. 1976. An Introduction to the Ecology and Zoo-geography of the Okavango Delta in Symposium of the Okavango Delta. Botswana Society, Gaborone, Botswana. Pyne, S.J. 1995 World Fire: The Culture of Fire on Earth. Henry Holt and Co., New York, USA. Roberts, S.J. 1995 Fire on the Galpag: Contemporary Aboriginal and Other Burning Patterns in Kakadu National Park, Northern Australia pp 31-38 in Cerulean, S.I. and Engstrom, R.T. (eds.) Fire in Weltands: A Management Perspective. Proceedings of the Tall Timbers Fire Ecology Conference, No. 19. Tall Timbers Research Station, Tallahassee, Florida, USA. Roy, D.P., Lewis, P.E. and Justice, C.O. 2002 Burned Area Mapping Using Multi-Temporal Moderate Spatial Resolution Data A Bi-directional Reflectance Model-based Expectation Approach in Remote Sensing of the Environment, Vol. 83, Issues 1-2, pp 263-386.

PAGE 225

209 Schapera, I. 1970 Tribal Innovators: Tswana Chiefs and Social Change 1795-1940. University of London, Athlone Press, London, UK. Scott Wilson Resource Consultants and The Environment and Development Group 2000 Environmental Impact Assessment of Veterinary Fences in Ngamiland: Vol 1 Strategic Environmental Assessment of the Veterinary Fences Policy in Ngamiland. Government of Botswana, DfID. Gaborone, Botswana. Spradley, J.P. 1979 The Ethnographic Interview. Harcourt Brace Jovanovich College Publishers. Orlando, Florida, USA. Stroppiana, D., Pinnock, S., Pereira, J.M.C. and Grgoire, J-M. 2002 Radiometreic analysis of SPOT-VEGETATION Images for Burnt Area Detection in Northern Australia in Remote Sensing of the Environment, Vol. 82, Issue 1, pp 21-37. Tlou, T. 1985 A History of Ngamiland 1750-1906: The Formation of an African State. Macmillan Botswana, Gaborone, Botswana. United Nations 2000 Map No. 4032, Okavango River Basin. Cartographic Section of the Department of Public Information of the United Nations for The Permanent Okavango Basin Comission (OKACOM), New York, USA. http://www.un.org/Depts/Cartographic/map/other/okavango.pdf. Site last visited March 2003. Valenti, M. 2000 Tracking Africas Inferno in Mechanical Engineering, Dec 2000. American Society of Mechanical Engineers, New York. http://www.memagazine.org/backissues/dec00/features/africas/africas.html. Site last visited February 2001. van Hoof, P.J.M., Kirkels, M.A.L.J., Riezebos, H.Th., Scheldorn, J.L.M. and de Wit, M.J.M. 1991 Ngamiland District CSDA (Western Part) Socio-economic Baseline Survey and Land Suitability Analysis. University of Utrecht, The Netherlands. van Hoof, P.J.M., Kirkels, M.A.L.J., Riezebos, H.Th., Scheldorn, J.L.M. and de Wit, M.J.M. 1993 Ngamiland District CSDA (Eastern Part) Socio-economic Baseline Survey and Land Suitability Analysis. University of Utrecht, The Netherlands. Wade, D., Ewel, J. and Hofstetter, R. 1980 Fire in South Florida Ecosystems. USDA Forest Service. Asheville, North Carolina. West, O. 1965 Fire in Vegetation and its Use in Pasture Management with Special Reference to Tropical and Subtropical Africa. Commonwealth Bureau of Pastures and Field Crops, Hurley, Berkshire, UK.

PAGE 226

210 World Commission on Environment and Development. 1987 Our Common Future. Oxford University Press, Oxford, UK.

PAGE 227

BIOGRAPHICAL SKETCH Lin Cassidy was born in South Africa. She gained her Bachelor of Arts degree at the University of Cape Town, with a double major in French and sociology. She then completed a 1-year honors degree in French literature, before moving to Maun, Botswana in 1987. After working in the wildlife-based eco-tourism industry in the Okavango Delta for several years, she attended the University of Zimbabwe, where she obtained a Bachelor of Science special honors degree in sociology. Lin returned to Botswana where she has been working as a freelance consultant in rural development, mainly on short-term projects for donor organizations, NGOs, or government departments. Increasingly she has focused her work on the community-based natural resources management sector, with a particular interest in the areas surrounding the Okavango. In 2001 she became a naturalized Botswana citizen. Lin came to the University of Florida to learn more about the ecological aspects of natural resources management, and to pursue the links between social and ecological systems. This thesis is part of her Master of Science degree in interdisciplinary ecology. She plans to complete her PhD in geography through the University of Floridas Land Use and Environmental Change Institute, before returning home to Botswana. 211


Permanent Link: http://ufdc.ufl.edu/UFE0000659/00001

Material Information

Title: Anthropogenic Burning in the Okavango Panhandle of Botswana: Livelihoods and Spatial Dimensions
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0000659:00001

Permanent Link: http://ufdc.ufl.edu/UFE0000659/00001

Material Information

Title: Anthropogenic Burning in the Okavango Panhandle of Botswana: Livelihoods and Spatial Dimensions
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0000659:00001


This item has the following downloads:


Full Text











ANTHROPOGENIC BURNING IN THE OKAVANGO PANHANDLE OF
BOTSWANA: LIVELIHOODS AND SPATIAL DIMENSIONS













By

LIN CASSIDY


A THESIS PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
MASTER OF SCIENCE

UNIVERSITY OF FLORIDA


2003




























Copyright 2003

by

Lin Cassidy















ACKNOWLEDGMENTS

I would like to thank my committee members (Dr Christy Gladwin, Food and

Resource Economics Department, Dr Michael Binford, Land Use and Environmental

Change Institute, Department of Geography; and Dr Mark Brown, Center for Wetlands,

Department of Environmental Engineering) for their support and guidance. Dean

Stephen Humphrey, College of Natural Resources and Environment and Dr Pete

Hildebrand, Food and Resources Economics Department gave me much encouragement.

Thanks are also due to Dr Jane Southworth, Geography Department and Dr Mickie

Swisher, Family, Youth and Community Science Department for assistance with data

interpretation. I am extremely grateful to Dr Helen-Jane Armstrong and the Map and

Imagery Library within the University of Florida's Marston Science Library for

purchasing the satellite images used in this study.

There are many people in Botswana whom I would like to acknowledge. Most

important are the people of Mogotho, Sekondomboro, Samochima and Nxamasere who

gave freely of their knowledge and time. I am grateful to Monty Montshiwa of Ecosurv

(Pty) Ltd for translating the questionnaire, and to Hannelore Bendsen of the Harry

Oppenheimer Okavango Research Centre for introductions, access to literature and

general support. I am indebted to my research assistants, Charlie John, Manyima

Manyima and Oteng Segakisi for their unflagging enthusiasm and meticulous data

collection.









I would like to express my gratitude to Steve Harpt and Bana ba Metsi School for

giving us a base on the east side of the panhandle, which made life in the study area a lot

richer and easier. I am also extremely grateful to Birgit and Reiner Kohler and family,

Robyn and Charles Sheldon, Eve and Hugh Murray-Hudson, and Jan and Eileen Drotsky

for their hospitality.

This thesis would not have been possible without funding from a Compton

Fellowship, a Tropical Conservation and Development Fellowship and a Tropical

Conservation and Development Field Research Grant. I greatly appreciate this support.
















TABLE OF CONTENTS
page

A C K N O W L E D G M E N T S ......... .................................................................................... iii

LIST OF TABLES ......... ..................... ........... ... .. ...... .............. viii

LIST OF FIGURES ......... ......................... ...... ........ ............ xi

ABSTRACT .............. .................. .......... .............. xiv

CHAPTER

1 IN TR OD U CTION ................................................ .. ....... .... .............. .

Statem ent of the Problem .................................................. ............................... 1
Research Objectives ...... ........ ..................... ...... ............ ........ 1
L literature R eview .................. .................. ... ............. .......... ...... .............. 2
Indigenous U se of Fire in W ildlands.................................... .......................... ........ 2
Fire in W wetlands ........................................................ ......... 4
Sustainable Livelihoods ........................................... ........ 5
B background and R research Setting........................................... ............................... 8
B o tsw an a ........................................................................ ................................. . 8
S tu d y A re a ......................................................................................... 1 0
F local V villages ......................................................................................................... 18
P lan of Study ........................................ 22

2 M E T H O D S .............................................................................2 4

Spatial Assessment ................................. ........................... .... ........ 24
D definition of Study A rea ....................................................................... 25
Location and Distribution of Key Vegetation Resources ................ ............ 25
Location and Distribution of Burnt Areas .............................................................. 26
Identification of Village Resource Areas ............................................... ............... 32
Calculation of Geographic Extent of Resource and Burn Areas ............................. 33
Socio-economic Assessment ..................... ............. ... ............. ...........33
Extent of Reliance on Key Natural Resources ............. ........... .............. 36
Variations in Reliance on Key Resources ....................... .......... ............. 40
Effect of Fire on Availability of Key Natural Resources...................... ............. 42
Conflict Surrounding Burning...................... .............. ..................... .............. 24




v









3 R E S U L T S ........................................................... ................ 4 7

Extent and Distribution of Fire in the Panhandle and Key Resource Areas .............. 47
K ey R sources A areas for Focal V illages............................................... ... ................. 51
Burning in the Key Resource Areas ................................ ......................... ....... 55
Tem poral D distribution of Fire .................................................................. ... 60
Discussion Theme: Local Knowledge of Fire Behavior ................. .................... 61
Reliance on the Wetlands Natural Resource Base .............................................. 62
Spatial Variations in Reliance on Key Resources.............. ............................ 70
W health Variations in Reliance on Key Resources................................................... 79
Gender Variations in Reliance on Key Resources .............................................. 81
Ethnic Variations in Reliance on Key Resources................................... .............. 85
Temporal Distribution of Resource Collection ...................................... ............... 93
Discussion Theme: Reasons for Temporal Variation in Collecting Plant
R e sou rces .................. ............. ...... ................... ......... .. 9 6
Effect of Fire on Access to Key Natural Resources......................... .............. 99
Effect of Fire on Grazing........................................ 100
Effect of Fire on Availability of Thatching Grass ............ ............. .......... .... 101
Effect of Fire on Availability of Reeds ................................................ .............. 102
Effect of Fire on Access to Palm Leaves, Water Lily Bulbs and Papyrus............. 103
Effect of Fire on Access to Fish ................................................... ................ 104
Effect of Fire on Access to W ildlife........ ....... ....................................... ....... 105
Effect of Fire on Livelihood Sustainability ......................................... ................ 106
Conflict Over Burning................................ ....... ............... .. .............. 124
Discussion Theme: Lack of Conflict Due to Common Purpose ........................... 124
Different Use of Panhandle Resources......................... .......................... ...... 125
T im in g o f F ires ..................................................................... ............... 12 7
Management and Consultation in the Use of Fire.................................................. 129

4 D ISCU SSION ..................................................................... ......... 133

Livelihoods A ssessm ent ..................................................................................... 133
V vulnerability Context ................................................ .. .. .... .. .......... .. 133
Livelihood A assets .................. ................................... ..... .............. 134
Policies, Institutions and Processes ................................... ................................... 138
Livelihood Strategies ..................................... .................. .. ........ .. 138
Livelihood Outcomes ................ .............................................. ............ .. 139
Conclusions and Recommendations............................................ 141
Extent and Distribution of Fire in the Panhandle and Key Resource Areas ........... 141
Reliance on the Wetlands Natural Resource Base .............................................. 142
Affect of Fire on Access to Key Natural Resources ............................................... 143
C on flict O v er B u rn in g ............................................................................... .............. 14 4
Recommendations for Further Study .............. ...... ....................................... 145
Recommendations for Improved Fire Management........................................... 145









APPENDIX

A HOUSEHOLD QUESTIONNAIRE .......................................................................148

B INFORMAL INTERVIEW TRANSCRIPTS ...................................................171

R eed/T hatch P erspectiv e................................................................. .................... 17 1
Livestock Perspective ........ .... ...................... .................. .. ..... .......... .. 184
Fishing Perspective .................. ..................................... .. ...... .... 187
Water Lily Perspective .......................................... 196
P alm L eaf P erspective............................................... .. ............ .. ............ 199
N on-U ser Perspective ........................................................................... .............. 202
"Wealthy" Man (owns Land Rover 110), Sekondomboro Village ................ ........... 202
K gotla M meetings ................. ....... ........................ ...... .... .......... 203
Kgotla M meeting Sekondomboro Village .............. ...... .......................................... 203
Kgotla M meeting Sam ochim a Village.................................... .......................... ........ 203
Kgotla M meeting Nxam asere Village.................................... ........................... ........ 204

L IST O F R EFER EN CE S ........................................................................... .............206

BIOGRAPHICAL SKETCH ............................................................. ............... 211
















LIST OF TABLES


Table page

1 Characteristics of focal villages ................................ ......... .................. 22

2 Contributions of spectral bands to principal components from subset of exposed
g ro u n d im a g e ............................................................................................................... 2 9

3 Pixel values for areas of low and high reflectance in exposed area subset ..................30

4 Population and sample sizes of focal villages..................................... ............... 34

5 Average sizes and prices for plant resources and fish ................................................37

6 Values of goods and flows used to create wealth index variable ................................39

7 Wetland resources included in natural resources value................... .................40

8 M measures of association by data type....................................... .......................... 41

9 Names of languages spoken by different ethnic groups .............................................41

10 Size of village resource ranges .............................................................................. 51

11 Extent of fire within village resource areas .............. ..... .........................56

12 Comparison of reliance on different resources.. ................................................... 63

13 Household reliance on different resources.......... ......... .........................64

14 Proportional contribution of wetland resources to households by village ...............70

15 Numbers of cattle owned by village. ........................................ ....................... 71

16 Thatching grass collected in 2001 by village........................................................... 72

17 Reeds collected in 2001 by village. ........................................ ........................ 73

18 Reed bundles sold in 2001 by village. .............................................. ............... 74

19 Households collecting palm leaves by village. .......................................................74









20 Bunches of palm leaves collected in 2001 by village................................................74

21 H households w ith basket-m akers. ............................................................................ 75

22 Source of palm leaves for basket-making........................................ ............... 76

23 Collection of water lily bulbs by village.................... ...... ...............77

24 Households usually fishing by village ....................................................................... 78

25 Fish caught in 2001 by village. ........................................ ......................................78

26 Fish sold in 2001 by village. ........................................... ........................................78

27 H household size by village. ...... ........................... ........................................80

28 Gender distribution of household heads. ........................................ ............... 82

29 Wealth index by gender of household head ............. ......................................82

30. Total household size by gender of household head. ............. .................................... 83

31 No. of household members 15 years and over by gender of household head ..............83

32 Bundles of thatching grass collected by household head...........................................84

33 Bundles of reeds collected by gender of household head..................... ................84

34 Proportional contribution of wetland resources to households in 2001 by ethnic
g ro u p ............................................................................................. 8 6

