There is increasing worldwide interest in agroforestry as a multifunctional land-use strategy that can improve farm productivity while generating multiple ecosystem services necessary for human populations trying to adapt to and mitigate climate change and restore degraded landscapes. However there are critical knowledge gaps in our scientific understanding of how a diversity of woody species under given management practices can be enhanced across different landscapes to deliver a range of these benefits. Reflections on the agenda for scaling up agroforestry suggest that there is no ‘one size fits all’ technology that can be promoted across large areas; instead menus of options have to be tailored to local contextual variations. With the increasing evidence that systematic acquisition of local knowledge is a valuable way of complementing scientific information, there is an urgent need to develop novel techniques to further elicit local knowledge and integrate it into all efforts to develop more inclusive agroforestry options, co-designing them to build resilience in landscapes and livelihoods. The four papers presented in this thesis draw from research in diverse smallholder landscapes in Sub-Saharan Africa. I used different participatory research methods to explore farmers’ tree management practices and knowledge of a diversity of trees, along with their functions and the agro-ecological interactions at play in land-use and livelihood dynamics. These contrasting contexts illustrate a diversity of major agricultural systems, encompassing cocoa farming in Côte d’Ivoire, coffee cultivation in Rwanda, and the management of native multiple purpose trees in the West African agroforestry parklands in Burkina Faso. The ‘options-by-context’ approach, applied through a multiple stakeholder engagement process to address the heterogeneity in the landscape and of land users in eastern Democratic Republic of Congo (DRC), holds general lessons for scaling up agroforestry. The key findings of the research show that farmers’ knowledge about a range of useful native and exotic trees, as well as the contextual variations associated with their management, is rich and complementary to science and can be articulated both qualitatively and quantitatively. The novel way of ranking trees by attributes and the explicit probability model, developed and tested in both the tropical highland context in Rwanda and in the semi-arid parklands in Burkina Faso, was found to be a quick and cost-effective way of classifying a broad range of trees managed by farmers based on ecological, management and utility attributes. The tree ranking estimates were consistent and in agreement with scientific assessments when they could be compared, thus allowing predictions for some of their agroecological effects. This knowledge can be explicitly integrated in tree-planting or agroforestry development initiatives to provide for more objective assessments of how a diverse range of tree species, largely unknown to science but important in farmers’ practice, might be expected to affect farm production and other ecosystem services. The stakeholder engagement approach used in eastern DRC was innovative as it built on explicit acquisition of local knowledge to facilitate a systematic consideration of trees at field, farm and landscape scales. This enabled the consideration of different options, in terms of practices/technologies but also market interventions and institutional reform against the contexts for which they were relevant (covering ecological, economic, social and cultural factors). We found that this approach led to a change in the attitudes and knowledge around tree planting by stakeholders, with an important shift away from the promotion of a handful of exotic tree species in woodlots, largely benefiting wealthier men, to recommendations for over 70 tree species, 30 of them native, with management practices that addressed the needs of women, various ethnic groups and different types of farmers. I conclude that a knowledge-intensive framework is required to design more inclusive, locally adapted and diversified tree-based options that aim to deliver both environmental and socio-economic benefits to a wider range of stakeholders. This framework explicitly looks at integrating local and scientific knowledge through the facilitation of broad-based stakeholder participation to identify agroforestry options for different contexts and the preconditions that may require interventions in the enabling environment. The next steps would be to investigate whether, by being sensitive to the needs and context of different smallholders, a knowledge-intensive framework leads to more effective scaling up of agroforestry than do conventional approaches to tree planting.