UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Forest dependence and forest degradation in southern Malawi Nerfa, Lauren 2018

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

Item Metadata

Download

Media
24-ubc_2018_september_nerfa_lauren.pdf [ 2.6MB ]
Metadata
JSON: 24-1.0371136.json
JSON-LD: 24-1.0371136-ld.json
RDF/XML (Pretty): 24-1.0371136-rdf.xml
RDF/JSON: 24-1.0371136-rdf.json
Turtle: 24-1.0371136-turtle.txt
N-Triples: 24-1.0371136-rdf-ntriples.txt
Original Record: 24-1.0371136-source.json
Full Text
24-1.0371136-fulltext.txt
Citation
24-1.0371136.ris

Full Text

FOREST DEPENDENCE AND FOREST DEGRADATION IN SOUTHERN MALAWI  by  Lauren Nerfa   A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF  THE REQUIRMENTS FOR THE DEGREE OF  MASTER OF SCIENCE in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Forestry)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  July 2018   Lauren Nerfa, 2018  ii The following individuals certify that they have read, and recommend to the Faculty of Graduate and Postdoctoral Studies for acceptance, a thesis entitled:  Forest Dependence and Forest Degradation in Southern Malawi   submitted by Lauren Nerfa in partial fulfillment of the requirements for the degree of Master of Science in Forestry.  Examining Committee:  Dr. Jeanine Rhemtulla, Faculty of Forestry Supervisor  Dr. Hisham Zerriffi, Faculty of Forestry Supervisory committee member  Dr. Sean Smukler, Faculty of Land and Food Systems Supervisory committee member  Dr. Terry Sunderland, Faculty of Forestry External examiner  Dr. Bianca Eskelson, Faculty of Forestry Chair     iii Abstract  Rural small-holder farmers in the tropics rely on forests for multiple ecosystem services, such as provisioning services for fuelwood, timber, wild foods and medicinal plants. Yet many of these forests are undergoing degradation and loss, thus jeopardizing long-term ecosystem functioning and services. Measuring levels of forest dependence in agricultural communities is key to understanding livelihood sustainability and potential approaches to forest-based poverty alleviation. Understanding the ecological changes in forests where communities collect forest products, particularly fuelwood, is important for identifying approaches to forest conservation. To address these issues, I conducted social and ecological research in southern Malawi. I conducted household surveys (n=157) in agricultural communities to assess levels of forest dependence. I developed a new index to measure forest dependence that incorporates: the diversity of forest products collected to meet household needs, the effort involved in collection, relative wealth, and alternative livelihood strategies. I compared the index values for the study area to relative forest income values, the proportion of total income comprised by forest-derived income, which is the commonly used measurement of forest dependence. I showed that the relative forest income approach may underestimate levels of forest dependence, and that my new index provides insights into additional livelihood aspects of household forest dependence.   I investigated tree species richness, abundance, diversity, composition and aboveground carbon (AGC) in forest plots (n=86) in the miombo woodlands where the farming communities harvest fuelwood, and compared them to reference sites in relatively undisturbed forests. I investigated whether proxies for harvesting access (elevation and distance to the main road) and harvesting pressure (number of settlements within a 3 km buffer) were correlated with the vegetation characteristics in the fuelwood harvesting sites. Tree species richness, abundance, diversity and AGC were lower in fuelwood harvesting sites than reference sites, species composition was significantly different, and the proxies for harvesting pressure and access were correlated with species abundance and AGC. The findings suggest that long-term sustainability of forest collection may be hindered due to forest degradation, which is problematic given the high forest dependence in the area. Interventions to increase sustainability of the social-ecological system could be explored.  iv Lay Summary  The overall goal of this research project was to understand how subsistence farmers depend on forest products, and how forest use has changed biodiversity and carbon stocks in tropical forests. The project took place in miombo woodlands in southern Malawi. I assessed levels of household forest dependence in agricultural villages using household surveys, and I developed a new index that incorporates the diversity of forest products collected, effort involved in collection, relative wealth, and alternative livelihood strategies. Forest dependence in the farming communities was higher than the traditional metric suggests. I also assessed ecological changes in forests where local communities harvest fuelwood using forest plots. Tree biodiversity and carbon stocks were lower in fuelwood harvesting areas than in reference sites in relatively undisturbed forests. Long-term sustainability of the social-ecological system is at risk due to the forest degradation and high forest dependence. Future work could explore interventions to increase sustainability.                        v Preface  The research project in Malawi was a pilot project initiated by my supervisor Dr. Jeanine Rhemtulla and her colleague Dr. Joleen Timko who had previously worked in the villages in the study area. I entered the pilot project as a research assistant in the summer of 2016, prior to the formal start of my Master’s program at UBC in September 2016, with the agreement that this field season would contribute data for my thesis. I co-designed the forest plot methodology with Jenny Liu, a BSc student at UBC at the time, and co-designed the household survey with Dr. Rhemtulla, Dr. Timko, and Abigail Dan, a PhD student at UBC at the time. Dr. Rhemtulla and Dr. Timko applied for and received ethics approval from the UBC research ethics board.  In the field I co-led the household surveys with Abigail Dan, which were enumerated by two Malawian translators. I led the fieldwork for the forest plots, assisted by local botanists. I had two research assistants, Caitlin Laidlaw and Hannah Crisp, both BSc students at UBC at the time, who assisted with field work for the household surveys and forest plots, and data entry of the household survey data.  Upon the formal start of my Master’s program in September 2016, I cleaned and analyzed the data from the household surveys and forest plots. I designed the research questions and thesis outline, with inputs from my supervisor and supervisory committee including Dr. Hisham Zerriffi and Dr. Sean Smukler. I completed the remainder of the data analysis and writing of the thesis over the course of my program. I received GIS assistance, such as uploading of data and creation of GIS layers, from Hannah Crisp and Caitlin Laidlaw.  Chapters 2 and 3 of the thesis will be submitted for publication following acceptance of the thesis. Chapter 2 will be co-authored by Dr. Hisham Zerriffi and Dr. Jeanine Rhemtulla who assisted in the development of the index and provided inputs on the text. Chapter 3 will be co-authored by Dr. Jeanine Rhemtulla, who provided inputs into the research design, analysis and text.   vi Table of Contents   Abstract .......................................................................................................................................... iii Lay Summary ................................................................................................................................. iv Preface............................................................................................................................................. v Table of Contents ........................................................................................................................... vi List of Tables ............................................................................................................................... viii List of Figures ................................................................................................................................. x Acknowledgements ....................................................................................................................... xii Chapter 1. Introduction.................................................................................................................. 1 1.1. Forest dependence in the global south and livelihood impacts of forest degradation and deforestation ............................................................................................................................ 1 1.2. Social and ecological consequences of forest degradation .............................................. 3 1.3. Tropical forests as complex social-ecological systems ................................................... 4 1.4. Policy applications ........................................................................................................... 5 1.5. Research on forest dependence and ecological consequences of forest degradation ...... 7 1.6. Case study in southern Malawi ........................................................................................ 9 1.6.1. Study area .............................................................................................................................................. 9 1.6.2. The research project ............................................................................................................................. 12 1.7. Research objectives ........................................................................................................ 14 Chapter 2. Measuring household forest dependence using a new index based on the collection of forest products ............................................................................................................................... 16 2.1. Introduction .................................................................................................................... 16 2.2. Defining forest dependence ........................................................................................... 18 2.3. Measuring forest dependence ........................................................................................ 19 2.4. Uses of indices ............................................................................................................... 21 2.5. Framework for a new forest dependence index ............................................................. 22 2.6. Testing the forest dependence index using the case of southern Malawi ...................... 26 2.6.1. Study area ............................................................................................................................................ 26 2.6.2. Data collection ..................................................................................................................................... 26 2.6.3. Calculation of the forest dependence index ......................................................................................... 27 2.6.4. Calculation of relative forest income ................................................................................................... 31 2.7. Levels of forest dependence in southern Malawi .......................................................... 32 2.8. Comparison of the forest dependence index to the relative forest income method ....... 36 2.9. Reasons for employing a forest dependence index ........................................................ 39  vii Chapter 3. Changes in tree species diversity, composition and aboveground carbon in areas of fuelwood harvesting in miombo woodland ecosystems of southern Malawi ............................... 44 3.1. Introduction .................................................................................................................... 44 3.2. Methods.......................................................................................................................... 46 3.2.1. Study area ............................................................................................................................................ 46 3.2.2. Forest plots........................................................................................................................................... 49 3.2.3. Statistical analysis ................................................................................................................................ 50 3.3. Results ............................................................................................................................ 53 3.3.1. Tree, sapling and seedling species diversity and composition ............................................................ 53 3.3.2. Aboveground carbon ........................................................................................................................... 60 3.3.3. Harvesting access and pressure ........................................................................................................... 61 3.4. Discussion ...................................................................................................................... 63 Chapter 4. Conclusion ................................................................................................................. 69 4.1. Measuring multiple livelihood aspects of forest dependence ........................................ 69 4.2. Ecological changes in forests with fuelwood harvesting ............................................... 70 4.3. Managing tropical forests as complex social-ecological systems.................................. 71 4.2. Future research ............................................................................................................... 73 Bibliography ................................................................................................................................ 76 Appendix A. Chapter 2 Additional Figures and Tables ............................................................... 86 Appendix B. Chapter 3 Additional Figures and Tables ............................................................... 87                 viii List of Tables  Table 2. 1. Summary statistics for the sub-indices (standardized using z-scores), for the study area in southern Malawi (n=157). Forest product use and effort values are for fuelwood and wild foods collected from village forests, forest reserves and private woodlots. ................................. 36 Table 2. 2. Annual forest, non-forest, and total household income in U.S. Dollars for the study area in southern Malawi (n=157). True forests are village forests, forest reserves and private woodlots; all locations include homesteads and farms in addition to true forests. Note that total income (forest and non-forest) is equal in the two cases because the income from fuelwood and wild foods collected from homesteads and farms is added to the total in each case, either to forest or non-forest income. .................................................................................................................... 37 Table 3. 1. Mean species richness, species abundance, Shannon index and Simpson’s diversity values for trees, saplings and seedlings in fuelwood harvesting sites (n=14) and reference sites (n=9). Values computed from the counts of individuals within each site. Standard errors shown. Significant p-values (=0.05) from Welch two sample t-test shown with an asterisk. ................ 56 Table 3. 2. Quasi-poisson generalized linear model (GLM) standardized coefficients and p-values testing elevation, distance to the main road and number of rooftops within 3 km as explanatory variables and species richness, species abundance, Shannon index, Simpson’s diversity and biomass as response variables for fuelwood harvesting sites (n=14). Significant p-values (=0.05) shown with an asterisk. ...................................................................................... 62 Table 3. 3. R2 values with p-values for site averages of distance to the main road, elevation and number of rooftops in 3 km proximity, plotted as vectors on the NMDS ordination for fuelwood harvesting sites (n=14). Significant p-values (=0.05) shown with an asterisk. .......................... 62 Table A. 1. Summary statistics of annual non-forest income sources with input costs, and total annual non-forest income in U.S. Dollars (USD) for the study area in Malawi (n=157). All minimum values were 0. ............................................................................................................... 86 Table B. 1. Ecological variables and proxies for harvesting access and pressure used in the generalized linear models (GLMs) averaged per site for fuelwood harvesting sites (n=14). Elevation is measured in meters above mean sea level and rooftop count is the number of rooftops within a 3 km radius circle around the forest plots, averaged per site. Elevation, distance to the main road and number of rooftops are the explanatory variables and species richness (number of species), species abundance (number of stems), Shannon index, Simpson’s diversity and AGC are the response variables used in the GLMs. .............................................................. 89 Table B. 2. Ecological variables and proxies for harvesting access and pressure averaged per site for reference sites (n=9). Species richness is the number of species and species abundance is the number of stems. Elevation is measured in meters above mean sea level and rooftop count is the number of rooftops within a 3 km radius circle around the forest plots, averaged per site. Note that GLMs were not conducted for reference sites but the data was compiled for observation. .. 89 Table B. 3. Presence of tree, sapling and seedling species by percentage of plots, as well as average tree DBH (in cm) and average tree basal area (in m2) calculated by summing the basal area for each tree of the species and dividing by the number of stems, for fuelwood harvesting sites (50 plots) and reference sites (36 plots). Note that two tree species present in one fuelwood harvesting plot each (Macaranga capensis (Baill.) Benth. ex Sim and Newtonia buchananii (Baker) G.C.C.Gilbert and Boutique) do not have DBH or basal area averages due to the inability to measure these trees in the field. Exotic species indicated with an (E). .................................... 90  ix Table B. 4. Minimum, mean and maximum aboveground carbon values for fuelwood harvesting sites (n=14) and reference sites (n=9), for five mixed-species allometric equations designed for the miombo woodlands. Biomass values of trees per plot were summed per site and converted to Mg per hectare then converted to carbon per hectare by multiplying by 0.47. Differences between means for fuelwood harvesting sites and reference sites were significant for each equation (Kuyah, p=0.0001; Mugasha, p=0.0002; Kachamba, p=0.0001; Chidumayo, p=0.0002; Ryan, p=0.0002)............................................................................................................................ 91                          x List of Figures  Figure 1. 1. Nested map of the study area in southern Malawi, Africa. Zomba district is in the lower section of the labelled study area map and Machinga in the upper right.  Forest reserves and national park labelled where forest assessments were conducted. Household surveys were conducted in nine villages to the east of the Zomba-Malosa forest reserve. ................................ 13 Figure 2. 1. Flowchart showing the steps involved in calculating the forest dependence index. Aggregation is addition-based, and occurs at the forest product level meaning that forest product use and effort values are combined per forest product first, then summed over all products at the household level. Abbreviations are as follows: F = forest product use, E = effort, W = relative wealth, L = non-forest livelihood sources, A = amount, T = time, n= total number of forest products. Subscripts are as follows: i = forest product level, j = household level, o = other source. The z-scores standardization method is used across the sample population for all intermediate steps, and the re-scaling normalization method is used in the final calculation of the index. ...... 25 Figure 2. 2. Histograms of household values (normalized using re-scaling) for the forest dependence index, for fuelwood and wild foods collected from true forests (village forests, forest reserves and private woodlots; mean=0.49, SD=0.14) and all locations (mean=0.47, SD=0.19) for the study area in southern Malawi (n=157). ............................................................................ 32 Figure 2. 3. Histograms of household values (standardized using z-scores) for the forest product sub-indices, including the forest product use and effort for fuelwood and wild foods collected from village forests, forest reserves and private woodlots, for the study area in southern Malawi (n=157). ......................................................................................................................................... 34 Figure 2. 4. Histograms of household values (standardized using z-scores) for the livelihoods sub-indices, including the asset-based wealth index and the number of non-forest livelihood strategies, for the study area in southern Malawi (n=157)............................................................ 35 Figure 2. 5. Histograms of household values (standardized using z-scores) for the forest product use and effort sub-indices combined – for fuelwood and wild foods collected from village forests, forest reserves and private woodlots – and the inverse of the relative wealth and livelihood strategies sub-indices combined, for the study area in southern Malawi (n=157). ..... 35 Figure 2. 6. Histograms of household values in the study area in southern Malawi (n=157) for forest dependence as measured using relative forest income, the proportion of total household income comprised by forest income, compared to the forest dependence index (normalized using re-scaling). Values for fuelwood and wild foods collected from “true forests” including village forests, forest reserves and private woodlots shown on the top (relative forest income mean=0.13, SD=0.18), where monetary values of forest products collected from homesteads and farms were added to the total non-forest income. Values from all locations shown on the bottom (relative forest income mean=0.28, SD=0.28). ........................................................................................... 38 Figure 3. 1. Study area in southern Malawi showing fuelwood harvesting sites (red) and reference sites (blue), as well as outlines of forest reserves and national parks. Study area indicated by the red box on the inset map of Malawi. Map created in ArcMap 10.5 (ESRI, 2016). ............................................................................................................................................ 47 Figure 3. 2. Dominant tree species in fuelwood harvesting sites (n=14; red) and reference sites (n=9; blue), by percentage of plots in which the species were found. Exotic species indicated with an (E). Note scale from 0 to 60 percent. ............................................................................... 54  xi Figure 3. 3. Dominant tree species in fuelwood harvesting sites (n=14; red) and reference sites (n=9; blue) by percentage of stems found in all plots. Exotic species indicated with an (E). Note scale from 0 to 30 percent. ............................................................................................................ 55 Figure 3. 4. Box plots showing site-level species abundance, species richness, Shannon index and Simpson’s diversity values for fuelwood harvesting sites (n=14) and reference sites (n=9), for counts of tree, sapling and seedling individuals. Significant differences between the means (Welch two sample t-test, =0.05) indicated with an asterisk...................................................... 56 Figure 3. 5. Rarefaction curves computed with 100 permutations, showing species richness for fuelwood harvesting sites (n=14; red) and reference sites (n=9; blue) by number of sites. ......... 58 Figure 3. 6. Non-metric multi-dimensional scaling (NMDS) ordinations showing species composition for counts of individuals per site, for trees, saplings and seedlings in fuelwood harvesting sites (n=14) and reference sites (n=9). NMDS1 and 2 are axes 1 and 2 respectively. Ordination constructed in R 3.4.0 using the package vegan, function metaMDS using Bray ecological distance and 4 dimensions for trees and seedlings, 5 for saplings to minimize stress (R Core Team, 2017). Stress values for trees, saplings and seedlings were 0.069, 9.67e-05 and 0.085 respectively. MRPP p-values shown............................................................................................. 59 Figure 3. 7. DBH values in cm for all trees (>4 cm DBH) in forest plots, for fuelwood harvesting sites (n=14) and reference sites (n=9). .......................................................................................... 60 Figure 3. 8. Boxplots showing the mean of the five allometric equations converted to aboveground carbon (Mg C/ha) per site for fuelwood harvesting sites (n=14) and reference sites (n=9) (Kuyah et al., 2014; Mugasha et al., 2013; Kachamba et al., 2016; Chidumayo, 2014; Ryan et al., 2011). .................................................................................................................................. 61 Figure A. 1. Histograms of household values of standardized (z-scores) amount and walking time data for the effort sub-index, for fuelwood and wild foods collected from village forests, forest reserves and private woodlots, for the study area in southern Malawi (n=157). ................ 86 Figure B. 1. Diagram of the size, layout and sampling of each subplot within the forest plots. Note that the percent covers of grass, shrubs and herbs were not analyzed in this thesis. ........... 87 Figure B. 2. Non-metric multidimensional scaling (NMDS) ordination for fuelwood harvesting sites (n=14) for counts of tree species per site, with mean elevation (R2 =0.52), mean distance to the main road (R2 =0.28) and mean number of rooftops in a 3 km radius (R2 =0.35), each averaged per site, plotted as vectors. Vectors scaled according to their R2 values.  NMDS1 and 2 are axes 1 and 2 respectively. Ordination constructed in R 3.4.0 using the package vegan, function metaMDS using Bray ecological distance and 4 dimensions (R Core Team, 2017). ..... 