Tobacco cultivation as a driver of land use change and degradation in the miombo woodlands of south‐west Tanzania

Miombo woodlands support agriculture, biodiversity, and multiple ecosystem services across an extensive part of sub‐Saharan Africa. Miombo is frequently overutilised with deforestation and degradation resulting in significant land use and land cover change (LULCC). Understanding the drivers of LULCC is essential to achieving sustainable land management in miombo woodland regions. Within a remote miombo area of south‐west Tanzania in the Kipembawe Division, Mbeya Region, social survey and ecological data were used to identify the direct and indirect drivers of LULCC. Our findings show that tobacco (Nicotiana tabacum) production results in an estimated annual deforestation rate of 4,134 ± 390 ha of undisturbed miombo woodland, of which 56.3 ± 11.8% is linked to the post‐harvest curing process. This deforestation represents 0.55 ± 0.06% of the wooded area of the Kipembawe Division. The perception of high incomes from tobacco cultivation has encouraged migration of both agriculturalists and pastoralists into the area, resulting in higher livestock numbers that lead to further degradation. Higher human populations need more woodland resources such as fuelwood and building materials and more farmland for food crops. Continued deforestation will reduce the long‐term profitability of tobacco cultivation due to a lack of fuel to cure the crop and could render production unviable. Action is urgently needed to conserve globally important biodiversity resources while enabling agricultural and pastoral activities to continue. Improved governance, together with sustainable land management strategies and diversification of livelihood strategies, can reduce dependence on tobacco cultivation and contribute to a sustainable future for this ecoregion.

Miombo woodlands are dominated by tree species of the legume subfamily Caesalpinioideae within three genera (Julbernardia, Brachystegia, and Isoberlinia;Frost et al., 2003). They are globally important owing to their capacity to store carbon and influence environmental and socioeconomic systems (Ribeiro, Syampungani, Matakala, Nangoma, & Ribeiro-Barros, 2015). They are locally important due to provisioning ecosystem services including medicinal plants, edible forest products, food for livestock, construction materials, and fuel sources (Dewees et al., 2010;Jumbe, Bwalya, & Husselman, 2008;Malambo & Syampungani, 2008). By 2050, sub-Saharan Africa's population is predicted to increase twofold (Eastwood & Lipton, 2011), leading to increasing pressure upon miombo woodland (Cabral, Vasconcelos, Oom, & Sardinha, 2011). Sustainable management of miombo woodlands is therefore needed, and they are receiving increasing global consideration (Williams et al., 2008). Presently, the greatest research focus in miombo woodland surrounds their role in carbon storage (e.g., Shirima et al., 2011;Williams et al., 2008), with limited understanding of the drivers of land use change.
Regionally, several direct anthropogenic drivers of LULCC have been identified in miombo systems, including overgrazing, agricultural expansion, charcoal, fuelwood, and timber extraction, rising urbanisation, unmanaged fires, and excessive exploitation of valuable animal and tree species (e.g., Cabral et al., 2011;Fisher, 2010;Ryan et al., 2016). Natural drivers of change that are likely to impact miombo woodlands include changes to rainfall patterns and volumes (Seth et al., 2013), rising temperatures (Pienaar, Thompson, Erasmus, Hill, & Witkowski, 2015), and altered fire regimes (Andela & van der Werf, 2014). General descriptions of drivers can provide information to inform regional land management policy, yet they do not identify localscale nuances necessary for land use and management decisions. To provide effective, enduring management solutions for miombo woodlands, it is necessary to understand both direct and indirect drivers (Nelson et al., 2006), especially as drivers differ substantially from region to region (Bond, Chambwera, Jones, Chundama, & Nhantumbo, 2010;Vinya, Syampungani, Kasumu, Monde, & Kasubika, 2011). This paper addresses this gap by providing empirical data from a miombo woodland landscape in south-west Tanzania, which is currently experiencing rapid land use change. The key anthropogenic drivers of land use change are identified through integrative quantitative and qualitative research methods.

| Study area
Miombo woodland represents 95% of forested area in Tanzania (MNRT, 2006). Between 1990 and 2000, it is estimated that 13% of Tanzanian miombo woodland was lost (FBD, 2008). Current estimations of Tanzanian woodland and forest loss range between 372,000 and 580,000 ha/year (FRELT, 2016;MNRT, 2015). The site for this study is located in the Kipembawe Division (8,766 km 2 ), within the Chunya District, Mbeya Region of south-west Tanzania (7°54′58.44″ S, 33°19′22.84″E, Figure 1). The study area is representative of other  However, this study found that the reserves are poorly managed owing to insufficient funding and limited capacity in terms of personnel and transport. Access to woodland is therefore largely unrestricted across both protected and unprotected areas. Average yearly precipitation is 933 ± 36 mm (n = 28 years). Rains typically start in October and occur frequently until May, with very little falling throughout the rest of the year. The soils are shallow and sandy, and the landscape is predominantly flat.

| Data collection
To identify the drivers of deforestation and degradation, a mixed methods approach was taken, combining social and ecological surveys.
This enabled a holistic examination of the drivers of land use change by drawing upon a range of complementary primary data sources.

