Resilient or vulnerable livelihoods? Assessing livelihood dynamics and trajectories in rural botswana

In this paper, we explore the resilience and vulnerability of livelihoods within two different social-ecological dryland contexts of Botswana over the last 30 years. We drew on primary field data sources, including oral histories, livelihood surveys, ecological surveys, as well as documented evidence of environmental, socioeconomic, and institutional dynamics to identify a broad range of activities that combine to create a range of different household livelihood outcomes. We used this information as a starting point to assess the ways in which livelihoods have changed over time, and evaluated whether they have become more resilient or more vulnerable, and considered the factors that have contributed to these outcomes. In the context of dynamic dryland social-ecological systems, we applied a livelihood trajectory approach to explore the shocks and stresses that affect livelihoods and to elucidate the characteristics of livelihood strategies that contribute to increased resilience or vulnerability. We used a vulnerability framework as a means of framing discussion about vulnerability and resilience and as a means of identifying broader insights. The research identified "accumulator", "diversifier", and "dependent" households and the ways in which they move between these categories. More resilient livelihood trajectories can be achieved if the important role of formal and informal institutions is recognized.


INTRODUCTION
Pastoralism is an important component of many rural livelihood strategies within Botswana (Dougill et al. 2010). However, livelihoods also comprise a number of other nonpastoral activities, many of which depend on a variety of different components of the natural resource base (Scoones 1996, Twyman 2001, Sallu et al. 2009). As such, many of the shocks and stresses that can destroy or damage the natural resource base can also adversely affect livelihood prospects over both the long and short term. In this paper, we investigate the resilience and vulnerability of rural livelihoods, and consider their relation to the dynamic natural resource base in two different social-ecological contexts of Botswana. In doing so, we first outline some of the key concepts related to livelihoods in terms of livelihood strategies, trajectories, resilience, and vulnerability. We next outline the research process and develop a background narrative of the environmental and livelihood systems in our study area, qualitatively determining those factors (environmental and nonenvironmental) that contribute to increased vulnerability and/or resilience. We apply Fraser's (2006Fraser's ( , 2010 vulnerability framework to help us understand these processes and to inform the direction of future interventions.

Livelihood approaches, resilience, and vulnerability
Chambers and Conway (1992) define a livelihood system as comprising the capabilities, assets (including both material and social resources), and activities required for a means of living. The chosen combination of assets and activities, undertaken usually at the household level, is often referred to as the household's "livelihood strategy". A livelihood strategy encompasses not only activities that generate income but many other kinds of elements, including cultural and social choices http://www.ecologyandsociety.org/vol15/iss4/art3/ (Ellis 2000). Livelihoods approaches illustrate how, in different contexts, sustainable livelihoods can be achieved through access to a range of livelihood assets (e.g., natural, social, financial, physical, and human capital) which, within the context of personal, institutional, and environmental provisions and constraints, are combined in the pursuit of different livelihood strategies. Within the sustainable livelihoods framework (Chambers andConway 1992, Scoones 1998), context is framed within the "vulnerability context", which includes issues of "seasonality", "trends", and "shocks".
Carney (1998) explains that "a livelihood is sustainable when it can cope with and recover from stresses and shocks and maintain or enhance its capabilities and assets both now and in the future, while not undermining the natural resource base". This interpretation of sustainability relates strongly to definitions that consider the "resilience" of social-ecological systems. Walker et al. (2006) define resilience as "the capacity of a system to experience shocks while retaining essentially the same function, structure, feedbacks, and therefore identity". As Marschke and Berkes (2006) suggest, "resilience offers a lens with which to explore stresses and shocks and to understand livelihood dynamics", and is "future oriented, and is used to characterize a system's ability to deal with change". By stresses, we mean "enduring shifts" (such as seasonality and trends) and shocks "transient disruption" (Leach et al. 2007). Incorporation of ideas surrounding resilience alongside understandings of vulnerability can contribute an essential temporal dimension to analysis, allowing the combinations of strategies and circumstances that move households towards more resilient outcomes over time to be identified, ultimately enabling them to embrace change as a result of shocks and/or stresses as opportunities for innovation and accumulation (Folke et al. 2002). In this paper, a focus on resilience can help us learn from the past to inform future planning. Fraser et al.'s (2010) vulnerability framework draws on several elements of the livelihoods approach, and in the context of this research stimulates the following questions: q Does the agroecosystem have the resilience to remain productive in a changing vulnerability context? q Do people have access to livelihood strategies that allow them to survive changes to the vulnerability context? q Do the institutions have the ability, capacity, and willingness to respond to a changing vulnerability context, especially in crisis situations?
In this paper, we use this framework to inform discussion of the direction and dynamics of livelihoods over a 30-year period. Through comparative research we provide a rich contextual narrative and use it to explore those factors that in isolation and combination push livelihoods along particular "trajectories" towards vulnerability or resilience. Bagchi et al. (1998) use the term "livelihood trajectories" to describe and explain the direction and pattern of livelihoods of individuals or groups of people (e.g., households). A livelihood trajectory approach allows the examination of an individual household's "strategic behavior that is embedded in a historical repertoire, in social differentiation" (de Haan and Zoomers 2005), and in perceptions of risk. Such an approach is sensitive to life histories (an individual's own "story" of their changing livelihoods). A focus on livelihood trajectories allows a deeper penetration into the beliefs, needs, aspirations, and limitations of people's lives, but one that is also contextualized in relation to power and institutions (de Haan and Zoomers 2005). An increasingly important application of the livelihood trajectory approach is in exploring the shocks and stresses that can affect livelihoods, as well as in elucidating the characteristics of the overall livelihood strategy that contribute to increased resilience or vulnerability.