88          xii Acknowledgements  I would like to thank my supervisor Dr. Jeanine Rhemtulla for enabling the many opportunities I have been fortunate to have during my Master’s program and for her continuous support throughout. I thank Dr. Joleen Timko for her assistance in initiating the project, especially for sharing her connections in Malawi from her past work in the area. I extend my gratitude to the community members of the Zomba-Malosa area for participating in the research project, and to our many local field assistants including McMillan Saddick, Willie Sagona, Silekiwe Mwenelupenbe, Steve, Kennedy, the local District Forest Officer and local forest guides in the Machinga district of Malawi. I thank the hardworking field team of students from the University of British Columbia who assisted with the field work including Abigail Dan, Jenny Liu, Hannah Crisp, Caitlin Laidlaw, and Robyn Clark. Thanks also to Hannah Crisp and Caitlin Laidlaw for their efforts with data entry, cleaning and GIS assistance. I would like to thank Dr. Sean Smukler and Dr. Hisham Zerriffi for their valuable contributions to the thesis throughout my Master’s program. Thank you to the Land and Lives lab for their suggestions and insights, including Libin Thaikkattil Louis, Gabriela Barragan, Laura Rasmussen, Hyeone Park, Tara Bergeson and Riley Finn. Lastly, I am thankful for the funding provided for the field project by a UBC Hampton New Faculty Award to Dr. Rhemtulla and for the funding for this Master’s project through a CGS-M scholarship from the National Sciences and Engineering Council of Canada.                1 Chapter 1. Introduction   1.1. Forest dependence in the global south and livelihood impacts of forest degradation and deforestation  Communities in the tropics living in mixed forest-agricultural landscapes rely on numerous regulating, provisioning and cultural ecosystem services provided by forests (Sunderlin et al., 2005; Vedeld et al., 2007; Wunder et al., 2014). In this thesis, I focus on provisioning ecosystem services, namely for the forest products that households collect from tropical forests such as fuelwood, building materials, wild foods and medicinal plants (Sunderlin et al., 2005; Vedeld et al., 2007; Wunder et al., 2014; Angelsen et al., 2014). Households use these products for subsistence purposes or as a form of income generation (Cavendish, 2000). The concept of forest dependence, defined in multiple ways in the literature, has been described as a reliance on forests for ecosystem services, subsistence needs, safety nets, gap fillers and opportunities for poverty alleviation (Sunderlin et al., 2005). It has been estimated that 350 million rural peoples live in or at the margin of forests and rely on forests for safety nets or supplemental income (Chao, 2012; World Bank, 2004). More than 500 million smallholder farmers grow trees on their farms or manage forest patches for subsistence and income (Chao, 2012). Smallholder farmers dwelling in these tropical forest landscapes are typified by limited resources, preference for economic benefits in the short term, and risk aversion (Erenstein, 2003). Marginalization, in terms of poor land quality, lack of assets and inputs, and limited market access are also key issues facing smallholders, who are a large portion of the world’s poor (Vignola et al., 2015). Therefore, small-holder farmers often have limited resources and few alternatives to collecting forest products.  Rates of poverty are generally high in tropical forest landscapes and forests provide for the needs of impoverished households (Sunderlin et al., 2005). Since forests are typically open access and have few barriers to entry, impoverished peoples are able to establish forest-based livelihoods relatively easily (Sunderlin et al., 2005). Those who lack alternatives often use forest products, particularly non-timber forest products (NTFPs), because of the low or moderate labour, capital and skill requirements, and open or semi-open access (Angelsen & Wunder, 2003). The role of forests as safety nets or gap fillers versus poverty traps has been discussed at length in the literature. Forests serve as safety nets in the sense that they provide forest products for consumption or sale  2 in unexpected times of economic shortfall, and as gap fillers by providing seasonal and periodic products such as fruits (Angelsen & Wunder, 2003). But they may also be poverty traps in that there are typically low income returns for the collection of forest products (Angelsen & Wunder, 2003). Some researchers argue that forests should only be considered poverty traps in situations where alternative livelihood strategies exist but policies or external interventions prevent the shift away from dependence on low-value forest extraction (Angelsen et al., 2014; Angelsen & Wunder, 2003).  Sunderlin et al. (2005) identified four different groups for which poverty and forests are linked: indigenous communities dwelling in forests who have historically lived outside of the market economy; forest dwellers in remote areas with limited access to the market economy such that their socio-economic systems change slowly; rural in-migrants who settle in forests to develop agriculture as they lack access to other land; and refugees from war and conflict who temporarily shelter in forests. It is important to note the degree of poverty for the group in question, which influences the extent of their forest use. The poorest of the poor, those who are landless, may depend on greater levels of forest exploitation (ex. charcoal production), than small-holder farmers who collect forest products for subsistence use (Angelsen & Wunder, 2003).   High rates of poverty also overlap with high rates of forest degradation and deforestation in the tropics (Sunderlin et al., 2005; Sunderlin at al., 2008). An estimated total of 500 million hectares of tropical forests have been degraded (Ghazoul et al., 2015). Gross annual rates of tropical deforestation have been estimated at 8 million hectares per year in the 1990s and 7.6 million hectares in the 2000s (Achard et al., 2014). Underlying causes of degradation and deforestation are manifold and include several categories of factors: economic, such as economic structures and market growth; demographic, such as population density and migration; technological, such as agricultural and forest sector technologies; policy, including formal policies and the policy climate; and cultural factors including public attitudes, values and beliefs (Geist & Lambin, 2002). The underlying causes influence the proximate causes, the direct human actions leading to deforestation and forest degradation. The most significant proximate driver of tropical deforestation is commercial agriculture (Hosonuma et al., 2012). Important proximate drivers of tropical forest degradation include over-harvesting of timber and other logging activities, fuelwood harvesting (charcoal or firewood), grazing of livestock, and uncontrolled burning (Hosonuma et al., 2012).   3 1.2. Social and ecological consequences of forest degradation   There are numerous consequences of forest degradation for people and ecosystems. One of the key social consequences of forest degradation is the impact on the forest dependence of small-holder farmers. Loss of tree cover and plant biomass occurring through forest degradation limits the availability of forest products, and thus the long-term sustainability of household harvesting for subsistence use. For example, in areas such as the Western Ghats of southern India where extensive forest degradation has occurred, households are no longer able to intensively harvest important forest products including fuelwood and fodder (Davidar et al., 2007). In response, plantations and managed forests will be required to supply forest products (Davidar et al., 2007). With high levels of forest degradation across India, linked to increased harvesting levels due to the growing population, continued forest harvesting has been identified as not conducive to forest conservation efforts (Davidar et al., 2010). In sub-Saharan Africa, extensive loss of woody vegetation is attributed primarily to increased population pressure on forests for agricultural land and fuelwood (firewood and charcoal) (Mitchard & Flintrop, 2013; Ryan et al., 2012). One response to ensure supplies of required forest products is to maintain desired tree species on farms and homesteads, or to plant trees to create agroforestry systems (Dewees, 1995). Governments have provided cash incentives or resources to plant trees on farms such as in Malawi through the Tree Planting Bonus Scheme or Agroforestry Extension program (Dewees, 1995; Kakhobwe et al., 2016).   Key ecological consequences of tropical forest degradation are losses of species diversity and reduction of carbon stocks, as well as reductions in ecosystem functioning. In this thesis, I focus on the ecological changes impacting plant communities, while acknowledging the impacts on animal species relying on forests for habitat. Plant biodiversity loss, which is also linked to changes in species composition, may cause the loss of functional groups and important ecosystem functions such as nutrient cycling, maintenance of soil structure and primary productivity (Díaz et al., 2004; Tilman  et al., 2014). Reduction in carbon stocks due to tree cover loss disrupts carbon cycles and contributes to climate change at the regional level (DeFries et a., 2004). Soil disturbance resulting from forest degradation also contributes to the losses of carbon stocks. Other ecosystem functions such as evapotranspiration and cloud formation may be altered, changing local  4 precipitation patterns which may lead to exacerbation of droughts (Lawton et al., 2001). Such changes in forest ecosystem function will most likely lead to a reduction in ecosystem services.  1.3. Tropical forests as complex social-ecological systems   In this thesis, I employ the complex systems framework, focusing on social-ecological systems. This framework is beneficial for synthesizing social and ecological aspects of tropical forests in areas of forest dependence and forest degradation. I invoke the complex systems framework for the purposes of framing the overall thesis and linking the social and ecological aspects of my research, but I do not conduct a formal analysis using the framework.  Complex social-ecological systems are a type of complex adaptive system. Complex systems are composed of multiple sub-systems and internal variables at multiple levels of organization and with multiple types of interactions (Ostrom, 2009). Complex adaptive systems have four basic properties: non-linearity, aggregation, diversity, and flows (Holland, 1995). Interactions and feedbacks at multiple scales enable complex adaptive systems to self-organize, based on the diversity and uniqueness of the components, interactions between components that are localized, and processes that enhance certain interactions (Holland, 1995; Levin, 1998). To organize the sub-systems and components of social-ecological systems to assess system sustainability, Ostrom (2009) developed a framework that identifies resource systems, resource units, governance systems and users, as well as interactions between them and outcomes that feed back into the sub-systems. Sustainability has been defined as the capacity for the continuation of a desired ecological or social condition or process (Tainter, 2006). In terms of long-term sustainability of complex social-ecological systems, the concepts of resilience, adaptability and transformability are useful (Walker et al., 2004). Resilience is a system’s capacity to absorb disturbance and re-organize to retain structure, function and feedbacks while allowing for new trajectories of the system; adaptability is the capacity of actors in a system to manage for resilience; and transformability is the capacity of actors to create a new system out of the existing one (Walker et al., 2004).    5 In the case of investigating household forest dependence and levels of forest degradation, the above described aspects of complex social-ecological systems assist in the organization of the ecological components and human actors and their interactions. In the context of forest dependence in tropical forest-agricultural landscapes, forests are the resource systems, forest products are the resource units, rural small-holder farmers are the users, and institutions at the community level or regional and national government are the governance systems. The key interaction between the users and the tropical forest resource systems in this case is the harvesting of forest products. Non-linearity is involved, for example, in the responses of the forest ecosystem to the harvesting of various forest products.  Increasing harvesting levels are not necessarily correlated with increasing detrimental effects on plant species and populations, rather impacts depend on life history and which plant parts are involved; some trees cannot survive low rates of harvesting whereas some perennial herbs may survive high harvesting rates (Ticktin, 2004).  Diversity is involved in the biological diversity within the ecosystem and the diversity of the stakeholders concerned with the forest. Aggregation is involved in species assemblages and the composition of the ecological community (Levin, 1998), and in the groupings of types of stakeholders or users of the resource system. Flows are involved in the flow of nutrients and energy in the ecological system, and the flow of goods and services (ex. forest products and other ecosystem services) to the social system. In a resilient system, levels of forest dependence and harvesting from forests would be sustainable without altering the system properties. Yet if degradation becomes high and the system properties change, such as through reduced plant species diversity and biomass, ecosystem functioning and services would likely be limited and the interactions between users and the resource system would need to change. The social-ecological system would be forced to transform into an alternative state.  1.4. Policy applications  Assessments of tropical forest degradation and levels of forest dependence can inform policies that jointly address forest conservation and poverty alleviation. As impoverished communities tend to live near or within forests (Sunderlin et al., 2008), combining the two strategies in policy approaches is wise. Poverty alleviation includes both poverty avoidance and poverty elimination (Angelsen & Wunder, 2003). Forest-based poverty alleviation has been  6 defined as the use of forest resources in the avoidance, mitigation and/or elimination of poverty (Sunderlin et al., 2004; Sunderlin et al., 2005). It can be accomplished in multiple ways: preventing the degradation of forest resources; redistributing forest resources for equitable access, and increasing the value of forest production (Sunderlin et al., 2004). Successful strategies can involve approaches to “people-centred forestry,” where local peoples are given power over management of local forest resources and the goal of management is to maintain sustainable sources of forest products (Sunderlin et al., 2004). One framework of ways in which communities, organizations and governments can invest towards poverty alleviation in forest landscapes is outlined by the acronym PRIME, which stands for productivity, rights, investment, markets and ecosystem services (Shyamsundar et al., 2018). The framework encourages: enhanced primary productivity and regeneration of plant species as well as effective forest management; user rights for forest dependence peoples; investments in institutions and services that ensure sustainable forest use;  increased access to markets for forest products for the rural poor; and enhanced flows of ecosystem services to forest dependent communities (Shyamsundar et al., 2018).   Forest landscape restoration and community-based forest management are two strategies that can combine poverty alleviation with forest conservation. Forest landscape restoration is advantageous in tropical forest social-ecological systems with forest degradation or loss where community forest dependence is high, because it involves restoring ecosystem integrity and enhancing human well-being, where local peoples are active agents and beneficiaries of restoration (Stanturf et al., 2012). One example of a large scale restoration project that helped to alleviate poverty was in the Poyang Lake Basin of China which lifted over 5 million farmers out of poverty over 20 years and increased forest cover from 27% to 60.5% through a combination of planting forests for conservation and establishing orchards and agroforestry (Huang et al., 2012).  In two other landscape-level restoration projects in Burkina-Faso and Niger, restoration involving reforestation and establishment of agroforestry increased food security for local communities while increasing biodiversity (Adams et al., 2016; Reij et al., 2010). Community-based forest management can also be involved in the conservation of forests and ecosystem services. While successful community-based management may be difficult to achieve, success has been shown where there is effective leadership and fair power relations between leaders and community members (Zulu, 2008). One example of successful community forest management from Zanzibar,  7 in the Jozani forest which includes a national park and is co-managed by local communities, generated revenue for all stakeholders, conserved endangered species such as the Zanzibar red colobus, and provided forest products for villages at the park periphery (Menzies, 2007). Another example of community forests in the Chittagong Hill Tracts of south-eastern Bangladesh had benefits both for the ecosystem, such as protecting biodiversity and enhancing forest regeneration in degraded areas, and for the local communities, such as sustainable consumption and sales of forest products (Nath et al., 2016). As exemplified by these cases, carefully designed and implemented strategies for forest landscape restoration and community-based forest management can have important local ecological and social benefits.  1.5. Research on forest dependence and ecological consequences of forest degradation  Given high rates of tropical forest degradation concurrent with high poverty, measuring forest dependence is important for informing sustainable livelihoods and forest use. Research on levels of household forest dependence have focused on measuring economic dependence using relative forest income, the contribution of the monetary value of forest products to total household income (Cavendish, 2000). Numerous studies have used this method (Campbell & Luckert, 2002; Cavendish, 2000; Vedeld et al., 2007; Wunder et al., 2014). In one of the seminal studies on forest income, Cavendish (2000) found that rural households in southern Zimbabwe derived on average 35.4% of their income from forests. Relative forest income was higher for more impoverished households: the poorest quintile derived approximately 40% of their income from forests as opposed to 10% derived by the wealthiest quintile (Cavendish, 2000). Absolute forest income was higher, however, for the wealthiest households (Cavendish, 2000). The global comparative analysis under the Poverty and Environment Network, conducted by the Centre for International Forestry Research (CIFOR), combined data on forest income from 8,000 households in 333 villages across 24 developing (Angelsen et al., 2014). The findings showed an average household relative forest income of 21.1% (Wunder et al., 2014). Income from forests was nearly as important to households as agricultural production, which comprised 28% of total income (Wunder et al., 2014).    8 While representing economic forest dependence is important for informing poverty alleviation policies, the relative forest income approach has a few limitations. Key limitations include: decreased accuracy of income estimates due to monthly, seasonal and annual income fluctuations (Nielsen et al., 2012), not accounting for the opportunity cost or labour for households of forest product collection (Pattanayak & Sills, 2001), and not being realistic in situations when households primarily consume rather than sell forest products (Kamanga et al., 2009). Additional livelihood aspects of forest dependence that the forest income method neglects include the labour involved in the collection of forest products, and whether or not households have alternative livelihood strategies to forest collection. In response to these limitations, in this thesis I propose the use of a new index of forest dependence that incorporates multiple livelihood factors of dependence, and which can be used complementary to the relative forest income method.  Investigating the ecological changes in degraded forests with high forest dependence is also important for understanding the sustainability of the ecosystem and local livelihoods. In order to measure ecological changes in degraded forests, degradation should be clearly defined. Forest degradation has been defined in numerous ways (Ghazoul et al., 2015; Sasaki & Putz, 2009; Simula, 2009). It has been argued that forest degradation is a social construct that depends on who creates the definition, whether it is a forester, ecologist, government official or local user (Lélé, 2000). One definition of degradation in accordance with the social-ecological systems framework is the loss of ecosystem resilience hence inhibiting the ability of the system to return to the state prior to disturbance (Ghazoul et al., 2015). A loss of resilience occurs due to reductions in canopy cover, species diversity, density, woody biomass and primary productivity, which impact ecosystem functioning (Chidumayo, 2013; Simula, 2009). Several studies have documented the relationship between anthropogenic disturbance and tropical forest degradation. One study using long term forest plots in Zambia, in and near the Lower Zambezi National Park, showed reduced stem density and woody biomass, and increasing species evenness in forest areas that have been degraded due to human caused fires and harvesting (Chidumayo, 2013). In the eastern Amazon, using both field plots and Landsat satellite data, selective logging was shown to cause a 35% reduction in aboveground carbon stocks, and logging together with fire caused a 57% reduction (Berenguer et al., 2014). In five regions across southern and central India in forests with fuelwood harvesting, remotely sensed data from the Forest Survey of India showed losses of dense and  9 moderately dense forest, and an increase in open forests, hence indicating forest degradation (Davidar et al., 2010).  Fuelwood is one of the most important forest products to the livelihoods of small-holder farmers in the tropics, yet the harvesting of fuelwood may cause forest degradation and even deforestation (Bailis et al., 2015). The ecological impacts of fuelwood harvesting are variable and localized, depending on population densities, demand for forest products, and management of forest harvesting (Heltberg, 2005). Several studies have used remotely sensed datasets to show losses of forest cover and woody biomass with fuelwood harvesting as one of the key drivers, such as in Mozambique where L-band synthetic aperture radar was used to detect a reduction in aboveground biomass in woodlands with fuelwood harvesting (Ryan et al., 2012). Spatio-temporal modelling tools have been developed and used to assess fuelwood supply and demand, such as the WISDOM tool that identified areas in southern and central Mexico where supply was not adequate to meet demand based on high fuelwood consumption and a deficit of forest cover (Masera et al., 2006). Yet recent studies have indicated the lack of knowledge of finer-scale changes in forests with fuelwood harvesting, such as changes in plant species diversity and composition (Sassen et al., 2015; Specht et al., 2015). One study in Mount Elgon National Park Uganda found a change in tree species composition in areas of fuelwood harvesting due to the reduction in species preferred and used for fuel (Sassen et al., 2015). Further assessments of forest composition changes in areas of fuelwood harvesting can be used to inform possible management interventions toward the maintenance of ecosystem resilience. In this thesis, I conduct an ecological assessment of tree composition in forests where households collect fuelwood and make a comparison to relatively intact forests to assess levels of degradation, and I investigate the correlation between the ecosystem changes and forest harvesting access and pressure.  1.6. Case study in southern Malawi  1.6.1. Study area   Malawi, in south-eastern Africa, is an ideal location in which to conduct research on forest degradation and household forest dependence. Malawi covers an area of 118,484 km2 and is landlocked. The main landform types include plains (42% of the country’s land area), lakes (20%),  10 and hills and ridges (12%) (Dijkshoorn et al., 2016). The main soil types are Lixisols (26% of the land surface), Luvisols (22%) and Cambisols (18%) (Dijkshoorn et al., 2016). A relatively dry, seasonal sub-tropical climate occurs in the country, with some elevational variation (Malawi Goverment, 2006). The warm, rainy season occurs from November to April with maximum temperatures ranging from 20C to >32C; the cool, dry season occurs from May to August, with minimum temperatures ranging from 4C to 10C (Malawi Goverment, 2006).   The miombo woodland, dominated by species of the Brachystegia, Julbernardia and Isoberlina genera, is the primary ecosystem type historically occurring in Malawi (Frost, 1996). Extending 2.7 million km2, the miombo woodlands are the most extensive dry forest in Africa (Frost, 1996). Miombo woodlands are categorized into dry and wet types (Frost, 1996). The dry miombo woodland – occurring in southern Malawi, Mozambique and Zimbabwe – is characterized by a tree canopy height typically less than 15 m and a tall grass understorey (Frost, 1996). Soils in the miombo woodlands are low in nutrients, well drained and acidic (Frost, 1996). In lower rainfall areas, Cambisols and Luvisols predominate in the woodlands, whereas in higher rainfall areas Ferralsols and Acrisols dominate (Frost, 1996). In Malawi the land tenure of the miombo woodlands is held either by the state (in forest reserves), by private owners (in private woodlots or woodlands), or by communities (in village forests) (Zulu, 2010). Many households rely on nearby forest reserves for the collection of forest products (Dewees et al., 2010; Kamanga et al., 2009). While felling of live trees is prohibited in the state-owned forest reserves, harvesting of certain forest products is allowed, such as medicinal plants, fruits and wild foods, fodder, thatch, and dry wood (Kamanga et al., 2009).  Population density in Malawi is high with a total population of 18 million and average density of 152 people per km2 in 2018. In southern Malawi, where the present research took place, the population density is the highest in the country at 184 people per km2 (Meijer et al., 2016). Annual population growth rates in the country are relatively high at 2.94% between 2010 and 2015 (United Nations, 2017b). Malawi is one of the world’s poorest countries, with the sixth lowest per capita purchasing power parity in 2017 (International Monetary Fund, 2017; United Nations, 2017a). The majority of the population are impoverished small-holder farmers. HIV/AIDs rates are among the highest in the world; in 2016 approximately one million people were living with the  11 disease, although rates of new incidence of HIV/AIDS and death have decreased since 2010 (UNAIDS, 2017). As a result of the death toll from HIV/AIDS, the number of orphans in Malawi is high: 16.7% of youth under the age of 18 are orphans or vulnerable children (USAID, 2016). Food insecurity and malnutrition are also key health issues in the country with the primary causes being poverty and environmental stressors (Misselhorn, 2005).  