| Ecological survey
Nine ecological survey sites were selected (described in Jew, Dougill, Sallu, O'Connell, & Benton, 2016), representing low to high levels of human utilisation of the woodland. Within each survey site, five transects were conducted to record land use type and utilisation levels.
Transects were 10 m wide and 1.5 km long and split into 20-m sections (Doggart, 2006), sampling 75,000 m 2 at each site. Within each section, all live, dead, and cut poles and timbers were recorded, and the main land cover type documented. Evidence of utilisation or removal of non-timber forest products and other disturbances was noted, for example logging, tree bark removal, and beehives.

| Social survey
The social survey consisted of household questionnaires, village-level focus groups, and semi-structured key informant interviews to obtain information on drivers of land use change and agricultural methods.
The five villages selected for involvement in the social survey were in close proximity to ecological survey sites with medium and low utilisation levels, allowing social and ecological survey data to be aligned by comparing quantitative data with qualitative data, particularly in terms of agricultural land cover. The four remaining ecological survey sites were not in close proximity to any village and therefore not suitable for comparable study. Villages were situated within three wards ("study" villages, Figure 1). A further village was selected for piloting the research methods ("pilot" village, Figure 1). Fieldwork took place February-September 2013, when the research team lived within the community, and therefore, field observations were an additional data source. Government census data were also used to determine demographic patterns within the district.
Within each of the five villages, 10% of households (n = 196) were chosen at random to engage in questionnaires (Meshack, Ahdikari, Doggart, & Lovett, 2006). These were undertaken with the head of the household, where a household was defined as containing people who eat at least one meal together and sleep in the same accommodation, and the head is the principal decision maker. Household farming activities were discussed. Questionnaires were con- Multiple focus group discussions took place in each village with identified sets of people (e.g., villagers, livestock keepers, and crop producers) determined through key informant interviews with village committee representatives (e.g., Participatory Forest Management Committee and Social Welfare Committee). Focus group discussions lasted for approximately 1 hr, with 2-8 people and an even number of men and women, subject to availability. Overall, 28 focus groups were conducted. The purpose of focus groups was to collect comprehensive qualitative information on relevant issues and to explore key themes and questions that had arisen in household questionnaires. A range of questions was presented, and all answers were considered between group participants with facilitation (Ritchie, Lewis, Nicholls, & Ormston, 2013). Each session was recorded, and the lead researcher took notes through translation.
Semi-structured interviews took place with 41 key informants at all governance levels from village to regional. Key informants were either involved with a particular programme or project or held extensive knowledge on a specific relevant topic (O'Leary, 2013). Snowball sampling was used to identify interviewees within the public, private, and voluntary sectors. Interviews explored key themes of relevance to each individual that had emerged through household questionnaires and focus groups. Interviews and focus groups were coded and grouped into themes for analysis, with direct and indirect drivers emerging from the data and subsequently undergoing comparison with the other data sources to determine validity.

| RESULTS
The main indirect drivers of LULCC were identified to be demographic (in-migration) and economic (rising tobacco prices). Direct drivers include the clearing of land for agriculture (in particular tobacco), energy demand for curing tobacco leaves, extraction of wood for household use and construction, and degradation and deforestation caused by livestock and livestock keepers.

| Economic: rising tobacco prices
Higher tobacco prices encourage in-migration as Ward B Officer 1 (2013) explained: "There is a lot of immigration for tobacco cultivation, when the price is high" (Jew, 2016, p. 95). Additionally, current residents may decide to cultivate tobacco or expand their cultivated area in response to rising prices. This is illustrated in Figure

| Direct drivers 3.2.1 | Agriculture
In the six ecological survey sites that experienced high and medium levels of woodland utilisation, transect data demonstrated that approximately 30% of land cover was agricultural, 7% was regenerating miombo woodland, and 62% was undisturbed vegetation (Table 1)

