Adoption of Agroforestry in Northwest Viet Nam: What Roles Do Social and Cultural Norms Play?

: This article presents research about the inﬂuences of social and cultural norms on the adoption of agroforestry in the northwest mountainous region of Viet Nam. The farming systems practiced by various ethnic groups in the northwest mainly occur on sloping land, which extends over 70% of the land area in the region. Decades of intensive monoculture of annual crops has resulted in severe soil erosion, contributing to soil degradation and decline in crop yields. Integrating agroforestry practices on sloping land has the potential to halt and reverse soil degradation and improve local livelihoods, but its adoption is conditioned by the diverse social and cultural norms of different ethnic groups. This research applies knowledge-based system methods in order to understand local opportunities, preferences, and constraints inﬂuencing the adoption of agroforestry practices, using a purposive, gender-balanced sample of sixty farmers from six villages across three provinces in the northwest region comprising people from Kinh, Thai and H’mong ethnic groups. Our results show that although farmers from all groups are aware of the economic and ecological beneﬁts of trees for soil conservation in general, they have different perceptions about the beneﬁts of particular agroforestry practices. Behavioural norms controlling agroforestry adoption vary amongst ethnic groups, and farmers’ individual social and cultural preferences inﬂuence their aspirations and adoption decisions. We conclude that developing appropriate agricultural interventions in a culturally rich environment such as northwest Viet Nam requires understanding of the context-speciﬁc needs and interests of socially and culturally disaggregated populations. Policies supporting agroforestry are more likely to contribute to more sustainable livelihoods and ecological beneﬁts where they are tailored to the requirements of different ethnic groups. adoption study hypothesises that agroforestry adoption occur when farmers have both the capacity and willingness to adopt agroforestry. "vuon-ao-chuong"-VAC (garden-ﬁshpond-livestock), which literally is the combination of a vegetable garden, ﬁshpond, and livestock recommended by government extension programs. Kinh farmers believe they are expected to adopt the modiﬁed VAC practice to help reduce soil erosion and prevent ﬂash ﬂoods. All Kinh farmers concurred that they learnt fruit tree management techniques and bought seedlings from their hometown to grow in their new upland environment; 26% of Kinh correspondents indicated that intercropping was a traditional technique to address limitations of land holding size. Farmers from all ethnic groups also mentioned that they had been trained and encouraged by local authorities and extension workers to gradually shift from maize/rice cultivation to perennial tree-cropping on slopes. Farmers often choose the most degraded plot with low crop yield to grow trees intercropped in the ﬁrst few years before later removing annual crops when the tree canopy closes. Today, there are a number of programs for growing avocado, macadamia, son tra ( Docynia indica ), Shan tea ( Camellia sinesis ), and various citrus species that are expected to shift the norm away from annual crop cultivation to tree-based cultivation on sloping lands.


Introduction
Agroforestry is where trees interact with agriculture and farmers manage interactions with the aim of achieving net economic and ecological gains [1,2]. "Agroforestry", therefore, is a broad term used to describe a wide range of practices at a range of scales [3]. At its simplest, agroforestry is a set of practices that combine woody components and agricultural crops and/or livestock, but this often involves quite complex interactions between people and trees, requiring systems analysis to understand why people adopt particular practices [4]. Based on spatial and temporal arrangements, agroforestry practices are defined as either simultaneous or sequential and in relation to the juxtaposition of tree and crop components. In the simultaneous practices, all components are intercropped at the same time, while in sequential practices, components occupy land at different times [5]. Sequential practices have varying degrees of overlap between the crop and tree components. Concomitant systems overlap in the beginning (such as in taungya systems), and

Theoretical Framework
Participatory approaches developed in the 1990s were viewed as a paradigm shift in research and development, providing tools to capture the views of local people and moving away from top-down prescriptions [28]. Earlier participatory approaches were, however, often applied without critically thinking about the issues of social barriers to farmers' decision-making. Several theories have been applied to understand farmers' behaviours. For example, the "Diffusion of innovation" theory [29] looks into how and why a technology is adopted and spread. Value-belief-norm theory [30] provides an approach to analyse social supports for the environmental movement. The theory of planned behaviour [31] is a conceptual framework that has often been used to explore the social dimensions of technology acceptance and adoption [32][33][34][35] and forms the basis for the methods in use in this paper. The key aspects of this theory are briefly discussed here.

Social Norms
Social norms are the expectations from a community on individuals to perform a behaviour in a specific situation. Social norms interact with social relationships to influence a farmer's adoption decision through, for example, their neighbours' adoption patterns, social expectation, and pressure or social status that may result from engaging in an activity. Current recommendations suggest that analysis of these norms should be incorporated in best practice adoption studies [36]. Adjusting agricultural intervention to socio-cultural factors can improve how farmers value innovations and their motivation to adopt them [37]. This has led to a major paradigm shift to support local innovation where farmers adapt practices to suit their local circumstances rather than the widespread promotion of prefabricated technology packages [27].