Deforestation rates in Malawi are high at 0.6 to 1% annual forest cover loss between 1990 and 2015, with similar or higher degradation rates (FAO, 2015; Kamanga et al., 2009). The key anthropomorphic stressors impacting the miombo woodlands are land conversion to agriculture and fuelwood harvesting (Fisher, 2004; Zulu, 2010). The land cover of southern Malawi has largely been converted from miombo woodlands into perennial or annual cropland (Haack et al., 2015). If faced by land scarcity, some households turn to illegally encroaching on the state-owned forest reserves to cultivate crops (Kishindo, 2004). Agricultural land tenure in the country is customary, where village chiefs distribute land holdings within the community and land is passed on within the family through generations (Place & Otsuka, 2001). The purchase of land is prohibited in this system yet occurs on a rare basis and land rentals are also rare, thus there are few options to respond to land scarcity, which is an issue especially in more densely-populated areas  (Kishindo, 2004; Takane, 2008). The other key stressor, fuelwood harvesting, occurs primarily in the form of charcoal production and secondarily for firewood for tobacco curing and fish smoking (Abbot & Homewood, 1999; Geist, 1999). The majority of Malawi’s population lack access to electricity and all households use fuelwood as either a primary or secondary energy source for cooking and/or heating (Dasappa, 2011). Charcoal production, which is currently illegal, has a 23% unit conversion from wood to charcoal, thus high amounts of live wood are required to produce charcoal and rapid tree cover loss ensues (Kammen & Lew, 2005). Charcoal production in southern Malawi is widespread and has caused degradation in many forest reserves (Dewees et al., 2010).       12 1.6.2. The research project  I joined a team to conduct a pilot study from June to August of 2016 in southern Malawi which focused on the themes of forest conservation and livelihood sustainability. The team included two Principle Investigators, Dr. Jeanine Rhemtulla and Dr. Joleen Timko, and six UBC students including myself. Our group was interested in the following questions: how dependent are farming communities on forest resources for their livelihoods; how has forest composition changed in areas of fuelwood harvesting; how is reforestation being used for social and ecological benefits; and what are the factors that have led to the failure of reforestation projects. We conducted surveys, interviews, focus groups and forest plots to investigate these topics.   The research involved in this thesis took place in the Zomba and Machinga districts of southern Malawi. I led household surveys (n=157), enumerated by Malawian translators, in nine agricultural villages adjacent to the Zomba-Malosa forest reserve in the Zomba district. I took measurements in forest plots (n=86) in the Zomba district in the Zomba-Malosa forest reserve where community members harvest fuelwood, and in reference sites of miombo woodlands nearby in the adjacent Machinga district in the Malosa and Liwonde forest reserves, and in Liwonde National Park (Figure 1.1). The study area was selected as Dr. Timko had conducted previous research in the area and thus we had local assistance from a community collaborator who is the leader of a local community-based organization called Tikambirane that is concerned with climate change and community well-being.            13  Figure 1. 1. Nested map of the study area in southern Malawi, Africa. Zomba district is in the lower section of the labelled study area map and Machinga in the upper right.  Forest reserves and national park labelled where forest assessments were conducted. Household surveys were conducted in nine villages to the east of the Zomba-Malosa forest reserve.   My team selected households to survey by tagging all rooftops in the study area in Google Earth Pro and using an algorithm to randomly select households (Google, 2013). At the end of the survey, our translators asked if the respondents would be willing to have my team accompany them on one of their fuelwood harvesting trips. From those who agreed, I generated a random sample of households to accompany into the Zomba-Malosa forest reserve to conduct the forest plots. To locate reference sites, I consulted the local District Forest Officer for Machinga, local forest guides and local botanists to select the least disturbed, most mature miombo woodlands near the Zomba-Malosa forest reserve.   14 1.7. Research objectives  With the overall objective of understanding the social-ecological system in southern Malawi that includes the miombo woodlands and the agricultural communities that are dependent on the forests, I undertook social and ecological research to address the following two questions:  1. What are the levels of forest dependence in agricultural villages, as measured using a proposed new index? 2. How do tree characteristics differ in miombo woodlands with fuelwood harvesting versus relatively undisturbed forests?   For the social assessment, I led a team to conduct household surveys (n=157) in agricultural villages adjacent to the Zomba-Malosa forest reserve and collected data on demographics, income, assets and forest use. In Chapter 2, I present a new index for measuring forest dependence. I outline the importance of the index, present its structure and calculation, and apply the data from the household surveys in southern Malawi to calculate the forest dependence of the households that I surveyed. I also compare the index values to the relative forest income approach of calculating forest dependence.  For the ecological assessment, I took measurements in forest plots (n=86) in the Zomba-Malosa forest reserve in areas where households harvest fuelwood and in reference sites of relatively undisturbed miombo woodlands.  In Chapter 3, I address the following questions:  1. How do tree species composition, diversity and aboveground carbon (AGC) differ in sites with fuelwood harvesting compared to reference sites? 2. To what extent are harvesting access (measured by elevation and distance to roads) and harvesting pressure (measured by proximity to settlements) predictors of tree diversity and AGC within fuelwood harvesting sites?    15 I conclude by summarizing and synthesizing the findings from the ecological and social research to discuss the social-ecological system as a whole, and I provide management suggestions in response to my findings. I discuss the importance of the research for the study area in southern Malawi, as well as for similar contexts in sub-Saharan Africa and elsewhere in the tropics.                                     16 Chapter 2. Measuring household forest dependence using a new index based on the collection of forest products  2.1. Introduction   Households in the global south rely on forests for numerous ecosystem services. Small-holder farmers living adjacent to forests collect multiple forest products for subsistence use and in some cases sale, particularly fuelwood, building materials, wild foods and medicinal plants (Angelsen et al., 2014; Byron & Arnold, 1999; Sunderlin et al., 2008). In impoverished communities household dependence on forests for provisioning ecosystem services may be high, especially in the absence of alternatives (Angelsen & Wunder, 2003). Forest-dependent households expend time and effort to harvest these forest products making it a labour-intensive livelihood strategy (Angelsen et al., 2014; Pattanayak & Sills, 2001). When deforestation and forest degradation occur, it is often the forest-dependent poor who have the most to lose.  Poverty is interrelated with dependence on forests in multiple ways. High forest dependence together with poverty may indicate a lack of access to livelihood alternatives, but high forest dependence may be a cause of poverty (Angelsen & Wunder, 2003). Forests may act as safety nets for the rural poor in times of scarcity or gap fillers in times of seasonal shortfalls (Angelsen & Wunder, 2003). Forest dependence may also be seen as a cause of poverty, however, because most forest products are economically marginal and have poor income generation potential (Angelsen & Wunder, 2003). Poverty and forest dependence are typically measured using household income (Angelsen et al., 2014; Angelsen & Wunder, 2003; Vedeld et al., 2007), a point on which I will elaborate later. When discussing levels of poverty and forest dependence it is important to differentiate between relative and absolute income that is derived from forests, absolute forest income being the sum total of the monetary value of products harvested from the forest, and relative forest income being the proportion of household income comprised by forest income. Case studies have shown that relative forest income decreases as total household income increases (Campbell & Luckert, 2002; Cavendish, 2000), yet this relationship was not found when many case studies were compared on an international scale (Angelsen et al., 2014). The use of forests for commercial purposes may increase relative forest income for wealthier households and limited access to forests may decrease relative forest income for the less wealthy (Angelsen et al.,  17 2014). Absolute forest income has also been shown to increase as total household income increases (Angelsen et al., 2014).   How we measure poverty also matters in assessments of the relationships between poverty and forest dependence. Measurements of poverty beyond income-based measures, for instance using asset-based measures, show slightly different trends than those described above for relative and absolute forest income. A case study in the Democratic Republic of Congo, for example, showed that households in intermediate wealth quintiles (determined using asset holdings) had the highest relative and absolute forest income (Nielsen et al., 2012). Measurements of income and assets together help to identify different types of poverty, including the chronic poor (low income, low assets), transient poor (low income, high assets), transient rich (high income, low assets), and chronic rich (high income, high assets) (Nielsen et al., 2012). The different groups depend on forests in different ways, for example with a higher proportion of transient and chronic rich households harvesting more valuable forest products (e.g. timber and poles) than poor households (Nielsen et al., 2012). Rural households with higher wealth generally have more options in preparing for and responding to economic shocks and thus may be less forest dependent (Dercon, 1998). In general, forests support the consumption needs of the chronic poor, act as a safety net for the transient poor, and may provide a pathway out of poverty for the transient rich (Nielsen et al., 2012).  Given high rates of deforestation and forest degradation in the tropics concurrent with community dependence on forests, quantifying the extent of forest dependence is important for informing approaches to forest-based poverty alleviation (Sunderlin et al., 2004). The objective of this chapter is to propose a new index that measures household forest dependence which incorporates multiple livelihood dimensions involved in the collection of forest products. I tested the efficacy of the index using a case study in Malawi and addressed the following questions: how do the forest dependence index values compare when forest products are included from forests versus other locations? How do the distributions of the sub-indices contribute to the distribution of the index values? How robust are the index values? Lastly, I compared the index values to the currently used relative forest income method for measuring forest dependence. In the next section, I review the literature on forest dependence and I introduce the use of indices. I then describe how  18 I developed the proposed index, tested it with data from the case study in Malawi and compared it to the relative forest income method. I present the findings and discuss the importance of the new index.  2.2. Defining forest dependence   Several definitions of forest dependence have been discussed in the literature and their disparities impact approaches to quantification. Byron and Arnold (1999) discussed the multiplicity of notions of “forest dependent peoples” by describing multiple types of forest dependence and identifying a spectrum of reasons that households depend on forests, from a purposeful choice to a last resort. There is also a range in the types of relationships communities have with forests, from those that are strictly economic to those with strong cultural and spiritual dimensions (Byron & Arnold, 1999). Sunderlin et al. (2005) described forest dependence in terms of reliance on ecosystem services, subsistence needs, safety nets, gap fillers and opportunities for poverty elimination. Angelsen and Wunder (2003) differentiated forest use for occasional safety nets or gap fillers with regular derivation of income. In discussing definitions of dependence for estimating global numbers of “forest dependent peoples”, the authors raised two types of dependence. The first is for “a dominant source of subsistence and cash income,” and the second is in “supplementary way[s].” The authors continued by describing five aspects of forest benefits: different groups of beneficiaries; types of forest products and services; the role of forests in the household economy or livelihood strategy; the extent of forest resource management; and use of high versus low value forest products. Non-timber forest products, timber, and additional ecosystem services were further identified as three economic benefits of forests  (Angelsen & Wunder, 2003). There may not exist a universally agreed upon definition of forest dependent peoples (Newton et al., 2016), thus when measuring forest dependence the relationship(s) that the communities involved have with the forest and the type of forest dependence should be clarified.   In the measurement of forest dependence, related concepts such as “forest use” and “reliance on forests” also need to be differentiated. Forest use includes: the practices involved in harvesting forest products, the use of other provisioning services from forests such as water, the use of forests as land for agriculture, and use for other cultural practices (Anthwal et al., 2010;  19 Campbell, 2005; Sunderlin et al., 2005). Forest reliance has been described in terms of economic standing as a form of insurance, support for consumption, and a means of poverty reduction (Angelsen & Wunder, 2003; Prado Córdova et al., 2013; Vedeld et al., 2007).  Reliance tends to be used when referring to economic reliance for income, and some researchers use the terms reliance and dependence interchangeably, or use the term reliance in their definition of forest dependence (Fisher, 2004; Sunderlin et al., 2005; Wunder et al., 2014).  I view the concept of forest dependence as subsuming use and reliance, to assess the broader scale of livelihood dependence on forests to meet household needs. Hence, measuring forest dependence should include assessing the provisioning services provided by forests as well as household economic standing related to livelihood dependence on forests.  Additionally, the terms forest and forest products should be defined, and the ecosystem from which forest products are collected should be specified. The FAO defines a forest as a treed ecosystem with a minimum of 0.5 hectare area, a minimum canopy height of 5 metres, and minimum canopy cover of 10% (Food and Agriculture Organization, 2006). Forest products are all physical goods of biological origin – either plant, animal or fungi – derived from forests (Belcher, 2003). Given forest degradation and conversion of forests to other land use/land cover types, households in rural mosaic landscapes collect “forest products” from several ecosystem types, including non-forest ecosystems (Angelsen & Wunder, 2003). Forest products will contribute to environmental income, the income derived from all “non-cultivated sources,” including forests and woodlands, wildlands (ex., savannahs and grasslands), bushlands, wetlands, fallows and wild (uncultivated or undomesticated) plants and animals harvested from croplands (Angelsen et al., 2014; Sjaastad et al., 2005). Forest environmental income is differentiated from non-forest environmental income, as the collection of forest products specifically from forest ecosystems. When measuring forest dependence, the focus should be on quantifying forest products collected from ecosystems that follow standard forest definitions, including woodlands.  2.3. Measuring forest dependence  Forest dependence is currently measured using forest income, which focuses on economic dependence. The forest income approach involves eliciting the quantities of forest products  20 collected by a household as well as the local monetary value of the products, to quantify the income equivalent for the household of harvesting forest products (Cavendish, 2000). Early seminal studies by Cavendish (2000) and Campbell and Luckert (2002) on forest and environmental incomes inspired several further studies, including the Poverty and Environment Network (PEN) project (Angelsen et al., 2014; Wunder et al., 2014). Significant for its breadth of 24 countries and roughly 8000 surveys, the Centre for International Forestry Research (CIFOR) led the PEN global comparative analysis using the forest income measurement (Angelsen et al., 2014). Forest income was determined to comprise on average 26.8%, 20.1% and 21.4% of household income in Latin America, Asia and Africa respectively (Angelsen et al., 2014).  The forest income approach stems from the history of income quantification in household economics research and policy. Traditional notions of poverty involved income and material wealth, following classical economists such as Adam Smith and David Ricardo (Angelsen & Wunder, 2003). Until the 1960s, policies on poverty alleviation largely emphasized increasing household income. In recent decades, however, the definition of poverty has been extended to include non-material aspects of well-being, such as health and education (Angelsen & Wunder, 2003). To understand the complexity of rural livelihoods, researchers acknowledged the need for measurements of poverty that extend beyond cash income (Sunderlin et al., 2005). Measuring, describing and analyzing poverty have hence expanded to become more holistic, such as with the sustainable livelihoods framework which incorporates multiple forms of capital – natural, human, social, physical and financial – to reflect the multiple resources available for livelihoods (Angelsen & Wunder, 2003; Scoones, 1998). In a similar vein, capturing the complex nature of forest dependence by going beyond income measurements may be beneficial for better understanding the livelihood implications of dependence, yet such approaches have not yet been attempted.   Measuring forest dependence using forest income has the key advantage of a common unit (income), yet the method has limitations. Household incomes fluctuate monthly, seasonally, and annually, and households may not keep rigorous accounts of their income flows, leading to decreased accuracy of income estimates through household surveys (Nielsen et al., 2012; Rutstein, 2008). Additionally, estimating forest income through amounts of forest products used does not account for the burden or inconvenience households face in the collection of forest products. The  21 effort involved is significant – household forest collection has been quantified as a function of household labour in a model that relates agricultural risk to the collection of forest products (Pattanayak & Sills, 2001). Here I assume forest product collection is perceived by households as a burden, although in some communities collection may not be perceived as such, where forest collection is undertaken as a part of life which contributes to customs and rituals, such as in the harvesting of medicinal plants (e.g. Cunningham, 1993; Lebbie & Guries, 1995). Time poverty is also an issue for household members, particularly for women who are responsible for much of household forest collection especially in Africa, because their opportunities to engage in other aspects of the household economy or personal development are limited by the lack of time (Blackden et al., 2006; Sunderland et al., 2014). Lastly, a key limitation of measuring forest income arises in situations when households primarily consume rather than sell products, which makes the monetization of forest products unrealistic. Given the limitations of the forest income method, and the additional factors involved in livelihood dependence for provisioning ecosystem services, I propose the use of a new index to measure forest dependence that is complementary to the forest income measurement.  2.4. Uses of indices  Indices, also called composite indicators, have been widely used in research and policy, particularly to assess economic development on the macro (international) level (Booysen, 2002). Indices are recognized as a valuable tool for their ability to illustrate complex issues from a variety of disciplines with interpretable values (Nardo et al., 2005). While countries are often evaluated and compared, the household level has also been used in indices, such as in the Energy Poverty Index (Mirza & Szirmai, 2010). I follow this approach and propose a new index to illustrate the complex issue of forest dependence at the household level.  The basic steps in constructing indices inform the design issues I highlight for a forest dependence index. The steps include selection, scaling, weighting and aggregation, and validation (Booysen, 2002). Selection involves the choice of relevant and measurable sub-indices and variables (Freudenberg, 2003). Scaling involves employing an approach to data normalization such as ranking, standardization or re-scaling so that data can be aggregated (Freudenberg, 2003; Nardo  22 et al., 2005). Weighting is used to assign the relative importance of components, whether equal or not (Freudenberg, 2003). Aggregation combines components using one of multiple possible methods: linear or geometric, additive or multiplicative, and multivariate (Nardo et al., 2005). Lastly, the validation process tests for robustness of the index, such as through uncertainty and sensitivity analyses (Saisana & Tarantola, 2002).   2.5. Framework for a new forest dependence index  A key advantage of using an index to measure forest dependence is the ability of indices to summarize complex issues into a single metric (Saisana & Tarantola, 2002). The index incorporates multiple livelihood aspects of forest dependence. In the present work, I focus on measuring provisioning services provided by forests. I nonetheless acknowledge the existence of numerous additional forest ecosystem services both qualitative and quantitative, including cultural, regulating and supporting services (Daily, 1997; Maass et al., 2005). I have identified four key livelihood aspects of household forest dependence for provisioning services that constitute the four sub-indices of the proposed forest dependence index: forest product use; effort expended in the collection of products; household relative wealth; and number of non-forest livelihood strategies.   The first two sub-indices concern the collection of forest products while the last two concern household livelihoods in terms of adaptive capacity and alternatives to forest collection. The diversity of collected forest products is quantified in the index, including the extent to which a household relies on each product. With increasing reliance on multiple forest products to meet household needs (Shackleton et al., 2011), dependence increases. The effort involved in collection is also quantified. With increasing labour (daily, weekly or monthly) spent on collecting forest products (Pattanayak & Sills, 2001), dependence increases. To incorporate relative household wealth given the importance of poverty in forest dependence, the index measures relative wealth using an asset-based approach. Using an asset as opposed to income measurement portrays long-term economic status more effectively, as assets are accumulated with savings over time (Filmer & Pritchett, 2001). With decreasing wealth, households lack alternatives to forest collection and dependence increases (Angelsen & Wunder, 2003). Lastly, the number of non-forest livelihood strategies a household employs is quantified to portray alternative strategies to forest collection.  23 Rural households tend to diversify their livelihoods to enhance income sources and reduce risk (Sunderlin et al., 2005). As the households in this context are primarily small-holder farmers, the key non-forest livelihood strategy is farming. Households may undertake additional livelihood strategies, such as wage labour, running a business or renting out land. The number of livelihood strategies will indicate economic standing in the short to medium term (Barrett et al., 2001). A decreasing number of non-forest livelihood strategies, hence a less diversified livelihood, means that household dependence on the forest increases because they have fewer livelihood alternatives to adapt and respond to economic shocks (Ellis, 1999; Wunder et al., 2014).   I will now identify the essential variables within each sub-index and discuss possible additional variables for use if datasets and contexts allow. The variables must be measured using a common unit for all forest products and all households.  The forest product use sub-index measures the amount of a product collected from the forest divided by the total amount of the product used by a household. For example, the amount of fuelwood collected from the forest is compared to the total use of fuelwood from all sources, including fuelwood that is purchased. The forest product use values will indicate the importance of collected forest products for meeting a household’s needs. The essential variables for the forest product use sub-index are the amounts of each forest product harvested from the forest and the amounts of each product from non-forest sources.  