Constraints for adopting climate-smart agricultural practices among smallholder farmers in Southeast Kenya

• Awareness and access are key factors in determining utilization of climate-smart agricultural practices and technologies. • Objective was to identify climate-smart agriculture constraints among farmers in Taita Taveta County of South-East Kenya. • The study applied Climate-Smart Agriculture Rapid Appraisal. • Climate-smart agriculture require more localized solutions in policies, extension services and development interventions. • This study increases understanding of constraints affecting adoption of new farm and land management practices. A R T I C L E I N F O Guest Editor: Jagadish Timsina.

• Awareness and access are key factors in determining utilization of climate-smart agricultural practices and technologies. • Objective was to identify climate-smart agriculture constraints among farmers in Taita Taveta County of South-East Kenya. • The study applied Climate-Smart Agriculture Rapid Appraisal. • Climate-smart agriculture require more localized solutions in policies, extension services and development interventions. • This study increases understanding of constraints affecting adoption of new farm and land management practices.

Introduction
The biophysical and socio-economic effects of anthropogenic climate change show large spatial variation globally (Wiebe et al., 2015), while responses to changes in global food security discourse emphasize homogeneous technical responses in regional and national policies. Rigorous scientific and political debate is ongoing concerning farmers' capability to adapt to changing climatic conditions and possibilities to mitigate climate change without threatening their livelihoods. Climatesmart agriculture (CSA) is a key concept in the present discourse of climate change mitigation and adaptation (FAO, 2017;IPCC, 2019;Klytchnikova et al., 2015). It can be defined as an agricultural activity that: 1) mitigates greenhouse gas emissions, 2) increases the adaptive capacity of farmers, and 3) sustainably intensifies agriculture for better livelihoods (Lipper et al., 2014).
The main critique of CSA concerns its technically oriented nature, as it fails to consider for example labour advances, consumption patterns, and land tenure questions (Taylor, 2018). Climate-smart agriculture also suffers from conceptual misunderstandings in the scientific community and resistance from civil society (Saj et al., 2017). Despite being in a restructuring stage, the concept has been proven to fulfil its three-win scenario worldwide, with positive impacts on the environment, agriculture, and socio-economic development (FAO, 2014;Neate, 2013). Recent studies have shown that common CSA technologies, such as no tillage, integrated soil fertility management, alternate wetting and drying, and nitrogen use efficiency, may bring benefits on a global scale (De Pinto et al., 2020). In addition, CSA technologies can substantially reduce women's labour burdens in agriculture (Khatri-Chhetri et al., 2019). Thornton et al. (2018) emphasized that CSA prioritization must also consider spatial and temporal scales to address producer and consumer needs. These findings increase the incentive to test their applicability in case studies.
As Andrieu et al. (2019) suggest, co-designing climate-smart farming systems is a complex process in which knowledge, technologies, and institutional environments interact, and a participatory and systems approach is therefore needed to generate local scientific knowledge that can identify appropriate solutions. In addition, Lan et al. (2018) have shown great variation in farm profitability with CSA practices across scales, meaning that similar practices can generate different profitability outcomes depending on the context. Understanding the constraints and local knowledge on CSA practices assists in formulating extension services, agricultural policies, and relevant interventions in the geographical context (Aryal et al., 2018). Empirical evidence from Western Kenya shows that locally driven adaptation incentives also provide implementation cost benefits (Chaudhury et al., 2016). Despite the importance of local knowledge in identifying appropriate climate-smart solutions, we found that the knowledge interactions on farm and land management practices between farmers and state actors are not clearly defined in the current CSA literature.
Our study investigates the spatial differences in climate change constraints of smallholder farmers in Taita Taveta County (TTC) of Southeast Kenya. The county's agriculture is drastically affected by climate change while also being a major source of greenhouse gas emissions exacerbating climate change and land-cover change due to agricultural expansion (Ge and Friedrich, 2020;Pellikka et al., 2018;Wachiye et al., 2020). This study uses the Climate-Smart Agriculture Rapid Appraisal (CSA-RA) -tool in TTC. The tool enables evaluating the state of agriculture to recognize challenges faced by smallholder farmers and to identify locally suitable practices that can be promoted in policymaking and development interventions (Mwongera et al., 2017). The tool is gender disaggregated, incorporating the fact that agricultural practices are also shaped by the cultural norms of societies, including gender-based roles in agriculture. Furthermore, by including gender when examining CSA adoption constraints, our aims are also to recognize women as active agents in adapting to change and contribute to filling the evident gap of research into the many gender dimensions of climate change, especially from a localized perspective (MacGregor, 2010;Ravera et al., 2016). On a broader level, we aimed to identify the primary constraints of smallholder farmers and CSA across TTC, to gain a view of possible areas requiring interventions on both policy and implementation levels.
Climate-Smart Agriculture Rapid Appraisal has previously been tested at a regional level in Tanzania and Uganda (Mwongera et al., 2015). In Kenya, a multicriteria decision support framework to target CSA intervention needs has been tested at the national level, although focusing on quantifiable data concerning vulnerability (Brandt et al., 2017). An earlier study in the Ndome and Ghazi areas of TTC has identified secured access to land as a significant disincentive for structural conservation measures for land improvement (Waswa et al., 2002). Previous research has also shown that cultivation of the most prominent agricultural land has occurred already a decade ago, and agriculture is expanding into areas with lower precipitation, stressing the need to adopt better soil conservation practices (Maeda et al., 2011;Maeda et al., 2010;Pellikka et al., 2018;Pellikka et al., 2013). Participatory methods for examining climate change impacts on traditional farming communities of the Taita Hills in TTC have previously been utilized by Capitani (2019). Our study broadens the analysis to cover both highlands and lowlands, which are dependent on ecosystem services such as water provisioning provided by the highland catchment areas. Earlier research on water governance in the area by Hohenthal et al. (2018) identified significant challenges in collaboration and dialogue-building between local people and state actors within power relations. Thus, there appears to be weak confidence in local resource management and knowledge in TTC.
Our research questions to investigate the CSA constraints are: 1) Which climate change -related challenges do farmers and state representatives identify for agriculture across TTC?
2) What factors restrain the adoption of climate-smart practices and use of climate-smart technologies across TTC?
3) Does spatial heterogeneity in agro-ecological conditions on a regional level necessitate locally appropriate responses to climate change adaptation?
Our hypotheses are that existing heterogeneity in the constraints to regional-level climate change adaptation prevents uniform upscaling of CSA practices and all-encompassing policies for small-scale agriculture in TTC (Abegunde et al., 2019), and that this heterogeneity is realized as a perceived knowledge gap in adapting CSA technologies. Heterogeneity in the constraints to climate change adaptation and suitable responses to it not only manifest spatio-temporally but also among the stakeholder groups depending on their socio-economic profiles, thus having both potential winners and losers (Notenbaert et al., 2017). Farmers lacking surplus land resources or investment capital have limited access to many of the climate change adaptation techniques, and possible intervention could come at the expense of disadvantaged farmers if this is not taken into consideration (Clay and King, 2019). Besides resource-related issues, farmer access to training, farm inputs, climate information technologies, markets, and credit have been identified as key challenges in adopting CSA technologies, particularly for women farmers (World Bank Group et al., 2015). Determining how these constraints differ by socioeconomic profiles and gender is crucial to understanding the adaptive capacities of agricultural systems.