METHODS
Data were collected in 2004 and 2005 when fieldwork was carried out as part of a larger research project that considered environmental, socioeconomic, and institutional dynamics in two of Botswana's remote rural settlements, Khawa and Kedia settlements in Central and Kgalagadi Districts, respectively (Fig. 1). These settlements were chosen for comparison because they were of similarly low economic status and were classified by the Ecology and Society 15(4): 3 http://www.ecologyandsociety.org/vol15/iss4/art3/ A mixed-method approach was taken in collecting the data. Methods used included oral histories and in-depth livelihood trajectory mapping exercises (n = 17), as well as household-level livelihood and resource use surveys (n = 98). These sought to identify the ways in which households use their environment, how environmental changes (drought, land degradation, etc.) affect livelihood decisions, Ecology and Society 15(4): 3 http://www.ecologyandsociety.org/vol15/iss4/art3/ and how environmental factors interact with broader socioeconomic and political processes to determine resource use outcomes and impacts on livelihood systems. Repeated vegetation and wild animal surveys were conducted before and after rains, and time-series sets of Landsat images and wild animal aerial count data records were collected from the Department of Surveys and Mapping and the Department of Wildlife and National Parks. Soil and climate data were collected from the Department of Surveys and Mapping and the Department of Meteorological Services, respectively (see Sallu [2007] for a more detailed outline of the methodology and data). Environmental change data were then analyzed in conjunction with livelihood trajectory results in order to elucidate the key dynamics of relationships between livelihoods and the natural resource base. The average time span covered by the investigation was limited to the 30 years between 1974 and 2005, primarily due to restrictions on the availability of climate, soil, vegetation, and large wild animal data. Consideration of this 30-year temporal frame nevertheless permitted incorporation of the periods of formal settlement establishment, which proved important in setting the boundaries of the livelihood context.
Data analysis was conducted throughout the period of information gathering. Initially, this was at a descriptive level in order to note any trends in the data, but it progressed to a more detailed level as both qualitative and quantitative social and environmental information was drawn together. Qualitative data were coded through processes of indexing the data under emerging themes. This permitted the identification of the factors that played an important role in the construction of livelihood strategies. Consistent triangulation of the results highlighted any contradictions and similarities in the different data sources. Where contradictions were found, further iterative reflection took place in the form of focus groups in order to ascertain why and how the conflicts in information may have occurred. This became a circular process that led to inductive interpretation and explanation as the ecological information was gradually juxtaposed within the emergent socioeconomic context. Quantitative data sets were analyzed using multivariate statistics. Livelihood and environmental data were classified using cluster analysis, and correlations were tested using principal components analysis. Landsat images were classified using ERDAS Imagine V.9 software and landscape-level changes were detected from raster attribute comparison (see Sallu [2007] for a more detailed outline of data analysis procedures). Based on this analysis, we aimed to identify contemporary strategies and the nature of trajectories to which they led. In doing this, we also identified the key changes to the vulnerability context and the combination of factors that have led to more resilient or vulnerable livelihood outcomes.

Social-ecological dynamics
The productivity of the natural resource base in dryland Botswana is exceptionally dynamic, with the provision of ecosystem goods and services largely determined by the extreme environmental conditions that affect water, soil, and landscape form. These include large diurnal and seasonal temperature variations, low average annual rainfall (ranging from a minimum of less than 200 mm in the southwest to a maximum of more than 800 mm in the northeast of the country), frequent and extended periods of drought (caused by cyclical [multidecadal] climate factors (Figs. 2 and 3; Shaw 1991, Batisani andYarnal 2010), sporadic heavy rainfall events (Bhalotra 1987, Goomes andPetrassi 1996), edaphic (soil related) variation (Ministry of Agriculture 1990), and hydrologic flux (Fig. 4;du Plessis andRowntree 2003, McCarthy et al. 2003). In particular, both natural and human-influenced riverine and lacustrine dynamics affect the ecology of landscapes associated with Okavango and Makgadikgadi systems in northern Botswana. These therefore have important impacts on our Kedia study area (Fig. 1). Soils in Khawa and in much of Kedia exhibit low fertility with limited key nutrient contents (nitrogen, phosphorous, potassium). Indeed, much of the country's northern, central, and southern regions are unsuitable for cultivation (Buckley et al. 1987, Tolsma et al. 1987, Ministry of Agriculture 1990. Despite this, soil heterogeneity can be diverse, leading to vegetative diversity equal to that of other savanna areas in Africa (Thomas and Shaw 1991). The resultant human responses to dynamism are manifest in the flexibility of livelihood activities. These are structured in such a way that people are able to take advantage of changing availabilities of and access to natural assets. In the rural settlements in our study areas, livelihoods are highly dependent on biodiversity (Sallu et al. 2009), and the environmental dynamics outlined above create both opportunities for and threats to the livelihood strategies that are followed at different times. For example, soil heterogeneity plays a significant role in determining landscape and species diversity. In Kedia, these dynamics have led to a diverse ecological landscape that is in stark contrast to Khawa where the landscape is edaphically homogeneous and less species diverse. The situation in Kedia thus offers more natural resource-based livelihood opportunities than are found in Khawa, across a similar sized landscape.
In both settlements, livelihood activities are strongly influenced by the spatial and temporal variability of rainfall. In Khawa and in far southwestern parts of Kedia, this results in patches, hot-or cold-spots of abundance, which punctuate the landscape and create spatially and temporally distributed opportunities for gathering and hunting. For example, after heavy rain in 2004, hotspots of water-rich veld fruits such as wild watermelon (Citrullus lanatus) and cucumbers (e.g., Cucumis africanus and Acanthosicyos naudinianus) were witnessed in patches of duneveld beyond 13 km from Khawa (Fig. 5). These particular hotspots provide water and nutrient-rich resources for wild animals, livestock, and humans during the dry season. If accessible, these hotspots provide valuable opportunities in times of inner-settlement food and water scarcity, thus making an important contribution to wider livelihood systems (Fig. 6).