The effort sub-index measures the potential burden a household faces by expending time and physical effort to collect forest products. The total number of forest products used will be reflected in the aggregation of the values of each. The essential variables in the effort sub-index include the amount of the forest product and the time spent walking to the collection site. Additional relevant variables that could be incorporated include the value of forest products compared to one another (Angelsen et al., 2014), time spent collecting forest products (Brouwer et al., 1997), frequency of collection (Liswanti et al., 2011), number or proportion of household members who collect forest products, number of genders who collect (Sunderland et al., 2014), and number of age groups who collect (Paumgarten, 2005).     24 The third sub-index is relative wealth, which is important because it impacts a household’s access to alternatives to forest product collection hence impacting their forest dependence. Following the DHS asset-based calculation of a wealth index, the essential variables are the household asset holdings, which are used to compare households’ wealth by constructing a wealth index (Rutstein, 2008).   The fourth sub-index is the number of non-forest livelihood strategies that a household engages in, which is a measure of whether households have livelihood alternatives to forest collection. The variable in this sub-index is the number of livelihood strategies that provide income or subsistence production (i.e., farming). Agriculture, wage labour, and businesses are examples, as well as remittances and cash allowances from governments or NGOs.   While several methods of aggregating a forest dependence index are possible, I propose an ideal metric composition and also mention alternatives. The proposed aggregation method uses an additive approach and weighting occurs at the forest product level (Figure 2.1). By aggregating at the forest product level, the forest product use and effort values remain together per forest product as they are combined first, followed by the summation of the values for all forest products per household. Because the data for the forest product use and effort are combined for individual forest products, weights could be assigned to forest products depending on their comparative value to a household. An alternative weighting approach would be to aggregate at the sub-index level. Such an approach would calculate each sub-index separately at first, then combine all of the sub-indices, which would therefore separate the forest product use and effort values for individual forest products, hence forest products could not be weighted separately. The sub-indices could, however, be weighted separately. With the proposed method one could assign weights to the two sub-indices that capture information on forest products (forest product use and effort) and the two sub-indices that capture information on livelihoods (relative wealth and non-forest livelihood strategies).    25  Figure 2. 1. Flowchart showing the steps involved in calculating the forest dependence index. Aggregation is addition-based, and occurs at the forest product level meaning that forest product use and effort values are combined per forest product first, then summed over all products at the household level. Abbreviations are as follows: F = forest product use, E = effort, W = relative wealth, L = non-forest livelihood sources, A = amount, T = time, n= total number of forest products. Subscripts are as follows: i = forest product level, j = household level, o = other source. The z-scores standardization method is used across the sample population for all intermediate steps, and the re-scaling normalization method is used in the final calculation of the index.  The proposed aggregation method is additive, which is used in cases where data are in a partially or fully comparable interval scale; if the interval scales are not comparable, a geometric weighting scheme must be applied (Ebert & Welsch, 2004; Nardo et al., 2005). Another aggregation method, the multi-criteria or multivariate approach can be used when index components have varying significance and require different weights (Nardo et al., 2005). An additive approach implies full compensability, where low performance in some variables can be overcome by higher performance in others, whereas a multi-criteria approach implies non- 26 compensability, and a geometric approach falls in between the two (Nardo et al., 2005). Standardization of the household values of the variables means that the data will be in a comparable interval scale.   2.6. Testing the forest dependence index using the case of southern Malawi   2.6.1. Study area  To assess the efficacy of the proposed forest dependence index, I used data from a case study in southern Malawi, Africa. Malawi is an ideal location in which to study levels of forest dependence. Among the most impoverished countries in the world, Malawi had the sixth lowest per capita PPP in 2017 (International Monetary Fund, 2017; United Nations, 2017a). Malawi is densely populated as a landlocked country with an area of 118,484 km2 and population of 18 million in 2018. The majority of the population are small-holder farmers (Fisher, 2004). Less than 10% of the population have access to electricity, and all households use firewood or charcoal for fuel, either as a primary or secondary energy source (Dasappa, 2011; Zulu, 2008). Rural households burn on average 5 kg of fuelwood per day (Nerfa et al., 2016, unpublished data).  Deforestation rates are high at 0.6 to 1.0% forest cover loss per annum from 1990 to 2015; forest degradation rates are similar if not higher (Food and Agriculture Organization, 2015; Kamanga et al., 2009). The main proximate causes of forest loss and degradation are land clearing for agriculture and charcoal production (Fisher, 2004; Zulu, 2008). Agricultural communities in Malawi collect fuelwood and other products from nearby forests, such as village forests on communal lands, state-owned forest reserves, and private woodlots (Zulu, 2010). In forest reserves, harvesting of certain products is permitted, including dry wood, thatch, fodder, medicinal plants, fruits and wild foods (Kamanga et al., 2009).  2.6.2. Data collection  Household surveys (n=157) were conducted from June to July in 2016 to quantify how forest products contribute to household livelihoods. The study boundary encompassed nine villages in one water catchment adjacent to the Zomba-Malosa forest reserve in the Zomba district of southern Malawi (See Figure 1.1). Our local collaborator facilitated initial meetings with village heads to receive permission for the current research. I randomly selected households to survey by  27 identifying and tagging rooftops in Google Earth Pro then used a random number algorithm to select the houses to visit (Google, 2013). Selection included 40% of all rooftops, providing a buffer in case of absentee households or incorrectly identified buildings found when navigating to the selected household for surveying. The final sample comprised approximately 30% of households per village.  The household survey was designed by adapting the PEN prototype questionnaire (Centre for International Forestry Research, 2008). Note that the forest dependence index was developed after the design and execution of the household surveys. The surveys were conducted with informed consent by research assistants hired as translators, in the local language Chichewa. The survey questions focused on household demographics, assets, income and production, and collection of forest products. Demographics included age, gender, education, occupation and special community roles (ex. village official) of all household members. Assets included buildings owned, valuable items owned, savings and land holdings. Income and production included all sources in the last year: agriculture, livestock, animal products, wage income, business, and other sources (ex. remittances), as well as the associated input costs (where applicable). Note that agricultural crops, livestock and animal products were included that generated income or that were used for subsistence purposes. Forest product categories included fuelwood, timber, wild food (ex. fruit, honey, mushrooms), wood for crafts, medicinal plants, fodder, and animals hunted. Forest product collection locations included forest reserves, village forests, private woodlots, homesteads, and the household’s own farm. For each forest product, the households were asked where they collect the product, how much they collected in the past month and how long it takes to walk to the collection site on a typical trip.   2.6.3. Calculation of the forest dependence index   I applied the data from the household surveys to calculate  my new forest dependence index. I used the z-scores standardization technique across the sample population (𝑥−𝜇𝜎 where  is the mean of the sample and  is the standard deviation of the sample) throughout the index calculation which results in a mean of 0 and standard deviation of 1, until the final step at which time I normalized the final index value using re-scaling across the sample population (𝑥−𝑚𝑖𝑛𝑚𝑎𝑥−𝑚𝑖𝑛   28 using the minimum and maximum values of the sample), which results in a minimum of 0 and maximum of 1. I used fuelwood and wild foods only, since all of the essential variables were available for these products, but not for timber, craft wood, medicinal plants, fodder, and animals, because we did not collect data on other sources of these forest products. Multiple meta-analyses have found that fuelwood and wild foods were the two most important forest products in economic forest dependence (Angelsen et al., 2014; Vedeld et al., 2007). I treated the forest products equally, although weights could be assigned. I first calculated the values for the collection of fuelwood and wild foods from village forests, forest reserves and private woodlots, which meet the FAO definition of forests. I also calculated the index using values from all locations where households collect forest products (village forests, forest reserves, private woodlots, homesteads and farms), to make a comparison and determine the importance of identifying the location where households are harvesting forest products. I conducted Kolmogorov-Smirnov tests to check for differences in the distributions of the household values for the study area using the two types of locations, and conducted paired t-tests comparing the mean values.   For the forest product use sub-index, I determined the average amount of fuelwood and wild food collected from the forest locations and the amount acquired from all other sources in the past month, in kilograms, for each household. ‘Other sources’ of fuelwood included amounts collected from the homestead and farm, and the amount purchased in Malawi Kwacha (MWK). I converted the monetary value in MWK to a kilogram value using the market value of 35 Kwacha per kilogram, as informed by our local community contact. ‘Other sources’ of wild foods were the amounts collected from the homestead and farm, wild foods purchased (only three households reported this), and the household’s crop production in the past month, where I divided the annual amount in kilograms of all crops produced by 12. The monthly crop estimate has two caveats: that households may not consume the entirety of their crop production and that crop production is not distributed evenly throughout the year. I used this calculation, however, as an estimate in the absence of data on amounts of foods consumed. For each forest product, I divided the amount collected from the forest by the sum of the amount collected from the forest and the amounts acquired from other sources. I then standardized the values using z-scores. Note that I also calculated the forest product use values with forest products collected from all locations (including the homestead and farm) for the separate computation of the index.  29 To calculate the effort sub-index, I began by standardizing the amounts of fuelwood and wild food in kilograms collected over the past month, and standardizing the walking times to the collection sites in minutes, across the sample population. I standardized the amounts and times per forest location for each forest product and each household then took the average, because the contexts of the locations were different. For example, walking distances to the village forest would be shorter than to the forest reserve and should not be directly compared. I then summed the standardized amount and the standardized walking time by product for each household, and standardized the values again. Although I used the product amount values in both the forest product use and the effort sub-indices, the values were used differently. Under the forest product use sub-index, the values were treated internally to the household as a proportion value followed by standardization across the sample population, whereas in the effort sub-index the values were only standardized across the sample population. I also calculated the effort values with forest products collected from all locations (including the homestead and farm) for the additional computation of the index.  To calculate the relative wealth sub-index, I calculated the wealth index following the DHS method (Rutstein, 2008). I conducted principal components analysis (PCA) on all assets reported for each household using SPSS (IBM Corp. Released, 2016). Asset units were either binary or continuous, the latter either counts of objects or values in hectares (for land holdings and buildings owned). I removed assets owned by either less than 5% or more than 95% of the population. For households with missing values I used mean substitution. All variables were individually standardized across all households prior to conducting the PCA. I then determined the household wealth index value by taking the sum of the asset value multiplied by the PCA loadings value for all assets, for each household. I organized households into wealth quintiles and tested for monotonicity of asset values (linear increase or decrease with increasing quintile). I removed any assets that did not exhibit monotonicity and re-ran the analysis to determine the final relative wealth values. The final assets included use of a community tap water source (binary), use of a river water source (binary), land area of buildings owned, number of buildings with brick walls, number of buildings with a natural roof, number of buildings with a metal roof, number of buildings with glass windows, number of buildings with a solid door, area of total land holdings, access to irrigation (binary), number of chickens owned, number of ducks owned, number of  30 bicycles, cell phone ownership (binary) and radio ownership (binary). The wealth index values were standardized when computed because the household asset values had been standardized.  For the non-forest livelihood strategies sub-index, I counted the number of non-forest livelihood sources from which households had derived income or produced for subsistence use in the past year. Possible types were: agriculture (crops produced for income or subsistence use), livestock sold, animal products sold, wage income, businesses owned, land rented out, remittances, and social assistance. The last category included social cash transfers, farming input vouchers and hunger relief vouchers from the government, as well as cash support from NGOs. I determined the total number of non-forest livelihood strategies per household, then standardized the values across the sample population.  To aggregate the sub-indices, I first summed the standardized forest product use and effort values per product then took the sum over all products for each household and standardized again. I took the sum of the wealth index and the livelihood strategies sub-indices, standardized the value, and took the inverse (1 minus x), to reflect a greater forest dependence for those households with lower wealth and fewer livelihood strategies. I took the sum of the standardized forest product use/effort value and the inverse standardized wealth index/livelihood strategies value. Lastly, I normalized the index values across the sample population using re-scaling to fix the values from 0 to 1.   To test the robustness of the index I computed Cronbach’s Alpha, and the first-order sensitivity index for each sub-index. I calculated Cronbach’s Alpha to test for internal consistency of the index: (𝐾𝐾−1) ∗ (1 −∑ 𝑣𝑎𝑟(𝑥𝑛)𝐾𝑛=1𝑣𝑎𝑟(𝑥𝑜)) where K is the number of sub-indices, var(xn) is the variance in sub-index n, and var(xo) is the variance of the sum of the sub-indices. To determine the contribution of the variance in the sub-indices to the variance in the final index value, I calculated the first-order sensitivity index for each sub-index by dividing the variance of the sub-index by the variance in the index. I did not perform uncertainty or sensitivity analyses following standard methods for indices because the forest dependence index calculation did not have variation in the typical sources of uncertainty, including data imputation, normalization, weighting,  31 and aggregation methods (Nardo et al., 2005). I used only one method for imputation of missing data (mean substitution), one normalization technique (standardization) other than the re-scaling of the final value, one method of aggregation (linear, additive), and did not apply weighting. If future users of the index wish to use multiple methods of the above steps, uncertainty and sensitivity analyses should be performed.   2.6.4. Calculation of relative forest income    To compare my proposed forest dependence index with the relative forest income method of measuring forest dependence, I used the household survey data to calculate forest income and total income for all households. Using the average local market values in MWK (informed by our local community contact), I converted the amounts of forest products collected into forest income. For fuelwood, I multiplied the kilogram value of the fuelwood collected monthly by 12 for an annual estimate, then by the MWK value per kilogram. For wild foods, I multiplied the amount collected in the past month by six – to account for the six-month dry season and six-month wet season which affects the availability of plant products – then multiplied by the market value per kilogram. Annual forest income was estimated from amounts collected from village forests, forest reserves and private woodlots, and separately from all forest locations for a comparison. In the former case, I added the monetary values of amounts collected from the homestead and own farm to the total non-forest income. I summed all annual income estimates of forest products. To estimate total household income in the past year, I summed all forest and non-forest income sources, and subtracted the input costs to non-forest income sources as applicable. For agriculture, the average local market value in MWK was applied to the total amount of each crop produced and the sum taken. For livestock and animal products, the sum of the income from sales in MWK was taken. If total income from non-forest sources was negative, due to higher input costs than income, the value was set to zero. I also calculated the U.S. Dollar equivalents of the Malawi Kwacha income values, estimated as 1 MWK = 0.0014 USD for the rate in 2018. Lastly, I calculated relative forest income by dividing forest income by total household income. I compared the relative forest income values to the forest dependence index values, for the values from village forests, forest reserves and private woodlots and separately from all forest locations, using Kolmogorov-Smirnov tests and paired t-tests.   32 2.7. Levels of forest dependence in southern Malawi  The distribution of the household forest dependence values from the new index was approximately normal, which suggests that the majority of households had intermediate levels of dependence (Figure 2.2). In the study area (n=157) the mean forest dependence value from village forests, forest reserves and private woodlots was 0.49 and the standard deviation was 0.14. While I expected a normal distribution following the central limit theorem, given that I added independent variables and normalized the sum, the distribution of the forest dependence index values was not perfectly normal. This is likely because the distributions of values for the variables in the sub-indices varied.    Figure 2. 2. Histograms of household values (normalized using re-scaling) for the forest dependence index, for fuelwood and wild foods collected from true forests (village forests, forest reserves and private woodlots; mean=0.49, SD=0.14) and all locations (mean=0.47, SD=0.19) for the study area in southern Malawi (n=157).   The distribution of forest dependence values calculated using forest products collected from village forests, forest reserves and private woodlots versus all forest locations was significantly different (Figure 2.2; Kolmogorov-Smirnov D=0.15, p=0.03). The difference in the means, however, was not significantly different (paired t-test p=0.06). The difference in the distributions highlights the importance of identifying which locations of forest products will be considered when measuring forest dependence. I think it is reasonable to prefer the index values using the village forest, forest reserve and private woodlot forest locations that follow the FAO Forest Dependence Index From True ForestsFrequency0.0 0.2 0.4 0.6 0.8 1.0010203040Forest Dependence Index From All LocationsFrequency0.0 0.2 0.4 0.6 0.8 1.00102030 33 definition since we are measuring forest dependence, rather than dependence on trees (which could be planted on farms, in agroforestry systems, agricultural systems or near homes).  The distributions of each of the sub-indices were different from one another, despite the standardization of each of the variables which gives a standard deviation of 1 (Figures 2.3, 2.4). The distribution of the final index was the most similar to the inverse of the relative wealth and number of livelihood strategies sub-indices combined, which is aggregated with the forest product use and effort sub-indices in the final step of the index calculation (Figure 2.5). Wild foods forest product use had the largest range while fuelwood forest product use had the smallest range (Table 2.1). The vast majority of households had low wild foods forest product use values and few had high values. Similar numbers of households had both high and low fuelwood forest product use values. Most households had intermediate values of fuelwood effort while most households had low values of wild foods effort. See Figure 1 in Appendix A for a breakdown of the effort sub-index for fuelwood and wild foods into the distributions of amounts and walking times.                    34                                             Frequency-1.0 -0.5 0.0 0.5 1.0 1.5 Fuelwood Forest Product Use (Standardized)010203040Frequency-2 -1 0 1 2 3Fuelwood Effort (Standardized)0102030405060Frequency0 2 4 6 8Wild Foods Forest Product Use (Standardized)020406080120Frequency-1 0 1 2 3 4 5Wild Foods Effort (Standardized)020406080Figure 2. 3. Histograms of household values (standardized using z-scores) for the forest product sub-indices, including the forest product use and effort for fuelwood and wild foods collected from village forests, forest reserves and private woodlots, for the study area in southern Malawi (n=157).   35 Figure 2. 4. Histograms of household values (standardized using z-scores) for the livelihoods sub-indices, including the asset-based wealth index and the number of non-forest livelihood strategies, for the study area in southern Malawi (n=157).        Figure 2. 5. Histograms of household values (standardized using z-scores) for the forest product use and effort sub-indices combined – for fuelwood and wild foods collected from village forests, forest reserves and private woodlots – and the inverse of the relative wealth and livelihood strategies sub-indices combined, for the study area in southern Malawi (n=157).    Wealth Index (Standardized)Frequency-2 -1 0 1 2 3 4 50102030405060Number of Non-Forest Livelihood Strategies  (Standardized)Frequency-3 -2 -1 0 1 2 3 40102030405060Frequency-2 0 2 4Forest Product Use and Effort Sub-indices (Standardized)6020406080Inverse of Relative Wealth andLivelihood Strategies Sub-indices (Standardized)Frequency-3 -2 -1 0 1 2 3051015202530 36 Table 2. 1. Summary statistics for the sub-indices (standardized using z-scores), for the study area in southern Malawi (n=157). Forest product use and effort values are for fuelwood and wild foods collected from village forests, forest reserves and private woodlots.   In the assessment of the robustness of the index, for the values computed from village forests, forest reserves and private woodlots, the Cronbach Alpha value was 0.24, indicating relatively low internal consistency. This is reasonable as forest dependence is not necessarily a unidimensional construct, and there are conceptual differences between the four sub-indices, where forest product use and effort pertain to the collection of forest products while relative wealth and number of livelihood strategies pertain to adaptability of the household economy. The Alpha value for the index calculated with all forest locations was lower at 0.15. Applying the first-order sensitivity index, the variance in the forest product use, effort (from village forests, forest reserves and private woodlots), relative wealth and number of livelihoods sub-indices contributed 0.55, 0.55, 0.52 and 0.55 respectively to total index variance, thus each sub-index contributed roughly equally to the total index variance.   2.8. Comparison of the forest dependence index to the relative forest income method   Household incomes in the study area in Malawi were low; 97% of households were below the poverty line of $1.90 USD per person per day (Roser & Ortiz-Ospina, 2017). Agriculture was the most important income source for the households, followed by wage income and income from businesses owned (Appendix A Table 1). Forest income values from village forests, forest reserves and private woodlots were lower than forest income values from all locations (Table 2.2).   37  Table 2. 2. Annual forest, non-forest, and total household income in U.S. Dollars for the study area in southern Malawi (n=157). True forests are village forests, forest reserves and private woodlots; all locations include homesteads and farms in addition to true forests. Note that total income (forest and non-forest) is equal in the two cases because the income from fuelwood and wild foods collected from homesteads and farms is added to the total in each case, either to forest or non-forest income.   The new forest dependence index values differed greatly from, and tended to be higher than, those derived from the relative forest income method (Figure 2.6). For forest products collected from village forests, forest reserves and private woodlots, the mean value of the forest dependence index (0.49) was significantly higher than the mean value of relative forest income (0.13; paired t-test p<2.2e-16; Figure 2.6).  