Farmer Capacity to Adopt
A farmer's capacity to adopt agroforestry is defined by both biophysical and socioeconomic conditions. Biophysical conditions are derived from understanding of the ecological conditions of the farm, including characteristics such as soil type, slope [38], farm size [39], or the geographical context of the plots [40]. On the other hand, socio-economic factors include: market incentives, household preferences [38], economic benefits of land tenure and available time [41], the amount of social capital and human capital (including knowledge), and the influence of local and national policies [42].
Applying the theory of planned behaviour [31], willingness to adopt is influenced by personal beliefs and attitudes, social norms, and perceived behavioural controls. Farmers' attitudes encompass how personal beliefs act on behaviour, which can be negative, positive, or neutral. Farmers' attitude towards adoption is strongly correlated with farmers' perceived behavioural control (how they perceive the level of difficulty in adopting) and their self-belief in the capacity to adopt and then maintain the practice [43]. For example, farmers' attitudes towards agroforestry were strongly related to the level of access to information and extension support in Bangladesh [44].

Farmer Perceptions
Farmers' adoption decisions regarding agroforestry practices are influenced by the degree of difficulty associated with acquiring accurate information about the benefits of such innovations [32,42], alongside perceptions of risks and barriers associated with tree planting [45] or knowledge of agroforestry techniques. The selection of tree species for agroforestry adoption highly depends on farmers' attitude or knowledge concerning the impact of trees on food, soil, water and crops from their experiences and observation [46].

Perceived Behavioural Controls
Perceived behavioural control is the personal perception of the difficulty involved in performing a specific behaviour [31]. High financial cost and lack of knowledge on tree management techniques are commonly perceived challenges of planting trees on farms [47]. Limited skills and techniques were highlighted in studies in Pakistan [48], while the unavailability of capital and quality seed were identified in Rwanda [49,50].
Social and cultural values maintain a strong link with ethnicity because different ethnic groups have their own religions, beliefs, values, and resources which influence their attitudes, social norms, and behavioural controls related to agricultural innovation [51]. If communities are viewed as homogenous groups, the result may be that only the voices of a small number of powerful people are included, and thus the design of interventions may not be appropriate for those who are supposed to be empowered [52].
In this study, we explore how social factors influence pathways to adoption for agroforestry practices designed to stabilise soils and support local livelihoods for different ethnic groups in northwest Viet Nam. This includes: (i) identifying and understanding the social factors including personal attitude, subjective norms and behavioural controls towards farmers' willingness to adopt agroforestry (and whether this varies with ethnicity); (ii) recording preferences for different forms of agroforestry systems; and (iii) identifying potential constraints to the adoption of agroforestry from three main ethnic groups, the Kinh, Thai, and H'mong.
This study contributes to the existing global literature in understanding cultural and ethnicity aspects of local ecological knowledge on agroforestry and other land uses [53][54][55][56]. Scaling up agroforestry adoption need to be adapted to fine-scale variation in ecological and social contexts, including local needs [57]. Adoption is more likely to happen when farmers have knowledge, labour, and secure land tenure [58,59]. Social factors including farmers' preferences, attitude, cultural or social constraints, and local knowledge strongly influence farmers' adoption decisions [60]. These factors can be categorised as farmers' capacity to adopt and farmers' willingness to adopt [61].
Farmers' capacity and willingness are two key elements for agroforestry adoption (Figure 1), based on the theory of planned behaviour [31] and the practical evaluation of agroforestry adoption [61]. This study hypothesises that agroforestry adoption will occur when farmers have both the capacity and willingness to adopt agroforestry.

