The study was carried out between the month of September 2023 and April 2024 in the two community units Kabulokor and Atala Kamusio that are located in Loima sub-county, Turkana County, Kenya (Figure 1). This region is primarily semi-arid and experiences unpredictable rainfall and families are vulnerable to severe droughts, limited food supplies, and limited sources of income (Kenya National Bureau of Statistics (KNBS), 2019).

Kabulokor (home to approximately 253 households with an average family size of about five persons (Kenya National Bureau of Statistics (KNBS), 2019) is mainly occupied by agropastoral households who grow some crops alongside herding. In contrast, Atala Kamusio which has about 246 households with a similar average family size to Kabulukor (Kenya National Bureau of Statistics (KNBS), 2019) has historically relied almost entirely on livestock. However, as grazing resources decline in Atala Kamusio, pastoralists have just gradually introduced short-season, drought-tolerant vegetable production as an alternative and complementary livelihood strategy in response to declining pastoral resources and changing climate conditions.

These two community units were purposively selected for this study because both communities carry out vegetable production, despite their differing livelihood systems. This presented a unique opportunity to understand how agroecological methods are adapted, implemented, and learnt by dryland communities with diverse livelihoods.

The research used a mixed method design, mixing qualitative data collection with a formative research approach and worked in three clear steps: (1) Focus Group Discussions, (2) Trials of Improved Practices (TIPs), and (3) Reflection Workshops. Figure 2 presents the overall sequence of these three phases, illustrating the progression from FGDs to TIP implementation and, finally, reflection workshops in each community.

Participant recruitment, progression, and attrition across the different study phases are summarized in Table 1. This overview provides transparency regarding participant numbers, selection processes, and reasons for drop-out.

The TIPs catalog of improved practices and TIPs tools were pre-tested with farmers at Napuu Farm in Lodwar for 2 days. Prior to data collection, the note-taker as well as the six enumerators received a 2 day training how to advise on the tools. Similarly, the follow-up TIPs questionnaire was pre-tested with Turkana University College students, and this helped the research team refine structure of the questions, their flow, and language use.

A detailed overview of the data management and analysis process is shown in Figure 3. The figure outlines key steps, for thematic development, coding structures, reliability checks, and the integration of qualitative and quantitative findings.

In order to enhance transparency of the qualitative analysis, Table 2 provides examples of final coding categories and analytical themes generated from the workshop discussions. The coding was done using MAXQDA software, allowing the study to systematically organize the discussions into meaningful themes. This presentation focuses on how the themes were developed, rather than the results themselves.

Ethical approval for the study was obtained from the School of Graduate Studies, Turkana University College, and the Institutional Review Board (IRB) of the Alliance of Bioversity International and CIAT (IRB51). Additional clearance was granted by the National Commission for Science, Technology and Innovation (NACOSTI) under License No. NACOSTI/P/23/29342.

Before data collection began, all participants were informed about the purpose and objectives of the study. Participation was entirely voluntary, and individuals were free to provide or decline informed consent, or to withdraw at any stage without any consequences. Throughout the study, participant confidentiality was ensured by safeguarding names, notes, and audio recordings.

Findings from FGDs and follow-up visits revealed that farmers faced a combination of agronomic, ecological, and socio-economic constraints. Participants consistently emphasized water scarcity, destruction of crops by free-roaming livestock, and limited access to quality seeds as the most pressing barriers to production. Although FGDs are primarily qualitative, we report the frequency with which these themes were mentioned because this information was used to prioritize practices for the subsequent TIPs. Quantitative data from follow-up visits provided additional confirmation of these trends. The main challenges identified by farmers are summarized in Figure 4.

Pest infestation and poor soil health were the most severe agronomic challenges, while water scarcity the most critical ecological limitation. Limited access to quality seeds, crop destruction by free roaming livestock was a limitation mentioned in half of the FGD. Market access was a constraint mentioned by one third, only.

One participant from Atala Kamusio explained,

Similarly, a farmer from Kabulokor noted,

Based on these prioritized challenges, TIPs were designed to focus on low-cost soil fertility management, pest control, and improved water-use efficiency, while remaining compatible with local livelihood systems and available resources.

Farmers' perceptions of trial plot performance and their willingness to continue using agroecological practices were assessed during the final TIP visit (third TIPs visit). The qualitative analysis identified indicators provide insight into perceived effectiveness and short-term sustainability of the tested practices. Overall, farmers' assessments of plot performance closely aligned with their stated intentions to continue adopting selected practices beyond the trial period (Table 4).

It should be noted that although most farmers set up control plots next to their trial plots, a proportion did not do so consistently due to challenges related to land availability, labor, water availability or other resources. The, comparisons between trial and control plots are based on farmer-reported assessments and participatory evaluation during the final workshop rather than standardized experimental measurements.