The distributions were also significantly different (Kolmogorov-Smirnov D=0.78, p<2.2e-16). The trend held for forest products collected from all forest locations, with the mean index value (0.47) higher than the mean relative forest income value (0.28; paired t-test p=7.30e-16; Figure 2.6). The distributions of relative forest income and forest dependence from all locations were significantly different (Kolmogorov-Smirnov D=0.48, p<2.2e-16). It is important to note that only fuelwood and wild foods were used in the forest dependence index and the forest income calculation, hence the distribution of values may change if more forest products were incorporated.   38  Figure 2. 6. Histograms of household values in the study area in southern Malawi (n=157) for forest dependence as measured using relative forest income, the proportion of total household income comprised by forest income, compared to the forest dependence index (normalized using re-scaling). Values for fuelwood and wild foods collected from “true forests” including village forests, forest reserves and private woodlots shown on the top (relative forest income mean=0.13, SD=0.18), where monetary values of forest products collected from homesteads and farms were added to the total non-forest income. Values from all locations shown on the bottom (relative forest income mean=0.28, SD=0.28).   I will now illustrate what leads to certain forest dependence values by discussing examples of households with low, medium and high index values. One household with a low forest dependence index value of 0.25 had a relative forest income of 0.005, a non-forest income in the past year of 537,500 MWK (753 USD), 4 non-forest livelihood strategies and a wealth index of 0.99. This household would spend 15 minutes walking to their private woodlot to collect fuelwood, collected 7 kg in the past month and had a standardized fuelwood forest product use value of 0.25. Relative Forest Income from True ForestsFrequency0.0 0.2 0.4 0.6 0.8 1.0020406080Forest Dependence Index from True ForestsFrequency0.0 0.2 0.4 0.6 0.8 1.0010203040Relative Forest Income from All LocationsFrequency0.0 0.2 0.4 0.6 0.8 1.0010203040Forest Dependence Index from All LocationsFrequency0.0 0.2 0.4 0.6 0.8 1.00102030 39 They collected no wild foods in the past month. A household with an intermediate forest dependence value of 0.55 had a relative forest income of 0.19, a non-forest income in the past year of 479,500 MWK (671 USD), 3 non-forest livelihood strategies, and a wealth index near the mean at 0.04. This household would spend 150 minutes walking to the forest reserve to collect fuelwood, collected 270 kg in the past month and had a fuelwood forest product use value of 1.36 (the highest of the standardized values). They collected no wild foods in the past month. Lastly, a household with the highest forest dependence index value of 1 had a relative forest income of 0.90, a non-forest income of 17,725 MWK (25 USD), 1 non-forest livelihood strategy and a wealth index of -0.92. This household would spend 200 minutes walking to the forest reserve to collect fuelwood, collected 240 kg in the past month and had a fuelwood forest product use value of 1.36. They spent 45 minutes walking to the forest reserve to collect wild foods, collected 15 kg in the past month and had a wild food forest product use value of 5.03 (near the maximum). Since the wealthiest of the households had a private woodlot from which to harvest their forest products, they did not expend as much time and effort collecting forest products as the households with higher forest dependence index values. The moderate and wealthiest households had more livelihood strategies than the least wealthy, which decreased their level of forest dependence comparatively. As we can see, the differential effort involved in collection of fuelwood and wild foods, the relative use of the forest products for fulfilling needs, and measures of non-forest wealth stratified the households into the different levels of forest dependence.   2.9. Reasons for employing a forest dependence index   Incorporating additional livelihood variables to measure forest dependence is important because the relative forest income method may have underestimated dependence when compared to the new index in the Malawi case study. The new index provides a comprehensive view of how dependence is involved in the livelihoods of small-holder farmers. Accounting for the diversity of forest products that contribute to household needs, the effort involved in collecting forest products, relative household wealth and the number of non-forest livelihood strategies sheds light on the extent of the multiple livelihood factors contributing to forest dependence. Including the time and physical effort of harvesting forest products, in addition to accounting for the amounts of forest products collected (which the relative forest income method also does) contributed to the increase in the forest dependence values. Without accounting for time and labor, the impact of forest  40 product collection on household livelihoods is less apparent. Of course, for some households whose time and effort were comparatively lower, their forest dependence values were reduced in the index. Including the degree to which households are dependent on forest products in the forest product use sub-index also contributes to the increase in dependence for households whose needs are primarily met from forest collection, and the reduction for others who have more alternative sources of forest products. The relative forest income method does not account for other sources of forest products. The relative wealth sub-index using the asset-based wealth index also impacts the forest dependence values, such as in the study by Nielsen et al. (2012) where amounts of forest products collected showed a different trend with respect to asset as opposed to income-based measures of household wealth. Incorporating the number of alternative livelihood strategies is also important for quantifying the extent to which households have alternatives to depending on forests. The relative forest income measure does this to a certain extent because the constituent values of total household income can be determined to compare non-forest sources to forest sources. Vedeld et al. (2007) compared the proportion of forest income to total household income using a diversification index based on Simpson’s index, which measured the inverse sum of the proportion of total income comprised by each livelihood activity. Their results showed an increase in the diversification index with increasing relative forest income values until 0.30, at which time the diversification index decreased with increasing relative forest income. This finding accords with the incorporation of the livelihood strategy sub-index since at higher levels of forest dependence, livelihood diversification is decreased. The forest dependence index has the advantage of directly incorporating the number of alternative livelihood strategies into determining dependence on forests.  For researchers and practitioners interested to apply a forest dependence metric, I encourage the use of the forest dependence index I have described or a variant thereof that maintains the overall index principles. The index must capture the variability of forest dependence in the sample population, reflect the diversity of forest products used to meet household needs, reflect the relative burden of the collection of forest products for households, and reflect how relative wealth and livelihood strategies impact levels of forest dependence. The index may reflect the relative importance of different forest products and account for the relative importance of the sub-indices if applicable, although the index values for the case study in Malawi did not do so.  41 The use of the additive, product-level weighting approach can be employed as in the Malawi case study. Variants could also be applied, in terms of aggregation techniques and selection of variables within the sub-indices. A geometric or multi-criteria aggregation technique could be used instead of the additive approach, to include information on the relative importance of variables (Nardo et al., 2005). Additional variables could be incorporated in the effort sub-index – for instance the number of household members who collect forest products – to increase the level of information on factors that influence levels of forest dependence. It is also important to note that households will have a range of sizes and the number of household members will impact levels of forest dependence, but I chose not to explicitly incorporate this into the index because the number of household members should be reflected in the amounts of forest products collected. An alternative to the product-level approach would be a sub-index level approach by calculating each sub-index separately first then aggregating the sub-indices, and assigning weights to the sub-indices. Such an approach would be useful where it is clear that there is differential importance of the sub-indices for determining household forest dependence.   In the application, interpretation and comparison of the forest dependence index values, it is important to recognize the relative nature of the values. The index values are similar to the DHS wealth index scores in that they can be compared within the study area but should not necessarily be compared across studies, because the scores have been constructed based on the sample population’s data and context (Howe et al., 2008). The relativity of the forest dependence index values stems from the incorporation of the DHS wealth index, and from the standardization of the data across the sample population. Care should be exercised if comparing across study sites. The index values could be compared across studies if the data has been collected in the same way and the aggregation methods are the same. Comparisons could be made between cases with similar social-ecological contexts, for example, comparing the values from the southern Malawi case study to other communities in Malawi or elsewhere in the miombo woodland region of southern Africa. An alternative approach to compare study sites with similar social-ecological contexts would be to pool the data from multiple sites such that the households would be compared directly in the standardization and normalization steps of the index construction. The resulting index values of households from different sites could be distinguished and compared using the histogram of the distribution of index values.  42  When comparing the forest dependence index to the relative forest income method, there are a few important caveats to discuss. The forest dependence index values for a given household are relative to other households in the study area, given the standardization of values throughout the calculation and normalization of values in the final step. The relative forest income values, on the other hand, are absolute in the sense that they are not calculated relative to other households – rather, they reflect the relative amount of forest income to total income for an individual household. When interpreting the values of the two metrics for two households with the same value, 0.2 for example, one household would have 20% of their total income comprised by forest income, while the other household would have an index value of 0.2 that stems from the aggregation of multiple aspects of dependence. The two metrics are in the same interval scale thus can be compared mathematically, and they both measure forest dependence, but in the income method dependence is measured strictly economically whereas in the index dependence is measured as a multifaceted livelihood strategy. Therefore, there is a conceptual difference between the two metrics, and their values cannot be directly compared but should be considered complementary.  Application of the index can assist in assessing levels of forest dependence and the relationship to levels of poverty through the additional livelihood information incorporated. By understanding the importance of forest products to household needs, the effort involved in collection, household relative wealth and the number of non-forest livelihood strategies, policies and management interventions can be designed and implemented to address local livelihood conditions towards poverty alleviation. Forest-based poverty alleviation strategies can be beneficial in contexts where high rates of poverty overlap with forest dependence (Sunderlin et al., 2004; Sunderlin et al., 2005). In such strategies, the role of forests in lifting households out of poverty versus preventing households from falling into deeper levels of poverty could be determined to understand what can realistically be accomplished in the local context (Sunderlin et al., 2005). Three components of the role of forests in poverty alleviation have been identified: forest products used to fulfill household consumption needs, forest products used in times of emergency as resources or income sources (the safety net concept), and forests helping to lift people out of poverty through the marketing or processing of forest products hence raising household income (Belcher, 2005).  The first two can be considered poverty mitigation and the  43 last poverty reduction (Belcher, 2005). With respect to enhancing poverty mitigation potential, a sustainable supply of forest products ought to be maintained (both for daily consumption and emergency consumption) (Belcher, 2005). As reflected in the forest dependence index, forest-based poverty alleviation approaches could include enhancing the availability and reducing the inconvenience of collecting a diversity of products – both timber and non-timber forest products – for households (Arnold & Perez, 2001; Belcher, 2005; Ticktin, 2004).  Regarding poverty reduction, opportunities in the production, processing and marketing of forest products can enhance sources of income (Belcher, 2005). Increasing commercialization of forest products may, however, increase harvesting practices that are detrimental to plant species and communities, thus strategies for joint conservation and development must be carefully implemented (Arnold & Perez, 2001).                   44 Chapter 3. Changes in tree species diversity, composition and aboveground carbon in areas of fuelwood harvesting in miombo woodland ecosystems of southern Malawi  3.1. Introduction   Fuelwood is the primary energy source for roughly 2.5 billion people across the planet and is especially important for impoverished households that lack energy alternatives (IEA, 2017). Fuelwood collection comprises 55% of global forest harvesting and is higher in certain areas such as sub-Saharan Africa where it comprises approximately 90% of roundwood production (Bailis et al., 2015; Food and Agriculture Organization of the United Nations, 2016). The harvesting of fuelwood, which involves collection of deadwood, live trees or tree branches, can contribute to forest degradation, a widespread issue in the tropics (Hosonuma et al., 2012; Sassen et al., 2015). The extent of degradation caused by fuelwood harvesting is highly localized and depends on harvesting rates (Heltberg, 2001). Degradation is likely to be high in areas with high population density, in fragile ecosystems, and where forest resources are poorly managed (Heltberg, 2001; Heltberg, 2005). Harvesting of deadwood and tree branches may cause minimal adverse effects on the forest in terms of tree composition and structure (Shackleton, 1993). Yet allowing forest harvesting may lead to more detrimental activities, such as harvesting of live trees and tree burning for charcoal (Sassen et al., 2015; Shackleton, 1993). Indeed, charcoal production is widely viewed as an activity that causes forest degradation (Mwampamba et al., 2013).  While fuelwood harvesting can cause forest cover loss and decreased woody biomass, knowledge of the finer-scale changes such as tree species diversity and composition is limited. Using a spatially-explicit model, Ghilardi et al. (2016) showed that biomass loss in forests in Honduras declined when households reduced fuelwood harvesting by switching to improved (fuel-saving) cookstoves. Several remote sensing studies have also shown that fuelwood extraction is an important driver of forest cover and woody biomass loss (DeFries et al., 2010; Ryan et al., 2012; Yiran et al., 2012). But few field studies have investigated more detailed changes in forest structure and composition due to fuelwood harvesting. One study in a national park in Uganda showed decreased basal area of tree species preferred for fuelwood near the park boundary (Sassen et al., 2015). In a modeling study of a montane cloud forest in Mexico, projections showed that an  45 increase in fuelwood harvesting intensity was correlated with increasing changes in forest structure and composition over hundreds of years (Rüger et al., 2008). Further understanding of the ecological changes, in terms of tree species composition and diversity, in areas with fuelwood harvesting could contribute to strategies for biodiversity conservation or ecological restoration in such forests; in the present work I use forest plots to investigate this in southern Malawi.  The impact of the harvesting of fuelwood and other forest products on forest structure and composition likely increases with harvesting access or pressure. Because measuring harvesting access and pressure directly is challenging, many researchers use proxies. For example, Serneels and Lambin (2001) used distance from forest sites to roads and settlements as proxies for human impacts on forests. One study in Mount Elgon National Park Uganda assessed forest plots and showed that in some of the areas where fuelwood harvesting occurs, stem density and species richness increased with distance from the park boundary hence distance from settlements (Sassen & Sheil, 2013). Another study using forest plots along transects showed increasing woody biomass, species richness and diversity with increasing distance from an urban centre in Tanzania (Ahrends et al., 2010). Elevation of forest sites is also a factor influencing forest access, where impacts of harvesting are likely higher in lower elevation sites due to proximity to settlements and ease of access (Sassen & Sheil, 2013). By assessing proxies for harvesting access and pressure, we can understand which factors impact changes in forest biodiversity and composition and apply this knowledge to sustainable forest management. I contribute to this literature by assessing the relationship between tree species diversity and aboveground carbon and the proxies for forest product harvesting access and pressure.  In order to understand the complex ecological effects of fuelwood harvesting, it is important to investigate the extent of local forest degradation. I addressed this issue by conducting a case study in southern Malawi, where the miombo (Brachystegia spp.) woodlands are a prime example of a social-ecological system undergoing rapid change. Between 1990 and 2010 Malawi lost 20% of its forest cover (Food and Agriculture Organization, 2010). Deforestation rates in Malawi are high at 0.6 to 1.0% forest cover loss per year from 1990 to 2015, with similar or higher rates of forest degradation (Food and Agriculture Organization, 2015; Kamanga et al., 2009). The main factors driving forest change in Malawi include land clearing for agriculture and over- 46 harvesting of fuelwood, including charcoal production (Fisher, 2004; Zulu, 2010). Households depend on the remaining forests for provisioning ecosystem services such as fuelwood, yet forests that have been heavily degraded may no longer be able to provide the required forest products.  Given high forest use as well as high deforestation and degradation rates, research into ecosystem change of the miombo woodlands in Malawi is critical.  In this study, I compared species composition, diversity and aboveground carbon in miombo woodlands of southern Malawi, in areas of high local use for fuelwood versus reference sites in relatively undisturbed forests. I investigated whether relative elevation and distance to roads (proxies for access to the forest), and proximity to settlements (proxy for harvesting pressure), predicted the differences in forest diversity and aboveground carbon in areas of fuelwood harvesting. I addressed two main questions:   1. How do tree species composition, diversity and aboveground carbon (AGC) differ in sites with fuelwood harvesting compared to reference sites? 2. To what extent are harvesting access (measured by elevation and distance to roads) and harvesting pressure (measured by proximity to settlements) predictors of tree diversity and AGC within fuelwood harvesting sites?  I hypothesize that tree species richness, abundance and diversity are lower, species composition is different, and AGC is lower in the areas of fuelwood harvesting compared to reference sites. I expect that elevation, distance to roads and proximity to settlements are predictors of species richness, abundance, diversity and AGC.   3.2. Methods   3.2.1. Study area  I conducted the study in southern Malawi [15°10'12.61"S, 35°18'0.28"E] in the Zomba-Malosa forest reserve in Zomba District, and in reference sites in the Malosa and Liwonde forest reserves and in Liwonde National Park in Machinga District (Figure 3.1). In Malawi the harvesting of certain plant parts is permitted in forest reserves, but harvesting was not currently occurring in  47 reference sites. The topography of the study area is characterized by a plain where the agricultural communities reside (approximately 750 m above mean sea level) punctuated by the Zomba plateau (2087 m above mean sea level) and additional nearby plateaus (1500 to 2000 m above mean sea level) where state-owned forest reserves are located (Ngongondo et al., 2011). The region is characterized by a savanna climate (Ngongondo et al., 2011). Annual average rainfall ranges from 700 mm in the lowlands to 2,500 mm in the highlands or plateaus, with 80% falling in the rainy season between November and April (Ngongondo et al., 2011). Mean annual minimum and maximum temperatures are 18C and 30C respectively (Malawi Goverment, 2006).                      Figure 3. 1. Study area in southern Malawi showing fuelwood harvesting sites (red) and reference sites (blue), as well as outlines of forest reserves and national parks. Study area indicated by the red box on the inset map of Malawi. Map created in ArcMap 10.5 (ESRI, 2016).  48 The historic land cover across most of Malawi is the miombo woodland ecosystem, dominated by species of the Brachystegia, Julbernardia and Isoberlina genera (Frost, 1996). The dry miombo woodland occurs in southern Malawi and is typified by a maximum tree canopy height of 15 m and understorey vegetation predominantly composed of grasses, followed by sedges and leguminous shrubs (Frost, 1996). Soils are acidic, low in nutrients and well drained (Frost, 1996). Cambisols and Luvisols predominate in the study area which is in relatively dry, low to moderate elevation areas of the forest reserves and National Park (Frost, 1996).  Poverty and human population pressures have led to land use/cover changes across the miombo woodlands in Malawi. Malawi is one of the poorest countries in the world, with the sixth lowest per capita purchasing power parity in 2017 (International Monetary Fund, 2017; United Nations, 2017a). Southern Malawi has the highest population density in the country with 184 people/km2 on average (Meijer et al., 2016; NSO, 2008). The Zomba district has a population of 579,639, with an area of 2,580 km2 and a density of 225 people/km2; the Machinga district has a population of 490,579 with an area of 3,771 km2 thus a density of 130 people/km2 (NSO, 2008). Most of the population are small-holder farmers (Fisher, 2004). Agricultural clearing of the woodlands in southern Malawi is widespread and the majority of the land cover is now annual or perennial cropland (Haack et al., 2015). Fuelwood harvesting, either firewood collection or charcoal production, has also caused forest change, thus grasslands or shrublands may also occur in previously forested areas (Fisher, 2004; Haack et al., 2015). Less than 10% of the population have access to electricity and all households use firewood or charcoal either as a primary or secondary energy source (Dasappa, 2011; Zulu, 2010). Rural households burn on average 5 kg of firewood per day (Nerfa et al., 2016, unpublished data). Charcoal production has a 23% unit conversion from wood to charcoal (Kammen & Lew, 2005), and due to its high demand with 82% of urban households using charcoal for energy (Smith et al., 2017),  rapid tree cover loss ensues. Although charcoal production is currently illegal in forest reserves in Malawi, it remains widespread (Smith et al., 2017).    Agricultural communities in Malawi are highly dependent on firewood and other forest products from nearby forests, particularly from state-owned forest reserves. Forest reserves originated in the 1911 Forest Ordination under British Colonial rule as an effort to protect  49 indigenous woodlands (Zulu, 2010). In 1926, the Communal Forest Scheme was established under the District Administration Ordinance, followed by the establishment of Village Forest Areas (VFAs) in the Forest Ordinance of 1931 where village heads were allocated areas in the reserves for local use and management (Kayambazinthu, 2000). The National Environmental Policy advocated for community participation in managing forest reserves and The Forestry Policy of 1996 included the formation of Village Natural Resources Management Committees (VNRMCs) to encourage co-management of forests by communities. There are currently 88 forest reserves across the country, the majority of which are in hilly miombo woodlands (Malawi Goverment, 2010). Today harvesting of certain products, including dry wood (ex. firewood), thatch grass, fodder, medicinal plants, fruits and wild foods, is permitted in the forests reserves (Jumbe & Angelsen, 2007; Kamanga et al., 2009).   3.2.2. Forest plots  I compared species diversity, composition and AGC where community members regularly harvest fuelwood in the Zomba-Malosa forest reserve, and in reference sites of miombo woodlands with minimal use (Figure 1). I chose fuelwood harvesting sites (n=14 with 50 plots in total) by accompanying local villagers collecting fuelwood in the Zomba-Malosa forest reserve. Household fuelwood harvesting in the forest reserve is high; my surveys show that households living in the adjacent villages collect on average 75 kg per month (Nerfa et al., 2016, unpublished data).  I chose reference sites (n=9 with 36 plots in total) by consulting the District Forest Officer in Machinga and local botanists to find the least disturbed, most mature miombo woodlands nearby, in Liwonde National Park, and Liwonde and Malosa forest reserves. Note that the fuelwood harvesting sites and reference sites were both located in forest reserves but that forest product harvesting was not, to my knowledge, currently occurring in reference sites due to management practices geared towards forest conservation. Harvesting of fuelwood and other forest products may have occurred previously in reference sites; the local District Forest Officer and botanists could not guarantee that the reference sites were in primary forests (DFO, personal communication, 2016). As the reference sites were selected to typify intact, mature miombo woodlands, the historic forest structure in the fuelwood harvesting sites (prior to degradation) is assumed to be similar to the forest structure in the reference sites. Historical imagery of the Zomba- 50 Malosa forest reserve from 1971 shows that the forest reserve was previously closed-canopy mature miombo woodland.   