Innovation in agronomy by large numbers of smallholder farmers will need to accelerate if global commitments to end hunger are to be achieved in the face of major climate and other global change that are both caused by, and impact, agriculture. Conventional research and development in agronomy have involved a research process that produces technologies, which are then promoted for uptake by large numbers of farmers through extension, with both research and extension phases being more or less participatory. Recent research, including key contributions to this special issue, reveals that the performance of many technology options varies hugely across the geographies over which development programmes operate, depending on social, economic and ecological context. This severely limits the value of attempting to produce recommendations for large areas and numbers of farmers and identifies the need for new ways of supporting innovation that address the real-world heterogeneity of farmer circumstances. Addressing this widespread phenomenon of option by context interaction (OxC) has profound implications for how agronomic research and development are organised. Papers in this special issue show the nature and implications of such interactions and suggest ways in which research and development systems need to respond in order to support locally relevant innovation. It is evident that a paradigm shift is well underway, with researchers embracing new modes of thinking and action required to address OxC interactions, but these also need to be taken up and further developed by extension and change agents in the public and private sector. It is only through continued co-development of methods involving both these constituencies, working closely with farmers that sufficient progress is likely to be made for smallholder farming to keep pace with global demand for food without further damaging the environmental resources upon which production is based.

Climbing bean is the key staple legume crop in the highlands of East and Central Africa. We assessed the impact of interactions between soil fertility characteristics, crop management and socio-economic factors, such as household resource endowment and gender of the farmer, on climbing bean productivity and yield responses to basal P fertiliser in northern Rwanda. Through a combination of detailed characterisations of 12 farms and on-farm demonstration trials at 110 sites, we evaluated variability in grain yields and responses to fertiliser. Grain yields varied between 0.14 and 6.9 t ha−1 with an overall average of 1.69 t ha−1. Household resource endowment and gender of the farmer was strongly associated with climbing bean yield, even though these were partly confounded with Sector. Poorer households and women farmers achieved lower yields than wealthier households and male farmers. Household resource endowment and gender were likely to act as proxies for a range of agronomic and crop management factors that determine crop productivity, such as soil fertility, current and past access to organic manure and mineral fertiliser, access to sufficient quality staking material, ability to conduct crop management operation on time, but we found evidence for only some of these relationships. Poorer households and female farmers grew beans on soils with poorer soil fertility. Moreover, poorer households had a lower density of stakes, while stake density was strongly correlated with yield. Diammonium phosphate (DAP) fertiliser application led to a substantial increase in the average grain yield (0.66 t ha−1), but a large variability in responses implied that its use would be economically worthwhile for roughly half of the farmers. For the sake of targeting agricultural innovations to those households that are most likely to adopt, the Ubudehe household typology – a Rwandan government system of wealth categorisation – could be a useful and easily available tool to structure rural households within regions of Rwanda that are relatively uniform in agro-ecology.

The extent to which coffee agroforestry systems provide ecosystem services depends on local context and management practices. There is a paucity of information about how and why farmers manage their coffee farms in the way that they do and the local knowledge that underpins this. The present research documents local agro-ecological knowledge from a coffee growing region within the vicinity of the Aberdare Forest Reserve in Central Kenya. Knowledge was acquired from over 60 coffee farmers in a purposive sample, using a knowledge-based systems approach, and tested with a stratified random sample of 125 farmers using an attribute ranking survey. Farmers had varying degrees of explanatory knowledge about how trees affected provisioning and regulating ecosystem services. Trees were described as suitable or unsuitable for growing with coffee according to tree attributes such as crown density and spread, root depth and spread, growth rate and their economic benefit. Farmers were concerned that too high a level of shade and competition for water and nutrients would decrease coffee yields, but they were also interested in diversifying production from their coffee farms to include fruits, timber, firewood and other tree products as a response to fluctuating coffee prices. A range of trees were maintained in coffee plots and along their boundaries but most were at very low abundances. Promoting tree diversity rather than focussing on one or two high value exotic species represents a change of approach for extension systems, the coffee industry and farmers alike, but is important if the coffee dominated landscapes of the region are to retain their tree species richness and the resilience this confers.