Environmental change
The inherent dynamics described above create diversity in otherwise homogeneous and speciespoor semi-arid landscapes, and provide spatially and temporally limited opportunities for different livelihood activities. However, longer term and larger scale environmental changes have altered the vulnerability context over the past 30 years and have influenced the livelihood trajectories that households have followed. This section considers five significant environmental changes that have occurred in either one or both of our study settlements over this time period.
1. Mid-1980s drought. Both settlements experienced prolonged drought in the 1980s, beyond the inherent rainfall variability that characterizes dryland environments (Fig. 2). This led to a significant reduction in the diversity and productivity of vegetation (Sallu 2007), and rapid declines in wild animal populations (Table 2; Williamson and Mbano 1988), which limited both the potential for hunting and the availability of grazing and fodder resources for pastoralism. While the ecology in these dryland systems has evolved to withstand drought, the prolonged nature of this event resulted in changes that exceeded the magnitude and extent of that associated with normal rainfall variability for the area. This drought created a shock within the dynamic system that affected both of our study settlements.
2. Late onset of rainfall. While no significant longterm change in average annual rainfall occurred between 1974 and 2004 in either settlement, data obtained from the Department of Meteorological Services show that peak monthly rainfall in Khawa started to fall an average of one month later than usual after 1984. Peaks in mean monthly rainfall between 1985 and 2004 fell in February, March, and April compared to January, February, andMarch in 1975 to 1984 (Fig. 3). Above average levels of rain fell in 2000 with the largest amounts of rainfall occurring in April. This has important implications for the productivity of vegetation growth (Tadross et al. 2005). If peak rainfall periods occur late in the season, low winter temperatures and frosts (usually in June, July, and August) may kill plants before they are fully mature. This has detrimental knock-on effects on the production of human foods, such as wild herbs and fruits, wild medicines, and plant-based materials used as building products and for crafts, and on rain-fed cultivation and the availability of grazing and browse resources for cattle and wild animals. This was an issue of particular concern in Khawa settlement, where diurnal temperatures were greatest. While it is difficult to conclusively link this stress to global climate change due to the inherent dryland rainfall dynamics described above and Fig. 4. Drying of the River Boteti as illustrated by the deviation from the mean annual volume flow (271 m 3 x 10 6 ) at Maun . (Source: Sharing Water Okavango/Kubango database http://www. sharing-water.net/) limited data with sparse spatial coverage, Tadross et al. (2007) show increasing evidence that there may be links between climate change and the reduced length of rainfall seasons in southern Africa.
3. Increased unpredictability of rainfall. In Kedia, there was a clear trend towards increased annual rainfall variability after 1996, with greatest volatility noted during the most recent time period, 1996-2004 (Fig. 2). Increased variability has resulted in increased unpredictability of rainfalldependent natural capital, which has resulted in years of either boom or bust with little in between. In particular, this stress on the system has increased the risks associated with rain-fed cultivation, which has resulted in significant impacts on the provision of livelihood opportunities. No such trend was seen in Khawa.
4. Drying of Lake Xau. During the mid-1980s, in the context of changes to rainfall patterns and prolonged drought, Kedia experienced significantly reduced water flows in the Boteti River (Fig. 4). Coupled with infrastructural developments in the river channel upstream (Zufferey 1983), this contributed to the complete drying of Lake Xau by 1984 (Cashdan 1985). Although in some years the lake has been known to dry up, to date the waters have not returned. This has resulted in the eradication of a seasonal surface water resource and the extinction of fish and water-dependent species such as hippopotamus (Hippopotamus amphibus), crocodile (Crocodylus niloticus), reeds ("lethaka"), and water lily (Nymphaea nouchali) in the area. It has also caused significant changes in vegetation composition, structure, and functioning, and has resulted in the cessation of flood recession cultivation (Sallu 2007). In turn, this added stress Ecology and Society 15(4): 3 http://www.ecologyandsociety.org/vol15/iss4/art3/ has had an important impact on the provision of livelihood opportunities. 5. Land degradation. In both settlements, land degradation (manifest as soil erosion and compaction, salination, and vegetation changes at a variety of scales) has also become an emerging environmental constraint through the associated decline in productivity it represents. Data from vegetation surveys illustrated that degradation was most typically recorded near settlements, cattle posts, watering holes (boreholes or wells), and transport routes (roads and tracks) (Sallu 2007). As was seen in Kedia during the mid-1980s, temporary deterioration of the land was also commonplace, particularly in areas seasonally frequented by large numbers of wild herbivores (Sallu 2007). This additional system stress meant that herders had to travel further afield to access water and suitable grasses for their cattle, while opportunities for hunting and arable cultivation as livelihood activities also decreased. http://www.ecologyandsociety.org/vol15/iss4/art3/ Fig. 6. Villagers separating Citrullus lanatus flesh and seeds at a temporary camp, 40 km from Khawa settlement, August 2004. Flesh is heated to form water, and the seeds are pounded into flour for porridge.