At each site, I sampled four 200 m2 circular plots (7.98 m radius). The first two fuelwood harvesting sites at the start of the study followed an alternative plot layout with fewer plots (two 400 m2 plots) but the same total area (800 m2). The only other exception was one of the fuelwood harvesting sites where I was only able to sample two 200 m2 plots. For fuelwood harvesting sites, the first plot was sampled where the community member I was following began harvesting. For reference sites, the first plot was sampled 100 m into the forest from the boundary of the reserve. I travelled along the foot paths, and alternated plot locations on either side of the path. To determine the plot centre, I generated a random compass bearing off the trail then walked 7.98 metres in the given direction.  I ensured the plots were completely off the path, repeating the random number generation if necessary until a suitable bearing was determined. Subsequent plots were spaced 100 m apart in the cardinal direction nearest to the direction of the foot path by following the bearing on GPS. I recorded walking time in minutes from the household’s homestead to the fuelwood harvesting location and all plot locations in fuelwood harvesting sites and reference sites were recorded using Trimble Juno 3B GPS units.  I employed a nested sub-plot design for vegetation identification and measurements (Appendix B Figure 1). I identified and measured the DBH of trees (>4 cm DBH) in the entire 200m2 plot. In one central 18m2 sub-plot (2.39 m radius), I identified and counted saplings (1-4 cm DBH) and seedlings (0-1 cm DBH). Lastly, I estimated percent cover of herbs, shrubs and grasses in one central 9 m2 nested sub-plot (0.8 m radius); I did not analyze the herb, shrub or grass data. Local botanists assisted with species identification in the field. If the plant could not be identified in the field, photographs were taken and shown to additional local botanists for identification.   3.2.3. Statistical analysis  I compared tree, seedling, and sapling species diversity and composition between fuelwood harvesting sites and reference sites. To understand species diversity, I calculated species richness  51 (number of species), abundance (number of individuals), Shannon index, and Simpson’s diversity (1-D where D=Simpson’s index) at the site level using counts of individuals, in the BiodiversityR package in R 3.4.0 with RStudio (Kindt & Coe, 2005; R Development Core Team, 2011). I calculated both the Shannon index and Simpson’s diversity since the former index is sensitive to both species richness and evenness while the latter is less sensitive to richness and is more effective at portraying evenness (Magurran, 2004). Tables 1 and 2 in Appendix B list the values for the ecological variables for reference sites and fuelwood harvesting sites. I used Welch two-sample t-tests to compare the mean richness, abundance and diversity values between fuelwood harvesting sites and reference sites. To compare species composition, I conducted non-metric multi-dimensional scaling ordinations (NMDS) at the site level, using Bray ecological distance, with the package vegan in BiodiversityR (Kindt & Coe, 2005; Oksanen et al., 2016). In ordination space the distance between sites reflects the differences in species composition between sites. I used multi-response permutation procedure (MRPP) to test if differences in composition between fuelwood harvesting sites and reference sites were significant.  I used five mixed-species allometric equations developed for the miombo woodlands and converted to carbon to compare AGC estimates between the fuelwood harvesting sites and reference sites. Because there is not one conventionally used allometric equation for the miombo woodlands, I used five equations:  AGB = 0.21691 x DBH2.318391
 (Kachamba et al., 2016) AGB = -2.5265 + 2.5553 x log(DBH) (Chidumayo, 2014) AGB = 0.1428 x DBH2.271               (Kuyah et al., 2014) AGB = 0.1027 x DBH2.4798  (Mugasha et al., 2013) AGB = -3.629 + 2.601 x log(DBH) (Ryan et al., 2011)  I applied the allometric equations to each tree in a plot. Tree aboveground biomass values per plot were summed per site and converted to Mg per hectare. I then multiplied by 0.47 to convert to Mg carbon per hectare (IPCC, 2006). I calculated the values using each equation for each fuelwood harvesting site and reference site. Additionally, I calculated the mean of all five  52 allometric equations for each site, and compared between fuelwood harvesting sites and reference sites using Welch two-sample t-tests. Finally, I compared summary statistics on DBH values for all trees in fuelwood harvesting sites and reference sites.  To analyze whether harvesting access and pressure were predictors of tree diversity and AGC within fuelwood harvesting sites, I measured elevation and distance to the main road for forest plots in the Zomba-Malosa forest reserve as the proxies for harvesting access, and counted the number of rooftops within a 3 km buffer as the proxy for harvesting pressure. Elevation data was gleaned from the GPS points for plots and averaged per site. I measured the Euclidean distance from each forest plot to the main road in ArcMap 10.5 then averaged the plot distances per site (ESRI, 2016). Using Google Earth Pro, I tagged domestic rooftops within a 3 km radius circle (estimated maximum walking distance households would typically travel) surrounding the locations of forest plots (Google, 2013). I tagged a total of 1551 rooftops in the proximity of the Zomba-Malosa forest reserve. I imported KML files of tagged rooftops into ArcMap 10.5, to count the number of rooftops within a 3 km circle of each of the plots, and took the average number per site (ESRI, 2016). Table 1 in Appendix B shows the values of the proxies of access and pressure for fuelwood harvesting sites. Note that I also determined the elevation, number of rooftops in a 3 km circle and distance to the main road for reference sites for observation (Appendix B Table 2), but I did not analyze the factors because harvesting pressure and access were not relevant in reference sites.  I then used the data in quasi-poisson (dispersion  1) generalized linear models (GLMs). Elevation, distance and number of rooftops were the explanatory variables and species abundance, richness, Shannon index and Simpson’s diversity were the response variables. In addition to the GLMs, using the function envfit in R 3.4.0 I fit the factors of mean distance, mean elevation and mean number of rooftops as vectors on the NMDS ordination of tree species composition and determined R2 values and significance of fit (R Core Team, 2017).      53 3.3. Results  3.3.1. Tree, sapling and seedling species diversity and composition   Diversity of indigenous tree species was higher in reference sites than fuelwood harvesting sites. In reference sites (n=9) I identified 53 indigenous tree species; in fuelwood harvesting sites (n=14) I identified 51 indigenous species and 5 exotic species (Appendix B Table 3). In fuelwood harvesting locations, the exotic species Eucalyptus camaldulensis Dehnh. and Eucalyptus tereticornis Sm. were present in 14% and 16% of sites and comprised 5% and 7% of total stems, respectively (Figures 3.2 and 3.3). Nine species occurred in greater than 10 plots for fuelwood harvesting sites and reference sites analyzed separately (20% of fuelwood harvesting site plots, 28% of reference site plots) (Figures 3.2 and 3.3). In terms of percentage of plots, Annona senegalensis and Pterocarpus angolensis DC. were the most common species in fuelwood harvesting sites while Brachystegia bussei Harms and Diplorhynchus condylocarpon (Müll.Arg.) Pichon were the most common species in reference sites. In terms of percentage of stems, Uapaca kirkiana Müll.Arg. and Eucalyptus tereticornis were the most common species in fuelwood harvesting sites and Brachystegia bussei and Diplorhynchus condylocarpon were the most common species in reference sites.   54  Figure 3. 2. Dominant tree species in fuelwood harvesting sites (n=14; red) and reference sites (n=9; blue), by percentage of plots in which the species were found. Exotic species indicated with an (E). Note scale from 0 to 60 percent.    55  Figure 3. 3. Dominant tree species in fuelwood harvesting sites (n=14; red) and reference sites (n=9; blue) by percentage of stems found in all plots. Exotic species indicated with an (E). Note scale from 0 to 30 percent.  The mean values of all four measures of tree biodiversity were significantly lower in fuelwood harvesting sites than in reference sites (Table 3.1), but the range of values across sites was often higher in fuelwood harvesting sites (Figure 3.4). Using rarefaction, I verified that the difference in tree species richness held despite the sampling effort discrepancy between reference sites (n=9) and fuelwood harvesting sites (n=14). When compared at 9 sites (maximum for reference sites), reference sites had significantly higher species tree richness with a total of 53 species compared to a total of 46.5 species for fuelwood harvesting sites (Figure 3.5).       56  Table 3. 1. Mean species richness, species abundance, Shannon index and Simpson’s diversity values for trees, saplings and seedlings in fuelwood harvesting sites (n=14) and reference sites (n=9). Values computed from the counts of individuals within each site. Standard errors shown. Significant p-values (=0.05) from Welch two sample t-test shown with an asterisk.                          57                                  Fuelwood Harvesting SitesReferenceSites2060Species Abundance *Fuelwood Harvesting SitesReferenceSites51015Species Richness *Fuelwood Harvesting SitesReferenceSites  0.5 1.5 2.5 *Fuelwood Harvesting SitesReferenceSites0.40.60.8Simpson's Diversity *Fuelwood Harvesting SitesReferenceSites0102030Species Abundance *Fuelwood Harvesting SitesReferenceSites1357Species RichnessFuelwood Harvesting SitesReferenceSitesShannon Index 0.0 1.0Fuelwood Harvesting SitesReferenceSites0.00.40.8Simpson’s DiversityFuelwood Harvesting SitesReferenceSites0200400Species AbundanceFuelwood Harvesting SitesReferenceSites51015Species Richness *Fuelwood Harvesting SitesReferenceSites0.51.5Shannon Index*Fuelwood Harvesting SitesReferenceSites0.20.6Simpson’s DiversityTreesSaplingsSeedlingsShannon IndexFigure 3. 4. Box plots showing site-level species abundance, species richness, Shannon index and Simpson’s diversity values for fuelwood harvesting sites (n=14) and reference sites (n=9), for counts of tree, sapling and seedling individuals. Significant differences between the means (Welch two sample t-test, =0.05) indicated with an asterisk.   58           Figure 3. 5. Rarefaction curves computed with 100 permutations, showing species richness for fuelwood harvesting sites (n=14; red) and reference sites (n=9; blue) by number of sites.  For saplings and seedlings there were fewer differences in species diversity and composition in fuelwood harvesting sites versus reference sites than for trees. For saplings, I identified 38 indigenous species and 3 exotic species in fuelwood harvesting sites, and in reference sites 17 indigenous species (Appendix B Table 3). For seedlings, I identified 47 indigenous species in fuelwood harvesting sites and 6 exotic species, and in reference sites 48 indigenous species (Appendix B Table 3). For saplings, the only measure to show a significant difference was species abundance, which was higher (p=0.0018) in fuelwood harvesting sites than reference sites (Table 3.1, Figure 3.4). For seedlings, the Shannon index and Simpson’s diversity were lower (p=0.0115; p=0.05) in fuelwood harvesting sites than reference sites but there was no significant difference for species abundance or Simpson’s diversity (Table 3.1, Figure 3.4).  NMDS ordination showed that tree species composition of fuelwood harvesting sites was both significantly different and more varied than in reference sites (Figure 3.6). Multi-response permutation procedure for Bray distance had a chance corrected within-group agreement A value of 0.0494 (p=0.001, stress=0.069), indicating significant separation between the two site types. Seedling and sapling species composition showed some overlap between fuelwood harvesting sites and reference sites. MRPP values using Bray distance suggested significant difference between groups for saplings (A=0.042, p=0.001, stress=9.67e-05) but not seedlings (A=0.0094, p=0.102, stress=0.085).  2 4 6 8 10 12 1401020304050Number of SitesSpecies RichnessFuelwood_Harvesting_elwood_Harvesting_SiteReference_Reference SitesFuelwood Harvesting Sites 59                                  Figure 3. 6. Non-metric multi-dimensional scaling (NMDS) ordinations showing species composition for counts of individuals per site, for trees, saplings and seedlings in fuelwood harvesting sites (n=14) and reference sites (n=9). NMDS1 and 2 are axes 1 and 2 respectively. Ordination constructed in R 3.4.0 using the package vegan, function metaMDS using Bray ecological distance and 4 dimensions for trees and seedlings, 5 for saplings to minimize stress (R Core Team, 2017). Stress values for trees, saplings and seedlings were 0.069, 9.67e-05 and 0.085 respectively. MRPP p-values shown.  60 3.3.2. Aboveground carbon  Aboveground carbon was lower in fuelwood harvesting sites than reference sites. DBH values in fuelwood harvesting sites were lower (mean=5.64 cm) than in reference sites (mean=12.91 cm; p-value < 2.2e-16). The largest tree in reference sites had a DBH value of 64.8 cm, compared to 31.1 cm in fuelwood harvesting sites (Figure 3.7). The average AGC value for all allometric equations was significantly lower for fuelwood harvesting sites (mean=2.50 Mg C/ha) than reference sites (mean=36.47 Mg C/ha; p=0.00016) (Figure 3.8). Appendix B Table 4 shows the summary statistics of each of the five allometric equations.                  Figure 3. 7. DBH values in cm for all trees (>4 cm DBH) in forest plots, for fuelwood harvesting sites (n=14) and reference sites (n=9).       Fuelwood Harvesting SitesDBH (cm)Frequency (number of stems)0 10 20 30 40 50 60050150Reference SitesDBH (cm)Frequency (number of stems)0 10 20 30 40 50 60050150 61              Figure 3. 8. Boxplots showing the mean of the five allometric equations converted to aboveground carbon (Mg C/ha) per site for fuelwood harvesting sites (n=14) and reference sites (n=9) (Kuyah et al., 2014; Mugasha et al., 2013; Kachamba et al., 2016; Chidumayo, 2014; Ryan et al., 2011).   3.3.3. Harvesting access and pressure  In fuelwood harvesting sites, elevation and distance to the main road had a significant positive effect on abundance and biomass, while number of rooftops had a significant negative effect on abundance and biomass (Table 3.2). When mean site elevation, distance to the main road and number of rooftops in a 3 km radius were plotted as vectors on the NMDS tree species ordination (Appendix B Figure 2), only elevation showed a significant R2 value (Table 3.3).                 Fuelwood Harvesting Sites Reference Sites0102030405060Average of Allometric Equations (Mg C/ha)p=0.0002 62 Table 3. 2. Quasi-poisson generalized linear model (GLM) standardized coefficients and p-values testing elevation, distance to the main road and number of rooftops within 3 km as explanatory variables and species richness, species abundance, Shannon index, Simpson’s diversity and biomass as response variables for fuelwood harvesting sites (n=14). Significant p-values (=0.05) shown with an asterisk.                   Table 3. 3. R2 values with p-values for site averages of distance to the main road, elevation and number of rooftops in 3 km proximity, plotted as vectors on the NMDS ordination for fuelwood harvesting sites (n=14). Significant p-values (=0.05) shown with an asterisk.             63 3.4. Discussion   The hypotheses that tree species diversity and AGC would be lower and composition would be significantly different between fuelwood harvesting sites in the Zomba-Malosa forest reserve and reference sites in nearby miombo woodlands were supported by the findings. The areas of the forest reserve with fuelwood harvesting can largely be considered to be in an earlier successional stage due to past disturbances that led to the reduction of species diversity and AGC, compared to the reference sites which were in the mature woodland stage (Frost, 1996). The fuelwood harvesting sites were in either the initial regrowth or tall sapling stage of succession (Frost, 1996), or due to tree planting, in a novel ecosystem state. In fuelwood harvesting sites, the dominant species tended to be early successional species that return following the disturbance regime of fire (occurring frequently in the dry season), such as Annona senegalensis Pers. and Pterocarpus angolensis  (Donfack et al., 1995; Frost, 1996; Strømgaard, 1986), or exotic species such as Eucalyptus camaldulensis and Eucalyptus tereticornis. In reference sites the most abundant species were Diplorhynchus condylocarpon which is important for forest recovery after fire, and Brachystegia bussei which requires protection from fire and is a mid-late successional species (Frost, 1996; Strømgaard, 1986). While some tree planting efforts occurred in the Zomba-Malosa forest reserve, such as with Eucalyptus camaldulensis and Eucalyptus tereticornis, species richness remained lower than in the reference sites. Another study looking at cloud forests in Ecuador that were cleared for timber and charcoal production and for agriculture, then had community tree planting occur (including exotic species), found that the planted areas were less diverse and had a significantly different composition compared to primary forests (Wilson & Rhemtulla, 2016). In the Zomba-Malosa forest reserve, past over-harvesting removed the majority of large trees (>10 cm DBH). The AGC values in the fuelwood areas were hence much lower than in the reference sites (mean of 2.42 Mg C/ha compared to 36.47 Mg C/ha). While I did not explicitly investigate the types of over-harvesting that led to the degradation, there was anecdotal evidence from community members that harvesting for timber and charcoal production were major causes (Whittaker & Rhemtulla, 2018, manuscript submitted). The trees which remain today were likely not selected for harvesting due to their undesirable form, such as Uapaca kirkiana which was one of the dominant species in the fuelwood harvesting sites (in 20% of plots and comprising 29.79% of stems) and which is a small and shrubby fruiting tree. A study in a cloud forest in Mexico found  64 that in areas of fuelwood harvesting, composition shifted to species that were not selected for fuelwood use (Rüger et al., 2008).   The findings on seedling and sapling diversity provide insight into the potential for natural regeneration in the Zomba-Malosa forest reserve. My results that seedling and sapling composition overlapped with reference sites are similar to other studies that have shown communities of both seedlings and saplings in planted tropical forests were most similar in composition to primary forests and least similar to unplanted forests (Wilson & Rhemtulla, 2016). Although seedling diversity values were lower or not significantly different between fuelwood harvesting sites and reference sites, sapling diversity was promising for regeneration. Since sapling species diversity in fuelwood harvesting sites was not significantly different from reference sites and abundance was higher, regeneration into a forest state similar to the “intact” miombo woodland ecosystem may be possible. The high abundance of saplings in the fuelwood harvesting areas may be due to the increased sunlight because of the reduction in canopy cover. With regards to planted trees, I found six exotic seedling species in fuelwood harvesting sites but only three exotic sapling species, which may suggest competitive exclusion. The presence of exotic species is not expected to hinder the regeneration of the forest given the reduced number of exotic species growing from the seedling into the sapling stage. The three exotic sapling species – Eucalyptus camaldulensis, Eucalyptus tereticornis, and Toona ciliata M.Roem. – only occurred in 6%, 4% and 6% of fuelwood harvesting plots respectively. Regarding species composition, seedling composition in fuelwood harvesting sites was not significantly different compared to reference sites, but sapling diversity was significantly different. Sapling composition in fuelwood harvesting sites did overlap with reference sites, however, thus there remains the possibility of growth into a similar composition to indigenous woodlands. Limitations to natural regeneration in the Zomba-Malosa forest reserve include the slow growth of tropical seedlings and juvenile saplings, and the potential suppression of saplings by the abundance of tall grasses which are characteristic of the miombo woodlands (Chidumayo, 2004; Higgins et al., 2000). Additionally, tree growth may be inhibited due to climate change, as sub-Saharan Africa is a vulnerable region to climate change impacts, including increased temperatures and drought incidence (IPCC, 2014; Li et al., 2009).   65 The hypothesis that harvesting access and pressure would be predictors of tree species diversity and AGC was partially supported since the proxies for harvesting access and pressure had effects on species abundance and biomass in the Zomba-Malosa forest reserve. Plots located further from the main road and at higher elevations are more difficult to access, which may explain the higher tree species abundance and woody biomass in those sites. As population size and density near forest sites increases the demand for forest products increases, which can also explain the change in tree species abundance and biomass. Yet elevation, distance to the main road and number of rooftops did not have significant effects on species richness, Shannon index or Simpson’s diversity. The people who over-harvested in the past may not have cut certain species preferentially across the landscape. Species richness and diversity tended to be high at the sites farther away from the villages and road at higher elevations, but some of the closer and lower sites also had relatively high richness and diversity, partly due to the presence of one or more exotic (planted) species. Of the proxies for harvesting access and pressure, only elevation had a significant R2 value when plotted on the NMDS ordination suggesting a trend in species composition with increasing difficulty of access due to elevational change. Because I did not directly assess harvesting pressure or intensity, I cannot identify all of the mechanisms that caused the ecological changes in the Zomba-Malosa forest reserve. Additional drivers of degradation have occurred in the forest reserve such as charcoal production and over-harvesting of timber (Whittaker & Rhemtulla, 2016, unpublished data). One important caveat with respect to the proxy of rooftop abundance is imprecision of the 3 km radius around forest plots. While studies have used the 3 km radius for an estimate of fuelwood collection areas (ex. Barve et al., 2005), the range where household members travel realistically deviates from a circular shape. Researchers have developed a valuable new approach to measure fuelwood collection areas by having community members carry GPS loggers on their harvesting trips; the shapes of the harvesting areas for most households were revealed to be polygonal (Singh et al., 2018). In the future, similar assessments to our study could be improved by assessing forest ecological changes in the specific fuelwood harvesting polygons.  An additional important factor affecting forest composition, which was not quantified, is the history of forest management. The forest reserves in the Zomba and Machinga districts were managed under the Improved Forest Management for Sustainable Livelihoods scheme, funded by the Malawi Government and the European Union, which was operational in two phases, from  66 2006-2009, and 2011-2013 (Senganimalunje et al., 2016). The program aimed to both alleviate poverty and conserve forests by supporting community-based forest management (Senganimalunje et al., 2016). Management strategies differed depending on the approaches the communities took, which impacted the extent of forest conservation hence the present ecological structure of the forests. In the Liwonde and Malosa forest reserves of our reference sites for instance, tree planting efforts were undertaken and more strict enforcement was employed to prevent over-harvesting for charcoal production (DFO, personal communication, 2016). Conservation of the woodlands also occurred in Liwonde National Park. Yet establishing park boundaries alone may be insufficient to conserve forest cover in some cases, for example, in Lake Malawi National Park. Park establishment was intended to reduce the loss of tree cover which was occurring. After establishment, however, the harvesting of wood continued for domestic fuel and construction materials and for commercial fish smoking, given a lack of alternatives; commercial harvesting caused the greatest degradation  (Abbot & Homewood, 1999). The forest product needs of local households should be considered to develop locally appropriate forest management strategies which will conserve forests while sustaining household livelihoods, including determining the ideal locations for strict protected areas versus sustainable use areas.  Ecosystem functioning may decline given the tree species diversity and AGC reduction in the Zomba-Malosa forest reserve. Losses of plant species diversity can result in losses of functional diversity and critical ecosystem functions such as primary productivity, nutrient cycling, soil structure and regulation of hydrological cycles (Díaz et al., 2004; Tilman et al., 2014). Tree cover loss and the reduction of mean DBH values, due to over-harvesting of large trees in the Zomba-Malosa forest reserve, reduce carbon stocks which disrupts local carbon cycles and contributes to regional climate change (DeFries et al., 2004). Additional detrimental effects on ecosystem function include decreased evapotranspiration and cloud formation, altering local precipitation patterns (Lawton et al., 2001).  Ecosystem services, particularly provisioning services, will also likely be impaired in the forest reserve. Quantities of fuelwood have already been reduced; households in the study area have noticed a decline in fuelwood in the past five years (Nerfa et al., 2016, unpublished data). Households prefer to use indigenous species for firewood as opposed to exotic species such as  67 Eucalyptus camaldulensis (J. Rhemtulla, personal communication, 2017), because indigenous species tend to have higher fuelwood quality making them preferable as a fuel source (Abbot et al., 1997).  