35 Livestock ownership and use of floodplain for grazing in 2001 by ethnicity. ............88

36 Bundles of thatching grass collected in 2001 by ethnicity. .............. ..................... 88

37 Collection of palm leaves by ethnic group. .............................................................89

38 Collection of water lily bulbs by ethnic group.................................. ............... 90

39 Involvement in fishing by ethnic group. ............................... .. ....................... 91

40 Seasonal Calendar of Flood and Collection of Main Resources...............................95

41 Extent of resources burnt within village resource areas in 2001 ..............................99

42 Effect of social characteristics on no. of adults in the household............................ 108

43 Effect of social characteristics on no. of wage jobs in household ...........................108

44 Regression table for factors influencing wealth in 2001 ........................ 110









45 Relative contribution of fire to range of resources accessed in 2001........................112

46 Relative contribution of fire to amount of fish caught in 2001............................... 114

47 Relative contribution of fire to amount of reeds collected in 2001 .......................... 116

48 Relative contribution of fire to amount of thatching grass collected in 2001............18

49 Rrelative contribution of fire to numbers of baskets made in 2001........................... 120

50 Relative effect of fire on proportional contribution of wetland resources to
livelihoods in 2001 ..................................................... ... .. .......... 122

51 Reasons believed for why people set fires experienced in 2001 .............................125

52 Resource affected by user held responsible for fire................................................126
















LIST OF FIGURES


Figure p

1 Sustainable livelihoods diagram .......................................................... .....................7

2 Map of Botswana showing location of study area ......................................................9

3 Map of the study area showing location of focal villages. ...........................................11

4 Map of average annual rainfall isohyets relative to the Okavango and its major
tributaries and basin. .......................... ......................................... .. ..... 12

5 Representative views of vegetation in the Okavango panhandle...............................14

6 Use of wetland resources in the Okavango panhandle. ............................................20

7 Savanna fire showing active fire front and burnt area............... ...............27

8 Signature plots of the 15-class unsupervised classification of the October scene
exposed area subset. ........................................................... .. ........ .... 31

9 Survey variables expected to influence access to wetland resources. .........................45

10 Stu dy area m ap ..................................................... ................ 4 8

11 Map showing main vegetation groups in vicinity of focal villages.............................49

12 M ap showing extent of burning in 2001. ......................................... ...............50

13 Map of grazing and wetland plant resource areas for each focal village...................52

14 Comparison of available wetland grazing for focal villages. .....................................53

15 Comparison of available plant resources for focal villages......................................54

16 Map showing graze resources burnt in the grazing areas of each focal village..........57

17 Proportion of grazing vegetation burnt in graze ranges.............................................58

18 Map showing papyrus /reed /thatch burnt in the plant collection areas of each focal
v illa g e .................................................................................................. . .5 9









19 Proportion of plant resources burnt in collection ranges. .........................................60

20 Proportional contribution of wetland resources to household livelihoods...................64

21 Frequency distribution of thatching grass bundles collected in 2001 ..........................66

22 Distribution of fish caught and fish sold in 2001.............................. ............... 69

23 Distribution of cattle ownership by village in 2001............................................. 71

24 Extent of water lily bulb collection by village in 2001..............................................76

25 Percentage distribution of wealth over households. ................ ..............................80

26 Distribution of ethnic groups by village. ........................................ ............... 85

27 Proportional contribution of wetland resources to household livelihood in 2001 by
ethnic group p ..........................................................................87

28 Quantities of water lily bulbs collected in 2001 by ethnic group.. ............................90

29 Numbers of fish caught and number sold in 2001 by ethnic group...........................92

30 Map showing change in open water between the 2001 wet and dry seasons ............94

31 Decision tree determining early or late collection of reeds and thatch......................98

32 Numbers of bundles of thatching grass collected by whether fire had been
experienced or not in 2001 ............... ................. ............................................. 101

33 Similarities in reports of fire in thatching grass and reeds in 2001..........................103

34 Numbers of fish caught categorized by fire experienced or not in 2001................. 105

35 Results of the Stage 1 multiple regression.............................. ............109

36 Results of the Stage 2 multiple regression........................ ..... .................11

37 Results of the first Stage 3 multiple regression ................................... ...............1. 13

38 Results of the second Stage 3 multiple regression...............................................115

39 Results of the third Stage 3 multiple regression .................................................1. 17

40 Results of the fourth Stage 3 multiple regression............................... ...............1.19

41 Results of the fifth Stage 3 multiple regression.. ......................... .................. 121

42 Results of the sixth Stage 3 multiple regression....................................................... 123









43. Proportion of 2001 fire events reported by month.................................................128

44. Differences in the role of fire in the livelihoods of rich and poor households.. .......140














Abstract of Thesis Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Master of Science

ANTHROPOGENIC BURNING IN THE OKAVANGO PANHANDLE OF
BOTSWANA: LIVELIHOODS AND SPATIAL DIMENSIONS

By

Lin Cassidy

May 2003

Chair: Christina H. Gladwin
Department: College of Natural Resources and Environment

This thesis examines how illegal, anthropogenic wildland fire affects people's

access to and use of resources in the wetlands of the panhandle region of Botswana's

Okavango Delta. The study focuses on two villages close to the main river channel, and

two near the shallower floodplains of the panhandle area. It addresses four specific

questions: What is the extent and distribution of fire in the Okavango panhandle in

general, and in key resource areas in particular, in a given year? How much do

households rely on wetland resources? How does fire affect the availability of these

natural resources? Is there any conflict among people over burning?

Two approaches to answering these questions are used. The first is a spatial

analysis based on the interpretation of two Landsat ETM satellite images. One scene is

from March 2001 at the end of the rainy season, which coincides with the peak of the

annual wetland flood. This provides information on the availability of wetland resources.

The other is from October 2001, at the end of the dry season, and provides the basis for









identifying burnt areas. The second part of the analysis is a livelihoods assessment based

on a quantitative household survey and informal theme-based interviews. Information on

the location, types, and quantities of resources used is examined in terms of household-

level socio-economic attributes as well as whether fire occurred in the resource areas or

not.

Results show that in 2001, fire covered about 5% of the study area as a whole.

More burning occurred in the resource areas of the villages near the floodplains than

those near the main river channel. Wealthier households, those from the dominant ethnic

groups, and those headed by men, tended to collect greater amounts of wetland resources.

However, the proportional contribution of these resources to their total livelihood was

significantly less than it was for poorer households with reduced socio-cultural status,

who had few alternative livelihood strategies. A quarter of all households experienced

fire in 2001; however fire appears overall to have a positive effect on the availability of

wetland resources. In addition, socio-economic factors affect access to the resources

more than fire does. There is little conflict due to burning because, although it is illegal,

most people see it as a way of improving the resource base. The limited conflict that

occurs is associated with poor timing, and the lack of warning that a fire is to be set. This

in turn is due to the secretiveness that the illegality of the practice induces.














CHAPTER 1
INTRODUCTION

Statement of the Problem

Every year, Africa's subequatorial savannas experience huge fires that consume a

total of approximately 1.3 million square kilometers (Valenti 2000). Fires in Botswana's

savannas and wetlands are no exception. Uncontrollable because of their remoteness and

intensity, fires often range for hundreds of kilometers, until they burn themselves out.

Nearly all these fires are started by people, just as they have been for millennia (Pyne

1995). The vast size of the fires suggests extensive destruction, and certainly every year

they consume resources on which many people depend. Aside from the potential

environmental impacts, this apparent contradiction raises several questions about the

social dimensions of anthropogenic fires. If fires really were a threat to the people's use

of natural resources, why would they set them? How do people view the role of fire in

their livelihoods?

Research Objectives

Specifically, this thesis examines how fire affects people's access to and use of

resources in the wetlands of the panhandle region of Botswana's Okavango Delta. In

this study the focus is narrowed to the Okavango panhandle because of the greater

concentration of people and natural resources there. It provides a bounded study area,

allowing fire management issues to stand out.

Until now, reports on the extent of burning have been mainly anecdotal and

speculative, with some estimates of around 75% of the total area of the Okavango Delta









being burnt each year (Patterson 1976). While it seems clear that large areas of the

Okavango panhandle, like the rest of the delta and other areas in Botswana, are burnt

every year, little information exists on the frequency, extent, or distribution of burning.

No data have been recorded about how fire affects people's access to natural resources in

the panhandle. In attempting to fill some of the information gaps, this thesis pursues the

following questions:

* What is the extent and distribution of fire in the Okavango panhandle in general, and
in key resource areas in particular, in a given year?

* How much do households rely on wetland resources?

* How does fire affect the availability of these natural resources?

* Is there any conflict among people over burning?

Literature Review

Indigenous Use of Fire in Wildlands

Fires in wildlands are inherently neither good nor bad. The desirability of fire is a

human evaluation of how it influences the environment's ability to meet the varied needs

of its inhabitants. People have been purposefully manipulating their environment with

fire for as long as they have been able to control it (Pyne 1995). In many developing

countries people continue to use fire to change their surroundings. This often leads to

conflict between the local population concerned with their own immediate needs and the

authorities who represent a constituency and who see the lack of control as a threat to the

environment and people's safety. And yet, as is discussed below, research shows that

local people have a clear understanding of fire behavior and how to use it for specific

purposes.









A study of Aboriginal communities living in the Northern Territory of Australia

documents the burning patterns caused by indigenous natural resource users in Kakadu

National Park (Roberts 1995). This study showed the reasons why people burned, their

methods for setting fire, and how the fires spread. Respondents showed clear preferences

for smaller patchier fires over large intense burs, and identified times of year when

burning would have less impact on the regeneration of plant species they used. Park

rangers were able to learn from the Aboriginal knowledge of the effects of burning on the

environment, and of the effect of season on burning.

Similarly, in north-eastern Luzon, in the Philippines, fire is used by different local

residents, such as farmers, ranchers and grass collectors, in ways that achieve specific

objectives for given land use practices (Masipiquefia et al. 2000). While they see fire as

an important tool, they understand the implications of resource loss. Again, knowledge

of the best season for certain types of burn was shown. Because optimal burning times

vary for different land uses, conflicts arise because the fires spread beyond the specific

resource area. An important addition to the many uses of fire that is noted here is that of

a weapon of injustice, jealousy or revenge (ibid.).

In sub-Saharan Africa, fire has been attributed to very specific aspects of pasture

management (West 1965). In Botswana, too, oral traditions show that tribal chiefs used

to regulate both the timing and location of veld fires. For example, fires could not be set

in the same place in consecutive years. Burning could also not be done until late winter,

after the thatching grass species had dropped their seeds, so ensuring the availability of

grass the following year. However, under pressure from white settlers accustomed to

European management practices, the colonial government convinced the chiefs that









burning was destructive. By the 1930s it was considered an offense to start a fire, or to

leave one burning (Schapera 1970). This restriction became further entrenched after

independence, with the promulgation of the Herbage Preservation (Prevention of Fires)

Act in 1978 (Government of Botswana 1978). This law, combined with the loss of power

of the chiefs under the post-independence centralized government, has led to the setting

of fires being done in secret and with no control.

The Okavango panhandle has been inhabited for thousands of years, sparsely at

first by the nomadic San peoples, with more intensive settlement occurring since the mid-

18th century with the influx of different Bantu peoples, such as the Hambukushu, Bayei,

Xereku, Batswana and Bakgalagadi (Afriye 1976, Tlou 1985). It is highly probable that

people have been setting fires in the Okavango for as long as they have lived there.

However, the short-lived nature of the Delta's wetland vegetation and the frequent

reworking of the sediments mean that there is very little historic record.

Fire in Wetlands

Fires are an important ecological component of wetlands. Most wetlands show

fluctuations in their moisture regimes, and where conditions support fire, marshy

wetlands (into which category the Okavango would fit) can burn frequently (Lugo 1995).

Experimental studies in wetlands in Florida show that wet periods tend to have a greater

impact on plant composition and the availability of nutrients than does fire

(Environmental Science and Engineering, Inc. 1994). The effects of fire on small

mammal and bird populations, and on plant biomass and composition, in Florida

wetlands tend to be short-term (less than 6 months), unless conditions are dry enough so

that the ground burns (Wade et al. 1980).









The number of small animals killed in Everglades sawgrass fires tends to be much

higher in unusually dry years. However, fire appears to benefit some species through

removing the dense accumulation of vegetation (ibid.). In the Kissimmee River marsh,

research showed that the number of both species and individuals were greater on burned

than on unburned plots. In addition, when fire was tied to changes in water levels, the

production of fish and macroinvertebrates increased (ibid.).

In the Okavango, research on the effects of fire has been limited to peat fires

(Ellery et al. 1989, Gumbricht et al. 2001). This is partly because by burning the soil,

peat fires bring about more permanent changes than do surface fires, and partly because

these persist for longer periods, making their study is more feasible. Peat fires are

associated with long-term reduction in flow of river channels, and resultant drying out of

peat soils. In comparison to surface fires, peat fires develop slowly, may persist for

several years, and are often not visible at the surface (Gumbricht et al., op. cit.). While

the University of Munich has recently begun to study the effect of fire on vegetation in

the seasonal areas of the Okavango Delta, no links have yet been made to how fire affects

the livelihoods of the people who rely on the Okavango for many resources.

Sustainable Livelihoods

The concept of sustainable livelihoods arose as part of the sustainable development

debate linking poverty and environmental degradation (World Commission on

Environment and Development 1987). There was a growing awareness of the need to

fully capture the perspective of the primary target of development objectives: the rural

poor. It was felt that to be more effective, development approaches needed to become

more people-centered. Definitions of poverty became broader and more

multi-disciplinary. The concept of sustainable livelihoods encompasses and links the









notions of capability, equity and sustainability. To clarify how these notions are linked,

a working definition was proposed as follows: "A livelihood comprises the capabilities,

assets (stores, resources, claims and access) and activities required for a means of living;

a livelihood is sustainable which can cope with and recover from stress and shocks,

maintain or enhance its capabilities and assets, and provide sustainable livelihood

opportunities for the next generation ... ." (Chambers and Conway 1992, pp 7-8). The

implication of this is that livelihoods should be both environmentally and socio-

economically sustainable.

The concept of sustainable livelihoods has been taken up as a practical tool by

various international organizations, which have refined the definition and prepared

different versions of an analytical framework (Carney et al. 1999, Neefjes 2000).

However, all are similar and contain the same three basic elements:

* An outcome, which is the sustainable livelihood,
* Strategies that rural people select to achieve this outcome, and
* The environmental and social context in which the livelihoods are located.
For most of the approaches two additional elements are broken out from the context

to be given specific emphasis: institutional processes and livelihood resources are

highlighted (Carney et al. 1999). The UK's Department for International Development's

(DflD) Sustainable Livelihoods Framework is shown diagrammatically in Figure 1. It

shows the vulnerability context on the left that is, the macro-level environment within

which people live, and over which they have little control. This may change suddenly,

presenting stresses or shocks to household attempting to survive at the micro level.

The diagram also shows the assets available to rural people. The pentagon should

be seen as flexible in shape, stretching out to those points of capital that are in greater










SUSTAINABLE LIVELIHOODS
FRAMEWORK


LIVELIHOOD
LIVELIHOOD ASSETS POLICIES, OUTCOMES
INSTITUTIONS More income
VULNERABILITY & PROCESSES Increased well-

Shocks Government LIVELIHOOD Reduced
Trends SN STRATEGIES vulnerability
Seasonality Private Laws Improved food
Sector Culture security
SPolicies More sustainable
N wInfluence Institutions use of NR base
& Access


Figure 1. Sustainable livelihoods diagram. Redrawn from DfID's diagram. (DflD 1999)
http ://www.livelihoods.org/info/guidance_sheetsrtfs.



abundance, and shrinking inwards where a resource is in short supply (DflD 1999). The

transforming structures and processes represent the administrative and legal context.