Study area
Taita Taveta County (TTC) in the Coast Region of Kenya ( Fig. 1) consists of Mwatate, Taveta, Voi, and Wundanyi sub-counties. Sixtythree per cent of TTC is covered by the Tsavo West and Tsavo East National Parks that border Taveta in the east and surround Mwatate, Voi, and Wundanyi. The Taita Hills that rise to over 1500 m.a.s.l., peaking at 2208 m, characterize the landscape, while the lower slopes of Mount Kilimanjaro characterize the landscape in Taveta. The county has a population of 340,671 (Kenya National Bureau of Statistics, 2019a). A large proportion of the land is communally owned, with approximately 35% of the population having title deeds for land and 55% with no formal land ownership documentation (County Government of Taita Taveta, 2018). Agriculture is the main source of livelihood, contributing to approximately 95% of household incomes and over 80% of employment (MoALF, 2016:1). In 2015-2016, the food poverty estimate for TTC was 39% which means that 139,000 people were unable to meet daily dietary recommendations (Kenya National Bureau of Statistics, 2018: 46). This exceeds the national average of 32%.
TTC has a bimodal rainfall pattern, with long rains occurring between March and May, and short rains between October and December (Ogallo et al., 2019). The County has nine agro-ecological zones (see Fig. 2 for Taita Hills) where microclimatic and soil conditions vary greatly, depending strongly on the altitude (Jaetzold et al., 2012). The average farm size is 0.4 ha in the highlands, 1.3 ha in the midlands, and 4.8 ha in the lowlands (County Government of Taita Taveta, 2018). Commercialized irrigated farming is practised in Taveta, while farming in Taita is mainly rain-fed domestic production. Agricultural products, such as tomatoes (Lycopersicon esculentum) and onions (Allium cepa), are also imported from Tanzania, while local cash crop production is exported e.g. to Mombasa.
This study was carried out in the wards of Kishushe, Maktau, Ngerenyi, and Chala, which represent different agro-ecological settings (see Table 1) that tend to also result in different adaptation strategies (Amare and Simane, 2017). Kishushe, Maktau, and Chala in the lowlands have low rainfall, while Ngerenyi in the highlands experiences more abundant rainfall. Kishushe and Maktau are characterized by predominantly subsistence agriculture with crops, such as maize (Zea mays), common beans (Phaseolus vulgaris L.), cowpeas (Vigna unguiculata), green grams (Vigna radiata), pigeon peas (Cajanus cajan), and sorghum (Sorghum bicolor), grown for domestic consumption. Ngerenyi and Chala have a stronger focus on commercial farming with the cultivation of e.g. bananas (Musa spp.), macadamia nuts (Macadamia tetraphylla), tomatoes, onions, and kales (Brassica spp.), in addition to subsistence crops.