This paper presents the main features of a unique decision-support tool developed for selecting tree species in coffee and cocoa agroforestry systems. This tool aims at assisting in the selection of appropriate shade trees taking into account local conditions as well as needs and preferences of smallholder farmers while maximizing ecosystem services from plot to landscape level. This user-friendly and practical tool provides site-specific recommendations on tree species selection via simple graphical displays and is targeted towards extension services and stakeholders directly involved in sustainable agroforestry and adaptation to climate change. The tool is based on a simple protocol to collect local agroforestry knowledge through farmers’ interviews and rankings of tree species with respect to locally perceived key ecosystem services. The data collected are first analysed using the BradleyTerry2 package in R, yielding the ranking scores that are used in the decision-support tool. Originally developed for coffee and cocoa systems of Uganda and Ghana, this tool can be extended to other producing regions of the world as well as to other cropping systems. The tool will be tested to see if repeated assessments show consistent ranking scores, and to see if the use of the tool by extension workers improves their shade tree advice to local farmers.

Agricultural development projects frequently promote new crop production technologies for adoption at scale on the basis of research and pilot studies in a limited number of contexts. The performance of these production technologies is often variable and dependent on context. Using an example from the Agroforestry for Food Security Project in Malawi, that promoted agroforestry technologies for soil fertility enhancement, we explore the nature and implications of variation in performance across farmers. Mean effects of these technologies, measured by differences in maize yield between agroforestry and sole maize plots, were modest but positive. However, there was large variation in those differences, some explained by altitude, plot management and fertilizer use but with much unexplained. This represents risk to farmers. Those communicating with farmers need to be honest and clear about this risk. It can be reduced by explanation in terms of contextual factors. This should be an aim of research that can often be embedded in scaling up the promotion of agronomic innovations.

In recent years, many studies have demonstrated the heterogeneity of the smallholder production environment. Yet agronomic research for development (R4D) that aims to identify and test options for increasing productivity has not consistently adapted its approaches to such heterogeneous conditions. This paper describes the challenges facing research, highlighting the importance of variation in evaluating the performance of soil management recommendations, integrating aspects of production risk management within the formulation of recommendations, and proposing alternative approaches to implement agronomic R4D. Approaches are illustrated using two multi-locational on-farm paired trials, each having one no-input control treatment and a treatment with fertilizer application for maize in Western Kenya and for beans in Eastern Rwanda. The diversity of treatment responses should be embraced rather than avoided to gain a better understanding of current context and its relation with past management.

Agriculture research often focuses on a technical problem. However, the most effective researchers usually intuit that this entry point is not sufficient to make themselves useful to farmers and nudge systems. Yet non-technical work frequently goes undocumented leaving many of the drivers of success unstudied. This paper attempts to understand the factors that contributed to the wide utilization of native crop varieties and species that were being promoted by the Ecuadorian National Agriculture Research Institution. The results show that what really made a difference in farmers' lives and the overall food system was increasing farmers' knowledge and capacity to produce quality seeds, promoting the consumption of these crops to national consumers, and linking farmers to outside groups. As a result, over a period of five years three case studies on three different farmer groups showed adoption rates of new varieties of between 20–50% and that they were able to produce approximately 7.5% of the annual demand for quinoa and lupin seed in Ecuador, from a starting point of virtually nothing. The research shows that the added value of a research institution might not be known at the beginning of the intervention, but rather will emerge over time through dialogue and negotiation based on systematic understanding of the context. Therefore, an appropriate stance for external organizations is to begin with an awareness of the existing assets of a specific farmer group and provide options that can be leveraged by local communities.