These include informal and formal organizations, laws and policies. They can operate at

all levels, from individuals to district level to the international arena (ibid.). For example,

it is as important to understand who in the household controls the allocation of income as

it is to understand the effect of the land use policies at individual, household and

community levels.

The strategies are the ways in which people meet or try to meet their goal of a

sustainable livelihood. Strategies will vary according to social factors such as gender,

class, and ethnicity, as well as according to spatial location. For example, the San have

traditionally relied on hunting game while groups of Bantu origin have focused on

keeping livestock. It is important to see how not only the types, but also the number, of









livelihood strategies available to different people vary. People with more livelihood

strategies may be more adaptable, and therefore less vulnerable to external shocks and

stresses. Different strategies may also lead to competition and conflict.

Background and Research Setting

Botswana

Botswana is a land-locked, semi-arid country in southern Africa, lying between 180

and 270 South and between 200 and 290 East. Much of the country comprises semi-arid

savanna, with average annual rainfall varying from about 650mm in the northeast to 250

mm in the southwest. Climate characteristics specific to the study area are discussed in

further detail below. It is sparsely settled, mainly because of its low rainfall and lack of

surface water. The country has had a stable, multi-party democracy since its peaceful

independence from the British Protectorate in 1966. Because of this it has attracted a lot

of international NGO and donor support, particularly during the 1980's and 1990's.

Diamonds provide about 80 percent of the country's foreign earnings, and form the

backbone of the economy, which as a result is highly centralized. Income from diamonds

is used to subsidize rural livelihoods that to a large extent are based on subsistence

agriculture. The traditional significance accorded to livestock ownership has supported a

strong livestock industry, subsidized heavily by Government and marketed to the

European Union through a preferential trade agreement. Most rural households grow

some crops, but in much of the country in many years the yields are insufficient to meet

annual household food requirements, and crop production is in almost all cases sustained

only with central government assistance. Households grow maize, sorghum and

millet,with melons and beans as intercrops.
































Kweneng


Kgalagadi
S i aborone
SI- SSouthern uth east
SOUTH AFRICA



SI Districts
.,"Major river courses
\/Major roads

0 200 400 600 800 1000 Kilometers



Figure 2. Map of Botswana showing location of study area.









In striking contrast to most of Africa, there are no traditional markets in any of the

settlements. This is primarily because there is very little specialization in livelihood

strategies, with most households producing the same goods. Agricultural productivity is

also low, so there is very little or no surplus to sell. All agricultural and natural produce

is typically obtained for subsistence consumption. Some informal trade takes place

between households, but this is typically on an adhoc basis. Any other goods that a

household needs to purchase are obtained from retail shops. Nearly all rural households

rely directly on natural resources as part of their livelihoods. This is most noticeable in

Ngamiland District, where the Okavango Delta is located.

The Okavango system is one of the most important economic and ecological

features of the country. As a Ramsar Convention Site wetland on tribal land, proper

management that includes resident communities is vital to its long-term sustainability. As

a permanent source of fresh water, it supports an abundance of plant and animal life that

contrasts starkly with the rest of the semi-arid northwestern region of the country.

Although the fresh water and concentration of resources of the Okavango Delta could

accommodate far greater settlement, the presence of vector-borne diseases that affect

both humans and livestock has until recently kept population levels low (Tlou, op. cit.).

Study Area

For the purposes of this research, the study area is defined as the wetland system of

the Okavango panhandle. It comprises the area reaching from the Namibian border in the

northwest to a point some 2.5 kilometers southeast of the village of Seronga a total

straight-line distance of approximately 92 km. About 10 km south of this point the

panhandle fans out to form the delta. In this study, the term panhandle will generally be

used to refer to the study area.







11




580000 590000 600000 610000 620000 630000 640000 650000

7980000 7980000





S-oi

7960000 790000




7960000 -Sa4o '- 7950000
























Scale
7950000 1 2 795000
7940000 -7940000

















7930000 Bright green indicates 7930000d
7920000 :a 7920000










580000 590000 600000 610000 620000 630000 640000 650000

season, vegetative growth is concentrated along the main channels of permanent water.
10 0 1 0 30 40 50 60
Projection: UT Coordinate System one 34 South
.. = 4 Datum: WCS.4










season, vegetative growth is concentrated along the main channels of permanent water.
Pink areas lining the edge of the panhandle are sandy areas associated with villages,
heavy grazing and bare fields at the end of the dry season.









Biogeophysical characteristics of the Okavango panhandle

The Okavango panhandle is a broad trough formed by parallel geological fault lines

between 12 and 15 km apart. These faults lie at right angles to the main fault lines that

limit the distal end of the delta itself (Hutchins et al. 1976). Since this is the southern

hemisphere, winters run from June to August. In Botswana they are dry and cool. The

hot and moist summers are from October to April. Rain falls from November to March,

peaking in January/February. The Okavango panhandle has a fairly arid climate, with an

average annual rainfall of 560 mm. The rainy season coincides with the arrival of the


Figure 4. Map of average annual rainfall isohyets relative to the Okavango and its major
tributaries and basin. Figures are given in millimeters. The 500 mm isohyet passes
through the Okavango delta just south of the panhandle. (Redrawn from Nicholson et al.
1997 and United Nations 2000.)









annual flood wave fed by an average of 1500 mm/annum of rainfall in the Planalto

Central highlands of Angola (Figure 4).

The Okavango river carries an average of 11 billion m3 of water a year. The river

is comprised of a main channel, which ranges from 75 to 250 m wide, and several side

channels and lagoons. With the arrival of the floodwaters the river fills up the extensive

floodplains that stretch between the fault lines. During the dry winter months the pulse of

the flood moves down through the Delta, and the water level in the panhandle drops,

drying out the outer reaches of the floodplains. Not only does the river loop tightly as it

flows southwards, but it also meanders from one side of the panhandle to the other. This

changes the water and vegetation characteristics close to the banks, with important

consequences for people's access to various resources, as is discussed below.

There is a distinct change in vegetation moving outwards from the channel margins

through the back swamp vegetation to the seasonal floodplains (Ellery et al. 2000).

Dense stands of papyrus (Cyperus papyrus) flank the river, while reeds (Phragmites

australis) and taller (thatching) grass (Miscanthusjunceus) grow on slightly elevated, but

flooded, land. Behind the papyrus, the floodplains are mainly covered in shorter aquatic

grasses and sedges (Cyperus articulatus, C. denudata, Cladium mariscus, Panicum

repens, inter alia). Small islands have formed in the panhandle where river channels

have moved, leaving perched ridges of sand. These are covered primarily with phoenix

palms (Phoenix reclinata).

The floodplains are very important breeding grounds for native tilapia fish (mainly

Oreochromis and Tilapia spp.). This is in part due to the slower flow of the water,

greater cover, and decreased vulnerability to predation by tiger fish (Hydrocynus



















I k4.
5': 9E~


A




























B
Figure 5. Representative views of vegetation in the Okavango panhandle. A) A
permanent river channel, showing reeds in the foreground, and papyrus on the far bank.
B) A seasonally wet floodplain. C) A seasonally flooded area where water lily bulbs are
collected and where basket fishing is done. D) Overlooking an area of dense aquatic
vegetation, taken from the sandveld bank. All pictures taken by L. Cassidy (June/July
2002).


L II
j I

1 .I'

























































Figure 5. continued.









vittatus). Mainly, however, the much higher productivity and ecological complexity

provide more food for both adults and young (pers. comm., Roger Bills, J.L.B Smith

Institute of Ichthyology, Rhodes University). Fish, reeds and thatching grass are key

resources on which rural households rely both for subsistence use and small-scale trade

(Mosepele 2001, Murray-Hudson et al. 1994). The panhandle floodplains are also

important for the grazing of livestock and, to a lesser extent, wildlife such as sitatunga

(Tragelaphus spekei), lechwe (Kobus leche) and hippopotamus (Hippopotamus

amphibious .

Socio-economic characteristics of the Okavango panhandle

Land tenure in the study area, as with most of Botswana, is communal. With the

centralization of government after independence, the authority of the chiefs has been

weakened, and traditional rules of access and obligations have been eroded. Use of

resources takes place under open access. Because population densities are low, the

negative consequences of this are not obvious, and the environmental "tragedy of the

commons" (Hardin 1968) is extremely localized around the larger settlements.1 Within

the wetlands, the ability for individuals to maximize use of resources is further limited by

the inaccessibility due to the deep channels and dense vegetation.

There are approximately 19,000 people living in the study area (Central Statistics

Office 2002). By far the majority are of Hambukushu origin. The Hambukushu are

traditionally matrilineal (Larson 1980), and without the rigorous social structure of the

Tswana that comprise 80% of the national population. Other ethnic groups present



1 Around larger villages in the study area, firewood and trees suitable for building poles may have been
used up within a radius of 2 km. Each year, grazing within a similar area results in all grass cover being
removed. For smaller villages, such as the four focused on in this study, the radius is probably less than 1
km.









include the Bayei, Bugakhwe and Xanikhwe San, Bakgalagadi, Batawana and Ovaherero

(Afriye 1976, van Hoof et al. 1991 and 1993). For all groups, the household is the

primary economic unit. Households tend to comprise three generations, and are generally

large. As with most of the other rural parts of Botswana, the proportion of women is

relatively high due to out-migration of men in search of wage labor (van Hoof et al. 1991

and 1993). Class, as a definition of how a household meets its livelihood needs, is only

beginning to emerge. That is, it is not possible to identify separate social classes based

on a single economic activity. Most households pursue several livelihood strategies

(Scott Wilson 2000). Wealth is usually determined by how many sources of income a

household has, rather than which type. Household size to some extent defines labor

availability, and the ability to diversify livelihood strategies (Cassidy 1997). There are

very few wage employment opportunities in the study area, and most households receive

cash in the form of remittances from absent household members. For this reason,

reliance on the natural resource base is still high.

The reason households need to maintain diverse livelihood strategies is to provide a

buffer against the unpredictable conditions under which they live. However, the recent

(1995/6) slaughter of all cattle in Ngamiland District to eradicate an outbreak of cattle

lung disease has caused irreversible economic dislocations (Fidzani et al. 1999). Six

years later, the district herd is still less than half its 1995 size (Scott Wilson, op. cit.), and

it is only the larger herd-owners who have been able to restock, which has reinforced

wealth disparities. Not only is cattle ownership a source of wealth in itself, but it is also

critical to the ability of households to plough fields for crops. As a result, the

proportional contribution of natural resources, particularly to poorer households, has









increased. Much of the increased resource use is within the panhandle, where important

foods and building materials are found.

Key natural resources from the panhandle

Fish are a large part of people's diets. Some 65% of the population depend on

fishing (Fidzani et al., op. cit.). House walls are constructed of reeds, then plastered with

mud. Outdoor enclosures are also built of reeds, while roofs are made with thatching

grass (Larson, op. cit.). Papyrus is used to make sleeping mats and baskets. Leaves from

young mokola palms (Hyphaene ventricosa) are used for making baskets that have a

range of uses. Water lily bulbs (Nymphaea capensis) and fruit from a variety of riparian

trees are collected and eaten (Murray-Hudson et al. 1994).

People's access to, and extent of use of, these resources are not determined by

socio-economic factors alone. Distance and accessibility are limiting factors, and may

determine whether people can collect for themselves, or may need to buy from others, or

even do without the resource altogether. Anecdotal evidence suggests that collection of

resources usually takes place within a regularly used collection area that reflects these

limiting factors. Similarly, grazing of livestock within the panhandle occurs according

to the availability of suitable fodder species and the maximum herding distance.

Focal Villages

Twelve villages are situated along the edge of the study area. Between these, cattle

posts and field areas are scattered. None of the settlements are within the panhandle

itself, they are all up on the sandveld, on the edge of the panhandle. Settlement is linear:

the roads run parallel to the panhandle, and the villages are strung out along the roads.

Only four villages were considered for this study: Mogotho, Sekondomboro, Samochima

and Nxamasere. These were selected on the basis of their location, and to some extent on









whether there was an indication that there had been fire in their collection and grazing

areas in 2001. Two villages represent each side of the panhandle. Two are "river"

villages that have the main channel pass close to their bank, while the other two are

"floodplain" villages. Maps show that Nxamasere is a relatively drier area. Its

floodplains contain several larger islands, and floods last a shorter period. The location

of the settlements is shown in Figure 3. Basic demographic and development information

is summarized in Table 1.

Remoteness increases for the villages in the following order: Nxamasere,

Samochima, Sekondomboro, Mogotho. The nearest major town and district headquarters

is Maun, where alternative food and building supplies, as well as wage employment, can

be obtained. By vehicle it is 3.5 hours (350 km) from Nxamasere, 4 hours (370 km) from

Samochima, 6 hours (420km) from Sekondomboro and 6.5 hours (450 km) from

Mogotho. However, by far the majority of people do not have vehicles, and traveling

times on foot or by donkey cart are considerably longer. In addition, road access to the

east bank villages of Mogotho and Sekondomboro is only possible after traveling north

and crossing with the river ferry 5 km south of the Namibian border, and 4x4 vehicles are

needed. This makes these two villages far more remote than distance alone suggests.

Remoteness often affects settlement size, while settlement size is a criterion for levels of

infrastructural development (Ngamiland District Council 1997). Remoteness is also

believed to play a part in determining extent of reliance on natural resources, because of

the lack of cash income opportunities and resource alternatives.






















































B
Figure 6. Use of wetland resources in the Okavango panhandle. A) Thatching grass
stored outside a reed enclosure. B) Hambukushu couple floating thatching grass down
the main channel on a papyrus raft. C) Cattle being swum across the main river. D) Girls
going basket-fishing. Picture A) taken by L. Cassidy, B) by R. Forrester, and C) and D)
by Love Botswana (www.lovebotswana.org).

























































Figure 6. continued.









Table 1. Characteristics of focal villages.


Village
Panhandle side
Ecosystem type
Population a
Sex ratio a
(proportion male)
Approx no. of
households b
Road surface to
village '
Road surface within
village
School

Health facilities




Water supplies


Postal facilities




Communications


a. Central Statistics


Mogotho
East
River
557
0.44


Sekondomboro
East
Floodplain
655
0.45


100

Dirt

Dirt


1 primary 1 primary


Health post, 1
nurse, 1
welfare

Reticulated to
communal
standpipes


Health post, 1
nurse, 1
welfare

Reticulated to
communal
standpipes


via bag sent via bag sent to
to neighboring
neighboring village
village
radio at radio at health
health post post
officee 2002


Samochima
West
River
847
0.46


130

Tar

Dirt


Nxamasere
West
Floodplain
1328
0.43


200

Tar

Dirt


none

Health post, 1
nurse, 1 welfare


none, collected
from river 1 km
away
via post office at
neighboring
village

radio at health post


1 primary, day
care facilities
Clinic, 1
matron, 1
nurse, 1
welfare
Reticulated to
communal
standpipes
post office




public
telephones


b. based on an average of 6.5 people per household (Cassidy 1997)
c. Ngamiland District Council 1997.