Contemporary livelihood strategies
In the context of these environmental dynamics and livelihood struggles, and despite significant socioeconomic and cultural differences within settlements (Table 1), three major groups of households, conducting similar or related livelihood activities, were identified in 2004/5 (Fig. 7). These strategy groups were classified using cluster analysis of household livelihood survey data. A description of the key characteristics of the households and livelihoods of each cluster group is presented in Table 3. Principal activities that determined strategy differentiation based on principal components analysis can be linked to Fraser et al.'s (2010) vulnerability framework, and included (1) ownership of livestock (Fraser et al.'s "access to assets"), (2) cultivation of arable crops (Fraser et al.'s "capacity of ecosystem to remain productive"), (3) reliance on government-provided social security benefits [1] (Fraser et al.'s "strength of formal and informal institutions"), and (4) permanent and temporary employment (Fraser et al. 's "access to assets") ( Table 3). Table 2. Population change of selected wild animal indicator species between the 1970s and 1994 for the southern Kalahari system associated with Khawa (data sources: DHV [1980] and DWNP [1994a, b]; modified from Boggs [2000]) and the northern Kalahari system associated with Kedia (data sources: Van Der Maas [1995] and Bonifica [1992]). CKGR = Central Kalahari Game Reserve; MPNP = Makgadikgadi Pans National Park. "-" refers to no data available.
A more varied strategy, undertaken by what we have called "diversifiers", was followed by 26% of households in Khawa and 25% of households in  Table 3 for full description of cluster group livelihoods. Livestock units: 1 cow = 1 LSU; 5 goats/sheep = 1 LSU (Sallu 2007).
Kedia (Table 3, Fig. 7). These households tended to distribute effort across multiple livelihood activities, with lesser tendency towards specialization (reflected by moderate livestock units, Fig. 7). While these households were typically composed of a range of ethnic groups, minority groups, such as Bakgalagadi in Khawa and Banajwa and Bakwena in Kedia, were absent.
Finally, a "dependency strategy", undertaken by "dependents", was followed by the remaining households. Households within this group were characterized as smaller than average in size, with low (e.g., 10) to zero livestock units, and were highly dependent on social security benefits. In Kedia, household members were also frequently employed as herders or laborers for other members of the community, often based outside the settlement (Table 3, Fig. 7). This category comprised the majority of households in both settlements (Fig. 7), and households were typically composed of a range of ethnic groups. In both settlements, this included minority groups, and in Kedia, involved a high proportion of Basarwa (Table 3).

Livelihood trajectory analyses
As a means of teasing out the relative importance of each of the multiple interacting factors that resulted in the contemporary livelihood strategies presented above, detailed investigation of livelihood trajectory data for 17 households across the two settlements was conducted for the 1974-2005 period. Some of the trajectories encountered are illustrated through the narrative cases presented in Table 4. While Table 4 clearly illustrates that trajectories were unique to households, some generic trends were evident. These trends are described below. http://www.ecologyandsociety.org/vol15/iss4/art3/  In 2004, due to low rainfall and degradation around the settlement, grazing and water resources were limited. Due to his access to transport in the form of a donkey cart and car (which facilitated the transport of water and food for those caring for the livestock), Thau sent his livestock 40 km west to access an area of rangeland that had benefited from rain earlier in the season (Fig. 5). This area provided sufficient wild watermelons (Citrullus sp.) and grazing to sustain his and other household herds for up to three months during the dry season. The extra nutrients gained from these resources enabled his goats to reproduce twice in that year, leading to rapid growth in numbers.
Accumulative trajectory -building resilience T1, 1985 -Limited assets, low capacity of agroecosystem to remain productive, some collective capacity to cope through social security benefits

R1. Diversification of livelihood activities
T2, 1990s -rapid accumulation of financial, physical, and human assets, moderate capacity of agroecosystem to remain productive (rainfall higher but degradation increasing), increasing individual and collective capacity to respond to crises as he became Councillor and his children got older

R2. Salaried employment R3. Investment in livestock accumulation R4. Investment in transport R5. Investment in access to water
T3, 2000s -continued accumulation of financial, physical, and human assets (children gained jobs), decreasing capacity of agroecosystem to remain productive, capacity to respond to agroecosystem decline high due to access to transport and private water source

Case 2 -Mr. Mpoelang, Khawa
Batlharo, aged 74, living with his wife, an adult son, and two children (one of whom has Down's syndrome) Mpoelang was one of the first settlers in Khawa in the late 1970s. At that time, the rains were so plentiful and appropriately timed that he was able to cultivate melons and beans near the settlement. He also hunted for wild meat and made leatherworks, and the household sold the leathers and dried meat (biltong) across the border into South Africa.

Fluctuating trajectory
T1, late 1970s -high capacity for agroecosystem to remain productive (high and appropriately timed rainfall, low numbers of people and livestock, and no degradation), high levels of access to natural assets (horses), moderate to high accumulation of financial or physical assets (car, livestock), moderate capacity to respond (social networks, no trade or resource-associated restrictions, no formal institutions, transport) (con'd) Ecology and Society 15(4): 3 http://www.ecologyandsociety.org/vol15/iss4/art3/ At that time, Mpoelang had 30 cows, 18 horses, and a car. In the mid-1980s, trade was no longer allowed across the border, and he ceased cultivation due to the drought and inappropriate timing of rainfall. In 1984, he also lost 15 cows and 18 horses to drought. In 1987, his household was not allocated a Special Game Licence, and as a consequence they were no longer able to hunt. Despite this, Mpoelang was still able to make leatherworks by purchasing skins from those who were hunting, and when the licence for selling leatherwork products was introduced, he purchased one. In the early 1990s, Mpoelang's livestock numbers slowly began to recover. There was, however, a setback in 1993 when eight of his cows were struck by lightening. Between 2000 and 2003, Mpoelang lost an additional 11 cows to lion predation. Financial compensation from the Wildlife Department, access to pensions, and the sale of horses for cows allowed some recovery of cattle stocks. In 2000, Mr. Mpoelang's household started to benefit from the quota system, with a small share of meat from the community-allocated hunt available for his household. In 2003, the household successfully cultivated rain-fed melons on a small scale near the house. In 2004, the household retained three horses and seven cows. The child with Down's Syndrome did not receive support from the government.