Perceived Behavioural Controls
Perceived behavioural control is the personal perception of the difficulty inv performing a specific behaviour [31]. High financial cost and lack of knowledge management techniques are commonly perceived challenges of planting trees o [47]. Limited skills and techniques were highlighted in studies in Pakistan [48], w unavailability of capital and quality seed were identified in Rwanda [49,50].
Social and cultural values maintain a strong link with ethnicity because diffe nic groups have their own religions, beliefs, values, and resources which influe attitudes, social norms, and behavioural controls related to agricultural innovatio communities are viewed as homogenous groups, the result may be that only the a small number of powerful people are included, and thus the design of interventi not be appropriate for those who are supposed to be empowered [52].
In this study, we explore how social factors influence pathways to adoptio roforestry practices designed to stabilise soils and support local livelihoods for ethnic groups in northwest Viet Nam. This includes: (i) identifying and understan social factors including personal attitude, subjective norms and behavioural con wards farmers' willingness to adopt agroforestry (and whether this varies with et (ii) recording preferences for different forms of agroforestry systems; and (iii) ide potential constraints to the adoption of agroforestry from three main ethnic gro Kinh, Thai, and H'mong.
This study contributes to the existing global literature in understanding cult ethnicity aspects of local ecological knowledge on agroforestry and other land u 56]. Scaling up agroforestry adoption need to be adapted to fine-scale variation in ical and social contexts, including local needs [57]. Adoption is more likely to when farmers have knowledge, labour, and secure land tenure [58,59]. Social fa cluding farmers' preferences, attitude, cultural or social constraints, and local kn strongly influence farmers' adoption decisions [60]. These factors can be catego farmers' capacity to adopt and farmers' willingness to adopt [61].
Farmers' capacity and willingness are two key elements for agroforestry a (Figure 1), based on the theory of planned behaviour [31] and the practical evalu agroforestry adoption [61]. This study hypothesises that agroforestry adoption w when farmers have both the capacity and willingness to adopt agroforestry.

Data Acquisition Methods
Data used in this study were collected from two surveys in 2016 and in 2017 in the northwest region of Viet Nam. Households covered in the surveys were in the sites of the Agroforestry for Livelihoods of Smallholder Famers in Northwest Viet Nam (AFLi) project, which is representative of the ecological landscape and social-cultural dynamic of the northwest region.
The surveys are described briefly below. The first survey explored farmers' agroforestry adoption preferences. This involved a purposive sampling of current adopters and non-adopters of agroforestry. The first group was current adopters, consisting of two sub-groups: (1) farmers who worked with the AFLi project, called "project adopters" (n = 166); and (2) farmers who adopted agroforestry but were not involved in the AFLi project, called "spontaneous adopters" (n = 7). The second group consisted of farmers who had not adopted agroforestry on their farms, classified as non-adopters (n = 56). Questions in the first survey were designed to understand the capacity of farmers to adopt agroforestry, and their perceptions of the degree to which their biophysical context (elevation, cultivation, traditional practice) and social-economic conditions (finance, labour, knowledge) affected their capacity to adopt agroforestry systems (for non-adopters) or expand these systems (for adopters).

Survey 2 (2017): In-Depth Household Survey
The second survey was an in-depth household survey which focused on acquiring farmers' local ecological knowledge regarding tree planting following a knowledge-based systems approach [62,63]. Indigenous knowledge is culturally specific [64]; therefore, this survey looked at how local knowledge was shaped by attitudes and perceptions, as well as behavioural controls towards agroforestry adoption and hence farmers' preferences with regard to potential agroforestry options and the degree to which these were influenced by their ethnicity. This survey used key informant interviews of six commune representatives, combined with farmer focus group discussions and semi-structured interviews. One focus group discussion was conducted in each village, with the same farmers participating in the semi-structured interviews.
Key informant interviews were designed to understand the overall context of the communes, including the distribution of ethnic groups across elevations, socioeconomic contexts of three ethnic groups, supporting policies of tree planting, and agroforestry development.
Farmer focus group discussions aimed to understand village history, culture, tradition, cultivation practices, agroforestry opportunities, and constraints using village map sketching, historic mapping, faming calendars, and strength-opportunity-weakness-threats analysis (SWOT).
Finally, semi-structured interviews with fifty-eight farmers (50% female) were used to explore their knowledge about agroforestry management and obtain a deeper understanding about agroforestry management, the values of trees, social norms related to tree planting, farmers' attitude on the benefits of agroforestry, and preferences relating to agroforestry adoption. The interviews were conducted together with farm visits. Farmers were again purposively selected for these interviews. All interviewees were agroforestry adopters who were currently not involved in the AFLi project (to avoid influence from project training). We consulted six village leaders to choose farmers implementing agroforestry practices by themselves. The total number of farmers interviewed was 58, with 10 farmers invited per village representing the minimum number of agroforestry adopters per village across the study sites. This was gender-balanced; thus, equal numbers of men and women were interviewed. Two villages were selected for each ethnic group (Kinh, Thai and H'mong), amounting to six villages in total ( Figure 2).