The majority of farmers (79 %) reported that trial plots performed better than adjacent control plots. They cited improved retention of the moisture in the soil, healthier crop appearance, and reduced crop stress as indicators of better performance. In contrast, 16% of the farmers felt that the control plots performed better and attributed this to the fact there were less pests, and their familiarity with traditional farming practices. Only 5% of the farmers felt that there was no observable difference between the trial plots and control plots.

These perceptions translated into a high level of willingness to continue with the practices. During the third visit (final visit), 79% of farmers expressed a clear intention to continue using at least one agroecological technique beyond the trial period. A further 18% were undecided, mainly due to uncertainty around input availability, while only 3% reported unwillingness to continue any of the tried practices.

The approaches that utilized local resources, such as ash and animal manure showed the greatest connection with the willingness to continue practicing. On the other hand, practices that required the development particular skills, like bottle irrigation technique, were less likely to be sustained in both systems and showed the need for specific trainings and/or manuals.

Table 5 summarizes the extent to which key themes identified in FGDs, TIPs, and reflection workshops converged, strengthened, or complemented one another, highlighting the most robust and analytically relevant findings across methods.

Overall, triangulation revealed strong convergence for water scarcity, pest management, and animal manure use, confirming these as central determinants of agroecological adoption. In contrast, practices such as bottle irrigation were initially perceived as promising but were later constrained by practical and technical barriers, illustrating how implementation-based methods strengthened earlier qualitative findings. Complementary insights from reflection workshops further highlighted the role of labor constraints and peer learning in shaping adoption trajectories.

Together, these triangulated findings illustrate how perceived challenges, practical feasibility, and social learning interact to shape short-term agroecological adoption in semi-arid and pastoralist contexts.

This study identifies several interconnected factors shaping vegetable production and agroecological adoption in semi-arid Turkana. These include agronomic and ecological constraints, access to resources, livelihood strategies, and social processes influencing farmer decision-making. Together, these factors explain why some practices were adopted and sustained, while others were discontinued.

Looking at the results through the Sustainable Livelihoods Framework (SLF) helps to understand why farmers adopt some agroecological practices and not others. Adoption depends not only on how well a practice works, but also on what resources farmers have, the time and labor they can invest, their knowledge and skills, and the support from their community. Practices using resources readily available at home, like animal manure or ash for pest control, were adopted more easily. In contrast, practices that needed extra materials, technical know-how, or more effort—like bottle irrigation—were harder to maintain.

Social connections and learning from other farmers played an important role, showing the value of social capital. Farmers were more likely to try and continue practices they could observe working successfully on neighboring plots. Differences between agropastoralists and pastoralists show that household resources and livelihood strategies shape which practices are feasible.

Using the framework makes it clear that adoption is influenced by a combination of resources, labor, skills, and social support—not just the technical performance of a technique. This perspective helps explain why some practices lasted while others dropped off and highlights how farmers make decisions to manage risk and sustain their livelihoods in challenging, semi-arid conditions.

The findings of this study suggest several practical and policy actions that can support sustainable vegetable production in semi-arid drylands:

Overall, integrating low-cost, context-specific agroecological practices with technical, social, and institutional support can enhance productivity, household food security, and resilience, providing a foundation for broader scaling and policy adoption.

This study relied on short-term Trials of Improved Practices (TIPs), descriptive statistics, and farmers' own observation rather than long-term yield monitoring, economic cost–benefit analysis, or multi-season performance indicators. The findings should thus not be interpreted as evidence that the agroecological practices are superior or that they will always lead to sustained productivity gains. Instead, the results show what is considered feasible by the agro-pastoralists and pastoralists to implement in their settings indicating that standard agricultural extension service may not be effective. The strength of this study is that it shows how farmers perceive the benefits, and the short-term patterns of adoption under real-world dryland conditions which provides them with limited amount of vegetable which they may not access otherwise (Jordan et al., 2022; Oduor, 2022; Oduor et al., 2023). This approach naturally limits how far the results can be generalized to other areas with different social, ecological, or resource conditions.

Although farmers were encouraged to set up control plots alongside their trial plots, not everyone was able to do so consistently due to the harsh environmental conditions like very low water availability. This made formal experimental comparisons difficult, so we relied on farmers' own observations and participatory assessments rather than statistically controlled measurements. The focus of the study was therefore on transparency and farmer-led evaluation to strengthen the self-help capacity rather than experimental precision as the access to the agriculture extension system is limited.

Finally, the adoption patterns we observed are shaped by context-specific factors such as livelihood type, labor availability, mobility, and access to local resources. While these factors are key to understanding adoption in semi-arid drylands, they also mean that the findings should be applied cautiously to other settings. Future research could build on these results by conducting longer-term trials that include farmer led quantitative yield measurements, economic assessments, and gender-focused analyses of labor and decision-making. This would provide stronger evidence of both effectiveness of agroecological practices, the generalizability of the results and the potential for scaling up these practices.