The agricultural research and development institutions in most developing countries are poorly equipped to support the needs of millions of smallholder farmers that depend upon them. The research approaches taken by these systems explicitly or implicitly seek simple, one-size-fits-all solutions for problems and opportunities that are extremely diverse. Radical change is needed to facilitate the agroecological intensification of smallholder farming. We propose that large-scale participatory approaches, combined with innovations in information and communications technology (ICT), could enable the effective matching of diverse options to the wide spectrum of socio-ecological context that characterize smallholder agriculture. We consider the requirements, precedents and issues that might be involved in the development of farmer research networks (FRNs). Substantial institutional innovation will be needed to support FRNs, with shifts in roles and relationships amongst researchers, extension providers and farmers. Where farmers’ organizations have social capital and strong facilitation skills, such alignments may be most feasible. Novel information management capabilities will be required to introduce options and principles, enable characterization of contexts, manage data related to option-by-context interactions and enable farmers to visualize their findings in useful and intelligible ways. FRNs could lead to vastly greater capacity for technical innovation, which could in turn enable greater productivity and resilience, and enhance the quality of rural life.

This research sought to link Andean soil knowledge and farmer categorization of soil fertility to soil science characterization of soils, and use these to understand the impacts of phosphorus (P) fertilization of legumes using rock phosphate and soluble P fertilizer in 17 smallholder-managed sites with varying soil properties. We found that farmer high/low categorization of soils corresponded to soil P fertility and distance from farmer dwellings. Measures of soil P fertility also were inversely related to mycorrhizal colonization of vetch roots and directly related to the potential for P release by legume residues (C:P ratio). However, particular soil properties (texture and calcium phosphate pools) were better in explaining the response of legume biological nitrogen fixation to P addition, with maximal impacts in low-clay soils and soils with low calcium phosphates, as assessed with a dilute HCl extraction. In these conducive contexts, legume BNF increased 67 and 150% for RP and TSP, respectively (p < 0.05), with similar increases in biomass P stocks that showed promise for soil regeneration. In low-P fields, added P increased potential P release from legume residues via lower C:P ratios (p < 0.05). Percent cover of soil increased from 35% in the control to 45 and 55% with RP and TSP (p < 0.05), which shows potential for reduced erosion risk under P fertilization of legumes. We discuss results with respect to the integration of local farmer knowledge systems with exogenous scientific knowledge.

The large diversity of farms and farming systems in sub-Saharan Africa calls for agricultural improvement options that are adapted to the context in which smallholder farmers operate. The socio-ecological niche concept incorporates the agro-ecological, socio-cultural, economic and institutional dimensions and the multiple levels of this context in order to identify which options fit best. In this paper, we illustrate how farming systems analysis, following the DEED cycle of Describe, Explain, Explore and Design, and embedding co-learning amongst researchers, farmers and other stakeholders, helps to operationalize the socio-ecological niche concept. Examples illustrate how farm typologies, detailed farm characterization and on-farm experimental work, in combination with modelling and participatory approaches inform the matching of options to the context at regional, village, farm and field level. Recommendation domains at these gradually finer levels form the basis for gradually more detailed baskets of options from which farmers and other stakeholders may choose, test and adjust to their specific needs. Tailored options identified through the DEED cycle proof to be more relevant, feasible and performant as compared to blanket recommendations in terms of both researcher and farmer-identified criteria. As part of DEED, on-farm experiments are particularly useful in revealing constraints and risks faced by farmers. We show that targeting options to the niches in which they perform best, helps to reduce this risk. Whereas the conclusions of our work about the potential for improving smallholders’ livelihoods are often sobering, farming systems analysis allows substantiating the limitations of technological options, thus highlighting the need for enabling policies and institutions that may improve the larger-scale context and increase the uptake potential of options.