R1. Accumulation of livestock in high rainfall years R2. Cultivation of crops in years of appropriately timed rainfall
T2, mid-1980s -low capacity of agroecosystem to remain productive (prolonged drought, inappropriate timing of rainfall), limited natural assets (drought), some physical and financial asset stores (leatherworks), moderate capacity to respond (reduced from high by permits restricting hunting and by border restrictions to trade)

V1. Loss of livestock V2. Loss of livelihood activity
T3, 1990s -moderate capacity of agroecosystem to remain productive, fluctuating financial, physical, and natural assets, changing institutional support to assist household's capacity to respond T4, 2000s -some capacity of agroecosystem to remain productive (ability to cultivate), some financial and physical assets (horses), moderate capacity to cope (trade horses for cows), high capacity to cope (access to pensions, compensation, quota meat, and pension)

Case 3 -Mr. Bakghotu, Khawa
Bakgalagadi, aged 66, living with his wife and no children Bakghotu first came to Khawa in 1974 with his wife. At that time he was hunting and he and his wife were making leatherworks from the skins of bateared fox (Otocyon megalotis), black-backed jackal (Canis mesomelas), and springhare (Pedetes capensis). Bakghotu and his wife were making cushions and rugs and selling them to people for money. In addition, they were collecting veld products for food and wild medicines for medicine and sale. Bakghotu was also engaged in woodcrafts. Between 1980 and 1992, as a destitute, he and his wife also received a Special Game Licence, which allowed them to hunt. Also, the remote area dweller (RAD) program provided his household with food rations, clothing, and blankets. Between 1987 and 1988, Bakghotu was also employed for a short time on piece jobs. These combined activities allowed him and his wife to purchase a few livestock -goats, donkeys, and horses. Bakghotu described himself as making "a very nice living up until the changes in licences". From 1989, Bakghotu was no longer able to sell his leatherworks without a licence, and from 1992 when the Quota Hunting Licenses were introduced, he was no longer able to hunt for himself. Even in more recent years, a licence has been introduced to prevent the sale of devil's claw (Sengaparile -a medicinal plant), and the other veld products Bakghotu used to collect for food, medicine, and crafts are now located far from the settlement due to increased degradation. These combined restrictions have severely limited the range of activities he is able to practice for his livelihood, and with the loss of his livestock due to drought and predation, he is now dependent on the government. He relies heavily on the old-age pension and destitute rations but complained of shortages in the destitute rations and reductions in the money provided to those in need.

Degenerative trajectory -vulnerable
T1, 1974 -high capacity for agroecosystem to remain productive, high levels of access to natural assets, moderate accumulation of financial and physical assets (livestock), moderate capacity to respond (asset stock)

R1. Engagement in diverse livelihood activities
T2, 1980s -low capacity of agroecosystem to remain productive, retained access to some natural assets (hunting permit), limited access to other natural assets (drought), some physical and financial asset stores, moderate capacity to respond (RAD program support)

V1. Loss of livestock
T3, 1990s -moderate capacity of agroecosystem to remain productive, loss of financial and physical asset stores, loss of access to natural assets (permit changes), increasing reliance on government support

Case 4 -Mr. Mathoa, Kedia
Bakalanga, aged 51, widowed, living with three youth (two present and one absent at time of survey), and three children Between 1976 and 1979, Mathoa worked in Orapa. At this time, he had very few livestock. In 1979, when he finished working, he invested the money he had saved into livestock, mainly cattle and goats, the digging of a well, and some cultivation. Between 1979 and 1984, the number of livestock units he owned increased from 19 to 57 due to the availability of good grazing, browse, and access to water. He also managed to accumulate land. However, between 1984 and 1994, he lost 50% of his livestock to drought, and no cultivation was possible. During this time, Mathoa gained employment with the Land Board and married a woman who was employed as a nurse at the clinic. The financial capital accumulated from both forms of employment allowed re-investment in livestock and land after 1994. By 2005, he had 15 hectares of land (12 hectares on one side of the settlement, three on the other side). The number of livestock units Mathoa accumulated peaked at 72 in 2000 after a particularly good rainfall year. Cultivation of sweet reed, maize, sorghum, watermelon, and beans was also possible that year. Food was generated for both subsistence and sale. During this period, Mathoa finished working with the Land Board and was retrained as a welder. He also received small amounts of money at this time in his role as Village Development Committee (VDC) chairman. The income generated through self-employment as a welder and from the VDC enabled him to maintain his livelihood status despite the death of his wife in 2004 and a drop in livestock units to 53 in 2005 after two successive years of below average rainfall. Some years (e.g., 2001), he was unable to harvest due to pest attack on crops.
Stable trajectory -retained resilience T1, late 1970s-early 1980s -high capacity of agroecosystem to remain productive, access to and accumulation of financial and physical assets, moderate capacity to respond to change R1. Salaried employment R2. Accumulation of livestock R3. Improved access to water R4. Accumulation of agricultural land T2, 1984-1994 -low capacity of agroecosystem to remain productive (prolonged drought and lakeside degradation), loss of some physical assets (livestock), high capacity to respond (employment, powerful social network)

V1. Loss of livestock
T3, 1994-2005 -moderate capacity of agroecosystem to remain productive (some rainfall but increasing unpredictability of rainfall, degradation), accumulation of financial and physical assets, high capacity to respond (self-employment, retained and expanding social networks)