Plan of Study

In examining how fire affects resources that play a key role in people's livelihoods

in the panhandle, it was necessary, firstly, to identify how much of key resources in the

focal villages' collection areas were burnt. Secondly, the ways in which people evaluated


(









the effect of this burning on their access to the resources needed to be determined. To

accomplish these tasks, two approaches were used.

The first was a geographic, or spatial, assessment of the location of fires, resources

and collections areas that people in the focal villages used. This was done by analyzing

satellite imagery, and by collecting geographic coordinates of resource types and

interviewing people in the field. Once the areas of interest had been defined, it was

possible to measure these using spatial modeling software.

The second approach was to ask the people living in the area about their

livelihoods, including how much they collected of each resource, and what their opinions

about the effects of fire and other aspects of burning were. Information from a

quantitative household survey was encoded and analyzed. Qualitative themes were

developed from informal interviews.

In order to be able to provide a quantified assessment, and to allow future

comparisons, it was necessary to define spatial and temporal boundaries to the study. To

this end, a defined area was clipped from the satellite imagery to reflect only the wetland

areas of the panhandle. Analysis was restricted to a one-year timeframe throughout the

research.














CHAPTER 2
METHODS

Primary data for this study were obtained through the analysis of satellite imagery

and from formal and informal interviews conducted in the study area.

Spatial Assessment

The spatial assessment is based largely on the interpretation of two Landsat

Enhanced Thematic Mapper (ETM) satellite scenes, from Path 175, Row 073. Landsat

ETM has a spatial resolution, or ground cell size, of 28.5 x 28.5 meters. These cells are

small enough to allow differentiation of features such as islands, river channels, clusters

of dense vegetation, and burn scars. The first scene was taken at the end of the rainy

season, and the peak of the flood, on 28 March 2001. The second was taken on 6

October 2001, at the end of the dry season. This date should capture most of the dry

season's burn scars, being late in the dry season, while still being early enough that

regrowth would have been minimal. Both scenes have less than 1 % cloud cover.

Image rectification, interpretation, analysis and map creation were done using

ERDAS Imagine version 8.5 software. Initial subsets focusing on the area immediately

surrounding the panhandle were created for ease of display. The subset of the October

image was then rectified to the March image with a total RMS error of 0.024 using 40

ground control points.









Definition of Study Area

The study area was created by performing a 2-class ISODATA1 unsupervised

classification of the March (wet season) initial subset. This was done by setting

clustering to initialize from statistics along the diagonal axis, and with standard

deviations of 1. Iterations were set to 20, with a convergence threshold of 0.95.

This separated wet from dry areas. The dry area was recorded to 0, leaving a single value

map layer of the wetland. The study area was consolidated by filling any islands that

showed up as dry areas within the boundaries, and by trimming any wet areas that had

been included from the dry land. This map layer was used to further subset both the

March and October images. All further analysis was performed on these final subsets,

hereafter referred to as the March scene or the October scene. This area is shown at the

beginning of the results chapter.

Location and Distribution of Key Vegetation Resources

A 15-class ISODATA unsupervised classification (parameters as above) was

carried out on the March image. In assigning vegetation categories to the classes,

information from ground control points and photographs taken in the field were used. In

the field it was clear that most of the vegetation types were mixed. There were few

accessible places where one of the dominant species occupied a full ground cell, let alone

the 100 m x 100 m that would allow one to be sure of having a true signal from the center

cell. However, by examining spectral signatures, it was possible to separate areas of

dense aquatic vegetation with high biomass, which comprised reeds, papyrus and

thatching grass the three wetland plant resources of greatest interest. These three


1 ISODATA stands for Iterative Self-Organizing Data Analysis Technique. This algorithm makes multiple
passes over the image, and after initial parameters are set by the user, organizes the clusters itself (Jensen
1996).









resources were therefore treated as one category, justifiable because respondents typically

obtained these resources from the same location. By comparing photographs and GPS

readings to coordinates on the classified image, it became clear that six of the classes

generated could be included in the dense aquatic vegetation category. This category is

considered as "plant resources". The six plant resources classes were recorded and

combined in a new single value map layer.

Areas of short grasses and sedges were also discernible, and verified by informants

as being important for grazing. Based on their spectral signatures, four classes were

combined to define vegetation areas suitable for grazing, even if these were in fact too

inaccessible to be used as such. A recorded map layer containing only grazing areas was

generated.

Location and Distribution of Burnt Areas

Burn scars in marshy wetlands are ephemeral, and are masked by the subsequent

flood and plant regrowth. By the time of the visit to the study area eight months after the

October scene was acquired, it was not possible to obtain any ground data to verify the

interpretations of the satellite image. For this reason, more rigorous analysis of the burn

areas as identified in the October scene was necessary.

The first step taken was a visual analysis of the October scene. In a MODIS-based

study of southern Africa, the 1.24 micrometer ([tm) band gave the best discrimination

between burned and unburned areas, followed by 0.86 itm, and then 1.64 |tm (Roy et al.

2002). This corresponds roughly to bands 5 (MIR1) and 4 (NIR). These two bands were

displayed with MIR2 to show the best contrast of burned areas in the October scene. Fire












scars, with active fire fronts or with distinctive shapes, on the savanna surrounding the


panhandle gave verification to this interpretation.


674000 675000 676000 677000 678000

-B'*^ 1. -


674000 675000 676000 677000


679000 680000 881000 682000 683000


OW


7912000


7911000


7910000


7909000


7908000


7907000


7906000


7905OOO


7904000


7903000


Scale
-i


Kilometers


Landsat ETM Image 6 October 2001
red, green. blue- 7. 4, 3


Projection: UTM Coordinate System Zone 34 South
Spheroid; GRS80
Datur: WGS84


Figure 7. Savanna fire showing active fire front and burnt area. This fire lies on the edge
of the wetland (bright green area), and, together with other similar burnt areas, was used
to provide verification of burnt areas within the panhandle. Smoke can also be seen north
of the flame pixels (red) in the upper right part of the image.


678000 679000 680000 681000 682000 683000


7912000


7911000


7910000


7909000


7908000


7907000


7906000


7905000


7904000


7903000









Figure 7 shows an active fire (in red, where cells have saturated pixel values1 (255)

in both MIR2 and thermal bands) and a fire scar behind it, about 20 km east of the study

area.

The next stage was to conduct a stage-by-stage hierarchical classification,

essentially as a process of eliminating known features to allow wider separation among

the remaining features. This was conducted as follows:

1. A 15-class ISODATA unsupervised classification (parameters as above) of the
October scene was done to allow for the creation of masks to remove obvious or
known features.

2. Open water (extremely low signatures in bands 2, 3, and particularly 4 and 5)
covered 2 classes, and this was recorded as a map layer and applied as a mask to the
October scene, creating a new image with no open water.

3. Next, the "no open water" image was classified into a new map layer using a 5-
class ISODATA unsupervised classification.

4. The classes numbered land 2 were identified as corresponding to healthy
vegetation, by three methods:

Firstly, the swipe utility was used over a viewer display of the image in r,g,b =
5,4,3 format. Classes 1 and 2 consistently covered the green (indicating high
reflecting in the NIR band) pixels. Nearly all uncovered pixels were shades of
pink, red and magenta.

Secondly, classes land 2 were swiped over a view display of a transformed NDVI
image, and were shown to correspond to areas of healthy growth.

Finally the spectral signatures generated for the 3-class map layer showed that
both Class 1 and Class 2 were still fairly wet areas, with low values in bands 2
and 3, and rising in bands 4 and 5, with band 7 dropping extremely low. In
contrast, Classes 3, 4 and 5 dropped sharply in the NIR band, and had extremely
high values in band 5 and high values in band 7.

5. Classes 1 and 2 were then recorded as a map layer and applied as a mask to the "no
open water" image.




1 In 8-bit data, pixel values range from 0-255, with 0 representing total absorbance, and 255 the maximum
measurable by the sensor.









6. Having now removed all open water and all healthy vegetation, by a process of
elimination all remaining areas would have to be senescent vegetation, or exposed
sand or soil due to burning or grazing.

7. A principal components analysis (PCA) was then carried out on this final masked "no
water, no healthy vegetation" scene. The first component accounted for 99.2% of the
variance in the scene. The major contribution to this band came from the thermal,
MIR1, and blue bands. The second component accounted for 0.62% of the variance,
while the third component accounted for 0.11%. The contributions of the various
spectral bands to each of the first 3 principal components are shown in Table 2. When
considering all these bands together, most of the variance of the non-vegetated or
sparsely vegetated areas is explained by the thermal, MIR, and NIR bands, as
suggested for the initial visual analysis.




Table 2. Contributions of spectral bands to principal components from subset of exposed
ground image.

PC 1 PC 2 PC 3
Band 1 (blue) 0.3122 -0.2045 -0.1214
Band 2 (green) 0.2593 -0.0959 -0.2605
Band 3 (red) 0.2800 0.0797 -0.3247
Band 4 (NIR) 0.2450 -0.2335 -0.6616
Band 5 (MIR1) 0.4424 0.5127 -0.1413
Band 6 (thermal) 0.6414 -0.4610 0.5493
Band 7 (MIR2) 0.2996 0.6424 0.2292

While it is possible that all the non-vegetated or sparsely vegetated area could be
burnt, the wetland conditions, with fire burning on vegetation on top of water, or
across areas where plants can start resprouting within days, make it difficult for the
sensor to define it as such (Stroppiana et al. 2002). However, charcoal and ash
deposits would suggest that burnt areas reflect less in NIR and MIR bands, and emit
more in the thermal band than sand with sparse vegetation. Conversely, sandy areas
with sparse vegetation would have higher reflectance in red, NIR and MIR bands, and
less so than burnt areas in the thermal band. Overall, therefore, burnt areas are likely
to appear darker than grazed or otherwise sparsely vegetated areas in the satellite
image.

8. Signature changes across the spectral bands for various sites on the exposed area "no
open water, no healthy vegetation" subset were examined. Consistent distinctions
could be made between two types of area, which corresponded roughly to whether









they appeared to have low or high reflectance. The variations in pixel values for the
two types of area are shown in Table 3.




Table 3. Pixel values for areas of low and high reflectance in exposed area subset.

Band Areas of Low Reflectance Areas of High Reflectance
Red < 80 > 90
NIR < 70 > 70
MIR1 < 130 > 150
MIR2 < 100 > 100
Thermal > 200 < 190

Based on these signatures, low reflectance areas are assumed to relate to burnt areas,
while areas of high reflectance are more likely to be areas of sparse or senescent
vegetation, consistent with grazing. This information was used to guide interpretation
of the final classification.

9. In the final stage of the hierarchical classification, a 15-class ISODATA unsupervised
classification was done on the exposed area subset (the image from which all open
water and healthy vegetation had been masked). The signatures of two of the classes
matched the low reflectance area pixel values identified above (see Figure 8). A third
class matched all the criteria except that the red pixel values were all over 90, while a
fourth class matched all the criteria except that the thermal pixel values were all
around 190. While these may also be burnt areas, it was decided that only the classes
matching the ranges tabulated above for all the bands should be included.

10. The two classes that most closely matched the pixel values in Table 3 were then
recorded as burned areas into a single value burn map layer, completing the
hierarchical classification process.

In spite of the limited ground data on burnt areas, a rough accuracy assessment of

the bum map layer was carried out. Coordinates of places repeatedly identified by

respondents as having burned in 2001, as well as those of 3 active burns in the October

image were used to create a set of 13 points for use in conducting an accuracy assessment

of the burnt areas. This was done on a composite of the burn map layer overlaid on the

March classification. Of the 13 points, 9 were classified as such in the map layer, while 3









































c 4 c----------- -----------U---------------------- f --I-- --------- v ----bs
Class
Class 9
Class 9
Class 10
-Class 11
S Class 12
120 C ls 13
Class 14





100





80






1 2 3 4 5 6 7
Layer

Figure 8. Signature plots of the 15-class unsupervised classification of the October scene exposed area subset, showing the two

classes selected for recoding into the burn map layer. Values on the Y-axis are mean pixel values, while the X-axis shows the 7

Landsat bands.









were classified as dense aquatic vegetation and 1 as sedge/short grass. This suggests

considerable under-classification of the burnt areas in the map layer, and supports the

likelihood of errors of omission in the classification process. It is highly probable that the

actual area burnt in 2001 was larger, because vegetation burnt earlier in the season would

have had longer to grow, and charcoal and ash would have dispersed. Similar low

misclassifications have been reported in northern Australia (Stroppiana et al., op. cit.).

Only one ground control point identified as vegetation was identified as burnt in the map

layer. Overall, an accuracy level of approximately 70% can be assessed, with the

reservation that the number of points is limited, and that the accuracy is related to rapid

changes in burnt areas. Although this assessment is statistically insufficient, it is

considered to be adequate as an indication. Once the burn map layer had been created

and verified, the area of bur in the papyrus/reed/thatch map layer and the grazing map

layer was calculated by using the burn map layer as a mask to subset those.

Identification of Village Resource Areas

Collection and grazing areas were defined on the basis of information gathered

from the household survey and informal interviews. In some instances, GPS co-ordinates

were taken. In the field, informants helped develop hand-drawn maps of places used for

various resources. These were then sketched onto 1:50 000 topographic sheets, and

modified to accommodate other information on accessibility. Using ERDAS Imagine,

areas of interest were developed to correspond to each village's grazing area and plant

resource collection area, as defined on the maps. These areas of interest were used to

subset the March image. The result was eight images, one for each of the grazing and

collection areas of the villages. Again, these were recorded to provide single value map

layers. It must be noted that these areas correspond to the range within which people









(and livestock) move, and within which the resources fall. The actual resources cover

only a proportion of these ranges.

Burning within these areas was calculated by using these range map layers as

masks to create subsets of the burn map layer. Burning of the resources i/ i/hi/ these

areas was also calculated by using the range map layers as masks this time of the

previously created bum-resources subsets referred to above.

Calculation of Geographic Extent of Resource and Burn Areas

ERDAS Imagine automatically measures the total area of a given map layer, and

reports the area in hectares in the Raster Attributes table under the image viewer. These

data were exported to a Microsoft Excel spreadsheet to calculate proportions.

Socio-economic Assessment

Most of the socio-economic data were collected in the study area between late May

and mid July 2002. Prior to starting the research, meetings were held with the village

chiefs to obtain their permission to work in the villages. With the exception of Mogotho,

the chiefs all called public meetings to introduce the study to the community. Three

research assistants helped administer a quantitative household survey, and to conduct

informal interviews on certain discussion topics.

Throughout this study the unit of analysis is the household, unless otherwise stated.