Study Site
The study was conducted in the AFLi project sites in Yen Bai, Son La and Dien Bien of the northwest region of Viet Nam. The northwest of Viet Nam (21-23° N and 103-105° E) is one of eight eco-regions in Viet Nam. It is the most mountainous, remote, and poorest area of Viet Nam. Forest cover accounts for about half of the area, although the region is prone to severe deforestation, land degradation, and soil erosion. Extreme weather such as landslides, droughts, frost, and hailstorms significantly affect agricultural production and the economic development of the region. Key agricultural products of three provinces include annual crops such as rice (

Study Site
The study was conducted in the AFLi project sites in Yen Bai, Son La and Dien Bien of the northwest region of Viet Nam. The northwest of Viet Nam (21-23 • N and 103-105 • E) is one of eight eco-regions in Viet Nam. It is the most mountainous, remote, and poorest area of Viet Nam. Forest cover accounts for about half of the area, although the region is prone to severe deforestation, land degradation, and soil erosion. Extreme weather such as landslides, droughts, frost, and hailstorms significantly affect agricultural production and the economic development of the region. Key agricultural products of three provinces include annual crops such as rice (

Analytical Method
In this paper, qualitative data analysis was used as the key analytical method to provide insight about farmers' perceptions about agroforestry adoption and how this affects their decisions, through looking at the reasons and supporting farmers' explanations of their adoption decisions [68].
Qualitative data analysis has been widely used in the literature regarding planned behaviour [31], which is a cognitive process involving subjective norms, perceived behavioural control, and attitudes, as well as in the acquisition and analysis of local agroecological knowledge [62,63].
Textual analysis of interview transcripts was the main tool used in the paper to analyse and describe the actors involved, their subjective norms, perceived behavioural control, and attitude. In addition, illustrations are provided through the use of matrix, chart, and figure formats for both descriptive and explanatory purposes.

Farmers' Perceptions of Biophysical Conditions Required for Agroforestry Expansion
Biophysical attributes include factors defining the suitability of an agroforestry practice such as elevation, rainfall, temperature, slope, and soil type. Elevation is an important biophysical condition linked to ethnic groups in northwest Viet Nam, with the H'mong group inhabiting areas over 800 m above sea level (asl), the Thai at 400-800 masl, and the Kinh below 400 masl. Farming practices and choice of tree-crops differ by elevation and ethnic group (Table 1).

Socio-Economic Conditions
Socio-economic attributes include financial capacity, size of landholding, distance of farm to market, education, and knowledge, awareness, and proficiency with techniques ( Table 2). The Kinh are lowland migrants that settled in the lower part of the region under the post-war migration policy of the Vietnamese government. The Kinh have advantages in terms of language, enabling easy access to education, technology, market information, and other productive resources. The Thai live just above the Kinh communities. They are also able to speak the Vietnamese language and have reasonable access to information and technology. In contrast, the H'mong live in high-elevation areas that are often far from the main roads and local markets. The H'mong have sparse access to external information and government support because of their social distance from other communities in Viet Nam, which is exacerbated by language barriers. Today, many H'mong farmers are still practicing shifting cultivation with short fallow periods, while some, possessing less land, have already shifted to sedentary farming. Poverty is manifest in the economic status and educational level of households, which influence the way farmers learn new techniques. Land holding influences the availability of land to be used under agroforestry. The Kinh people have long been using agroforestry as a traditional practice, while the H'mong people rarely did so in the past. Although farmers from all groups were aware of the benefits of using trees in soil conservation, they had different perceptions on the benefits of specific agroforestry practices, which was likely to influence the types of agroforestry adopted (see Figure 3). All groups stated that agroforestry practices had some provisioning function relating to income generation but had differing needs in relation to regulating functions. The H'mong group were interested in increased land, labour and fertiliser use efficiency, while the Thai highlighted soil erosion reduction, and the Kinh were motivated by the idea of soil fertility improvement. This study suggests that farmers' specific social circumstances influence their aspirations and constraints related to adopting agroforestry interventions. When it comes to tree planning, H'mong farmers have myriad considerations even with increased demand for fuelwood and timber for house construction-first is to save time walking from and to their farms and homes; second is their limited family labour; and third is the high-cost of fertilisers. Meanwhile, Thai farmers think that timber and firewood, food for humans and animals (buffalo and cow), and soil erosion mitigation are important benefits that can be derived from agroforestry. Most Thai households have confined animals, so fodder grasses in agroforestry practices are useful to augment their cut-and-carry animal feeding strategies. However, reluctance to adopt agroforestry also grew from concerns that trees grow slowly, affecting cash flow. Finally, apart from income, Kinh farmers think that agroforestry is an efficient strategy to address having a limited size of land holding.
Having more access to technical information from local extension workers, the Kinh are more willing to intercrop various tree species, especially in home-gardens. In Tan Que district, Son La province, many Kinh farmers are already intercropping peanuts and beans with perennials to improve soil fertility and optimise land productivity.
(2) Subjective Norms Influencing Agroforestry Adoption Subjective norms constructed by the beliefs of ethnic groups about others' expectations toward forest protection, tree-growing and agroforestry were revealed in the in-depth interviews and focus group discussions. Farmers hold the same subjective norms, i.e., expectations of their behaviours regarding forest protection. Government reforestation programs supported the establishment of new forest plantations, and natural forest regeneration and protection [69]. In these programs, production forest lands were allocated for farmers to grow timber trees such as Acacia spp., Melia azedarach, Manglietia conifera, Pinus spp., and son tra (Docynia indica). In turn, households and communities were expected to protect the forest. In Thai villages, however, cultural norms exist wherein community forests are considered "ghost forests" or burial grounds that families come to visit every year. The Thai also believe they have to protect the old wild trees in their village because they are holy and revered for use in traditional ceremonies. When the Kinh migrated to the uplands, they were expected to retain forest patches on hill tops while annual crops, fruit trees or cash crops could be planted in the mid-lower portions of the hill, resulting in a forest-garden-fishpond-livestock system. This comes from a farming design called "vuon-ao-chuong"-VAC (garden-fishpond-livestock), which literally is the combination of a vegetable garden, fishpond, and livestock recommended by government extension programs. Kinh farmers believe they are expected to adopt the modified VAC practice to help reduce soil erosion and prevent flash floods. All Kinh farmers concurred that they learnt fruit tree management techniques and bought seedlings from their hometown to grow in their new upland environment; 26% of Kinh correspondents indicated that intercropping was a traditional technique to address limitations of land holding size. Farmers from all ethnic groups also mentioned that they had been trained and encouraged by local authorities and extension workers to gradually shift from maize/rice cultivation to perennial tree-cropping on slopes. Farmers often choose the most degraded plot with low crop yield to grow trees intercropped in the first few years before later removing annual crops when the tree canopy closes. Today, there are a number of programs for growing avocado, macadamia, son tra (Docynia indica), Shan tea (Camellia sinesis), and various citrus species that are expected to shift the norm away from annual crop cultivation to tree-based cultivation on sloping lands. sts 2021, 12, x FOR PEER REVIEW 9 o called "vuon-ao-chuong"-VAC (garden-fishpond-livestock), which literally is the com nation of a vegetable garden, fishpond, and livestock recommended by government tension programs. Kinh farmers believe they are expected to adopt the modified VA practice to help reduce soil erosion and prevent flash floods. All Kinh farmers concurr that they learnt fruit tree management techniques and bought seedlings from th hometown to grow in their new upland environment; 26% of Kinh correspondents in cated that intercropping was a traditional technique to address limitations of land holdi size. Farmers from all ethnic groups also mentioned that they had been trained and e couraged by local authorities and extension workers to gradually shift from maize/r cultivation to perennial tree-cropping on slopes. Farmers often choose the most degrad plot with low crop yield to grow trees intercropped in the first few years before later moving annual crops when the tree canopy closes. Today, there are a number of progra for growing avocado, macadamia, son tra (Docynia indica), Shan tea (Camellia sinesis), a various citrus species that are expected to shift the norm away from annual crop cultiv tion to tree-based cultivation on sloping lands. (