Materials and methods
Climate-Smart Agriculture Rapid Appraisal is a research tool developed to evaluate agricultural conditions in adapting to climate change, mitigating greenhouse gases, and providing sustainable livelihoods (Mwongera et al., 2017). It combines both participatory rural appraisal (PRA) and rapid rural appraisal (RRA) methods to form a holistic view of CSA practices and technologies. Practices are understood broadly as ways of doing things, for example, precision farming, tillage, and fertilization. Technologies are new farm products, services, or applications introduced into new or existing practices, and include e.g. new drought-tolerant varieties; a hardy breed of cattle; or a new slow-release fertilizer. The RRA methods include farmer and key informant interviews to gather basic socio-economic information. The PRA methods engage community members into participatory exercises. In this study, data were gathered through household surveys, farmer workshops, key informant interviews, and transect walks. The CSA-RA tools and their validity for the local community were discussed and modified together with local stakeholders before the actual fieldwork. This included adjusting the questions asked based on prior knowledge of local conditions, such as contemplating whether a clan-based land tenure system still played a role in agricultural activities or how to discuss the interconnectedness of the lowlands and highlands. Preliminary fieldwork was conducted in July-August 2018 and the actual CSA-RA in February-March 2019.

Workshop exercises
Altogether 102 farmers participated in four two-day workshops held at each location: 31 in Kishushe, 24 in Maktau, 27 in Ngerenyi, and 20 in Chala. Gender and age balances were emphasized in the participant mobilization, with preconditions of farming activities, local residence, and not belonging to the same household. Of the participants, 50 were women and 47 were men, while five did not identify their gender. Participants came from 36 villages in the four sites, and the villages were distributed roughly within each site's agro-ecological zone. The average age of participants identifying their age (n = 72) was 45 years; the youngest was 20 years and the oldest 75 years. The genderdisaggregated workshop exercises included cropping calendars, organization mapping, climate calendars, historical calendars, and the use/ awareness of CSA technologies in CSA ranking. Mixed-gender exercises included crop listing, participatory resource mapping, and final ranking of the CSA practices. In participatory resource mapping, participants drafted a map of their locality including settlements, roads, rivers, boundaries, and the locations of various services. By drafting climate calendars, the farmers identified rainfall patterns and typical weather events and conditions at different times of the year. Within the historical calendar exercise, farmers rated the quality and quantity of conditions regarding climatic, resource, and agricultural activities at various temporal scales. The last exercise was a pairwise ranking of CSA practices, in which the awareness and use of various practices were examined according to gender.
Informed consent was acquired from all participants, and genderdisaggregated workshop activities were facilitated for women groups by women facilitators and for men groups by men facilitators. Facilitators were prepared for situations with more dominant participants and how to de-escalate these situations. All two-day workshop sessions ended with a wrap-up, during which the results were presented and the exercises discussed together with the participants.
Key comments made by farmers in the workshops were transcribed and quantified, and organization and local resource maps alongside tables, such as rainfall distribution patterns, filled by the workshop groups, were digitized. The transcribed data were analysed and compared with simple queries to the etic coded key informant data to form coherent thematical entities that express the farmers' personal  Jaetzold et al. (1983Jaetzold et al. ( , 2012, and map data © OpenStreetMap contributors (2019).

Table 1
Agro-ecological conditions in the case study sites (Jaetzold et al., 2012;Kenya National Bureau of Statistics, 2019b Very short to short (55-74 days) * Growing period of more than 45 days possible with run-off catching techniques.
views on climate change challenges and adaptation methods in the county. Furthermore, the analyses compared trends across the study sites to recognize spatial variability in agriculture and responses to climate change.