This is considered the most appropriate scale for quantification of resources use, as well

as for comparisons between types and extent of use. Due to time constraints, it was not

possible to conduct a survey of all the households in each village. For statistical validity

ii i/hi/ each village, a minimum sample of 30 per village would be preferred. According

to the Central Limit Theorem, the shape of the sampling distribution more closely

resembles the shape of a normal distribution as sample size increases. Where the sample









size is greater than or equal to 30, a normal distribution shape is reached if the population

itself is normally distributed (Agresti and Finlay 1999). Within each village sampling

was done using a systematic random approach (ibid.). The estimated total number of

households in each village was divided by 30 to identify the skip number (k). The

sample was then randomized by selecting the first household at random for any value

between 1 and k. After that every kth household was selected by walking transects

through the village that were perpendicular to the panhandle bank, thus avoiding any bias

associated with the linear features of the settlements. The number of households was in

fact not known, and these were estimated by dividing the population by 6.5, the 1997

estimated mean household size for the Okavango (see Chapter 1). Thirty-five

questionnaires were administered in Mogotho and 36 in each of the other villages, giving

a total of 143 cases.




Table 4. Population and sample sizes of focal villages.

Village Mogotho Sekondomboro Samochima Nxamasere Total
Population a 557 655 847 1328 3387
Est. # of 85 100 130 200 515
households
Sample 35 36 36 36 143
a. Central Statistics Office 2002.


The results of the survey are considered to be representative of all households

within the focus villages. For a 95% confidence interval and assuming maximum

heterogeneity, the total sample provides a margin of error of 0.08 %, while the samples

for each village have margins of error of 0.16%.









The survey was tested on five households in Mogotho and a few minor

modifications were made to the questionnaire as a result. The questionnaire (see

Appendix A) was conducted in the national language Setswana, which is understood by

the researcher and almost all respondents. Where necessary, the research assistants

provided translations into Thimbukushu, Xanikhwe or Seyei or Sekgalagadi. The

questionnaire gathered information on household characteristics, tools and assets owned,

economic activities, quantities of resources used and the effects of fire on these resources.

A strict one-year point of reference was adhered to for all questions. The data were all

coded and entered into a database, and analyzed using SPSS version 10.0 software. Two

types of variables were recorded: nominal (qualitative) data and interval (quantitative)

data.

Forty-three informal interviews were also conducted 8 in Mogotho, 11 in

Sekondomboro, 15 in Samochima and 9 in Nxamasere (see Appendix B). Information

was solicited in as neutral and open a manner as possible, and informants were asked to

look at both positive and negative aspects of the topics raised. Respondents were

encouraged to focus on one particular resource for the discussion, even if they used

several. Discussions were held across several themes, to provide deeper contextual

information about people's views regarding access to resources and about burning.

According to Spradley (1979), themes form part of the system of meaning that underpins

a given cultural context. By focusing the informal interviews on specific topics,

recurring themes were able to emerge. Ethnographic analysis of these themes was used

to provide a systematic interpretation of the perspective of those reliant on the resource









base (ibid.). This allowed the social constructs of the respondents to be presented without

being filtered by external theories or value judgments.

Extent of Reliance on Key Natural Resources

Information on the extent of reliance on natural resources is taken mostly from the

quantitative household survey. Information for the panhandle as a whole was first

assessed. The proportions of households relying on each of the key resources, as well as

the central tendencies of the amounts of these resources, were examined. Descriptive

reports such as variability and distribution were generated.

Quantification of resource use

Some households had a clear idea of exactly how much of a particular resource

they had collected in 2001. However, for many, particularly with regard to resources

collected sporadically, it was necessary to calculate this based on number of collecting

days, and amounts per day. In addition, an attempt was made to standardize units. For

reeds and thatching grass, small bundles were taken as the unit, and larger bundles were

counted as double. Wherever possible, actual bundles were examined. Water lily bulbs

were difficult to assess. Typically people would go out with an enamel dish, but

sometimes they would use plastic shopping bags. These quantities were estimated in

terms of a 20 cm diameter dish. A recent fisheries survey (Mosepele 2001) that involved

keeping diaries meant that many respondents were more aware of how much fish they

caught. While reeds and thatching grass were collected within a very short time-frame,

fish tended to be caught all year, or for several months at a time.

For respondents who said they fished every day of the year, the number of days

used was 300, as this was more probable, considering other social and economic factors.

For those fishing for several months at a time, months were calculated at 25 days each.









However, when assessing the number of fish, size was not taken into consideration. An

exception to this was fish caught by traditional basket. These fish are typically less than

10 cm long, and are collected in bowls. Because the focus in this study is on the




Table 5. Average sizes and prices for plant resources and fish.

Item Unit Size a Prices b
Thatching grass large bundle 20 cm 0 to 30 cm o P 20
Thatching grass small bundle 10 cm o to 15 cm o P 10
Reeds large bundle 30 cm o to 60 cm o P 20
Reeds small bundle 15 cm o to 30 cm o P 10
Mabinda (reed) mats each large, 2 m x 3 m P 20
Mabinda (reed) mats each small, 1.5 x 3 m P 10
Water lily bulbs C dish 20 cm o P 2-P 3
Palm leaves wrist o bundle P5 for 5 cm o to7 cm o P 5
Baskets tatana flat 25 cm o P30
Baskets big closed 50 cm tall P 100
Baskets small closed 25 cm tall P 50
Baskets seteko dish-shaped 75 cm o P 50
Papyrus shoots for eating a stalk P0.25
Papyrus for mats large bundle P 20
Papyrus sleeping mats each large P 100
Papyrus sleeping mats each small P 85
Tilapia Fish (fresh) each or kg 30cm, P4.50 or 1 kg P9-P10
Tilapia Fish (fresh) each or kg 15 cm or 500g P5
Tilapia Fish (dried) kg -P2.50
Barbel Fish (fresh) each large fish P 6
a. 0 =diameter
b. Values are given in Botswana Pula. At the time of writing US$1 = Pula 6. Prices are
based on average costs given by respondents in informal interviews.
c. These items are very rarely sold.









contribution to livelihoods, and not on the impact of extraction on the ecosystem, these

bowls were quantified in terms of their equivalent size of net-caught fish. Each bowl

collected was considered to equal 2 fish. The approximate sizes and local values of key

resources are given in Table 5.

Quantification of wealth

In order to allow comparisons between households in terms of wealth, a wealth

index (WI) variable was developed. This reflected both goods, such as the list of tools

and assets owned, and flows, such as cash income opportunities from employment or

small businesses. Values ascribed to the various assets are important more in terms of

magnitude than in the actual value given. For each household, the value of each item or

activity (as shown in Table 6) was summed into a new variable based on whether the item

or activity was present (1) or absent (0) in the past year, independently of the number of

items or extent of activity. An important exception to these are livestock owned and

wage employment, where actual numbers were multiplied by the representative values.

An example from a random household would be:

WI for Household X = (beer selling: 1*200) + (street vendor: 0*500) + (carpentry:

0*500) + (thatching: 0*200) + (brick-making: 0*500) + (well-digging: 1*200) +

(smithy: 0*200) + ... + (grows crops: 1*200) + (cattle: 3*2000) + (goats: 5*250)

+ (donkeys: 0*100) + (chickens: 7*5) + (wagejobs: 1*5000) + (drought relief

jobs: 4*100).

The inclusion of livestock is important because people base social status to

livestock ownership (particularly cattle). The relationship between wealth and cattle

ownership is interesting because people in the study area do not regularly trade their

livestock, only selling or slaughtering for special events such as weddings or funerals, or









if there is an urgent need for cash. The role of cattle in determining wealth is explored

further in a causal path analysis discussed below.




Table 6. Values of goods and flows used to create wealth index variable.

Item Value a Item Value
Beer selling 200 Hoe 20
Street vendor 500 Plough 100
Carpentry 500 Yokes 20
Thatching 200 Harrows 20
Brick-making 500 Spade 20
Well-digging 200 Axe 20
Smithy 200 Gun 2000
Bakery 500 Spear 20
Poultry project 500 Fish net 200
Knitting for sale 100 Fishing basket 20
Sewing for sale 100 Large pots for paid cooking 100
Bicycle 200 Jerry cans for water 100
Cart 1000 Drums for water 200
Vehicle 50000 Mosquito net 50
Tractor 75000 Grows crops 200
Dug-out canoe 1000 Cattle no.*2000
Sledge 500 Goats no.*250
Wheelbarrow 100 Donkey no.*100
Standpipe in yard 200 Chickens no.*5
Water pump 500 No. of full-time jobs no.*5000
Generator 1000 No. of drought relief jobs no.*100
Radio 100
Lamps 20
Stove 100
a. Values are given in Botswana Pula. Prices are based on current average costs of items,
allowing some devaluation. Annual income from small businesses was estimated
based on the purchase potential in the villages. The value of cattle is accurate at
current prices.









Proportional contribution of key wetland resources to livelihoods

Another measure of assessing the importance of wetland resources to individual

households is in terms of the proportion they contributed to the total household livelihood

in 2001. To calculate this, the amounts of the resources listed in Table 7 that households

collected in 2001 were multiplied by the values presented in Table 5.




Table 7. Wetland resources included in natural resources value.

reeds thatching grass
palm leaves water lily bulbs
papyrus for eating papyrus for mats
fish *

These figures were then summed into to provide a natural resource value (NRV)

and measured against the wealth index (WI) as follows:

Proportional contribution = NRV/(NRV + WI),

where NRV = natural resources value and WI = wealth index.

This provided another variable for assessing variations in reliance on the key

resources. It must be noted that there are also dryland resources such as poles, timber and

firewood that are not considered in the total livelihoods. In addition, some wetland

resources are not collected every year. However, it is felt that the likelihood of any

household collecting a particular resource in a given year is the same, and so the

proportional contribution for a year is indicative of most years.

Variations in Reliance on Key Resources

Differences in extent of reliance on certain resources were assessed in terms of

spatial location, gender of household head, ethnicity and wealth. These were primarily









tests of measures of association, with all tests set to a 0.05 significance level (95%

confidence interval). The combination of data types determined the tests conducted, as

shown in Table 8.




Table 8. Measures of association by data type.

Nominal Interval
Nominal Crosstabs (Chi-square Phi Comparisons of means (independent sample
and Cramer's V; Fisher t-test for 2 samples; ANOVA for more
exact test for small than 2 samples, F-value as measure of
samples) association), Fisher's LSD to follow up
which samples vary significantly.
Interval Correlations (Pearson's correlation co-
efficient)

Spatial variation is assessed on the basis of the 4 focal villages. Analysis of gender

differences was limited to household head. Households in the study were categorized as

male-headed, female (dejure) headed, and female (de facto) headed. This last category

was defined where the male head was absent for longer than 6 months of each year.


Table 9. Names of languages spoken by different ethnic groups.

Ethnic Group Language Comment
Hambukushu Thimbukushu
Xereku Xereku Related to the Hambukushu
Bakgalagadi Sekgalagadi
Bugakhwe Bugakhwe A San group
Xanikhwe Xanikhwe Sometimes known as "River San"
Bayei Seyei
Batswana Setswana









Main language spoken was used to characterize the ethnicity of a household. This

was done because the government has avoided any emphasis on ethnic or tribal identities.

As a result, people are unused to talking about this. The groups and their languages are

shown in Table 9. In the text of this thesis, the ethnic group name will be used in

preference to the language.

Effect of Fire on Availability of Key Natural Resources

Figures were taken from the spatial analysis of the satellite imagery to determine

what proportion of resources were burnt. This information was supplemented by the

responses of people in households that had experienced fire where they normally collect

resources, as to whether fire had improved or reduced their access to specific resources.

Only households that usually collected a given resource were asked if there had been fire

in the resource area in 2001. Although there is no direct evidence to support this, it is felt

that the number of "yes" responses was lower than the number of households that

actually knew there had been a fire there. This feeling is due to some hesitancy

witnessed as some respondents answered the question, and may in part be because people

were wary of talking about an illegal activity, or embarrassed to say they didn't know,

even if this was because they themselves weren't the collector, or the household hadn't

collected that year. There was also a sense that, if the fire hadn't affected some one, for

them it had not occurred.

Qualitative responses on how fire affects resources were taken from the informal

interviews. These were categorized and built into ethnographic themes as described

above. They were also used to provide contextual information for the analysis of the

household survey responses relating to this issue.









Causal path linking sustainable livelihoods, resource use and the effects of fire

During the initial stages of data analysis it became clear that socio-economic

factors and not just fire played an important role in determining access to resources.

In order to be able to bring the focus back to livelihoods and their sustainability, and to

compare the relative effects of fire and livelihood sustainability on access to resources, a

causal path analysis was developed using both original and derived variables from the

household survey. An important assumption is that the wealth index (WI) is an adequate

proxy for livelihood sustainability, at least in the short term and at household level. As is

shown in Table 6, both assets (goods) and income opportunities (flows) were considered

when creating the WI variable. While bivariate tests showed that wealth was central to

resource use, it was clear that it was not the sole factor. Significant associations, as show

in Chapter 3, revealed that other factors influence wealth, and that wealth is not the sole

factor determining access to resources. Given the complex range of variables involved,

three stages of multiple regressions were used to test the likely causal path, presented

diagrammatically in Figure 9. This diagram is based on the survey variables that had

strong associations in bivariate tests.

Social characteristics, as represented by gender of household head and ethnicity,

were assumed to be the primary factors influencing the number of adults in the household

(people aged 15 years and over), as well as the number of jobs. Bivariate analysese did

not show significant associations between gender of household head and number of cattle

owned, nor with ethnicity or cattle owned. Nevertheless, these relationships are tested

again here to confirm this finding. The predicted influence of social characteristics is

shown in Stage 1 in the diagram in Figure 9. Number of wage jobs is thought to be the

most important flow, while cattle and adult labor are thought to be the most important









assets, in determining wealth. Indeed, cattle and wage employment have already been

assumed to be part of wealth, hence their inclusion in the WI variable in the first place.

However, in order to show their importance, the WI variable was altered for this exercise.

Three key items (values for canoe ownership, cattle owned and wage jobs) that

individually had strong associations with resource use were taken out to create a modified

WI variable (referred to in the diagram as MVI) so that their relative input to wealth (as a

proxy for livelihood sustainability) could be determined.

The second stage of the multiple regressions tested the MVI for dependency on

number of adults in the household, number of cattle owned and number of wage jobs in

the household. The third and final stage of the analysis returned the focus to the role of

wealth relative to other factors notably fire in determining various aspects of resource

use.

Several of the variables contained nominal data with more than 2 categories. This

meant that they could not simple be recorded to numeric "presence/absence" values. In

order to include them in the analysis, it was necessary to create dummy variable by

moving each category to a separate variable. This was done for ethnicity and location

(village). Bugakhwe and Xanikhwe were combined into a single group San because

their individual sample sizes were small. Hambukushu was chosen as the referent

variable for ethnicity, and Mogotho as the referent for location. These two variables were

therefore not entered into the regression equations. Female defacto households were

recorded as male-headed, and the variable was assigned new values to allow gender of

household head to be considered as either male or not. Ownership of a canoe and

experience of fire were recorded in a similar manner to indicate presence or absence.













Stage 2 Stage 3


No. of adults in
household


No. of adults
in household


Ethnicity


No. of cattle
owned


Note:
"No. of adults" and "Ethnicity" are considered
to affect access to resources both directly and
through "Wealth"


Figure 9. Survey variables expected to influence access to wetland resources. The variables were selected on the basis of significant bivariate
associations. This causal path diagram shows the 3 stages for which multiple regressions will be run. In the first stage, no. of adults, no. of
wage jobs and no. of cattle are separately tested as dependent on ethnicity and gender of household head. In the second stage the modified
wealth index (MVI) is tested as dependent on no. of adults, no. of wage jobs, and no. of cattle. In the third and final stage, various variables
that comprise access to wetland resources are separately tested as dependent on no. of adults, ethnicity, MVI, canoe ownership, location/village
(where presence of resources may differ), and whether fire was experienced or not. This final stage allows the effect of fire on access to
resources to be evaluated relative to other factors.


