A pervasive challenge to ending hunger and reducing poverty is that farming practices known to sustainably increase production and reduce and reverse land degradation processes are not widely adopted by smallholder farmers. Current approaches to development-focused agricultural research typically focus on maximising the productivity of one component of the farming system and evaluating innovations based on averages and plot-scale metrics such as yield per hectare. Yet, for restorative farming practices to be adopted, they not only need to be productive and profitable they also need to be attractive within the broader context of smallholder livelihood systems. This thesis responds to the need for systemic approaches to evaluating innovations and attempts to embrace both the complexity and diversity of smallholder livelihood systems through its methods. Conducted in the context of a large-scale land restoration project working with over 2,500 farmers in the eastern drylands of Kenya, my overarching goal was to assess the impact of on-farm restoration practices (planting basins and tree planting) on the livelihoods of participating households, so as to improve the specificity of recommendations and scaling of restoration efforts.

In Chapter 2, I present various analytical approaches to assessing the efficacy of planting basins for growing maize. In doing so, I demonstrate how even simple arithmetic and translating plot- scale metrics into farm-scale metrics can be a helpful first step towards more farmer-relevant assessments. In Chapter 3, I develop a farm-scale model that extends the results from Chapter 2 across multiple years and explores the impact of planting basins for two households with contrasting resource endowments. Chapters 4 and 5 take a more qualitative and inductive approach, broadening the scope of my assessment to include the role of intrahousehold gender dynamics and aspirations. Combining survey data on decision-making and labour participation with interviews and focus group discussions, I investigate how gender roles and relations influence the uptake of planting basins and tree planting. In Chapter 5, I explore the changing livelihood aspirations of rural women amidst intensifying male outmigration using a novel narrative-based survey tool, contributing to a more nuanced understanding of rural aspirations and the socio-cultural context within which they are embedded.

Although planting basins provided impressive gains in average maize yield, there was strong variability in their performance. For some households, they provided substantial production increases, extra days of food and a potential income boost, while other households faced substantial losses. This variability in performance presents substantial risks for farmers when basins are promoted as a generalised recommendation. Furthermore, plot-scale metrics such as yield and intensity of adoption overlooked the wider role planting basins play within people’s livelihoods, other than maximising yield (e.g., yield stability) and share of benefits amongst household members. Chapters 4 and 5 revealed that heterogeneity also exists in the aspirations of those within households and that women are likely to be important catalysts of agricultural innovation amid the increasing outmigration of men and feminisation of farm management.

Attempting to apply a more systemic approach to evaluating restoration practices, this thesis is interdisciplinary in nature and broad in its scope and use of methods. It brings together alternative approaches to assessing agricultural innovations, and in doing so, stresses the need for development-focused agricultural research to step away from a fixation on differences in mean yield and to embrace variation in innovation performance and complexity of smallholder livelihoods, not avoid it.