Transect walks, household surveys, and key informant interviews
Transect walks were carried out to better understand the environmental conditions and challenges faced by the farmers and to verify the data collected from the workshops, household survey, and key informant interviews (Fig. 3). The transect walks and household surveys took place within a 5-km radius from the workshop locations. The walks were 2-4 km long, stopping approximately every 100 m to record a Global Positioning System (GPS) coordinate point, and to photograph and examine the environmental and socio-economic conditions. All transects were conducted together with a local expert who could identify farming practices, built structures, soil types, flora and fauna in the area.
Household surveys were conducted in the workshop localities. Participants were randomly sampled using geospatial building data traced from satellite imagery. Altogether 65 interviews were conducted in the household surveys: 12 in Kishushe, 21 in Maktau, 21 in Ngerenyi, and 11 in Chala. The household survey data comprised farming activities and the land tenure system. We used the quantified household survey data to build an economic gross margin analysis of the agricultural activities in the four case study sites taking into consideration the production costs for each product (e.g. input costs and hired labour) and their selling price and quantities. The survey data collected were compared with the workshop data to find similarities and irregularities in the responses regarding environmental challenges and social constraints to CSA practices and technologies. In addition, awareness and use of various CSA technologies and practices were quantified and compared.
Key informants were chosen to represent local agricultural stakeholders in the administrative and civil society sectors. Sixteen key informants were interviewed during 10 interview sessions. These included individuals from e.g. non-governmental organizations (NGOs) that worked with farmers, and government and county officers from the departments of the environment, livestock, agriculture, drought management, gender, social development, and co-operatives. The recorded audio data were transcribed and then analysed with ATLAS.ti 9 -qualitative analysis software to find regularities, co-occurrence, and interlinkages between hierarchical codes (N = 166). The coding framework contained categories and subcategories based on predefined and emerging codes. Predefined codes were categorised into thematical semi-structured interview questions, respondent background codes, location codes, crop variety and livestock codes, CSA practices and technologies, land tenure related codes, and other descriptive and meta codes. Emerging codes included relevant topics that were frequently mentioned in the interviews such as fall armyworm. Combined codes were used in queries to record patterns in the responses.

Perceived changes in the climate and environment
The household interviews showed that 92% of respondents had observed a change in weather conditions during the past 20-year period, with increased variability, temperatures, and either reduced or unreliable rainfall across the sites. The perceived trends of increasing temperature and shifting rainfall patterns are in line with the temperature and rainfall distribution analyses of historical climate data collected from local weather stations in the county but slightly contradict the actual precipitation quantity, which appears to have only a minor decreasing trend and high spatial variability in the county (MoALF, 2016;Ogallo et al., 2019). The largest challenges regarding climate variability were identified as water scarcity, prolonged droughts, and unpredictable or reduced rainfall. Of the survey respondents, 60% had experienced either low or non-existent yields due to these challenges. Only 20% of respondent households had been self-sufficient in terms of food they would normally produce throughout the previous year (two seasons), and 57% of respondents identified no significant income activities outside of farming.
Forty-five per cent of respondents identified pests as their most significant natural disaster challenge, with fall armyworm (Spodoptera frugiperda) infestations being the most cited. Other key challenges were human-wildlife conflicts, mentioned by 31% and mostly caused by elephants (Loxodonta africana) and monkeys (such as Cercopithecus mitis, Chlorocebus pygerythrus), and strong winds causing erosion and structural breakages, mentioned by 14%. These challenges were also noted in the transect walks (see Fig. 3). Key informant interviews noted human-wildlife conflicts to be rampant due to the human population increase, which has forced farmers to encroach closer to conservation areas and sanctuaries. Floods and droughts were also identified as key challenges.
Due to these challenges, farmers have had to source water from outside their localities and buy food they would normally produce, which has required seeking alternative sources of income such as casual labour, beekeeping and increasing livestock numbers. Farm-level responses to the above-mentioned challenges included finding water sources for irrigation, planting smaller areas to prevent risks, changing seed varieties and livestock breeds to more drought-tolerant ones, and increasing pesticide use. Especially in Maktau, water shortage was perceived as moderate and constant, which has forced farmers to cultivate more drought-tolerant crops and to shift pastures more frequently. In Chala, water shortages have led to farmers implementing shallow wells to irrigate their crops. A reduction in soil fertility was experienced elsewhere, except Maktau, forcing farmers to apply fertilizers. Minimum or zero tillage has increased in Kishushe due to soil quality depletion. However, newly adopted soil conservation and crop rotation practices were noted to have improved soil quality in Ngerenyi. Soil erosion was considered an issue in the highlands, causing landslides and eroding roads, which has restricted market accessibility.