Access to wetland resources (as
assessed separately for each of
the following dependent
variables):

* Range of resources collected

* No. of fish caught

* Amount of reeds collected

* Amount of thatch collected

* No. of baskets made

* Proportional contribution of
wetland resources to livelihood


Stage 1









Temporal variations in effects of fire

Given that the timing of the fires (how early in the season) appeared to be a key

factor in determining the negative effect on livelihoods, it seems likely that the timing of

collection (how late in the season) by a given household would also make them more

susceptible to the negative effects of fire. A decision tree model based on factors

determining whether households could go early to collect resources or not was therefore

developed from the relevant theme. Ethnographic decision tree modeling allows one to

predict group behavior in situations where decisions are made by individuals or in this

study, individual households (Gladwin 1989). A set of alternatives or "if- then" rules

determine the path of the tree, and reveal differences between households in their ability

to access resources soon enough to avoid loss due to fire.

Conflict Surrounding Burning

Measuring conflict between people also proved difficult, because for many

respondents, conflict meant confrontation. This did not happen because, as became clear

during the course of the survey, the identity of those responsible was normally not

known. Substitute questions such as whether people lost access to resources because of

fire, or whether they believed that fire was bad, and why, were asked. Some post hoc

statistical analysis was possible based on the timing of reported fires, and which resource-

user types were typically held responsible for setting fires. Comparing cross-tabulations

of month of fire versus origin of fire for each resource type allowed the overall number of

fire events experienced to be identified. The informal interviews provided several themes

linking conflict to the lack of control over burning.















CHAPTER 3
RESULTS

Extent and Distribution of Fire in the Panhandle and Key Resource Areas

The wetlands of the panhandle, as defined in Chapter 1, measure some 88790

hectares (ha), including permanently dry islands within its boundaries. The spatial extent

of the study area is shown in Figure 10. Open water (with no surface vegetation)

accounts for about 10% of this area. 1 Dense aquatic vegetation, such as reeds, papyrus

and tall grasses, covers more than 40% of the panhandle's area. Short grasses and sedges

(floodplain vegetation) account for some 30% of the study area, including those parts

which may be covered in shallow floodwaters for part of the year, but which dry out

seasonally. These vegetation groups are shown graphically in Figure 11. Figure 11

shows only that portion of the panhandle in the vicinity of the focal villages to provide

greater resolution for the different classes.

By early October, at least 3690 ha (approximately 4.1%) of the wetland area had

burnt (Figure 12) during the 2001 dry season. Most of the burning took place on the

floodplain areas. While 4.7% of the floodplains' short grass and sedge vegetation burnt,

only 2.2% of the dense aquatic vegetation did.










1 All the measurements and proportions in this section are derived from interpretation of the satellite
imagery.











590000 600000 610000 620000 630000 640000 650000


Sekondombuor


7980000




7970000




7960000




7950000




7940000




7930000




7920000


640000 650000
590000 600000 610000 620000 630000 640000 650000


Kilometers


Projection: UTM Coordinate System Zone 34 South
Spheroid: GRS80
Datum: WGS84


Figure 10. Study area map. Green areas denote healthy vegetation, while areas that
appear black are open water, and pink areas are exposed soil. All 12 villages along the
panhandle are located on the dry land, but only the four villages that are the focus of this
study are shown.


Samochima






Nxamasere











Okavango Panhandle Wetlands Area
Area = 88.790 hectares

Landsat ETM Image 28 March 2001
red. green, blue = 5. 4. 3


gotho


7980000




7970000




7960000




7950000




7940000




7930000




7920000


~--'-- I







49




590000 600000 610000 620000 630000




7970000 .. 7970000






7960000 7* 0000
7950000 790000


-, ";'. IIJ EJF
.'. & .. :'









7940000 7 5 o 7960000
,' r f.; I 'l






P j li A 1 '.I-, l.h" .. .
Hippo .. ,







7930000 S 7930000




590000 600000 610000 620000 630000
< 0Kilometers
10 0 10 20 30
Projection: UTM Coordinate System Zone 34 South
Spheroid: GRS80
Datum: WGS84

Figure 11. Map showing main vegetation groups in vicinity of focal villages. Grazing
areas correspond to the "Open water seasonal" and "Sedge / short grass" classes.
Classes were derived from the March 2001 Landsat scene using an unsupervised 15-class
classification. The sedge / short grass category was merged from four of these classes,
while the papyrus/reeds/ thatch class was merged from 6 of them. This map shows why
Mogotho and Samochima are considered to be "river" villages and Nxamasere and
Sekondomboro "floodplain" villages.

















7980000





7970000





7960000





7950000





7940000





7930000





7920000


Kilomreters


Projection: UTM Coordinate System -
Spheroid: GRS80
Datum: WGS84


Zone 34 South


Figure 12. Map showing extent of burning in 2001. Burnt areas are concentrated near
the "floodplain" villages, while the "river" villages experienced less fire.


590000 600000 610000 620000 630000 640000 650000








i .
S,'-. Sekondornboro


6. .1




Samochima



4'-, ..' Mogotho
,.

Nxamascre
*

















Areas Burnt in 2001
Study Area


590000 600000 610000 620000 630000 640000 650000
t" .. .*




[; '. ,




'<,..








i Areas Burnt in 2001
Study Area .


590000 600000 610000 620000 630000 640000 650000


7980000





7970000





7960000





7950000





7940000





7930000





7920000


.----;









Key Resources Areas for Focal Villages

The spatial extent of the plant collection ranges and the livestock grazing ranges, as

developed in the methods section, for the four villages is shown in Figure 13, while Table

10 summarizes the values. It is striking to note that for both range types, the two

floodplain villages have roughly double the area of the other two. The larger size is

primarily accounted for by the greater distance from the panhandle banks to the main

river channels, as the length along the banks is similar. There is considerable overlap

between the grazing and plant collection ranges of each village, but generally within the

ranges these two main resource types are discrete.


Table 10. Size of village resource ranges

"River" villages

Village Mogotho Samochima

Plant collection range (ha) 678.6 764.8

Proportion of total study 0.76% 0.86%


area

Length along banks (km)
Width (km)


12.3
0.58


11.7
0.98


"Floodplain" villages

Nxamasere Sekondomboro

1463.5 1306.75

1.65% 1.47%


12.4
1.10 (2.72 at
extreme)


8.1
6.87


Grazing range (ha)
Proportion of total study
area
Length along banks (km)

Width (km)


1243.1
1.40%


14.2

1.05


1157.3
1.30%


14.6

0.85 (2.02 at
extreme)


3407.39
3.84%


14.6

2.55


2858.63
3.22%


11.7

2.40 (5.34 at
extreme)







52




590000 600000 610000 620000 630000 640000 650000
7980000 7980000



Sekondornboro
("floodplain" village)
7970000.. 7970000


d ef

Sarnochima L
7960000 ("rivor" village) 7960000



Mogotho
("river" village)
7950000 7950000
Nxamasere
("floodplain" village)



7940000 7940000




7930000 7930000
Sekondombnhor Gra2ing Area
SarVochirTia Grazingc Area
| NNaasere GraTing Area
Mogotho Grazing Arco
7920000 Sekondomboro Plant Collection Area 7920000
Samochima Plant Collection Area
Nxamrasere Plant Collection Area
Mogotho Plant Collection Area

590000 600000 610000 620000 630000 640000 650000

I 10 0 10 20 30 40 50 60

Projection: UTM Coordinate System Zone 34 South
Spheroid: GRS80
Datum: WGS84

Figure 13. Map of grazing and wetland plant resource areas for each focal village.
Resource areas were based on information provided by informants during the survey, and
from GPS coordinates collected in the field. There is considerable overlap between graze
areas and plant collection areas for each village. Both types of area extend much further
into the panhandle for the "floodplain" villages, while those of the "river" villages are
constrained close to the panhandle edge. This is because dense aquatic vegetation limits
access.










Grazing areas

The importance of floodplains for grazing is shown by both the size and the

proportion of short grass/sedge vegetation available to Sekondomboro and Nxamasere

villages. These data are presented in Figure 14.


Figure 14. Comparison of available wetland grazing for focal villages. The Y-axis gives
square kilometer values for size of grazing area and percent of grazing area for available
graze. Only in the case of Sekondomboro do the actual graze resources cover more than
half of the grazing area.



Plant collection areas

The availability of resources within these ranges also varies considerably. The

river villages have the greatest concentration of reeds, thatching grass and papyrus. In

terms of spatial extent, these three resources are considered together (see for example, the


70.0


6 'Size of grazing -re-
60.0 i

P'er: ent of grazing
50.0 3re-; 3 aiab 1 graze


40.0


30.0


20.0


10.0 -


0.0
Mogotho Nxamasere


Sekondomboro Samochima







54


papyrus/reed/thatch class shown in Figure 11). Together these three wetland plant

resources cover 59.7% of Samochima's, and 45.2% of Mogotho's plant collection area.


This is in contrast to the floodplain villages, where these resources cover 44.5% of


Sekondomboro's and only 23.4% of Nxamasere's areas, due to its drier conditions, as is


shown in Figure 15.


Figure 15. Comparison of available plant resources for focal villages. The Y-axis gives
square kilometer values for size of plant collection area and percent of the collection area
covered in papyrus/reeds/thatch.




Water lily bulb collection was not mapped. These areas are related to shallow open


water, of which the largest patches are found at Sekondomboro. Hyphaene palms grow


on larger islands where floodwaters do not reach. These islands are scarce except around


Nxamasere. "Other resources we use are the riparian trees, such as the fruit of mochaba


70 0



60 0
,H,, ,-illi-J n- 1 1 ill i, -,l

50 0



40 0



30 0



20 0



100



00
Mogotho Nxamasere


Sekondomboro Samochima









[Ficus sycomorus], mokochum [Diospyros mespiliformis], and motsaodi [Garcinia

livingstonei]. There are also many leafy plants that are important as food in our lives. If

these things are found close to the village, we sometimes make specific trips to collect

them, but if they are far away we will just look for them when we are harvesting reeds."

(Water lily bulb collector, Sekondomboro, June 2001)

Fishing

Fishing takes place in several different types of locations, depending on the

techniques used. Traditional Hambukushu basket fishing is done by women in the

shallow vegetated floodplains. According to responses from the survey, the vegetation is

important as it harder for fish to escape. Because access is on foot, basket fishing usually

takes place within 500 m of the edge of the panhandle. Some women walk several

kilometers along the bank to find suitable places. This was particularly evident at

Samochima.

Net fishing is mainly done in areas of unvegetated water on the floodplains,

although nets are also strung along quieter lees of channels and lagoons. Hook-and-line

fishing is done on the big areas of permanent water, and, like net fishing, is dependent on

access by dugout canoe or motorboat. Motor boats are used by small artesanal fishermen,

who have a co-operative based at Samochima. This co-operative is supported by

subsidies and technical advice from the fishing division of the government's agricultural

extension agency. The boats give them the ability to range up to 50 km away along the

main rivers.

Burning in the Key Resource Areas

The percentage burnt within all the total villages' plant collection areas (8.65%)

and grazing areas (10.8%) is much higher than the 4.16% for the study area as a whole.









Nevertheless, this is still a small proportion of the resources areas. There were differences

between the villages, as Table 11 indicates. Notice how greater areas and proportions

were burnt in the plant collection areas of the floodplain villages.


Table 11. Extent of fire within village resource areas

"River" villages


Village

Area burnt in plant
collection range (ha)
Proportion of collection
range burnt


Area burnt in grazing range
(ha)
Proportion of grazing range
burnt


Mogotho Samochima

46.57 11.05


6.86%



76.83


6.18%


1.45%



34.26


2.96%


"Floodplain" villages
Nxamasere Sekondomboro

252.84 53.86


17.28%



613.48


18.0%


4.12%



209.80


7.34%


Fire in grazing areas

Nxamasere had the highest proportion of its graze area burnt (see Figure 16 and

Figure 17). In terms of actual graze resources, the two river villages both had less than

2% burnt, while the floodplain villages had more than double this proportion.







57



600000 610000 520000




.. SekondombLoro
7970000 t 7970000










7960000 Sam-nchimn 7960000








iA og ot ho

7950000 7950000



Study Area Nxamnsram
Sokondomboro Gro7ing Aron
Samochima Grazinig Area
SNxamascro Groaing Area
Mogolho Gra.ing Area
Gra2e Resources Burnt
7940000 Non-graze Resources Burnl 7940000

600000 610000 620000

Kilometers
10 0 10 20

Projection: UTM Coordinate System Zone 34 South
Spheroid: GRS80
Datum: WGS84


Figure 16. Map showing graze resources burnt in the grazing areas of each focal village.
Maroon indicates actual graze resources burnt, while tan indicates burning of other
vegetation types.







58



20.0

S*P. r.: 1i.l .:.1 ,|r i -.r-, L.iri.
18.0
M pared to-I ._.i tre_- rver v_
t I I h. I .| q -l r murh l s h flrllml
16.0


14.0


12.0


10.0


8.0


6.0


4.0


2.0


0.0
Mogotho Nxamasere Sekondomboro Samochima


Figure 17. Proportion of grazing vegetation burnt in graze ranges. Nxamasere and
Sekondomboro, the "floodplain" villages, had about twice the area of graze resources
burnt compared to the "river villages.




Fire in plant collection areas

The proportion of the actual resources that burnt within the collection areas appears


to be much less than for the collection areas themselves. All the villages had less than


2% of their reeds/thatch/papyrus burnt, with the exception of Nxamasere. This village


had 5.6% of these plant resources burnt in their collection area. Samochima appears to


have experienced very little fire at all (see Figure 19 and Figure 18).













ES 0000 600000 610000 620000


7970000











7960000











7950000











7940000


600000


Projection: UTM Coordinate System -
Spheroid: GRS80
Datum: WGS84


Zone 34 South


Figure 18. Map showing papyrus /reed /thatch burnt in the plant collection areas of each
focal village. Red indicates burning of actual resources, while beige indicates burning of
other types of vegetation.


Study Aron
Sckondomboro Plant Collection Area
SSarnochima Plant Collection Arca
Nxamasere Plant Collection Area
Mogotho Plant Collcction Area
Papyrus /Reeds /Thatch Burnt
Non-resource VegetaLion Burnil


7970000











7960000











795000











7940000


590000


610000


620000


Kil orrnters


I _ I


590000


G00000


610000


620000







60



200


180 *


160 M 'i : i j


140


120


100


80


60


40


20


00
Mogotho Nxamasere Sekondomboro Samochima

Figure 19. Proportion of plant resources burnt in collection ranges.





Temporal Distribution of Fire

Because only one dry season satellite image was analyzed, it is not possible to


know exactly when fires occurred. However, responses from the household survey


indicate that more than half the fires people observed in their resource areas were in


September. No fires were reported for the months of February to July. According to


informal interview respondents, this is because the flood waters are still too high at this


time of year, and so it is too wet to burn. Some people noted that fishermen did in fact


set many small fires in the vegetation immediately around the openings where they set


their nets, and that these fires did not spread at all because of the water.