3) Behavioural Controls Affecting Agroforestry Adoption
Behavioural controls or challenges to adopting agroforestry varied amongst the th ethnic groups; however, lack of land, labour, and financing are common problems. M farmers also lacked quality planting materials and market links. Land as a limitation, ho ever, manifests in different ways-the Kinh have smaller farmlands and therefore ha nothing to spare for agroforestry. In contrast to the Kinh, the Thai and H'mong have larg landholdings, but prefer to establish agroforestry on plots near their homestead and fertile soils. In addition, H'mong farmers had concerns about the technical manageme of agroforestry because of their low access to technical information exacerbated by th language barrier with Vietnamese-speaking extension workers. The study also found ge  (

3) Behavioural Controls Affecting Agroforestry Adoption
Behavioural controls or challenges to adopting agroforestry varied amongst the three ethnic groups; however, lack of land, labour, and financing are common problems. Most farmers also lacked quality planting materials and market links. Land as a limitation, however, manifests in different ways-the Kinh have smaller farmlands and therefore have nothing to spare for agroforestry. In contrast to the Kinh, the Thai and H'mong have larger landholdings, but prefer to establish agroforestry on plots near their homestead and on fertile soils. In addition, H'mong farmers had concerns about the technical management of agroforestry because of their low access to technical information exacerbated by their lan-guage barrier with Vietnamese-speaking extension workers. The study also found genderspecific behavioural controls. Most of the concerns about marketing fruit amongst the Thai were from female respondents, while only H'mong men had concerns about market. Both male and female farmers from all ethnic groups were concerned about the technical aspects of agroforestry and expressed a strong desire to learn management techniques.