Impacts of extreme weather conditions
The climate calendars indicated that women and children are disproportionately affected during perceived extremely dry years, as they typically travel long distances to fetch water. Crops that are not normally under irrigation require watering. When irrigation by any means is impossible, farmers must reduce the land size under cultivation and change their farming practices or abandon them. Nevertheless, crops may end up drying in the fields, creating severe food shortages. Human and livestock diseases and pests increase alongside human--wildlife conflicts. Livestock deaths occur, and animals are badly affected by dust. Cattle are sold at throwaway prices to buy food. While food is bought from outside the locality, communities must also depend on government support for food relief and charcoal burning for income increases. Heavy winds also occur during some years, ripping off roofs, destroying storage facilities, and felling trees.
During wet years, the harvest may be bountiful, but significant yield losses are stated to occur, as either crops cannot be collected or cereals become impossible to dry. Crops can be attacked by pests, such the fall armyworm, and challenges in post-harvesting create a risk for aflatoxin poisoning. Blight disease can affect tomatoes. Farmers therefore tend to use crude methods, such as excessive pesticide, herbicide, and fungicide application, to control crop pests and plant diseases, especially when confronted with unfamiliar ones. Although livestock benefits from abundant pastures, cattle may contract diseases sensitive to moisture, such as hoof diseases. Nevertheless, fruit trees produce well during extremely wet years, and large production may be gained from livestock. Floods and gully erosion may wash away houses, break structures, and destroy farms in the lowlands, and, according to women, repairing and cleaning the damages and finding firewood increases their workload significantly. In the following season, planting crops becomes more arduous due to overgrown weeds.

Shifting agricultural practices and technology transfer
Regarding the second research question on CSA constraints, government officials in Mwatate and Wundanyi sub-counties emphasized that lack of knowledge is the main reason for the low utilization level of several CSA technologies such as intercropping, green manure, cover crops, and composting. By contrast, Taveta was said to be technologically more advanced, and challenges there were more related to a lack of capital and resources. An NGO representative noted that while farmers are aware of many CSA technologies, they are confused with which of them to invest in, as e.g. conservation agriculture equipment are costly.
The extension services provided by the county were perceived as inadequate by some workshop participants in Maktau, even though the respondents in all four locations had identified receiving extension services. Government agencies and parastatals, along with informal, local, national, and international NGOs, have implemented agricultural development interventions in the county. For example, World Vision, Plan International, and Action Aid have worked with farmers on issues such as water conservation techniques and women empowerement. The Kenya Climate Smart Agriculture Project conducted research and farmer workshops simultaneously with our fieldwork in the area.
According to county officials, the past decades have seen a dramatic shift from manual land preparation to machinery use, e.g. communally rented tractors. Land in the highlands suitable for agricultural expansion has been exhausted and farmers are cultivating very small parcels of land, but cultivated land is increasing in the lowlands, as also evidenced by Pellikka et al. (2018). This has led to the abandonment of fallowing and to increased fertilizer use in the highlands. Land scarcity in the hills has also encouraged zero-grazing of livestock, and stricter control has been implemented to prevent grazing in national parks and conservation areas. Taveta has experienced a shift from cultivating maize to highvalue horticultural crops, such as chillies and bananas, while horticulture has decreased in the hills.

Gendered roles in agriculture
The gendered roles in agricultural practices appear to depend on the perceived physicality of the task that derive from cultural assumptions concerning gender differences (Little, 2002), but even more so on whether the produce is meant for home consumption or sale. Table 2 describes the perceived five most important crops and their gendered associations. Produce that requires sourcing from the markets or creates sales revenue is typically associated with men, while produce for home consumption is associated with women. This was expressed more clearly in the key informant interviews. A practice changing from subsistence to commercial or vice versa may also have implications for the agricultural roles of families.
The key informant interviews further disclosed that decision-making power in the household over agricultural issues, such as the type of crops to be grown, livestock to be reared, and where to sell the products and at what price, was dominated by the men. The following quotes are by county-level, government-level, and parastatal officers working directly with women farmers on issues such as gender, agriculture, and development. 2 A key informant (#1, a woman) expressed the following concerning intrahousehold decision-making: 'The few times I've gone to the field and posed this question, the women will say they have to consult their husbands before they engage in anything. So, decisions are made, based on the agreement that they have with the husband. --I think the issue [,] of course [,] is when you want to engage in 2 To increase readability, we have slightly changed the punctuation and omitted or added individual words in the quotes with respect to the original phrases, but without changing the original meaning of the sentences.

any agricultural activity[,] [a] woman has to first consult the husband if she can go ahead and do the agriculture activity. --I think the biggest problem the woman faces is 1) the right to own land, 2) she faces a lot of difficulties even in making personal choices when it comes to what do I want for myself, because she is tied to her family and to the society.'
Another respondent (#2, a woman) stated:

'The main decision maker is the man. There could be a few houses where you have consultations, but I [ . . . ] believe that in most households the man makes the decision on [ . . . ] what is to be farmed, the size of the farm to be used, the type of inputs to be used. Because you see in most of the houses, it's the man who has the money.'
However, non-farm employment outside the community is common for men in TTC, and practical agricultural decisions are often made by women in the two-headed households. Agricultural information provided by extension services was identified to reach both men and women equally in the county. Furthermore, women were said to participate more in communal meetings regarding agriculture. Additionally, in the organizational mapping exercises in all four locations, women identified more agriculture-related organizations, social groups, and especially women-targeted small-loan providers than the men did.
Despite women playing a major role in agricultural production, the key informant interviews revealed that women rarely hold titles to their plots due to the traditional land tenure system that favours men in inheriting, selling, and bequeathing land, regardless of whether households are single-or two-headed households. This complicates investing in new technologies, as loans from major financial institutions typically necessitate land titles for approving credit. A key informant (#1, a woman) stated the following regarding the land tenure system: 'Men own land. And you find [in] our traditions and our cultures with the patriarchal society, it was very [pause], women just don't own land. If a woman will own land, [she] will own land for example [ . . .] [al] woman who stays at home, even the cash that she has started saving if she's going to buy a land, you know she has to make the decision with the husband. But of course who is more probable to take the title deeds? It's the man.' Another key informant (#2, woman) stated the following on acquiring loans: '

They [women] are not major decision makers on how the land will be used. They don't make that decision. They cannot even use the title deed to access credit. They cannot use the title deed as security to get loans if the title deed is in the name of the husband, they can't.'
A third key informant (#3, man) expressed the following statement on the challenges that women face in agriculture: '

Socio-economic constraints in agriculture
The key informant interviews identified marketing structures to be challenging for farmers, as prices fluctuate unexpectedly. Farmers have been discouraged from forming strong co-operatives to gain agglomeration benefits because of competition by large-scale farmers selling their products, brokers exploiting the farmers, and poor management that has been claimed to have occurred in previous years, for example in coffee and aloe production. Value addition was noted to be minimal and mainly focused in Chala, with its horticultural products and fruits. Food distribution centres operated by e.g. brokers and co-operatives were also considered ineffective. The key hindering factors for value addition include lack of technical know-how, lack of a suitable market, and insufficient capital to invest in the required infrastructure. Officials stated that despite general improvements of the road networks in Taveta and Taita Hills, the poor feeder road network hindered transporting products to the markets.
Attracting youth to agriculture was identified as a main challenge, as farmers are ageing and the land tenure system does not support intergenerational transitions well. In the lowlands, some farmers are squatting on communal land, whereas title deeds may be registered to earlier generations in the hills. Acquiring title deeds was described to be expensive outside government-led land adjudications.
The economic analysis shows positive net returns from crop production in Kishushe and Chala and negative returns in Maktau and Ngerenyi (Table 3). Livestock production was excluded from the analysis based on the high diversity of livestock production systems and Table 2 Lists of the five most highly ranked crops and livestock, and their gendered association in the crop-listing exercise of the workshops. F = women, M = men, C = consumption, S = sale, D = dairy, and B = beast of burden. short-term viewing cycle of one year. This underestimates the variable production costs. The higher crop returns in Chala are likely attributed to more commercialized farming techniques, land availability, and farmer focus on high-value horticultural crops, such as bananas, compared with subsistence crops. Renting land from irrigation schemes is reflected in the variable costs in Chala. In Ngerenyi, the lower crop income may be attributed to limited land holdings. Farm inputs were mentioned to be readily available from agrovets, but prices were considered high, and seed quality was also identified as a problem.