Discussion Theme Local Knowledge of Fire Behavior

People are very aware of fire and how it moves within the panhandle, as the

following excerpts from the informal interviews show. However, their expressions are a

reflection not only of their experiences, but also of whether they see fire as good or bad.

"There is no way that fire can be controlled, especially in the delta because it is

very difficult to stop, it burns from the first place to the last one, and results in the whole

ecosystem burning." "Whether fire can be controlled depends on how dry it is. If it is

very dry, no-one can stop it. Generally fires in the delta cannot be controlled." "Delta

fires are usually stopped by water or hippo paths."

"In the old days we used to bum regularly, but now it is less frequent and this is

why the fires are so big these days. If you burn every other year, the fires don't go so

far." "Even if you burn every year, you can still get big fires. There is no difference in

the size or intensity of fire whether you bum every year or once in seven years."

"Fires go according to the wind." "Usually fires do bum everything because of the

wind." "It is the wind which always brings the fires here." "Wind and the amount of

grass are the cause of big dangerous fires that go long distances." "The direction of the

fire usually depends on the direction of the wind."

"The size of the fire depends on the amount of old vegetation or twigs burning.

The smaller the amount, the smaller the fire." "Fires are bigger the year after a big

flood, because there is more to burn."

"Even early in the winter [June July], there is so much wind and the fuels are so

light and hot, they spread easily." "Burning before winter isn't possible because of the

floodwater, which stops burning." "Flood doesn't affect fire, the vegetation always gets

dried out."









"Fire never kills everything, wherever there is water patches of vegetation do

remain. These things all re-grow from their base or roots so they are not damaged by

fire. "The interaction of fire and water is essential. When there is drought, even the

roots bur because there is no water in the soil. "We expect rain in October and

November, so September is the best time for fire.'

Reliance on the Wetlands Natural Resource Base

Only three households out of the 143 surveyed (2.1%) did not use any of the

wetland resources asked about in the survey. An unexpected finding was that very few

people (8.39%) grew crops in the floodplains, in spite of the greater level of nutrients

compared to the conditions on the sandveld. Respondents explained that the land

authority had removed all rights to floodplain fields because of increased conflicts with

hippos. Where floodplain fields did exist, these were typically on the banks beyond the

reach of floodwaters.

The association between use of one resource and that of another is measured in two

ways whether a household usually uses each of the resources (nominal data), and how

much of each of the resources they use (interval data). The nominal data relating to

whether a household usually used one resource compared to its usual use of others is

shown in Table 12. Significant values are shown in bold.

In 2001, wetland resources contributed on average 14.7% of household livelihoods,

as measured as a proportion of the household's total wealth (which will be discussed in

greater detail below), with a median value of 5.2%. For 10% of households, however,

wetland resources contribute at least 50% of the total livelihood. The range of values is

shown graphically in Figure 20. The proportion of households relying on various

resources are shown in Table 13.










Table 12. Comparison of reliance on different resources. The table presents the results
of cross-tabular analyses where usual use of one resource is compared against
usual use of another. Significant values are shown in bold. Use of reeds,
thatching grass and fish have strong associations with each other. Use of palm
leaves is significantly related to use of water lily bulbs and papyrus for eating.


Grazing Reeds a Palm Thatching Water Papyrus Papyrus Fishing
leaf grass lily for for
bulbs eating mats


.094 .161 .157
.354 .110 .121


-.139 .437
.097 .000


.050
.551


.140 .010 .005 .068
.165 .919 .960 .501


.098 .066 .124 .177
.239 .429 .137 .035


.275 .199 .125 .056
.001 .017 .136 .503


.080 .128
.337 .125


.175 .163
.036 .051


- .177 .190 .246
.035 .023 .003


Grazing



Reeds



Palm leaf



Thatching
grass


Water lily
bulbs


Papyrus
for eating


Papyrus
for mats


Fishing


.268
.001


.138
.098


a. The first figure is the value of Phi and Cramer's V (see Methods). The second is the
significance level.


.188
.024









Table 13. Household reliance on different resources.
Proportion of Proportion of Mean amount
HH generally HH collecting collected Median amount Std.
collecting 2001 2001 collected 2001 Dev.
Grazing 37.1% n/a n/a n/a n/a
Reeds 95.1% 69.2% 31.0 bundles 20 bundles 37.51
Palm leaf 27.3% 16.8% 3.9 bunches 3 bunches 2.9
Thatching
grass 95.8% 74.1% 30.2 bundles 20 bundles 39.4
Water lily
bulbs 35.0% 22.4% 21.1 dishes 14.3 dishes 17.9
Papyrus
for
eating 27.3% 18.9% 18.1 bundles 10 bundles 28.2
Papyrus
for mats 41.3% 21.7% 4.3 bundles 2 bundles 5.7
Fishinga 37.8% 29.4% 1168 fish 550 fish 1976.7
a. It is believed that the proportion of households consuming fish is much higher, and
that the relatively low value reflects a growing commercialization of the resource.


0.000 .125 .250 .375 .500 .625 .750 .875


Figure 20. Proportional contribution of wetland resources to household livelihoods. The
proportion of total household livelihood that was derived from wetland resources in 2001
is shown for the survey sample (N = 143). The mean proportion was 0.147, with a
standard deviation of 0.21. The proportional contribution is calculated by NRV/(NRV +
WI), where NRV = natural resources value and WI = wealth index.









Grazing of livestock in the floodplain is determined by several factors. Not all

households own livestock, and not all types of livestock graze. Trying to assess the

extent of use is more difficult. While 51.7% of households owned goats, these are

browsing animals and use the sandveld instead of the wetlands. It is for cattle that the

grazing resources of the panhandle are mainly used. Only 39.8% of households own

cattle, and the distribution of the numbers across the households is extremely skewed, as

will be discussed below. Nevertheless, 37.1% of all households, and 61.4% of all cattle

owners, graze their livestock in the panhandle when conditions are dry enough. Donkeys

and horses account for the difference.

Thatching grass appears to be the most important key resource collected 95.8% of

households normally collect thatching grass. This resource is used exclusively for

making roofs, which are typically replaced every three years. For this reason, the

proportion of households collecting in 2001 was slightly lower 74.1%. The mean

amount collected for all households who normally collect grass, including those that did

not collect in 2001, was 23.4 bundles. If the number of traditional (as opposed to

government employee) households is taken at 2500 for the study area as a whole, this

suggests a level of extraction of between 55 000 and 60 000 bundles.

While the mean number of thatching grass bundles collected may provide a broad

scale idea of levels of extraction, it does not accurately reflect the variance or skewedness

of use. Seven outliers have skewed the data to the right. Five of these cases have

extreme values, of which two households collected 270 and 300 bundles last year. The

mode (25.5% of 2001 collectors) was 20 bundles per household. Even without the two

extreme outliers there is still considerable skewedness and kurtosis (greater clustering), as









is shown in relation to the normal curve in Figure 21. Only seven households (6.6%) sold

thatching grass last year, all selling 20 bundles or less.


Figure 21. Frequency distribution of thatching grass bundles collected in 2001. For the
households who collected grass that year (N = 106), the mean number of bundles was
30.2, with a standard deviation of 39.4.



Reed collection follows a very similar pattern to thatching grass. Ninety-five % of

households regularly collect reeds, which are used to build the walls of both huts and

yard enclosures. Less commonly nowadays, reeds are also flattened and woven into rigid

mabinda mats that are used for sitting on or for wall or floor coverings. In 2001 69.2% of

households collected reeds, with the mean number of bundles 22.59. Assuming the 2500

households referred to above, this once again suggests approximately 55 000 to 57 000

bundles of reeds for the study area as a whole.


10

(D

LL 0
0-10


160-170 240-250


80-90









As with thatching grass, there were two extreme outliers one of which was the

same household fro both resource types. The outlying households in reed collection

harvested 200 and 275 bundles last year, while both the median and mode for bundles of

reeds collected were 20, with a standard deviation of 37.5. This shows how skewed to

the right distribution is, as two standard deviations to the left would give negative values.

Twenty-three percent of those who collected reeds in 2001 sold bundles, compared to the

6.6% selling thatching grass. Nevertheless, the quantities sold remained very small

(mode = 10 bundles, mean = 10.7 bundles). There is no significant correlation between

amount of reeds collected and amount sold (Pearson's r =0.066, p = 0.514). Given this,

and the lack of markets and the small quantities sold, it is likely that the sale of reeds

reflects a once-off need for cash rather than increasing commercialization of the resource.

Only 15.4% of households had members who still made mabinda mats, typically

done by men. During the interviews several respondents said that few people knew how

to do this any more. In the past year, only 7% actually made mats, while 2.8% of

households sold mats. One man made and sold 100 under commission from a nearby

safari lodge, but the other households sold between 6 and 12 mats each.

Palm leaf collection and basket-making will be discussed in the section on spatial

variation in reliance, as this resource is mainly found in one area. Note that papyrus for

eating is in most cases not collected in bundles. Instead people eat it while out fishing or

collecting other resources. When brought home it is usually for children, and must be

eaten that day or else it dries out. Ten households in the sample (7%) sold papyrus

sleeping mats in 2001. All sold 6 or less. Some respondents showed how empty grain

sacks were now being used for sleeping mats, and explained they were easier to make. It









is not sure if sufficient quantities of these sacks are permanently available, or if their

presence and use is limited to how long welfare relief is provided after the cattle

slaughter.

Reliance on fish needs to be examined both in terms of the proportion of

households collecting, as well as the amount of fish being sold. The proportion of

households who said they usually catch fish in the delta was 37.8%. Those actually

fishing in 2001 comprised 29.4%. Most households caught enough fish (mode = 600

fish, median = 550 fish) for their households. However, the mean value (1168.1) reflects

considerable right-skewedness, which is explained by 35.7% of those fishermen catching

in 2001 who caught between 800 and 10000 fish each. Fish sale quantities are typically

low, implying local reliance on a resource collected by others within the study area.

These data, together with the low proportion of households catching for

themselves, suggest that a large number of households buy fish caught by others.

Increasingly, though, the fishermen of the Samochima co-operative are selling outside the

study area as they have both the quantities and the equipment to make transportation

viable. .There is a significant correlation between size of catch and numbers of fish sold

(Pearson's = 0.652, one-tailed significance < 0.001). Of the households selling fish, 3

sold less than 250. The mean number offish sold by the 42.9% of fishermen who caught

in 2001 was 1274. The comparative distribution of fish caught and fish sold is shown in

Figure 22.







69



25




20




15




10




0



LL
250 2750 5250 7750 10250
A


6



5



4



3



2



0 1




250 1250 2250 3250 4250 5250 6250
B
Figure 22. Distribution of fish caught and fish sold in 2001. A) For those households
catching fish that year (N = 42), the mean number caught was 1168, with a standard
deviation of 1976.7. B) For those selling fish (N = 18) the mean number sold was 1274,
with a standard deviation of 1427.6.









Spatial Variations in Reliance on Key Resources

There are strong differences between villages in the proportional contribution of

wetland resources to household livelihoods (F = 2.657, p = 0.051). In both Mogotho and

Sekondomboro wetland resources comprise almost 20% of the total household livelihood,

compared to 8.8% in Nxamasere (see Table 14). To some extent, this could reflect

Nxamasere's distance away from the edge of the panhandle, as well as the fact that it is

closest to Maun where other types of building materials can be obtained. Given that

ethnicity is also associated with village location, cultural differences may also be

contributing to this association. This is discussed in greater detail below.

In all villages the proportion of households owning livestock of any kind (including

chickens) is similar ranging from 77.8% in Sekondomboro to 94.3% in Mogotho.

Those owning cattle were less, as is shown in the box-plots in Figure 23 (where each box

shows the central 50% for each group, while whiskers indicate the bottom and top

percentiles. Larger boxes reflect greater spread of response values, with the sample size

given in the X-axis above the name of the group), and in Table 15.




Table 14. Proportional contribution of wetland resources to households by village.

Village Mean proportion of Median proportion of household
household livelihood in 2001 livelihood in 2001
Mogotho 19.3% 10.5%
Sekondomboro 19.6% 11.3%
Samochima 11.2% 3.8%
Nxamasere 8.8% 2.6%

































Mogotho Sekondomboro Samochima Nxamasere


Figure 23. Distribution of cattle ownership by village in 2001. The boxes represent the
interquartile ranges that contain 50% of the values for each village. The line across the
box is the median value. The whiskers extend to the lowest and highest values, excluding
outliers. Open circles indicate outliers, while extreme outliers are shown by stars






Table 15. Numbers of cattle owned by village.

Village Proportion of households Mean # owned by Median # owned by
owning cattle these households these households
Mogotho 37.1% 24.6 6
Sekondomboro 33.3% 12.8 5.5
Samochima 30.5% 17.7 5
Nxamasere 41.7% 16.6 12


Total numbers of cattle are very low. It is clear that people are taking time to

replace their herds, and more than half the people in all villages own no cattle at all.









Ninety-three percent of all cattle owners in Nxamasere, and 63.6% in Samochima grazed

their livestock in the panhandle at some point in 2001. This compares with 47.4% for

both Mogotho, and 41.7% for Sekondomboro, the east bank villages. The differences

between the villages are significant (Phi and Cramer's V = 0.419, P = 0.018), and the

figure for Sekondomboro is surprisingly low, given the availability of wetland grazing. It

may be that given the current low stocking rates, people prefer to graze their cattle on the

sandveld where grasses are more palatable.

Direct access to thatching grass shows some spatial variation. As Table 16 shows,

almost all households in Mogotho collected thatching grass in 2001, compared to the

west bank villages of Samochima and Nxamasere, where less than two-thirds did. While

the villages of Mogotho and Sekondomboro are perched right on the edge of the

panhandle, Nxamasere and Samochima are 6 km and 1 km away respectively. This

increased distance may make collection difficult for some households. In Samochima,

the mean number of bundles collected is highest. This fact, combined with fewer

households collecting, suggests the possibility of greater trade in the resource in this

village.


Table 16. Thatching grass collected in 2001 by village.

Village Proportion of Mean no. of Median no. of Sample sum
households bundles bundles amount
collecting collected collected
Mogotho 94.3% 31.4 20 1035
Sekondomboro 83.3% 25.6 20 767
Samochima 58.3% 40.4 20 848
Nxamasere 61.1% 25.2 26.5 555
Total 74.1% 30.2 20 3205









The concentration of grass collection into fewer households in Samochima would

appear to suggest greater trade in this resource. However, this notion is not borne out by

data on quantities of thatching grass sold. Only one household in the sample sold

thatching grass in 2001, and that was only 10 bundles. Figures for the other villages were

similarly low: Mogotho two households each selling 7 bundles; Sekondomboro one

household selling 20 bundles; and Nxamasere two households selling 10, and one

selling 5 bundles.

Only 52.8% of households in Nxamasere collected reeds last year, compared to

82.9%, 77.8% and 63.9% for Mogotho, Sekondomboro and Samochima respectively.

There was a significant difference in the mean number of bundles collected in Mogotho

compared to Nxamasere (t = 2.287, p = 0.028). This reflects the proportion of resources

available in the collection areas for the villages as identified above.