Case 5 -Mr. Mmegwa, Kedia
Bakalanga, aged 76, married and living with one wife, four youth, and five children Mr. Mmegwa worked for six years in South African mines. Between 1951 and1975, he lived in Xhumo and Beetsao. He was a healthy man who owned land and small shops. His household used to cultivate enough to sell and managed to dig a well from the money this generated. Since their move to Kedia in 1975, they have not experienced a good harvest. Instead, livelihood activities at this time specialized in livestock farming and small amounts of hunting around the lake. Livestock units in this household peaked at 163 in 1991. Mr. Mmegwa stopped hunting around this time because he considered himself to be rich. In 1993, however, Mr. Mmegwa's well was stolen and sold by someone else without his knowledge. This resulted in the death and/ or loss of all of his livestock and the reliance on the only other livelihood activity that contributed significantly to the household at that time -his wife's brewing and sale of alcohol. This activity protected the household livelihood from collapse, and over time, in combination with the collection and sale of medicinal plants and Mr. Mmegwa's monthly receipt of an oldage pension, generated enough money to purchase livestock and access water once more. By 2005, the numbers of livestock in the household had reached 11 cows, 16 goats, six donkeys, two horses, and 10 chickens. Despite several of Mmegwa's children being away from the settlement, either working or studying, none of them send remittances.
Fluctuating trajectory with shift from accumulatordiversifier strategy T1, 1975-1991 -fluctuating capacity of agroecosystem to remain productive, access to and accumulation of natural assets, high capacity to respond to change (financial and physical capital stores, social network outside settlement)

R1. Accumulation of livestock
T2, 1993 -moderate capacity of agroecosystem to remain productive (some rainfall but increasingly unpredictable, degradation, lake dry), loss of access to water, moderate capacity to respond (skills and knowledge to diversify activities)

V1. Loss of livestock R2. Diversification of livelihood activities
T3, 2005 -moderate capacity of agroecosystem to remain productive (some rainfall but increasingly unpredictable, degradation, lake dry), moderate access to assets, moderate capacity to respond (diverse skills, access to government support)

Case 6 -Mr. Baitsile, Kedia
Basarwa, aged 57, married and living with one wife, one other adult, three youth, and eight children During the 1970s, Baitsile's household livelihood was composed of a range of livelihood activities including fishing, hunting, cultivation, livestock farming, and veld product collection. Of greatest significance at that time was cultivation and veld product collection. In a good season, he was able to cultivate up to ten 50-kg bags of maize, ten 50-kg bags of sorghum, and eight 50-kg bags of beans, as well as plentiful supplies of pumpkin (maputse), green melon (marotse), and watermelon (moghapo). Veld products collected after rain included the fruits of Grewia species (mogwana, moretlwa/ moseme, motsotsojane) and moretologa (Ximenia americana), as well as the Mophane worm. Baitsile was able to sell some of these natural products. However, he described how in more recent time periods (1990 onwards) veld products had become increasingly scarce near the settlement due to the numbers of people collecting such resources. During the mid-1980s, Baitsile was unable to cultivate due to poor rainfall, he was no longer able to fish due to the lake drying, veld products were significantly reduced due to low rainfall, and he was no longer able to hunt due to permit changes. His livestock unit declined from 13 in 1981 to none in 1991. In 1991, the only activity that kept the household going was government-provided employment in the form of piece jobs laying water pipes. Between 1997 and 1999, Baitsile was employed by the Community Hunting Project. In 2000, after this project had ceased, the government gave Baitsile 15 cows. The number of livestock has remained similar since. In 2005, Baitsile gained employment as a night watchman at the settlement shop.
Fluctuating trajectory T1, 1970s-early 1980s -high capacity of agroecosystem to remain productive, access to and accumulation of natural and financial assets, moderate capacity to respond to change R1. Accumulation of a variety of natural asset stocks R2. Cultivation of range of crops R3. Supplementary collection of wild food T2, mid-1980s -low capacity of agroecosystem to remain productive (drought, degradation, dry lake), reduced access to natural assets (permit changes, drought, dry lake), no ability to respond to change (low to no asset stocks)

V1. Loss of livelihood activity(ies)
T3, 1990s-2005 -moderate capacity of agroecosystem to remain productive (some rainfall but increasingly unpredictable, degradation, lake dry), some access to natural assets, moderate capacity to respond (government support and employment)

Case 7 -Mr. Charlie, Kedia
Basarwa, aged 34, married and living with one wife and two young children; lives adjacent to his sick mother and elderly stepfather When Charlie was growing up (late 1970s-early 1980s), the family's food came from the lake. They used to eat reeds and fish. When the lake dried, life changed, and he can remember being hungry as a child. Charlie was given one cow and one goat by relatives when he started his own household, and by 2005 there were five cows and four goats. Some died that year due to a lack of water. Charlie currently lives approximately 10 km outside the center of the settlement in an area of Mopane woodland. He spends much of his time now (2005) helping his sick mother care for her livestock, collecting and preparing medicines and woodcrafts, and illegally hunting for small game. Cultivation has not been possible since 2000 due to the irregular rainfall, and even when it is possible, his household has only a small amount of land to cultivate. Charlie's household has access to a well, which is a shared resource with other Basarwa families. His children are cared for by the government during school term time. They come home during holidays. The elder parents both receive a pension.

Diversified trajectory -increasing resilience
T1, late 1970s-early 1980s -high capacity of agroecosystem to remain productive, access to and accumulation of natural assets, low capacity to respond to change

V1. Use of wild foods (lake products)
T2, mid-1980s -low capacity of agroecosystem to remain productive (drought, dry lake), reduced access to natural assets (drought, dry lake), some ability to respond to change (family support, livestock)

V2. Loss of livestock V3. Loss of wild lake foods
T3, late 1980s-2005 -moderate capacity of agroecosystem to remain productive (some rainfall but increasingly unpredictable, degradation, lake dry), good access to natural assets (living outside settlement), high capacity to respond (diverse skills, family support, access to government support)