Typologies of Agroforestry Adopters and Non-Adopters
Farmer typologies were derived based on criteria identified for the capacity and willingness to adopt agroforestry practices. (Table 3). We examined the capacity and willingness of non-adopters, wherein more than half (53%), and mostly Thai and Kinh were Type 4 farmers who were unwilling to adopt or expand agroforestry due to limited capacity in land, labour, financing and technical management (Figure 4), and also because of a lack of evidence on agroforestry benefits. Type 3, comprising 14% of non-adopters, mostly belonged to the H'mong ethnic group who expressed a lack of know-how on agroforestry.
Thirty percent of non-adopters had resources and were willing to adopt agroforestry on their farms. Most of these were H'mong, because they saw the benefits of agroforestry from project adopters. One of the techniques they wanted to adopt was planting fruit trees together with grass so that they had more food for cattle in the winter. Cows and buffalos play an important role in H'mong livelihoods because they help farmers in land preparation and the transport of materials between homes and farms. Two H'mong farmers did not have enough land but they were interested in agroforestry and willing to adopt it, whereas none of the Kinh or Thai farmers who lacked land wanted to adopt agroforestry.
The majority of the non-adopters (53%) did not want to adopt agroforestry because of a perceived lack of capacity. Most of these were Thai and Kinh farmers. Only a few H'mong farmers were in this group, primarily because they lacked cultivatable land (which, in this context, meant land that was fertile and less than 10 km from their home). For the Kinh and Thai farmers, they stated that they did not have spare land for agroforestry.
Of those that were unwilling but had capacity (Type 3: 14%), a large proportion were H'mong; the reasons given were a lack of knowledge about agroforestry. For the Kinh and Thai, they had not yet seen the benefits, therefore they were not willing to adopt agroforestry at present.
In Type 4, the number of non-adopters was much higher compared to adopters, clearly because they did not receive direct support from the AFLi project. This is consistent with the impact of low access to information leading to low rates of adoption [70], although in other contexts farmers owning more resources were found to adopt agroforestry more readily [41]. H'mong farmers expressed willingness to adopt agroforestry despite their limited capacity in comparison to the Kinh and Thai, consistent with positive perception being an important factor in agroforestry adoption [71]. The research also shows that ethnicity linked to socio-economic contexts highly influenced agroforestry adoption. For example, the Thai displayed the highest potential for adopting or expanding agroforestry, which can be linked to the wider range of tree-species suitable in mid-level altitudes and socio-economic factors such as the access to market, seedlings, and information. The choice of agroforestry practices also differed with access to market by the same ethnic group-the H'mong in Dien Bien province were more interested in fruit trees than the Hmong in Yen Bai because of the former's greater access to the highway and local markets. Furthermore, despite having the same farming tradition, the Thai in Son La have less concern about marketing agroforestry products in contrast to those in Dien Bien, whose primary concern is market distance and transportation.

Positive capacity
Have available land, labour, finance, techniques Do not like to adopt agroforestry (for non-adopter farmers) or only want to maintain current agroforestry adoption (for adopter farmers) Have available land, labour, finance, techniques Have positive attitude towards agroforestry, willing to adopt (for non-adopter farmers) or expand agroforestry (for adopter farmers) Negative capacity Type 4: Lack of available land, labour, finance, techniques Do not like to adopt agroforestry (for non-adopter farmers) or only want to maintain current agroforestry adoption (for adopter farmers) Type 2: Lack of available land, labour, finance, techniques Have positive attitude towards agroforestry, willing to adopt (for non-adopter farmers) or expand agroforestry (for adopter farmers) We examined the capacity and willingness of non-adopters, wherein more than half (53%), and mostly Thai and Kinh were Type 4 farmers who were unwilling to adopt or expand agroforestry due to limited capacity in land, labour, financing and technical management (Figure 4), and also because of a lack of evidence on agroforestry benefits. Type 3, comprising 14% of non-adopters, mostly belonged to the H'mong ethnic group who expressed a lack of know-how on agroforestry.

Preferred Agroforestry Options for Different Ethnic Groups
Based on the understanding of capacity, willingness and motivation relating to agroforestry adoption, farmers were able to identify suitable tree-based options which fitted their local social-ecological context and that were economically viable. Fruit tree intercropping systems were the common interest of all groups because fruits have high selling prices. The other systems were identified based partly on individual farmer motivations for agroforestry. For example, H'mong farmers tended to prefer an annual crop component in their practices to provide food. In contrast, Thai farmers liked to have grass for livestock, and Kinh farmers wanted to improve soil by growing nitrogen-fixing species such as peanut or soybean (Table 4).
Understanding local preferences and motivation to adopt agroforestry helps project teams to modify trials and match them with the local interest. For example, the grass component was removed from practices for H'mong farmers, while peanut/pineapple was added into the practices offered to Kinh farmers. Grass was maintained for the Thai group.