Compared to thatching grass, there was greater trade in reeds in 2001, particularly

in Mogotho, where 41.4% of households that collected reeds sold some of their bundles.

Quantities again are low, with the median values at 10 bundles for Nxamasere and

Mogotho, and 5 bundles for Sekondomboro and Samochima.




Table 17. Reeds collected in 2001 by village.

Village Proportion of Mean no. of Median no. Sample sum
households bundles of bundles amount
collecting collected collected
Mogotho 82.9% 34.1 26 988
Sekondomboro 77.8% 33.4 20 936
Samochima 63.9% 32.3 20 744
Nxamasere 52.8% 21.3 20 404
Total 69.25 31.0 20 3072









Table 18. Reed bundles sold in 2001 by village.

Village Proportion of Mean no. of Median no. of Sample sum
households bundles sold bundles sold amount
selling
Mogotho 34.2% 13.25 10 159
Sekondomboro 2.8% 5.0 5 5
Samochima 8.3% 5.77 5 17
Nxamasere 19.4% 9.3 10 65
Total 16.1% 10.7 10 246



Table 19. Households collecting palm leaves by village.

Village Usually collecting Collecting in 2001
Mogotho 0.0% 0%
Sekondomboro 5.6% 2.8%
Samochima 25.0% 8.3%
Nxamasere 77.8% 55.6%



Table 20. Bunches of palm leaves collected in 2001 by village.

Village Proportion of Mean no. of Median no. of Sample sum
households bunches bunches amount
collecting collected collected
Mogotho 0% 0 0 0
Sekondomboro 2.9% 2.0 2 2
Samochima 5.56% 5.3 4 16
Nxamasere 55.56% 3.7 3 75
Total 66.67% 3.9 3 93

As suggested earlier, there are significant variations in the participation by different

villages in palm leaf collection (Pearson chi-square = 68.07, p < 0.001). The proportion

of households that regularly collect palm leaves is shown in Table 19. All of the

Samochima households that regularly collect palm leaves travel to Nxamasere to do so,

supporting information from the informal interviews that this is where the resource is

concentrated. In 2001, only 33.3% of Samochima palm leaf collectors actually collected,









compared to 72.4% of Nxamasere collectors. Only 2 households from Nxamasere sold

palm leaves in 2001. Amounts collected are shown in Table 20.

In spite of the low levels of palm leaf collection, a high proportion of households in

all villages have members who make baskets (Table 21). The significant difference

shown for Nxamasere compared to the other villages (Pearson chi-square = 19.91, p <

0.001) is probably related to the availability of the resource. T he ways in which

households making baskets in 2001 obtained palm leaves is shown in Table 22.

Mohembo, the northern-most village in the panhandle was regularly mentioned as a place

for buying palm leaves. The number of baskets made by these households is not

significantly different, but the number sold is (F = 2.86, p = 0.045). No households sold

baskets in Sekondomboro last year, while 58.3 % of Nxamasere basket-makers sold a

mean of 4.7 baskets each. The market for baskets is through a grassroots co-operative,

Thokadi Women, who sell to a wholesaler in Botswana's capital city as well as

informally to tourists from the nearby safari lodge.




Table 21. Households with basket-makers.

Village name Makes Mean no. of Median no. of
baskets baskets made baskets made
per household per household
Mogotho 45.7% 3.2 2
Sekondomboro 38.9% 2.7 2
Samochima 47.2% 3.1 2
Nxamasere 86.1% 4.8 3
Total 54.5% 3.6 2






76


Table 22. Source of palm leaves for basket-making.

Collected by Bought Both collected Gift Total
HH member and bought
Sample number 25 30 1 4 60
Proportion 41.7% 50.0% 1.7% 6.% 100%

The number of households relying on water lily bulbs for food varies only slightly,

as is shown in Table 23. However, there are significant differences (F = 2.89, p =0.05) in

the extent to which households in the different villages use the resource (see boxplot in

Figure 24). Of those that collected water lily bulbs, households in Samochima collected a

mean of 52.5 dishes, compared to 18.6, 22.5 and 15.7 for Mogotho, Sekondomboro and

Nxamasere respectively.


Mogotho Sekondomboro Samochima Nxamasere

Figure 24. Extent of water lily bulb collection by village in 2001. The boxes represent
the interquartile ranges that contain 50% of the values for each village. The line across
the box is the median value. The whiskers extend to the lowest and highest values,
excluding outliers. Open circles indicate outliers, while extreme outliers are shown by
stars.


* 0


O









Table 23. Collection of water lily bulbs by village.

Village name Usually collect Collecting in 2001
Mogotho 25.7% 25.7%
Sekondomboro 38.9% 30.6%
Samochima 30.6% 5.6%
Nxamasere 44.4% 27.8%


The proportion of households that usually collects papyrus for mat-making is

remarkably similar for all villages, ranging from 38.9 % to 42.9%. However, there were

marked differences between the proportions that actually collected in 2001. While 42.9%

of Mogotho households collected papyrus for mats in 2001, only 16.7% in

Sekondomboro, 5.5% in Samochima, and 22.2% in Nxamasere did. Quantities in

Sekondomboro were lowest, with a mean of 2.3 bundles. It was in this village that

people mentioned that they were making sleeping mats out of grain sacks nowadays. The

highest mean number of bundles collected was in Nxamasere (5.1 bundles).

With regards to papyrus collected for eating there was a significant difference for

the different villages (Phi and Cramer's V = 0.371, p < 0.001). While 48.6 % of

Mogotho households usually collect papyrus shoots for eating, only 2.8% in Nxamasere

do. Many people that collected papyrus for eating said they did so when out fishing or

collecting other resources. In 2001, no households in Nxamasere collected, and only 11.1

% (N = 1) of those that rely on the resource in Samochima did. The median bundle

equivalent collected in Sekondomboro was 13, compared to 5 in Mogotho.

As expected, fewer households in Nxamasere were involved in fishing than in the

other villages (see Table 24). In Sekondomboro and Samochima, the proportion of

households actually catching fish in 2001 was much lower. The mean number of fish









caught by households in Nxamasere in 2001 was lower, but not significantly so at the

0.05 level. Mean and median values for 2001 fishing households are shown in Table 25.

There were no significant differences between villages.


Table 24. Households usually fishing by village.

Village Households Households fishing Households selling
involved in fishing in 2001 fish in 2001
Mogotho 31.4% 31.4% 20.0%
Sekondomboro 50.0% 38.9% 13.8%
Samochima 41.7% 19.4% 11.1%
Nxamasere 27.8% 27.8% 5.6%


Table 25. Fish caught in 2001 by village.

Village Proportion of Mean no. of Median no. of Sample sum
households fish caught fish caught amount
fishing
Mogotho 31.4% 1209.6 1200 13305
Sekondomboro 38.9% 1498.5 280 20979
Samochima 19.4% 1294.3 600 9060
Nxamasere 27.8% 571.4 75 5714
Total 29.4% 1168.0 550 49058


Likewise, there were no significant differences between villages in amounts of fish

sold, although Samochima's higher mean figures may reflect that some of the households

are members of the fishing co-operative.









Table 26. Fish sold in 2001 by village.

Village Proportion of Mean no. of Median no. of Sample sum
households fish sold fish sold amount
selling
Mogotho 20% 1082.9 1200 7580
Sekondomboro 13.8% 940.0 850 4700
Samochima 11.1% 1912.5 800 7650
Nxamasere 5.6% 1505.0 1505 3010
Total 12.6% 1274.4 925 22940

Wealth Variations in Reliance on Key Resources

Wealth, as defined using the wealth index described in Chapter 2, is not equally

distributed across households. Most households have a small proportion of the total

wealth within the study area, as is shown in Figure 25. Because other studies have shown

that wealth is related to household size, the first thing to test is whether there is a positive

correlation between the wealth index and the numbers of family members. While there is

no significant correlation for overall household size (Pearson = 0.068, one-tailed p =

0.200)., the correlation figures for wealth index vs. number of adults (15 years and over)

are stronger (Pearson = 0.120, one-tailed p = 0.076), although not significant at the 95%

level. Household sizes in all villages were in fact slightly larger than recorded in earlier

studies (see Chapter 1), and are shown in Table 27. If this were due to household

consolidation as a result of AIDS, that would explain the limited correlations with wealth.

For any particular resource, there are no significant differences in the mean wealth

of households that are involved in the collection of it, compared to the mean wealth of

households that are not. Only the sale of thatching grass was significantly related to

wealth (Pearson = 170, p = 0.041).













90

80

70

S 60O
60

0 50-
0
4 40

30

20

10


1 11 21 31 41 51 61 71 81 91 101 111 121 131 141

Number of households

Figure 25: Percentage distribution of wealth over households. The blue line represents
how equal distribution of the percentage of total wealth (the summation of the WI for all
cases in the survey) would appear, while the pink line shows the actual distribution. The
distribution has a Gini co-efficient of 0.7303, where equal distribution = 0 and
completely unequal distribution = 1. Fifty percent of the total wealth is owned by 7 of
the 143 surveyed households.






Table 27. Household size by village.


Village Mean Median
Mogotho 7.54 7.00
Sekondomboro 8.50 8.00
Samochima 8.08 7.00
Nxamasere 7.06 7.00
Total 7.80 7.00









There are no significant correlations between wealth and the amounts of each of the

resources that households collected in 2001. However, there is a significant negative

correlation between wealth and the proportional contribution of wetland resources to the

household livelihood (t = -2.361, p = 0.021). Given the importance of cattle to wealth, it

is not surprising that there was also a strong inverse association between number of cattle

owned and proportional contribution to household livelihood from wetland resources (t =

-0.172, p = 0.058).

Household size was significantly linked to households that usually collected reeds

(F = 4.583, p = 0.034) and that usually made baskets (F = 6.536, p = 0.012). While there

was a significant correlation between fish collection and household size (Pearson = 0.403,

p = 0.019), there were no strong relationships between household size and plant resources

collected or sold.

Gender Variations in Reliance on Key Resources

The high proportion of households headed by women in their own right (dejure)

was unexpected (see Table 28) given patterns elsewhere in Botswana. For example, the

dominant Tswana tribes are patrilineal and traditionally women have not been recognized

as decision-makers. This means that until recently, there have been extremely few

households with dejure female heads, although females heading households in their

husbands' absence (de facto) have been more common. In the study area, respondents

revealed that in many instances, male partners to the female heads were present, but since

the couple was not married, and lived in the woman's house, she was considered both the

head and the decision-maker that is, the dejure head. While this may in part be due to

the breakdown in marriage traditions, it has an original basis in the groom living with the

bride at her parents' homestead for the period (often years) during which bridewealth was









paid (Larson 1980). The average wealth (see Table 29) of male headed households was

greatest, followed by female defacto and then female dejure, but the differences were

not significant. However, one aspect of wealth that is important is ownership of a dugout

canoe, because of the role it plays in access within the panhandle. A higher proportion of

male-headed households than female headed households own canoes (Phi and Cramer's

V = 0.203, p = 0.053).




Table 28. Gender distribution of household heads.

Head of Household Proportion
Male 46.2%
Female dejure 51.0%
Female de facto 2.8%



Table 29. Wealth index by gender of household head.

Mean Median Min Max Std Dev
Male 26252.5 9272.5 360.0 441905.0 61195.5
Female de facto 24117.5 8457.5 6305.0 73250.0 32816.9
Female dejure 22055.34 6270 0.0 302875.0 51666.6
Total 24050.17 7650 0.0 441905.0 55630.7


Looking at labor-availability as determined by household size, on average male-

headed households are largest, and female dejure households are smallest. In this case,

while the difference may not seem much, it is significant (t = 2.598, p = 0.010). Again,

household size is considered important as it plays a role in maximizing access to

resources. The significance of the difference increases when looking at adult members (t









= 3.147, p = 0.002). Gender of household head does not vary significantly across the

different villages or language groups.




Table 30. Total household size by gender of household head.

Mean Median Min Max Std Dev
Male 8.8 7.0 1.0 32.0 5.2
Female defacto 7.3 7.0 3.0 12.0 3.8
Female dejure 7.0 6.0 1.0 12.0 2.7
Total 7.8 7.0 1.0 32.0 4.1



Table 31. No. of household members 15 years and over by gender of household head.

Mean Median Min Max Std Dev
Male 5.26 5.0 1 14 2.5
Female de facto 3.75 3.5 2 6 1.7
Female dejure 4.10 4.0 1 7 1.8
Total 4.62 4.0 1 14 2.2

Neither does gender of household head appear to play a role in determining the

proportional contribution of wetland resources to the household. However, there are

some differences in the extent of use of some resources. For example, there is very little

difference in the mean number of cattle owned by the different household types, although

there is a significant difference in whether livestock are grazed on the floodplains or not

(Phi and Cramer's V = 0.298, p = 0.013). Some respondents stated that floodplain

grazing is for "poorer" people who don't have cattleposts located away from the village.

While 70.5% of female dejure headed households graze their livestock on the floodplain,

only 41.2% of male-headed, and 33.3% of female defacto headed do.









Since nearly all households usually collect thatching grass, it is not surprising that

there is no significant gender difference regarding participation in this activity. Amongst

those who collected in 2001, however, there were strong differences in the mean amounts

of thatching grass collected (F = 3.066, p = 0.051), with male-headed households

collecting almost twice as much on average as female headed households.

Reed collection is similar to that of thatching grass. However, although the mean

number of reed bundles collected was greater for men, this was not significantly so.




Table 32. Bundles of thatching grass collected by household head.

Household head % h'holds Mean no. Median no. Sample sum
collecting collected collected amount (bundles)
Male 75.8% 40.1 25.0 2004
Female de facto 100% 20.5 20.0 82
Female dejure 71.2% 21.5 20.0 1119
Total 74.1% 30.2 20.0 3205



Table 33. Bundles of reeds collected by gender of household head.

Household head % h'holds Mean no. Median no. Sample sum
collecting collected collected amount (bundles)
Male 77.3% 37.2 30.0 1898
Female de facto 100% 23.5 24.0 94
Female dejure 60.3% 24.6 20.0 1080
Total 69.3% 31.0 20.0 3072


There was no discernable association with selling either thatching grass or reeds

and gender of household head. While a higher proportion of male-headed households

than female-headed households usually collected palm leaves, papyrus, and fish, this was









not significantly so. More female dejure headed households usually collected water lily

bulbs, but the difference was not significant. In terms of quantities both collected and

sold, there were no s associations between gender of household head and mean amounts.

Ethnic Variations in Reliance on Key Resources

There is a very strong relationship between ethnicity and spatial distribution (Phi =

0.624, Cramer's V = 0.361, p < 0.001). This is important because it is often the location

that explains use of a resource more than do different cultural practices. The

Hambukushu are most numerous in all settlements (see Figure 26).


40




30 7

Bakgalagadi

20I IXereku
20
SBugakhw e

II Xanikhw e
10 IHl Batsw ana


I IBayei

0,_______ I Hambukushu


Figure 26. Distribution of ethnic groups by village. The Xereku live only in the
northernmost villages, while the Bayei and Batswana live in the villages on the southwest
side of the panhandle. The distribution of the ethnic groups reflects the direction from
which they migrated. Ethnicity of the household was determined on the basis of main
language spoken in the household.