Khawa settlement
Between 1974 and 2005, trajectories of elite accumulator households in Khawa were primarily aimed at building up asset bases, with periodic peaks and troughs due mainly to a gain or loss of employment, livestock disease, and/or drought (e. g., Case 1, Table 4). Asset accumulation in Khawa focused on the conversion of employment income into physical assets, primarily the expansion of livestock herds, and investment in improved access to water and transport. It is well known that in some arid and semi-arid environments (in particular those where stock mortality is density independent) the maximization of stocking densities helps ensure long-term survival after drought stock loss (Campbell et al. 2006, Barrow et al. 2007). With only a small number of elite accumulators and an expansive communal rangeland landscape, which is largely functionally intact beyond 4 km of the settlement, cattleposts, or boreholes, this accumulative strategy led these household livelihoods to become increasingly resilient to environmental changes over time. In addition, investment in improved access to water increased household resilience to drought-induced water stress. Access to transport (in particular a vehicle) improved resilience to degradation as well as drought. The availability of transport improved access to areas outside the settlement that were rain-fed and/or most ecologically intact. Transport also facilitated access to external institutions and distant markets for the direct sale or purchase of goods. As socially and politically elite members of the community, members of accumulator households were best placed to predict, monitor, and adapt to economic and institutional changes, and therefore were most likely able to position themselves to achieve the most resilient trajectories in the face of environmental change.
Over the same time period, many households in Khawa that followed a dependency strategy had not demonstrated such resilience in their livelihoods (e. g., Case 3, Table 4). These households were especially vulnerable to the combined impacts of drought and degradation, particularly over the past 15 to 20 years, with many failing to recover from the prolonged mid-1980s drought. A lack of financial income and limited access to water and transport coupled with a range of institutional and policy-related changes (one of the factors that Fraser et al.'s vulnerability framework identifies as a key for the successful management of environmental shocks) meant that these households had not been able to overcome drought shock or degradation stress. The most significant institutional and policy changes that have affected these households in Khawa have been (1) changes to the hunting licensing system on hunting-associated livelihoods, (2) settlement-specific trends towards elite capture of productivity hotspots and water resources in the settlement, (3) failure in the effective provision of government support, and (4) breakdown of social capital within the community. Both (1) and (2) have reduced nonaccumulator households' access to water and ecological diversity, which has limited their options with regard to the livelihood activities they can pursue and has increased their vulnerability to drought and degradation. Shifts to a new quota licensing system in Khawa have provided more equitable community-wide access to wild animal (hunting) resources but have preferentially benefited elite accumulator households, who were previously expected to purchase licences, and negatively affected the less successful households, who now experience more limited access to this resource and more limited opportunities to generate income and/or obtain food from it (e.g., Cases 1, 2, and 3, Table 4). While government support prevented many dependent households from experiencing a total loss of resilience (e.g., a shift from a dependent strategy to a strategy that might be labeled as a "leaver" -a household that is forced to leave the area and settle elsewhere to gain a new livelihood), the failure to effectively provide destitute (Case 3, Table 4) and disabled community members (Case 2, Table 4) with government support in this community (the reasons for which are unclear) failed to create opportunities for improved livelihoods or trajectory shifts among many households. In the context of fluctuating agroecosystem conditions, ineffective distribution of support from the government, elite capture of natural assets, and the general breakdown of social capital within the community (e.g., breakdown in inter-generational and intra-family support [e.g., Case 5, Table 4]), several diversifier households also experienced a downward trajectory shift from diversifier to dependent over the study period.

Kedia settlement
Household livelihood trajectory trends in Kedia differed from those of Khawa due to both social and environmental differences between sites. Since the 1970s in Kedia, trajectories of accumulator households, composed entirely of the Bakalanga http://www.ecologyandsociety.org/vol15/iss4/art3/ ethnic group (Table 3), have consistently retained a tendency to specialize in livelihoods built upon pastoralism and arable cultivation. Similar to accumulator households in Khawa, these households have been subject to peaks and troughs in their livelihood activities over time. They have invested income from salaried employment in accumulating livestock and gaining access to water and land as a means to build up asset stocks (e.g., Case 4, Table 4). The trajectories of the accumulators in Kedia, however, demonstrated greater vulnerability to climate shocks over the same time period. With less opportunity for elite capture, only one accumulator household studied in Kedia (Case 4, Table 4) was able to re-accumulate livestock to pre-1980s drought levels by 2000, and even though re-accumulation had been possible, it had occurred at a much slower rate than in Khawa. Livestock assets of all accumulator households declined due to drought-induced starvation and/or thirst, which was compounded by the drying lake and increased degradation. Over time, restrictions to the area available for grazing due to the introduction of veterinary fencing dictated by government policy (1988 and 1996) and increased competition from nearby settlements has led to increased grazing pressure, which has reduced the ability of the agroecosystem to remain productive. This has created or increased the likelihood of density dependent livestock mortality. Coupled with the amplified risks associated with cultivation under less predictable rainfall regimes and dry lake conditions, it is not surprising that accumulator household livelihoods have become increasingly vulnerable to climate-induced shocks. For example, since the mid-1980s drought, successful large-scale cultivation in Kedia has been achieved only once, in 2000 (Cases 4 and 7, Table 4).
The impacts of this vulnerability on livelihood trajectories in Kedia has caused some accumulator households to shift to a more diversified strategy (e. g., Case 5, Table 4). In some cases, however, reduced human capital (e.g., as a result of withinhousehold sickness, ageing, death, or outmigration of household members) and/or reduced social capital (e.g., reduced family support, trust, and/or reciprocity), and policy changes have compounded the extent to which a strategy can remain resilient. As such, factors beyond the environment clearly also exacerbate the stress and cause shifts in livelihood trajectories. It is therefore apparent that Fraser et al.'s (2010) "capacity of agro-ecosystems to remain productive", "capacity for individuals to adapt based on access to assets", and the "strength of formal and informal institutions" have all proved relevant in determining the direction of these households' trajectories.
Households in Kedia that were characterized as following diversified and dependency strategies also followed post-1980s downward trajectories but showed greatest resilience to climate-induced shock and/or stress. These less accumulative and less specialized livelihood strategies and/or those that were reliant on the effective functioning of social security mechanisms (either or both government and traditional) had fewer accumulated assets to lose and had access to a wider range of livelihood activity or substitution options, which facilitated the absorption of shock effects and prevented strategy shifts. Indeed, over the historical period covered by our study, the livelihood strategies of Basarwa in Botswana experienced a long-term shift towards diversification. Since formal settlement establishment in 1978, many Basarwa in Kedia have become increasingly involved in pastoralism and cultivation as well as in maintaining their more traditional focus on hunting, gathering, and fishing (e.g., Case 7, Table 4). As opportunities for fishing ceased when the lake dried in the mid-1980s, and when restrictive hunting permits were introduced, skills in crafts, gathering, and inconspicuous (illegal) forms of hunting, which are typically common among the Basarwa, provided opportunities for greater livelihood resiliency. In contrast, livelihoods that were solely dependent on more climate-sensitive activities -e.g., strategies that specialized in cultivation and/or pastoralism, which are common among the Bakalanga -were typically more vulnerable. The practice of a diverse range of activities therefore helped buffer the stresses and shocks of the 1980s and limited the overall impact on livelihood trajectories.
The government's effective provisioning of financial, nutritional, and educational support to children, destitutes, orphans, and the elderly in Kedia, coupled with strong traditional social security mechanisms (which were of particular importance to the livelihoods of Basarwa and Bateti groups -e.g., Cases 6 and 7, Table 4) led none of the diversifier or dependent households to experience such a radical loss of resilience that they were pushed to leave the settlement. Family and friendship sharing networks (Silberbauer 1981, Kent 1995 outside the settlement, and in many cases beyond the veterinary fences surrounding the http://www.ecologyandsociety.org/vol15/iss4/art3/ settlement (e.g., Case 7, Table 4), buffered the livelihood impacts of agroecosystem stress caused by degradation and drought. Such mechanisms allowed family members or friends to provide support through the sharing of each other's advantage or disadvantage (e.g., a gain or loss of physical, financial, or natural assets). This opportunity to access and utilize social networks and thus benefit from high social capital maintained the resiliency of such livelihoods because it allowed risks to be spread over a wider geographical area. It is clear, therefore, that in Kedia, formal and informal institutions (Fraser et al.'s [2010] vulnerability framework) have played a key role in reducing vulnerability among dependents during the study period.