Yet the carbon stocks of native tree species in the forest reserve has declined, thus household harvesting of indigenous species may be unsustainable. As woody biomass becomes scarce, households spend more time collecting fuelwood, a trend already apparent in our study area (Nerfa et al., 2016, unpublished data). While some studies have also shown an increased collection time in cases of fuelwood scarcity, the relationship may be more nuanced (Brouwer et al., 1997). In one study in Malawi, due to fuelwood scarcity households increased their collection time as their distance to woodlands increased to a point, beyond which households used closer sources and switched from larger wood to twigs (Brouwer et al., 1997). In the Zomba-Malosa forest reserve currently households collect wood products largely from the ground (Nerfa & Rhemtulla, 2016, unpublished data) as these are the biomass sources available and harvesting of live trees in forest reserves is currently restricted. Investigating abundance of deadwood in the Zomba-Malosa forest reserve could contribute to understanding the ecological impacts of fuelwood collection. Volume of deadwood has been shown to increase with distance from park boundaries and settlements, in forests where community members harvest fuelwood (Sassen et al., 2015).   In the context of forest degradation, high household forest dependence and climate change, ecological restoration could be employed to remedy the social and ecological challenges in the Zomba-Malosa forest reserve. Following the Bonn challenge to restore 350 million hectares of forest by 2030, in 2016 the government of Malawi pledged to restore 4.5 million hectares (48% of the total land mass) by 2030 (Bonn Challenge, 2016). To make tree planting effective in southern Malawi, the needs of local communities ought to be considered because household use of local forests is high, especially for fuelwood collection. Restoration of the forest reserves could be undertaken with consideration for multiple forest functions and uses across the landscape (Chazdon et al., 2016). The forest landscape restoration approach would be suitable, as a process which aims to restore ecological integrity and human well-being at the landscape level (Lamb et al., 2012). Multiple restoration strategies could be used in the area, including the combination of active and passive restoration. Based on our analysis of harvesting pressure and access, fast-growing species could be planted in the forest reserve closest to the villages to provide woody  68 biomass for fuelwood, reserving the farther locations for passive restoration of the indigenous woodlands or active restoration as needed. Amongst the villages and agricultural lands, tree cover could be increased using agroforestry techniques, to enhance ecosystem functioning and services (Jose, 2009). For successful restoration, desired ecological and social outcomes could be investigated in the mixed forest-agricultural landscapes of southern Malawi. Similar research could be conducted across the country of Malawi and elsewhere in the miombo woodland ecosystem region of Southern Africa, to benefit forest conservation and contribute to poverty alleviation.                                69 Chapter 4. Conclusion  In tropical forest-agricultural landscapes, small-holder farmers rely on forests for multiple ecosystem services (Sunderlin et al., 2005). Yet in many places, these forests are being degraded rapidly (Ghazoul et al., 2015), with numerous social and ecological consequences. The ecosystem services on which communities depend, especially provisioning services for essential forest products such as fuelwood, are hindered by the degradation of forests (Brosius, 1997). Livelihood sustainability for tropical small-holder farmers, who are impoverished and lack alternatives to forest dependence (Angelsen & Wunder, 2003), is hence at risk. Quantifying levels of forest dependence in rural communities, therefore, can contribute to poverty alleviation strategies. Ecologically, forest degradation involves the loss of biodiversity and carbon stocks, reducing ecosystem functioning and resilience (Ghazoul et al., 2015). Understanding ecological changes in tree species diversity and composition resulting from the harvesting of forest products (particularly fuelwood) is also critical to inform forest restoration and conservation for ecological and social benefits. In this thesis, I conducted social and ecological assessments in a case study in southern Malawi to understand the state of the social-ecological system of the miombo woodlands where forest dependence and forest change are high.   4.1. Measuring multiple livelihood aspects of forest dependence  Household dependence on forests in rural landscapes in the tropics is generally understood to be high, and quantifying levels of dependence is important for implementing policies that contribute to livelihood sustainability (Wunder et al., 2014). Currently, household forest dependence is measured using relative forest income, a measurement that determines the proportion of household income constituted by the monetary value of collected forest products (Cavendish, 2000; Vedeld et al., 2007). This method, however, has limitations. It does not capture income fluctuations or account for the burden of time and effort that households expend in collecting forest products, and it is unrealistic in situations where households primarily consume rather than sell forest products. Using complementary measures of forest dependence may therefore be beneficial, to capture additional livelihood information related to dependence. In Chapter 2, I created a new index to measure forest dependence. The index portrays multiple  70 livelihood aspects of forest dependence, including the use of multiple forest products to meet household needs, the effort involved in forest collection, and relative household wealth and number of non-forest livelihood strategies to measure access to alternatives to forest product collection. I calculated the index for the study area in southern Malawi and compared the values from the new index with those from the relative forest income approach.   I showed that the relative forest income approach may have underestimated household forest dependence. Incorporating multiple livelihood aspects of forest dependence into the index tended to increase the levels of dependence, indicating that forest collection has important impacts on the livelihood strategies of local households. The index illustrates that forest dependence is not only an economic issue but is often a labour-intensive livelihood strategy that contributes significantly to meeting household needs. In social-ecological systems research, the index can help to illustrate the interactions between forest users and forest resource systems through the collection of forest products, and it can be combined with ecological data to understand the extent to which dependence is sustainable. Researchers and practitioners can use the index to measure forest dependence in the global south to contribute to forest-based poverty alleviation strategies.  4.2. Ecological changes in forests with fuelwood harvesting   Fuelwood is a key forest product on which rural farming communities rely (Bailis et al., 2015). The harvesting of fuelwood may, however, cause deforestation and degradation (Heltberg, 2001), producing a negative feedback effect on the ability for communities to continue harvesting fuelwood and other forest products for subsistence use. Therefore, assessing ecosystem changes in forests with fuelwood harvesting is important for understanding what is needed for forest conservation. Studies on the ecological impacts of fuelwood harvesting are largely conducted using remote sensing analyses of forest cover and biomass loss; field studies that assess biodiversity and forest composition at a finer scale are scarce. In Chapter 3, I investigated the changes in miombo woodlands in terms of tree species diversity, composition and aboveground woody biomass in the study area in southern Malawi where agricultural communities collect fuelwood and made comparisons to reference sites in relatively undisturbed woodlands nearby. I assessed whether there was a correlation between the ecological variables and proxies for  71 harvesting access (elevation and distance to the main road) and pressure (number of rural houses in a 3 km buffer) in the forest locations with fuelwood harvesting.  Fuelwood harvesting areas had lower tree species diversity and biomass and a different species composition than the relatively undisturbed woodlands. The proxies for harvesting access and pressure were correlated with species abundance and biomass in the fuelwood harvesting areas. Thus, ecosystem functioning and ecosystem services may be reduced in the forests on which the agricultural communities rely for provisioning ecosystem services. Given the degradation of the forest, harvesting of fuelwood and other forest products may be unsustainable. The resilience of the forest ecosystem has hence been reduced, and interventions to transform the social-ecological system will likely be necessary since households will need to continue harvesting forest products in the area.  4.3. Managing tropical forests as complex social-ecological systems  Given the forest degradation in the Zomba-Malosa forest reserve and the dependence of local households on the forest for provisioning ecosystem services, the sustainability of the social-ecological system is at risk. With reductions in tree species diversity, aboveground biomass, and a change in species composition, the diversity, flows and aggregation of the system have changed from prior to the forest degradation. Household forest dependence in the area remains high, yet the reduction in biomass will limit the available flow of resources. Households in the communities have begun to experience a reduction in available fuelwood in the past five years and most households expend more time collecting fuelwood than in the past. Households will most likely need to change their collection patterns to find sustainable harvesting practices. Interventions such as forest-landscape restoration and community-based forest management could be undertaken to increase system resilience, to enhance the ability of the system to respond to disturbances due to climate change such as drought and anthropogenic pressures such as increasing population density.   Forest landscape restoration, a process that seeks to restore ecological processes at the landscape scale (Lamb et al., 2012), could be used to restore the degraded forest and provide the forest product needs for the local peoples while conserving plant species diversity and carbon  72 stocks in the miombo woodlands. A combination of restoration activities could take place across the forest-agricultural landscape, such as assisted natural regeneration of the native woodlands by protecting seedlings and saplings, complemented by tree planting of fast-growing species for fuelwood provisioning near the villages and trees incorporated on farms and homesteads (Dewees, 1995). To maintain diversity and flows, plant biological diversity and carbon stocks could be enhanced through tree planting of various native species and potentially fast-growing exotic species such as Eucalyptus and Senna species, which have been previously planted in the area, near the villages or in agroforests. In order for successful restoration to take place, suitable species would need to be identified, land allocated, and resources secured such as seedlings, tools, and labour. Members of Tikambirane (the local community-based organization) have already begun planting Albizia lebbeck (Linnaeus) Benth. and Senna siamea (Lam.) Irwin and Barneby, two exotic species selected because they are fast-growing, in available land owned by willing farmers and in an area of the Zomba-Malosa forest reserve near the villages (M. Saddick, personal communication, 2017).  Community-based forest management could also be strengthened in the area, including enhanced benefits for the community members involved. While forest co-management schemes have been attempted in the study area, government support was often short-term and community members were left without adequate resources to maintain seedlings that had been planted and without adequate training to enforce forest management practices (Whittaker & Rhemtulla, 2018, manuscript submitted). In a study investigating factors that have led to failure of forest restoration projects in southern Malawi, community members expressed that they felt the government had retracted their responsibility to enforce forest management through the creation of co-management schemes, creating tension in the communities due to unsuccessful attempts at self-enforcement (Whittaker & Rhemtulla, 2018, manuscript submitted). Community members felt that they should receive financial incentives to participate in training and management activities, given their levels of poverty and lack of resources; ultimately, community-based forest management in the area should have community benefits and assist in alleviating poverty so that household pressure on forests for resources and agricultural land can shift away from heavy forest use (Whittaker & Rhemtulla, 2018, manuscript submitted), towards forest conservation. Through interventions in the social-ecological system in terms of restoration and community-based management, the system  73 of the forest-agricultural landscape in southern Malawi would transform. The goal would be to shift the system to an alternate state that aims to enhance ecosystem functioning and services in a resilient landscape of forest, agriculture and agroforestry land use and land cover types, where the community members can maintain sustainable livelihoods.  4.2. Future research  Given the levels of forest dependence and forest degradation in the southern Malawian landscape, the integration of further social and ecological research could assist in enhancing the sustainability of the social-ecological system. Few studies have combined social and ecological research when investigating forest dependence or fuelwood harvesting (ex., Jagger & Perez-Heydrich, 2016; Mahamane et al., 2017), but given the linkages between components in social-ecological systems, the integration of multi-disciplinary knowledge is crucial for understanding these forestry issues in complex systems. The research on changes in tree species composition in the forest reserve could be accompanied by research into community member species preferences for fuelwood and other forest products in the reserve including non-tree plant species, to understand how the forest degradation will impact household preferences and collection patterns. Investigations into the social pressures that have led to local forest degradation could also be beneficial for understanding the historical and present ecological conditions in the area and for outlining management interventions to combat future degradation. Additional analyses could use the social-ecological systems framework, such as following Ostrom’s (2009) framework for assessing sustainability. By identifying the key interactions between resource systems and units and governance systems and users, as well as the outcomes of the interactions, scenarios to enhance the livelihood sustainability of the forest dependent communities could be envisioned. The suggested studies for the site in southern Malawi could be scaled up to the miombo woodland region of southern Africa, and could be applied in similar social-ecological contexts in tropical dry forests.   Following the ecological research presented in this thesis, additional research could further investigate past, present, and future ecological dynamics in the forest reserve. Historical studies could identify how the changes in tree composition and biomass occurred over the past few  74 decades. Further studies on present conditions could include quantifying deadwood and the contributions of deadwood to ecosystem functioning (Sassen et al., 2015). Non-tree plant species diversity and productivity could also be quantified, such as the abundant grass species, to understand ecosystem function and regeneration. With respect to future projections, the natural regeneration potential of the degraded miombo woodlands could be investigated, to understand whether passive restoration would be adequate or if active ecosystem restoration would be necessary (Chazdon, 2008). Assisted natural regeneration or active restoration techniques could be explored in terms of plant species selection and growth rates, and methods for optimizing the spatial allocation of land use and land cover types across the landscape to maximize benefits for conservation and livelihoods (Boedhihartono & Sayer, 2012).  To follow the quantification of forest dependence in southern Malawi, additional social research could investigate perspectives of community members on forest restoration and forest management. Attitudes of individuals towards tree planting in the forest reserve and agroforestry implementation could be explored. Group discussions could be used to determine what successful restoration and forest management would look like in the area, and how success could be achieved. Community member perceptions of forest dependence could also be explored, to understand whether the local community members are interested to maintain their livelihood dependency on the forest or whether they would prefer to seek alternative livelihood strategies. The quantitative data of the research in this thesis could be complemented with qualitative data that gives voice to the views of community members, such as through household interviews and focus group interviews.  In this thesis, I introduced a new composite index to measure household forest dependence and I conducted an analysis on the changes in tree species composition in miombo woodlands with fuelwood harvesting, using a case study in southern Malawi. I found that levels of forest dependence in the study area were high, and that the currently used metric (relative forest income) may have underestimated dependence. Ecologically, I found reduced tree species diversity and aboveground carbon as well as a different species composition in areas of fuelwood harvesting versus reference sites of intact miombo woodlands, and a significant relationship between harvesting access and pressure and species abundance and aboveground carbon. The findings of  75 high forest dependence and forest degradation indicate that sustainability of the social-ecological system is at risk. Strategies to increase the sustainability and resilience of the system could be explored, such as forest landscape restoration and community-based forest management. Further ecological research, for instance on the regeneration capacity of the degraded forest, and social research, for example on community member perspectives towards forest use and management, could be conducted in the study area of southern Malawi or in similar social-ecological systems towards forest conservation and community well-being.                           76 Bibliography  Abbot, J. I. O., & Homewood, K. (1999). A history of change: Causes of miombo woodland decline in a protected area in Malawi. Journal of Applied Ecology, 36(3), 422–433. Abbot, P., Lowore, J., Khofi, C., & Werren, M. (1997). Defining firewood quality: A comparison of quantitative and rapid appraisal techniques to evaluate firewood species from a Southern African Savanna. Biomass and Bioenergy, 12(6), 429–437. Achard, F., Beuchle, R., Mayaux, P., Stibig, H. J., Bodart, C., Brink, A., Carboni, S., Desclée, B., Donnay, F., Eva, H. D., Lupi, A., Raši, R., Seliger, R., & Simonetti, D. (2014). Determination of tropical deforestation rates and related carbon losses from 1990 to 2010. Global Change Biology, 20(8), 2540–2554. Adams, C., Rodrigues, S. T., Calmon, M., & Kumar, C. (2016). Impacts of large-scale forest restoration on socioeconomic status and local livelihoods: what we know and do not know. Biotropica, 48(6), 731–744. Ahrends, A., Burgess, N. D., Milledge, S. A. H., Bulling, M. T., Fisher, B., Smart, J. C. R., Clarke, G. P., Mhoro, B. E., & Lewis, S. L. (2010). Predictable waves of sequential forest degradation and biodiversity loss spreading from an African city. Proceedings of the National Academy of Sciences, 107(33), 14556–14561. Angelsen, A., Jagger, P., Babigumira, R., Belcher, B., Hogarth, N. J., Bauch, S., Börner, J., Smith-Hall, C., & Wunder, S. (2014). Environmental Income and Rural Livelihoods: A Global-Comparative Analysis. World Development, 64(S1), S12–S28. Angelsen, A., & Wunder, S. (2003). Exploring the forest--poverty link: key concepts, issues and research implications. Economist, (40), 68. Anthwal, A., Gupta, N., Sharma, A., Anthwal, S., & Kim, K. H. (2010). Conserving biodiversity through traditional beliefs in sacred groves in Uttarakhand Himalaya, India. Resources, Conservation and Recycling, 54(11), 962–971. Arnold, J. E. M., & Perez, M. R. (2001). Can non-timber forest products match tropical forest conservation and development objectives? Ecological Economics, 39, 437–447. Bailis, R., Drigo, R., Ghilardi, A., & Masera, O. (2015). The carbon footprint of traditional woodfuels. Nature Climate Change, 5(3), 266–272. Barrett, C. B., Reardon, T., & Webb, P. (2001). Nonfarm income diversification and household livelihood strategies in rural Africa: Concepts, dynamics, and policy implications. Food Policy, 26(4), 315–331. Barve, N., Kiran, M. C., Vanaraj, G., Aravind, N. A., Rao, D., Uma Shaanker, R., Ganeshaiah, K. N., & Poulsen, J. G. (2005). Measuring and mapping threats to a wildlife sanctuary in southern India. Conservation Biology, 19(1), 122–130. Belcher, B. (2005). Forest product markets, forests and poverty reduction. International Forestry Review, 7(2), 82–89. Belcher, B. M. (2003). What isn’t an NTFP? International Forestry Review, 5(2), 161–168. Berenguer, E., Ferreira, J., Gardner, T. A., Aragão, L. E. O. C., De Camargo, P. B., Cerri, C. E., Durigan, M., De Oliveira, R. C., Vieira, I. C. G., & Barlow, J. (2014). A large-scale field assessment of carbon stocks in human-modified tropical forests. Global Change Biology, 20(12), 3713–3726. Blackden, C. M., Wodon, Q., & Shetty, S. (2006). Gender, time use, and poverty in Sub-Saharan Africa: Foreword. World Bank Working Paper, (73). Boedhihartono, A. K., & Sayer, J. (2012). Forest Landscape Restoration: Restoring What and  77 For Whom? In J. Stanturf, D. Lamb, & P. Madsen (Eds.), Forest Landscape Restoration: Integrating Natural and Social Sciences (pp. 309–323). Bonn Challenge. (2016). Malawi Commitments. Retrieved from http://www.bonnchallenge.org/malawi Booysen, F. (2002). An overview and evaluation of composite indices of development. Social Indicators Research, 59(2), 115–151. Brosius, J. P. (1997). Endangered forest, endangered people: Environmentalist representations of indigenous knowledge. Human Ecology, 25(1), 47–69. Brouwer, I. D., Hoorweg, J. C., & van Liere, M. J. (1997). When households run out of fuel: Responses of rural households to decreasing fuelwood availability, Ntcheu District, Malawi. World Development, 25(2), 255–266. Byron, N., & Arnold, M. (1999). What futures for the people of the tropical forests? World Development, 27(5), 789–805. Campbell, B., & Luckert, M. (2002). Uncovering the Hidden Harvest: valuation methods for woodland & forest resources. Valuation Methods for Woodlands and Forest Resources. London; Sterling, VA; Earthscan Publications. Campbell, M. O. N. (2005). Sacred groves for forest conservation in Ghana’s coastal savannas: Assessing ecological and social dimensions. Singapore Journal of Tropical Geography, 26(2), 151–169. Cavendish, W. (2000). Empirical regularities in the poverty-environment relationship of rural households: Evidence from Zimbabwe. World Development, 28(11), 1979–2003. Centre for International Forestry Research. (2008). PEN Prototype Questionnaire, Version 4.4 September 2008. Retrieved from http://www.cifor.org/fileadmin/fileupload/PEN/pubs/pdf_files/PEN_Prototype_Questionnaire_-_version_4-4_-_September_2008.pdf Chao, S. (2012). Forest peoples: Numbers across the world. Forest Peoples Programme, 27. Chazdon, R. L. (2008). Beyond deforestation: Restoring forests and ecosystem services on degraded lands. Science, 320(5882), 1458–1460. Chazdon, R. L., Brancalion, P. H. S., Laestadius, L., Bennett-, A., Buckingham, K., Kumar, C., Moll-rocek, J., & Célia, I. (2016). When is a forest a forest? Forest concepts and definitions in the era of forest and landscape restoration. AMBIO A Journal of the Human Environment, (March), 1–36. Chidumayo, E. N. (2004). Development of Brachystegia-Julbernardia woodland after clear-felling in central Zambia: Evidence for high resilience. Applied Vegetation Science, 7(2), 237–242. Chidumayo, E. N. (2013). Forest degradation and recovery in a miombo woodland landscape in Zambia: 22 years of observations on permanent sample plots. Forest Ecology and Management, 291, 154–161. Chidumayo, E. N. (2014). Estimating tree biomass and changes in root biomass following clear-cutting of Brachystegia-Julbernardia (miombo) woodland in central Zambia. Environmental Conservation, 41(1), 54–63. Cunningham, A. B. (1993). African medical plants: Setting priorities at the interface between conservation and primary healthcare. People and Plants Working Paper, 1(March), 50. Daily, G. C. (1997). Nature’s Services: Societal dependence on natural ecosystems. Island Press. Washington, DC. Dasappa, S. (2011). Potential of biomass energy for electricity generation in sub-Saharan Africa.  