| Tobacco cultivation and energy demand for curing
Tobacco cultivation within Kipembawe was introduced in the 1960s, and people were moved to the area through government relocation schemes to grow tobacco, originally through communal systems. In the recent years, inputs (seeds, fertiliser, and pesticides) have been supplied by two tobacco merchants and are distributed by a Primary Co-operative Society.
Seeds are distributed free of charge, and inputs including fertiliser and pesticides are received on loan, which is repaid at the end of the season.
The nature of tobacco cultivation is described in the following quote: The normal pattern with land clearance is that tobacco is planted and harvested, and to dry that tobacco another area of land is cleared. The following year the tobacco will be planted on that cleared land, and another crop such as maize is planted on the old land. Then more trees must be harvested to cure that harvest. The year after that the farmer will grow tobacco on the first field. However, the farmer will still need more wood to cure the tobacco, so each year they must remove some trees, even if it is not always a larger block. If they wish to expand their farm they must clear land.
Tobacco Company 1, 2013 (Jew, 2016, pp. 89-90) To preserve the tobacco leaves, they need to be dried, or cured. In this area, tobacco is flue cured, where the leaves are hung in burners and the flues are lit below. This process has a high-energy demand, for which wood is used. Farmers tend to build their tobacco burners   0.55 ± 0.06% of the total woodland. If this rate continues, the woodland in Kipembawe will be entirely lost in 180 ± 15 years ( The tobacco industry is conscious of this energy demand and the impact that woodland loss will have on the sustainability of production: We need a lot of firewood for tobacco. If they cut the trees it means that trees will be finished and tobacco production will not be there anymore because we need a lot of wood to cure the tobacco. If we don't have wood we don't have tobacco. Kipembawe used to be a very big forest. So it will be in 10-15 years there will be very big problems here, the tobacco production will diminish. Tobacco Company 1, 2013 (Jew, 2016, p. 90) To reduce the extraction of native trees, tobacco companies are encouraging the planting of fast-growing eucalyptus trees as an alternative energy source and the use of "modern" fuel-efficient tobacco burners over traditional burners (Company 1 and 2, 2013). However, throughout this research, only two "modern" burners and one small eucalyptus lot were observed.
In addition to the conversion of undisturbed vegetation to farmland to cultivate tobacco, curing of leaves uses 200,000 ha of woodland a year globally (Geist 1999), accounting for 1.7% of global net forest cover loss. The figure estimated here is that 1 kg of cured tobacco requires 12.06 ± 2.52 kg firewood is similar to that found in other studies (Otanez, 2008;Siddiqui & Rajabu, 1996). This suggests that, in the absence of empirical data from curing barns, 12.06 ± 2.52 kg is a suitable estimate for the amount of firewood required to cure tobacco in Kipembawe. Traditional tobacco burners lose ∼98% of the energy supplied; "modern" tobacco burners are 44% more thermally efficient, losing 55% of energy supplied (Musoni, Nazare, Manzungu, & Chekenya, 2013). Should demand remain the same, encouraging the building of "modern" tobacco burners will significantly reduce the use of fuelwood for curing in Kipembawe. Rapidly increasing populations in poor rural areas without corresponding out migration as seen in this study leads to agricultural expansion (Rudel, 2013) and puts further demands onto the surrounding environment for ecosystem services such as wood for building materials and fuel, leading to further degradation. Within Southern Africa, fuelwood or charcoal provides 70% of the energy consumed (Syampungani, Chirwa, Akinnifesi, Sileshi, & Ajayi, 2009).
In other parts of Chunya, 88.3% of harvested timber is converted to charcoal (Sawe, Munishi, & Maliondo, 2014). The remoteness of Kipembawe and high availability of firewood contribute to the low extraction of charcoal. However, this is likely to change with increasing urban demands, few affordable alternatives (Ahrends et al., 2010), and an improving road network.
Similarly, increasing livestock numbers in Kipembawe have resulted in further demands upon the surrounding environment. Livestock numbers have increased due to the in-migration of members from the agro-pastoralist Sukuma tribe, searching for grazing and water due to displacement from their traditional lands in northern Tanzania (Charnley, 1997). The expansion of the Ruaha National Park has led to further displacement of Sukuma from the Mbarali District in Mbeya, adjacent to Kipembawe (Sirima & Backman, 2013). Such sociopolitical drivers are an additional indirect driver of land use change.
Miombo woodlands are well known for their ability to regenerate (Kalaba et al., 2013), and this is a vital component of the dynamics within miombo woodlands. In this study area, there was little evidence of regenerating land being "reused" for cultivation, and the continued deforestation of undisturbed vegetation leads to losses of endemic flora and fauna Jew, Loos, Dougill, Sallu, & Benton, 2015). A study by Prins and Kikula (1996), also within the Chunya District, found that tobacco cultivation ceased in some areas during the 1980s when tobacco prices dropped. However, the land was not left to regenerate but was used for the cultivation of other crops. They also found that areas that had been cultivated for over 7 years did not regenerate even when they had been left fallow for at least 15 years (Prins & Kikula, 1996) as a result of damage to the rootstocks (Boaler & Sciwale, 1966). As such, a reliance on the regrowth properties of miombo woodland to mitigate woodland loss is inadvisable.

| CONCLUSION
Land use change in Kipembawe results from increases in tobacco cultivation driven by rising prices and in-migration to cultivate the crop.
This has led to an increase in population, which drives further direct land use change through the extraction of wood resources to provide housing and firewood, in addition to clearance and degradation of woodland for cultivation and livestock keeping. Given that tobacco cultivation is linked to the majority of LULCC changes within Kipembawe, it is the main driver of land use change leading to woodland degradation. Due to the Tanzanian government's current positive support for tobacco production, it is probable that tobacco cultivation will continue to increase, driving further woodland degradation and deforestation. Action is required to avoid these impacts. Tobacco companies and government forestry and livestock departments have policies in place (e.g., "modern" burners, woodlots, livestock movement, and logging restrictions), but there is little evidence of their implementation. This must be addressed, in addition to the development of land management strategies that regulate woodland utilisation and alternative methods for drying the tobacco crop. Encouraging the development of diverse livelihood approaches and limiting the top price of tobacco could reduce the incentive to cultivate tobacco and limit further degradation of the miombo woodland system.