DISCUSSION -TOWARDS FUTURE LIVELIHOOD RESILIENCE
In both Kedia and Khawa settlements there were some generic choices that, if taken, increased the likelihood of a livelihood trajectory increasing in resilience between 1974 and 2005: 1. accumulation of financial assets (through waged-or self-employment, business, and/or the sale of natural and/or physical assets); 2. investment in and accumulation of physical assets (e.g., land and diverse herds of livestock); 3. opportunistic and strategic diversification of livelihood activities; 4. diversity within livelihood activities (e.g., investment in a range of stock types and planting of a mixture of crop types);

investment in improved access to water;
6. investment in transport; and 7. access to government support for elders, destitutes, and the disabled.
The following factors typically precipitated a shift towards increased vulnerability: (1) loss of livestock, (2) reduced access to natural assets (environmentally or policy determined), and (3) reduced diversity of livelihood activities conducted/ loss of livelihood activity/option.
In light of the continued heavy reliance of remote rural households on natural assets, the impacts of current and past dynamics, and the potential future impacts of climate change in the Kalahari (Thomas et al. 2005) on agroecosystems, there is likely to be an increasing role for formal and informal institutions in reducing vulnerability in Botswana. Clearly, in the context of these two settlements, ensuring access to a diversity of assets is vital. In the face of increasing climatic uncertainty, the key challenges to maintaining the effective functioning of this social-ecological system include (1) maintaining agroecosystem health to ensure adequate future supplies of natural resources (most essentially water, plant, and wildlife resources), (2) preventing elite capture and accumulation of opportunities, and (3) ensuring opportunities for diversification and generation of financial capital.
Many of these challenges could be addressed through improved functioning of formal and informal institutions (Twyman et al. 2004). Developments that facilitate (1) improved efficiency and accuracy of the distribution of government support, (2) adaptive management of dryland agroecosystems to ensure accommodation of dynamics rather than the imposition of stability and control, (3) more equitable access to natural resources, and (4) more equitable access to diversification opportunities and accumulation of financial and physical assets, will assist. As was illustrated in both settlements, collective action within communities is required along with government incentives and programs. If such developments cannot be achieved, and the predicted impacts of climate change continue, households are likely to increasingly face the need to move in search of better functioning and more resilient socialecological systems.

CONCLUSION
This paper has drawn on the concepts of livelihood trajectories and resilience to assist in the exploration of vulnerability in the drylands of Botswana. We used a combination of primary and secondary data to examine the inherent social-ecological dynamics in the study area and to categorize households into three different groups according to the ways in which their livelihood strategies exploited these inherent dynamics. Based on this information, we qualitatively assessed those factors that contributed to the emergence of vulnerability and/or resilience, http://www.ecologyandsociety.org/vol15/iss4/art3/ and elucidated five environmental changes operating independently of the inherent environmental dynamics that in many cases altered the vulnerability context and the overall direction of livelihood trajectories. In returning to the questions outlined earlier in relation to Fraser et al.'s (2010) framework, we illustrated the combined influence of environmental change and formal and informal institutions in determining a household's access to and use of assets and therefore its ability to create more resilient livelihood outcomes. In some cases, the agroecosystem remained productive in a changing vulnerability context, and for some people, their survival was supported by the combination of the livelihood strategies they pursue and the institutional capacity and willingness related to their particular context. Our paper has nevertheless indicated that the everyday details in each narrative have a profound influence on households' livelihood trajectories and resilience.
In view of projected climate changes in this part of southern Africa and their potential impacts, these findings have highlighted the importance of formal and informal institutions in building resilience and the need for increased effort to ensure the most vulnerable households have access to a diversity of assets.