CSA applicability
The pairwise ranking of CSA farm management practices showed similarities between the sites, while each site prioritized differently their most important practice. The use of zaï pits was ranked as the most important practice in the dry agro-ecological zone of Kishushe, while Maktau, suffering from a constant water shortage, emphasized the importance of rainwater harvesting. Farmyard manure application was prioritized in Ngerenyi, where soils were eroded. The more commercial agriculture in Chala focused on irrigation (Table 4).
The lack of awareness stated in the key informant interviews contradicts our findings from the household survey and the workshop data. A further examination of the household interviews regarding awareness and use of CSA technologies related to the third research question on spatial heterogeneity reveals that farmers are aware of most available technologies, but their utilization typically follows the agro-ecological division between the lowlands and highlands (Fig. 4). Major knowledge gaps may be seen in the awareness of zero tillage, green manure, and organic agriculture. Notably, agroforestry and mulching are practised more commonly in Kishushe than in Maktau, despite both representing relatively similar agro-ecological zones. Maktau, on the other hand, is the only site where organic agriculture is practised more commonly.
The household data concerning technological awareness and where agricultural information is acquired from (44% of women and 60% of men mentioned extension services) suggest, alongside the statements made in the key informant interviews, that women and men have relatively equal access to agricultural information regarding CSA technologies. However, men are more aware of practices such as composting, green manure use, mulching, and zero tillage (Fig. 5). Fallowing, farmyard manure, intercropping, and organic agriculture were practised more frequently in households where women farmers were interviewed.
The constraints of adopting CSA technologies expressed in the pairwise ranking indicate that the challenges differ greatly depending on the agro-ecological zone (Table 5). Moreover, the ranking exercise demonstrates that farmers are aware of the common CSA technologies and they have opinions of their applicability in their local context. Lack of knowledge or awareness of CSA technologies was most common in Maktau, where workshop participants also indicated inadequate extension services. In general, the main hindering factors across the scale of CSA technologies are characterized by a lack of resources such as water, land, labour, time, money, knowledge, or training. Other constraints are mainly related to human-wildlife conflicts, pests, and diseases. Societal pressure affecting the adoption of terracing, farmyard manure, and contour ploughing may exist, as these technologies have been heavily promoted by the government and even by the colonial administration before the independency of Kenya (Thurston, 1987). According to key informants, farmers have also been encouraged by the Ministry of Agriculture, Livestock and Fisheries of Kenya to engage in agroforestry and intercropping and to avoid over-grazing.
Agricultural water management methods in the lowlands are emphasized by rainwater harvesting, whereas they focus more on water retention, irrigation, and erosion avoidance in the highlands. Irrigation is commonly used in Chala, while it was severely hampered by water scarcity in other locations. Irrigation-related conflicts between farmers are also common in Chala and in the highlands. Zaï pits have been introduced to the lowland people, but a lack of continued awareness creation and training has led to less people utilizing them after the first stage of adoption. Land tenure restrictions limit practice and technology adoption, such as fallowing and crop rotation, in the Taita Hills.
Several CSA practices and technologies, namely composting, mulching, building ridges or bunds, and terracing, are also considered laborious. Composting and green manure use are restricted by plant material scarcity. According to participants, agroforestry practices are hindered by wildlife attacks and a general lack of knowledge. Practising organic agriculture was disrupted by pest attacks and soil infertility in the highlands, forcing farmers to use chemicals. Inorganic fertilizers are readily available, and they are being widely used in the highlands and in Chala, where workshop participants had also indicated a decline in soil fertility. In contrast, soil is considered fertile in Kishushe and Maktau, which eliminates the need for fertilizers. Green manure and zero tillage had the lowest awareness levels and were considered risky practices amid uncertain climatic conditions.

Discussion
We examined the constraints faced by smallholder farmers in TTC that prevent them from effectively adapting to climate change and improving their livelihoods amid changing environmental conditions. The results provide similar evidence as the conclusions obtained with CSA-RAs conducted in Tanzania and Uganda by Mwongera et al. (2017): great heterogeneity exists in agricultural conditions on a regional level, guiding the adoption of CSA technologies. Further, in line with the research by Belay et al. (2017) in Ethiopia and Elum et al. (2017) in South Africa, the majority of farmers in TTC have acknowledged changes in the local climate and are reacting autonomously to these changes. The model by Boitt et al. (2014) revealed that climate change will likely shift agro-ecological zones in the Taita Hills, forcing farmers to change cultivation patterns, and indeed, farmers in Taita Taveta are both knowledgeable and utilize the most common climate-smart farm and crop management practices. They use improved seed and livestock Table 3 Gross margin analysis in Euros (EUR), based on the household survey in 2019 in Kishushe (n = 12), Maktau (n = 21), Ngereyni (n = 21), and Chala (n = 11). Crop income was calculated for all crops grown and 1 EUR was equivalent to an average of 113.64 Kenyan Shillings in 2019.  varieties and constantly seek new information from agricultural extension officers, NGOs, and other farmers. Farmers have changed their ways of agricultural production to better meet market requirements. Key informant interviews in Taita emphasized that low adoption rates of CSA technologies are due to farmers' lack of knowledge. In contrast, during both the workshops and household surveys, the farmers themselves acknowledged strong awareness of these technologies across the agro-ecological zones. The adoption gap for CSA technologies and practices in TTC can be seen in the difference between awareness and adoption rates. Despite farmers being aware of many CSA practices and technologies, a smaller proportion implement these technologies. Moreover, the farmers in TTC perceived inequality in the extension services provided based on geographical location but not according to gender. Kpadonou et al. (2017) and Holden et al. (2018) Wekesa et al. (2018), which indicates that even CSA practices that require low capital investments may have low adoption rates, possibly due to non-monetary resource constraints.
This study further shows that unfavourable market conditions and limited value addition, especially outside Taveta, are also constraints to farming in the area. This is in accord with Kirui et al. (2017), who noted that many farmers in rural Kenya are constrained by poor marketing facilities characterized by high transportation costs, poor storage, and high wastage. The results of our gross margin analysis of crops further indicate that agriculture in Maktau and Ngerenyi must be sustained more by keeping livestock and other forms of income, such as casual labour, as the gross margins are negative. This could mean that there are less resources to invest in agriculture and to respond quickly to extreme weather events and pest infestations. The high prevalence of pests, such as fall armyworm, in the data is likely linked to the first major fall armyworm outbreak in Kenya, which occurred in 2017, causing annual losses of over 3 billion KSh (ca. 2,640,000€) to maize farmers (De Groote et al., 2020). This may also explain the increased adoption of pesticides and changes in crop varieties. The rationale behind the shifts in land and farm management practices is thus highly complex when interlinked with such phenomena as climate change.
Further, women's marginalization in TTC from decision-making processes, as described in the household roles, lack of property rights, and limited access to credit and other production inputs, hinder them from investing in CSA technologies and practices to achieve their optimal production potential. This hindrance may, for example, be reflected in agroforestry and cover cropping that require monetary inputs, in building ridges and terraces that increase on-farm labour and crop rotation, and in fallowing and intercropping that require the acquisition of larger land parcels. According to FAO and CARE (2019), the agricultural sustainability can best be enhanced by ensuring that men and women have equal access to productive resources, CSA, and laboursaving technologies, services, and institutions within their localities. Tsige et al. (2020) also suggest that while CSA presents opportunities for economic development and food security, technological success has been continuously hindered by gender constraints. This emphasizes the need for institutions working on climate change to partner with local women's organizations to adequately tap the potential of women as adapters of CSA technologies and practices. Tapping into this potential is important, especially when considering women's central role in agriculture and their high social capital within the organizational network. Gender transformative actions are also needed within the policy and legal framework to ensure better access to resources when promoting CSA practices and technologies.
Despite advances, TTC significantly lags behind the rest of Kenya when it comes to food security (Kenya National Bureau of Statistics, 2018). Nearly all respondents described growing irregularity in weather conditions and having faced challenges due to climate change. Their adaptation capacity seems to be hindered by a myriad of sources besides farm and crop management practices such as access to land and resources, land ownership issues, human-wildlife conflicts, pest infestations, and unpredictable market mechanisms. Dhanya and Ramachandran (2016) have also recognized that in South India, farmer perceptions are consistent with observed changing weather patterns, except for rainfall quantity and distribution, and that farmers are able to clearly express their specific adaptation needs. However, farmers have relatively low resources to invest in CSA innovations, and adopting multiple technologies may impose a considerable burden on an individual farmer (Makate et al., 2019). Moreover, CSA technologies have proven to be labour-, knowledge-and capital-intensive, and often adopting these technologies is highly dependent on land tenure-related issues (Kpadonou et al., 2017). The socio-economic perspectives in our results are in line with Hohenthal et al. (2017) in recognizing the complexity of local resource management and empowering the farmers from the narrative in which their lack of knowledge of farm and land management practices is the main constraint on technology adoption. This may further contribute to imagining alternative intervention strategies that are based on dialogue-building between state actors and farmers (Hohenthal et al., 2018).