Understanding farm diversity is essential to delineate recommendation domains for new technologies, but diversity is a subjective concept, and can be described differently depending on the way it is perceived. Historically, new technologies have been targeted primarily based on agro-ecological conditions, largely ignoring socioeconomic conditions. Based on 273 farm households' surveys in Ethiopia, we compare two approaches for the delineation of farm type recommendation domains for crop and livestock technologies: one based on expert knowledge and one based on statistical methods. The expert-based typology used a simple discriminant key for stakeholders in the field to define four farm types based on Tropical Livestock Unit, total cultivated surface and the ratio of these two indicators. This simple key took only a few minutes to make inferences about the potential of adoption of crop and livestock technologies. The PCA-HC analysis included a greater number of variables describing the farm (land use, household size, cattle, fertilizer, off-farm work, hiring labour, production). This analysis emphasized the multi-dimensional potential of such a statistical approach and, in principle, its usefulness to grasp the full complexity of farming systems to identify their needs in crop and livestock technologies. A sub-sampling approach was used to test the impact of data selection on the diversity represented in the statistical approach. Our results show that diversity structure is significantly impacted according to the choice of a sub-sample of 15 of the 20 variables available. This paper shows the complementarity of the two approaches and demonstrates the influence of data selection within large baseline data sets on the total diversity represented in the clusters identified.

The new Brazilian Forest Code obliges farmers to restore degraded lands on conservation set-asides required on all rural properties, known as Permanent Preservation Areas and Legal Reserves. While farmers can do so through agroforestry systems (AFS) to improve their livelihoods and offset the costs of restoration, these areas must also maintain basic ecological functions. How this can be done in practice, however, stills need to be regulated at the state level, but there are few scientific studies in Brazil on the costs and benefits of restoration through AFS. These knowledge and policy gaps have left ample room for interpretation and thus discouraged technicians from recommending — and policymakers from regulating — AFS for restoration. Aiming to bridge these gaps, this study draws from a comprehensive literature review, expert knowledge, stakeholder participation and analyses of agroforestry-based restoration efforts in the field. First, we discuss the social and ecological benefits of AFS as well as key constraints to upscaling agroforestry-based restoration. Focusing on different biophysical and social contexts within the Brazilian Cerrado and Caatinga biomes, we conclude that AFS can provide win–win solutions for reconciling livelihoods needs with conservation goals and propose practical options to do so, which vary according to farmer objectives, resource constraints and opportunities. In order to guide implementation of these agroforestry options, we propose design principles and criteria as well as key species and management practices most suited to different contexts and farmer objectives.

The success of scaling out depends on a clear understanding of the factors that affect adoption of grain legumes and account for the dynamism of those factors across heterogeneous contexts of sub-Saharan Africa. We reviewed literature on adoption of grain legumes and other technologies in sub-Saharan Africa and other developing countries. Our review enabled us to define broad factors affecting different components of the scaling out programme of N2Africa and the scales at which those factors were important. We identified three strategies for managing those factors in the N2Africa scaling out programme: (i) testing different technologies and practices; (ii) evaluating the performance of different technologies in different contexts; and (iii) monitoring factors that are difficult to predict. We incorporated the review lessons in a design to appropriately target and evaluate technologies in multiple contexts across scales from that of the farm to whole countries. Our implementation of this design has only been partially successful because of competing reasons for selecting activity sites. Nevertheless, we observe that grain legume species have been successfully targeted for multiple biophysical environments across sub-Saharan Africa, and to social and economic contexts within countries. Rhizobium inoculant and legume specific fertiliser blends have also been targeted to specific contexts, although not in all countries. Relatively fewer input and output marketing models have been tested due to public–private partnerships, which are a key mechanism for dissemination in the N2Africa project.