78 Energy for Sustainable Development, 15(3), 203–213. Davidar, P., Arjunan, M., Mammen, P. C., Garrigues, J. P., Puyravaud, J. P., & Roessingh, K. (2007). Forest degradation in the Western Ghats biodiversity hotspot: Resource collection, livelihood concerns and sustainability. Current Science, 93(11), 1573–1578. Davidar, P., Sahoo, S., Mammen, P. C., Acharya, P., Puyravaud, J. P., Arjunan, M., Garrigues, J. P., & Roessingh, K. (2010). Assessing the extent and causes of forest degradation in India: Where do we stand? Biological Conservation, 143(12), 2937–2944. DeFries, R., Karanth, K. K., & Pareeth, S. (2010). Interactions between protected areas and their surroundings in human-dominated tropical landscapes. Biological Conservation, 143(12), 2870–2880. DeFries, R. S., Foley, J. A., & Asner, G. P. (2004). Land-use choice : balancing human needs and ecosystem function. Frontiers in Ecology and the Environment, 2(5), 249–257. Dercon, S. (1998). Wealth, risk and activity choice: Cattle in Western Tanzania. Journal of Development Economics, 55(1), 1–42. Dewees, P. A. (1995). Trees on farms in Malawi: Private investment, public policy, and farmer choice. World Development, 23(7), 1085–1102. Dewees, P. A., Campbell, B. M., Katerere, Y., Sitoe, A., Cunningham, A. B., Angelsen, A., & Wunder, S. (2010). Managing the miombo woodlands of Southern Africa: Policies, incentives and options for the rural poor. Journal of Natural Resources Policy Research, 2(1), 57–73. Díaz, S., Hodgson, J. G., Thompson, K., Cabido, M., Cornelissen, J. H. C., Jalili, A., Montserrat-Martí, G., Grime, J. P., Zarrinkamar, F., Asri, Y., Band, S. R., Basconcelo, S., Castro-Díez, P., Funes, G., Hamzehee, B., Khoshnevi, M., Pérez-Harguindeguy, N., … Zak, M. R. (2004). The plant traits that drive ecosystems: Evidence from three continents. Journal of Vegetation Science, 15(3), 295–304. Dijkshoorn, J. A., Leenaars, J. G. B., Huting, J. R. M., & Kempen, B. (2016). Soil and Terrain database of the Republic of Malawi (No. 2016/01). ISRIC-World Soil Information. Donfack, P., Floret, C., & Pontanier, R. (1995). Secondary Succession in Abandoned Fields of Dry Tropical Northern Cameroon. Source Journal of Vegetation Science Journal of Vegetation Science, 6(6), 499–508. Ebert, U., & Welsch, H. (2004). Meaningful environmental indices: A social choice approach. Journal of Environmental Economics and Management, 47(2), 270–283. Ellis, F. (1999). Rural Livelihood Diversity in Developing Countries: Evidence and Policy Implications. Diversity, 40(40), 1–10. Erenstein, O. (2003). Smallholder conservation farming in the tropics and sub-tropics: A guide to the development and dissemination of mulching with crop residues and cover crops. Agriculture, Ecosystems and Environment, 100(1–3), 17–37. ESRI. (2016). ArcGIS Desktop: Release 10.5. Redlands CA. Filmer, D., & Pritchett, L. (2001). Estimating Wealth Effects Without Expenditure Data--Or Tears: An Application to Educational Enrollments in States of India. Demography, 38(1), 115–132. Fisher, M. (2004). Household welfare and forest dependence in Southern Malawi. Environment and Development Economics, 9(2), 135–154. Food and Agriculture Organization. (2006). Global Forest Resources Assessment 2005: Progress towards sustainable forest management. FAO Forestry Paper 147. Retrieved from http://www.fao.org/documents/show_cdr.asp?url_file=/DOCREP/004/Y1997E/y1997e0d.ht 79 m%5Cnwww.fao.org/forestry/site/fra2005/en%5Cnhttp://www.fao.org/forestry/index.jsp Food and Agriculture Organization. (2010). Global Forest Resources Assessment 2010: Country Reports, Malawi. Retrieved from http://www.fao.org/docrep/013/al557E/al557e.pdf Food and Agriculture Organization. (2015). Global Forest Resources Assessment 2015. FAO Forestry. Retrieved from http://www.fao.org/forest-resources-assessment/current-assessment/en/ Food and Agriculture Organization of the United Nations. Mechanical Wood Products Branch. (2016). FAOSTAT. Retrieved from http://www.fao.org/faostat/en/#data/FO Freudenberg, M. (2003). Composite indicators of country performance: a critical assessment. OECD Science, Technology and Industry Working Papers, 16, 35. Frost, P. (1996). The Ecology of Miombo Woodlands. The Miombo in Transition: Woodlands and Welfare in Africa. CIFOR, Bogor, Indonesia. Geist, H. J. (1999). Global assessment of deforestation related to tobacco farming. Tobacco Control, 8(1), 18–28. Geist, H. J., & Lambin, E. F. (2002). Proximate Causes and Underlying Driving Forces of Tropical Deforestation. BioScience, 52(2), 143. Ghazoul, J., Burivalova, Z., Garcia-Ulloa, J., & King, L. A. (2015). Conceptualizing Forest Degradation. Trends in Ecology and Evolution, 30(10), 622–632. Ghilardi, A., Bailis, R., Mas, J. F., Skutsch, M., Elvir, J. A., Quevedo, A., Masera, O., Dwivedi, P., Drigo, R., & Vega, E. (2016). Spatiotemporal modeling of fuelwood environmental impacts: Towards improved accounting for non-renewable biomass. Environmental Modelling and Software, 82. Google. (2013). Google Earth Pro, release 7.1.2.2041. Santa Clara, CA. Haack, B., Mahabir, R., & Kerkering, J. (2015). Remote sensing-derived national land cover land use maps: a comparison for Malawi. Geocarto International, 30(3), 270–292. Heltberg, R. (2001). Determinants and impact of local institutions for common resource management. Environment and Development Economics, 6(02). Heltberg, R. (2005). Factors determining household fuel choice in Guatemala. Environment and Development Economics, 10(3), 337–361. Higgins, S. I., Bond, W. J., & Trollope, W. S. W. (2000). Fire, resprouting and variability: A recipe for grass-tree coexistence in savanna. Journal of Ecology, 88(2), 213–229. Holland, J. H. (1995). Hidden Order: How Adapation Builds Complexity. Basic Books (Vol. 75). Hosonuma, N., Herold, M., De Sy, V., De Fries, R. S., Brockhaus, M., Verchot, L., Angelsen, A., & Romijn, E. (2012). An assessment of deforestation and forest degradation drivers in developing countries. Environmental Research Letters, 7(4), 044009. Howe, L. D., Hargreaves, J. R., & Huttly, S. R. A. (2008). Issues in the construction of wealth indices for the measurement of socio-economic position in low-income countries. Emerging Themes in Epidemiology, 5. Huang, L., Shao, Q., & Liu, J. (2012). Forest restoration to achieve both ecological and economic progress, Poyang Lake basin, China. Ecological Engineering, 44, 53–60. IBM Corp. Released. (2016). IBM SPSS Statistics for Windows, Version 24.0. 2016. IEA. (2017). Energy Access Outlook 2017. Retrieved from https://www.iea.org/publications/freepublications/publication/WEO2017SpecialReport_EnergyAccessOutlook.pdf International Monetary Fund. (2017). International Monetary Fund World Economic Outlook Database, April 2017. Retrieved from  80 https://www.imf.org/external/pubs/ft/weo/2017/01/weodata/index.aspx IPCC. (2006). 2006 IPCC Guidelines for National Greenhouse Gas Inventories. 2006 IPCC Guidelines for National Greenhouse Gas Inventories, 3(Chapter 2: Mineral Industry Emissions), 1–40. IPCC. (2014). Climate Change, Adaptation, and Vulnerability. Organization & Environment, 24(March), 1–44. Retrieved from http://oae.sagepub.com/cgi/content/abstract/24/3/269 Jagger, P., & Perez-Heydrich, C. (2016). Land use and household energy dynamics in Malawi. Environmental Research Letters, 11(12). Jose, S. (2009). Agroforestry for ecosystem services and environmental benefits: An overview. Agroforestry Systems, 76(1), 1–10. Jumbe, C. B. L., & Angelsen, A. (2007). Forest dependence and participation in CPR management: Empirical evidence from forest co-management in Malawi. Ecological Economics, 62(3–4), 661–672. Kachamba, D. J., Eid, T., & Gobakken, T. (2016). Above- and belowground biomass models for trees in the miombo woodlands of Malawi. Forests, 7(2). Kakhobwe, C. M., Kamoto, J. F., Njoloma, J. P., & Ozor, N. (2016). Scaling Up Agroforestry Farming Systems: Lessons from the Malawi Agroforestry Extension Project. Journal of Agricultural Extension, 20(1), 153–162. Kamanga, P., Vedeld, P., & Sjaastad, E. (2009). Forest incomes and rural livelihoods in Chiradzulu District, Malawi. Ecological Economics, 68(3), 613–624. Kammen, D. M., & Lew, D. J. (2005). Review of Technologies for the Production and Use of Charcoal. Renewable and Appropriate Energy Laboratory Report, 1–19. Kayambazinthu, D. (2000). Empowering communities to manage natural resources: where does power lie? The case of Malawi. In: Shackleton, S.E., Campbell, B. (Editors), Empowering Com- Munities to Manage Natural Resources: Case Studies from Southern Africa., (SADC Wildlife Sector Natural Resources Management Program, Lilongwe, Malawi, CSIR, Pretoria and WWF, Southern Africa, Zimbabwe), 37–58. Kindt, R., & Coe, R. (2005). Tree diversity analysis; A manual and software for common statistical methods for ecological and biodiversity studies. World, (January), 196. Retrieved from http://www.worldagroforestry.org/treesandmarkets/tree_diversity_analysis.asp Kishindo, P. (2004). Customary land tenure and the new land policy in Malawi. Journal of Contemporary African Studies, 22(2), 213–225. Kuyah, S., Sileshi, G. W., Njoloma, J., Mng’omba, S., & Neufeldt, H. (2014). Estimating aboveground tree biomass in three different miombo woodlands and associated land use systems in Malawi. Biomass and Bioenergy, 66, 214–222. Lamb, D., Stanturf, J., & Madsen, P. (2012). What Is Forest Landscape Restoration? In Forest Landscape Restoration: Integrating Natural and Social Sciences (pp. 3–23). Lawton, R. O., Nair, U. S., Pielke R.A., S., & Welch, R. M. (2001). Climatic impact of tropical lowland deforestation on nearby montane cloud forests. Science, 294(5542), 584–587. Lebbie, A. R., & Guries, R. P. (1995). Ethnobotanical value and conservation of sacred groves of the Kpaa Mende in Sierra Leone. Economic Botany, 49(3), 297–308. Lélé, S. (2000). Degradation, sustainability or transfomation. In Environment: Myth or Reality, (Seminar 486), 31–37. Levin, S. A. (1998). Ecosystems and the Biosphere as Complex Adaptive Systems. Ecosystems, 1(5), 431–436. Li, Y., Ye, W., Wang, M., & Yan, X. (2009). Climate change and drought: a risk assessment of  81 crop-yield impacts. Climate Research, 39, 31–46. Liswanti, N., Sheil, D., Basuki, I., Padmanaba, M., & Mulcahy, G. (2011). Falling back on forests: how forest-dwelling people cope with catastrophe in a changing landscape. International Forestry Review, 13(4), 442–455. Maass, J. M., Balvanera, P., Castillo, A., Daily, G. C., Mooney, H. a, Ehrlich, P., Quesada, M., Miranda, A., Jaramillo, V. J., García-oliva, F., Martínez-yrizar, A., Cotler, H., López-blanco, J., Pérez-jiménez, A., Búrquez, A., Tinoco, C., Ceballos, G., … Ayala, R. (2005). Ecosystem Services of Tropical Dry Forests: Insights from Long- term Ecological and Social Research on the Pacific Coast of Mexico. Ecology And Society, 10(1), 17. Magurran, A. E. (2004). Measuring Biological Diversity. Measuring Biological Diversity, 256. Mahamane, M., Zorrilla-Miras, P., Verweij, P., Sitoe, A., Ryan, C., Patenaude, G., Grundy, I., Nhantumbo, I., Metzger, M. J., Ribeiro, N., Baumert, S., & Vollmer, F. (2017). Understanding Land Use, Land Cover and Woodland - Based Ecosystem Services Change, Mabalane, Mozambique. Energy and Environment Research, 7(1). Malawi Goverment. (2006). Climate of Malawi: Temperature Maps. Retrieved from http://metmalawi.com/climate/temperature.php Malawi Goverment. (2010). Malawi State of Environment and Outlook Report. Environment for Sustainable Economic Growth. Lilongwe, Environmental Affairs Department. Masera, O., Ghilardi, A., Drigo, R., & Angel Trossero, M. (2006). WISDOM: A GIS-based supply demand mapping tool for woodfuel management. Biomass and Bioenergy, 30(7), 618–637. Meijer, S. S., Sileshi, G. W., Catacutan, D., & Nieuwenhuis, M. (2016). Agroforestry and deforestation in Malawi: inter-linkages between attitudes, beliefs and behaviours. Agroforestry Systems, 90(4), 645–658. Menzies, N. (2007). Our forest, your ecosystem, their timber: Communities, conservation, and the state in community-based forest management. New York: Columbia University Press. Mirza, B., & Szirmai, A. (2010). Towards a new measurement of energy poverty: A cross-community analysis of rural Pakistan. UNU-MERIT Working Papers. Misselhorn, A. A. (2005). What drives food insecurity in southern Africa? a meta-analysis of household economy studies. Global Environmental Change, 15(1), 33–43. Mitchard, E. T. A., & Flintrop, C. M. (2013). Woody encroachment and forest degradation in sub-Saharan Africa’s woodlands and savannas 1982-2006. Philosophical Transactions of the Royal Society B: Biological Sciences, 368(1625). Mugasha, W. A., Eid, T., Bollandsås, O. M., Malimbwi, R. E., Chamshama, S. A. O., Zahabu, E., & Katani, J. Z. (2013). Allometric models for prediction of above- and belowground biomass of trees in the miombo woodlands of Tanzania. Forest Ecology and Management, 310, 87–101. Mwampamba, T. H., Ghilardi, A., Sander, K., & Chaix, K. J. (2013). Dispelling common misconceptions to improve attitudes and policy outlook on charcoal in developing countries. Energy for Sustainable Development, 17(2), 75–85. Nardo, M., Saisana, M., Saltelli, A., Tarantola, S., Hoffman, A., & Giovannini, E. (2005). Handbook on constructing composite indicators. OECD Statistics Working Papers. Nath, T. K., Jashimuddin, M., & Inoue, M. (2016). Community-Based Forest Management (CBFM) in Bangladesh (Vol. 22, Vol. 22). Springer. Nerfa, L., Rhemtulla, J., Timko, J., & Dan, A. (2016). Forest Use & Livelihoods survey – Malawi 2016. Unpublished data.  82 Nerfa, L., & Rhemtulla, J. (2016). Forest Follows Data Collection. Unpublished data. Newton, P., Miller, D. C., Byenkya, M. A. A., & Agrawal, A. (2016). Who are forest-dependent people? A taxonomy to aid livelihood and land use decision-making in forested regions. Land Use Policy, 57, 388–395. Ngongondo, C. S., Xu, C. Y., Tallaksen, L. M., Alemaw, B., & Chirwa, T. (2011). Regional frequency analysis of rainfall extremes in Southern Malawi using the index rainfall and L-moments approaches. Stochastic Environmental Research and Risk Assessment, 25(7), 939–955. Nielsen, M. R., Pouliot, M., & Kim Bakkegaard, R. (2012). Combining income and assets measures to include the transitory nature of poverty in assessments of forest dependence: Evidence from the Democratic Republic of Congo. Ecological Economics, 78, 37–46. NSO. (2008). 2008 Population and Housing Census. Retrieved from http://www.nsomalawi.mw/index.php?option=com_content&view=article&id=107%3A2008-population-and-housing-census-results&catid=8&Itemid=6 Oksanen, J., Blanchet, F., Kindt, R., Legendre, P., & O’Hara, R. (2016). Vegan: community ecology package. R Package 2.3-3. Retrieved from https://cran.r-project.org/package=vegan Ostrom, E. (2009). A general framework for analyzing sustainability of social-ecological systems. Science. Pattanayak, S. K., & Sills, E. O. (2001). Do Tropical Forests Provide Natural Insurance? The Microeconomics of Non-Timber Forest Product Collection in the Brazilian Amazon. Land Economics, 77(4), 595–612. Paumgarten, F. (2005). The role of non-timber forest products as safety-nets: A review of evidence with a focus on South Africa. GeoJournal, 64(3), 189–197. Place, F., & Otsuka, K. (2001). Population, tenure, and natural resource management: The case of customary land area in Malawi. Journal of Environmental Economics and Management, 41(1), 13–32. Prado Córdova, J. P., Wunder, S., Smith-Hall, C., & Börner, J. (2013). Rural income and forest reliance in highland guatemala. Environmental Management, 51(5), 1034–1043. R Development Core Team, R. (2011). R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing (Vol. 1). Retrieved from http://www.r-project.org Reij, C., Tappan, G., & Smale, M. (2010). Agroenvironmental transformation in the Sahel: another kind of “green revolution.” In Proven successes in agricultural development a technical compendium (pp. 161–189). Roser, M., & Ortiz-Ospina, E. (2017). Global Extreme Poverty. OurWorldInData.Org, 1. Retrieved from https://ourworldindata.org/extreme-poverty/ Rüger, N., Williams-Linera, G., Kissling, W. D., & Huth, A. (2008). Long-term impacts of fuelwood extraction on a tropical montane cloud forest. Ecosystems, 11(6), 868–881. Rutstein, S. O. (2008). The DHS Wealth Index: Approaches for rural and urban areas. Demographic and Health Research, (60). Ryan, C. M., Hill, T., Woollen, E., Ghee, C., Mitchard, E., Cassells, G., Grace, J., Woodhouse, I. H., & Williams, M. (2012). Quantifying small-scale deforestation and forest degradation in African woodlands using radar imagery. Global Change Biology, 18(1), 243–257. Ryan, C. M., Williams, M., & Grace, J. (2011). Above- and belowground carbon stocks in a miombo woodland landscape of mozambique. Biotropica, 43(4), 423–432. Saisana, M., & Tarantola, S. (2002). State-of-the-art Report on Current Methodologies and  83 Practices for Composite Indicator Development. Joint Research Centre. Italy: European Commission, (July), 1–72. Sasaki, N., & Putz, F. E. (2009). Critical need for new definitions of “forest” and “forest degradation” in global climate change agreements. Conservation Letters, 2(5), 226–232. Sassen, M., & Sheil, D. (2013). Human impacts on forest structure and species richness on the edges of a protected mountain forest in Uganda. Forest Ecology and Management, 307, 206–218. Sassen, M., Sheil, D., & Giller, K. E. (2015). Fuelwood collection and its impacts on a protected tropical mountain forest in Uganda. Forest Ecology and Management, 354, 56–67. Scoones, I. (1998). Sustainable Rural Livelihoods a Framework for Analysis. Analysis, 72, 1–22. Senganimalunje, T. C., Chirwa, P. W., Babalola, F. D., & Graham, M. A. (2016). Does participatory forest management program lead to efficient forest resource use and improved rural livelihoods? Experiences from Mua-Livulezi Forest Reserve, Malawi. Agroforestry Systems, 90(4), 691–710. Serneels, S., & Lambin, E. F. (2001). Proximate causes of land-use change in Narok District, Kenya: a spatial statistical model. Agriculture, Ecosystems & Environment, 85(1–3), 65–81. Shackleton, C. M. (1993). Fuelwood harvesting and sustainable utilisation in a communal grazing land and protected area of the eastern transvaal lowveld. Biological Conservation, 63(3), 247–254. Shackleton, S., Delang, C. O., & Angelsen, A. (2011). Non-Timber Forest Products in the Global Context. Non-Timber Forest Products in the Global Context, 7, 55–81. Shyamsundar, P., Ahlroth, S., P, K., & Onder, S. (2018). Understanding Forests’ Contribution to Poverty Alleviation: A Framework for Interventions in Forested Areas. World Bank Working Paper. Simula, M. (2009). Towards defining forest degradation: comparative analysis of existing definitions. Forest Resources Assessment Programme working Paper. Retrieved from http://www.ardot.fi/Documents_2/Degradationdefintions.pdf Singh, D., Aung, T., & Zerriffi, H. (2018). Resource Collection Polygons: A spatial analysis of woodfuel collection patterns. Energy for Sustainable Development, 45, 150–158. Sjaastad, E., Angelsen, A., Vedeld, P., & Bojö, J. (2005). What is environmental income? Ecological Economics, 55(1), 37–46. Smith, H. E., Hudson, M. D., & Schreckenberg, K. (2017). Livelihood diversification: The role of charcoal production in southern Malawi. Energy for Sustainable Development, 36, 22–36. Specht, M. J., Pinto, S. R. R., Albuqueque, U. P., Tabarelli, M., & Melo, F. P. L. (2015). Burning biodiversity: Fuelwood harvesting causes forest degradation in human-dominated tropical landscapes. Global Ecology and Conservation, 3, 200–209. Stanturf, J., Lamb, D., & Madsen, P. (2012). Forest Landscape Restoration. Forest Landscape Restoration - Integrating Natural and Social Sciences, 15, 322. Strømgaard, P. (1986). Early secondary succession on abandoned shifting cultivator’s plots in the Miombo of south central Africa. Biotropica, 18(2), 97–106. Sunderland, T., Achdiawan, R., Angelsen, A., Babigumira, R., Ickowitz, A., Paumgarten, F., Reyes-García, V., & Shively, G. (2014). Challenging Perceptions about Men, Women, and Forest Product Use: A Global Comparative Study. World Development, 64(S1), S56–S66. Sunderlin, William D., Angelsen, Arild and Wunder, S. (2004). Forests and poverty alleviation. In State of the World’s Forests 2003 (p. 10).  84 Sunderlin, W. D., Belcher, B., Santoso, L., Angelsen, A., Burgers, P., Nasi, R., & Wunder, S. (2005). Livelihoods, forests, and conservation in developing countries: An overview. World Development, 33(9 SPEC. ISS.), 1383–1402. Sunderlin, W. D., Dewi, S., Puntodewo, A., Müller, D., Angelsen, A., & Epprecht, M. (2008). Why forests are important for global poverty alleviation: A spatial explanation. Ecology and Society, 13(2). Tainter, J. A. (2006). Social complexity and sustainability. Ecol. Complex., 3(2), 91–103. Takane, T. (2008). Customary land tenure, inheritance rules, and smallholder farmers in Malawi. Journal of Southern African Studies, 34(2), 269–291. Team, R. D. C. (2017). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Ticktin, T. (2004). The ecological implications of harvesting non-timber forest products. Journal of Applied Ecology, 41(1), 11–21. Tilman, D., Isbell, F., & Cowles, J. M. (2014). Biodiversity and Ecosystem Functioning. Annual Review of Ecology, Evolution, and Systematics, 45(1), 471–493. UNAIDS. (2017). Malawi Country Overview. Retrieved from http://www.unaids.org/en/regionscountries/countries/malawi/ United Nations. (2017a). List of Least Developed Countries. Retrieved from http://unctad.org/en/PublicationsLibrary/ldcr2017_en.pdf United Nations. (2017b). World Population Prospects 2017. Retrieved from https://esa.un.org/unpd/wpp/DataQuery/ USAID. (2016). Malawi Orphans and Vulnerable Children Fact Sheet. Retrieved from https://www.usaid.gov/malawi/fact-sheets/malawi-orphans-and-vulnerable-children-fact-sheet Vedeld, P., Angelsen, A., Bojö, J., Sjaastad, E., & Kobugabe Berg, G. (2007). Forest environmental incomes and the rural poor. Forest Policy and Economics, 9(7), 869–879. Vignola, R., Harvey, C. A., Bautista-Solis, P., Avelino, J., Rapidel, B., Donatti, C., & Martinez, R. (2015). Ecosystem-based adaptation for smallholder farmers: Definitions, opportunities and constraints. Agriculture, Ecosystems and Environment, 211, 126–132. Walker, B., Holling, C. S., Carpenter, S. R., & Kinzig, A. (2004). Resilience, Adaptability and Transformability in Social – ecological Systems. Ecology and Society, 9(2), 5. Whittaker, A., & Rhemtulla, J. (2016). Focus groups on factors leading to the failure of reforestation projects. Unpublished data. Whittaker, A., & Rhemtulla, J. (2018). Why We Fail: Stakeholder Perceptions of the Social and Ecological Barriers to Reforestation in Southern Malawi. Manuscript Submitted for Publication. Wilson, S. J., & Rhemtulla, J. M. (2016). Acceleration and novelty: Community restoration speeds recovery and transforms species composition in Andean cloud forest. Ecological Applications, 26(1), 203–218. World Bank. (2004). Sustaining Forests: a development strategy. Journal of Development Economics. Retrieved from http://siteresources.worldbank.org/INTFORESTS/Resources/SustainingForests.pdf Wunder, S., Angelsen, A., & Belcher, B. (2014). Forests, Livelihoods, and Conservation: Broadening the Empirical Base. World Development, 64(S1), S1–S11. Wunder, S., Börner, J., Shively, G., & Wyman, M. (2014). Safety Nets, Gap Filling and Forests: A Global-Comparative Perspective. World Development, 64(S1), S29–S42.  85 Yiran, G. A. B., Kusimi, J. M., & Kufogbe, S. K. (2012). A synthesis of remote sensing and local knowledge approaches in land degradation assessment in the Bawku East District, Ghana. International Journal of Applied Earth Observation and Geoinformation, 14(1), 204–213. Zulu, L. C. (2010). The forbidden fuel: Charcoal, urban woodfuel demand and supply dynamics, community forest management and woodfuel policy in Malawi. Energy Policy, 38(7), 3717–3730. Zulu, L. E. O. C. (2008). Community Forest Management in Southern Malawi: Solution or Part of the Problem? Society and Natural Resources, 21(8), 687–703.                                   86 Appendix A. Chapter 2 Additional Figures and Tables  Figure A. 1. Histograms of household values of standardized (z-scores) amount and walking time data for the effort sub-index, for fuelwood and wild foods collected from village forests, forest reserves and private woodlots, for the study area in southern Malawi (n=157).   Table A. 1. Summary statistics of annual non-forest income sources with input costs, and total annual non-forest income in U.S. Dollars (USD) for the study area in Malawi (n=157). All minimum values were 0.                     Fuelwood Amount (Standardized)Frequency-1 0 1 2 30102030405060Fuelwood Walking Time (Standardized)Frequency-1 0 1 2 3 40102030405060Frequency0 1 2 3 4 Wild Foods Amount (Standardized)020406080100140Frequency-1 0 1 2 3 Wild Foods Walking Time (Standardized)020406080100 87 Appendix B. Chapter 3 Additional Figures and Tables                        Figure B. 1. Diagram of the size, layout and sampling of each subplot within the forest plots. Note that the percent covers of grass, shrubs and herbs were not analyzed in this thesis.       r = 7.98 m, A = 200 m2 r = 2.39 m, A = 18 m2 r = 0.8 m, A = 9 m2 Trees (>4cm DBH) Percent cover grass, shrubs, herbs Seedlings (0-1 cm  DBH), Saplings (1-4 cm DBH)  88  Figure B. 2. Non-metric multidimensional scaling (NMDS) ordination for fuelwood harvesting sites (n=14) for counts of tree species per site, with mean elevation (R2 =0.52), mean distance to the main road (R2 =0.28) and mean number of rooftops in a 3 km radius (R2 =0.35), each averaged per site, plotted as vectors. Vectors scaled according to their R2 values.  NMDS1 and 2 are axes 1 and 2 respectively. Ordination constructed in R 3.4.0 using the package vegan, function metaMDS using Bray ecological distance and 4 dimensions (R Core Team, 2017).           89 Table B. 1. Ecological variables and proxies for harvesting access and pressure used in the generalized linear models (GLMs) averaged per site for fuelwood harvesting sites (n=14). Elevation is measured in meters above mean sea level and rooftop count is the number of rooftops within a 3 km radius circle around the forest plots, averaged per site. Elevation, distance to the main road and number of rooftops are the explanatory variables and species richness (number of species), species abundance (number of stems), Shannon index, Simpson’s diversity and AGC are the response variables used in the GLMs.    Table B. 2. Ecological variables and proxies for harvesting access and pressure averaged per site for reference sites (n=9). Species richness is the number of species and species abundance is the number of stems. Elevation is measured in meters above mean sea level and rooftop count is the number of rooftops within a 3 km radius circle around the forest plots, averaged per site. Note that GLMs were not conducted for reference sites but the data was compiled for observation.       90                                                  Table B. 3. Presence of tree, sapling and seedling species by percentage of plots, as well as average tree DBH (in cm) and average tree basal area (in m2) calculated by summing the basal area for each tree of the species and dividing by the number of stems, for fuelwood harvesting sites (50 plots) and reference sites (36 plots). Note that two tree species present in one fuelwood harvesting plot each (Macaranga capensis (Baill.) Benth. ex Sim and Newtonia buchananii (Baker) G.C.C.Gilbert and Boutique) do not have DBH or basal area averages due to the inability to measure these trees in the field. Exotic species indicated with an (E).    91                                 92               93 Table B. 4. Minimum, mean and maximum aboveground carbon values for fuelwood harvesting sites (n=14) and reference sites (n=9), for five mixed-species allometric equations designed for the miombo woodlands. Biomass values of trees per plot were summed per site and converted to Mg per hectare then converted to carbon per hectare by multiplying by 0.47. Differences between means for fuelwood harvesting sites and reference sites were significant for each equation (Kuyah, p=0.0001; Mugasha, p=0.0002; Kachamba, p=0.0001; Chidumayo, p=0.0002; Ryan, p=0.0002).  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.24.1-0371136/manifest

Comment

Related Items