Ethnicity and Agroforestry Adoption
Results of this research clearly demonstrate that ethnicity associated with specific contexts highly influences agroforestry adoption as well as the relevance of potential designs. The results suggest that the Thai ethnic group had the highest potential for adopting agroforestry and for expanding their current practices in the near future. This is a combination of the advantages of living at medium elevations, which were suitable for more tree species coupled with moderate access to market, seedlings, and information. The Kinh were more technologically advanced and had good access to markets, but their land size was restricted. The H'mong were more isolated and had the most difficult agricultural conditions, although they had the largest land holding sizes.
The number of non-adopters in Group 4 (i.e., those with no capacity and unwillingness to adopt) was much smaller than the number of the adopter group (Figures 4 and 5). Adopter farmers had received training by the project teams about agroforestry benefits. This is consistent with low access to information leading to low rates of adoption [70], although in other contexts farmers owning more resources have been found to adopt agroforestry more readily [41]. In this research, a higher percentage of H'mong farmers were willing to adopt agroforestry compared to Kinh and Thai, although they had more limited resources. However, only the H'mong farmers who had spontaneously adopted wanted to expand their agroforestry practices (Figures 4 and 5). Famers' positive perceptions towards agroforestry is important for adoption [71]; therefore, this brings up a high potential for scaling out agroforestry adoption for the H'mong, who are particularly vulnerable to poverty, climate change, and degraded landscapes.
Farmers' perspectives on benefits from agroforestry generally align with scientific findings about positive impacts of agroforestry such as soil loss reduction [10], improving soil fertility [8,71], providing food and fodder [18,71], and reducing the impact of extreme weather [72], but it was difficult for farmers to recognise some benefits that they could not observe, such as climate resilience or carbon sequestration. People's attitudes to future agroforestry options were also heavily context-and ethnic-specific. For example, H'mong farmers in two different provinces, Dien Bien and Yen Bai, had different preferred options. This can be explained by variations in their access to the highway (i.e., a context variable rather than an ethnic variable). The options that farmers wanted to have in Table 4 were consistent with their motivations to adopt agroforestry in Figure 3, and fits with their context. H'mong famers preferred cash crops or fruit trees with short-term crops for food and high income in a sufficiently short time period (3-4 years). Thai farmers wanted to have fodder grass for their livestock, and Kinh farmers wanted to have legume plants in their systems. However, it is clear that market factors play an overwhelmingly important role in the choices of farmers and their likelihood of change over the short term [73]. Except for the Thai in Dien Bien and the H'mong in Yen Bai whose fruit trees do not grow well, all three ethnic groups in other areas liked to grow fruit trees with the species dependent on market availability. However, it is clear that market factors play an overwhelmingly important role in the choices of farmers and their likelihood of change over the short term [73]. Except for the Thai in Dien Bien and the H'mong in Yen Bai whose fruit trees do not grow well, all three ethnic groups in other areas liked to grow fruit trees with the species dependent on market availability. Farmers adopt tree species and modify how they adopt them based on their own needs [74]. Various adaptations were observed from the survey, such as planting one or double grass strips between tree rows, and growing trees at different spacings and densities. This was quite similar to the adoption patterns in Nepal [71], where farmers in the lowlands did not opt to grow fodder trees with paddy rice, but upland farmers integrated fodder trees with rainfed crops to feed livestock, which then provide manure to fertilise crops. Unlike the farmers planting trees to improve soil fertility in Malawi [75] or for fuel and timber in Ethiopia [76], most farmers in northwest Vietnam wanted high-value fruit trees suitable for their biophysical condition.
The complexity and diversity of agroforestry adoption makes it difficult to monitor and evaluate the benefits of agroforestry accrued to adopters and non-adopters. The longterm nature of trees makes this even more challenging, because farmers intercrop tree and crops for the first few years but may focus entirely on the trees after a few years once they have become productive. Attitudes to adoption may change when the context changes (such as project support and market availability); our observations represent farmers attitudes at the time we conducted our survey, although this will include their accumulated experience.

How Farmers Transition from Being Non-Adopters to Adopters
Previously, adopters have been classified as "real adopters", "testers" or "pseudo adopters" depending on the permanency of their use of particular practices [8] conditioned by how farmers may have benefited from projects socially or materially beyond Farmers adopt tree species and modify how they adopt them based on their own needs [74]. Various adaptations were observed from the survey, such as planting one or double grass strips between tree rows, and growing trees at different spacings and densities. This was quite similar to the adoption patterns in Nepal [71], where farmers in the lowlands did not opt to grow fodder trees with paddy rice, but upland farmers integrated fodder trees with rainfed crops to feed livestock, which then provide manure to fertilise crops. Unlike the farmers planting trees to improve soil fertility in Malawi [75] or for fuel and timber in Ethiopia [76], most farmers in northwest Vietnam wanted high-value fruit trees suitable for their biophysical condition.
The complexity and diversity of agroforestry adoption makes it difficult to monitor and evaluate the benefits of agroforestry accrued to adopters and non-adopters. The long-term nature of trees makes this even more challenging, because farmers intercrop tree and crops for the first few years but may focus entirely on the trees after a few years once they have become productive. Attitudes to adoption may change when the context changes (such as project support and market availability); our observations represent farmers attitudes at the time we conducted our survey, although this will include their accumulated experience.