There is a lot of interest in the contribution that agroforestry can make to reverse land degradation and create resilient multifunctional landscapes that provide a range of socio-economic benefits. The agroforestry research agenda has been characterized by approaches that promote a few priority tree species, within a restricted set of technological packages. These have often not spread widely beyond project sites, because they fail to take account of fine scale variation in farmer circumstances. New methods are needed to generate diverse sets of agroforestry options that can reconcile production and conservation objectives and embrace varying local conditions across large scaling domains. Here, we document a novel approach that couples local knowledge acquisition with structured stakeholder engagement to build an inclusive way of designing agroforestry options. We applied this approach in the eastern part of the Democratic Republic of Congo (DRC) where armed conflict, erratic governance and poverty have resulted in severe pressure on forests in the Virunga National Park, a global biodiversity hotspot. Around the park, natural resources and land are severely degraded, whereas most reforestation interventions have consisted of exotic monocultures dominated by Eucalyptus species grown as energy or timber woodlots mainly by male farmers with sufficient land to allocate some exclusively to trees. We found that structured stakeholder engagement led to a quick identification of a much greater diversity of trees (more than 70 species) to be recommended for use within varied field, farm and landscape niches, serving the interests of a much greater diversity of people, including women and marginalized groups. The process also identified key interventions to improve the enabling environment required to scale up the adoption of agroforestry. These included improving access to quality tree planting material, capacity strengthening within the largely non-governmental extension system, and collective action to support value capture from agroforestry products, through processing and market interventions. Integrating local and global scientific knowledge, coupled with facilitating broad-based stakeholder participation, resulted in shifting from reliance on a few priority tree species to promoting tree diversity across the Virunga landscape that could underpin more productive and resilient livelihoods. The approach is relevant for scaling up agroforestry more generally.

Rapid climatic and socio-economic changes challenge current agricultural R&D capacity. The necessary quantum leap in knowledge generation should build on the innovation capacity of farmers themselves. A novel citizen science methodology, triadic comparisons of technologies or tricot, was implemented in pilot studies in India, East Africa, and Central America. The methodology involves distributing a pool of agricultural technologies in different combinations of three to individual farmers who observe these technologies under farm conditions and compare their performance. Since the combinations of three technologies overlap, statistical methods can piece together the overall performance ranking of the complete pool of technologies. The tricot approach affords wide scaling, as the distribution of trial packages and instruction sessions is relatively easy to execute, farmers do not need to be organized in collaborative groups, and feedback is easy to collect, even by phone. The tricot approach provides interpretable, meaningful results and was widely accepted by farmers. The methodology underwent improvement in data input formats. A number of methodological issues remain: integrating environmental analysis, capturing gender-specific differences, stimulating farmers' motivation, and supporting implementation with an integrated digital platform. Future studies should apply the tricot approach to a wider range of technologies, quantify its potential contribution to climate adaptation, and embed the approach in appropriate institutions and business models, empowering participants and democratizing science.

In the past two decades, a growing body of work on research stations and farmers' fields in Southern Africa has provided evidence that fertilizer trees can improve the productivity of land, increase crop yields and contribute towards climate change mitigation and adaptation. In a recent issue of Experimental Agriculture, Coe et al. (2019) published analysis of risks associated with adopting agroforestry in Malawi. The article contains several factual inaccuracies about agroforestry and misinterpretations of earlier work. Our aim in this correspondence is, therefore, to point out the key problems, seek clarification from Coe and co-workers, and stimulate wider scientific debate on the perceived risks of adopting agroforestry.

Our paper ‘Loading the dice in favour of the farmer: reducing the risk of adopting agronomic innovations’ revealed mean increases but also large variation in the impact of four agroforestry practises on maize yield, as experienced by farmers in Malawi. This prompted a response from Sileshi and Akinnifesi that was critical of the data and methods used. Their main concern was that farmers did not necessarily manage crops identically in plots with and those without trees, so the yield differences that we measured may be partly caused by these differences in crop management. We argue here that it is valid and useful to look at the actual effect on crop yield of farmers having trees intercropped with maize, rather than controlling for how the crop is managed, because this is what happens in the real world. Farmers respond to having trees in their field by treating their crop differently, so this is part of the system response to having trees in fields. Attempts to eliminate this will result in measuring an artefact rather than the real impact of trees on crop yield. By doing this, we revealed important variation in the impact of trees on crop yield amongst farmers, and we argue that it is important to explore, assess and communicate to farmers and development actors the extent and implications of this variation. Understanding the contextual factors that determine who is likely to benefit most from an innovation and for whom it is less suitable can then be incorporated in scaling up, so that targeting of innovations and the appropriateness of messages given to farmers are continuously refined.