Study limitations
Participatory methods, such as CSA-RA, can provide insight to local complexities that shape the agricultural production and livelihoods of a community or a district-level administrative unit. The benefit of these methods is that they provide a fair amount of flexibility. Nevertheless, generalizability and representativeness tend to suffer from typically smaller sample sizes and the subjectivity of perceived applicability and impacts of CSA practices. Sampling may exclude the most marginalized farmers, and research settings with focus groups, such as farmer workshops, may produce biased data if unequal power structures in knowledge production are not taken into consideration. The descriptiveness of the data allows for bringing forth the voices of farmers and other local stakeholders but simultaneously makes detecting significant drivers of agricultural knowledge and technology uptake more difficult and unsure.

Conclusion
The diversity of agro-ecological zones and socio-economic settings in TTC creates a complex food production system that is significantly affected by climate change. Farmers have acknowledged a shift towards uncertainty in weather patterns and are attempting to find ways to secure their livelihoods amid these processes. Upscaling CSA technologies is challenging due to heterogeneity in both biophysical and socioeconomic conditions, but there is evidence that locally appropriate responses to climate change are utilized in agriculture. Moreover, development interventions have been conducted in the case study area, and farmers have found many of these interventions useful. Nevertheless, the civil society sector may not produce adequate solutions for climate change adaptation among the farmers, and government extension services should provide technical support suitable for the differing agroecological zones in the region.
Dissonance also occurs in the perceived awareness and utilization of CSA technologies between state actors and farmers, which requires dialogue-building across diverse agricultural settings to find the most effective focus for extension services and interventions. Our results also highlight the need to invest in rural public education to provide farmers with training in CSA practices and technologies after the local focus has been identified through dialogue. Further research is needed to examine the limitations that gender and the current land tenure system in TTC, with its land-use conflicts, create for the climate change adaptation efforts of smallholder farmers. The findings of our research are based on the contextual settings of TTC, but the results indicate that the adoption of CSA practices and utilization of technologies especially in sub-Saharan regions that are heavily based on subsistence agriculture with heterogenous agro-ecological zones, require localized solutions in policy formation and planning of both agricultural extension services and development interventions that consider the agency of the farmers.

Declaration of Competing Interest
None.
corresponding author is also grateful for advice from Dr. Matti Räsänen and Sheila Wachiye in the research work.