How Farmers Transition from Being Non-Adopters to Adopters
Previously, adopters have been classified as "real adopters", "testers" or "pseudo adopters" depending on the permanency of their use of particular practices [8] conditioned by how farmers may have benefited from projects socially or materially beyond what they derive from the agroforestry practices that they may be using. Therefore, the adopters might implement the technologies just because of the incentives from a project rather than the intrinsic value of the practice. In this study, it is unclear how permanently the farmers in the project adopter group will use agroforestry practices, although 61% expressed a capacity and willingness to adopt agroforestry. On the other hand, non-adopter farmers who want to adopt agroforestry might be more committed because they are willing to do it on their own without any support. This requires further research to track adoption and adaptation over long periods of time, encompassing the entire length of agroforestry rotations.
The social condition that enables the scaling out of diversified farming systems requires changes in people's aspirations and actions [77]. In order to scale out agroforestry adoption, it is necessary to shift farmers from other types into Type 1 (have capacity and willingness to adopt) from both adopters and non-adopters. Moving from Type 2 to Type 1 requires them to improve their capacity factors, such as labour, finance, or land, which is quite difficult and dependent on external support. The most feasible option is to increase their access to social capital in order to hire more labour or rent more land. Moving from Type 3 to Type 1 requires a change in farmers' attitudes and perceptions, addressing behavioural controls constraining agroforestry adoption. Farmers can change by themselves if they see the success from project agroforestry trials, or learn from neighbours, friends, and social media. This is sustainable because farmers combine the new information and techniques with their indigenous knowledge to generate locally adapted practice, consistent with farmers adopting agroforestry in Malawi [78]. Successful agroforestry trials should be promoted widely through different channels which farmers can easily access. The proportion of Type 4 farmers who are adopters is much smaller than the non-adopters, meaning that more farmers wanted to continue adopting after some time working with the project.

Policy Implications for Scaling out Agroforestry
Policy plays an important role in the widespread adoption of agroforestry [79]. This study suggests that for the effective adoption of agroforestry, government policy must respond to the needs at farmer level and fit into their specific contexts, which is consistent with recommendations from various authors [80]. Findings from this study show that addressing perceived behaviour controls enables the conditions for farmers to adopt agroforestry. Strong correlation often exists between farmers' intention to adopt or maintain agroforestry and their behavioural controls [43], but behavioural controls vary amongst ethnic groups and are related to their contexts, which differ from one community to another [81]. Therefore, policies from governmental or development projects should be tailored to different ethnicities in different locations and modified in light of their local knowledge and practices. This research also revealed that men and women often appeared to have different concerns towards agroforestry adoption. For example, Thai women were concerned about the market for fruits, while H'mong men wanted to learn more about agroforestry techniques. Effective policies will, therefore, need to be based on a clear understanding of the needs of men and women. Further research on gender would help shed light on design principles for scaling out agroforestry. Furthermore, understanding different motivation for agroforestry adoption and the preferred options of different farmers could help policy makers and development projects design best-fit agroforestry practices for specific contexts. If an option does not fit with their existing practices, farmers are unlikely to be willing to adopt it [80].

Conclusions
This research has revealed contrasting social contexts for agroforestry adoption amongst three ethnic groups in northwest Viet Nam, the Kinh, Thai, and H'mong. Nonadopter and adopter farmers were categorised into four groups with different levels of capacity and willingness to adopt agroforestry. Farmers' behavioural controls constraining the adoption of agroforestry also varied among the three ethnic groups and were influenced by their location, accessibility to market, and different cultivation traditions. Most farmers lacked access to high-quality tree seedlings and connections to markets. Kinh and Thai farmers in lowland areas were concerned about climate change and the high cost of managing agroforestry and preferred high-value fruit trees. H'mong people, in contrast, were more concerned about the efficiency of using fertiliser and labour and accessing required financial resources to support the purchase of seedlings and fertilisers. This understanding contributes to selecting which agroforestry interventions are most likely to be suitable for different ethnic groups, and what supporting policies are required to enable adoption.  In particular all participation in the research was voluntary, and all information, including any confidential personal data, was anonymised and stored in appropriately secure facilities, following current World Agroforestry protocols.
Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author. The data are not publicly available due to privacy.